TOPIC T3: DIMENSIONAL ANALYSIS AUTUMN 2012

Objectives

(1) Be able to determine the dimensions of physical quantities in terms of fundamental

dimensions.
(2) Understand the Principle of Dimensional Homogeneity and its use in checking
equations and reducing physical problems.
(3) Be able to carry out a formal dimensional analysis using Buckingham’s Pi Theorem.
(4) Understand the requirements of physical modelling and its limitations.

1. What is dimensional analysis?

2. Dimensions
2.1 Dimensions and units
2.2 Primary dimensions
2.3 Dimensions of derived quantities
2.4 Working out dimensions
2.5 Alternative choices for primary dimensions

3. Formal procedure for dimensional analysis

3.1 Dimensional homogeneity
3.2 Buckingham’s Pi theorem
3.3 Applications

4. Physical modelling
4.1 Method
4.2 Incomplete similarity (“scale effects”)
4.3 Froude-number scaling

5. Non-dimensional groups in fluid mechanics

References

White (2006) – Chapter 5

Hamill (2001) – Chapter 10
Chadwick and Morfett (2004) – Chapter 11
Massey (2005) – Chapter 5

Hydraulics 2 T3-1 David Apsley

1. WHAT IS DIMENSIONAL ANALYSIS?

Dimensional analysis is a means of simplifying a physical problem by appealing to

dimensional homogeneity to reduce the number of relevant variables.

It is particularly useful for:

• presenting and interpreting experimental data;
• attacking problems not amenable to a direct theoretical solution;
• checking equations;
• establishing the relative importance of particular physical phenomena;
• physical modelling.

Example.
The drag force F per unit length on a long smooth cylinder is a function of air speed U,
density , diameter D and viscosity . However, instead of having to draw hundreds of
graphs portraying its variation with all combinations of these parameters, dimensional
analysis tells us that the problem can be reduced to a single dimensionless relationship
c D = f (Re)
where cD is the drag coefficient and Re is the Reynolds number.

In this instance dimensional analysis has reduced the number of relevant variables from 5 to
2 and the experimental data to a single graph of cD against Re.

2. DIMENSIONS

2.1 Dimensions and Units

A dimension is the type of physical quantity.

A unit is a means of assigning a numerical value to that quantity.

SI units are preferred in scientific work.

2.2 Primary Dimensions

In fluid mechanics the primary or fundamental dimensions, together with their SI units are:
mass M (kilogram, kg)
length L (metre, m)
time T (second, s)
temperature (kelvin, K)

In other areas of physics additional dimensions may be necessary. The complete set specified
by the SI system consists of the above plus
electric current I (ampere, A)
luminous intensity C (candela, cd)
amount of substance n (mole, mol)

Hydraulics 2 T3-2 David Apsley

2.3 Dimensions of Derived Quantities

Dimensions of common derived mechanical quantities are given in the following table.

Quantity Common Symbol(s) Dimensions

Area A L2
Geometry Volume V L3
Second moment of area I L4
Velocity U LT–1
Acceleration a LT–2
Angle 1 (i.e. dimensionless)
Kinematics
Angular velocity T–1
Quantity of flow Q L3T–1
Mass flow rate m& MT–1
Force F MLT–2
Moment, torque T ML2T–2
Dynamics Energy, work, heat E, W ML2T–2
Power P ML2T–3
Pressure, stress p, ML–1T–2
Density ML–3
Viscosity ML–1T–1
Kinematic viscosity L2T–1
Fluid properties Surface tension MT–2
Thermal conductivity k MLT–3 –1
Specific heat cp, cv L2T–2 –1
Bulk modulus K ML–1T–2

2.4 Working Out Dimensions

In the following, [ ] means “dimensions of”.

Example.
dU
Use the definition = to determine the dimensions of viscosity.
dy

Solution.
From the definition,
force / area
= =
dU/dy velocity / length
Hence,
MLT −2 /L2
[ ]= = ML−1T −1
LT −1 /L

Alternatively, dimensions may be deduced indirectly from any known formula involving that
quantity.

Hydraulics 2 T3-3 David Apsley

Example.
UL
Since Re = is known to be dimensionless, the dimensions of must be the same as

those of UL; i.e.

[ ] = [ UL] = (ML−3 )(LT −1 )(L) = ML−1T −1

2.5 Alternative Choices For Primary Dimensions

The choice of primary dimensions is not unique. It is not uncommon – and it may sometimes
be more convenient – to choose force F as a primary dimension rather than mass, and have a
{FLT} system rather than {MLT}.

Example.
Find the dimensions of viscosity in the {FLT} rather than {MLT} systems.

Answer: [ ] = FL–2T

Hydraulics 2 T3-4 David Apsley

3. FORMAL PROCEDURE FOR DIMENSIONAL ANALYSIS

3.1 Dimensional Homogeneity

The Principle of Dimensional Homogeneity

All additive terms in a physical equation must have the same dimensions.

Examples:
s = ut + 12 at 2 – all terms have the dimensions of length (L)

p+ 1
2
V 2 + gz = p 0 – all terms have the dimensions of pressure (ML–1T–2)

p V2
+ +z=H – all terms have the dimensions of length (L)
g 2g

Dimensional homogeneity is a useful tool for checking formulae. For this reason it is useful
when analysing a physical problem to retain algebraic symbols for as long as possible and
only substitute numbers right at the end. Note, however, that dimensional analysis cannot
determine numerical factors; e.g. it cannot distinguish between ½at2 and at2 in the first
formula above.

Dimensional homogeneity is the basis of the formal dimensional analysis that follows.

3.2 Buckingham’s Pi Theorem

Experienced practitioners can do dimensional analysis by inspection. However, the formal

tool which they are unconsciously using is Buckingham’s Pi Theorem1:

Buckingham’s Pi Theorem

(1) If a problem involves n relevant variables and m independent dimensions then it can be
reduced to a relationship between n – m non-dimensional parameters 1, ..., n-m.

(2) These non-dimensional groups can be constructed as follows.

(i) Choose m dimensionally-distinct scaling variables (aka repeating variables).
(ii) For each of the n – m remaining variables construct a non-dimensional of the form
= (variable)( scale1 ) a ( scale2 ) b ( scale3 ) c L
where a, b, c, ... are chosen so as to make each non-dimensional.

Note. It is common to choose for scaling variables a geometric quantity (e.g. a length), a
kinematic (time-containing) quantity (e.g. velocity or acceleration) and a dynamic (mass- or
force-containing) quantity (e.g. density).

1
Buckingham, E., 1914. The use of comes from its use as the mathematical symbol for a product.

Hydraulics 2 T3-5 David Apsley

3.3 Applications

Example.
Obtain an expression in non-dimensional form for the pressure gradient in a horizontal pipe
of circular cross-section. Show how this relates to the familiar expression for frictional head
loss.

Step 1. Identify the relevant variables.

dp/dx, , V, D, ks,

Step 2. Write down dimensions.

dp [ force / area] MLT −2 × L−2
= = ML−2 T −2
dx length L
–3
ML
V LT–1
D L
ks L
ML–1T–1

Step 3. Establish the number of independent dimensions and non-dimensional groups.

Number of relevant variables: n=6
Number of independent dimensions: m = 3 (M, L and T)
Number of non-dimensional groups ( s): n – m = 3

Step 4. Choose m (= 3) dimensionally-independent scaling variables.

e.g. geometric (D), kinematic/time-dependent (V), dynamic/mass-dependent ( ).

Step 5. Create the s by non-dimensionalising the remaining variables: dp/dx, ks and .

dp a b c
1 = D V
dx
Considering the dimensions of both sides:
M 0 L0 T 0 = (ML−2 T −2 )(L) a (LT −1 ) b (ML−3 ) c
= M 1+c L−2+ a +b−3c T − 2−b
Equate powers of primary dimensions. Since M only appears in [ ] and T only
appears in [V] it is sensible to deal with these first.
M: 0=1+c ⇒ c = –1
T: 0 = –2 – b ⇒ b = –2
L: 0 = –2 + a + b –3c ⇒ a = 2 – b + 3c = 1
Hence,
dp
D
dp
1 = DV − 2 −1 = d2x (Check: OK – ratio of two pressures)
dx V

ks
2 = (by inspection, since ks is a length)
D

Hydraulics 2 T3-6 David Apsley

 3= D aV b c
In terms of dimensions:
M 0 L0 T 0 = (ML−1T −1 )(L) a (LT −1 ) b (ML−3 ) c
= M 1+c L−1+ a +b −3c T −1−b
Equating exponents:
M: 0=1+c ⇒ c = –1
T: 0 = –1 – b ⇒ b = –1
L: 0 = –1 + a + b –3c ⇒ a = 1 – b + 3c = –1
Hence,

3 = (Check: OK – this is the reciprocal of the Reynolds number)

VD

Step 6. Set out the non-dimensional relationship.

1 = f ( 2, 3)
or
dp
D
dx = f ( k s , ) (*)
V2 D VD

Step 7. Rearrange (if required) for convenience.

We are free to replace any of the s by a power of that , or by a product with the
other s, provided we retain the same number of independent dimensionless groups.
In this case we recognise that 3 is the reciprocal of the Reynolds number, so it looks
better to use ′3 = ( 3 ) −1 = Re as the third non-dimensional group. We can also write
dp hf
the pressure gradient in terms of head loss: = g . With these two modifications
dx L
the non-dimensional relationship (*) then becomes
gh f D k
2
= f ( s , Re)
LV D
or
L V2 k
hf = × × f ( s , Re)
D g D
Since numerical factors can be absorbed into the non-specified function, this can
easily be identified with the Darcy-Weisbach equation
L V2
hf =
D 2g
where is a function of relative roughness ks/D and Reynolds number Re, a function
given (Topic 2) by the Colebrook-White equation.

Hydraulics 2 T3-7 David Apsley

Example.
The drag force on a body in a fluid flow is a function of the body size (expressed via a
characteristic length L) and the fluid velocity V, density and viscosity . Perform a
dimensional analysis to reduce this to a single functional dependence
c D = f (Re)
where cD is a drag coefficient and Re is the Reynolds number.

What additional non-dimensional groups might appear in practice?

Notes.
(1) Dimensional analysis simply says that there is a relationship; it doesn’t (except in the
case of a single , which must, therefore, be constant) say what the relationship is.
For the specific relationship one must appeal to theory or, more commonly,
experimental data.

(2) If 1, 2, 3, ... are suitable non-dimensional groups then we are liberty to replace
some or all of them by any powers or products with the other s, provided that we
retain the same number of independent non-dimensional groups; e.g. ( 1)–1, ( 2)2,
2
1/( 3) .

(3) It is extremely common in fluid mechanics to find (often after the rearrangement
mentioned in (2)) certain combinations which can be recognised as key parameters
such as the Reynolds number ( Re = UL/ ) or Froude number ( Fr = U/ gL ).

(4) Often the hardest part of the dimensional analysis is determining which are the
relevant variables. For example, surface tension is always present in free-surface
flows, but can be neglected if the Weber number We = U2L/ is large. Similarly, all
fluids are compressible, but compressibility effects on the flow can be ignored if the
Mach number (Ma = U/c) is small; i.e. velocity is much less than the speed of sound.

(5) Although three primary dimensions (M,L,T) may appear when the variables are listed,
they do not do so independently. The following example illustrates a case where M
and T always appear in the combination MT–2, hence giving only one independent
dimension.

Hydraulics 2 T3-8 David Apsley

Example.
The tip deflection of a cantilever beam is a function of tip load W, beam length l, second
moment of area I and Young’s modulus E. Perform a dimensional analysis of this problem.

Step 1. Identify the relevant variables.

, W, l, I, E.

Step 2. Write down dimensions.

L
W MLT–2
l L
I L4
E ML–1T–2

Step 3. Establish the number of independent dimensions and non-dimensional groups.

Number of relevant variables: n=5
Number of independent dimensions: m = 2 (L and MT–2 - note)
Number of non-dimensional groups ( s): n – m = 3

Step 4. Choose m (= 2) dimensionally-independent scaling variables.

e.g. geometric (l), mass- or time-dependent (E).

Step 5. Create the s by non-dimensionalising the remaining variables: , I and W.

These give (after some algebra, not reproduced here):

1 =
l
I
2 = 4
l
W
3 = 2
El

Step 6. Set out the non-dimensional relationship.

1 = f ( 2, 3)
or
I W
= f( 4 , 2 )
l l El

This is as far as dimensional analysis will get us. Detailed theory shows that, for small elastic
deflections,
1 Wl 3
=
3 EI
or
−1
1 W   I 
=  2 × 4 
l 3  El   l 

Hydraulics 2 T3-9 David Apsley

4. PHYSICAL MODELLING

4.1 Method

If a dimensional analysis indicates that a problem is described by a functional relationship

between non-dimensional parameters 1, 2, 3, ... then full similarity requires that these
parameters be the same at both full (“prototype”) scale and model scale. i.e.
( 1 )m = ( 1) p
( ) =(
2 m 2 )p
etc.

Example.
A prototype gate valve which will control the flow in a pipe system conveying paraffin is to
be studied in a model. List the significant variables on which the pressure drop across the
valve would depend. Perform dimensional analysis to obtain the relevant non-dimensional
groups.

A 1/5 scale model is built to determine the pressure drop across the valve with water as the
working fluid.

(a) For a particular opening, when the velocity of paraffin in the prototype is 3.0 m s–1
what should be the velocity of water in the model for dynamic similarity?

(b) What is the ratio of the quantities of flow in prototype and model?

(c) Find the pressure drop in the prototype if it is 60 kPa in the model.

(The density and viscosity of paraffin are 800 kg m–3 and 0.002 kg m–1 s–1 respectively. Take
the kinematic viscosity of water as 1.0 ×10–6 m2 s–1).

Solution.
The pressure drop p is expected to depend upon the gate opening h, the overall depth d, the
velocity V, density and viscosity .

List the relevant variables:

p, h, d, V, ,

Write down dimensions:

p ML–1T–2
h L
d L
V LT–1
ML–3
ML–1T–1

Number of variables: n = 6
Number of independent dimensions: m = 3 (M, L and T)
Number of non-dimensional groups: n – m = 3

Hydraulics 2 T3-10 David Apsley

Choose m (= 3) scaling variables:
geometric (d); kinematic/time-dependent (V); dynamic/mass-dependent ( ).

Form dimensionless groups by non-dimensionalising the remaining variables: p, h and .

1 = pd aV b c

M L0 T 0 = (ML−1T −2 )(L) a (LT −1 ) b (ML−3 ) c

0

= M 1+c L−1+ a +b −3c T − 2−b

M: 0=1+c ⇒ c = –1
T: 0 = –2 – b ⇒ b = –2
L: 0 = –1 + a + b – 3c ⇒ a = 1 + 3c – b = 0
− 2 −1 p
⇒ 1 = pV =
V2

h
2 = (by inspection, since h is a length)
d

3 = d aV b c
(probably obvious by now, but here goes anyway ...)
M L T = (ML T )(L) a (LT −1 ) b (ML−3 ) c
0 0 0 −1 −1

= M 1+c L−1+ a +b −3c T −1−b

M: 0=1+c ⇒ c = –1
T: 0 = –1 – b + 0 ⇒ b = –1
L: 0 = –1 +a + b – 3c ⇒ a = 1 + 3c – b = –1
⇒ 3 = d −1V −1 −1
=
Vd
Recognition of the Reynolds number suggests that we replace 3 by
Vd
′3 = ( 3 ) −1 =

Hence, dimensional analysis yields

1 = f ( 2,
′3 )
i.e.
p h Vd
2
= f( , )
V d

(a) Dynamic similarity requires that all non-dimensional groups be the same in model and
prototype; i.e.
 p   p 
1 =
  = 
2 

2 
 V  p  V m
h h
2 =  =  (automatic if similar shape; i.e. “geometric similarity”)
 d  p  d m

Hydraulics 2 T3-11 David Apsley

  Vd   Vd 
′3 =   =  
 p  m
From the last, we have a velocity ratio
Vp ( / ) p dm 0.002 / 800 1
= = × = 0 .5
Vm ( / ) m d p 1.0 × 10 − 6 5
Hence,
Vp 3.0
Vm = = = 6.0 m s −1
0.5 0.5

(b) The ratio of the quantities of flow is

Q p (velocity × area ) p
2
Vp  d p 
= =   = 0 .5 × 5 2 = 12.5
Qm (velocity × area ) m Vm  d m 

(c) Finally, for the pressure drop,

2
 p   p  ( p) p  Vp  800
=   =   =   = × 0.5 2 = 0.2
p
1 2  2 
⇒
 V  p  V m ( p) m m  Vm  1000
Hence,
p p = 0 .2 × p m = 0.2 × 60 = 12.0 kPa

Hydraulics 2 T3-12 David Apsley

4.2 Incomplete Similarity (“Scale Effects”)

For a multi-parameter problem it is often not possible to achieve full similarity. In particular,
it is rare to be able to achieve full Reynolds-number scaling when other dimensionless
parameters are also involved. For hydraulic modelling of flows with a free surface the most
important requirement is Froude-number scaling (Section 4.3)

It is common to distinguish three levels of similarity.

Geometric similarity – the ratio of all corresponding lengths in model and prototype are the
same (i.e. they have the same shape).

Kinematic similarity – the ratio of all corresponding lengths and times (and hence the ratios
of all corresponding velocities) in model and prototype are the same.

Dynamic similarity – the ratio of all forces in model and prototype are the same; e.g.
Re = (inertial force) / (viscous force) is the same in both.

Geometric similarity is almost always assumed. However, in some applications – notably

river modelling – it is necessary to distort vertical scales to prevent undue influence of, for
example, surface tension or bed roughness.

Achieving full similarity is particularly a problem with the Reynolds number Re = UL/ .
• Using the same working fluid would require a velocity ratio inversely proportional to
the length-scale ratio and hence impractically large velocities in the scale model.
• A velocity scale fixed by, for example, the Froude number (see Section 4.3) means
that the only way to maintain the same Reynolds number is to adjust the kinematic
viscosity (substantially).

In practice, Reynolds-number similarity is unimportant if flows in both model and prototype

are fully turbulent; then momentum transport by viscous stresses is much less than that by
turbulent eddies and so the precise value of molecular viscosity is unimportant. In some
cases this may mean deliberately triggering transition to turbulence in boundary layers (for
example by the use of tripping wires or roughness strips).

Surface effects

Full geometric similarity requires that not only the main dimensions of objects but also the
surface roughness and, for mobile beds, the sediment size be in proportion. This would put
impossible requirements on surface finish or grain size. In practice, it is sufficient that the
surface be aerodynamically rough: u k s / ≥ 5 , where u = w / is the friction velocity and
ks a typical height of surface irregularities. This imposes a minimum velocity in model tests.

Other Fluid Phenomena

When scaled down in size, fluid phenomena which were negligible at full scale may become
important in laboratory models. A common example is surface tension.

Hydraulics 2 T3-13 David Apsley

4.3 Froude-Number Scaling

The most important parameter to preserve in hydraulic modelling of free-surface flows driven
by gravity is the Froude number, Fr = U / gL . Preserving this parameter between model (m)
and prototype (p) dictates the scaling of other variables in terms of the length scale ratio.

Velocity
(Fr ) m = (Fr ) p
1/ 2
 U    U m  Lm 
  = U  ⇒ =
 gL   gL  U L 
 m  p p  p
i.e. the velocity ratio is the square root of the length-scale ratio.

Quantity of flow
5/ 2
Qm  Lm 
Q ~ velocity × area ⇒ =
Q p  L p 
Force
3
Fm  Lm 
F ~ pressure × area ⇒ =
F p  L p 
This arises since the pressure ratio is equal to the length-scale ratio – this can be seen from
either hydrostatics (pressure ∝ height) or from the dynamic pressure (proportional to
(velocity)2 which, from the Froude number, is proportional to length).

Time
1/ 2
t m  Lm 
t ~ length ÷ velocity ⇒ =
t p  L p 

Hence the quantity of flow scales as the length-scale ratio to the 5/2 power, whilst the
time-scale ratio is the square root of the length-scale ratio. For example, at 1:100 geometric
scale, a full-scale tidal period of 12.4 hours becomes 1.24 hours.

Example.
The force exerted on a bridge pier in a river is to be tested in a 1:10 scale model using water
as the working fluid. In the prototype the depth of water is 2.0 m, the velocity of flow is
1.5 m s–1 and the width of the river is 20 m.

(a) List the variables affecting the force on the pier and perform dimensional analysis.
Can you satisfy all the conditions for complete similarity? What is the most important
parameter to choose for dynamic similarity?

(b) What are the depth, velocity and quantity of flow in the model?

(c) If the hydrodynamic force on the model bridge pier is 5 N, what would it be on the
prototype?

Hydraulics 2 T3-14 David Apsley

5. NON-DIMENSIONAL GROUPS IN FLUID MECHANICS

Dynamic similarity requires that the ratio of all forces be the same. The ratio of different
forces produces many of the key non-dimensional parameters in fluid mechanics.

(Note that “inertial force” means “mass × acceleration” – since it is equal to the total applied
force it is often one of the two “forces” in the ratio.)

UL inertial force
Reynolds number Re = = (viscous flows)
viscous force
1/ 2
U  inertial force 
Froude number Fr = =   (free-surface flows)
gL  gravitational force 
U 2L inertial force
Weber number We = = (surface tension)
surface tension
U inertial force
Rossby number Ro = = (rotating flows)
L Coriolis force
1/ 2
U  inertial force 
Mach number Ma = =   (compressible flows)
c  compressibility force 

These groups occur regularly when dimensional analysis is applied to fluid-dynamical

problems. They can be derived by considering forces on a small volume of fluid. They can
also be derived by non-dimensionalising the differential equations of fluid flow (see White,
Chapter 5), or the online notes for the 4th-year Computational Hydraulics unit.

Hydraulics 2 T3-15 David Apsley