You are on page 1of 127

Department of Mechanical Engineering

The University of Hong Kong

Mechanics of Fluids MECH2008 (2008 – 2009)

Lecturer: Dr. C.O. Ng (office: HW7-1; phone: 28592622; email:

Required Text: Fundamentals of Fluid Mechanics 5th Ed., B.R. Munson, D.F. Young & T.H.
Okiishi, Wiley Asia Student Edition.

References: 1) Fluid Mechanics: Fundamentals and Applications, Y.A. Cengel & J.M.
Cimbala, McGraw-Hill.
2) Fluid Mechanics 6th Ed., F.M.White, McGraw-Hill.
3) Multi-Media Fluid Mechanics CD-ROM, Cambridge University Press.

Assessment: In-course continuous assessment 10%

• a mid-term test to be announced later
Examination in December 90%

Topics Covered:

1. Flow kinematics with differential vector calculus

2. Differential equations of motion

3. Unidirectional viscous flow and hydrodynamic lubrication

4. Potential flow and stream function

5. Boundary layer and drag

6. Open-channel flow and fluid machines


This is a Level-II Mechanics of Fluids course demanding the knowledge you acquired in the
first-year fluid and mathematics courses. In particular, the following topics are relevant and
should be reviewed if you have already forgotten the stuffs:-
1) Properties of fluid (density, viscosity);
2) Principles of fluid statics;
3) Fluid dynamics by control volume approach
¾ continuity equation
¾ energy equation (or Bernoulli equation)
¾ momentum equation
¾ head, head loss
4) Differentiation and integration;
5) Vector differential calculus (grad, div, curl, Gauss theorem, etc);

Lecture Notes and Worked Examples
(The corresponding section numbers in the textbook or other references are noted wherever


A. Description of Fluid Motion (Section 4.1.1)

• Lagrangian description: fluid particles are “tagged” or identified; rate of change of
flow properties as observed by following a fixed particle; variables are functions of
the initial position of particles and time.
• Eulerian description: fluid properties and variables are field variables, which are
functions of position in space (with respect to a fixed frame of reference) and time.
The Eulerian description, which is comparable to the data recorded by a measuring
device fixed in position, is more convenient to use in fluid mechanics.

Eulerian and Lagrangian descriptions In the Lagrangian description, one

of temperature of a fluid discharging must keep track of the position and
from a smoke stack velocity of individual particles

• Rectangular (Cartesian) coordinates:

z x = ( x, y, z ) = ( x1 , x2 , x3 ) = xi ( i = 1, 2,3)
V = (u , v, w) = (u1 , u2 , u3 ) = ui ( i = 1, 2,3)
⎛ ∂ ∂ ∂ ⎞ ⎛ ∂ ∂ ∂ ⎞ ∂
y ∇=⎜ , , ⎟=⎜ , , ⎟= ( i = 1, 2,3)
⎝ ∂x ∂y ∂z ⎠ ⎝ ∂x1 ∂x2 ∂x3 ⎠ ∂xi
x ⎛ ∂ ∂ ∂ ⎞ ∂u ∂v ∂w ∂ui
∇ iV = ⎜ , , ⎟i(u , v, w) = + + =
⎝ ∂x ∂y ∂z ⎠ ∂x ∂y ∂z ∂xi

pressure p ( x , t ) - scalar (0th order tensor)

• Primitive variables:
velocity V ( x, t ) - vector (1st order tensor)

Deduced variable stress τ ( x, t ) - 2nd order tensor

In the Eulerian description, one
defines field variables, such as the
pressure field and the velocity field,
at any location and instant in time.

B. Kinematics (Sections 4.2 and 6.1)

• Total (a.k.a. material, substantial) derivative = local rate of change + convective (or
advective) rate of change = the rate of change as observed following a particle of
fixed identity. It is an operator that can be applied to any scalar or vector quantity.

d ( ) ∂( )
= + (V i∇ )( )
dt ∂t
∂( ) ∂( ) ∂( ) ∂( )
= + u +v +w
∂t ∂x ∂y ∂z
local rate convective rate of change
of change

e.g., local acceleration =
∂V ∂V ∂V
convective acceleration = (V i∇ )V = u +v +w
∂x ∂y ∂z
- The local rate of change, also called the unsteady term, vanishes identically for
a steady flow. Therefore a flow is steady if and only if ∂ / ∂t ≡ 0 .
- The quantity (V i∇ ) is a scalar convective operator that determines the time
rate of change of any property (e.g., velocity, density, concentration,
temperature) of a particle by reason of the fact that the particle moves from a
place where the property has one value to another place where it has a
different value.

The total derivative is defined by

following a fluid particle as it moves
throughout the flow field. In this
illustration, the fluid particle is
accelerating to the right as it moves
up and to the right. A velocity field with respect to a fixed
frame of reference (x, y). A point fixed
in space is occupied by different fluid
particles at different time.

⎧Translation ⎫
⎪ ⎪
⎪ + ⎬ (rigid body motion)
⎪ Rotation ⎪⎭

General motion = ⎨ +
⎪Dilatation (change in volume)

⎪ +

⎩Angular deformation (change in shape)

As illustrated by:-

The various modes of deformation can be expressed in terms of the velocity gradients.

• Divergence of velocity is the volumetric strain/dilatation rate (rate of change of

volume per unit volume)
∂u ∂v ∂w
∇ iV ≡ + +
∂x ∂y ∂z
where ∂u ∂x , ∂v ∂y and ∂w ∂z are the components of the volumetric strain rate due
to elongation of a fluid element in the x-, y-, and z-directions, respectively.
Consider a small element of dimensions δ x × δ y × δ z :

Because of the velocity differential δ u over a distance δ x , the element is lengthened
in the x-direction by δ u ⋅ δ t over a small period of time δ t . The corresponding
change in volume is therefore
δ Vx = δ u ⋅ δ t ⋅ δ y ⋅ δ z ,
and the volume strain rate (change in volume per volume per time) is
δ Vx δ u ⋅ δ t ⋅ δ y ⋅ δ z δ u ∂u
= = = as δ x, δ t → 0 .
V ⋅ δ t δ x ⋅ δ y ⋅ δ z ⋅ δ t δ x ∂x
Similarly, for the lengthening of the element in the y- and z-directions
δ Vy ∂v δ Vz ∂w
= , = as δ y, δ z , δ t → 0
V ⋅ δ t ∂y V ⋅ δ t ∂z
The total volume strain rate is hence given by the divergence of the velocity.
δV ∂u ∂v ∂w
= + + as δ x, δ y, δ z , δ t → 0
V ⋅ δ t ∂x ∂y ∂z

In this incompressible flow, in which the

velocity divergence is identically zero, an
initially square parcel of marked fluid will
deform into a long thin shape (stretch in x-
direction, but shrink in the y-direction) in the
course of movement shown in the figure. The
flow is irrotational is this case.

• Any shear deformation can be decomposed into rigid body rotation and angular
deformation. Consider a small element undergoing shear deformation

Because of the velocity differential δ v over a distance δ x , the face OA rotates
counterclockwise by an angle δα = δ v ⋅ δ t / δ x over a small period of time δ t .
Therefore the angular velocity of OA is
dα ∂v
α= = as δ x, δ t → 0
dt ∂x
Similarly, the face OB rotates clockwise at an angular velocity given by
d β ∂u
β= = as δ y, δ t → 0
dt ∂y

O x

The deformation can be decomposed into a rigid body rotation at an angular velocity
1 ⎛ ∂v ∂u ⎞
( )
ω = α − β = ⎜ − ⎟ , where counterclockwise rotation is taken to be positive,
2 2 ⎝ ∂x ∂y ⎠

(α −β )

O x
α −β )
and an angular deformation, where the corner angle decreases at a rate given by
∂v ∂u
γ = α + β = + , where a positive rate means a decreasing angle,
∂x ∂y

(α +β )

O x
α +β )

• Rate of angular deformation of a 2-D fluid element moving in the x-y plane (angular
deformation is considered to be positive if it is to decrease the original right angle) is
hence defined to be
δα + δβ ∂v ∂u
γ xy = lim = + .
δ t →0 δt ∂x ∂y
For a 3-D element in general, the rate of change of the corner angle that is initially a
right angle between the i-j axes
∂u ∂u
γ ij = i + j ( i ≠ j ) ,
∂x j ∂xi

which is symmetric, i.e., γ ij = γ ji .

• Rotation of a fluid element (about an axis which is perpendicular to the plane of the
fluid motion) is the average of the angular velocities of the two mutually
perpendicular sides of the element, where counterclockwise rotation is considered to
be positive:
1 ⎛ ∂v ∂u ⎞
rotation about z -axis: ωz = ⎜ − ⎟
2 ⎝ ∂x ∂y ⎠
1 ⎛ ∂w ∂v ⎞
rotation about x-axis: ωx = ⎜ − ⎟
2 ⎝ ∂y ∂z ⎠
1 ⎛ ∂u ∂w ⎞
rotation about y -axis: ωy = ⎜ − ⎟
2 ⎝ ∂z ∂x ⎠

Rotation (or angular velocity) vector ω = (ω x , ω y , ω z )

Note that for a 2-D flow in the x-y plane, ωx and ω y vanish identically; hence the
rotation vector is always perpendicular to the x-y plane.

• To generalize, we may define

1 ⎛ ∂u ∂u ⎞ 1 1
shear rate tensor eij = ⎜ i + j ⎟ = angular deformation rate = γ ij , and
⎜ ⎟
2 ⎝ ∂x j ∂xi ⎠ 2 2

rate of rotation ω = vorticity, where

vorticity ζ = ∇ ×V (curl of velocity)

i j k
∂ ∂ ∂
∂x ∂y ∂z
u v w
⎛ ∂w ∂v ⎞ ⎛ ∂u ∂w ⎞ ⎛ ∂v ∂u ⎞
=⎜ − ⎟i + ⎜ − ⎟ j +⎜ − ⎟k
⎝ ∂y ∂z ⎠ ⎝ ∂z ∂x ⎠ ⎝ ∂x ∂y ⎠
2ω x 2ω y 2ω z

ζ = 2ω

The direction of a vector cross product is The vorticity vector is equal to twice the
determined by the right-hand rule. angular velocity vector of a rotating fluid

The difference between a rotational and irrotational flow: fluid elements in a rotational
region of the flow rotate about their own axis, but those in an irrotational region of the
flow do not.

In this incompressible and rotational flow, an

initially square fluid parcel will not only
elongate, but also rotate about its axis as it moves
over the time periods shown in the figure.

C. The Reynolds Transport Theorem (Section 4.4)

Two approaches of analyzing a problem. The Reynolds transport theorem

(a) System approach: follow the fluid as (RTT) provides a link between
it moves and deforms; no mass crosses the system approach and the
the boundary. (b) Control volume control volume approach.
approach: consider the changes in a
certain fixed volume; mass crosses the

- Material Volume: a volume that contains the same fluid as it moves and deforms
following the motion of the fluid
- Material Surface: enclosing surface of a material volume; by definition no fluid
particles can cross it.
- Control Volume: a volume of fluid in a flow field, usually fixed in space, to be
occupied by different fluid particles at different times.
- Control Surface: imaginary or physical enclosing surface of a control volume.
- Flux: amount of property (e.g., mass, momentum, energy) crossing a unit area of a
surface per unit time.

We state without proof the Reynolds transport theorem, which provides a basis for
developing differential equations for the various conservation laws:

d ∂
dt ∫∫∫MV ∂t ∫∫∫CV ∫∫
ρ bdV = ρ bdV + ρ bV i n d A

rate of change of the local rate of change of the net out-flux of the
property within property within the fixed property across the
the material volume control volume that happens entire control surface
to coincide with the material
volume at that instant


ρ = density of fluid
b = an intensive property B of fluid (property per unit mass)
MV = material volume that happens to coincide with CV at time t

CV = control volume (fixed in space)
CS = control surface
n = unit outward normal to CS

The integral of ρ bV indA over the

A moving system (hatched region)
and a fixed control volume (shaded control surface gives the net amount
region) in a diverging portion of a of the property B flowing out of the
flow field at times t and t + t . control volume (into the control
volume if it is negative) per unit time.

D. Conservation of Mass (Section 6.2)

If the property is mass, then b = 1, and

d d
L.H.S. ∫∫∫ ρ dV = ( mass in MV ) = 0
dt MV dt
(by definition of MV , which always contains the same fluid)

∂t ∫∫∫CV
R.H.S. ρ dV + ∫∫ ρV ind A

= ∫∫∫ dV + ∫∫∫ ∇ i( ρV )dV
CV ∂t CV

CV is stationary by Gauss theorem

Equating L.H.S. and R.H.S., and removing the volume integral since CV is arbitrary, we
get the differential form of Continuity Equation

+ ∇ i( ρV ) = 0
or, using the identity ∇ i( ρV ) = V i∇ρ + ρ∇ iV,

+ ρ∇ iV = 0
dρ ∂u
or in index form +ρ i =0 ( i, j = 1, 2,3, summation over repeated index )
dt ∂xi

INCOMPRESSIBLE FLOW is defined as one in which the density of a fluid particle

is invariant with time ⇔ = 0 , which implies
∇ iV = 0
(ie, divergence of velocity is zero for incompressible flow)
In Cartesian coordinates, the continuity equation for incompressible flow reads

∂u ∂v ∂w
+ + =0
∂x ∂y ∂z

Note that a flow with constant density is always incompressible, but an incompressible
flow does not necessarily have a constant density (e.g., flow in a stratified sea).

E. Applied Forces
• Body force due to gravity on a small fluid element = ρ gdV
• Surface stress s = τ in , where n is the unit outward normal vector to the surface, and

⎡τ xx τ xy τ xz ⎤
⎢ ⎥
τ = ⎢τ yx τ yy τ yz ⎥ = τ ij ( i, j = 1, 2, 3)
⎢τ zx τ zy τ zz ⎥
⎣ ⎦

are the stress components on an infinitesimal cubic fluid element.

τ ij is a second order tensor, where

the first index i denotes the face (on which the stress acts) being normal to xi ,
and the second index j denotes the stress component being in the x j direction.

In the textbook, the normal stress is denoted by σ ii in order to distinguish it from the
shear stress τ ij ( i ≠ j ) .
It can be shown that τ ij is symmetric, ie, τ ij = τ ji . Therefore there are only 6
independent stress components.

F. Conservation of Linear Momentum (Section 6.3)

Apply Newton’s second law of motion to a material volume of fluid:

dt ∫∫∫ ∫∫ sdA ∫∫∫ ρ gdV
ρVdV = +

rate of change surface body

of momentum stress force

The L.H.S. can be converted, using the transport theorem and the continuity equation,
d dV
into ∫∫∫
dt MV
ρVdV = ∫∫∫ ρ
dV .

The first term on the R.H.S. is ∫∫ sdA = ∫∫ τ indA = ∫∫∫ ∇iτdV

on using Gauss theorem.

Plugging these terms back, and removing the volume integral since the volume is
arbitrary, we get the differential form of momentum equation

ρ = ∇ iτ + ρ g

The left hand term is a total derivative, which can be expanded into the Eulerian form:

⎛ ∂V ⎞ ⎛ ∂ui ∂u ⎞ ∂τ
ρ⎜ + V i∇V ⎟ = ∇ iτ + ρ g or ρ⎜ + u j i ⎟ = ij + ρ gi
⎝ ∂t ⎠ ⎜ ∂t ∂x j ⎟⎠ ∂x j

By now, there are more unknowns than equations. To close the problem, we need to
introduce CONSTITUTIVE (stress vs strain-rate) relations to relate the stress and the
If the fluid is Newtonian, a linear relationship is followed

⎛ ∂ui ∂u j ⎞
τ ij = − pδ ij + μ ⎜ + ⎟
⎜ ∂x ∂xi ⎟⎠
⎝ j

⎧1 for i = j
where δ ij = ⎨ , μ = dynamic viscosity coefficient
⎩0 for i ≠ j

Finally, on substituting the above relationship, we obtain the Navier-Stokes equations

⎛ ∂V ⎞
ρ⎜ + V i∇V ⎟ = −∇p + ρ g + μ∇ 2V
⎝ ∂t ⎠

or in index form,

⎛ ∂ui ∂u ⎞ ∂p ∂ 2u
⎜ ∂t
+uj i ⎟ = − + ρ gi + μ 2i ( i, j = 1, 2,3, summation over repeated index )
⎝ ∂x j ⎟⎠ ∂xi ∂x j

∂ui ∂u 1 ∂p ∂ 2u
or +uj i = − + gi +ν 2i
∂t ∂x j ρ ∂xi ∂x j
(I) (II) (III) (IV) (V)

where ν = μ/ρ is the kinematic viscosity

Meanings of the five terms:-

(I) – local acceleration;

(II) – convective acceleration (inertia), nonlinear term of the equation;
(III) – pressure gradient;
(IV) – gravity;
(V) – viscous diffusion of momentum owing to molecular viscosity of the fluid.

Now, we have 4 equations (1 continuity + 3 components of momentum) for the four

variables u x , u y , u z , and p as functions of space and time ( x, y, z , t ) . Note that it is the
pressure gradient, rather than the pressure itself that drives the flow.

The Equations of Motion for an Incompressible Newtonian Fluid

In Rectangular Coordinates (x, y, z)

∂u x ∂u y ∂u z
Continuity: + + =0
∂x ∂y ∂z

∂u x ∂u ∂u ∂u 1 ∂p ⎛ ∂ 2u ∂ 2 u ∂ 2u ⎞
x-component: + ux x + u y x + uz x = − + ν ⎜ 2x + 2x + 2x ⎟ + g x
∂t ∂x ∂y ∂z ρ ∂x ⎝ ∂x ∂y ∂z ⎠

∂u y ∂u y ∂u y ∂u y 1 ∂p ⎛ ∂ 2u ∂ 2u ∂ 2u ⎞
y -component: + ux + uy + uz =− + ν ⎜ 2y + 2y + 2y ⎟⎟ + g y
∂t ∂x ∂y ∂z ρ ∂y ⎜ ∂x ∂y ∂z
⎝ ⎠

∂u z ∂u z ∂u z ∂u z 1 ∂p ⎛ ∂ 2 u z ∂ 2u z ∂ 2u z ⎞
z -component: + ux + uy + uz =− +ν ⎜ 2 + 2 + 2 ⎟ + g z
∂t ∂x ∂y ∂z ρ ∂z ⎝ ∂x ∂y ∂z ⎠

( g x , g y , g z ) are the components of the acceleration due to gravity in the x, y , and z directions.
If, say, x and y are horizontal axes and z is positive upward, then g x = g y = 0, and g z = − g .
Also, the gravity can be combined implicitly with the pressure term by introducing
p* ≡ p − ρ g i x = p − ρ ( g x x + g y y + g z z ) .

In Cylindrical Coordinates (r, θ, z)

1 ∂ ( rur ) 1 ∂uθ ∂u z
Continuity: + + =0
r ∂r r ∂θ ∂z

∂ur ∂u u ∂u u 2 ∂u 1 ∂p
r -component: + ur r + θ r − θ + u z r = −
∂t ∂r r ∂θ r ∂z ρ ∂r
⎡ ∂ ⎛1 ∂ 1 ∂ 2u 2 ∂u ∂ 2u ⎤
+ν ⎢ ⎜ ( rur ) ⎞⎟ + 2 2r − 2 θ + 2r ⎥ + g r
⎣ ∂r ⎝ r ∂r ⎠ r ∂θ r ∂θ ∂z ⎦

∂uθ ∂u u ∂u u u ∂u 1 ∂p
θ -component: + ur θ + θ θ + r θ + u z θ = −
∂t ∂r r ∂θ r ∂z ρ r ∂θ
⎡ ∂ ⎛1 ∂ 1 ∂ 2u 2 ∂u ∂ 2u ⎤
+ν ⎢ ⎜ ( ruθ ) ⎞⎟ + 2 2θ + 2 r + 2θ ⎥ + gθ
⎣ ∂r ⎝ r ∂r ⎠ r ∂θ r ∂θ ∂z ⎦

∂u z ∂u u ∂u ∂u 1 ∂p
z -component: + ur z + θ z + u z z = −
∂t ∂r r ∂θ ∂z ρ ∂z
⎡ 1 ∂ ⎛ ∂u z ⎞ 1 ∂ uz ∂ uz ⎤
2 2
+ν ⎢ ⎜r ⎟+ 2 + 2 ⎥ + gz
⎣ r ∂r ⎝ ∂r ⎠ r ∂θ ∂z ⎦

G. Scaling and Approximation
• Because of the inertia terms (convective acceleration), the Navier-Stokes (NS)
equations are non-linear equations.
• Except for simple flow geometry, analytical solutions do not exist in general.
• Fortunately, for many practical applications, not all terms in the equations are equally
important, and therefore some subdominant terms can be dropped in favor of a first
approximation of the problem. The approximate equations can then be solved
(analytically or numerically) with much greater ease than the full-blown ones.
• It is important to judge, for a particular problem, the relative significance of the
individual terms in the NS equations, which can be reflected from the magnitude of
the corresponding non-dimensional parameters.

For illustration, consider incompressible unsteady flow past a body:


Characteristic scales:
Length (L); Time scale of unsteadiness (T); Velocity (U); Pressure (P)

Introduce dimensionless variables (distinguished by *):

V* = V / U , t* = t / T , x* = x / L, p* = p / P, g* = g / g
the normalized NS can be expressed as

⎛ L ⎞ ∂V * ⎛ P ⎞ ∗ ⎛ gL ⎞ ⎛ ν ⎞ ∗2
⎜ ⎟ + V * i∇ ∗V* = − ⎜ 2 ⎟
∇ p * + ⎜ 2 ⎟ g* + ⎜ ⎟∇ V *
⎝ UT ⎠ ∂t * ⎝ ρU ⎠ ⎝U ⎠ ⎝ UL ⎠

The scales have been chosen to be representative of the variables so that all the
dimensionless terms are order unity. Now, the importance of each term (relative to
the inertia) is carried by its bracketed coefficient.

L temporal acceleration
= Strouhal number (St) =
UT convective accelertion
P pressure force
= Euler number (E) =
ρU 2
UL inerita
= Reynolds number (Re) =
ν viscous force
U inerita
= Froude number (Fr) =
gL gravity force

Possible Cases of Simplification:-

Large Re
Re 1 ⇒ negligible viscous effect,

∂V 1
NS reduces to Euler equations + V i∇V = − ∇p + g
∂t ρ

Small St
St 1 ⇒ negligible unsteady effect ⇒ quasi-steady flow,
∴ The local (temporal) acceleration term can be dropped.

Small Re
Re 1 ⇒ negligible inertia effect (good news!)
Viscous force is significant, and is to be balanced by pressure gradient.
For slow and viscous flow and negligible gravity, the flow is
called Creeping Flow
0 = − ∇p + ν∇ 2V
Nonlinear inertia terms are now gone, analytical solutions are possible
if the flow geometry is simple enough.

Spatial Dimension
Also, it is often the case that the flow varies only in one or two spatial dimensions,
and therefore the problem can be reduced to a one- or two-dimensional problem, for
which only one or two velocity components need to be solved. Some common cases
of one-dimensional flow:
• fully developed pipe or channel flow: axial velocity as a function of radial
distance from center of pipe u = u (r ) , or longitudinal velocity as a function of
distance from the bottom of channel u = u ( y ) ;
• axi-symmetrical flow: velocity is symmetrical about an axis (e.g., point
source/sink, vortex).

VISCOUS FLOW (Section 6.9)

A. Mathematical Formulation for a Fluid Dynamics Problem

• constant fluid properties (density ρ , viscosity μ )
• Newtonian fluid (linear, isotropic and purely viscous material)

Basic Variables:
Velocity V = (u , v, w) = V ( x, y, z , t ) (3 variables)
Pressure p = p ( x, y, z , t ) (1 variable)

Basic Governing Equations:

Continuity ∇ iV = 0 (1 equation)
∂V 1
Navier-Stokes + V i∇V = − ∇p + g +ν∇ 2V (3 equations)
∂t ρ

Other derived variables:

τ ( x, y, z, t ) = − pI + μ ⎡∇V + ( ∇V ) ⎤ , I = isotropic tensor

⎣ ⎦
with stress components (see the definition on page 11):
τ xx = − p + 2μ , (normal stress)
τ yy = − p + 2μ , (normal stress)
⎛ ∂u ∂v ⎞
τ xy = τ yx = μ ⎜ + ⎟ (shear stress)
⎝ ∂y ∂x ⎠

⎛ ∂w ∂v ⎞ ⎛ ∂u ∂w ⎞ ⎛ ∂v ∂u ⎞
Vorticity ζ = ∇ ×V = ⎜ − ⎟i + ⎜ − ⎟ j +⎜ − ⎟k
⎝ ∂y ∂z ⎠ ⎝ ∂z ∂x ⎠ ⎝ ∂x ∂y ⎠

Boundary Conditions:
• No-slip boundary condition: the velocity of a fluid in contact with a solid
impermeable wall must equal that of the wall
Vfluid = Vsolid along a fluid-solid interface

If in particular the wall is stationary, the fluid adjacent to the wall must have zero

The development of velocity profiles due to the no-slip condition as a fluid flows past
a blunt nose and a flat plate.

• Interface boundary condition between two fluids: when fluid A and fluid B meets at
an interface, the velocity and stress must match between the two fluids at the interface
VA = VB , τA = τB along a fluid-fluid interface

If, say, the interface is flat (along x-direction) and the fluids are moving parallel to the
interface, the continuity of stress implies the continuity of pressure and shear stress at
the interface
du du
pA = pB , μA = μB
dy A dy B

• Free-surface boundary condition: a degenerate form of the above interface boundary

condition occurs at the free-surface of a liquid, meaning that fluid A is a liquid (say,
water, oil) and fluid B is a gas (usually air). By virtue of the fact μair μliquid , the
shear stress at the air-liquid interface is negligibly small, and it is reasonable to
approximate the shear stress to be at the interface, which is hence called a free surface,
pliquid = patmosphere , μliquid =0 along the free surface
dy liquid

• Other boundary conditions, such as inlet condition, outlet condition, periodic
condition and symmetry, may also apply to certain types of boundaries, depending on
the problem.

Boundary conditions along a plane of symmetry are defined so as to ensure that the
flow field on one side of the symmetry plane is a mirror image of that on the other
side, as shown above for a horizontal symmetry plane.

Initial Condition If the problem is time dependent (i.e., unsteady), an initial condition also
needs to be specified.

Let us consider in the following sections a few applications of the Navier-Stokes equations,
in which the flow configuration is simple enough for analytical solutions (exact or
approximate) to be deduced. The assumptions are that the flow is steady ( ∴ ∂ / ∂t = 0 ),
laminar, and incompressible and the fluid is Newtonian.

B. Plane Poiseuille-Couette Flow

Upper plate moving at a constant speed U



Lower fixed plate

Note that this is a unidirectional flow u = u ( y ), v = 0 . Therefore there is no dependence

on x for all variables: ∂ / ∂x = 0 .

The flow is driven by three forcings: (1) motion of the upper plate; (2) pressure gradient
in the x-direction, ∂p / ∂x = a constant ; (3) gravity, if x is not in a horizontal direction.

Recall the momentum equations:

∂u ∂u ∂u 1 ∂p ⎛ ∂ 2u ∂ 2u ⎞
x-component: +u +v =− +ν ⎜ 2 + 2 ⎟ + g x
∂t ∂x ∂y ρ ∂x ⎜ ∂x ∂y ⎟⎠

∂ 2u 1 ∂p
⇒ν 2 = − g x ....................... (1)
∂y ρ ∂x

∂v ∂v ∂v 1 ∂p ⎛ ∂ 2v ∂ 2v ⎞
y -component: +u +v =− ν
+ ⎜ 2 + 2 ⎟⎟ + g y
∂t ∂x ∂y ρ ∂y ⎜ ∂x ∂y
⎝ ⎠
1 ∂p
⇒− + g y = 0 ....................... (2)
ρ ∂y

Note that the inertia terms are identically zero, which is true for all unidirectional flows
irrespective of the Reynolds number.

Equation (2) simply gives that the pressure p = p( x) + ρ g y y .

The R.H.S. of equation (1) is constant, so the equation can be integrated twice with
respect to y, giving
⎛ ∂p ⎞y
u( y) = ⎜ − ρ g x ⎟ + C1 y + C2
⎝ ∂x ⎠ 2μ
where C1 and C2 are integration constants that can be determined using the boundary
conditions that
u ( y = 0) = 0 (no slip at the lower plate), and
u ( y = h) = U (speed of the upper plate).

Solving for these constants, we obtain the solution for the velocity profile (see Fig. 6.31

⎛ ∂p ⎞ h ⎡y ⎛ y⎞ ⎤
2 2
u( y) = ⎜ − + ρ g x ⎟ ⎢ −⎜ ⎟ ⎥ + U
⎝ ∂x ⎠ 2 μ ⎣⎢ h ⎝ h ⎠ ⎦⎥ h
Couette Flow
Poiseuille Flow

Couette flow is caused by the motion of a boundary wall moving in its own plane, while
Poiseuille flow is caused by axial pressure gradient or gravity in the direction of flow.

The shear stress in the flow is

du ( y ) ⎛ ∂p ⎞⎡h ⎤ U
τ xy = μ = ⎜− + ρ gx ⎟ ⎢ − y⎥ + μ
dy ⎝ ∂x ⎠⎣2 ⎦ h
linear stress distribution due to constant stress
Poiseuille flow, due to Couette flow
zero stress at the centerline

The discharge (flow-rate) per unit width of channel is given by

⎛ ∂p ⎞ h
h hU
Q = ∫ udy = ⎜ − + ρ g x ⎟ +
⎝ ∂x ⎠ 12 μ 2

The volume flow averaged (mean) velocity

⎛ ∂p ⎞ h
u = Q / h = ⎜ − + ρ gx ⎟ +
⎝ ∂x ⎠ 12 μ 2

It is left as an exercise for you to show the following

Given that −∂p / ∂x is a positive constant and g x = 0 , determine the location of the
maximum velocity. It is also the point where the shear stress vanishes (why?).
Hence, find the minimum value of U such that the shear stress will not vanish
throughout the flow.

C. Circular Poiseuille Flow

We now consider laminar flow through a circular tube:
• The objective to find the relationship between volumetric flow rate and pressure
change along a pipe of circular section.
• Examples include blood flow in capillaries, air flow in lung alveoli, where the
Reynolds number is not high enough for the flow to become turbulent.
• Navier-Stokes equations in cylindrical coordinates are to be used, where
∂ / ∂θ = 0 , since the flow is axially-symmetric (i.e., no dependence on angular
position in a cross-section of the flow).
• We have seen that the gravity can be combined with the pressure gradient in a
trivial manner, so let us ignore gravity in the following analysis.

uz (r)

Circular pipe of radius R

Again, this is a unidirectional flow: ur = uθ = 0, u z ≠ 0 is driven by a constant and

steady pressure gradient dp/dz in the axial direction.

The continuity equation reduces to
1 ∂ ( rur ) 1 ∂uθ ∂u z
+ + = 0 ⇒ u z must not depend on z , ∴ uz = uz (r )
r ∂r r ∂θ ∂z

The z -component momentum equation is simplified to

∂u z ∂u u ∂u ∂u 1 ∂p
+ ur z + θ z + u z z = −
∂t ∂r r ∂θ ∂z ρ ∂z
⎡ 1 ∂ ⎛ ∂u z ⎞ 1 ∂ uz
∂ 2u z ⎤
+ν ⎢ ⎜r ⎟ + + ⎥ + gz
⎢⎣ r ∂r ⎝ ∂r ⎠ r ∂θ ∂z 2
2 2
d ⎛ du z ⎞ r dp
⇒ ⎜r ⎟= , which can be integrated twice with respect to r to give
dr ⎝ dr ⎠ μ dz
r 2 dp
u z (r ) = + C1 ln r + C2
4 μ dz
The two integration constants C1 and C2 can be determined using the boundary
u z (r = 0) is finite ⇒ C1 = 0
R 2 dp
u z (r = R) = 0 (no slip at boundary wall) ⇒ C2 = −
4μ dz
Plugging back, we get the expression for the velocity profile
dp R 2 ⎡ r 2 ⎤
uz ( r ) = − 1−
dz 4 μ ⎢⎣ R 2 ⎥⎦
which is a parabolic distribution with the maximum at the center:
R 2 dp
umax = u z (r = 0) = −
4 μ dz
The flow-rate is
R π R 4 dp
Q = ∫ u z dA = 2π ∫ u z rdr = −
A 0 8μ dz
The mean velocity is half the maximum velocity
π R 4 dp

u =Q/ A= 8 μ dz =−
R 2 dp umax
= ..................... (1)
π R2 8μ dz 2
The shear stress at wall is given by
du R dp u
τw = − μ z =− = 4μ
dr r = R 2 dz R

For a given length L of the pipe, the pressure drop is Δp = − ( dp / dz ) L and the head loss
due to friction is h f = Δp / ρ g . Hence we may obtain from equation (1) the Darcy-
Weisbach equation

L u2
hf = f (D = 2 R = diameter of pipe)
D 2g
64 ρ Du
where the Darcy friction factor f = , and Re = is the Reynolds number.
Re μ

Recall that the Moody diagram (attached below) provides the graphical functional
dependence of the friction factor on the Reynolds number and the relative wall
roughness. The above relation is represented by the straight line near the left end of the
figure, where the flow is laminar, or the Reynolds number Re < 2,000. Note that for
laminar flow, the friction loss is not affected by wall roughness ε .

The pipe flow becomes turbulent when Re > 4,000, for which the friction factor is given by
the empirical Colebrook formula

1 ⎛ε /D 2.51 ⎞
= −2.0 log10 ⎜

+ ⎟⎟ ,
f ⎝ 3.7 Re f ⎠
which is graphically represented in the Moody diagram above.

D. Nearly One-Dimensional Flow by Lubrication Approximation
Lubrication in a Slider Bearing
A slider bearing is designed as a thrust bearing to support very large loads. To carry these
loads, the fluid film between the solid surfaces must develop normal stresses, so we are
interested in predicting the pressure distribution and thus the load-carrying capacity of the
bearing. Typical examples of slider bearings are found in the shafts of screw-propelled
ships and in the high-speed turbines of electricity-generating stations. For example, the
thrust of the ship’s propellers may be transmitted through a series of pads (see the figure
below) to the hull of the ship. Each pad (slider) may be tilted slightly to account for the
relative effects of pressure, speed and viscosity, and thus maintain the fluid film between
the two surfaces (the slider and its guide) which are in relative motion, and thereby reduce

In most lubrication problems the relevant Reynolds number is so small that viscous terms
in the Navier-Stokes equation dominate completely. The reason is not necessarily that the
coefficient of viscosity is large; it is more due to the fact that the thickness of the film is
extremely small compared to the lateral dimensions of the bearing. The Reynolds number
may be defined as
ρ × slider speed × film thickness
Re = 1

Let us now formulate a model of the slider bearing with a planar face, with the further
assumptions that
1. The lubricant is an incompressible Newtonian viscous fluid with constant viscosity.
2. The bearing has infinite length into the paper, and the bearing guide is flat. The gap
height h(x) between the slider and its guide varies so gently that the flow is nearly
one-dimensional through a section of the bearing.
3. Gravity can be ignored.
4. The flow has settled down and we need consider only the steady problem.

The local film thickness is
⎛h −h ⎞
h( x) = h1 + ⎜ 2 1 ⎟ x (1)
⎝ L ⎠
where h1 < h2 L so that the film is extremely thin.

The flow is quasi-one-dimensional, and we may recall the equation for discharge for
combined Poiseuille-Couette flow (on page 31):

dp h3 hU
Q=− −
dx 12 μ 2

where the gravity has been ignored and U is now in the negative direction. By
conservation of mass, Q must be a constant. If there were Couette flow alone, the
discharge would decrease down the slider as the gap height decreases from h2 to h1 .
Therefore in order to balance the flow, the pressure gradient must not be zero.
Rearranging the terms the above equation gives

dp 12 μ ⎛ hU ⎞
=− 3 ⎜ +Q⎟ (2)
dx h ⎝ 2 ⎠

We further suppose that both ends of the bearing are exposed to surrounding lubricant,
or to the atmosphere. Then we have

p ( L) = p(0) = p0

as the boundary condition for the pressure. It follows that integrating (2) from 0 to L is
equal to zero
L dp
∫0 dx dx = p( L) − p(0) = 0
L⎛ U Q⎞
⇒ − ∫ ⎜ 2 + 3 ⎟dx = 0
⎝ 2h h ⎠

Now, on substituting (1) for h(x), the above integral can be carried out to give

Q= U
( h1 + h2 )
We may put Q back into (2), which is integrated again, but the upper limit is now a
general position “x”
x dp x⎛ U Q⎞
∫0 dx dx = p ( x ) − p0 = −12 μ ∫0 ⎜⎝ 2h2 + h3 ⎟⎠dx
After some algebra, we get
p( x) = p0 − ( h − h1 )( h − h2 )
h (h22 − h12 )

Note that since h1 ≤ h( x) ≤ h2 , p ( x) ≥ p0 or a positive pressure distribution is

established within the bearing fluid to support the normal load. It can be readily shown
that, by setting dp/dx = 0 in (2), the pressure reaches a maximum

3μUL ( h2 − h1 ) h1 L
pmax − p0 = at x =
2h1h2 ( h1 + h2 ) ( h1 + h2 )
It may be shown that in the left-hand section of the bearing, the pressure gradient is
positive so that it drives fluid in the flow direction, and in the right-hand section the
pressure gradient is negative so that it drives fluid against the flow direction. In this way,
the Poiseuille flow will balance the Couette flow to result in a constant discharge
throughout the pad.

Lubrication performance is found to be favorable, since by using orders of magnitude

we may estimate that


τ dx

shear stress

μU / h

μUL / h

Bearing load pdx pressure L

By virtue of the small film thickness, the bearing can support a large load with only
small frictional resistance.

Analysis can be considerably simplified if the flow under consideration can be regarded as

A. Inviscid (Nonviscous) Flow (Section 6.4)

• Flow of an ideal fluid with zero viscosity ( μ = 0 ) would be inviscid exactly.
• In practice, flow is approximately inviscid when the effects of shear stresses on the
motion are small as compared to other influences. One guiding condition is that the
Reynolds number Re must be very large:
viscous force 1 ρVL
∼ 1, where Re =
inertia force Re μ
• Many flows involving water or air, whose viscosity is small, can practically be
considered as inviscid as long as the viscous effects are not dominant (e.g., far from a
• When the viscous force becomes negligible, the Navier-Stokes equations reduce to
Euler’s equations
∂V 1
+ V i∇V = − ∇p + g (viscous term is missing here)
∂t nonlinear term ρ
is still here

• For incompressible flow, Euler’s equations of motion can be integrated along a

streamline to yield the Bernoulli equation (which you learnt already in Year I; read
Section 6.4.2 for a review)
p V
+ + z = constant along a streamline
ρ g 2g
• It is remarkable that the Bernoulli equation provides an algebraic (rather than vector
differential) relationship between pressure, velocity and position in the earth’s
gravitational field.

B. Irrotational (Potential) Flow (Section 6.4.3)

• Recall that vorticity (curl of velocity) is twice the rotation (angular velocity) of a fluid
• A fluid element will acquire vorticity when acted upon by a couple to cause it to
rotate. One source of rotation is unbalanced shear stresses acting on its periphery.
When shear stresses are absent, it is possible that the flow is irrotational.
• A flow field is irrotational if, at every point, the vorticity vanishes or
∇ ×V = 0 .
• It can be shown that the flow of an inviscid fluid which is irrotational at a particular
instant of time remains irrotational for all subsequent times. That means, the motion
of an inviscid fluid which is started from rest is always irrotational (provided the flow
lies outside a boundary layer).
• This result is known as the Persistence of Irrotational Motion of an inviscid fluid. It
is because the setting up of a rotation would require forces tangential to the boundary;
and such forces, which arise through the viscous properties of the fluid, are non-
existent in the inviscid fluid model.

• The constant in the Bernoulli equation becomes universal (i.e., not specific to a
streamline) when the flow is irrotational (Section 6.4.4). Therefore, for
incompressible irrotational flow, the Bernoulli equation can be applied between any
two points in the flow field:
p1 V12 p V2
+ + z1 = 2 + 2 + z2
ρ g 2g ρ g 2g
• The procedures of finding a solution for an irrotational flow field are typically:
o Firstly, solve for the kinematics (velocity components) from an equation
derived from the condition of zero vorticity, which is the subject matter of the
following sections;
o Secondly, find the pressure from the Bernoulli equation.
You should appreciate that solving irrotational flow equations is usually much simpler
than solving the full Navier-Stokes equations.
• You are cautioned that irrotationality fails to apply to a boundary layer, which is a
thin layer that develops next to a solid wall owing to no-slip of the flow at the wall.
No matter how small its viscosity is, a real fluid cannot “slide” past a solid boundary.
The flow in a boundary layer is always viscous and highly rotational (a rapid change
in velocity from zero at wall to the free stream value over a short distance); real fluid
behavior must be accounted for in a boundary layer (Chapter 9).

C. The Velocity Potential (Section 6.4.5)
• For any scalar field φ , curl(grad φ ) = 0 is an identity. See a proof below.

• Alternatively speaking, a velocity field V is irrotational or curl V = 0 if and only if

there exists a scalar field φ such that V = grad φ .

• The scalar function is called velocity potential

V ≡ ∇φ

∂φ ∂φ ∂φ
Cartesian coordinates: u= , v= , w=
∂x ∂y ∂z

∂φ 1 ∂φ ∂φ
Cylindrical coordinates: ur = , uθ = , uz =
∂r r ∂θ ∂z

Irrotational flow is therefore also called potential flow.

• The velocity potential satisfies Laplace’s equation on substituting the above relation
into the continuity equation:
∇ iV = 0 ⇒ ∇ i∇φ = 0, or ∇ 2φ = 0

∂ 2φ ∂ 2φ ∂ 2φ
Cartesian coordinates: + + =0
∂x 2 ∂y 2 ∂z 2

1 ∂ ⎛ ∂φ ⎞ 1 ∂ 2φ ∂ 2φ
Cylindrical coordinates: ⎜r ⎟+ + =0
r ∂r ⎝ ∂r ⎠ r 2 ∂θ 2 ∂z 2

• The immediate upshot is: for irrotational flow, one only needs to solve for a scalar
function (instead of a vector with 2 or 3 components) from one single equation in
order to determine the kinematics (good news!). However, the differential equation
for the scalar function is one order higher than that for the vector function (no free

lunch!). Once the potential is found, its spatial gradients will give the velocity

D. Equipotential Lines and Streamlines (Sections 6.2.3 and 6.5)

• A two-dimensional potential flow field can be graphically represented using a flow
net composed of equipotential lines and streamlines.
• Equipotential lines are (contour) lines of constant velocity potential, while streamlines
are lines in the flow field that are everywhere tangent to the velocity. It can be shown
that these two sets of lines are orthogonal (i.e., they intersect each other at right

• You may recall the following mathematical statement:

The grad of a scalar function, say ∇φ , gives the maximum rate of

spatial change of the function, and is in a direction normal to the local
line along which the function is constant..

It follows that:
∇φ ≡ V ⊥ equipotential lines
⇒ streamlines ⊥ equipotential lines

Figure 6.15 shows a flow net for a 90o bend. A
flow net is useful in the visualization of a flow
pattern. To further understand what information a
flow net can provide, we need to know
something about stream function.

Stream Function
• For 2-D incompressible flow, another scalar function, viz stream function can be
introduced to identically satisfy the continuity equation.

A stream function ψ ( x, y ) or ψ ( r ,θ ) is defined such that

∂ψ ∂ψ
u= , v=− for 2-D flow in Cartesian coordinates,
∂y ∂x
∂u ∂v
which satisfies + = 0 identically.
∂x ∂y
1 ∂ψ ∂ψ
ur = , uθ = − for 2-D flow in Polar coordinates,
r ∂θ ∂r
∂ ∂u
which satisfies ( rur ) + θ = 0 identically.
∂r ∂θ

Note that the stream function is introduced based on kinematics consideration only. It
is definable for any two-dimensional incompressible flow fields, irrespective of the
flow being inviscid or not.

• Physically, ψ is constant along a streamline since

∂ψ ∂ψ
dψ = dx + dy = −vdx + udy
∂x ∂y
That means, a line of constant ψ (along which dψ = 0 ) will have its slope in the same
direction of flow: dy / dx = v / u . This is nothing but the defining property for a

Note that a solid boundary is always a V

streamline. At a particular instant of time, ψ is constant
there is no fluid crossing any streamline,
and distinct streamlines cannot cross.

• Given any two points in space whose stream function values are known, then the
volume flow rate across any line joining these two points is equal to the difference in
values of their stream functions.

One can readiy see from the above figure that

∂ψ ∂ψ
dq = udy − vdx = dy + dx = dψ
∂y ∂x
∴ q = ∫ dψ = ψ 2 −ψ 1

• If the 2-D flow is irrotational, the stream function also satisfies Laplace’s equation,
∂v ∂u
∇ ×V = 0 ⇒ − =0
∂x ∂y
∂ ⎛ ∂ψ ⎞ ∂ ⎛ ∂ψ ⎞
⇒ ⎜− ⎟− ⎜ ⎟=0
∂x ⎝ ∂x ⎠ ∂y ⎝ ∂y ⎠
∂ 2ψ ∂ 2ψ
⇒ + =0
∂x 2 ∂y 2
Therefore, for two-dimensional irrotational flow, both the velocity potential φ and the
stream function ψ satisfy Laplace’s equation. They are called harmonic functions,
and they are harmonic conjugates of each other. These functions are related, but
their origins are different:

– The stream function is defined by continuity; the Laplace equation for ψ results
from irrotationality.

– The velocity potential is defined by irrotationality; the Laplace equation for φ

results from continuity.

By now, referring back to Figure 6.15, you should understand that in a flow net the
velocity is roughly given by
Δφ Δψ
V≈ ≈
Δn Δs

where Δn is the spacing between two adjacent equipotential lines, and Δs is the
spacing between two adjacent streamlines. Therefore, the velocity is higher in a
region where the mesh is finer, and lower where the mesh is coarser.

E. Some Simple Plane Potential Flows (Sections 6.5.1-6.5.4)
1) Uniform Flow with constant velocity U

For case (a) where the flow is purely in the x-direction:

Velocity potential φ = Ux = Ur cos θ

(∴ equipotential lines are parallel to the y -axis )

Stream function ψ = Uy = Ur sin θ

(∴ streamlines are parallel to the x-axis )
Can you write down the corresponding φ and ψ for case (b) where the flow is at an
angle α with the x-axis?

2) Source and Sink

A 2-D source is a line (from a mathematical perspective)
that runs perpendicular to the plane of flow and injects
fluid equally in all directions. The figure shows the flow
field of a source at the origin, from which fluid particles
emerge and follow radial pathlines. The strength of a
source, denoted by m, is the volume rate of flow
emanating from unit length of the line.

m =V / L

By conservation of mass, m = 2π rur for any radial distance r from the source located at the
∂φ 1 ∂ψ m 1 ∂φ ∂ψ
origin. Hence, ur = = = , uθ = =− = 0 . On integrating,
∂r r ∂θ 2π r r ∂θ ∂r

Velocity potential φ= ln r

(∴ equipotential lines are concentric circles centered on the origin )

Stream function ψ= θ

(∴ streamlines are radial lines )
The radial and tangential velocities are:
ur = uθ = 0
2π r
o when m > 0 , the flow is radially outward, the origin is a SOURCE
o when m < 0 , the flow is radially inward, the origin is a SINK
o the origin is a singularity where ur → ∞
o conservation of mass is satisfied everywhere except the origin

3) Vortex
In contrast to a source, a vortex has the pathlines
being circles centered on the origin, and fluid
particles move along these circles. The vortex can
be used to model the flow round the plughole in a
bathtub. An irrotational vortex is called a free
vortex. The strength of a vortex is measured by the
circulation Γ = ∫ V ids around a closed curve C

that encloses the center of the vortex. Hence,

∂φ 1 ∂ψ 1 ∂φ ∂ψ Γ
ur = = = 0, uθ = =− = .
∂r r ∂θ r ∂θ ∂r 2π r
On integrating,
Velocity potential φ= θ

(∴ equipotential lines are radial lines )

Stream function ψ =− ln r

(∴ streamlines are concentric circles centered on the origin )
The radial and tangential velocities for a free vortex are
ur = 0 uθ =
2π r

o the flow is not defined at the origin
o the vorticity curl V = 0, except at r = 0, where V is not defined
o free vortex (a) is irrotational flow, tangential velocity decreases radially uθ ∝ r −1
o forced vortex (b) is rotational flow, tangential velocity increases radially uθ ∝ r

4) Doublet
Consider a combination of a source and a sink of equal strength m and separated at a
distance 2a (left figure):

If the source and sink are moved indefinitely closer together ( a → 0 ) in such a way
that the product 2am (distance apart × strength) is kept finite and constant, then we
obtain a doublet. The streamline pattern for a doublet is shown in the right figure
above. The line joining the source to the sink is called the axis of the doublet, and is
taken to be positive in the direction from sink to source. The strength of the doublet
is K = ma / π .

K cos θ
Velocity potential φ=
(∴ equipotential lines are circles through the origin tangent to the y -axis )

K sin θ
Stream function ψ =−
(∴ streamlines are circles through the origin tangent to the x-axis )
The basic potential flows that have been discussed so far are more mathematical
constructions than physically realistic entities (although a source/sink may represent
the flow field of an injection/withdrawing well, and so on). However a combination of
these basic potential flows may provide a representation of some flow fields of
practical interest. This is the subject matter for the next section.

F. Superposition of Basic Potential Flows (Sections 6.6)

Let us first be reminded that for inviscid flow, a solid boundary is a streamline, and
conversely, a streamline can be considered as a solid boundary. The kinematic
conditions along the two are the same: normal velocity = 0. In fact, we may replace any
streamline in a flow field by an impermeable surface without disturbing the flow.

Since the governing equation (Laplace’s equation) for potential flow is linear,
superposition of solutions gives the solution to the combined effect. In the following

examples, you will see how ideal flows can be described by a combination of basic
solutions. The key thing is to locate the dividing streamline. The general procedures are
as follows. (1) Sketch some streamlines for the combined flow. (2) Find the location of a
stagnation point where the velocity vanishes (you may expect that when flow past a body,
there is a point somewhere on the body surface where the flow velocity is zero). (3)
Evaluate the stream function at the stagnation point ψ stagnation . (4) The dividing streamline,
which passes through the stagnation point, can be determined by letting the stream
function be equal to the stagnation stream function ψ ( r ,θ ) = ψ stagnation .

1) Source + Uniform Flow = Flow Past a Half Body

It is more convenient to use polar coordinates ( r ,θ ) where r is the radial distance

from the source. It is along the negative x-axis where the flow due the uniform flow is
directly opposite to that due the source. At a point x = −b the velocities due to the two
flows cancel each other, and this is identified as the stagnation point.

∴ uniform flow velocity U = radial outward flow due to source ur ( r = b )

m m
⇒U = ⇒b=
2π b 2π U

Combined stream function ψ ( r ,θ ) = ψ uniform flow + ψ source

= Uy + θ

= Ur sin θ + θ

The stream function at the stagnation point has the value
ψ stagnation = ψ ( r = b, θ = π ) = = π bU
Therefore the dividing streamline that passes through the stagnation point is given by

ψ ( r ,θ ) = ψ stagnation
m m
or Ur sin θ + θ=
2π 2
b (π − θ )
or r=
sin θ
or y = b (π − θ )

This streamline, which has the shape shown above, can be considered as a solid
boundary of a half-body that extends from x = −b to x → +∞ . The flow exterior to
this streamline represents the flow past a half-body, whose thickness at large x can be
estimated to be 2π b , since

⎧ πb ⎧0
y→⎨ as θ → ⎨
⎩−π b ⎩2π
Note that the every fluid particle emanated from the source is completely enclosed
within the dividing streamline. The flow pattern around the half-body is described by
streamlines ψ > ψ stagnation . The velocity components and the pressure can then be
determined as described in earlier sections.

2) Source + Sink + Uniform Flow = Flow Past a Rankine Oval

The source and the sink are of the same strength: any mass of fluid injected by the
source is eventually drawn into the sink. The dividing streamline is now a closed
curve. This finite body, called Rankine Oval, has two stagnation points, one at the
front end and the other at the rear end of its boundary.

3) Doublet + Uniform Flow = Flow Past a Circular Cylinder

As the source and the sink combine to become a doublet, the Rankine Oval becomes a
circular cylinder. As the flow past a circular cylinder is of fundamental interest, let us
examine the flow in some detail.

The combined stream function is
K sin θ
ψ = Ur sin θ −
(K = strength of doublet)

1 ∂ψ ⎛ K⎞
The radial velocity is ur = = ⎜ U − 2 ⎟ cos θ
r ∂θ ⎝ r ⎠

Obviously, the radial velocity vanishes at a circular surface with the radius
r = a = ( K /U )
1/ 2

This defines the dividing streamline, which represents the surface of a circular
cylinder of radius a. Substituting a for K, the stream function can be written as
⎛ a2 ⎞
ψ = Ur ⎜ 1 − 2 ⎟ sin θ
⎝ r ⎠
from which we obtain the velocity components
1 ∂ψ ⎛ a2 ⎞
ur = = U ⎜1 − 2 ⎟ cos θ
r ∂θ ⎝ r ⎠
∂ψ ⎛ a2 ⎞
uθ = − = −U ⎜1 + 2 ⎟ sin θ
∂r ⎝ r ⎠
On the cylinder surface r = a, the tangential velocity is uθ s = −2U sin θ . As expected,
there are 2 stagnation points, at θ = 0, π .

The pressure distribution on the cylinder surface can be found from the Bernoulli
1 1
p0 + ρU 2 = ps + ρ uθ2s
2 2
ps − p0 = ρU 2 (1 − 4sin 2 θ )


where p0 is the far upstream pressure. It is

remarkable that the pressure distribution is
symmetrical about the horizontal and the
vertical diameters. Therefore there is no net
force arising from the pressure distribution
around the cylinder in both streamwise and
lateral directions. In other words, both drag
and lift forces are exactly zero, as predicted
from the potential flow theory.

This zero drag prediction is contrary to what has been observed in reality. There is
always a significant drag developed on a cylinder when it is placed in a stream of
moving fluid. This discrepancy is called d’Alembert’s Paradox, which was not
explained until the concepts of boundary layer and flow separation were developed.
A comparison between the inviscid and the real pressure distributions is shown above.

4) Free Vortex + Doublet + Uniform Flow = Flow Past a Rotating Circular Cylinder

The effect of adding a vortex is to upset

the symmetry of flow about the
horizontal diameter. Therefore, the
pressure in the upper half of the cylinder
is not balanced by the pressure in the
lower half. This results in a net lift
force acting laterally on the cylinder.

5) Sink + Free Vortex = Spiral Flow

6) Two separated sources of equal strength = source flow with a neighboring wall

(Chapter 9)

A. Introduction (Section 9.1.2)

In 1904, Prandtl developed the concept of the boundary layer, which provides an
important link between ideal-fluid flow (inviscid irrotational flow) and real-fluid flow
(viscous rotational flow). It was accepted that for fluids with relatively small viscosity (or
more exactly, flow with a high Reynolds number), the effect of internal friction in the
fluid is appreciable only in a narrow region surrounding the fluid boundaries. Therefore
the flow sufficiently far away from the solid boundaries may be considered as ideal flow
(in which effects of viscosity are neglected). However, flow near the boundaries suffers
retardation by the boundary shear forces and at the boundaries the velocity is zero (no-
slip condition). A steep velocity gradient is therefore resulted in a thin layer adjacent to
the boundaries, which is known as the boundary layer. It is of great significance when
behavior of real fluid is considered. For example, it explains the d’Alembert’s paradox –
the drag force experienced by a cylinder in stream that cannot be predicted with a
potential theory.

Flow of a uniform stream parallel to

a flat plate. The larger the Reynolds
number, the thinner the boundary
layer along the plate at a given x-

Flow past a circular cylinder; the boundary layer separates from the surface of the body in
the wake for large Reynolds number.

B. Description of the Boundary Layer (Section 9.2.1)

(1) Development of the Boundary Layer

nominal limit
of boundary
u = 0.99 U


• On-coming flow is irrotational and has a uniform velocity U.

• The boundary layer starts out as a laminar boundary layer, in which fluid particles
move in smooth layers and the velocity distribution is approximately parabolic. As the
flow moves on, the continual action of shear stress tends to slow down additional
fluid particles, causing the boundary layer thickness to increase with distance
downstream from the leading edge. See below for a definition of the boundary layer
• The flow within the boundary layer is subject to wall shear, and dominated by viscous
forces. The velocity gradient (hence the rotation of fluid particles) is the largest at the
wall, and decreases with distance away from the wall, and tends to zero on matching
with the main stream flow. Roughly speaking, the flow is said to be rotational within
the boundary layer, but is irrotational outside the boundary layer.
• As the thickness of laminar boundary layer increases, it becomes unstable and some
eddying commences. These changes take place over a short length known as the
transition zone.
• It finally transforms into a turbulent boundary layer, in which particles move in
haphazard paths. Due to the turbulent mixing, the velocity distribution is much more
uniform than that in the laminar boundary layer. The increase of thickness along the
plate continues indefinitely but with a diminishing rate. If the plate is smooth (i.e.,
negligible roughness size), laminar flow persists in a very thin film called the viscous
sub-layer in immediate contact with the plate and it is in this sub-layer that the greater
part of the velocity change occurs.

Comparison of laminar and

turbulent flat plate
boundary layer profiles
(left: non-dimensionalized
by the boundary layer
thickness; right: in physical

(2) Thicknesses of the Boundary Layer
i) Boundary Layer Thickness δ
The velocity within the boundary layer increases to the velocity of the main stream
asymptotically. It is conventional to define the boundary layer thickness δ as the distance
from the boundary at which the velocity is 99% of the main stream velocity.

There are other ‘thicknesses’, precisely defined by mathematical expressions, which are
measures of the effect of the boundary layer on the flow.

ii) Displacement Thickness δ∗

It is defined by
∞  u 
δ * = ∫ 1 −  dy
 U 

δ∗ is the distance by which the boundary surface would have to be shifted outward if the
fluid were frictionless and carried at the same mass flowrate as the actual viscous flow. It
also represents the outward displacement of the streamlines caused by the viscous effects
on the plate. Conceptually one may ‘add’ this displacement thickness to the actual wall
and treat the flow over the ‘thickened’ body as an inviscid flow.

iii) Momentum Thickness θ

It is defined by
∞ u  u 
θ =∫ 1 −  dy
0 U  U 

θ is the thickness of a layer of the main stream whose flux of momentum equals the
deficiency in the boundary layer, equivalent to the loss of momentum flux per unit width
divided by ρU 2 due to the presence of the growing boundary layer. The momentum
thickness is often used when determining the drag on an object.

Note that when evaluating the above integrals for δ * and θ, the upper integration limit
can practically be replaced by δ.

C. Laminar Boundary Layer Over a Flat Plate
• Heuristic Analysis


leading edge
of plate

Consider steady flow past a flat plate at zero incidence. The effect of viscosity is to
diffuse momentum normal to the plate. Consider a fluid element that is close enough
to the wall to be influenced by viscosity. In travelling a distance x, it has been
influenced by viscosity for a time t ∼ x / U . The influence of viscosity will have
spread laterally to a distance
1/ 2
ν x 
δ ∼ (ν t ) ∼  
1/ 2

U 
1/ 2
ν 
∼   ≡ ( Re x )
−1/ 2
x  Ux 
The above analysis is rather crude, and does not yield a full equation for the growth of
the boundary layer thickness. It however correctly describes one important
relationship for the laminar boundary layer: δ / x ∝ ( Re x )
−1/ 2
where Re x ≡ Ux / ν is
the local Reynolds number in terms of the distance from the leading edge x. This
relationship is found to be valid at a distance far behind the leading edge: δ / x 1.

The heuristic analysis can be further carried on to find relations for the wall stress:
1/ 2
∂u τw U  νU 3 
τ w = ρν or ∼ν ∼  
∂y y =0
ρ δ  x 
The wall shear stress τ w decreases with increase of x until the boundary layer turns
turbulent. The local friction coefficient, which is defined as follows, is given by
∼ 2 ( Re x )
−1/ 2
Cf ≡ 1
ρU 2

While the numerical factor of 2 is far from the true value, the functional dependence
of Cf on Rex is correctly predicted.

• Exact Solution by Blasius (Section 9.2.2)
A more rigorous analysis, using the technique of similarity solution, was developed
by Blasius for the laminar boundary layer over a flat plate. While the details of the
analysis are beyond the scope of this course, it is important to note the following
results derived from Blasius’ solution.

δ ( x) 5
boundary layer thickness =
( Re x )
1/ 2
τw 0.664
friction coefficient Cf ≡ =
ρU ( Re x )
1 2 1/ 2

D 1.328
drag coefficient CD ≡ =
ρU L ( Re L )1/ 2

where D is the skin friction drag force on unit width of a plate of length L:
D = ∫ τ w dx .

D. The Boundary Layer Momentum-Integral Equation (Section 9.2.3)

By virtue of the property that the boundary layer thickness δ is much smaller than the
streamwise length scale (say, L): δ / L 1 , one may simplify the Navier-Stokes equations
to obtain the boundary-layer approximation:

 ∂u ∂v
continuity + =0
∂x ∂y

 ∂u ∂u 1 ∂p ∂ 2u
 x -momentum u + v = − + ν
 ∂x ∂y ρ ∂x ∂y 2
 1 ∂p
 y -momentum 0 = −
 ρ ∂y

with the boundary conditions:

(u, v) = (0, 0) at y = 0
u =U, p = P as y → ∞
(where U , P are the velocity and pressure
of the inviscid flow just outside the boundary layer)

From the y-momentum equation, it is clear that the pressure in the boundary layer is
constant laterally across the layer and equal to the near-wall pressure of the inviscid flow
outside the boundary layer.

On integrating the x-momentum equation with respect to y from y = 0 to y = δ, and after

some algebra including the use of the continuity equation, one may obtain the Karman
momentum integral equation

d dU
τw = ρ
(U 2θ ) + ρU δ *

where τ w = wall shear stress

ρ = density
U = near-wall velocity of the outer inviscid flow
u u
θ = momentum thickness = ∫ 1 − dy
U U

δ u
δ * = displacement thickness = ∫  1 − dy
 U
δ = boundary layer thickness

This momentum integral equation is applicable to laminar, transitional or turbulent boundary layer.

In particular, in the absence of pressure gradient (e.g., flow over a flat plate), the free stream velocity
U = constant and dU / dx = 0 , and therefore the momentum integral equation reduces to

τ w = ρU 2

by which the skin friction drag and drag coefficient are simply given by

L L dθ D ρU 2θ L θ
D = ∫ τ w dx = ρU 2 ∫ dx = ρU 2θ L , CD = = =2 L
0 0 dx 1
ρU 2 L 1
ρU L

where θ L is the momentum thickness at x = L .

(1) Laminar Boundary Layer Over a Flat Plate Revisited – approximate solution by
momentum integral equation

It is remarkable that approximate solutions, which are reasonably close to the exact
ones, can be obtained for the boundary layer thickness and drag coefficients from the
momentum integral equation on adopting an assumed velocity profile
= f (η )
where η = is the y -coordinate normalized with respect to the local boundary
δ ( x)
layer thickness.

The steps are as follows:-

a) Find the relation between θ and δ
δ u  u 1
θ =∫  1 − dy = δ ∫ f (1 − f ) dη = aδ , ( a is a numerical constant )
0 U
 U 0

b) Find the wall shear stress from Newton’s law of viscosity
du U ∂u / U
τ w = ρν = ρν
dy y =0 δ ∂y / δ y =0
U df  ρν U 
= ρν = b  ( b is another numerical constant )
δ dη η =0  δ 

c) Substitute θ and τ w into the momentum integral equation

 ρν U  2 dδ
b  = a ρU
 δ  dx
bν dδ 1 dδ 2
⇒ =δ =
aU dx 2 dx

Integrating the above equation with respect to x, assuming that δ = 0 at x = 0:

δ2 = x
δ 2b / a
⇒ =
x ( Re x )1/ 2
τw 2ab
friction coefficient Cf ≡ =
ρU ( Re x )
1 2 1/ 2

D 2 2ab
drag coefficient CD ≡ =
ρU L ( Re L )1/ 2

It turns out that the values of a and b are rather insensitive to the choice of the
approximate velocity profile u / U = f (η ) as long as it is a reasonable one
satisfying the boundary conditions.

Some assumed velocity profiles are

 2η − η 2 parabolic

3 η3
f (η ) =  η − cubic
2 2
  πη 
sin  2  sine
  
which satisfy

f ( 0 ) = 0, (no-slip at y = 0)
f (1) = 1, (u = U at y = δ ) .
f ' (1) = 0 (no stress at y = δ )

(2) Turbulent Boundary Layer Over a Flat Plate (Section 9.2.5)
The one-seven-power law, suggested by Prandtl, is used for the velocity profile in the
turbulent boundary layer with zero pressure gradient:
1/ 7
u  y
=  or f (η ) = η1/ 7
U δ 
by which the momentum thickness is
θ = δ ∫ f (1 − f ) dη = δ ∫ η1/ 7 (1 − η1/ 7 ) dη =
1 1
0 0 72
The one-seven-power law fails to describe the velocity profile at y = 0 , where
∂u / ∂y → ∞ . The following empirical formula obtained for pipe flow can be adopted
1/ 4
 ν 
τ w = 0.0225 ρU  

 Uδ 
Substituting θ and τ w into the momentum integral equation, and integrating with
respect to x:
1/ 4
4 5 / 4  72 × 0.0225   ν 
δ =   x + constant
5  7 U 
It is assumed that the turbulent boundary layer starts from x = 0. (This is a
contradiction to the fact that the boundary layer starts out as a laminar one, but this
assumption has given good results.) Therefore, the constant = 0. Further simplification
δ 0.370 τw 0.0288 0.072
= , = , CD = .
( Re x ) ρU ( Re x ) ( Re L )
1/ 5 2 1/ 5 1/ 5

These results are valid for smooth flat plates with 5 × 105 < Re L < 107 .

Note that for the turbulent boundary layer flow the boundary layer thickness increases
with x as δ ∼ x 4 / 5 and the shear stress decreases as τ w ∼ x −1/ 5 . For laminar flow these
dependencies are x1/ 2 and x −1/ 2 , respectively.

E. Effect of Pressure Gradient (Section 9.2.6)
The pressure in the streamwise direction (i.e., along the body surface) will not be constant
if the body is not a flat plate. Consequently, the free stream velocity at the edge of the
boundary layer U is also not a constant but a function of x. Whether the free-stream flow
is accelerating or decelerating along the body surface will have dramatically different
effects on the development of the boundary layer.

Let us re-examine flow past a circular cylinder, and find out what causes d’Alembert’s

You may recall that inviscid flow past a circular cylinder has a symmetrical pressure
distribution around the surface of the cylinder about the vertical axis. This results in a
zero pressure drag, which is however not true in reality for any fluid with a finite
viscosity. Such discrepancy is now referred to as d’Alembert’s paradox.

Despite the discrepancy, the potential theory helps to reveal that the pressure and hence
the free-stream velocity Ufs on the cylinder’s surface are not constant. From A to C, the
pressure gradient is negative and the flow is accelerating, and from C to F, the opposite is

The real fluid flow past a circular cylinder is like this:

Flow Past A-B-C

• the streamlines are converging, i.e., flow is accelerating, and the free-stream velocity
U reaches a maximum at C.
• the pressure is decreasing along the cylinder surface, i.e., ∂p / ∂x < 0 , net pressure
force is in forward direction, and the pressure gradient is said to be ‘favorable’.
• the accelerating flow tends to offset the ‘slowing down’ effect of the boundary on the
fluid. Therefore, the rate of boundary layer thickening decreases and flow remains

Flow Past C-D

• the streamlines are diverging, and the flow is retarding.
• the pressure is increasing along the cylinder surface, i.e., ∂p / ∂x > 0 , net pressure
force opposes the flow, and the pressure gradient is said to be ‘adverse’ or
• it reduces the energy and forward momentum of the fluid particles in proximity to the
surface, causing the thickness to increase sharply and fluid near the surface be brought
to a standstill ( ∂u / ∂y at the surface is zero) at D. See figure (b).

Flow Past D-E-F
• flow close to the cylinder surface starts to reverse at D (separate point), i.e., fluid no
longer to follow the contour of the surface. The phenomenon is termed separation.
• large irregular eddies formed in the reverse flow (the wake), in which much energy is
lost to heat.
• the pressure in the wake remains approximately the same as at the separation point D,
and is therefore lower than that predicted by the inviscid theory (see figure c). This
lowering of pressure behind the cylinder resulting from flow separation leads to a net
pressure drag on the cylinder. This explain d’Alembert’s paradox. Note that the wider
the wake, the larger the pressure drag, and vice versa.

Influence of the pressure gradient



Influence of a strong pressure gradient on a turbulent flow: (a) flow is relaminarized by a

negative (favorable) pressure gradient; (b) the boundary layer is thickened by a positive
(unfavorable) pressure gradient.

Further remarks about flow separation

• separation can occur only under an adverse pressure gradient and when the fluid is
• separation occurs with both laminar and turbulent
boundary layers. Laminar boundary layer is more prone
to separation than turbulent boundary layer. Thus, as
shown in figure (c) on page 85, the turbulent boundary
layer can flow farther around the cylinder before it
separates than can the laminar boundary layer.
Therefore the wake size will be narrower if the flow is
turbulent at the separation point than if it is laminar.
This explains why it is desirable to have dimples on a
golf ball, which can effectively reduce the drag by
inducing a narrower turbulent wake behind the ball.

Turbulent boundary layers are more resistant

to flow separation than are laminar boundary
layers exposed to the same adverse pressure
gradient. The laminar boundary layer (upper)
cannot negotiate the sharp turn of 20o, and
separates at the corner (flow is from left to
right). The turbulent boundary layer (lower)
on the other hand manages to remain
attached around the sharp corner.

F. Drag (Section 9.3)
Any object moving through a fluid (or a stationary object immersed in a viscous flow)
will experience a drag, D – a net force in the direction of flow due to the pressure and
shear forces on the surface of the object.

Drag = Pressure Drag + Skin Friction Drag


Pressure Drag = resultant force arising from the non-uniform and asymmetrical pressure
distribution around the surface the body. It is also called form drag as it
depends on the form or the shape of the body.

Skin Friction Drag = resultant force due to fluid shear stress on the surface of the object.

D = ∫ p cos θ dA + ∫ τ w sin θ dA

The drag coefficient CD is given by the ratio of the total drag force to the dynamic force
CD = 1
ρU 2 A
where U = relative velocity of fluid far upstream of the object,
A = frontal area – the projected area of the object when viewed from a direction
parallel to the oncoming flow if it is a blunt (or bluff) object (e.g., a cylinder);
or the planform area – the projected area of the object when viewed from
above it if it is a streamlined object (e.g., a flat plate).


Typically the drag coefficient depends on
(i) the shape of the object,
(ii) orientation of the object with the flow (e.g., a flat plate normal to flow has a
different CD than a flat plate parallel to flow),
(iii) the Reynolds number Re = UD / ν where D is a characteristic dimension of the
(iv) surface roughness if the drag is dominated by skin friction and the boundary
layer is turbulent.

Flow Past a Flat Plate

When a flat plate is held normal to flow, the flow is

separated upon past over the plate. A region of
eddying motion (wake) is formed at the rear of the
plate, the pressure there being much reduced.
Therefore the pressure drag is dominant, and the
plate is a bluff body in this position. The drag shows
little dependence on the Reynolds number.
When a flat plate is held parallel to flow, formation of the boundary layer over the plate is
appreciable and flow separation is
negligible. Therefore the skin friction
drag is significant. The plate is a
streamlined body in this position.
The drag coefficient increases when the boundary layer becomes turbulent.

Flow Past a Circular Cylinder/Sphere

• Re ≤ 1
– creeping flow
– no flow separation
– CD decreases with increasing Re ( CD = 64 / Re for a sphere)
(Note that a decrease in the drag coefficient with Re does not necessarily imply a
corresponding decrease in drag. The drag force is proportional to the square of the
velocity, and the increase in velocity at higher Re will usually more than offset the
decrease in the drag coefficient.)
• Re = 10
– separation starts occurring on the rear of the body forming a pair of vortex bubbles
– vortex shedding begins at Re ≅ 90, leading to an oscillating Karman vortex street
wake (see next page)
– region of separation increases with increasing Re
– CD continues to decrease with increasing Re until Re = 103, at which pressure
drag dominates
• 10 < Re < 105

– CD remains relatively constant, which is a characteristic behavior of blunt bodies

– flow in the boundary layer is laminar, but the flow in the separated region is
highly turbulent, thereby a wide turbulent wake
• 10 < Re < 106

– a sudden drop in CD somewhere within this range of Re

– this large reduction in CD is due to the flow in the boundary layer becoming
turbulent, which moves the separation point further on the rear of the body,
reducing the size of the wake and hence the magnitude of the pressure drag. This
is in sharp contrast to streamlined bodies, which experience an increase in the drag
coefficient (mostly due to skin friction drag) when the boundary layer turns

Laminar boundary layer separation with
a turbulent wake for flow past a circular
cylinder at Re = 2000.

(a) (b)

Flow over (a) a smooth sphere at Re = 15,000, and (b) a sphere at Re = 30,000 with a trip
wire; the delay of boundary layer separation is clearly seen by comparing these two
photographs. The delay of separation in turbulent flow is caused by the rapid fluctuations
of the fluid in the transverse direction, which enables the turbulent boundary layer to
travel farther along the surface before separation occurs, resulting in a narrower wake and
a smaller pressure drag. Recall also that turbulent flow has a fuller velocity profile as
compared to the laminar case, and thus it requires a stronger adverse pressure gradient to
overcome the additional momentum close to the wall.

Karman Vortex Streets

The Karman vortex street is one of the best-known vortex patterns in fluid mechanics.
The vortex street is just a special type of unsteady separation over bluff bodies such as a
cylinder. The vortex street is highly periodic having a frequency which is proportional to
U/D, where D is the length of the bluff body measured transverse to the flow and U is the
incoming flow speed. This periodicity is responsible for the "singing" of telephone wires.
In fact, vortex streets are almost always involved when the wind generates a fairly pure
tone as it blows over obstacles.

A practical consequence of the regular,

periodic flow is that the forces on the
body are also periodic. Because the flow
is asymmetric fore and aft as well as in
the direction transverse to the flow, the
body will experience both an oscillating
drag and lift. If the frequency of the
shedding is close to a structural
frequency, resonance can occur, usually
with unpleasant results.

(V) OPEN-CHANNEL FLOW (Chapter 10)
A. Introduction
• Studies of open channel flow are important for design and planning of river control,
inland navigation, surface drainage, irrigation, water supply and urban sanitation.
• An open channel is a conduit in which the liquid flows with a free surface subjected to
atmospheric pressure. It includes river channels (natural water courses), constructed
channels (e.g., canals, flumes) and enclosed conduits (e.g., sewers, culverts) operating
partially full.
• Open channels normally have a very small slope S < 0.01 , and the pressure variation
with depth is nearly hydrostatic. The hydraulic grade lines for all the streamtubes are
the same and coincide with the free surface of the flow. Open channel flow is
essentially caused by slope of the channel and self-weight of the liquid.

B. Types of Flow
1. Uniform and Non-Uniform Flow
Uniform flow – mean velocity does not change (in both magnitude and direction) from
one section to another along the channel. With a free surface this implies
a constant cross section and flow depth, which is called the normal
depth. Hence the liquid surface is parallel to the base of the channel.
Uniform flow results from an exact balance between the gravity and
frictional effects. (NB: uniform channel flow however does not require
uniformity of velocity across any one section of the flow.)

Non-uniform (varied) flow – the mean velocity V and the fluid depth y change with
distance x along the channel. Non-uniform channel flow can be classified
into gradually, or rapidly varying flow when the flow depth changes
slowly ( dy / dx 1) , or rapidly ( dy / dx ∼ 1) with distance along the channel.

2. Steady and Unsteady Flow

The flow is unsteady or steady depending on whether or not the depth at a given location
changes with time. Surface waves propagating on a channel is in fact unsteady, but may
appear steady to an observer who travels at the same speed as the waves.

3. Laminar and Turbulent Flow

The flow is laminar or turbulent depending on the magnitude of the Reynolds number,
which for open channels can be defined as Re ≡ VRh /ν , where Rh is the hydraulic radius

equal to the ratio of the cross section area to the wetted perimeter. Open-channel flows in
rivers, culverts, surface drainage, etc typically involve water as the fluid (with fairly small
viscosity) and have relatively large dimensions so that the flows are invariably turbulent.
For turbulent channel flow, the velocity profile is rather uniform (except close to the
bottom and lateral boundaries) across a section of the flow, and therefore it is a common
practice to use the section-averaged velocity V = V(x) as the primary variable in open
channel flow.

4. Tranquil (Sub-critical) and Rapid (Super-critical) Flow

Using the momentum principle, one may show that a small-amplitude surface wave (which
is, say, caused by some small disturbance to the flow) will travel in a shallow pool of
liquid at the speed c = gy , where y is the local fluid depth. An open channel flow may
have sharply different behaviors when its flow velocity V is smaller or larger than this
wave speed. The ratio of these two velocities is known as the Froude number, the
magnitude of which corresponds to the following types of flow:

⎧< 1 sub-critical flow

V ⎪
Fr ≡ ⎨= 1 critical flow
gy ⎪
⎩> 1 super-critical flow

Briefly speaking, a sub-critical flow is so low in speed (thereby called tranquil) that a
disturbance to the flow may send waves both upstream and downstream of the channel. In
sharp contrast, a super-critical flow is a high speed flow (thereby called rapid) and a wave
cannot be transmitted upstream.

C. Energy Considerations

Energy line
V 2 / 2g

Free surface = hydraulic grade line

p / ρg


Channel bed

For channels under consideration, the bottom slope is very small ( ∼ 0.001). Streamlines are
therefore virtually straight and parallel, and pressure variation is hydrostatic. This implies

that a point with pressure p is at a depth p / ρ g below the free surface, so the sum of the
pressure and elevation heads ( p / ρ g + z ) represents the height of the free surface above the
datum level, or the hydraulic grade line coincides with the free surface.

Referring to the above figure, the energy equation may be applied to the two points (1) and (2)
on the channel bed, giving
p1 V12 p V2
+ + z1 = 2 + 2 + z2 + hL
ρ g 2g ρ g 2g
where hL is the head loss due to frictional effects between the two sections. Clearly,
p1 / ρ g = y1 , p2 / ρ g = y2 , z1 − z2 = So l , and hL = S f l , where S0 and S f are respectively the
slopes of the channel bed and the energy line. S f is also called the energy gradient, as it is
the rate at which energy head is lost to friction. On substituting, the energy equation becomes

y1 − y2 =
− V12 )
+ ( S f − S0 ) l
E1 = E2 + ( S f − S0 ) l

where the specific energy, E, is defined as

E = y+ .

For uniform flow, the flow is invariant with distance along the channel, so y1 = y2 and
V1 = V2 , and the energy equation gives

S f = S0 (for uniform flow)

That is, the energy gradient is exactly equal to the geometrical gradient of the channel when
the flow is uniform.

D. Uniform Flow: Chezy-Manning Equation

Normal depth yn = depth of uniform flow
flow sectional area A
Hydraulic radius Rh = =
wetted perimeter P

There is no change in momentum along the channel when the flow is uniform, and therefore
the net force acting on a control volume, as shown above, must be zero. By uniformity, the
two end forces are equal F1 = F2 . Consequently, the downward gravity force is exactly
balanced by the bottom friction:
τ w Pl = W sin θ = ( ρ gAl ) S0
⇒ τ w = ρ gRh S0
By analogy with pipe flow, the wall stress τ w can be expressed as
τ w = ρV 2
where f is the friction factor, which for complete turbulent flow depends only on the
roughness of channel. Combining these equations, we get the so-called Chezy equation for
uniform flow
V = C Rh S0

where C = 8 g / f is the Chezy coefficient. This coefficient is found to be not a constant,
but depends on the size and shape of the channel section, and on the roughness of the

The simplest and most widely used relation for C is the Manning formula:
Rh1/ 6
where n is the Manning roughness coefficient. On substituting Manning formula into the
Chezy equation, we get the so-called Chezy-Manning equation:

2 / 3 1/ 2
Rh2 / 3 S01/ 2 Rh S f
V= =
n n

for the velocity of a uniform channel flow. It follows that the flow rate is given by
ARh2 / 3 S01/ 2 A5/ 3 S01/ 2
Q = VA = = 2/3 .
n P n

Note that the Manning roughness coefficient is NOT dimensionless. When SI units are used,
it has the following values for different types of channels.

For a channel of rectangular cross section with a width b, the hydraulic radius
yb y
Rh = =
2 y + b 2 y / b +1
In the limiting case of a very wide rectangular channel such that b y , then Rh ≈ y , or
the hydraulic radius is approximately equal to the flow depth. In a very wide channel, the
uniform flow velocity and the discharge per unit width of channel are therefore

2 / 3 1/ 2 5 / 3 1/ 2
yn2 / 3 S01/ 2 yn S f Q yn5/ 3 S01/ 2 yn S f
V= = , q = = Vyn = =
n n b n n

E. Specific Energy: Alternative Depth of Flow

Let us recall the specific energy that has been introduced in Section C. The specific energy,
E, is the sum of the pressure and velocity heads
E = y+ .
Specific energy is also the energy head referred to the base of the channel. It is a very
important concept in open-channel flow. While the specific energy is constant along the
channel when the flow is uniform, it may increase or decrease down the channel when the
flow is non-uniform.

For the convenience of discussion, let us consider only rectangular cross section from here
onward. The results can be extended to an arbitrary cross section, but will not be considered

1. Variation of Specific Energy with Flow Depth

Since the velocity is related to the unit width discharge by V = q / y , we may write the specific
energy as
E = y+
2 gy 2
Here the equation consists of three variables: E, y, q. It would be of interest to examine the
cases: i) q is constant, and E varies with y; and ii) E is constant, and q varies with y.

If q is kept constant, E varies with y in the following manner.

This is known as the specific energy diagram, in which the following points are notable.

• For y → 0, V = → ∞, ∴ E → ∞ (asymptotic to the E axis) .
• For y → ∞, V = → 0, ∴ E → h (asymptotic to the line y = E ) .
• Between these two extremes, E declines to a minimum Emin at the critical point:
o depth and velocity at this point are called the critical depth yc and critical
velocity Vc ;
o critical depth represents the least possible specific energy with which the fixed
discharge q is able to flow in the channel of given slope.
• For each other value of E greater than the minimum, there are two possible values of y,
one greater and one less than yc . The two corresponding depths for a given value
E > Emin are known as the alternative depths.
• Flow can be classified into
y > yc , V < Vc flow is tranquil or sub-critical
y = yc , V = Vc flow is critical
y < yc , V > Vc flow is rapid or super-critical

2. Criterion for Minimum Specific Energy

Since the minimum E occurs when dE/dy = 0, we may readily obtain, using the relation for E
given above, the following condition for critical flow
dE q2
=1− 3 = 0
dy gy
q2 V 2
⇒ Fr 2 ≡ = =1
gy 3 gy

Being consistent with our earlier discussion in Section B.4, the flow is critical, sub-critical or
super-critical depending on the value of the Froude number. Combining all of the above,

⎧< 1 y > yc , V < Vc sub-critical flow, E > Emin

V ⎪
Fr ≡ ⎨= 1 y = yc , V = Vc critical flow, E = Emin
gy ⎪
⎩> 1 y < yc , V > Vc super-critical flow, E > Emin

Also, the critical flow depth yc is given by

1/ 3
q2 ⎛ q2 ⎞
Fr ≡ 3 = 1
⇒ yc = ⎜ ⎟ ,
gyc ⎝ g ⎠
and hence the minimum specific energy is
q2 3
Emin = yc + 2
= yc .
2 gyc 2

It is remarkable that, for a given discharge, the critical flow occurs when the specific energy
is the lowest possible value Emin .

3. Criterion for Maximum Discharge
The equation for the specific energy may be rearranged into
q = 2g y ( E − y )
1/ 2

If E is held constant, q varies with y as follows.

sub-critical flow

yc critical flow

super-critical flow


The discharge per unit width, q, reaches a maximum value qmax at a particular depth. For
each other value of q < qmax , there are two possible values of flow depth y. Again, the
criterion for this maximum q is given by dq / dy = 0 , which after some algebra leads to
= 1,
gy 3

which is identical to the critical flow condition. The corresponding flow depth is therefore
the critical depth yc . Hence, we may make another remark that for a given specific energy E,
the critical flow occurs when the discharge is the largest possible value qmax .

4. Significance of Bed Slope

The bed slope which is required to produce uniform flow in a channel operating at the critical
depth is called the critical slope S0 c . This critical slope can be found by using the Chezy-
Manning equation and the critical flow condition
Rhc2 / 3 S01/c2 gyc n 2
Vc = = gyc ⇒ S0 c = 4 / 3
n Rhc
It can be seen that the critical slope depends on the discharge and the boundary roughness.
If uniform slope occurs in a channel with bed slope less than the critical ( S0 < S0c ) , the flow
must be tranquil (sub-critical), and the slope is said to be mild (or sub-critical).
Likewise, with S0 > S0c , the flow must be rapid (super-critical) and the slope is said to be
steep (or super-critical).

F. Gradually Varied Flow

1. Introduction
• Uniform flow, in which a uniform flow depth is maintained, requires constancy of all
channel characteristics (i.e., cross-sectional shape, bed slope, roughness) throughout the
flow. In natural streams, this condition is hardly attained and flow is invariably non-
uniform in nature. Even for artificial channels with uniform cross section, etc., uniform
flow is only a condition to be approached asymptotically. The surface of a varied (i.e.,
non-uniform) flow is not parallel to the bed and takes the form of a curve.
• Steady varied flow is broadly divided into two kinds:-
o Gradually varied flow, in which changes of depth and velocity take place over a
long distance and degree of non-uniformity is very slight. Boundary friction is
significant and is to be accounted for.
o Rapidly varied flow, in which the sectional area of flow changes abruptly within a
short distance. Turbulent eddying loss is more important than boundary friction in
this case. Hydraulic jump, which is a typical example of rapidly varied flow, will
be examined in the next section.
• Gradually varied flow may result from
o a change in cross-sectional shape, bed slope, boundary roughness of the channel.
o the installation of control structures (e.g., sluice gate, weirs, etc.).

2. General Equation for Gradually Varied Flow

When the flow is non-uniform, the specific energy is no longer a constant, but varies with
distance, x, along the channel. It has been derived in Section C that
E ( x + Δx ) − E ( x ) = ( S0 − S f ) Δx

where S0 and S f are respectively the slopes of the channel bed and the energy grade line.
This relation implies that the specific energy gradient is equal to the difference between these
two slopes.
= S0 − S f
On the other hand, by the definition of the specific energy,
dE d ⎛ V2 ⎞ d ⎛ q2 ⎞ ⎛ q 2 ⎞ dy
= (1 − Fr 2 )
= ⎜y+ ⎟ = ⎜ y + 2 ⎟
= ⎜ 1 − 3 ⎟
dx dx ⎝ 2 g ⎠ dx ⎝ 2 gy ⎠ ⎝ gy ⎠ dx dx
Combining the two equations above, one may get the general equation governing the free
surface gradient in a gradually varied flow:

dy S0 − S f
dx 1 − Fr 2

Note that S f and Fr are functions of x, and the geometrical bed slope S0 may or may not
change with x depending on construction. The equation above may be integrated numerically
to yield the surface profile for any type of gradually varied flow. This kind of profile
evaluation is normally carried out by a commercial package nowadays and will not be
discussed here.

3. Profile Classification
The general surface profile equation can be utilized in establishing the various forms of
varied flow profile. The equation may be rearranged into
dy ⎛ 1 − S f / S0 ⎞
= S0 ⎜ ⎟.
⎝ 1 − Fr ⎠
Let us consider a very wide rectangular channel. As noted earlier in Section D, when the flow
is uniform,
⎛ qn ⎞
S0 = S f = ⎜ 5/ 3 ⎟
⎝ yn ⎠
where yn is the normal depth.
An assumption is made here that for gradually varied flow the energy gradient S f can be
related to the local flow depth y in the same manner as in the above formula for uniform
flow. Therefore, when the flow is gradually varied
2 2
⎛ qn ⎞ ⎛ qn ⎞
S0 = ⎜ 5 / 3 ⎟ , S f = ⎜ 5/3 ⎟ ⇒ S f / S 0 = ( yn / y )
10 / 3
⎝ yn ⎠ ⎝y ⎠
On the other hand, the Froude number can be written as
q2 yc3 ⎛ q2 ⎞
Fr 2 = = ⎜∵ = 1⎟ .
gy 3 y 3 ⎝ gyc

Putting these relations back into the surface profile equation

1 − ( yn / y )
10 / 3
= S0 .
1 − ( yc / y )
With the equation above, we may outline the various forms of varied flow profiles, with a
classification based on the following features.

Type of Bed Slope

Mild Slope (M): S0 < S0 c , yn > yc
Steep Slope (S): S 0 > S0 c , yn < yc
Critical Slope (C): S 0 = S0 c , yn = yc
Horizontal Slope (H): S0 = 0, yn = ∞
Adverse Slope (A): S0 < 0, yn = ∞

Flow Depth ( y ) Relative to Normal Depth ( yn ) and Critical Depth ( yc )

Type 1: y > yn , and y > yc > 0 ⇒ Backwater curve
Type 2: y is between yn and yc < 0 ⇒ Dropdown curve
Type 3: y < yn , and y < yc > 0 ⇒ Backwater curve

Hence, a flow occurring on a mild slope has an M1 curve when its depth is greater than the
normal depth, and has an M3 curve when its depth is less than the critical depth, and so on.

Further notes:-
The critical and normal depth lines (CDL, NDL) and the channel bed form the boundaries of
3 zones. The curves of gradually varied flow surface profile approach each of these zone
boundaries in a specific manner:
a) Upper limits of depth – the curves tend to become asymptotic to a horizontal water line.
b) Normal depth line – approached asymptotically (except for C curves).
c) Critical depth line – intersected at right angles (except for C curves).
d) Bed – intersected at right angles.
There are altogether 12 possible profiles of gradually varied flow, as depicted below.

Channel Depth dy Type of Type of

Symbol Form of Profile
slope Relations dx Profile Flow

y > yn > yc M1
+ Backwater

Mild yn > y > yc M2

- Dropdown

yn > yc > y M3
+ Backwater

y > yc = yn C1
+ Backwater
yc = y = yn
Parallel to Uniform,
Critical C2
bed Critical
yc = yn > y C3
+ Backwater
y > yc > yn S1
+ Backwater

yc > y > yn S2
Steep - Dropdown

yc > yn > y S3
+ Backwater

y > yc H2
- Dropdown
yc > y H3
+ Backwater
y > yc A2
- Dropdown
yc > y A3
+ Backwater

G. Hydraulic Jump

1. Introduction

When a change from rapid (super-critical) to tranquil (sub-critical) flow occurs in open
channel, a hydraulic jump appears, through which the depth increases abruptly in the
direction of flow.

In engineering practice, the hydraulic jump frequently appears downstream from overflow
structures (spillways) or underflow structures (sluice gated) where velocities are high. It may
be used as an effective dissipation of kinetic energy (and thus prevent scour) of channel
bottom) or as a mixing device in water or sewage treatment designs where chemicals are
added to the flow.

2. General Equation of Hydraulic Jump

In spite of the complex appearance of a hydraulic jump with

its turbulence and air entrainment, it may be analyzed by
application of the momentum equation.

Consider flow in a rectangular channel, and apply

momentum equation for the control volume of unit width
between sections 1 and 2:
F1 − F2 = ρ q (V2 − V1 )

Substituting the following relations

1 1
F1 = ρ gy12 , F2 = ρ gy22 (∵ hydrostatic pressure variation with depth ) ,
2 2
q q
V1 = , V2 = ( by continuity ) ,
y1 y2

and after rearranging

q 2 y12 q 2 y22
+ = +
gy1 2 gy2 2
2q 2
⇒ y1 y22 + y12 y2 − =0
⎛y ⎞ ⎛y ⎞ ⎛ q2 ⎞
⇒ ⎜ 2 ⎟ + ⎜ 2 ⎟ − 2Fr12 = 0 ⎜ where Fr = 3 ⎟

⎝ y1 ⎠ ⎝ y1 ⎠ ⎝ gy ⎠
⎛y ⎞ ⎛y ⎞
or ⎜ 1 ⎟ + ⎜ 1 ⎟ − 2Fr2 2 = 0
⎝ y 2 ⎠ ⎝ y2 ⎠
which can be solved for y2 / y1 and y1 / y2 respectively
y2 1
= −1 + 1 + 8Fr12
y1 2
y1 1
= −1 + 1 + 8Fr22
y2 2
These are the general equations for a hydraulic jump.

• The depths y1 and y2 on either side of a jump are called conjugate depths for the
• It can be shown by noting the Froude numbers that the upstream and downstream
flow of a jump is always rapid (super-critical) and tranquil (sub-critical) respectively.
From the equations above
1/ 2
⎡ ( y2 / y1 ) + ( y2 / y1 )2 ⎤
Fr1 = ⎢ ⎥ >1 (∵ y2 > y1 )
⎢⎣ 2 ⎥⎦
1/ 2
⎡ ( y1 / y2 ) + ( y1 / y2 )2 ⎤
Fr2 = ⎢ ⎥ <1 (∵ y1 < y2 )
⎣⎢ 2 ⎦⎥
• A super-critical flow may turn into sub-critical only through an abrupt change in the
form of a hydraulic jump. Conversely, a sub-critical flow may turn into super-critical
through a gradual and smooth transition.

3. Energy Loss in Hydraulic Jump

As an excellent energy dissipator, the energy lost or power dissipated in a hydraulic jump
needs to be quantified.

If the head loss is hL , the balance of energy heads before and after the jump gives

V12 V2
y1 + = y2 + 2 + hL
2g 2g

⇒ hL = y1 − y2 +
(V12 − V22 )

q2 ⎛ 1 1 ⎞
= y1 − y2 + ⎜ 2− 2⎟
2 g ⎝ y1 y2 ⎠
⎛ y y 2 + y12 y2 ⎞ ⎛ 1 1 ⎞ ⎛ 2q 2 ⎞
= y1 − y2 + ⎜ 1 2 ⎟⎜ 2 − 2 ⎟ ⎜∵ y y
1 2
+ y 2
1 2y − = 0⎟
⎝ 4 ⎠⎝ y1 y2 ⎠ ⎝ g ⎠
(y − y )

= 2 1
4 y1 y2
which also equals the difference in specific energy across the jump.
To ensure a positive magnitude, y2 > y1 and not vice versa.
Therefore, a hydraulic jump is irreversible.

The power dissipated per width of channel is P = ρ gqhL .

H. Some Examples of Composite-Flow Profiles
• Downstream of a sluice gate

• A change of bed slope from steep to mild: the hydraulic jump may be formed either
on the steep slope or on the mild slope depending on whether the downstream
conjugate depth y2 is smaller or greater than the normal depth on the mild slope.

• Other types of change of bed slope

• Free fall at the end of a channel

• Flow over a bump (bottom friction is ignored): the free surface over the bump is
depressed or elevated when the flow is sub-critical or super-critical, respectively.

When the flow before the bump is sub-critical (state 1a), the flow depth y2 decreases
(state 2a). Since the decrease in flow depth is always greater than the bump height
(why?), the free surface will be suppressed. But if the flow before the bump is super-
critical (state 1b), the flow depth rises over the bump (state 2b), creating a bigger
bump over the free surface. The situation is reversed if the channel has a depression
in its bed: the flow depth increases if the approach flow is sub-critical and decreases if
it is super-critical.

As the bump height Δzb is increased, point 2 continues shifting to the left on the
specific energy diagram (while point 1 remains unaffected), until finally reaching the
critical point at which the specific energy is the minimum, and the flow over the bump
is critical. This critical height of a bump is given by the difference between the
original and the minimum specific energy levels Δzbc = E1 − Emin . Since the specific
energy has already reached the minimum level at this state, the flow over the bump
will only remain critical even when the bump height is further increased. To
overcome a bump of height greater than Δzbc , the approach flow must adapt itself
(say, either to increase the upstream energy level E1 or to reduce the flow rate so that
Emin is decreased) so that the bump height = E1 − Emin is always satisfied. When this
happens, the flow is said to be choked.

The fact that flow over a sufficiently high obstruction in an open channel is always
critical is the working principle of weirs (broad-crested or sharp-crested), which are
used to measure the volume flow rate in open channels.