Dimensional Analysis

A. Application of Dimensional Analysis:

1. To reduce the number of variables in an experimental program
(numerical or laboratory)
2. To convert from one set of units to another
3. Model testing
4. To check equations

B. Dimensional Homogeneity:
The concept of dimensional homogeneity says that all terms in equations
must have the same unit, e.g. The Bernoulli Equation is in units of m – N/N.
From Newton’s 2nd Law two basic systems of units may be used, F , L , T , θ
and the M , L , T , θ system. All other properties and fluid variable are
expressed from these.
F = Force L = length θ = temperature
M = Mass T = time

C. Dimensional Analysis:
The current and historical motivation for dimensional analysis is that data
collection and analysis is minimized greatly.
For illustrative purposes consider the problem of determining the drag F on a
sphere of diameter D moving through a viscous ( µ ) incompressible fluid with
velocity V . The fluid density is ρ . We hypothesize that F is a function of
these variables, F = f (D , V ,µ,ρ ) . Since to determine the relationship
experimentally requires an investigation of each variable holding the others
constant a great number of plots will result. This entails many spheres of
various diameters, and a variety of fluids of varying viscosity and density. For
instance, if we had 10 values for each of the four independent variables we
would need 10 4 = 10,000 laboratory measurements to collect information on
the phenomenon.
Let’s see how we can drastically reduce the amount of laboratory work
needed to study the problem.

1
D. Buckingham Pi Theorem:
We can now try dimensional analysis to form a sequence of dimensionless π
groupings to reduce the amount of data taking.

1. List all variables

Quantity Reasoning
F , drag force primary quantity of interest
D , sphere size bigger the sphere the greater the drag
µ , viscosity higher friction, more drag

ρ, fluid density the sphere must accelerate fluid past it

and more force is require for a higher
density fluid
V , velocity affects the shear forces

2. For the M , L , T system the units of the quantities are

ML M M L
FD = Fdrag = ; D = L; µ = ; ρ = 3 ; V =
T2 LT L T

3. The π theorem requires an exponent balance.

MoLo T o = FDv ρ w µ x D y V z (any chain rule sequence of terms will do)

4. Substitute dimensions
v w x z
 ML   M   M  y L 
M L T =  2   3    [L ]  
o o o

 T   L   LT  T 

2
5. Expand and equate exponents
on M : 0 = v + w + x
on L : 0 = v − 3w − x + y + z
on T : 0 = − 2v − x − z
Note: We now have 3 equations and five unknowns. The best we can do
is to solve for three unknowns in terms of the other two. Which 3 we solve for
is optional however some intuition will often provide a more useful result. The
following guidelines will help in selecting which 3 “repeating variables” to
choose.
i) From a set of variables that has m (1, 2 or 3) of the basic dimensions

(M, L, T ) . Select m variables that have amongst them all the m basics.
ii) Look at those of only major importance to the study.
iii) Usually inertial terms are important (mass × accel.) or (momentum sec )
ρ , Q , V are good choices.

6. For this problem choose FD , D and V as repeaters, which

means we’ll solve v , y and z in terms of w and x
v = −w−x  v = − w − x
 
y = − v + 3w + x − z  → y = 2w + x  reduced set
z = − 2v − x  z = 2w + x 
 

7. Substitute exponents into original theorem of step 3 and

obtain:
MoLo T o = FD(-w -x ) ρ w µ x D (2w + x ) v (2w + x )
Collecting terms with the same exponent, w and x here, we get:
w x
 ρv 2D 2   µvD 
o o o
MLT =    
 FD   FD 
The bracketed terms are dimensionless and called “ π ” terms.
ρv 2D 2 µvD
π1 = and π 3 =
FD FD

3
The original problem FD = f (D, V, ρ, µ ) has now been reduced to f (π1 , π 3 ) =0.

We can now rearrange this equation to suit our needs if necessary. Since π
terms are nondimensional they can be manipulated by cross multiplication
and/or division. An infinite number of π terms is possible but a minimum
number is computed by the following formula.
 Number independen t 
   Number of basic   Minimum number of 
 variables involved in  −   =  
 phenonema   dimensions used   terms for complete set 
 

In the example we had 5 independent variables and used all three dimensions
(M, L, T ) so we’ll need (5) – (3) = 2 π terms which will be independent of each
other and therefore form a complete set.

Finally for our problem, since we are interested in the behavior of FD as a

function of π1 and π 3 lets us divide it out. So let’s form a new variable

π1 ρv 2D 2 µvD ρvD
π2 = = = = π 2 = Reynolds Number,
π3 FD FD µ

which is measure of inertia forces to viscous forces. Our final nondimensional

parameter equation is f (π1 , π 2 ) = 0 which can be plotted on one graph (as
opposed to many we saw earlier.

ρV 2D 2
π1 =
FD

π1aπ
2a

π1a ρVD
π2a π2 =
µ

4
If we desire the drag for the conditions Va , D a , ρ a , µ a we need go to the

ρ a v aDa
graph and evaluate π 2 a = . Then π1a is read off the graph and Fa is
µa

ρv 2D2
computed from the equation Fa = .
π1a

5
Class Example:
In studying the thrust of a screw propeller it is judged that the following
variables are involved.
Thrust force F MLT-2 Make fnctional
Propeller Diam D L relationship of D, V
Ship Velocity V LT-1 and ρ .
Angular Velocity That is, these
N T-1
(RPM) three are the
Gravity g LT-2 repeaters
Fluid Density ρ ML-3
Kinematic Viscos. υ L2T-1
We have 7 independent variables and we have used all 3 dimensions so we
expect 7 − 3 = 4 π terms. We need not write all 7 simultaneous equations;
instead we’ll consider one new variable combined with the repeaters.
–––––“Tag on” Method.
x
(
π1 = D x1 v y1 ρ z1F → (L ) 1 LT -1 ) (ML ) (MLT )
y1 -3 z -2 1
= MoLo T o
Now solve:
M: 0 = z1 + 1  x1 = − 2
 F
L : 0 = x 1 + y1 − 3z1 + 1 ⇒ y 1 = − 2 ∴ π1 =
 z = -1 D v 2ñ
2

T: 0 = - y1 − 2  1

Similarly
DN Dg υ
π2 = ; π3 = ; π4 = ← This is reciprocal of R, Reynolds
V V2 DV
number
Let π 2 = D x 2 v y 2 ρ z2 N , etc

The solution could be π1 as some function of the other 3 π terms.

F DN Dg υ 
π1 = φ(π 2 , π 3 , π 4 ) or = φ , 2, 
D V ρ
2
2
 V V DV 
experiments would now find φ , the functional relationship amongst the π
terms or groups of variables.

6
7
Similitude
If we are to use a model to analyze a large fluids problem similitude between
the model (m) and the prototype (p ) must exist. Similitude means that all flow
details in m and p must be the same. Two similarities exist.
A. Geometric: i.e. a scaled model
B. Kinematic: velocity, force, and accelerations are identical.

Dynamic similarity exists when A & B exists i.e.

Dynamic Similarity = Geometric Similarity + Kinematic Similarity
Consider a prototype (p ) and model (m) sphere moving at v op & v om

respectively through a fluid

v op dVp
ρp vρopm
dVm
vµomp ρp
Dp
Dm
v om
µm

Model
Prototype

for corresponding element masses, ρ dV , dynamic similarity requires that

force polygons be similar.

Pressure
force (fp )p
(fp )m
(fi)m
inertial
friction force (fi)p (fτ )m
force (fτ )p

8
From similar polygons it follows that
∂v
(fi)m (fi)p (fi)m (fi)p Recall : τ=µ
= and = ∂n
(fτ )m (fτ)p (fp )m (fp)p dV = (dn)(dm)(ds )
This holds for all mass elements dV in dynamic similarity. Referring to the
figure let’s be precise about the forces.

ds
∂τ  ∂v 
τ + dn inertial force: fi = ma = ρ dV  v 
∂n  ∂s 
∂τ
+ s friction force: fτ = τA = dn (dm ds )
dm ∂n
∂  ∂v 
dm or by substitution: fτ =  µ dn(dm ds)
τ ∂n  ∂n 
The force ratios become:

fi
=
ρdV v ∂v
∂s
(=
)
ρu ∂u s
∂ = intertial force
fτ  ∂  ∂v   ∂ v
2
viscous force
  µ  dndm ds µ 2
 ∂n  n   ∂n

Since similarity exists the free stream, Vo , velocity ratios also represent the

point velocity ratios and all the ratios of the velocity derivatives. Hence v can
be replaced by Vo . From geometric similarities the distances between

corresponding points must also be in the same ratio as the ratio of the
boundary lengths. Therefore dn and ds may be replaced by the sphere
diameter D . For dynamic similarity
 ρv ∂v   ρv ∂v     Vo 
ρV Vo
 ∂s  =  ∂s  ⇒  o D  =  ρVo D 
 µ ∂ 2v   µ ∂ 2v   Vo   Vo 
 ∂n  m 
2
∂n 2  p  µ D 2  

µ 2 
D p
m

which reduces to

 ρVoD   ρVoD  R m = R p high R implies high inertia forces and

 µ  =  
  m  µ  p ↑ Reynolds Number relatively unimportant viscous force

9
A similar analysis for
(fi) gives
(fp )
 po   po  The Euler number is the ratio of pressure to
 ρV 2  =  ρV 2  Em = Ep
 o m  o p inertial forces.

 F   ρvD 
Earlier we hound that π1 = f (π 2 ) ⇒  2 2  = f  
 ρv D   µ 
F
If ~ p and V ~ Vo , p ~ p o then
D2
 po   ρv oD 
 ρV 2  = G µ  ⇒ E = function G[R] ;
 o  

therefore, if R duplicated then E is also.

10
Common Dimensionless Groups and their Interpretation
In fluid mechanics five π groups are recognized in studies without heat
transfer considerations.

1. Euler Number: =
Pressure force
⇒
p
∝
pA
, (E )
Inertia force ρV 2
ρQV
p
In testing work the “pressure coefficient” , equal to twice the E , is
1
2 ρV 2
used in pipe flow, orifice flow, open channel flow and drag.
ρVL
2. Reynolds Number: =
Inertial forces
⇒ , (R )
Viscous forces µ
Used to classify flow regimes; also pipeflow, boundary layers and
immersed flow.

3. Froude Number: =
Inertial force
⇒
V2
, (F )
gravity force gL
This is sometimes called the “kinetic flow factor.” It is most often used in
river hydraulics. All problems involving a free surface use the F number.

Open channel flow classification, ship design, hydraulic jumps and

spillway models.
Inertial forces ρV 2
4. Cauchy Number: = ⇒
Elastic forces K

this is also known as the Mach number

V
(M)
K ρ

applies to compressible flow

ρV 2L
5. Weber Number: =
Inertial forces
⇒ , (W)
Surface Tensio Forces σ
applies to capillary waves, fluid interfaces, orifices and weirs.

11
Model Testing
If analytical methods are too difficult or unreliable, the hydraulic engineer
resorts to model studies. However, model studies are usually costly.

Similarity:
Similarity exists when the prototype and model π terms are equal.
π1m = π1p , LL, πnm = πnp . In hydraulic engineering the majority of model

studies are based on the Reynolds or Froude criteria as being the major π
terms. The other π terms are of minor importance and are neglected except
in cases where the particular phenomenon they represent is under study.

12
Example 1
A 0.15m φ sphere models a 0.6m φ sphere. The prototype travels at 100 kph
(28 m/sec) in air.

a) For dynamic similarity what should the wind tunnel test speed be?
Solution:The object is immersed so R p modeling applies.

 ρVD   ρVD 
Rp = Rm ,   =  
 µ p  µ m
We’ll assume that the airs are the same:
i.e., ρ m = ρp and µ m = µ p

The effects of compressibility are ignored for now.

∴ [VD ]m = [VD]p

Dp  0.6 
Vm = Vp = 100  = 400 kph (112 m/sec)
Dm  0.15 

b) Is the incompressibility assumption valid?

In air compressibility must be considered when the fluid speed exceeds
3/10 the speed of sound in air or when M > 0.3 . For our problem the limit

would be 370 kph (0.30 x 1230 kph). We are just over it.

C) How can we fix this? By changing the fluid. What if we used a water
tunnel instead of air? Rewrite the Reynolds modeling statement as:
 ρVD   ρVD   VD   VD 
 µ  =  µ   υ  =  υ 
 m  p m p

Since υair = 1.0(10-5), υH2O = 1.0(10-6), 10:1 ratio

υm D υ  0.6 
Hence: Vm = Vp = 0.1 100 = 40 kph (11 fps)
υp D m  0.15 
Another solution would be to increase the 0.15m φ sphere slightly and
use the air tunnel.

13
Given: Vertical (Z ) component of Navier-Stokes equation

∂w ∂w ∂w ∂w 1 ∂p  ∂2W ∂2W ∂2W 

+u +v +w = FB − + v 2 + + 
2 
∂t ∂x ∂y ∂z ρ ∂z  ∂x ∂y 2
∂z 
if FB = - g (body force)
w v u x y z
W = , V = ,U= , X = , Y = , Z =
Ur Ur Ur L L L
p t
P = , T =
ρU 2r L/U r

L – characteristic reference length; Ur – characteristic reference velocity

Write the z-component equation in a dimensionless form.
L
Multiply by
ρUr2

∂W ∂W ∂W ∂W gL ∂P µ  ∂ 2W ∂2W ∂2W 
+U +V +W = − − +  + + 
∂T ∂X ∂Y ∂Z Ur2 ∂z ρUrL  ∂X 2 ∂Y 2 ∂Z 2 

1 ∂P 1  ∂ 2W ∂ 2 W ∂ 2 W 
= − − +  + + 
2 
Fr2 ∂Z R e  ∂X 2
∂Y 2
∂Z 
Ur ρUr L
Fr = ; Re =
gL µ

The equation is now only dependent on Fr & R e and the geometry associated

with the flow.

∴ If two flows have same Fr & R e number with similar geometry. The flows

are equivalent.

14
15