74

Viscous Flow

Up to this point in your fluid dynamics education, from Basic Fluid

Mechanics, to Fundamental Aerodynamics, to Gas Dynamics to
this class, you have studied inviscid flows (with the exception of
the first days of BFM). Unfortunately, there are no inviscid flows,
all flows are viscous and most practical flows are turbulent
(unsteady). In this section we turn our attention to viscous flows
and begin the study of boundary layer theory.

We’ll begin by reviewing what you learned at the beginning of

Basic Fluid Mechanics, move quickly through the basic definitions
and variables associated with viscous flows and boundary layers,
and then discuss an airfoil analysis tool that incorporates viscous
effects. The goals of this effort is to complete the drag picture via
computation of the skin friction. To do this let us return to the first
few days of Basic Fluid Mechanics.

Consider the flow between two infinite parallel plates; the bottom
is stationary while the top moves to the right with speed U.

u(y)
y

We were told that viscosity causes the fluid to stick to the plate,
hence its velocity is equal to that of the plate. If the flow has zero
 75

dp
pressure gradient, i.e., = 0 , then a linear velocity profile
dx
emerges. This is called Couette flow. One is usually asked to
calculate the viscous resistance of the fluid on the walls or vice
versa given the knowledge that
du
τ =µ (15.1)
dy
where τ is the shear stress, which has units of pressure, force/area,
and µ is the molecular viscosity, a function of the working fluid
and temperature. The velocity gradient is the inverse of the slope
of the line drawn in the figure.

High Shear Profile

Low Shear Profile

Zero Shear Profile

Reverse Flow Profile

 76

Facts about Couette flow:

1. The shear stress always acts in the direction opposite to the
flow.
2. The bottom wall sees fluid moving in the positive x-
direction.
3. The top wall sees fluid moving in the negative x-direction.
This is at first confusing, but some order is brought about by the
right hand rule.

If you use your right hand and allow the thumb to point in the
direction of the surface normal, the index finger will point in the
direction of positive shear stress.

Consider the plates

U

τ +ve
n̂
n̂
τ +ve

du
Is this consistent with ’s sign?
dy
du
Yes, all along the profile u grows as y increases, so > 0.
dy

Is this consistent with the effect of the fluid on the walls?

Yes, the fluid tends to pull the bottom wall to the right but
slows down or retards the motion of the top wall.
 77

What about from the perspective of the fluid, does the convention
still make sense?
n̂
τ +ve
fluid control volume

τ +ve
nˆ
du
Is this consistent with ’s sign and the direction of the forces on
dy
the fluid?

Yes, the top wall imparts onto the fluid a force in the +x
direction while it imparts a force in the –x direction on the bottom
wall.

Common Error:

dy du
We are used to thinking of slope in terms of , but we have
dx dy
and tend to illustrate it physically, that is with u in the +x-direction
and y in its natural direction. Hence,

u(y)
u(y)

y y
A B

τ A < τ B at the wall

 78

To keep it straight just remember the discrete approximation

du ∆u
≈ and ask yourself the question, “how much does u change
dy ∆y
as y increases?”

The Couette flow discussed during the first classes of Basic Fluid
Mechanics provides some good insight into the effects of viscosity,
however, it is way over simplified in its presentation.
1. τ is not just a property that exists at the walls, it exists
everywhere in the fluid. τ = τ ( y) for this problem.
2. τ is not a scalar or a vector, it is actually a tensor, that is, it
has 9 components and is defined only once a plane surface is
provided in the flow.

In its tensor form τ is associated with 2 direction and given 2

subscripts. The first subscript refers to the normal direction of the
plane defining it and the second refers to the direction in which the
force/area acts. In the Couette flow example, τ = τ yx , a force
defines by a plane whose normal is in the y-direction and which
acts in the x-direction.
 79

Note that
 ∂v ∂u 
τ xy = τ yx = µ  +  (15.2)
 ∂x ∂y 
we had v = 0 in the Couette flow so τ simplified considerably. In
the general case
 ∂w ∂v 
τ yz = τ zy = µ  +  (15.3)
 ∂y ∂z 
∂w ∂w 
τ zx = τ xz = µ  + 
 ∂ z ∂x 
Normal stresses are different from but in the same direction as the
pressure.

 ∂u ∂v ∂w   ∂u 
τ xx = λ  + +  + 2µ   (15.4)
 ∂x ∂y ∂z   ∂x 
r
∇ ⋅V
r  ∂v 
τ yy = λ (∇ ⋅ V ) + 2 µ   (15.5)
 ∂y 
r ∂w
τ zz = λ (∇ ⋅ V ) + 2 µ   (15.6)
 ∂z 
λ is the bulk viscosity which by assumption is
2
λ=− µ (15.7)
3
via Stokes hypothesis.

Newton’s second law can now be applied to a viscous fluid

element by considering the shear stresses in addition to pressure to
derive the governing, Navier-Stokes equation.
 80

Skin Friction Coefficient

As with lift and drag, it is useful to nondimensionalize the shear

stress. We recall that τ has units of pressure, i.e., force/area.
However, unlike the pressure coefficients, which is
nondimensionalized by the dynamic pressure of the freestream,
q∞ , the edge properties are the more important parameters.
Therefore, the skin friction coefficient is the shear stress
normalized by the boundary layer edge dynamic pressure:

τw
Cf = (16.1)
1
ρ e ue2
2

Clearly C f is a function of the location along the surface.

Reynolds Number

Using the idea of edge properties it is possible to define a Reynolds

number:

ρ e ue L
Re L = (16.2)
µe
where L is the length along the surface. This idea can be extended
and used as a way of nondimensionalizing the axial coordinate:

ρ e ue x
Re x = (16.3)
µe
 81

For a linear velocity profile like that in Couette flow with plates
separated by a distance D, we see that

ue
τ w = µe (16.4)
D
The Re x and Re L definitions are useless in this case, since the
plates are of infinite extent. Instead we introduce

ρ e ue D
Re D = (16.5)
µe
Combining Eqs. (16.4) and (16.5) for Couette flow, we see

µ (ue D ) 2µ
Cf = =
1
ρ e ue2 ρ e ue D
2 (16.6)

Cf = 2
ReD
Which is, of course, true only for Couette flow, but demonstrates a
basic idea that

1
Cf ∝ (16.7)
Re
showing that as Reynolds number goes up, the skin friction goes
down.
 82

Boundary Layer Concept

The Couette flow example is one of a fully viscous flow. In that

case, the shear stress is constant throughout the region. This is not
true in the general case of viscous flows, particularly external
flows, and leads to the viscous flow concept: the boundary layer.

Boundary Layer – The region of the flow close to a wall in

which viscous effects are dominant.

The boundary layer essentially divides regions of the flow

dominated by viscous effects from those dominated by inviscid
effects. Unfortunately, the definition of a boundary layer is a little
fuzzy, in that its extent depends upon the effects one is pursuing,
i.e., velocity, mass flow, momentum or temperature. To that end,
several boundary layer thickness definitions exist:

Boundary layer thickness - δ 99 . The normal direction

distance from the surface at which the local velocity achieves 99%
of the edge value.

This definition is clearly arbitrary, there exists a δ 90 , a δ 95 , etc.,

but is useful in defining the extent of the so-called velocity
boundary layer.
 83

Thermal boundary layer thickness -δ T . The distance from

the surface at which the local temperature achieves 99% of the
edge temperature.

This definition is even harder to quantify because the wall

boundary condition is more flexible. That is the temperature can be
higher or lower than the edge conditions, or the wall can be
insulated, i.e., adiabatic. In general, the temperature increases in
the velocity boundary layer because viscous effects convert the
energy in the flow to heat. However, wall heat transfer can still
suck that energy out, allowing for a wide variety of possible
profiles.

The arbitrary nature of the velocity and thermal boundary layer

definitions left researchers searching for more quantitative
measures. A general observation of viscous flows based on
conservation of mass ideas is that the velocity “deficit” caused by
the viscous walls forces incoming mass to exit above the plate, i.e.,
 84

inducing a v-velocity component. This, in effect, creates a

streamline displacement effect that moves the external flow
streamlines a finite distance. This distance is not the same as δ 99 .

Displacement thickness - δ - The normal direction distance

*

from the surface that an otherwise undisturbed streamline would be

displaced because of the effect of viscous walls.

The basic idea comes from control volume theory and utilizes
velocity deficit ideas.

If
y1
A = ∫ ρudy - mass flow between 0 and y1 .
0

y1
B = ∫ ρ e ue dy - mass flow between 0 and y1 if no viscous
0
effects are present
y1
B− A= ∫ (ρ e ue − ρu)dy - mass flow deficit
0
(17.1)

This mass flow deficit is equated to an inviscid mass flow across a

new distance from the wall called the displacement thickness.

B − A = ρ e ueδ * (17.2)

Equating Eqs. (17.1) and (17.2) gives

 85

y1
ρ e ueδ = ∫ ( ρ e ue − ρu)dy
*

 y1
ρu 
δ ≡ ∫ 1 −
*
dy (17.3)
0 ρu 
 e e 

It’s physical interpretation is given in the figures below

Unfortunately, the story does not end here, since the deficit
associated with mass flow is not the same as the deficit associated
with momentum flow. A similar thickness is defined for
momentum by using the idea of the difference in momentum
carried by the edge velocity.

y1
A = ∫ ρu 2 dy - momentum flow between 0 and y1 carried
0
by the actual velocity..
y1
B = ∫ ρuue dy - momentum flow between 0 and y1 carried
0
by a fictitious edge velocity.
 86

y1
B − A = ∫ ρu(ue − u)dy - momentum flow deficit
0

B − A = ρ e ue2θ - inviscid momentum flow

ρu 
y1
u
θ≡∫ 1 − dy (17.4)
0 ρ e ue  ue 
Momentum Thickness

The momentum thickness, θ , is an important parameter for drag

prediction and skin friction because it represents a momentum
deficit, i.e., drag. It is proportional to the skin friction.

Reynolds Numbers

The three new distances lead to three new Reynolds numbers

ρ e ue δ
Re δ ≡ Boundary layer thickness Reynolds number
µe

ρ e ue δ *
Re δ * ≡ Displacement thickness Reynolds number
µe

ρ e ue θ
Re θ ≡ Momentum thickness Reynolds number
µe

These numbers are particularly useful for correlating boundary

layer parameters and the transition to turbulence.
 87

Shape Factor

Another important parameter is the shape factor.

δ*
H≡
θ

It is generally true that δ > δ > θ .

*
 88

Derivation of the Boundary Layer Equations

The governing equations of viscous fluid mechanics are the

Navier-Stokes equations. A greatly simplified set of governing
equations can be derived in the boundary layer region of a viscous
flow. This descends from knowledge of the flow field variables in
this region. Start with the Navier-Stokes equations, written in
nondimensional form for the x-momentum equation:

∂u′ ∂u′ 1 ∂p′ 1 ∂   ∂v′ ∂u′ 

ρ ′u′ + ρ ′v′ =− + µ ′ +  (17.5)
∂x′ ∂y′ γM ∞2 ∂x′ Re ∞ ∂y′   ∂x′ ∂y′ 

This form of the governing equations comes about by utilizing

nondimensionalized variables

ρ u v p µ x y
ρ′ = , u ′ = , v′ = , p ′ = , µ′ = , x′ = , y ′ =
ρ∞ V∞ V∞ p∞ µ∞ c c

Start with the dimensional form

∂u ∂u ∂p ∂   ∂v ∂u 
ρu + ρv = − +  µ  +  (17.6)
∂x ∂y ∂x ∂y   ∂x ∂y 

Substitution of the nondimensional variables gives

∂u ′ ∂u ′  p  ∂p′  µ ∞  ∂   ∂v′ ∂u ′ 
ρ ′u′ + ρ ′v′ = − ∞ 2  +    µ ′ ∂x′ + ∂y′ 
∂x′ ∂y′ ρ V
 ∞ ∞ ∂x ′ ρ V c
 ∞ ∞  ∂y ′   
(17.7)
 89

where
p∞ γp∞ a∞2 1
= = =
ρ ∞V∞2 γρ ∞V∞2 γV∞2 γM ∞2

and
µ∞ 1
=
ρ ∞V∞ c Re ∞

The idea of using nondimensional variables is to compare the size

of specific terms in the equations. In this way, one can determine
which terms to eliminate.

Fundamental assumption: The Boundary Layer is very thin

compared to the body.

If c is the chord length then δ << c .

Variable Variation in BL
u′ 0 →1 ⇒ u′ = O(1)
ρ ′ = O(1)
x′ = O(1)
( c ) = O(δ )
y′ = O δ
 90

The continuity equation then says

∂ρ ′u ′ ∂ρ ′v′
+ =0 (17.8)
∂x′ ∂y′
O(1)O(1) O(1)v′
+ =0
O(1) O(δ )

A B

Therefore for term A to be about the same size as term B, we must

have

v′ = O(δ ) (17.9)

Hence terms in Equation (17.5) become

O(1) O(1) 1 O(1) 1  1   O(δ ) O(1)  

O(1)O(1) + O(1)O(δ ) =− +  O (1) +  
O(1) O(δ ) γM ∞ O(1) Re ∞
2
 O (δ )   O (1) O (δ )  

1
We then assume the Re is very large such that = O (δ 2 ) , so
Re ∞
that

 1   O (1)  
O (1) + O (1) = O (1) + O (δ ) 
2
O (1) O (δ ) + O (δ )  
 O (δ )    
O (1) + O (1) = O (1) + O (δ 2 ) + O (1)

There is then one term in the equation that is much smaller than the
others and can be eliminated, thereby simplifying the equations
greatly.
 91

Upon simplification the equations become

∂u ′ ∂u ′ 1 ∂p′ 1 ∂  ∂u ′ 
ρ ′u ′ + ρ ′v′ =− +  µ′  (17.10)
∂x ′ ∂y ′ 2
′ ′
γM ∞ ∂x Re ∞ ∂y  ∂y′ 

A similar analysis can be performed for the y-momentum equation.

∂v′ ∂v′ 1 ∂p′ 1 ∂   ∂v′ ∂u ′ 

ρ ′u′ + ρ ′v′ =− + µ ′ + 
∂x′ ∂y′ γM ∞2 ∂y′ Re ∞ ∂x′   ∂x′ ∂y′ 

O(δ ) O(δ ) O(1) 1   O(δ ) O(1)  

O(1)O(1) + O(1)O(δ ) = + O(δ 2 ) O (1)  O(1) + O(δ )  
O(1) O(δ ) O(δ ) O(1)   

Taken together we get

1
O (δ ) + O (δ ) = O  + O (δ 3 ) + O (δ )
δ 

Where there is only one dominant term. Which implies to first

approximation that that term is the only one that remains.

∂p
=0 (17.11)
∂y

Which is the y-momentum equation for a boundary layer.

 92

A similar analysis can be done for the energy equation resulting in

the complete set of boundary layer equations:

∂ (ρu ) ∂ (ρv )
+ =0
∂x ∂y
∂u ∂u dp ∂  ∂u 
ρu + ρv = − e + µ 
∂x ∂x dx ∂y  ∂y 
∂p (17.12)
=0
∂y
2
∂h ∂h ∂  ∂T  dp ∂  ∂u 
ρu + ρv =  k  + u e +  
∂x ∂x ∂y  ∂y  dx ∂y  ∂y 

with boundary conditions

wall: y=0 u=0 v=0 T=Tw

edge: y→∞ u→ue T→Te
 93

Similarity Solutions

The development of the boundary layer equations simplifies

considerably the governing equations and in some cases makes
them tractable to analytical solution. One such approach to
defining solutions is to employ a transformation of variables so
that in some sense every profile is similar to another. This type of
solution is called a similarity solution. One place where this leads
to useful solutions is the laminar flat plate boundary layer.

Blasius Laminar Flat Plate Solution

The governing equations are transformed by the introduction of the

similarity parameters η and ξ, where

V∞ u
ξ = x, η=y , f ′(η ) =
υx V∞
 94

Under this transformation the boundary layer equations reduce to:

2 f ′′′ + ff ′′ = 0 (18.1)

A much simpler ODE.

With BCs η=0: f=0 f’=0
η→∞: f’=1

The Blasius boundary layer profile for a laminar flat plate in these
coordinates is then

The solution to Equation (18.1) is done in a numerical fashion and

values of f, f’ and f” are provided in tabulated form (next page).
The text demonstrates how the profiles of f and its derivatives can
be used to determine properties of the boundary layer such as the
skin friction coefficient and the boundary layer thickness. It is
suggested that you review that material so that you will be able to
calculate these parameters.
95
 96

From this tabulated data the following can be determined:

0.664
Cf = Local Skin Friction Coefficient
Re x
1.328
Cf = Total Skin Friction Coefficient for a flat plate of length C
Re c
5.0 x
δ=
Re x Boundary Layer Thickness

1.72 x
δ* = Displacement Thickness
Re x
0.664 x
θ= Momentum Thickness
Re x

Note that the total skin friction for the flat plate is related to the
momentum thickness via
2θ
C f = x =c
c
Indicating that the skin friction drag coefficient is directly
proportional to the value of the momentum thickness at the trailing
edge.
 97

The Blasius solution and its derivatives are shown below.

Experimental laminar flat plate boundary layer data are plotted

below:
 98

Turbulent Flat Plate Solution

Laminar flow is characterized by a smooth, layered appearance

(laminate). Flows generally behave this way when the Reynolds
number is low. As the Reynolds number increases the flow
transitions from a smooth state to one with random perturbations
about some mean. The figure above depicts experiments in which a
flat plate boundary layer transitions from laminar to turbulent flow.

Characteristics of turbulent flow include fluctuations in pressure,

temperature and velocity superimposed about a mean value. Note
the thickening of the boundary layer in this case. These behaviors
manifest themselves as a significant increase in skin friction as the
flow transitions from laminar to turbulent and a reduced slope in
skin friction decay with x, as illustrated in the figures shown next.
 99

Some simple formulae for turbulent flat plate flows are

0.37 x
δ≈ (19.1)
Re1x 5

0.074
Cf ≈ (19.2)
Re1c 5

C f ≈ 0.246 Reθ−0.268 10 −0.678 H (19.3)

C f ≈ 0.058 Reθ− 0.268 (0.93 − 1.95 log10 H )

1.705
(19.4)
 100

The first feature of Eq. (19.1) is the fact that turbulent boundary
layers grow more quickly than laminar:

δ ∝ x4 5 turbulent flows

δ ∝ x1 2 laminar flows

Entire careers have been and are now devoted to the study of
turbulent flows, as such, we can only touch briefly on some
important first topics.

Turbulent Velocity Profiles

Dimensionless turbulent velocity profiles often correlate with the

introduction of several new scaling parameters:

τW
v* = wall friction velocity (19.5)
ρ
u
u+ = turbulent inner-law velocity (19.6)
v*
yv *
y+ = turbulent inner-law wall distance (19.7)
υ
These variables are considered important because they scale the
experimental data nicely.
 101

Linear Sublayer
+
The linear or laminar sublayer is defined as y < 10 we find

u+ = y + (19.8)
 102

Law of the Wall/Logorithmic Region

1
u+ = ln y + + B (19.9)
κ
where κ ≈ 0.41 and B ≈ 5.0 are the generally accepted values of
the constants.

Wake Region

The wake region becomes a much more difficult region to

characterize as it is strongly dependent on external flow conditions
like the pressure gradient. One modification of Eq. (19.9) that can
be used in this region is

1 2Π  y 
u+ = ln y + + B + f  (19.10)
κ κ δ 
where Π is called the Coles Wake Parameter.
 103

Transition to Turbulence

The transition from laminar to turbulent flow is often studied

through the use of stability theory, i.e., the tendency of the flow to
cause a disturbance to grow or decay. However, transition is much
more involved as many events occur in the course of the transition
process.
 104

The stability of boundary layers is often tied to the velocity

profiles. Inflection points in these profiles are often an indication
of instability.

The N crit variable used in XFOIL is a method in the class of

e N methods for predicting instability and the transition to
turbulence. They are notoriously inaccurate and must be used with
caution and more than a little physical insight. The XFOIL manual
describes some potential choices for N crit and a suggestion given
for typical airfoil problems. However, be careful to check the
definitions for everything used in XFOIL as they may be
nonstandard!!!!
 105

Momentum Integral Relation

A compelling set of methods for computing both laminar and

turbulent boundary layers is found from the Momentum Integral
Relation. This relation is found by starting with the boundary layer
equations, multiplying continuity by (u − U e ) and then subtracting
the result from the momentum equation, this leaves:

1 ∂τ ∂ ∂
− = (ue − u) + (uue − u 2 )
ρ ∂y ∂t ∂x
(19.11)
du ∂
+ (ue − u) e + (vue − vu)
dx ∂y

Eq. (19.11) is then integrated from the wall to infinity.

τw ∂ ∞ ∂ ∞
= ∫ (ue − u)dy + ∫ (uue − u 2 )dy
ρ ∂t 0 ∂x 0
(19.12)
∞
du
+ e
dx
∫(ue − u) dy − v wue
0

This is called the Karman Integral Relation, which can be

rewritten:

τw Cf 1 ∂ dθ * 1 due vw
= = (u δ *
) + + (2θ + δ ) −
ρue2 2 ue2 ∂t
e
dx ue dx ue

(19.13)
 106

for steady flow with an impermeable wall

Cf dθ θ due
= + (2 + H ) (19.14)
2 dx ue dx

which states that given a δ and θ distribution, C f can be

*

determined. It is made particularly useful by finding a suitable

velocity profile and/or correlation with experimental data.

Thwaites Method

Thwaites rewrote the momentum integral relation using a new

parameter
θ 2 ue′  θ 
2

λ= =  Λ (19.15)
υ δ 
ueθ
The momentum integral relation when multiplied by
υ
becomes:

τ wθ ueθ dθ θ 2 ue′
= + (2 + H ) (19.16)
µue υ dx υ

Thwaites idea was to group these terms and seek appropriate

correlations with experimental data.

τ wθ
≈ S (λ ) shear correlation
µue

δ*
H= ≈ H (λ ) shape-factor correlation
θ
 107

Equation (19.16) then becomes

d λ
ue   ≈ 2[S ( λ ) − λ ( 2 + H )] = F ( λ ) (19.17)
dx  ue′ 

The data all fall along a single line which is astounding

Thwaites proposed

F ( λ ) ≈ 0.45 − 6.0λ (19.18)

Given this form the momentum integral relation has a closed form
solution:

0.45υ x 5
θ ≈ 6 ∫ ue dx
2
(19.19)
ue 0
 108

µue
τw = S (λ )
θ (19.20)
δ * = θH ( λ )

Thwaites further suggests

S ( λ ) ≈ (λ + 0.09 )
0.62
(19.21)

and

H ( λ ) ≈ 2.0 + 4.14 z − 83.5z 2 + 854 z 3 − 3337 z 4 + 4576 z 5

(19.22)

where z = (0.25 − λ ) .
109
 110

Thwaites is applied once ue ( x ) is known. The solution process is:

1. Compute θ from Eq. (19.19)

2. Compute λ from Eq. (19.15)
3. Compute S ( λ ) from Eq. (19.21)
4. Compute τ w from Eq. (19.20a)
5. Compute H ( λ ) from Eq. (19.22)
Compute δ from Eq. (19.20b)
*
6.

Separation can be predicted by using τ w = 0 which implies

S ( λ ) = 0 ⇒ λ = −0.09