2010 08 14 NotesOnFluidMechanicsAndGasDynamics Wassgren PDF

Notes on Fluid Mechanics and Gas Dynamics
Carl Wassgren, Ph.D.

School of Mechanical Engineering
Purdue University
wassgren@purdue.edu
16 Aug 2010
Chapter 01:
Chapter 02:
Chapter 03:
Chapter 04:
Chapter 05:
Chapter 06:
Chapter 07:
Chapter 08:
Chapter 09:
Chapter 10:
Chapter 11:
Chapter 12:
The Basics
Fluid Statics
The Integral Approach
The Differential Approach
Potential Flows
Dimensional Analysis
Navier-Stokes Solutions
Turbulence
Boundary Layers
Pipe Flows
Fluid Machinery
Gas Dynamics
Chapter 01:
The Basics
1.
2.
3.
4.
5.
6.
7.
8.
Symbolic vs. Numeric Approach

Dimensions and Units
Taylor Series Expansion Approximation
Experimental Uncertainty
Statistical vs. Continuum Approach
Fluid Properties
Flow Visualization
Some Basic Definitions and Concepts
C. Wassgren
Chapter 01: The Basics
Last Updated: 05 Sep 2008
1.
Symbolic vs. Numeric Calculations
Consider the following example. You need to determine the trajectory of a projectile fired from a cannon.
The projectile has a mass of 10 kg and the cannon is tilted at an angle of 30 from the horizontal. The
initial velocity of the projectile from the cannon is 100 m/s. Determine:
1. the distance the projectile will travel and
2. how long the projectile is in flight.
We can approach this problem a couple of different ways. The first is to start with the given numbers and
immediately begin the calculations. The second approach is to solve the problem symbolically and then
substitute the numbers at the end.
Numerical Solution:
Draw a free body diagram (FBD) of the projectile.
y
100 m/s
30
10 kg*9.81 m/s2=
98.1 N
Assume this height is negligible.
Use Newtons 2nd law to determine the acceleration of the projectile:

x = 0 m/s 2
Fx = mx 0 N = (10 kg ) x
= my
(10 kg ) ( 9.81 m/s 2 ) = (10 kg ) y

y = 9.81 m/s 2
Integrate with respect to time to determine the projectiles velocity and position given the projectiles initial
x and y velocities and positions:
x = x0 = (100 m/s ) ( cos 30 ) = 86.6 m/s
y = ( 9.81 m/s 2 ) t + y 0 = ( 9.81 m/s 2 ) t + (100 m/s ) ( sin 30 ) = ( 9.81 m/s 2 ) t + 50 m/s
x = ( 86.6 m/s ) t
(1.1)
y = ( 4.91 m/s ) t + ( 50 m/s ) t

2
(1.2)
The projectile will hit the ground when y = 0 m so that by rearranging Eqn. (1.2) we find that the time aloft
is:
50.0 m/s
t=
= 10.2 s
(1.3)
4.91 m/s 2
Substituting into Eqn. (1.1) gives the distance traveled as
x = ( 86.6 m/s )(10.2 s ) = 883 m
(1.4)
As you can see, weve made a number of calculations along the way to finding the answers. Now lets
address some additional questions based on these answers. How does the maximum time aloft depend on
the mass of the projectile? If the initial speed from the cannon doubles, how is the range affected? What
angle maximizes the distance the projectile travels? The answers to these questions are not obvious from
Eqns. (1.1) (1.4); we would need to perform additional calculations. Also, consider how many
calculations would need to be made if we had to determine the range and time aloft for a variety of cannon
angles, initial velocities, and cannon ball masses.
C. Wassgren
Now lets try working the same problem using symbols rather than numbers. Well plug in the numbers at
the very end of the problem.
Symbolic Solution:
Draw the FBD as before:
y
V
mg
x
Assume this height is negligible.
Follow the same approach as before but now with symbols rather than with numbers.
Fx = mx 0 = mx x = 0
= my
mg = my
y = g
x = x0 = V cos
y = gt + y0 = gt + V sin
x = (V cos ) t
y=
(x0 = 0)
(1.5)
gt + (V sin ) t (y0 = 0)
(1.6)
The time aloft is found by setting y = 0:

2V sin
t=
g
and the distance traveled is:
2
2V 2 cos sin V sin ( 2 )
=
x=
g
g
(1.7)
(1.8)
We can now plug in the given numbers to get our numerical answers:
2 (100 m/s ) sin ( 30 )
t=
= 10.2 s
9.81 m/s 2
Same answers as before!
2
(100 m/s ) sin ( 60 )
x=
= 883 m
9.81 m/s 2
Using these results for t and x we can easily calculate the time aloft and distance traveled for a variety of
values of , V, and m. Note that nowhere in Eqns. (1.5) (1.8) does the mass appear so we conclude that
the mass of the cannon ball is unimportant to our calculations. We also observe that if we double the initial
velocity, the time aloft will double and the distance traveled will quadruple. This information is easily lost
in our calculations where numbers were used right away (refer to Eqns. (1.3) and (1.4)).
C. Wassgren
Lastly, if we wanted to determine the angle that will maximum the distance traveled for a given velocity,
we observe from Eqn. (1.8) that we want sin(2) to be as large as possible. Thus, we should tilt our cannon
at an angle of = 45. Substituting this back into Eqns. (1.7) and (1.8) gives:
2V
tmax =
g
xmax =
V2
g
We can also easily double-check the dimensions of the equations and verify that they are dimensionally
homogeneous.
L
[t ] = L T = T
T2
OK!
2
L
[ x] = L T = L
T2
where L and T represent length and time, respectively.
( )
We can conclude from this exercise the following:

1. More information is contained in our solutions when using the symbolic approach than when
using the numeric approach.
2. If several calculations must be made using different values of the parameters, solving the problem
first symbolically rather than starting the problem immediately with the numbers can save
considerably on the number of computations required. Furthermore, its much easier to correct
numerical mistakes at the end of the problem rather than at the beginning or in the middle of the
problem.
Youre almost always better off working out a problem using symbols than with numbers!
Be Sure To:
1. Work out problems symbolically and wait to substitute numerical values until the final relation has
been derived.
2. Try to physically interpret your equations.
3. Make sure any relations you derive and the numbers you calculate are physically reasonable.
4. Double check that the dimensions (or units) of your answers are correct.
C. Wassgren
2.
Dimensions and Units
A dimension is a qualitative description of the physical nature of some quantity.

Notes:
1.
2.
3.
4.
A basic or primary dimension is one that is not formed from a combination of other dimensions. It is
an independent quantity.
A secondary dimension is one that is formed by combining primary dimensions.
Common dimensions include:
[M] = mass
[L] = length
[T] = time
[] = temperature
[F] = force
If [M], [L], and [T] are primary dimensions, then [F = ML/T2] is a secondary dimension. If [F], [L],
and [T] are primary dimensions, then [M = FT2/L] is a secondary dimension.
A unit is a quantitative description of a dimension. A unit gives size to a dimension.

Common systems of units include:
dimension
[L], length
[T], time
[], temperature
[M], mass
[F = ML/T2], force
SI
(Systme International d Units)
m
s
K
kg
N
BG
(British Gravitational)
ft
s
R
slug
lbf
EE
(English Engineering)
ft
s
R
lbm
lbf
Notes:
1. 1 N = 1 kgm/s2 ; 1 lbf = 1 slugft/s2 ; 1 lbf = 32.2 lbmft/s2
2. The kilogram-force (kgf) is (unfortunately) a commonly used quantity. The conversion between kgf
and Newtons is: 1 kgf = 9.81 N.
3. A helpful reference for unit conversions and constant values is:
Avallone, E.A. and Baumeister, T., Marks Standard Handbook for Mechanical Engineers,
McGraw-Hill.
Dimensional homogeneity is the concept whereby only quantities with similar dimensions can be added (or
subtracted). It is essentially the concept of You cant add apples and oranges.
For example, consider the following equation:
10 kg + 16 C = 26 m/s
This equation doesnt make sense since it is not dimensionally homogeneous. How can one add mass to
temperature and get velocity?!?
Note that dimensional homogeneity is a necessary, but not sufficient, condition for an equation to be
correct. In other words, an equation must be dimensionally homogeneous to be correct, but a
dimensionally homogeneous equation isnt always correct. For example,
10 kg + 10 kg = 25 kg
The equation has the right dimensions but the wrong answer!
Be Sure To:
1. Verify that equations are dimensionally homogeneous.
2. Carefully evaluate unit conversions. (A unit conversion error caused the loss of the $125M Mars
Climate Observer spacecraft in 1999!)
C. Wassgren
3.
Taylor Series Expansion Approximation
If we know the value of some quantity, y, at some location, x, then how can we determine the value of y at a
nearby location x+dx?
y
y(x+dx)
y(x)
x+dx
x
SOLUTION:
We can use a Taylor series expansion for y about location x:
y ( x + x) = y ( x) +
dy
dx
( x ) +
x
d2y
dx 2
( x )
x
2!
+
(1.9)
As x becomes very small, (x)(dx), and the higher order terms become negligibly small: (x) >> (x)2
>> (x)3:
dy
y ( x + dx ) = y ( x ) +
(1.10)
( dx )
dx x
Note that Eqn. (1.10) is simply the equation of a line.
We can see whats happening more clearly if we consider the following plot:
y
y(x+dx) y(x) + dy/dx|x(dx)
y(x+dx)
dy/dx|x
y(x)
x+dx
Well use this approximation often, especially when examining how quantities vary over small distances.
Be Sure To:
1. Make sure you understand how this procedure works. It will be used frequently in the remainder of the
notes.
C. Wassgren
4.
Experimental Uncertainty
In any experimental (or even computational) study, attention must be paid to the uncertainties involved in
making measurements. Including the uncertainty allows one to judge the validity or accuracy of the
measurements. Uncertainty analysis can also be useful when designing an experiment so that the
propagation of uncertainties can be minimized (this will be covered in an example).
Consider a measurement of a flow rate through a pipe. Lets say that one measures a flow rate of 1.6 kg/s.
Now consider a theoretical calculation that predicts a flow rate of 1.82 kg/s. Are the theory and
measurement inconsistent? The answer depends upon the uncertainty in the measurement. If the
experimental uncertainty is 0.3 kg/s, then the true measured value could very well be equal to the
theoretical value. However, if the experimental uncertainty is 0.1 kg/s, then the two results are very likely
to be inconsistent.
There are two parts to uncertainty analysis. These include:
1. estimating the uncertainty associated with a measurement and
2. analyzing the propagation of uncertainty in subsequent analyses.
Both of these parts will be reviewed in the following sections. There are many texts (such as Holman, J.P.,
Experimental Methods for Engineers, McGraw-Hill) that can be referred to for additional information
concerning experimental uncertainty.
Estimation of Uncertainty
There are three common types of error. These include blunders, systematic (or fixed) errors, and random
errors.
1.
Blunders are errors caused by mistakes occurring due to inattention or an incorrectly configured
experimental apparatus.
Examples:
Blatant blunder: An experimenter looks at the wrong gauge or misreads a scale and, as a result,
records the wrong quantity.
Less blatant blunder: A measurement device has the wrong resolution (spatial or temporal) to
measure the parameter of interest. For example, an experimenter who uses a manometer to
measure the pressure fluctuations occurring in an automobile piston cylinder will not be able to
capture the rapid changes in pressure due to the manometers slow response time.
Subtle blunder: A measurement might affect the phenomenon that is being measured. For
example, an experimenter using an ordinary thermometer to make a very precise measurement of a
hot cavitys temperature might inadvertently affect the measurement by conducting heat out of the
cavity through the thermometers stem.
2.
Systematic (or fixed) errors occur when repeated measurements are in error by the same amount.
These errors can be removed via calibration or correction.
Example:
The error in length caused by a blunt ruler. This error could be corrected by calibrating the ruler
against a known length.
C. Wassgren
3.
Random errors occur due to unknown factors. These errors are not correctable in general.
Blunders and systematic errors can be avoided or corrected. It is the random errors that we must account
for in uncertainty analyses. How we quantify random errors depends on whether we conduct a single
experiment or multiple experiments. Each case is examined in the following sections.
1.
Single Sample Experiments

A single sample experiment is one in which a measurement is made only once. This approach is
common when the cost or duration of an experiment makes it prohibitive to perform multiple
experiments.
The measure of uncertainty in a single sample experiment is 1/2 the smallest scale division (or least
count) of the measurement device. For example, given a thermometer where the smallest discernable
scale division is 1 C, the uncertainty in a temperature measurement will be 0.5 C. If your eyesight
is poor and you can only see 5 C divisions, then the uncertainty will be 2.5 C. One should use an
uncertainty within which they are 95% certain that the result lies.
Example:
The least count for the ruler to

the left is 1 mm. Hence, the
uncertainty in the length
measurement will be 0.5 mm.
Example:
You use a manual electronic stop watch to measure the speed of a person running the 100 m dash. The
stop watch gives the elapsed time to 1/1000th of a second. What is the least count for the
measurement?
SOLUTION:
Although the stop watch has a precision of 1/1000th of a second, you cannot respond quickly enough to
make this the limiting uncertainty. Most people have a reaction time of 1/10th of a second. (Test
yourself by having a friend drop a ruler between your fingers. You can determine your reaction time
by where you catch the ruler.) Hence, to be 95% certain of your time measurement, you should use an
uncertainty of 1/2(0.1 sec) = 0.05 sec.
Be Sure To:
1. Always indicate the uncertainty of any experimental measurement.
2. Carefully design your experiments to minimize sources of error.
3. Carefully evaluate your least count. The least count is not always 1/2 of the smallest scale division.
C. Wassgren
2.
Multiple Sample Experiment

A multiple sample experiment is one where many different trials are conducted in which the same
measurement is made.
Example:
- making temperature measurements in many identical hot cavities (shown below) or making
temperature measurements in the same cavity many different times
many identical cavities and thermometers

We can use statistics to estimate the random error associated with a multiple sample experiment. For
truly random errors, the distribution of errors will follow a Gaussian (or normal) distribution which has
the following qualitative histogram:
# of measurements
with a particular
measurement value
measurement value
(e.g. temperature)
To quantify the set of measurement data, we commonly use the mean of the data set and its standard
deviation or variance. For example, consider N measurements of some parameter x: x1, x2, , xN.
sample mean, x (a type of average)
1 N
x = xn
N n =1
(1.11)
sample standard deviation, (a measure of how precise the measurements are: as , the
precision )
=
( xn x )
N 1 n =1
C. Wassgren
(Note: The variance is 2.)
(1.12)
Notes:
1. It is not possible to comprehensively discuss statistical analyses of data within the scope of these
notes. The reader is encouraged to look through an introductory text on statistics for additional
information (see, for example, Vardeman, S.B., Statistics for Engineering Problem Solving, PWS
Publishing, Boston).
2. The square of the standard deviation (2) is known as the variance.
3. The coefficient of variation, CoV or CV (also rsd = relative standard deviation), is defined as the
ratio of the standard deviation to the mean, i.e. CoV x . A small CoV means that the scatter
in your measurements is small compared to the mean.
4. For random data (a Gaussian or normal distribution) and a very large number of measurements:
68%
x 1
95% of measurements fall between x 2
x 3
99%
used in most engineering

situations as a measure of
the uncertainty
# of measurements
with a particular
measurement value
measurement value
(e.g. temperature)
x
1
2
3
4.
If the number of measurements is not very large (N < 30 for example), it is better to use the
Students t-distribution for estimating the uncertainty (refer to an introductory text on statistics
such as Vardeman, S.B., Statistics for Engineering Problem Solving, PWS Publishing, Boston):
x t
(1.13)
where t is a factor related to the degree of confidence desired (again, a 95% uncertainty is typically
desired in engineering applications), is the standard deviation given in Eqn. (1.12), and n is the
number of measurements made. The following table gives the value of t for various values of N
and a 95% confidence level. Note that as N the t factor approaches the large sample size
value (1.96).
N
t95%
2
12.71
3
4.30
4
3.18
5
2.78
6
2.57
7
2.45
8
2.36
9
2.31
10
2.26
15
2.14
20
2.09
30
2.04
1.96
5. Often one presents data in terms of its true (rather than sample) mean, , and a confidence interval.
For random data, the true mean lies within the interval
t
=x
(1.14)
N
where x and are the sample mean (Eqn. (1.11)) and standard deviation (Eqn. (1.12)),
respectively, t is the confidence interval factor (found from the Student t-distribution as in the
table above), and N is the number of data.
Be Sure To:
1. Use a t-distribution for evaluating the uncertainty when you do not have sufficient data for using the
standard deviation as a meaningful measure of the uncertainty.
C. Wassgren
10
Example:
Eight measurements are made of a fluids density and are presented in the following table. Determine the
interval in which the mean density lies with a confidence level of 95%.
Measurement #
1
2
3
4
5
6
7
8
density [g/cc]
1.20
1.24
1.25
1.22
1.17
1.19
1.21
1.20
SOLUTION:
The sample mean and standard deviation are:
1
x = (1.20 + 1.24 + 1.25 + 1.22 + 1.17 + 1.19 + 1.21 + 1.20 ) g/cc = 1.21 g/cc
8
2
2
2
2
1 (1.20 1.21) + (1.24 1.21) + (1.25 1.21) + (1.22 1.21)
2
( g/cc ) = 0.03 g/cc

=
8 1 + (1.17 1.21)2 + (1.20 1.19 )2 + (1.20 1.21)2 + (1.20 1.20 )2
The number of samples is N = 8 and the confidence interval is 95%. Thus, the t factor should be 2.36 as
given on the table from the previous page. The true mean then lies within the interval (to within 95%
confidence):
( 2.36 )( 0.03)
t
= 1.21
g/cc (95% CI)
=x
N
8
= 1.21 0.02 g/cc (95% CI)
C. Wassgren
11
Propagation of Uncertainty
Let R be a result that depends on several measurements: x1, , xN, or, in mathematical terms:
R = R ( x1 , , xN )
For example, the volume of a cylinder is:
V = r 2h
V = V ( r, h )
How do we determine the uncertainty in the result R due to the uncertainties in the measurements x1, ,
xN? In the example above, what is the uncertainty in the volume V given the uncertainties in the radius, r,
and height, h?
Consider how a small variation in parameter, xn, call it xn, causes a variation in R, call this variation Rxn:
Rxn = R ( x1, , xn + xn , xN ) R ( x1, , xn , xN )
Rx =
R ( x1, , xn + xn , xN ) R ( x1, , xn , xN )
xn
Rx
n
R
xn

xn
xn
(Note that an = is only strictly true as xn dxn.)
uncertainty
partial derivative in measurement xn
of R w/r/t xn
uncertainty
in R due to
uncertainty in xn
The total uncertainty in R, R, due to uncertainties in all measurements x1,,xN, assuming that the xn are
independent so that the variations in one parameter do not affect the variations in the others, is estimated as:
1
1
2
2
N R

(1.15)
= x xn
n =1 n

The relative uncertainty in R, uR, is given by:
R
uR =
(1.16)
R
2
For example, the uncertainty in the cylinder volume, V=r h, due to uncertainties in the radius, r, and
height, h, is:
N
R = Rxn
n =1
2
2
V
V

V =
r +
h
h

r
2
2
2
= ( 2 rh r ) + ( r 2 h )
and the relative uncertainty is:

1
2
2
V
1
2
uV =
= 2 ( 2 rh r ) + ( r 2 h )
V
r h
r 2 h 2
= 2 +
r h
2
2
= ( 2ur ) + ( uh )
Note:
1. Use absolute quantities when calculating the uncertainty. For example, use R or K as opposed to F
or C for temperature, and use absolute pressures rather than gage pressures.
2. In an uncertainty analysis the uncertainty of some quantities may be so small compared to the
uncertainties in the remaining quantities that they can be considered exactly known. This is
generally the case for well characterized constants and material parameters, e.g. the acceleration due to
gravity.
C. Wassgren
12
Be Sure To:
1. Use absolute quantities when evaluating uncertainties, e.g. absolute temperature and pressure.
2. Review your uncertainty analyses to determine which measurements result in the greatest error in a
derived quantity. Design your experiments to reduce these uncertainties.
C. Wassgren
13
Example:
A resistor has a nominal stated value of 100.1 . A voltage difference occurs across the resister and the
power dissipation is to be calculated in two different ways:
a. from P=E2/R
b. from P=EI
In (a) only a voltage measurement will be made while both current and voltage will be measured in (b).
Calculate the uncertainty in the power for each case when the measured values of E and I are:
E = 1001 V (for both cases)
I = 100.1 A
R
E
I
SOLUTION:
Perform an uncertainty analysis using the first formula for power.
P=E
(1.17)
R
The relative uncertainty in P is:
1
2
uP = uP2 , E + uP2 , R
where
1 P
R 2E
E
E = 2
uP , E =
= 2u E
E = 2
P E
R
E
E
(1.18)
(1.19)
1 P
R E 2
R
R = 2 2 R =
= u R
P R
R
E R
Substitute into Eqn. (1.18).
uP , R =
(1.20)
2
uP = 4uE2 + uR2
The relative uncertainties in the voltage and resistance are:
E
1V
uE =
=
= 1%
E 100 V
R 0.1
uR =
=
= 1%
R
10
uP = 2.24%
C. Wassgren
14
(1.21)
(1.22)
(1.23)
Now perform an uncertainty analysis using the second relation for power.
P = EI
The relative uncertainty in P is:
(1.24)
2
uP = uP2 , E + uP2 , I
where
1 P
1
E
E =
uP , E =
( I ) E = = uE
P E
EI
E
1 P
1
I
R =
uP , I =
( E ) I = = uI
P R
EI
I
(1.25)
(1.26)
2
uP = uE2 + uI2
The relative uncertainties in the voltage and resistance are:
E
1V
uE =
=
= 1%
E 100 V
I 0.1 A
uI =
=
= 1%
I
10 A
uP = 1.41%
(1.27)
(1.28)
(1.29)
We observe that using the second relation (P = EI) gives a smaller uncertainty for the given values.
C. Wassgren
15
Example:
A certain obstruction-type flowmeter is used to measure the flow of air at low velocities. The relation
describing the flow rate is:
1/ 2
2p
m = CA 1 ( p1 p 2 )
RT1
where C is an empirical discharge coefficient, A is the flow area, p1 and p2 are the upstream and
downstream pressures, T1 is the upstream temperature, and R is the gas constant for air.
Calculate the relative uncertainty in the mass flow rate for the following conditions:
C = 0.920.005 (from calibration data)
p1 = 250.5 psia
T1 = 5302 R
p = p1-p2 = 1.40.005 psia
A = 1.00.001 in2
What factors contribute the most to the uncertainty in the mass flow rate?
SOLUTION:
The relative uncertainty in the mass flow rate is given by:
um = um2 ,C + um2 , A + um2 , p1 + um2 ,T1 + um2 ,p

where
1 m
C
C =
um ,C =
= uC
m C
C
1 m
A
A=
um , A =
= uA
m A
A
1 m
1 p1 1
p1 =
um , p1 =
= 2 u p1
m p1
2 p1
um ,T1 =
(1.30)
(1.31)
(1.32)
(1.33)
1 m
1 T1
T1 =
= 12 uT1
m T1
2 T1
(1.34)
1 m
1 ( p ) 1
= 2 up
(1.35)
( p ) =
m p
2 p
Note that the there is negligible uncertainty in the gas constant R since it is presumed to be known to a high
degree of accuracy.
um ,p =
C. Wassgren
16
um = uC2 + u A2 + 14 u 2p1 + 14 uT21 + 14 u2p

where the relative uncertainties are:
C 0.005
uC =
=
= 0.54%
C
0.92
(1.36)
(1.37)
A 0.001 in 2
=
= 0.10%
A
1.0 in 2
p 0.5 psia
u p1 = 1 =
= 2.0%
p1
25 psia
uA =
uT1 =
T1
T1
2 R
530 R
(1.38)
(1.39)
= 0.38%
(1.40)
( p ) 0.005 psia
=
= 0.36%
1.4 psia
p
um = 1.2%
up =
(1.41)
Examine the contributions of each term on the right hand side of Eqn. (1.36) to determine which
uncertainty has the greatest influence on the uncertainty in m .
(
) = 2.9 *10
u = (1.0 *10 ) = 1.0 *10
u = ( 2.0 *10 ) = 1.0 *10
u = ( 3.8*10 ) = 3.6 *10
u = ( 3.6 *10 ) = 3.2 *10
2
3 2
uC2 = 5.4 *103

2
A
2
p1
1
4
1 2
4 T1
1
4
1
4
1
4
2
p
1
4
2 2
3 2
3 2
The uncertainty in the p1 measurement contributes the most to the uncertainty in m .
C. Wassgren
17
5.
Statistical vs. Continuum Approach
Most substances consist of a collection of molecules or atoms. In studying how substances behave, we can
either explicitly account for the molecular behavior, referred to as the statistical approach, or instead treat
the substance as being continuous, referred to as the continuum approach.
Statistical Approach
In the statistical approach we treat the substance of interest as a collection of individual objects (e.g.
molecules). The interactions between molecules are modeled (e.g. using Newtons laws) and the
macroscopic behavior of the substance is determined by utilizing probability and statistics. This
approach is useful in understanding the behavior of material properties (e.g. fluid viscosity) but it is not
very practical for modeling typical engineering applications. Statistical Mechanics and Kinetic Theory
utilize the statistical approach.
Continuum Approach
The continuum approach ignores individual molecules and instead treats the substance of interest as
being continuously distributed in space. The macroscopic behavior of the substance is modeled using
basic conservation laws (e.g. mass, momentum, energy). The continuum method requires that the
smallest length scale of interest, referred to as the macroscopic length scale (e.g. the length scale over
which significant changes in properties occur), be much larger than the microscopic length scale,
typically the mean free path of a molecule (for gases). The mean free path, , of a molecule is the
average distance a molecule travels before colliding with another molecule. By requiring that the
(macroscopic length scale) >> (microscopic length scale), there will be enough molecules at a point
(the smallest macroscopic length scale of interest) so that meaningful averages of the molecular
behavior can be made at that point.
Consider the following experiment. Lets measure the density, , of a fluid in a small cube of length, L.
The local density of the fluid is defined as the total mass of molecules in the cube, mi, divided by the
cubes volume, L3, as the volume approaches the smallest macroscopic length scale of interest, : The
following WWW page demonstrates this experiment:
http://widget.ecn.purdue.edu/~meapplet
mi
lim 3
L
L
mi/L3
statistical
fluctuations
local value
of the density
spatial
fluctuations
L
At large length scales, L, the density will not have very good spatial resolution and, hence, may not be
a very useful quantity. At very small length scales the number of molecules within the box will vary
significantly with time since molecules continuously enter and exit the box. Since the small box can
contain only a few molecules to begin with, the density fluctuations will be very large. There is an
intermediate region between the previously discussed extremes that will have few fluctuations but
good spatial resolution.
C. Wassgren
18
Notes:
1.
In most engineering applications the length scale condition is easily satisfied. The mean free path of a
gas can be estimated using the following simple analysis. Consider gas molecules with an effective
cross-sectional area of A and a mean (molecular) speed of c . The volume swept out by a molecule per
A
unit time is:

V = cA
(1.42)
c
The expected number of collisions with other molecules per unit time is:
n = V
(1.43)
where is the number of molecules per unit volume. The average time between collisions, t, is the
inverse of the collision rate:
t = 1 = 1 = 1
(1.44)
n
( V ) ( cA)
Thus, the typical distance between collisions, aka the mean free path , is:
= ct = 1
( A)
(1.45)
For air at standard conditions:
2.7 *1019 cm-3 and A 1*1015 cm 2 3.7 *105 cm

This mean free path is certainly much smaller than any length scale in which we are typically
interested.
Examples of where the continuum assumption may not be valid include:
a. flow in the upper atmosphere where the mean free path is large (e.g. at an altitude of 100 miles,
the mean free path is approximately 80 m!),
b. flow within a shock wave where the macroscopic length scale is very small (the width of a shock
wave is on the order of 1 m), and
c. granular flows (e.g. flowing sand) where the microscopic length scale the macroscopic length
scale.
2.
The Knudsen number, Kn, is a dimensionless parameter that indicates when the continuum assumption
is valid:
(microscopic length scale)
Kn
1 the continuum assumption valid
(1.46)
(macroscopic length scale)
Four flow regimes are typically defined based on the Knudsen number. These are (according to
Zucrow, M.J. and Hoffman, J.D., Gas Dynamics Vol. 1, Wiley):
Kn < 0.01
continuum flow
This is the flow regime that occurs in most engineering applications.
0.01 < Kn < 0.1 slip flow
In this regime, the fluid may still be treated as a continuous substance but
the no-slip condition at boundaries does not hold. Instead, fluid may slip at
boundaries.
0.1 < Kn < 3.0 transitional flow
The flow in this regime is very difficult to analyze since the fluid cannot be
considered a continuum and molecules still interact to a considerable degree.
3.0 < Kn
free molecular flow
In this regime molecules are spaced so far apart that they rarely interact.
The flow may be modeled as a collection of non-interacting molecular
impacts.
3.
The continuum assumption is valid in the vast majority of engineering applications. For example, the
Knudsen number for a typical engineering flow in which the mean free path is ~1*10-5 cm (refer to the
calculation in Note #1) and the macroscopic length of interest is 1 mm, is Kn ~ 0.0001.
C. Wassgren
19
4.
The Knudsen number can be related to the Reynolds and Mach numbers using some results from
kinetic theory. From kinetic theory, the kinematic viscosity of a gas, , is related to the mean free path,
, and mean molecular speed, c , by:
= 12 c
(1.47)
The mean molecular speed is related (via kinetic theory) to the speed of sound, c, by:
c=c
(1.48)
8
where is the specific heat ratio (= cp/cv). Substituting Eqn. (1.47) into Eqn. (1.48) and simplifying
gives:
c=
(1.49)
2 c
From the definition of the Knudsen, Reynolds, and Mach numbers:
Kn L =

2 cL
V
(where L is the macroscopic length scale of interest)
c
2 VL

= 1 Re = Ma
L
Kn L =
Ma
2 Re L
Hence, flows occurring at large Mach numbers and small Reynolds numbers tend toward noncontinuum flows.
(1.50)
Note that for large Reynolds numbers over an object, the significant length scale that should be used in
the defining the Reynolds number is typically the boundary layer thickness, . A later set of notes will
show that for (laminar) boundary layers, the boundary layer thickness is related to the macroscopic
length scale, L, by:
Re Re L (where ReL = VL/ and Re = V/)

(1.51)
L
Re L
Hence, Eqn. (1.50) for boundary flows (occurring at large Reynolds numbers) is:
Kn =
5.
Ma
2 Re
Ma
(1.52)
Re L
To learn more about non-continuum flows, refer to the following texts:

a. Schaaf, S.A. and Chambr, P.L., Flow of Rarefied Gases, Princeton University Press.
b. Nguyen, N-T. and Wereley, S.T., Fundamentals and Applications of Microfluidics, Artech House.
C. Wassgren
20
6.
Fluid Properties
Some good fluid property references include:
a. Avallone, E.A. and Baumeister III, T., Marks Standard Handbook for Mechanical Engineers,
McGraw-Hill.
b. Kestin, J., and Wakeham, W.A., Transport Properties of Fluids, CINDAS Data Series on Material
Properties, C.Y. Ho, ed., Hemisphere Publishing.
density, {M/L3}, [kg/m3, slugs/ft3, lbm/ft3]
- The density of a substance is a measure of how much mass there is of the substance per unit
volume.
H 0@4C = 1000 kg/m3 = 1.94 slugs/ft 3 = 62.4 lb m /ft 3
2
air@15 C ,1atm = 1.23 kg/m3 = 2.38*103 slugs/ft 3 = 7.68*102 lb m /ft 3
-
Density does not vary greatly with temperature for liquids in general.
Density does change considerably with temperature and pressure for gases.
A substance where the density remains constant for all conditions is considered incompressible.
- A good engineering rule of thumb is that if there are no significant temperature changes and
for fluid velocities less than approximately 1/3 the speed of sound in the fluid, the fluid can be
approximated as incompressible (the proof of this is examined when discussing compressible
fluid flow).
- In air, the speed of sound at standard conditions (p = 1 atm, T = 59 F = 15 C), is
approximately 1100 ft/s (340 m/s).
- The speed of sound in water is approximately 4800 ft/s (1500 m/s).
- The speed of sound in steel is approximately 16400 ft/s (5000 m/s).
- In most instances, liquids and gases flowing at low speeds can be approximated as
incompressible.
specific gravity, SG {no units dimensionless}

- The specific gravity of a liquid is the ratio of the liquids density to the density of water at
some specified condition (typically at 4C).
SGliquid
liquid
H O @ 4 C
For example, the density of mercury (Hg) at 20C is 13.6*103 kg/m3. Hence, SGHg = 13.6.
-
specific weight, {F/L3} [N/m3, lbf/ft3]

- The specific weight of a substance is the weight of the substance per unit volume.
W mg Vg
=
=
V
V
V
= g
For example, the specific weight of water at 4C is 9.81*103 N/m3 = 62.4 lbf/ft3.
specific volume, v {L3/M} [m3/kg, ft3/slug, ft3/lbm]

- The specific volume of a substance is how much volume the substance occupies per unit mass.
- The specific volume is simply the inverse of the density:
1
v=
In thermodynamics, the specific volume is commonly used in place of density.
Be Sure To:
1. Be careful when using the incompressible flow assumption. Make sure that the assumption is
reasonable for your flow situation. For liquids, the assumption is usually reasonable. For gases, you
need to check the flow velocities and temperatures.
C. Wassgren
21
pressure, p {F/L2} [Pa=N/m2, psf=lbf/ft2, psi=lbf/in2]

-
Pressure is the inward acting force per unit area acting normal to a surface.
dFn
=
* dA
n

*
p

negative sign
outward pointing
area of surface
small normal force
acting on surface
since pressure
acts inward on
surface
pressure acting
on surface
unit normal vector

for the surface
pressure force is normal (at right

angles) to the surface
dFn = pdAn
n
surface with small area, dA, and outward
pointing unit normal vector, n
-
Pressure is a scalar quantity. It has no direction (its a number just like temperature). The surface
orientation is what gives the pressure force its direction.
Proof: Consider the pressure forces acting on a small triangular wedge of fluid at rest. Assume the
wedge extends a depth of 1 (unit depth) into the page. Also assume that the pressure depends
on direction.
p dy
= mx
sin
) (1) sin p ( dy )(1) = (

x
p(dy/sin)(1)
1
dxdy )(1)
x
( p px ) dy = ( 1 2 dxdy ) x
y
dy/sin
As dx, dy 0, p = px
A similar approach can be taken in the y-direction

to find: p = py
dy
dx
py(dx)(1)
Thus, we observe that the pressure is the same in every direction.

-
Pressure is the result of molecules colliding against (a real or imaginary) surface. Every time a particle
bounces off a surface, its momentum, mu, changes (u changes direction). From Newtons 2nd Law,
F=d/dt(mu), there must be a force exerted on the molecule by the wall (and conversely, on the wall by
the molecule) in order to cause the change in the molecules momentum. These collisions result in
what we call pressure.
Absolute pressure [psia] is referenced to zero pressure (a perfect vacuum has pabs=0).
- There are no molecules in a perfect vacuum thus there is no pressure (due to molecular collisions).
- Atmospheric pressure at standard conditions:
patm, abs = 14.696 psia = 101.33*103 Pa (abs) = 1 atm = 1.0133 bar
-
Gage pressure [psig] is referenced to atmospheric pressure: pgage = pabs - patm

- Atmospheric pressure at standard conditions: patm = 0 psi (or psig).
- A perfect vacuum has p = -14.7 psig (referencing to standard atmospheric conditions).
An important equation of state for an ideal gas is the ideal gas law:
p = RT
Note that p and T are absolute quantities, e.g. [p] = psia or Pa (abs), [T] = R or K
C. Wassgren
22
px(dy)(1)
The vapor pressure, pv of a liquid is the pressure at which the liquid is in equilibrium with its own
vapor.
The pressure in the vapor is

the vapor pressure, pv.
vapor
liquid
A closed container with the liquid and
vapor in equilibrium.
-
The vapor pressure for a liquid will increase with increasing temperature.
If the pressure in the liquid falls below the vapor pressure, then the liquid will turn to vapor. This
can occur via boiling or cavitation.
- Boiling occurs when the temperature of the liquid increases so that the vapor pressure equals
the surrounding atmospheric pressure. For example, the vapor pressure of water at 20 C is
0.023 atm while the vapor pressure at 100 C is 1 atm. Hence, one method of turning liquid
water to vapor is to bring the water temperature to 100 C while holding the surrounding
pressure at 1 atm. This is known as boiling and is shown schematically in the phase plot
shown below on the left-hand side.
- Cavitation occurs when the surrounding pressure drops below the vapor pressure. Using the
previous example, we can also turn liquid water to water vapor by dropping the surrounding
pressure to 0.023 atm at 20 C. This is shown schematically in the phase plot shown below
on the right-hand side.
pressure
pressure
liquid
liquid
solid
solid
cavitation
(T =constant, p )
boiling
(p=constant, T )
gas
gas
temperature
-
temperature
Cavitation can cause considerable damage to surfaces. When cavitation occurs in a liquid,
pockets of vapor form (either as bubbles or large voids). As the vapor pockets travel into a
region where the surrounding pressure is greater than the vapor pressure, the vapor region
rapidly collapses. This collapse can be so rapid that shock waves propagate from the
collapsing region and impact on nearby surfaces causing small bits of the surface to erode
away. Hence, cavitation is typically avoided when designing pumps, pipe bends, and
underwater propellers.
Be Sure To:
1. Use an absolute pressure, and not a gage pressure, in the ideal gas law.
2. Be careful not to mix gage and absolute pressures when evaluating pressure forces.
3. Use the correct area when calculating a pressure force.
4. Integrate to find a pressure force when the pressure is not uniform over the area over which the
pressure acts.
5. Check for cavitation in low pressure flows of liquids.
C. Wassgren
23
temperature, T {} [R, K, F, C]
An objects temperature is a quantitative way of describing how hot the object is (temperature is a
measure of the agitation or random kinetic energy of the molecules). We typically measure an objects
temperature using a device called a thermometer.
Experience tells us that when two objects are placed in contact with each other and have different
temperatures, the hotter object (i.e., the one with a larger temperature) will become cooler while the cooler
object becomes hotter. When the two objects have the same temperature, they no longer change
temperature. The objects are then said to be in thermal equilibrium.
A simple but fundamental concept concerning thermal equilibrium is the:
Zeroth Law of Thermodynamics: If two bodies are in thermal equilibrium with a third body, then
the two bodies will also be in thermal equilibrium.
This concept is key when comparing the temperatures of two objects not in contact using a thermometer
since the thermometer acts as the third body.
To use the concept of temperature, we must first define some scale on which well measure temperatures.
Perhaps the easiest scale to define would be one where we reference all temperatures to some reproducible
and unique physical phenomena such as the freezing or boiling points of water.
Two point scales use two phenomena to define the temperature scale. For example, lets define the ice
point of water (when ice and water are in equilibrium at a pressure of 1 atm) as our 0 temperature and the
steam point of water (when water and water vapor are in equilibrium at a pressure of 1 atm) as our 100
temperature. We can now measure all temperatures relative to this scale. Examples of two-point scales
include the Celsius and Fahrenheit scales:
ice point of H2O steam point of H2O
Celsius scale
0 C
100 C
Fahrenheit scale
32 F
212 F
Long ago, researchers noticed something curious when measuring the temperature of various gases at
different pressures (and constant volume). They found that the temperature of a gas at low pressures is
proportional to its pressure at a constant volume:
T = a + bp
(1.53)
where a and b are constants.
p
gas #1
gas #2
gas #3
gas #4
-273.15 C
= -459.67 F
By extrapolating the temperature of the gases at zero pressure, we find that the lowest possible temperature,
or the absolute zero temperature, is 273.15 C on the Celsius scale which corresponds to 459.67 F on
the Fahrenheit scale (i.e., a= -273.15 C = -459.67 F). This temperature scale, based on the behavior of
ideal gases, is referred to as the ideal gas temperature scale.
C. Wassgren
24
To make things a bit easier, lets redefine our scale so that absolute zero is the zero point of our temperature
scale, i.e.:
T = bp
(a=0)
p
gas #1
gas #2
gas #3
gas #4
0K
This is called an absolute temperature scale since the lowest temperature is zero.
Note that so far our temperature scales have been based on the behavior of a particular substance (e.g.
water) or a particular class of substances (e.g. gases). A better way to define a temperature scale is to make
the scale independent of substances. Such a scale is called a thermodynamic temperature scale.
In order to define such a temperature scale, we would need to first learn about the 2nd Law of
Thermodynamics; a topic not covered in these notes (refer to [3] for this topic). Suffice it to say here that
the scale using this method gives the same result as that using the ideal gas temperature scale.
To summarize, the lowest possible temperature is:
0 R = 0 K
and
(1 K) = (1 C)
(1 R) = (1 F)
Some additional helpful conversions are given below (the refers to temperature):
(K) = 1.8 (R) (1.8 = 9/5)
(C) = [(F)-32]/1.8
(C) = (K) 273.15
(F) = (R) 459.67
Another convenient conversion formula:
10 C = 50 F (for every 5 C increase, add 9 F )
Be Sure To:
1. Use an absolute temperature when using the ideal gas law.
References
3. Moran, M.J. and Shapiro, H.N., Fundamentals of Engineering Thermodynamics, Wiley.
C. Wassgren
25
(dynamic) viscosity, {FT/L2} [Ns/m2, lbfs/ft2]

10 Poise = 1 Ns/m2 = 1 kg/(ms)
1 centipoise = 0.01 Poise = 0.001 Ns/m2
(Note: Poise is pronounced pwz.)
Another common unit:
Viscosity is the internal friction within the fluid. Its a measure of how easily a fluid flows.
The viscous stresses in a fluid will be related to the deformation rate of a small element of fluid.
Recall that for solids, the force is related to the amount of deformation, e.g. Hookes Law for springs.
For fluids, however, the forces are related to the deformation rates.
Consider the deformation of a small piece of fluid with area (dxdy) as shown below. The top of the
fluid element is subject to a shear stress, , over a short time, dt. During this time, the top of the fluid
element moves with a small velocity, du, with respect to the bottom of the element. The total distance
the top moves relative to the bottom will be (du*dt). The angular deformation of the fluid element can
be measured by the angle the vertical sides of the element have deformed.
(du*dt)
dy
dx
Here, the small angle, d, is found from simple trigonometry to be:

dudt
tan d =
dy
Since the angle is very small, tan(d) d. The rate at which the element deforms is found by
dividing both sides by dt:
d du
=
shear or angular deformation rate
dt dy
Hence, the rate of angular deformation of the element, d/dt, is equal to the velocity gradient, du/dy, in
the fluid. Furthermore, since the shear stress, , will be a function of deformation rate (as stated
previously), we have:
du
d
= fcn = fcn
dt
dy
where the term fcn refers to the fact that we dont yet know how the shear stress depends on the
deformation rate, we just know that it does.
-
A Newtonian fluid is one in which the shear stress varies proportionally with the deformation rate.
The constant of proportionality is called the dynamic viscosity, .
du
=
shear stress relation for a Newtonian fluid
(1.54)
dy
- Air and water are two examples of Newtonian fluids.
- A more precise definition of the shear stress in Eqn. (1.54) is:
du
yx = x
dy
C. Wassgren
26
where the subscript on the stress indicates that the shear stress acts on a y-plane in the xdirection. The x-subscript on the velocity indicates that it is the x-component. We will review
this sign convention in greater detail later in the notes when discussing stresses (Chapter 03).
-
In a non-Newtonian fluid the shear stress is not proportional to the deformation rate but instead
varies in some other way.
- Non-Newtonian fluids are further classified by how the shear stress varies with deformation
rate. The apparent viscosity, app, is the viscosity at the local conditions as shown in the plot
below (for a Newtonian fluid the apparent viscosity remains constant).
non-Newtonian
shear thinning
Newtonian
shear stress,
non-Newtonian
shear thickening
apparent viscosity, app
rate of shearing strain
(deformation rate), du/dy
-
In shear thinning (aka psuedoplastic) fluids, the apparent viscosity decreases as the shear
stress increases. Examples of shear thinning fluids include blood, latex paint, and cookie
dough.
In shear thickening (aka dilatant) fluids, the apparent viscosity increases as the shear stress
increases. An example of a shear thickening fluid is quicksand or a thick cornstarch-water
mixture.
Viscosity is weakly dependent on pressure but is sensitive to temperature. For liquids, the
viscosity generally decreases as the temperature increases and increases as pressure increases.
Changes in temperature and pressure can be very significant in lubrication problems. For gases,
the viscosity increases as the temperature increases (in fact, from kinetic theory one can show that
T1/2). The following plot shows the variation in dynamic viscosity with temperature for
several fluids.
H20@20 C = 1.00*10-3 Ns/m2 = 1 cP

air@20 C = 1.81*10-5 Ns/m2 = 0.018 cP
(Plot from Fox, R.W. and

McDonald, A.T., Introduction to
Fluid Mechanics, 5th ed., Wiley.)
C. Wassgren
27
The kinematic viscosity, , is a quantity that appears often in fluid mechanics. It is defined as:
H20@20 C = 1.00*10-6 m2/s = 1 cSt
air@20 C = 1.50*10-5 m2/s = 15 cSt
The dimensions of kinematic viscosity are {L2/T} with common units of [m2/s, ft2/s]. Another
common unit for kinematic viscosity is the Stoke: 1 Stoke = cm2/s. Note that kinematic viscosity
has the dimensions of a diffusion coefficient. The kinematic viscosity is a measure of how rapidly
momentum diffuses into a flow. This will be discussed again in a later set of notes when
reviewing solutions to the Navier-Stokes equations (Chapter 06).
There are several techniques for measuring the viscosity of a fluid. Several devices include the:
- Couette viscometer
- capillary tube viscometer
- sedimentation rate viscometer
- falling ball or falling cylinder viscometer
- rotating disk or oscillating disk viscometer
A good reference on experimental viscometry is: Dinsdale, A. and Moore, F., Viscosity and its
Measurement, Chapman and Hall.
Lets examine a common flow situation shown in the figure below. A fluid is contained between
two, infinitely long parallel plates separated by a distance H. The bottom plate is fixed while the
top plate moves at a constant velocity V. There are no pressure gradients in the fluid. This type of
flow situation is called a planar Couette flow.
top plate moves at
constant velocity, V
fluid
fluid velocity,
u(y) = (V/H)y
y
x
fixed bottom plate

-
One of the first things we would notice when conducting this experiment is that the fluid sticks to
the solid boundaries, i.e. there is zero relative velocity between the fluid and the boundary. This is
referred to as the no-slip condition. The no-slip condition occurs for all real, viscous fluids under
normal conditions. (The no-slip condition may be violated in rarefied flows where the motion of
individual molecules becomes significant, i.e. when the Knudsen number is Kn > 0.1.)
The second thing we would notice is that the fluid velocity profile is linear with the velocity given
by:
y
u =V
H
Note that the velocity at the bottom plate is zero and at the top plate the velocity is V; thus, the
no-slip condition is satisfied. We will derive this velocity profile later in these notes when
discussing solutions to the Navier-Stokes equations (Chapter 06).
C. Wassgren
28
If the fluid is Newtonian, then the shear stress acting on the fluid is:
du
V
yx = x =
dy
H
Note that the shear stress is a constant everywhere in the fluid, i.e. there is no y dependence.
Additionally, the shear stress will act to resist the motion of the top plate and try to carry the
bottom plate along with the fluid.
shear stress acting on wall

due to fluid
positive shear stress acting

on fluid element next to wall
(refer to the stress sign
convention in Chapter 03)
-
An inviscid fluid is one that has zero viscosity ( = 0). Although there are no typical fluids that
are completely inviscid (a counter-example: superfluid helium is completely inviscid!), there are
many cases where assuming that the fluid viscosity is negligible is a reasonable approximation. In
addition, many analyses of fluid flow rely on an inviscid assumption in order to make the
mathematics of the analyses tractable without the use of computers. Note that the no-slip
condition does not hold when the fluid is inviscid.
An ideal fluid is one that is incompressible and inviscid. The ideal fluid approximation is often
reasonable and is commonly used in fluid mechanics analyses. Well consider ideal fluid flow in
Chapter 04.
Be Sure To:
1.
Get your shear stress directions correct. Eqn. (1.54) is the shear stress acting on the fluid element.
2.
Use the correct area when evaluating shear forces.
3.
Integrate to determine a shear force on an area where the shear stress is not uniform over the area.
4.
Evaluate the velocity gradient in Eqn. (1.54) at the location where youre interested in determining
the shear stress.
C. Wassgren
29
Example:
Determine the magnitude and direction of the shear stress that the water applies:
a. to the base
b. to the free surface
free surface
U
u
y y
= 2
U
h h
water
y
u
SOLUTION:
The shear stress, yx, acting on a Newtonian fluid is given by:
du
yx =
(1.55)
dy
where
du
1 2y
=U 2 2
(1.56)
dy
h h
Evaluating the shear stress at the base and free surface gives:
2U
yx y = 0 =
base (y = 0):
(1.57)
h
This is the stress the wall exerts on the fluid. The fluid will exert an equal but
opposite stress on the wall.
y
x
free surface (y = h):
yx
y =h
fluid element adjacent to the

wall with positive yx shown
=0
(1.58)
The air at the free surface does not exert a stress on the water. Although in
reality the air will exert a small shear stress on the water, assuming that the shear
stress is negligible is reasonable in most engineering applications.
C. Wassgren
30
Example:
Magnet wire is to be coated with varnish for insulation by drawing it through a circular die of 0.9 mm
diameter. The wire diameter is 0.8 mm and it is centered in the die. The varnish, with a dynamic viscosity
of 20 cP, completely fills the space between the wire and the die for a length of 20 mm. The wire is drawn
through the die at a speed of 50 m/s. Determine the force required to pull the wire.
SOLUTION:
Apply Newtons 2nd Law in the x-direction to the wire shown in the diagram below.
R0 R
i
r x
Since the gap width is small compared to the wire radius,

curvature effects may be neglected and a linear velocity
profile may be reasonably assumed (i.e. (Ro Ri)/Ri << 1).
F, V
= F + rx ( 2 Ri L ) = 0 (Note that the wire is not accelerating.)
(1.59)
where rx is the shear stress the fluid exerts on the wire and (2RiL) is the area over which the shear stress
acts. The shear stress the wire exerts on the fluid (assumed to be Newtonian) is:
du
rx =
(1.60)
dr
where
du
0 V
V
=
=
(Note that u(r = Ro) = 0 and u(r = Ri) = V.)
(1.61)
dr Ro Ri Ro Ri
Recall that the shear stress the fluid exerts on the wire will be equal to, but opposite, the value given by Eqn.
(1.60). Substitute Eqns. (1.60) and (1.61) into Eqn. (1.59) and simplify.
V
F
( 2 Ri L ) = 0
Ro Ri
2 Ri L V
F =
(1.62)
Ro Ri
Use the numerical values given in the problem statement.
Ri = 4.0*10-4 m
Ro = 4.5*10-4 m
L = 2.0*10-2 m
= 20 cP = 0.02 kg/(ms) (Note: 100 cP = 0.1 kg/(ms).)
V = 50 m/s
F = 1.0 N
C. Wassgren
31
bulk modulus, E {F/L2} [Pa=N/m2, psi=lbf/in2]

-
The bulk modulus is a measure of the compressibility of a substance. The larger the bulk modulus, the
less compressible the substance. An incompressible substance has an infinite bulk modulus.
The bulk modulus is defined as the ratio of the differential pressure, dp, over the relative change in the
differential volume, dV/V, caused by the differential pressure change:
dp
E
bulk modulus
dV
dp
V
original volume, V
- If dp>0 (increase in pressure) and dV<0 (decrease in volume), then E>0.
change in volume, dV
- The bulk modulus can also be written in terms of density rather than volume:
dp
E =
d
( )
since m=V and dm=0=dV+dV d/ = -dV/V.

Water at 20 C, E = 2.21 GPa.
For an ideal gas under isothermal conditions:
dp
p = RT
= RT
E T = RT = p
d T
For an ideal gas under isentropic conditions:
dp
p = ( const )
= ( const ) 1
d S
= p = RT
where =cp/cv.
C. Wassgren
32
Example:
What is the fractional change in density of water in the ocean over a depth of 1000 m (3281 ft)? The
change in pressure over this depth, p, is 9.81*106 Pa.
SOLUTION:
Assume that the bulk modulus for water remains constant at E = 2.2*109 Pa (found from a fluid properties
table). Recall that the bulk modulus is defined as:
dp
E =
(1.63)
d
1
=
E
pf
dp
p0
p f p0
f
p
ln
=
=
E
E
0
p
= exp
0
E
Note that f = 0 + so that:
f 0 +
=
= 1+
(1.64)
(1.65)
Substitute Eqn. (1.65) into Eqn. (1.64):

p
1+
= exp
0
E
p
= exp
1
E
In the given problem:

p = 9.81*106 Pa
E = 2.2*109 Pa
= 4.5*10 = 0.45%
0
(1.66)
Water remains essentially incompressible over this depth!
Note that in most engineering problems (where pressure variations are not enormous), it is reasonable to
assume that water is incompressible.
C. Wassgren
33
The speed of sound in a substance is related to how compressible the substance is. It can be
shown that the speed of sound, c, is given by:
c=
(note that the S indicates that the process occurs isentropically)

S
(This derivation is covered in the notes regarding compressible flows in Chapter 11.) Since the
bulk modulus can be written as:
dp
dp
E =
=
d
d
( )
we have:
c=
Hence, we observe that the speed of sound in a substance is related to the ratio of the substances
compressibility to its density.
-
Recall that for an isentropic process involving an ideal gas: E = p = RT, so that:
c = RT
speed of sound for an ideal gas
C. Wassgren
34
surface tension, {F/L} [N/m, lbf/ft]

-
Surface tension is caused by unbalanced molecular forces occurring at the interface between a liquid
and a solid surface, two dissimilar liquids, or between a liquid and a gas.
- For example, water beads up on a waxed surface because the intermolecular forces between the
water molecules is greater than the intermolecular forces between the water and the wax.
Examples where we observe surface tension include:

bubbles
bugs walking on the surface of water
trees
soda in a straw
Surface tension effects become much more significant as length scales (L) decrease since:
Fsurface tension ~ L (proportional to interface length)
Finertia, body forces ~ L3 (proportional to volume)
C. Wassgren
35
Example:
Determine the relationship between the surface tension in a soap bubble and the pressure difference
between the inside and outside of the bubble.
SOLUTION:
Analyze the problem by cutting the spherical bubble with radius, R, in half. The free body diagram of the
half-bubble is shown in the figure below. The forces acting on the bubble include the surface tension force
holding the two bubble halves together, and the pressure force acting to push the two bubble halves apart.
surface tension force,

FST = (2R)
exterior pressure force,

FPE = pE (R2)
F =
interior pressure force,

FPI = pI (R2)
FPE + FST FPI = 0

= pE R 2
( pI pE ) =
= 2 R
= pI R 2
2
R
C. Wassgren
36
The contact angle between a liquid and a surface is the angle the liquid surface makes at the contact
point with the surface as shown in the figure below.
liquid
contact angle (always measured through the liquid)
If > 90, then the surface is considered hydrophobic (it doesnt like the fluid).
If < 90, then the surface is considered hydrophilic (it does like the fluid).
- For example:
water on metal:
hydrophilic
mercury on metal
hydrophobic
Precise measurements of contact angle are difficult to make due to the sensitivity of the
contact angle to surface chemical variations.
Be Sure To:
1. Consider the effects of surface tension at fluid interfaces, especially at small length scales.
C. Wassgren
37
Example:
Determine the height a liquid will rise in a cylindrical tube due to surface tension. The contact angle of the
liquid on the tube surface is .
SOLUTION:
Draw a free body diagram of the liquid contained in the tube above the free surface. The two forces acting
on the liquid will be its weight and the surface tension force. These forces are indicated in the figure on the
bottom right.
2R
surface tension force in y-direction:

= (2R)cos
H
patm
weight (R2H)g
liquid
2R
From Newtons 2nd Law in the y-direction:

Fy = 0 = ( 2 R ) cos ( R 2 H ) g
H =
2 cos
Note that as R , H .
gR
C. Wassgren
38
7. Flow Kinematics
A good reference on the experimental aspects of this topic is: Merzkirch, W., Flow Visualization,
Academic Press.
In 1986, the Chernobyl Nuclear Power Plant in the Soviet Union released
radioactive fallout into the atmosphere as a result of an explosion in one of
the reactors (see the figure to the right). The radioactive plume covered
regions of the western Soviet Union, Europe, and even parts of eastern
North America. If such an accident were to happen again, how would you
know which communities would be affected by the drifting radioactive
cloud? If one had predictions of wind velocity measurements as a function
of location and time, i.e. u = u(x, t), could you figure out what areas will be
covered by the cloud? In this section, well present three forms of flow
kinematics: streamlines, streaklines, and pathlines. Each of these lines
provides different information on the movement of fluid. For the toxic
cloud release, one would be most interested in determining the streakline
passing through the location of the damaged reactor.
Damaged reactor at Chernobyl (photo from
http://en.wikipedia.org/wiki/Chernobyl_disaster).
Streamlines
A streamline is a line that is everywhere tangent to the velocity field vectors.
Note that since streamlines are parallel to the
velocity vectors, there will be no flow across a
streamline.
streamline
Experimentally, one can visualize streamlines using the Particle Image Velocimetry (PIV) technique. In
PIV, fluid particles are tagged, usually by mixing in very small, neutrally buoyant bits of paint, and
taking two photographs in rapid succession. Velocity vectors can then be produced by connecting the
dots.
photo at time t
photo at time t+t
approximate velocity
vectors at time t
Note that the values of approximate velocity vectors can be found from the definition of a derivative:
x (t + t ) x (t )
x ( t )
t
C. Wassgren
39
We can determine the equation of a streamline given a velocity field by simply using the definition of a
streamline. Lets zoom in on a small part of a streamline (well consider only a 2D flow for simplicity):
Since the streamline is tangent to the velocity vector, the slope of
the streamline will be equal to the slope of the velocity vector:
uy
dy
(1.67)
=
dx
ux

u
uy
dx
dy
ux
slope of
streamline
slope of
velocity vector
Similarly, in the x-z and y-z planes we have:

dz u z
dz u z
=
=
dx u x
dy u y
(1.68)
We can combine Eqns. (1.67) and (1.68) into a more compact form:
dx dy dz
=
=
Equation for a streamline.
ux u y uz
(1.69)
Notes:
1. There is no flow across a streamline.
2. A stream tube is a tube made by all the streamlines passing through a closed curve. There is no flow
through a stream tube wall.
streamlines
3.
A stream filiment is a stream tube with infinitesimally small cross-sectional area.
Streaklines
A streakline consists of a line that connects all fluid particles that have passed through the same point in
space at a previous (or later) time.
Experimentally a streamline can be visualized by injecting dye into a fluid flow at a particular point.
fluid particle that passed
through (x0, y0, z0) at time t0=t1
dye
(x0, y0, z0)
fluid particle that passed

through (x0, y0, z0) at time t0=t2
streakline at time t passing

through (x0, y0, z0)
To determine the equation of a streakline at time, t, passing through the point (x0, y0, z0), we solve the
differential equation describing a particles position:
dx
u=
(1.70)
dt
where u is the velocity field subject to the initial condition that a particle pass through the point x0 = (x0, y0,
z0) at some previous (or later) time, t0:
x ( t = t0 ) = x 0
(1.71)
C. Wassgren
40
Note that t0 will be different for each particle, i.e. it varies (refer to the fluid particles in the figure above).
The solution of Eqn. (1.70) subject to initial condition in Eqn. (1.71) will consist of a set of parametric
equations in t0 (note that t is a known value since we want to know the streakline at a particular time, t).
Pathlines
A pathline is the line traced out by a particular particle as it moves from one point to another. It is the
actual path a particle takes.
Experimentally a pathline can be visualized by tagging a particular fluid particle and taking a longexposure photograph of the fluid motion.
pathline of a particle passing
through (x0, y0, z0) at time t0
particle passes
through (x0, y0, z0)
at time t0
location of tagged fluid particle

at time t
To determine the equation of a pathline at time, t, for a particle passing through the point (x0, y0, z0) at some
previous time t0, we solve the differential equation describing the particles position:
dx
u=
(1.72)
dt
where u is the velocity field subject to the initial condition that the particle pass through the point x0 = (x0,
y0, z0) at some time, t0:
x ( t = t0 ) = x 0
(1.73)
Note that t0 will be a particular value, i.e. its fixed. The solution of Eqn. (1.72) subject to initial condition
in Eqn. (1.73) will consist of a set of parametric equations in t (note that by varying t we can trace out the
location of the particle for various times).
Notes:
1. The streamline, streakline, and pathline passing through a particular location can be different in an
unsteady flow, but will be identical in a steady flow.
2. The quantity t0 is the time when a fluid particle passes through the point x0. Hence, for a pathline t0 is
fixed since there is only one fluid particle. However, for a streakline t0 varies since there are many
fluid particles passing through the point x0, each at a different t0.
3. Why dont we use a Lagrangian derivative (Chapter 03 material) when solving Eqn. (1.72) for a
particles pathline (since the pathline is a Lagrangian concept)? It turns out that the Lagrangian
derivative of a particles position is equal to its Eulerian derivative. Consider, for example, the change
in the x position of the particle as we follow it. Note that the position, x, is an Eulerian quantity.
Dx x
x
x
x
x
(1.74)
=
+ ( u ) x =
+ ux
+ uy
+ uz
= ux
Dt t
t
x
y
z

=0
4.
=1
=0
=0
A demonstration of the differences between streamlines, streaklines, and pathlines can be found at:
https://engineering.purdue.edu/~wassgren/applets
Be Sure To:
1. Understand the definitions for streamlines, streaklines, and pathlines.
2. Understand what initial conditions to use when evaluating streaklines and pathlines.
3. Draw the direction of flow on the streamlines, streaklines, and pathlines.
4. Its perfectly correct to represent the position of a fluid particle parametrically, i.e. x = x(t) and y = y(t).
C. Wassgren
41
Example:
A velocity field is given by:

V y
V x
u = 2 0 2 1 i + 2 0 2 1 j
2
(x + y )
(x + y ) 2
where V0 is a positive constant, i.e. V0 > 0. Determine:
a. where in the flow the speed is V0
b. the equation and sketch of the streamlines
c. the equations for the streaklines and pathlines
SOLUTION:
The speed is given by:
u = u x2 + u y2
(1.75)
where
ux =
uy =
V0 y
2
(x + y2 )
V0 x
(1.76)
2
(1.77)
( x2 + y 2 ) 2
Substituting into Eqn. (1.75) gives:
u =
V02 y 2
( x2 + y2 )
V02 x 2
( x2 + y2 )
u = V0
(1.78)
The flow speed is everywhere equal to V0.

The slope of the streamline is tangent to the slope of the velocity vector:
dy u y
=
dx u x
Substitute Eqns. (1.76) and (1.77) and solving the resulting differential equation.
V0 x
(1.79)
dy ( x 2 + y 2 ) 2
x
=
=
V0 y
dx
y
1
( x2 + y2 ) 2
y
xdx
y0
x0
ydy =
) (x
12 y 2 y02 =
2
(where (x0, y0) is a point located on the streamline)
x02
1
2
x02
y02
x +y = +
= constant
(1.80)
The streamlines are circles! Note that when x > 0 and y > 0, Eqns. (1.76) and (1.77) indicate that ux < 0 and
uy > 0 (note that V0 > 0) so that the flow is moving in a counter-clockwise direction.
x
Since the flow is steady, the streaklines and pathlines will be identical to the streamlines.
C. Wassgren
42
Example:
Consider a 2D flow with a velocity field given by:
u = x (1 + 2t )i + yj
Determine the equations for the streamline, streakline, and pathline passing through the point (x,y)=(1,1) at
time t=0.
SOLUTION:
The slope of a streamline is tangent to the velocity vector.
dy u y
=
dx u x
where
u x = x(1 + 2t )
(1.81)
(1.82)
uy = y
(1.83)
Substitute Eqns. (1.82) and (1.83) into Eqn. (1.81) and solve the resulting differential equation.
dy
y
=
dx x(1 + 2t )
y
dy
(1 + 2t )
=
y
y0
x0
dx
(where (x0, y0) is a point passing through the streamline)
x
y
x
(1 + 2t ) ln = ln
y0
x0
(1+ 2t )
y
x
=

y
0
x0
For the streamline passing through the point (x0, y0) = (1, 1) at time t = 0:
y=x
(1.84)
(1.85)
A streakline is a line that connects all of the fluid particles that pass through the same point in space. The
equation for the streakline can be found parametrically using Eqns. (1.82) and (1.83).
dx
ux =
(1.86)
= x(1 + 2t )
dt
dy
uy =
=y
(1.87)
dt
Solve the previous two differential equations.
x
x0
y
y0
dx
= (1 + 2t )dt
x
x
ln = t + t 2 t0 t02
x0
(1.88)
y
ln = t t0
y0
(1.89)
t0
t
dy
= dt
y
t0
where t0 is the time at which a fluid particle passes through the point (x0, y0) on the streakline. Hence, the
streakline passing through the point (x0, y0) = (1, 1) at time t = 0 is given parametrically (in t0) as:
ln ( x ) = t0 t02
x = exp t0 t02
ln ( y ) = t0
y = exp ( t0 )
(1.90)
(1.91)
Recall that t0 is the time when a fluid particle passes through the point (x0, y0).
C. Wassgren
43
A pathline is a line traced out by a particular fluid particle as it moves through space. The equation for the
pathline can be found parametrically using Eqns. (1.82) and (1.83).
dx
ux =
= x(1 + 2t )
(1.92)
dt
dy
uy =
=y
(1.93)
dt
Solve the previous two differential equations.
x
x0
y
y0
dx
= (1 + 2t )dt
x
x
ln = t + t 2 t0 t02
x0
(1.94)
y
ln = t t0
y0
(1.95)
t0
t
dy
= dt
y
t0
where t0 is the time at which a fluid particle passes through the point (x0, y0) on the pathline. Hence, the
pathline for a particle passing through the point (x0, y0) = (1, 1) at time t0 = 0 is given parametrically (in t)
as:
ln ( x ) = t + t 2
x = exp t + t 2
ln ( y ) = t
y = exp ( t )
(1.96)
(1.97)
Note that the streamline, streakline, and pathline are all different. A plot of these lines through (1, 1) at t =
0 is shown below.
16
streamline
streakline
pathline
14
12
10
8
6
4
2
0
0
10
C. Wassgren
44
8.
Some Basic Definitions and Concepts
Before we begin to study the behavior of fluids, we need to first define a few commonly used terms.
fluid
A fluid is a substance that deforms continuously when subject to non-isotropic stresses.
Notes:
1.
An isotropic stress state is one in which the stresses are the same in all directions.
2.
Liquids and gases are considered fluids. Their properties are different (e.g. the density of air at
standard conditions is 1.23 kg/m2 while the density of water at standard conditions is 1000 kg/m3),
but their bulk motion is similar.
3.
Solids will not deform continuously when subject to unbalanced forces. They will deform up until
a point and then resist further deformations (e.g. behavior similar to that of a spring).
4.
Some substances are difficult to classify, e.g. slurries and granular materials. The study of
material behavior (deformation and flow) is known as rheology. For additional information, refer
to Steffe, J.F., Rheological Methods in Food Process Engineering, Freeman Press and Bird, R.B.,
Armstrong, R.C., and Hassager, O., Dynamics of Polymeric Liquids, Wiley.
system and surroundings

A system is a particular quantity of matter chosen for study. The surroundings include everything that
is not the system.
control volume (CV), control surface (CS), and outward-pointing unit normal vector
A control volume (CV) is a particular volume or region in space. A control surface (CS) is the surface
enclosing the control volume. The orientation of the CS at a particular location is given by the
direction of its outward-pointing unit normal vector, n , at that location. The outward-pointing unit
normal vector has a magnitude of one, is perpendicular to the control surface, and always points out of
the CV.
CV
n
CS
C. Wassgren
45
property (extensive, intensive, and specific)

Properties are macroscopic characteristics of a system. An extensive property is one that depends on
the amount of mass in the system. An intensive property is one that is independent of the mass in the
system. A specific property is an extensive property per unit mass (a specific property is also an
intensive property). Some examples: mass, m, is an extensive property, pressure, p, is an intensive
property, and specific volume, vV/m, is a specific property. An easy way to determine whether a
property is extensive or intensive is to divide the system into two parts and see how the property is
affected.
state
The state of a system is the systems condition or configuration as described by its properties in
sufficient detail so that it is distinguishable from other states. Often a state can be described by a
subset of the systems properties since the properties themselves may be related.
scalar and vector fields
A field representation of a quantity gives the value of that quantity at all locations and times, i.e. the
field representation of the density, , is:
= ( x, y , z , t )
A scalar is a quantity that has only a magnitude, e.g. temperature, T. A vector is a quantity that has
both magnitude and direction, e.g. velocity, u.
steady and unsteady
A steady flow is one in which flow properties at each location do not change with time, i.e.:
() = 0
t
An unsteady flow is a flow where properties do change with time.
flow dimensionality
The dimension of a flow is equal to the number of spatial coordinates required to describe the flow.
For example:
0D flow:
u=u(t) or u=constant
1D flow:
u=u(t; x) or u=u(t; x)
2D flow:
u=u(t; r, ) or u=u(t; r, )
3D flow:
u=u(t; x, y, z) or u=u(x, y, z)
A flow is uniform if it does not vary in a spatial direction. A flow is non-uniform if it does vary in that
spatial direction. For example, the velocity profile in the figure below at left is non-uniform in the ydirection. The figure at the right is uniform in the y-direction.
y
Be Sure To:
1. Understand the previous definitions since they will be used frequently in the discussion of fluid
mechanics.
C. Wassgren
46
Review Questions
1. What is the conversion between lbm and slugs? What is the conversion between lbf and lbm? Estimate,
without a calculator, what 20 C is in F.
2. What are the dimensions (in M, L, T) of power?
3. Do you understand the physical concept behind the Taylor series approximation given below?
dy
y ( x + dx ) = y ( x ) +
( dx )
dx x
Can you draw a sketch corresponding to your explanation?
4. What is meant by least count? Give an example of a situation where the least count for a
measurement may be different for two experimenters using the same equipment.
5. Set up and work out an example concerning propagation of experimental uncertainty.
6. What is the difference between relative and absolute uncertainty?
7. Why should uncertainty be included in experimental measurements?
8. Under what conditions can a fluid be considered a continuum?
9. Give an example of where the continuum approximation is not appropriate.
10. What is the no-slip assumption?
11. Discuss the definition and meaning of the Knudsen number.
12. What is meant (in words) by a Newtonian fluid? Write the shear stress-shear strain rate relationship
for a Newtonian fluid. Is this the shear stress that acts on the fluid or that the fluid exerts on its
neighbors?
13. Give an example of a shear thickening fluid.
14. How does viscosity vary with temperature for liquids? with pressure?
15. What is the difference between dynamic and kinematic viscosities?
16. What is meant by an inviscid fluid?
17. What is meant by an incompressible fluid? Under what conditions can a flow be considered
incompressible?
18. What is meant by an ideal fluid?
19. What are the pressure and temperature at standard conditions?
20. What is the density of air at standard conditions (in kg/m3)? What is the approximate density of water
under ordinary conditions (in kg/m3)? What is the specific gravity of mercury?
21. What is the difference between absolute and gage pressures?
22. An engineer wants to minimize the resistance on a block sliding over a thin film of liquid. Would it be
better to increase the thickness of the film, decrease the thickness, or does it matter? Explain your
reasoning in words and provide analytical support for your argument (i.e. supporting equations).
23. Estimate the absolute and kinematic viscosities of water under ordinary conditions.
24. What is meant by a Couette flow? Under what geometric conditions would it be reasonable to apply
the Couette flow assumption to flow between two rotating concentric cylinders?
25. What is cavitation? Can cavitation be a problem in a pipeline containing steam? Give some examples
of where cavitation occurs. Will a submarine propeller be more likely to cavitate near the ocean surface
or deep in the ocean? Why?
26. How do you expect the speed of sound in water containing small air bubbles to compare with the speed
of sound in water and the speed of sound in air? (Hint: Consider the mixtures bulk modulus and
density to the pure substance values.)
27. What is the approximate speed of sound in air at standard conditions?
28. Which bubble has greater internal pressure: one with a smaller radius or one with a larger radius?
29. What are the dimensions of surface tension?
30. Draw a picture defining the contact angle measurement.
31. A small cubicle object is place in water. Determine the maximum specific gravity of the object, in
terms of the problems other pertinent parameters, that will keep the object afloat.
32. Describe, in words and equations, what is meant by a streamline, streakline, and pathline.
33. Under what conditions will a streamline, streakline, and pathline be identical, in general?
34. Experimentally, how would one produce a streamline, streakline, and pathline?
35. Describe analytically the difference between a streakline and pathline.
36. What is the definition of a fluid? What is the fundamental difference between fluids and solids? Give
an example of a substance that has similarities with both fluids and solids.
C. Wassgren
47
37. What is meant by a system? What is meant by a control volume?

38. What is meant by uniform? What is meant by steady? Give examples of velocity fields that are
uniform/non-uniform and steady/unsteady.
39. What is meant by an extensive property? Give an example of an extensive property. What is meant
by an intensive property? Give an example of an intensive property.
40. How is the dimension of a flow determined? Give an example of a 2D, unsteady velocity field. Is it
possible to have a 2D velocity field but a 3D temperature field?
41. What is meant by a uniform flow? Give an example of a velocity profile that is uniform in the xdirection but is unsteady.
C. Wassgren
48
Chapter 02:
Fluid Statics
1.
2.
3.
4.
Pressure Distribution in a Static Fluid

a.
Incompressible Fluids
b.
Compressible Fluids
Manometers
Forces on Submerged Surfaces
a.
Planar Surfaces
b.
Curved Surfaces
Pressure Distribution Due to Rigid Body Motion
C. Wassgren
Chapter 02: Fluid Statics
49
Chapter 03:
Integral Analysis
1.
2.
3.
4.
5.
6.
7.
8.
9.
Lagrangian and Eulerian Perspectives

The Reynolds Transport Theorem
Conservation of Mass
The Linear Momentum Equation
The Angular Momentum Equation
Basic Thermodynamic Definitions
Discussion of Energy, Work, and Heat
Conservation of Energy
The 2nd Law of Thermodynamics
C. Wassgren
Chapter 03: Integral Analysis
50
Last Updated: 25 Aug 2008
1.
Lagrangian and Eulerian Perspectives
There are two common ways to study a moving fluid:

1. Look at a particular location and observe how all the fluid passing that location behaves. This is
called the Eulerian point of view.
2. Look at a particular piece of fluid and observe how it behaves as it moves from location to location.
This is called the Lagrangian point of view.
Example:
Lets say we want to study migrating birds. We could either:
1. stand in a fixed spot and make measurements
as birds fly by (Eulerian point of view), or
2.
bird
tag some of the birds and make measurements

as they fly along (Lagrangian point of view).
radio transmitter
Lagrangian (aka Material, Substantial) Derivative
(Go to http://widget.ecn.purdue.edu/~meapplet for an interactive Java applet on this topic.)
If we follow a piece of fluid (Lagrangian viewpoint), how will some property of that particular piece of
fluid change with respect to time?
same piece of fluid
piece of fluid
time:
t0+t
location: x(t0+t)
time:
t0
location: x(t0)
Lets say were interested in looking at the time rate of change of temperature, T, that the particle observes
as it moves from location to location. The particle may experience a temperature change because the
temperature of the entire field of fluid may be changing with respect to time (i.e. the temperature field may
be unsteady). In addition, the temperature field may have spatial gradients (different temperatures at
different locations, i.e. non-uniform) so that as the particle moves from point to point it will experience a
change in temperature. Thus, there are two effects that can cause a time rate of change of temperature that
the particle experiences: unsteady effects, also known as local or Eulerian effects, and spatial gradient
effects, also known as convective effects. We can describe this in mathematical terms by writing the
temperature of the entire field as a function of time, t, and location, x:
T = T (t, x )
(3.1)
Note that the location of the fluid particle is a function of time: x=x(t) so that
T = T (t, x (t ))
(3.2)
Taking the time derivative of the temperature, expanding the location vector into its x, y, and z components,
and using the chain rule gives:
dT
T T dx T dy T dz
=
+
+
+
(3.3)
dt following a
t x
dt y
dt z
dt
fluid particle
=ux
=u y
=u z
Note that dx/dt, dy/dt, and dz/dt are the particle velocities ux, uy, and uz respectively. Writing this in a more
compact form:
C. Wassgren
51
DT T
T
T
T
=
+ ux
+ uy
+ uz
Dt
t
x
y
z
(3.4)
T
=
+ (u )T
t
The notation, D/Dt, indicating a Lagrangian (also sometimes referred to as the material or substantial)
derivative, has been used in Eqn. (3.4) to indicate that were following a particular piece of fluid. More
generally, we have:
D
+
u )()
() =
()
(

Dt
t

Lagrangian rate of
change (changes as we
follow a fluid particle)
local or Eulerian
rate of change (changes
due to unsteady effects)
convective rate
of change (changes due to
a change in particle position)
(3.5)
() + u x () + u y () + u z ()

t
x
y
z
where () represents any field quantity of interest.
=
Notes:
1.
2.
The Lagrangian derivatives in cylindrical and spherical coordinates are:
D
= + ur
+
+ uz
cylindrical:
Dt t
r r
z
u
D
spherical:
= + ur
+
+
Dt t
r r r sin
(3.6)
(3.7)
The acceleration experienced by a fluid particle is given by:

Du u
u
u
u
u
Cartesian:
=
+ (u ) u =
+ ux
+ uy
+ uz
Dt
t
t
x
y
z
ur
t
u
a =
t
u z
az =
t
u
ar = r
t
ar =
cylindrical:
spherical:
a =
a =
C. Wassgren
u
t
u
t
(3.8)
ur u ur
u
u2
+
+ uz r
r
r
z
r
u
u u
u
uu
+ ur + + u z + r
r
r
z
r
u z u u z
u z
+ ur
+
+ uz
r
r
z
u
u ur
u
ur 1 2
+ ur r +
+
u + u2
r
r r sin r
u u 1
u
u u
+ ur + +
+ ur u u2 cot
r
r r sin r
u u u
u u 1
+ ur
+
+
+ ur u + u u cot
r
r r sin r
+ ur
52
(3.9)
(3.10)
Example:
A fluid velocity field is given by:
u = (cy 2 )i + (cx 2 ) j
where c is a constant. Determine
a. the components of the acceleration and
b. the points in the flow field where the acceleration is zero.
SOLUTION:
The acceleration of a fluid element is given by:
Du u
u
u
a=
=
+ ux
+ uy
Dt t
x
y
where
u
= 0 (steady flow)
t
u
ux
= cy 2 2cxj = 2c 2 xy 2 j
x
u
uy
= cx 2 2cyi = 2c 2 x 2 yi
y
a = 2c 2 x 2 yi + 2c 2 xy 2 j
( )(
( )(
(3.11)
(3.12)
Set the acceleration equal to zero.

a = 0 = 2c 2 x 2 yi + 2c 2 xy 2 j
either x = 0 or y = 0 (This is the locus of points where the total acceleration is zero.)
C. Wassgren
53
(3.13)
Example:
A fluid velocity field is given by:
u = 2te x
Will a fluid particle accelerate in this flow? Why?
SOLUTION:
The acceleration is given by:
Du u
u
u
u
a=
=
+ ux
+ uy
+ uz
Dt
t
x
y
z
= 2 e x
2t
=0
=0
=0
Hence, for the given flow:

a = 2e x Yes, fluid particles will accelerate due to the local (or Eulerian) derivative.
Example:
Now consider the following flow:
u = xe x
Will a fluid particle accelerate in this flow? Why?
SOLUTION:
The acceleration is given by:
Du u
u
u
u
a=
=
+ ux
+ uy
+ uz

Dt
t
x
y
z

=0
= e x
=0
=0
Hence, for the given flow:

a = xe x Yes, fluid particles will accelerate due to the convective derivative.
C. Wassgren
54
Example:
The market price, P (in dollars), of used cars of a certain model is found to be:
P = $1000 + ( $0.02 / mile ) x ( $2 / day ) t
where x is the distance in miles west of Detroit, MI and t is the time in days. If a car of this model is driven
from Detroit at t = 0 towards the west at a rate of 400 miles per day, determine:
a. whether the value of the car is increasing or decreasing, and
b. how much of this change is due to depreciation and how much is due to moving into a better market.
SOLUTION:
To determine if the value of the car is decreasing, take the Lagrangian derivative of the market price.
DP P
P
=
+u
= ( $2/day ) + ( 400 miles/day )( $0.02/mile ) (where u is the speed of the car) (3.14)

Dt
t
x
= $8 / day
DP
= $6/day Hence, the value of the car is increasing.
Dt
The car depreciates at a rate of -$2/day (this is P/t). The change in the cars value increases at a rate of
$8/day due to moving into a different market (this is uP/x).
C. Wassgren
55
2.
Reynolds Transport Theorem (RTT)
Recall that we can look at the behavior of small pieces of fluid in two ways: the Eulerian perspective or the
Lagrangian perspective. Often were interested in the behavior of an entire system of fluid (many pieces of
fluid) rather than just an individual piece. How do we analyze this situation? We can use Eulerian and
Lagrangian approaches for analyzing a macroscopic amount of fluid but we need to first develop an
important tool called the Reynolds Transport Theorem.
Why do we want to do this? It turns out that the behavior of fluids (most substances in fact) can be
described in terms of a few fundamental laws. These laws include:
conservation of mass (COM),
Newtons 2nd Law,
the angular momentum principle,
conservation of energy (COE, aka the First Law of Thermodynamics), and
the second law of thermodynamics.
These laws are typically easiest to apply to a particular system of fluid particles (Lagrangian perspective).
However, the Lagrangian forms of the laws are typically difficult to use in practical applications since we
cant easily keep track of many individual bits of fluid. Its much easier to apply the laws to a particular
volume in space instead (referred to as a control volume, an Eulerian perspective). For example, tracking
the behavior of individual bits of gas flowing through a rocket nozzle would be difficult. Its much easier
to just look at the behavior of the gas flowing into, out of, and within the volume enclosed by the rocket
nozzle. The Reynolds Transport Theorem is a tool that will allow us to convert from a system point of
view (Lagrangian) to a control volume point of view (Eulerian).
Lets consider a system of fluid particles that is coincident (occupying the same region in space) as our
control volume (CV) at some time, t:
at time t
system
CV
the fluid is moving
At some later time, t+t, the system may have moved relative to the CV.
at time t+t
system
CV
C. Wassgren
56
Let B be some transportable property (i.e., some property that can be transported from one location to
another, e.g. mass, momentum, energy) and be the corresponding amount of B per unit mass, i.e.:
small volume, dV,
containing: dBsys=dV
Bsys =
dV
Vsys
BCV =
(3.15)
dV
CV
system with total volume,

Vsys, containing Bsys
where Bsys and BCV refer to the total amount of B in

the system and control volume, respectively.
Note that at time, t, the total amounts of B in the system and control volume are equal since the system and
CV are coincident:
Bsys ( t ) = BCV ( t )
(3.16)
However, at time, t+t, the system and CV no longer occupy the same region in space so that, in general,
Bsys(t+t) BCV(t+t).
Note that B may be changing with time so that,
in general, Bsys(t+t) Bsys(t).
at time t+t
Bout(t+t)
In the figure at the left, Bout is the amount of B

that has left the CV and Bin is the amount of B
that has entered the CV.
system
Bin(t+t)
CV
Utilizing the figure shown above, we see that:

BCV ( t + t ) = Bsys ( t + t ) Bout ( t + t ) + Bin ( t + t )
(3.17)
Subtracting Bsys(t) from both sides and dividing through by t gives:

BCV ( t + t ) Bsys ( t ) Bsys ( t + t ) Bsys ( t ) Bout ( t + t ) Bin ( t + t )
=
+
(3.18)
t
t
t
t
Now lets substitute BCV(t) = Bsys(t) on the left hand side, subtract Bout(t)/t and Bin(t)/t on the right-hand
side (note that Bout(t)=Bin(t)=0), and then take the limit of the entire equation as t0:
BCV ( t + t ) BCV ( t )
lim
=
t 0
t
Bsys ( t + t ) Bsys ( t )
Bout ( t + t ) Bout ( t )
Bin ( t + t ) Bin ( t )
lim
lim
+ lim
(3.19)
t
0
t
0
t 0
t
t
t
DBsys dBout dBin
dB
CV =
+
dt
Dt
dt
dt
C. Wassgren
57
Note that the D/Dt notation has been used to signify that the first term on the right hand side of Eqn. (3.19)
represents the time rate of change as we follow a particular system of fluid (Lagrangian perspective).
Re-arranging the equation and substituting in for BCV and Bsys using Eqn. (3.15):
d (B B )
D
d
out
in
dV = dV +
(3.20)
dt
Dt
dt
CV
Vsys
The last term on the right hand side of Eqn. (3.20) represents the net rate at which B is leaving the control
volume through the control surface (CS). Lets examine this term more closely by zooming in on a small
piece of the control surface and observing how much B leaves through this surface in time t.
The component of the fluid velocity out of the
control volume through surface, dA, is given by:
u rel n
where urel=usys-uCS is the velocity of the fluid
small piece of control
relative to the control surface. The volume of fluid
surface, dA, with outward leaving through surface dA in time t is then:
point normal vector, n
dV = ( u rel n ) tdA = ( u rel dA ) t
(urel n )t
urel
dA
Thus, the volumetric flowrate, dQ, (volume per

unit time) through surface dA is given by:
dQ = u rel dA
(3.21)
system
CV
Now use Eqn. (3.21) to write the net rate at which B leaves the control volume:
d ( Bout Bin )
= dQ = ( u rel dA )
dt
CS
CS
Combining Eqn. (3.22) with Eqn. (3.20) gives:
D
d
dV = dV +
( u rel dA )
dt
Dt
CS
CV
Vsys

net rate at which B leaves
rate of increase of
B within the system
rate of increase of
B within the CV
(3.22)
(3.23)
the CV through the CS
The Reynolds Transport Theorem!
C. Wassgren
58
3.
Conservation of Mass (COM)
In words and in mathematical terms, COM for a system is:
D
The mass of a system remains constant.
dV = 0
(3.24)
Dt
Vsystem

mass
of the system
where D/Dt is the Lagrangian derivative (implying that were using the rate of change as we follow the
system), V is the volume, and is the density. Using the Reynolds Transport Theorem, Eqn. (3.24) can be
converted into an expression for a control volume:
D
d
dV = dV + ( u rel dA ) = 0
(3.25)
dt
Dt
V
CV
CS
system
d
dV +
dt
CV

rate of increase of
mass inside the CV
( u
rel
dA )
=0
COM for a CV
(3.26)
CS

net rate at which mass leaves
Helpful Hints:
1.
Carefully draw your control volume. Dont neglect to draw a control volume or draw a control
volume and then use a different one.
2.
Make sure you understand what each term in conservation of mass represents.
3.
Carefully evaluate the dot product in the mass flux term.
4.
You must integrate the terms in conservation of mass when the density or velocity are not uniform.
Lets consider a few examples to see how COM for a CV is applied.
C. Wassgren
59
Example:
Calculate the mass flux through the control surface shown below. Assume a unit depth into the page.
control surface
SOLUTION:
dA = Rd
The mass flux through the surface is given by:
m =
= 2
u rel dA =
()
Vi cos i + sin j ( Rd )

= u rel
= 2
CS
cos d = VR sin
= 2
= dA
=n
= 2
= VR
n = cos i + sin j
= 2 VR
m = VD
We could have also determined the mass flux by noticing that any mass passing through the curved control
surface must also pass through a vertical control surface as shown below.
n = i
V
y=R
m =
CS
u rel dA =
()
Vi i ( dy ) = 2 VR = VD (The same answer as before!)
y = R
C. Wassgren
=u rel
=n
= dA
60
Example:
Consider the flow of an incompressible fluid between two parallel plates separated by a distance 2H. If the
velocity profile is given by:
y2
u = uc 1 2
H
where uc is the centerline velocity, determine the average velocity of the flow, u . Assume the depth into
the page is w.
H
H
SOLUTION:
The volumetric flow rate using the average velocity profile must give the same volumetric flow rate using
the real velocity profile.
y =+ H
= dA
2
u(y)
4
dy
Qreal = u dA = dQ =
uc 1 y
dyw
= 3 uc wH
(3.27)
2
H

A
A
y = H
= dQ
The velocity, u(y), is nearly constant
over the small distance dy so we can
write the volumetric flowrate over this
small area as dQ = u(y)dA = u(y)(dyw).
Qaverage = u dA = u ( 2 Hw ) (There is no need to integrate since the velocity is uniform over y.) (3.28)
A
Qreal = Qaverage
4
u wH
3 c
= u ( 2 Hw )
(3.29)
u = 23 uc
C. Wassgren
(3.30)
61
Example
An incompressible flow in a pipe has a velocity profile given by:
r2
u (r ) = u c 1 2
R
where uc is the centerline velocity and R is the pipe radius. Determine the average velocity in the pipe.
SOLUTION:
The volumetric flow rate using the average velocity profile must give the same volumetric flow rate using
the real velocity profile.
dr
dA
r=R

2
2
dA
=
2rdr
1
Qreal = u dA = dQ =
uc 1 r 2 ( 2 rdr ) = 2 uc R
(3.31)
R

A
A
r =0
r
= dQ
The velocity, u(r), is nearly constant

over the small annulus with radius dr so
we can write the volumetric flow rate
over this small area as dQ = u(r)dA =
u(r)(2rdr).
Qaverage = u dA = u R 2
(There is no need to integrate since the velocity is uniform over r.) (3.32)
Qreal = Qaverage
1
uc R 2
2
= u R2
(3.33)
u = 12 uc
C. Wassgren
(3.34)
62
Example:
Water enters a cylindrical tank through two pipes at volumetric flow rates of Q1 and Q2. If the level in the
tank remains constant, calculate the average velocity of the flow leaving the tank through a pipe with an
area, A3.
Q1
Q2
h=constant
V3 = ?
A3
SOLUTION:
Apply conservation of mass to the fixed control volume shown below.
Q1
Q2
h=constant
V3 = ?
A3
d
dt
where
d
dt
dV + u
CV
rel
dA = 0
(3.35)
CS
dV = 0
(steady flow, the mass in the control volume isnt changing with time)
CV
rel
dA = Q2 Q1 + V3 A3
CS
Substitute and re-arrange.

Q2 Q1 + V3 A3 = 0
V3 =
Q1 + Q2
A3
C. Wassgren
(3.36)
63
Example:
Water enters a cylindrical tank with diameter, D, through two pipes at volumetric flow rates of Q1 and Q2
and leaves through a pipe with area, A3, with an average velocity, V 3 . The level in the tank, h, does not
remain constant. Determine the time rate of change of the level in the tank.
D
Q1
Q2
hconstant
V3
A3
SOLUTION:
Apply conservation of mass to a control volume that deforms to follow the free surface of the liquid.
D
Q1
Q2
hconstant
V3
A3
d
dt
dV + u
CV
rel
dA = 0
(3.37)
CS
where
d
dt
dV =
CV
d D2
dh D 2
h
=
dt
4
dt 4

= M CV
rel
dA = Q2 Q1 + V3 A3
CS
dh D 2
Q2 Q1 + V3 A3 = 0
dt 4
dh Q2 + Q1 V3 A3
=
dt
D2
4
C. Wassgren
(3.38)
64
We could have also chosen a fixed control volume through which the free surface moves. Using this type
of control volume, conservation of mass is given by:
d
dV + u rel dA = 0
(3.39)
dt
where
d
dt
CV
CS
dV = 0
(the mass of fluid in the fixed control volume remains constant)
CV
=Vtop
CS

dh D 2
u rel dA = Q2 Q1 + V3 A3 +
dt
4

m top

dh Q2 + Q1 V3 A3
=
2
dt
D 4
dh Q2 + Q1 V3 A3
=
(This is the same answer as before!)
dt
D2
4
C. Wassgren
65
(3.40)
Example:
A spherical balloon is filled through an area, A1, with air flowing at velocity, V1, and constant density, 1.
The radius of the balloon, R(t), can change with time, t. The average density within the balloon at any
given time is b(t). Determine the relationship between the rate of change of the density within the balloon
and the rest of the variables.
R(t)
1, V1, A1
b
SOLUTION:
Apply conservation of mass to a control volume that deforms to follow the interior surface of the balloon.
R(t)
1, V1, A1
b
d
dt
where
d
dt
dV + u
CV
rel
dA = 0
(3.41)
CS
R = R
dV = dt
3
3
b
CV
d b
dR
+ 4b R 2
dt
dt
= M CV
rel
dA = 1V1 A1
CS
Substitute and simplify.

d
4
dR
R3 b + 4b R 2
1V1 A1 = 0
3
dt
dt
dR
1V1 A1 4b R 2
d b
dt
=
4 R3
dt
3
C. Wassgren
(3.42)
66
Example:
A box with a hole of area, A, moves to the right with velocity, ubox, through an incompressible fluid as
shown in the figure. If the fluid has a velocity of ufluid which is at an angle, , to the vertical, determine
how long it will take to fill the box with fluid. Assume the box volume is Vbox and that it is initially empty.
ufluid
area, A
ubox
volume, Vbox
SOLUTION:
Apply conservation of mass to a control volume fixed to the interior of the box. Change our frame of
reference so the box appears stationary.
ufluid
area, A
ubox
y
x
volume, Vbox
d
dt
dV + u
CV
rel
dA = 0
(3.43)
CS
where
d
dt
dV =
CV
d ( VCV )
dt
dVCV
dt
u rel dA = ubox i ufluid sin i ufluid cos j

Ai = ( ubox + ufluid sin ) A
= dA
=
CS
u rel

dV
CV ( ubox + ufluid sin ) A = 0
dt
dVCV
= ( ubox + ufluid sin ) A
dt
VCV =VCV
t =T
dVCV = ( ubox + ufluid sin ) A
VCV = 0
dt
(Note that ubox, ufluid, , and A dont change with time.)
t =0
VCV = ( ubox + ufluid sin ) AT

T =
( ubox
VCV
+ ufluid sin ) A
C. Wassgren
(3.44)
67
Example:
Determine the rate at which fluid mass collects inside the room shown below in terms of , V1, A1, V2, A2,
Vc, R, and . Assume the fluid moving through the system is incompressible.
2R
V1
room
r2
V = Vc 1 2
R
A1
A2
V2
SOLUTION:
Apply conservation of mass to a control volume fixed to the interior of the room.
2R
V1
room
r2
V = Vc 1 2
R
A1
y
A2
d
dt
where
d
dt
dV + u
CV
dV =
CV
rel
V2
dA = 0
(3.45)
CS
dM CV
dt
r=R
u rel dA = V1i A1i + V2 sin i cos j A2 j +
Vi dAi
r =0
CS
r =R
= V1 A1 + V2 A2 cos +
V (1 r
c
r =0
R2
) 2 rdr
= V1 A1 + V2 A2 cos + 2 Vc R 2
dM CV
V1 A1 + V2 A2 cos + 2 Vc R 2 = 0
dt
dM CV
= V1 A1 V2 A2 cos 2 Vc R 2
dt
C. Wassgren
(3.46)
68
Example:
Construct from first principles an equation for the conservation of mass governing the planar flow (in the xy
plane) of a compressible liquid lying on a flat horizontal plane. The depth, h(x,t), is a function of position,
x, and time, t. Assume that the velocity of the fluid in the positive x-direction, u(x,t), is independent of y.
Also assume that the wavelength of the wave is much greater than the wave amplitude so that the
horizontal velocities are much greater than the vertical velocities.
free surface
liquid
h(x,t)
u(x,t)
SOLUTION:
Apply conservation of mass to the fixed control volume shown below. Assume a unit depth into the page.
h(x,t)
Note that the mass flow
rate at the center of the
control volume is uh.
u(x,t)
y
x
dx
d
dt
where
d
dt
dV + u
CV
dA = 0
(3.47)
CS
dV = t (hdx) = t ( h ) dx
= M CV
CV
u
CS
rel
rel

dA = ( uh ) + ( uh ) 12 dx + ( uh ) + ( uh ) 12 dx = ( uh ) dx
x
x

x
= m left
(Note that: m left/ = m center

right
= m right
m
+ center 12 dx
x
where m center = uh )
( h ) dx + ( uh ) dx = 0
t
x
( h ) + ( uh ) = 0
t
x
(3.48)
If the fluid is incompressible, then Eqn. (3.48) simplifies to:

h
+ ( uh ) = 0
t x
C. Wassgren
69
(3.49)
4.
Linear Momentum Equation (LME)
In this section well consider Newtons 2nd law applied to a control volume of fluid. Recall that linear
momentum is a vector quantity (it has both magnitude and direction) and is given by the mass*velocity. In
words and in mathematical terms, Newtons 2nd Law for a system is:
The rate of change of a systems linear momentum is equal to the net force acting on the system.
system
D
u XYZ dV = Fon system
Dt
Vsystem
uXYZdV
Y

LM of a small
LM of the system
Z
X
piece of fluid
(3.50)
system), V is the volume, and is the density. The quantity, uXYZ, represents the velocity of a small piece
of fluid in the system with respect to an inertial (aka non-accelerating) frame of reference XYZ (recall
that Newtons 2nd law holds strictly for inertial frames of reference). Note that a frame of reference moving
at a constant velocity in a straight line is non-accelerating and thus is inertial.
The term, Fon system, represents the net forces acting on the system. These forces can be of two different
types. The first are body forces, Fbody, and the second are surface forces, Fsurface. Body forces are those
forces that act on each piece of fluid in the system. Examples include gravitational and electro-magnetic
forces. Surface forces are those forces acting only at the surface of the system. Examples of surface forces
include pressure and shear forces. Thus,
Fon system = Fbody on system + Fsurface on system
(3.51)
Using the Reynolds Transport Theorem to convert the left hand side of Eqn. (3.24) from a system point of
view to an expression for a control volume gives:
D
d
u XYZ dV =
u XYZ dV + u XYZ ( u rel dA )
(3.52)
Dt
dt
V
CV
CS
system
Since the Reynolds Transport Theorem is applied to a coincident system and control volume, the forces
acting on the system will also act on the control volume. Thus, the LME for a CV becomes:
d
Fbody on CV + Fsurface on CV =
u XYZ dV + u XYZ ( u rel dA )

dt
CV
net surface force
net body force
CS

acting on the CV
acting on the CV
C. Wassgren
rate of increase of
LM inside the CV
LME for a CV
(3.53)
net rate at which LM leaves

70
Notes:
1. Recall that the LME is a vector expression. There are actually three equations built into Eqn. (3.53).
For example, in a rectangular coordinate system (Cartesian coordinates) we have:
d
FX ,body on CV + FX ,surface on CV =
u X dV + u X ( u rel dA )
dt
FY ,body on CV + FY ,surface on CV =
FZ ,body on CV + FZ ,surface on CV
d
dt
d
=
dt
CV
CS
Y dV
CV
( u rel dA )
(3.54)
CS
Z dV
CV
( u rel dA )
CS
2. It is important to distinguish between the two velocities uXYZ and urel. The velocity, uXYZ, represents the
fluid velocity with respect to an inertial frame of reference XYZ (e.g. a frame of reference fixed with
respect to the stars or moving at a constant velocity in a straight line). The velocity, urel, is the velocity
of the fluid with respect to the control surface, i.e. urel = ufluid, XYZ uCS, XYZ.
3. So far weve only discussed the LME for inertial (aka non-accelerating) frames of reference. We can
also apply the LME to non-inertial (aka accelerating) frames of reference but we need to add additional
acceleration terms. Well consider accelerating frames of reference a little later.
4. In order to avoid mistakes, be sure to do the following when applying the LME:
a. draw the CV that the LME is being applied to,
b. indicate the frame of reference that is being used,
c. state your significant assumptions,
d. draw a free body diagram (FBD) of the pertinent forces, and
e. write down the LME and then indicate the value of each term in the LME
While these things may seem trivial and unnecessary, writing them down in a clear and concise manner
can greatly reduce the likelihood of mistakes.
Helpful Hints:
1.
Carefully draw your control volume. Dont neglect to draw a control volume or draw a control
volume and then use a different one.
2.
Clearly indicate your frame of reference. Dont neglect to indicate a frame of reference or draw a
frame of reference and then use a different one.
3.
Carefully draw free body diagrams for your control volume. This will facilitate your evaluation of
forces in the linear momentum equation.
4.
Make sure you understand what each term in the linear momentum equation represents.
5.
Carefully evaluate the dot product in the mass flux term.
6.
Be sure to use the correct velocity components in the CV and CS integral terms.
7.
You must integrate the terms in the linear momentum equation when the density or velocity are not
uniform.
8.
Dont forget to include pressure and shear forces in the surface force term.
9.
Dont forget to include the weight of everything inside the control volume when gravitational body
forces are significant.
Now lets consider some examples to see how we apply the LME to a CV.
C. Wassgren
71
Example:
A jet of water is deflected by a vane mounted on a cart. The water jet has an area, A, everywhere and is
turned by an angle with respect to the horizontal. The pressure everywhere within the jet is atmospheric.
The incoming jet velocity with respect to the ground (axes XY) is Vjet. The cart has mass M. Determine:
a. the force components, Fx and Fy, required to hold the cart stationary, and
b. the horizontal force component, Fx, if the cart moves to the right at the constant velocity, Vcart
(Vcart<Vjet)
A
Vjet
Vcart
Y
Fx
X
Fy
SOLUTION:
Part (a):
Apply conservation of mass and the linear momentum equation to a control volume surrounding the cart.
Use an inertial frame of reference fixed to the ground (XY).
Vout (this velocity is currently an
unknown quantity)
A
Vjet
Fx
X
Fy
First apply conservation of mass to the control volume to determine Vout.

d
dV + u rel dA = 0
dt
CV
CS
C. Wassgren
72
(3.55)
where
d
dt
dV = 0
(the mass within the control volume doesnt change)
CV
= u rel
=u rel = A
=A

u rel dA = Vjet i Ai + Vout cos i + sin j A cos i + sin j
CS

) (
left side
right side
= V jet A + Vout A cos + sin

2
=1
= V jet A + Vout A
(Note that the jet area remains constant.)
V jet A + Vout A = 0
Vout = V jet
(3.56)
Now apply the linear momentum equation in the X-direction:

d
u X dV + u X ( u rel dA ) = FB , X + FS , X
dt
where
d
dt
CV
CS
(3.57)
dV = 0 (the momentum within the control volume doesnt change with time)
CV
= u rel
=u X
= u X = u rel = A
=A

u X ( u rel dA ) = Vjet Vjet i Ai + V jet cos V jet cos i + sin j A cos i + sin j
CS

( )
left side
) (
right side
2
= Vjet
A + Vj2et A cos cos 2 + sin 2

=1
2
Vjet
A
( cos 1)
FB , X = 0 (no body forces in the x-direction)

FS , X = Fx
(all of the pressure forces cancel out)

2
Vjet
A ( cos 1) = Fx
2
Fx = Vjet
A (1 cos )
C. Wassgren
(3.58)
73
Now look at the Y-direction:

d
uY dV + uY ( u rel dA ) = FB ,Y + FS ,Y
dt
where
d
dt
CV
CS
Y dV
(3.59)
= 0 (the momentum within the control volume doesnt change with time)
CV
=u rel
uY
=A

=

uY ( u rel dA ) = Vjet sin Vjet cos i + sin j A cos i + sin j
CS

) (
right side
2
A sin
V jet
cos + sin )
(
2
=1
2
= V jet
A sin
FB ,Y = Mg (assume that the fluid weight in the CV is negligible compared to the cart weight)
FS ,Y = Fy

2
Vjet
A sin = Mg + Fy
2
Fy = Vjet
A sin + Mg
C. Wassgren
(3.60)
74
Part (b):
Apply the linear momentum equation to a control volume surrounding the cart. Use a frame of reference
fixed to the cart (xy). Note that this is an inertial frame of reference since the cart moves in a straight line at
a constant speed. In addition, in this frame of reference, the cart appears stationary and the jet velocity at
the left is equal to Vjet-Vcart.
Vout (this velocity is currently an
unknown quantity)
A
Vjet - Vcart
y
x
Fx
Fy
First apply conservation of mass to the control volume to determine Vout
d
dV + u rel dA = 0
dt
where
d
dt
CV
CS
dV = 0
(3.61)
(the mass within the control volume doesnt change)
CV
= u rel
u rel
=A
= A

=

u rel dA = Vjet Vcart i Ai + Vout cos i + sin j A cos i + sin j
CS

) (
left side
right side
= V jet Vcart A + Vout A cos 2 + sin 2

=1
= V jet Vcart A + Vout A

(Note that the jet area remains constant.)
Vjet Vcart A + Vout A = 0

Vout = Vjet Vcart
(3.62)
Now apply the linear momentum equation in the x-direction:

d
u x dV + u x ( u rel dA ) = FB , x + FS , x
dt
where
d
dt
CV
CS
u dV = 0
x
(3.63)
(the momentum within the control volume doesnt change with time)
CV
C. Wassgren
75
=u rel
=u x
=u rel
=u X
A
= A

=

u x ( u rel dA ) = V jet Vcart V jet Vcart i Ai + V jet Vcart cos V jet Vcart cos i + sin j A cos i + sin j
CS

) (
)(
left side
= V jet Vcart
right side
) (
A + V jet Vcart
A cos cos 2 + sin 2

=1
= V jet Vcart
A ( cos 1)
FB, x = 0 (no body forces in the x-direction)

FS , x = Fx (all of the pressure forces cancel out)
(Vjet Vcart ) A ( cos 1) = Fx

2
Fx = Vjet Vcart
2
) A (1 cos )
(3.64)
Now solve the problem using an inertial frame of reference fixed to the ground (frame XY). From Eqn.
(3.62) we know that using a frame of reference fixed to the cart, the jet velocity on the right hand side is:
V
= V V
cos i + sin j
(3.65)
out,
relative to cart
jet
cart
)(
Hence, relative to the ground the jet velocity on the right hand side is:
V
=V
+ V = V V
cos i + sin j + V i
out,
relative to
ground
out,
relative to
cart
cart
jet
cart
)(
(3.66)
cart
Now consider the linear momentum equation in the X direction.

d
dt
where
d
dt
CV
CS
(3.67)
dV = 0 (the momentum within the control volume doesnt change with time)
CV
=u
= u rel
= u rel
X
=A
= A

u X ( u rel d A ) = V jet V jet Vcart i A i + V jet Vcart cos + Vcart V jet Vcart cos i + sin j A cos i + sin j
CS

=u X
)(
left side
) (
right side
= V jet V jet Vcart A + V jet Vcart cos + Vcart V jet Vcart A cos 2 + sin 2

=1
2
2
2
= V jet
+ V jetVcart + V jet Vcart cos + VcartV jet Vcart
A
= V jet Vcart
= V jet Vcart
( cos 1) A
cos V jet Vcart
FB, X = 0 (no body forces in the x-direction)

FS , X = Fx
(Vjet Vcart ) A ( cos 1) = Fx

2
Fx = Vjet Vcart
2
) A (1 cos )
(Same answer as before!)
(3.68)
Note that using a frame of reference that is fixed to the control volume is easier than using one fixed to the
ground. This is often the case.
C. Wassgren
76
Example:
A fluid enters a horizontal, circular cross-sectioned, sudden contraction nozzle. At section 1, which has
diameter D1, the velocity is uniformly distributed and equal to V1. The gage pressure at 1 is p1. The fluid
exits into the atmosphere at section 2, with diameter D2. Determine the force in the bolts required to hold
the contraction in place. Neglect gravitational effects and assume that the fluid is inviscid.
bolts
D2
V1
D1
p1
atmosphere
SOLUTION:
Apply the linear momentum equation in the X-direction to the fixed control volume shown below.
bolts
D2
V1
D1
p1
Fbolts
The CS cuts through the bolts. So that Fbolts is the

force one side of the bolts applies to the other side.
d
dt
where
d
dt
dV + u X ( u rel dA ) = FB , X + FS , X
CV
atmosphere
(3.69)
Cs
dV = 0 (steady flow)
CV
=A
= A

=u
u
=u X =
rel D 2
rel D 2
D12
D22
2
u X ( u rel dA ) = V1 V1i 1 i + V2 V2 i
i = V12
+ V22
4
4
4
4
Cs
(Note that V2 is unknown for now.)

FB, X = 0
=u X
FS , X = p1,gage
D12
+ Fbolts
4
(Note that p2,gage = 0 since p2,abs = patm. We could have also worked the problem using absolute
pressures everywhere. The pressure force on the left hand side would be p1,absD12/4 and the pressure
force on the right hand side would be patmD12/4 (note that the diameter is D1 and not D2).)
C. Wassgren
77

D12
D22
D12
V12
+ V22
= p1,gage
+ Fbolts
4
4
4
D12
D22
D12
Fbolts = V12
+ V22
p1,gage
4
4
4
(3.70)
To determine V2, apply conservation of mass to the same control volume.

d
dV + u rel dA = 0
dt
where
d
dt
CV
Cs
(3.71)
dV = 0
CV
u rel dA = V1
D12
4
Cs
+ V2
D22
4
V1
D12
4
+ V2
D
V2 = V1 1
D2
D22
4
=0
(3.72)
Substitute Eqn. (3.72) into Eqn. (3.70) and simplify.
Fbolts =
V12
D12
4
V12
D1 D22
D12
p1,gage
4
D2 4
D12
(3.73)
1 p1,gage
4 D2
4
Note that Fbolts was assumed to be positive when acting in the +X direction (causing compression in the
bolts). If Fbolts < 0 then the bolts will be in tension.
Fbolts =
V12
2
D12 D1
C. Wassgren
78
Example:
Water is sprayed radially outward through 180 as shown in the figure. The jet sheet is in the horizontal
plane and has thickness, H. If the jet volumetric flow rate is Q, determine the resultant horizontal
anchoring force required to hold the nozzle stationary.
Q
SOLUTION:
Apply the linear momentum equation in the X direction to the fixed control volume shown below.
Y
Fy
Fx
Fx
Q
H
R
side view
top view
d
dt
where
d
dt
dV +
CV
( u rel dA ) = FB , X
+ FS , X
(3.74)
CS
CV
uX
= dA
=

=

2
u X ( u rel dA ) =
(V sin ) V Rd H = V RH sin d = V 2 RH cos 0
CS
=0
=0
= V 2 RH ( 1 1)
= 2 V 2 RH
(Note that there is no X-momentum at the control volume inlet. Also, V is an unknown quantity at the
moment.)
FB, X = 0
FS , X = Fx
(All of the pressure forces cancel and only the anchoring force remains.)
C. Wassgren
79
Substitute.
Fx = 2 V 2 RH
(3.75)
To determine V, apply conservation of mass to the same control volume.

d
dV + u rel dA = 0
dt
where
d
dt
CV
CS
dV = 0
(3.76)
(steady flow)
CV
= dA

u rel dA =
Q +

inlet
CS
V Rd H = Q + V RH
= 0

outlet

Q + V RH = 0
V=
Q
RH
(3.77)
Substitute Eqn. (3.77) into Eqn. (3.75).

2
Q
Fx = 2
RH
RH
(3.78)
Note that Fy = 0 due to symmetry.
C. Wassgren
80
Example:
A variable mesh screen produces a linear and axi-symmetric velocity profile as shown in the figure. The
static pressure upstream and downstream of the screen are p1 and p2 respectively (and are uniformly
distributed). If the flow upstream of the mesh is uniformly distributed and equal to V1, determine the force
the mesh screen exerts on the fluid. Assume that the pipe wall does not exert any force on the fluid.
variable mesh
screen
Section 1
Section 2
p1
2R
p2
V1
SOLUTION:
First, note that the linear velocity profile at the outlet may be written as:
r
V = Vmax
(3.79)
R
where Vmax is the flow velocity at r = R. Now apply conservation of mass to the fixed control volume
shown below to find Vmax in terms of the upstream properties.
d
dV + u rel dA = 0
(3.80)
dt
where
d
dt
CV
CV
dV = 0
(steady flow)
dr
CV
=V

dA
r=R

=

r
2
u rel dA = V1 R +
Vmax ( 2 rdr )

R
CV
r
=0
left side

r
dA = 2rdr
right side
= V1 R 2 + 23 Vmax R 2
V1 R 2 + 23 Vmax R 2 = 0
Vmax = 32 V1
Section 1
variable mesh
screen
(3.81)
Section 2
F
p1
2R
X
p2
V1
C. Wassgren
81
The control volume weaves in

and out of the mesh so that
the mesh is not part of the
control volume, and instead
exerts a force, F, on the
control volume.
Now apply the linear momentum equation in the X-direction to the same control volume.
d
u X dV + u X ( u rel dA ) = FB, X + FS , X
dt
where
d
dt
CV
CV
(3.82)
dV = 0
CV
dA
2

=

3 r
V
2 1 ( rdr )
R
r =0

r=R
u X ( u rel dA ) = V12 R 2 +

CV
left side
right side
V12 R 2
18 V12 R 2
9 2 2
V R
8 1
FB, X = 0
FS , X = F + p1 R 2 p2 R 2
18 V12 R 2 = F + p1 R 2 p2 R 2
F = 8 V12 R 2 + ( p1 p2 ) R 2
(3.83)
This is the force the mesh applies to the control volume (i.e. the fluid). The fluid applies an equal and
opposite force to the mesh.
C. Wassgren
82
Example:
Incompressible fluid of negligible viscosity is pumped, at total volume flow rate Q, through a porous
surface into the small gap between closely spaced parallel plates as shown. The fluid has only horizontal
motion in the gap. Assume uniform flow across any vertical section. Obtain an expression for the pressure
variation as a function of x.
L
V(x)
Assume a depth w
into the page.
SOLUTION:
Apply conservation of mass to the following differential control volume.
L
d
dV + CS u rel dA = 0
dt CV
where
d
dt CV
d
d
Q
dA = Vh ( w ) + ( Vhw ) ( 12 dx ) + Vhw + ( Vhw ) ( 12 dx )

wdx
dx
dx
Lw
CS
d
Q
dV
Q
wdx = hw
dx
wdx
= ( Vhw ) dx
dx
Lw
dx
Lw
Substituting and simplifying gives:
dV
Q
hw
dx =
wdx
dx
Lw
dV
Q
=
dx Lhw
V =V
x=x
Q
dV
=
dx (V(x = 0) = 0 due to symmetry.)
Lhw
V =0
x=0
rel
Vhw x
Q x
=
or V =

Q
L
hw L
C. Wassgren
(3.84)
(3.85)
83
Now apply the linear momentum equation in the X-direction to the same control volume.
d
u X dV + CS u X ( urel dA ) = FBX + FSX
dt CV
where
d
u X dV = 0 (steady flow)
dt CV
d
d
dA ) = V 2 hw + ( V 2 hw ) ( 12 dx ) + V 2 hw + ( V 2 hw ) ( 12 dx )
dx
dx
CS
d
dV
= ( V 2 hw ) dx = 2 V
hwdx
dx
dx
(Note that the flux of mass from the porous surface has no X-momentum.)
FBX = 0
u ( u
X
rel
d
d
FSX = phw + ( phw ) ( 12 dx ) phw + ( phw ) ( 12 dx )

dx
dx
d
dp
= ( phw ) dx = hwdx
dx
dx
dV
dp
2 V
hwdx = hwdx
dx
dx
dV
dp
2 V
=
dx
dx
Substituting Eqns. (3.84) and (3.85) gives:
2
dp
Q
2
x=
dx
Lhw
p = patm
p= p
Q
dp = 2
Lhw
2 x = 12 L
xdx
x= x
Q 1 2
2
patm p =
(4 L x )
Lhw
2
2
Q 1 x
p patm =
4
hw L
p patm
( hw)
1 x

4 L
C. Wassgren
(3.86)
84
The LME for Non-Inertial Frames of Reference

Recall that Newtons 2nd law holds strictly for inertial (non-accelerating) frames of reference. Now lets
consider frames of reference that are non-inertial (accelerating). First lets examine how we can describe
the motion of a particle in an accelerating frame of reference, call it frame xyz, in terms of a nonaccelerating frame of reference, call it frame XYZ.
particle
y
rxyz
xyz/XYZ
rXYZ
rxyz/XYZ
X
The position of a particle in frame XYZ is given by rXYZ and in frame xyz the particles position is given by
rxyz. The two position vectors are related by the position vector of the origin of frame xyz in frame XYZ,
rxyz/XYZ:
rXYZ = rxyz / XYZ + rxyz
(3.87)
The velocity of the particle in frame XYZ can be found by taking the time derivative of the position vector,
rXYZ, with respect to frame XYZ (as indicated by the subscript XYZ in the equation below):
drxyz / XYZ
drxyz
drXYZ
=
+
(3.88)
dt XYZ
dt
dt
XYZ
XYZ
The time derivative of rxyz/XYZ is simply the velocity of the origin of frame xyz with respect to frame XYZ,
uxyz/XYZ, i.e.:
drxyz / XYZ
= u xyz / XYZ
(3.89)
dt
XYZ
We must be careful, however, when evaluating the time derivative of rxyz in frame XYZ since both the
magnitude of rxyz and the basis vectors of frame xyz can change with time (the basis vectors of xyz can
change due to rotation of the frame xyz with respect to frame XYZ). To calculate the time derivative of rxyz
in frame XYZ, lets first write rxyz as a magnitude, rxyz, multiplied by the basis vectors of frame xyz, e xyz ,
then use the product rule to evaluate the time derivative:
drxyz
dt
=
XYZ
Note that
drxyz
dt
d rxyz e xyz
dt
XYZ
drxyz
dt
e xyz + rxyz
XYZ
e xyz = u xyz
de xyz
dt
(3.90)
XYZ
(3.91)
XYZ
is the velocity of the particle in frame xyz.
C. Wassgren
85
The time derivative of the xyz basis vectors is found from geometric considerations. Consider the drawing
shown below of the change in the x-basis vector as a function of time. For simplicity, well assume that the
rotation only occurs in the xy plane (i.e. x = y = 0) :
The time derivative of the basis vector is given by:
e x ( t + t ) e x ( t )
de x
= lim
0
t
dt
t
Note from the diagram that:
e x ( t + t ) e x ( t ) = e x ( t ) cos z + e y ( t ) sin z e x ( t )
e x ( t + t )
e x ( t + t ) e x ( t )
e x ( t )
= e x ( t )( cos z 1) + e y ( t ) sin z
In addition, as t 0, z 0 and:
( cos z 1) 1 ( z )2 / 2 1 = 12 ( z )2
and sin z z
so that:
1 ( z ) e x ( t ) + z e y
e x ( t + t ) e x ( t )
de x
= lim
= lim 2
t 0
t 0
dt
t
t
d
= z e y
dt
de x
= z e y (where z = dz/dt)
dt
In general, it can be shown that:
de xyz
= xyz / XYZ e xyz
dt XYZ
2
(3.92)
(3.93)
so that
de xyz
rxyz
dt
= rxyz xyz / XYZ e xyz = xyz / XYZ rxyz
(3.94)
XYZ
Combining Eqns. (3.88) (3.91), and (3.94) we find that the velocity of a fluid particle in the inertial frame
XYZ is:
u XYZ
= u xyz / XYZ
+
u xyz
+ xyz / XYZ rxyz
(3.95)

velocity of particle
in frame XYZ
velocity of frame xyz

w/r/t frame XYZ
velocity of particle
in frame xyz
velocity of particle in XYZ

due to rotation of frame xyz
w/r/t frame XYZ
where uxyz is the particle velocity in non-inertial frame xyz, xyz / XYZ is the angular velocity of frame xyz
with respect to frame XYZ, and rxyz is the position vector of the particle from the origin of frame xyz.
The acceleration of a particle in frame XYZ in terms of xyz quantities can be found in a similar manner:
du xyz / XYZ
du xyz
du XYZ
d
=
+
+
xyz / XYZ rxyz
(3.96)
dt XYZ
dt
dt
dt
XYZ
XYZ
XYZ

=a XYZ
= a xyz / XYZ
d
u xyz e xyz
dt
)
XYZ
xyz / XYZ rxyz + xyz / XYZ

=
d rxyz e xyz
dt
XYZ
where the results from Eqns. (3.90), (3.91), (3.93), and (3.94) are used to simplify the last two expressions
in Eqn. (3.96):
du xyz
de xyz
d
u xyz e xyz
=
e xyz + u xyz
dt
dt
dt
(3.97)
XYZ
= a xyz + xyz / XYZ u xyz

and
C. Wassgren
86
xyz / XYZ
d rxyz e xyz
= xyz / XYZ u xyz + xyz / XYZ rxyz
dt
(3.98)
XYZ
= xyz / XYZ u xyz + xyz / XYZ xyz / XYZ rxyz
Substituting Eqns. (3.97) and (3.98) into Eqn. (3.96) and simplifying gives:
a XYZ

rectilinear acceleration of
particle in frame XYZ
a xyz / XYZ

frame xyz w/r/t frame XYZ
2 xyz / XYZ u xyz )

(

Coriolis acceleration of
particle in frame XYZ due to
rectilinear motion of particle in
frame xyz
xyz / XYZ rxyz )
(

a xyz

particle in frame xyz
tangential acceleration of
particle in frame XYZ due to
rotational acceleration of frame xyz
+ xyz / XYZ xyz / XYZ rxyz

(3.99)
centripital acceleration of particle in

frame XYZ due to rotation of frame xyz
Now lets use these relations to determine an expression for the LME using a non-inertial frame of
reference. Recall that the Lagrangian statement for the LME is (refer to Eqn. (3.50)):
D
u XYZ dV = Fon system
(3.100)
Dt
Vsystem
Substitute Eqn. (3.95) into Eqn. (3.100) and re-arrange:
D
Fon system =
u xyz / XYZ + u xyz + xyz / XYZ rxyz dV
Dt
Vsystem
(3.101)
D
D
=
u xyz dV +
u xyz / XYZ + xyz / XYZ rxyz dV
Dt
Dt
Vsystem
Vsystem
Now use the Reynolds Transport Theorem to convert the first term on the right hand side to a control
volume perspective and re-arrange:
D
Fbody on CV + Fsurface on CV
Dt
Vsystem
(3.102)
d
dt
xyz dV
CV
C. Wassgren
xyz
( u rel dA )
CS
87
The remaining Lagrangian term can be simplified by changing the volume integral to a mass integral and
noting that the mass of the system doesnt change with time:
D
Dt
Vsystem
D
=
u xyz / XYZ + xyz / XYZ rxyz dm
Dt
M system
D
=
u xyz / XYZ + xyz / XYZ rxyz dm
Dt
(3.103)
M system
D
Dt
Vsystem
Since uxyz/XYZ and xyz/XYZ are functions only of time (these variables describe the motion of the reference
frame xyz and not the fluid field), and because Drxyz/Dt=uxyz1, we can replace the Lagrangian time
derivative with an Eulerian time derivative and substitute in our result from Eqn. (3.99):
D
Dt
Vsystem
Vsystem
(3.104)
d
dt
a xyz / XYZ +
xyz / XYZ rxyz + 2 xyz / XYZ u xyz + xyz / XYZ xyz / XYZ rxyz dV
Vsystem
Substituting Eqns. (3.103) and (3.104) back into Eqn. (3.102) and noting that when we apply the Reynolds
Transport Theorem the control volume and system volume are coincident (so that the system volume
integral in Eqn. (3.104) can be replaced by a control volume integral), we find that the LME can be applied
using a non-inertial frame of reference, xyz, if the following form is used:
Fbody on CV + Fsurface on CV
{a
xyz / XYZ
) (
)}
+
(3.105)
CV
d
dt
xyz dV
CV
xyz
( u rel dA )
CS
Lets consider a few examples to see how this form of the LME is applied.
Drxyz
Dt
rxyz
t

=0
+ ux
rxyz
x
+ uy
= e x
rxyz
y

=e y
+ uz
rxyz
z
= u xyz
= e z
where rxyz = xe x + ye y + ze z
C. Wassgren
88
Example:
A jet of water is deflected by a vane mounted on a cart. The water jet has an area, A, everywhere and is
turned an angle with respect to the horizontal. The pressure everywhere within the jet is atmospheric.
The incoming jet velocity with respect to the ground (axes XY) is Vjet. The cart has mass M. Determine the
horizontal acceleration of the cart at the instant when the cart moves with velocity Vcart (Vcart<Vjet) if no
horizontal forces are applied
A
Vjet
Vcart
Y
Fx
X
Fy
SOLUTION:
Apply the linear momentum equation to a control volume surrounding the cart. Use a frame of reference
fixed to the cart (xy). Note that this is not an inertial frame of reference since the cart is accelerating. As in
part (b) of this example problem worked out previously, in this frame of reference the cart appears
stationary and the jet velocity at the left is equal to Vjet-Vcart. Following the analysis given in part (b) of this
example problem worked out previously, conservation of mass indicates that the velocity on the right of the
control volume is Vjet Vcart.
Vjet - Vcart
A
Vjet - Vcart
y
x
Fy
Apply the linear momentum equation in the x-direction:
d
u x dV + u x ( u rel dA ) = FB, x + FS , x ax / X dV
dt
CV
CS
CV
C. Wassgren
89
(3.106)
where
d
dt
u dV 0
x
CV
(The cart has zero velocity in this frame of reference. The fluid in the control volume does accelerate
in this frame of reference; however, its mass is assumed to be much smaller than the cart mass. Hence,
the rate of change of the control volume momentum in this frame of reference is assumed to be zero.)
=u rel
=u x
=u rel
=ux
A
= A

=

u x ( u rel dA ) = V jet Vcart V jet Vcart i Ai + V jet Vcart cos V jet Vcart cos i + sin j A cos i + sin j
CS

) (
)(
left side
= V jet Vcart
) (
right side
A + V jet Vcart
A cos
()
cos 2 + sin 2
=1
= V jet Vcart
A ( cos 1)
FB, x = 0 (no body forces in the x-direction)

FS , x = 0 (all of the pressure forces cancel out)
x/ X
dV Ma (the mass within the CV is approximately equal to the cart mass)
CV
(Vjet Vcart ) A ( cos 1) = Ma

2
(Vjet Vcart ) A (1 cos )

2
a=
(3.107)
Now solve the problem using an inertial frame of reference fixed to the ground (frame XY). The velocity
out of the right side of the cart is given by Eqn. (3.66). The linear momentum equation in the X direction
gives:
d
(3.108)
dt
where
d
dt
CV
CS
dV Ma
CV
(The mass within the control volume is approximately equal to the cart mass since the fluid mass is
assumed to be negligible.)
=u X
= u rel
A

=

u X ( u rel dA ) = V jet V jet Vcart i Ai + V jet Vcart cos + Vcart V jet Vcart cos i + sin j A cos i + sin j
CS

= u rel
=u X
=A

)(
left side
) (
right side
= V jet V jet Vcart A + V jet Vcart cos + Vcart V jet Vcart A cos 2 + sin 2

=1
2
2
2
= V jet
+ V jetVcart + V jet Vcart cos + VcartV jet Vcart
A
2
2
= V jet Vcart cos V jet Vcart A
= V jet Vcart
( cos 1) A
FB, X = 0 (no body forces in the x-direction)

FS , X = 0 (all of the pressure forces cancel out)
C. Wassgren
90
Ma + Vjet Vcart
2
) A ( cos 1) = 0
(Vjet Vcart ) A (1 cos )

2
a=
(Same answer as before!)

(3.109)
M
As in part (b), using a frame of reference that is fixed to the control volume is easier than using one fixed to
the ground.
C. Wassgren
91
Example:
The tank shown rolls along a level track. Water received from a jet is retained in the tank. The tank is to
accelerate from rest toward the right with constant acceleration, a. Neglect wind and rolling resistance.
Find an algebraic expression for the force (as a function of time) required to maintain the tank acceleration
at constant a.
initial mass of cart and water, M0

A
V
F
U
SOLUTION:
First apply conservation of mass to a control volume surrounding the cart (shown below) in order to
determine how the cart mass changes with time.
The frame of reference xy is

fixed to the cart.
jet velocity relative to the

y
cart = (V U)
x
d
dt
where
d
dt
dV + u
CV
dV =
CV
rel
dA = 0
(3.110)
CS
rel
dM CV
dt
dA = (V U ) A
CS

dM CV
(V U ) A = 0
dt
dM CV
= (V U ) A
dt
C. Wassgren
(3.111)
92
Since the cart acceleration is constant (= a), we may write:

U = at (Note that U(t = 0) = 0 since the cart starts from rest.)
(3.112)
Note that Eqn. (3.112) is only true when a = constant. Otherwise, if a = a(t) one must write the velocity as:
t
U = U 0 + adt
(3.113)
Substitute Eqn. (3.112) into Eqn. (3.111) and solve the resulting differential equation.
dM CV
= (V at ) A
dt
M CV = M CV
t =t
dM CV =
M CV = M 0
(V at ) Adt
t =0
(
+ (Vt
)
)A
M CV M 0 = Vt 12 at 2 A
M CV = M 0
1
2
at 2
(3.114)
Now apply the linear momentum equation in the x direction to the same control volume. Note that the
frame of reference xy is not inertial since the cart is accelerating.
d
u x dV + u x ( u rel dA ) = FB, x + FS , x ax / X dV
(3.115)
dt
where
d
dt
CV
CS
CV
u dV 0
x
(most of the mass inside the CV has zero velocity in the given frame of reference)
CV
( u rel dA ) = (V U )2 A
CS
FB, x = 0
FS , x = F
x/ X
dV aM CV
CV
(V U ) A = F aM CV
2
F = (V U ) A aM CV
2
(3.116)
Now substitute Eqns. (3.112) and (3.114) into Eqn. (3.116).

2
F = (V at ) A a M 0 + Vt 12 at 2 A
(3.117)
Note that we could have also used a frame of reference fixed to the ground (inertial) to solve this problem.
C. Wassgren
93
Example:
A model solid propellant rocket has a mass of 69.6 gm, of which 12.5 gm is fuel. The rocket produces 1.3
lbf of thrust for a duration of 1.7 sec. For these conditions, calculate the maximum speed and height
attainable in the absence of air resistance. Plot the rocket speed and the distance traveled as functions of
time.
SOLUTION:
Assume that the mass flow rate from the rocket is constant. Also assume that the thrust remains constant
over the burn duration.
Apply the linear momentum equation in the y-direction to the CV shown using a frame of reference
attached to the rocket.
y
x
g
MCVg
(pe patm)Ae
Ve
d
u y dV + CS u y ( u rel dA ) = FB, y + FS , y CV a y / Y dV
dt CV
where
d
u y dV 0 (Most of the fluid has zero velocity in this frame of reference.)
dt CV
u ( u
y
rel
dA ) = Ve ( eVe Ae ) = eVe2 Ae
CS
FB , y = M CV g (weight)
FS , y = ( pe patm ) Ae (The exit pressure may be different from atmospheric pressure.)
y /Y
dV aM CV (Were using an accelerating frame of reference.)
CV
Substituting and simplifying:

eVe2 Ae = M CV g + ( pe patm ) Ae M CV a
a = g +
eVe2 Ae + ( pe patm ) Ae
(3.118)
M CV
C. Wassgren
94
Note that the thrust, T, is the force required to hold the rocket stationary (neglecting gravity).
T
Ve
(pe patm)Ae
x
d
ux dV + CS ux ( u rel dA ) = FB, x + FS , x
dt CV
where
d
ux dV 0 (Most of the fluid has zero x-velocity.)
dt CV
u ( u
x
rel
dA ) = Ve ( eVe Ae ) = eVe2 Ae
CS
FB , x = 0
FS , x = ( pe patm ) Ae + T
eVe2 Ae = ( pe patm ) Ae + T
T = eVe2 Ae + ( pe patm ) Ae
(3.119)
Substitute Eqn. (3.119) into Eqn. (3.118):

T
a = g +
M CV
(3.120)
Apply COM to the same CV:

d
dV + CS ( u rel dA ) = 0
dt CV
where
dM CV
d
dV =
dt CV
dt
( u
rel
dA ) = eVe Ae = m
CS

dM CV
+ m = 0
dt
Assuming the mass flow rate is a constant, solve Eqn. (3.121) subject to initial conditions:
M CV
M0
(3.121)
dM CV = m dt
0

M CV = M 0 mt
where M0 is the initial mass of the CV.
C. Wassgren
(3.122)
95
Substitute Eqn. (3.122) into Eqn. (3.120) and solve the differential equation for the velocity:
dU
T
a=
= g +

dt
M 0 mt
U
dU = gdt + M
Tdt

0 mt
U = gt

T M 0 mt
ln
m M 0
U = gt

T
mt
ln 1

m M0
(3.123)
Solve the differential equation given in Eqn. (3.123) for the height of the rocket.

dh
T
mt
U=
= gt ln 1
dt
m M 0
h

T
mt
ln 1
dt
M0
0 m
t
dh = gtdt
h = 12 gt 2 +

T M0
mt
ln 1
m m
M
0

mt
t ln 1
M
0

+ t

(3.124)
Note that Eqns. (3.120), (3.122) (3.124) are written specifically for when the fuel is burning. When the
fuel has been expended, the rocket equations of motion are:
a = g
(3.125)
U = U t =t g ( t t )
(3.126)
h = 12 g ( t t ) + U t = t ( t t ) + ht = t
(3.127)
where t is the time at which the fuel has been expended.

For the given problem were told:
M0 = 69.6 g
Mfuel = 12.5 g
T = 1.3 lbf = 5.79 N
t = 1.7 sec
giving a mass flow rate of:
M
m = fuel = 7.35 g/sec = 7.35*10-3 kg/sec
t
The maximum velocity will occur at the moment the fuel has been expended (neglecting the velocities as
the rocket falls back to the ground). The maximum height will occur when the velocity is zero.
Umax = U(t = t = 1.7 sec) = 139.2 m/s
hmax = h(t = tm = 15.9 sec) = 1100 m
(h(t = t) = 114 m)
The maximum height occurs when:

U = U t = t g ( tm t ) = 0
tm = t +
U t =t
g
C. Wassgren
96
The rocket speed and height are plotted below:
200
1200
150
1000
800
50
0
600
0
10
15
20
-50
25
30
35
height, h [m]
velocity, U [m/s]
100
400
-100
200
-150
-200
0
time, t [sec]
C. Wassgren
97
Example:
Explain why hurricanes rotate in a counter-clockwise direction in the northern hemisphere and in a
clockwise direction in the southern hemisphere. Do you think the flow down a bathtub drain or toilet will
follow the same trend? Why or why not?
SOLUTION:
The acceleration of a fluid particle in a rotating frame of reference is:
xyz / XYZ rxyz + 2 xyz / XYZ u xyz + xyz / XYZ xyz / XYZ rxyz
a XYZ = a xyz / XYZ + a xyz +
For the given problem:
(3.128)
a xyz / XYZ = 0
e z
(the hurricane eye isnt accelerating)

a xyz = 0
e r
(fluid particles arent accelerating)

xyz / XYZ = 0
(constant Earth rotation speed)

xyz / XYZ = eY
(constant Earth rotation speed)
u xyz = Vr e r + V e
e Y
re r
e u = V e + V e
xyz
r r

fluid particle
rxyz
(velocity of a fluid particle assuming 2D motion,

since fluid is pulled toward the low pressure
hurricane eye, Vr < 0)
= re r
(radial distance out to a fluid particle)
Note:
e Y = cos cos e r + sin cos e + sin e z
looking at the hurricane from above

C. Wassgren
98
Substitute into Eqn. (3.128):

a XYZ = 2e Y Vr e r + V e + e Y ( e Y re r )

(3.129)
centripital acceleration
Coriolis acceleration
Note that:
e Y e r = cos cos e r + sin cos e + sin e z e r
= sin cos e z + sin e
e Y e = cos cos e r + sin cos e + sin e z e
= cos cos e z sin e r
e Y e z = cos cos e r + sin cos e + sin e z e z
= cos cos e + sin cos e r

Hence, Eqn. (3.129) becomes (substituting s = sin , c = cos, etc.):
(
)
(
)
(
)
2
= 2Vr ( s c e z + s e ) + 2V ( c c e z s e r ) + r s c ( c c e + s c e r ) + s ( c c e z s e r )
a XYZ = 2Vr ( eY e r ) + 2V eY e + 2 r eY ( eY e r )
= 2Vr s c e z + s e + 2V ( c c e z s e r ) + 2 reY s c e z + s e
(
( 2V c c 2V s c rc s c ) e
a XYZ = 2V s 2 r s s c 2 2 rs 2 e r + 2Vr s 2 r s c c 2 e +
2
(3.130)
Note that since the rotation rate of the Earth is very small, the Coriolis acceleration effects will be much
larger than the centripetal acceleration effects, i.e.:
2Vr 2 r
(3.131)
Looking at first e term in Eqn. (3.130) (this is the one that will dominate based on Eqn. (3.131)), we
observe that when Vr < 0 (fluid moves toward the hurricane eye), > 0 (northern hemisphere), and > 0
(the Earth rotates in a counter-clockwise direction when looking from above), the e term will be negative.
When < 0 (southern hemisphere), the e term will be positive. Recall from the linear momentum
equation:
d
u xyz dV +
dt
CV
xyz
( u rel dA ) = FB + FS
CS
XYZ dV
CV
dFCoriolis 2Vr s e dV
(3.132)
Hence, the effective Coriolis effect produces an effective force acting on fluid particles pushing them in a
counter-clockwise direction ( +e ) in the northern hemisphere ( > 0) and in a clockwise direction ( e ) in
the southern hemisphere ( < 0).
C. Wassgren
99
Will the flow down a bathroom drain follow the same trend? Lets investigate the ratio of a typical
(convective) inertial force magnitude to a typical Coriolis force magnitude (this ratio is known as the
Rossby #, Ro, and shows up often in geophysical flows).
F
L2V 2
V
Ro inertial ~ 3
~
(3.133)
FCoriolis
L V L
where V is a typical flow velocity and L is a flow length scale. When the Rossby number is small, then the
Coriolis effect becomes significant. However, when the Rossby number is large then we may neglect the
Coriolis effect.
For a typical hurricane:
L 550 km (340 mi), V 10 m/s (23 mi/hr), 2 rad/24 hr = 7.3*10-5 rad/s
Ro 0.25 the Coriolis effect must be considered!
For flow down a bathtub drain:
L 1 m, V 1 m/s (2.2 mi/hr), 2 rad/24 hr = 7.3*10-5 rad/s
Ro 13,800 the Coriolis effect can be neglected The Earths rotation has essentially no effect
on the flow going down a bathtub drain.
C. Wassgren
100
5.
Angular Momentum Equation (AME)
In words and in mathematical terms, the angular momentum principle for a system is:
The rate of change of a systems angular momentum (AM) is equal to the net moment (or torque) acting on
the system.
hXYZdV
AM of a small
piece of fluid
system
rXYZ
uXYZdV
LM of a small
piece of fluid
Y
Z
D
( h XYZ + rXYZ u XYZ ) dV = M on system
Dt
Vsystem

AM of the system
(3.134)
system), V is the volume, and is the density. The quantity hXYZ is the intrinsic specific angular
momentum of a small piece of fluid resulting from the spin of fluid molecules contained within that small
piece of fluid. In typical fluids (e.g. non-polar, non-magnetic fluids), the angular momentum vectors of the
individual molecules are randomly oriented so that the sum of the intrinsic angular momentum vectors in a
region containing many molecules is zero. Hence, we will neglect this contribution to the angular
momentum of the fluid in the remainder of these notes. The quantity, uXYZ, represents the velocity of a
small piece of fluid in the system with respect to an inertial (aka non-accelerating) frame of reference
XYZ (recall that Newtons 2nd law holds strictly for inertial frames of reference) and rXYZ is the distance
from the inertial coordinate system origin to the fluid element. Note that a frame of reference moving at a
constant velocity in a straight line is non-accelerating and thus is inertial.
The term, Mon system, represents the net moments (or torques) acting on the system. These moments can be
due to both body and surface forces, i.e.
M body = rXYZ Fbody
M surface = rXYZ Fsurface

Note that if the fluid is magnetic, it is also possible to have an additional body moment that would induce
the fluid molecules to change their intrinsic angular momentum (h). As stated before, we wont consider
such fluids in these notes. The study of magnetic fluids is known as magnetohydrodynamics.
Using the Reynolds Transport Theorem to convert the left hand side of Eqn. (3.134) from a system point of
view to an expression for a control volume gives:
D
d
rXYZ u XYZ dV =
rXYZ u XYZ dV + rXYZ u XYZ ( u rel dA )
(3.135)
Dt
dt
V
CV
CS
system
Since the Reynolds Transport Theorem is applied to a coincident system and control volume, the moments
acting on the system will also act on the control volume. Thus, the AME for a CV becomes:
d
M body on CV + M surface on CV =
rXYZ u XYZ dV + rXYZ u XYZ ( u rel dA )

dt
CV
net
moment
due
net moment due

CS

to surface forces
to body forces
acting on the CV
acting on the CV
rate of increase of
AM inside the CV
(3.136)
net rate at which the CV AM changes

due to fluid leaving through the CS
AME for a CV
C. Wassgren
101
Notes:
1.
Recall that the AME is a vector expression. There are actually three equations built into Eqn. (3.136).
AME for Rotating (and not accelerating in translation) Frames of Reference

Recall that Newtons 2nd law holds strictly for inertial (non-accelerating) frames of reference. Often it is
more convenient to use a rotating (non-inertial) reference frame when applying the AME.
Y
uXYZ, uxyz
rXYZ, rxyz
x
xyz/XYZ
The Lagrangian statement for the AME is (refer to Eqn. (3.134)):
D
M on system =
rXYZ u XYZ dV
(3.137)
Dt
Vsystem
The mass of the system remains constant so the Lagrangian derivative can be brought inside the integral:
D
D
rXYZ u XYZ dV =
(3.138)
( rXYZ u XYZ ) dV
Dt
Dt
Vsystem
V
system
Since we are considering a reference frame that is only rotating and not accelerating in translation, the
position and velocity vectors may be written as2:
rXYZ = rxyz
(3.139)
u XYZ = u xyz + xyz / XYZ rxyz
Substituting into equation (3.138) gives:

D
rxyz u xyz + xyz / XYZ rxyz dV
Dt
Vsystem
Vsystem
Vsystem
D
D
rxyz u xyz +
rxyz xyz / XYZ rxyz dV
Dt
Dt
(3.140)
D
rxyz u xyz dV
Dt
+
Drxyz
D
xyz / XYZ rxyz + rxyz
xyz / XYZ rxyz dV
Dt
Dt
Vsystem
Again, since the mass of the system is constant, the Lagrangian derivative can be brought outside of the
first integral. In addition, we know from previous work1 that Drxyz/Dt=uxyz and:
D
xyz / XYZ rxyz + xyz / XYZ u xyz + xyz / XYZ xyz / XYZ rxyz
xyz / XYZ rxyz =
Dt

2
Refer to an earlier set of notes regarding the LME for non-inertial reference frames for the derivation of
these quantities.
C. Wassgren
102
Vsystem
+
=
D
rxyz u xyz dV
Dt
D
Dt
Drxyz
D
xyz / XYZ rxyz + rxyz
xyz / XYZ rxyz dV
Dt
Dt
Vsystem
(3.141)
(rxyz u xyz ) dV
Vsystem
u xyz xyz / XYZ rxyz +
dV
xyz / XYZ rxyz + xyz / XYZ u xyz + xyz / XYZ xyz / XYZ rxyz
rxyz
Vsystem
To simplify things further, we can re-arrange the first term in the second integral and incorporate it into the
second term:
u xyz xyz / XYZ rxyz +
(
)

= rxyz
r + 2 xyz / XYZ u xyz + xyz / XYZ ( xyz / XYZ rxyz )
xyz / XYZ xyz

r + xyz / XYZ u xyz + xyz / XYZ xyz / XYZ rxyz
rxyz
xyz / XYZ xyz
(3.142)
Substituting Eqns. (3.139) (3.142) into Eqn. (3.137) gives:
D
M on system =
rxyz u xyz dV
Dt
Vsystem
{r
xyz
)}
(3.143)

r + 2 xyz / XYZ u xyz + xyz / XYZ xyz / XYZ rxyz dV
xyz / XYZ xyz
Vsystem
Now use the Reynolds Transport Theorem to convert the first term on the right hand side to a control
volume perspective and re-arrange. Also note that since the CV and system are coincident, the moments
acting on the CV will be the same as the moments acting on the system and the CV mass will be the same
as the system mass.
M body on CV + M surface on CV
{r
xyz
)}

r + 2 xyz / XYZ u xyz + xyz / XYZ xyz / XYZ rxyz dV
xyz / XYZ xyz
(3.144)
CV
d
dt
xyz
CV
u xyz dV +
xyz
u xyz ( u rel dA )
CS
AME for a rotating frame of reference

Lets consider some examples to see how this form of the AME is applied.
C. Wassgren
103
Example:
A lawn sprinkler is constructed from pipe with an inner diameter of d with each arm having a length of R.
Water flows through the sprinkler at a volumetric flow rate of Q. A force, F, is applied a distance, l, from
the sprinkler hub on one of the sprinkler arms. If the water stream leaving the sprinkler arm is at an angle
with respect to the tangent of the circle traced out by the sprinkler arms, determine:
a. the force, F, required to hold the sprinkler stationary,
b. the force, F, required to have the sprinkler rotate at a constant angular velocity, ,
c. the angular acceleration of the sprinkler if the sprinklers moment of inertia (including the fluid inside
the sprinkler) is I and it is rotating with angular velocity , and
d. the maximum angular velocity, max, of the sprinkler if no force is applied
V
F
pipe diameter, d
TOP VIEW
R
SIDE VIEW
Q
C. Wassgren
104
SOLUTION:
First consider the case where the sprinkler does not rotate ( = 0). For the fixed frame of reference and
control volume shown, the angular momentum equation is:
d
(3.145)
( r u ) dV + ( r u )( u rel dA ) = M S + M B
e
dt
where
d
dt
CV
CS
( r u ) dV = 0
top view
(steady flow)
e r
CV
( r u ) ( urel dA ) =
CS
d2
2 Re r V ( sin e r + cos e ) V
4
two
arms
= 2 RV 2 cos
d2
e z
4
(Neglect the small bend in the pipe end when estimating the radius, r.)
M S = le r Fe = Fle z
MB = 0
Substitute and simplify:
2 RV 2 cos
F = 2 V 2
d2
4
e z = Fle z
d2
R
cos
4
l
(3.146)
Note that from conservation of mass on the same control volume:
Q = 2V
d2
4
V=
2Q
(3.147)
d2
C. Wassgren
105
If the sprinkler is rotating, then use the same control volume (attached to and surrounding the sprinkler
arms) but use a frame of reference that rotates with the control volume. The angular momentum equation
for a rotating frame of reference is:
d
dt
(r u )
xyz
dV +
CV
(r u)
xyz (
u rel dA ) = M S + M B
CS
xyz
CV
where
d
dt
(r u )
xyz
dV = 0 (the flow is steady in the rotating frame of reference)
CV
(r u )xyz ( urel dA ) = 2 Re r V ( sin e r + cos e ) V
CS
d2
= 2 V
4
d2
4
R cos e z
(Neglect the small bend in the pipe end when estimating the radius, r.)
M S = le r Fe = Fle z
MB = 0
xyz
xyz / XYZ rxyz + 2 xyz / XYZ u xyz + xyz / XYZ xyz / XYZ rxyz dV =

CV
e re + 2e Ve + e ( e re ) dr d
2 re r
z
r
z
r
z
z
r

4
= re
r =0
= 2 Ve
=re

=2 re r
r=R
=2
( r e
2
+ 2Vre z
d2
e 2 d r 2 dr + d 4Ve rdr
dr =
z
z
4
4
4
0
0

=I
e + 2 V d R 2 e
= I
z
z
4
where I is the sprinklers moment of inertia. Note that the small, bent portion of the pipe has been neglected
in the control volume integral. Substitute and simplify.
2 V 2
e + 2R 2 V d e + Fle
R cos e z = I
z
z
z
4
4
d2
2
= Fl 2 V d R (V cos + R )
I
4
(3.148)
From Eqn. (3.148) the force required to maintain the sprinkler at a constant angular velocity is:
F = 2 V
d2 R
4
(V cos + R ) (Note that if = 0 this simplifies to Eqn. (3.146).)
(3.149)
If the force, F, is removed, the angular acceleration is:

2
= 2 V d R (V cos + R )
4 I
(3.150)
The maximum angular velocity of the sprinkler if no force is applied is:

V cos
max =
C. Wassgren
106
(3.151)
Now lets solve the same problem but using a fixed frame of reference as shown in the figure below.
Consider the sprinkler in an arbitrary orientation.
e Y
l
e X
top view
d
dt
( r u ) dV + ( r u )( u
CV
rel
dA ) = M S + M B
CS
where
d
dt
r=R
d
2
r ( cos e X + sin eY ) V ( cos e X + sin eY ) + r ( sin e X
dt
r =0
r=R
d
rV ( cos sin e sin cos e ) + r 2 cos 2 e + sin 2 e
= 2
Z
Z
Z
Z
dt
r =0
r=R
d
2 2
d r dr
= e Z 2
4
dt
r = 0

=I
d
= {I e Z }
dt
e
= I
Z
( r u ) dV =
CV
+ cos eY ) d 2 dr
4
d 2 dr
( r u ) ( u
(3.152)
} V 4 d 2 )
rel dA ) = 2 R ( cos e X + sin e Y ) V sin ( ) e X + cos ( ) e Y + R [ sin e X + cos e Y ]
CS
= 2 R V cos cos ( ) e Z sin sin ( ) e Z + R cos 2 e Z + sin 2 e Z V d 2

4
(3.153)
= 2 V d 2 R V cos ( + ) + R e Z
4
= 2 V d 2 R (V cos + R ) e Z
4
M S = l ( cos e X + sin eY ) F ( sin e X + cos eY )
= Fl cos 2 e Z + sin 2 e Z
(3.154)
= Fle Z
MB = 0
Substitute and simplify (considering only the Z-direction).
+ 2 V
I
d2
4
R (V cos + R ) = Fl
C. Wassgren
(3.155)
This is the same result as Eqn. (3.148)!
107
Example:
A pipe branches symmetrically into two legs of length, L, and the whole system rotates with angular speed,
, around its axis. Each branch is inclined at angle, , to the axis of rotation. Liquid enters the pipe
steadily, with zero angular momentum, at the volume flow rate Q. The pipe diameter, D, is much smaller
than L.
a.
Obtain an expression for the external torque required to turn the pipe.
b.
What additional torque would be required to impart angular acceleration ?
V
r
L
D
z frame of reference
rotates with the arms
SOLUTION:
Apply the angular momentum equation to the CV shown above using a frame of reference rotating with the
arms (this is an accelerating frame of reference). Let V be the velocity of the fluid in the pipe arms (V
D2/4= 1/2Q).
d
( rxyz u xyz ) dV + CS ( rxyz u xyz ) ( urel dA )
dt CV
= MB + MS
xyz
xyz / XYZ rxyz + 2 xyz / XYZ u xyz + xyz / XYZ ( xyz / XYZ rxyz ) dV

CV
where
d
( rxyz u xyz ) dV = 0 (steady problem in the given frame of reference)
dt CV
(r
xyz
u xyz ) ( u rel dA ) =
CS
+ L ( cos e z + sin e r ) V ( cos e z + sin e r ) ( 12 Q )

incoming flow
0
top arm
+ L ( cos e z sin e r ) V ( cos e z sin e r ) ( 12 Q )

bottom arm
=0
M B = 0 (neglect gravity)
M S = T (this is the torque we must apply to the rotating arm)
C. Wassgren
108
xyz
xyz / XYZ rxyz + 2 xyz / XYZ u xyz + xyz / XYZ ( xyz / XYZ rxyz ) dV =

CV
e z s ( cos e z + sin e r ) +
D 2
s =0 s ( cos e z + sin e r ) 2e z V ( cos e z + sin e r ) + 4 ds +
e z e z s ( cos e z + sin e r )

s=L
top arm
e z s ( cos e z sin e r ) +
s=L
D 2
s
e
V
e
e
+
ds
cos
sin
2
cos
sin
(
)
(
)
z
r
z
z
r
4
s =0
e z s ( cos e z sin e r )
e z

bottom arm
s= L
=
=
=
D2
4
D2
4
2
s ( cos e z + sin e r ) s sin e + 2V sin e s sin e r ds +
s=0
s= L
s cos e sin e s sin e 2V sin e + s 2 sin e ds

z
r)
r
s=0
2
2
2
2
L
s sin cos e r + s sin e z 2sV sin cos e r + 2sV sin e z s 2 2 sin cos e
0 + s 2 sin cos e + s 2 sin 2 e + 2sV sin cos e + 2sV sin 2 e + s 2 2 sin cos e ds
r
z
r
z
2s sin
2
e z + 4sV sin 2 e z ds
0
2
2 L3 sin 2 + 2VL2 sin 2 e z

4 3
Combining together all of the terms in the angular momentum equation:
0 = T
D2
23 L3 sin 2 + 2VL2 sin 2 e z

4
D2
23 L3 sin 2 + 2VL2 sin 2 e z

4
or, since V D2/4= 1/2Q:
T=
D2
L3 sin 2 + Q L2 sin 2 e z
T =
6
The torque required to turn the pipe at a constant angular velocity, , is:
T = Q L2 sin 2 e z
The additional torque required to impart an angular acceleration is:
Tadditional =
D2
6
L3 sin 2 e z
C. Wassgren
109
6.
Basic Thermodynamic Definitions
Before beginning our discussion of conservation of energy, we need to first review some basic
thermodynamic definitions.
system and surroundings

A system is a particular quantity of matter chosen for study. The surroundings include everything
that is not the system.
control volume (CV), control surface (CS), and outward-pointing unit normal vector
A control volume (CV) is a particular volume or region in space. A control surface (CS) is the
surface enclosing the control volume. The orientation of the CS at a particular location is given by
the direction of its outward-pointing unit normal vector, n , at that location. The outward-pointing
unit normal vector has a magnitude of one, is perpendicular to the control surface, and always points
out of the CV.
n
n
n
n
CV
n
CS
n
n
property (extensive, intensive, and specific)

Properties are macroscopic characteristics of a system. An extensive property is one that depends
on the amount of mass in the system. An intensive property is one that is independent of the mass in
the system. A specific property is an extensive property per unit mass (a specific property is also an
intensive property). Some examples: mass, m, is an extensive property, pressure, p, is an intensive
property, and specific volume, vV/m, is a specific property. An easy way to determine whether a
property is extensive or intensive is to divide the system into two parts and see how the property is
affected.
state
The state of a system is the systems condition or configuration as described by its properties in
sufficient detail so that it is distinguishable from other states. Often a state can be described by a
subset of the systems properties since the properties themselves may be related.
process
A process is a transformation from one state to another. A few common processes include:
isothermal process: A process that occurs at constant temperature.
isobaric process: A process that occurs at constant pressure.
isochoric or isometric process: A process that occurs at constant volume.
adiabatic process: A process in which there is no heat transfer between the system and
surroundings.
isentropic process: A process that occurs at constant entropy.
path
A process path is the series of states that a system passes through during some process.
C. Wassgren
110
cycle
A cycle is a sequence of processes that begins and ends at the same state. At the conclusion of a
cycle, all properties have the same values they had at the beginning of the cycle. Thus, there is no
change in the systems state at the end of a cycle.
equilibrium
An equilibrium state is one in which there are no unbalanced potentials (or driving forces) within the
system driving the system to another state. Note that there are many types of equilibrium, e.g.
thermal, mechanical, chemical, and phase equilibriums.
quasi-equilibrium process
A quasi-equilibrium process is one where the process proceeds in such a manner that the system
remains infinitesimally close to an equilibrium state at all times. One can interpret a quasiequilibrium process as occurring slowly enough so that the system has time to adjust internally such
that properties in part of the system do not change any faster than those properties in other parts of
the system.
reversible and irreversible processes
A reversible process is one in which the system is in a state of equilibrium at all points in its path.
In a reversible process, the system and the surroundings can be restored exactly to their initial states.
An irreversible process is one where the system is not in a state of equilibrium at all points in its
path. The system and surroundings cannot be returned to their exact initial states in an irreversible
process. Note that all natural processes are irreversible. Several effects causing irreversibility
include viscosity, heat conduction, and mass diffusion.
equation of state
An equation of state is a relationship between properties of a particular substance or class of
substances. Equations of state cannot be obtained from thermodynamics but are obtained either
from experimental measurements or from some molecular model. Note that there can be various
types of equations of state, e.g. two equations of state for an ideal gas include a thermal equation of
state which is the ideal gas law, p=RT , and a caloric equation of state which describes the
relationship between the internal energy and temperature, du=cvdT.
C. Wassgren
111
7.
Discussion of Energy, Work, and Heat
Now lets move our discussion to the three basic thermodynamic concepts of energy, work, and heat.
Energy
The energy associated with some phenomenon is not a physical quantity but is, in fact, just a number
resulting from a formula containing physically measurable quantities related to that phenomenon. For
example, the energy associated with the macroscopic motion of a system of mass, m, moving with a speed,
V, is equal to 1/2mV 2. By itself, the energy associated with a phenomenon is not a very useful quantity.
However, experiments examining the total energy of a system, i.e. the sum of all the various energies, have
resulted in a very remarkable observation. When the system does not interact with its surroundings, the
total energy of the system remains constant. The energy associated with a particular phenomenon may
change; however, it can only change at the expense of the energy associated with some other phenomenon.
Well examine this observation in greater detail a little later but for now we will define the various types of
energy that are most common in fluid flows.
Kinetic Energy, KE
The energy associated with the macroscopic motion of a system relative to a reference frame xyz is known
as the kinetic energy, KE:
2
KE = 1 2 mVxyz
(3.156)
where m is the mass of the system and Vxyz is the velocity of the system in the reference frame, xyz.
Potential Energy, PE
The energy associated with a systems ability to do work in an external force field such as a gravity field is
known as the potential energy, PE. For example, the gravitational potential energy for a mass, m, located
in a gravitational field with gravitational acceleration, g, pointing in the z-direction is:
(3.157)
PE = mgz
where z is the height of the mass above some reference plane.
Internal Energy, U
The internal energy of a system is comprised of a number of sub-classes of energy which include:
a. sensible energy
This is the energy associated with the internal molecular translational, rotational, and vibrational
motion. Temperature is a measure of this type of internal energy. The larger the temperature of a
system, the greater its sensible energy.
b. latent energy
This is the energy associated with the attraction between molecules. We concern ourselves with latent
energy most often when examining processes that involve a change of phase such as going from a solid
to a liquid or from a liquid to a gas (or vice versa).
c. chemical energy
This is the energy associated with the attraction between atoms.
d. nuclear energy
This is the energy associated with the attraction between particles within an atom such as the attraction
between protons and neutrons.
There are other forms of internal energy (e.g., the energy associated with electric and magnetic dipole
moments) but we rarely encounter these in typical engineering applications. In these notes well only
concern ourselves with sensible energy.
C. Wassgren
112
The total energy of a system is the sum of these various types of energy:
E = U + KE + PE
(3.158)
Note that the total, internal, kinetic, and potential energies are extensive properties, i.e., the magnitude of
these energies depends on the system mass. In terms of specific quantities (extensive properties per unit
mass) we have:
e = u + 12V 2 + G
(3.159)
where e is the specific total energy, u is the specific internal energy, V is the velocity magnitude, and G is a
(conservative) potential energy function. Note that the force per unit mass resulting from a conservative
potential energy function is found by taking the negative of its gradient, i.e. if G = gz where g is the
gravitational acceleration and z is the height of the system above some reference plane, then fgravity = G =
g.
The values of the specific internal energy, u, at different states for various substances are tabulated in
thermodynamic property tables. Most introductory thermodynamics books have such tables for steam,
refrigerants, and a variety of gases. Of particular interest in these notes is the specific internal energy for
ideal gases and incompressible substances. Well examine how to evaluate u for these substances a little
later.
C. Wassgren
113
Work
Work is an energy interaction (a way to transfer energy) occurring at the boundary between a system and
its surroundings. Thus, work is not a property of a system but rather is associated with a process that the
system is undergoing. The work done on the system by its surroundings depends on the path of the
process. A quantity that is also commonly encountered when discussing work is the power, defined as the
work done per unit time.
The work done on a system is equal to the dot product of the force acting on the system, Fon system, and the
distance over which the force acts, ds:
dWon system = Fon system
ds
(3.160)

small distance
over
small amount
of work on the
system
force acting
on the system
which the force

acts
s2
W12 = F ds
(3.161)
s1
where W12 is the total work in going from state 1 (indicated by s1) to state 2 (s2). Note that the work
depends on the path taken from s1 to s2. For example, consider the situation shown below:
Example:
A block with weight, w, is pushed on a frictional surface. The friction coefficient between the block and
the surface is . Determine the amount of work done by the friction force on the block when moving the
block from state 1 to state 2 using the paths shown.
1
L
(a)
(b)
SOLUTION:
The work done by the friction force, w, in case (a) is:
W12 = w ( L ) (note that the friction force is in the direction opposite to the displacement)
The work done by friction in case (b) is:
W12 = w ( 3L )
C. Wassgren
114
Types of Work
Now lets consider a few different types of work that can be done by or on a system. The types of work
well present here include work due to gravity, acceleration, pressure, electricity, springs, and rotating
shafts. In the drawings below, the system is enclosed by a dashed line.
Gravitational Work (aka Potential Energy)
Consider the work required to move an object with mass, m, to a higher elevation in a gravity field (assume
a quasi-static process so that accelerations can be neglected):
object with
weight, mg
Won system,12 =
on system
0
h
=
1
( mge
ds
) ( dze z )
Won system,1 2 = mg h
(3.162)
This is just the change in the potential energy of the

system! Note that the work on the surroundings is
equal to, but has the opposite sign, of the work done
on the system.
Acceleration Work (aka kinetic energy)
Consider the work required to accelerate an object with mass, m, from velocity, V1, to velocity, V2:
1
V2
V1
x
Won system,12 = Fon system ds

1
V2
dV

m
=
e x Vdt
e x
dt

= dx
V1
Newton's 2nd law
object with
mass, m
V2
= m VdV
V1
Won system,12 =
2m
(V
2
2
V12
(3.163)
This is just the change in the kinetic energy of the

system!
C. Wassgren
115
Pressure Work
Consider the work done by the expansion of a fluid (a gas or liquid) in a piston:
2
Won surr,12 = Fon surr ds
1 2
fluid
1
x2
( pAe
) ( dxe x )
x1
x2
dx
pAdx
x1
pressure force the system

exerts on the face of the
piston: F = pA
where p is the pressure at
the piston face and A is the
area of the piston face
V2
Won surr,12 =
pdV
(3.164)
V1
Note that dV=Adx. Note also that in this example, the

work on the surroundings has been calculated instead
of the work acting on the system. To get the work
done on the system, we simply have:
Won system = -Won surr
If we plot how the pressure changes with volume we get a p-V diagram:
p
The area under the curve is equal to the work done by
path
2
p2
the fluid on the surroundings in going from state 1 to
1
area
=
pdV
state 2.
p
1
V2
Won surr,12 =
V1
V2
pdV
V1
dV
Note that different paths from state 1 to state 2 will give different amounts of work:
p
p
2
(a)
(b)
W12(a) > W12(b)
One example of a pressure-volume relationship is known as a polytropic process where the pressure and
volume are related by:
pV n = constant
(3.165)
C. Wassgren
116
Example:
A gas in a piston assembly undergoes a polytropic expansion from an initial volume, Vi=0.1 m3, and initial
pressure, pi=2 bar (1 bar = 1*105 Pa), to a final volume of Vf=0.5 m3. Determine the work the gas does on
the piston for n=1.5 and n=1 (where pV n=constant).
SOLUTION:
The work the gas performs on the piston is given by:
V = 0.5 m3
Wi f =
(3.166)
pdV
V = 0.1 m
where, for a polytropic expansion,

pV n = constant=c
where n is a constant. Substitute Eqn. (3.167) into Eqn. (3.166).
c 1 n 0.5 m3
V = 0.5 m3
V
n 1
0.1 m3
Wi f =
cV n dV = 1 n
c ln V 0.5 m3
V = 0.1 m3
n =1
0.1 m3
(3.167)
(3.168)
When n = 1.5, the constant is
)(
c = 1*105 Pa * 0.1 m3

= pi
1.5
= 3.2 *103 N m-0.5
(3.169)
=Vi
and the work performed by the gas, using Eqn. (3.168), is:
Wi f =
3.2*103 N m-0.5
3
0.5 m
0.5
0.5
0.1 m3
0.5
Wi f = 1.1*104 N m
When n = 1, the constant is:
)(
(3.170)
c = 1*105 Pa * 0.1 m3 = 1*104 N m

= pi
(3.171)
=Vi
and the work performed by the gas, using Eqn. (3.168), is:
0.5 m3
Wi f = 1*104 N m ln
0.1 m3
Wi f = 1.6 *104 N m
C. Wassgren
(3.172)
117
Electric Work
Electrons moving across a system boundary can do work on a system since in an electric field, a force acts
on an electron. When N Coulombs of electrons pass through a potential difference, V (the voltage), the
electric work done on the system is:
Won system = NV
(3.173)
The corresponding power is:
Won system = VN
= VI = I 2 R = V
(3.174)
R
where I is the current and R is the resistance of the system. Note that Ohms Law (V=IR) has been used in
deriving the last two expressions on the right hand side of Eqn. (3.174).
Example:
A 12 V automotive battery is charged with a constant current of 1.5 A for 3 hrs. Determine the work done
on the battery.
SOLUTION:
The work done on the battery is:
t =T
Won battery =
VIdt = VIT = (12 V )(1.5 A ) ( 3 hr 3600 s hr )
t =0
Won battery = 0.2 kJ

Spring Work
Now lets examine the work required to compress a spring with stiffness, k:
2
Won system,1 2 = Fon system ds
1
x2
( kxe
) ( dxe x )
x1
x
x2
x1
Won system,1 2 =
2k
(x
2
1
x22
(3.175)
Note k is assumed constant in the Eqn. (3.175).

Shaft Work
Another method of transferring energy between a system and the surroundings is through shaft work. Shaft
work is most often associated with rotating fluid machines such as compressors, pumps, turbines, fans,
propellors, and windmills. The power acting on a system is given by:
Won system = Ton system
(3.176)
where Ton system is the torque acting on the system (assumed constant here) and is the angular velocity of
the shaft.
C. Wassgren
118
Heat
Heat is another form of boundary energy interaction occuring between a system and its surroundings. The
difference between heat and work is that heat transfer occurs due to differences in temperature. Heat
moves from regions of high temperature to regions of low temperature. Like work, heat is not a property of
a system but rather is associated with a process. Furthermore, the amount of heat transferred during a
process depends on the path taken during the process.
Modes of Heat Transfer
Heat can be transferred between the system and surroundings via three methods: conduction, convection,
and radiation.
Conduction
Conduction is the transfer of energy from the more energetic particles of a substance to the adjacent less
energetic ones as a result of interactions between the particles.
, (this is a vector quantity since the heat travels in a particular direction) due to
The rate of heat transfer, Q
conduction through an area, A, of a substance is given by Fouriers Law of Heat Conduction:
= kAT
Q
(3.177)
where k is a material property of the substance known as the thermal conductivity, and T is the thermal
gradient in the substance. Note that the negative sign in the equation is required so that heat moves from
regions of higher to lower temperature.
Convection
Convection is the mode of energy transfer between and solid surface and an adjacent fluid that is in motion;
it involves the combined effects of conduction and relative fluid motion (also known as advection).
The rate of heat transfer, Q , leaving a surface with area, Asurface, and entering the fluid due to convection is
given by Newtons Law of Cooling:
Q = hAsurface (Tsurface Tfluid )
(3.178)
where h is the heat transfer coefficient for the situation, Tsurface is the temperature of the surface which is
in contact with the fluid with temperature, Tfluid. The heat transfer coefficient will depend on the surface
and fluid properties as well as the flow characteristics. It is generally an experimentally determined
property for all but the simplest flow situations.
Radiation
Radiation is the energy emitted by matter in the form of electromagnetic waves as a result of changes in the
electronic configurations of the atoms or molecules. Unlike conduction and convection, radiation does not
require an intervening medium.
The rate at which heat is emitted from a surface with area, Asurface, depends on the absolute temperature of
the surface, Tsurface, as indicated by the modified Stefan-Boltzmann Law:
4
Q emitted = AsurfaceTsurface
(3.179)
where Q
is the rate at which heat is emitted from the surface, is the emissivity of the surface
emitted
(01), and is the Stefan-Boltzmann constant (=5.67*10-8 W/(m2K4)=0.1714*10-8 Btu/(hrft2R4)).

A blackbody is an object with an emissivity of one: blackbody=1, i.e. a blackbody is a perfect emitter of
radiation.
C. Wassgren
119
Surfaces can also absorb radiation. The heat flux absorbed by a surface via radiation is given by:
Q absorbed = Qincident
(3.180)
where is the absorptivity of the surface (01). Note that a blackbody has =1; it is both a perfect
emitter and absorber of radiation.
Actual determination of the rate at which radiation is emitted and absorbed by a surface can be very
complicated since the rate depends on factors such as surface orientation, the effects of the intervening
medium, and the surface spectral characteristics.
For the special case where a small surface interacts with a much larger surface, the intervening fluid has no
affect on the radiation transfer, and = (termed a grey body), the rate of heat transfer from the surface to
the surroundings via radiation is:
Q
= A
T4
T 4
(3.181)
emitted
surface
surface
C. Wassgren
120
Example:
An insulated frame wall of a house has an average thermal conductivity of 0.0318 Btu/(hrftR). The
thickness of the wall is 6 in. At steady state, the rate of energy transfer by conduction through an area of
160 ft2 is 400 Btu/hr, and the temperature decreases linearly from the inner surface to the outer surface. If
the outside surface temperature of the wall is 30 F, what is the inner surface temperature in F?
SOLUTION:
A
TC
TH
k
x
A
Q
TC
Q x
=
=
=
=
=
0.0318 Btu/(hrftR)
6 in = 0.5 ft
160 ft2
400 Btu/hr
30 F = 490 R
x
x
From Fouriers Law, the heat transfer through the wall is:
dT
T TH
Q x = kA
= kA C
x
dx
(assuming a linear temperature change)
(3.182)
Re-arrange to solve for TH.

TH = TC +
Q
x
kA
(3.183)
Using the given parameters:

k = 0.0318 Btu/(hrftR)
x = 6 in = 0.5 ft
A = 160 ft2
Q = 400 Btu/hr
TC = 30 F = 490 R
TH = 529 R = 69 F
C. Wassgren
121
Example:
A cartridge electrical heater is shaped as a cylinder of length 200 mm and outer diameter of 20 mm. Under
normal operating conditions the heater dissipates 2 kW while submerged in a water flow which is at 20 C
and provides a convection heat transfer coefficient of 5000 W/(m2K). Neglecting heat transfer from the
ends of the heater, determine the heaters surface temperature. If the water flow is inadvertently terminated
while the heater continues to operate, the heater surface is exposed to air which is also at 20 C but for
which the heat transfer coefficient is 50 W/(m2K). What is the corresponding surface temperature? What
are the consequences of such an event?
SOLUTION:
Q
TS
L
Determine the surface temperature using Newtons Law of Cooling.

Q = hA (TS T )
(3.184)
where A = DL. Re-arranging gives:

TS = T +
Q
hA
(3.185)

T
= 20 C = 293 K

Q
= 2000 W
h
= 5000 W/(m2K)
D
= 20*10-3 m
L
= 200*10-3 m
A = 1.3*10-2 m2
TS = 325 K = 52 C
If instead, h = 50 W/(m2K), then:
= 3500 K = 3200 C
TS
This temperature is probably large enough to melt the cartridge heater!
C. Wassgren
122
Example:
An uninsulated steam pipe passes through a room in which the air and walls are at 25 C. The outside
diameter of the pipe is 70 mm, and its surface temperature and emissivity are 200 C and 0.8, respectively.
If the coefficient associated with free convection heat transfer from the surface to the air is 15 W/(m2K),
what is the rate of heat loss from the surface per unit length of pipe?
SOLUTION:
The convective heat transfer rate, QC , is given by Newtons Law of Cooling:
Q C = hAS ( TS T )
(3.186)
where h is the convection heat transfer coefficient, AS is the surface area of the pipe, TS is the surface
temperature of the pipe and T is the ambient temperature.
The radiative heat transfer rate, Q R , is given by:
Q R = AS (TS4 T4 )
(3.187)
where is the surface emissivity and is the Stefan-Boltzmann constant.

The total heat transfer rate from the pipe is:
QT = Q C + Q R
(3.188)

h
= 15 W/(m2K)
D
= 70*10-3 m AS = DL AS/L = 0.22 m
TS
= 200 C = 473 K
= 15 C = 288 K
T
= 0.8
= 5.67*10-8 W/(m2K4)
QC / L = 580 W/m

Q / L = 420 W/m
R
QT / L = 1 kW/m
C. Wassgren
123
8.
Conservation of Energy (COE) (aka 1st Law of Thermodynamics)
Conservation of Energy for a System

In words and in mathematical form, the first law of thermodynamics for a system is:
The increase in total energy of a system is equal to the heat added to the system plus the work done on the
system.
Won system
dEof system = Qinto system + Won system
system
(3.189)
Qinto system
where dE is a small increase in the total energy of the system, Qinto system is a small amount of heat
transferred into the system, and Won system is a small amount of work done on the system by the
surroundings.
Notes:
1.
Since energy is a property of a system, an exact differential (the d operator in dE) is used to specify
the small change in the energy. In other words, the difference in energy between two states depends
only upon the endpoint states and is independent of the path between the two states. The small change
in heat and work are indicated using an inexact differential (the operator in Q and W) to signify
that both heat and work are path dependent processes.
2.
Note that different disciplines have different notations for COE. In particular, in thermodynamics
work is usually discussed in terms of the work done by the system on the surroundings so that the 1st
Law becomes:
dEof system = Qinto system Wby system
(3.190)
In order to avoid confusion regarding the proper sign for work, these notes will try to clearly specify
whether work is being done on or by the system. Understanding that if one does work on a system, its
energy will increase is generally sufficient to avoid most sign convention problems.
3.
We can also write COE in terms of time rates of changes by taking the limit of the changes in the
properties over a short amount of time as the time approaches zero:
E
= Q
+ W
(3.191)
of system
into system
on system
Now lets consider a few simple examples.
C. Wassgren
124
Example:
A rigid tank contains a hot fluid that is cooled while being stirred. Initially the internal energy of the fluid
is 800 kJ. During the cooling process, the fluid loses 500 kJ of heat and the stirring propeller does 100 kJ
of work on the fluid. What is the final internal energy of the fluid?
fluid
stirring propeller
rigid tank
SOLUTION:
Apply conservation of energy to the system of fluid contained within the tank.
Esys = E f Ei = Qinto system + Won system
where
Ei = Ui = 800 kJ
Qinto system = -500 kJ
Won system = 100 kJ
Ef = Uf = 400 kJ
C. Wassgren
125
Example:
Four kilograms of a certain gas is contained within a piston-cylinder assembly. The gas undergoes a
polytropic process where: pV1.5=constant. The initial pressure is 3 bars, the initial volume is 0.1 m3, and
the final volume is 0.2 m3. The change in the specific internal energy of the gas in the process is u = -4.5
kJ/kg. There are no significant changes in the kinetic or potential energies of the gas. What is the net heat
transfer for the process?
gas
SOLUTION:
Apply conservation of energy the system of gas as shown in the figure below.
pA
gas
Esys = Qadded + Won sys
(3.192)
to sys
where
V =V2
Won sys =
V =V1
V =V2
pdV = p1V11.5

=c
dV
=
1.5
V =V1 V
1
0.5
p1V11.5 (V20.5 V10.5 ) (c is the constant in pV1.5 = c)
(3.193)
and
Esys = msys esys = msys usys (The kinetic and potential energy changes are negligible.)
(3.194)
Re-arranging Eqn. (3.192) and substituting Eqns. (3.193) and (3.194) gives:
1
Qadded = msys usys 0.5
p1V11.5 (V20.5 V10.5 )
(3.195)
to sys
Using the given values:

msys = 4 kg
usys = -4500 J/kg
p1
= 3*105 Pa
V1
= 0.1 m3
V2
= 0.2 m3
Qadded = -0.4 kJ (heat is leaving the system)
C. Wassgren
126
Example:
A passive solar house that is losing heat to the outdoors at an average rate of 50,000 kJ/hr is maintained at
22 C at all times during a winter night for 10 hr. The house is to be heated by 50 glass containers each
containing 20 L of water that is heated to 80 C during the day by absorbing solar energy. A thermostatcontrolled, 15 kW back-up electric resistance heater turns on whenever necessary to keep the house at
22 C.
a. How long will the electric heating system need to run during the night?
b. How long would the electric heater run during the night if the house did not incorporate solar heating?
heat loss of
50,000 kJ/hr
house
maintained at 22 C
50 glass containers filled with 20 L of water

each at an initial temperature of 80 C
SOLUTION:
Apply conservation of energy to the house.
Qinto
house
house
maintained at 22 C
Won
house
Esystem = Qinto + Won

system
(3.196)
system
The change in total energy of the house will consist of the change in the internal energy (potential and
kinetic energy changes will be negligible). Furthermore, the total internal energy change will include the
total energy change in the house structure, house air, and water tanks.
Esystem = U system = U house + U air + U water
(3.197)
structure
Since the house structure and air are maintained at a constant temperature, Uhouse = Uair = 0. Hence, Eqn.
(3.196) can be re-written as:
U water = Qinto + Won
(3.198)
system
system
C. Wassgren
127
The total change in the internal energy of the water (assuming an incompressible fluid) is given by:
U water = mwater cwater (T f ,water Ti ,water )
(3.199)
The total heat added to the house is:
Qinto = ( 50, 000 kJ/hr )(10 hr ) = 500,000 kJ
(3.200)
system
and the total work done on the house by the electric heater is:
Won = (15 kW ) t
(3.201)
system
where t is the time over which the heater operates.

Substitute Eqns. (3.199) (3.201) into Eqn. (3.198).
mwater cwater (T f ,water Ti ,water ) = 500, 000 kJ+ (15 kW ) t
(3.202)
Using the given parameters in Eqn. (3.202).

mwater = 50(20 L)(0.001 m3/L)(1000 kg/m3) = 1000 kg
cwater = 4.179 kJ/(kgK) (from a thermodynamics table)
Tf,water = 22 C
Ti,water = 80 C
t = 4.8 hrs Hence, the heater must be on for 4.8 hrs at night with the water tanks.
If the water containers were not present, then the left-hand side of Eqn (3.202) would be zero (Uwater = 0)
and:
t = 9.3 hrs Hence, the heater must be on for 9.3 hrs at night without the water tanks.
C. Wassgren
128
Conservation of Energy for a Control Volume

To write COE for a control volume, we utilize the Reynolds Transport Theorem (RTT) to convert our
system expression to a control volume expression. Lets first rewrite Eqn. (3.191) using the Lagrangian
derivative notation (were interested in how things change with respect to time as we follow a particular
system of fluid) and write the total energy of a system in terms of an integral:
D
(3.203)
e dV = Qinto system + Won system
Dt
Vsys

Esystem
Applying the Reynolds Transport Theorem and noting that the system and control volume are coincident at
the time we apply the RTT gives:
d
dt
e dV + e ( u
CV
rel
dA ) = Qinto CV + Won CV
(3.204)
COE for a CV
CS
Note: e=u+1/2V2+G where G is a conservative potential energy function with the specific gravitational
force given by fgravity = -G. For the remainder of these notes, G will be assumed to be G = gz (
fgravity = -g) where g is the acceleration due to gravity which points in the negative z-direction.
Now lets expand the rate of work (power) term into rate of pressure work (pdV power), shaft power, and
the power due to other effects such as viscous forces, electromagnetic forces, etc.
Won CV = Wpressure,on CV + Wshaft,on CV + Wother,on CV
(3.205)
In particular, we can write the rate of pressure work term in the following way:
urel
dWpressure,on CV = dFon CV u rel
n
= pdAn u rel
pressure force, pdA
= p ( u rel dA )
surface area, dA
The rate of pressure work over the entire CS is:
W
= p ( u dA )
(3.206)
pressure, on CV
rel
CS
Equation (3.206) is the rate at which pressure work is performed on the fluid flux through the control
surface. The rate at which pressure work is done on a moving solid boundary is included in the Wother
term.
Substituting Eqns. (3.206) and (3.205) into Eqn. (3.204), expanding the specific total energy term in the
surface integral, and bringing the rate of pressure work term to the LHS gives:
d
e dV +
u + p + 1 2 V 2 + gz ( u rel dA )
dt
(3.207)
CV
CS
= Qinto CV + Wshaft, on CV + Wother, on CV

The quantity (u+p/) shows up often in thermal-fluid systems and is given the special name of specific
enthalpy, h:
h u + p = u + pv
specific enthalpy
(3.208)
where v=V/m is the specific volume

Note that just as with internal energy, tables of thermodynamic properties typically list the value of the
specific enthalpy for various substances under various conditions.
Substituting Eqn. (3.208) into Eqn. (3.207) gives:
d
e dV +
h + 1 2 V 2 + gz ( u rel dA ) = Qinto CV + Wshaft, on CV + Wother, on CV
dt
CV
(3.209)
CS
C. Wassgren
129
Notes:
1.
For a flow where the total energy within the CV does not change with time (a steady flow), we have:
h + 1 V 2 + gz ( u dA ) = Q
+ W
+ W
(3.210)
rel
into CV
shaft, on CV
other, on CV
CS
Note that flows may be unsteady at the local level (e.g. detailed flow within a pump), but may be
steady on a larger scale (e.g. the average conditions within the pump housing).
2.
For a steady flow with a single inlet (call it state 1) and outlet (call it state 2) we can write COE as:
(h +
2V
+ gz
) (h +
2V
+ gz
= qinto CV + wshaft, on CV + wother, on CV
(3.211)
where q = Q m and w = W m (note that from COM the mass flow rate is a constant.)
i.
The quantity, , is known as the kinetic energy correction factor. It is a correction factor
accounting for the fact that the average velocity, V , does not contain the same kinetic energy as a
non-uniform distribution. For example, consider the kinetic energy contained in the two flow
profiles shown below:
The average specific kinetic energy flux is:
ke =
2V
2
( V dA ) 1 2 mV
We define the kinetic energy correction factor, :
pipe with
cross-sectional
area, A
The average velocity is found by:

1
V =
VdA
A
2V
( V dA )
A
1
2 mV
(3.212)
so that
ke =
2 mV
For a laminar flow in a circular pipe, the velocity profile is parabolic resulting in =2. For a
turbulent flow, 1 as the Reynolds number, defined as Re D VD , increases (suggesting
that the velocity profile becomes more uniform as the Reynolds number increases).
ii. The quantity:
hT = h0 h + 1 2 V 2 + gz
(3.213)
is referred to as the total specific enthalpy or the stagnation specific enthalpy. Note that for
gases, the gz term is often much smaller than the other terms and thus is often neglected.
iii. If the flow is adiabatic ( q = 0 ) and the rate of work by forces other than pressure can be neglected,
then:
hT = constant
adiabatic flow with no shaft or other work
(3.214)
3.
Now lets re-write Eqn. (3.211) but expand the specific enthalpy terms:
u + p + 1 V 2 + gz u + p + 1 V 2 + gz = q
+w
) (
2
into CV
shaft, on CV
+ wother, on CV
(3.215)
Re-arranging terms and dividing through by the gravitational acceleration gives:
C. Wassgren
130
V2
V2
+
+ z =
+
+ z
2g
2g
g
2 g
1
( u2 u1 qinto CV )
Wshaft, on CV
(3.216)
Wother, on CV
+
+

g
mg
mg
Note that each term in this equation has the dimensions of length. The terms are also referred to as
head quantities as defined below:
V2
velocity head
2g
p
pressure head
g
z elevation head
( u2 u1 qinto CV ) head loss, H (head lost due to heating of the fluid)
g

Wshaft on CV
shaft head, HS (head added due to shaft work; recall that W = T )

mg
Wother on CV
other head, HO (head added due to other work on the fluid)

mg
The equation in this form is also known as the Extended Bernoulli Equation:
V2
V2
+
+ z =
+
+ z H L + HS + H O
g
2
g
2
g
2
1
4.
(3.217)
Now lets examine the other work term more closely. Specifically, lets concern ourselves with the
work done by viscous effects. Consider the rate of viscous work done on the CV shown below:
urel
n
surface area, dA
dWviscous,on CV = dFviscous on CV u rel
viscous force, dFviscous
so that the total rate of viscous work acting on the

CS is:
W
= dF
u
viscous,on CV
viscous on CV
rel
CS
i.
Note that at a solid boundary, urel = 0 (due to the no-slip condition) so that the rate of viscous work
is zero at solid surfaces. If the flow is inviscid then urel 0 but dFviscous = 0 and so the rate of
viscous work is also zero.
ii. If the control volume is oriented so that the velocity vectors are perpendicular to the normal
vectors of the CS, then the rate of viscous work done on the CV will be zero.
dFviscous on CV u rel = 0
since the viscous force will be perpendicular to the velocity vector.
iii. The rate of viscous work may not be negligible if the control volume is chosen as shown below:
Viscous forces along streamline surfaces may be
significant if the shear stress, , is large:
Fv2
u
=
n
Fv1
where n is the direction normal to the streamlines.
streamlines
C. Wassgren
131
5.
The rate of work due to body and surface forces could have been written in a different form:
Wbody, on CV = ( u f B ) dV
CV
Wsurface, on CV =
( u f ) dA
S
CS
where fB is the body force per unit mass acting on the CV and fS is the surface force per unit area.
COE for a control volume using these relations is:
d
e dV + e ( u rel dA ) = Qinto CV + ( u f B ) dV + ( u f S ) dA + Wother, on CV
dt
CV
CS
CV
CS
where e=u+1/2V2.
Note that the total energy does not include the potential energy since the potential energy is contained
in the rate of work due to body forces term. The rate of gravitational work can be included in the LHS
integral (i.e. we can write e=u+1/2V2+gz) assuming that the gravitational acceleration remains steady
and is conservative (i.e. derivable from a potential function).
Lets examine a few examples.
C. Wassgren
132
Example:
Consider a large classroom on a hot summer day with 150 students, each dissipating 60 W of heat. All the
lights, with 4.0 kW of rated power, are kept on. The room has no external walls, and thus heat transfer
through the walls and the roof is negligible. Chilled air is available at 15 C and the temperature of the
return air is not to exceed 25 C. Determine the required flow rate of air, in kg/s, that needs to be supplied
to the room to keep the average temperature of the room constant.
SOLUTION:
Apply conservation of energy to a control volume that encloses the classroom.
d
e dV +
h + 1 2 V 2 + gz ( u rel dA )
m inlet@15 C
dt
CV
CS
= Qinto CV + Wshaft, on CV + Wother, on CV

where
d
dt
e dV = 0
(steady state)
Q lights
Qstudents
m return@25 C
CV
(h +
2V
)
)
+ gz ( u rel dA ) = ( mh
+ ( mh
= m ( hreturn hinlet )
inlet
return
CS
(Note that in steady operation: m inlet = m return . In addition the changes in the airflow velocity and
elevation will be negligible since the velocities and elevation differences are expected to be small.)
Qinto CV = Qstudents + Q lights
Wshaft, on CV = Wother, on CV = 0
m ( hreturn hinlet ) = Qstudents + Q lights
m =
Qstudents + Q lights
(3.218)
hreturn hinlet
Assuming air is a perfect gas:
hreturn hinlet = c p (Treturn Tinlet )
(3.219)
where cp is the specific heat of air at constant pressure. Note that if we dont consider air as a perfect gas
then we can look up the specific enthalpy of air (at 1 atm) in a thermodynamics table.
Using the given data:
Qstudents = 150 * 60 W = 9 kW
Q
= 4 kW
lights
Treturn = 25 C = 298 K
Tinlet = 15 C = 288 K
c p = 1000 J
( kg K )
m = 1.3 kg/s
C. Wassgren
133
Example:
Determine the maximum pressure increase across the 10 hp pump shown in the figure.
d2 = 1.5 in
water
V1=30 ft/s
d1 = 1 in
pump
10 hp
SOLUTION:
Apply conservation of energy, in the form of Eqn. (3.217), across the pump.
p
V2
V2
+
+ z =
+
+ z H L + HS + H O
2g
2g
g
2 g
1
where
p = p2 p1 (This is the pressure rise were trying to determine.)
2
d
V2 = V1 1 (from conservation of mass)
d2
2 1 1 (assuming turbulent flow)
z2 z1 0 (negligible elevation difference between the inlet and outlet)
H L = 0 (no losses this will give the maximum pressure rise across the pump)
W
H S = S (from the definition of shaft head)

mg
H O = 0 (there is no other work being performed other than shaft work)
4
V12 WS
p V12 d1
+
+
=

g 2 g d2
2 g mg
4
p V12 d1 WS
1 +
=

g 2 g d 2 mg
C. Wassgren
(3.220)
134

V1 = 30 ft/s
d1 = 1.0 in = 8.3*10-2 ft
d 2 = 1.5 in = 1.3*10-1 ft
V2 = 13 ft/s
H 2 0 = 1.1*10-5 ft 2 /s
Re1 =
V1d1
Re 2 =
V2 d 2
m = V1
H 2 0 =230,000 (turbulent flow assumption ok!)

H 2 0 =150,000 (turbulent flow assumption ok!)
d12
= 1.94 slug/ft
) ( 30 ft/s )
8.3*102 ft 2
) = 3.8 slug/s
4
4

10
hp
550
ft
lb
/
s
hp
))
(
W
f (
HS = S =
= 45 ft (Note: 1 lbf = 1 slugft/s2)
2

mg
( 3.8 slug / s ) 32.2 ft/s
= 56 ft
p = 3500 psf = 24 psi
C. Wassgren
135
Example:
The velocity profile for a particular pipe flow is linear from zero at the wall to a maximum of uc at the
centerline. Determine the average velocity and the kinetic energy correction factor.
r
uc
R
SOLUTION:
The average velocity is found by setting the volumetric flow rate using the average velocity profile equal to
the volumetric flow rate using the real profile.
Qavg = Qreal
(3.221)
profile
profile
r =R
u R2 =
r =0
1 ( 2 rdr )
R
r=R
= 2 uC
r2
r
R
r =0
dr
r =R
1
1 r3
= 2 uC r 2
3 R r =0
2
u = 13 uC
(3.222)
The kinetic energy correction factor, , is found by equating the kinetic energy flux using the average
velocity with the kinetic energy flux using the actual velocity profile.
1
2
( u R2 ) u 2 =

r =R
1
2
r =0
= m
r

r
uC 1 R ( 2 rdr ) uC 1 R

= dm
r =R
1
2
uC3 1
r =0
r=R
= uC3
r
( 2 rdr )
R
(3.223)
r =0 1 R ( rdr )
where u = 13 uC . Solving the previous equation for gives:
27
= 2.7
10
C. Wassgren
(3.224)
136
Example:
Air, treated as an ideal gas, flows through the turbine and heat exchanger arrangement shown in the figure
with the data for the flow streams also indicated. Heat transfer to the surroundings can be neglected, as can
all kinetic and potential energy effects. Determine the temperature T3, in K, and the power output of the
second turbine, in kW, at steady state.
power out = 10,000 kW
T1
T3 = ?
p3 = p2
T2 = 1100 K
p2 = 4 bars
air
T1 = 1400 K
p1 = 12 bars
T2
power out = ?
T4 = 980 K
p4 = 1 bar
heat
exchanger
air
T5 = 1480 K
p5 = 1 bar
mass flow rate = 1200 kg/min
T6 = 1200 K
p6 = p5
SOLUTION:
Apply conservation of energy to each component. Assume steady, 1D flow with negligible heat transfer to
the surroundings and negligible kinetic and potential energy changes.
power out = 10,000 kW

T1
T2 = 1100 K
p2 = 4 bars
air
T1 = 1400 K
p1 = 12 bars
From conservation of energy:
m 12 ( h2 h1 ) = Winto
(3.225)
CV
where
h1
=
h2
=

Winto =
1515.4 kJ/kg (from thermodynamics tables for air at T1 = 1400 K)

-10,000 kW
CV
and m 12 is the mass flow rate at stations 1 and 2.

Substitute and solve for the mass flow rate.
m 12 = 28.2 kg/s
C. Wassgren
137
T3 = ?
p3 = p2
T2 = 1100 K
p2 = 4 bars
heat
exchanger
air
T5 = 1480 K
p5 = 1 bar
mass flow rate = 1200 kg/min
T6 = 1200 K
p6 = p5

m 12 ( h3 h2 ) = m 56 ( h5 h6 )
where
h3
h2
m 12
h6
h5
m 12
=
=
=
=
=
=
(3.226)
?
28.2 kg/s
1200 kg/min = 20 kg/s
Substitute and solve for the enthalpy at station 3.

h3 = 1398.0 kJ/kg T3 = 1300 K (from interpolation of thermodynamics tables for air)
T3 = ?
p3 = p2
T2
power out = ?
T4 = 980 K
p4 = 1 bar

m 34 ( h4 h3 ) = Won
(3.227)
CV
where
h4
h3
m 34
=
=
=

1398.0 kJ/kg (from previous work)
28.2 kg/s (= m 12 )
Substitute and solve for the mass flow rate.

Won = 10,570 kW (work is extracted from the turbine)
CV
C. Wassgren
138
Example:
In a proposed jet propulsion system for an automobile, air is drawn in vertically through a large intake in
the roof at a rate of 3 kg/s, the velocity through this intake being small. Ambient pressure and temperature
are 100 kPa (abs) and 30 C. This air is compressed, heated, and then discharged horizontally out a nozzle
at the rear of the automobile at a velocity of 500 m/s and a pressure of 140 kPa (abs). If the rate of heat
addition to the air stream is 600 kW, find the nozzle discharge area and the thrust developed by the system.
SOLUTION:
The nozzle discharge area can be determined from the mass flow rate. Note that since the flow is steady,
the outlet and inlet mass flow rates are the same.
m 0 = oVo Ao = m i
(3.228)
m i
Ao =
(3.229)
oVo
The outlet air density can be expressed in terms of the outlet pressure and temperature using the ideal gas
law.
m RTo
Ao = i
(3.230)
Vo po
The outlet temperature can be determined by applying COE to the following CV.
mi , pi , Ti
Q into
CV
Vo , po
F
x
d
+ Wother
e dV + CS hT ( u rel dA ) = Qinto
dt CV
CV
on CV
where
d
e dV = 0 (steady flow)
dt CV
h ( u
T
rel
(3.231)
dA ) = m o ( ho + 12 Vo2 ) m i ( hi )
CS
Note that: (1) since the flow is steady, m o = m i = m , (2) the potential energy differences between
the inlet and outlet are negligible, and (3) the inlet kinetic energy is negligible.
Q into = Q into (this quantity is given in the problem statement)
CV
CV
Wother = 0
on CV

m ( ho + 12 Vo2 hi ) = Q into
(3.232)
CV
Assuming perfect gas behavior (h = cpT):

Qinto
V2
CV
To = Ti +
o
c p m 2c p
C. Wassgren
(3.233)
139

Ti
= 30 C = 303 K
m i
= 3 kg/s
Q
= 600 kW
into
CV
cp
V0
= 1005 J/(kgK)
= 500 m/s
T0 = 378 K
R
To
po
V0
=
=
=
=
0
0
287 J/(kgK)
378 K
140 kPa
500 m/s
= 1.29 kg/m3
= 4.6*10-3 m2
To determine the thrust, apply the linear momentum equation in the x-direction to the same control volume.
d
(3.234)
ux dV + CS ux ( u rel dA ) = FB, x + FS , x
dt CV
where
d
ux dV = 0
dt CV
u ( u
x
rel
dA ) = Vo m o
CS
FB , x = 0
FS , x = F ( po patm ) Ao (the pressures are given in terms of absolute pressures)

Vo m o = F ( po patm ) Ao
F = Vo m o + ( po patm ) Ao
(3.235)

V0
= 500 m/s
m o
= 3 kg/s
po
= 140 kPa
pi
= 100 kPa
0
= 4.6*10-3 m2
F = 1670 N
C. Wassgren
140
9.
The Second Law of Thermodynamics
While the first law of thermodynamics tells use that energy must be conserved during a process, it doesnt
tell us whether or not the process will actually occur. The Second Law of Thermodynamics states that
processes occur in a certain direction. A process cannot take place unless it satisfies both the First and
Second Laws of Thermodynamics.
A detailed analysis of the Second Law is beyond the scope of these notes. A brief review of some of the
more significant points will be given however.
The Second Law of Thermodynamics can be stated in many ways. For example, two common and
equivalent ways to state the 2nd Law are:
Kelvin-Planck Statement of the 2nd Law:
It is impossible for any device that operates on a cycle to receive heat from a single reservoir and
produce a net amount of work.
Clausius Statement of the 2nd Law:
It is impossible to construct a device that operates on a cycle and produces no effect other than the
transfer of heat from a lower-temperature body to a higher-temperature body.
An important corollary to the Clausius Statement is the Clausius Inequality:
Qinto system
0
Clausius Inequality
(3.236)
T
where Qinto system is the heat added to the system and T is the absolute temperature of the system. The
equality sign holds strictly for reversible processes. This inequality is used in the derivation of the
property known as entropy. Rather than show the derivation for this property, we will simply define it
below.
entropy, S
Entropy, S, is an extensive system property defined in the following manner:
Qinto system
dSsystem
T
rev
(3.237)
where Q is the heat added to the system, T is the absolute temperature of the system, and the process
by which the heat is added is reversible.
The change in a systems entropy as it undergoes a change of state is found by integrating Eqn. (3.237)
:
2
S2 S1 =
Qinto system
T
C. Wassgren
(3.238)
rev
141
Notes:
1. reversible and irreversible processes
A reversible process is one in which the system is in a state of equilibrium at all points in its path.
In a reversible process, the system and the surroundings can be restored exactly to their initial
states. An irreversible process is one where the system is not in a state of equilibrium at all points
in its path. The system and surroundings cannot be returned to their exact initial states in an
irreversible process. Note that all natural processes are irreversible. Several effects causing
irreversibility include viscosity, heat conduction, and mass diffusion.
2.
Entropy is a system property so that the change in the entropy depends only on the endpoints of
the process and not on the path taken during the process. Thus, the change in entropy between
two states is independent of whether or not the process is reversible or irreversible. Equation
(3.238) simply gives a method for calculating the entropy change using a reversible process. Note
that if a process is irreversible, we have from the Clausius inequality (Eqn. (3.236)):
Qinto system
T
Qinto system
Qinto system
T
rev
2

= S1 S 2
S2 S1
Qinto system
(3.239)
We can combine Eqns. (3.238) and (3.239) to give the general equation:
2
S2 S1
Qinto system
(3.240)
or in differential form:
Qinto system
dSsystem
T
where the equality holds strictly for reversible process paths.
(3.241)
3.
A reversible, adiabatic process is also an isentropic process since dS=Q/T (reversible) and Q=0
(adiabatic) so that dS=0.
4.
The value of entropy for various substances at various states can typically be found in
thermodynamic property tables.
5.
Statistical thermodynamics shows us that entropy can also be interpreted as a measure of the
molecular disorder of a system.
C. Wassgren
142
Second Law of Thermodynamics for a Control Volume

As with COM, the LME, the AME, and COE, we will convert our system form of the 2nd Law to a control
volume form using the Reynolds Transport Theorem. To do so, we should first write the 2nd Law in terms
of time rates of change (refer to Eqn. (3.241)):
DSsystem Qinto system
(3.242)
Dt
T
where the Lagrangian derivative notation has been used to remind us that were following a system. Now
lets write the equation in terms of an integral so that we can have variations in the properties of the system:
qinto system
D
s dV
dV
(3.243)
Dt
T
Vsys
Vsys
where s is the specific entropy and qinto system is the rate of heat addition into the system per unit volume.
The RHS of Eqn. (3.243) represents the sum of the heat transfer input to absolute temperature ratio for
every particle in the system.
After applying the Reynolds Transport Theorem to convert to a control volume perspective, we have:
qinto CV
d
s dV + s ( u rel dA )
dV
(3.244)
dt
T
CV
CS
CV
where the equality holds strictly for reversible heat addition processes.
C. Wassgren
143
Example:
An inventor claims to have developed a device requiring no work input or heat transfer, yet able to produce
steady state hot and cold air streams as shown in the figure. Evaluate this claim assuming the ideal gas
model for air and ignoring kinetic and potential energy effects.
air at 60 C, 2.7 bars
air at 20 C, 3 bars
no heat or work addition
SOLUTION:
Apply conservation of mass, conservation of energy, and the 2nd Law to the control volume shown below.
2
1
air at 20 C, 3 bars
3
no heat or work addition
Conservation of Mass:
d
dt CV
where
d
dt CV
rel
(3.245)
dA = m 3 + m 2 m 1
CS

m 3 + m 2 m 1 = 0
m 3
m
= 1 2
m 1
m 1
C. Wassgren
(3.246)
144
Conservation of Energy:
d
2
+ Won
e dV + CS ( h + 12 V + gz ) ( u rel dA ) = Qinto
dt CV
CV
CV
where
d
e dV = 0 (steady flow)
dt CV
(h +
1
2
(3.247)
V 2 + gz ) ( u rel dA ) = m 3 h3 + m 2 h2 m 1h1
CS
Q into = Won = 0 (no heat or work addition)

CV
CV

m 3 h3 + m 2 h2 m 1h1 = 0
m 3
m
h3 = h1 2 h2
m 1
m 1
m 2
m 2
h2
1
h3 = h1
m 1
m 1
(3.248)
m 2 h1 h3
=
m 1 h2 h3
(3.249)
From thermodynamics tables for air at the given inlet and outlet temperatures:
h1 = 293.2 kJ/kg (T1 = 20 C = 293 K)
h2 = 333.3 kJ/kg (T2 = 60 C = 333 K)
h3 = 273.1 kJ/kg (T3 = 0 C = 273 K)
Hence,
m 2
= 0.334 and
m 1
m 3
= 0.666
m 1
(3.250)
Note that the outgoing mass flow rates are both positive, consistent with the problem description.
C. Wassgren
145
2nd Law of Thermodynamics:

q
d
s dV + s ( u rel dA ) into CV dV
dt CV
T
CS
CV
where
d
s dV = 0 (steady flow)
dt CV
s ( u
rel
(3.251)
dA ) = m 3 s3 + m 2 s2 m 1 s1
CS
qinto CV
dV = 0 (no heat added to the control volume)
T
CV

?
m 3 s3 + m 2 s2 m 1 s1 0
?
m 3
m
s3 + 2 s2 s1 0
m 1
m 1
m 2
1
m 1
m 2
s2 s1 0
s3 +
m 1
?
m
( s3 s1 ) + 2 ( s2 s3 ) 0
m 1
(3.252)
Note that for an ideal gas, the specific entropy at a given temperature and pressure can be determined
(derived in Chapter 11) by:
p
s (TB , pB ) s (TA , p A ) = s 0 (TB ) s 0 (TA ) R ln B
(3.253)
pA
0 0
p3 m 2 0 0
p2 ?
(3.254)
s2 s3 R ln 0
s3 s1 R ln +
p1 m 1
p3
From thermodynamics tables for air at the given conditions:

s01 = 1.6783 kJ/(kgK) (T1 = 20 C = 293 K)
s02 = 1.8069 kJ/(kgK) (T2 = 60 C = 333 K)
s03 = 1.6073 kJ/(kgK) (T3 = 0 C = 273 K)
and from the given conditions:
p1 = 3.0 bar
p2 = 2.7 bar
p3 = 2.7 bar
Thus, Eqn. (3.254) simplifies to:
0.0259 kJ/ ( kg K ) > 0
(3.255)
nd
and the 2 Law of Thermodynamics is satisfied.

Since conservation of mass, conservation of energy, and the 2nd Law are all satisfied, the claim of the
inventor is not unreasonable.
C. Wassgren
146
Review Questions
1. What is meant by the Eulerian and Lagrangian perspectives?
2. Describe the Reynolds Transport Theorem in words? Why is it used?
3. State, both in words and in mathematics, the Lagrangian forms of conservation of mass, Newtons 2nd
Law, and conservation of energy.
4. Why is it important to draw a well-defined control volume when applying conservation of mass, the
linear momentum equations, conservation of energy, or the 2nd Law of Thermodynamics?
5. What do each of the terms represent in the Lagrangian and Eulerian statements of conservation of
mass?
6. Does conservation of mass depend upon the frame of reference?
7. Why is it important to draw a well-defined frame of reference when applying the linear momentum
equations?
8. What does each of the terms represent in the Eulerian form of the linear momentum equation?
9. What restrictions are placed on the frame of reference when applying the LMEs?
10. In order to change the momentum of a flow, what must act on the flow?
11. Give examples of body and surface forces.
12. Explain what urel is. For what circumstances will urel and u be the same?
13. Why is the dot product used (ureldA) in determining the flow rate out of a control volume?
14. Can one apply the non-inertial form of the LME to an inertial frame of reference? How about applying
the inertial form of the LME to a non-inertial frame of reference?
15. What types of frames of reference can be considered inertial? Give examples of frames of reference
that are not inertial.
16. Given the following where MCV is the mass in a control volume, m is a mass flow rate, and t is time:
dM CV
= m
dt
will the following always be true (where M0 is the control volume mass at t = 0)? Explain your
answer.

M CV = M 0 mt
17. Describe what each term represents in the Eulerian form of the angular momentum equation.
18. Why isnt the intrinsic angular momentum of the fluid included in the angular momentum equation?
19. Consider a precessing, spinning top. Is its angular momentum conserved?
20. Describe what each term represents in the Eulerian form of conservation of energy.
21. What is meant by the term adiabatic?
22. Why are shear work terms often (but not always!) neglected in conservation of energy?
23. What is the definition of enthalpy?
24. In the following form of conservation of energy, where are the terms involving the work due to
movement in a gravity field and the work due to pressure forces?
d
2
+ Won
e dV + CS ( h + 12 V + gz ) ( u rel dA ) = Qinto
dt CV
CV
CV
25. Describe what each term represents in the Eulerian form of the 2nd Law of Thermodynamics.
26. What is the definition of entropy in terms of heat and temperature?
27. What is meant by the term reversible? Give some examples of physical processes that result in
irreversibility.
28. How are adiabatic, reversible processes related to isentropic ones?
C. Wassgren
147
Chapter 04:
Differential Analysis
1. Index Notation
2. Continuity Equation
3. Review of Stress
4. Momentum Equations
5. Fluid Element Deformations
6. Stress-Strain Rate Relations for a Newtonian Fluid
7. Acceleration in Streamline Coordinates
8. Eulers Equations in Streamline Coordinates
9. Energy Equation
10. Entropy Equation
11. Vorticity Dynamics
12. Vorticity Transport Equation
13. Bernoullis Equation
14. Kelvins Theorem
15. Croccos Equation
C. Wassgren
Chapter 04: Differential Analysis
148
1.
Index Notation (aka Tensor Notation)
Index notation is a compact way of writing numbers and equations. Index notation is extremely useful for
representing the equations of fluid mechanics, derivations, etc.
Examples:
1. a1, a2, a3 can be written as: ai where i=1, 2, 3
2.
A matrix of numbers can be written as:

a11 a12 a13
a
where i = 1, 2, 3 and j = 1, 2, 3
21 a22 a23 = aij
a31 a32 a33
3.
ai+bi represents three numbers:

a1+b1, a2+b2, a3+b3
4.
Aij = Bij represents nine equations:

A11=B11, A12=B12, A13=B13, A21=B21, , A33=B33
5.
Aijk=Bijk represents 27 equations:

A111=B111, A112=B112, A113=B113, A211=B211, , A333=B333
Free Indices
A free index is an index that appears exactly once in a term. Each term in an equation must have the
same free indices. A repeated index is one that appears twice in a term. No index may appear more
than twice in a term.
Examples:
1. aijk bj= cik
CORRECT
(two free indices: i and k, one repeated index: j)
2.
aijbjk = clm
INCORRECT
(two free indices on each side of the equation (i, j and l, m) but
theyre not the same!)
3.
aijbj = cij
INCORRECT
(i is the only free index on the LHS while i and j are free indices on
the RHS)
C. Wassgren
149
Summation Convention
If a subscript appears exactly twice in a term (i.e. its a repeated index), then summation over that
subscript from 1 to 3 is implied.
Examples:
3
1.
aii =
ii
= a11 + a22 + a33
i =1
3
2.
aij b j =
a b
ij j
= ai1b1 + ai 2 b2 + ai 3b3
j =1
3
3.
Aij Bij =
A B
ij ij
= ( A11 B11 + A12 B12 + A13 B13 ) + + ( A31 B31 + A32 B32 + A33 B33 )
i =1 j =1
Notes:
1. Repeated indices are dummy indices:
2.
No index may appear more than twice in a term:

INCORRECT
ai bi ci
Aiii
3.
aii = ajj = akk
INCORRECT
The summation convention is suspended by writing no sum or by underlining one of the

repeated subscripts.
ii ( no sum ) = 11 , 22 , 33
C. Wassgren
150
Kroneckers Delta
Kroneckers Delta, ij, is the 2 index symbol defined by:
0 i j
ij =
1 i = j
Notes:
1.
2.
12 = 13 = 21 = 23 = 31 = 32 = 0
11 = 22 = 33 = 1
The Kronecker Delta is the identity matrix:
1 0 0
ij = 0 1 0
0 0 1
Examples:
1. Show that ij a j = ai .
i = 1: 1 j a j = 11 a1 + 12 a2 + 13 a3 = a1

=1
=0
=0
i = 2 : 2 j a j = 21 a1 + 22 a2 + 23 a3 = a2

=0
=1
=0
i = 3 : 3 j a j = 31 a1 + 32 a2 + 33 a3 = a3

=0
2.
=0
=1
Show that ii = 3 .
ii =
11 +
22 +
33 = 3

=1
=1
=1
C. Wassgren
151
Permutation (aka Alternating) Unit Tensor

The permutation tensor, ijk, is the 3-index symbol defined as:
+1 for 123, 231, 312
ijk = 1 for 321, 213,132

0 for all other permutations
Notes:
1. Its convenient to remember the following pictures for determining the proper sign of the
permutation tensor:
1
3
1
3
2.
Switching any two indices changes the sign of the permutation tensor, e.g.: ijk = - ikj
3.
A convenient identity:
ijk ilm = jl km jm kl
Example:
1. Show that ijk ijk = 6 .
1k k1 + 2 k k 2 + 3k k 3
ijk ijk = jj
kk jk kj = 9

= 3 =3
=1

=1
=1
= 93
=6
C. Wassgren
152
Tensors
A tensor of rank r is a quantity having nr components in n-dimensional space (e.g. a tensor of rank 2 in
3D has 32 = 9 components). The components of a tensor expressed in two different coordinate systems
are related by:
Tijk m = is jt ku mvTstuv
where is are the direction cosines between the ei and es axes.
Notes:
1. A tensor of rank 2 is often called a dyad, e.g. Aij (two free subscripts).
2.
A tensor of rank 1 is called a vector, e.g. ai (one free subscript).
3.
A tensor of rank 0 is called a scalar, e.g. c (zero free subscripts).
4.
The vector notation for a dyad is often given as: A
C. Wassgren
153
Basic Mathematical Operations

Addition:
Two tensors of equal rank can be added to yield a tensor of the same rank.
Cijk = Aijk + Bijk
Multiplication:
If a tensor, A, having rank, a, is multiplied by tensor, B, having rank, b, then a

tensor, C, of rank, a+b, results.
Cijkrst = Aijk Brst
Example:
Aij
Brs = Cijrs

2nd order 2nd order
Transpose:
4th order
The transpose of a tensor is given by:

TijTk = Tk ji
Example:
AijT = A ji
If you represent the components of a 2nd order tensor in matrix form, then the
transpose is equivalent to swapping the off-diagonal components.
Symmetric:
A symmetric tensor is one that has the property:

Tijk = Tk ji
A symmetric tensor is equal to its transpose.
Anti-Symmetric:
An anti-symmetric tensor is one that has the property:

Tijk = Tk ji
Notes:
1. A tensor, symmetric in ijk, is often indicated using the notation: T(ijk).
2.
A tensor, anti-symmetric in ijk, is often indicated using the notation: T[ijk].
3.
T( ijk ) =
1
2
(Tijk + Tk ji )
4.
T[ijk ] =
1
2
(Tijk Tk ji )
5.
Tijk = T( ijk ) + T[ijk ]
6.
Tii = T11 + T22 + T33 = trace (Tij )
C. Wassgren
154
Basic Mathematical Operations (cont.)

Dot Products
(aka Inner Products):
a b = ai bi
a B = ai Bij
A b = Aij b j
A B = Aij B jk
A : B = Aij B ji
A : BT = Aij Bij
T = ab Tij = ai b j
ci = ijk a j bk
Cross Product:
c = ab
Gradient:
( )i
Divergence:
a =
ai
= ai ,i
xi
A =
(Note: ab = ( ba ) .)
= ,i
xi
Aij
xi
= Aij ,i
Curl:
( a )i = ijk
Laplacian (scalar):
2 =
ak
= ijk ak , j
x j
2
= ,ii
xi xi
Gauss Theorem (aka Divergence Thm):
a n dS = adV
dS
a n dS = a
i ,i dV
i i
where S is the surface enclosing the volume V and n is the outward pointing
unit normal vector for the surface area element dS.
Stokes Theorem:
a dl = ( a ) n dS
C
dl
a dl =
i
dS
ijk ak , j ni dS
where the curve C defines the surface S and dl is a vector tangent to the curve C at
a particular point. (Note that the shape of the surface on which Stokes Theorem is
to be applied must be known so that the relation between the surface area and
contour is well defined.)
References:
1.
2.
3.
4.
Borg, S.F., Matrix-Tensor Methods in Continuum Mechanics, Van Nostrand, 1963.

Myklestad, N.O., Cartesian Tensors, Van Nostrand, 1967.
Synge, J.L. and Schild, A., Tensor Calculus, University of Toronto Press, 1949.
Aris, R., Vectors, Tensors, and the Basic Equations of Fluid Mechanics, Prentice-Hall, 1962.
C. Wassgren
155
Example:
Prove that the following are true using index notation:
(a b) c = a (b c) = (c a) b
t (u v ) = u (t v ) v ( t u )
u v = v u
SOLUTION:
(a b) c = ijk a j bk ci
= jki bk ci a j = (b c) a
= kij ci a j bk = (c a) b
[ t (u v)]i = ijk t j klmul vm

= ijk klm t j ul vm
= kij klm t j ul vm
= ( il jm im jl ) t j ul vm
= t j ui v j t j u j vi
= u ( t v ) v ( t u ) i
[u v ]i = ijk u j vk
= ikj vk u j
= [ v u ]i
C. Wassgren
156
Example:
Show, using index notation, that:
= 0
SOLUTION:
( )i
= ijk ( , k ), j
= ijk , kj
(4.1)
= ikj , kj
= ijk , jk
= ijk , kj (the order of the differentiation doesnt matter)
(4.2)
The only way for Eqns. (4.1) and (4.2) to be equal is if they both equal zero, i.e.:
ijk , kj = 0 = ijk , kj
(4.3)
Therefore:
( )i = 0 = 0
C. Wassgren
157
Example:
Solve:
ai = ijk b jk
for b[jk].
SOLUTION:
imn ai = imn ijk b jk

= ( mj nk mk nj ) b jk
= bmn bnm
= 2 12 ( bmn bnm )
= 2b[mn]
b[ jk ] = 12 ijk ai
C. Wassgren
158
2.
Continuity Equation (aka COM for a differential CV )
The continuity equation, which is simply conservation of mass for a differential fluid element or control
volume, can be derived several different ways. Two of these methods are given below.
Method 1:
Apply the integral approach to a differential control volume:
y
dy
z
dz
dx
Assume that the density and velocity are and u, respectively, at the control volumes center. The
mass fluxes through each of the side of the control volume are given by:
m x,center
m in through left = m x,center +
12 dx
x
= ( u x dydz ) + ( u x dydz ) 12 dx
x
= u x + ( u x ) 12 dx ( dydz )
(where m x ,center is the mass flux in the x-direction at the center of the control volume (Recall
the Taylor Series approximation discussed in Chapter 01.))
m out through right = u x + ( u x ) 12 dx ( dydz )
m in through bottom = u y +
u y 12 dy ( dxdz )
y
)(
m out through top = u y +

u y ( 12 dy ) ( dxdz )
y
m in through back = u z + ( u z ) ( 12 dz ) ( dxdy )

z
m out through front = u z + ( u z ) ( 12 dz ) ( dxdy )
Thus, the net mass flow rate into the control volume is given by:
m net,into CV = ( u x ) +
u y + ( u z ) ( dxdydz )
y
z
x
The rate at which mass is increasing within the control volume is given by:
m
= ( dxdydz )
t within CV t
(4.4)
where is the density at the center of the control volume. Note that since the density varies linearly
within the CV (from the Taylor Series approximation), the average density in the CV is .
C. Wassgren
159
Conservation of mass states that the rate of increase of mass within the control volume must equal the
net rate at which mass enters the control volume:
m
= m net,into CV
t within CV
( dxdydz ) = ( u x ) + u y + ( u z ) ( dxdydz )
t
+ ( ux ) +
u y + ( uz ) = 0
t x
y
z
Written in a more compact form:
+ ( u) = 0
t
Or in index notation form:

+
( ui ) = 0
t xi
(4.5)
(4.6)
(4.7)
Method 2:
Recall that the integral form of COM is given by:
d
dV + u rel dA = 0
dt
CV
CS
Consider a fixed control volume so that:

d
dV =
dV and u rel = u
dt
t
CV
CV
By utilizing Gauss Theorem (aka the Divergence Theorem), we can convert the area integral into a
volume integral:
u dA = ( u ) dV
CS
CV
Substitute these expressions back into COM to get:

t + ( u ) dV = 0
CV
Since the choice of control volume is arbitrary, the kernel of the integral must be zero, i.e.:
+ ( u) = 0
same result as before!
t
C. Wassgren
160
Notes:
1. For a fluid in which the density remains uniform and constant ( = constant), the continuity
equation simplifies to:
u = 0 continuity equation for a constant density fluid
2.
An incompressible fluid is one in which the density of a particular piece of fluid remains
constant, i.e.:
D
= 0 definition of an incompressible fluid
Dt
Note that an incompressible fluid does not necessarily imply that the density is the same
everywhere in the flow (i.e. its not uniform). An example of such a flow would be a stratified
flow in the ocean where the density of various layers of ocean water varies due to salinity and
temperature variations.
The density of fluid particles
varies from layer to layer but
remains constant within a layer.
A flow with a constant and uniform density fluid, however, is an incompressible flow.
The continuity equation for an incompressible fluid can be found by using the definition of an
incompressible flow:
D
=0=
+ ( u )
= ( u )
t
t
Dt
Substituting into the continuity equation:
+ ( u ) = 0
t
( u ) + ( u ) = 0
( u ) + ( u ) + ( u ) = 0
u = 0
3.
continuity equation for an incompressible flow
(4.8)
Another useful form of the continuity equation is as follows:

u

+
( ui ) = 0 = + ui + i
t xi
t
x
xi
i
= D
Dt
u
D
= i
Dt
xi
(4.9)
4.
The continuity equation (Eqn. (4.7)) is valid for any continuous substance (e.g., a solid as well as a
fluid).
5.
Equation (4.7) is referred to as the conservative form of the continuity equation while Eqn. (4.9) is
the non-conservative form. The conservative form implies that the equation represents an Eulerian
viewpoint of the continuity equation. The non-conservative form represents the Lagrangian
viewpoint.
C. Wassgren
161
Example:
The y-velocity component of a steady, 2D, incompressible flow is given by:
u y = 3 xy x 2 y
Determine the most general velocity component in the x-direction for this flow.
SOLUTION:
Consider the continuity equation:
u x u y
+
=0
x
y
(4.10)
u y
u x
=
=
3xy x 2 y = 3 x + x 2
x
y
y
Integrate ux with respect to x.
u x = 32 x 2 + 13 x3 + f ( y )
(4.11)
where f(y) is an unknown function of y.
C. Wassgren
162
Example:
A piston compresses gas in a cylinder by moving at a constant speed, V. The gas density and the piston
length are initially 0 and L0, respectively. Assume that the gas velocity varies linearly from velocity, V, at
the piston face to zero velocity at the cylinder wall (at L). If the gas density varies only with time,
determine (t).
V gas with
density, (t)
L(t)
SOLUTION:
As given in the problem statement, assume the gas velocity, u, varies linearly with distance x from the
piston face with the boundary conditions: u(x = 0) = V and u(x = L(t)) = 0.
x
u ( x, t ) = V 1
(4.12)
L (t )
However, the piston moves at a constant speed so that:
L ( t ) = L0 Vt
(4.13)
x
u ( x, t ) = V 1
L0 Vt
(4.14)
Apply the continuity equation assuming 1D flow.

+ ( u ) = 0
t x
d
u
=
(Note that = (t).)
dt
x
(4.15)
1
=V
dt
L0 Vt
t =t
=V
dt
0 Vt
t =0

L Vt
ln = ln 0
0
L0
Vt
= 1
0 L0
C. Wassgren
(4.16)
163
3.
Review of Stress
Traction Vector (aka Stress Intensity, Stress Vector)

Consider a small area on or within a deformable body subject to both surface and body forces as shown
below.
M
A
The net force and moment acting on the area A

, where A is the magnitude of the area and is its
corresponding unit normal vector, are denoted by F and M, respectively. The traction vector (aka stress
intensity or stress vector) on the surface is defined as:
F
T lim
(a vector with dimensions of force per unit area)
(4.17)
A0 A
and the couple stress vector on the surface is defined as:
M
C lim
(a vector with dimensions of torque per unit area)
(4.18)
A0 A
Notes:
1. Usually C = 0 since body moments are rare. An example of a case in which body moments (and thus
the couple stress vector) is not zero is in a material comprised of polar or magnetic elements (i.e.
molecules or domains) subject to an external electric or magnetic field. In such a case, the material
elements will try to orient themselves in a preferred direction. The couple stress vector may also be
non-zero in powders where the component particles have an aspect ratio greater than one and, hence,
tend to re-orient when under load.
2. To completely describe the traction at a point, we need to know T for all orientations, , of the
differential surface area at that point.
3. Let ij be the components of the traction vector T. Consider, for example, the tractions on the faces of
a differential cube as shown below. The traction on each face in terms of its components is:
x2
T1 = 11e1 + 12 e 2 + 13e 3
T2
T2 = 21e1 + 22 e 2 + 23e 3
T3 = 31e1 + 32e 2 + 33e 3

where e i are the unit direction vectors of the
22
axes. The quantity ij is known as the stress

tensor.
21
23
12
32
11
x1
13
31
33
T1
x3
T3
C. Wassgren
164
Stress Sign Convention

The sign conventions for stresses are as follows:
1. A positive face is a face that has a normal vector pointing in a positive direction.
2. Positive stresses on positive faces point in the positive direction.
3. Positive stresses on negative faces point in the negative direction.
4. The first subscript on the stress refers to the face on which the stress acts. The second subscript
refers to the direction in which the stress acts.
The figures below show positive stresses on all of the cubes faces.
x2
face
22
21
23
32
31
x2
direction
12
13
11
x1
33
13
11
32
12
x3
33
31
x1
23
21
Positive Stresses on Positive Faces
x3
22
Positive Stresses on Negative Faces
Notes:
1. ii (no sum) are referred to as normal stresses.
2. ij (ij) are referred to as shear stresses.
3. negative normal stresses compression
4. positive normal stresses tension
C. Wassgren
165
Cauchys Formula
Cauchys formula is used to determine the traction vector on an arbitrarily oriented surface with an
orientation vector, , given the stress tensor. Consider the small tetrahedral element shown below.
x2
T
The areas of each face are:

dA = dA = dA ( 1e1 + 2 e 2 + 3e 3 )
dA = dA
31
13
11
12
dAe1 = dA e1 = dA 1
33
32
dAe 2 = dA e 2 = dA 2
23
dAe3 = dA e 3 = dA 3
x1
21
22
x3
Apply Newtons 2nd Law to the element in the x1 direction:
dVx1 =
F1 = T e1 dA 11 dA e1 21 dA e 2 31 dA e3 + f1 dV
= 11dA 1 + 21dA 2 + 31dA 3 + (

x1 f1 ) dV
T1 dA
T1 = 11 1 + 21 2 + 31 3 + (
x1 f1 )
dV
but lim
dA 0 dA
=0
dV
dA
T1 = 11 1 + 21 2 + 31 3
T1 = j1 j
A similar approach may be followed to derive expressions for the x2 and x3 directions. In general, we have:
Ti = ji j
Cauchys Formula
Cauchys formula may be used to determine the traction, T, on a surface with an orientation, , given the
stress tensor, ji, at the point of interest.
yy
e y
yx
yz
x
z
C. Wassgren
166
Symmetry of the Stress Tensor

Consider the Angular Momentum Principal for the small element of material shown below (only the
stresses acting on the positive faces and causing rotations in the x3 axis are shown for clarity). Also note
that no body couples are assumed to act on the element.
x2
21 +
21
x2
( 12 dx2 )
dx3
12 +
12
x1
( 12 dx1 )
x1
dx2
x3
dx1
Using Newtons 2nd Law for rotational moments in the x3-direction:
3 = 12 + 12 12 dx1 dx2 dx3 12 dx1 + 12 + 12 12 dx1 dx2 dx3

I
3
x1
x1
2
2
( 12 dx1 )
O dx1dx2 dx3 dx1 + dx2
21 + 21 12 dx2 dx1dx3 12 dx2 21 + 21 12 dx2 dx1dx3

x2
x2
Dividing through by the volume (dx1dx2dx3) and taking the limit as dx1,dx2,dx30 gives:
21 = 12
A similar approach can be taken in the x1 and x2 directions to arrive at the general result:
ij = ji
The stress tensor is symmetric (when no couple stresses are present).
C. Wassgren
167
( 12 dx2 )
Another approach to proving symmetry of the stress tensor is to write the AME explicitly for the small
element (again assuming no body couples):
D
ijk r j f k dV + ijk r j
ijk r j uk dV
(4.19)
lk nl dA =
Dt
=Tk
Vsys
Ssys
Vsys

torque due to
body forces
torque due to
surface forces
time rate of change of

the angular momentum
of the element
Using the divergence theorem the surface integral may be written as a volume integral:
ijk r j lk nl dS = ijk
r j lk dV = ijk jl lk + r j lk ,l dV
(4.20)

xl
Ssys
Vsys
Vsys
jk
Substituting into equation (4.19) and simplifying (note that the mass of the element remains constant):
D
ijk rj f k dV + ijk jk + rj lk ,l dV =
ijk rj uk dV
Dt
Vsys
Vsys
Vsys
D
r j uk dV = 0
ijk rj f k + jk + rj lk ,l
Dt

Vsys
Duk Dr j
rj
uk
+
Dt Dt
Note that Drj/Dt=uj and ijkujuk=0 (i.e. uu=0). Re-arranging the previous equation gives:

Du
ijk r j f k + lk ,l k + jk dV = 0
Dt

Vsys
As the volume of the cube becomes very small (dV0), the r vector also becomes very small. Hence:
Duk

lim ijk r j f k + lk ,l
+
dV
= ijk jk dV = 0
jk
dV 0
Dt

V
Vsys
sys
Since the volume is arbitrary:
ijk jk = 0
(4.21)
Now apply the permutation tensor to both side of Eqn. (4.21) and utilize an identity:
ilm ijk jk = 0
(lj mk lk mj ) jk = 0
lm ml = 0
ml = lm
Thus, the stress tensor is symmetric (again, assuming no body couples).
C. Wassgren
168
Example:
The material element shown below has the following stress tensor components:
y
(0, 2, 0) m
0 4
3 0 kPa
4 0 5
1
[ ] = 0
x
(1, 0, 0) m
z
a.
b.
c.
d.
(0, 0, 1) m
Find the components of the traction vector, T, on the plane described by the unit normal vector, .
Determine the component of T parallel to .
Determine the component of T perpendicular to .
Determine the angle between T and .
SOLUTION:
First determine the components of the unit normal vector, , by taking the cross product of the vector
pointing from (1, 0, 0) to (0, 2, 0) with the vector pointing from (1, 0, 0) to (0, 0, 1) then normalizing.
( 0, 2, 0 ) (1, 0, 0 ) ( 0, 0,1) (1, 0, 0 )
( 1, 2, 0 ) ( 1, 0,1) ( 2,1, 2 )
=
=
=
( 1, 2, 0 ) ( 1, 0,1) ( 2,1, 2 )
( 0, 2, 0 ) (1, 0, 0 ) ( 0, 0,1) (1, 0, 0 )
= ( 32 , 13 , 23 )
Next, use Cauchys formula to determine the traction vector, T.
Ti = ji j
T1 = 11 1 + 21 2 + 31 3 = (1 kPa ) ( 32 ) + ( 0 kPa ) ( 13 ) + ( 4 kPa ) ( 23 )

T2 = 12 1 + 22 2 + 32 3 = ( 0 kPa ) ( 23 ) + ( 3 kPa ) ( 13 ) + ( 0 kPa ) ( 23 )
T3 = 13 1 + 23 2 + 33 3 = ( 4 kPa ) ( 23 ) + ( 0 kPa ) ( 13 ) + ( 5 kPa ) ( 23 )

T = ( 2,1, 32 ) kPa
The component of T parallel to is found by taking the dot product T .
T|| = T = Ti i = ( 2 kPa ) ( 23 ) + (1 kPa ) ( 13 ) + ( 32 kPa )( 32 )
T|| = 59 kPa
The component of T perpendicular to is found by taking the difference between T and T||.
44 32 28
T = T T|| = ( 2,1, 23 ) ( 59 )( 32 , 13 , 32 ) kPa = ( 27
, 27 , 27 ) kPa
T = 2.27 kPa
The angle between T and can be determined from the dot product between the two vectors.
T
T = T cos cos =
T
cos =
95 kPa
5 kPa
T
=
= 79
= 215
2
T
( 2,1, 3 ) kPa (1) 3 kPa
= 103.8
C. Wassgren
169
4.
Momentum Equations (aka the LME for a differential CV)
The momentum equations, which are the simply the linear momentum equations for a differential fluid
element or control volume, can be derived several different ways. Three of these methods are given below.
Method 1:
y
dy
z
dz
dx
Assume that the density and velocity are and u, respectively, at the control volumes center.
Consider only the x-momentum equation first. The x-momentum fluxes through each of the side of the
control volume are given by:
x )center
( mu
x )in through left = ( mu
x )center +
12 dx
( mu
x
= ( u x dydzu x ) + ( u x dydzu x ) ( 1 2 dx )
x
= u x u x + ( u x u x ) ( 1 2 dx ) ( dydz )
x)
= u x u x + ( u x u x ) ( 1 2 dx ) ( dydz )
( mu
out through right
x
x)
= u yux +
( mu
in through bottom
u y ux
y
) ( 1 2 dy ) ( dxdz )
) ( 1 2 dy ) ( dxdz )
x)
= u y ux +
( mu
out through top
u yux
y
x)
= uz ux +
( mu
in through back
( uz u x ) ( 1 2 dz ) ( dxdy )
z
( u z ux ) ( 1 2 dz ) ( dxdy )
z
Thus, the net x-momentum flux out of the control volume is given by:
x)
= ( uxux ) +
u y u x + ( u z u x ) ( dxdydz )
( mu
net, out of CV
y
z
x
x)
= uz ux +
( mu
out through front
(4.22)
The rate at which the x-momentum is increasing within the control volume is given by:
= ( u x dxdydz ) = ( u x ) ( dxdydz )
( mux )
t
t
t
within CV
(4.23)
where and ux are the density and x-velocity, respectively, at the center of the control volume. Note
that since these quantities vary linearly within the CV (from the Taylor Series approximation), the
averages within the CV are and ux.
C. Wassgren
170
The forces acting on the control volume include both body and surface forces. The body force acting
on the CV in the x-direction, FB,x, can be written as:
FB, x = f B, x ( dxdydz )
(4.24)
where fB,x is the body force per unit mass acting in the x-direction (e.g., for weight, the body force per
unit mass acting in the x-direction is simply gx).
The surface forces acting on the control volume include both normal and tangential forces. Writing the
surface force acting in the x-direction, FS,x, in terms of stresses gives:
xx
FS , x = xx +
( 1 2 dx ) ( dydz ) + xx + xxx ( 1 2 dx ) ( dydz )

normal force on left face
normal force on right face
yx
yx +
( 1 2 dy ) ( dxdz ) + yx + yx ( 1 2 dy ) ( dxdz )
y
y

shear force on bottom face
shear force on top face
zx
zx 1
zx +
1 2 dz ) ( dxdy ) + zx +
(
(
2 dz ) ( dxdy )
z
z

shear force on back face
shear force on front face
yx zx

FS , x = xx +
+
( dxdydz )
y
z
x
(4.25)
The LME states that the rate of increase of linear momentum within the control volume plus the net
rate at which momentum leaves the control volume must equal the net force acting on the control
volume:
x)
+ ( mu
= FB , x + FS , x
(4.26)
( mux )
net, out of CV
t
within CV
Substituting Eqns. (4.83)-(4.25) into Eqn. (4.26) gives:
( u x ) ( dxdydz ) + ( u x ux ) + u y ux + ( uz ux ) ( dxdydz ) =
t
y
z
x
yx zx

f B, x ( dxdydz ) + xx +
+
( dxdydz )
y
z
x
(4.27)
( u x ) + ( ux u x ) + u y u x + ( uz u x ) = f B , x + xx + yx + zx
t
x
y
z
x
y
z
A similar approach can be taken to determine the y- and z-components of the momentum equations.
All three components of the momentum equations can be written in the following compact (index
notation) form:
ji
u j ui = f B,i +
(4.28)
( ui ) +
t
x j
x j
In vector notation, Eqn. (4.28) is written as:
( u ) + ( u )( u ) = f B + T
t
(4.29)
Note that T = since the stress tensor is symmetric.
C. Wassgren
171
Expand the left hand side of Eqn. (4.28) and utilize the continuity equation:
u
u
+ i + ui
u j ui = ui
u j + u j i
( ui ) +
t
x j
t
t
x j
x j
u
u
u j + i + u j i
= ui
+
t
t x j
x j

= 0 (continuity eqn)
Dui
Dt
Substituting into back into Eqn. (4.28) gives:

ji
Du
i = f B ,i +
momentum equations
Dt
x j
(4.30)
Method 2:
Apply Newtons 2nd Law directly to a small piece of fluid:
D
( ui dxdydz ) = f B,i ( dxdydz ) + ji ( dxdydz )
Dt
x j
(4.31)
where the determination of the body and surface forces are described in the previous method.
Expanding the Lagrangian derivative gives:
Du
D
D
( ui dxdydz ) = i ( dxdydz ) + ui ( dxdydz )
Dt
Dt
Dt
but the second term on the RHS of this equation will be zero since the mass of the fluid element
remains constant. Thus, Eqn. (4.31) can be simplified to:
ji
Du
i = f B ,i +
Dt
x j
Method 3:
Recall the integral form of the LME:
d
ui dV + ui ( u rel dA ) = FB ,i + FS ,i
dt
CV
CS

( ui )
d
ui dV =
dV
and
dt
t
CV
CV
u rel = u
Note that the body force can be written as:
FB,i =
f B,i dV
CV
and the surface force can be written as:
FS ,i =
ji n j dA
CS
C. Wassgren
172
By utilizing Gauss Theorem (aka the Divergence Theorem), we can convert the area integrals into
volume integrals:
ui ( u dA ) = ( ui u ) dV =
u j ui dV
x j
CS
CV
CV
CS
ji n j dA
CV
ji
x j
dV
Substitute these expressions back into the LME to get:
ji
u j ui f B ,i
( ui ) +
dV = 0
x j
x j
t
CV
ji
u j ui f B,i
=0
( ui ) +
t
x j
x j
But this is the same expression as Eqn. (4.28) so we see that the final result will be the same!
ji
Du
i = f B ,i +
Dt
x j
Notes:
1. In order to be more useful to us, we need to have some way of relating the stresses acting on the fluid
element (or CV) to other properties of the flow (namely the velocities). This is accomplished using a
constitutive law which relates the stresses to the strain rates for a particular fluid or class of fluids.
2.
Equation (4.30) is valid for any continuous substance.
3.
Equation (4.28) is the conservative form of the LME. Equation (4.30) is the non-conservative form of
the LME.
C. Wassgren
173
Example
Consider the flow of a mixture of liquid water and small water vapor bubbles. The bubble diameters are
very small in comparison to the length scales of interest in the flow so that the properties of the mixture can
be considered point functions. For example, the density of the mixture at a point can be written as:
M = V + (1 ) L
where M is the mixture density, L is the liquid density, V is the vapor density, and is the void
fraction or the fraction of volume that is vapor in a unit volume of the mixture. Assume that evaporation
occurs at the bubble surface so that the liquid water turns to water vapor at a mass flow rate per unit volume
denoted by s.
a.
b.
c.
What is the continuity equation for the mixture?

What is the continuity equation for the liquid water phase?
What are the momentum equations for the liquid water phase?
SOLUTION:
The continuity equation for the mixture will be the normal continuity equation:
M
+
( M ui ) = 0
t
xi
To show that this relation is true, consider the control volume shown below.
(4.32)
dx
y
dy
The rate of change of mass within the control volume is:
( M dxdydz ) = M dxdydz
t
t
The net mass flux into the CV in the x-direction is:
m x , net = M ux dydz M u x + ( M u x ) dx dydz = ( M ux ) dxdydz
x
into CV
Following a similar approach in the y and z directions gives:
m y ,net = ( M u y ) dxdydz
y
into CV
m z,net
into CV
( M uz ) dxdydz
z
(4.34)
(4.35)
(4.36)
Thus, from conservation of mass:

M
dxdydz = ( M u x ) dxdydz ( M u y ) dxdydz ( M u z ) dxdydz

t
x
y
z
+
( M ui ) = 0
t
xi
C. Wassgren
(4.33)
(4.37)
(4.38)
174
To determine the continuity equation for the liquid water phase, consider the control volume drawn below
where the CV surrounds each vapor bubble.
dx
y
x
dy
The rate of change of liquid mass within the control volume is:
L (1 ) dxdydz = (1 ) L dxdydz
t
t
(4.39)
The net liquid mass flux into the CV in the x-direction is:
m x , net = (1 ) L u x dydz (1 ) L u x + (1 ) L u x dx dydz = (1 ) L u x dxdydz

x
x
into CV
(4.40)
Following a similar approach in the y and z directions gives:
m y ,net = (1 ) L u y dxdydz
(4.41)
y
into CV
(1 ) L u z dxdydz
z
The rate at which liquid mass is being converted to vapor mass is:
m out of CV = s (1 ) dxdydz
=
m z,net
(4.42)
into CV
(4.43)
due to evap.
Thus, from conservation of mass:
(1 ) L dxdydz =
t
(4.44)
(1 ) L ux dxdydz (1 ) L u y dxdydz (1 ) L u z dxdydz s (1 ) dxdydz

x
y
z
(1 ) L +
(1 ) L ui = (1 ) s (continuity eqn. for liquid phase)
t
xi
(4.45)
To determine the momentum equations for the liquid phase, apply the momentum equation to the same
control volume used to derive the liquid phase continuity equation. The change in momentum of liquid
within the CV is:
d
(4.46)
u dV = t ui L (1 ) dxdydz = t ui L (1 ) dxdydz
dt CV
The net flux of linear momentum out of the CV through the sides of the CV is:
u ( u
rel
dA ) =
CS
ui (1 ) Lu x dxdydz + ui (1 ) L u y dxdydz + ui (1 ) L u z dxdydz + ui s (1 ) dxdydz

x
y
z
ui (1 ) Lu j dxdydz + ui s (1 ) dxdydz
x j
(4.47)
(Note that the term involving s is the rate at which momentum leaves the liquid phase due to the fact
that the liquid is evaporating.)
C. Wassgren
175
The surface forces acting on the control surface are:

ji
FS + FB =
dxdydz + fVonL ,i L (1 ) dxdydz + g i L (1 ) dxdydz
x j
(4.48)
Note that the stress terms are the surfaces forces acting on the sides of the CV. The term fVonL,i is the force
per unit mass that the vapor phase exerts on the liquid phase, and the last term in Eqn. (4.48) is the body
force acting on the liquid phase where gi is the body force per unit mass.
Substituting into the linear momentum equation and simplifying results in:
ji
ui (1 ) L u j =
ui L (1 ) +
+ fVonL ,i L (1 ) ui s (1 ) + g i L (1 )
t
x j
x j
(4.49)
The continuity equation derived previously for the liquid phase (Eqn. (4.45)) could be used to further
simplify the momentum equation, if desired.
C. Wassgren
176
5.
Fluid Element Deformations
In order to relate the stresses acting on a fluid element to the other variables in the momentum equations,
we need to determine a constitutive law. The constitutive law should relate the stresses to the rates of
strain (or deformation rates) of a fluid element. For a solid, the necessary constitutive law relates the
stresses to the strains (or deformations). In order to derive an appropriate constitutive law, we must first
discuss the general types of deformations that can occur for a fluid element and then describe, in
mathematical terms, the rates at which these deformations occur.
Any general deformation can be decomposed into a combination of translation, dilation (aka dilatation),
rigid body rotation, and angular deformation as shown in the figure below.
general deformation
translation
dilation
angular deformation
rigid body rotation
Now lets describe the rate at which each of these deformations occurs.
Translation
dx
The rate of translation is described by the time rate of change of the position of the
element, i.e. the velocity.
dx
=u
(4.50)
rate of translation =
dt
C. Wassgren
177
Dilation (aka Dilatation)

B
u2
dx2 dt
x2
The rate of dilation can be described by the rate at which the relative
volume of the element increases with time.
1 dV
volumetric dilation rate =
V dt
dx2
dx1
u1
dx1dt
x1
The velocity of point A relative to point O in the x1 direction is:

u
u1 = 1 dx1
x1
Thus, point A stretches the element in the x1 direction over time dt a distance of:
u1
dx1dt
x1
The increase in volume of the element due to the relative movement of point A is:
u
dVA = 1 dx1dt dx2 dx3

x1
A similar approach can be followed for stretching in the x2 and x3 directions. The total increase in volume
of the element is:
This volume is comprised of
u u
u
C
dV = 1 + 2 + 3 dx1dx2 dx3 dt
B
H.O.T.s and, hence, is
x1 x2 x3
neglected.
u2
dx2 dt dx1dx3
x2
u1
dx dt dx dx
A x 1 2 3
1
Note that higher order volume terms have been neglected in deriving the previous result. The volumetric
dilation rate is thus:
u1 u2 u3
+
+
dx1dx2 dx3 dt
1 dV x1 x2 x3
=
V dt
dx1dx2 dx3 dt
u1 u2 u3 ui
+
+
=
= u
x1 x2 x3 xi
(4.51)
Notes:
1. For an incompressible fluid, the volumetric dilation rate is zero since if the volume of the element
changes, the density must also change (COM).
C. Wassgren
178
Angular Deformation
u1
dx2 dt
x2
The rate of angular deformation in the 1-2 plane can be

described as the average rate at which the sides of the
element approach one another, i.e. the average rate at
which the angles AOA and BOB increase.
B
C
dx2
O
d
dx1
u 2
d x1 d t
x1
The angle AOA (d) is:

u2
dx1dt
x1
tan ( d ) =
dx1
But since the angle d is very small, tan(d) = d:
u
d = 2 dt
x1
Similarly, the angle BOB (d) is:
u
d = 1 dt
x2
Define the rate of angular deformation, S12, (aka rate of shearing strain) in the 1-2 plane as the average time
rate of change of these two angles:
d d 1 u2 u1
+
+
S12 12
=
dt 2 x1 x2
dt
Similarly, we can determine the angular deformation rate in the 1-3 and 2-3 planes:
u u
u u
S13 = 12 3 + 1 and S23 = 12 3 + 2
x1 x3
x2 x3
Combine the angular deformation rate and the dilation rate into one tensor quantity called the shearing
strain tensor, Sij:
u u j
Sij = 12 i +
(4.52)
x j xi
Notes:
1. Dilation rate is given as the trace of Sij:
= trace ( Sij ) = Sii = S11 + S22 + S33 =

2.
u1 u2 u3
+
+
x1 x2 x3
The shearing strain tensor is symmetric! i.e. Sij = Sji
C. Wassgren
179
Rigid Body Rotation
u1
dx2 dt
x2
The rate at which the fluid element rotates about the 3axis in rigid body motion can be described as the average
rate at which the sides of the element rotate in the same
direction.
C
B
d
dx2
O
d A
dx1
u2
dx1dt
x1
The rotation rate about the 3-axis, 3, is given by:

d d 1 u2 u1
3 12
+
=
dt 2 x1 x2
dt
Rotations about the 1 and 2 axes can be found in a similar manner:
u u
u u
1 = 12 3 2 and 2 = 12 1 3
x2 x3
x3 x1
The rate of rotation of the element can be summarized using the rotation rate vector, :
u u
u u
u
u
= 12 3 2 e1 + 12 1 3 e 2 + 12 2 1 e 3
x
x
x
x
x
x2
3
1
1
2
3
(4.53)
= 12 u
Notes:
1. The rotation rate vector, , is written in index notation form as:
u
i = 12 ijk k
x j
2.
The rotation rate vector can also be written as an anti-symmetric rotation rate tensor, Rij:
u u j
(4.54)
Rij 12 i
= ijk k (Note: Rij = R ji )
x j xi
where Rij is the rotation rate in the i-j plane. Note that the diagonal elements of the tensor Rij are zero.
3.
The vorticity, , of a fluid element is defined to be twice the rotation rate of the element:
2 = u
or in index notation:
u
i = ijk k
x j
(4.55)
An irrotational flow is one in which = 0. A rotational flow is one in which 0.
C. Wassgren
180
Now that we have described the deformation rate components (e.g. dilation, angular deformation, and rigid
body rotation; translations are treated separately) of a fluid element, lets combine these into a single tensor
quantity known as the deformation rate tensor, eij:
u u j
u u j
u
eij i = Sij + Rij = 12 i +
(4.56)
+ 12 i
x j
x j xi
x j xi
We will use the deformation rate tensor when deriving the constitutive relations between stress and strain
(deformation) rates in a fluid.
C. Wassgren
181
Example
A fluid has a velocity field given by
u = 2 xi 3 yj + zk
At the location (x, y, z) = (-2, -1, 2), calculate:
a. the normal and shearing strain rates at the location, and
b. the rotational velocity of the fluid.
SOLUTION:
The strain rate tensor is given by:
u u j
Sij = 12 i +
x j xi
so that the normal strain rates are:

u u y u z
S xx , S yy , S zz = x ,
,
x y z
S xx , S yy , S zz = ( 2, 3,1)
and the shearing strain rates are:
( S xy = S yx , S xz = S zx , S yz = S zy ) = 12 uyx +

u y 1 u x u z 1 u y u z
+
,
+
,
x 2 z
x 2 z
y
S xy = S yx , S xz = S zx , S yz = S zy = ( 0, 0, 0 )
The rotational velocity of a fluid element is given by:
u
.)
i = 12 ijk k (Note that the vorticity is twice the rotation rate, i.e. = 2
x j
and, thus, for the given case:
1 u u y u x u z u y u x
x , y , z = z
,
,
2 y
z z
x x
y
x , y , z = ( 0, 0, 0 ) (Fluid elements are not rotating anywhere! The flow is irrotational.)
C. Wassgren
182
Stress-Strain Rate Relations for a Newtonian Fluid

The following assumptions are based on observation and intuition.
The key assumptions in deriving the stress-strain rate constitutive relations for a Newtonian fluid are:
1. When the fluid is at rest, the pressure exerted by the fluid is the thermodynamic pressure, p.
2. For a Newtonian fluid, the stress tensor, ij, is linearly related to the deformation rate tensor, ekl,
and depends only on that tensor.
3. There are no preferred directions in the fluid so that the fluid properties are point functions. This
is the condition of isotropy.
Now lets examine how these assumptions aid us in deriving the appropriate constitutive law.
Assumption 1:
When the fluid is at rest, the pressure exerted by the fluid is the thermodynamic pressure, p.
This assumption implies the following:
ij = p ij + ij
(4.57)
where ij is referred to as the viscous stress tensor (aka deviatoric stress tensor) and it is only a
function of the fluid motion (i.e. ij=0 for a static fluid). Note that the pressure term is negative since
compression of the fluid element is indicated by a negative normal stress.
Assumption 2:
For a Newtonian fluid, the stress tensor, ij, is linearly related to the deformation rate tensor, ekl, and
depends only on that tensor.
The 9 elements of ij can be written as a linear combination of the 9 elements of ekl:
ij = Aijkl ekl
(4.58)
where Aijkl is a tensor of rank 4 (81 elements) that depends only on the local state of the fluid.
Notes:
1. Recall that the deformation rate tensor is given as:
u
u
u
u
u
ekl k = Skl + Rkl = 12 k + l + 12 k l
xl
xl xk
xl xk

symmetric,
S kl = Slk
2.
anti-symmetric,
Rkl = Rlk
Air and water are common examples of Newtonian fluids.
Since the stress tensor, ij, is symmetric (refer to the notes reviewing stress), ij must also be
symmetric. And since ij is symmetric, the components of the Aijkl tensor multiplied by the antisymmetric part of the deformation rate tensor, ekl, must be zero. Thus,
u
u
ij = 12 Bijkl k + l
(4.59)
xl xk
where Bijkl is the Aijkl tensor with the Aijkl components multiplied by the components of the antisymmetric part of ekl set equal to zero.
C. Wassgren
183
Assumption 3:
There are no preferred directions in the fluid so that the fluid properties are point functions. This is the
condition of isotropy.
The condition of isotropy means that the fluid properties are the same in all directions. Examples of
non-isotropic materials include fluids comprised of long chain molecules or oriented fibrous solids
such as wood.
It can be shown (out of the scope of these notes) that the most general 4th order isotropic tensor can be
written as:
) (
Bijkl = ij kl + ik jl + il jk + ik jl il jk
(4.60)
where , , and are scalar quantities.

Substitute Eqn. (4.60) into Eqn. (4.58) and simplify:
ij = 12 ij kl + ( ik jl + il jk ) + ( ik jl il jk ) k + l
xl xk
(4.61)
where
1
2
1
2
( ij kl ) uxk + xul = 12 ij 2 uxk = ij uxk
( ik jl + il jk )
uk ul 1 ui u j
+
+
= 2
x j xi
xl xk
u j ui
+
+
xi x j

u u j
= i +
x j xi
u u
(ik jl il jk ) uxk + xul = 12 xui + x j x j + xui = 0
so that:
ij =
u u j
uk
ij + i +
x j xi
xk
(4.62)

ij = p ij + ij
= p ij +
u u j
uk
ij + i +
x j xi
xk
u u j
u
ij = p + k ij + i +
x j xi
xk
stress-strain rate constitutive relation for a Newtonian fluid
(4.63)
Notes:
1. The quantity, , is referred to as the dynamic viscosity.
2.
The quantity, , is referred to as the 2nd coefficient of viscosity.
C. Wassgren
184
3.
How is the thermodynamic pressure related to the normal stresses? Define the mechanical pressure,
p , as the average of the normal stresses:
( )
p 13 trace ij = 13 ii = 13 ( 11 + 22 + 33 )
For a Newtonian fluid:
p = 13 ( 11 + 22 + 33 )
u
u
u
u
= 13 p + k + 2 1 + p + k + 2 2
x
x2
k
1
k
= p

uk
u
+ 2 3
+p +
xk
x3

uk 2 u1 u2 u3

+
+
xk 3 x1 x2 x3
uk
uk
p = p + 23
or p = p + + 23

xk

xk
=K
=K
where K bulk viscosity = +2/3. In general, the thermodynamic pressure is not the same as the
mechanical pressure. What then is the physical significance of the bulk viscosity, K, term? The
mechanical pressure is a measure of the translational energy only. The thermodynamic pressure,
however, is a measure of the total energy (translational, rotational, vibrational, etc.) The bulk
viscosity, K, is a measure of the transfer of energy from the translational mode to the other modes. For
example, when fluid flows through a shock wave, there is a considerable transfer of energy between
the translational mode and the other modes; hence, the bulk viscosity cannot be neglected for such a
flow process.
For typical flows, however, the bulk viscosity is often neglected. For example:
1. For monatomic gases the only energy mode is the translational mode so that:
K =0 p= p
For polyatomic gases and liquids, the bulk viscosity is often small so that we usually assume:
K 0 p= p
The assumption that the bulk viscosity is zero (or equivalently, = 2/3) is known as Stokes
Relation.
2.
For an incompressible fluid the velocity divergence term is zero (from the continuity equation) so
that the bulk viscosity is irrelevant:
uk
=0 p= p
xk
3.
The bulk viscosity term is generally not negligible when there is a rapid expansion or contraction
of the fluid such as when fluid passes through a shock wave or when considering acoustic
absorption.
C. Wassgren
185
4.
The stress tensor given in Eqn. (4.63) can be substituted into the momentum equations to give the
Navier-Stokes Equations:
ji
Du
i =
+ fi
Dt
x j
u u j
i +
x j xi
Navier-Stokes equations for a Newtonian fluid
Dui
p
uk
=
+
+
Dt
xi xi xk x j
+ fi
(4.64)
For an incompressible fluid with constant dynamic viscosity:
u
Dui
u
p
j
i
+ fi
=
+
+
Dt
xi
x j x j xi x j
=0
Dui
2 ui
p
Du
=
+
+ fi or
= p + 2 u + f
xi
x j x j
Dt
Dt
(4.65)
Navier-Stokes equations for an incompressible, Newtonian fluid with constant dynamic

viscosity
For an inviscid (=0) fluid that follows Stokes Relation or is incompressible:
Dui
p
Du
=
+ fi or
= p + f
xi
Dt
Dt
These relations are known as Eulers Equations.
(4.66)
As a reminder, the Navier-Stokes equations are the momentum equations for a Newtonian fluid.
Eulers equations are the momentum equations for an inviscid fluid.
C. Wassgren
186
Example
Consider a 3D steady flow of an incompressible, Newtonian liquid with a velocity field given by:
u = axi + ayj 2azk
There are no body forces acting on the flow and the pressure at the origin is p0.
a. Show that the continuity equation is satisfied,
b. Determine the pressure field.
c. Determine the vorticity field.
SOLUTION:
The continuity equation is:
u x u y u z
+
+
= a + a 2a = 0
x
y
z
Continuity is satisfied!
The pressure field may be found using the Navier-Stokes equations. Note that the body forces are zero.
Du
2ui
p
i =
+
Dt
xi
x j x j
2u
u
u
2u
2u
u
p
= u x x + u y x + u z x + 2x + 2x + 2x
x
x
x
y
z
y
z
p
= a2 x

p ( x, y, z ) = 12 a 2 x 2 + f ( y , z )
2u y 2 u y 2u y
u y
u y
u y
p
2 +
= ux
+ uy
+ uz
+
+ 2
2
x
y
x
y
z
z
y
p

= a2 y
p ( x, y, z ) = 12 a 2 y + g ( x, z )
2u
u
u
u
2u
2u
p
= u x z + u y z + u z z + 2z + 2z + 2z
x
z
x
y
z
y
z
p
= 4 a 2 z

p ( x, y, z ) = 2 a 2 z 2 + h ( x, y )
Combining the previous expressions and noting that p(0, 0, 0) = p0:
p ( x, y, z ) = p0 12 a 2 x 2 12 a 2 y 2 a 2 z 2
The vorticity field is:
u
i = ijk k
x j
x =
u z u y
=0
y
z
u x u z
=0
z
x
u y u x
z =
=0
x
y
y =
( x, y, z ) = 0 The flow is irrotational!

Note that the viscous force terms in the Navier-Stokes equation are zero (ui,jj = 0 ).
C. Wassgren
187
6.
Acceleration of a Fluid Particle in Streamline Coordinates
Often its helpful to use streamline coordinates (s,n) instead of Cartesian coordinates (x, y) when describing
the motion of a fluid particle.
Lets determine a fluid particles acceleration parallel (s-direction) and normal (n-direction) to a streamline
for a steady, 2D flow. Consider the figure shown below.
fluid particle
s = s3
s = s2
streamlines
s = s1
Notes:
1. The coordinates (s, n) are just like (x, y)
coordinates. They specify the location
of the fluid particle.
2. Lines of constant s and n are .
3. n points toward the center of curvature.
n = n1
n = n2
n = n3
The acceleration of the fluid particle is given by:

Du
a=
Dt
where u = us . Substituting and expanding gives:
D ( us )
Du
Ds
a=
= s
+u
Dt
Dt
Dt
Now lets expand the Lagrangian derivative terms keeping in mind that u=u(s, n):
Du
u
u
u
u
=
+ un
+ us
=u

Dt
t
n
s
s

= 0 (flow
tangent to
streamline)
= 0 (steady)
(4.67)
(4.68)
(4.69)
= u (flow
tangent to
streamline)
and
Ds
=
Dt
s
t
= 0 (steady)
un
= 0 (flow
tangent to
streamline)
s
+
n
C. Wassgren
us
=u (flow
tangent to
streamline)
s
s
=u
s
s
(4.70)
188
To determine how s varies with the s-coordinate, consider the following figures:
O
d
radius of
curvature, R
streamline
n ( s )
B
ds
s ( s + ds )
s ( s + ds )
O
d
B
ds
s ( s ) A
s ( s )
Note that the triangles OAB and OAB are similar. Hence,
ds 1
ds ds
=
= ds
=
R
ds R
s
(4.71)
=1
Also as ds0, ds points in the n direction so:

ds 1
= n
ds R
Ds u
= n
Dt R
(4.72)
(4.73)
Substituting Eqns. (4.73) and (4.69) into equation (4.68) gives the fluid particle acceleration in streamline
coordinates:
u2
u
a = u s + n
(4.74)
R
s

tangential
acceleration
normal
acceleration
C. Wassgren
189
Example:
Water flows through the curved hose shown below with an increasing speed of u = 10t ft/s, where t is in
seconds. For t = 2 s determine:
a. the component of acceleration along the streamline,
b. the component of acceleration normal to the streamline, and
c. the net acceleration (magnitude and direction).
u
R = 20 ft
SOLUTION:
The acceleration component in the streamline direction is:
u
u
as =
+u
t
s
where
u
= 10 ft 2 (The flow is unsteady.)
s
t
u
= 0 (The flow velocity doesnt change with respect to position along the streamline.)
s
as = 10 ft 2
s
(4.75)
The acceleration component normal to the streamline is:
an =
u2
R
(4.76)
where
u 2 (10 * 2 ft/s )
=
= 20 ft 2 (The velocity is evaluated at t = 2 s.)
s
R
20 ft
an = 20 ft 2 (The acceleration is toward the center of curvature.)
s
2
The net acceleration is:

a = an n + as s
a = ( 20n + 10s ) ft
s2
as
(4.77)
an
a
a = 22.4 ft
s2
C. Wassgren
190
7.
Eulers Equations in Streamline Coordinates
Recall from previous analyses that the differential equations of motion for a fluid particle in an inviscid
flow in a gravitational field are:
Du
= p + g
(Eulers Equations)
(4.78)
Dt
For simplicity, further assume that were dealing with a 2D, steady flow. Now write Eqn. (4.78) in
streamline coordinates (s, n):
fluid particle
streamlines
p
+ gs
s
p
n-direction: an = + g n
n
s-direction:
as =
(4.79)
(4.80)
Recall that in streamline coordinates:
u2
u
and an =
R
s
so that Eqns. (4.79) and (4.80) become:
u
1 p
u
=
+ gs
s
s
2D, steady Eulers equations in streamline coordinates
1 p
u2
=
+ gn
n
R
as = u
(4.81)
We can draw an important and very useful conclusion from the normal component of Eqn. (4.81). For a
flow moving in a straight line (R ) and neglecting gravity (gn = 0) we have:
p
=0
i.e., the pressure does not change normal to the direction of the flow!
n
This is very helpful when considering the pressure in a free jet (shown in the figure below). Since free jets
typically have negligible curvature and gravitational effects, the pressure everywhere normal to the free jet
will be the same!
p = patm
g
free jet
p = patm
R
In n direction: 0 =
1 p
p
+g
= g
n
n
u2
1 p
p
u2
=
=
p as n
R
n
n
R
The largest pressure is on the outside bend while the smallest
pressure is on the inside bend. If the fluid is a liquid and the
inside bend pressure reaches the vapor pressure of the liquid,
cavitation will occur.
In n direction:
C. Wassgren
191
8.
Energy Equation (aka COE for a differential CV)
The energy equation, which is simply conservation of energy for a differential fluid element or control
volume, can be derived several different ways. Two of these methods are given below.
Method 1:
y
dy
z
dz
dx
Assume that the density, specific total energy, and velocity are , e, and u, respectively, at the control
volumes center. The total energy fluxes through each of the side of the control volume are given by:
( m x e )center 1
)in through left = ( m x e )center +
2 dx
( me
x
= ( u x dydze ) + ( u x dydze ) ( 1 2 dx )
x
= u x e + ( u x e ) ( 1 2 dx ) ( dydz )
)
= u x e + ( u x e ) ( 1 2 dx ) ( dydz )
( me
out through right
x
)
= u y e +
( me
in through bottom
u y e ( 1 2 dy ) ( dxdz )
y
) ( 1 2 dy ) ( dxdz )
)
= u y e +
( me
out through top
u y e
y
)
= uz e +
( me
in through back
( u z e ) ( 1 2 dz ) ( dxdy )
z
( uz e ) ( 1 2 dz ) ( dxdy )
z
where the specific total energy (not including the potential energy), e, is given by:
e = u + 12 u u
and u is the specific internal energy.
)
= uz e +
( me
out through front
Thus, the flux of total energy out of the control volume is:
)
= ( uxe) +
u y e + ( u z e ) ( dxdydz )
( me
net, out of CV
y
z
x
C. Wassgren
192
(4.82)
(4.83)
The rate at which the total energy is increasing within the control volume is given by:
= ( e dxdydz ) = ( e )( dxdydz )
( me )
t
t
t
within CV
(4.84)
where and e are the density and specific internal energy, respectively, at the center of the control
volume. Note that since these quantities vary linearly within the CV (from the Taylor Series
approximation), the averages within the CV are and e.
The rate at which heat is added to the control volume is given by:
Qinto CV = qinto CV ( dxdydz )
(4.85)
where qinto CV is the rate of heat transfer into the control volume per unit volume. Note that the mode
of heat transfer is not indicated at this point in the derivation.
The rate at which work is done on the control volume due to body forces is given by:
W B,on CV = ( f B u )( dxdydz ) = f B ,i ui ( dxdydz )
(4.86)
Note that the potential energy is not included in Eqn. (4.82) since that term is included in the rate of
body force work term in Eqn. (4.86). The rate at which work is done on the control volume due to
surface forces is given by:
WS ,on CV = dF u
xydA
( xx u x )
( u )
WS ,on CV = xx u x +
( 1 2 dx ) xx ux + xx x ( 1 2 dx ) ( dydz )
x
x

u
xxdA
xy u y
+ xy u y +
x

( 1 2 dx ) xy ux +
xy u y
x
( 1 2 dx ) ( dydz )

( xz u z )
( xz u z )
1 dx ) u +
+ xz u z +
1 2 dx ) ( dydz )
(
(
xz z
2
x
x

+
=
ji ui ( dxdydz )
x j
(4.87)
Conservation of energy states that the rate of increase of total energy within the control volume plus
the net rate at which total energy leaves the control volume must equal the rate at which heat is added
to the control volume plus the rate at which work is done on the control volume.
)
+ ( me
= Qinto CV + W B,on CV + WS ,on CV
(4.88)
( me )
net, out of CV
t
within CV
Substituting Eqns. (4.83) (4.87) into Eqn. (4.26) gives:
( e )( dxdydz ) + ( u x e ) + u y e + ( uz e ) ( dxdydz ) =
t
y
z
x
+ qinto ( dxdydz ) + f B,i ui ( dxdydz ) +
e u j = qinto CV + ui f B ,i +
ji ui
( e ) +
t
x j
x j
C. Wassgren
193
ji ui
x j
) ( dxdydz )
(4.89)
e
u j + u j
e u j = e
+
+e
( e ) +
t
x j
t
t
x j
x j
e
u j + + u j
= e
+
t
t x j
x j

Substituting back into Eqn. (4.89) gives:

De
= qinto CV + ui f B,i +
ji ui
Dt
x j
C. Wassgren
De
Dt
energy equation
194
(4.90)
Method 2:
Recall the integral form of COE:
d
e dV + e ( u rel dA ) = Q into CV + W B, on CV + WS , on CV
dt
CV
CS

( e )
d
e dV =
dV
and
dt
t
CV
CV
u rel = u
Note that the heat transfer into the CV can be written as:
Q
= q
dV
into CV
into CV
CV
The work on the CV due to body forces is written as:

W
= ( f u ) dV =
f u dV
B ,on CV
CV
B ,i i
CV
and the work on the CV due to surface forces is written as:

W
= ( f u ) dA =
f u dA = n u dA
S ,on CV
CS
S ,i i
CS
ji j i
CS
where the surface forces have been written in terms of the stresses.
By utilizing Gauss Theorem (aka the Divergence Theorem), we can convert the area integrals into
volume integrals:
e ( u dA ) = ( e u ) dV =
e u j dV
x j
CS
CV
CV
ji n j ui dA =
CS
ui ji
x j
CV
) dV
Substitute these expressions back into COE to get:
e u j qinto CV f B,i ui
ji ui dV = 0
( e ) +
x j
x j
t
CV
e u j qinto CV f B ,i ui
ji ui = 0
( e ) +
t
x j
x j
De
= qinto CV + ui f B,i +
ji ui
Dt
x j
C. Wassgren
195
Notes:
1. The rate of heat transfer may be re-written in terms of the rate of heat transfer per unit area out of the
control volume through the control surface, q. In terms of the Method 1 approach, the rate of total heat
transfer into the control may be written as:

q
q

Qinto CV = qx + x ( 1 2 dx ) + qx + x ( 1 2 dx ) ( dydz )
qx 1
x
x
x

qx + x 2 dx

q y
q y

q y +
( 1 2 dy ) + q y + ( 1 2 dy ) ( dxdz )
y
y

q
qx + x 12 dx

qz
qz 1

1
x
qz +
( 2 dz ) + qz + z ( 2 dz ) ( dxdy )

Simplifying the previous relation gives:
q y qz
q j
q
Qinto CV = x +
+
( dxdydz )
( dxdydz ) = ( q )( dxdydz ) =
y
z
x j
x
The rate of total heat transfer in terms of the rate of heat transfer per unit area may also be derived
using the Method 2 approach and the Divergence Theorem:
q j
Qinto CV = ( q dA ) = q j n j dA =
dV
x j
CS
CS
CV
Substituting the previous expressions into the energy equation (Eqn. (4.90)) gives:
q
De
ji ui
= i + ui f B ,i +
Dt
xi
x j
(4.91)
energy equation in terms of the heat transfer per unit area

2.
The energy equation (either Eqn. (4.90) or (4.91)) may be simplified further by noting that:
De D
u
u
=
( u + 1 2 u u ) = t + t 12 ui ui + u j x + u j x 12 ui ui
Dt Dt
j
j
u
u
u
u
=
+uj
+ ui i + u j ui i
t
x j
t
x j

=
Du
Dt
= ui
(4.92)
Dui
Dt
and
ji
u
ji ui = ui
+ ji i
x j
x j
x j
(4.93)
Substituting these expressions into Eqn. (4.90) (or (4.91)) gives:

ji
Dui
u
Du
+ ui
= qinto CV + ui f B ,i +ui
+ ji i
Dt
x j
x j
Dt
ji
u
Dui
Du
= qinto CV + ji i + ui f B,i +
Dt
x j
x j
Dt
but the terms in brackets are just the momentum equations!
C. Wassgren
196
The dot product of the momentum equations with the velocity is known as the mechanical energy
equation.
ji
Dui
ui
=
ui f B ,i
+
ui
mechanical energy equation
(4.94)

Dt
x j

rate at which work is
rate of increase
of kinetic energy of
fluid element
done on the fluid element

due to body forces
rate at which work is

done on the fluid element
due to stress gradients
Note that:
Du
D 1
( 2 ui ui ) = ui Dti
Dt
The energy equation without the mechanical energy equation terms is known as the thermal energy
equation and is given by:
u
Du
= qinto CV + ji i
Dt
x j
or, in terms of the heat transfer per unit area:
q
Du
=
i
+
Dt
x

i
rate of increase
of internal energy
within the fluid element
3.
rate at which
heat is added
to the fluid element
through the surface area
ui
x j

ji
thermal energy equation
(4.95)
rate at which mechanical

energy is converted to thermal
energy due to deformations
of the fluid element
The heat flux term, q, may be written in terms of a temperature gradient using Fouriers Law of
Conduction (assuming that conduction is the dominant mode of heat transfer):
T
q = k T or qi = k
(4.96)
xi
where k is the thermal conductivity (in its most general form, the thermal conductivity is a tensor
quantity) of the substance and T is the temperature. Note that the negative sign is included in Eqn.
(4.96) to account for the fact that heat flows from regions of high temperature to regions of low
temperature. Thus, the thermal energy equation can be written as:
ui
Du
T
=
(4.97)
k
+ ji
Dt xi xi
x j
Thermal Energy Eqn using Fouriers Law of Conduction
C. Wassgren
197
4.
The rate of work term in the thermal energy equation (Eqn. (4.95)) includes both reversible and
irreversible work terms. Consider the rate of work term using the stress tensor for a Newtonian fluid:
u j ui ui
u
u
ji i = p + k ij +
+

xi x j x j
x j
xk
u j
u j ui ui
u
= p
+
+ k +
x j
xi x j x j
xk

reversible
pressure work
irreversible viscous work
The irreversible rate of work term (the rate at which mechanical energy is being converted into thermal
energy) is often referred to as the energy dissipation function, :
2
u j ui
u
= k +
+
xi x j
xk
u
i
x j
energy dissipation function
Thus, the thermal energy equation can be written as:

u j
Du
= qinto CV p
+
x j
Dt
or, if conduction is the significant mode of heat transfer:

u j
Du
T
=
k
p
+
Dt x j x j
x j
5.
(4.98)
(4.99)
(4.100)
Note that for an incompressible fluid, the continuity equation:

ui
=0
xi
can be used to simplify the thermal energy equation and the energy dissipation function to the
following forms:
Du
= qinto CV +
(4.101)
Dt
thermal energy equation for an incompressible fluid
where the energy dissipation function is given by:
u j ui ui
=
+
(4.102)
xi x j x j
energy dissipation function for an incompressible, Newtonian fluid
C. Wassgren
198
6.
The energy dissipation function for a Newtonian fluid is a positive definite quantity which means that
viscosity always acts to convert mechanical energy into thermal energy.
2
u
i
x j
1 u u j
i +
2 x j xi

u j ui
u
= k +
+
xk
xi x j
u j ui
u
= k +
+
xk
xi x j
1 u u j
+ i
2 x j xi
1 u j ui u j u j ui ui ui u j
2 xi x j xi xi x j x j x j xi
1
+
u 1 u j ui
= k +
+
2 xi x j
xk
u 1 u j ui
= k +
+
2 xi x j
xk
u j u j ui ui
+
xi xi x j x j

= 0( since i and j are dummy indices )

2
u 1 u j ui
= k +
+
2 xi x j
xk
(4.103)
Note that for an incompressible fluid:

uk
=0
xk
so that:
1 u j ui
+
2 xi x j
>0
since >0.
For a compressible, Newtonian fluid, Stokes hypothesis states that:
= 23
After simplifying Eqn. (4.103) it can also be shown that >0.
C. Wassgren
199
6.
The total energy equation may also be written in a form using the specific enthalpy, h, which is defined
as:
p
hu+
(4.104)
Consider the thermal energy equation (Eqn. (4.99)) using the definition of the energy dissipation
function (equation (4.98)):
u j
Du
= qinto CV p
(4.105)
+
x j
Dt
Re-write the pressure term in the previous equation utilizing the continuity equation:
u j
1 D
D p Dp
(4.106)
p
= p

=
x j
Dt
Dt
Dt
Du
D p Dp
= qinto CV
+
+
Dt
Dt Dt
D
p
Dp
+
u + = qinto CV +
Dt
Dt
Dh
Dp
= qinto CV +
+
Dt
Dt
(4.107)
For an incompressible fluid, Eqn. (4.105) is:

Du
= qinto CV +
Dt
For an incompressible fluid the internal energy is a function only of temperature, i.e., u = u(T) = cT.
Thus, the energy equation for an incompressible fluid becomes:
DT
c
= qinto CV +
energy eqn for an incompressible fluid
(4.108)
Dt
Note: For an incompressible substance, cv = cp = c.
Thus, the energy equation is uncoupled from the continuity and momentum equations for an
incompressible flow. In other words, the unknowns of velocity, ui , and pressure, p, (4 unknowns) can
be solved using the continuity and momentum equations (4 equations). Once these quantities are
determined, the internal energy (i.e. temperature) can be calculated using the thermal energy equation
(Eqn. (4.108) ).
C. Wassgren
200
7.
For an adiabatic, inviscid flow, Eqn. (4.107) reduces to:

Dh Dp
=
(4.109)
Dt
Dt
The mechanical energy equation (Eqn. (4.94)) can be used to re-write the Lagrangian derivative of the
pressure:
Dui
p
ui
= ui
+ ui f B ,i
Dt
xi
ui
p
Dui p
p
=
+ ui
+ ui f B ,i
Dt t
t
xi

=
Dp
Dt
Dui
Dp p
=
ui
+ ui f B ,i
Dt t
Dt
Substituting into Eqn. (4.109):
Dui
p
Dh p
D 1
=
ui
+ ui f B ,i =
( 2 ui ui ) + ui f B,i
Dt t
Dt
t
Dt
(4.110)
D
p
( h + 1 2 ui ui ) = + ui f B,i
t
Dt
If the body force is conservative, i.e. fB = -G, then:
G
DG
G
ui f B,i = ui
=
+
(4.111)
xi
Dt
t
Furthermore, if the body force is also independent of time (obviously a good assumption if gravity is
the only body force considered), then Eqn. (4.111) may be substituted into Eqn. (4.110) and then
simplified to give:
D
p
DG
( h + 1 2 ui ui ) = t Dt
Dt
D
1 p
h + 1 2 ui ui + G =
(4.112)

Dt
t
= h0 , stagnation enthalpy
energy equation for the adiabatic flow of an inviscid fluid with
conservative, time-independent body forces
With the additional assumption that the flow is steady, we observe that the total specific enthalpy of a
fluid particle will remain constant, i.e. Dh0/Dt = 0. In particular, the total specific enthalpy will remain
constant along a streamline.
C. Wassgren
201
9.
Entropy Equation (aka the 2nd Law for a differential CV)
The entropy equation, which is simply the 2nd Law of Thermodynamics for a differential fluid element or
control volume, can be derived several different ways. Three of these methods are given below.
Method 1:
y
dy
z
dz
dx
Assume that the density, specific entropy, and velocity are , s, and u, respectively, at the control
volumes center. The entropy fluxes through each of the side of the control volume are given by:
( m x s )center 1
)in through left = ( m x s )center +
2 dx
( ms
x
= ( u x dydzs ) + ( u x dydzs ) ( 1 2 dx )
x
= u x s + ( u x s ) ( 1 2 dx ) ( dydz )
)out through right = u x s + ( u x s ) ( 1 2 dx ) ( dydz )

( ms
x
)in through bottom = u y s +

u y s ( 1 2 dy ) ( dxdz )
( ms
)out through top = u y s +

( ms
u y s
y
) ( 1 2 dy ) ( dxdz )
)in through back = u z s + ( u z s ) ( 1 2 dz ) ( dxdy )

( ms
)out through front = u z s + ( u z s ) ( 1 2 dz ) ( dxdy )

( ms
z
Thus, the net entropy flux out of the control volume is given by:
)net, out of CV = ( u x s ) +
u y s + ( u z s ) ( dxdydz )
( ms
y
z
x
The rate at which the entropy is increasing within the control volume is given by:
= ( s dxdydz ) = ( s )( dxdydz )
( ms )
t
t
within CV
(4.113)
(4.114)
where and s are the density and specific entropy, respectively, at the center of the control volume.
Note that since these quantities vary linearly within the CV (from the Taylor Series approximation),
the averages within the CV are and s.
C. Wassgren
202
The 2nd Law states that the rate of increase of entropy within the control volume plus the net rate at
which entropy leaves the control volume must be greater than or equal to the rate of heat transfer into
the control volume divided by the temperature of the fluid element to which the heat is transferred:
q
)
+ ( ms
into CV ( dxdydz )
(4.115)
( ms )
net, out of CV
t
T
within CV
where qinto CV is the rate of heat transfer into the CV per unit volume. Substituting Eqns. (4.113) and
(4.114) into Eqn. (4.115) gives:
q
( s )( dxdydz ) + ( u x s ) + u y s + ( u z s ) ( dxdydz ) into CV ( dxdydz )

t
y
z
T
x
q
( s ) + ( u x s ) + u y s + ( u z s ) into CV
x
y
z
T
t
This equation can be written in the following compact forms:

q
( s ) + ( us ) into CV
t
T
or
q
u j s into CV
( s ) +
T
t
x j
(4.116)
(4.117)
s
u j s = s + + s
u j + u j
( s ) +
t
x j
t
t
x j
x j
s
= s +
u j + + u j
t x j
t
x j

Ds
Dt
Substituting back into Eqn. (4.117) gives:

Ds qinto CV
entropy equation
Dt
T
(4.118)
Method 2:
Apply the 2nd Law of Thermodynamics directly to a small piece of fluid:
q
D
(4.119)
( s dxdydz ) into CV ( dxdydz )
Dt
T
Expanding the Lagrangian derivative gives:
D
Ds
D
( s dxdydz ) = ( dxdydz ) + s ( dxdydz )
Dt
Dt
Dt
but the second term on the RHS of this equation will be zero since the mass of the fluid element
remains constant. Thus, Eqn. (4.119) can be simplified to:
Ds qinto CV

Dt
T
C. Wassgren
203
Method 3:
Recall the integral form of the 2nd Law:
qinto CV
d
s dV + s ( u rel dA )
dV
dt
T
CV
CS
CV

( s )
d
s dV =
dV
and
dt
t
CV
CV
u rel = u
By utilizing Gauss Theorem (aka the Divergence Theorem), we can convert the area integral into a
volume integral:
s ( u dA ) = ( s u ) dV =
u j s dV
x j
CS
CV
CV
Substitute these expressions back into the 2nd Law to get:
q
u j s into CV dV = 0
( s ) +
x j
T
t
CV
q
u j s into CV = 0
( s ) +
t
x j
T
Ds qinto CV
Dt
T
Notes:
1. For a reversible, adiabatic flow (aka an isentropic flow):
Ds
=0
Dt
2.
Recall that for a simple, compressible system where the only surface forces are reversible pressure
forces (so that the = may be used in Eqn. (4.118)), the 1st and 2nd Laws of Thermodynamics may be
combined to give:
De 1
Dv
= qinto sys p
Dt
Dt
(i.e. de = vq pdv where q is the heat per unit volume, not per unit mass)
where v is the specific volume and for a simple, compressible system the total specific energy is equal
to the specific internal energy, i.e. e = u. Re-writing the specific volume in terms of the density,
utilizing Eqn. (4.118), and simplifying gives:
Du
Ds
D1
=T
p
Dt
Dt
Dt
Ds Du p D
=
(4.120)
Dt Dt 2 Dt
This expression may also be written in terms of the specific enthalpy, h, by re-writing the 2nd term on
the RHS of the equation:
D 1 1 Dp D p
Ds Du D p 1 Dp
p
=
+
=
T

Dt Dt Dt
Dt Dt Dt Dt

T
= Dh Dt
Ds Dh 1 Dp
=
Dt Dt Dt
C. Wassgren
(4.121)
204
10. Vorticity Dynamics

Recall that the vorticity, , of a fluid element is equal to twice the rotation rate of the element:
u
= u or i = ijk k
x j
Notes:
1. A rotational flow is defined as one in which the vorticity is not zero.
2.
An irrotational flow is defined as one in which there is no vorticity, i.e.

irrotational = u = 0
A useful concept when discussing vorticity is the vortex line. A vortex line is a line that is everywhere
tangent to the flows vorticity vectors.
Notes:
1. A vortex line is analogous to a streamline.
2.
A vortex tube is a tube made by all the vortex lines passing through a closed curve.
vortex lines
3.
A vortex filament is a vortex tube with infinitesimally small cross-section.
4.
There are no vortex lines in an irrotational flow.
5.
There can be no sources or sinks of vorticity in a flow. This follows from the following vector
identity:
( u ) = 0
= 0
(4.122)
Zero divergence of vorticity means that there are no sources or sinks of vorticity which in turn means
that vorticity is neither created nor destroyed in a flow.
So how then is vorticity generated in a flow? It must be introduced at a boundary (e.g. a fluid or solid
boundary). According to Eqn. (4.122), vortex lines must either form closed curves or start and end at
boundaries.
C. Wassgren
205
6.
Another useful quantity for the discussion of vorticity dynamics is the circulation, :
u ds
ds
(4.123)
u
flow
The relationship between the vorticity and the circulation about a curve, C, enclosing an area, A, with
unit normal, n , is found using Stokes Theorem:
u ds = ( u ) n dA
C
d
= n
dA
= n dA or
A
(4.124)
Notes:
a. The circulation around any cross-section of the same vortex tube remains constant. Recall that:
= 0
so that
2
( ) dV = 0
V
( n ) dA = 0
divergence theorem S
where V is the volume enclosed within the vortex tube

and S is the surface area of this volume.
A2
1
A1
Aside
Breaking the total area into the area of the top, bottom, and sides:
( n ) dA = 0 = ( n ) dA + ( n ) dA + ( n ) dA
S
A1
A2
Aside
On the sides of the vortex tube, the normal vector for the area is perpendicular to the vorticity
vectors (from the definition of a vortex tube) so that:
( n )sides = 0
Thus,
0=
( n ) dA + ( n ) dA
A1
A2
( n ) dA = ( n ) dA
A1
A2
Using Eqn. (4.124) and noting that the outward pointing normal vector on area A1 points in the
opposite direction as the vorticity vector there, we have:
1 = 2
Hence, the circulation around any cross-section of the same vortex tube remains constant. This
observation is also known as Helmholtzs Third Law.
C. Wassgren
206
11. Vorticity Transport Equations (aka Helmholtz Eqn)

The vorticity transport equation is an alternate expression of the Navier-Stokes equations. Consider the
Navier-Stokes equations for an incompressible fluid with constant dynamic viscosity:
Du
= p + 2 u + f
Dt
Divide through by the density, , (note that it is a constant here since were considering an incompressible
fluid) and also write the body force, f, as the gradient of a potential function, G (allowable if f is a
conservative body force):
g
p
Du
= + 2u G ( f B = G , e.g. Let G = gz so that f B = ge z .) z
(4.125)
Dt

=
Note that the kinematic viscosity, , is the ratio of the dynamic viscosity to the density. Now expand the
acceleration term on the LHS:
Du u
=
+ ( u ) u
Dt t
The second term in the previous equation can be expanded using the following vector identity:
u)
( u ) u = 12 ( u u ) u (
=
Substituting these relations into equation (4.125) gives:

p
u 1
+ ( u u ) u = + 2 u G
t 2
Now take the curl of Eqn. (4.126) and simplify:

u
p
+ 12 ( u u ) u = + 2 u + G
where
(4.126)
= ( u ) =
t t
t
12 ( u u ) = 0 (From the vector identity: =0.)
(u ) = u
)
(

= 0, vorticity
is divergence free
u)
(

( u ) + ( ) u (using a vector identity)
= 0, continuity
equation
p
= 0
2 u = 2 ( u ) = 2
G = 0
+ ( u ) ( ) u = 2
t

=
D
Dt
D
= ( ) u + 2
(4.127)
Dt
Vorticity Transport Equation for an incompressible, Newtonian fluid (aka Helmholtz Eqn)
C. Wassgren
207
Notes:
1. Lets interpret what each of the terms in the vorticity transport equation means:
D
= ( ) u +
2

Dt
diffusion of

rate of change
of fluid element
vorticity
stretching and
turning of a
vortex line
vorticity
2.
The vorticity transport equations do not contain pressure or body force terms explicitly. Assuming
uniform density, the pressure and body forces act through the center of mass of the element and thus
cannot produce rotation. Only the shear stresses may produce vorticity. Note that in a stratified flow
where the density gradient results in a non-coincident geometric center and center of mass, the pressure
forces can produce rotation of the fluid element. Hence, Eqn. (4.127) should not be used for stratified
flows.
3.
The vorticity transport equations are sometimes used in numerical calculations in place of the NavierStokes equations.
4.
For a 2D flow, the Vorticity Transport Equation simplifies to:

D
= 2 (since the vorticity points in a direction perpendicular to the stream lines)
Dt
Hence, vorticity can only diffuse (and not stretch) in a 2D flow. If the flow is inviscid, then:
D
= 0 (2D inviscid flow)
Dt
and the vorticity remains constant for each fluid element. Hence, in a 2D inviscid flow, if the flow
starts off irrotational, then it must remain irrotational! This is a very important result that will be
explored more fully when discussing Kelvins Theorem.
C. Wassgren
208
12. Bernoullis Equation

Eulers equations (the momentum equations for an inviscid fluid) can be simplified to an expression known
as Bernoullis Equation for the conditions given below:
Steady flow of an inviscid fluid in a conservative force field along either a streamline or a vortex line:
dp 1
+ 2 ( u u ) + G = constant
(4.128)
Irrotational flow of an inviscid fluid in a conservative force field:
dp 1
+
+ 2 ( ) + G = F ( t )
t
where u= and F(t) is a function only of time.
(4.129)
Derivation of Bernoullis Equation:

To begin, first consider Eulers equations (recall that Eulers equations are the momentum equations for an
inviscid fluid):
Du
1
= p G
Dt
where a conservative body force (fB = -G) has been assumed.

Re-write the convective acceleration term using the following vector identity:
(u ) u = ( 12 u u ) u ( u )
u
1
+ ( 1 2 u u ) u ( u ) = p G
t
Collect gradient terms on the left hand side of the equation:

1
u
p + ( 1 2 u u ) + G =
+ u ( u )
(4.130)
t
Note that the pressure gradient term can be re-written in a slightly different form as shown below.
1
dp
dp
dp
Note:
=d
p dx =
=
dx

a
a
a
a dx = e x + e y + e z ( dxe x + dye y + dze z )
1
dp
y
z
x
p =
a
a
a
dx + dy + dz
=
x
y
z
= da
Substituting this expression into Eqn. (4.130), simplifying, and noting that the vorticity, , is given by
=u:
dp 1
+ 2 u u + G =
+ u
(4.131)
Now consider two particular cases.
C. Wassgren
209
Steady flow along a streamline or a vortex line

A steady flow results in:
u
=0
t
Taking the dot product of Eqn. (4.131) with a little length of line, dx, that is along either a streamline or a
vortex line gives:
dp 1
+ 2 u u + G dx = ( u ) dx
(4.132)
Since the vector (u

) is perpendicular to both the streamline and the vortex line, the dot product with dx
will be zero:
( u ) dx = 0
Furthermore, the dot product of the gradient on the left hand side of Eqn. (4.132) with dx results in an
ordinary differential:
dp 1
dp 1
+ 2 u u + G dx = d
+ 2 u u + G

Thus,
dp 1
d
+ 2 u u + G = 0

dp 1
+ 2 u u + G = constant
(4.133)
Irrotational flow
In an irrotational flow, the velocity can be written as the gradient of a velocity potential function, , i.e.,
u=, since in an irrotational flow, =u=0, and from the vector identity, =0, thus u can be written
as u=. Substituting into Eqn. (4.131) and noting that the vorticity is zero in an irrotational flow:
( )
dp 1
+ 2 + G =
=
(4.134)
Combining gradient terms and simplifying:

dp 1
+
+ 2 + G = 0
(4.135)
Now take the dot product of the previous equation with a short distance in any direction, dx, and integrate
the resulting expression along that path:

dp 1
+
+ 2 + G dx = 0
dp 1
+ 2 + G = 0
d +
dp 1
+
+ 2 + G = F ( t )
(4.136)
t
where F(t) is a function only of time (this is introduced in the integration since the terms in the equation
may vary with both position and time).
C. Wassgren
210
Notes:
1. For a fluid with constant density, i.e. =constant:
dp 1
=
dp
2.
dp
fluid with constant density
For an ideal gas (p=RT):

isothermal case: T = T0
dp
d ( RT0 )
= RT0

= RT0 ln
ideal gas with isothermal conditions
0
where T0 and 0 are a reference temperature and density, respectively
dp
isentropic case (constant specific heats, i.e. a perfect gas):
dp
d p0 0
) = p
0 0
2 d =
p0
1 0
p p0 = ( 0 )
1 =
p0

1 0 0
where p0 and 0 are a reference pressure and density, respectively

This expression can be simplified further by noting that for a perfect gas:
p0
R
= RT0
cp =
0
1
and for a perfect gas undergoing an isentropic process:
1
T
=
T0 0
so that
dp
= c pT
C. Wassgren
perfect gas with isentropic conditions
211
Example:
A water tank has an orifice in the bottom of the tank:
g
h
area of orifice, A(0)
cross-sectional
area, A(y)
The height, h, of water in the tank is kept constant by a supply of water which is not shown. A jet of water
emerges from the orifice; the cross-sectional area of the jet, A(y), is a function of the vertical distance, y.
Neglecting viscous effects and surface tension, find an expression for A(y) in terms of A(0), h, and y.
SOLUTION:
Apply conservation of mass to the following CV:
0
y
2
d
dt CV
where
d
dV = 0 (The flow is steady.)
dt CV
rel
dA = V0 A0 + V2 A2
CS

A
V2 = V0 0
A2
(4.137)
Now apply Bernoullis equation from point 1 to point 0 and from point 1 to point 2:
( p + 12 V 2 gy ) = ( p + 12 V 2 gy ) = ( p + 12 V 2 gy )
1
where
p1 = p0 = p2 = patm (These points are all at free surfaces.)
V1 = 0 and V0 and V2 are related through Eqn. (4.137).

y1 = h, y0 = 0, y2 = y
C. Wassgren
212

gh = 12 V02 = 12 V22 gy
2
A
gh = V = V 0 gy
A2
The first two equations in the previous expression state that:
V0 = 2 gh
Equation (4.138) combined with the second two equations gives:
1
2
2
0
1
2
2
0
(4.138)
A0
gy
= 1+ 1
2
A2
2 V0
2
A0
gy
y
= 1+
= 1+
A
gh
h
2
A2
1
=
A0
y
1+
h
C. Wassgren
213
Example:
a.
b.
c.
Using an integral approach, write the differential equation governing the motion of an inviscid,
incompressible fluid (with density ) oscillating within the U-tube manometer shown. The
manometer cross-sectional area is A.
What is the natural frequency of the fluid motion?
What are the implications of this result for making time-varying pressure measurements using a
manometer?
tube ends are open to the
atmosphere
g
h1
h2
incompressible, inviscid fluid

with density
Assume this distance is negligible

compared to h1+h2.
SOLUTION:
Apply the LME in the y-direction to the following two CVs.
h1
CV 1
CV 2
C. Wassgren
h2
214
CV 1:
d
uY dV + CS uY ( u rel dA ) = FBY + FSY
dt CV
where
d 2 h1 dh1 2
d
d dh1
dV
=
Ah
=
A
h1 2 +
Y
1
dt CV
dt dt
dt
dt
u ( u
Y
rel
dA ) = 0
CS
FBY = Ah1 g
FSY = pbottom A
d 2 h dh 2
A h1 21 + 1 = Ah1 g + pbottom A
dt
dt
2
d 2h
p
dh
h1 21 = h1 g 1 + bottom
dt
dt
(4.139)
CV 2:
d
uY dV + CS uY ( u rel dA ) = FBY + FSY
dt CV
where
d 2 h2 dh2 2
d
d dh2
u
dV
Ah
A
h2 2 +
2

Y
dt CV
dt dt
dt
dt
u ( u
Y
rel
dA ) = 0
CS
FBY = Ah2 g
FSY = pbottom A
d 2 h dh 2
A h2 22 + 2 = Ah2 g + pbottom A
dt
dt
2
h2
d 2 h2
p
dh
= h2 g 2 + bottom
2
dt
dt
C. Wassgren
(4.140)
215
From conservation of mass considering both CVs combined together:

dh2
dh
= 1
dt
dt
(One side moves down at the same rate that the other side moves up.)
(4.141)
Subtract Eqn. (4.140) from (4.139) and make use of Eqn. (4.141).
2
h1
p
p
d 2 h1
d 2 h2
dh
dh
h
= h1 g 1 + bottom + h2 g + 2 bottom
2
2
2
dt
dt
dt
dt
h1
d 2 h1
d dh
dh dh
h2 1 = ( h2 h1 ) g 1 + 1
2
dt dt
dt
dt dt
( h1 + h2 )
d 2 h1
+ g ( h1 h2 ) = 0
dt 2
(4.142)
Let:
L = h1 + h2
z = h1 h2
dh
dz dh1 dh2 dh1 dh1
=
=
+
=2 1
dt
dt
dt
dt
dt
dt
2
2
d h
d z
= 2 21
dt 2
dt
so that Eqn. (4.142) becomes:
d 2 z 2g
+
z=0
dt 2
L
(4.143)
The general solution to this differential equation is:

2g
2g
z = A sin t
+ B cos t
L
L
Specifying the following initial conditions:

z ( t = 0 ) = z0
(4.144)
(4.145)
dz
(t = 0) = V
dt
gives the solution:
z =V
2g
2g
L
sin t
+ z0 cos t
2g
L
L
(4.146)
The natural (radian) frequency of the manometer, , is:
2g
L
(4.147)
The practical implication of this result is that one must make sure that the manometer fluid length, L, is
sufficiently small so that the manometers natural frequency is large enough to accurately capture the
temporal variations in the pressure measurements. In other words, a manometer with a large L will not be
able to capture rapid pressure fluctuations.
C. Wassgren
216
This problem may also be solved using a accelerating frames of reference.

y1
y2
These FORs are fixed to

the free surfaces.
h1
CV 1
CV 2
h2
Y
LMEs in the y-direction using the indicated accelerating FORs:

d
u y dV + CS u y ( u rel dA ) = FBy + FSy CV ay / Y dV
dt CV
where
d
d
u y dV = 0
u y dV = 0
dt CV1
dt CV2
u ( u
y
rel
dA ) = 0
u ( u
y
rel
dA ) = 0
CS1
CS2
FBy1 = h1 Ag
FBy 2 = h2 Ag
FSy1 = pbottom A
FSy 2 = pbottom A
CV1
a y / Y dV =
d 2 h1
h1 A
dt 2
d 2h
0 = h1 Ag + pbottom A 21 h1 A
dt
Subtract the two previous equations and simplify:
d 2h
d 2h
0 = ( h2 h1 ) g 21 h1 + 22 h2
dt
dt
Make use of the Eqn. (4.141) and re-arrange to get:
d 2h
( h1 + h2 ) 21 + g ( h1 h2 ) = 0
dt
This is the same equation as Eqn. (4.142)!
C. Wassgren
CV2
a y / Y dV =
d 2 h2
h2 A
dt 2
0 = h2 Ag + pbottom A
217
d 2 h2
h2 A
dt 2
This problem may also be solved using the unsteady Bernoulli equation. Assuming the flow is irrotational,
inviscid, and incompressible, Bernoullis equation may be written at an instant in time as:
y
p 1 2
p 1 2
g
(4.148)
+ + 2 V + gy =
+ + 2 V + gy
t
1 t
2
where the point 1 is on the free surface of the left leg of the manometer and point 2 is on the free surface of
the right leg of the manometer. In addition:
V=
=
t 1 t
=
2
y = h1
dV
dh
d 2h
dh
d
(V1h1 ) = h1 1 + V1 1 = h1 21 + 1
dt
dt
dt
dt
dt
Vdy =
dV
dh
d 2h
dh
d
(V2 h2 ) = h2 1 + V1 2 = h2 22 + 2
dt
dt
dt
dt
dt
y=0
y = h2
Vdy =
y =0
p1 = p2 = patm
V12 = V22
y1 = h1
y2 = h2
2
2
d 2 h dh
d 2 h dh
h1 21 + 1 + gh1 = h2 22 + 2 + gh2
dt
dt
dt
dt
Making use of Eqn. (4.141) gives:
d 2h
(4.149)
( h1 + h2 ) 1 + g ( h1 h2 ) = 0
dt
Eqn. (4.149) is identical to Eqn. (4.142) so the solution will be the same as that derived previously.
C. Wassgren
218
Another Approach to Deriving Bernoullis Equation

We can also derive Bernoullis equation by considering LME and COM applied to a differential control
volume as shown below. In the following analysis, well make the following simplifying assumptions:
1. steady flow
2. inviscid fluid
3. incompressible fluid
Note that the control volume shown follows the streamlines.
streamlines
A+1/2dA
(pA)
(p+1/2dp)(A+1/2dA)
V+1/2dV
A-1/2dA
dsAg
(p-1/2dp) (A-1/2dA)
V-1/2dV
-1/2d
+1/2d
y
z
(pA)
ds
x
dz = ds sin
First apply COM to the CV shown above:

d
dV + ( u rel dA ) = 0
dt
where
d
dt
CV
CS
dV = 0
(steady flow)
CV
( u
rel
dA ) = (V + 1 2 dV )( A + 1 2 dA )
CS
(V 1 2 dV )( A 1 2 dA )
= VdA + AdV + H .O.T .
(Note that there is no flow across a streamline.)
VdA = AdV
C. Wassgren
219
Now apply the LME in the s-direction to the same CV:

d
us dV + us ( u rel dA ) = FB , s + FS , s
dt
where
d
dt
CV
CS
u dV = 0
(steady flow)
CV
u ( u
s
rel
dA ) = (V 1 2 dV )
( A 1 2 dA)
CS
+ (V + 1 2 dV )
( A + 1 2 dA)
= 2 VAdV + V 2 dA + H .O.T .
FB, s = dsA ( g sin ) = Ag ds
sin
= Agdz

= dz
FS , s = ( p
2 dp
)( A
2 dA
) ( p + 1 2 dp )( A + 1 2 dA) + pdA
= Adp + H .O.T .
2 VAdV + V 2 dA = Agdz Adp
Substitute the result from COM into the result from the LME and simplify:
2 VAdV + V 2 dA = Agdz Adp

= VAdV
dp
+ VdV + gdz = 0
We can integrate this equation to get:

p 1 2
+ 2 V + gz = constant
(4.150)
Again, its important to review the assumptions built into the derivation for Eqn. (4.150):
1. steady flow
2. inviscid fluid
3. incompressible fluid
4. flow along a streamline
C. Wassgren
220
13. Kelvins Theorem

In an inviscid flow of a fluid with constant density, or a fluid where the pressure is a function only of
the density, where the only body forces are conservative, the vorticity of each fluid element is
preserved.
Notes:
1. A conservative body force is one that can be written as the gradient of a potential function, i.e.:
f = G
The force due to gravity is an example of a conservative body force:
f = ( gz ) = ge z
2.
A fluid in which the pressure is a function only of the density, i.e., p=p(), is called a barotropic
fluid. An example of such a fluid would be an ideal gas undergoing an isentropic flow process:
3.
p
=
p0 0
An important result of Kelvins Theorem is that if a flow starts off irrotational, viscous forces are
negligible, and the fluid has either constant density or is a barotropic fluid, then the flow will
always remain irrotational.
Proof of Kelvins Theorem:

Consider the flow of an inviscid fluid where only conservative body forces are considered. The time rate of
change of the circulation about a specific collection of fluid particles is given by:
D D
D
=
u dl =
ui dxi
(4.151)
Dt Dt
Dt

C

C
where is the circulation, C is the contour about the fluid particles, u is the fluid velocity, and dl=dxiei is a
small displacement along the contour. The Lagrangian time derivative can be brought inside the contour
integral (refer to Pneuli, D. and Gutfinger, C., Fluid Mechanics, pp. 310-312 for the proof to this) to give:
Du
D ( dxi )
D
= i dxi + ui
(4.152)
Dt
Dt
Dt

C
Note that:
D ( dxi )
x
x
Dx
= d i = d i + u j i = dui
Dt
t
x
Dt

j
=0
= ij
We can also substitute in for the fluid particle acceleration using Eulers equations:
Dui
1 p G
=
Dt
xi xi
Note that in the previous equation conservative body forces have been assumed where:
G
fi =
xi
Substituting Eqns. (4.153) and (4.154) into Eqn. (4.152) gives:
1 p
D
G
=
dxi
dxi + ui dui
Dt
xi
xi
(4.153)
(4.154)
(4.155)
C. Wassgren
221
Several of the terms in this equation may be simplified since:

p
dxi = dp
xi
G
dxi = dG
xi
(4.156)
ui dui = d ( 1 2 ui ui )
dp
D
dp
= dG + d ( 1 2 ui ui ) =
Dt
dG + d (
C
2 ui ui
(4.157)
The second and third terms in this equation are zero since these functions are single-valued (i.e. at each
location the quantities have a unique value) and the contour C is a closed curve:
dG = 0
C
d (
2 ui ui
)=0
Thus,
D
=
Dt

C
dp
(4.158)
If the density is a constant then:

D
1
=
dp
Dt

C
and since the pressure is a single-valued function:

D
=0
Dt
Since the circulation about the fluid particles remains unchanged, the vorticity of the fluid particles will
also remain unchanged.
Also, if the pressure is a function only of the density then:
dp
p = p ( ) dp =
d dp = f ( ) d
d
D
dp
d
=
= f ()
Dt

C

C
D
=0
Dt
since the density is a single-valued function.
Therefore we see that for the flow of an inviscid fluid in a conservative force field where either the density
of the fluid is constant or where the pressure is a function only of the density, the vorticity of a collection of
fluid particles will remain unchanged.
C. Wassgren
222
Notes:
1. When the pressure is a function of variables other than the density, the contour integral will not, in
general, be zero.
2.
As might be anticipated, the vorticity of a fluid element may be changed through the action of
viscosity, non-conservative forces, or density variations that are not a function solely of the pressure
variations.
3.
Kelvins Theorem applies strictly to a simply-connected region, i.e. a contour that does not intersect
itself and contains only fluid. A contour that surrounds some object, e.g. an airfoil, is not a simplyconnected region and therefore Kelvins Theorem does not hold for such a contour. This fact is
significant when examining the lift on an airfoil since it is possible to have circulation around an airfoil
(i.e., around a non-simply connected region) in an otherwise irrotational flow.
C. Wassgren
223
14. Croccos Equation

Croccos equation relates the vorticity of a flow field to the gradients in the entropy and stagnation enthalpy
of the fluid in a flow where viscosity and body forces are negligible. Croccos equation is given as:
u
u + T s = ( h + 1 2 u u ) +
(4.159)
t
Derivation of Croccos Equation:

Consider the momentum equations for an inviscid fluid (Eulers equations) for a flow in which body forces
are negligible:
Du u
1
=
+ ( u ) u = p
(4.160)
Dt t
Re-write the convective derivative term using the following vector identity:
(u ) u = ( 12 u u ) u ( u )
Utilize the definition of vorticity, :
= u
and substitute into equation (4.160) to get:
u
1
+ ( 1 2 u u ) u = p
t
(4.161)
Now consider the 1st Law of Thermodynamics for a fluid element (assumed to be a pure substance) where
only reversible pressure work is considered:
1
1
du = q pd
(4.162)
Note that q is the amount of heat added to the fluid element per unit volume. Substituting the definition of
entropy for a reversible (zero viscosity has been assumed) process:
1
Tds = q
and enthalpy:
p
1 dp
dh = du + d = du + pd +

into Eqn. (4.162) and simplifying gives:
dp
dh = Tds +
(4.163)
Note that we can write Eqn. (4.163) in a slightly different form by utilizing the following:
dh = h dx
ds = s dx
dp = p dx
where dx is a small length in any direction. Substituting these relations into Eqn. (4.163) and simplifying
gives:
p
= T s h
(4.164)
C. Wassgren
224

u
+ ( 1 2 u u ) u = T s h
t
Re-arranging this equation results in Croccos Equation:
u
u + T s = ( h + 1 2 u u ) +
t
Recall that this equation holds for a flow in which viscous and body forces are negligible.
(4.165)
Notes:
1. Consider a steady, inviscid flow in which body forces are negligible so that Eqn. (4.159) becomes:
u + T s = h0 where h0 = h + 1 2 u u
(4.166)
Lets restrict our investigation to flow along a streamline by taking the dot product of the previous
equation with u.
u ( u ) + T ( u ) s = ( u ) h0
(4.167)

= 0 (vector identity)
Note that since were concerned here with steady flows,

Dh0
Ds
= ( u ) s and
= ( u ) h0
(4.168)
Dt
Dt
and Eqn. (4.167) becomes:
Ds Dh0
=
(4.169)
T
Dt
Dt
Hence, if there is no dissipation along a streamline (i.e., Ds/Dt = 0), then the stagnation enthalpy, h0,
must remain constant along the streamline (since Dh0/Dt = 0).
2.
Now consider the case where the stagnation enthalpy in a steady flow is uniform along the streamlines
so that h0 = 0. For this case, Eqn. (4.159) becomes:
u = T s
(4.170)
For such a flow we can conclude the following important statement. For the steady flow of an
inviscid fluid in which body forces are negligible and where the stagnation enthalpy is constant,
an irrotational flow will be isentropic and an isentropic flow will be irrotational.
3.
Consider the uniform, supersonic flow in front of a blunt-nosed object. The incoming flow will have a
constant stagnation enthalpy and will also be irrotational (and thus isentropic). A curved shock wave
will stand in front of the object. Across the shock wave the stagnation enthalpy will remain constant
but the entropy will change (flow across a shock wave is a non-isentropic process). Since the shock
wave is curved, there will be a gradient in the entropy normal to the downstream streamlines. Thus,
from Croccos equation we observe that vorticity will also be generated downstream of the curved
shock wave and the flow will, by definition, be rotational.
Although there is a change in entropy across a normal shock wave and an oblique shock wave, there is
no entropy gradient in the normal direction (the entropy change across the shock is uniform along the
length of the shock) and, thus, the flow will remain irrotational downstream of normal and oblique
shocks if the upstream flow is irrotational. Note that Croccos equation does not strictly apply across a
shock wave since the large velocity gradient within the shock means that viscous effects are also
significant there. However, Croccos equation can be applied just upstream and downstream of the
shock wave.
C. Wassgren
225
Review Questions
1. Describe what each term represents in the Lagrangian derivative.
2. What is the formal definition of an incompressible fluid? Give an example of an incompressible flow
where the fluid density is not uniform?
3. Write the continuity equation for an incompressible fluid?
4. Write the continuity equation for a fluid with constant and uniform density?
5. Describe the naming and sign convention for stresses, ij.
6. Will the stress tensor always be symmetric?
7. Describe the various ways in which a fluid element can deform.
8. What is the vorticity of a fluid element (in words and in mathematical form)?
9. What is the deformation rate tensor (in words and in mathematical form)?
10. What is meant by an irrotational flow?
11. What three key assumptions are made in deriving the stress-strain rate constitutive relation for a
Newtonian fluid?
12. How is the mechanical pressure related to the thermodynamic pressure?
13. Why doesnt the bulk viscosity enter into incompressible fluid flow problems?
14. Describe what each term represents in the Navier-Stokes equations.
15. What are Eulers equations?
16. Describe the various ways in which a fluid element can deform.
17. What is the vorticity of a fluid element (in words and in mathematical form)?
18. What is the deformation rate tensor (in words and in mathematical form)?
19. What is meant by an irrotational flow?
20. What three key assumptions are made in deriving the stress-strain rate constitutive relation for a
Newtonian fluid?
21. How is the mechanical pressure related to the thermodynamic pressure?
22. Why doesnt the bulk viscosity enter into incompressible fluid flow problems?
23. What does each term represent in the energy equation?
24. What does each term represent in the mechanical energy equation?
25. What does each term represent in the thermal energy equation?
26. What is Fouriers Law of Conduction (in mathematical terms)?
27. What does the energy dissipation function represent?
28. Is the energy dissipation function always positive?
29. Is the thermal energy equation required to solve for the flow velocity and pressure in incompressible
flows? How about for compressible flows?
30. Under what conditions can vorticity be generated within a flow?
31. How are vorticity and circulation related?
32. What are the assumptions that go into the following form of Bernoullis equation?
p V2
+
+ z = constant
g 2g
33. What are the assumptions that go into the following form of Bernoullis equation?
p V 2
+ +
+ gz = F ( t )
t 2
34. What does Bernoullis equation look like for an ideal gas flowing isentropically?
35. What is Kelvins Theorem? What is its significance?
C. Wassgren
226
Chapter 05:
Potential Flows
1. Stream Functions
2. Potential Functions
3. Complex Variable Methods for Potential Flows
4. Blasius Integral Laws
5. Kutta-Joukowski Theorem
6. Conformal Mappings
7. Joukowski Transformation
8. Method of Images
9. Added Mass
10. Numerical Methods for Solving Potential Flows
11. Doublet Distributions
C. Wassgren
Chapter 05: Potential Flows
227
1.
Stream Functions
A stream function is a special scalar function that is useful when analyzing 2D flows. As will be shown, a
stream function has the following properties:
1. A stream function satisfies the continuity equation.
2. A stream function is a constant along a streamline.
3. The flow rate between two streamlines is equal to the difference in the streamlines stream
functions.
First, lets define the stream function.
Define a scalar function, , called a stream function, such that the continuity equation is automatically
satisfied for 2D (planar and axi-symmetric) flows.
For a 2D incompressible flow in rectangular coordinates, define =(x, y) such that:
ux
and u y
(5.1)
y
x
If the stream function is defined in this manner, then the continuity equation will automatically be
satisfied:
u u y 2 2
u x
0
x
y xy yx
For a 2D incompressible flow in polar coordinates, =(r, ):
1
and u
ur
r
r
The continuity equation for a 2D, incompressible flow in polar coordinates is:
1
1 u
1 2 1 2
0
rur
r r
r r r r r
(5.2)
Now lets consider some of the additional properties of the stream function.
C. Wassgren
228
The Stream Function is Constant Along a Streamline

Lets determine the curve along which the stream function remains constant. Well only consider an
incompressible flow in rectangular coordinates here for simplicity (the same result holds for polar
coordinates and compressible flows).
The total change in the stream function, d, where =(x, y), over some displacement (dx, dy) is given by:
x, y d
dx
dy u y dx u x dy
x
y
where the definition of the stream function has been used to write d in terms of the velocities. To find
the slope of the curve along which =constant, we let d=0.
d 0 u y dx u x dy
uy
dy
dx constant u x
Notice that the slope of the curve along which the stream function is constant is exactly the same as the
slope of a streamline. Thus, we conclude that the stream function is constant along a streamline!
=1
=2
=3
where 1, 2, and 3 are constants.

Note that these constants may vary
from streamline to streamline.
streamlines
C. Wassgren
229
Example:
A particular planar, incompressible flow can be described with the following stream function:
Axy
where A is a constant.
a. Sketch the streamlines for the flow.
b. Determine the velocity components for the flow.
SOLUTION:
The stream function is a constant along a streamline so the equation of the streamlines will be:
1
y
(hyperbolas!)
Ax
A plot of the streamlines is shown below. Note that A has been assumed to be a positive constant (i.e. A >
0) in determining the direction of the flow.
y
A>0
The velocities are determined from the definition of the stream function.
Ax and u y
Ay
ux
x
y
C. Wassgren
230
The Flow Rate Between Two Streamlines is Equal to the Difference in their Stream Functions
Now lets examine how the flow rate between two streamlines is related to the stream function. Consider
the sketch below.
A
dy
dA
dx
dx=dA sin
dy=dA cos
n cos e x sin e y
The volumetric flow rate passing between the two streamlines, and thus crossing through a line drawn
between the two streamlines can be found by calculating the volumetric flow rate through a small piece of
the line and then integrating from one streamline to the other.
dx
dy
dQ u dA u x e x u y e y cos e x sin e y dA u x e x u y e y e x
e y dA
dA
dA
dy
dx
u x dy u y dx
y
x
dQ d
Integrating from streamline A to streamline B gives:
QAB B A
The volumetric flow rate between two streamlines is equal to the difference in the streamline stream
functions!
Note that if B>A then the flow is from left to right. If B<A then the flow is from right to left.
C. Wassgren
231
Example:
The velocity field for a planar, incompressible flow is given by:
u 2( x 2 y 2 )e x 4 xye y
a.
b.
Determine the stream function for this flow field if (0,0)=0.

Determine the volumetric flow rate across the line AB shown in the figure.
y
B: (0,1)
A: (1,0)
SOLUTION:
Recall that the velocities are determined from the stream function in the following manner.
ux
2 x2 y 2
y
uy
4 xy
x
2 x 2 y 13 y 3 f x
(5.3)
2x 2 y g y
(5.4)
where f and g are, at this point, unknown functions of x and y, respectively. Comparing Eqns. (5.3) and
(5.4) indicates that:
f x c and g y 23 y 3 c
where c is a constant. Hence, the stream function is:
2 x 2 y 13 y 3 c
(5.5)
Knowing that (0, 0) = 0 we can conclude that c = 0 and:

2 x 2 y 13 y 3
(5.6)
Recall that the volumetric flow rate between two streamlines is equal to the difference in the streamline
stream functions.
QAB B A
(5.7)
where
B 0,1 23
A 1, 0 0
so that
QAB 23
C. Wassgren
(5.8)
232
Building Block Stream Functions

The properties of stream functions described previously are enough to justify their use. There are
additional reasons to use stream functions, however. Models of real flows can be produced by combining
building block stream functions. The significance of this topic will be discussed in greater detail when
examining the potential function (especially the complex potential); a topic covered later in the notes. For
now, however, it is sufficient to present some building block stream functions and discuss how they can
be combined to produce models of actual flows. First, lets examine a few basic building block stream
functions.
uniform stream
shown for U0,V0> 0

V0
y
x
V0 x U 0 y
u x U 0 ; u y V0
U0
line source (m>0) or sink (m<0)

y
r
free line vortex

y
shown for m> 0

m is referred to as the
source (sink) strength.
forced line vortex

y
ln r
2
shown for > 0
is referred to as the
circulation.
ur 0; u
1
2 r
shown for > 0

Kr 2
2
ur 0; u Kr
m
0 2
2
m 1
; u 0
ur
2 r
shear flow
shown for > 0
u x 2 Ay; u y 0
extensional flow
y
shown for > 0
C. Wassgren
Ay 2
Can also be used to

model flow in a corner
and stagnation point
flow.
233
Axy
u x Ax; u y Ay
Superposition of Stream Functions

Because the continuity equation is a linear PDE, the building block stream functions just presented can
be combined together to produce new stream function flows.
Proof:
LetT=1+2 where 1 and 2 are stream functions that satisfy the continuity equation. The
velocities determined from the new stream function are given by:
T 1 2 1 2
ux
y
y
y
y
1 2
T
2
1
x
x
x
x
Substitute these velocities into the continuity equation:
u u y
2 1 2
u x
1
x
y x y
y y x
x
1 2 1 2
xy xy yx yx
0
Thus, the new stream function, T, formed by the superposition of the original stream functions also
satisfies the continuity equation.
uy
C. Wassgren
234
Example:
The doublet is formed by superposing a source and sink of equal strength separated by an infinitesimal
distance.
y
r2
2
r1
1
The stream function for the source/sink pair is given by (m>0):

m
m
m

2 1
2 1
2
2
2
Re-arrange and take the tangent of both sides:
tan 2 tan 1
2
tan
(5.9)
tan 2 1
m
1 tan 21
Note that a trig identity has been used in deriving the previous expression. From the figure we observe
that:
r sin
r sin
tan 1
and tan 2
r cos a
r cos a
Substitute these expressions into Eqn. (5.9) and simplify:
r sin r sin
2 r cos a r cos a
tan
m 1 r sin r sin
r cos a r cos a
r 2 sin cos ar sin r 2 sin cos ar sin
r 2 cos 2 a 2
r 2 sin 2
1 2
r cos 2 a 2
2ar sin
2
r
cos 2 a 2
2
2
r cos a 2 r 2 sin 2
r 2 cos 2 a 2
2ar sin
2
r a2
m
2ar sin
tan 1 2
(5.10)

2
2
r a
Stream function for a source/sink pair of equal strength each located a distance a from the origin
along the x-axis.
C. Wassgren
235
The streamlines for the stream function given in Eqn. (5.10) are shown in the following figure.
y
Note that as a0, Eqn. (5.10) becomes:

m 2ar sin
lim
a 0
2 r 2 a 2
since the tangent of a very small angle approaches the value of the angle. If we let the source and sink
approach each other (a0) while we let the source/sink strength approach infinity (m) such that the
ratio ma/=K=constant, then the stream function becomes:
K sin
doublet oriented
Note: 0<2
r
along x -axis
The streamlines for the doublet are circles passing through the origin as shown in the figure below.
y
shown for K>0
Note:
1. Doublets have an orientation. The stream function for a doublet oriented in the y-direction is given by:
K cos
y
doublet oriented
r
along y-axis
shown for K>0
x
C. Wassgren
236
Example:
The flow of a frictionless fluid (there can be slip at solid surfaces) around a non-rotating cylinder can be
modeled as a uniform stream superimposed with a doublet:
flow around uniform doublet
cylinder
stream
If the cylinder radius is R, determine the velocity of the fluid on the surface of the cylinder as a function of
angular position, .
y
uniform stream with velocity, U
doublet with strength k
Note that inside the cylinder the

streamlines look like:
C. Wassgren
237
SOLUTION:
The stream function for flow around a non-rotating cylinder is given by combining the stream function for
a uniform stream with the stream function for a (horizontally-oriented) doublet.
flow around uniform doublet
cylinder
stream
Uy
K sin
r
K sin
r
cylinder
At the moment, K is an unknown constant. It can be determined by noting that there is no flow through the
cylinder surface. Hence, the radial velocity, ur, at r = R should be zero.
1
K cos
ur
U cos
r
r2
K sin
u
U sin
r
r2
No flow through the cylinder surface at r = R:
K cos
ur r R 0 U cos
R2
2
K UR
Hence, the stream function and flow velocities for flow around a non-rotating cylinder are:
R 2
flow around Ur sin 1
cylinder
r
flow around Ur sin
ur
R 2
1
Ur cos 1
r
r
R 2
U sin 1
r
r
On the cylinder surface (r = R):

ur r R 0 and u r R 2U sin
Note that there are stagnation points at = 0, A maximum speed of 2U occurs at = /2, -/2.
C. Wassgren
238
Notes:
1. Stream functions can also be defined for steady, compressible, 2D flows. For example, in rectangular
coordinates:

ux 0
and u y 0
where 0 is a reference density
y
x
The continuity equation for these conditions is:
u x u y 0
0
u

0
x
x
y
x y y
2.
Stream functions also exist for axi-symmetric, incompressible flows (referred to as Stokes stream
functions): =(r, z)
1
1
ur
and u z
r z
r r
The continuity equation for these conditions is:
1 rur u z
0
r r
z
3.
Stream functions cannot be defined for arbitrary 3D flows.
C. Wassgren
239
2.
Potential Functions
The velocity field for an irrotational flow can be written as the gradient of a potential function, :
u
since, from a vector identity:
0
and because an irrotational flow is defined as one with zero vorticity, i.e.:
0
(5.11)
Now lets ensure that the continuity equation is satisfied for an incompressible fluid (compressible potential
flows will be considered in a separate set of notes dedicated specifically to compressible flows):
u 0 0
2 0
(5.12)
This is Laplaces Equation!, a well studied, linear, elliptic partial differential equation that appears in many
other disciplines such as electromagnetics and conduction heat transfer.
The momentum equations for a potential flow simplify to Bernoullis equation since the flow is everywhere
irrotational (refer to an earlier set of notes concerning Bernoullis equation):
p 1
2 G F t
(5.13)
t
where G is a conservative body potential (e.g. for gravity, G = gz where g and e z point in opposite
directions) and F(t) is a function of time. For a steady potential flow, Eqn. (5.13) simplifies to:
p 1
2 G constant
(5.14)
Note that the momentum equation (Eqn. (5.13) or (5.14)) need not be solved to determine the fluid
kinematics. Solving Eqn. (5.12) subject to appropriate boundary conditions is sufficient to determine the
flow velocity field. This occurs because we placed two restrictions on the flow field: the continuity
equation and the irrotationality assumption. The momentum equation can be solved to determine the fluid
pressure field once the velocity field is known.
The appropriate boundary conditions for Laplaces equation are either Dirichlet (the functions value is
specified), Neumann (the functions gradient is specified), or mixed. At solid surfaces the appropriate
boundary condition for the flow is that the flow velocity normal to the surface is equal to the surface
velocity, i.e.:
u n U n
(5.15)
where u is the fluid velocity, U is the boundary velocity, and n is the normal vector to the boundary. This
is a Dirichlet or kinematic boundary condition. Note that the no-slip condition is not satisfied for potential
flows. This occurs because potential flows have no viscous force contributions since the viscous terms in
the Navier-Stokes equations (i.e. momentum equations) drop out due to the irrotationality assumption. As
a result, the Navier-Stokes equations, which are normally 2nd order PDEs, simplify to the 1st order Eulers
equations (and can be simplified further to Bernoullis equation). Hence, only a single boundary condition
must be specified.
Neumann boundary conditions are specified at free surfaces, i.e. surfaces where the pressure is defined.
These are sometimes called dynamic boundary conditions. Bernoullis Equation (Eqns. (5.13) or (5.14)) is
used to relate free pressure boundary conditions to the velocity field.
C. Wassgren
240
Notes:
1.
Incompressible potential flows are often referred to as ideal fluid flows since the fluid is
incompressible and viscous forces are negligible.
2.
Potential functions can be defined for 3D flows (as long as theyre irrotational). Recall that stream
functions could only be specified for 2D flows.
3.
The governing equation for potential flows (Laplaces equation) is a linear PDE so that the principle of
superposition can be used to combine potential flow solutions. The approach is similar to that
discussed previously for stream functions.
4.
Potential functions and stream functions are intimately related. This will become clear in the following
section of notes concerning the complex potential function.
5.
Streamlines ( = constant) and equi-potential lines ( = constant) are perpendicular everywhere in the
flow. Consider the curves along which = constant (a streamline) and = constant:
d 0 dx
d 0 dx
where dx is a small distance along the curves. Re-write these relations in terms of the velocities:
dy u y
dx 0 u y dx u x dy
dx u x
dx 0 u x dx u y dy
u
dy
x
dx
uy
From analytical geometry, two curves are perpendicular to each other if the slopes of the curves
multiplied together equals -1. Hence, we see that the streamlines and equi-potential lines will always
be perpendicular to each other. The resulting mesh of streamlines and equi-potential lines is known as
the flow net.
= 1
= 2
= 3
= 2
= 3
= 1
C. Wassgren
241
3.
Complex Variable Methods for Investigating Planar, Ideal, Irrotational Flows

A good mathematics reference for this topic is: Churchill, R.V. and Brown, J.W., Complex Variables
and Applications, McGraw-Hill.
Lets define the complex potential, f(z):

where
f ( z ) i
z = x+iy = rexp(i) where 0 < 2and exp(i)=cos()+isin()

is the velocity potential
is the stream function
y
r
Why do this? Because it allows us to present information in a compact manner and because we can use
tools from complex variable mathematics to analyze fluid flows.
Notes:
1.
A few complex variables preliminaries:

a. z = x+iy = rexp(i) where 0 < 2and exp(i)=cos()+isin()
b. |z| = (x2+y2)1/2 = r
c. arg(z) = tan-1(y/x) =
zz z x 2 y 2
e.
f.
log(z) = ln(r) + i
A function f of the complex variable z is analytic on an open set if it has a derivative at each point
in that set. (counter-example: f(z) = |z|2 is not analytic anywhere since its derivative exists only at
z = 0.)
A function, h, is harmonic if it has continuous partial derivatives of the first and second order and
satisfies Laplaces equation:
2 h 0
If a function, f(z)=a(x,y)+ib(x,y), is analytic in D, then the first order partial derivatives of its
component functions, a and b, must satisfy the Cauchy-Riemann equations throughout D.
a b
a
b
and
x y
y
x
If two functions, a and b, are harmonic in a domain D and their first-order partial derivatives
satisfy the Cauchy-Riemann equations throughout D, b is said to be a harmonic conjugate of a.
g.
h.
i.
2.
d.
If a function, f(z)=a(x,y)+ib(x,y), is analytic in a domain D then its component functions, a and b, are
harmonic conjugates in D.
Proof:
Since f(z) is analytic in a domain D, then the first order partial derivatives of its component functions,
a(x,y) and b(x,y), satisfy the Cauchy-Riemann equations throughout D (Note #1h).
Differentiating the Cauchy-Riemann equations gives:
2 a 2b
2a
2b
x 2 xy
xy
x 2
and
2 a 2b
2a
2b
2
2
yx y
yx
y
But from advanced calculus:
2a
2a
2b
2b
and
xy yx
xy yx
C. Wassgren
242
2a
2a
2b
2b
and
x 2
y 2
y 2
x 2
2 a 0 and 2 b 0
Thus, a and b are harmonic (Note 1g). Since a and b are harmonic and satisfy the Cauchy-Riemann
equations in D, they are harmonic conjugates of each other in D (Note 1h). Therefore, if
f(z)=a(x,y)+ib(x,y) is analytic in a domain D then its component functions, a and b, are harmonic
conjugates in D.
3.
Any analytic function, f(z)=(x,y)+i(x,y), is a valid 2D, incompressible, irrotational flow field.
Proof:
Recall that for an irrotational flow, the velocity may be written as the gradient of a potential function,
:
u 0
from a vector identity: 0 for any
u
For the flow to satisfy the continuity equation for an incompressible fluid:
u 0 2 0
is a harmonic function!
Also recall that the stream function, , is defined for 2D flows such that continuity for an
incompressible fluid is automatically satisfied:
ux
and u y
y
x
2 2
0
xy xy
continuity is automatically satisfied!
If the flow is also irrotational, the stream function must also satisfy:
u y u x
0
u 0
x
y
so that u
2 2
2 0
x 2 y 2
is a harmonic function!
In addition,

ux
x y
uy
y
x
Cauchy-Riemann equations
and are harmonic conjugates!
From Note 2, the components of any analytic function are harmonic conjugates. Thus, since the
governing equations for the fluid are Laplaces equation and since and are harmonic conjugates,
then any analytic f(z)=+iwill be a valid flow field.
Thus, by choosing various forms of f(z) that are analytic, we can produce various valid (incompressible
and irrotational) flow fields. Whether or not the flow fields are interesting from an engineering
perspective is another matter.
C. Wassgren
243
Notes:
4.
Some building block flows and their complex potentials:
uniform stream
f ( z ) U 0 iV0 z
V0
y
U 0 x V0 y
V0 x U 0 y
U0
shown for U0,V0> 0
line source (m>0) or sink (m<0)
z0
r
f ( z)
m
log z z0
2
m
m
ln r
2
m

i
log z z0
2

ln r
2
shown for m> 0
free line vortex
z0
f ( z)
shown for > 0
line doublet (x-orientation)
f ( z)
z0
r
c
z z0
c cos
r
c sin

r
shown for c > 0
line doublet (y-orientation)
z0
r
f ( z)
ic
z z0
c sin
r
c cos

r
shown for c > 0
Note that in the table above:

z=x+iy and z0=x0+iy0
0 < 2
1/ 2
2
2
r x x0 y y0
C. Wassgren
y y0
and tan 1
x x0
244
Notes:
5.
Fluid velocities are found via differentiation of the complex potential:

df
f ( z )
u x iu y
dz
f df z
where f z x, y i x, y
x dz x
f
i
u x iu y
and
x x
x
z
1
and
since z x iy
x
df
f df z
u x iu y (an identical result occurs if we consider
)
dz
y dz y
A rectangular coordinates example:
c
f z
(line doublet oriented in the x-direction and centered at the origin)
z
c x 2 2 xyi y 2 c x 2 y 2
df
c
cz 2
2cxy
2
i
2
2
2
2
2
2
2
2
dz
z
zz
x y
x y x2 y 2
ux
c x 2 y 2
2 2
and u y
2cxy
y2
A polar coordinate example:

m
log( z )
(source/sink at origin)
f ( z)
2
1
df
m 1 m
m exp(i )
m
cos i sin
2 r
dz 2 z 2 r exp(i ) 2
r
m
m
cos and u y
sin
ux
2 r
2 r
or in polar coordinates using some geometry and trig.:
usin
u
tan-1(u/ur)
2
x
2
y
C. Wassgren
-1
245
ur
tan-1(uy/ux)
uy
u tan
uy
u y ur
-1 u
u u u u and tan tan

ur
ux
u x 1 u tan
ur
Substituting in for our values of ux and uy and simplifying:
u tan
2
u
m
u
r
2
2
1 tan 2 0
ur u
and tan
ur
u
1 tan
u
r
m
u 0 and ur
2 r
2
r
ucos
ux
urcos
ursin
In general the relation between the velocity components expressed in rectangular and polar coordinates
is given by (refer to the figure shown above):
u x ur cos u sin
u y ur sin u cos
df
u x iu y ur cos u sin i ur sin u cos
dz
ur cos i sin iu cos i sin
ur iu cos i sin
6.
df
u x iu y ur iu exp i
dz
(5.16)
We can use the principle of superposition to combine complex potentials and form new complex
potentials since if two functions are analytic in a domain D, then their sum is also analytic.
C. Wassgren
246
Notes:
7.
A few example flows created by superposition:
Flow over a Rankine half-body:

Combine the complex potentials for a uniform stream and a source (m>0):
m
f ( z ) Uz
log z
2
y
x
source
Flow over a Rankine oval:

Combine the complex potentials for a uniform stream, a source, and a sink:
m
m
f ( z ) Uz
log( z a )
log( z a)
2
2
where m>0 and a > 0.
y
x
sink
source
C. Wassgren
247
Flow around a non-rotating cylinder of radius R:

Combine the complex potentials for a uniform stream and a doublet:
c
f ( z ) Uz
z
where the constant c is found by not allowing any flow through the cylinder walls, i.e.
ur r R 0
c cos
c sin
f ( z ) Ur cos
i Ur sin r
r
c cos
1
U cos
r2
r r
c cos
ur r R 0 U cos
c UR 2
R2
so that the complex potential becomes:
R2
f ( z) U z
ur
(5.17)
y
x
doublet
Notes:
1.
Real (viscous) flow over a sphere (a golf ball in the figure below) is shown below. The
streamlines for flow over a cylinder look much the same.
The streamlines over the front half of a cylinder are similar to those predicted by the potential flow
analysis. In fact, the velocity and pressures field on the front half of the cylinder are also accurate
(the pressures will be discussed in a moment.) The flow field downstream of the cylinder is not
accurately predicted. The discrepancy between the potential flow analysis and real life occurs due
to the formation of a viscous boundary layer on the cylinder surface. The boundary layer
separates near the top/bottom points of the cylinder and forms a wake. Assuming irrotational flow
in the boundary layer and wake are poor assumptions. However, outside the boundary layer and
wake, the potential flow assumption is reasonable. Well discuss boundary layers in a later section
of notes.
C. Wassgren
248
2.
The pressure distribution on the cylinder surface can be predicted using Eqn. (5.17) and
Bernoullis equation:
R2
f ( z) U z
From Eqn. (5.16), the flow velocity field is:
ur iu exp i
R2
df
U 1 2
dz
z
R2
U
1
r exp 2i
R2
R2
U 1 2 exp 2i U exp i 2 exp i exp i
r
r
R2
U cos i sin 2 cos i sin exp i
r
2
2
R
R
U 1 2 cos i 1 2 sin exp i
r
r
(Note that for this case it would be easier to determine the velocity using or directly rather
than the complex potential). On the cylinder surface (r = R):
ur r R 0
u
r R
2U sin
The pressure distribution on the cylinder surface is found via Bernoullis equation, and expressed
in terms of a dimensionless pressure coefficient, cp:
p p
c p 1s 2 1 4sin 2
(5.18)
2 U
Notes:
a. The total drag (FD) and lift (FL) on the cylinder may be found by integrating the pressure
distribution over the entire cylinder surface:
FL
p
2
FD
p cos Rd
0
2
FL
FD
(5.19)
p sin Rd
Either by actually evaluating Eqn. (5.19) or noting that the velocity field is symmetric over
the front and back and upper and lower surfaces, the drag and lift forces on the cylinder are
both zero. Of course in real flows we know that the drag is not zero. The fact that the
potential flow model predicts zero drag while real flows have non-zero drag is known as
dAlemberts Paradox. We, of course, now know that the discrepancies are explained by the
formation of a boundary layer and boundary layer separation. dAlemberts paradox will be
discussed again when reviewing Blasius integral law.
C. Wassgren
249
b.
Equation (5.18) is compared to experimental data in the plot below (from Fox, R.W. and
McDonald, A.T., Introduction to Fluid Mechanics, 5th ed., Wiley.)
Again, the potential flow analysis predicts the pressure distribution reasonably well over the
upstream part of the cylinder but does a poor job over the back half due to boundary layer
separation.
Flow around a rotating cylinder of radius R:

Combine the complex potentials for a uniform stream, a doublet, and a free vortex:
c i
f ( z ) Uz
log z
z 2
where the constant c is found by not allowing any flow through the cylinder walls, just as in the
previous example. Note that the addition of a vortex will not change the value of c since a vortex
only produces tangential flow and not radial flow. As a result, the complex potential becomes:
R 2 i
f ( z) U z
log z
z 2
FL
y
FD
doublet and
free vortex
C. Wassgren
250
Notes:
1.
The corresponding velocity field is:

R 2 i
df
ur iu exp i U 1 2
dz
z 2 z
2
R
i
exp i
U 1 2 exp 2i
r
2 r
Using the previous results for a non-rotating cylinder:
R2
ur U 1 2 cos
r
(5.20)
R2
1
u U 1 2 sin
2 r
r
On the cylinder surface (r = R):

ur r R 0
1
2 R
Using Bernoullis equation, the pressure coefficient over the surface is:
2
1
1
c p 1 4sin 2 4
sin
2 UR
2 UR
The corresponding drag, FD, and lift, FL, are:
FD 0
u
r R
2U sin
FL U
(5.21)
(5.22)
(5.23)
Notes:
a. The drag again is zero and is not unexpected due to the fore/aft symmetry of the velocity
field.
b. The lift is non-zero and is related to the flow circulation. This type of lift is referred to as
Magnus lift. Both drag and lift for potential flows will be discussed in detail when
reviewing Blasius integral law and the Kutta-Joukowski theorem.
c. The photo below shows the flow past a rotating golf ball. The flow is from left to right
and the golf ball rotates in a clockwise manner ( < 0). From Eqn. (5.23), the lift on the
golf ball will be in the positive vertical direction.
In real (i.e. viscous) flows the lift on a rotating object comes primarily from deflection of
the downstream wake (the fluid momentum is directed downward resulting in an upward
force on the ball) rather than from the unbalanced pressure distribution on the object.
The Magnus effect is often mistakenly referred to as the primary source of the lift force.
C. Wassgren
251
Notes:
7.
Flow in and around corners of varying angles can be modeled using the following complex potential:
f ( z ) Az n where A and n are constants
a.
This produces flows between boundaries intersecting at an angle /n (only flows with n1/2 are of
interest):
n=2
n=3
n=1
x
or
or
y
x
y
y
x
x
n=2/3
n=1/2
b.
The potential and stream functions are given by:

f ( z ) Az n A r exp i Ar n exp in Ar n cos n iAr n sin n
Ar n cos n and Ar n sin n
n
ur
c.
Anr n 1 cos n and

r
1
Anr n 1 sin n
r
The fluid speed at the origin is:

lim u lim f ( z ) lim nAr n 1 exp i (n 1) lim nAr n 1
r 0
r 0
lim u A
r 0
C. Wassgren
r 0
r 0
n 1
n 1
n 1
252
d.
In a real fluid (one with viscosity), the flow along the surface streamline would:
for n>1:
separate before reaching the corner and produce a standing eddy
for n<1:
separate after reaching the corner unless the corner angle is small
C. Wassgren
253
4.
Blasius Integral Law
Consider the 2D, incompressible, inviscid, steady, irrotational flow around an arbitrary closed body:
Coutside
y
Cinside
x
D
uy
dy
dx
ux
p
Using the LMEs, determine the lift, L, and drag, D, acting on the body:
D
pdy
Coutside
Coutside
pdx
Coutside
Coutside
u x u x dy u y dx
u dA
u y u x dy u y dx
u dA
From Bernoullis equation (neglecting gravity):

p 1 2 u x2 u y2 c p c 1 2 u x2 u y2
where c is a constant. Substituting and re-arranging:
D cdy 1 2 u x2 u y2 dy u x2 dy u x u y dx
Coutside
Coutside
cdx 1 2 u x2 u y2 dx u x u y dy u y2 dx
Noting that:
cdy
Coutside
cdx 0
Coutside
and simplifying:
D 1 2 u x2 u y2 dy u x u y dx
Coutside
Coutside
1 2 u x2 u y2 dx u x u y dy
As shown below, the previous drag and lift relations may be written in terms of the complex potential.
C. Wassgren
254
df
dz dz i 2 C ux iu y dx idy
2 Coutside
outside
i
i
2 Coutside
2
x
2iu x u y u y2 dx idy
2
2
2
2
ux dx u y dx 2ux u y dy i ux dy u y dy 2ux u y dx
2 Coutside
1 2 u 2 u 2 dy u u dx i 1 2 u 2 u 2 dx u u dy
x y
x
y
x y
x
y

Coutside
D
L
Substituting the expressions for lift, L, and drag, D, found previously:

i
df
dz dz D iL
2 Coutside
BLASIUS INTEGRAL LAW
How is this result used? Typically, it is applied using a theorem from complex variables referred to as the
Residue Theorem (Churchill, R.V. and Brown, J.W., Complex Variables and Applications, McGraw-Hill,
5th ed., pg. 169) which states:
n
w( z )dz 2 i Res w( z )
C
k 1
z zk
A residue is the coefficient in front of the 1/(z-z0) term (the b1 term in the series below) in the Laurent series
expansion of an analytic complex function about the point z0 (Churchill and Brown, pg. 144):
bn
n
w( z ) an z z0
n
n 0
n 1 z z0
where the coefficients an and bn are given by
1
w( z )dz
1
w( z )dz
an
and bn
n 1
2 i C z z0
2 i C z z0 n 1
The details of the expressions above wont concern us here and are only presented for completeness.
Blasius Integral Law is used in deriving the Kutta-Joukowski theorem given in the following section
which relates the lift (and drag) around any arbitrary, closed object to the circulation, , caused by the
object.
C. Wassgren
255
Example:
Determine, using the Blasius Integral Law, the lift acting on a rotating cylinder.
SOLUTION:
The complex potential function for flow around a rotating cylinder is:
R 2 i 1
df
R 2 i
f z U z
log z
U 1 2
dz
z 2 z
z 2
(5.24)
R 2 i 1 R 2 2 1
df
2
U
1
U 1 2 2 2

z2 z
z 4 z
dz
Let the contour in the integral law, C, be the circle defined as:
z R exp i dz iR exp i d
(5.25)
(5.26)
where 0 < 2and R is an arbitrary radius greater than R. Thus, the Blasius Integral Law for this
problem is:
i
2
R2
1
R2
1
2
U
1
0 R2 exp 2i R exp i U 1 R2 exp 2i 4 2 R2 exp 2i iR exp i d

2
2
4
2
i R
2 exp i
R
R
U 1 2 exp 2i 2
R U 2 exp i 1 2 2 exp 2i 4 exp 4i
d
R2
2 0
R
R
R R
4
df
dz dz i 2
2
C
R
U 2
R
df
dz i U D 0 and L U
2 C dz
These are the same results that we found previously (Eqn. (5.23))!
D iL i
C. Wassgren
256
(5.27)
5.
Kutta-Joukowski Theorem
Now consider the flow around an arbitrary closed body (centered at the origin) in a uniform stream of
horizontal velocity, U. Far from the body (z) the complex potential will be of the following form (a
Laurent series expansion):
bn
m i
f ( z ) Uz
(5.28)
log z n
n 1 z
Note that the coefficients, an, for the terms involving zn (n 2) in the Laurent series (refer to the previous
set of notes on the Blasius integral law) are all zero since we are considering external flows (recall that the
velocity field is given by df/dz so that terms involving zn where n 2 will approach as z).
Furthermore, since we are concerned only with closed bodies, the net source term, m, should also be zero.
We, however, will continue to include the source term until the end of this analysis.
Given the complex potential above, lets apply Blasius Integral Law to determine the lift and drag about an
arbitrary object:
df
m i 1
n 1
U
nbn z
dz
2 z n 1
2
df
m i U m i 1
1
2
U 2

2 O 3
dz
2 z 2 z
z
Using the Residue Theorem to evaluate Blasius Integral Law:

D iL i
df
df
m i
dz dz i 2 2 i Res
i 2 i 2
U
z 0
dz
2 Coutside
2

2
mU i U
Thus, we see that for a closed object (m=0):

D 0 and L U
KUTTA-JOUKOWSKI THEOREM
For an object that is not closed (e.g. a Rankine half-body), we have:
D mU and L U
(5.29)
L
Notes:
1.
The result given above indicates that there is no drag around an arbitrary closed object in an a steady,
incompressible, irrotational, inviscid flow. In real life of course there is always some drag on an
object. The conflict between the derived value of zero drag and the real-life value of non-zero drag is
referred to as dAlemberts Paradox. There is no paradox, in fact, if one realizes that it is viscous
effects (skin friction drag and form, aka pressure, drag resulting from the formation of a wake (which
in turn is a result of boundary layer separation)) which produces drag on an object.
2.
Bodies of semi-infinite extent (e.g. a Rankine half-body) do have drag on them due to the fact that the
net source term, m, is not zero. The drag is a result of a non-zero flux of horizontal momentum out
through the control surface.
3.
The Kutta-Joukowski theorem states that the lift on an object is directly proportional to the net
circulation, , caused by the object. This is an important observation that is especially useful in
aerodynamics when calculating the lift on an airfoil. As will be shown later, the circulation around an
airfoil is dependent on the free stream velocity so that the lift turns out to be proportional to the
circulation squared.
C. Wassgren
257
6.
Conformal Mappings
Conformal maps are analytic functions that transform curves (e.g. equi-potential lines and streamlines) in one
complex plane, call this the z-plane, to similar curves, but expanded or contracted and rotated at each point, in a
different complex plane, call this the -plane.
Conformal maps are useful because they allow us to use the complex potential for a straightforward flow (e.g. flow
around a rotating cylinder), after the proper mapping, as the complex potential for a more complex flow (e.g. flow
around an airfoil). The complexity comes into play when trying to find the proper mapping that will give us the
desired transformation.
f(z) where z=x+iy
C. Wassgren
mapping function:
=F(z)
258
f() where =+i
First, lets examine some properties of a conformal map. Let =+i be an analytic function of z=x+iy given by:
=F(z)
Because the transforming function, F, is analytic, there is a connection between curves in the z-plane and
corresponding curves in the -plane.
Proof:
Determine the derivative of the function, =F(z), at a point, z, by approaching the point from two different
directions:
z+z
y
z+z
where:
F z z
F z z
The length ratios in the z and planes are:
z

and
z

and the angles separating the lines are:
arg( ) arg( )
arg( z ) arg( z )
and

z
arg
arg
z

Also, because an analytic function has a unique derivative:
d
2
z O z
dz
d
2
z O z
dz
So that the length ratios and angles between the lines are the same in each plane:

z
z
and arg
arg
as z , z 0
Thus, lengths in the neighborhood of z are stretched by a scale factor, d dz , and are rotated by an angle,
arg d dz , into the -plane.
C. Wassgren
259
Notes:
1. Curves of small linear dimension in the z-plane are mapped into curves of similar shape, but expanded or
contracted and rotated, in the -plane.
2.
3.
4.
5.
Note that a large region may be transformed into a region that bears no resemblance to the original one since the
scale factor and angle of rotation vary, in general, from point to point.
d
At singular points of the map, i.e. points where
0 or , the mapping is not conformal.
dz
Since lines of constant and are in the z-plane, they will also be in the -plane, except at singular points.
(Refer to Note 1.)
Since an analytic function of another analytic function is also analytic (Churchill and Brown, p. 56), we are
assured that the function resulting from the conformal map of a complex potential will also be a valid complex
potential (it will be a 2D, incompressible, irrotational flow).
6. Velocities in the -plane are proportional to the velocities in the z-plane by the inverse of the scale factor:
df
df dz
dz
u x iu y
u iu
d dz d
d
7.
Singularities such as vortices and sources/sinks in the z-plane map to identical singularities in the -plane. This
can be seen by considering the flow in a neighborhood of the singularity as the neighborhood shrinks to an
infinitesimally small radius.
m u dA u x dy u y dx
C
u ds u x dx u y dy
C
Consider the integral of the complex velocity around the contour C:

df
dz dz ux iu y dx idy ux dx u y dy i ux dy u y dx
C
C
C
C
df
dz im

dz
C
Thus,
z imz
Cz
df
df d
df
dz
dz
d im
dz
C z d dz
C d
Source/sink and free vortex singularities in the z-plane map to similar singularities in the -plane.
8.
Doublet singularities in the z-plane map to doublet singularities in the -plane but with the strength changed in
d
d
magnitude by
and an orientation rotated by arg
(recall that a doublet is formed by bringing a
dz
dz
source and sink of equal strength infinitesimally close to each other while keeping the product m/a constant
where a is the separation distance between the source and sink).
9.
Even though the occurrence of boundary layer separation in the real flow may limit the usefulness of a potential
flow model, transformation of the flow field to a different flow field may produce a realistic flow. For example,
although in a real flow boundary layer separation occurs for flow over a rotating cylinder, the mapping to an
airfoil shape appears realistic.
10. Conformal maps are another tool we can use to produce realistic-looking flows using potential functions.
C. Wassgren
260
7.
Joukowski Transformation
An example of a particular conformal mapping is the Joukowski transformation:

c2
z
z
where c. This mapping will transform flow around a rotating cylinder
in the z-plane to flow around an airfoil in the -plane. This airfoil is
referred to as a Joukowski airfoil.
(5.30)
y
Consider a circle of radius R centered at the point z0 such that the

circumference of the circle passes through the point z=c:
z0 c Rexp i
z0
The points defining the circle are given by:

z z0 Rexp i c Rexp i Rexp i
If we map the points of the circle in the z-plane to the -plane using the transformation given in Eqn. (5.30),
the resulting figure looks like an airfoil:
c =1.0, R /c =1.1, =5.0 deg
3.0
2.0
y or
1.0
0.0
-3
-2
-1
-1.0
-2.0
-3.0
x or
C. Wassgren
261
Notes:
1.
The geometry of the Joukowski airfoil is determined by the quantities R/c and. The camber of the
airfoil is proportional to (camber as ). The thickness of the airfoil increases as (R/c) increases.
The chord length of the airfoil is approximately equal to 4c (the exact chord length will also depend
on the airfoil thickness).
camber
thickness
mean camber line
leading edge (LE)
trailing edge (TE)

chord
An airfoil with no camber (=0):
c =1.0, R /c =1.1, =0 deg

1.0
0.5
0.0
-3
-2
-1
-0.5
-1.0
An airfoil with no camber (=0) and larger thickness (R/c=2.0):
c =1.0, R /c =2.0, =0 deg

2.0
1.0
0.0
-4
-2
-1.0
-2.0
C. Wassgren
262
2.
The trailing edge of the Joukowski airfoil will be cusp-shaped. Real airfoils typically end in a finite
angle.
angle > 0
2
3.
Joukowski airfoils are not commonly used in practice; however, they provide a good model for
predicting the general behavior of airfoils at small angles of attack (so that boundary layer separation
wont occur in the real-world airfoils).
4.
Note that the trailing edge of the airfoil corresponds to the location where the cylinder intersects the
location z=c. The transformation is not conformal at z=c and thus the angle between intersecting lines
in the -plane is not necessarily the same angle between intersecting lines in the z-plane at the point
z=c.
C. Wassgren
263
Now consider, in the z-plane, a uniform flow with velocity U around a rotating cylinder with circulation
(an unknown value at this point) and radius R. The complex potential for this flow is given by:
R 2 i
z
f z U z
log
(5.31)
z 2
R
Lets rotate the flow so that the incoming stream is at an angle of attack, , with respect to the horizontal:
z z exp i z z exp i
Lets also translate the origin of the cylinder so that it is centered at the position z0=c-Rexp(-i):
z z z0 z z z0
The new complex potential is given by
i
z z0
R2
f z U z z0 exp i
exp i
log
exp i
z z0
2
R
(5.32)
Note that we havent yet determined the value of the circulation, . This will be found in the section
below. First, however, lets plot some streamlines for the case with zero circulation (=0):
Of particular interest in the plot is the condition at the trailing edge of the airfoil. The streamlines at the
trailing edge make a very sharp turn (the streamlines are not smooth at the very tip of the airfoil) resulting
in infinite fluid accelerations and velocities at the trailing edge. This is not a very realistic flow and does
not match what we observe in flows around real airfoils. We can avoid this infinite velocity problem by
adjusting the circulation around the airfoil so that the flow leaves smoothly from the trailing edge. This is
equivalent to moving the rear stagnation point to the tip of the trailing edge. This adjustment is referred to
as the Kutta Condition.
C. Wassgren
264
To quantitatively determine what the value of the circulation must be to satisfy the Kutta Condition, lets
examine the complex velocity of fluid along the airfoil surface:
df
df dz dz dz
u iu
d dz dz dz d
where
df
R2
i
U 1
2
dz
z
2
z
dz
exp i
dz
dz
1
dz
2
dz c
1
d z
z z z0 exp i
z0 c Rexp i
so that the complex velocity in the -plane is given by:

1
c 2
df
R2
i
U 1
u iu
exp i 1
d z 2 2 z
z
(5.33)
Note that at z=c the magnitude of the complex velocity approaches infinity due to the (dz/d term. To
prevent infinite velocities from occurring, the term within the curly brackets {} must equal zero at z=c:

R2
i
U 1
2
2 z
z
z c
R2
R2
2 iUz 1
2
exp
iU
R
i
2
Rexp i
z
z c
where z z c c z0 exp i c c Rexp i exp i Rexp i
2 iUR cos i sin cos i sin
4 UR sin
Thus, to prevent infinite velocities from occurring at the trailing edge of the airfoil, the circulation must be
given by:
4 UR sin
(5.34)
Now that we know the circulation about the airfoil, we can use the Kutta-Joukowski Theorem to determine
the lift of the airfoil:
L U
(5.35)
L 4U 2 R sin
C. Wassgren
265
The lift is often presented in dimensionless form as the lift coefficient, cL:
L
R
cL
2 sin
1 U 2 4c

c
2
where 4c is the approximate chord length of the airfoil.
(5.36)
We can also determine the pressure distribution on the airfoil surface by using Bernoullis equation (recall
that were dealing with an incompressible, irrotational flow so the same Bernoulli constant is used
everywhere):
ps 1 2 us2 p 1 2 U 2
where ps and us are the pressure and speed on the airfoil surface, and p and U are the pressure and speed
far from the airfoil.
The magnitude of the velocity on the surface of the airfoil can be found using the complex velocity given in
Eqn. (5.33) with:
z z z0 exp i Rexp i exp i Rexp i
where defines the location on the airfoil surface. After some algebra, we arrive at:
2U sin sin
us
2
c
1
z
(5.37)
with z=c-Rexp(-i)+Rexp(i).
The pressure is often expressed non-dimensionally as the pressure coefficient, cP:
cP
p p
u
1 s
1 U 2
U
2
(5.38)
Notes:
1.
The lift coefficient predicted by our potential flow analysis of a Joukowski airfoil is reasonably close
to values found experimentally at small angles of attack and small camber (to avoid boundary layer
separation).
2.
Flow over a flat plate can be found by letting R/c=1 and =0. The resulting lift coefficient is:
cL 2 sin
The Joukowski transformation can also produce flow around curved plates (let R/c=0 and >0).
3.
Note that increasing the angle of attack, , the camber, , and the thickness, R/c, all act to increase the
lift of an airfoil.
C. Wassgren
266
4.
The Joukowski transformation can also produce flow around ellipses. To produce this type of flow, we
center the cylinder at the origin and choose R>c. The points on the cylinder surface are given by
z Rexp i
c =1.0, R =2.0
2.0
1.0
0.0
-3
-2
-1
-1.0
-2.0
where 0<2
C. Wassgren
267
8.
Method of Images
In much of our previous work concerning potential flows, we investigated external flows in an infinite
expanse of fluid. Since there are a number of phenomena that are of interest when there is flow near a
boundary, we should find a method for modeling flows near walls. The Method of Images is such a
method.
Consider how we can model the flow from a source near a wall. To produce a horizontal streamline
representing the wall, we can add to our original source, an image source an equal distance away from
where we want our wall to be.
m
ln r1 ln r2
y
2
y
2
where r1 x 2 y a
r1
r2 x 2 y a
a
x
r
r2
Notes:
1.
There is a net upward velocity at the location of the original source of:
m 1
V
2 2a
where m is the source strength due to the flow induced by the image source.
2.
The vertical force acting on the source can be determined by first calculating the pressure force acting
on the wall using Bernoullis equation and then noting that the force acting on the wall is equal, but in
the opposite direction, to the force acting on the source. The pressure at the wall is given by:
pw 1 2 Vw2 p
where pw and Vw are the pressure and velocity magnitude at the wall and p is the pressure far from the
wall (U approaches zero as we move far from the wall). The velocity along the wall is found from the
potential function:
m
2
2
ln x 2 y a ln x 2 y a
2
ux
uy
y 0
y 0
y 0
y 0
x2 a2
0
2
x
m
pw p 1 2
2
2
x a
Note that without the sources, the pressure acting on the wall would be p. Hence, the increase in the
force acting on the wall is:
C. Wassgren
268
(x, y)
Fon wall
due to source
pw p dx
x
m
1
dx
2
2
2
x a
x
1 m x
1
x
2
tan 1
2
a
4 x a
a
Fon wall
due to source
m2
4 a
Consequently, the force on the source will be:

m2
Fon source
4 a
A particularly interesting application of the method of images is investigating the effect that the ground has
on the lift of an airfoil. Lets use a crude model consisting of a free vortex combined with a uniform stream
to investigate this effect. Recall that in order to satisfy the Kutta condition, an airfoil must have some
circulation which in turn develops lift (from the Kutta-Joukowski Theorem). The potential flow model is
given below (drawn for >0).
y
Ux 1 2
2
y
r1
ya
where 1 tan 1
U
x
r
a
r2
y
a
2 tan 1
a
x
a
x
a
As before with the source example, the force acting on the source will have the same magnitude, but with
opposite sign, as the force acting on the wall. The force acting on the wall is found by integrating the
pressure force over the entire wall:
x
U a
U
ux y 0
2
x y 0
2
x2 a2
a
1

x
uy
y 0
0
y 0
2 aU
a
pw p 1 2 U 2 1 2 U 2
x 2 a 2 x 2 a 2
The resulting force acting on the wall due to the vortices (again, subtracting out the pressure when no
vortices are present) is then:
C. Wassgren
269
(x, y)
Fon wall
due to vortex
pw p dx
2
1 2 aU
a
dx
2
2
2
2
x a x a
x
2
2U
1
x 1 x
x
12
tan 1
tan 1 2
2
a
a 2 x a
a

2 2U
2a
U 1
aU
Thus, the force acting on the vortex is:

Fon vortex U 1
aU
Fon wall
due to vortex
Recall that the lift force acting on an object in an infinite expanse of fluid with circulation is given by the
Kutta-Joukowski theorem as:
Linfinite U
expanse
Keep in mind that from the previous discussion regarding Joukowski airfoils, the circulation around an
airfoil is negative ( < 0) in order to satisfy the Kutta condition at the trailing edge of the airfoil.
The difference between the lift generated when the wall is nearby versus the lift in an infinite expanse of
fluid is:

L Lwall Linfinite U 1
U
aU
4
present
expanse
4 a
Hence, the wall acts to decrease the lift. Experience shows, however, that an airfoil near the ground, aka in
ground effect, actually has increased lift (and decreased drag) rather than decreased lift. Why do we have
this discrepancy? Its because our analysis considers an infinitely long airfoil, i.e. one with no wing tips.
At the end of a finite wing, trailing vortices (as opposed to the vortex bound to the airfoil resulting
from the Kutta condition) are generated by the wing tips as shown below. These wingtip vortices occur
because air in the high pressure region underneath the airfoil is pushed around the wing tips to the low
pressure above the airfoil.
U
bound vortex
trailing vortices
starting vortex
(This will be discussed later.)
When viewed from behind, the trailing vortices appear as shown below.
wingtip vortex
wingtip vortex
low pressure side of wing
high pressure side of wing
C. Wassgren
270
The wingtip vortices induce a downwash along the wing surface and thus reduce the effective angle of
attack that the airfoil sees.
Dind
ind
wind
eff
ind
eff
wing
The lift acting on the wing per unit span (i.e. distance into the page), L, will be the lift calculated for the
local effective angle of attack, eff, where:
w
eff ind with tan ind ind
U
Here, is the nominal angle of attack and ind is the induced angle of attack resulting from the wingtip
vortices which induce a local downwash velocity of wind. Since the effective angle of attack is reduced, the
lift on the airfoil will also be reduced.
There is also an induced drag on the wing Dind since the local flow is at an angle of ind from the free stream
which tilts the actual lift vector slightly downstream
Dind L tan ind
Again, Dind is the induced drag per unit span of the wing. Note that the induced drag on the wing is not due
to viscous effects, but is due solely to the induced angle of attack resulting from the induced downwash.
Hence, there is a drag on a finite wing, even in an ideal flow, due to the trailing vortices.
The trailing vortices also drift downwards over time due to the flow induced at the center of each vortex by
the other vortex. If the airfoil was turned upside down, the vortex orientation would be reversed and the
vortices would drift upwards over time!
b
Vdownward
Vdownward
If this vortex has a circulation of , the

induced velocity at the center of the other
vortex is:
1
Vdownward
2 b
where b is the distance between the wingtips.
Near the ground, two image vortices must also be included in the analysis in order to make the horizontal
ground streamline. As a result of the flow induced by the image vortices, the wingtip vortices drift outward
when reaching the ground.
downwash induced by original vortices
original vortices
upwash induced by image vortices
image vortices
C. Wassgren
271
The image vortices also contribute an upwash along the wing which helps to counteract the downwash
caused by the original vortices. This reduction in the downwash helps increase the lift and reduce the drag
on the finite wing when it is located near the ground. This is the source of the observed ground effect.
Notes:
1.
Recall that from previous discussions regarding vorticity, vortex lines must either form closed loops or
terminate on a boundary ( = 0). So how then do the vortex lines corresponding to the trailing
vortices terminate? The bound vortex/trailing vortex lines actually form a closed loop through a
starting vortex line (shown in a previous figure). The starting vortex occurs during the transient
when the lift on the airfoil changes (e.g. at start up). Since the starting vortex is typically located far
behind the bound vortex, its effects are typically neglected in steady airfoil analyses and the bound
vortex/trailing vortex combination is treated as a horseshoe vortex.
2.
The trailing vortices are not actually concentrated solely at the wingtips. Instead, there is a distribution
of trailing vortices along the wing due to variations in the circulation which result from changes in
airfoil geometry and local flow conditions. These variations must be included a finite wing analyses
(see, for example, Kuethe, A.M. and Chow, C-Y., Foundations of Aerodynamics, Wiley.)
bound vortex
trailing vortices
3.
Trailing vortices have been the source of several airline disasters (see, for example,
http://www.asy.faa.gov/safety_products/wake.htm and http://aviation-safety.net/events/EFV.shtml). If
an aircraft flies behind a preceding aircraft too closely, it can be caught in the trailing vortices and
cause the pilot to lose control of the aircraft (this phenomenon is sometimes mistakenly referred to as
wake turbulence). The strength of the vortices is proportional to the lift generated by the airfoil
which in turn is related to the weight of the aircraft. Hence, the spacing between aircraft (near an
airport for example) is a function of their relative size.
4.
Ground effect has been used as a significant component in the design of several aircraft. The
following web site lists a number of WIG (wing-in-ground-effect) aircraft: http://www.setechnology.com/wig/index.php).
Even pelicans take advantage of ground effect!
C. Wassgren
272
Notes:
1.
We may sometimes need an infinite number of reflections to properly model a flow. Consider for
example, the flow resulting from a source located midway between two walls. An infinite number of
images are required for perfect symmetry.
C. Wassgren
273
Example:
a.
b.
Write the potential function that simulates the flow of a line source placed asymmetrically between
two parallel walls as shown in the figure.
Compute the dimensionless velocity, u = au/(4m), on the lower wall at (x/a, y/a) = (1,0) accurate
to three significant digits.
y
2a
m
SOLUTION:
Use the Method of Images to create the given flow. The sequence of images is shown below.
repeat reflecting images to
a
a
y
2a
2a
m
a
a
2a
2a
repeat reflecting images to

The potential function for the given flow field is:
2
2
2
2
2
2
2
2
m ln x y a ln x y a ln x y 5a ln x y 5a
(5.39)
2
2
2
2
2
2
2
2
2
ln x y 7a ln x y 7a ln x y 11a ln x y 11a
C. Wassgren
274
ln x 2 y 6k 1 a 2 ln x 2 y 6k 1 a 2
2
2
2
2
k 0 ln x y 6k 5 a ln x y 6k 5 a

or, in dimensionless terms:
2
2
2
2
k ln x
y 6k 1 ln x y 6k 1

2
2
2
2
2
k 0 ln x y 6k 5 ln x y 6k 5 4 ln a
(5.40)
(5.41)
where the dimensionless potential function is = /(4m) and the dimensionless positions are x = x/a and
y = y/a.
The dimensionless velocities resulting from this potential function are:
1
1
2
2
2
k x 2 y 6k 1
x y 6k 1
2 x
u x
1
1
x
k 0
x 2 y 6 k 5 2 x 2 y 6 k 5 2
y 6k 1
y 6k 1
2
2
2
2
k x y 6k 1
x y 6k 1

u y
y k 0
y 6k 5
y 6k 5
2
2
2
2
x
y
k
x
y
k
6
5
6
5
where ux = aux/(4m) and uy = auy/(4m).
The velocity at (x, y) = (1, 0) is:

k
1
1
u x 1, 0 4
2
2
1 6k 1 1 6k 5
k 0
(5.43)
(5.44)
u y 1, 0 0 (as expected since the point is on a wall)
C. Wassgren
(5.42)
275
(5.45)
The value of the horizontal velocity component at (x, y) = (1,0) as a function of k is given in the table
below. Note that %diff from prev is the percentage change in the value of ux from the previous value of
ux, i.e. % diff = (ux,k+1 ux,k)/ux,k * 100%.
k
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
u'x
% diff from prev
1.962
2.042
4.08%
2.066
1.15%
2.077
0.53%
2.083
0.31%
2.087
0.20%
2.090
0.14%
2.092
0.10%
2.094
0.08%
2.095
0.06%
2.097
0.05%
2.097
0.04%
2.098
0.04%
2.099
0.03%
2.099
0.03%
2.100
0.02%
2.100
0.02%
2.101
0.02%
2.101
0.02%
2.101
0.01%
Hence, the velocity components at (x,y) = (1,0) are (ux, uy) = (2.10, 0).
C. Wassgren
276
9.
Added Mass
Added mass (aka apparent or virtual mass) is the concept whereby we add an extra mass to an object
when we accelerate the object through a fluid. This added mass term accounts for the force required to
accelerate the surrounding fluid to a higher velocity.
To study this concept, lets consider the potential function for a cylinder in a fluid that is stagnant far from
the cylinder. To form this potential function, we first form the potential function for a uniform stream with
velocity U flowing around a stationary cylinder of radius R.
R2
stationary cylinder in uniform stream Ur cos 1 2
r
To change our frame of reference so that the fluid far from the cylinder is stationary, we add in a uniform
stream of velocity U in the opposite direction:
cylinder moving through stagnant fluid stationary cylinder in uniform stream of velocity U Ur cos
UR 2 cos
r
The resulting potential function describes the flow produced by a cylinder of radius R moving at velocity U
through an otherwise quiescent fluid.
cylinder moving through stagnant fluid
Lets now determine the total kinetic energy in the fluid (outside of the cylinder):
r
KEtotal
2 rdr ur2 u2
r R
where the fluid velocity components are given by:

dA = 2rdr
UR 2 cos
1 UR 2 sin
ur
and u
r
r2
r
r2
U 2 R4
ur2 u2 4
r
Note that the in the kinetic energy formula is the fluid density. Substituting the speed into the expression
for the total kinetic energy:
r
KEtotal
r R
U 2 R 4 U 2 R 4
2 2 rdr
r4
2r 2
R2
rR
U 2
If we apply a force such that it increases the velocity of the cylinder by a small amount U, the total kinetic
energy of the fluid will increase by an amount (neglecting higher order terms):
R2
R2
2
KEtotal
U U
U 2 R 2 U U
2
2
The average force we must apply to the cylinder over time t (the time over which the velocity goes from
velocity U to velocity U+U) to increase the total kinetic of the fluid is:
FU t
KEtotal
total work done in time t
required to increase KE of fluid
R 2 U U
U
R2
U t
t
Thus, the total force required to accelerate a cylinder of mass, M, through a quiescent fluid is given by:
2 dU
F M
R

dt
added
cylinder
mass
mass
C. Wassgren
277
The term added to the cylinder mass in the previous equation is referred to as the added mass (aka
apparent or virtual mass). Thus, the added mass for a cylinder is Madded=R2. Again, the is the fluid
density.
Notes:
1.
Note that added mass is only a factor for unsteady flows. There is no added mass term for a steady
flow.
2.
Added mass terms are typically only significant in flows of liquids since the added mass for gases is
often small compared to the objects mass (gas is typically very small). Added mass terms in a gas can
be significant however if the object is large and has small mass (e.g. a parachute).
3.
We could have also found the force on the cylinder by integrating the pressure force around the
cylinder surface which is found using the unsteady Bernoulli equation (neglecting gravitational
effects):

p 1 2 V 2
p
on surface (r R )
dU
dt
Integrating the pressure force along the surface of the cylinder to find the total force on the cylinder:
2
2
2
2
dU
F ps cos Rd R p cos d 1 2 U 2 cos d R
cos 2 d
dt
0
0
0
0
dU
F R 2
(the same answer as before)
dt
ps p 1 2 U 2 R cos
4.
The added mass is dependent on the shape of the object. Its possible to have different values for the
added mass depending on the orientation of the object (e.g. an ellipse will have different added masses
depending on its orientation.) We can also have added mass effects due to rotational acceleration of an
object.
5.
An additional reference concerning added mass is: Yih, C.S., 1969, Fluid Dynamics, McGraw-Hill
(Now published by West River Press, Ann Arbor, MI).
6.
The added mass presented here was calculated for an inviscid fluid. For unsteady viscous flows an
additional term referred to as the Bassett force also appears which takes into account unsteady viscous
force terms.
C. Wassgren
278
10. Finite Difference Methods

Recall that the governing equation for an incompressible, irrotational flow is:
2 0
(5.46)
where is the velocity potential. Our goal here is to re-write Eqn. (5.46) using a finite difference
approximation so that we can solve the equation numerically. Well assume a 2D flow in Cartesian
coordinates to make the following analyses more straightforward; however, the same ideas can be applied
to 3D and non-Cartesian (but still orthogonal!) coordinate systems. In particular, well solve Eqn. (5.46) at
the grid points shown in the figure below. Note that for simplicity, neighboring grid points are assumed to
be separated by a distance h in both the x- and y-directions. The derivations given below may also be
extended to non-uniform grid spacings.
streamlines
i,j+1
i,j i+1,j
i-1,j
finite difference grid
h
i,j-1
The values of the second order partial derivatives in Eqn. (5.46), e.g. 2i,j/x2, may be written in terms of
the neighboring values of by using Taylor series expansions about the point i,j. For example, for
determining 2i,j/x2, express i+1,j and i-1,j in terms of Taylor series expansions about i,j and then add the
two Taylor series together.
i 1, j i , j
i 1, j i , j
h
1! x
h 2 2
2! x 2
i, j
h
1! x
h 3 3
2! x 2
h 2 2
2! x 2
2
i, j
3! x3
i, j
h 2 2
i, j
i 1, j i 1, j 2i , j 2
i, j
h 4 4
4! x 4
h 3 3
i, j
h 4 4
4! x 4
3! x3
(5.47)
i, j
h 4 4
i, j
4! x 4
(5.48)
i, j
(5.49)
i, j
The previous expression may be re-arranged to solve for the 2nd order derivative.
i 1, j i 1, j 2i , j
h 2 4
2
4! x 4 i , j
h2
x 2 i , j
A similar approach may be used in the y-direction to determine 2i,j/y2.
i , j 1 i , j 1 2i , j
2
h 2 4
4! y 4 i , j
y 2 i , j
h2
(5.50)
(5.51)
Hence, at point (i,j), the solution to Eqn. (5.46) may be written as:
0 2
i, j
2
x
2
i, j
2
y 2
i 1, j i 1, j 2i , j
h
i, j
h 2 4
4! x 4
i , j 1 i , j 1 2i , j
i, j
0 i 1, j i 1, j i , j 1 i , j 1 4i , j 2
C. Wassgren
h
h 4 4
4! x 4
2
i, j
279
h 4 4
4! y 4
h 2 4
4! y 4
i, j
(5.52)
i, j
The previous equation is an exact solution to Eqn. (5.46) at the point (i,j) as long as all of the higher order
terms are included in Eqn. (5.52). If the value of h is sufficiently small, then we may approximate Eqn.
(5.52) by neglecting terms of order h4 and higher since they will be small in comparison to the remaining
terms (as long as the higher order derivatives dont simultaneously become very large). The resulting
truncated equation is now only an approximate solution to Eqn. (5.46).
(5.53)
0 i 1, j i 1, j i , j 1 i , j 1 4i , j
where the truncation error in the previous equation is of order h4.
Notes:
1.
Note that at a vertical solid boundary, the horizontal velocity (ux) is zero. In terms of the velocity
potential:
i,j+1 y
u x i, j 0
h
x i , j
i-1,j i,j
x
h
i,j-1
Determining the potential at (i-1,j) in terms of the Taylor series expansion about point (i,j) gives:
i 1, j i , j
h
1! x i , j
h 2 2
2! x 2
h 3 3
i, j
3! x3
i, j
h 4 4
4! x 4
(5.54)
i, j
Re-arranging this equation gives:

3
2 i 1, j i , j
2
h
O
x3
x 2 i , j
h2
i, j
Combine Eqn. (5.55) with Eqn. (5.51) to solve for 2i,j = 0.
0 2i , j
2
x 2
2 i 1, j i , j
h2
i, j
2
y 2
i, j
i, j 1 i, j 1 2i, j O h 3
O h2
x3
y 4
i, j
h2
3
4
0 i , j 1 i , j 1 2i 1, j 4i , j O h3 3 O h 4 4
x
y
i, j
If the previous equation is truncated, then it becomes:

0 i , j 1 i , j 1 2i 1, j 4i , j
A similar approach can be used at a horizontal boundary to give:

3
4
0 i 1, j i 1, j 2i , j 1 4i , j O h3 3 O h 4 4
x
y
i, j
0 i 1, j i 1, j 2i , j 1 4i , j
C. Wassgren
(5.55)
280
i, j
i, j
(5.56)
(5.57)
i, j
y
(5.58)
h
i-1,j i,j
i,j-1
i+1,j
x
(5.59)
or at a corner:
0 2i 1, j 2i , j 1 4i , j
3
O h3 3
x
O h4
y 4
i, j
0 2i 1, j 2i , j 1 4i , j
i, j
y
i,j+1
0 i+1,j
h
(5.60)
x
(5.61)
2.
Equation (5.46), i.e. Laplaces equation, is an elliptic partial differential equation. In order to have a
well-posed problem, i.e. the equation has a unique solution that depends continuously on the
boundary and/or initial data, the gridded flow domain must be finite and continuous boundary
conditions must be specified along the entire boundary. The boundary conditions may be either
Dirichlet boundary conditions (where the value of is specified), Neumann boundary conditions
(where the gradient of is specified), or a combination of both types of boundary conditions (known
as mixed boundary conditions).
3.
There are two common methods to solving the resulting finite difference approximations to Eqn.
(5.46) at every point in the flow domain. Non-iterative, or direct, methods solve the equations
directly (in one step) while iterative methods solve the equations after repeated calculations that
(hopefully) converge on the answer. Examples of each of these methods are given in the following
discussions using the simple flow field and grid shown below.
4
3
2
1
= 0
uniform grid spacing
= 1
1 2 3 4
i
Before numerically solving for the values of at each of the grid points, we can easily observe that
for the uniform grid spacing shown, we anticipate that values for in row 2 (i,j = 2) should be 2/3
and the values for in row 3 (i, j = 3) should be 1/3.
C. Wassgren
281
For the given example, write the finite difference equations for each point on the grid using the
expressions derived previously.
At (i,j) = (1,1):
1,1 1
(a given boundary condition)
At (i,j) = (2,1):
2,1 1
At (i,j) = (3,1):
3,1 1
At (i,j) = (4,1):
4,1 1
At (i,j) = (1,2):
1,1 1,3 22,2 41,2 0
At (i,j) = (2,2):
1,2 3,2 2,1 2,3 42,2 0
At (i,j) = (3,2):
2,2 4,2 3,1 3,3 43,2 0
At (i,j) = (4,2):
23,2 4,1 4,3 44,2 0
At (i,j) = (1,3):
22,3 1,2 1,4 41,3 0
At (i,j) = (2,3):
1,3 3,3 2,2 2,4 42,3 0
At (i,j) = (3,3):
2,3 4,3 3,2 3,4 43,3 0
At (i,j) = (3,3):
23,3 4,2 4,4 44,3 0
At (i,j) = (1,4):
1,4 0
At (i,j) = (2,4):
2,4 0
At (i,j) = (3,4):
3,4 0
At (i,j) = (4,4):
4,4 0
Re-write the previous equations in matrix form.

1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1,1 1
0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1
2,1
0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 3,1 1

0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 4,1 1
1 0 0 0 4 2 0 0 1
0 0 0 0 0 0 0 1,2 0

0 1 0 0 1 4 1 0 0 1 0 0 0 0 0 0 2,2 0
0 0 1 0 0 1 4 1 0 0 1 0 0 0 0 0 0
3,2
0 0 0 1 0 0 2 4 0 0 0 1 0 0 0 0 4,2 0

0 0 0 4 2 0 0 1 0 0 0 1,3 0
0 0 0 0 1
0 0 0 0 0 1 0 0 1 4 1 0 0 1 0 0 2,3 0

0 0 0 0 0 0 1 0 0 1 4 1 0 0 1 0 3,3 0
0 0 0 0 0 0 0 1 0 0 2 4 0 0 0 1 4,3 0

0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1,4 0
0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0
2,4
0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 3,4 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 4,4 0
C. Wassgren
282
(5.62)
A direct method for solving Eqn. (5.62) is Gauss Elimination. The algorithm for Gauss Elimination
is not presented here and instead the reader is encouraged to review the method in a numerical
methods text (see, for example, Hoffman, J.D., Numerical Methods for Engineers and Scientists, 2nd
ed., Marcel-Dekker).
Solving Eqn. (5.62) using Gaussian Elimination gives:
1,1 1

2,1 1
3,1 1

4,1 1
1,2 2
2,2 2 3
2
3,2 3
4,2 2 3
These are the results we expected!
1
1,3 3
2,3 1 3

3,3 1 3
1
4,3 3
1,4 0
0
2,4
3,4 0

4,4 0
(5.63)
Notes:
a. Gaussian elimination is the preferred method for solving systems of linear equations.
Modifications to the Gaussian elimination algorithm have been proposed that are optimized for
banded matrices (where non-zero entries in the matrix occur in diagonal bands) such as the one
in Eqn. (5.62). Thomas algorithm is one such algorithm that is particularly efficient for tridiagonal matrices.
b.
Direct methods, as opposed to iterative methods, will always converge to a solution (assuming
that the given [A] matrix is non-singular, i.e. it has a non-zero determinant).
c.
For very large systems, direct methods are generally less efficient than iterative methods.
C. Wassgren
283
Another approach for determining the values of i,j in Eqn. (5.62) is to use an iterative scheme. With
an iterative method, an initial guess for the solution to i,j is assumed. These initial values for i,j are
then used to generate new values for i,j using a scheme that reduces the value between the current
values of i,j and the actual solution values. The scheme is repeated using the new values for i,j
until the values converge to the solution. Note that convergence of the iterative algorithm is not
always guaranteed, which is the major drawback to iterative methods.
One commonly used iterative algorithm is Gauss-Seidel Iteration with Successive Over-Relaxation.
In this algorithm, the new value for i,jn+1, where the superscript n+1 indicates the new value, is
determined using the previous values, found at iteration step n, of at the surrounding points (this
is actually known as the Jacobi Iteration Method Gauss-Seidel Iteration will be discussed in a
moment). For example, Eqn. (5.53) may be written in iterative form as:
in, j 1
1
4
n
i 1, j
in1, j in, j 1 in, j 1
(5.64)
Iteration on all points (i,j) continues until the error between the current iteration value for and the
previous iterations value for is less than some tolerance, i.e.:
Repeat iterations until in, j 1 in, j tolerance for all (i, j).
(5.65)
The difference between Gauss-Seidel iteration and Jacobi iteration is that Jacobi iteration determines
the value for i,jn+1 based on all of the previous iteration values whereas Gauss-Seidel iteration makes
use of the new values for as they become available. For example, if we iterate in the previous
example moving in the direction of increasing i and increasing j, then the value for 3,3 will be:
n 1
3,3
1
4
n 1
2,3
n
n 1
n
4,3
3,2
3,4
(5.66)
By using the already updated values at the neighboring grid points, convergence is accelerated.
Often, the rate of convergence of the iterations can be improved by implementing a relaxation
scheme. With relaxation the value of i,jn+1 is found using a linear combination of Eqn. (5.64) (or
rather an equation similar to Eqn. (5.66) depending on the direction of iteration) and the previous
value for i,jn, i.e.:
in, j 1 in, j in, j 2 in, j

1
(5.67)
where the superscript n+1/2 refers to the intermediate value of i,j calculated using Eqn. (5.64) (or
Eqn. (5.66)) and is referred to as the relaxation parameter. The effect of relaxation can be most
easily understood when presented graphically. In many instances the iterative values of i,j approach
the actual value of i,j from one direction as shown in the plot below where a particular i,j is shown
as a function of the number of iterations, n. We observe from the plot that by using over-relaxation
we are extrapolating the value of i,jn+1 using i,jn and i,jn+1/2 to help reach the converged value more
quickly.
actual value of i,j
i1, j
1
i0,j 2
i0, j
C. Wassgren
i,j
over-relaxation
no relaxation
i0, j 2 i0, j
1
number of iterations, n
284
Notes:
a. If > 1 then the process is known as over-relaxation, < 1 is under-relaxation, and when = 1
there is no relaxation. Under-relaxation is typically used when the iterations produce oscillatory
values for i,j.
4.
b.
For over-relaxation, the iterative scheme can be shown to diverge if 2.
c.
Relaxation can reduce the convergence rate considerably, often by one to two orders of
magnitude!
d.
The optimal choice for the relaxation parameter is not known a priori, in general, and multiple
computations using differing values of need to be performed to determine opt. Despite the
additional computations, determining opt is still a worthwhile effort, especially if the system of
equations must be solved multiple times (if the boundary conditions change for example). As a
rule of thumb, larger systems usually have a larger value for opt.
Additional issues such as the effects of round-off and truncation errors should be considered in more
depth for a better understanding of numerical solutions. The reader is encouraged to study numerical
methods texts for more information on these topics.
C. Wassgren
285
11. Doublet Distributions

So far weve approached potential flow problems by choosing potential functions and observing what types
of flows result. Lets now look at a method of specifying an object shape and determining what the
potential function should be.
Well just examine a simple method here but it should be noted that more sophisticated methods (although
based on the same concepts) are addressed in most books on aerodynamics (see, for example, Kuethe, A.M.
and Chow, C.-Y., Foundations of Aerodynamics, Wiley).
Recall that when we combined a uniform flow with a doublet, flow around a cylinder resulted. Now lets
imagine combining on the x-axis a large number of doublets with varying strength.
y
U
x
The stream function evaluated at a point (x, y) for such a flow is given by
L
K yd
y
Uy
2
2
0 x y
(x,y)
r
where K()d is the total strength of the

doublets over a very small distance d, is
the distance from the origin, and L is the
total length of the line of doublets.
Since we generally solve these types of problems numerically, re-write the integral as a summation:
jN
K j y
Uy
2
2
j 1 x y
j
where =L/N .
Were usually interested in determining what the potential function should be for a specific object. To
solve this inverse problem, we note that the object surface is a streamline so the stream function remains
constant on the surface. Since we can arbitrarily adjust the value of the stream function (by adding in a
constant remember that only differences or derivatives of the stream function are of interest to us), we can
adjust the stream function so that its value is zero on the object surface:
jN
i 0 Uyi cij K j
j 1
where i is a point on the object surface and cij is referred to as the influence coefficient (the contribution
of a doublet of unit density at the location j to the point i):
yi
cij
(5.68)
2
xi j yi2
C. Wassgren
286
The result is a system of equations we can solve numerically to determine the appropriate values of Kj:
c11 K1 c12 K 2 c1N K N Uy1
cN 1 K1 cN 2 K 2 cNN K N Uy N
Here the yi are known since the geometry is known, U is known, and the cij are known as discussed
previously.
Notes:
1.
We could also have used a potential function in the previous analysis but instead of specifying the
value of the potential function on the surface of the object, we would instead require that there is no
flow through the surface:
0
un
n
This approach is generally more involved than if we use stream functions.
2.
We can extend these ideas to asymmetric objects by distributing doublets along curved paths.
3.
Instead of using a line of doublets (aka doublet panel), we could also use lines of sources (aka source
panels), or lines of vortices (aka vortex panels).
4.
There will be no lift on objects generated with doublets or sources since they produce no net
circulation. Only vortex panels will produce circulation and lift.
5.
These ideas can be extended to 3D using 3D source/doublet/vortex potentials.
C. Wassgren
287
Review Questions
1. Describe three properties of stream functions.
2. Can stream functions be used for rotational flows? How about irrotational flows?
3. What restrictions are there when using stream functions?
4. Can stream functions be superposed?
5. What is the governing equation for an incompressible potential flow?
6. What are the requirements for modeling a flow as a potential flow?
7. What are the appropriate boundary conditions for a potential flow?
8. Can potential functions be written for 3D flows? How about stream functions?
9. Can stream functions be written for rotational flows?
10. How are streamlines related to equipotential lines?
11. Under what conditions can one write a complex potential function to describe a flow?
12. How is a fluid velocity field determined from a complex potential function?
13. Describe the potential flow model for ideal fluid flow around a non-rotating cylinder.
14. Describe the potential flow model for ideal fluid flow around a rotating cylinder.
15. What is dAlemberts paradox?
16. What causes Magnus lift? Is this what causes baseballs or golf balls to curve?
17. Why does potential flow modeling fail to capture the behavior of real flows downstream of a cylinder?
18. What is the Blasius integral law?
19. What is the Kutta-Joukowski theorem?
20. What is dAlemberts paradox?
21. Describe the method of images.
22. What is meant by ground effect?
23. What is meant by added mass? Under what conditions will the added mass on an object be
significant?
C. Wassgren
288
Chapter 06:
1.
2.
3.
4.
5.
6.
Buckingham-Pi Theorem
Method of Repeating Variables
Dimensionless Governing Equations
Modeling and Similarity
Stokes Number for Small Particles in a Flow
C. Wassgren
Chapter 06: Dimensional Analysis
289
1.
Dimensional analysis is a method for reducing the number and complexity of variables used to describe a
physical system. Its a technique that can be applied to all fields, not just fluid mechanics. The mechanics
of dimensional analysis are simple to learn and apply, and the benefits from using it are significant.
Dimensional analysis can be used to present data in an efficient manner, reduce the number of experiments
or simulations one needs to perform to investigate the relationship between variables, and scale results.
However, dimensional analysis cannot tell us what the functional relationship is between variables.
Additional experiments or theoretical analyses are required to determine this information.
Motivating Example #1
To motivate the use of dimensional analysis, lets consider a simple example involving a
ball falling under the action of gravity in a vacuum. From basic physics, we know that
the vertical position of the ball, y, is given by:
y 12 gt 2 y 0 t y0
(6.1)
where g is the acceleration due to gravity, t is the time from when the ball was released,
y
y 0 is the initial speed of the ball, and y0 is the initial position of the ball. Note that Eqn.
(6.1) is dimensional. In other words, each term in the equation has dimensions of length
[L]. For example, the dimension of the first term on the right hand side is length, [1/2gt2]
= L, where the square brackets indicate dimensions of. If we were to plot the position, y, as a function of
time, t, for varying g, y 0 , and y0, we would have plots that look like the following.
35
35
y0 = 10 m, ydot0 = 0
y0 = 20 m, ydot0 = 0
y0 = 30 m, ydot0 = 0
y0 = 10 m, ydot0 = -1 m/s
y0 = 20 m, ydot0 = -1 m/s
y0 = 30 m, ydot0 = -1 m/s
y0 = 10 m, ydot0 = -2 m/s
y0 = 20 m, ydot0 = -2 m/s
y0 = 30 m, ydot0 = -2 m/s
y0 = 10 m, ydot0 = 0
g = 9.81 m/s 2
g = 4.91 m/s 2
y0 = 20 m, ydot0 = 0
30
30
y0 = 30 m, ydot0 = 0
y0 = 10 m, ydot0 = -1 m/s
y0 = 20 m, ydot0 = -1 m/s
25
y0 = 30 m, ydot0 = -1 m/s
position, y 0 [m]
position, y 0 [m]
25
y0 = 10 m, ydot0 = -2 m/s
20
y0 = 20 m, ydot0 = -2 m/s
y0 = 30 m, ydot0 = -2 m/s
15
20
15
10
10
Clearly
there is
0
0.0
0.5
1.0
1.5
2.0
2.5
3.0
0.0
0.5
1.0
time, t [s]
1.5
2.0
2.5
time, t [s]
35
y0 = 10 m, ydot0 = 0
y0 = 20 m, ydot0 = 0
y0 = 30 m, ydot0 = 0
y0 = 10 m, ydot0 = -1 m/s
y0 = 20 m, ydot0 = -1 m/s
y0 = 30 m, ydot0 = -1 m/s
y0 = 10 m, ydot0 = -2 m/s
y0 = 20 m, ydot0 = -2 m/s
y0 = 30 m, ydot0 = -2 m/s
g = 3.27 m/s 2
30
position, y 0 [m]
25
20
15
10
0
0.0
0.5
1.0
1.5
2.0
2.5
3.0
time, t [s]
C. Wassgren
290
3.0
Now lets present the same information, but in dimensionless form. Starting with Eqn. (6.1), divide all
terms by y0 (a length), to make each term dimensionless:
y
y
g
1 g
t2
0 t
1
(6.2)
y0
y0
2 y0
gy0
or, in a slightly more compact form,
y 12 t 2 y 0 t 1
(6.3)
where
y
y
, y 0
y0
y 0
gy0
, and t t
g
y0
(6.4)
are the dimensionless position, initial speed, and time, respectively. Note that Eqns. (6.3) and (6.1) are
identical; theyre just written in dimensionless or dimensional form. Now if we were to plot Eqn. (6.3) for
all of the various combinations of variables, we would have the following plot.
1.0
dimensionless position, y' = y /y 0
0.9
ydot0' = 0
ydot0' = -1
0.8
ydot0' = -2
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.0
0.5
1.0
1.5
0.5
dimensionless time, t' = t (g /y 0)
This dimensionless plot contains all of the information that was contained in the previous dimensional
plots. As you can see, presenting data in dimensionless form is very efficient!
Now lets assume we didnt know that Eqn. (6.1) existed and we had to perform a series of experiments to
try to find the functional relationship between the variables
(6.5)
y fcn1 g , t , y 0 , y0
where fcn1 is the unknown function were trying to determine. Lets say that we perform a series of
experiments where we vary each of the variables independently five times. Since we have four
independent variables (g, t, y 0 , y0), this means we have a total of 54 = 625 experiments to perform! Not
only is this a lot of experiments, but some of these experiments are likely to be difficult and expensive to
carry out (e.g. varying g is not trivial).
Now if we instead performed a dimensional analysis on Eqn. (6.5) (which you will learn how to do later in
this set of notes), we could show that Eqn. (6.5) can be written in dimensionless form as:
g
y
y
, 0
fcn2 t
(6.6)
y0 gy
y0
0
(Compare Eqn. (6.6) to Eqn. (6.2) to verify.) Equation (6.6) contains only two independent variables,
hence, varying each parameter five times gives a total of 52 = 25 total experiments. Clearly, performing a
dimensional analysis can reduce the number of experiments one needs to perform! Not only are the
number of experiments reduced, but the experiments can be much easier to perform. For example, varying
the two independent parameters in Eqn. (6.6) can be achieved by simply letting time vary, and varying the
C. Wassgren
286
initial drop speed. We neednt worry about varying gravity, g, independently since g is contained within
the term t(g/y0)1/2.
Finally, now lets say that we are interested in launching an object on the Moon (indicated by the subscript
M) from a specified height, y0,M = 1 m, with a specified speed, y 0,M = 1 m/s, and want to know how long
it will take for the object to impact the ground, yM(tM = ?) = 0. We know that the acceleration due to gravity
on the Moon is gM = 1.62 m/s2.
Again, assuming we dont know that Eqn. (6.1) exists, we can still determine this time by performing a
similar experiment on Earth, and then scaling the result. If were to perform this similar experiment on
Earth, where gE = 9.81 m/s2, we need to first determine the initial drop height, yE,0, and speed, y 0,E , for the
Earth experiment. Since the same physical process holds for both the Moon and Earth, the dimensionless
terms describing the process will be identical, i.e. Eqn. (6.6) will be the same for the Earth and Moon.
Thus, we can equate dimensionless terms to determine the values that should be used on the Moon:
lim
y
y
yE
0 m y E , yM 0
y0,E = 1 m
y0, E y0, M
1 m

0 m
yM
y0 E y0 M
(6.7)
y 0
y
y0, E g E
1 m 9.81 m s 2
y 0,E = 2.46 m/s(6.8)
1 m s
0 y 0, E y 0, M
2
gy
gy
1
m
1.62
m
s
y
g
M
M
0,
0
0
E
M
When performing the drop test on Earth with the given initial conditions, the time required for the ball to
hit the ground is tE = 0.77 s (which can be verified using Eqn. (6.1)). To determine the corresponding time
for the Moon, we equate the last dimensionless term in Eqn. (6.6).
g
g
g E y0, M
9.81 m s 2 1 m
tM = 1.89 s
(6.9)
0.77 s
t
t
tM t E
g M y0, E
1.62 m s 2 1 m
y 0 E y0 M
This time is exactly what one would calculate from Eqn. (6.1) using y0,M = 1 m, y 0,M = 1 m/s, and gM =
1.62 m/s2. Thus, we see that dimensional analysis can be used for scaling!
Hopefully, youre convinced that dimensional analysis is a worthwhile topic to study and apply. The
remainder of this chapter presents the mechanics of performing a dimensional analysis along with
examples. In addition, similarity and scaling issues are discussed.
Motivating Example #2
Most fluids engineering problems are too complex to be amenable to analytic, closed-form solutions. As a
result, experiments are used to determine relationships between the variables of interest (e.g. pressure and
velocity).
Lets consider the following example. Say we want to measure the pressure difference, p = p2-p1,
between two points separated by a distance, L, in a pipe.
p2
p1
C. Wassgren
287
On what variables do we expect the average pressure gradient, p/L, to depend? From experience and
intuition we might expect the following parameters to be important:
V average flow velocity
D pipe diameter
fluid density
fluid dynamic viscosity
We can write this relationship in the following, more mathematical form:
p/L = fcn(V, D, , )
In order to determine the form of this function, it would be logical to design experiments where we vary
just one of the parameters while holding the others constant and observe how p/L varies. For example:
p/L
D, , constant
p/L
V, , constant
V
p/L
D
V, D, constant
p/L
V, D, constant
This procedure, although logical, can be very time consuming, expensive, and difficult (if not impossible)
to perform (e.g. Can you find fluids that have the same viscosity but varying density?) As with the first
motivating example, using dimensional analysis will greatly simplify our experimental procedure. This
example will be used while presenting the various steps of performing a dimensional analysis in the
following sections.
C. Wassgren
288
2.
Buckingham Pi Theorem
The key component to dimensional analysis is the:

If an equation involving k variables is dimensionally homogeneous, it
Buckingham Pi Theorem:
can be reduced to a relationship among k-r independent dimensionless
products (referred to as terms), where r is the minimum number of
reference dimensions required to describe the variables, i.e.
# of variables # of reference dimensions
# of terms

(The proof to this theorem will not be presented here.)
Notes:
1. Dimensionally homogeneous means that each term in the equation has the same units. For
example, the following form of Bernoullis equation:
p V2
z constant
g 2g
is dimensionally homogeneous since each term has units of length (L).
2. A dimensionless product, also commonly referred to as a Pi () term, is a term that has no
dimensions. For example,
p
V 2
is a dimensionless product since both the numerator and denominator have the same dimensions.
3. Reference dimensions are usually basic dimensions such as mass (M), length (L), and time (T) or
force (F), length (L), and time (T). Well discuss the usually modifier a little later when
discussing the method of repeating variables.
C. Wassgren
289
3.
Method of Repeating Variables
The Buckingham Pi Theorem merely states that a relationship among dimensional variables may be
written, perhaps in a more compact form, in terms of dimensionless variables (terms). The Pi Theorem
does not, however, tell us what these dimensionless variables are. The method of repeating variables is an
algorithm that can be used to determine these dimensionless variables.
The method of repeating variables algorithm is as follows:
1.
List all variables involved in the problem.

a. This is the most difficult step since it requires experience and insight.
b. Variables are things like pressure, velocity, gravitational acceleration, viscosity, etc.
c. List only independent variables. For example, you can list:
(density) and g (gravitational acceleration), or
and (specific weight), or
g and
but you should not list, g, and since one of the variables is dependent on the others.
d. If you include variables that are unimportant to the system, then youll form terms that wont
have an impact in practice. This situation is the same one youd have if dimensional variables
were used.
e. If you leave out an important variable, then youll find that in practice your relationship between
dimensionless terms cant fully describe the system behavior. Again, this situation is the same one
youd have if you used dimensional terms.
2.
Express each variable in terms of basic dimensions.

a. For fluid mechanics problems we typically will use mass (M), length (L), and time (T) or force
(F), length (L), and time (T) as basic dimensions. We may occasionally need other basic
dimensions such as temperature ().
b. For example, the dimensions of density can be written as:
M FT 2
(Note: The square brackets are used to indicate dimensions of.)
3 4
L
L
3.
Determine the number of terms using the Buckingham-Pi Theorem.

a. (# of terms) = (# of variables) (# of reference dimensions)
b. Usually the # of reference dimensions will be the same as the # of basic dimensions found in step
2. There are (rare) cases where some of the basic dimensions always appear in particular
combinations so that the # of reference dimensions is less than the # of basic dimensions. For
example, say that the variables in the problem are A, B, and C, and their corresponding basic
dimensions, are:
M
M
MT
A 3 B 3 2 C 3
L
LT
L
The basic dimensions are M, L, and T (there are 3 basic dimensions). Notice, however, that the
dimensions M and L always appear in the combination M/L3. Thus, we really only need two
reference dimensions, M/L3 and T, to describe all of the variables dimensions.
4.
Select repeating variables where the number of repeating variables is equal to the number of reference
dimensions.
a. The repeating variables should come from the list of variables. In our previous example, the list of
variables is p/L, V, D, , and .
b. Each repeating variable must have units independent of the other repeating variables.
c. Dont make the dependent variable one of the repeating variables. In our previous example, p/L
is the dependent variable.
d. All of the reference dimensions must be included in the group of repeating variables.
C. Wassgren
290
5.
Form a term by multiplying one of the non-repeating variables by the product of the repeating
variables, each raised to an exponent that will make the combination dimensionless.
a. This step is most clearly illustrated in an example and will not be discussed here.
b. Repeat this step for all non-repeating variables.
6.
Check that all terms are dimensionless.

a. This is an important, and often overlooked, step to verify that your terms are, in fact,
dimensionless.
7.
Express the final form of the dimensional analysis as a relationship among the terms.
a. For example, 1 = fcn(2, 3, , k-r).
C. Wassgren
291
Example:
Lets use our pipe flow experiment as an example.
Step 1:
The variables that are important in this problem are:
p/L average pressure gradient over length L
V average flow velocity
D pipe diameter
fluid density
fluid dynamic viscosity
p/L = fcn1(V, D, , )
Step 2:
The basic dimensions of each variable are:
p F M
L L3 L2T 2
L
V
T
D L
M
L3
FT M
2
LT
L
Step 3:
(# of variables) = 5 (p/L, V, D, , )
(# of reference dimensions) = 3 (F, L, T or M, L, T)
(# of terms) = (# of variables) (# of reference dimensions) = 2
Thus, instead of having a relation involving 5 terms, we actually have a relationship involving just 2
terms!
Step 4:
Select the following 3 repeating variables (We require three since the # of reference dimensions is
three.):
, V, D
Notes:
1. These three repeating variables have independent dimensions.
2. The dependent variable (p/L) is not one of the repeating variables.
3. We could have also selected (,, V) or (,, D) or (D,, V) as repeating variables. The choice
of repeating variables is somewhat arbitrary (as long as they have independent reference
dimensions and do not include the dependent variable).
C. Wassgren
292
Step 5:
Form terms from the remaining, non-repeating variables:
1 p aV b D c
L
0
M M L L
MLT 2 2 3
L T L T 1
0
1
M M Ma
0 1 a
2 3 a
a 1
L L L L L 0 2 3a b c b 2
T 0 T 2 T b
0 2 b
c 1
0
b c
p D
L

1
V 2
2 aV b D c
a
0
M M L L
MLT
3
LT L T 1
M 0 M 1M a
0 1 a
a 1
0
1 3 a b c
L L L L L 0 1 3a b c b 1
T 0 T 1T b
2
0 1 b
c 1
VD
Step 6:
Verify that each term is, in fact, dimensionless.
p D
L
1
2
V
M
2
VD LT
M L L3 T 2
M 0 L0T 0
2 2
2
LT 1 M L
OK!
L3 T 1
M 0 L0T 0
M LL
OK!
Step 7:
Re-write the original relationship in dimensionless terms.
p D
L

fcn2
2
V
VD
C. Wassgren
293
Notes:
1. Instead of having to run four different sets of experiments as was discussed at the beginning of this
chapter, we only really need to run one set of experiments where we vary:
VD
p L D
and measure:
p D
L
V 2
V 2
Note: We can use 1/ in place of

since we havent actually
specified what the function, fcn2,
looks like.
This is a Reynolds number!

VD
All of the information contained in the previous four plots is contained within this single plot! This
reduces the complexity, cost, and time required to determine the relationship between the average
pressure gradient and the other variables.
2.
Dimensional analysis is a very powerful tool because it tells us what terms really are important in an
equation. For example, we started with the relation:
p fcn V , D, ,
1
L
leading us to believe that V, D, , and are all important terms by themselves. However, dimensional
analysis shows us that instead of the terms by themselves, it is the following grouping of terms:
p D
L

fcn2
V 2
VD
that is important in the relationship. This is a subtle but very important point.
3.
Dimensional analysis tells us how many dimensionless terms are important in a relation. It does not
tell us what the functional relationship is.
4.
The dimensionless terms found via dimensionless analysis are not necessarily unique. Had we
chosen different repeating variables in the previous example, we would have ended up with different
terms. One can multiply, divide, or raise their set of terms to form the terms found by another.
The number of terms, however, is unique.
5.
After a bit of practice, one can quickly form terms by inspection rather than having to go through the
method of repeating variables.
C. Wassgren
294
Example:
An open cylindrical tank having a diameter D is supported around its bottom circumference and is filled to
a depth h with a liquid having a specific weight . The vertical deflection, , of the center of the bottom is a
function of D, h, d, , and E where d is the thickness of the bottom and E is the modulus of elasticity of the
bottom material. Form the dimensionless groups describing this relationship.
SOLUTION:
1.
Write the dimensional functional relationship.

f1 D, h, d , , E
2.
Determine the basic dimensions of each parameter.

L
h L
D L
d L
M
F
3
2 2
LT
L
M
F
E 2 2
LT
L
3.
Determine the number of terms required to describe the functional relationship.

# of variables = 6 (, D, h, d, , E)
# of reference dimensions = 2 (L, F/L2 or L, M/T2)
(Note that the number of reference dimensions and the number of basic dimensions are not the
same for this problem!)
(# terms) = (# of variables) (# of reference dimensions) = 6 2 = 4
4.
Choose two repeating variables by which all other variables will be normalized (same # as the # of
reference dimensions).
D, (Note that the dimensions for D and are independent.)
5.
Make the remaining non-repeating variables dimensionless using the repeating variables.
1 D a b
1 L 1 F L
F 0 L0 L
F: 0 b
L: 0 1 a 3b a 1
1
2 hD a b
1 L1 F L
F 0 L0 L
F: 0 b
L: 0 1 a 3b a 1
2 h
D
C. Wassgren
295
3 dD a b
1 L1 F L
a
F 0 L0 L
F: 0 b
L: 0 1 a 3b a 1
3 d
D
4 ED a b
L L1 F L
F 0 L0 F
F: 0 1 b b 1
L: 0 2 a 3b a 1
4 E
D
6.

1 D L L 1 OK!
2 h D L L 1
OK!
3 d D L L 1
OK!
4 E D F L2 1 L L
7.
OK!

h d E
f2 , ,
D
D D D
C. Wassgren
296
Example:
A viscous fluid is poured onto a horizontal plate as shown in the figure. Assume that the time, t, required
for the fluid to flow a certain distance, d, along the plate is a function of the volume of fluid poured, V,
acceleration due to gravity, g, fluid density, , and fluid dynamic viscosity, . Determine an appropriate set
of dimensionless terms to describe this process.
volume, V
SOLUTION:
1.

t f1 d ,V , g , ,
2.

t T
d L
V L3
g LT 2
M L3
M LT
3.

# of variables = 6 (t, d, V, g, , )
# of reference dimensions = 3 (T, L, M)
(Note that the number of reference dimensions and the number of basic dimensions are equal
for this problem.)
4.
Choose three repeating variables by which all other variables will be normalized (same # as the # of
d, g, (Note that the dimensions for these variables are independent.)
C. Wassgren
297
5.
1 td a g b c
T L
M 0 L0T 0
1 1
M:
0c
L : 0 a b 3c
0 1 2b
T:
1 t
L M
2 3
T L
a 12
b
1
2
c0
g
d
2 Vd a g b c
L3 L L M
M 0 L0T 0 2 3
1 1 T L
M:
0c
a 3
L : 0 3 a b 3c b 0
0 2b
T:
c0
V
d3
3 d a g b c
M L L M
M 0 L0T 0
2 3
LT 1 T L
a 32
M:
0 1 c
L : 0 1 a b 3c b 12
0 1 2b
T:
c 1
3
6.
d gd

g T L1 2 1
1 OK!
1 t
1
2
d 1 T L
V L3 1
1 OK!
3
3
d 1 L
M L3 1 T 1
1 OK!
1
1
2
2
d gd LT M L L L
3
7.

V
g

t
f2 3 ,
d d gd
d
C. Wassgren
298
4.
Dimensionless Form of the Governing Equations
Consider the dimensional form of the governing equations for an incompressible fluid with constant
viscosity in a gravity field:
u j
Continuity (COM)
x j
Navier-Stokes Eqns (LME):
Thermal Energy Eqn:
ui
u
2 ui
p
uj i
gi
t
x j
xi
x j x j
T
u j ui
T
2T
c
uj
k
t
xi x j
x
x
x
j
j
j
ui
x j
Note that in the thermal energy equation the internal energy has been written as the specific heat (assumed
constant, note that for an incompressible flow cv = cp = c) multiplied by the temperature and the heat
transfer has been assumed to be due solely to conduction (Fouriers Law with a constant conduction
coefficient). Lets re-write these equations in dimensionless form using some characteristic flow quantities
(to be discussed in a moment). The variables in the equations will be normalized using the following
quantities:
x
xi* i xi Lxi*
L
u
*
ui i ui Uui*
U
t
t*
t t*
p*
p
p0
p p0 p*
T
T T0T *
T0
where the superscript * refers to a dimensionless quantity. The quantity L represents a characteristic
length for the flow of interest (e.g. a pipe diameter or the diameter of a sphere), U is a characteristic
velocity (e.g. the free stream velocity or the average velocity in a pipe), is a characteristic time scale (e.g.
the period of an oscillating boundary), p0 is a characteristic pressure (e.g. the free stream pressure or the
vapor pressure), and T0 is a characteristic temperature (e.g. the free stream temperature). These
characteristic quantities give us an estimate of the typical magnitude of the various terms in the equations.
They wont be exact values but they will give us a feel for how a parameter might scale in a flow, e.g. we
might expect the fluid velocities in a flow to scale with the incoming free stream velocity.
T*
Now lets rewrite the governing equations using these dimensionless parameters.
Continuity:
u *j
x*j
Navier-Stokes Eqns:
p0 p* U 2 ui*
U ui* U 2 * ui*
gi
j
t *
L
L xi* L2 x*j x*j
x*j
Thermal Energy Eqn:
cT0 T * U cT0 * T * kT0 2T * U 2
2
uj * 2
t *
L
x j
L x*j x*j
L
u*j u* u*
i i
xi* x*j x*j
The continuity equation simplifies to an identical equation but in dimensionless form. The Navier-Stokes
and thermal energy equations are still in dimensional form due to the dimensional terms in front of each
dimensionless term. Note that each dimensional term represents a characteristic force/energy magnitude.
C. Wassgren
299
For example, the term in front of the convective acceleration term in the momentum equations, i.e. the
(U2/L) in front of u j ui x j , represents the magnitude of a typical (convective) inertial force (recall
that each term in the N-S equations has dimensions of force per unit volume). Similarly, the (U/L2) term
represents a characteristic viscous force. In order to make the N-S and thermal energy equations
dimensionless, it is customary to divide through by the characteristic (convective) inertial force term in the
N-S equations and by the convective term in the thermal energy equation:
u*j
continuity:
x*j
*
*
p0 p* ui*
gi L
L ui
* ui
u
* * 2
j
*
*
2
*
U
UL
x j
t
x j x j U
U xi
Navier-Stokes:
*
*
*
k 2T * U 2 u j ui* ui*
L T
* T
j
U *
x*j c UL x*j x*j cT0 UL xi* x*j x*j
t
Now each term in the continuity, N-S, and thermal energy equations is dimensionless. The dimensionless
quantities in front of each term have special meaning:
Thermal Energy:
Strouhal #, St
L
represents the ratio of (local or Eulerian) inertial forces to (convective) inertial forces.
U
The Strouhal number is often significant in unsteady, periodic flows.

Euler #, Eu
p0
represents the ratio of pressure forces to (convective) inertial forces.
U 2
The Euler number is typically significant in flows where large changes in pressure occur. The Euler
number is also often written as a pressure coefficient, cP:
cP
p p0
U 2
or in flows where cavitation occurs, as the cavitation number, Ca:
Ca
p pv
U 2
where pv is the vapor pressure of the fluid.

Reynolds #, Re
UL
represents the ratio of (convective) inertial forces to viscous forces.
The Reynolds number is significant in virtually all fluid flows.

Froude #,
Fr
represents the ratio of (convective) inertial forces to gravitational forces
gL
The Froude (pronounced frd) number is typically significant in flows involving a free surface.
Prandtl #,
Pr
c
k
represents the ratio of the momentum diffusivity, , to the thermal diffusivity,
k/(c).
The Prandtl number gives a measure of how rapidly momentum diffuses through a fluid compared
to the diffusion of heat. Most gases have a Prandtl number near one (heat and momentum diffuse at
nearly the same rate) while water has a Prandtl number near ten (momentum diffuses faster than
heat).
C. Wassgren
300
Thus, the dimensionless forms of the governing equations are:

continuity:
Navier-Stokes Eqns:
Thermal Energy Eqns:
u*j
x*j
ui*
St
St
t *
T *
t
u*j
ui*
u*j
*
x*j
Eu
p*
xi*
1 ui*
1 g
2 i
* *
Re x j x j Fr gi
*
2T *
U 2 1 u j ui* ui*
1
x*j Pr Re x*j x*j c pT0 Re xi* x*j x*j
T *
Additional dimensionless quantities occur when dealing with other equations of significance (e.g. the
equations for a compressible fluid) and with the boundary conditions (e.g. surface tension effects or surface
roughness).
C. Wassgren
301
5.
Modeling and Similarity
Models are often used in fluid mechanics to predict the kinematics and dynamics of full-scale (often
referred to as prototype) flows. From previous discussions of dimensional analysis, we observe that we can
write the governing equations and boundary conditions of our flow in dimensionless terms ( terms).
Thus, if we have two different flows (e.g., a large-scale, prototype flow and a small scale, model flow) that
have identical dimensionless parameters, then the same solution, also in terms of dimensionless parameters,
will hold for both. This is extremely helpful when modeling fluid systems.
When a model and the prototype have the same dimensionless parameters, we say that they are similar. We
typically discuss similarity in three categories: geometric, dynamic, and kinematic.
Geometric similarity occurs when the model is an exact geometric replica of the prototype. In other words,
all of the lengths in the model are scaled by exactly the same amount as in the prototype.
For example:
prototype
WP
model
WM
LM
LP
LP/LM = WP/WM
Note that surface roughness may even need to be scaled if it is a significant factor in the flow.
Dynamic similarity occurs when the ratio of forces in the model is the same as the ratio of forces in the
prototype, i.e.
[ReP = ReM]
(ratio of inertial to viscous forces)P = (ratio of inertial to viscous forces)M
[EuP = EuM]
(ratio of pressure to inertial forces)P = (ratio of pressure to inertial forces)M
[FrP = FrM]
(ratio of inertial to grav. forces)P = (ratio of inertial to grav. forces)M
etc.
Kinematic similarity occurs when the prototype and model fluid velocity fields have identical streamlines
(but scaled velocities). Since the forces affect the fluid motion, geometric similarity and dynamic similarity
will automatically ensure kinematic similarity.
Note:
1. When modeling, we need to maintain similarity between all of the dimensionless parameters that are
important to the physics of the flow. This means that we do not necessarily need to have similarity
between all terms, just the ones that significantly affect the flow physics. Knowing a priori what
dimensionless terms are important can be difficult but with experience the task is often easier.
2.
It is not uncommon to have the important physics of a system change at different scales. For example,
surface tension forces become more pronounced at smaller geometric scales. If one was scaling up a
small system in which surface tension was an important effect, but didnt consider the dynamic
similarity of the surface tension force at the larger scale, then the scaling experiments would result in
incorrect results.
C. Wassgren
302
Example:
The drag characteristics of a blimp 5 m in diameter and 60 m long are to be studied in a wind tunnel. If the
speed of the blimp through still air is 10 m/s, and if a 1/10 scale model is to be tested, what airspeed in the
wind tunnel is needed for dynamic similarity? Assume the same air temperature and pressure for both the
prototype and model.
SOLUTION:
For dynamic similarity, equate the model and prototype Reynolds numbers.
Re P Re M
VD
VD

P M
Since both the model and prototype use air at the same temperature and pressure as the working fluid, P =
M.
D
10
VM VP P 10 m
s 1
D
M
VM 100 m
s
Note that the model speed is still low enough that Mach number effects (i.e. compressibility effects) do not
come into play.
C. Wassgren
303
Example:
The height of the free surface, h, in a tank of diameter, D, that is draining fluid through a small hole at the
bottom with diameter, d, decreases with time, t. This change in free surface height is studied
experimentally with a half-scale model. For the prototype tank:
H = 16 in. (the initial height of the free surface)
D = 4.0 in.
d = 0.25 in.
Experimental data is obtained from the prototype and half-scale model and is given below:
Model Data
h [in.]
8.0
7.0
6.0
5.0
4.0
3.0
2.0
1.0
0.0
1.
2.
3.
Prototype Data
h [in.]
16.0
14.0
12.0
10.0
8.0
6.0
4.0
2.0
0.0
t [s]
0.0
3.1
6.2
9.9
13.5
18.1
24.0
32.5
43.0
t [s]
0.0
4.5
8.9
14.0
20.2
25.9
32.8
45.7
59.8
Plot, on the same graph, the height data as a function of time for both the model and the prototype.
Develop a set of dimensionless parameters for this problem assuming that: h = f(H, D, d, g, t)
Replot, on the same graph, the height data as a function of time in non-dimensional form for both the
model and prototype.
D
g
H
hole with diameter, d
C. Wassgren
304
SOLUTION:
h [in.]
First plot the model and prototype dimensional data.
18
16
14
12
10
8
6
4
2
0
model
prototype
20
40
60
80
t [s]
Now perform a dimensional analysis to determine the dimensionless terms describing the relationship.
1.

h f1 H , D, d , g , t
2.

h L
H L
D L
d L
g LT 2
t T
3.

# of variables = 6 (h, H, D, d, g, t)
# of reference dimensions = 2 (L, T)
(Note that the number of reference dimensions and the number of basic dimensions are equal
for this problem.)
4.
Choose two repeating variables by which all other variables will be normalized (same # as the # of
H, g (Note that the dimensions for these variables are independent.)
C. Wassgren
305
5.
h
1
(Found via inspection.)
H
D
2
H
d
3
H
4 tH a g b
a
T L L
L0T 0 2
1 1 T
a 12
L: 0 ab
T : 0 1 2b
b 12
4 t
6.
g
H

h
L1
1 OK!
1
H
1L
D
L1
1 OK!
2
H 1 L
d
L1
1 OK!
3
H 1 L
g T L12 1
1 OK!
4 t
1
2
H 1 T L
7.

D d
h
g
f 2 , , t
H
H
H H
Now plot the model and prototype data in dimensionless form. Note that since there is geometric similarity
(the model is one-half the size of the prototype):
d
d
D
D
and
H M H P
H M H P
C. Wassgren
306
1.2
model
prototype
1.0
h/H
0.8
0.6
0.4
0.2
0.0
0
100
200
300
400
t*sqrt(g/H)
Notice that the data collapse to a single curve when plotted in dimensionless terms.
C. Wassgren
307
Partial Similarity
True similarity may be difficult to achieve in practice. In such cases, one must either: (a) acknowledge
that model testing may not be possible, or (b) relax one or more similarity requirements and use a
combination of experimentation and analysis to scale the measurements.
For example, in modeling the flow around ships both Reynolds number and Froude number similarity are
important; however, both are difficult to achieve simultaneously. In such cases, one of the similarity
requirements is relaxed (in boat modeling its the Reynolds number similarity requirement) and a
combination of experiments and analysis is utilized to scale the measurements.
The two primary components of drag on a ships hull are viscous drag (i.e., the friction of the water against
the hulls surface) and wave drag (i.e., the force required to create the waves generated by the hull). The
two significant dimensionless parameters corresponding to these phenomena are:
VL
Re
(ratio of inertial to viscous forces)
Reynolds number:
Froude number:
Fr
(ratio of inertial to gravitational forces)

gL
Maintaining both Reynolds number and Froude number similarity is difficult to achieve in practice.
V
V
L
g M gM g P
L
FrM FrP
(6.10)
VM VP M
VM VP M
gL
gL
L
g
LP
P
P
M
P
using previous
result
L 2
V L
VL
VL
Re M Re P M P M M
M P M
(6.11)
VP LP
M P
LP
As an example, consider a scale model which has LP = 100LM, P = H2O = 1 cSt M = 0.001 cSt. There
is no such common model fluid available! Thus, we cannot easily maintain both Froude number and
Reynolds number similarity.
How do we resolve this difficulty? In practice, Froude number similarity is maintained with water as both
the prototype and model fluid (i.e., Eqn. (6.10) holds). Reynolds number similarity is neglected in the
experiment and instead analysis or computation is used to estimate the viscous drag contribution. The
procedure is as follows:
1. The total drag acting on the model is measured in the experiment. This drag force is usually
expressed in terms of a dimensionless resistance coefficient.
2. The viscous drag contribution to the total drag is calculated using analysis (e.g., boundary layer
analysis) or computation (e.g., computational fluid dynamics).
3. The difference between the total drag and the viscous drag is the wave drag.
4. The wave drag is a function of the Froude number, which is held constant during scaling. So, the
experimental wave drag data can be scaled between the model and prototype.
5. Estimate the viscous drag contribution for the prototype using analysis or computation.
6. Sum the predicted viscous drag force (step 5) with the scaled wave drag force (step 4) to get the
total prototype drag force.
C. Wassgren
308
As a demonstration of how well the procedure works, consider the resistance coefficient data from a 1:80
scale model test of the U.S. Navy guided missile frigate Oliver Hazard Perry (FFG-7) as shown in the plots
below. (Note that these plots are from Figs. 7.2 and 7.3 in Fox, R.W., Pritchard, P.J., and McDonald, A.T.,
2008, Introduction to Fluid Mechanics, 7th ed., Wiley.) The error between the scaled and actual total drag
force measurements is approximately 5%.
from scale model testing
predicted from boundary

layer theory
predicted from boundary

layer theory
Scale Model
Prototype
Experimental observations have shown that in many (but not all!) cases, Reynolds number similarity may
be neglected for sufficiently large Reynolds numbers.
(Figure from Fox, R.W.,
Pritchard, P.J., and McDonald,
A.T., 2008, Introduction to Fluid
Mechanics, 7th ed., Wiley.)
Moody diagram Friction factor as a function of Reynolds number and pipe relative roughness. Note that
the friction factor becomes independent of Reynolds number for sufficiently large Reynolds numbers.
C. Wassgren
309
(Figure from Fox, R.W.,

Pritchard, P.J., and McDonald,
A.T., 2008, Introduction to Fluid
Mechanics, 7th ed., Wiley.)
Drag coefficients for a sphere and circular disk as a function of Reynolds number. Note that the drag
coefficient is insensitive to Reynolds number over a wide range of Reynolds numbers.
C. Wassgren
310
The Stokes Number (St) for Small Particles in a Flow
The Stokes number, St, is defined as the ratio of the particle response time, p, to the fluid response time, f:
p
St
(6.12)
f
A response time is a measure of how rapidly a quantity responds to rapid changes. The Stokes number for
a particle is essentially a measure of how well the particle follows fluid streamlines. If St<<1 then the
particle will be able to follow the fluid streamlines whereas if St>>1 then the particle will not be able to
follow sudden changes in the fluid velocity. For example, consider driving down a country road late at
night during the summer when a lot of bugs are out. If the Stokes number for a bug is small, then it will
follow the fluid streamlines as you drive past it and it wont impact your car. However, if the Stokes
number for the bug is large, it will end up hitting your windshield since it wont be able to follow the fluid
streamlines that contour around your car.
Splat!
St >> 1
St << 1
The particle response time can be found by considering the particle equation of motion (assuming spherical
particles) for a particle with a speed slower than the surrounding fluid (so the particle accelerates):
2
2 dp
3 du p
1
(6.13)
d
C
u
u
p
D 2
f
f
p
p
4
6 dt
where p an f are the particle and fluid densities, dp is the particle diameter, up and uf are the particle and
fluid velocities, t is time, and CD is the particle drag coefficient. Define the Reynolds number for the
particle using the local relative velocity:
f u f u p d p
Re d
(Note that uf > up is assumed.)
(6.14)
f
Substitute Eqn. (6.14) into Eqn. (6.13) and simplify:
du p
f
2
3
f u f u p
CD Re d
dt
f u f u p d p
4 p d p
3 f
CD Re d
u f u p
4 p d p2
For small Reynolds numbers the drag coefficient approaches the Stokes drag:
24
CD
(Were now assuming that were dealing with small particles.)
Re d
so that the particle equation of motion becomes:
du p 18 f
u f u p
dt
p d p2
The solution to this equation, assuming a constant fluid velocity and a particle released from rest (up(t = 0)
= 0), is:
u p u f 1 exp t / p
where p is the particle response time given by:
d2
p p p
18 f
C. Wassgren
311
Let the fluid response time, f, for the flow geometry be:
L
f
uf
where L is a typical flow dimension (e.g. the effective frontal diameter of the car in the car/bug example
discussed previously).
Hence, the Stokes number for the particle is:
p p d p2 u f
St
f 18 f L
Re-writing in terms of the Reynolds number, ReL, based on the typical flow dimension, L:
f uf L
Re L
gives:
St
p p d p2 u f
f
Re L
f 18 f L
f uf L
Re
St L
18
dp

L
For very small particles compared to the flow dimension, i.e. (dp/L << 1), and moderate flow Reynolds
numbers and density ratios, we observe that St<<1 and the particle should follow the fluid streamlines.
C. Wassgren
312
Review Questions
1. Describe some of the benefits to performing a dimensional analysis of a problem.
2. What does the Buckingham-Pi theorem state? Are the dimensionless terms resulting from the theorem
unique?
3. Describe the method of repeating variables. Must this method always be followed to determine
dimensionless terms?
4. What is the difference between basic dimensions and reference dimensions?
5. Describe the three types of similarity.
6. Must there be exact similarity between a model and prototype in order to perform engineering
modeling?
7. In words, define the Reynolds, Froude, Strouhal, and Euler numbers.
C. Wassgren
313
Chapter 07:
Navier-Stokes Solutions
1. A Few Comments Regarding Exact Solutions to the Navier-Stokes Equations

2. Planar Couette-Poiseuille Flow
3. Poiseuille Flow
4. Starting Flow in a Circular Pipe
5. Impulsively Started Flat Plate (aka Stokes 1st Problem)
6. Oscillating Flat Plate (aka Stokes 2nd Problem)
7. Planar Stagnation Point Flow (aka Heimenz Flow)
8. Low Reynolds Number Flows
9. Stokes Flow Around a Sphere
10. Lubrication Flow
C. Wassgren
Chapter 07: Navier-Stokes Solutions
314
1.
A Few Comments Regarding Exact Solutions to the Navier-Stokes Equations
Because there is no general method for solving a system of non-linear, partial differential equations, there
are only a few exact solutions to the governing equations of fluid mechanics. For an incompressible fluid
with constant viscosity, the equations governing the fluid motion are the continuity and Navier-Stokes
equations:
u 0
u
u u p 2 u f
t
In general, we must make a number of assumptions to simplify the governing equations so that they
become manageable analytically. In particular, we often simplify the equations so that the non-linear
convective term in the Navier-Stokes equations, (u)u, is zero. Although we need to make many
assumptions in determining exact solutions, the resulting solutions are still of great engineering value.
They are often good models for real-world flows and they are commonly used to validate numerical codes
and experimental methods.
One assumption well make in all of the solutions is that the flow is laminar as opposed to being turbulent
or transitional. A laminar flow means that the fluid moves in smooth layers (or lamina). A turbulent flow
is one in which the fluid flows in an almost chaotic manner with a number of vortices of different size and
nearly random spatial and temporal variations in the fluid velocity. A transitional flow is one between the
laminar and turbulent states where the flow is mostly laminar but with occasional turbulent fluctuations.
Boundary Conditions (BCs)
When solving the governing equations of fluid dynamics, well need to apply boundary conditions (BCs)
for specific flow geometries. Two common types of BCs include kinematic and dynamic boundary
conditions. Kinematic boundary conditions specify the fluid velocity. One example is the no-slip
boundary condition which states that at either a solid boundary or fluid interface, the fluid velocity must be
continuous:
u fluid U boundary
Another common kinematic boundary condition is that fluid velocities must remain finite. Dynamic
boundary conditions specify that stresses must be continuous across solid or fluid interfaces:
nn ,fluid nn ,boundary
ns ,fluid ns ,boundary
where the subscripts nn and ns refer to the normal and shear stresses at the boundary with the normal
vector n .
Now lets investigate some exact solutions.
Note that there are many graduate level texts that review more exact solutions than will be presented in
these notes. Several good references include:
White, F.M., Viscous Fluid Flow, McGraw-Hill.
Panton, R.L., Incompressible Flow, Wiley.
Currie, I.G., Fundamental Mechanics of Fluids, McGraw-Hill.
C. Wassgren
315
2.
Planar Couette-Poiseuille Flow
Consider the steady flow of an incompressible, constant viscosity Newtonian fluid between two infinitely
long, parallel plates separated by a distance, h.
fluid
x
Well make the following assumptions:

u z constant and
1. The flow is planar.
0
z
0
t
u y
u
x
0
x
x
f x 0 and f y g
2. The flow is steady.
3. The flow is fully-developed in the x-direction.

4. The only body force is that due to gravity in the y-direction.
Lets first examine the continuity equation:

u x u y
0
x
y
From assumption #3 we see that:
u y
u x
0
0
x
y
Based on this result and assumptions #1 and #3 we see that the y-velocity is a constant:
u y constant
Since there is no flow through the walls, the y-velocity must be zero.
(call this condition #5)
uy 0
Now lets examine the Navier-Stokes equation in the y-direction:
2u
u
u
2u
u
p
y u x y u y y 2y 2y f y
x
y
y
y
t
x
We can simplify this equation using our assumptions:
2u y
u y
u y
u y
2u y
p
fy

ux
uy
t
x
y
y

x 2
y 2
g (#4)
0 (#2,#5)
0 (#3,#5) 0 (#5)
0 (#3,#5)
0
(#5)
g
y
p x, y f x gy
(7.1)
where f(x) is an unknown function of x.
C. Wassgren
316
Now lets examine the Navier-Stokes equation in the x-direction:

2u 2u
u
u
u
p
x u x x u y x 2x 2x f x
x
y
x
y
t
x
After simplifying:
2
u x
u x
p
u x
ux 2ux
ux
uy
2 2 fx
y
t
x
x
x
y
0 (#4)
0 (#5)
0 (#3)
0 (#2)
0 (#3)
2
u x 1 p
x
y 2
Based on assumptions #1, #2, and #3 we can write:
2ux d 2ux
y 2
dy 2
so that the simplified Navier-Stokes equation in the x-direction becomes:
d 2 u x 1 p
dy 2
x
Integrating twice with respect to y (note that from Eqn. (7.1) we observe that p/x is not a function of y):
1 p 2
ux
y c1 y c2
2 x
where c1 and c2 are unknown constants that we find using our boundary conditions. Note that this is the
equation of a parabola.
Lets examine the following case:
fixed bottom boundary:
ux y 0 0
top boundary moving with velocity, U: u x y h U

After applying these boundary conditions to determine the constants c1 and c2 we find that the fluid velocity
in the x-direction is given by:
y h 2 p y
y
ux U
1
h 2 x h h
This type of flow is often referred to as a planar Couette-Poiseuille flow (pronounced pwz I)
Notes:
1.
The stress acting on the fluid at any point can be found from the stress-strain rate constitutive relations
for a Newtonian fluid.
u u
ij p ij i j
x
j xi
2.
If we remove the pressure gradient and move the fluid using just the moving upper boundary, the
velocity profile becomes linear:
y
ux U
h
This type of flow is referred to as a planar Couette flow.
C. Wassgren
317
3.
If we hold both boundaries stationary and move the fluid using only a pressure gradient (note that flow
in the positive x-direction occurs for dp/dx < 0), the velocity profile becomes:
h 2 p y
y
ux
1
2 x h h
This type of flow is referred to as a planar Poiseuille flow.
4.
The average flow velocity may be found by setting the volumetric flow rate using the average velocity
equal to the volumetric flow rate using the real velocity profile. For example, for planar Poiseuille
flow the average velocity is:
y h
h 2 p y
y
h3 p
Q uh
1 dy

2 x h h
12 x
y 0
u
5.
h 2 p 2
umax
12 x 3
(7.2)
Recall that we assumed that these solutions only hold for laminar flows (the uy component is zero).
Experimentally we observe that planar Couette-Poiseuille flow remains laminar for:
uh
Re
1500
where Re is the Reynolds number of the flow and u is the average flow velocity. It should be noted
that the value of 1500 is only approximate and can vary considerably depending on how carefully the
experiment is conducted. Its value is given only as an engineering rule-of-thumb.
5.
Velocity profiles for the various conditions are sketched below:

U
dp/dx > 0
y
x
C. Wassgren
Couette flow (dp/dx = 0)

dp/dx < 0
318
3.
Poiseuille Flow
Consider the steady flow of an incompressible, constant viscosity, Newtonian fluid within an infinitely
long, circular pipe of radius, R.

1.
The flow is axi-symmetric and there is no swirl velocity.
2.
The flow is steady.
3.
The flow is fully-developed in the z-direction.
4.
There are no body forces.
0 and u 0
0
t
ur u z
0
z
z
f r f f z 0

1 rur 1 u u z
0
r r
r
z
From assumptions #1 and #3 we see that:
rur constant
Since there is no flow through the walls, the constant must be equal to zero and thus:
ur 0
Now lets examine the Navier-Stokes equation in the z-direction:
1 u z 1 2 u z 2 u z
u u u
u
p
u
z ur z z u z z
2 fz
r
2
2
r
r
z
z
z
t
r r r r
2
2
u z
u z u u z
u z
p
1 u z 1 u z u z
r
ur
uz
2 fz

r
r r r r 2
t 0(#5) r
z
z
2
z
0 (#4)
0 (#3)
0 (#1)
0 (#3)
0 (#1)
0 (#2)
d du z r dp
r
dr dr dz
du
r 2 dp
r z
c1
dr 2 dz
r 2 dp
c1 ln r c2
4 dz
Note that in the previous derivation the fact that uz is a function only of r has been used to change the
partial derivatives to ordinary derivatives. Furthermore, examining the Navier-Stokes equations in the r
and directions demonstrates that the pressure, p, is a function only of z and thus ordinary derivatives can
be used when differentiating the pressure with respect to z.
uz
C. Wassgren
319
Now lets apply boundary conditions to determine the unknown constants c1 and c2. First, note that the
fluid velocity in a pipe must remain finite as r0 so that the constant c1 must be zero (this is a type of
kinematic boundary condition). Also, the pipe wall is fixed so that we have uz(r=R)=0 (no-slip condition).
After applying boundary conditions we have:
R 2 dp
r2
Poiseuille Flow in a Circular Pipe
uz
1 2
4 dz R
Notes:
1.
The velocity profile is a paraboloid with the maximum velocity occurring along the centerline. The
average velocity in the pipe is found from:
u
1
R2
rR
u z 2 rdr
r 0
R 2 dp

8 dz
umax
where umax is the maximum fluid velocity.

2.
As with planar Couette-Poiseuille flow, we can determine stresses using the constitutive relations for a
Newtonian fluid. The shear stress that the pipe walls apply to the fluid, w, is:
R dp 4 u
w
2 dz
R
where u is the average velocity in the pipe. Note that an alternate method for determining the average
wall shear stress, which in this case is equal to the exact wall shear stress, is to balance shear forces
and pressure forces on a small slice of the flow as shown below.
dp
dz R 2
p
dz
p R 2
dz
w 2 Rdz
dp 2
dz R w 2 Rdz
0 p R 2 p
dz
R dp
w
(The same answer as before!)
2 dz
z
In engineering applications it is common to express the average shear stress in terms of a (Darcy)
friction factor, fD, which is defined as:
4
64
f D 1 w 2 64
uD Re
2
where D=2R is the pipe diameter and Re is the Reynolds number. The Darcy friction factor commonly
appears in the Moody chart for incompressible, viscous pipe flow. Note again that this solution is only
valid only for a laminar flow. The condition for the flow to remain laminar is found experimentally to
be:
uD
Re
2300
3.
We can also use the general solution (before applying boundary conditions) to determine the flow
between two concentric cylinders by applying different boundary conditions. For example, two fixed
cylinders will have the boundary conditions: uz(r=RI)=0 and uz(r=RO)=0 where RI and RO are the
inner and outer cylinder radii.
C. Wassgren
320
4.
Laminar flow in an elliptical cross-section pipe can be determined by considering the simplified
Navier-Stokes equation in the z-direction but using Cartesian coordinates (assuming ux=uy=0):
2 u z 2 u z 1 dp
(Poissons equation!)
2
x 2
y
dz
where z is the coordinate along the centerline of the pipe. Note that the pipe wall boundary is the
y
ellipse given by:
2
x y
1
a b
2b
2a
where a and b are the lengths of the major and minor axes. Since we must satisfy the no-slip boundary
condition at the pipe walls, lets guess that the solution has the form:
x 2 y 2
u z 1
a b
since this profile automatically satisfies the boundary condition. The quantity is an unknown
constant. To determine if this is indeed a valid solution to the fluid equations, we first note that it
satisfies the continuity equation (ux=uy=0 and uz is not a function of z). If we substitute into the zcomponent of the Navier-Stokes equations (Poissons equation above) we find that our guess for the
velocity distribution is valid if the constant is given by:
2 1 dp
2
2 2
b dz
a

a 2b2
dp
2 a 2 b 2 dz
which means that the velocity profile for an elliptical pipe is given by:
2
2
a 2b2
dp x y
uz
1
2
2
2 a b dz a b
velocity profile in a pipe of elliptical cross-section

For very complex cross-sections, we can determine the velocity profile by solving Poissons equation
numerically.
C. Wassgren
321
4.
Starting Flow Between Two Parallel Plates
Consider a flow starting from rest between two parallel flat plates. The bottom plate is fixed while the top
plate moves impulsively at t > 0 with constant velocity, U. There are no pressure gradients in the flow.
top plate moves with velocity, U
fluid
x
fixed bottom plate

u z constant and
1. The flow is planar.
0
z
u x u y
0
x
x
f x f y =0
2. The flow is fully-developed in the x-direction.

3. There are no body forces.

u x u y u z
0
x
y
z
u y
0
y
Based on this result and assumptions #1 and #2 we see that the y-velocity can be at most a function of time:
uy f t
Since there is no flow through the walls at any time, the y-velocity must be zero.
uy 0
Now lets examine the Navier-Stokes equation in the x-direction:
2ux 2ux 2ux

u x
u x
u x
u x
p
fx
ux
uy
uz
2 2 2
y
x
z
x
x
y
z
0 (#3)
0 (#4)
0 (#2)
0
0 #1
0 #1
0 (#2)
2
u x
ux
t
y 2
The initial and boundary conditions for the flow are:
no flow initially
u x y, t 0 0
(7.3)
(7.4)
no slip at y = 0
u x y 0, t 0
(7.5)
no slip at y = h
u x y h, t 0 U
(7.6)
C. Wassgren
322
Note that as t , the flow profile should approach the Couette flow profile derived previously, i.e.:
y
u x y, t U
(7.7)
h
Hence, lets investigate a solution of the form:
y
u x u x U
(7.8)
h
Substituting back into Eqn. (7.3) and the boundary and initial conditions gives:
u x
2 u x
(7.9)
t
y 2
u x y, t 0 U
u x y 0, t 0
y
h
(7.10)
(7.11)
u x y h, t 0 0
(7.12)
To solve Eqn. (7.3), lets try a separation of variables approach where:

u x y, t Y y T t
(7.13)
so that, upon substitution into Eqn. (7.3), we have:

YT Y T
T
Y
2
(7.14)
T
T
where is a constant since the only way the T and Y sides of the equation can be equal for any t and y is if
both sides are equal to a constant. Solving for each part of the equation gives:
T
2
T t c1 exp 2 t
(7.15)
T
Y
2

Y y c2 sin
y c3 cos
y
(7.16)
Y

Hence the solution has the form:

u x y, t c2 sin
y c3 cos
y c1 exp 2 t

u x y, t exp 2 t c4 sin
y c5 cos
y

(7.17)
In order to satisfy the boundary condition at y = 0 (Eqn. (7.11)), the constant c5 must equal zero and Eqn.
(7.17) becomes:

(7.18)
u x y, t c exp 2 t sin
y

In order to satisfy the boundary condition at y = h (Eqn. (7.12)) without having c4 = 0, we must have:
n
where n is an integer
(7.19)
h
Hence:
t
y
u x , n y, t cn exp n 2 2 2 sin n
(7.20)
h
h
C. Wassgren
323
Since Eqn. (7.9) and the boundary and initial conditions (7.10) - (7.12) are linear, we can add the together
the solutions in Eqn. (7.20) so that they satisfy the given initial condition (Eqn. (7.10)). Note that we can
add together the constants for the negative values of n with the positive values of n since the magnitude of
the exponential and sine terms are identical, i.e.:
2
2
t
y
t
y
c n exp n 2 2 sin n c n exp n 2 2 sin n

h
h
h
h
t
y
c n c n exp n 2 2 sin n
h
h
2
t
y
d n exp n 2 2 sin n
h
h
Furthermore, we neednt include n = 0 since it will give ux,n = 0 = 0 which doesnt contribute to the
summation. Hence, the solution to Eqn. (7.9) subject to the given boundary conditions (Eqns. (7.11) and
(7.12)) is:
t y
u x y, t d n exp n 2 2 2 sin n
(7.21)
h
h
n 1
where the constants dn are found by forcing Eqn. (7.21) to satisfy the given initial condition (Eqn. (7.10)).
Fourier sine series analysis at t = 0 gives the constants as:
yh
2
y
y
2U
2U
n
d n U sin n dy
cos n
(7.22)
1
h y 0
h
h
n
n
Combining Eqns. (7.8), (7.21), and (7.22) gives:
u x y, t
U
y 2 1
t

exp n 2 2 2
h n 1 n
h
y

sin n
h

(7.23)
A plot of the dimensionless velocity profile for various dimensionless times is shown below.
dimensionless position, y /h
1.0
0.9
0.8
0.7
0.6
t ' = t /h
0.5
0.4
t'=0.025
t'=0.100
t'=0.250
t'=0.500
0.3
0.2
0.1
0.0
0.0
0.2
0.4
0.6
0.8
1.0
dimensionless velocity, u x /U
C. Wassgren
324
6.
Starting Flow in a Circular Pipe
Consider the unsteady flow of an incompressible, constant viscosity, Newtonian fluid within an infinitely

1.
2.
3.
0 and u 0
u
u
r z 0
z
z
f r f f z 0

1 rur 1 u u z
0
r r
r
z
rur constant
ur 0
1 u z 1 2 u z 2 u z
u u u
u
p
u
2 fz
z ur z z u z z
r
2
2
r
z
z
z
r
t
r r r r
2
2
u z
u z u u z
u z
p
1 u z 1 u z u z
r
ur
uz
2 fz

r
r r r r 2
z
z
2
z
t 0 (#4) r
0 (#3)
0
(#3)
0
(#1)
0
(#2)
0
(#1)
u z
1 d du z
dp
(7.24)

r
t
dz
r dr dr
The initial and boundary conditions for the flow are:
no flow initially
uz r, t 0 0
no slip
u z r R, t 0
C. Wassgren
(7.25)
(7.26)
325
We know that as t the flow should approach the Poiseuille flow solution found in the previous section,
i.e.:
uz r , t
R 2 dp
r2
1 2
4 dz R
(7.27)
Hence, lets investigate a solution of the following form:

R 2 dp
r2
u z u z
1 2
4 dz R
(7.28)
THE REMAINDER OF THIS DERIVATION IS INCOMPLETE.
C. Wassgren
326
5.
Impulsively Started Flat Plate (aka Stokes First Problem, aka the Rayleigh Problem)
Now lets consider the incompressible, constant viscosity, Newtonian fluid flow resulting from the sudden
movement of an infinitely long flat plate. The geometry of the problem is shown below:
fluid
x
0
U t
U
t 0
t 0
Well make the following assumptions for this unsteady flow:

1.
The flow is planar.
2.
The flow is fully-developed in the x-direction.
3.
0 and u z constant
z
u y
u
x
0
x
x
fx f y fz 0
4.
There is no pressure gradient in the x-direction.
p
0
x
Noting that the continuity equation gives uy=0 and simplifying the Navier-Stokes equations using the given
assumptions we find:
u x
2ux
t
y 2
where =/ is the kinematic viscosity (dimensions of L2/T). The (kinematic) boundary and initial
conditions for this flow are:
u x y 0, t U (for t 0)
(no-slip condition at the plate)
and
u x y , t remains finite
(kinematic condition far from the plate)
u x y, t 0 0
(fluid is initially at rest)
Note that there is no geometric length scale in the problem which suggests that we can use a similarity
variable, , to reduce the number of independent variables from two (t and y) to one (=(y,t)) (i.e.
convert the PDE into an ODE). We may anticipate this reduction in the number of variables by considering
where and when the fluid velocity reaches some value, e.g. ux(y,t)=0.4U. It is reasonable to expect that the
location, y, where the velocity reaches 0.4U will vary depending on t (e.g. the location y gets farther from
the plate as t gets larger). Thus, ux will not depend on the parameters y and t independently but will instead
depend on some combination of y and t.
To determine this combination, lets first re-write the velocity in dimensionless form using the plate
velocity, U:
u
u* x
U
so that the original PDE becomes:
u
2u
u *
2u*
2
t
y
t
y 2
with the boundary conditions:
C. Wassgren
327
0 for t 0
u * y 0, t
1 for t 0
*
u y , t remains finite
Note that since the velocity is dimensionless, it must depend only on dimensionless quantities. The only
dimensional quantities in the PDE are t, y, and . We can form only one dimensionless variable, call it ,
the similarity variable, from these quantities:
y
4 t
Note that a 4 is added to the similarity variable, , for convenience in solving the resulting differential
equation. Thus the dimensionless velocity will be a function only of the similarity variable:
u
u * x f
U
Re-writing the original PDE in terms of this similarity variable gives:
f
t
y
where
f
df
d
and f
d2 f
d 2
and
t
y
2t
4 t
so that the final equation becomes:
1

f f
4t
2t
f 2 f 0
subject to the boundary conditions:
0 for
(Note that the initial condition is subsumed into the condition.)
f
1 for 0
Thus, we see that by using a similarity variable (justified based on dimensional arguments), the PDE with
two independent variables is reduced into a (linear) ODE. Now we must solve the ODE. Fortunately, we
can solve the resulting ODE without much difficulty:
d
f 2 f 0
ln f 2
d
df
c1 exp 2
d
f c1 exp 2 d c2
0
where is a dummy variable of integration. Applying the boundary conditions to determine the constants
c1 and c2:
f 0 1 c2
f 0 c1 exp 2 d 1 c1
0
where the indefinite integral has been evaluated. Thus, the velocity distribution for this flow is given by:
C. Wassgren
328
y
4 t
ux
2
y
1
exp 2 d 1 erf
impulsively started flat plate flow
U
0
4 t
where the integral is also known as the error function (erf).
Notes:
1.
A plot of the flow profile in terms of dimensional quantities:
2.0
t=t1
t=t2
t=t3
1.5
t=t4
1.0
0.5
0.0
0.0
0.5
1.0
u/U
and in dimensionless form:
2.0
=y /(4 t )
1/2
1.5
1.0
0.5
0.0
0.0
0.5
1.0
u /U
C. Wassgren
329
2.
As we can see from the plots shown in Note #1, the effect of the plate diffuses into the remainder of
the fluid. An estimate of the depth of fluid that is affected by the movement of the plate may be found
by determining the distance from the plate, , where the velocity is 1% that of the plate velocity, i.e.
u/U=0.01. This can be found either by estimating from the plot or calculating:
ux
0.01 when 1.8 3.6 t
U
Thus, we see that the thickness of the affected layer is proportional to the square root of the kinematic
viscosity and to the square root of the time (note that it is not a function of the plate velocity, U, or the
absolute viscosity, ). The distance, , which is also referred to as the shear layer thickness, is an
important parameter that gives us an estimate of how far into the flow the effects of the boundary are
felt. We will come across this parameter again, in terms of a boundary layer thickness, in a later set of
notes.
The shear layer thickness after 1 minute in:
air is:
=10.8 cm
(air=0.150 cm2/sec)
water is: =2.8 cm
(water =0.010 cm2/sec)
Thus we see that the effects of a boundary are felt further into a flow of air than into a flow of water in
a given amount of time!
3.
We can also use this solution to examine the flow resulting from a fluid with a uniform velocity U over
a plate that has come to a sudden stop from an initial velocity U. To produce the resulting velocity
profile, we note that this flow can be produced via a Galilean transformation of the problem we just
investigated, (u/U)stopped plate = 1-(u/U)moving plate. The resulting flow profile is:
ux
y
erf
U
4 t
4.
The vorticity in the flow is found via:

u

y
z x U 1 erf
y
y
4 t
y2
U

exp
t

4 t
distance from plate, y
3.0
2.5
t1
2.0
t2
1.5
t3
1.0
t4
0.5
0.0
0.0
0.5
1.0
vorticity, z
Vorticity is created at the wall through the no-slip condition and diffuses through the rest of the fluid
through the action of viscosity.
C. Wassgren
330
6.
Oscillating Flat Plate (aka Stokes Second Problem, aka the Rayleigh Problem)
Consider the incompressible, constant viscosity, Newtonian fluid flow resulting from the sinusoidal
oscillation of an infinitely long flat plate. The geometry of the problem is shown below:
fluid
x
U(t) = Ucos(t)
Well make the following assumptions for this unsteady flow:
0 and u z constant
z
u y
u
x
0
x
x
fx f y fz 0
1.
The flow is planar.
2.
The flow is fully-developed in the x-direction.
3.
Noting that the continuity equation gives uy=0 and simplifying the Navier-Stokes equations using the given
assumptions we find:
u x
2ux
t
y 2
where =/ is the kinematic viscosity (dimensions of L2/T). The (kinematic) boundary and initial
conditions for this flow are:
u x y 0, t U cos t
(no-slip condition at the plate)
u x y , t remains finite
(kinematic condition far from the plate)
u x y, t 0 0
(fluid is initially at rest)
Since the boundary condition is time dependent, we might expect that the fluid velocity will have the
following (separation of variables) form:
u x y, t f y exp i t
where only the real part of the velocity is relevant to the solution.
Substituting into the PDE and simplifying gives:
i f exp it f exp it
f exp it i f exp it 0
f
f 0
Solving for f we find:
i
f y A exp y
i
i
u x y, t A exp y
it
exp it A exp y

which can be simplified to:
C. Wassgren
331

1 i
u x y, t A exp
y
it A exp y
t
exp i y
2
2
2

u x y, t A exp y
t i sin y
t
cos y
2
2
2

u x y, t A exp y
cos t y
2
2
where in the last step of the previous analysis, only the real part of the velocity component is relevant to the
solution. Applying the boundary conditions we find:

flow due to an oscillating plate
u x y, t U exp y
cos t y
2
2
Notes:
1.
A plot of the flow looks like:
y *sqrt( /(2 ))
5.0
wt = 0
4.0
wt = (1/5)*2pi
wt = (2/5)*2pi
wt = (3/5)*2pi
3.0
2.0
1.0
wt = (4/5)*2pi
0.0
-1.0
0.0
1.0
u /U
2.
The velocity decreases exponentially with the distance from the plate. Also note that there is a phase
lag in the velocity profile compared to the plate which is a function of distance from the plate.
3.
The region of fluid affected by the plate can be estimated by determining the y location (y = ) at
which u/U = 0.01. Well also assume the maximum value for the cosine function.
ux

0.01 exp
cos t
U
2
2
ln 0.01
6.51
Again notice that .

4.
We can also use this solution to investigate flow oscillating far from the plate and having a fixed plate
by performing a Galilean transformation on the velocity profile.
C. Wassgren
332
7.
(Planar) Stagnation Point Flow (aka Hiemenz Flow)
Consider the flow in the vicinity of a stagnation point:
y
x
Well make the following assumptions in the analysis of this flow:
0 and uz constant
z
u y
u
x
0
t
t
fx f y fz 0
1.
The flow is planar.
2.
The flow is steady.
3.
Recall from our earlier discussion of potential flows that the complex potential model for this type of flow
is:
f z Az 2 A x 2 y 2 i 2 Axy
where A is a constant that is proportional to the velocity far from the body, U , divided by a characteristic
length of the body, L:
U
A
L
The constant of proportionality depends on the exact shape of the body. The velocity components for the
flow are given by:
f z u x iu y 2 Az 2 Ax i 2 Ay
u x 2 Ax and u y 2 Ay
The pressure distribution at any point (x,y) in the flow is given by:
p( x, y ) p0 1 2 u x2 u y2 p0 2 A2 x 2 y 2
where p0 is the pressure at the stagnation point.
Note that this potential flow solution satisfies our governing equations of fluid dynamics (continuity and
Navier-Stokes) and it satisfies part of the no-slip condition (no flow through the surface). It does not,
however, satisfy the tangential component of the no-slip condition. Thus, the potential flow solution is of
limited use since it wont be a good model close to the plate surface.
To determine a valid solution close to the plate surface, lets try modifying the potential flow model (well
work with the stream function since close to the surface the flow will be rotational) so that it does satisfy
the no-slip boundary conditions. Lets try the following stream function:
2 Axf
u x 2 Axf and u y 2 Af
where f=f(y) and f =df/dy. The function, f, is unknown at this point. Well place several constraints on the
function, f, so that it satisfies the governing fluid equations (continuity and Navier-Stokes) and well also
make sure that far from the plate the flow has the same form as the original potential flow solution.
C. Wassgren
333
We know that since were using a stream function the continuity equation is automatically satisfied. To
make sure we satisfy the momentum equations we substitute the velocity components into the NavierStokes equations (simplified using our assumptions):
ux
ux
2u 2u
u x
u
1 p
uy x
2x 2x
x
y
y
x
x
u y
x
uy
u y
y
2u y 2u y
1 p
2 2
x
y
y
4 A2 x f 4 A2 xff
1 p
4 A2 ff
1 p
2 A xf
2 A f
We need to say something about the pressure distribution before proceeding further. Lets integrate the ycomponent of the Navier-Stokes equations with respect to y:
1 p
4 A2 ff
2 A f
y
p
4 A2 ff 2 A f
y
p x, y 2 A2 f 2 A f g ( x)
where g(x) is an unknown function of x (since p is a function of both x and y). To determine the form of
g(x) we recall that far from the plate the current solution should approach the potential flow solution where
the pressure distribution is given by:
p( x, y ) p0 2 A2 x 2 y 2
2
and the function f(y)y (which gives the original potential flow function). Thus the unknown function
of x should be given by (as y becomes very large):
p x, y 2 A2 y 2 A 1 g ( x) p0 2 A2 x 2 y 2
2
g ( x) p0 2 A2 x 2 2 A
and the pressure distribution becomes:
p x, y p0 2 A2 f 2 A 1 f 2 A2 x 2
2
Substituting this pressure distribution into the x-component of the Navier-Stokes equations gives:
p
-4 A2 x
x
2
4 A2 x f 4 A2 xff 4 A2 x 2 A xf
f ff f 1 0
2A
Thus, the function f must satisfy the non-linear, 3rd order ODE given above in order to satisfy the xmomentum equation (note that weve already established that the original stream function will satisfy the
continuity equation and y-momentum equation). The boundary conditions for the ODE are:
no-slip (x -component): u x x, y 0 0
f y 0 0
no-slip (y -component): u y x, y 0 0
potential flow far from plate: f y y
C. Wassgren
334
f y 0 0
f y 1
Currently the ODE and boundary conditions are in dimensional form. To make the solution to the ODE
general, lets re-write it in terms of dimensionless parameters:
f y
2A
F where y
df
df d
dy d dy
d 2 f d

d 2 dy
2A
2A
F
F
2 A
2A
2A
F
2A
3/ 2
d 3 f d
2A
2A
F
F
f

3
d dy
2A
so that the original dimensional ODE is given in dimensionless form by:
F FF F 1 0
2

F 0 0
F 0 0
F 1
An exact analytical solution to this ODE has not been found so we solve it numerically (using, for example,
a Runge-Kutta numerical scheme). Even though we solve the equation numerically, we still consider the
result an exact solution since we can find the solution numerically to any precision.
The velocity components and pressure distribution are found using the original assumed stream function:
2 A xF
u x 2 AxF
and
u y 2 A F
p x, y p0 A F 2 2 A 1 F 2 A2 x 2
Notes:
1.
Plots of the functions F and F as a function of look like (plot from Panton, R.L., Incompressible
Flow, Wiley.)
= (2A/)0.5y
C. Wassgren
F, F, F
335
2.
Note that for this flow the non-linear convective terms in the Navier-Stokes equations, (u)u, did not
drop out as they have in the previous exact solutions.
3.
In flows around objects with surface curvature (e.g. a torpedo-shaped object), this solution will still be
valid in some vicinity of the stagnation point. As we zoom in very close to the stagnation point, the
local object surface will be approximately flat.
4.
Recall that far from the plate, the viscous flow solution should approach the potential flow solution.
We can estimate this distance by determining the location, y=, at which the x-velocity is 99% that of
the velocity far from the plate at the same x-location, U (for y> the vorticity will be very small since
the velocity gradients are small):
u x 2 AxF and u x U 2 Ax
ux
2A
F 0.99 when
2.4
U
2.4
2A
The distance, , is referred to as the (99%) boundary layer thickness. Note that here the boundary
layer thickness is a constant value and proportional to the square root of the kinematic viscosity. Since
the boundary layer thickness is constant, we can interpret that the shear layer displaces the outer
potential flow a constant distance from the surface. From the plot we note that as , F(-0.65)
(recall that in the potential flow region far from the boundary, F is linear) so that this displacement
thickness, D, is given by:
2A
D
0.65 D 0.65
2A
Well address the concept of a displacement thickness again when discussing boundary layer flow.
5.
The pressure gradients for the flow are given by:

dU
p
(using the U defined in the previous note)
4 A2 x U
dx
x
p
2 A FF F
y
The pressure gradient in the x-direction is the same as that given by Bernoullis equation using the
outer potential flow velocity while the pressure gradient in the y-direction will be small if the
kinematic viscosity is small (F, F , and F are all of order 1 near the surface). Thus, the pressure in
the shear layer is nearly constant in the y-direction and it has the same magnitude as the pressure in the
outer potential flow. This is an important result that will be discussed again when investigating
boundary layer flows.
6.
An exact solution for axi-symmetric stagnation point flow can also be found. Its solution was first
presented by Homann (1936). The approach for the axi-symmetric problem is very similar to what
was presented here for planar flow except a different stream function is used. Refer to White, F.M.,
Viscous Fluid Flow, McGraw-Hill for more details. The resulting velocity profiles, pressure, and shear
stress distributions for the axi-symmetric case are similar to those found for the planar case but with
slightly different magnitudes.
C. Wassgren
336
8.
Very Low-Reynolds Number (aka Creeping, aka Stokes) Flows (Re << 1)
Consider the governing equations for an incompressible fluid, neglecting body forces, in dimensional form:
u 0
(7.29)
u
u u p 2 u
t
Recall that when the Reynolds number is very small, viscous forces dominate the inertial forces. Lets rewrite these equations in dimensionless form keeping in mind that well be investigating flows where
viscous forces dominate (or where the fluid inertia is negligible). The variables in the equations are
normalized in the following manner:
x
u
x*
x Lx *
u*
u Uu *
L
U
(7.30)
p
tU
Lt *
t
p U p *
t*
p*
L
L
U
U
L
where the superscript * refers to a dimensionless quantity and L and U represent, respectively, a
characteristic length and velocity for the flow of interest. Note that the pressure has been made
dimensionless using a characteristic viscous stress, U/L, rather than the usual dynamic pressure term, i.e.
U2. This is because here were investigating flows where viscous forces dominate (or fluid inertia is
negligible). Also note that weve assumed that the time scale is set by the flow velocity and length scale.
This assumption is fine unless there is some other well-defined time scale in the problem such as an
oscillation period (for acoustic applications, for example). Now lets rewrite the governing equations using
these dimensionless parameters.
* u* 0
U 2 u*
U * * U *2 *
*
*
*
* u u 2 p 2 u
L t
L
L
Dividing through by the characteristic viscous force term gives:

UL u*
*2 *
*
*
*
* *
* u u p u
u*
u* * u* * p* *2 u*
*
Re
(7.31)
(7.32)
where Re is the Reynolds number which is a ratio of typical fluid inertial forces to viscous forces in a flow.
If the viscous forces dominate, then the Reynolds number should be small. For creeping flows we
investigate the limit when Re0, i.e., the fluid has no inertial terms. Thus, for creeping flows the
governing fluid equations simplify to:
* u* 0
(7.33)
* p* *2 u*
or in dimensional form:
u 0
Stokes Equations
(7.34)
p 2 u
Note that does not appear in Eqn. (7.34) indicating that Stokes flows behave the same regardless of the
surrounding fluid density!
C. Wassgren
337
Two additional useful relations can be found if we take the curl of both sides of the momentum equation:
p 2 u
2 0
and if we take the divergence of both sides of the momentum equation (and using continuity):
p 2 u
(7.35)
2 p 0
Both the vorticity and pressure fields satisfy Laplaces equation for a creeping flow.
(7.36)
Notes:
1.
Examples where creeping flow might occur (Re << 1):

a. small length dimensions (flow in small pipes or channels, around small particles, flow through
small pores)
b. very viscous fluids
c. small velocities
A good reference for this topic is Happel and Brenner (1965).
2.
Since Laplaces equation is a linear PDE, we can add together solutions to form new solutions (the
principle of superposition). This is very similar to what we observed in potential flows where we
could add together valid velocity fields (in the form of a potential function) to form new velocity
fields. The difference however is that here we can also add together pressure or vorticity fields. Note
that in potential flows we couldnt add together pressure fields since the pressure was found using the
non-linear Bernoullis equation.
3.
If we consider a 2D flow and use a stream function to describe the velocity field, we find that the
vorticity can be written in terms of the stream function as:
u
u
2 2
z y x 2 2 2
(7.37)
x
y
x
y
2 2 4 0
governing equation for a 2D Stokes flow
(7.38)
(In 2D Cartesian coordinates: 4
4
x 4
4
x 2 y 2
4
y 4
.)
This is referred to as the biharmonic equation and is a common equation found other fields of study
(e.g. solid mechanics where the Airy stress function is used to solve plane problems in elasticity).
4.
The pressure increases proportionally with the dynamic viscosity of the fluid assuming that the
viscosity is independent of pressure (refer to Eqn. (7.34)). Note that when the pressures become very
large, such as in lubrication flows, the viscosity becomes a function of pressure (recall that the
viscosity is also a function of temperature).
5.
There are several approaches to finding solutions to creeping flow problems. These include:
a. forming building block solutions that we can add together to form new solutions (we used a
similar approach with potential flow problems) (Note that we can add together pressure fields
since Eqn. (7.36) is linear! We cant do this for potential flows since Bernoullis equation is nonlinear in terms of the velocities.)
b. solving the boundary value problem for the given geometry and boundary conditions,
c. borrowing solutions from other disciplines which have the same governing equations (e.g. the
Airy stress function for solid mechanics), and
d. using computational methods to solve the governing equations.
C. Wassgren
338
6.
Note that if u is a Stokes flow solution, then u = -u is also a solution since:

u 0 and p 2u p p
In addition, 2 0 and 2 p 0 where u u .
Hence, Stokes flows are kinematically reversible and flow around symmetric objects will produce
symmetric streamlines.
C. Wassgren
339
9.
Stokes Flow Around a Sphere
Now lets examine the creeping flow around a sphere of radius, R, in a uniform stream of velocity, U. For
axi-symmetric creeping flows it is convenient to use a stream function in spherical polar coordinates, (r, ,
), to describe the fluid velocity. The angle is zero when aligned with the incoming free stream. Since
the flow is axi-symmetric, the stream function will be a function only of r and . After substituting the
stream function into the biharmonic equation (in spherical coordinates and noting that for an axi-symmetric
problem there is no variation in the -direction):
r
4 0
2
2
1 2 cot
U
2 2
0
2
2
r
r
r
The velocity components are related to the spherical stream function by:
R
1
1
and u
ur 2
r sin r
r sin
These forms of the velocity in terms of the stream function can be verified by substituting them into the
(incompressible) continuity equation in spherical polar coordinates (recall that the stream function is
defined such that it automatically satisfies the continuity equation):
1 2
1
u sin 0
r ur
2
r sin
r r
1 2
1
1
r 2
r 2 r
r sin r sin
ur
1 1
1
r 2 r sin r sin
1
r 2 sin
1
2
r
sin
1
sin
r
r
sin

r r
0
r
The no-slip condition at the surface means that:

u r r R u r R 0
r R 0 and r R 0
r
and far from the sphere, as r, the stream function approaches the stream function for a uniform stream:
r2
r Usin 2 constant
2
ur U cos and u U sin
Solve the differential equation with the given boundary conditions using separation of variables. Based on
the form of the stream function far from the origin, lets assume that the solution has the form: =f(r)sin2
2
1 2 cot
2
2
2 2
f r sin 0
2
r
r
r
After simplifying we get:

2
d2
2
2 2 f r 0
r
dr
In trying to solve this ODE, lets try a solution of the form: f=rn:
C. Wassgren
340
d2
2 n
n4
2 2 r n 2 n 3 2 n n 1 2 r 0
r
dr
n 1,1, 2,3
A
f (r ) Br Cr 2 Dr 3
r
The corresponding stream function and velocities are:
A
(r , ) Br Cr 2 Dr 3 sin 2
r
A B
ur 2 3 C Dr cos
r
r
A B
u 3 2C 3Dr sin
r
r
Applying the boundary conditions we find that:

UR 3
3UR
U
A
B
C
D0
4
4
2
Consequently:
R 2U R 3r 2r 2 2
(r , )

sin
4 r R R2
R 3 3R
ur U cos 1 3
2r
2r
R 3 3R
u U sin 1 3
4r
4r
The pressure, found using the momentum equation (p = 2u), is:

2
3 U R
p p
cos
2 R r
The viscous stresses are found using the constitutive relations for a Newtonian fluid:
u
rr 2 r
r
1 u ur
2

r
r
1 u ur u cot

r
r
r sin
r r
r r
1 ur
r
u
1 ur
r
r r
r sin
sin
r
u
1 u
sin r sin
Evaluating at the spheres surface (r=R) gives:
rr r R r R r R r r R r R 0
r R
3 U
sin
2 R
C. Wassgren
341
The drag force acting on the sphere surface (r=R) is found by integrating the pressure and viscous forces in
the horizontal direction over the entire spheres surface:
r r R sin
R
dr R cos d
F r r R sin dA p r R cos dA where dA 2 R 2 sin d
dr
dA 2 r
2 R 2 sin d
0
0
cos
F 4UR 2UR
Thus, the total force acting on the sphere consists of 2/3 viscous force and 1/3 pressure force giving a total
force of:
F 6UR
Stokes Drag
Notes:
1. Stokes drag is strictly valid only when Re0 but it is found experimentally to be a reasonable estimate
up to Re=1.
2.
Note that the drag is independent of the fluid density and is proportional to the velocity (and not
velocity squared).
3.
Stokes drag is usually presented in dimensionless form as a drag coefficient, cD. The usual form of this
is:
F
6UR
12
24
cD
1 U 2 R 2
1 U 2 R 2
2 UR UD
2
where D=2R is the sphere diameter. Note that the expression can be simplified further by using the
Reynolds number based on the sphere diameter:
24
cD
Stokes Drag Coefficient
Re D
4.
Oseen (1910) included first-order inertial effects in the drag analysis and found a drag coefficient of:
24
3
cD
1 Re D
Re D 16
This drag coefficient is found to give good results up to ReD 5
5.
Although the streamlines for flow around a sphere look similar between a potential flow and a Stokes
flow (in the spheres frame of reference (FOR)), there are some important differences. The streamlines
for potential flows are grouped closer together near the sphere than they are for a Stokes flow. More
strikingly, if we plot the streamlines using a frame of reference where the fluid is at rest far from the
sphere (and the sphere moves with a velocity -U), we find that in a potential flow the fluid is pushed
out of the way while in a Stokes flow the fluid is dragged along with the sphere.
FOR fixed to cylinder
FOR fixed to ground
creeping flow
potential flow
C. Wassgren
342
6.
We can also use the solution approach presented here to determine the drag on a spherical droplet of a
fluid (with dynamic viscosity i) in a different fluid (of dynamic viscosity o). The general differential
equation is the same but the boundary conditions are different. For the spherical droplet problem, the
boundary conditions at the sphere radius consist of continuous velocity components (no-slip but the
tangential velocity is not zero) and stresses between the droplet fluid and the outer fluid. The resulting
drag force acting on the droplet becomes:
1 2 o 3 i
F 6 R oU
1 o i
For i >> o (e.g. solid droplet in a gas or liquid) we get the original Stokes drag
equation: F 6oUR
For i << o (e.g. gas bubble in a liquid) we get a smaller drag force (since the outer fluid can slip at
the boundary surface): F 4 R oU
7.
It can be shown that the drag on an irregular object in a Stokes flow is bounded by the drag on a
sphere that inscribes the object and the drag on a sphere that circumscribes the object (refer to Hill and
Power, 1956). This is a useful result for practical applications.
Dinscribed sphere < Dirregular object < Dcircumscribed sphere
8.
An interesting observation, referred to as the Stokes Paradox, can be made using dimensional analysis.
Assuming that inertia is negligible for a Stokes flow, the force, F, on an object is:
F fcn , U , L
(7.39)
where , U, and L are the dynamic viscosity, velocity, and characteristic length for the flow. For a 2D
flow, the dimensions of the force will be F/L (force per unit depth) while for a 3D flow the dimension
of the force will simply be F. Thus, from dimensional analysis:
F
(7.40)
F2D 2 D constant
U
F
(7.41)
F3D 3 D constant
UL
where F is the dimensionless force. Equation (7.40) indicates that in a 2D Stokes flow the force on an
object is independent of the object size. This contradicts what we observe in reality. Hence, our initial
assumption that the fluid inertia is negligible must be incorrect. For 2D flows, the density must be a
factor in determining the force on an object:
F2 D fcn , , U , L
(7.42)
F2D
F2 D
UL
fcn
C. Wassgren
(7.43)
343
10. Lubrication Flow
One very important application of creeping flows is in the study of lubrication problems. Lets consider the
example of a simple, stationary, planar slipper pad bearing as shown in the figure below:
L
stationary slipper pad bearing
p
flow direction
h0
y
h1
bottom plate moving

at constant horizontal velocity
U
To analyze this flow, lets examine the typical magnitudes of various terms in the Navier-Stokes equations.
First lets consider the x-component of the Navier-Stokes equations for the steady flow of an
incompressible fluid with negligible body forces (the gravitational body force term in lubrication problems
is typically very small in comparison to the other term and so is neglected):
2u 2u
u
u
p
u v 2 2
(7.44)
y
x
y
x
x
Characteristic magnitudes:
u ~ U x ~ L y ~ h0
The characteristic y-velocity can be determined from the continuity equation:
Uh
u v
v
u U
v~ 0
~
x y
y
x L
L
(7.45)
(7.46)
If we examine the magnitudes of the convective inertial forces we find:

u U 2
u U 2
u ~
and v
~
(7.47)
L
L
x
y
The magnitudes of the viscous forces are:
2 u U
2 u U
2 ~ 2
2 ~ 2
and
(7.48)
L
h0
x
y
Since were investigating flows where h0/L<<1, the second term will dominate the magnitude of the
viscous forces.
Lets examine the case where convective inertial terms can be neglected in comparison to the viscous terms
( a creeping flow):
U 2
U
2
L
h0
UL h0
1
L
(7.49)
h
Re L 0 1
L
for a creeping flow
(7.50)
To check the value of this ratio for a typical lubrication problem, consider the following parameters:
U=10 m/s, L=4 cm, h0=0.1 mm, =5*10-4 m2/s (SAE 30 oil)
gives:
ReL=800 but ReL(h0/L)2=0.005
C. Wassgren
344
Thus, this flow could be considered a creeping flow.

Using the simplifications just discussed, the Navier-Stokes equation in the x-direction reduces to:
p
2u
2
x
y
Note that we expect the magnitude of the pressure gradient in the x-direction to be of the order:
p U
~ 2
x
h0
based on the previous dimensional arguments.
(7.51)
(7.52)
Considering the Navier-Stokes equation in the y-direction we find that the pressure gradient in the ydirection should be of the order:
p U
p U
~
~ 2
(7.53)
y
h
L
h0
0
x
where h0/L<<1 has been assumed. Thus, it is reasonable to assume that the pressure remains essentially
constant in the y-direction (in comparison to how the pressure changes in the x-direction).
Solving the differential equation given in equation (7.51) gives:
1 p 2
u
y c1 y c2
2 x
Applying no-slip boundary conditions at the top and bottom walls:
u y 0 U and u y h 0
(7.54)
(7.55)
results in the following velocity profile:

1 p 2
y
u
y yh U 1
h
2 x

(7.56)
Couette Flow
Poiseuille Flow
We see that the velocity profile is a combination of a Poiseuille flow and a Couette flow.
We now have a relation relating two unknowns, the velocity profile and the pressure distribution. We must
use another relation solve for the pressure distribution (or velocity distribution) in terms of known
quantities. So far weve used the momentum equations to derive equation (7.56), now lets consider the
continuity equation; specifically, the mass flow rate at any cross-section must remain the same:
y h x
m
(7.57)
udy 0
x x y 0
(Note that the fluid was assumed incompressible when writing Eqn. (7.57).) Substituting in for the velocity
using equation (7.56) gives:
y h x
y 0
udy
y h x
y 0
1 p 2
y
2 x y yh U 1 h dy 0
h
1 p h h3
U h 0
2
x 2 x 3 2
p 3
h
h 6U
x x
x
Reynolds Equation for lubrication in a planar channel
C. Wassgren
345
(7.58)
Notes:
1.
We can use Reynolds Equation to solve for the pressure distribution, p(x), assuming we know the
bearing geometry, h(x). Lets consider the simple example using the slipper pad bearing design shown
in the figure below:
L
stationary slipper pad bearing
p
h0
bottom plate moving

at constant horizontal velocity
h1
U
The bearing geometry, which is a straight line, is described by:
x
(7.59)
h x h0 h1 h0
L
Substituting into equation (7.58) and solving for the pressure gradient (using the boundary conditions:
p(x=0) = p(x=L) = p) gives:
6 x L 1 x L 1 h1 h0
p p
(7.60)
2
UL
h1
h1
x
2
1
1
1
h0
h0
L
h0
A plot of this dimensionless pressure distribution is shown below.
4.0
(p -p inf)/(UL /h 02)
3.5
As h1/h0 1, the
flow approaches a
Couette flow.
3.0
2.5
h1/h0=0.90
h1/h0=0.75
2.0
h1/h0=0.50
h1/h0=0.25
1.5
1.0
0.5
0.0
0.0
0.2
0.4
0.6
0.8
1.0
x /L
The magnitude of the maximum pressure can be quite large. Consider for example the following
parameters (the same parameters as used previously):
U=10 m/s, L=4 cm, h0=0.1 mm, =4*10-1 N-sec/m2 (SAE 30 oil)
C. Wassgren
346
UL
1.6 MPa 160 atm!

h02
A more accurate analysis of the flow would also include variations in the fluid viscosity due to the
large pressure variations.
2.
Note that a truly symmetric bearing and flow would result in zero lift on the bearing since the pressure
increase on the upstream side of the bearing would be offset by an equivalent pressure reduction on the
downstream side (the solid line in the drawing below).
pgage
x
The reason real bearings can support a load is because the pressure distribution is, in fact, not
symmetric. When a liquid flows past the centerline the pressure begins to decrease below zero gage
pressure as shown in the figure above. However, the minimum pressure is limited by the cavitation
(i.e. vapor) pressure of the liquid (dashed line in the figure above). Hence, there is a net positive gage
pressure force acting on the bearing surface.
C. Wassgren
347
Review Questions
1. Describe several common assumptions used to simplify the Navier-Stokes equations.
2. Describe several common boundary conditions used when solving the Navier-Stokes equations.
3. At what (rule of thumb) Reynolds number does transition from laminar to turbulent flow occur for
planar Couette flow?
4. At what (rule of thumb) Reynolds number does transition from laminar to turbulent flow occur for
Poiseuille flow?
5. Sketch the velocity profiles for a planar Couette-Poiseuille flow with different pressure gradients.
6. What is meant by the shear layer thickness?
7. How does the shear layer thickness typically depend on the kinematic viscosity for laminar flows?
C. Wassgren
348
Chapter 08:
Boundary Layers
1.
2.
3.
4.
5.
6.
7.
8.
9.
Boundary Layer Introduction

Boundary Layer Equations
Solution for Laminar Boundary Layer Flow Over a Flat Plate (Blasius Solution)
Falkner-Skan Solutions
Approximate Methods: The Krmn Momentum Integral Equation
Approximate Methods: Thwaites Correlation
Turbulent Boundary Layer over a Flat Plate
Boundary Layer Separation
Forces on Objects Immersed in a Fluid Flow
C. Wassgren
Chapter 08: Boundary Layers
349
1.
Boundary Layer Introduction
Boundary layers are the regions near a boundary in which rotational effects are significant. The various
flow field regions are indicated in the figure below.
outer, potential flow
(irrotational flow, negligible
viscous effects)
typical streamline
wake region
(viscous forces negligible since
velocity gradients are small,
vorticity is convected in from the
boundary layer)
boundary layer thickness
inner, boundary layer flow
(strong viscous effects,
no-slip condition
velocity gradients0
vorticity0)
C. Wassgren
350
boundary layer
separation point
Boundary Layer Thickness Definitions

Before continuing further, we should define what we mean by the thickness of a boundary layer. There
are three commonly used definitions:
1.
(99%) boundary layer thickness, :

This thickness definition is the most commonly used definition. The boundary layer thickness, , is
defined as the distance from the boundary at which the fluid velocity, u, is 99% that of the outer
velocity, U:
u y 0.99U
(8.1)
y
U
u=0.99U
2.
displacement thickness, D or *:
The displacement thickness, D, is the distance at which the undisturbed outer flow is displaced from
the boundary by a stagnant layer of fluid that removes the same mass flow as the actual boundary layer
profile.
y
mass flow deficit
effective displaced
boundary
U u dy
y 0
mass flow deficit

U D
u
U u dy U
D 1
u
u
dy 1 dy
U
U
0
C. Wassgren
(8.2)
351
Example:
Consider the mass flow rate between two parallel plates in which a boundary layer has formed:
free stream
velocity, U
H/2
H/2
boundary layer velocity, u(y)

core velocity, U
Lets just consider the lower half of the flow since the upper and lower halves are symmetric. The
mass flow rate in the lower half is:
H
udy
m 1 2
U dy
mass flow rate in BL
mass flow rate in outer flow
Note that the uniform, outer flow velocity at this cross-section is U which is larger than U (to conserve
mass). The second integral can be re-written as:
H
U dy
U dy U dy
0
so that the mass flow rate is:

H
m 1 2 udy
0
U dy U dy u U dy U dy
0
U 1 dy U H 2
U
0
m 1 2
D U
Thus, if we replace the real velocity profile by a uniform velocity profile at that cross-section (which
has velocity U), we must decrease the effective cross-sectional area (per unit depth) of each half of the
pipe at that particular location by the displacement thickness, D, to maintain the same mass flow rate.
Note that we would need to have another relation for D in order to solve for U and visa-versa.
C. Wassgren
352
3.
momentum thickness, M or :
The momentum thickness, M, is the thickness of a stagnant layer that has the same momentum deficit,
relative to the outer flow, as the actual boundary layer profile. The concept is similar to that for the
displacement thickness except instead of a mass deficit, the momentum thickness considers the
momentum deficit.
momentum deficit
y
effective displaced
boundary
u U u dy
y 0
2
u U u dy U M
momentum deficit
U 2 M
u
u
u
u
1 dy 1 dy
U
U
0U
0U
(8.3)
Example:
Consider the boundary layer flow over a flat plate. Well assume a steady flow with the pressure
everywhere equal.
top of CV follows a streamline so that there is
y
no flow out from the top of the CV
U
U
u
x
D
x
Lets apply the LMEs in the x-direction to the CV shown to determine the drag acting on the fluid (or
the plate) over the distance x.
d
2
2
D
u
dV
u
d
D
x
CS x rel
0 u dy U h
dt CV
where the height, h, is found via COM on the same CV to be:
u
Uh udy h dy
U
0
0
so that the drag is:
u
u
u
D u 2 dy U 2 dy
D U 2 1 dy
U
U
U
0
0
0
D U M
2
(8.4)
Thus we see that the drag is related to the momentum thickness. Note that this particular example is for
a situation with no pressure gradients (the pressure is a constant). The drag expression will be different
if the pressure varies for the flow but the resulting drag expression will still be in terms of the
momentum thickness.
Notes:
1. Usually: D M
C. Wassgren
353
2.
Boundary Layer Equations
To determine the boundary layer velocity profile for a given flow, we need to go back to the governing
equations of fluid motion. Lets consider steady, incompressible, laminar flow over a sharp-edged flat plate
as shown in the figure below.
y
U
u
x
x
In the following analysis, well assume that the boundary layer thickness, , is much smaller than the
distance from the leading edge of the plate, x:
To help simplify the governing equations, lets make some estimates of the magnitudes of some of the other
parameters:
u ~U
x~x
y~
From continuity:
u v
v u U
U
~
~
0
v~
x
x y
y x x
We can use these estimates in the momentum equations (Navier-Stokes equations) to determine if any
terms are small in comparison to the other terms in the equations (note that body forces are neglected in the
following equations since they are typically very small in comparison to the other terms).
2u 2u
1 p
u
u
x-dir: u
v
2 2
x
y
y
x
x
y -dir:
2v 2v
1 p
v
v
v
2 2
x
y
y
y
x
x-dir:
U2 U2
1 p U U
x
x
x x 2 2
y -dir:
U2 U2
1 p U U

x x x x
y x 2 x 2 x
Since were assuming that <<x, we see that the second viscous term in the x-momentum equation is much
greater than the first viscous term:
U
U
2
2
x
so that the first viscous term can be neglected. Also, since the fluid in the boundary layer is being
accelerated in the x-direction despite strong viscous forces, well assume that the inertial terms are of the
same order as the viscous term:
C. Wassgren
354
x
U 2 U
~ 2 ~
U
x
Thus, we expect the boundary layer thickness to scale with

Furthermore, since we assumed that <<x, we have:
x for a laminar flow over a flat plate.
1 Re x 1
(8.5)
x
Ux
The assumption that <<x is the same as saying that the Reynolds number based on the x-position must be
much greater than one.
1
Now lets examine the magnitude of the pressure gradient in the x-direction to the magnitude of the
pressure gradient in the y-direction. From the x-momentum equation we see that:
p U 2 U
~
~ 2
x
x
From the y-momentum equation we see that:

p U 2 U
~
~

y
x x 2 x
Since <<x we see that:
p
p
x
y
so that the pressure remains essentially constant in the y-direction in comparison to how the pressure
changes in the x-direction, i.e. p=p(x). Thus, we can use the pressure in the outer, potential flow,
determined using Bernoullis equation, to determine how p varies with x (since the pressure is essentially
constant in the y-direction):
p
U
p 1 2 U 2 constant
U
x
x
Substituting this relation into the x-momentum equation and simplifying gives the following:
u
u
U
2u
v
U
2
x
y
x
y
u v
continuity:
0
x y
boundary conditions:
momentum: u
assuming:
u x, y 0 0
(no-slip at boundary)
v x, y 0 0
(no flow through boundary)
u x, y U
(inner flow matches outer flow far from wall)
Re x
Ux
These are known as the laminar boundary layer equations.

Note that U is the outer flow velocity, i.e. the velocity just outside the boundary layer. It is not necessarily
the upstream velocity, U.
C. Wassgren
355
3.
Exact Solution to Laminar Boundary Layer Flow over a Flat Plate with no Pressure Gradient
(aka the Blasius Solution)
Lets consider steady, incompressible, laminar flow over a sharp-edged flat plate as shown in the figure
below.
y
U
u
x
x
From the previous notes, the boundary layer equations are:
u
u
U
2u
v
U
2
x
y
x
y
u v
continuity:
0
x y
momentum: u
assuming:
u x, y 0 0
v x, y 0 0
u x, y U
Re x
Ux
For flow over a flat plate, the outer potential flow velocity, U, will remain constant if the displacement
thickness remains small (Rex>>1 so that the outer flow is not perturbed much):
U
p
U constant
0
0
x
x
so that the boundary layer equations simplify to:
u
u
2u
momentum: u
v
2
x
y
y
u v
continuity:
0
x y
assuming:
u x, y 0 0
v x, y 0 0
u x, y U
Re x
Ux
Notice that there is no characteristic length scale in the x-direction so that we might expect that the flow
profiles will have similar shape (but scaled in magnitude) as we move downstream. This being the case,
lets look for a solution to the boundary layer equations of the form:
u
y
f
U

C. Wassgren
356
where the boundary layer thickness, , for this laminar flow will scale with (refer to the previous notes):
x
~
U
Thus, lets try looking for a similarity solution to the original PDEs using the similarity variable, , defined
as:
u
U
f where y
U
2 x
Note that a factor of 2 has been included in the similarity variable merely for convenience. It will make
the resulting differential equation a little easier to work with. One other simplification we can make since
we are investigating a planar flow is to write the velocity components u and v in terms of a stream function:
d
u
Uf
y
d y
d
y
U
f
d
2 x
f d constant
U
Since the constant is arbitrary (we really only care about the velocities so it doesnt matter what the
constant is), set it equal to zero. The resulting stream function becomes:
2 UxF
= U
(8.6)
where F f . The velocities are found from the stream function:

U
2UxF
2UxF
2 x
y
y
u UF
u
x
2
2U
1
F 2UxF
x
x
2
(8.7)
2U
1
U
F
2UxF y
x
2
2 x3
U
F F
2x
(8.8)
Note that f = F . Also,

UF y
U
U
u UF
F
3
2
2x
x
x
2 x
U
u UF
UF
2 x
y
y
2u
y 2
2 UF
U
UF
2 x
2 y
2
U
v
U
F F F
y 2 x
y 2 x
C. Wassgren
357
The continuity equation is automatically satisfied after substitution (as expected since a stream function has
been used) and the original momentum PDE becomes:
U
U
U
U
UF ' F
F F UF
UF
2x
2 x
2 x
2x
F FF 0
and the boundary conditions become:
u x, y 0 0
F 0 0
v x, y 0 0
F 0 0
u x, y U
F 1
(8.9)
(8.10)
In summarizing these results, we see that the original boundary layer PDEs for laminar flow over a flat
plate with no pressure gradient can be simplified to a non-linear ODE by using a similarity variable. The
resulting equations and boundary conditions become:
F FF 0
F 0 0
F 0 0
F 1
Blasius Equation for Boundary Layer Flow over a Flat Plate
u UF '
v
2x
F F
U
2 x
C. Wassgren
358
Notes:
1.
There is no known closed-form solution to this ODE so we resort to solving it numerically (using a
Runge-Kutta method for example). A plot of the solution looks like (plot from Panton, R.L.,
Incompressible Flow, 2nd ed., Wiley):
u
U
U
x
The analytical results found here match experimental data very well as shown in the figure. Note that
the similarity variable in the plot above does not include the square root of two term, i.e.:
plot 2 what's used in these notes
2.
The boundary layer thickness, , is found from the numerical solution to occur at =3.5:
u
F ' 3.5 3.5 0.99
U
The boundary layer thickness is then:
U
5.0
3.5
1
x Re x2
2 x
3.
The displacement and momentum thicknesses can also be found numerically using their definitions:
D 1.72
M 0.664
and
1
1
2
x
x
Re x
Re x2
4.
The shear stress at the plate surface in dimensionless form is known as the friction coefficient, cf:
d u U
U
2 F
yx
d y y 0
0.664
y 0
0
cf
1
1 U 2
1 U 2
Re x
Re x2
2
2
Note that the friction coefficient is a ratio of the shear stress to the dynamic pressure in the flow.
C. Wassgren
359
5.
The drag coefficient, cD, defined as the dimensionless drag acting on the plate between x=0 and x=L,
is found by integrating the shear force over the plate area:
xL
cD
x 0
yx y 0
dx
1.328
(where cD is the drag coefficient per unit depth)

1
Re L2
Note that although the boundary layer assumptions break down near the leading edge of the plate (Rex
is not >> 1), the distance over which this is the case is small in comparison to the typical lengths of
interest. This discrepancy is generally neglected in engineering applications.
6.
2 U
This solution is only valid for laminar boundary layers. As a rule of thumb, the transition from laminar
flow to turbulent flow occurs at: Rex 500,000.
7. Recall that he boundary layer equations on which the Blasius solution is based are valid only when Rex
>> 1. In practice, it has been found that the Blasius solution is accurate when ReL > 1000. For 1 ReL
1000, the following relation developed by Imai (1957) is more appropriate:
1.328 2.3
cD
1
Re L2 Re L
8.
In summary, for flow over a flat plate with no pressure gradient, the exact Blasius solution gives:
D 1.72
M 0.664
5.0
1
1
1
x Re x2
x
x
Re x2
Re x2
cf
0.664
1
Re x
cD
1.328
1
Re L2
Re x 106
C. Wassgren
360
Example:
A thin flat plate 55 by 110 cm is immersed in a 6 m/s stream of SAE 10 oil at 20 C. Compute the total
skin friction drag if the stream is parallel to (a) the long side and (b) the short side.
SOLUTION:
U = 6 m/s
W
SAE 10 oil = 1.20*10-4 m2/s

SAE 10 oil = 870 kg/m3
Determine the Reynolds number at the trailing edge of the plate to see if its laminar.
UL
Re L
(The flow is considered laminar if Re < 500,000.)
(8.11)
When L = 1.10 m then ReL = 55,200 laminar flow. When L = 0.55 m then ReL = 27,500 laminar
flow.
Now determine the drag on the plate using the drag coefficient, cD, for laminar flat plate flow (the Blasius
solution).
D 12 U 2 2 LW cD
top and bottom

faces
1.328
D 12 U 2 2 LW
1
2
Re L
(8.12)
When L = 1.10 m, W = 0.55 m, ReL = 55,200, and D = 107 N.

When L = 0.55 m, W = 1.10 m, ReL = 27,500, and D = 152 N.
Note that the drag is greater when the short side is aligned with the flow. Why? Because from Eqn. (8.12)
we observe that the drag varies with L but is proportional to W. Hence the drag will increase more
rapidly with increasing width than with increasing length.
C. Wassgren
361
4.
Falkner-Skan Boundary Layer Solutions
Recall that a similarity approach was used to find solutions to stagnation point flow (Hiemenz Flow) and
boundary layer flow over a flat plate (Blasius solution). Falkner and Skan (1931) investigated what other
flows can be solved using similarity solutions. In their analysis they assumed that:
(8.13)
u x, y U x f
where
(8.14)
Here, u is the velocity within the boundary layer, U is the outer flow (potential flow) velocity, is the
similarity variable, and is a distance scaling function. Note that at this point, U and are unknown
functions of x.
Its convenient to work in terms of a stream function rather than velocity components so lets re-write the
velocity in terms of a stream function, :
y
U x f U x f
u
d constant
y
U x x f
(8.15)
Note that the constant in the previous equation has been set to zero since only differences (or derivatives) in
the stream function have any significance. Note also that the stream function automatically satisfies the
continuity equation. The stream function must also satisfy the boundary layer momentum equation:
u
u
U
2u
u
v
U
2
x
y
x
y
(8.16)
2
2

U
3
U
3
y xy x y 2
x
y
where
d
y d dU
d
d
dU
f U
f U f 2
f U
f U
f
dx
dx
dx
dx
x
dx dx
1
U f Uf
dU
y d
dU
U d
2
f
f U 2
f
f
dx
dx
dx
xy dx
2
y
3
y
(8.17)
Uf
Uf
C. Wassgren
362
Substituting and simplifying

dU
dU
U d
d
d
dU
Uf
Uf
2
f U
f
f U
f
Uf f
U
dx
dx
dx
dx

dx
dx
U
dU
U 2 d
dU
U 2 d
U 2 d
dU
Uf
f f U
ff
ff
f f U
2
f 2
dx
dx
dx
dx
dx
dx
U d
U ff
dx
dU
U d
dU
Uf
f 2
U ff U 2
dx
dx
dx
(8.18)
Multiply through by 2/(U) and re-arrange to get:

d
2 dU
2
f
1 f 0
U ff
dx
dx
(8.19)
If a similarity solution exists, Eqn. (8.19) should be an ODE for the function f in terms of . Thus, the
coefficients in front of the second and third terms should be, at most, constants, i.e. for a similarity solution
to exist, we must have:
d
2 dU
U and
dx
dx
where and are constants.
(8.20)
Note that these two equations can be combined to form an expression that is more convenient to work with
later in our analysis:
d
d
dU
U 2 2U
2
dx
dx
dx
dU 2 dU
d
2 U
dx
dx
dx
d
dU
2
U 2
dx
dx
U 2 2
dx
(8.21)

2
f ff 1 f 0
(8.22)
The boundary conditions for this ODE are:

u x, y 0 0 f 0 0
v x, y 0 0
f 0 0
u x, y U f 1
C. Wassgren
(8.23)
363
The procedure for determining the exact solutions to the boundary layer equations can be found via the
following procedure:
a. Select values for the constants and . Note that at this point we dont yet know what geometry
we are investigating. The geometry will be determined in the next step.
b. Determine the corresponding form for U(x) and (x) using Eqn. (8.20), or, more conveniently, part
of Eqn. (8.20) and Eqn. (8.21):
d
dU
U 2 2 and 2

(8.24)
dx
dx
c. Determine the function f() from the ODE given in equation (8.22) subject to the boundary
conditions specified in equation (8.23). This part is usually solved numerically.
2
(8.25)
f ff 1 f 0
f 0 0
f 0 0
(8.26)
f 1
d. Determine the stream function from Eqn. (8.15) and velocity components from the stream function.
The wall shear stress may also be determined.
U x x f
(8.27)
u
Uf
y
(8.28)
dU
d
d
f U
f
f U
x
dx
dx
dx
U f 0
u
2
2
w
y y 0
y y 0
v
(8.29)
(Refer to Eqn. (8.17).)
(8.30)
Notes:
1.
Recall that U(x) is the flow profile for the outer, potential flow and (x) is a distance scaling parameter.
2.
Consider the case when =1 and is left arbitrary. From Eqn. (8.24) we have:
d
U 2 2
dx
U 2 x 2 constant
The scaling factor will be zero at x=0 so the constant will also be zero. Now divide the equation by the
other equation given in Eqn. (8.24):
1
1
U 2 x 2
dU
2
dx
1 dU
U dx
x 2
ln U
U x cx
ln x constant
C. Wassgren
(8.31)
364
This outer flow velocity distribution has the same form as the potential flow over a wedge (refer back
to an earlier set of notes). Recall that the complex potential for flow over a wedge is given by:
f z Az n
(8.32)
df
n 1
U iV
Anz n 1 An x iy
dz
where A is a constant and the angle between the walls of the wedge is equal to /n as shown below.
Note that this f (the complex potential) is not the same as the f in Eqn. (8.22).
y
/n
x
line of symmetry
Making use of symmetry we can produce flow over a wedge shape with a wedge angle denoted by, :
y
/n
wedge shape
x
line of symmetry
wedge angle,
Along the surface of the wedge, y=0, the potential flow horizontal velocity is:
U Anx n 1
Comparing Eqn. (8.33) to Eqn. (8.31) we find that and n are related by:
n 1
n
2
The full angle of the wedge, , will be:

1
2 - 2 1
n
(8.33)
(8.34)
(8.35)
The scaling function,(x), is found using Eqns. (8.24) and (8.31):

dU

dx
2
U f 0
c 2 1

x
2
2
(8.36)
x 2
cx 2
2
c
C. Wassgren
1
2
f 0
365
c3 221
x
f 0
2
(8.37)
Now lets consider two special cases for =1.

Flow over a Flat Plate (Blasius flow): n=1 (=0, =1):
The outer flow velocity is given by equation (8.31):
U x c U0
(8.38)
where U0 is a constant velocity. The scaling function is found from Eqn. (8.36) to be:
2 x
x
U0
This is precisely the same scaling function found in our earlier investigation of the Blasius
solution. The governing ODE is found from Eqn. (8.25):
f ff 0
Again, this is identical to what was found during our investigation of the Blasius solution.
(8.39)
(8.40)
Stagnation Point Flow (Heimenz flow): n=2 (=1, =1):

The stagnation point flow solution can be recovered using =1 and =1. Using these parameters
we find:
U x cx
(8.41)
f ff 1 f 0
which are the same results we found in our previous analysis of stagnation point flow.
2
C. Wassgren
366
Notes:
a.
The figure below shows the velocity profile for various flows over wedge shapes (produced by
letting =1 and varying .)
(Figure 20.4 from Panton, R.L.,
Incompressible Flow, 2nd ed,
Wiley. Note that in the figure m =
n-1= /(2 - ).
n>1
(accelerating flow)
decelerating flows
Blasius soln
n<1
(decelerating flow)
accelerating flows
One item of significance observed from the figure is the fact that for accelerated flows (n>1 dU/dx > 0)
there is no inflection point in the boundary layer profile. When the flow decelerates (n<1 dU/dx < 0)
there is an inflection point. Of particular interest is the case when (n-1) = -0.0904. For this case we
observe that the inflection point occurs at the wall and, as a result, the corresponding wall shear stress is
zero. It is at this point when boundary layer separation occurs. Thus, we see that a laminar boundary
layer is able to support only a very small deceleration without separation occurring. We will address the
topic of boundary layer separation later in the notes.
/n
n = 1 (= 0)
C. Wassgren
n = 2 (= 1)
367
n = 2/3 (= -1)
Example:
A uniform stream of incompressible fluid flows over a planar wedge of half-angle, (/2), side length, L,
and base length, H, as shown in the figure below. The upstream flow velocity and pressure are U and p.
L
U, p
a.
b.
c.
(/2)
Use a Faulkner-Skan boundary layer solution to determine the skin friction drag acting on the wedge
assuming laminar flow.
Check your solution to part (a) by calculating the drag for a flat plate (= 0) and comparing with the
Blasius solution.
Determine the form drag on the wedge assuming that the pressure at the back of the wedge is the
same as the free stream pressure.
SOLUTION:
Recall that the Faulkner-Skan boundary layer solution for flow over a wedge with total angle, (/2), is:
2
f ff 1 f 0
(8.42)
x
y
L
(8.43)
f 0 0
f 0 0
U, p
(/2)
f 1
U x cx 2
x 2
(8.46)
(8.47)
(8.44)
(8.45)
c3 221
x
f 0
2
(8.48)
The skin friction drag over both surfaces of length, L, (assuming unit depth into the page), and also taking
into account the angle of the surfaces, is:
xL
Dskin
2 cos
friction
Dskin
2 cos
friction
2 w dx
2
x 0
xL
(8.49)
c3 221
c3
x
f 0 dx
f 0
2
2
x 0
1 2 1 1 L
c3
f 0 2 1 x 2
2
1
0
2
Dskin
friction
2 cos
x 0
2 1
x 2 dx
(8.50)
2 12
c
f 0
L
2
1
C. Wassgren
xL
368
(8.51)
Note that when = 0, we get Blasius flow over a flat plate.

Dskin
c3
friction
f 0 2 L 2
1
(8.52)
U x c (from Eqn. (8.45))
(8.53)
Dskin friction 8U 3 L f 0 32
1
2
U L
UL
Blasius
f 0
(8.54)
To determine f 0 , we need to solve the ODE:

f ff 0
(8.55)
numerically. Performing this calculation (or using Table 4.1 from White, Viscous Fluid Flow, for example)
gives:
f 0 0.46960
(8.56)
and thus:
2.656
Dskin friction 12 U 2 L
Re 1 2
Blasius
L
(8.57)
The Blasius drag coefficient (considering two sides of the plate) is:
Dskin fraction,
1.328
2.656
two-sides
2
Dskin fraction, 12 U 2 L
CD ,Blasius,
1
1 U 2 L
Re 1 2
2
two-sides
two-sides
Re L
L
2
This is the same as Eqn. (8.57)!
(8.58)
The form drag may be found by integrating the pressure force acting on the surface and taking the
component in the direction of the incoming flow. Use gage pressures to simplify the calculation. The
pressure may be found using Bernoullis equation in the outer, potential flow.
Dform 2sin
xL
pgage dx where pgage 12 U 2 U 2
(8.59)
x 0
(Note that since a gage pressure is being used, the pressure force on the back of the wedge neednt be
considered.)
xL
U2
1
U 2 1 2 dx
Dform 2sin 2
2
U
x 0
xL
U2
1 2
U
x 0
dx sin
2
2
sin
sin
2 U 2 L Uc 2
Dform sin
U 2
L2
C. Wassgren
x L
2
2 2
c
x
1
dx
2
x 0
(8.60)
2 U 2 L 1 Uc 2 22 L
(8.61)
369
2
2
U 2
5.
Approximate Methods: The Krmn Momentum Integral Equation
So far weve only examined boundary layer flows that lend themselves to similarity solutions. This, of
course, is very restrictive. There are many non-similar boundary layer flows that we would also like to
investigate. Since the majority of fluid mechanics problems are extremely complex, we often have to
resort to empirical or semi-empirical methods for investigating the flows in greater detail. Here well
discuss one such semi-empirical method used for investigating boundary layers called the Krmn
Momentum Integral Equation (KMIE).
The idea is straightforward and relies on the Linear Momentum Equation. Consider a differential control
volume as shown in the figure below. The top of the control volume is defined by the line separating the
boundary layer region from the outer flow region (this is not a streamline).
dp 1
p dx 2 dx d
d p
dx
dx
+d
w dx
dx
Lets apply the linear momentum equation in the x-direction to the control volume:
Fx,on CV
d
dt
u dV u u
x
CV
rel
dA
(8.62)
CS
where
d
dp 1

Fx,on CV
p p p dx p
dx (assuming unit depth)
2 dx d
w
dx
dx

left
bottom
right
d
dt
CS
u dV 0
x
top
(steady flow)
CV
y

y
d
d
u 2 dy dx U udy dx
dx
dx
y 0
y 0
y 0

y 0
u x u rel dA
u 2 dy
left
C. Wassgren
u 2 dy
right
370
top
Note that the mass flow rate through the top is found via conservation of mass on the same control volume:

d
m top udy udy udy dx 0
dx
0
0

in through top
0
in at left
(8.63)
out through right
m top
d
udy dx
dx
0
Substituting and simplifying (and neglecting higher order terms):
dp
d
d
u 2 dy dx U
udy dx
dx w dx
dx
dx
dx
(8.64)
dp
d
d
u 2 dy U
udy
w
dx
dx
dx
Recall that the pressure at a given x location remains constant with y position so that we can find dp/dx in
terms of the outer (potential) flow velocity using Bernoullis equation outside of the boundary layer:
dp
dU
p 1 2 U 2 constant
U
0
dx
dx
(8.65)
dp
dU
U
dx
dx
In addition, well re-write the boundary layer thickness in terms of an integral so that:
dp
dU
dU
dy
U
dx
dx
dx
Udy
(8.66)
Additional re-arranging gives:
d
dU
d
udy
uUdy
udy
U
dx
dx 0
dx 0
0
d
dU
dU
d
2
udy
Udy w u dy uUdy
dx
dx
dx 0
dx 0
0
0
Additional re-arranging and simplifying gives:
dU
d
w u U u dy
U u dy
dx
dx
0
0
d 2 u
u dU
u
U
U 1 dy
1
dy
dx
U U dx
U
0
0
M
D
(8.67)
(8.68)
Thus, if the fluid has constant density:

w d 2
dU
U M DU
dx
dx
This is known as the Krmn Momentum Integral Equation (KMIE).
C. Wassgren
(8.69)
371
(8.70)
Notes:
1.
2.
If the pressure remains constant, then dU/dx=0 and

d
w U 2 M
dx
(8.71)
The typical methodology for using the KMIE is as follows:

a. Obtain an approximate expression for U=U(x) from inviscid flow theory (e.g. potential flow
theory). Recall that Bernoullis equation can be used to relate the pressure and U.
b. Assume a velocity profile in the boundary layer subject to the appropriate boundary conditions, i.e.
assume a form for:
u
y
(8.72)
f
U

uy
uy
(8.73)
0 0 and
1 1
U
U
The form of the approximate velocity profile is typically found based on curve fits to experimental
measurements of the boundary layer velocity profile.
Note that higher order profiles will have additional boundary conditions. For example, a cubic
curve fit will also have a boundary condition that matches the slope of the velocity profile at the
freestream boundary.
c.
The shear stress at the wall for a laminar flow can also be determined from the Newtonian stressstrain rate constitutive relations to be:
u
d
U U
(8.74)
w
d y
y
0
(For a turbulent flow experimental data for the wall shear stress are used. This will be discussed in
greater detail later in these notes.) This laminar wall shear stress must be the same shear stress as
that found using the KMIE (Eqn. (8.70)). Thus, we can equate the two shear stress expressions.
The resulting differential equation can then be solved for the boundary layer thickness, , as a
function of x.
C. Wassgren
372
Example:
Consider laminar flow over a flat plate (U=constant). Approximate the boundary layer velocity distribution
using a parabolic profile:
y y 2
y
for 0 1
2
u
(8.75)
U
y
1
1
Evaluate the momentum thickness, M:

1
u
u
1 d 2 2 1 2 2 d
U U
0
0
(8.76)
2
M
15
where in the above equation the substitution =y/ has been used. Now substitute this momentum
thickness into Eqn. (8.70). Note that dU/dx=0 so that the second term on the right hand side in Eqn. (8.70)
is zero.
w d 2
dU
U M DU
dx
dx
d
d
2
w U 2 M U 2
dx
dx
15
This should be the same shear stress found using Eqn. (8.74):
U d
U
2
w
2 2
d
(8.77)
(8.78)
so that by equating the two we have:

U
2
d
2 U 2
15
dx
d
0
15
dx
15
12
x
U
2
(8.79)
30
5.5
Ux
Ux
5.5
1
Re x2
Recall that the exact (Blasius) solution for flow over a flat plate gave:
5.0
1
x Re x2
(8.80)
so that the approximate solution is off by only 10%!
C. Wassgren
373
Example:
A measured dimensionless laminar boundary layer profile for flow past a flat plate is given in the table
below. Use the momentum integral equation to determine the 99% boundary layer thickness. Compare
your result with the exact (Blasius) result.
y/
0.00
0.08
0.16
0.24
0.32
0.40
0.48
0.56
0.64
0.72
0.80
0.88
0.96
1.00
u/U
0.00
0.133
0.265
0.394
0.517
0.630
0.729
0.811
0.876
0.923
0.956
0.976
0.988
1.000
SOLUTION:
Apply the Krmn Momentum Integral Equation:
w d
dU
M U 2 DU
dx
dx
Assuming a flat plate flow with no pressure gradient:
dU
U constant
0
dx
Simplifying Eqn. (8.81) gives:
d
w U 2 M
dx
(8.81)
(8.82)
(8.83)
The momentum thickness is given by:

y
u
y 0 U
1
U
dy
U
y
1
U
y
d
Integrating the data numerically using the trapezoidal rule gives:

M 0.131
d
w 0.131U 2
dx
C. Wassgren
(8.84)
(8.85)
374
For a laminar flow, the shear stress can also be expressed as:
du
dy
y 0
u
U d U
d y
(8.86)
y 0
Differentiating the data numerically using a 1st order finite difference scheme:
U
w 1.66
(8.87)
Equating Eqns. (8.85) and (8.87) gives:

d
U
0.131U 2
1.66
dx

x x
U
x
1 2
12.67
2
U
dx
d 12.67
5.034
x0
5.034
Ux Re x
1
(8.88)
Equation (8.88) is within 1% of the exact Blasius solution of x 5.0 / Re x2 .

1
Another approach to this problem is to fit a polynomial curve to the given data rather than numerically
differentiating and integrating the data.
3.
This approximate technique can be used for either laminar or turbulent flows. In fact, this method is
especially useful for analyzing turbulent boundary layer profiles (to be discussed in later notes).
C. Wassgren
375
6.
Approximate Methods: Thwaites Correlation
Thwaites method (1949) is generally considered the best available one parameter method for describing
laminar boundary layers. The correlation uses the Krmn Momentum Integral Equation along with
dimensionless experimental laminar boundary layer data.
Recall from the previous notes that the momentum thickness, M, the displacement thickness, D, and the
shear stress at the wall, w, can be related in an average sense via the Krmn Momentum Integral Equation
(KMIE):
W
d
dU
(8.89)
U 2 M DU
dx
dx
(Note again that U is the outer velocity, i.e. the flow velocity just outside the boundary layer.) We can rearrange this equation to the following form:
W
d
dU
dU
2 M U
U 2 M DU
dx
dx
dx
(8.90)
d
dU
2 M D U
U 2 M
dx
dx
Lets write the KMIE using two dimensionless shape factors, H and T, defined in the following manner
(following the approach of Holstein and Bohlen, 1940):
D
M
and
W M
U
(8.91)
U
M
where H is the shape correlation and T is the shear correlation. These shape factors only depend on the
shape of the velocity profile. Re-writing the KMIE using these shape factors gives:
d
UT
dU
U 2 M 2 H M U
(8.92)
M
dx
dx
Multiplying through by M/(U) and re-writing the dM/dx term gives:
d
UT
dU
U 2 M 2 H M U
M
dx
dx
T U
M d M
2 dU
2 H M
dx
dx
U d M2 M2
dU
T
2 H
2 dx
dx
d M2
dU M2
U
(8.93)
2 T 2 H
dx
dx
After plotting several known velocity profiles, researchers found that H and T depend almost entirely on
another dimensionless quantity, , where:
2 dU
M
(8.94)
dx
so that the shape of the boundary layer velocity profile will be determined entirely by . Substituting Eqn.
(8.94) into Eqn. (8.93) gives:
d M2
dU M2
U
2 T 2 H
dx
dx
d
U
(8.95)
F 2 T 2 H
dx U
0
C. Wassgren
376
Lets plot F() as a function of for a number of known profiles.

(Figure scanned from Figure 4-22, White, F.M., Viscous Fluid Flow, 2nd ed., McGraw-Hill. Note that in
the figure below is the momentum thickness, M.)
= 2[T-(2+H)]
Thwaites used a simple linear curve fit to the data:

d M2
U
F 0.45 6.0
dx
dU M2
0.45 6.0
dx
0.45 U
6.0
d M2
dx
1 d
5
M2 U 6
U dx
2
M
dU
dx
M2 U 6 0.45 U 5 dx constant
M x 0 0 constant 0
(8.96)
so that the momentum thickness is related to the outer flow velocity by:
M2
0.45
U6
U dx
5
(8.97)
C. Wassgren
377
Notes:
1.
The functions H() and T() are found from plots of several known velocity profiles. The figure below
shows Thwaites correlation data for the shape factors.
The following two curve fits to the data are given by White, F.M., Viscous Fluid Flow, 2nd ed.,
McGraw-Hill.
T 0.09
0.62
H 2.0 4.14 z 83.5 z 2 854 z 3 3337 z 4 4576 z 5
(8.98)
where z 0.25
4.0
0.6
3.5
0.5
3.0
())
HH(
2.0
0.3
1.5
TT(() )
0.4
2.5
0.2
1.0
H(lambda)
T(lambda)
0.5
0.0
-0.15 -0.10 -0.05
0.1
0.0
0.00
0.05
0.10
0.15
0.20
0.25
0.30
//)(dU
/dx )
(M2M
)(dU/dx)
2.
The typical procedure for using Thwaites method is as follows:

a. Determine U=U(x) for the outer potential flow.
b. Determine M using Eqn. (8.97).
c. Determine using Eqn. (8.94).
d. Determine W and D using Eqn. (8.91) and the curve fits or figure shown in the first note.
3.
This method is considered one of the best methods for predicting the behavior of laminar boundary
layers. It is accurate to about 5% for favorable or mild adverse pressure gradients and is accurate to
about 15% near the separation point. When more accurate calculations are necessary, one typically
turns to numerical methods for solving the boundary layer equations.
4.
Recall that the boundary layer separation point occurs when w=0 T()=0 so that
at separation point 0.090
C. Wassgren
378
(8.99)
5.
Thwaites method, as presented here, is restricted to laminar, planar flows. A similar type of method
can also be derived for laminar, axi-symmetric flows.
6.
Note that the outer flow velocity, U, may in fact be significantly different than the expected potential
flow velocity profile. For example, for flow around a cylinder (or any bluff body), boundary layer
separation results in a large wake region. Thus, the potential flow prediction (using a uniform stream
and a doublet to model flow around a cylinder) for the outer flow velocity may be greatly in error.
Often we must resort to experimental data to obtain the outer flow velocity profile.
C. Wassgren
379
Example:
Consider the simple decelerating non-similar outer flow described by
x
U U 0 1
L
Using Thwaites method, determine the point at which flow separation occurs.
(8.100)
SOLUTION:
Using Thwaites correlation we have:
M2
0.45
U 05 1 x
L
0.075
1 x L
U
U 06
1 x
dx
(8.101)
0
so that the dimensionless parameter, , is:
6
2 dU
M
0.075 1 x
1
L
dx
Separation will occur when =-0.090:

6
0.090 0.075 1 x
1
L
xsep
0.123
L
This value is within 3% of the exact result xsep/L=0.120 which is found numerically.
6
C. Wassgren
380
(8.102)
(8.103)
Example:
Consider the boundary layer flow resulting from a sink located at the trailing edge of a thin, flat plate as
shown in the figure.
y
sink strength, m
a.
b.
Using Thwaites method, determine and plot the dimensionless momentum thickness, M/(a2/m)1/2 ,
over the top surface of the plate as a function of dimensionless distance along the plate, x/a.
Will boundary layer separation occur on the plate? If so, determine the location of the separation
point. If not, explain why. [Hint: No calculations are necessary for this part.]
SOLUTION:
Determine the outer flow velocity profile by modeling the outer, potential flow as a single line sink located
at x = a.
m
f z
log z a (where m > 0)
(8.104)
2
The fluid velocity is given by:
z a
z a
x a iy
df m 1
m
m
m
u x iu y
2
2
2 x y 2 2 xa a 2
dz 2 z a 2 z a z a 2 z za za a 2
u x iu y
x a
m
m
y
i
2
2
2 x a y
2 x a 2 y 2
(8.105)
On the plate surface, y = 0 so that:

m 1
and u y 0 (Note that 0 x a on the plate surface.)
ux
2 x a
(8.106)
Use Thwaites method to determine the momentum thickness, M:
M2
0.45
U6
x x
x x
0.45
U dx m
5
x 0
2 x a
2 0.45 x a 6 x x
x a
x 0
m 1
dx
2 x a
dx
x 0
2 0.45 x a 6
x a a 4
4m
2
6
0.45 a 2 x x
1 1
2m
a a
0.45 a 2
2m
x 2 x 6
1 1
a a
C. Wassgren
381
M
a2
x 2 x 6
0.225 1 1
a a
(8.107)
Boundary layer separation will not occur on the plate since there is a favorable pressure gradient along the
surface. This can be shown by examining Bernoullis equation in the outer flow.
p
U
p 12 U 2 constant
U
0
x
x
p
U
U
(8.108)
x
x
The velocity on the plate surface is given by Eqn. (8.106) so that:
p
m

x
2
(8.109)
x a 3
Since 0 x a, p/dx < 0 (a favorable pressure gradient).

Another approach to showing that there will be no boundary layer separation is to use Thwaites method
(applicable to laminar boundary layers). Boundary layer separation occurs when the wall shear stress is
zero, i.e. w = 0. In Thwaites method this occurs when:
2 dU
0.090
(8.110)
M
dx
Substituting Eqn. (8.107) for U gives:
2 m
1
0.090 M
(8.111)
2 x a 2
Noting that the right hand side will always be positive (note that m > 0 as defined in Eqn. (8.104)), we must
conclude that boundary layer separation will not occur.
C. Wassgren
382
7.
Turbulent Boundary Layer over a Flat Plate (no pressure gradient)
To analyze a turbulent boundary layer we must use the momentum integral approach coupled with
experimental data since no exact solutions are known. To approximate the velocity profile in a turbulent
boundary layer, recall the Law of the Wall (from a previous set of notes on turbulence):
*
u * yu
for yu*/ 5
1 yu *
ln
for yu*/ > 5
c
u
K
where u* = (w/)1/2 is the friction velocity.
We could substitute this velocity profile into the momentum thickness integral:
y
u
u
M
1
dy
U U
y 0
u
(8.114)
and solve. Using the stated velocity profile, however, can get cumbersome. Instead, Prandtl suggested
approximating the logarithmic turbulent velocity profile using a 1/7th power-law curve fit:
1
u
y 7
for y

U
u
1
for y
U
Using this velocity profile, the momentum thickness becomes:
1
1
y
y
7
7
u
u
y y
M
1
dy 1 dy

U U
y 0
y 0
72
To determine the shear stress, recall that from the Karman momentum integral equation:
d
d
7
w U 2 M U 2
dx
dx
72
so that the friction coefficient becomes:
w
7 d
Cf
1 U 2
36 dx
2
(8.117)
(8.118)
Experimental wall friction data for turbulent boundary layers can be fit using:
1
C f 0.020 Re 6
(8.119)
where Re = (U/). Equating the two friction coefficients gives:

7 d
U
0.020
36 dx

16
laminar
U
d 0.103

0
1
dx
x0
x x0
Note: The shear stress relationship

holds strictly for the part of the BL
that is turbulent. If the laminar BL
thickness and distance downstream
are small in comparison to the
current thickness and location, then
we may assume that the turbulent
BL starts approximately at the
leading edge, i.e. 0 = 0 at x0 = 0.
U 6
x x0

Assuming 0 = 0 at x0 = 0 gives:
7
0.120
x
U x
turbulent
16 x x
0 0.120
7
C. Wassgren
383
0.163
1
Re x7
(8.120)
From this relation we can also determine the displacement thickness, momentum thickness, friction factor,
and drag coefficient. These relations are summarized below.
0.16
1
Re x7
Cf
(8.121)
0.027
1
Re x7
(8.124)
D
x
CD
0.02
1
Re x7
(8.122)
0.031
1
Re L7
(8.125)
M
x
Re x
0.016
1
Re x7
(8.123)
U x
Notes:
1. The boundary layer thickness grows as ~ x6/7 for a turbulent boundary layer whereas it grows as ~
x1/2 for a laminar boundary layer. Hence, a boundary layer grows more rapidly with distance for
turbulent flow than for a laminar flow. The momentum and displacement thicknesses also increase
more rapidly for turbulent boundary layers.
2. The shear stress decreases more rapidly for laminar flow than for a turbulent flow. The drag does not
increase as rapidly in a laminar flow as compared to a turbulent flow.
3. Another experimental friction curve fit that is commonly used is:
0.0466
Cf
(8.126)
1
Re 4
which gives:
D 0.0478
M 0.0371
0.382
(8.127)
(8.128)
(8.129)
1
1
1
5
5
x Re x
x
x
Re x
Re x5
Cf
0.0594
1
Re x5
(8.130)
CD
0.0742
1
Re L5
(8.131)
Re x
U x
nd
White (in White, F.M., Viscous Fluid Flow, 2 ed., McGraw-Hill) states that the experimental curve fit
given by Eqn. (8.126) is based on limited data and is not as accurate as the curve fit given by Eqn.
(8.119). This argument is supported by the plot shown below (plot from White, F.M., Viscous Fluid
Flow, 2nd ed., McGraw-Hill).
a more accurate analytical relation not
presented here due to its complexity
Eqn. (8.130)
Eqn. (8.124)
C. Wassgren
384
Example:
A thin smooth sign is attached to the side of a truck as shown. Estimate the skin friction drag on the sign
when the truck speed is 55 mph.
5 ft
20 ft
3 ft
4 ft
GO BOILERS !!
SOLUTION:
Assume that the boundary layer forms at the front of the trailer.
b
L1
region 1
L2
region 2
L1 = 5 ft
L2 = 25 ft
b = 4 ft
U = 55 mph = 80.7 ft/s
air = 1.57*10-4 ft2/s
air = 2.38*10-3 slugs/ft3
To find the drag on the sign, determine the drag on region 2 and subtract the drag from region 1.
Dsign D2 D1
(8.132)
where
Di cDi 12 U 2 Li b (i = 1 or 2)
(8.133)

Dsign 12 U 2 b cD 2 L2 cD1 L1
(8.134)
The drag coefficients are determined from the Reynolds numbers at each regions trailing edge.
80.7 ft/s 5 ft
UL
2.6 *106 (turbulent!)
Re1 1
1.57*10-4 ft 2 /s
Re 2
UL2
80.7 ft/s 25 ft
1.57*10-4 ft 2 /s
1.3*107 (turbulent!)
(8.135)
(8.136)
Assume that the flow is fully turbulent throughout regions 1 and 2 (neglect any laminar flow contribution)
so that:
0.0742
0.0742
cD1
3.87 *103
(8.137)
1
1
5
5
6
Re1
2.6 *10
C. Wassgren
385
cD 2
0.0742
1
Re 2
0.0742
1.3*10
7
2.80 *103
(8.138)
Substitute into Eqn. (8.134) and evaluate.

Dsign
1
2
2.38*10
slugs/ft 3 80.7 ft/s
4 ft 2.80*10-3 25 ft 3.87*10-3 5 ft
Dsign 1.57 lbf
C. Wassgren
(8.139)
386
Example:
The flat plate formulas for turbulent flow over a flat plate assume that turbulent flow begins at the leading
edge (x = 0). In reality, there is an initial region of laminar flow as shown in the figure.
y
x
turbulent flow
laminar flow
1.
Derive an expression for the 99% boundary layer thickness in the turbulent region by accounting for
the laminar part of the flow.
2. Plot the dimensionless boundary layer thickness, /x, as a function of Reynolds number (104 Rex
108, use a log scale for the Rex axis) for your derived relation and for the turbulent relation that does
not consider the laminar part.
Assume a 1/7th power law velocity profile for the turbulent boundary layer and an experimental friction
1
coefficient correlation of C f 0.020 Re 6 .
SOLUTION:
First determine the boundary layer thickness in the laminar flow region using the Blasius solution:
5.0
(Rex < 500,000)

(8.140)
1
x Re x2
Assume that the transition to turbulence occurs at a Reynolds number of 500,000 so that condition at the
transition point is:

trans 3.536 *103
(8.141)
U
where
500, 000
xtrans
(8.142)
U
Now use the Karman Momentum Integral Equation to determine the boundary layer characteristics for the
turbulent region. Assume that the velocity profile follows the following form:
1
y 7 y
1

u
(8.143)

U
y
1
1
Using this velocity profile, the momentum thickness is:

1
1
y
y
7
7
u
u
7
y y
M 1 dy 1 dy
72
U
U

y 0
y 0
(8.144)
To determine the shear stress, recall that from the Karman Momentum Integral Equation, with a constant
outer velocity:
d
d
7
w U 2 M U 2
(8.145)
72
dx
dx
so that the friction coefficient is:
C. Wassgren
387
Cf
w
7 d
2
36 dx
2 U
(8.146)
Using the given experimental wall friction correlation:

1
C f 0.020 Re 6
(8.147)
where Re = (U/), equate the two friction coefficients to give:

16
7 d
U
0.020
36 dx

U
d 0.103

trans
1
16
(8.148)
x x
dx
(8.149)
x xtrans
16
U
(8.150)
x xtrans

where Eqns. (8.141) and (8.142) are used for trans and xtrans, respectively. Substituting and simplifying
results in:
trans 0.120
7
6

U
0.120
U

1.380 *104
7
16
x 500, 000
U

0.120 x 6.000 *104 1.380 *104
U
U
U
7

4
0.120 4.620 *10
x
Ux
Ux
6
0.120 4.620 *104

1
7
x Re x6
Re x6
(8.151)
7
(8.152)
(8.153)
Re x 500, 000
(8.154)
Compare this result to one that assumes that the turbulent boundary layer starts from the leading edge:
0.16
(Rex > 500,000)

(8.155)
1
x Re x7
dimensionless boundary layer thickness, /x
0.06
no laminar part
includes laminar part
0.05
0.04
0.03
0.02
0.01
0.00
1.E+04
1.E+05
1.E+06
1.E+07
1.E+08
Reynolds number, Rex
C. Wassgren
388
8.
Boundary Layer Separation
Consider flow around a cylinder as shown below:
Let s be the distance from the

stagnation point along the cylinder
surface.
wake
outer potential flow

thin BL
BL separates
On the front side of the cylinder the boundary layer grows with increasing distance as we might expect.
(Note: The pressure near the cylinder surface decreases with distance from the stagnation point on the front
half dp/ds < 0.) At a particular location on the cylinder surface, the boundary layer no longer remains
attached to the surface. This point is termed the boundary layer separation point. The occurrence of
boundary layer separation is a result of energy dissipation in the boundary layer due to viscous effects and
adverse pressure gradients.
Define:
an adverse pressure gradient as one in which dp/ds > 0, and
a favorable pressure gradient as one in which dp/ds < 0
Consider the pressure near the surface of the cylinder determined from potential flow theory:
adverse
favorable
p
C
p+ /2U2 A
1
potential flow theory
p-3/2U2
0
Imagine a fluid particle in the BL. It experiences approximately the same pressure as that in the outer
potential flow (refer to the notes discussing the BL equations). As the fluid particle moves in the BL, it
loses energy due to viscous effects. As a result, it does not have enough energy to coast up the pressure
hill from point B to C. The fluid moves up the pressure hill as far as possible and then BL separation
occurs.
C. Wassgren
389
(From White, F.M., Fluid Mechanics, 3rd ed., McGraw-Hill.)
Inviscid analysis works reasonably well on the upstream side but fails miserably on the downstream side
due to boundary layer separation. Notice that the pressure in the wake is relatively uniform and equal to the
pressure at the boundary layer separation point. Hence, delaying boundary layer separation helps to
recover the pressure in the wake as a well as decreasing the area of the wake. As a result, the drag due to
pressure forces (aka form drag) decreases as boundary layer separation is delayed.
C. Wassgren
390
Lets look at the velocity profile at different points along a plate for a flow with an adverse pressure
gradient (dp/dx > 0):
recirculation
region
velocity profile
inflection point, i.e.
2u
0
y 2
w = 0
(zero slope at the wall)
This is defined as the separation
point.
backflow at the wall (recirculation). The

boundary layer thickens considerably
the BL assumption that v << u is no
longer valid
In an adverse pressure gradient flow the boundary layer velocity profile will always have an inflection
point. This can be shown by considering the boundary layer momentum equation:
u
u
1 dp
2u
1 dp
dU
.)
2 (Bernoullis equation has been used to write:

u v
U
dx
dx
x
y
dx
y
Note that at the wall boundary (y = 0), u = v = 0 so that:
dp
2u
2
dx
y y 0
Thus, at the wall boundary in an adverse pressure gradient (dp/dx > 0), we must have:
2u
0
y 2 y 0
However, at the free stream interface (y = ) we have:
2u
0
y 2 y
in order for the boundary layer profile to merge smoothly with the outer flow velocity profile. The change
in sign of the boundary layer curvature indicates that somewhere within the boundary layer there must be a
point of inflection. The inflection point moves toward the outer flow boundary as the flow moves
downstream in an adverse pressure gradient flow.

C. Wassgren
391
C. Wassgren
392
Notes:
1. Turbulent BLs have more momentum than laminar BLs (consider the velocity profiles)
fluid in a turbulent BL will travel further into an adverse pressure gradient than fluid in a
laminar BL
BL separation occurs later for a turbulent BL than for a laminar BL
y/
Note that w, laminar < w, turbulent since

the velocity gradient for laminar
flow is less than the velocity
gradient for turbulent flow.
u/U
(From Van Dyke, M., An Album of Fluid Motion, Parabolic Press.)
C. Wassgren
393
C. Wassgren
394
9.
Forces on Objects Immersed in a Fluid Flow
The force acting on an object immersed in a fluid flow is comprised of the force due to pressure variations
over the surface and the force due to viscous shear stresses.
If we know the pressure (p) and shear stress () distribution over the object, then:
FL
FP pn dA
FS ,i ji ni dA
dA
FD
where FP is the force due to the pressure component, FS is the force due to the shear stress component, and
A is the surface area of the object.
The component of the force acting in the direction parallel to the incoming flow is known as the drag force,
FD, and the component perpendicular to the incoming flow is known as the lift force, FL.
Notes:
1. The pressure force component of the drag is known as form drag while the shear stress drag component
is known as the skin friction drag.
2.
A streamlined body is one in which the (skin friction drag) >> (form drag).
small wake
A bluff body is one in which the (form drag) >> (skin friction drag).
large wake
3.
The lift and drag are often expressed in dimensionless form as a lift and drag coefficient, CD and CL:
FL
FD
CL
CD
1 U 2 A
1 U 2 A
2
2
where A is usually the frontal projected area (area seen from the front) for a bluff body or the planform
area (the area seen from above) for a streamlined body.
bluff body
(use the frontal projected area)
C. Wassgren
streamlined body
(use the planform area)
395
Flow Around a Cylinder (or Sphere) As a Function of Reynolds Number
Note: Re
VD
Re << 1
(creeping or Stokes flow)
5 < Re < 50
(fixed eddies)
60 < Re < 5000

(Karman Vortex Street,
periodic shedding of vortices)
5000 < Re < 200,000
Re 200,000
(drag crisis)
Re > 200,000
C. Wassgren
396
(From Batchelor, G.K., An Introduction to Fluid Dynamics, Cambridge University Press.)

C. Wassgren
397
C. Wassgren
398

C. Wassgren
399
Notes:
1.
The periodic shedding of vortices off the object results in periodic forces exerted on the object in the
cross-stream direction. The Tacoma Narrows bridge disaster (figure shown below) occurred because
a structural natural frequency of the bridge matched the frequency of the shedding vortices.
The Tacoma Narrows bridge collapsed in 1940.

2.
Experimental measurements have shown that the dimensionless frequency of the shedding vortices,
f, expressed as a Strouhal number, i.e. St = fD/V, remains relatively constant at 0.2 for 100 < ReD <
1*106.
St = fV/D
Re = VD/
(Figure from White, F.M., Fluid Mechanics, McGraw-Hill.)
The fact that the Strouhal number is insensitive to the Reynolds number over a wide range of
Reynolds numbers has been used to design a flow velocity meter known as the vortex flow meter
(shown below).
By measuring the frequency of the forces acting on the obstruction (of known size) and knowing that
the Strouhal number is approximately equal to 0.2, the flow velocity can be estimated.
C. Wassgren
400
drag crisis
Commonly used curve fits to the curve shown above are:

ReD < 1:
CD = 24/ReD (Stokes drag law)
CD = 24/ReD (1+3/16ReD) (Oseens approximation)
ReD < 5:
0 ReD 2*105:
CD = 24/ReD + 6/(1+ReD0.5) + 0.4
5
ReD < 2*10 :
CD = 0.44 (Newtons Law)
drag crisis
C. Wassgren
401
From the data we observe that increasing the roughness of the surface causes the drag crisis to
occur at a smaller Reynolds number.
Note:
The Reynolds number for a 95 mph baseball and a 170 mph golf ball are approximately 206,000 and
213,000, respectively. Hence, both are near the drag crisis!
C. Wassgren
402
The following is from White, F.M, Fluid Mechanics, 3rd ed, McGraw-Hill.
C. Wassgren
403
The following is from White, F.M, Fluid Mechanics, 3rd ed, McGraw-Hill.
C. Wassgren
404
Review Questions
1. What scaling arguments are used in deriving the boundary layer equations?
2. What are the appropriate boundary conditions for the boundary layer equations?
3. What restrictions are there on the Reynolds number for using the boundary layer equations?
4. Describe how the pressure within a boundary layer is determined.
5. Describe, in words, the approach used in the Blasius solution to the boundary layer equations.
6. What assumptions are made in the Blasius boundary layer solution? (e.g. Reynolds number
limitations, pressure gradients, free stream conditions, surface curvature, etc.)
7. At what Reynolds number (an engineering rule of thumb estimate) does a laminar boundary layer
transition to a turbulent boundary layer?
8. How does the boundary layer thickness vary with the distance from the leading edge of the boundary
layer for a flat plate, no pressure gradient boundary layer flow?
9. What is the expression for the 99% boundary layer thickness resulting from the Blasius solution?
10. What do the Falkner-Skan boundary layer solutions represent?
11. What are the boundary conditions used in the Falkner-Skan boundary layer solution?
12. Give two examples of practical boundary layer solutions that are embedded within the Falkner-Skan
general solution.
13. Can the Karman momentum integral equation (KMIE) be used for flows with non-uniform pressure
gradients? Turbulent flows? Compressible flows? Unsteady flows?
14. How might one find the outer flow velocity, U, when using the KMIE?
15. Describe the typical methodology used when applying the KMIE.
16. What is the 1/7th power law profile for a turbulent boundary layer?
17. In which type of boundary layer flow does the shear stress decrease most rapidly? Laminar or
turbulent? In which type of flow does the drag increase most rapidly?
18. Give a physical description of why boundary layer separation occurs.
19. What defines the point at which boundary layer separation occurs?
20. Why cant the boundary layer equations be used downstream of boundary layer separation point?
21. Why do turbulent boundary layers separate further downstream than laminar boundary layers?
22. What is meant by favorable and adverse pressure gradients?
23. Can a boundary layer separate in a favorable pressure gradient flow?
24. Must boundary layers always separate in an adverse pressure gradient flow?
25. What are the restrictions in using Thwaites correlation?
26. Describe the flow behavior as a function of Reynolds number for flow over a cylinder.
27. Sketch a plot of drag coefficient as a function of Reynolds number for flow over a sphere. Indicate
points of particular interest on the plot. Identify whether the axes are linear or logarithmic.
C. Wassgren
405
Chapter 09:
Turbulence
1. Introduction to Turbulence
2. Time-Averaged Continuity and Navier-Stokes Equations and Reynolds Stresses
3. Law of the Wall
C. Wassgren
Chapter 09: Turbulence
406
1.
Introduction to Turbulence
Lets consider the following simple experiment (this thought experiment is similar to the famous dye
injection experiment performed by Osbourne Reynolds). At a particular point in a pipe flow, lets measure
a velocity component of the fluid as a function of time for varying Reynolds numbers.
Re D
uD
measure velocity
at this point
At Reynolds numbers less than approximately 2300 we would find the following:
velocity, u
u
laminar flow:
smooth flow with no fluctuations
ReD < 2300
time
For 2300 < ReD < 4000 we would find a slightly different behavior:
velocity, u
u
transitional flow:
smooth laminar flow with random
bursts of turbulent behavior
2300 < ReD < 4000
time
For ReD>4000 we find a very different behavior:

velocity, u
u
turbulent flow:
complex flow with random
fluctuations about some mean velocity
ReD > 4000
time
C. Wassgren
407
Notes:
1.
Turbulence is a very difficult phenomenon to analyze. It is typically studied using semi-empirical

analyses, i.e. those analyses that use both theory and experimental data. The transition region is even
more difficult to analyze.
2.
The uncertainties associated with the transitional regime are also reflected in the value for the friction
factor, f, shown below in the Moody chart. Note that for 2300 < ReD < 4000 the value for the friction
factor is not well defined since the friction factor varies considerably as the flow transitions between
laminar behavior and turbulent behavior. (The plot below came from Fox, R.W. and McDonald, A.T.,
Introduction to Fluid Mechanics, 5th ed., Wiley).
C. Wassgren
408
2.
Time-Averaged Continuity and Navier-Stokes Equations
Since the time-varying velocity data shown in Section 1s example appears to consist of a fluctuating part
superimposed on a mean value, lets make the following definitions. First, express the instantaneous
velocity component, ui, as the sum of a mean velocity, ui , and a fluctuating velocity, ui , i.e.
ui ui ui
(9.1)
where the mean velocity over the time interval from t0 to t0+T is given by:
1
ui
T
t0 T
(9.2)
ui dt
t0
Note that the time average of the fluctuating velocity will be zero:
1
ui
T
t0 T
t0
t0 T
t0 T
1
1
ui dt
ui dt 0
ui ui dt
T
T
t0
t0

ui
(9.3)
ui
However, the mean of the square of the fluctuating velocities will, in general, be greater than zero:
ui
t0 T
ui ui
dt 0
(9.4)
t0
Similarly,
1
uiu j
T
t0 T
ui ui u j u j dt 0
(in general) (i j)
(9.5)
t0
Now lets look at the governing equations for an incompressible, Newtonian fluid (continuity and NavierStokes) where well use the mean and fluctuating parts for our unknown variables:
ui ui ui
(9.6)
p p p
Substituting into the continuity equation gives:
ui ui ui ui ui
(9.7)
0
xi
xi
xi xi
Now lets take the time-average of the continuity equation:
1
T
t0 T
t0
ui ui
dt 0
xi xi
(9.8)
where
1
T
1
T
t0 T
t0
t0 T
t0
t0 T
ui
u
1
dt
ui dt i
xi
xi
xi T
t0
(9.9)
t0 T
ui
1
dt
uidt 0
xi
xi T
t0
Thus, the time-averaged continuity equation becomes:

ui
0
xi
(9.10)
Note that the time-averaged continuity equation looks essentially the same as the instantaneous continuity
equation except it is in terms of time-averaged velocities.
C. Wassgren
409
Now lets take the same approach with the Navier-Stokes equations:
u
u
2u
p
2i fi
i uk i
xk
xi
xk
t
(9.11)
To help with the upcoming analysis, lets re-write the left-hand-side using the continuity equation:
u
u
u
2u
p
2i fi
i uk i ui k
t
xk
xk
xi
xk
(9.12)
ui uk
xk
Write the velocities and pressure in terms of mean and fluctuating parts:
ui ui uk uk
ui ui
p p
2 ui ui
fi
t
xk
xi
xk2
u
u
2u
2u
p p
i i
2i 2i fi
ui uk ui uk uk ui uiuk
t
t
xk
xi xi
xk
xk
Time-average the entire previous equation noting that for any term F:
F F
x x
and
(9.14)
ui
p 2ui
ui uk uk ui
2 0
t
t
xk
so that we have:
u
2u

p
i
ui uk uiuk
2i fi
t
x
x
x
k
(9.13)
(9.15)
(9.16)
We can further simplify this equation by using the time-averaged continuity equation:
u
u
ui uk ui k uk i
xk
xk
xk
(9.17)
Thus, the time-averaged Navier-Stokes equations become:

u
u
p
ui
uiuk fi
i uk i
xk
xi xk xk
t
C. Wassgren
410
(9.18)
If we compare Eqn. (9.18) to the instantaneous Navier-Stokes equation, we see that an extra term appears
on the right hand side with the same dimensions as the laminar shear stress term:
uiuk
(9.19)
xk
These terms are referred to as Reynolds stresses (in fact, they are momentum flux terms).
Notes:
1.
There are both Reynolds normal and shear stresses. For example, the x-component of the timeaveraged Navier-Stokes equations is given by:
Du
p
fx
Dt
x
(9.20)
u
2
u
u
u
u v
u w
y y
z z
x x
Reynolds "normal stress"
Reynolds "shear stress"
Reynolds "shear stress"
2. What, physically speaking, is a Reynolds stress?

First lets examine what causes viscosity in a laminar flow. Recall that a fluid is comprised of a
collection of molecules. In a flow with a velocity gradient, the molecules in a particular layer will
have an average velocity in addition to some random (thermal) motion. Since there is a velocity
gradient in the fluid (u constant), molecules in adjacent layers will not have the same average
velocity. Due to their random motion, molecules starting in a particular layer (with a particular
average velocity) will move into layers with a different average velocity. When a molecule moves into
a region of higher velocity, the molecules in that layer must accelerate the incoming molecule up to the
new speed. In order to accelerate the new molecule, the molecules already in the layer must exert a
force on the new molecule (via molecular collisions). From Newtons 3rd law, the new molecule exerts
an equal but opposite force on the layer. An identical process happens when a particle moves into a
region of slower average velocity. Thus, we see that the laminar shear stress is a result of the random
flux of molecules into neighboring layers with different average velocities.
In a turbulent flow, we have this microscopic effect plus a macroscopic flux of fluid due to the random
motion of the fluid. This macroscopic random fluid motion into regions of differing average motion
gives rise to Reynolds stresses.
y
u y
u
v
random macroscopic
fluid motion
total
turbulent
laminar
C. Wassgren
411
3.
Near a wall, the Reynolds shear stresses are small due to the wall restricting the random motion of the
fluid. This region is termed the viscous sub-layer and
(9.21)
viscous sub-layer laminar
Far from the wall, turbulent motion dominates. This region is termed the turbulent core and:
turbulent core turbulent
(9.22)
The region between the laminar sub-layer and the turbulent core is referred to as the transitional
region where the laminar and turbulent shear stresses are of the same order of magnitude.
y
turbulent core
total
transitional region
laminar sub-layer
4.
The relative turbulence intensity is defined as the magnitude of the Reynolds stresses relative to
some characteristic flow speed, U, e.g. an outer flow velocity:
TI
5.
uiu j
(9.23)
If the turbulence intensity is the same in all directions, then the turbulence is considered isotropic.
Otherwise, the turbulence is anisotropic. As shown in the following figure for turbulent flow over a
flat plate with no pressure gradient and Rex 107, the turbulence near the wall is found to be
anisotropic with the relative turbulence intensity normal (v) to the wall being smaller than the
streamwise (u) and lateral (w) turbulence intensities. This effect is the result of the geometric
constraint posed by the wall. Farther from the wall at approximately y/ = 0.8, the turbulence becomes
nearly isotropic.
u
U
(Figure 6.4, White, F.M., Viscous Fluid Flow, McGraw-Hill, 3rd ed.)
C. Wassgren
412
6.
The turbulent kinetic energy, K, is defined as the kinetic energy of the normal turbulent fluctuations:
K 12 uiui
(9.24)
The turbulent kinetic energy is used in models that examine the balance of energy associated with
turbulent motion (beyond the scope of these notes).
3.
Law of the Wall
Lets write the quantities that affect the mean velocity profile, u , near the wall:
u f y, , , w
(9.25)
where y is the distance from the wall, and are the fluid density and kinematic viscosity, and w is the
shear stress that the wall exerts on the fluid.
By performing a dimensional analysis on these variables we find a relation termed the Law of the Wall:
yu*
u
g
(9.26)
u*

where g is an unknown function and
u*
(9.27)
where u* is referred to as the friction velocity.

Nearest the wall, in the region known as the laminar sub-layer ( 0 yu*/ 5), the wall stress is given by:
du
w
(9.28)
dy
Dividing through by the density and integrating, the mean velocity profile in the vicinity of the wall is:
y
w
dy du u*
u*
u
u
y u
(9.29)
u* y
yu *
mean velocity profile in the laminar sub-layer 0
5
C. Wassgren
413
(9.30)
Now lets examine the turbulent core region. One simple model, known as Prandtls Mixing Length
Hypothesis, assumes that the fluctuating velocities, e.g. u and v, are approximately equal to some typical
eddy length, l, multiplied by the velocity gradient, du / dy :
y
u , v l
du
dy
Thus, the Reynolds shear stress in the turbulent core is:

du
dy
u v l 2
(9.31)
Note that l cannot be constant since u, v 0 as y 0. This stipulation implies that l 0 as y 0.

Lets assume a simple relation that satisfies this condition:
l Ky
(9.32)
where the constant K is known as Karmans Universal Mixing Length Constant.
Substituting equation (9.32) into equation (9.31) and simplifying gives:
2
du
2 2 du
K y
dy
dy
l 2
u
*2
du
K y
dy
2
u * Ky
since u *
du
dy
dy
K
K
u
1
* du ln y * u c * ln y c
y
u
u
u
K
u
1 yu *
yu*
c
ln
(for
70 i.e. in the turbulent core region)
u* K
(9.33)
In summary, the Law of the Wall states:

yu*
u
u*

0
yu *
yu *
yu*
5:
70
70 :
C. Wassgren
u* y
(laminar sub-layer)
(9.34)
(often use Eqn. (9.36) here)
(transition region)
(9.35)
u
1 yu *
ln
c
u* K
(turbulent core)
(9.36)
u*
414
Notes:
1.
Experimental curve fits indicate that: K 0.41 and c = 5.0.
2.
Comparing laminar and turbulent pipe flows we observe that the turbulent velocity profile (a
logarithmic curve) is blunter than the laminar profile (a parabolic curve).
laminar
3.
turbulent
(The profile becomes
blunter as Re.)
A plot of the Law of the Wall velocity relations and corresponding experimental data is shown below.
Fit 1 refers to:

u
u* y
u*
Fit 2 refers to the
transition from laminar
to turbulent behavior.
Fit 3 refers to:
yu *
u
2.5ln
5.5
*
u

Fit 4 refers to:
1
u* y 7
8.74
u*

Fit 5 refers to:
u
u* y 10
11.5
u*

u
laminar
sub-layer
transition
region
turbulent region
(Figure 20.4 from Schlichting, H., Boundary Layer Theory, McGraw-Hill.)
C. Wassgren
415
Chapter 10:
Pipe Flows
NOT YET COMPLETE
1.
2.
3.
4.
5.
6.
7.
Entrance Region
Fully Developed Laminar Circular Pipe Flow
Fully Developed Turbulent Circular Pipe Flow
Moody Chart
Other Losses
Extended Bernoulli Equation
Pipe Systems
C. Wassgren
Chapter 10: Pipe Flows
416
1.
Entrance Region
inviscid core
pipe
diameter, D
entrance
length, L
fully developed flow
The shaded regions are where viscous stresses are

important (the boundary layer).
The flow in the entrance region is complex and will not be investigated here. Experiments have shown that
the dimensionless length of the entrance region depends on whether the entering flow is laminar or
turbulent, with:
laminar flow:
L/D 0.06 ReD
(10.1)
turbulent flow: L/D 4.4 ReD1/6
(10.2)
For many engineering flows:
(10.3)
104 < ReD < 105 20 < L/D < 30
2.
Fully Developed Laminar Circular Pipe Flow (Poiseuille Flow)
Consider the steady flow of an incompressible, constant viscosity, Newtonian fluid within an infinitely

1.
2.
The flow is steady.
3.
4.
There are no body forces. f r f f z 0
0 and u 0
0
t
ur u z
0
z
z

1 rur 1 u u z
0
r r
r
z
rur constant
ur 0 (call this condition #5)
C. Wassgren
417
(10.4)
(10.5)
(10.6)

1 u z 1 2 u z 2 u z
u u u
u
p
u
2 fz
z ur z z u z z
r
2
2
r
z
z
z
r
t
r r r r
u z
u z u u z
u z
p
1 u z
ur
uz
r r r r
r
t
r
z
z
0 (#5)
0 (#3)
0 (#1)
0 (#2)
2
2
1 uz uz
fz
r 2 2
z 2

0 (#4)
0 (#1)
0 (#3)
r dp
dz
du z r 2 dp
r
c1
dr 2 dz
(10.7)
d du z
r
dr dr
(10.8)
r 2 dp
c1 ln r c2
(10.9)
4 dz
uz
Now lets apply boundary conditions to determine the unknown constants c1 and c2. First, note that the
fluid velocity in a pipe must remain finite as r0 so that the constant c1 must be zero (this is a type of
kinematic boundary condition). Also, the pipe wall is fixed so that we have uz(r=R)=0 (no-slip condition).
After applying boundary conditions we have:
R 2 dp
r2
(10.10)
uz
1 2 Poiseuille Flow in a Circular Pipe
4 dz R
Notes:
1.
The velocity profile is a paraboloid with the maximum velocity occurring along the centerline. The
average velocity in the pipe is found from:
u
1
R2
r R
u z 2 rdr
r 0
R 2 dp D 2 dp

8 dz 32 dz
umax
(10.11)
where umax is the maximum fluid velocity and D is the pipe diameter.
2.
The volumetric flow rate through the pipe is:

D 4 dp
Q u 4 D 2

128 dz
C. Wassgren
(10.12)
418
3.
We can determine stresses using the constitutive relations for a Newtonian fluid. The shear stress that
the pipe walls apply to the fluid, w, is:
R dp 4 u
(10.13)
w
2 dz
R
where u is the average velocity in the pipe. Note that an alternate method for determining the average
wall shear stress, which in this case is equal to the exact wall shear stress, is to balance shear forces
and pressure forces on a small slice of the flow as shown below.
dp
dz R 2
p
dz
p R 2
dz
w 2 Rdz
z
dp 2
dz R w 2 Rdz
0 p R 2 p
dz
R dp
w
(The same answer as before!)
2 dz
(10.14)
(10.15)
In engineering applications, it is common to express the average shear stress in terms of a

dimensionless (Darcy) friction factor, fD, which is defined as:
4
64
(10.16)
f D 1 w 2 64
uD Re
2
where D=2R is the pipe diameter and Re is the Reynolds number. The Darcy friction factor commonly
appears in the Moody chart for incompressible, viscous pipe flow. Note again that this solution is only
valid only for a laminar flow. The condition for the flow to remain laminar is found experimentally to
be:
uD
(10.17)
Re
2300
4.
Lets re-write Eqn. (10.11):

p
p+p
D 2 dp
D 2 p
(10.18)
u
L
32 dz
32 L
where, in the fully developed region, the pressure gradient remains constant and so we may write dp/dz
as p/L where p is the pressure drop over a length L of the pipe. Re-arranging Eqn. (10.18) and
dropping the absolute value symbol for convenience:
32 uL
(10.19)
p
D2
Make the previous equation dimensionless by dividing through by the dynamic pressure based on the
average flow speed.
64 L 64 L
p
(10.20)

2
1
uD D Re D D
2
The dimensionless pressure drop is also referred to as the loss coefficient, k. Hence, for a laminar
flow, the loss coefficient corresponding to the viscous stresses due to the pipe walls is:
klaminar,
wall stresses
64
Re D
C. Wassgren
L
L
fD
D
D
(10.21)
419
3.
Fully Developed Turbulent Circular Pipe Flow
Turbulent Flow in a Smooth (but Frictional) Pipe

The volumetric flow rate in a smooth pipe for turbulent flow may be estimated by integrating the time
averaged velocity profile, modeled using the Law of the Wall, over the cross-sectional area of the pipe. As
an engineering approximation, well neglect the influence of the pipe curvature and the presence of the
opposite side of the pipe in this estimation.
r R
R
y
1 yu *
R
r
(10.22)
Q u 2 rdr 2 u * ln
c R y dy

r 0
0
where 0.41 and c = 5.0. Note that weve switched coordinates so we can integrate out from the wall to
the centerline. Also, were neglecting the laminar sub-layer velocity profile since it is typically very thin.
The average velocity in the pipe is thus:
Ru *
V
Q
2.44
ln
V
1.34
u*
R2

(10.23)
Recall that
8 u*
8
f D w2
V
V2
Thus,
V
8
u*
fD
since u *
(10.24)
(10.25)
Also note that:

Ru *
Thus,
Ru *
R w
1
Re D
2
f DV 2 1 2 RV
8
2
fD
8
(10.26)
fD
8
(10.27)

1
fD
8
2.44 ln Re D
1.34
fD
8
2
1
1.99 log10 Re D
fD
(10.28)
f D 1.02
(10.29)
where a log10 is used in place of the ln in the previous equation. Prandtl derived Eqn. (10.29) but
altered the constants slightly to better fit the data:
1
2.0 log10 Re D
fD
f D 0.8 Friction factor for turbulent flow in a smooth pipe.
(10.30)
Equation (10.30) is implicit in fD, which means that an iterative approach must be taken in order to solve
for fD as a function of ReD. A number of approximations to this relation have been proposed that are easier
to solve. For example, Blasius, a student of Prandtls, suggested the following approximation:
0.316
fD
Valid for 4000 < ReD < 105.
(10.31)
1
Re D4
C. Wassgren
420
Turbulent Flow in a Very Rough Pipe

The roughness of the pipe walls can significantly affect the friction factor for turbulent flows (roughness
has a negligible effect on the friction factor for laminar flows). Recall from the Law of the Wall that the
time averaged velocity in the laminar sub-layer is:
u
yu *
yu *
for
0
5
(10.32)
u*
Thus, the thickness of the laminar sub-layer, LSL, is:
LSL u *
5
5 LSL *
(10.33)
u
Since,
f
u * w V D (Refer to Eqn. (10.25).)
(10.34)
8
we have
5 8
5
8
5
8
LSL
LSL
(10.35)
V fD
D
VD f D Re D f D
LSL
D
14.1
Re D
(10.36)
fD
Thus, if the wall roughness, , is much smaller than the laminar sub-layer thickness, then well still have a
laminar sub-layer and the flow wont be significantly affected by the wall roughness, i.e. we may treat the
wall as being smooth (but still frictional).
However, if >> LSL, then the laminar sub-layer will be destroyed and the wall roughness becomes the
new length scale for use in the Law of the Wall, i.e.
u
y
(10.37)
fcn
*
u

Following the same analysis as that for turbulent flow in a smooth pipe, but using y/ in place of yu*/, we
obtain:

1
2.0 log10 D Friction factor for turbulent flow in a very rough pipe.
(10.38)
3.7
fD
Note that Eqn. (10.38) is independent of ReD.

Turbulent Flow in a Rough Pipe
For the transitional regime where /D is between smooth and very rough, empirical formulas in which
the friction factor is a function of both /D and ReD have been developed:
1
2.51
2.0 log10 D
Friction factor for turbulent flow in a rough pipe.
(10.39)
3.7 Re D f D
fD
or, the explicit empirical formula,

1.11

1
6.9
1.8log10
D Friction factor for turbulent flow in a rough pipe.
(10.40)
Re
3.7
fD
D
C. Wassgren
421
4.
Moody Chart
The previous friction factor relations have been summarized into a single plot known as the Moody chart,
which is shown below.
(Figure from Fox, et al., Introduction to Fluid Mechanics, Wiley.)
C. Wassgren
422
Notes:
1.
For Reynolds numbers less than 2300, one may use either the analytical expression for the friction
factor:
64
fD
Re D
or the Moody chart.
2.
Reynolds numbers between approximately 2300 and 4000 correspond to the transitional regime
between laminar and turbulent flow. The gray region in the Moody chart reflects the fact that the
friction factor can vary significantly in this region.
3.
The fully rough zone in the Moody chart is a region where the friction factor is a weak function of
the Reynolds number, but a strong function of the relative roughness (refer to the second half of
Section 3 of these notes). If the Reynolds number of a flow is unknown, but is expected to be large,
it is often helpful to assume that the flow is in the fully rough zone as an initial first guess.
4.
The roughnesses of various types of pipe materials have been compiled into tables such as the
following one.
Material (new)
[ft]
[mm]
riveted steel
0.003 0.03
0.9 9.0
concrete
0.001 0.01
0.3 3.0
wood stave
0.0006 0.003 0.18 0.9
cast iron
0.0085
0.26
galvanized iron
0.0005
0.15
asphalted cast iron
0.0004
0.12
commercial steel or wrought iron
0.00015
0.046
drawn tubing
0.000005
0.0015
glass
smooth
smooth
C. Wassgren
423
Example:
1.
2.
3.
Using the Moody chart, determine the friction factor for a Reynolds number of 105 and a relative
roughness of 0.001.
What is the friction factor for a Reynolds number of 1000?
What is the friction factor for a Reynolds number of 106 in a smooth pipe?
SOLUTION:
1.
2.
3.
The friction factor is fD 0.0225 (Follow the red lines in the following figure.)
Since the Reynolds number is less than 2300, we can use the exact laminar flow relation:
64
fD
fD = 0.064.
(10.41)
Re D
Alternately, we could use the Moody chart by following the blue lines in the following figure.
The friction factor is fD 0.012. (Follow the green lines in the following figure.)
C. Wassgren
424
5.
Other Losses
The loss due to the viscous resistance caused by the pipe walls is referred to as a major loss. Pressure
losses may occur due to viscous dissipation resulting from fluid interactions with other parts of a pipe
system such as valves, bends, contractions/expansions, inlets, and connectors. These losses are known as
minor losses. The names can be misleading since its not uncommon in pipe systems to have most of the
pressure loss resulting from the minor losses (e.g., a pipe system with a large number of bends and valves,
but short sections of straight pipe).
What causes these minor losses? The pressure loss results primarily from viscous dissipation in regions
with large velocity gradients, such as in a recirculation zone as shown in the figure below.
recirculation zone
A closely related phenomenon known as the vena contracta acts to effectively reduce the diameter at
entrances and bends. The recirculation zone also results in a pressure loss.
recirculation zone
Reffective
Although minor loss coefficients can be determined analytically for certain situations, most frequently the
loss coefficient for a particular device is found experimentally. Essentially, one measures the pressure drop
across the device, p, and forms the loss coefficient, k, using:
p
k 1
(10.42)
2
2 V
where is the fluid density and V is the average speed through the device. Many tables with
experimentally determined loss coefficients have been generated.
Notes:
1.
When using a loss coefficient, it is important to know what velocity has been used to form the
coefficient. For example, the loss coefficient for a contraction is typically based on the speed
downstream of the contraction, while the loss coefficient for an expansion is based on the speed
upstream of the expansion.
2.
Minor losses are sometimes given in terms of equivalent lengths of pipe. An equivalent minor loss
of 10 pipe diameters worth of a particular type of pipe means that the major loss caused by a pipe of
that type, 10 diameters in length will give the same pressure loss as the minor loss. Thus, a loss
coefficient and equivalent pipe length, Le, can be related by:
L
(10.43)
k fD e
D
C. Wassgren
425
6.
Extended Bernoulli Equation
Recall from previous notes that conservation of energy may be written as:
z
p
p
V2
V2
z H L ,1 2 H S ,1 2
2g
2g
g
2 g
1
g
(10.44)
where each of the terms in the equation has dimensions of length. The 1 and 2 subscripts refer to the
inlet and outlet conditions, respectively. The individual terms are referred to as:
p
pressure head
g
V2
2g
velocity or dynamic head
z
elevation head
H L ,1 2 head loss
H S ,1 2 shaft head
Recall that the in the velocity head term is the kinetic energy correction factor which accounts for the fact
that an average speed is used in the Extended Bernoulli Equation rather than the real velocity profile (refer
back to earlier notes concerning conservation of energy). A value of = 2 is used for laminar flows while
= 1 is typically assumed for turbulent flows (actually, 1 as ReD .)
The head loss term, HL, accounts for both major and minor pressure losses, and may be written as:
V2
(10.45)
H L ,1 2 ki i
2g
i
where the subscript i accounts for every loss in the pipe system. Recall that the major loss coefficient
may be written as:
L
(10.46)
kmajor f D
D
The shaft head term, HS, accounts for the pressure addition (reduction) resulting from the inclusion of
devices such as pumps, compressors, fans, turbines, and windmills. Those devices that add head to the
flow are positive (e.g. pumps), while those that extract head are negative (e.g. turbines). The shaft head
term may be written in terms of shaft power, W S , as:
W
(10.47)
H S ,1 2 S ,1 2
mg
where m is the mass flow rate through the device.
C. Wassgren
426
7.
Pipe Systems
Most pipe flow problems can be classified as being of one of three types:
Type I:
The desired flow rate is specified and the required pressure drop must be determined.
Type II:
The desired pressure drop is specified and the required flow rate must be determined.
Type III:
The desired flow rate and pressure drop are specified and the required pipe diameter must
be determined.
Serial Systems
Parallel Systems
C. Wassgren
427
Chapter 11:
Fluid Machinery
1.
2.
3.
4.
5.
6.
Types of Fluid Machinery

Elementary Pump Theory
Net Positive Suction Head
Pump Similarity
Specific Speed
System Characteristic Curves and Pump Selection
C. Wassgren
Chapter 11: Fluid Machinery
428
1.
Types of Fluid Machinery
Two basic categories of fluid machines:

1. those that do work on the fluid:
a. pumps
(used for liquids)
b. fans
(used for gases/vapor; p < a few inches of H20)
c. blowers
(used for gases/vapor; a few inches of H2O < p < 1 atm)
d. compressors
(used for gases/vapor; p > 1 atm)
2.
those that extract work from the fluid:

a. turbines
Types of Pumps
1. Positive Displacement Pumps (PDPs)
a. force fluid movement using changes in volume
b. e.g. reciprocating piston engines, heart, gear pumps, rotating screw pumps, bellows
c. typically produce a periodic flow rate
d. large p (pressure rise) possible but usually have a small Q (flow rate)
2.
Dynamic Pumps
a. no closed volumes as in PDPs
b. p due to changes in fluid momentum
c. e.g. axial flow and radial flow pumps (aka turbomachines), jet pumps, electromagnetic pumps
d. (pdynamic pumps) typically < (pPDP)
e. (Qdynamic pumps) typically > (QPDP)
The operating principles for a wide variety of pumps may be viewed at:
http://www.animatedsoftware.com/elearning/All%20About%20Pumps/aapumps.html
C. Wassgren
429
Some Examples of Positive Displacement Pumps

This section remains incomplete.
Gear Pumps
External Gear Pump
(Image from: http://www.pumpschool.com/principles/external.htm )

Internal Gear Pump
(Image from: http://media-2.web.britannica.com/eb-media/56/3656-004-5CA4041D.gif )

Gear pumps are often used in automatic transmissions.
Lobe Pump
(Image from: http://www.megator.co.uk/lobe_pump.htm )

Two and three lobes
Commonly used in diesel superchargers.
C. Wassgren
430
Vane Pump
(Image from: http://www.pumpschool.com/principles/vane.htm )

Centrifugal force or springs are used to push out the vanes.
Often used in power steering pump, automatic transmissions
Screw Pump
(Image from: http://en.wikipedia.org/wiki/File:Archimedes_screw.JPG )

Archemiedes Screw Pump 2000 yrs old, often used for irrigation
C. Wassgren
431
(Progressive) Cavity Pump
(Top image from: http://www.animatedsoftware.com/pumpglos/progrssv.htm

Bottom image from: http://www.roymech.co.uk/Related/Pumps/Rotary%20Positive%20Displacement.html )
Piston Pump
Wobble Plate Piston Pump
(Image from: http://www.roymech.co.uk/Related/Pumps/Rotary%20Positive%20Displacement.html )
C. Wassgren
432
Wolfhart Principle Pump
(Images from: http://www.allstar.fiu.edu/aero/wolfhart_pump_principle.htm )

Ball Piston Pump
(Image from: http://www.animatedsoftware.com/pumpglos/ballpist.htm )

Bent-Axis Piston Pump
C. Wassgren
433
Radial piston pump

Rotary cam pump
(Image from: http://www.labpump.co.kr/data/aboutpump.htm )

Swash plate piston pump
C. Wassgren
434
Diaphragm Pump
(Image from: http://en.wikipedia.org/wiki/File:Bomba_diafragma.jpg )
Finger pump
(Image from: http://www.animatedsoftware.com/pumpglos/fingerpu.htm )
Peristaltic pump
(Left image from: http://en.wikipedia.org/wiki/File:Eccentric_pump.gif

Right image from: http://www.roymech.co.uk/Related/Pumps/Rotary%20Positive%20Displacement.html )
Often used as fuel pumps.
C. Wassgren
435
Some Examples of Dynamic Pumps

Axial Pump
Propeller pump
motor
discharge elbow
propeller
inlet plenum
(Image from:
http://www.sulzerpumps.com/Portaldata/9/Resources/brochures/power/vertical/JP_Vertical_E00635.pdf )
Radial Pump
(Left image from: http://commons.wikimedia.org/wiki/File:CetriFugal_Pump.jpg

Right image from: http://www.motorera.com/dictionary/pics/r/Radial-flow_pump.gif )
C. Wassgren
436
Mixed Pump
guide vanes
volute
(Image from: http://www.fao.org/docrep/010/ah810e/AH810E07.htm )
C. Wassgren
437
Jet Pumps
(Image from: http://www.fao.org/docrep/010/ah810e/AH810E07.htm )
Ram Pump
(Image from: http://www.lifewater.ca/ram_pump.htm )
C. Wassgren
438
Air Lift Pumps
(Images from: http://www.airliftpump.com/index.htm )
C. Wassgren
439
Elementary Pump Theory

To determine the work that the pump does on the fluid passing through it, well use the Moment of
Momentum Equation which relates torque to momentum fluxes.
Consider flow through the following rotating pump:
Vrb2
front view of pump impeller
V2
V2 , outgoing fluid velocity

w/r/t a fixed FOR
U2 = r2
Vrb1 V1
V1, incoming fluid velocity

w/r/t a fixed FOR
U1 = r1
r1
r2
Here:
1, 2
V1, V2
U1, U2
Vrb1, Vrb2
r2 r1
entrance/exit blade angles w/r/t the hub

fluid velocities w/r/t a FOR fixed to the ground
blade velocities w/r/t a FOR fixed to the ground
fluid velocities w/r/t to the blade
From geometry, we have:

V1 U1 Vrb1
(11.1)
V2 U 2 Vrb2
To determine the torque, T, that must be applied to the shaft in order to rotate the impeller at angular
velocity, , we use the moment of momentum relation:
V2
d
dV
d
r
A
(11.2)
rel
on CV
dt CV
CS
V
1
Assume the following:

1. steady flow
2. uniform flow between blades
3. incompressible flow
4. the only moment applied to the CV is that due to shaft torque, Ton CV (which points in the same
direction as the shaft rotation, )
5. the fluid velocities are measured w/r/t an inertial FOR
Using these assumptions and simplifying the moment-of-momentum equation:
Ton CV m r2Vt 2 rV
Eulers Turbomachinery Equation
1 t1
(11.3)
where Vt1 and Vt2 are the absolute fluid velocities tangent to the impeller hub.
C. Wassgren
440
Notes:
1. The power required to drive the impeller is:
Won CV Ton CV
W
m r V rV
on CV
2 t2
1 t1
but U1 = r1 and U2 = r2 so that:

Won CV m U 2Vt 2 U1Vt1
In terms of the head added to the fluid:
W
U V U1Vt1
H added on CV 2 t 2
mg
g
to CV
(shaft head)
(11.4)
2.
Only the absolute fluid velocity tangential to the impeller contributes to the increase in fluid head.
3.
For an idealized centrifugal pump, the incoming flow has no tangential component Vt1 = 0
UV
H added 2 t 2
(shaft head for an idealized centrifugal pump)
g
to CV
4.
(11.5)
Consider the fluid velocity vector geometry at the exit of the blade:
Vrb2
V2
Vt2
Vn2
Vrb2
U2 = r2
U2-Vt2
Vn 2
tan 2
U 2 Vt 2
Vt 2 U 2 Vn 2 cot 2
(11.6)
Substituting into the shaft head relation for an idealized centrifugal pump (equation (11.5)) gives:
U U Vn 2 cot 2 U 22 U 2Vn 2 cot 2
(11.7)
H added 2 2
g
g
g
to CV
Note that the volumetric flow rate through the pump is related to the radial fluid velocity from COM:
Vn2
Vn2
Q Vn 2 2 r2
b2
(11.8)
impeller impeller
circumference thickness
b2
C. Wassgren
441
Substituting and noting that U2 = r2gives:

2
r2 r2Q cot 2
H added
g
g 2 r2 b2
to CV
r2
cot 2
Q
g
to CV
g 2 b2
theoretical head rise across an idealized centrifugal pump
H added
(11.9)
Notes:
a. Equation (11.9) is an equation of a line.
H
shut-off r2
g
head
2 > 90 (forward curved)
Forward curved blades are not

used very often due to flow
instabilities.
2 = 90 (radial)
2
2 < 90 (backward curved)

Q
b.
In an actual flow, losses occur within the pump due to friction with the blades (which varies with
Q2), flow separation, impeller blade-shroud clearance flows, and other 3D flow effects.
H
ideal case
friction losses (~ Q2)
actual
other losses
Q
A quadratic curve is often used to fit experiment pump head curves: H = H0 A Q2.
5.
Pump efficiency is defined as:
mgH
water or hydraulic horsepower (power you get out)
p
T
brake horsepower (power you put in)
a.
b.
typical pump efficiencies: p = 85% (well-designed) 60% (poorly-designed)

As pump size , the ratio of surface area to volume frictional losses p.
C. Wassgren
442
best efficiency point

(BEP)
power put in
(From Munson, B.R., Young, D.F., and Okiishi, T.H., Fundamentals of Fluid Mechanics, 3rd ed., Wiley.)
C. Wassgren
443
Example:
An idealized centrifugal water pump is shown below. The volumetric flow rate through the pump is 0.25
ft3/s and the angular speed of the impeller is 960 rpm. Calculate the power required to drive the pump.
55
960 rpm
3 in.
11 in.
0.75 in.
0.25 ft3/s
SOLUTION:
Apply conservation of angular momentum to the fixed control volume shown below.
2
r2
r1
b2
V2
d
dt
r u dV r u u
CV
CS
C. Wassgren
rel
dA
r F
CV,CS
444
where
d
dt
r u dV 0
(steady flow)
CV
exiting fluid velocity w/r/t ground
V2 sin 2 2 r2 b2
r u urel dA r2e r V2 sin 2e r cos 2e r2e

CS
velocity of fluid exit area

exiting the CV
r22 r2V2 cos 2 V2 sin 2 2 r2b2 e z
(Note that there is no flux of angular momentum at the inlet since r1 and V1 are parallel.)
r F Te z
CV,CS
T r22 r2V2 cos 2 V2 sin 2 2 r2b2
(11.10)
The velocity V2 can be related to the volumetric flow rate, Q, by considering the flow at the impeller exit.
Q V2 sin 2 2 r2b2
V2
Q
2 r2 b2 sin 2
(11.11)
gQ
Q
T r22
2 b2 tan 2 g
(11.12)

(g)H20 = 62.4 lbf/ft3
g
= 32.2 ft/s2
Q
= 0.25 ft3/s
b2
= 0.75 in = 6.25*10-2 ft
= 960 rpm = 100.5 rad/s

2
= 55
r2
= 5.5 in = 4.58*10-1 ft
T = 10.0 ftlbf
The power required to drive the impeller is:
W T
W 1005 ftlbf/s = 1.83 hp
C. Wassgren
445
The problem could have also been worked out using velocity polygons. Since the absolute inlet velocity
has no tangential component:
UV
H 2 t 2 where U 2 r2
g
and
W mgH
r2Vt 2
mg
W
g
Use a velocity polygon at the exit to determine Vt2.
Vrb2
V2
Vn2
Vrb2
Vt2
U2 = r2
U2-Vt2
From the geometry:

Vt 2 U 2 Vn 2 cot 2
where Vn2 is found from conservation of mass:
Q
Vn 2
2 r2 b2
U2
= 46.3 ft/s
= 1.39 ft/s
Vn2
H
= 65.1 ft
W 1005 ftlbf/s = 1.83 hp (Same answer as before!)
C. Wassgren
446
2.
Net Positive Suction Head (NPSH)
Along the suction side of the impeller blade near the pump inlet are regions of low pressure.
rotation
direction
blade
suction side, region

of low pressure
fluid velocity at inlet
relative to blade
If the local pressure is less than the vapor pressure of the liquid, then cavitation will occur:
p pvapor cavitation
Recall that cavitation is boiling (liquid turning to vapor) occurring when the pressure is less than the
liquids vapor pressure. Cavitation can not only significantly decrease the performance of a pump, but it
can also cause pump damage, vibration, and noise. Vapor bubbles caused by cavitation move into regions
of higher pressure, collapse violently, and produce localized regions of very high pressure that can chip
away at surfaces. The result is that the pump material erodes away. One can often hear when cavitation
occurs because of the noise generated by the collapsing vapor bubbles.
Lets define a quantity that will aid us in determine when cavitation in a pump might occur:
p V2
pv
net positive suction head, NPSH

g 2 g s g
(11.13)
Notes:
1. The first term in NPSH is the head at the suction side of the pump near the impeller inlet. This is the
region where we expect to have the lowest head. (Note that we define our reference plane for elevation
along the centerline of the pump z = 0.)
2.
The second term in NPSH is the vapor pressure head. This is the head when the liquid turns to vapor.
The vapor pressure is typically given in terms of an absolute pressure so the suction pressure should
also be an absolute pressure. Note that vapor pressure increases as temperature increases.
3.
(NPSH)R Net Positive Suction Head Required to avoid cavitation. This quantity is a pump property
and is determined experimentally.
4.
(NPSH)A Net Positive Suction Head Available to the pump. This quantity is a system property and
can be determined via analysis or experiments. NPSHA is related to the total head available to the
pump at the pump inlet (minus the vapor head).
5.
We must have (NPSH)A > (NPSH)R to avoid cavitation. Regulations typically recommend at least a
10% margin for safety. For critical applications such as for power generation or flood control, a 100%
margin is often used.
6.
NPSHR increases with increasing flow rate since the pressure at the suction side of the pump blade near
the pump inlet will decrease (consider Bernoullis equation). Similarly, the pressure will decrease with
increasing blade rotation rate resulting in an increased NPSHR.
C. Wassgren
447
Example:
Determine the NPSHA for the following system
P
H
SOLUTION:
Choose point 1 to be on the surface of the tank and point 2 to be just upstream of the pump.
Apply the EBE from 1 to 2:
p
p
V2
V2
z H L12 H S 12
2g
2g
g
2 g
1
where
p2 ps
p1 patm
V2 Vs (assume turbulent flow = 1)
V1 0
z2 z1 H
H S 12 0
H L12 H L12
Substitute and simplify to get:

p V2
patm
z1 z2 H L12
g
g 2g
S
From the definition of NPSH we have:

p
p
NPSHA atm H H L12 v
g
g
Notes:
1. Increasing H, HL12, or pv decreases NPSHA and increases the likelihood of cavitation since the
difference between NPSHA and NPSHR is reduced.
2. Increasing patm increases NPSHA and decreases the likelihood of cavitation.
3. pv varies with temperature
4. pv is usually given in terms of absolute pressure
C. Wassgren
448
3.
Pump Similarity
Most pump performance data (H-Q curves) are given only for one value of the pump rotational speed and
one pump impeller diameter. Is there some way to determine the pump performance data for other speeds
and diameters without requiring additional testing? There is using dimensional analysis!
Recall that we typically are interested in knowing the head rise across a pump, H (=h/g where h is the
specific energy rise across the pump), power required to operate the pump, W (bhp), and pump efficiency,
, as a function of the volumetric flowrate through the pump, Q:
(11.14)
h, W , fcns Q, , , D,
where and are the fluid density and dynamic viscosity, D is the pump impeller diameter, and is the
pump rotational speed.
Performing a dimensional analysis we find the following:
, , fcns , Re
(11.15)
where
gH
dimensionless head coefficient =
dimensionless power coefficient =
efficiency =
dimensionless flow coefficient =
Re
2 D2
W
3 D 5
QgH
(11.17)
(11.18)
Reynolds number =
(11.16)
Q
D3
(11.19)
D 2
(11.20)
Notes:
1. In most pump flows, Re is very large
the variations in viscous effects from one flow to another are small
Re similarity can be neglected
a.
b.
2.
If Re is considerably different from one flow to another, e.g. pump water vs. pumping molasses,
then Re effects cannot be ignored. The flow physics for large Re, where viscous forces inertial
forces, are different than for small Re where viscous forces inertial forces.
Thus, for large Re:
, , fcns
(11.21)
QgH
W
C. Wassgren
(11.22)
449
3.
For similarity between geometrically similar flows (assuming large Re), we have the following pump
scaling laws:
Q Q
1 2

(11.23)
3
3
D 1 D 2
gH gH
1 2
2 2 2 2
(11.24)
D 1 D 2
1 2
1 2
a.
W W

3 5
3 5
D 1 D 2
(since 1 = 2, 1 = 2, 1 = 2 and = /)
(11.25)
For a given pump (D = constant) using the same fluid (, = constant) and the same gravity (g1 =
g2):
Q1 1
Q Q

Q2 2
1 2
H1 1
H H

2 2
H 2 2
1 2
(11.26)
2
(11.27)
3
W W
W1 1

3 3
W2 2
1 2
b.
(11.28)
For a given pump speed ( = constant) but varying diameters (assuming a geometrically similar
family of pumps), and using the same fluid (, = constant) and the same gravity (g1 = g2):
Q1 D1
Q2 D2
H1 D1
H 2 D2
5
W1 D1
W2 D2
(11.29)
Note that we are assuming that all length scales within the pump are scaled in the same way to
maintain geometric similarity. This is not always true in practice since pump impellers with
different diameters are often put in the same pump casing. Also, surface roughness isnt scaled
proportionally. The result is that the pump scaling laws are only approximations.
C. Wassgren
450
Example:
A centrifugal pump with a 12 in. diameter impeller requires a power input of 60 hp when the flowrate is
3200 gpm against a 60 ft head. The impeller is changed to one with a 10 in. diameter. Determine the
expected flowrate, head, and input power if the pump speed remains the same.
SOLUTION:
Since the pump speed remains the same and assuming geometrically similar pumps, the correct pump
scaling laws are:
Q1 D1
Q2 D2
H1 D1
H 2 D2
5
W1 D1
W2 D2

Q1 = 3200 gpm
D1 = 12 in
D2 = 10 in
H1 = 60 ft
W1 = 60 hp
Q2 =
H2 =
W2 =
1850 gpm
41.7 ft
24.1 hp
C. Wassgren
451
4.
Specific Speed, Ns
A useful dimensionless term results from the following combination of previously defined terms:
1
1
2
Q 2
Ns specific speed = 3
3
4 gH 4
(11.30)
Combining the terms in this manner eliminates the impeller diameter, D.
Notes:
1. It is customary to characterize a machine by its specific speed at the design point, i.e. Ns is usually only
given for the BEP operating conditions.
a. low Q, high H low Ns centrifugal pumps
b. high Q, low H high Ns axial pumps
2.
In practice (especially in the US), a combination of units are used to describe , Q, and H such that Ns
is dimensional (signified by Nsd):
N sd
rpm Q gpm
H ft
Ns and Nsd have the same physical meaning but are different in magnitude by a constant factor:
Nsd = 2733 [rpm(gpm)1/2/(ft)3/4] Ns
3.
(11.31)
Given , Q, and H, we can calculate Ns (or Nsd) and, using the following chart, determine which type
of pump would be most efficient for the given conditions.
Following are some rules of thumb:
1.
Positive displacement pumps are used for small flow rates, Q, and large head rises, H.
2.
Centrifugal pumps are for moderate H and large Q.
3.
For very large head rises, pumps are often combined in series (aka multi-stage).
4.
Axial flow pumps are for large Q and low H.
C. Wassgren
452
Example:
A small centrifugal pump, when tested at 2875 rpm with water, delivered a flowrate of 252 gpm and a head
of 138 ft at its best efficiency point (efficiency is 76%). Determine the specific speed of the pump at this
test condition. Sketch the impeller shape you expect. Compute the required power input to the pump.
SOLUTION:
The dimensional specific speed is given by:
N sd
rpm Q gpm
3
H ft 4
Nsd = 1130 rpmgpm1/2/ft3/4
The dimensionless specific speed is:

N sd
Ns
1
rpm gpm 2
2733
3
ft 4
Ns
= 0.414
The expected impeller shape is radial as shown in the figure below.
(Figure from Munson, B.R., Young, D.F., and Okiishi, T.H., Fundamentals of Fluid Mechanics, 3rd ed.,
Wiley.)
The power input to the pump is given by:
W
Wshaft fluid
where
Wfluid mgH
QgH
Wfluid = 8.80 hp
W
= 11.6 hp
(Note: 1 ft3 = 7.48 gal, 1 hp = 550 lbfft/s, and 1 lbf = 1 slugft/s2.)
shaft
C. Wassgren
453
5.
System Characteristic Curves and Pump Selection
How do we select a pump for a given system? Analyze the system to determine the shaft head required to
give a specified volumetric flow rate. Compare this result to a given pump performance curve (H-Q curve)
to determine if the pump operates efficiently at this Q. If so, then the choice of pump is appropriate.
Consider the following example:
z2
z1
P
Apply the EBE from 1 to 2:
p
p
V2
V2
z H L12 H S 12
2g
2g
g
2 g
1
where p1 = p2 = patm and V1 = V2 = 0. Thus, the head required from the pump is:
H S 12 z2 z1 H L12
(11.32)
(11.33)
Recall that:
H L12 K i
i
Vi 2
Q2
Ki i 2
2g
2 gAi
i
(11.34)
so that:
H S 12 z2 z1 CQ 2
(11.35)
where C is a constant that incorporates the loss coefficients and area ratios, and an is used since the loss
coefficients may depend on the flow velocities.
C. Wassgren
454
The conditions at which the system will operate will depend on the intersection of the system head curve
with the pump performance curve as shown in the figure below.
H
system curve
pump curve
operating point
Notes:
1. Ideally we would want the operating point to occur near the BEP for the pump.
2.
3.
For laminar flow,

64 L 64 L c
where c is a constant
K major

Re D VD D Q
HL ~ Q
system curve is: HS = c1 + c2 Q (a line instead of a parabola!)
(11.36)
The system curve may change over time due to fouling of the pipes and other factors increased
losses system curve becomes steeper
system curve after fouling

original system curve
Q
The pump curve may also change due to wear on the bearings, impeller, etc.
H
original pump curve
pump curve after wear
C. Wassgren
455
4.
Stability issues become significant when the pump has a flat or falling (defined as a performance curve
where H as Q ) performance curve.
H
rising (H as Q )
falling (H as Q )
Q
Consider perturbations to the

system from the operating point.
Hpump > Hsystem
Q
H
Hpump < Hsystem
Q
stable
system curve
pump curve
unstable
Hpump > Hsystem Hpump < Hsystem

Q
Q
Q
H
system curve
falling pump curve
pump curve
For the system shown at the left

there are three operation points!
The system may oscillate
between these points!
H
system curve
flat pump/system curves
pump curve
For the system shown at the

left, the system may drift
over a wide range of Q!
C. Wassgren
456
Example:
Water is to be pumped from one large open tank to a second large open tank. The pipe diameter throughout
is 6 in. and the total length of the pipe between the pipe entrance and exit is 200 ft. Minor loss coefficients
for the entrance, exit, and the elbow are shown on the figure and the friction factor can be assumed constant
and equal to 0.02. A certain centrifugal pump having the performance characteristics shown is suggested
as a good pump for this flow system.
a. With this pump, what would be the flow rate between the tanks?
b. Do you think this pump would be a good choice?
Kelbow = 1.5
pump
10 ft
pipe diameter = 6 in
total pipe length = 200 ft
Kexit = 1.0
Kentrance = 0.5
SOLUTION:
Apply the extended Bernoullis equation from point 1 to point 2.
pump
H
z
1
C. Wassgren
457
V2
V2
g 2 g z g 2 g z H L H S
2
1
where
p2 p1 patm (free surface)
(11.37)
V2 V1 0 (large tanks)
z2 z1 H
V2 L
f K entrance K exit K elbow (where V is the mean velocity in the pipe)

2 g D
Note that the mean pipe velocity can be expressed in terms of the volumetric flow rate.
Q
V
D2
4
8Q 2 L
H S H 2 4 f K entrance K exit K elbow

gD D
HL
(11.38)
(11.39)
For the given problem:

H
= 10 ft
g
= 32.2 ft/s2
f
= 0.02
D
= 6 in = 0.5 ft
L
= 200 ft
(Note: Kmajor = f(L/D) = 8.0)
Kentrance = 0.5
Kexit
= 1.0
Kelbow = 1.5
HS = (10 + 4.43Q2) ft
Note that [Q] = ft3/s.
(11.40)
This is the head that must be added to the fluid by the pump in order to move the fluid at the
volumetric flow rate Q.
With [Q] = gpm, Eqn. (11.40) becomes:
HS = (10 + 2.25*10-5Q2) ft
Note that [Q] = gpm.
C. Wassgren
458
(11.41)
Plot Eqn. (11.41) on the pump performance curve to determine the operating point.
system curve (Eqn. (11.41))
From the figure we observe that the operating point occurs at:
Q 1600 gpm
corresponding to a head rise and efficiency of
H 67 ft
84%
The operating efficiency is close to the optimal efficiency of 86% so this is a good pump to use.
C. Wassgren
459
Chapter 12:
Gas Dynamics
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
Introduction to Gas Dynamics

Equations of State
One-Dimensional Flow
Speed of Sound and the Mach Cone
Adiabatic, 1D Compressible Flow of a Perfect Gas
Effects of Area Change on Steady, 1D, Isentropic Flow
Adiabatic Flow with Friction (Fanno Flow)
Flow with Heat Transfer (Rayleigh Flow)
Normal Shock Waves
Flow in Converging-Diverging Nozzles
Flows with Mass Addition
Generalized Steady, One-Dimensional Flow
Oblique Shock Waves
Expansion Waves
Reflection and Interaction of Oblique Shock Waves
Reflection and Interaction of Expansion Waves
Equations of Motion in Terms of the Velocity Potential
Small Perturbation Theory
Method of Characteristics
Flow Past a Wavy Wall Using Small Perturbation Theory
Thin Airfoils in Supersonic Flow
Unsteady, 1D Compressible Flow
Description of a Shock Tube
C. Wassgren
Chapter 12: Gas Dynamics
460
1.
Introduction to Gas Dynamics
What is Gas Dynamics?

Gas dynamics is a branch of fluid mechanics that examines the dynamics of compressible fluid flows and
of gases in particular.
What is the motivation for studying compressible fluids?
Although topics regarding compressible fluid mechanics have been studied since the 1800s, few scientists
and engineers were interested in the topic apart from those studying ballistics and steam turbine design. It
wasnt until WW II with the development of high speed planes, rockets, and energetic explosives that the
study of compressible flows became widespread. Ever since, the understanding of compressible fluid
mechanics has been important in the development of not only the previously mentioned topics, but also of
jet engines, rocketry, re-entry spacecraft, gas pipelines, combustion, and gas turbines.
What is special about compressible fluids?
Compressibility of a fluid results in several important phenomena that are not observed in incompressible
fluids. Two of the most significant of these phenomena are shock waves and choked flow conditions.
Both of these phenomena are the result of the fact that in compressible fluids, pressure disturbances
propagate at a finite speed. For example, if one claps their hands, the pressure disturbance caused by the
colliding hands propagates into the surrounding atmosphere with a finite speed (equal to the speed of
sound). Thus, a finite amount of time passes before the surrounding air recognizes the effects of the
clapping hands. In contrast, in a truly incompressible fluid, pressure disturbances propagate at an infinite
speed. Thus, pressure disturbances are felt instantly everywhere in the fluid domain.
The fact that disturbances travel at finite speed raises the question of what happens if the cause of the
pressure disturbance travels faster than the pressure disturbance itself? As an example, lets consider an
aircraft flying in the atmosphere. As the aircraft moves slower than the disturbances propagate, pressure
disturbances travel ahead of the aircraft and inform the air in front that the aircraft is about to arrive.
Thus, the air can move smoothly out of the way as the aircraft approaches. However, if the aircraft travels
faster than the speed of propagation, then the air in front of the aircraft cant move out of the way and
begins to pile up in front of the aircraft. The result is the formation of a shock wave across which there is
a rapid change in the air pressure, temperature, density and velocity.
Now lets consider a different situation. Imagine a large, pressurized tank with a converging nozzle which
empties into another large tank (refer to the figure below). While holding the pressure in the left tank
constant, lets begin to reduce the pressure in the right hand tank.
pressure
in this
tank
remains
constant
pressure
in this
tank can
vary
When we lower the pressure in the right hand tank, a pressure disturbance propagates upstream to the
constant pressure tank and informs the fluid upstream that the pressure in the right hand tank has
dropped. As a result, the flow rate between tanks increases. As we continue to lower the pressure in the
right hand tank, the flow rate continues to increase until we reach a velocity in the converging nozzle where
the fluid velocity is equal to the speed at which pressure disturbances propagate. Now if we continue to
lower the pressure in the right hand tank, that pressure information can no longer propagate upstream since
C. Wassgren
461
the fluid is flowing in the opposite direction at the same speed. Thus, we have a choked flow condition
where we can no longer increase the flow rate between the tanks.
In addition to these two phenomena, compressible flows have other counter-intuitive behaviors regarding
how the fluid velocity varies with the area through which the fluid flows and how the velocity is affected
by frictional effects. Well investigate all of these phenomena later in this course.
What tools are required to study compressible fluid mechanics?
Several basic concepts are used in studying compressible fluid mechanics. These include:
conservation of mass
linear momentum equation
conservation of energy (1st law of thermodynamics)
2nd law of thermodynamics
equations of state (e.g., the ideal gas law)
various concepts from thermodynamics
In addition, well require knowledge of calculus, vector calculus, and differential equations (ODEs and
PDEs).
C. Wassgren
462
2.
Equations of State
Data for the various properties of a particular substance or class of substances can typically be found in
thermodynamic property tables. Often, it is more convenient and informative to establish analytical
relations between these various data. Equations of state describe the relations between the various
properties of a substance or class of substances. Note that there can be more than one equation of state.
For example, when discussing an ideal gas, well consider both a thermal and caloric equation of state.
Before discussing equations of state for various substances, we should first define a few terms. First, well
be concerned only with pure substances in these notes:
A pure substance is one that has a fixed chemical composition throughout. A pure substance can
exist in more than one phase, but its chemical composition must be the same in each phase. Examples
of pure substances include distilled water, nitrogen, helium, and carbon dioxide. A uniform mixture of
non-reacting gases can be regarded as a pure substance. Note that if one of the component gases
changes into a liquid phase (for example if the mixture is cooled), the liquid would have a different
composition than the gas phase and thus the system could no longer be considered a pure substance
since the system would not have a uniform composition.
Another important point to address is the question of how many properties we need to specify before we
uniquely define the system state. We know from experience that not every system property must be
specified before we can identify the state of a system. In fact, we observe that many properties are related.
Thus, only a subset of properties needs to be given to uniquely determine a systems state. In our studies,
well consider simple, compressible systems, defined as those systems where electrical, magnetic, surface
tension, gravitational, and motion effects are negligible. The state postulate for a simple, compressible
system states that only two independent properties are required to uniquely define the systems state. If
additional effects are significant, e.g. gravitational forces and accelerations, then additional properties are
required, e.g. elevation and velocity.
C. Wassgren
463
Relations for an Ideal Gas

One particularly important class of substances that is very important in gas dynamics is the ideal gas. An
ideal gas is a model describing the behavior of real gases in the limit of zero pressure and infinite
temperature (i.e., zero density). It does not account for the interaction between molecules of the gas (e.g.
inter-molecular forces). Nevertheless, the ideal gas model is a reasonably accurate model for gases where
the system pressure is less than 0.05 times the critical pressure or the system temperature is more than twice
the critical temperature (to be discussed).
thermal equation of state
The thermal equation of state for an ideal gas is given by what is commonly referred to as the
ideal gas law:
p RT
(12.1)
where p is the absolute pressure of the gas, is the gas density, R is the gas constant for the gas of
interest, and T is the absolute temperature.
Notes:
1. Absolute pressures and temperatures must be used when using the ideal gas law.
2. The gas constant, R, will be different for different gases. The gas constant can be determined
in terms of the universal gas constant, Ru.
R
R u
(12.2)
M
where Ru=8314 (Nm)/(kgmoleK)=1545.4 (ftlbf)/(lbmmoleR) and M is the molecular mass
of the gas. The gas constant for air is R=287 (Nm)/(kgK)=53.3 (ftlbf)/(lbmR). Note that
Mair = 28.98 kg/kgmol.
3. The compressibility factor, Z, is defined as:
pv
p
(12.3)
Z
RT RT
If Z 1 for a gas, then it can be modeled well with the ideal gas model. The compressibility
factor, Z, is plotted below for a variety of substances as a function of the reduced pressure,
p/pcr, and reduced temperature, T/Tcr, where pcr and Tcr are the critical pressure and
temperature. Note that the critical temperature is the temperature above which a gas cannot
be liquefied no matter how large a pressure is applied. The critical pressure is the minimum
pressure for liquefying a gas at the critical temperature.
Notes:
1. For values of p/pcr < 0.05 or
T/Tcr > 2, Z 1 so that in this
range the ideal gas model
works well.
2. The table below gives the
critical temperature and
pressure values for various
substances:
TR = T/Tcr
PR = p/pcr
C. Wassgren
464
Gas
air
He
H2
N2
O2
CO2
CO
Tcr [K]
132.41
5.19
33.24
126.2
154.78
304.20
132.91
pcr [atm]
37.25
2.26
12.80
33.54
50.14
72.90
34.26
T/Tcr > 2
Tair > 265 K
p/pcr < 0.05
pair < 1.86 atm
caloric (aka energy) equation of state

Before discussing the caloric equation of state, we must first define the property known as specific
heat.
specific heat
Recall that earlier it was mentioned that internal energy, sensible energy in particular, is
related to temperature. Furthermore, we know from experience that some materials heat up at
different rates than others. For example, 4.5 kJ of energy added to a 1 kg mass of iron will
raise the irons temperature from 20 C to 30 C. To raise 1 kg of water from 20 C to 30 C,
however, requires 41.8 kJ; about 9 times the amount of energy than is required to raise the
irons temperature an equivalent amount.
The property that quantifies the energy storage capability of matter is called the specific heat.
The specific heat of a substance is the energy required to raise the temperature of a unit mass
of a substance by one degree. In general, the energy required to raise the temperature of a
substance will depend on the process path. Two particular processes of interest are where the
systems volume is held constant while energy is added and where the pressure in the system
is held constant while energy is added.
The specific heat at constant volume, cv, is the energy required to raise the temperature of a
unit mass by one degree during a constant volume process. The specific heat at constant
volume is defined as:
u
cv
(12.4)
T v
The specific heat at constant pressure, cp, is the energy required to raise the temperature of a
unit mass by one degree during a constant pressure process. The specific heat at constant
pressure is defined as:
h
cp
(12.5)
T p
Note that cp is always greater than cv since at constant pressure, the system is allowed to
expand and the energy for this expansion work must also be supplied to the system.
Now lets return to our discussion of the caloric equation of state for an ideal gas. Since an ideal
gas is considered a simple, compressible system, the internal energy, u, is uniquely determined by
two properties. Here well use the properties of temperature, T, and specific volume, v(=1/):
u u T , v
(12.6)
so that any change in the internal energy is given by:
u
u
(12.7)
du
dT
dv
T v
v T
Utilizing our definition for specific heat given in equation (12.4) we have:
u
(12.8)
du cv dT
dv
v T
The second term on the right hand side of equation (12.8) is zero for an ideal gas. (This can be
shown using Maxwells relations; a topic not addressed in these notes. See for example, Moran
and Shapiro, Fundamentals of Engineering Thermodynamics, 3rd ed., Wiley, Section 11.4.2.)
Thus, the internal energy of an ideal gas, u, is a function only of the temperature:
du
u u T du
dT
(12.9)
dT
Using the definition given in equation (12.4) we have:
C. Wassgren
465
du cv dT
(12.10)
Integrating both sides and noting that the specific heat can be a function of temperature in general:
T
u uref
c T dT
(12.11)
Tref
The enthalpy is also only a function of temperature for an ideal gas as shown below:
p
h u u RT h h(T )
(12.12)
Taking the derivative of enthalpy with respect to the temperature and using the definition given in
equation (12.5) we see that:
dh
h
dT c p dT
(12.13)
dT
T
h href
T dT
(12.14)
Tref
where cp is the specific heat at constant pressure which can, in general, be a function of
temperature. In addition,
dh du RdT cv dT RdT
cv R dT
(12.15)
Comparing equations (12.13) and (12.15) we see that the specific heats for an ideal gas are related
in the following manner:
c p cv R
(12.16)
The specific heat ratio, , defined as
cp
cv
appears frequently in gas dynamics. Other helpful relations include:
R
R
cp
cv
1
1
C. Wassgren
466
(12.17)
(12.18)
1st and 2nd Law Considerations for an Ideal Gas (Gibbs Equation)
Now lets consider the 1st and 2nd laws for an ideal gas. Recall that the 1st Law (COE) for a system
is:
(12.19)
de qinto system won system
From the 2nd law we have:

qinto system
ds
(12.20)
T
where the equality holds only for a reversible process. If we consider a simple, compressible
system (so that de = du) where only reversible pdv work is considered, we can substitute equation
(12.20) into equation (12.19) (note that we use the equality in equation (12.20) since we are
considering a reversible process):
du Tds pdv
Tds du pdv
du pdv dh vdp
(Note: dh d u pv du pdv vdp )

T
T
T
T
This equation is also known as the Tds equation or Gibbs equation.
ds
(12.21)
Notes:
1. Even though Eqn. (12.21) was derived for a reversible process, it holds true even for
irreversible processes. This is because entropy is a property and is thus independent of the
process taken between the end states.
For an ideal gas we can substitute in Eqns. (12.10) and (12.13) for du and dh :
dT
dv
dT
d
dT
dp
R
cv
R
cp
R
ds cv
T
v
T
T
p
(12.22)
The Perfect Gas

A perfect gas is an ideal gas with constant specific heats. The previous relations given for an ideal
gas simplify to the following for a perfect gas:
equation (12.11)
u2 u1 cv T2 T1
(12.23)
equation (12.14)
h2 h1 c p T2 T1
(12.24)
s2 s1 cv ln
equation (12.22)
c p ln
T2
v
T
R ln 2 cv ln 2 R ln 2
1
T1
v1
T1
T2
p
R ln 2
T1
p1
(12.25)
isentropic process for a perfect gas

Recall that for an isentropic process (a reversible, adiabatic process), ds = 0, so that Eqn.
(12.25) becomes (note that density will be used instead of specific volume):
T
p
T
c p ln 2 R ln 2
cv ln 2 R ln 2
T1
p1
1
T1
R
T2 2 cv
T1 1
Using the relations in Eqn. (12.18) we have:
C. Wassgren
cp
R
T2 2
T1 1
p2 T2

p1 T1
p2 T2 1

p1 T1
(12.26)
467
(12.27)
Substituting Eqn. (12.26) into Eqn. (12.27):
p2 2
p1 1
(12.28)
Equations (12.26) through (12.28) will be very useful in our analysis of gas dynamics.
C. Wassgren
468
The Imperfect Gas
For most real gases, the compressibility factor is not equal to one, i.e. Z = pv/(RT) 1. However, in most
engineering applications the pressure ratio, p/pcr, is small enough and the temperature ratio, T/Tcr, large
enough so that the ideal gas model can be used with reasonable accuracy. The specific heats, though, may
not be constant so that a perfect gas model cannot be assumed.
The specific heats are a strong function of temperature and only weakly dependent on the pressure (over a
typical range found in most engineering applications) as shown in the plots below for air.
for air
for air
The figure on the left plots cp/R as a function of

temperature for various substances. Note that
for air the specific heat is a weak function of
temperature for temperatures less than ~1000K.
Hence, a perfect gas assumption for air flows
with temperatures less than 1000 K is
reasonable.
The specific heat temperature dependence can
be predicted from statistical mechanics models.
The variations in specific heat are due to the
activation of different energy storage modes
(e.g. vibration and rotation) within the gas
molecules at different temperatures. More
detail on this topic can be found in Callen,
H.B., Thermodynamics and an Introduction to
Thermostatistics, Wiley.
There are several relations used for describing the variation in the specific heats with temperature. It is
often easier, however, to use tabulated values for determining the various thermodynamic properties such
as the specific internal energy, enthalpy, and entropy.
C. Wassgren
469
Notes:
1.
Recall that that for an ideal gas:

T
dh c p dT h T c p dT where h(T = 0) = 0
(12.29)
u h pv h RT
(12.30)
Values for h and u can typically be found in tables given at the back of most thermodynamics texts.
2.
For an ideal gas we also have:

T
dT
dp
dT
p
ds c p
R
s sref c p
R ln
T
p
T
pref
Tref
(12.31)
Choosing absolute zero as the reference temperature and defining:

T
dT
s0 cp
T
0
gives:
T
T
T2 dT Tref dT
p 1 dT ref dT
p
s2 sref s1 sref c p c p R ln 2 c p c p R ln 1
T
T
pref 0
T
T
pref
0
0
0
p
p
dT 1 dT
cp
R ln 2 ln 1
T 0
T
pref
pref
0
p
s2 s1 s20 s10 R ln 2
p1
0
Values for s can be found in thermodynamics tables.
T2
s2 s1 c p
C. Wassgren
(12.32)
470
(12.33)
Relations for an Incompressible Substance

An incompressible system is one in which the density (or specific volume) of the system remains
constant:
d dv 0
(12.34)
Since the state of a simple system can be determined using two properties, we can write: u=u(T,v) so
that:
u
u
(12.35)
du
dT
dv
cv dT
T v
v T 0
cv
From the definition of enthalpy we have:

dh du p dv
vdp du vdp
(12.36)
The enthalpy can also be written as: h=h(T,p) so that:

h
h
dh
dT
dp
T p
p T
(12.37)
Equating Eqn. (12.37) with Eqn. (12.36) we have:

h
h
dh
dT
dp du
vdp
T
p T
cv dT
p
c p
c p cv
for an incompressible substance
Equation (12.21) for an incompressible substance reduces to:

du dh vdp
ds
T
T
T
C. Wassgren
471
(12.38)
(12.39)
3.
One-Dimensional Flow
flow dimensionality
The dimension of a flow is equal to the number of spatial coordinates required to describe the
flow. For example:
0D flow:
u=u(t) or u=constant
u=u(t; x) or u=u(x)
1D flow:
2D flow:
u=u(t; r,) or u=u(r,)
3D flow:
u=u(t; x, y, z) or u=u(x, y, z)
A flow is uniform if it does not vary in a spatial direction. A flow is non-uniform if it does vary in
that spatial direction. For example, the velocity profile in the figure below at left is non-uniform
in the y-direction. The figure at the right is uniform in the y-direction.
y
Notes:
1. The flow of a real fluid through a pipe is not one-dimensional due to the no-slip condition at
the pipe walls. If the Reynolds number of the flow is large enough, the flow may be
approximated to be 1D within reasonable accuracy. As a flows Reynolds number increases,
the velocity profile becomes more blunt-shaped and more closely approaches that of a
uniform profile.
Re
small Reynolds number large Reynolds number

(laminar)
(turbulent)
2.
VD
where Re is the Reynolds

number, V is the average
velocity, D is the pipe
diameter, and is the
kinematic viscosity of the
fluid.
The one-dimensional flow assumption is exact for a stream filament.

a. Recall that there is no flow across a streamline.
b. A stream tube is a tube made by all the streamlines passing through a closed curve.
There is no flow through a stream tube wall.
streamlines
c.
A stream filiment is a stream tube with infinitesimally small cross-sectional area.
C. Wassgren
472
Governing Equations for One-Dimensional, Steady Flow
Lets write our governing equations (COM, LME, COE, 2nd Law) for a one-dimensional, steady flow.
Conservation of Mass (COM)
VA
VA+d(VA)
dx
d
dt
where
d
dt
dV u
CV
rel
dA 0
CS
dV 0
(steady flow)
CV
rel
dA VA VA d VA
CS
d VA 0
d VA
C. Wassgren
m constant
(12.40)
473
Linear Momentum Equation (LME)

(p+1/2dp)dA
V2A+d(V2A)
V A
pA
Let P be the pipe perimeter

pA+d(pA)
Pdx
dx
x
Since the flow is 1D, consider only the x-component of the LME:
d
u x dV u x u rel dA Fx, body Fx , surface
dt
on CV
on CV
CV
where
d
dt
CS
u dV 0
(steady flow)
CV
CS
u rel dA V 2 A V 2 A d V 2 A d V 2 A
but, from COM we have: m VA =constant
mdV
VAdV
u rel dA d V 2 A d mV
CS
Fx, body 0
on CV
Fx, surface pA pA d pA p 1 2 dp dA Pdx

on CV
d ( pA) pdA Pdx

Adp Pdx
Note that higher-order terms have been neglected in the previous expression and that the friction force acts
only in the x-direction since the boundaries vary smoothly (the slope is small, no discontinuities).
Well re-write the friction force term using a hydraulic diameter, DH:
4A
DH
P
and a friction factor, f:
f F 1 2 V 2
fF, Fanning friction factor or
4 fD
2 V
fD, DArcy friction factor
so that the momentum equation becomes, after substitution and re-arranging:

4f
dp VdV 1 2 V 2 F dx 0
DH
C. Wassgren
474
(12.41)
(12.42)
(12.43)
(12.44)
Notes:
1.
Gravitational effects have been neglected in the previous analysis since when dealing with gases,
gravitational effects are typically very small compared to other terms in the momentum equation.
2.
The DArcy friction factor, fD, is the friction factor found in the Moody diagram for pipe flows.
3.
For a frictionless flow, equation (12.44) simplifies to:

dp VdV 0
dp
2V
constant
(12.45)
Bernoullis Equation!
The integral occurs because the pressure can, in general, be a function of the density.
a.
For an incompressible fluid,=constant, so that, after integrating along a streamline:

p 1 2
2 V constant
BE for incompressible fluid
b.
(12.46)
For an ideal gas: p=RT, so that, for an isothermal process:

RT ln
1 2
2 V constant
0
BE for ideal gas (isothermal)
(12.47)
where 0 is a reference density.

c.
For a perfect gas undergoing an isentropic process, p-=constant:

p c
dp
c
1
1
1
1

dp c
p
p
but the constant, c, is given by:

1
c p p0 0
c p 1 p0 01
Substituting and simplifying we find that the momentum equation can be simplified to
p 1 2
2 V constant
1
Note that from the ideal gas law, p/ = RT and from Eqn. (12.18):
c pT 1 2 V 2 constant
BE for perfect gas (isentropic)
C. Wassgren
475
(12.48)
Conservation of Energy (COE)
Won CV
Q into CV
dx
d
dt
where
d
dt
e dV h
CV
2V
gz u rel dA Q into CV Wshaft, Wother,

on CV
CS
e dV 0
CV
2V
gz
on CV
(steady flow)
rel
dA m h
CS
h
md
2V
2V
2V
d h
2V
Note that the gravitational potential term has been neglected since this term is typically much smaller
than the other terms in the equation for gases.
Qinto CV Qinto CV
W
W
on CV
on CV
Substituting and simplifying we have:
d h 1 2 V 2 qinto won
CV
(12.49)
CV
Notes:
1.
For an adiabatic flow ( qinto CV 0 ) with no work ( won CV 0 ):

h 1 2 V 2 constant COE for adiabatic flow with no work (i.e. isentropic flow)
This is the same as Eqn. (12.48) since for a perfect gas, h = cpT.
C. Wassgren
476
(12.50)
2nd Law of Thermodynamics (aka Entropy Equation)
Q into CV
d ms

ms
ms
dx
d
dt
where
d
dt
s dV s u
CV
rel
CS
dA
qinto CV
T
CV
s dV 0
CV
s u
rel
dA m s s ds mds
CS
CV
qinto CV
T
qinto CV
T

q
ds into CV (where dqinto CV is the heat added to the CV per unit mass)
T
(12.51)
Notes:
1.
For an adiabatic flow ( qinto CV 0 ): ds 0 where the equality holds only for a reversible flow
(adiabatic and reversible isentropic).
C. Wassgren
477
4.
Speed of Sound
The speed of sound, c, in a substance is the speed at which infinitesimal pressure disturbances propagate
through the surrounding substance. To understand how the speed of sound depends on the substance
properties, lets examine the following model.
Consider a wave moving at velocity, c, through a stagnant fluid. Across the wave, the fluid properties such
as pressure, p, density, , temperature, T, and the velocity, V, can all change as shown in the figure below.
downstream
upstream
V=0
p
p+p
+
T+T
c
Now lets change our frame of reference such that it moves with the wave:
downstream
upstream
V=c
c-V
p+p
+
T+T
stationary wave
Now apply COM and the LME on a very thin CV surrounding the wave:
downstream
upstream
V=c
p
C. Wassgren
c-V
p+p
+
T+T
478
COM:
d
dt
where
d
dt
dV u
CV
rel
dA 0
CS
dV 0
CV
rel
dA cA c V A
CS

cA c V A
c c V c V
c
V

LME:
d
dt
where
d
dt
u dV u
x
CV
(12.52)
u rel dA Fx, body Fx, surface
CS
on CV
on CV
u dV 0
x
CV
m c V m V
u rel dA mc
CS
cAV
Fx, body 0
on CV
Fx , surface pA p p A pA
on CV

cAV pA
p
c
V
Substituting the result from equation (12.52):
p
c
c
c2
p
1
C. Wassgren
(12.53)
479
For a sound wave, the changes across the wave are infinitesimal so that:
p p
c 2 lim
1
p dp

(12.54)
We also need to specify the process by which these changes occur. Since the changes across the wave are
infinitesimal, we can regard the wave as a reversible process. Additionally, the temperature gradient on
either side of the wave is small so there is negligible heat transfer into the control volume so the process is
adiabatic. Thus, the changes across a sound wave occur isentropically (an adiabatic, reversible process is
also isentropic):
p
c2
speed of sound in a continuous substance
(12.55)
s
Notes:
1.
Note that equation (12.55) is the speed of sound in any continuous substance. Its not limited to fluids.
2.
If the wave is not weak (i.e. the changes in the flow properties across the wave are not infinitesimal),
then viscous effects and temperature gradients within the wave will be significant and the process can
no longer be considered isentropic. We will discuss this situation later when examining shock waves.
3.
Note that according to equation (12.54) the stronger the wave, i.e. the greater , the faster the wave
will propagate. This will also be examined when discussing shock waves.
4.
For an ideal gas undergoing an isentropic process (ds = 0):

R d R dp
p
p
=
RT
cv c p p
s
Substituting into equation (12.55), the sound speed for an ideal gas is:
c RT
speed of sound in an ideal gas
(12.56)
(12.57)
Notes:
a. The absolute temperature must be used when calculating the speed of sound.
b. The speed of sound in air (=1.4, R=287 J/(kgK)) at standard conditions (T=288 K) is 340 m/s
(=1115 ft/s 1/5 mile/s).
c. It is not unexpected that the speed of sound is proportional to the square root of the temperature.
Since disturbances travel through the gas as a result of molecular impacts, we should expect the
speed of the disturbance to be proportional to the speed of the molecules. The temperature is
equal to the random kinetic energy of the molecules and so the molecular speed is proportional to
the square root of the temperature. Thus, the speed of sound is proportional to the square root of
the temperature.
C. Wassgren
480
5.
Equation (12.55) can also be written in terms of the bulk modulus. The bulk modulus, E of a
substance is a measure of the compressibility of the substance. It is defined as the ratio of a differential
applied pressure to the resulting differential change in volume of a substance at a given volume:
p
p
E
(12.58)
V
V
Notes:
1. dp>0 dV<0 E>0
2. From COM: dV/V = -d
3. E compressibility
dp
volume before:
volume after:
V
V-dV
The isentropic bulk modulus, E|s is defined as:

p
p
E s
V
V s
(12.59)
Thus, the speed of sound can also be written as:

E s
c2
alternate sound speed expression
(12.60)
Notes:
a. The isentropic bulk modulus for air is E|s = RT.
b. The isentropic bulk modulus for water is 2.19 GPa so that the speed of sound in water (=1000
kg/m3) is 1480 m/s (=4900 ft/s 1 mile/s 5X faster than the speed of sound in air at sea level).
c. For solids, the bulk modulus, E, is related to the modulus of elasticity, E, and Poissons ratio, ,
by:
E
3 1 2
E
For many metals (e.g. steel and aluminum), Poissons ratio is approximately 1/3 so that EE
1. The speed of sound in stainless steel (E = 163 GPa; = 7800 kg/m3) is 4570 m/s (= 15000 ft/s
3 mile/s 3X faster than the speed of sound in water).
C. Wassgren
481
6.
The Mach number, Ma, is a dimensionless parameter that is commonly used when discussing
compressible flows. The Mach number is defined as:
V
Ma
(12.61)
c
where V is the flow velocity and c is the speed of sound in the flow.
Notes:
a. Compressible flows are often classified by their Mach number:
Ma < 1
subsonic
Ma = 1
sonic
Ma > 1
supersonic
Additional sub-classifications include:
Ma < 0.3
incompressible
transonic
Ma 1
Ma > 5
hypersonic
b. The square of the Mach number, Ma2, is a measure of a flows macroscopic kinetic energy to its
microscopic kinetic energy.
7.
The change in the properties across the sound wave can be found from the following:
COM (note that A = constant)
cA d c dV A 0
dV cd 0
dV d
Speed of Sound (isentropic process):

dp
c2
d
(12.62)
(12.63)
Ideal Gas:
p RT
dp d dT
p
T
(12.64)
Combine Eqns. (12.62) and (12.63) and simplify.

2 dp
RT dp c
c
p
p d
RT
p dV
dV d 1 dp
c
p
(12.65)
Now combine Eqns. (12.65) and (12.64).

dp 1 dp dT
p p T
dT 1 dp
T
p
(12.66)
Thus, across a compression wave (dp > 0): dV > 0, d > 0, and dT > 0.
Across a rarefaction, or expansion wave (dp < 0): dV < 0, d < 0, and dT < 0.
C. Wassgren
482
The Mach Cone
Consider the propagation of infinitesimal pressure waves, i.e., sound waves, emanating from an object at
rest. The waves will travel at the speed of sound, c.
object
(V=0)
location of sound pulse
after time t
c(t)
c(3t) c(2t)
Now consider an object moving at subsonic velocity, V<c:
V(2t)
The pressure pulses are more closely

spaced in the direction of the objects
motion and more widely spaced behind
the object. Thus, the frequency of the
sound in front of the object increases,
while the frequency behind the object
decreases. This is known as the
Doppler Shift.
V(3t)
V(t)
c(3t)
c(2t)
c(t)
direction of object motion

Now consider an object traveling at the sonic speed, V=c:
Since no wave fronts propagate ahead
of the object, an observer standing in
front of the object will not hear it
approaching until the object reaches the
observer.
V(3t)
V(2t)
V(t)
c(2t)
c(3t)
Note that the infinitesimal pressure

changes in front of the object pile up
on one another producing a sudden,
finite pressure change (a shock wave!)
c(t)
locus of all
wave fronts
C. Wassgren
483
Lastly, consider an object travelling at supersonic speeds, V>c:

The object out runs the pressure pulses
it generates.
V(3t)
V(2t)
V(t)
c(3t)
c(2t)
The locus of wave fronts forms a cone

which is known as the Mach Cone.
The object cannot be heard outside the
Mach Cone. This region is termed the
zone of silence. Inside the cone, the
region known as the zone of action, the
object can be heard.
c(t)
Mach Cone
(locus of all wave fronts)
zone of action
zone of silence
sin 1 1 Ma
C. Wassgren
The angle of the cone, known as the

Mach angle, , is given by:
ct
c
1

sin
V t V Ma
484
(12.67)
5.
Adiabatic, 1D Compressible Flow of a Perfect Gas
Now lets consider the 1D, adiabatic flow of a compressible fluid. Recall that from the energy equation we
have:
h 1 2 V 2 constant
(12.68)
For a perfect gas we can re-write the specific enthalpy in terms of the specific heat at constant pressure and
the absolute temperature: h=cpT:
c pT 1 2 V 2 constant
T
V2
constant
2c p
(12.69)
We can re-write this equation in terms of the Mach number:

V
V
Ma
c
RT
V 2 RT Ma 2
T
RTMa 2
2c p
constant
Substituting the following relation:

R 1
cp
results in:
1
T 1
Ma 2 constant
2
Adiabatic, 1D Flow of a Perfect Gas
(12.70)
If the flow can also be considered reversible, thus making the flow isentropic (adiabatic + reversible =
isentropic), we can use the isentropic relations for a perfect gas:
p constant T 1
and
constant T 1
to give:
1
1
p 1
Ma 2
constant
2
(12.71)
1
1
1
Ma 2
constant
2
Isentropic, 1D Flow of a Perfect Gas
C. Wassgren
(12.72)
485
Stagnation and Sonic Conditions
It is convenient to choose some significant reference point in the flow where we can evaluate the constants
in equations (12.70)-(12.72). Two such reference points are commonly used in compressible fluid
dynamics. These are the stagnation conditions and sonic conditions.
Stagnation Conditions
Stagnation conditions are those conditions that would occur if the fluid is brought to rest (zero velocity
Ma=0). We use the isentropic stagnation conditions where the flow is brought to rest isentropically. These
conditions are typically indicated by the subscript 0. Thus, equations (12.70)-(12.72) can be written in
terms of stagnation conditions:
T 1
1
Ma 2
T0
2
adiabatic, 1D flow of a perfect gas
(12.73)
isentropic, 1D flow of a perfect gas
(12.74)
(12.75)
p 1
1
Ma 2
1
p0
2
1 2 1
1
Ma
0
2
We can also determine the speed of sound at the stagnation conditions using the fact that c=(RT)1/2:
c 1
1
Ma 2
c0
2
12
(12.76)
Notes:
1. Stagnation conditions are also commonly referred to as total conditions (given by the subscript T).
2. Stagnation conditions can be determined even for a moving fluid. The fluid doesnt necessarily have
to be at rest to state its stagnation conditions. To determine stagnation conditions we only need to
consider the conditions if the flow were brought to rest.
3. Equations (12.74) and (12.75) are for a flow brought to rest isentropically.
4. Tables listing the values of equations (12.73)-(12.76) for various Mach numbers are typically given in
the back of most textbooks concerning compressible fluid flows.
5. Note that the stagnation temperature is greater than the flow temperature. This is because when the
flow is decelerated to zero velocity, the macroscopic kinetic energy is converted to internal energy
(microscopic kinetic energy) and thus the temperature increases.
1.2
1
ratio
0.8
T/T0
p/p0
0.6
r/r0
0.4
c/c0
0.2
0
0
3
Ma
C. Wassgren
486
=1.4
6.
The stagnation pressure is a significant property for a flow because it is directly related to the amount
of work we can extract from the flow. For example, imagine we bring a flow to rest so that we have
stagnation conditions within the tank shown below.
tank with stagnation
pressure, p0
piston used for doing work
flow
The larger the stagnation pressure in the tank, the greater the force we can exert on a piston that can be
used to perform useful work.
C. Wassgren
487
Sonic Conditions
Another convenient reference point is where the flow has a Mach number of one (Ma=1). Conditions
where the Mach number is one are known as sonic conditions and are typically specified using the
superscript *.
Equations (12.73)-(12.76) evaluated at sonic conditions are:
T * 1
1
T0
2
(12.77)
(12.78)
(12.79)
(12.80)
T*
0.8333
T0
adiabatic, 1D flow of air
(12.81)
p*
0.5283
p0
isentropic, 1D flow of air
(12.82)
*
0.6339
0
isentropic, 1D flow of air
(12.83)
c*
0.9129
c0
adiabatic, 1D flow of air
(12.84)
p* 1 1
1
p0
2
1
* 1 1
1
0
2
c* 1
1
c0
2
12
Notes:
1. For air, =1.4 so that:
On the Isentropic Flow Assumption
In many engineering gas dynamics flows, the assumption that the entropy of the fluid remains constant (an
isentropic process) is a good one. If viscous and heat transfer effects can be neglected, then we can
reasonably assume that the flow is isentropic (isentropic = reversible + adiabatic). This is often the case for
flows through short, insulated ducts or through stream tubes not passing through a boundary layer or a
shock wave (strong viscous effects occur in both cases). Experiments have verified that the isentropic
assumption under these conditions is reasonable.
C. Wassgren
488
Mollier (aka h-s) Diagrams
Mollier diagrams are diagrams that plot the enthalpy (h) as a function of entropy (s) for a process. They are
often useful in visualizing trends.
Notes:
1. Sketches of constant pressure and constant volume (or density) curves are shown in the figure below.
h
p
curves of constant p
curves of constant v
s
2.
For a perfect gas, curves of constant volume (or density) and constant pressure have slopes given by:
Tds du pdv
Tds dh vdp
cv
dh
Tds
c p dT pdv
T
cp
ds p
dh
3.
dh
T
ds v
The difference between the flow enthalpy and the stagnation enthalpy for an isentropic process is equal
to the specific kinetic energy:
h
p0
stagnation state
h0
1
/2V2
h0 h 1 2 V 2
s0
isentropic process
C. Wassgren
489
6.
Effects of Area Change on Steady, 1D, Isentropic Flow
Mass conservation states that for a steady, 1D, incompressible flow, a decrease in the area will result in an
increase in velocity (and visa-versa). This is not necessarily true, however, for a compressible flow as will
be shown below.
Consider mass conservation for a steady, 1D flow:
m AV constant
d AV 0
AVd VdA AdV 0
d
dA dV
0
A V
(12.85)
Notes:
1. If the flow is incompressible, then d=0 and we see that: dV/V = -dA/A. Thus, if the area decreases
(dA<0), then the velocity must increase (dV>0).
2. For a compressible fluid, the density may change so we need an additional relationship between
density and either area or velocity to draw any conclusions about how changes in area affect changes
in velocity.
Recall that the speed of sound is given by:
p
c2
s
Lets concern ourselves with an isentropic flow, s=constant, so we can re-write this expression as:
dp
d 2
c
Well also make use of Bernoullis equation (COLM):
dp
VdV 0
(12.86)
(12.87)
Substituting equations (12.86) and (12.87) into equation (12.85) gives:

1 dp dA dV
0
A V
c2
V 2 dV dA dV
0
A V
c2 V
V2
dV dA
2 1
A
c
V
Ma 1 dVV dAA
2
(12.88)
Note that the trends for pressure and density are opposite the trends for velocity:
dp
dV
d
dV
From equations (12.88) and (12.85):

From equation (12.87):
Ma 2
2
V
V
V
For an ideal gas: ds 0 cv
dT
d
dT R d
dT
dV
R
1 Ma 2
T
T
cv
T
V
dV 0 dT 0 d Ma 0
dV 0 dT 0 d Ma 0
C. Wassgren
490
Lets interpret equation (12.88). Consider the following the cases:

Ma < 1 (subsonic flow)
dA < 0 dV > 0
dA > 0 dV < 0
(decreases in area result in increases in velocity)

(increases in area result in decreases in velocity)
A subsonic nozzle should have a decreasing area.

A subsonic diffuser should have an increasing area.
The area-velocity relationships for subsonic flow are similar to that for incompressible flow.
Ma > 1 (supersonic flow)
dA < 0 dV < 0
(decreases in area result in decreases in velocity)
dA > 0 dV > 0
(increases in area result in increases in velocity)
A supersonic nozzle should have an increasing area.
A supersonic diffuser should have a decreasing area.
The area-velocity relationships for supersonic flow are opposite to that for subsonic flow.
Ma = 1 (sonic flow)
dA = 0
(sonic conditions must occur at an inflection point in the area)
The area at which Ma=1 must be a minimum based on the previous relationships for subsonic and
supersonic flow.
For example, if we assume that Ma=1 occurs at a maximum, then if the flow starts off subsonic
and we increase the area, the Mach number will decrease and diverge from Ma=1. If the flow
starts off supersonic and the area increases, then the Mach number will increase and diverge from
Ma=1. Thus, we see that the Mach number will not occur at the maximum area.
Notes:
1. Nothing can be said about how the velocity is changing when the Ma=1. The velocity can
either be decreasing, remaining constant, or increasing.
2. A minimum area does not necessarily imply that the Mach number will be one, i.e.:
Ma 1 minimum area
(12.89)
minimum area Ma 1
This is because if dA=0 in equation (12.88), then either Ma=1 or dV=0.
Ma=1
Ma=1
Ma<1
Ma>1
Ma<1 or Ma>1
Ma<1 or Ma>1
area must be a minimum

Now lets examine some other consequences resulting from mass conservation.
C. Wassgren
491
Since the mass flow rate must be a constant in 1D, steady flow, we can write:
m AV * A*V *
where the * quantities are the conditions where Ma=1. Lets re-arrange this equation and substitute the
isentropic flow relations we derived previously.
A * V *
A* V
* 0 c*
cMa
0
* 0 c* c0 1
c c Ma
0
0
where
* 1
1
0
2
1
1
1 2
1
Ma
0
2
c* 1
1
2
c0
1
1
12
c 1
Ma 2
1
2
c0
12
Substituting and simplifying we have:

1
2 1
6
5
A/A*
1 2
1
Ma
A
1
2
A* Ma 1 1
(12.90)
4
3
2
1
0
0
Ma
C. Wassgren
492
1.4 4
Notes:
1.
2.
3.
4.
Equation (12.90) tells us what area we would need to contract to to get sonic conditions (Ma=1, A=A*)
given the current Mach #, Ma, and area, A.
We could also interpret equation (12.90) as saying, given the area for sonic conditions, A*, the Mach
number, Ma, and area, A, are directly related for an isentropic flow. Recall that this relationship results
from conservation of mass and the assumption of an isentropic flow.
Values for A/A* as a function of Mach # are typically included in compressible flow tables found in the
appendices of most fluid mechanics textbooks.
What happens if we constrict the area to a value less than A*? For a subsonic flow, the new area
information can propagate upstream and downstream and, as a result, the conditions everywhere
change (i.e., the Mach numbers change according to equation (12.90) where the new area would be
A*.) If the upstream flow is supersonic, then some non-isentropic process must occur upstream (a
shock wave) so that the constricted area is no longer less than A*.
C. Wassgren
493
Choked Flow
Consider the flow of a compressible fluid from a large reservoir into the surroundings. Let the pressure of
the surroundings, called the back pressure, pB, be controllable:
stagnation
conditions
p0, T0, 0
throat conditions
Ath, pth, Vth
back pressure, pB
When pB=p0 there will be no flow from the reservoir since there is no driving pressure gradient. When the
back pressure, pB, is decreased, a pressure wave propagates through the fluid in the nozzle and into the
tank. Thus, the fluid in the tank is informed that the pressure outside has been lowered and a pressure
gradient is established resulting in fluid being pushed out of the tank.
pressure wave
propagating upstream
Thus, when pB < p0, the fluid will begin to flow out of the reservoir. Furthermore, as pB/p0 , Vth , and the
mass flow rate . Note that the flow through the nozzle will be subsonic (Ma<1) since the fluid starts off
from stagnation conditions and does not pass through a minimum until reaching the throat. Additionally,
since the flow is subsonic, the pressure at the throat will be the same as the back pressure, i.e., pth=pB. That
this is so can be seen by noting that if pth>pB, then the flow would expand upon leaving the nozzle and as a
result, the jet velocity would decrease, and the pressure would increase. Thus, the jet pressure would
diverge from the surrounding pressure. But the jet must eventually reach the surrounding pressure so the
assumption that pth>pB must be incorrect. A similar argument can be made for pth<pB.
As we continue to decrease pB/p0, well eventually reach a state where the velocity at the throat will reach
Ma=1 (Vth=V*=c*). The pressure ratio at the throat will then be:
pth
p
p* 1
B
1
2
p0
p0
p0
and the fluid velocity will be:
(12.91)
Vth V * c* RT *
C. Wassgren
(12.92)
494
Any further decrease in pB has no effect on the velocity at the throat since the pressure information can no
longer propagate upstream into the reservoir. The fluid velocity out of the tank is the same as the speed of
the pressure wave into the tank so that the pressure information cant propagate upstream of the throat.
Thus, all flow conditions upstream of the throat will remain unchanged. As a result, we can no longer
increase the mass flow rate from the tank by changing the back pressure. This condition is referred to as
choked flow conditions. The maximum, or choked, mass flow rate will be the same as the mass flow rate
at the throat where sonic conditions occur:
p* * *
m choked *V * A*
V A
RT *
p*
A*
RT *
where equation (12.92) has been utilized. Substituting the following isentropic relations:
p*
1
p
p0 1
2
p0
p0
T*
1
T0 1
T0
2
T0
and simplifying results in:

T*
m choked
*
1 21
1
p0
A
2
RT
(12.93)
Notes:
1. The choked mass flow rate (equation (12.93)) is the maximum mass flow rate that can be achieved
from the reservoir.
2. A quick check to see if the flow will be choked or not is to check if the back pressure is less than or
equal to the sonic pressure, i.e.:
p
pB p* 1 1
p*
1
, then the flow will be choked and th
.
(12.94)
p0
p0
p0
p0
2
What happens outside of the nozzle if the back pressure is less than the sonic pressure? The flow must
eventually adjust to the surrounding pressure. It does so by expanding in a two-dimensional process
known as an expansion fan; a topic well address later in the course.
If
3.
Ma=1
pth/p0=p*/p0
expansion fans
pB/p0<p*/p0
C. Wassgren
495
4.
The previously described processes are sketched in the following plots:
pth/p0
m choked
p*/p0
p*/p0
pB/p0
p*/p0
pB/p0
p/p0
1
pB/p0
p*/p0
pressure variations in
expansion fan
throat
C. Wassgren
496
Steady, 1D, Isentropic Flow of an Imperfect Gas
Imperfect gas effects become significant when dealing with high temperature and high Mach number flows.
Examples of imperfect gas effects include:
1. variations in specific heats due to the activation of intra-molecular energy modes (i.e. rotational,
vibration, electronic modes),
2. dissociation of molecules, e.g.
for T > 2000 K, O2 2O and
for T > 4000 K, N 2 2 N
3. ionization of atoms, e.g.
for T > 9000 K, O O e and N N e
4. chemical reactions.
Non-equilibrium effects can also be significant at high Mach numbers.
In these notes, well focus only on the variations in specific heats due to temperature.
Recall that the governing equations are:
COM:
d VA 0
LME:
COE:
2nd Law:
(12.95)
dp VdV 0
(12.96)
(12.97)
(12.98)
dh VdV 0
ds 0
Notes:
1. These equations are independent of the type of fluid being considered.
2.
The COE and LME statements are equivalent for an isentropic flow. Recall that combining the first
and second laws for a simple, compressible system with reversible pdv work gives:
du Tds pdv
du d pv Tds pdv d pv
(12.99)
dh Tds vdp
but since ds = 0 and v = 1/:
dp
dh
(12.100)
Substituting into COE gives:

dp VdV 0
(12.101)
This is the same relation that we had from the LME! Hence, the COE and LME expressions are
equivalent.
3.
Equations (12.95)-(12.98) (3 equations, recall from note 2 that equations (12.96) and (12.97) are
equivalent) have a total of 5 unknowns (, p, V, h, s). Note that area, A, is typically a known quantity.
We need two additional extra relations to close the system of equations. These relations are the
equations of state for the substance of interest.
p, s
h h p, s
The equations of state may be given in equation, tabular, or graphical form. To solve the system of
equations, one must generally employ an appropriate numerical scheme.
a. Note that only two independent properties are required to fix the state of a simple compressible
system at equilibrium.
C. Wassgren
497
4.

Empirical specific heat data is often expressed in terms of curve fits. Exponential and polynomial
curve fits are common. For example, Zucrow and Hoffman use a 4th-order polynomial curve fit:
cp
(12.102)
a bT cT 2 dT 3 eT 4
R
where [T] = K and the constants a e are given in the table below for various substances and for
various temperature ranges.
gas
O2
N2
CO
CO2
A
H2
H2O
CH4
C2H4
a.
b*103
-1.87822
0.736183
-1.20815
1.51549
-1.61910
1.48914
8.73510
3.09817
0
0
2.67652
0.511195
-1.10845
2.94514
-3.97946
10.4168
11.3831
11.4918
a
3.62560
3.62195
3.67483
2.89632
3.71009
2.98407
2.40078
4.46080
2.50000
2.50000
3.05745
3.10019
4.07013
2.71676
3.82619
1.50271
1.42568
3.45522
c*106
7.05545
-0.196522
2.32401
-0.572353
3.69236
-0.578997
-6.60709
-1.23926
0
0
-5.80992
0.0526442
4.15212
-0.802243
24.5583
-3.91815
7.98900
-4.36518
d*109
-6.76351
0.0362016
-0.632176
0.0998074
-2.03197
0.103646
2.00219
0.227413
0
0
5.52104
-0.0349100
-2.96374
0.102267
-22.7329
0.677779
-16.2537
0.761551
e*1012
2.15560
-0.00289456
-0.225773
-0.00652236
0.239533
-0.00693536
0.000632740
-0.0155260
0
0
-1.81227
0.00369453
0.807021
-0.00484721
6.96270
-0.0442837
6.74913
-0.0501232
h0*10-3
-1.04752
-1.120198
-1.06116
-0.905862
-14.3563
-14.2452
-48.3775
-48.9614
-0.745375
-0.745375
-0.988905
-0.877380
-30.2797
-29.9058
-10.1450
-9.97871
5.33708
4.47731
s00
4.30528
3.61510
2.35804
6.16151
2.95554
6.34792
9.69515
-0.986360
4.36600
4.36600
-2.29971
-1.96294
-0.322700
6.63057
0.866901
10.7071
14.6218
2.69879
T range [K]
300-1000
1000-5000
300-1000
1000-5000
300-1000
1000-5000
300-1000
1000-5000
300-1000
1000-5000
300-1000
1000-5000
300-1000
1000-5000
300-1000
1000-5000
300-1000
1000-5000
Note that table also includes data for calculating the specific enthalpy, h, as a function of
temperature:
T
c dT
p
T0
h
h0 aT 12 bT 2 13 cT 3 14 dT 4 15 eT 5
R
where h = h0R when T = T0.
b.
(12.103)
The table also includes data for calculating s0, a quantity that will be discussed in the following
note (note #5).
T
0
T0
dT
T
s
s00 a ln T bT 12 cT 2 13 dT 3 14 eT 4
R
where s0 = s00 R when T = T0.
c.
(12.104)
The specific internal energy can be determined using the definition for enthalpy and the ideal gas
law:
u h pv h RT
(12.105)
C. Wassgren
498
d.
The specific heat at constant volume, cv, and the specific heat ratio, , are determined from:
cv c p R
(12.106)
e.
cp
(12.107)
cv
Include calculations for dry air incomplete.

The composition of clean dry atmospheric air near sea level.
Component
% by Volume
N2
78.084
O2
20.9476
A
0.934
CO2
0.0314
H2
0.00005
Ne
0.001818
Kr
0.000114
Xe
0.0000087
He
0.000524
CH4
0.0002
NO
0.00005
f.
5.
Alternate cp expressions for monatomic and diatomic molecules incomplete.
In order to simplify matters when dealing with isentropic flow of an ideal, but imperfect gas (one
where the specific heats vary with temperature), well define some additional useful properties.
Note that from equation (12.99):
dh dp
ds
T T
For an ideal gas (dh = cpdT and p = RT):
dT
dp
ds c p
R
T
p
Integrating and using some reference state as the lower limit of integration gives:
T2
s2 s1
T1
p
dT
R ln 2
T
p1
(12.108)
(12.109)
(12.110)
Define s0:
T
s0
Tref
dT
T
(12.111)
where Tref is some reference state so that equation can be written as:
p
s2 s1 s20 s10 R ln 2
p1
(12.112)
Note that equation (12.111) is a function only of temperature. Tabulated values of s0 are commonly
given in the tables of most thermodynamics or gas dynamics texts.
For an isentropic process, equation (12.112) can be written as:
s0 s0
p2
exp 2 1
R
p1
C. Wassgren
499
(12.113)
If we define state 1 as the reference state, then equation (12.113) simplifies to:
s0
p
pr
exp
(12.114)
pref
R
where pr is known as the relative pressure. Tabulated values of pr are commonly given in the tables
of most thermodynamics or gas dynamics texts. Note that for an isentropic process:
p2 pr 2
(12.115)
p1
pr1
C. Wassgren
500
Example:
Consider the steady, 1D, isentropic flow of air from a tank. Within the tank the pressure and temperature
are 10 MPa and 3000 K, respectively. Determine the temperature and velocity at a location downstream of
the tank where the pressure is 100 kPa if:
a. the air is assumed to behave as a perfect gas, and
b. the air is ideal but has variable specific heats.
C. Wassgren
501
Example:
A wind tunnel using air operates with a stagnation pressure and temperature of 10 atm and 1000 K,
respectively. The test section of the tunnel is designed to operate at a Mach number of 5. Determine the
pressure and temperature in the wind tunnel at this design Mach number:
a. assuming the perfect gas behavior, and
b. assuming imperfect gas behavior.
C. Wassgren
502
7.
Adiabatic Flow with Friction (Fanno Flow)
So far weve examined adiabatic, reversible flows (isentropic flows). For flows in long ducts, the
frictional, or viscous, effects can be significant and the flow can no longer be modeled as isentropic.
long duct:
need to include friction
non-isentropic (irreversible)
short duct:
friction negligible
isentropic (reversible)
To examine the effects of friction, lets consider the case of a steady, 1D, adiabatic flow in a constant area
duct that has irreversible effects due to friction (viscous) effects.
First lets derive the governing equations for the following differential control volume:
V+dV
V
pA
P is the perimeter of the duct
(p+dp)A
wPdx
dx
x
From COM we have:

d
dV u rel dA 0
dt
CV
where
d
dt
CS
dV 0
(steady flow)
CV
rel
dA VA VA d VA
CS
d VA
Thus,
d VA AdV Vd A 0 (recall that A=constant)
d
dV
0
V
C. Wassgren
(12.116)
503
From the LME in the x-direction we have:

d
u x dV u x u rel dA Fx,body Fx,surface
dt
on CV
on CV
CV
where
d
dt
CS
u dV 0
(steady flow)
CV
m V dV
u rel dA mV
CS
mdV
AVdV
Fx ,body 0
on CV
Fx ,surface pA p dp A w Pdx
on CV
dpA
4 w A
dx
DH
VAdV dpA
4 w A
dx
DH
Thus,
dp VdV
4 w
dx 0
DH
(12.117)
From COE we have:

d
e dV
h 12V 2
dt
CV
where
d
dt
rel
dA Q into CV Won CV
CS
e dV 0
CV
where DH4A/P is a hydraulic diameter
2V
CS
(steady flow)
rel
dA m h 1 2 V 2 m h 1 2 V 2 d h 1 2 V 2
h 1 2 V 2 m dh VdV
md
Qinto CV 0
W
0
on CV
(adiabatic conditions)
(no work is done on the control volume, recall that although the walls
exert a force on the control volume, they do no work since there is no
displacement at the walls due to the no-slip condition)
Thus,
dh VdV 0
C. Wassgren
(12.118)
504
The Second Law gives:

d
s dV s u rel dA
dt
CV
CS
qinto CV
CV
where
d
dt
s dV 0
(steady flow)
CV
s u
rel
m s ds mds
dA ms
CS
qinto CV
CV
(adiabatic conditions)
Thus,
ds 0
Note that only the > has been retained since the friction results in irreversible conditions.
(12.119)
Now lets also include the equations of state. Well concern ourselves only with an ideal gas.
The Thermal Equation of State (the ideal gas law) gives:
p RT
dp RTd RdT
so that
dp d dT
p
T
The Caloric Equation of State for an ideal gas can be written as:
dh c p dT
We can also combine the 1st and 2nd Laws with the ideal gas law to write:
dT
dp
ds c p
R
T
p
(12.120)
(12.121)
(12.122)
Finally, from the Mach number relation for an ideal gas we have:
V2
V2
Ma 2 2
RT
c
so that
2VdV V 2 dT
2Mad Ma
RT RT 2
d Ma
Ma
d Ma
Ma
VdV
RTMa 2
V 2 dT
2 RT 2 Ma 2
dV dT
2T
V
C. Wassgren
(12.123)
505
Now lets combine equations (12.116) through (12.118) and (12.120) through (12.123) (we wont go
through all of the algebra here) to get the following relations:
1
Ma 2 1
Ma 2
d Ma
2
4 w dx
(12.124)
2
2
Ma
1 Ma
V DH
Ma 2
dV
V
2 1 Ma 2
4 w
dx
V DH
4
dT 1 Ma
T
1 Ma 2
(12.125)
4 w
dx
V DH
(12.126)
2
2
dp Ma 1 1 Ma 4 w
dx
2
2
p
1 Ma
V DH
(12.127)
Ma 2 4 w
dx
2
2
1 Ma V DH
(12.128)
4 w
ds
1 Ma 2
dx
2
cp
V DH
(12.129)
It is useful to consider how the isentropic stagnation pressure, i.e. the pressure we would have if we brought
the flow to rest isentropically, varies in a Fanno flow. Recall that the isentropic stagnation pressure is
given by:
p 1
1
1
Ma 2
p0
2
so that, after differentiating and doing some algebra,
dp0
Ma 2 4 w
dx
2
p0
2 V DH
(12.130)
Note that we could also derive equation (12.130) by noting that equation (12.122) may be written in terms
of stagnation quantities:
dT
dp
ds c p 0 R 0
T0
p0
and incorporating COE (equation (12.118)) written for an ideal gas:
dh VdV dh0 dT0 0
dT0 0
gives:
dp0
ds
p0
R
C. Wassgren
(12.131)
(12.132)
506
Now lets examine the trends indicated by these equations. Note that:
4 w
dx 0
V 2 DH
d(Ma) > 0
dV > 0
dT < 0
dp < 0
d< 0
ds > 0
dT0 = 0
dp0 < 0

d(Ma) < 0
dV < 0
dT > 0
dp > 0
d > 0
ds > 0
dT0 = 0
dp0 < 0
eqn (12.124)
eqn (12.125)
eqn (12.126)
eqn (12.127)
eqn (12.128)
eqn (12.129)
eqn (12.131)
eqn (12.130)
Notes:
1.
Usually we write the shear stress in terms of a friction factor, f:
w fF
w
4 fD
fF
1
4
2 V
2 V
Fanning friction factor

2
DArcy friction factor
fD
2.
The DArcy friction factor is used most often in the analysis of incompressible pipe flows while the
Fanning friction factor is often used for compressible flows.
3.
In general, the friction factor, f, is a function of the flow Reynolds number, Re, the relative roughness
of the pipe walls, /DH, and the Mach number, Ma (recall the Moody chart from undergraduate
incompressible pipe flow problems). Typically, Mach number effects are small in comparison to the
effects of Re and /DH so they are often neglected. Furthermore, the Reynolds number for
compressible flows is often quite large. As a result, the friction factor remains essentially constant for
a compressible flow.
4.
The stagnation temperature remains constant and the stagnation pressure always decreases with
friction.
C. Wassgren
507
Substituting the (Fanning) friction factor into equation (12.124) and integrating over a particular length of
pipe:
Ma 2
Ma1
2 1 Ma 2 d Ma
Ma 3 1
Ma 2
2
x2
4 fF
x1
dx
1
2
1 2 Ma 2
4 fF
L12
(12.133)
DH
1
2
1 2 Ma1
where L12 = x2-x1. Note that the overbar (implying a mean value) on the friction factor will be dropped in
subsequent equations for convenience.
1 1
1

Ma12 Ma 22
1 Ma12
ln

Ma 22
2
Equation (12.133) relates the conditions between two arbitrary points in a frictional duct flow. Its often
more convenient to make one of the points a well-defined reference point. The most logical reference
would be where sonic conditions occur, i.e. Ma2=1, since in Fanno flow the Mach number approaches
unity. Using this reference point equation (12.133) becomes:
1 Ma 2
4 f F * 1 1 Ma 2 1
L
ln
(12.134)
Ma 2 2
DH
1
2 1
Ma 2
2
where L* is the length of duct required, real or imaginary, for the flow to reach Ma = 1.
Equations (12.125) through (12.129) can be integrated in a similar manner:
1
1 2 1
2
Ma
2 1 2 Ma
*
V
12
(12.135)
c
1 1
2

1 2 Ma
T * c*
2
1 1 2 1
1
Ma 2
*
Ma 2
2
p
1 1
* Ma 2
12
(12.136)
12
1
2
1 2 Ma
(12.137)
(12.138)
2 1
s s*
1 2 1
2
Ma
ln
1
R
2
Ma 1
1 1
*
p0 Ma 2
p0
2 1
C. Wassgren
1
2
1 2 Ma
2 1
508
(12.139)
(12.140)
Notes:
1. The length of duct, L12, required for the flow to go from a given initial Mach number, Ma1, to a given
final Mach number, Ma2, is found by:
4 fF
4f
4f
L12 F L*1 F L*2
DH
DH
DH
*
*
where 4fFL 1/DH and 4fFL 2/DH are found from Eqn. (12.134) for the given Mach numbers.
Ma=1
Ma2
Ma1
L2*
L12
L1*
2.
In order to find the change in some flow property, e.g. the pressure, between two locations where the
Mach numbers are known, simply take the ratio:
p2 p*
p2
p1
p1 p*
3.
4.
5.
6.
Equations (12.134) through (12.140) are tabulated as a function of Mach number for air in the
appendices of most compressible flow texts.
What happens if the duct is longer than L*. Assuming the back pressure is low enough, for a subsonic
flow the flow adjusts upstream so that the length of duct becomes L*. If the flow is supersonic, then a
shock wave forms upstream at a location where the length from the shock to the end of the duct is L*.
Well discuss this in more detail in the next set of notes.
L* is typically on the order of 10-100 duct diameters for a supersonic flow, depending on the friction
factor. For example, for a friction factor of fF = 0.0025, the length of duct required to reduce a flow
with Ma = to Ma = 1 is 82 pipe diameters. Thus, supersonic flows reach sonic conditions in a short
distance. Furthermore, large losses in stagnation pressure also occur.
In a Fanno flow with no shock waves, there can be no transition between supersonic and subsonic
flows since both types of flow tend toward sonic conditions.
C. Wassgren
509
The Fanno Line
The Fanno Line is the locus of all possible Fanno flow states shown in a T-s plot. To determine how T and
s are related in a Fanno flow, we can combine equations (12.118), (12.120), (12.121), and (12.122) then
integrate to give:
1
1
2
T0 T
s s1
T
(12.141)
ln
T1 T0 T1
cp
Equation (12.141) allows us to relate the entropy, s, to an arbitrarily chosen temperature, T, for a given
stagnation temperature, T0. A plot of this relation is shown below:
p01
p02
T0 = constant
1
Ma<1
Ma>1
1
Ma=1
(maximum entropy)
2
s
The Fanno line shows all possible combinations of entropy and temperature that can exist in an adiabatic
flow with friction in a constant area duct at a given stagnation temperature (recall that the stagnation
temperature remains constant in an adiabatic flow).
Notes:
1. The processes always move in a direction that increases the entropy.
2. Again for a Fanno flow:
d(Ma) > 0
eqn (12.124)
d(Ma) < 0
dV > 0
eqn (12.125)
dV < 0
dT < 0
eqn (12.126)
dT > 0
dp < 0
eqn (12.127)
dp > 0
d< 0
eqn (12.128)
d > 0
ds > 0
eqn (12.129)
ds > 0
eqn (12.131)
dT0 = 0
dT0 = 0
eqn (12.130)
dp0 < 0
dp0 < 0
C. Wassgren
510
Example:
Air flows in a 0.100 m ID duct under adiabatic conditions. Calculate the length of duct required to raise the
Mach number of the air from Ma1=0.5 to Ma2=0.9 if the average value of the Fanning friction coefficient is
0.005.
C. Wassgren
511
Example:
An experiment is designed to measure friction coefficients for the supersonic flow of air. Measurements
from an experimental apparatus, consisting of a well-insulated, converging-diverging nozzle attached to a
smooth, round tube, give the following data:
pressure upstream of the nozzle = 516 cm Hg abs
temperature upstream of the nozzle = 107.3 F
throat diameter = 0.2416 in
diameter of nozzle exit and of tube = 0.5009 in
pressure at a point 29.60 diameters from the tube inlet = 37.1 cm Hg abs
What is the average Fanning friction factor for the tube with these conditions?
C. Wassgren
512
Choking in Fanno Flow
Consider what happens if the duct length has its maximum value for a given duct inlet Mach number (so
that the exit Mach number is unity) and then we increase the duct length beyond this value. Well assume
that the back pressure remains low enough to produce sonic exit conditions, i.e. pB p*.
Subsonic Flow:
The duct inlet Mach number will decrease until a steady-state solution again becomes possible with
sonic exit conditions. This decrease results in a reduction in the flow rate, i.e. the flow is choked by
friction. The mass flow rate through the duct can be found by the conditions at the duct inlet which is
assumed to have isentropic flow.
Mai < 1
isentropic
duct inlet
Mae = 1
L*
1

c
1
21
Ma 2
m VA 0 Ma c0 A 0 c0 AMa 1
2
c0
0
(As the Mach number decreases, the mass flow rate also decreases, i.e. m Ma 0 .)
Further increases in the duct length result in decreases in the duct inlet Mach number and mass flow
rate.
C. Wassgren
513
Supersonic Flow:
- First consider the case where the duct length has its maximum value for the inlet Mach number. The
flow within the duct will be supersonic and the exit conditions will be sonic (refer to curve A shown in
the plots below).
- Now increase the duct length to position B as shown in the figure. A shock wave forms in the duct at a
location such that the exit conditions are sonic. The mass flow rate through the device remains
unaffected since the flow at the nozzle throat is sonic.
- As the duct length is increased further, the shock wave moves upstream until finally at the duct length
indicated by position C, the shock wave is located at the duct inlet. Now the flow throughout the duct
is subsonic. The mass flow rate, however, remains unaffected since the throat conditions are sonic.
- Further increases in the duct length cause the shock wave to move into the diverging section of the
nozzle toward the nozzle throat. The duct inlet Mach number will continue to decrease but the mass
flow rate is unaffected.
- If the duct length is increased further, the shock vanishes at the nozzle throat and the flow throughout
the entire device becomes subsonic and further increases in the duct length will result in decreasing
duct inlet Mach number and decreasing mass flow rate through the device (refer to the previous
discussion concerning subsonic flow).
A
Ma
A,B
A
B
B
C
C
p/p*
C
A,B
C. Wassgren
x
A
514
Performance of Long Ducts at Various Pressure Ratios
Now lets consider the flow in long ducts with differing inlet nozzle conditions and varying back pressures.
Ducts Fed by Converging Nozzles
Consider the flow through a long duct fed by a converging nozzle as shown in the figure below. The
stagnation conditions are assumed fixed while the back pressure can be varied. For a high enough
back pressure (cases A and B shown in the plots below), the flow through the entire device, including
the exit, will remain subsonic and the exit pressure will equal the back pressure. If the back pressure is
decreased to precisely the sonic pressure (case C), the flow throughout the device will be subsonic
except at the exit where sonic conditions exist. For this particular case the exit pressure is equal to the
back pressure. If the back pressure is decreased further, the flow conditions within the device will
remain the same as in case C since the flow is choked at the exit; however, upon leaving the exit the
flow will expand through expansion fans to reach the back pressure.
p0, T0
fixed
pB (variable)
pE
frictionless
nozzle
constant area,
frictional duct
p/p0I
1
A
B
C
subsonic (MaE < 1)

subsonic (MaE < 1)
sonic (MaE = 1)
sonic (MaE = 1) with

expansion fans
x
C. Wassgren
515
Ducts Fed by Converging-Diverging Nozzles

Now consider flow through a duct with a converging-diverging inlet nozzle. The flow through this
device is considerably more complicated than through the duct fed by a converging nozzle. In the
following discussion well assume that the stagnation conditions feeding the nozzle are constant and
that the ratio of the nozzle exit/duct inlet area to the throat area is fixed, i.e. Ai/At = constant.
Subsonic flow at the nozzle exit / duct inlet:
Well first consider the cases where the Mach number at the nozzle exit/duct inlet is subsonic, i.e. MaI
< 1. Refer to the following figure for the various cases described below.
p0, T0
fixed
pE
frictionless
nozzle
pB (variable)
constant area,
frictional duct
p/p0I
1
1a, 1bi
1bii
p*/p0I
2a, 2bi
2bii
3a, 3bi
3bii
throat
exit
inlet
As the back pressure decreases:

1.
2.
3.
The flow in the nozzle and duct are subsonic. Consider cases (a) and (b) described below.
The flow in the nozzle is subsonic except at the nozzle throat where sonic conditions occur: MaT
= 1, pT = p*. The flow is now choked at the nozzle throat. Further decreases in back pressure will
not affect conditions upstream of the throat. Note that the Mach number at the duct inlet, MaI, for
this case is determined by the area ratio, AI/A* = AI/AT. Consider cases (a) and (b) described below.
The flow downstream of the throat is supersonic for some distance then passes through a normal
shock wave and becomes subsonic for the remainder of the nozzle. Hence, the flow into the duct
is subsonic. Consider cases (a) and (b) described below.
a. If the back pressure is greater than the sonic pressure for the inlet Mach number, then the exit
Mach number will be subsonic and the back pressure will equal the exit pressure: MaE < 1
p E = p B > p *.
b. If the back pressure is less than or equal to the sonic pressure, then the exit will be at sonic
conditions: MaE = 1 pE = p* pB. Note that the flow is now choked at the duct exit.
Further decreases in the back pressure will not affect the flow within the duct or nozzle.
i. If pB = p* then the flow equals the back pressure when exiting the duct.
ii. If pB < p* then expansion fans will form outside of the duct through which the flow
pressure will equilibrate with the back pressure.
C. Wassgren
516
Supersonic flow at the nozzle exit / duct inlet:

Now consider the cases when the Mach number at the nozzle exit/duct inlet is supersonic, i.e. MaI > 1.
Refer to the following figure for the various cases described below. Note that the Mach number at the
nozzle entrance is determined solely by the area ratio, AI/A*. Furthermore, since the Mach number is
fixed, the sonic length is also fixed.
p0, T0
fixed
pE
frictionless
nozzle
pB (variable)
constant area,
frictional duct
p/p0I
1
p* is in this range
p*/p0I
pE,supersonic
throat
exit
inlet
1
2
3
4
5
First consider the cases where the duct length is less than the sonic length, L < L*I, i.e. the exit
pressure, assuming supersonic flow throughout the duct, is less than the sonic pressure, pE,supersonic < p*.
(Recall that the pressure in a supersonic Fanno flow increases with increasing distance down the duct).
1.
2.
3.
4.
5.
The back pressure is sufficiently large that the flow cannot leave the duct under supersonic
conditions and come into equilibrium with the back pressure. Instead, a normal shock wave forms
within the duct causing a transition to subsonic flow. The shock wave forms at a location such
that the exit pressure equals the back pressure: MaE < 1 pE = pB. Decreasing the back pressure
causes the shock wave to move closer to the duct exit.
The entire duct flow is supersonic and the flow pressure equilibrates with the back pressure
through a normal shock wave at the duct exit. The flow downstream of the shock is subsonic so
the pressure just downstream of the shock is equal to the back pressure.
The flow within the duct is supersonic but pE < pB (also note that pE < p* as stated in the paragraph
above). Oblique shock waves form outside the duct through which the flow pressure equilibrates
with the back pressure.
The flow within the duct is supersonic and the pressure at the duct exit is equal to the back
pressure, pE = pB. Note that pE < p*.
The flow within the duct is supersonic but pE > pB (also note that pE < p*). Expansion fans form
outside the duct through which the flow pressure equilibrates with the back pressure.
C. Wassgren
517
p0, T0
fixed
pE
frictionless
nozzle
pB (variable)
constant area,
frictional duct
p/p0I
1
As pB, the shock

moves toward the exit.
p*/p0I
1
pE,supersonic
2
p*
throat
exit
inlet
Now consider the cases where the duct length is greater than or equal to the sonic length, L L*I, i.e.
the exit pressure, assuming supersonic flow throughout the duct, is greater than or equal to the sonic
pressure, pE,supersonic p*. (Recall that the pressure in a supersonic Fanno flow increases with
increasing distance down the duct).
1.
2.
3.
For the case where the back pressure is greater than the sonic pressure, i.e. pB > p*, a normal shock
wave will form at a location within the duct such that the exit flow (which is subsonic) will equal
the back pressure: MaE < 1 pE = pB > p*.
For the case where the back pressure is equal to the sonic pressure, i.e. pB = p*, a normal shock
wave will form at a location within the duct such that the exit flow is sonic: MaE = 1 pE = pB =
p*.
For the case where the back pressure is less than the sonic pressure, i.e. pB < p*, a normal shock
wave will form at a location within the duct such that the exit flow is sonic: MaE = 1 pE = p* >
pB. Expansion fans will form outside the duct through which the flow pressure equilibrates with
the back pressure.
Notes:
1. Once the flow chokes at either the nozzle throat or duct exit, the mass flow rate through the system will
no longer be a function of the downstream conditions and will depend only on the upstream stagnation
conditions.
2.
Shock waves within the duct will not be sharp discontinuities as idealized here. Instead, boundary
layer effects will tend to smear the transition from supersonic to subsonic flow over a considerable
distance. Similar shock smearing is observed when shock waves form in the diverging section of a
converging-diverging nozzle.
C. Wassgren
518
Example:
This example is not yet complete.
Air flows through a converging-diverging nozzle, with an exit-to-throat area ratio of 2.0, into a long, well
insulated duct that has a diameter of 0.10 m, (Fanning) friction factor of 0.001, and a length of 5 m.
Determine:
C. Wassgren
519
Isothermal Flow with Friction
In long pipelines there is too much surface area for the flow to be considered adiabatic. Instead a more
reasonable assumption is that the flow is isothermal. The Mach numbers in such flows tend to be small but
the pressure change, p, can be large we cant treat the flow as being incompressible.
Consider isothermal (dT=0), 1D, steady flow of an ideal gas in a constant area (dA=0) duct. Using the
same approach as the differential analysis for adiabatic frictional flow, we determine the following
relations:
d dV
0
(12.142)
COM:
V
Note: dA=0
4 w
dx 0
DH
LME:
dp VdV
COE:
c p dT0 VdV qinto
(12.143)
(12.144)
Notes:
1. dh=0 since dT=0.
2. dh0 = cpdT + VdV = qinto = Q into m
3. We are not assuming adiabatic conditions so heat transfer must be included.
2nd Law:
Ideal Gas Law:
ds
qinto
T
Since the flow is frictional, it is also irreversible.
dp d
p
Note: dT=0
d Ma
(12.145)
(12.146)
Mach # relation:
dV
Ma
V
Note: dT=0
(12.147)
Entropy relation:
dp
p
Note: dT=0
(12.148)
ds R
The local adiabatic stagnation temperature is given by:

1
T0 T 1
Ma 2
2
so that
dT0 1 Mad Ma T
where dT=0
T0
1
2
T 1
Ma
2
1 Ma 2 d Ma
dT0
Ma
T0
1
2
1 2 Ma
C. Wassgren
(12.149)
520
The local stagnation pressure is given by:

1
p0 p 1
Ma 2
2
so that
dp 1
Ma 2
dp0
2
p0
1 Mad Ma p 1
Ma 2
2
p 1
Ma 2
2
dp0 dp
1
Ma 2
1 Mad Ma 1
p0
p
2
2 1
1
Combine equations (12.142) through (12.149) leaving (4fFdx/D) as the independent variable:
2
dp d
dV
ds
Ma 2 4 f F
1 d Ma
dx
p
V
R 2 1 Ma 2 DH
2 Ma 2
dp0
p0
dT0
T0
(12.150)
Ma 2
Ma 2 1
2
4 f F dx
1
D
2 1 Ma 2 1
Ma 2 H
2
(12.151)
1 Ma 4
1
2
2 1 Ma
4 fF
dx
2 DH
1
Ma
Ma2 < 1/(subsonic)

d(Ma) > 0
dV > 0
dp < 0
dT = 0
d< 0
ds > 0
dT0 > 0 (heat added)
dp0 < 0
C. Wassgren
(12.152)
Ma2 > 1/ (subsonic/supersonic)

d(Ma) < 0
dV < 0
dp > 0
dT = 0
d > 0
ds < 0 (Note: ds > qinto/T)
dT0 < 0 (heat removed
dp0 > 0 for Ma2 < 2/(+1)
dp0 < 0 for Ma2 > 2/(+1)
eqn (12.150)
eqn (12.150)
eqn (12.150)
isothermal
eqn (12.150)
eqn (12.150)
eqn (12.152)
eqn (12.151)
521
Working equations can be found by integrating equations (12.150) through (12.152) using the location
where Ma2=1/ as a reference point (indicated by the superscript *t to signify limiting conditions for
isothermal, frictional flow).
4 f F L*t 1 Ma 2
ln Ma 2
DH
Ma 2
V
V *t
p0
p0*t
T0
T0*t
(12.153)
*t p*t
Ma
1 2
3 1
1+
Ma 2
1
2
Ma
(12.154)
(12.155)
2 1
1+
Ma 2
3 1
2
(12.156)
Notes:
1.
Lets determine the amount of heat that must be added (in the case of subsonic flow) or removed (in
the case of supersonic flow) for the flow to go from a given Mach number, Ma, to the choked flow
Mach number, Ma*t = 1/1/2. From equation (12.144) we have:
c p dT0 qinto
and from equation (12.152) we have
1 Ma 4
4 fF
dT0
dx
T0
2 1 Ma 2 1
Ma 2 H
2
Combining these two relations gives:

c pT0 1 Ma 4
4 fF
dx
qi nto
(12.157)
1
D
2 1 Ma 2 1
Ma 2 H
2
From equation (12.157) we see that as the Mach number approaches its limiting value of 1/1/2, the
local heat transfer required to maintain isothermal conditions becomes very large. Hence, the
assumption of isothermal flow for Mach numbers in the neighborhood of the limiting Mach number
may not be a good one since there may not be sufficient heat transfer to maintain isothermal
conditions.
C. Wassgren
522
2.
Consider the following flow situation.
p0, T0
pback
L
Assume that for the given back pressure, pback, the duct length is the sonic length, i.e. L=L*t and
Mae=1/1/2. Note that since the exit Mach number is subsonic, the exit pressure will equal the back
pressure, i.e. pe = p*t = pback. What happens if we now increase the duct length or drop the back
pressure? It can be shown that there is no solution using the isothermal, frictional flow equations for
the new conditions. The isothermal flow assumption breaks down since we would need to supply an
infinite amount of heat transfer to get to the new conditions (refer to equation (12.157)).
C. Wassgren
523
Example:
Natural gas (=1.3, R=75 (ftlbf)/(lbmR)) flows through a pipeline of diameter 3 ft and (Fanning) friction
factor 0.001. At a particular point in the pipeline the pressure is 200 psia, the temperature is 500 R, and
the velocity is 50 ft/s. The pipeline is kept at constant temperature.
a. Determine the maximum length of the pipe that one could have from this point given these conditions.
b. What would be the pressure at the maximum length?
C. Wassgren
524
8.
Flow with Heat Transfer (Rayleigh Flow)
Now lets consider compressible flows where heat transfer is significant. Examples of such flows include
those in which combustion, evaporation, condensation, or wall heat exchange occurs. Note that we wont
worry about how the heat gets into (or out of) the flow. Well just assume that we know its value,
otherwise we would need to include heat transfer analyses into our discussions.
To analyze the effects of heat transfer on a compressible flow, consider frictionless, 1D, steady flow of an
ideal gas in a constant area (dA=0) duct.
qinto
V+dV
(p+dp)A
pA
dx
x
Using the same approach as the differential analysis for frictional flow gives:
d dV
0
COM:
V
Note: dA=0
(12.158)
LME:
dp VdV 0
(12.159)
COE:
dh VdV qinto
Note: q Q m (heat addition per unit mass flow rate)
(12.160)
2nd Law:
ds
qinto
(12.161)
T
Note: The flow may be reversible if the temperature gradients are very small.
Ideal Gas Law:
dp d dT
p
T
(12.162)
caloric eqn of state:
dh c p dT
(12.163)
c p dT0 c p dT VdV qinto

d Ma
Mach # relation:
Entropy relation:
Ma
ds c p
C. Wassgren
dV dT
2T
V
(12.164)
dT
dp
R
T
p
(12.165)
525
Local stagnation pressure and temperature:

T 1
Ma 2
1
2
T0
(12.166)
p 1
1
Ma 2
p0
2
(12.167)
Combining these equations gives and using dT0 (= qinto/cP) as the independent variable:
1
1 Ma 2 1
Ma 2
d Ma
2
dT0
2
Ma
T0
2 1 Ma
dp
Ma 2
2
dT0
2
1 Ma
T0
Ma 2 1
1 Ma 1 2 1 Ma
1 Ma
2
dT
(12.168)
(12.169)
dT
T0
(12.170)
1
Ma 2
dV
d
2
dT0
2
V
1 Ma
T0
(12.171)
dp0
Ma 2 dT0
p0
2 T0
(12.172)
ds 1
dT
1
Ma 2 0
2
cp
T0
(12.173)
C. Wassgren
526
Note:
for qinto > 0 dT0 > 0

for qinto < 0 dT0 < 0
Ma > 1
Ma < 1
heat
heat
addition
removal
dT0 < 0
dT0 > 0
d(Ma) > 0
d(Ma) < 0
dV > 0
dV < 0
dp < 0
dp > 0
dT < 0 for Ma2<1/
dT > 0 for Ma2<1/
dT < 0 for Ma2>1/
dT > 0 for Ma2>1/
d< 0
d > 0
dp0 > 0
dp0 < 0
ds > 0
ds < 0
heat
addition
dT0 > 0
d(Ma) < 0
dV < 0
dp > 0
dT > 0
heat
removal
dT0 < 0
d(Ma) > 0
dV > 0
dp < 0
dT < 0
d> 0
dp0 < 0
ds > 0
d < 0
dp0 > 0
ds < 0
Notes:
1. dT0 = qinto/cp
2. Heating always decreases the stagnation pressure a loss in pressure recovery and thus efficiency.
Removing heat results in an increasing stagnation pressure; however, in practice other effects act to
decrease the stagnation pressure.
3. For 1/0.5<Ma<1, heat addition decreases the flow temperature while heat removal increases the flow
temperature. The energy input from the heat goes into the kinetic energy of the flow rather than the
thermal energy for this range of Mach numbers.
4. Mathematically, it appears that we could transition from subsonic to supersonic flow by controlling the
heat transfer rate. In practice, transitioning from subsonic to supersonic flow using heat transfer has
not been observed.
C. Wassgren
527
Integrating equations (12.168) through (12.173) using sonic conditions as a reference point gives the
following working equations:
p
p
1
1 Ma 2
(12.174)
V * 1 Ma 2
* V +1 Ma 2
T
T*
(12.175)
12 Ma 2
1 Ma 2
(12.176)
1 2 1 2
Ma
1
*
2
p0 1 Ma 2 1
p0
T0
T0*
(12.177)
2 1 Ma 2 1
1
Ma 2
2
2
1 Ma 2
(12.178)
s s*
1
2
ln Ma
2
cp
1 Ma
(12.179)
q12 c pT01 02 1
T01
(12.180)
Note that the maximum amount of heat that can be added to a flow for a given initial state is:
T*
q12,max c pT01 0 1
T01
where, using Eqn. (12.178),
T0*
T01
2
2 1 Ma
2
2
1 Ma
1
2
1
Ma
so that
q12,max
2
2 1 Ma
c pT01
2
1 Ma 2
C. Wassgren
1
2
1
Ma
1
528
(12.181)
Plotting the locus of all possible states for Rayleigh flow on a T-s diagram gives the Rayleigh Line:
T *
T
s* s1
1
Ma 1
ln
1
cp
*
T
p
1
Ma=1
(maximum entropy)
Ma<1
Ma>1
s
The arrows are drawn for heat addition.
Reverse the arrows for heat removal.
C. Wassgren
529
Example:
Air flows with negligible friction through a duct of area 0.25 ft2. At section 1, flow properties are T1=600
R, p1=20 psia, and V1=360 ft/s. At section 2, p2=10 psia. The flow is heated between sections 1 and 2.
Determine the properties at section 2, the energy added, and the entropy change. Finally, plot the process
on a T-s diagram.
C. Wassgren
530
Example:
Air flows in a constant-area duct. At the inlet the Mach number is 0.2, the static pressure is 90 kPa, and the
static temperature is 27 C. Heat is added at a rate of 120 kJ/(kg of air). Assuming a perfect gas with
constant specific heats, determine the properties of the air at the end of the duct. Assume also that the flow
is frictionless and that cp=1000 J/(kgK).
C. Wassgren
531
Example:
A gaseous mixture of air and fuel enters a ram-jet combustion chamber with a velocity of 200 ft/s, at a
temperature of 120 F, and at a pressure of 5 psia. The heat of reaction H of the mixture for the particular
fuel-air ratio employed is 500 Btu/(lbm of the mixture). It is desired to find the stream properties at the exit
of the combustion chamber. It will be assumed that friction is negligible, that the cross-sectional area is
constant, and that the properties of both the reactants and the products are equivalent to air in respect to
molecular weight and specific heat.
C. Wassgren
532
9.
Normal Shock Waves
Consider the movement of a piston in a cylinder:
stagnant gas
compression wave
When we first move the piston, an infinitesimal (compression) pressure wave travels down the cylinder at
the sonic speed. Behind the wave, the pressure, temperature, density, and increase slightly and the fluid has
a small velocity following the wave.
If we continue to increase the piston velocity, additional pressure waves will propagate down the cylinder.
However, these waves travel at a slightly increased speed relative to a fixed observer due to the increased
fluid temperature and fluid movement. The result is that the waves formed later catch up to the previous
waves. When the waves catch up to the first wave, their effects add together so that the small changes
across the individual waves now become a sudden and finite change called a shock wave.
1st wave
2nd wave
u1
c2
T1
c1
u0 = 0
T0
c2 > c1 since T1 > T0 and u1 moves to the right

t, time
piston
shock wave
pressure wave
(sound wave)
x, position
C. Wassgren
533
Notes:
1.
2.
3.
4.
5.
The velocity of a shock wave is greater than the speed of sound since:
p
c2
1

For a sound wave, d 0. For a shock wave, so that cshock wave > csound wave.
A shock wave is a pressure wave across which there is a finite change in the flow properties.
Shock waves only occur in supersonic flows.
Shock waves are typically very thin with thicknesses on the order of 1 m. Thus, we consider the
changes in the flow properties across the wave to be discontinuous.
The sudden change in flow properties across the shock wave occurs non-isentropically since the
thermal and velocity gradients are large within the shock wave itself.
To analyze a shock wave, well use an approach similar to that used to examine a sound wave. Lets
consider a fixed shock wave across which flow properties change:
very thin CV so there is no flow
out of the top and bottom
p1, T1, 1, V1
upstream
(isentropic flow)
p2, T2, 2, V2
downstream
(isentropic flow)
fixed shock wave

(non-isentropic process)
COM:
1V1 A 2V2 A
1V1 2V2
(12.182)
LME (in x-direction):

2 mV
1 p1 A p2 A
mV
1V1 V2 V1 2V2 V2 V1 p1 p2
(12.183)
COE:
h1 1 2 V12 h2 1 2 V22
(12.184)
Also, h01 h02 . Note that no heat is added to the CV, i.e. the process is adiabatic.
2nd Law:
s2 s1 (since the process is adiabatic but irreversible)
Thermal Equation of State (ideal gas law):
p1
p
2 R
1T1 2T2
(12.186)
Caloric Equation of State (for a perfect gas):

h c pT
C. Wassgren
(12.185)
(12.187)
534
Combining equations (12.182) and (12.183):

p1
p
2 V2 V1
1V1 2V2
Substituting equation (12.186):
RT1 RT2
V2 V1
V1
V2
Substituting equations (12.184) and (12.187) then re-arranging:
V2 R
V2
R
T0 1 T0 2 V2 V1
2c p V2
2c p
V1
RV2T0
(12.188)
RV2V12
RV V 2
RV1T0 1 2 V1V2 V2 V1
2c p
2c p
VV
V1V2 V2 V1 R V2 V1 T0 V2 V1 1 2
2c p
VV
V1V2 V2 V1 R V2 V1 T0 1 2
2c p
VV
V1V2 R T0 1 2
2c p
R
V1V2 1
RT0
2c p
RT0
V1V2
R
1
2c p
Finally, substituting the relation:

R c p cv 1
cp
cp
and re-arranging we have:

2 RT0
V1V2
1
C. Wassgren
Prandtls Equation
(12.189)
535
Dividing both sides of the equation by the sound speed on either side of the shock wave and utilizing the
definition for the Mach # for a perfect gas:
RT0 RT0
V1
V2
2
RT1 RT2 1 RT1 RT2

T0 T0
2
1 T1 T2
Recall that for the adiabatic flow of a perfect gas:
Ma1Ma 2
T 1
1
Ma 2
T0
2
So that:
T0
2
Ma1Ma 2
1 T1
T0
T2
1
2 1
2 1
1
Ma12 1
Ma 22
1
2
2

After additional algebra we can reduce this equation to the following:

Ma 22
1 Ma12 2
2 Ma12 1
(12.190)
Relation between upstream Mach # (Ma1) and downstream Mach # (Ma2) across a normal shock
wave.
Notes:
1. When Ma1>1, then Ma2<1 (supersonic to subsonic flow) and when Ma1<1, then Ma2>1 (subsonic to
supersonic flow).
2. From experiments, we observe that shock waves never form in subsonic flows (Ma1<1) even though
equation (12.190) does not give any indication of this. Well use the 2nd law in a moment to show that
shock waves can only form in supersonic flows (Ma1>1).
The temperature ratio across the shock wave can be determined using the adiabatic stagnation temperature
relation for a perfect gas and noting that the stagnation temperature remains constant across a shock:
1
T2 1 1 Ma 2
2
T
2
0
1
T1 1
2
T 1
Ma
1
0
2
1
Ma12
T2
2
T1 1
2
1 2 Ma 2
C. Wassgren
(12.191)
536
The pressure ratio across the shock can be determined by combining equations (12.191), (12.186), and
(12.182) along with the definition of the Mach number for a perfect gas:
p2 2T2 V1T2
1T1 V2T1
p1
Ma1
Ma 2
RT Ma T
RT1 Ma1 T2
2
T2
T1
1 2
1
Ma1
p2 Ma1
2
p1 Ma 2 1 1 Ma 2
2
2
(12.192)
1 2
2 V1 p2T1 Ma1 1 2 Ma1
1 V2 p1T2 Ma 2 1 1 Ma 2
2
2
1 2
2 V1 Ma1 1 2 Ma1
1 V2 Ma 2 1 1 Ma 2
2
2
1 2
1 2 Ma1
1 1 Ma 22
2
(12.193)
We can also determine the ratio of the isentropic stagnation pressures and densities across the shock wave:
p 1
1
Ma 2
p0
2
1 2
p1

Ma1
p 1
01
1 2
p2
Ma 2
p 1
2
02
p02
p01
1 2
p2 1 2 Ma1
p1 1 1 Ma 22
2
p02
p01
1 2
1
Ma1
Ma1
2
Ma 2 1 1 Ma 2
2
2
21
C. Wassgren
1 2
1
Ma1
Ma1
2
Ma 2 1 1 Ma 2
2
2
1 2
1 2 Ma1
1 2
Ma 2
1
2
(12.194)
537
02 p02 T01
01 p01 T02
but since T01 = T02
02 p02
01 p01
1 2
1
Ma1
Ma1
2
Ma 2 1 1 Ma 2
2
2
21
(12.195)
The sonic area ratio across the shock can be determined from the fact that the mass flow rate across the
shock must remain constant (COM):
m 1 m 2
1*V1* A1* 2*V2* A2*

A2*
1* V1*
A1* 2* V2*
The sonic ratios can be determined from the following:
1*
* 1 1
2 1
2
01 02
1* 01
2* 02
1 2
1
Ma1
Ma1
2
Ma 2 1 1 Ma 2
2
2
2 1
and
T1*
T
0
V1* c1*
T1*
1
*
*
*
*
V2 c2
T2
T2
T
0
Note that T01 = T02 has been used in the previous equation. Substituting these two sonic ratios and
simplifying:
1 2
1
Ma1
A2* Ma 2
2
A1* Ma1 1 1 Ma 2
2
2
C. Wassgren
2 1
(12.196)
538
Note that we could have also used the isentropic area ratios on either side of the shock wave:
1 2
1
Ma1
A1
1
2
A1* Ma1 1 1
so that:
2 1
1 2
1
Ma 2
A2
1
2
A2* Ma 2 1 1
and
A1
1 2
*
1
Ma1
*
A2 A1 Ma 2
2
A1* A2 Ma1 1 1 Ma 2
*
2
A2
2 1
2 1
where A1=A2
Notes:
1. Equations (12.191)-(12.196) may be written only in terms of Ma1 by substituting equation (12.190).
The resulting equations are:
Ma 22
1 Ma12 2
2 Ma12 1
(12.197)
2 Ma 2 1
T2
1
2 1 Ma12
T1
12 Ma12
(12.198)
1 Ma1
2 V1
1 V2 1 Ma12 2
(12.199)
p2
2
1
Ma12
p1 1
1
(12.200)
T02
1
T01
(12.201)
p02 02
p01 01
Ma12
A1* 2
*
A2 1 1 Ma 2
1
C. Wassgren
2
1
Ma12
1
1
539
1
1
(12.202)
2.
Now lets examine the change in entropy across the shock using:
T
p
s2 s1 c p ln 2 R ln 2
T1
p1
If we substitute equations (12.198) and (12.200) into equation (12.203) and plot:
(12.203)
0.10
0.05
(s2 -s 1)/c p
0.00
-0.05
0.5
1.0
1.5
2.0
-0.10
-0.15
-0.20
-0.25
-0.30
Ma1
We observe that for Ma1 < 1 the entropy decreases across the shock. The 2nd law, however, states that
the entropy must increase across the shock (refer to equation (12.185)). Thus, shock waves can only
form when Ma1>1.
Also note that as the upstream Mach number approaches one (Ma11), the flow through the shock
approaches an isentropic process. An infinitesimally weak shock wave, one occurring when Ma1=1,
results in an isentropic process. This type of shock is, in fact, just a sound wave.
C. Wassgren
540
3.
Plots of equations (12.198)-(12.202) as a function of Ma1 are shown below:
values
Ma2
T2/T1
p2/p1
r2/r1
p02/p01
A2*/A1*
3
2
V2/V1
1
0
1.0
1.5
2.0
2.5
3.0
Ma1
The plot shows the following relations:
T2 > T1 and T2/T1 as Ma1
p2 > p1 and p2/p1 as Ma1
2 > 1 and 2/1 as Ma1
V2 < V1 and V2/V1 as Ma1
T02 = T01
p02 < p01 and p02/p01 as Ma1
02 < 01 and 02/01 as Ma1
A2* > A1* and A2*/A1* as Ma1
1
Ma 2
Ma1
2
Ma2 as Ma1 Furthermore, lim

4.
The shock strength is defined as the change in pressure across the shock wave relative to the upstream
pressure: p/p1=p2/p1-1. Viewing the trends shown in the previous plot, the larger the incoming Mach
number the stronger the shock wave.
C. Wassgren
541
5.
On a T-s diagram, the states across a shock wave correspond to the intersection of the Fanno and
Rayleigh lines for the flow:
T, temperature
downstream
state (state 2)
Fanno line
Rayleigh line
upstream state
(state 1)
s, entropy
The reason for this is because the flow across the shock satisfies the Fanno relations for COM
(equation (12.182)), COE (equation (12.184)), and the ideal gas relations (equations (12.186) and
(12.187)). The shock also satisfies the Rayleigh relations for COM, COLM (equation (12.183)), and
the ideal gas relations. The shock states must therefore occur at the intersection of the Fanno and
Rayleigh lines in order for the shock to satisfy all of the basic relations simultaneously. Furthermore,
state 2 lies to the right of state 1 in the T-s diagram since entropy must increase across the shock (2nd
law).
C. Wassgren
542
Example:
According to a newspaper article, at the center of a 12,600 lbm Daisy-Cutter bomb explosion the
overpressure in the air is approximately 1000 psi. Estimate:
a.
the speed of the resulting shock wave into the surrounding air,
b.
the wind speed following the shock wave,
c.
the temperature after the shock wave has passed, and
d.
the air density after the shock wave has passed.
C. Wassgren
543
10. Flow in Converging-Diverging Nozzles
Consider flow through a converging-diverging nozzle (aka a deLaval nozzle) as shown below.
p0, T0, 0
V0
pB
pE
Athroat, pthroat
Lets hold the stagnation pressure, p0, fixed and vary the back pressure, pB. The plot below shows how the
static pressure ratio, p/p0, varies with location, x, for various values of the back pressure ratio, pB/p0:
p/p0
1
2
3
5
6
p*/p0
7
4
8
x
throat
exit
Cases:
1. There is no flow through the device since pB = p0.
2. There is subsonic flow throughout the device and pE = pB.
3. There is subsonic flow throughout the device except at the throat where pthroat = p* (Ma=1). The flow
is now choked; further decreases in pB will not affect the flow upstream of the throat. The exit
pressure will equal the back pressure, pE = pB, since the flow is subsonic at the exit.
4. Subsonic flow will occur in the converging section, sonic flow will occur at the throat (pthroat = p*), and
supersonic flow will occur in the diverging section. This type of flow is called correctly expanded
flow or flow at design conditions since no shock waves form anywhere in the device and pE = pB.
5. Subsonic flow will occur in the converging section and sonic flow will occur at the throat (pthroat = p*).
A portion of the diverging section will be supersonic with a normal shock wave occurring at a location
such that the subsonic flow downstream of the flow will have an exit pressure equal to the back
pressure: pE = pB. As the back pressure decreases, the shock wave moves downstream of the throat
and toward the exit. The pressure rise across the shock wave also increases as the back pressure
decreases.
6. This case is similar to case 5 except that the shock wave is precisely at the nozzle exit. The pressure
just downstream of the shock wave equals the back pressure since the flow is subsonic there. The flow
everywhere within the C-D nozzle is isentropic except right at the exit.
7. The flow within the C-D nozzle (and the exit) is isentropic. The normal shock that was located at the
exit for case 6 has moved outside the device to form a complicated sequence of oblique shock waves
alternating with expansion fans (these are 2D phenomena to be discussed in a following section of
notes). This case is called the overexpanded case since the diverging section of the device has an area
that overexpands the flow to a pressure that is lower than the back pressure. External shock waves are
required to compress the flow to match the back pressure.
8. This case is similar to case 7 except that the flow outside of the device forms a sequence of expansion
fans alternating with oblique shock waves (a sequence out of phase with the sequence mentioned in
case 7). This case is called the underexpanded case since the diverging section of the device has an
area that is not large enough to drop the exit pressure to the back pressure. External expansion waves
are required expand the flow to match the back pressure.
C. Wassgren
544
Notes:
1. The critical back pressure ratio corresponding to case 3 can be found from the isentropic relations (the
flow throughout the entire device is isentropic). Assume that the geometry, and hence the exit-tothroat area ratio, Ae/At, is given. Since for case 5 the flow is choked we know that At = A*.
Furthermore, since the exit flow is subsonic we also know that pe = pb. From the area ratio we can
determine the exit Mach number, Mae:
1
1
2 1
1
Ma e2
Ae Ae
1
2
(where the subsonic Mae is chosen)
A* At Ma e 1 1
The back pressure ratio, pb/p0, for case 5 can be determined given the exit Mach number.
pb pe 1
1
1
Ma e2
2
p0 p0
2.
The critical back pressure ratio corresponding to case 4 can be determined in a manner similar to that
described above in Note 1. For case 4 however, the supersonic value for Mae should be used when
determining the exit Mach number from the area ratio.
3.
The critical back pressure ratio corresponding to case 6 can be found by combining the isentropic
relations with the normal shock wave relations. When the shock wave occurs right at the exit of the
device, the flow just upstream of the exit can be found from the isentropic relations:
1
1
2 1
1
Ma e21
Ae Ae
1
2
A1* At Ma e1 1 1
(where the supersonic Mae1 is chosen)
pe1 1
1
1
Ma e21
2
p01
Note that the subscript 1 denotes the conditions just upstream of the shock wave. To determine the
conditions just downstream of the shock we use the normal shock wave relations:
pe 2
2
1
Ma e21
1
pe1 1
where pe2 is the pressure just downstream of the shock. Since the downstream flow is subsonic and
because were at the exit of the device, the downstream pressure, pe2, must also equal the back
pressure, pb. Thus,
pb
p
p p
2
1 1 2 1
Ma e21
1
Ma e1
e 2 e 2 e1
1
2
p01 p01
pe1 p01 1
C. Wassgren
545
4.
The location of a shock wave for a back pressure in the range corresponding to case 3 and case 5 can
be determined through iteration.
a. Assume a location for the shock wave (e.g. pick a value for A/At since the geometry is known).
b. Determine the Mach number and pressure just upstream of the shock, Ma1 and p1, using the
isentropic relations as discussed in Note 2.
1
1
Ma12
1
A
A
2
A1* At Ma1 1 1
2 1
(where the supersonic Ma1 is chosen)
c.
p1 1
1
1
Ma12
p01
2
Calculate the stagnation pressure ratio and sonic area ratio across the shock using the normal
shock relations:
d.
1
1
1
Ma12
*
p02 A1
2
1 1
2
Ma
* 2

1
p01 A2 1 1 Ma 2 1
1
1
2
Determine the exit Mach number and exit pressure ratio using the isentropic relations and the
downstream sonic area and stagnation pressure:
1
1
2 1
1
Ma e2
Ae Ae A1*
1
2
(where the subsonic Mae is chosen)
A2* At A2* Ma e 1 1
Note that since the flow is choked, the throat area is equal to the upstream sonic area, i.e. At = A1*.
pe 1
1
1
Ma e2
2
p02
e.
f.
Note that since the exit Mach number is subsonic, the exit pressure will equal the back pressure,
i.e. pe = pb.
Calculate the ratio of the back pressure to the upstream stagnation pressure:
pb
p p
e 02
p01 p02 p01
Check to see if the back pressure ratio calculated in step (e) matches with the given back pressure
ratio. If so, then the assumed location of the shock is correct. If not, then the go back to step (a)
and repeat. If the back pressure ratio calculated in part (e) is less than the given back pressure
ratio, then the assumed shock location is too far downstream. If the back pressure ratio calculated
in part (e) is greater than the given back pressure ratio, then the assumed shock location is too far
upstream.
p/p01
actual
pb/p01
actual
location
exit
C. Wassgren
546
(Figure from: Liepmann, H.W. and Roshko, A., Elements of Gasdynamics, Wiley.)
C. Wassgren
547
5.
In real nozzles flows, the flow will typically separate from the nozzle walls as a result of the large
adverse pressure gradient occurring across a shock wave. Interaction of the shock with the separated
boundary layer results in a more gradual pressure rise than what is expected for the ideal, normal shock
analysis.
It is also possible that downstream pressure information can propagate upstream in the diverging
section even when the core flow is supersonic. In a real flow, a boundary layer will form along the
wall with the flow in part of this boundary layer being subsonic. Thus, pressure information can
propagate upstream within the subsonic part of the boundary layer and affect the flow in the diverging
section. When the back pressure is in the range corresponding to Case 7 (back pressure less than the
exit pressure when a shock stands at the exit, and greater than the isentropic case corresponding to
supersonic diverging section flow), oblique shocks will typically form within the diverging section and
flow separation occurs as shown in the figures below. The exact pressure and location of the
separation point are dependent on the boundary layer flow.
C. Wassgren
548
A converging-diverging nozzle with pressure taps along the

length of the device. The flow is from left to right.
The pressure ratio as a function of the axial distance in the CD nozzle for various back pressures.
Note the gradual

pressure rise due to the
interaction between the
shock and the
separated boundary
layer.
C. Wassgren
549
Example:
A converging-diverging nozzle, with Ae/At = 1.633, is designed to operate with atmospheric pressure at the
exit plane. Determine the range(s) of stagnation pressures for which the nozzle will be free from normal
shocks.
C. Wassgren
550
Example:
A converging-diverging nozzle, with an exit to throat area ratio, Ae/At, of 1.633, is designed to operate with
atmospheric pressure at the exit plane, pe = patm.
a. Determine the range(s) of stagnation pressures for which the nozzle will be free from normal shocks.
b. If the stagnation pressure is 1.5patm, at what position, x, will the normal shock occur?
The converging-diverging nozzle area, A, varies with position, x, as:
2
A x Ae
2x
1 1 1
At
At
L
throat area, At
exit area, Ae
stagnation
conditions
/2L
L
C. Wassgren
551
Supersonic Wind Tunnel Design
There are three common designs for supersonic wind tunnels:

1. high-pressure gas storage tanks (and/or vacuum tanks) for blowdown wind tunnels,
2. a compressor and diffuser for continuous-duty wind tunnels, and
3. shock tubes for high-enthalpy wind tunnels.
Here well study only the first two categories: blowdown and continuous-duty wind tunnels.
Blowdown Wind Tunnels
A schematic for a typical blowdown wind tunnel is shown below. Another possible design would be to use
atmospheric conditions at the inlet and use a vacuum tank at the exit (such a design is effectively the same
as the one shown below).
C-D nozzle with
throat area, At
model to be tested
tank with
pressure , p01,
temperature, T01,
and volume, V
back pressure, pb
test section with
design Mach number, MaTS > 1,
and area, ATS
The wind tunnel will have supersonic flow in the test section as long as the back-to-tank pressure ratio,
pB/p01, is less than the back pressure ratio corresponding to case 6 shown in the figure below (no shock
waves anywhere within the device). If the back pressure ratio becomes too high, then a shock wave will
form in the diverging section of the nozzle and there will be subsonic flow in the test section.
p/p0
1
2
3
5
6
p*/p0
throat
C. Wassgren
exit
552
7
4
8
x
Image from: http://history.nasa.gov/SP-440/ch5-6.htm

Notes:
1.
There is a fixed amount of time for which the device will operate at the design test section Mach
number, MaTS, since the tank mass will decrease with time. To extend the duration of the test, a
diverging section can be added to the exit of the device as shown below.
C-D nozzle with

throat area, At
tank with
pressure , p01,
temperature, T01,
and volume, V
diverging section with exit area, AD

back pressure, pb
test section with

design Mach number, MaTS > 1,
and area, ATS
The presence of the diverging section allows the tank to drop to a lower pressure before a shock wave
appears ahead of the test section.
C. Wassgren
553
Example:
A blowdown wind tunnel exhausting to atmospheric pressure (14.7 psia) is to be designed. The test section
cross-sectional area is specified to be 1 ft2, and the desired test section Mach number is 2.0. The supply
tank can be pressurized to 150 psia and heated to 150 F. Determine the throat area and supply tank
volume required for a testing time of 30.0 s.
If a diverging section with an area ratio equal to 3.375 times that of the throat is added downstream of the
test section, what is the new testing time?
C. Wassgren
554
Continuous-Duty Wind Tunnels

Continuous duty wind tunnels utilize a compressor to produce the driving pressure gradient for the flow. In
order to minimize the required compressor power, the wind tunnel should operate as efficiently as possible,
i.e. as close to isentropic conditions as possible. Continuous-duty wind tunnels can be either open-circuit
where air is drawn in and exhausted to the surroundings, or closed-circuit where the working gas is
recycled through the system. The following schematics show examples of both types of systems.
C. Wassgren
555
Again, in order to minimize the compressor power requirements, the losses in the system should be
minimized. The ideal case (shown below) is to have an isentropic deceleration from supersonic to subsonic
speeds.
Mat2 = 1
Mat1 = 1
MaTS > 1
p01
Ma < 1
Ma < 1
pback
p/p01
Of course in reality there would
be viscous boundary layer losses
in the duct which would result in
a loss of stagnation pressure in
the duct. Well ignore those
losses here and assume isentropic
flow.
1
p*/p01
1st throat
2nd throat
TS
Consider what happens if we design the wind tunnel such that the downstream throat has a smaller area
than the upstream throat, i.e. At2 < At1. As we decrease the back pressure ratio, pback/p01, then we will have
subsonic flow throughout the device until at a critical back pressure the flow through the 2nd throat will
become choked. As we decrease the back pressure further, a shock wave will form in the diverging section
of the downstream diffuser (refer to the figure shown below). The test section is considered blocked since
further reductions in the back pressure will not cause any changes upstream of the second throat. Since the
test section was subsonic before blocking occurred, it will remain subsonic after blocking.
Mat2 = 1
Mat1 < 1
p01
MaTS < 1
Ma < 1
Ma < 1
pback
At2 = A1* < At1
p/p01
1
p*/p01
1st throat
C. Wassgren
2nd throat
TS
556
Now consider what happens if we make the 2nd throat just a little bit larger than the 1st throat. As we
decrease the back pressure we will reach a case where the flow in the 1st throat becomes choked and a
shock wave forms in the diverging section of the first throat. The flow in the test section will be subsonic.
As the back pressure decreases, the shock wave in the first throat moves further downstream and becomes
stronger. Recall that as the shock becomes stronger, A2*/A1* increases. If the second throat area is smaller
than the A2* for the strongest shock wave, i.e., one that stands at the test section entrance, then a second
shock will appear downstream of the second throat and the flow is once again blocked. The figure shown
below illustrates this condition.
Mat2 = 1
Mat1 = 1
MaTS < 1
Ma < 1
p01
Ma < 1
pback
At2 = A2*
At1 > At2

At1 = A1*
p/p01
1
p*/p02
p*/p01
1st throat
2nd throat
TS
If the downstream throat has an area greater than the sonic area downstream of the shock when the shock
wave stands at the entrance of the test section (see the figure below)
Mat2 < 1
Mat1 = 1
p01
MaTS < 1
Ma < 1
Ma < 1
pback
At1 = A1*
A2*
At1 where A2*/A1* is found from MaTS
A1*
For example, if MaTS = 2.0, At2 > 1.39 At1.
At 2 A2*
p/p01
1
p*/p01
p*/p02
1st throat
C. Wassgren
TS
557
2nd throat
and we decrease the back pressure further, then the shock will be swallowed by the 2nd throat and the flow
within the test section will, at last, be supersonic (shown below).
Mat2 > 1
Mat1 = 1
p01
MaTS > 1
Ma < 1
Ma < 1
pback
At1 = A1*
p/p01
1
p*/p01
1st throat
TS
2nd throat
Now lets get back to the original discussion. Once the shock has been swallowed by the second throat, the
shock will stand in the diverging section of the downstream diffuser. The wind tunnel is now considered
running or started. In order to isentropically decelerate the flow we should now decrease the area of the
2nd throat so that it is approximately the same as the upstream throat area. Since there are no longer any
shock waves upstream of the 2nd throat, we theoretically could approach the 1st throat area; however, due to
boundary layer effects we will always have to make the 2nd throat slightly larger than the first. Decreasing
the 2nd throat area too much results in shock waves in the diverging sections of both the 1st and 2nd throats
(discussed previously) and the wind tunnel is once again blocked.
Photo from: http://history.nasa.gov/SP-440/ch5-5.htm
C. Wassgren
558
Once the wind tunnel is running and weve decreased the 2nd throat area, we should try to minimize the
stagnation pressure loss through the shock wave in the 2nd diverging section (and, hence, increase the tunnel
efficiency). To do this we increase the back pressure, pback, so that the shock wave will move further
toward the second throat thus decreasing the shocks strength.
Mat2 > 1
Mat1 = 1
p01
MaTS > 1
Ma < 1
Ma < 1
pback
p/p01
1
p*/p01
1st throat
TS
2nd throat
The ideal case is to have the shock positioned exactly at the second throat. In practice, however, a shock
standing exactly at the second throat is unstable and could disgorge, i.e. move back into the diverging
section of the first throat, and block the test section once again.
Note:
An excellent reference on the history of wind tunnel development at NACA/NASA can be found at:
http://www.hq.nasa.gov/office/pao/History/SP-440/contents.htm
C. Wassgren
559
Example:
Consider a supersonic wind tunnel starting as shown in the figure below. The upstream nozzle throat area
is 1.25 ft2, and the test section design Mach number is 2.50. As the tunnel starts, a normal shock stands in
the divergence of the nozzle where the area is 3.05 ft2. Upstream stagnation conditions are T01 = 1080 R
and p01 = 115 psia. Find the minimum possible diffuser throat area at this instant. Calculate the entropy
increase across the shock. What would be the minimum possible diffuser throat area to start this wind
tunnel?
p01, T01
C. Wassgren
560
Supersonic Diffuser Design
Another application where the efficient deceleration of a supersonic flow is of interest is a supersonic
diffuser at the inlet of aircraft jet engines. The flow entering a jet engine typically needs to be subsonic in
order to avoid shocks in the compressor section and give efficient combustion in the combustor. Obviously
the most efficient deceleration of the incoming flow is desired since the thrust out of the device will
decrease if the upstream stagnation pressure decreases.
Many of the same ideas discussed previously for the design of supersonic wind tunnels are pertinent here as
well. Consider a diffuser with a fixed inlet and throat area, Ai and At, respectively, in a supersonic flow
with an upstream Mach number of Ma > 1. At design conditions, i.e. Ma = MaD, the flow through the
diffuser will be shockless (isentropic) as shown in the figure below.
(Ma > 1) = MaD
Mat = 1
Ma < 1
At
Ai
In the ideal case the inlet-to-throat area ratio will be related to the design Mach number, MaD, by the
isentropic relation for the sonic area ratio, i.e.:
1
1
Ma 2D
Ai
Ai
1
2
A* At Ma D 1 1
2 1
(12.204)
However, we need to consider what happens as the aircraft comes up to the design Mach number from rest
(Ma = 0). Consider the plot shown below:
A/A*
1
1
MaD
Ma
For Ma < MaD, (Ai/A*) < (Ai/A*)D so that At = A*D < A*Ma < MaD. Thus, the diffuser cannot swallow all of
the air flowing toward the inlet. A shock wave forms in front of the inlet to produce subsonic flow so that
some of the air can spill over the inlet as shown in the figure below.
(Ma > 1) < MaD
C. Wassgren
561
As the upstream Mach number increases, the sonic area approaches the throat area, i.e. A* At, and the
shock moves closer to the inlet (the shock gets weaker and less flow needs to be diverted around the
diffuser). Eventually well reach design conditions but a normal shock will still appear ahead of the inlet
since the sonic area after the shock, A2*, will be greater than the throat area, At, at design conditions.
Ai
Ai A1* Ai A1*
A2* A1* A2* At A2*
(Ma > 1) = MaD

1 2
A2* At
If we continue to increase Ma the shock wave will move closer to the inlet until at a critical upstream
Mach number, Ma, crit, the shock will be positioned exactly at the inlet such that the diffuser can
accommodate all of the mass flow heading toward it (no spill-over). This occurs when A2* = At:
Ai Ai
A A*
* *i 1* fcn Ma , crit
At A2 A1 A2
where Ai/A1* is found from the isentropic relations and A1*/A2* is found from the normal shock relations.
Ma,crit > 1
1 2
A further increase in the upstream Mach number will cause the shock wave to be swallowed by the diffuser
where it will come to a steady state position within the diverging section.
Ma > Ma,crit
Now the upstream Mach number can be decreased back down to the design Mach number so that the shock
wave will travel back upstream toward the throat and, hence, become weaker (less stagnation pressure drop
across the shock). The ideal case is to bring the Mach number down exactly to the design Mach number so
that the shock occurs exactly at the throat and has zero strength.
Notes:
1. In practice we dont want to operate the diffuser exactly at design conditions since any small decrease
in the upstream Mach number will cause the shock wave to disgorge from the diffuser and the entire
process for swallowing the shock must be repeated again.
2.
One measure of the diffuser performance is the stagnation pressure recovery coefficient defined as:
p
0,exit of diffuser
p0
At design conditions under ideal conditions this coefficient should be unity.
C. Wassgren
562
3.
Over-speeding the diffuser is often impractical. For example, consider a diffuser designed to operate
at a Mach number of 1.7 (Ai/A* = Ai/At = 1.338). The critical Mach number for swallowing the shock
will be:
Ma,crit > 1
Note: A2* = At
1 2
1
1
Ma 22
Ai
Ai
1
2
1.338
Ma 2 1 1
A2* sub At
2
Ma ,crit 2.65
2 1
Ma 2 0.5 Ma1 2.65
Thus, to achieve isentropic flow through the diffuser, we would need to operate our diffuser at a Mach
number just greater than 2.65 then decrease the Mach number down to just greater than 1.7. Designing
an aircraft to achieve this over-speed Mach number is often impractical.
4.
Since fixed geometry diffusers are often impractical, other diffuser types have been designed. These
include designs that have variable areas so that the throat area can be increased to swallow the shock,
then decreased again to the design conditions (very similar to what was discussed for supersonic wind
tunnels).
variable area
actuators
movable plug
Oblique shock wave diffusers are also often used. The stagnation pressure loss across an oblique
shock is less than that across a normal shock wave. The weaker the oblique shock wave, the smaller
the stagnation pressure loss. Normal shock waves may still appear in the device but theyll be weaker
than if there wasnt an oblique shock since the oblique shock helps to decelerate the flow.
C. Wassgren
563
C. Wassgren
564
Note that engine inlets typically bleed off or remove boundary layers as shown in the F-16 engine
inlet design. This is done in order to avoid exposing engine components to the unsteady conditions
resulting from wakes formed by separated boundary layers.
C. Wassgren
565
11. Flows with Mass Addition
Now lets consider compressible flows where mass addition (or removal) occurs. Examples of such flows
include those in which solid rocket propellant is burned, a gas coolant is added to the flow such as in film
cooling of turbine blades, or the boundary layer flow is removed such as in some wind tunnels.
Well make the following assumptions for our analysis of flow with mass addition:
- no heat transfer
- no other work
- no friction
- no area change
- no significant elevation changes
- steady flow
- 1D flow (This isnt a very good assumption in general, but well make the assumption here for
simplicity.)
dm i , pi , i , Ti , Vix , Viy , si
m , p, , T , V , s
m dm , p dp, d , T dT , V dV , s ds
fluid is assumed to be
completely mixed
in this region
Using the same control volume approach as in previous analyses gives:

COM:
m dm m dm i 0
dm d VA dm i
d
dV dm
V
m
Note: dA=0
C. Wassgren
(12.205)
566
LME in x-direction:
d mV
mV
dm iVix pA pA d pA
mV
dm iVix dpA
d mV
Vdm mdV
dm iVix dpA
dp
V V Vix dm VmdV
0
VA
VA
dm
0
m
Note: y Vix/V, and dm dm i
dp VdV V 2 1 y
(12.206)
COE:
0 d mh
0 mh
0 dm i h0i 0
mh
dm
dh0 h0 h0i
0
m
dh0 0
(12.207)
Note: For simple mass addition we assume that the inlet fluid has the same composition
and stagnation enthalpy as the main stream, i.e. h0i = h0. In general, however, the
addition stream may have a different stagnation enthalpy than the main flow. The general
case will be considered when we investigate general 1D flows.
2nd Law:
d ms
ms
dm i si 0
ms
dm
ds s si
0
(12.208)
m
Note: The mass addition stream may have, in general, a different entropy than the main
stream. Moreover, mixing processes are generally irreversible.
Ideal Gas Law:
dp d dT
(12.209)
caloric eqn
dh c p dT
(12.210)
of state
When combined with COE (Eqn. (12.207)):

c p dT0 0 dT0 0
(12.211)
Mach # relation:
Gibbs Eqn:
d Ma
Ma
dV dT
V
2T
(12.212)
dp
dT
dp
R
R 0
T
p
p0
Note: dT0 = 0 from eqn (12.211).
ds c p
C. Wassgren
(12.213)
567
Local stagnation pressure and temperature:

1
T0 T 1
Ma 2
2
p0 p 1
Ma 2
2
(12.214)
(12.215)
Combining these equations so that dm is the driving potential, and using the substitution, y Vix/V, gives:
d Ma
Ma
1
2
1 2 Ma
2
2 dm
1 Ma y Ma
2
m
1 Ma
dp Ma 2
p 1 Ma 2
d
1
dm
Ma 2 1 y y
2 1
2
1 Ma 2 y Ma 2 dm
m
1 Ma
2
dT 1 Ma
dm
1 Ma 2 y Ma 2
2
T
m
1 Ma
(12.217)
(12.218)
(12.216)
(12.219)
dV
1
1 Ma 2 y Ma 2 dm
2
m
V
1 Ma
(12.220)
dp0
dm
Ma 2 1 y
p0
m
(12.221)
dT0 0
(12.222)
ds
dm
1 Ma 2 1 y
cp
m
(12.223)
C. Wassgren
568
Notes:
1. The trends in equations (12.216)-(12.223) will depend on dm , y, and Ma. Equation (12.221) shows
that for mass addition, the stagnation pressure will decrease if y < 1 and will increase if y > 1.
Equation (12.223) shows that entropy has the opposite trend.
2.
For y < 1, all factors that involve y are positive so that for mass addition ( dm 0 ) we have:
Ma < 1
Ma > 1
d(Ma) > 0
d(Ma) < 0
dp < 0
dp > 0
d < 0
d > 0
dT < 0
dT > 0
dV > 0
dV < 0
dp0 < 0
dp0 < 0
ds > 0
ds > 0
The opposite trends occur for mass removal ( dm 0 ). Note that choking is possible with mass
addition since the Mach number approaches unity for both subsonic and supersonic flows.
3.
We cannot simply add mass to transition from a subsonic to a supersonic flow. Equation (12.216)
indicates that as the flow approaches a sonic Mach number, the addition stream must have a large mass
flow rate to continue approaching the sonic Mach number. In the limit as Ma 1, dm .
C. Wassgren
569
4.
For y = 0 (i.e. the flow comes in normal to the stream), equations (12.216)-(12.223) become exact
differentials. Integrating using sonic conditions (denoted by the superscript *) as a reference gives:
m
Ma
*
m
1 Ma 2
p
p
Ma 2
2 +1 1
2
(12.224)
1
1 Ma 2
*
1 Ma 2
(12.225)
1
2
1 2 Ma
2 1
1
Ma 2
*
2
T
1
T
*
p0 1 Ma 2
p0
(12.226)
2 1
Ma
1
Ma 2
*
2
V
+1
V
5.
(12.227)
12
(12.228)
2 1
Ma 2
1
2
(12.229)
T0
1
T0*
(12.230)
p
s s*
ln 0*
R
p0
(12.231)
The relationship between T and s may be found by considering equations (12.227), (12.229), and
(12.231). Below is a T-s diagram for a flow with mass addition.
arrows are drawn for mass addition (y = 0)
T
at Ma = 0, T/T* = (+1)/2
T*
Ma < 1
Ma = 1
Ma > 1
as Ma , T/T* 0
C. Wassgren
s*
570
Example:
Air with initial stagnation conditions of 600 K and 1 MPa (abs) flows at a Mach number of 0.3 at the
entrance to a constant-area, porous-walled duct. During passage through the duct, the mass flow rate is
increased by 50%. Determine the exit Mach number, pressure, temperature, and stagnation pressure.
C. Wassgren
571
12. Generalized Steady, One-Dimensional Flow
In our previous analyses we have only considered simple flows where there is only one potential affecting
the flow (e.g. area change, friction, heat transfer, or mass addition). Now well discuss how to analyze
flows where multiple effects are considered. Our approach is very similar to the approach we have used
many times before: well draw a control volume, apply our conservation laws, utilize definitions, and
simplify our resulting equations.
Qinto
dm
D
z
x
m dm
Ff
gx
gz
In our analysis, well include the effects of:

- area change, dA
- friction, Ff
- heat transfer, Q
- mass addition, dm
- other work, W
- other forces, D
- gravity body force, gAdx
Well further assume that the flow is steady, one-dimensional, and that the fluid is well-mixed within the
control volume.
Now lets apply our conservation laws:
COM:
m dm m dm i 0
dm d VA dm i
dm d dV dA
V
A
C. Wassgren
(12.232)
572
LME (in x-direction):

d mV
mV
dm iVix
mV
pA pA d pA p 1 2 dp dA D Ff g x A 1 2 dA dx
mdV
dmV
dm iVix dpA D Ff g x Adx
dpA mdV
g x Adx Ff D dm V Vix 0
Substituting the following:

m VA , y Vix/V, and Ff w Pdx f F
V 2 Pdx 1 2 V 2 4 Af F DH :
4 f dx D
dm
V 2 1 y
0
dp VdV g x dx 1 2 V 2 F
m
DH A
(12.233)
COE:
0 d mh
0 mh
0 dm i h0i Q into CV Won CV
mh
dm
qinto CV won CV
(12.234)
m
The term in parentheses represents the difference in the stagnation enthalpy of the main stream and the
incoming flow. Note that: h0 h 12 V 2 g z z .
dh0 h0 h0i
2nd Law:
d ms
ms
dm i si
ms
CV
ds s si
dm
Q into CV
T
qinto CV
CV
(12.235)
Now lets specify that were dealing with an ideal gas:

p RT
dh c p dT
ds c p
Ma
(12.236)
(12.237)
dT
dp
R
T
p
(12.238)
(12.239)
RT
Since were concerned with a gas, well also assume that gravitational effects are negligible compared to
the other terms in the equations (i.e. g 0).
C. Wassgren
573
Well also utilize the definitions of the isentropic stagnation pressure and the adiabatic stagnation
temperature:
p 1
1
Ma 2
p0
2
T 1
1
Ma 2
T0
2
(12.240)
(12.241)
Equations (12.232)-(12.241) are a system of equations that can be combined and solved1 for the dependent
variables:
d Ma dp d dT dV dp0 ds
, ,
,
,
,
,
Ma
p T V p0 c p
in terms of the independent variables, or driving potentials:
dA 4 f F dx
2 D dT0 dm
,
,
,
A DH
Ma 2 pA T0 m
The following table summarizes the resulting equations.
For details, refer to Zucrow and Hoffman, Gas Dynamics: Volume I, Wiley.
C. Wassgren
574
Table 1. Change in flow properties in terms of driving potentials.
Driving Potentials
Change in
Flow
Property
d Ma
Ma
4 f F dx
2 D
2
DH Ma pA
dA
A
Ma 2
dp
Ma 2
1 Ma 2
Ma 2
1 Ma 2
dT
1 Ma 2
1 Ma 2
dV
V
dp0
1 Ma 2
2 1 Ma
1 Ma
ds
Ma 2
Ma 2
1 Ma
Ma 2
Ma 2
2 1 Ma 2
1 Ma 2
Ma 2
Ma 2 2 1 y y
1 Ma 2
1 Ma 2 y Ma 2
1 Ma 2
1 Ma 2
2 1 Ma 2
1 Ma 2
1 Ma 2
1 Ma 2 1 Ma 2 y Ma 2
1 Ma 2
1 Ma 2 y Ma 2
1 Ma 2
Ma 2 1 y
2
0
y Ma 2
1 Ma 2
1 Ma 2
cp
where
1 Ma
1 Ma 4
2 1 Ma 2
p0
1
Ma 2
2 1 Ma 2
dF
1 Ma 2 y Ma 2
2 1 Ma 2
T0
1 Ma
2 1 Ma
Ma 2 1 1 Ma 2
1
1 Ma
dm
m
dT0
1 Ma 2 1 y
1
y
1
2
Ma 2
Vix
F impulse function pA mV
C. Wassgren
575
Notes:
1.
The terms in the table are often referred to as influence coefficients.
2.
, is a convenient definition that is helpful in determining the

The impulse function, F pA mV
reaction force for one-dimensional steady flows. For example, the thrust on the jet engine shown
below can be determined from the difference between the outgoing and incoming impulse functions:
m 1V1
m 2V2
p1A1
p2A2
T
m 2V2 m 1V1 T p1 A1 p2 A2
T m 2V2 p2 A2 m 1V1 p1 A1
T F2 F1
3.
As can be seen from equations (12.234) and (12.237), the effects of heat transfer, other work, and the
difference in the main and incoming stream stagnation enthalpies are all included in the change in the
stagnation temperature, dT0.
4.
How one uses the table is best shown by example. Lets say were interested in determining how the
Mach number varies as a function of the driving potentials. From the table we see that the Mach
number variation is given by:
d Ma
dA
Ma 2 4 f F dx
2 D
2
2
Ma
1 Ma A 2 1 Ma DH Ma 2 pA
(12.242)
2
2
1 Ma 2 dT0 1 Ma y Ma dm
m
1 Ma 2
2 1 Ma 2 T0
We would now need to know how to model the driving potentials as we move downstream in flow
(e.g. How does the area vary as we move downstream in the duct?) Finally, we would integrate the
resulting equation (numerically if necessary) to solve for the Mach number variation.
C. Wassgren
576
5.
We can also extract trends from the influence coefficients. For example, equation (12.242) may be
written as:
dA Ma 2 4 f F dx
2 D
2
2 DH Ma pA
d Ma
A
2
Ma
1 Ma 1 Ma 2 dT0
dm
1 Ma 2 y Ma 2
m
T0
2
Note that is always positive.

a. If < 0, then the Mach number will decrease for subsonic flow and increase for supersonic flow,
i.e. the Mach number diverges from one.
b. If > 0, then the Mach number will increase toward one for subsonic Mach numbers while the
Mach number will decrease toward one for supersonic Mach numbers. Hence, choking is possible
if > 0.
c. If = 0, then the Mach number does not change since d(Ma) = 0 (the flow is at an inflection
point).
It is possible that a flow may have changes in the sign of as the flow moves downstream. For
example, simple isentropic flow in a converging-diverging nozzle has > 0 in the converging section,
= 0 at the throat, and < 0 in the diverging section.
Note that sonic conditions (Ma = 1) can only occur where = 0 otherwise d(Ma) would be infinite.
For example, consider the special case where only friction and area changes are present (D = dT0 =
dm = 0) so that:
dA Ma 2 4 f F dx
A
2 DH
Since the friction term will always be positive, and since sonic conditions must occur when = 0, the
sonic point must occur in a diverging section (dA > 0). The exact location of the sonic point can be
determined since we know how the area varies with x:
2
1 4 f F dx
1 dA
dx
0
Adx
2 DH
dA
A
1 dA 4 f F

(Since A(x), fF, and DH are given, one can solve for x.)
A dx 2 DH
C. Wassgren
577
6.
Note that many of the relations in Table 1 have a (1 Ma2) in the denominator. Hence, one must
proceed with care when integrating the relations near the sonic point since the gradients become very
large there. At the sonic point in particular, terms with (1 Ma2) in the denominator are undefined so
they can not be integrated directly. If the independent variables can be expressed explicitly as a
function of x, gradients at the sonic point can be resolved using lHopitals rule. For example, lets
again consider the case for a frictional flow with area change (D = dT0 = dm = 0) so that the Mach
number relation is given by:
d Ma
Ma 1 dA Ma 2 4 f F Ma x

2 DH
dx
1 Ma 2 A dx
1 Ma 2
d Ma
d Ma
d Ma
dx
lim
Ma
1
d Ma *
dx
dx
2Ma
xx
dx
*
Ma 1
d Ma
dx
1 d Ma

2
dx
x x*
Ma 1
1 d
4
d Ma
3 1
x x*
Ma
*
1
dx x x 4
dx
Ma 1
*
d Ma * 3 1
d Ma 1 d
*

xx
dx 4 Ma 1 dx
4 dx
where
4f
1 dA
(12.243)
(12.244)
F
xx
Ma 1
A dx x x 2 DH
*
x x*
Ma 1
d
dx
4 f d Ma
d 1 dA
F
dx A dx x x*
DH dx
x x*
Ma 1
(12.245)
Equations (12.243), (12.244), and (12.245) can be combined to give a single quadratic equation in
terms of the unknown sonic Mach number gradient. Two solutions can be found with the appropriate
one being determined by the downstream boundary conditions (Recall our previous discussions
concerning isentropic flow through a C-D nozzle. Whether the flow remains subsonic or supersonic
after the throat depends on the back pressure.)
7.
The previous generalized flow table (Table 1) can be simplified to give our previous results for simple
flows as shown in the following table.
C. Wassgren
578
Simple Area Change

A
1 2
T0
1 Ma
Ma
DH
m *
2
1

2 1 Ma 2
1 Ma
1
1 2 2
ln Ma

2
1
2
1

T
*
p0
Ma 1
1
1
1 2
Ma 2 1
F
*
2
1

1 2
Ma 1
1 Ma
1
2
Ma

1
Ma 2
2
Ma 2 1
Ma 2 1
cp
1 Ma
1 Ma
1 2
ln Ma

1
2
1 Ma 1
1
2
ln Ma
2
1 Ma
2
Ma

1
1
1
2 1
1 Ma
1 Ma
1 Ma 2
1 Ma
1 Ma 2
1 Ma
1 Ma 2
*
Ma 1
p0*
2
Ma

1
1 2
2
1

ss
1 Ma
4 fF L
2 1
*
T0
Simple Mass Addition, y 0
Ma 1
Simple Diabatic
Simple Frictional Flow
2
1 Ma 1
1
1
1
2 1 1 Ma 2
ln

1 1
Table 2. Change in flow properties for simple flows.
C. Wassgren
579
Example:
Consider steady air flow through a constant area circular duct which has a diameter of 10 cm and a length
of 1 m. The Mach number, pressure, and temperature at the inlet to the duct are 0.3, 200 kPa, and 80 C.
Heat is added at a uniform rate to the air as it flows through the duct causing the stagnation temperature to
increase by 300 K. If the (Fanning) friction factor can be assumed to be 0.003, find the Mach number,
pressure, and temperature at the outlet of the duct.
C. Wassgren
580
13. Oblique Shock Waves
An oblique shock wave is a shock wave that forms at an angle with respect to the incoming flow (note that
normal shock waves are a special case of an oblique shock wave). As with normal shock waves, the flow
properties abruptly change when passing through an oblique shock. In addition, the flow is turned through
the shock as shown in the figure below.
p1, 1, T1
V2
V1
p2, 2, T2
Well refer to as the shock angle (the angle of the shock with respect to the upstream velocity) and as
the flow turning (or deflection) angle.
The working equations for oblique shock waves can be determined by analyzing a very thin control volume
that straddles the shock wave (an approach identical to what we used to analyze normal shock waves).
Well also break the upstream and downstream velocities into components that are normal and tangential to
the shock wave.
p1, 1, T1
V1
VN1
VT2
VT1
VN2
V2
p2, 2, T2
COM:
1VN 1 2VN 2
(12.246)
LME:
N-dir: 2VN2 2 1VN21 p1 p2
T 2 mV
T 1 0 VT 2 VT 1 VT
T-dir:
mV
(The tangential velocity components on either side of the wave are equal!)
COE:
h1 1 2 V12 h2 1 2 V22
C. Wassgren
h1 1 2 VN21 h2 1 2 VN2 2 (since VT1 = VT2)
581
(12.247)
(12.248)
(12.249)
2nd Law:
s2 s1 (The large gradients within a shock wave result in an irreversible process.)
(12.250)
Equations of state for a perfect gas:

p RT
h c pT
(12.251)
(12.252)
For an ideal gas:

V
Ma
RT
(12.253)
p 1
Ma 2
1
p0
2
T 1
Ma 2
1
T0
2
(12.254)
(12.255)
From geometry:
V12 VN21 VT2
(12.256)
V22 VN2 2 VT2

VN 1 V1 sin
VN 2 V2 sin
(12.257)
VT V1 cos V2 cos
C. Wassgren
582
Combining the previous equations and, after much simplifying, we have:

Ma N 1 Ma1 sin
Ma N 2 Ma 2 sin
Ma 2N 2
Ma 2N 1
2 Ma 2N 1
(12.258)
(12.259)
(12.260)
p2
2
1
Ma 2N 1
p1 1
1
(12.261)
1 Ma 2N1
2 VN 1
tan
1 VN 2 tan 2 1 Ma 2N 1
(12.262)
2 Ma 2N 1 1
T2
2 1 Ma 2N 1
2
T1
1 Ma N1
(12.263)
V2
sin
2
1
V1 sin 1 Ma 2N 1 1
(12.264)
2
p02 1 Ma N 1
p01 2 1 Ma 2N 1
2
2 Ma N 1 1
(12.265)
T02
1
T01
(12.266)
tan
2 1
1
2
tan 1 Ma N 2
2
(12.267)
1 Ma12
1 tan
2
tan 2 Ma N 1 1
(12.268)
C. Wassgren
583
Notes:
1. Equations (12.260)-(12.266) are the same relations as for a normal shock wave except that the normal
component of the velocity is used. Thus, we can use normal shock relations to determine flow
properties across an oblique shock as long as we use the normal component of the velocity.
2.
There are two free parameters in the equations (12.260) - (12.268). These typically are the incoming
Mach number, Ma1, and the flow turning angle, . Note that the flow is always turned toward the
shock wave.
3.
Analysis using the 2nd Law (refer to the notes on normal shock waves) states that an oblique shock will
only form if the incoming normal Mach number is greater than or equal to one, i.e. MaN1 1. The
downstream normal Mach number will be less than one, i.e. MaN2 1. Note that the total downstream
Mach number, Ma2, may not be less than one due to the tangential velocity component.
4.
Based on the previous note, the minimum value for (the angle of the shock to the incoming flow) is
sin-1(1/Ma1) which corresponds to a Mach wave (i.e. a sound wave). The maximum value for is 90
corresponding to a normal shock wave. For both of these shock angle limits we find (from equation
(12.268)) that the turning angle of the flow is = 0. Note that will have a maximum value since it is
zero for min = sin-1(1/Ma1), is positive for larger values of , then returns to zero for max = 90. We
will discuss this maximum value for in a later note.
C. Wassgren
584
5.
Since the equations relating the incoming Mach number, Ma1, the wave angle, , and the turning angle,
, are complicated, its often more instructive and useful to present the data in graphical form rather
than in equation form.
The following figure plots the wave angle, , as a function of the incoming Mach number, Ma1, for
different values of the turning angle, , for =1.4 using Eqn. (12.268) [plot from Zucrow and Hoffman, Gas
Dynamics: Vol. I, Wiley]. The lines corresponding to the maximum turning angle, max, and where the
downstream flow is sonic, Ma2 = 1, are also shown in the figure. Note that for a given Ma1, there are two
values for , the larger corresponding to a strong shock with Ma2 < 1 and the smaller corresponding to a
weak shock with Ma2 > 1. Well discuss these observations in greater detail in a subsequent note.
strong shocks (Ma2 < 1)

maximum turning
angle, max
weak shocks (Ma2 < 1)
weak shocks (Ma2 > 1)
C. Wassgren
585
The following figure plots the downstream Mach number, Ma2, as a function of the incoming Mach
number, Ma1, for different values of the turning angle, , for =1.4 using equations (12.267) and (12.268)
[plot from Zucrow and Hoffman, Gas Dynamics: Vol. I, Wiley]. The line corresponding to the maximum
turning angle, max, is also shown in the figure.
weak shocks
(Ma2 > 1)
strong shocks
(Ma2 < 1)
C. Wassgren
586
6.
The plots shown in Note 5 indicate that there are two possible wave angles, , corresponding to a
particular incoming Mach number, Ma1, and turning angle, . The smaller value of corresponds to a
weak oblique shock and the larger value corresponds to a strong oblique shock. Flow downstream
of the strong shock is always subsonic while the flow downstream of the weak shock is usually
supersonic, except for a region where is close to max.
Ma1
Ma2,strong < 1
Ma2,weak usually > 1
strong
weak
Both types of shocks occur in practice with the weak shock being more prevalent. If there is a
blockage or a high-pressure condition downstream of the shock, the strong shock solution will
typically occur (e.g. at the inlet of a supersonic jet engine diffuser where the internal flow is at a high
pressure). For flows occurring in the atmosphere where the pressure far downstream of the deflection
can only be infinitesimally different from the pressure far upstream of the deflection, the weak shock
will occur (e.g. supersonic flow over the surface of an aircraft). To further complicate matters, since
the governing equations of the fluid motion are nonlinear, it is possible to have multiple, stable flow
solutions. Which solution is observed will depend on the path taken to get to the solution. The
hysteresis associated with the starting/overspeeding of a supersonic jet engine diffuser is a good
example of a flow situation having multiple, stable solutions (e.g. operating at the design speed with a
shock in front of the inlet, or after overspeeding where the shock has been swallowed). The solution
that occurs in this flow situation depends on the path taken to get to that solution. If no other
information is available, it is generally reasonable to assume that the weak shock occurs.
7.
The figures in Note 5 also indicate that there is a maximum turning angle, max, that can be achieved
through an oblique shock (this maximum turning angle separates the weak and strong shock solutions).
It can be shown that the maximum wave angle corresponding to the maximum turning angle for a
given upstream Mach number is (refer to Ferri, A., Elements of Aerodynamics of Supersonic Flow,
Macmillan, NY):
1 1
1 4
1
sin 2 max
Ma12 1 1 1
Ma12
Ma1
(12.269)
2
2
16
Ma1 4

The corresponding maximum turning angle may be found using Eqns. (12.258) and (12.268).
C. Wassgren
587
The following figure plots the maximum turning angle, max, and the wave angle corresponding to this
maximum turning angle, max, as a function of the incoming Mach number, Ma1 [plot from Zucrow and
Hoffman, Gas Dynamics: Vol. I, Wiley].
C. Wassgren
588
What happens if the flow is deflected by an angle larger than max? A curved, detached shock wave
occurs with a strong shock below the sonic line and a weak shock above sonic line. Analysis of curved
shock waves is very difficult due to the existence of subsonic, transonic, and supersonic flow, each of
which has very different governing differential equations. In addition, the flow downstream of the
shock system will have curved streamlines and be irrotational (well discuss this in a different set of
notes when investigating Croccos Theorem).
weak
Ma2 > 1
Ma1 > 1
Ma2 = 1
strong
Ma2 < 1
>max
Note that max increases with increasing upstream Mach number. Hence, its possible that a flow with
a detached, curved shock at a low Mach number may produce an attached, oblique shock at larger
Mach numbers.
Ma1 > 1
C. Wassgren
> max
Ma1 > Ma1
589
< max
Example:
A uniform supersonic air flow traveling at Mach 2.0 passes over a wedge. An oblique shock, making an
angle of 40 with the flow direction, is attached to the wedge for these flow conditions. If the static
pressure and temperature in the uniform flow are 5 psia and 0 F, determine the static pressure and
temperature behind the wave, the Mach number of the flow passing over the wedge, and the wedge half
angle.
C. Wassgren
590
Example:
Air having an initial Mach number of 3.0, a free stream static pressure of 101.3 kPa, and a free stream static
temperature of 300 K is deflected by a wedge by an angle of 15. Assuming that a weak oblique shock
occurs, calculate:
a. the downstream static pressure,
b. the downstream density,
c. the downstream temperature,
d. the change in the stagnation pressure through the shock,
e. the downstream velocity, and
f. the downstream Mach number.
C. Wassgren
591
Example:
Consider a compression corner with a deflection angle of 28. How does the pressure ratio across the
oblique shock change if the incoming Mach number is doubled from 3 to 6?
Ma1, p1
Ma2, p2
28
C. Wassgren
592
Example:
An aircraft is to cruise at a Mach number of 3. The stagnation pressure in the flow ahead of the aircraft is
400 kPa. Compare the stagnation pressure recovery ratio for three possible intake scenarios.
a. An intake that involves a normal shock in the free stream ahead of the intake followed by an isentropic
deceleration of the subsonic flow behind the shock wave to an essentially zero velocity.
b. An oblique shock wave diffuser in which the air flows through an oblique shock wave and then an
isentropic deceleration of the subsonic flow behind a normal shock wave to an essentially zero
velocity.
c. An ideal shockless convergent-divergent diffuser in which the air is isentropically brought to an
essentially zero velocity.
Ma=3
p=400 kPa
isentropic
deceleration
15
Ma=3
p=400 kPa
Ma=3
p=400 kPa
C. Wassgren
isentropic
deceleration
isentropic
deceleration
593
14. Expansion Waves
Recall the piston experiment in our previous discussion regarding the formation of shock waves. Now lets
consider what happens when we move the piston toward the left (as shown in the figure below) so that an
expansion (or rarefaction) wave propagates down the length of the tube.
dV
dV
stagnant gas
rarefaction or
expansion wave
When we first move the piston, an infinitesimal strength pressure wave travels down the cylinder at the
sonic speed. Behind the wave the pressure, temperature, and density decrease (refer to a previous set of
notes concerning property changes across a sound wave). In addition, the flow velocity behind the wave
will move in the direction of the piston (away from the wave).
If we continue to increase the piston velocity, additional pressure waves will propagate down the cylinder.
However, these waves travel at a slightly lower speed relative to a fixed observer due to the decreased fluid
temperature and leftward fluid velocity behind each wave. Hence, the waves start to spread out. This is the
opposite of what occurred for compression waves.
1st wave
2nd wave
2dV
2dV
c2
dV
T1
c1
V=0
T0
c2 < c1 since T1 < T0 and dV moving to left

t, time
piston
pressure wave
(sound wave)
x, position
Notes:
1. Since the waves do not coalesce, the change in the properties across each wave are infinitesimal.
Hence, the flow through each wave is considered isentropic.
2. It is impossible to form an expansion shock wave since each subsequent wave travels slower than the
previous wave. The waves will never coalesce. This can also be proven mathematically by showing
that entropy would decrease across an expansion shock.
C. Wassgren
594
Prandtl-Meyer Expansion Fans
Now lets consider expanding a steady supersonic flow around a gradual, outward-turning corner as shown
in the figure below.
Ma1
Mach waves
2< 1
Ma2 > Ma1
The gradual curve can be approximated by a series of very small, discrete turns, each of which results in a
small expansion Mach wave as shown in the previous figure. Recall that Mach waves exist only for
supersonic flows, are infinitesimally weak pressure waves, and are inclined at an angle (known as the Mach
angle), = sin-1(1/Ma), with respect to the flow. Note that across an expansion wave the Mach number
will increase (refer to a previous set of notes concerning sound waves) so that successive expansion waves
will have smaller Mach angles (Ma2 > Ma1 2 < 1 in the figure above). As a result, the waves diverge,
remain infinitesimally weak, and thus the flow across the waves is isentropic. This type of expansion is
sometimes referred to as a non-centered expansion fan. For a sharp corner, the waves comprising the
expansion fan start at the corner point then diverge outward as shown in the figure below. This type of fan
is known as a centered expansion fan.
oblique shock
centered expansion fan
Notes:
1. The flow into and out of each Mach wave is uniform (i.e. 1D).
2. It is also possible to have isentropic compression waves resulting from gradual (infinitesimal), noncentered turns as shown in the figure below. Non-centered compression fans, however, will eventually
converge to form oblique shock waves which are non-isentropic. A centered compression fan resulting
from a finite angle corner is an oblique shock wave and thus is non-isentropic.
non-isentropic
oblique shock
isentropic
compression region
gradual compression turn

3.
A compression wave turns the flow toward the wave while an expansion wave turns the flow away
from the wave.
C. Wassgren
595
Now lets analyze the steady, 2D flow through a single Mach wave as shown in the figure below. Our
analysis will be for a compression Mach wave (d > 0 in the figure below) but the analysis will also hold
for an expansion Mach wave (d < 0). Note that the upstream Mach number must be supersonic in order
for Mach waves to exist.
p, T,
VT + dVT
VN
VN + dVN
VT
p+dp, T+dT, +d
V+dV
Note: d is defined as being positive for

counter-clockwise (i.e. compression)
turns. For an expansion, d < 0.
From geometry:
VT V cos
(12.270)
VT dVT V dV cos d
(12.271)
From the LME in the tangential direction:

T 0 dVT 0
m VT dVT mV
(12.272)
Combining equations (12.270)-(12.272):

VT V cos V dV cos d
V cos V dV cos cos d sin sin d
Since the turning angle, d, is very small we can write:

V cos V dV cos d sin
V cos Vd sin dV cos dVd sin
Neglecting H.O.T.s and simplifying:
dV
d tan
V
(12.273)
Since the expansion wave is a Mach wave (with wave angle, sin = 1/Ma), we have:
1
tan
Ma
Ma 2 1
1
(1-Ma2)1/2
d
dV
(Note: Since Ma > 1, d > 0 dV < 0 and d < 0 dV > 0.)

V
Ma 2 1
C. Wassgren
596
(12.274)
For a perfect gas:

V 2 RT Ma 2
1
RT0 Ma 1
Ma 2
2
1
dV 1
d Ma
Ma 2
1
V
2
Ma
(12.275)
Substituting (12.275) into equation (12.274) gives:

Ma 2 1 d Ma
d
1 2 Ma
1
Ma
2
Note: Since Ma > 1, d > 0 d(Ma) < 0 and d < 0 d(Ma) > 0.
Integrating equation (12.276) gives:
1 1 1
tan
Ma 2 1 tan 1 Ma 2 1 constant

1
1
(12.276)
(12.277)
The constant of integration can be determined if the initial Mach number, Ma1, and flow deflection angle,
1, are known. For convenience, we define a reference state where, Ma = 1, and = 0. For these
conditions, equation (12.277) becomes:
1 1 1
tan
Ma 2 1 tan 1 Ma 2 1

Prandtl-Meyer Angle
(12.278)
1
1
Note: The symbol has been changed to a in order to signify that this is the Prandtl-Meyer
angle. The Prandtl-Meyer angle is the angle, , that the flow needs to be turned to go from sonic
conditions to get to the new Mach number, Ma.
C. Wassgren
597
Notes:
1. The angle, , is positive for counter-clockwise (compressive) rotations and negative for clockwise
(expansive) rotations. The convention, however, is to drop the negative sign when reporting PrandtlMeyer angles.
The Prandtl-Meyer angle is plotted as a function of Mach number for = 1.4 in the figure below.
Prandtl-Meyer angle, | [deg]
2.
120
= 1.4
100
80
60
40
20
0
0
Mach number, Ma
3.
For an arbitrary incoming Mach number (Ma1 > 1, 1 = fcn(Ma1)) we can imagine that there is some
imaginary upstream corner that expands the flow from Ma0 = 1 (0 = 0) to the current Mach number,
Ma1 (1 = fcn(Ma1)), as shown in the figure below.
Ma1 > 1
1 0
imaginary
expansion
Ma0 = 1
0 = 0
To expand a flow from Ma1 to Ma2 (Ma2 > Ma1), we need to turn the flow by an angle of 2-1. For
example, to get from sonic conditions (Ma0 = 1) to Ma1 = 2.0, we need to turn the flow by an angle of
1-0 = -26.38 - 0 = -26.38 and to get the flow from sonic conditions to Ma2 = 3.0 we need to turn the
flow by an angle of 2-0 = -49.76 - 0 = -49.76. To go from Ma1 = 2.0 to Ma2 = 3.0, we need to turn
the flow by an angle of 2-1 = -49.76 - (-26.38) = -23.38. Note that the negative signs indicate that
we need to turn the flow away from the Mach wave, i.e. expand the flow.
4.
Compressive Mach waves turn the flow toward the Mach wave while expansive Mach waves turn the
flow away from the Mach wave.
d
expansive Mach wave
compressive Mach wave

C. Wassgren
598
5.
Since the expansion process is isentropic, the isentropic flow relations can be used throughout the
expansion fan. The isentropic relations may also be used in a limited region within a non-centered
compression fan where the Mach waves do not intersect. An oblique shock forms where the
compression Mach waves merge.
6.
To expand the flow of air (= 1.4) from sonic conditions (Ma = 1) to an infinite Mach number, i.e. Ma
, we must turn the flow by max = -130.4.
1
1
(12.279)
max lim
Ma
2 1
If the corner has an angle greater than this maximum angle, then a vacuum region forms. Note that the
continuum assumption would break down in the region adjacent to the vacuum.
Ma = 1
130.4
Ma = 1
> max
vacuum
Ma
C. Wassgren
Ma
599
Example:
A uniform supersonic flow at Mach 2.0, static pressure 10 psia, and temperature 400 R expands around a
10 corner. Determine the downstream Mach number, pressure, temperature, and the fan angle.
fan angle
10
C. Wassgren
600
Example:
A wind tunnel nozzle is designed to yield a parallel uniform flow of air with a Mach number of 3.0. The
stagnation pressure of the air supply reservoir is 7000 kPa, and the nozzle exhausts into the atmosphere
(100 kPa). Calculate the flow angle at the exit lip of the nozzle.
Ma=3.0
7000 kPa
100 kPa
angle = ?
C. Wassgren
601
Example:
Calculate the lift and drag coefficients (per unit depth into the page) for a flat-plate airfoil with a chord
length of 1 m. The plate is at an angle of attack of 6 degree with respect to the incoming flow which has a
Mach number of 2.5. Clearly sketch the wave patterns at both the leading and trailing edges of the airfoils.
Note that the lift and drag coefficients are based on the planform area of the airfoil.
1m
Ma=2.5
6
C. Wassgren
602
15. Reflection and Interaction of Oblique Shock Waves
In this set of notes well consider the interaction between oblique shocks and solid boundaries, free
surfaces, and with other oblique shocks.
Reflection with a Solid Boundary
When an oblique shock intersects a solid, straight boundary, it will reflect as another oblique shock in order
for the flow to remain parallel to the boundary as shown in the figure below.
Ma2, p2
1
Ma3, p3
Ma1, p1
1
wall
Image from: Shapiro, A.H., The Dynamics and
Thermodynamics of Compressible Fluid Flow Vol. I, Wiley.
The flow properties through the incident and reflected shocks can be determined using the following
procedure.
1. For the given Ma1 and , determine Ma2 and p2/p1.
2. For this value of Ma2 and since the turning angle of the second wave is also (in order to keep the
flow parallel to the flat wall), one can determine Ma3,2, and p3/p2.
3. The pressure ratio across both waves, p3/p1, is found using:
p3 p3 p2
p1 p2 p1
4. The angle that the reflected wave makes with the wall is 2-.
Notes:
1. The wave angle, , and the turning angle, , are measured with respect to the incoming flow direction.
2.
Without viscous effects the pressure change is discontinuous across the shock at the wall. Due to the
boundary layer, the flow drops to zero velocity at the wall and so the flow adjacent to the wall is
subsonic and thus cannot sustain pressure discontinuities. Thus the boundary layer causes the pressure
distribution to spread out as shown in the figure below.
1
p3
p1
without BL
p3
p1
with BL
The details of the shock interaction with the wall depend on whether the boundary layer is laminar or
turbulent, the thickness of the boundary layer, and the shock strength. A flow separation bubble may
also occur due to an adverse pressure gradient.
C. Wassgren
603
3.
Recall that there is a maximum possible angle through which the flow can be turned using an oblique
shock. This maximum angle (max) decreases as the incoming Mach number decreases. Thus, when the
flow passes through the initial oblique shock, there exists the possibility that the reflected flow must be
turned by an angle that is greater than the maximum turning angle in order to remain parallel to the
wall. In this case, a Mach reflection appears where a curved strong shock occurs adjacent to the wall
behind which there is subsonic flow. Since this subsonic flow need not be parallel to the wall, the flow
above the wall shock layer also does not have to be parallel to the wall. The flow behind the curved
wall shock is divided from the flow behind the reflected oblique shock by a slipline (aka contact
surface, slipstream, vortex sheet). Across the slipline there are changes in velocity, temperature, and
entropy. Since the flow is subsonic, the details of the flow may be affected by downstream conditions
and, as a result, the analytical modeling of the Mach reflection is difficult.
incident
shock
reflected
incident
shock
shock
Ma > 1
flow is not parallel to the wall
Ma < 1
slipline
reflected
shock
flow direction

4.
Instead of a flat wall, consider the reflection from a wall that is turned at some angle. The flow must
still remain parallel to the downstream wall. The oblique shock reflected from a wall turned away
from the flow will be weaker than one reflected from a flat wall. It is even possible to cancel the
reflected oblique shock by turning the downstream shock at the angle through which the flow is turned
by the initial shock. Turning the wall by an angle greater than this results in an expansion fan.
C. Wassgren
604
Example:
Air flowing with a Mach number of 2.5 with a pressure of 60 kPa (abs) and a temperature of 253 K passes
over a wedge which turns the flow through an angle of 4 leading to the generation of an oblique shock
wave.
a. If this oblique shock wave impinges on a flat wall, which is parallel to the flow upstream of the wedge,
determine the pressure and Mach number behind the reflected shock wave.
b. If the wedge is 1 m long, what is the minimum height above the wall, hmin, that the wedge must be in
order to not intersect the reflected shock.
c. If the reflected shock did intersect with the wedge, what type of shock pattern would appear?
d. How could we avoid producing the reflected shock?
1m
4
Ma1=2.5
p1=60 kPa (abs)
hmin
T1=253 K
C. Wassgren
605
Oblique Shock Reflection from a Free Surface

When an oblique shock intersects a free surface, the reflection must be an expansion fan so that the flow
pressure remains equal to the free surface pressure as shown in the figure below.
free surface
p0
free surface
p0
p3 = p0
p1 = p0
Ma3
Ma1
oblique shock
C. Wassgren
Ma2
p2 > p0
606
expansion fan
Example:
A symmetric converging-diverging nozzle is designed for an exit Mach number of 2.0. With the nozzle
exhausting to a back pressure of 15 psia, however, and a reservoir pressure of 78.2 psia, the nozzle is
overexpanded as is shown in the figure below.
pb= 15 psia
R1
p0=78.2 psia
a.
b.
R2
R3
Determine the Mach number in regions R1, R2, and R3.

Determine the angle of the flow with respect to the horizontal (in degrees) in regions R1, R2, and R3.
C. Wassgren
607
Interaction of Oblique Shock Waves

When two oblique shock waves intersect, each wave will be transmitted through, but be affected by, the
other wave. For example, consider the interaction of two oblique shocks as shown in the following
diagram.
slipline
The flow in regions 2 and 3 are found given the conditions in region 1 and the wall turning angles. The
flow in regions 4 and 5 must be parallel to one another and, hence, from the linear momentum equation the
pressures in regions 4 and 5 must be the same. An iterative procedure can be used to solve for the flows
from regions 2 and 3 into regions 4 and 5. A slipline occurs between regions 4 and 5 since these regions, in
general, have different entropy, temperature, and velocity.
Notes:
1. There may be oblique shock intersection situations where no solution exists using oblique shocks. In
this case, more complicated flow patterns may occur that include normal shock waves (Mach
reflections).
slipline
Ma < 1
slipline

C. Wassgren
608
Example:
Air is flowing down a wide channel with a Mach number of 3, pressure of 30 kPa, and a temperature of 263
K. The upper wall of this channel turns through an angle of 4 towards the flow while the lower wall turns
through an angle of 3 towards the flow leading to the generation of two oblique shock waves which
intersect each other. Find the pressure and flow direction downstream of the shock intersection.
Ma1=3
p1=30 kPa
T1=263 K
C. Wassgren
609
Now consider flow in a corner with several small, discrete changes in the wall angle as shown in the figure
below.
sliplines
weak reflected waves required in
order to equalize the pressure on both
sides of the slipstream
(can be compression or expansion
depending on the flow situation)
stronger oblique shock
weak oblique shocks
The oblique shocks will intersect and coalesce into a single oblique shock which is stronger than any of the
initial oblique shock waves. Sliplines and weak reflected waves appear at the intersection of the waves.
Notes:
1. An oblique shock is generally considered to form when two or more compression waves coalesce.
2.
For a continuously inward curving wall, the slipstreams are spaced an infinitesimal distance apart so
the downstream flow has continuously changing entropy, velocity, and temperature. The oblique
shock that forms from the interacting compression waves will have a strength such that it turns the
flow by the wall angle, , for the given incoming Mach number.
Region of continuously varying
velocity, entropy and temperature
in the direction perpendicular to
the wall.
3.
When waves travel in the same direction as viewed by an observer oriented so that theyre looking
downstream, the waves are said to be of the same family. The waves are further classified as being
either left-running or right-running depending on what direction the waves are oriented. Examples
of this notation are given below.
flow direction
observer looking
downstream
left-running
compression wave
right-running
expansion wave
These waves are of different families since one

is left-running and the other is right running.
C. Wassgren
610
All of the compression waves shown in the

figure above are of the same family since they
are all left-running waves.
16. Reflection and Interaction of Expansion Waves
Reflection with a Solid Boundary

When an expansion wave intersects a solid boundary, it will reflect as another expansion fan (of opposite
family) in order for the flow to remain parallel to the boundary as shown in the figure below.
Ma3, p3
Ma2, p2
Ma1, p1
non-simple region
The flow through the incident and reflected expansion fans can be determined using the following
procedure.
1. For the given (Ma1, 1) and 12, determine (Ma2, 2) using 2 = 1 + .
2. For this value of (Ma2, 2) and since the turning angle of the second wave is also (in order to
keep the flow parallel to the flat wall), one can determine (Ma3,: 3 = 2+ .
3. The pressure ratio across both waves, p3/p1, is found using the isentropic relations:
p2 1
1
Ma 22
p01
2
p3 1
1
Ma 32
p01
2
Notes:
1. The flow through the expansion fans is isentropic.
2.
The region where the expansion fans interact is known as a non-simple region. This region is not
amenable to our Prandtl-Meyer expansion fan analysis. Instead, we use can the Method of
Characteristics (a topic that will be discussed in a later set of notes) to analyze the flow in this region.
3.
An expansion wave may be canceled by turning the flow inward by the same amount as the flow
turning angle.
C. Wassgren
611
Expansion Fan Reflection from a Free Surface

When an expansion fan intersects a free surface, the reflection must be an oblique shock (of opposite
family) so that the flow pressure remains equal to the free surface pressure as shown in the figure below.
non-simple region
p0
free surface
p1 = p0
p0
Ma1
expansion fan
Ma3
p2 < p0
free surface
p3 = p0
Ma2
oblique shock
Interaction of Expansion Waves

When expansion fans intersect, each expansion wave will be transmitted through, but be affected by, the
other waves. There can be many regions of non-simple flow as shown in the figure below.
Notes:
1.
No slipstreams occur with expansion fans since the flow properties change continuously through the
fan.
C. Wassgren
612
Mach Diamonds
A phenomenon known as Mach Diamonds (aka shock diamonds) can form when a supersonic stream exits
a device and interacts with the surrounding atmosphere. Consider supersonic flow from an exit, with
pressure pe, into the surrounding atmosphere with pressure, pb.
Over-Expanded Case
pe < pb
Under-Expanded Case
pe > pb
Notes:
1. At design conditions, pe = pb, so additional expansion or compression of the flow is not required.
Hence, Mach diamonds do not appear at design conditions.
2. If the flow turning angle is sufficiently large, Mach reflections (aka Mach disks in 3D) will appear as
shown in the figure below.
pe < pb
3.
4.
5.
The bright regions in the Mach diamond shown in the photograph above are caused by heating of the
gas as it passes through the oblique shocks.
Viscous interaction with the external fluid results in dissipation of the Mach diamond pattern.
Even without viscous effects, the sequence of oblique shocks and expansion fans will eventually
dissipate since the stagnation pressure decreases after each shock. This can be shown by considering
the Mach number in the regions downstream of an oblique shock (referred to by the subscript i) which
are in contact with the free surface such that:
pb 1
1
1
Ma i2
2
p0i
Since p0i decreases after passing through an oblique shock and pb is a constant, the ratio pb/p0i increases
and Mai decreases. Thus, the sequence of oblique shocks gets weaker.
C. Wassgren
613
Interaction Between An Oblique Shock and an Expansion Fan

When an oblique shock interacts with an expansion fan of the same family, the shock will be weakened and
become curved. Behind the shock wave the flow becomes rotational. In addition, the incident expansion
waves will be reflected.
straight shock
curved, weakened shock
region of rotational flow
straight shock
slip line
Notes:
1. Waves will also be reflected from where waves intersect the slip line. These waves were left off the
previous schematic for the sake of clarity.
2. The region of rotational flow is not isentropic (the entropy varies continuously behind this curved
section of shock wave). This will be discussed in a later set of notes concerning Croccos Theorem.
C. Wassgren
614
17. Equations of Motion in Terms of the Velocity Potential
Now lets consider the irrotational, isentropic flow of a compressible fluid where body forces are
negligible. Recall that the momentum equations for a fluid in which viscous and body forces are negligible
are given by Eulers Equations:
u
u u p
t
We can re-write the pressure gradient term in terms of the speed of sound, c, as is shown below:
dp
dp
d
p dx dp
dx
d
d
dp
p
d
but since the flow is isentropic, we have:
dp
dp
p

c 2
d
d s
Thus, the momentum equations can be written as:
u
u u c 2
t
Now take the dot product of this equation with the velocity and re-arrange:
u
c2
u u u u
u
t
1
c2
u u u u u u
2 t
The continuity equation can be used to re-write the RHS of the previous equation:
u
u u 0
t
t
u
u
t
Thus, the momentum equations become:
1
c 2
c2 u
(12.280)
u u u u u
t
2 t
The density in the previous equation may be eliminated using Bernoullis equation. Since the flow is
irrotational, we can write Bernoullis equation as:
dp 1
2 F t
t
where the velocity has been written in terms of a velocity potential, .
C. Wassgren
615
Taking the time derivative of Bernoullis equation gives:

2 dp 1
u u F ' t
t 2 t 2 t
But,
dp dp d
d
c2
t t d
t
so that:
2
t

c2
d
c 2
t
c 2 1
u u F ' t
t 2 t
2 1
c 2
'
F
t
u u
t
t 2 2 t

1
2 1
u u u u u F ' t 2
u u c 2 u
2 t
2 t
t
u
1 2
u u u u u F ' t
2
2
t
c t
u u F t u u u
2
c t t
Re-writing the velocities in terms of the potential function gives:
2 2 F t
c t t
(12.281)
Governing equation for the isentropic, irrotational flow of a compressible fluid where body forces
are negligible
C. Wassgren
616
Notes:
1. Consider the flow of an incompressible fluid. If the flow is incompressible, then the speed of sound in
the fluid will be infinite. Thus, the governing equation for the irrotational flow of an incompressible
fluid becomes:
2 0
(12.282)
Governing equation for the irrotational flow of an incompressible fluid in which body forces
are negligible
Note that this is just Laplaces equation! A significant point regarding Eqn. (12.282) is that it is a
linear PDE which means that the principle of superposition may be used to add together building
block solutions to form new and more complex solutions.
2.
For a steady, compressible flow, Eqn. (12.281) simplifies to:

1
2 2
(12.283)
c
Governing equation for the steady, isentropic, irrotational flow of a compressible fluid in
which body forces are negligible
3.
Equations (12.281) and (12.283) are non-linear PDEs which are complex to solve by hand. There is
currently no known method for analytically solving these equations in a general way (computational
techniques can be used, however). Instead, we must resort to special cases for solving these equations.
One of these methods, known as small-perturbation theory, is described in the following section.
C. Wassgren
617
18. Small Perturbation Theory
Recall that the equation of motion for an irrotational flow where body and viscous forces are negligible is:
2 2 F t
(12.284)
c t t
where the velocity is given in terms of a potential function, u=, and c is the speed of sound. For a
steady flow, this equation simplifies to:
1
2 2
(12.285)
c
Suppose that a uniform flow approaches an object that is sufficiently slender so that it produces only a
small perturbation to the incoming stream as shown in the figure below.
The potential function for such a flow can be written as:
U x
(12.286)
U
where U is the velocity of the incoming flow and is the
y
potential function for the velocity perturbations, u, i.e.:
u '
x
(Note: U x U e x
)
u '

1
2 2 U e x U e x U e x
(12.287)
c
The speed of sound, c, should also be written in terms of a velocity perturbation. This is accomplished by
recalling that for a steady, isentropic flow, the stagnation enthalpy will remain constant:
h0 h0, h 1 2 U 2 h 1 2 U 2
where h and U are the local enthalpy and velocity. If we consider the fluid to be a perfect gas, then:
R T c
1
1
h c p T
so that:
c2
c2
1 2 U 2 1 2 U 2
1
1
1
U 2
2
c 2 c2 1
Ma 2
2
c
where
U 2 U e x U e x
C. Wassgren
(12.288)
618
Substituting Eqn. (12.288) into Eqn. (12.287) and expanding (refer to the Appendix):
2
1 Ma x y z
2
u x'
U
+
u 'y
2 2
2
2
2
Ma 1 2 1 2 2 2 Ma
U
x
z
Ma 2
'
u x
U
2 u 'y
2
U
x
2 1
Ma 2
'
u x
u'
2Ma 2 x
u 'y
2 u'
z
y 2 U
2 2 u 'y
2 2
z U
y
2
u 'y
2Ma 2
yx
U
u'
z
U

2
u'
2 Ma 2 z
U
xy

xz
2
z
2 2 u z'
2 2
z U
x
u'
2
2Ma 2 x
zy
U
2 2
2 2
y
x
u z'

U
zx
Note that the small perturbation assumption has not been used in deriving the previous equation. Now, if
we assume that the velocity perturbations are indeed small, i.e.:
u 'y
u'
u'
, Ma 2 z 1
Ma 2 x , Ma 2
U
U
U

Ma 2
u x'
u 'y
2
, Ma
U
u'
, Ma 2 z
U
u ' u 'y
u'
, Ma 2 x
Ma 2 x
U U
U

then the PDE simplifies to:
2
u z'
(12.289)
u 'y
2
, Ma
U
1 Ma x y z 1 Ma
2
u'
x
U
u'
z
U
2
2
x
(12.290)
Eqn of Motion for Small Perturbations when Ma1
Note that the 2/x2 term on the RHS has been retained in the previous equation. This is because when
Ma is near unity, the 2/x2 term on the LHS may be of the same order of magnitude as the RHS and
thus the RHS can not be neglected. Thus, Eqn. (12.290) is the appropriate form for the equation of motion
when the free stream Mach number is near unity.
Note that Eqn. (12.290) is non-linear. If we can assume further that:
Ma 2 u x'
1
1 Ma 2 U
then Eqn. (12.290) simplifies to:
2
1 Ma x y z 0
2
(12.291)
(12.292)
Eqn of Motion for Small Perturbations, Ma not near unity
C. Wassgren
619
Notes:
1. Note that the assumptions (Eqn. (12.289)) used in deriving Eqn. (12.292) also indicate that the Mach
number cannot be too large (e.g. we cant use Eqn. (12.292) to model hypersonic flows).
2.
Equation (12.292) is a linear PDE. This means that we can use the principle of superposition to add
together solutions of the equation to form new solutions.
3.
It is instructive to examine the mathematics of Eqn. (12.292) in greater detail. For simplicity, lets
consider only 2D flows. The general form for a 2nd order, linear PDE (with x and y as the independent
variables) is:
2
2
2
C 2 D
E
F G
A 2 B
xy
x
y
x
y
The behavior of the PDE will vary significantly depending on the value of its principle part which is
given by:
0 elliptic PDE
2
B 4 AC 0 parabolic PDE
0 hyperbolic PDE
The details of the differences in behavior between the three types of PDEs will not be examined here
except for dependence on boundary conditions. For an elliptic PDE, the solution at a particular point
in the domain will depend on all of the boundary conditions. In the language of mathematics, this is
the same as saying that there are no real characteristic directions (we will discuss characteristic curves
later in the course). The prototype elliptic PDE is Laplaces equation:
2 0
For a hyperbolic PDE, the solution at a particular point in the domain will depend only on a particular
region of the boundary conditions (termed the zone of dependence). Furthermore, the solution at that
point will have an effect only on a particular region termed the zone of influence. Mathematically
speaking, there are two real characteristic directions for hyperbolic PDEs. The prototype equation for
hyperbolic PDEs is the wave equation:
2
2u
2 u
c
t 2
x 2
We will not discuss parabolic PDEs here since we will only be concerned with elliptic and hyperbolic
PDEs in the following analyses. In addition, we already stated that Eqn. (12.292) is for the case when
Ma is not near a value of one, and -4AC only equals zero when Ma = 1.
How does all of this relate to the small perturbation equation of motion? Consider Eqn. (12.292)
simplified for 2D flow:
2
1 Ma x y 0
2
(12.293)
The behavior of the flow will vary significantly depending on the Mach number:
1 elliptic PDE
Ma
A = 1-Ma2, B = 0, C = 1 B2-4AC = 4(Ma2 - 1)
1
hyperbolic
PDE
Recall that when Ma1 (the transonic region), we must use equation (12.290) rather than equation
(12.292). The analysis in the transonic region is complex due to the non-linearity of the governing
equation.
Based on our previous analyses of 1D, compressible flow, the behavior of subsonic and supersonic
flows can be quite different. The same holds true here even though the governing equation (Eqn.
(12.293)) looks fairly simple. A deeper investigation of the differences between subsonic and
supersonic flows will be given in a following section.
C. Wassgren
620
4.
Since we linearized the governing PDE using the small perturbation assumption, we should also
linearize the boundary conditions. The appropriate boundary condition at a solid boundary for an
inviscid flow is not the no-slip condition since the fluid can slip tangent to the surface. Instead, we
specify that the flow will not penetrate the solid object, i.e.:
0
n
where n is the direction normal to the objects surface. This is the same as stating that the surface of
the object is a streamline for the flow (recall that there is no flow across a streamline). Thus, we can
write:
u 'y
'
y
u
U
dy
y
U
dx surface U u x'
u'
x
1 x
U
Since were assuming that the perturbation velocities are small in comparison with the free stream
velocity, we have:
u 'y
u x'
U
surface
U
Since the object is considered to be slender, we can use a Taylor series to show that the y-perturbation
velocity at y=0 can be used rather using the velocity on the surface:
dy
dx
surface
u 'y xs , ys u 'y xs , 0
u 'y
u 'y
y
ys
xs ,0
xs , 0
where (xs, ys) are the coordinates of the object surface and ys is assumed to be very small. Thus, the
appropriate boundary condition at the object surface becomes:
dy
dx
surface
u 'y
xs ,0
U
C. Wassgren
(12.294)
621
5.
An additional quantity that is often helpful when analyzing external flows is the pressure coefficient,
Cp, which is defined as:
p p
Cp
1 U2
2
This can be written for a perfect gas as:
p
1
p
(12.295)
Cp
1 Ma 2
2
The pressure ratio can be written in terms of the free stream and perturbation velocities by using the
energy equation for an isentropic (2D) flow:
2
2
1
1
T
U u x' u 'y T
U 2
2c p
2c p
where
cp
R
1
so that
T
1 2
U 2U u x' u x'
2 R
u T
2
'
y
'
u'
u
T
1 1 2 1 Ma 2 2 x x
U U
T
For an isentropic flow of a perfect gas:
u 'y

U
1 2
U
2 R
2 1
2
'
u x' u 'y
p T
1
2 ux
1 2 1 Ma 2
(12.296)
U U U
p T
The previous relation may be simplified further by use of the binomial theorem which states that for
x<1:
n n 1 x 2
1 x n 1 nx
2!
Since the perturbations are small:
'
u x' u 'y
1
2 ux
1
Ma
2
2
U U U
C. Wassgren
2
2 Ma
2 1
2
'
'
'
2 u x u x u y
U U U
622
Using the assumption that the perturbation velocities are small in comparison to the free stream
velocity, the remaining terms in the series can be neglected. Substituting this result into Eqn. (12.296)
and then substituting this result back into the definition of the pressure coefficient (Eqn. (12.295))
gives:
2
2
'
u x' u 'y
u
2
x
1 2 Ma 2
2
U U U
2
'
'
u 'y
u
u
x
x

2
Cp
U U U
1 Ma 2
Again, using the small perturbation assumption we observe that the squared terms in the previous
equation will be very small compared to the remaining term so that the pressure coefficient is given by:
u'
C p 2 x
(12.297)
U
C. Wassgren
623
6.
Equation (12.293) for a subsonic flow may be reduced to Laplaces equation using an appropriate
transformation of variables. This is useful since there has been a large body of work in determining
the solution to Laplaces equation for various boundary conditions. In particular, the equation of
motion for an incompressible flow is Laplaces equation.
To transform Eqn. (12.293), define a new coordinate system using the variables (, ) and a
transformed perturbation potential, , in the following manner:
x
y
where 1-Ma 2
(12.298)

Note that:
1 1

x
x
x
x
x 2 x x x x
1
0
1
0
(12.299)
1 2
1 1

y
y
y
y
y 2 y y y y
0

0

(12.300)
2
2

1 2 2
2
2
2

0

2
0
(12.301)
2 2
This is Laplaces Equation in the (, ) plane! This is the same equation of motion as for an
incompressible fluid (2= 0).
Since we transformed the governing equation, we also need to transform boundary conditions. Let the
shape of a boundary surface in the (x, y) plane be described by:
y s f xs
(12.302)
and the corresponding surface in the (, ) plane be:
s f s
C. Wassgren
624
(12.303)
Recall from Eqn. (12.294) that the boundary condition at a surface in the actual flow (in the (x, y)
plane) is:
dy
(12.304)
u y
U
xs ,0
y x ,0
dx surface
s
The same boundary condition expressed in the transformed flow, i.e. in the (, ) plane, is:
df
u
U
s ,0
d surface
,0
(12.305)
Note however that:

y x ,0
y
y
,0
s
s
s ,0
Hence, since the left hand sides of Eqns. (12.304) and (12.305) are equal, we also have:
dy
df
df
dx surface dx surface d surface
(12.306)
(12.307)
Since the slope of the actual boundary surface is (df/dx)surface, Eqn. (12.307) tells us that the shape of
the boundary surface in the transformed plane ((, )) is the same as that in the real ((x, y)) plane.
Furthermore, since the boundary surface is the same in the transformed plane, and since Eqn. (12.301)
is the equation of motion for an incompressible fluid, then must be the perturbation velocity
potential for an incompressible flow past the same boundary surface. Hence, if we can determine
assuming incompressible flow over the surface, the corresponding compressible flow solution can be
determined using the scaling relationships given in Eqn. (12.298).
Recall that the pressure coefficient (Eqn. (12.297)) is given by:
u'
2
2 1 1 2
C p 2 x
U x
U x U
U
(12.308)
C p
Note that the last term in parentheses in the previous equation is the pressure coefficient determined
from the incompressible (i.e., transformed plane) solution, C p 0 . Hence, the pressure coefficient for
the subsonic, compressible flow can be determined by scaling the incompressible pressure coefficient
by:
Cp
Cp
(12.309)
1 Ma 2
This scaling relationship is known as the Prandtl-Glauert Rule.
C. Wassgren
625
Notes:
a. It can also be shown that the lift and moment coefficients are scaled in a similar manner for
linearized, subsonic compressible flow (note that the drag coefficient is always zero for subsonic
potential flow):
CL
CL
(12.310)
1 Ma 2
CM
b.
CM
(12.311)
1 Ma 2
The effect of compressibility on the flow perturbation velocities can be determined from Eqn.
(12.298):
1
u
u
(12.312)
u
1 Ma 2
Compressibility acts to increase the magnitude of the perturbations (since the denominator is
always less than one). Hence, a disturbance to a compressible flow reaches further into the flow
than for an incompressible flow.
c.
The Prandtl-Glauert rule tends to underpredict the pressure coefficient for real flows. It is only
reasonably accurate up to a Mach number of about 0.7. This is because the rule is based on
linearization of the governing equations. Other rules have been proposed (e.g. the Karman-Tsien
rule and the Laitone rule) that incorporate non-linear flow effects to give improved predictions to
the pressure coefficient.
d.
Recall that for an isentropic flow that
1
c
p cT 1 dp
dT 1
1
so that a decrease in pressure corresponds to a decrease in temperature. So, for subsonic flow over
the suction side of an airfoil near sonic conditions, not only will the pressure drop, but the
temperature will drop as well. This drop in temperature in humid conditions can result in water
vapor turning to liquid water, i.e. condensation. This effect is often visible on the surface of high
speed aircraft in humid environments.
(Images from: http://en.wikipedia.org/wiki/Prandtl%E2%80%93Glauert_singularity )
C. Wassgren
626
Appendix
1
U e x U e x U e x
c2

1
2 U
e x y e y z e z
x
c
e x
e y
e z
U
U
x x y y z z
x
y
z

1
2 U
e x y e y z e z
x
c
1
c2
1
c2
1
c2
2
2
2
x U
y U
e
e
e z
U
x x 2
x xy
x xz
2
2
2
y
e x
e
e z
y y 2
y yz
y yx
2
2
2

y
e z
e
e
x
z zx
z zy
z z 2

2 2 2
U
U

x
x x 2
y yx z zx
2 2 2

2
x xy y y
z zy
y
2
2
2
U
z x xz y yz z z 2
2
U
2
x 2
2
2
2

U
yx
z zx x z zx
2
2 2 2
xy y y 2
y z zy
2 2
2
2

xz z y yz z z 2
2
2U
2 2
x x 2 x x 2
y yx x y
2
y xy y x
2
z xz z x
2 2
2 2 2
U 2 2U

2
y xy z xz
x
x x
2
2
2
2 2 2
2
2
2
z z
y y
x x
2
2
2
2
2
x y yx
y z zy
x z zx
C. Wassgren
627
2 2
2 2 2
U 2 2U

2
y xy z xz
x
x x
2 2
2 2
2 2
1

2
2
2
2
2
x
y
z
c
x
2
2
2
2 2 2
x y yx
y z zy
x z zx
Recall that from the speed of sound equation:

1
U 2
2
c 2 c2 1
Ma 2
2
c
1 2 U e x
Ma
c2 1
2
c2
(12.313)
2
2
2
U 2 2U
x x y z
1
c2 1
Ma 2
2
c2
2
2
1 2U 1
c2 1
2 c2 x c2 x y z
c2
Ma 2 1
2 1 2 2 2
U x U 2 x y z
2
C. Wassgren
628
(12.314)
Substituting (12.314) into Eqn. (12.313) gives:

Ma 2 1
2 1 2 2 2 2
c2 1

U x U 2 x y z
2
2 2
2 2 2
U 2 2U

2
y xy z xz
x
x x
2 2
2 2
2 2

2
2
2
y
z
x
x
2
2
2
2 2 2
x y yx
y z zy
x z zx
Ma 2 1
2 1 2 2 2 2
U x U 2 x y z
2
2 2 2 2

2
y xy z xz
U x x
x
2
2
2
1 2 2 2
2

Ma 2
2
2
z z 2
y y
U x x
2 2 2 2
2
U x y yx y z zy x z zx
2
Ma 2 1 2 Ma 2 1 2 Ma 2 1 2
U
U
U
x x 2
x y 2
x z 2
Ma 2 1 2 2 Ma 2 1 2 2 Ma 2 1 2 2
x
x
x
2U 2
2U 2
2U 2
x 2
y 2
z 2
Ma 2 1 2 Ma 2 1 2 Ma 2 1 2
2
2
2
2U 2
2U 2
2U 2
y z
y x
y y
Ma 2 1 2 2 Ma 2 1 2 2 Ma 2 1 2 2
z
z
z
2U 2
2U 2
2U 2
x 2
y 2
z 2
+Ma 2
2
x 2
2 Ma 2 2 2 Ma 2 2 2 Ma 2 2
U x x 2
U y xy
U z xz
Ma 2 2 Ma 2
2
U 2 x x 2
U
2Ma 2 2 2Ma 2 2 2Ma 2 2
U 2 x y yx
U 2 y z zy
U 2 x z zx
C. Wassgren
2 Ma 2 2
2
2
U z z 2
y y
629
2 Ma 2
2
2
2
2
2
2 Ma 1 2 Ma 1 2 2 2
2
2
2
2
U
x 2
2U 2
z z
x x
x x
y y
Ma 2 1 2 Ma 2 1 2 2 Ma 2 2 2 Ma 2 2
U
U
U y xy
U z xz
x y 2
x z 2
2
Ma 2 1 2 2 2 2 2 2 2 2
2
2

2 2

2U 2
z y x 2
z z x 2
y
x y
2Ma 2 2 2Ma 2 2 2Ma 2 2
U 2 x y yx
U 2 y z z y
U 2 x z z x
1 Ma x y z
2
Ma 2
u'
1 x
U
'
2
u
2
2 1 2 1 Ma x
x
U
u'
1 Ma 2 x
U
u 'y
2 2
2
2 2 2 Ma
U
z
y

'
u
1 2 1 Ma 2 x
u'
2Ma 2 x
U
2
2 u 'y
2
U
x
u 'y

U
2 2 u 'y
2 2
z U
y
2
u 'y
2Ma 2
yx
U
2 u'
z
y 2 U
2
u'
2 Ma 2 z
xy
U
2 2 u'

2 z
x 2
z U
u'
z
U

u'
2
2Ma 2 x
U
zy

2
z
xz
2
2 2
y 2
x
u z'

U
zx
1 Ma x y z
2
u x'
u 'y
2 2
2
2
2
Ma 1 2 1 2 2 2 Ma
U
x
z
Ma 2
'
u x
U

2 u 'y
2
U
x

1 2 1
Ma 2
'
u x
u'
2Ma 2 x
U

u 'y

U
C. Wassgren
2 u'
z
y 2 U
2 2 u 'y
2 2
z U
y
2
u 'y
2Ma 2
yx
U
u'
z
U
630
2
u'
2 Ma 2 z
xy
U
xz
2
z
2 2 u z'
2 2
z U
x
u'
2
2Ma 2 x
U
zy

2 2
y 2
x
u z'

U
z x
19. Method of Characteristics
The method of characteristics is a procedure by which non-linear, hyperbolic PDEs may be solved in an
algorithmic fashion.
Before beginning with the algorithm, we must first discuss a few details regarding hyperbolic PDEs. To
begin, lets first define a characteristic curve. A characteristic curve is a curve across which the value
of some parameter is continuous, but the derivatives of that parameter are indeterminate. Consider
for example the flow through a Mach wave. Recall from our previous analysis that velocity changes across
a Mach wave are very small, i.e. the velocity across a Mach wave goes from V to V+dV. This infinitesimal
velocity change (dV) occurs over zero distance since the Mach wave has no thickness. Thus, although the
velocity changes continuously across the Mach wave, the velocity gradient is indeterminate (dV/0=?).
Hence, a Mach wave is a characteristic curve according to our definition. Characteristic curves have
additional useful properties but we will not discuss them at this time.
Our first task will instead be to determine the slope of a characteristic curve in a general flow situation.
Recall from our previous analysis that the equation of motion for a steady, irrotational, isentropic
compressible flow with negligible body and viscous forces is given by:
1
2 2
c
Expanding using Cartesian coordinates:

2 2 2

1
e x
e y
e z
2 2 2
2
y
z
x
y
z
c
x x y y z z x

1
2
e x
e y
e z
y
z x
c x
2
x
2
y
2
z
1
c2
1
c2
C. Wassgren
2 2
e x
y yx z zx
x
2
2
2

e y
2
xy y y
z zy
2
2
2

ez

xz y yz z z 2
2
2
2 2 2 2
2
x y yx x z zx
x x
2 2
2
2

2
y z zy
y x xy y y
2 2
2
2

z x xz z y yz z z 2
2
2
2
2 2 2
2
2
2
x x
z z
y y
2
2
2

2
2
2
y z zy
x z zx
x y yx
631
(12.315)
For 2D flow, the previous equation can be written as:

2
2
2
2 2 2

2
c
0
c 2
2
x y xy
x x 2
y y
This non-linear equation has the form:

2
2
2
A 2 2B
C 2 0
xy
x
y
(12.316)
(12.317)
where A, B, and C depend on /x and /y.

For an isentropic flow, the velocities, i.e. /x and /y, are continuous. Since were interested in
finding the velocity characteristic curves, we should consider differential changes in the velocities:
2
2

du x d 2 dx
dy
(12.318)
xy
x x
2
2
du y d
dx 2 dy
y
y xy
Now lets rewrite equations (12.317) through (12.319) as a system of linear equations where the
independent variables are the velocity gradients (2/x2, 2/xy, and 2/y2):
2
2
2
C 2
0
A 2 2 B
xy
x
y
dx
2
x
dy
2
xy
(12.319)
(12.320)

2
2
dy 2 d
xy
y
y
Recall from linear algebra that Kramers rule states that a system of equations has a unique solution if and
only if the determinant of the system is not equal to zero, i.e.:
A 2B C
(12.321)
dx dy 0 Ady 2 2 Bdxdy Cdx 2 0
0 dx dy
0
dx
Thus, if this determinant is equal to zero, then the velocity derivatives (or gradient) (2/x2, 2/y2, and
2/xy) will be indeterminate. The slope of such a curve is given by:
Ady 2 2 Bdxdy Cdx 2 0
dy 2 B 4 B 2 4 AC B B 2 AC
dx
2A
A
where
2

A c 2
c 2 u x2

B
u x u y
x y
2

2
2
C c 2
c uy
y
C. Wassgren
632
Thus,
2 2
2
2
2
2
dy u x u y u x u y c u x c u y
dx
c 2 u x2
dy
dx
u x u y u x2u 2y c 4 c 2 u x2 c 2u 2y u x2 u 2y
c
u x2
u x u y c 2 u x2 u 2y c 4
c
uxu y
c
u x2
Ma 2 1
slope of characteristic curves
u x2
(12.322)
c2
Recall from our previous discussion that a curve across which the velocity derivatives (or gradient) are
indeterminate are referred to as characteristic curves.
Notes:
1. For Ma<1 there are no real characteristic curves (elliptic PDE). For Ma=1 there is one characteristic
curve (parabolic PDE), and for Ma>1 there are two characteristic curves (hyperbolic PDE).
2.
Recall from our previous, linearized (small perturbation) analysis that the governing PDE was linear:
2
2 2
0
A 2 2B
C
x
xy y 2
where
A 1 Ma 2
B0
C 1
Thus, the slope of the characteristic curves are given by:
dy B B 2 AC
dx
A
Ma 2 1
Mach lines!
Thus, the characteristic curves for linearized flow are straight Mach lines.
C. Wassgren
633
3.
It can be shown that the characteristic curves are inclined at the Mach angle to the local flow direction.
To show this, first re-write the velocity components in terms of the velocity magnitude, V, and the
local flow inclination, :
y
V
u x V cos
(12.323)
u y V sin
x
Substituting into Eqn. (12.322) and noting that, sin= 1/Ma (where is the Mach angle), gives:
dy Ma 2 sin cos Ma 2 1
dx
1 Ma 2 cos 2
1
Ma
sin cos
Ma 2 1
2
1
Ma
1
cos 2
Ma 2
(Ma2-1)1/2
sin 2
tan = (Ma2-1)-1/2
sin cos
tan
sin = 1/Ma
2
2
sin cos
After considerable rearrangement and simplification using trig identities we find that:
dy
tan
(12.324)
dx
The - corresponds to the right-running characteristic while the + corresponds to the left-running
characteristic.
Thus, the slope of the characteristic lines will be inclined at the Mach angle to the local flow
direction.
left-running
characteristic curve
left-running
Mach line
V
y
right-running
characteristic curve
right-running
Mach line
x
Notes:
a. The direction of a characteristic curve or Mach line is determined by facing downstream. The leftrunning waves are on the LHS while the right-running waves are on the RHS.
b.
The characteristic curves are the Mach curves for the flow.
C. Wassgren
634
c.
Although we now have an expression for the slope of the characteristic curves (Eqn. (12.324)), we
still need to have some relation for determining how the velocities change along a characteristic
curve. Recall from the system of equations from Eqn. (12.320):
2
2
2
C 2
0
A 2 2 B
xy
x
y
dx
2
x
dy
2
xy
dx
2
xy

d
du x
x
0
dy
2
y

d
du y
y
where
2

2
2
A c 2
c ux
x

B
u x u y
x y
2

2
2
C c 2
c uy
y
The solution for the gradient of the x-velocity component in the x-direction (using Kramers rule)
is:
u x 2
2
x
x
0
du x
du y
2B C
dy 0
dx dy
A 2B C
dx dy 0
0 dx dy
Cdxdu x Cdydu y 2 Bdydu x

Ady 2 2 Bdxdy Cdx 2
(Note: We could also consider the other velocity gradients such as uy/x and would get the same
result given below.) Recall that along a characteristic curve the denominator in the previous
expression will be zero. Thus, in order for ux/x to remain finite (were assuming that the
velocities in the flow will be finite so that velocity gradients will also be finite), the numerator
must also be zero:
Cdxdu x Cdydu y 2 Bdydu x 0
dx 2 B
du x dy C
Substituting in for the characteristic curve slope (Eqn. (12.324)) and simplifying gives:
uxu y
Ma 2 1
2
du y
c
du x
u 2y
1 2
c
du y
C. Wassgren
635
(12.325)
Re-write the velocity components in terms of the velocity magnitude, V, and direction, :
u x V cos
du x dV cos V sin d
u y V sin
du y dV sin V cos d
Substituting into Eqn. (12.325) and simplifying:

1 dV
1
tan
V d
Ma 2 1
(12.326)
We now need to integrate the previous expression as we move along a characteristic line. Weve
performed this same integration before when discussing Prandtl-Meyer expansion fans (refer to
Eqn. (12.275) and (12.276)). Instead of re-deriving the result again, well just copy the result
here:
b
d Ma 2 1
a
dV
b a
V
b a b a
(12.327)
where a andb are the Prandtl-Meyer angles for the flow at locations a and b, both of which lie
on the same characteristic curve. Again, a and b are the flow velocity directions at points a and
b. If we let the point a be some reference point on the characteristic curve, then:
constant
along a right-running characteristic curve
(12.328)
constant
along a left-running characteristic curve
(12.329)
Equations (12.328), (12.329), and (12.324) can be combined together in an algorithm for solving
steady, supersonic, irrotational compressible flows.
C. Wassgren
636
Method of Characteristics Algorithm
Suppose that the flow conditions are known along some curve CD:
C
A
B
D
Since the flow conditions are known on this curve, the flow angle, , and Prandtl-Meyer angle, , are
known at the two points A and B. We now move along the right-running characteristic curve from point A
until it intersects with the left-running characteristic curve from point B. Call this intersection point P. At
point P we have:
P P A A CR
P P B B CL
where CR and CL are constants. Re-arrange the previous expressions to get:
P 1 2 CR CL
CR C L
Since we now know the Prandtl-Meyer angle and flow direction at point P, we can determine the remainder
of the flow properties for this isentropic flow (e.g., Mach number, pressure, temperature, velocity
components, etc.) However, we still dont know the location of point P. Since the Mach number changes,
in general, from point to point in the flow field, the characteristics curves will not necessarily be straight
lines. However, if we choose the points A and B to be very close together, then we can approximate the
position of point P by the intersection of the left- and right-running Mach lines passing through A and B
(call this point P).
P
Mach lines
C
A
B
D
Obviously the location of point P approaches P as the distance between A and B becomes smaller.
We can determine the conditions at other points in the flow field by repeating this procedure for other
points where the flow conditions are known.
C. Wassgren
637
Note that along a right-running characteristic, CR = constant so:
12 CR CL 12
CR CL 12 CL
C CL 12 CL
12 CR CL 12
R
Similarly, along a left-running characteristic, CL = constant so:
12 CR CL 12 CR
C 1 CR
12 CR CL 12 CR
L 2
1
2
(12.330)
CR CL
C. Wassgren
638
(12.331)
Example:
Uniform radial flow at Mach 2.0 enters a 2D diverging channel with straight walls. Compute the variation
of Mach number in this radial flow field, assuming isentropic, steady flow. The walls are inclined at a total
angle of 12. Compare your results at the centerline with those using 1D isentropic flow analysis.
Ma=2.0
12
C. Wassgren
639
Method of Characteristics
Types of Points:
Interior Points:
CL and CR are known, determine and
Solid Boundary Points:

either CL or CR is known and is known
Free Boundary Points:
either CL or CR is known and is known
(Since p is known can determine Ma can determine )
Interaction of Waves
reflections off solid boundaries
cancellation of waves
reflections off free boundaries
intersection of waves
Region-to-Region Method
Can cross characteristics too. Recall from our previous work that across a Mach wave (a characteristic
curve), we have:

where is the Prandtl-Meyer angle and is the flow orientation. These expressions can be written as:
2 2 1 1
across a right-running characteristic
and
2 2 1 1
across a left-running characteristic
(same as moving along the opposite characteristic)
Angle of lines separating regions
ij 12 i i j j (across R-running)
ik 12 i i k k (across L-running)
L
k
i
u
C. Wassgren
640
Example:
Flow at the exit of a Mach 1.5 supersonic nozzle is expanded from an exit-plane pressure of 200 kPa (abs)
to a back pressure of 100 kPa (abs). Determine the flow just downstream of the nozzle exit in the regions
indicated.
pb=100 kPa (abs)
Mae=1.50
pe=200 kPa (abs)
4
7
1
2
C. Wassgren
641
10
Design of a Supersonic Wind Tunnel Nozzle
One simple application of the Method of Characteristics is in the design of a supersonic wind tunnel nozzle.
Recall that the function of a supersonic nozzle is to accelerate a flow from Ma=1 to some final supersonic
Mach number. In order to simulate free flight, the flow through the test section of the wind tunnel should
be parallel and uniform.
Consider a symmetric nozzle so that only the upper half of the nozzle must be considered. For simplicity,
assume that the incoming flow at the throat is uniform and at Mach 1.0 (of course in a real nozzle the
incoming flow would not be uniform). The flow will first expand in the region from points a through b as
shown in the figure below. Four of the expansion waves (right-running characteristics) are shown in the
sketch.
f
c
b
a
test section
The expansion waves reflect off the centerline at points g through j as expansion waves which act to turn
the flow back toward the horizontal (these are left-running characteristics). These waves impinge on the
nozzle wall at points c through f. In order to avoid further reflections, the wall contour at points c through f
should be sloped at the wave turning angle so that the waves are cancelled.
Notes:
1. As more waves are included in the analysis, the nozzle contour will become smoother.
2.
The initial expansion from a to b is arbitrary in the design. The critical points in the analysis are points
c through f which must designed to provide wave cancellation.
3.
To design a nozzle with the shortest length, the expansion from points a to b should take place as a
centered Prandtl-Meyer expansion fan:
d
a,b
g
4.
The design of a real nozzle should also factor in boundary layer effects in order to give the correct flow
area and avoid boundary layer separation.
C. Wassgren
642
Example:
Design the diverging section of a supersonic nozzle to produce uniform Mach 1.8 flow. Assume that the
length of the nozzle should be kept to a minimum.
C. Wassgren
643
20. Flow Past a Wavy Wall Using Small Perturbation Theory
Recall that for a steady, irrotational, 2D flow with negligible body and viscous forces, the equation of
motion assuming small velocity perturbations, is:
2 2
(12.332)
1 Ma 2 x 2 y 2 0
The linearized boundary condition at the object surface is:
'
dy
dx
surface
uy
x ,0
(12.333)
Far from the object (y), the velocity perturbations must remain finite.
Recall that the pressure coefficient at the object surface is given by:
2u x'
Cp
U
(12.334)
Now consider flow past a wavy wall. Let the profile of the wall be given by:
2 x
ys A sin
Ma
Well assume that the wall causes small perturbations in the flow, i.e.:
A
Thus, the boundary condition at the wall is:

2 AU
dy
2 x
'
cos
U
uy
x ,0
dx surface

C. Wassgren
(12.335)
644
Subsonic Flow
First lets examine the case when Ma2 < 1 so that (1-Ma2) > 0. Equation (12.332) will be an elliptic PDE
for this case. One method for solving elliptic, linear PDEs is to use separation of variables where we
assume that the solution can be written as some function of x multiplied by some function of y:
x, y X x Y y
Substituting into Eqn. (12.332) and simplifying gives:
1 Ma X Y XY 0
2
X
X
1 Ma Y
2
Since the LHS of the previous equation is a function only of x, and the RHS is a function only of y, then in
order for the two sides to be equal, they must equal a constant, which well call -k2. Thus, we can write the
following two equations:
X k 2 X 0
Y k 2 1 Ma 2 Y 0
The solution to the first differential equation involving X is:

X c1 cos kx c2 sin kx
and the solution to the differential equation involving Y is:
Y c3 exp ky 1 Ma c4 exp ky 1 Ma
2
Note that the square root term results in a real quantity since the incoming flow is subsonic. Substituting
these functions into the perturbation potential and determining the corresponding velocity perturbations
gives:
2
2
c1 cos kx c2 sin kx c3 exp ky 1 Ma c4 exp ky 1 Ma
u x'
'
uy
k c sin kx c cos kx c exp ky 1 Ma c exp ky 1 Ma
k 1 Ma c cos kx c sin kx c exp ky 1 Ma c exp ky 1 Ma
y
1
The constants c1, c2, c3, and c4 are determined from the boundary conditions at the wall and at y. In
order for the perturbation velocities to remain finite as y we must have c3=0. At the wall (y = 0) we
have (Eqn. (12.335)):
2 AU
2 x
2
u y x ,0
cos
c4 k 1 Ma c1 cos kx c2 sin kx

Thus we see that c2=0, k=(2/), and
U
AU
2 A
c1c4

2 2
1 Ma 2
1 Ma
C. Wassgren
645
Thus, the perturbation potential and perturbation velocities become:

AU
2 x
2 y
2
cos
1 Ma
exp
2

1 Ma
2 AU
ux
'
1 Ma
uy
'
2 AU
2 x
2 y
2
1 Ma
exp

(12.336)
sin
2 x
2 y
2
1 Ma
exp

cos
The pressure coefficient at the wall (refer to Eqn. (12.334)) is:

4 A
2 x
Cp
sin
2

1 Ma
(12.337)
Notes:
1. The disturbance of the wall dies out as y.
y
streamlines
x
2.
The pressure peaks occur in the troughs of the wall and vice versa, i.e., the pressure is in-phase with
the wall. As a result, there will be no drag force on the wall.
y, Cp
pressure profile is symmetric net

horizontal force is zero
C. Wassgren
646
3.
We can also examine this compressible subsonic flow using the coordinate transformation discussed in
a previous set of notes concerning the Prandtl-Glauert rule. Recall that if we examine flow in the (,
) plane where:
x
y where 1-Ma 2
(12.338)

then the governing equation becomes Laplaces equation, i.e.:
2 2
0 (This is the governing equation for an incompressible flow!)

(12.339)
2 2
Furthermore, the shape of the boundary surface in the (, ) plane is the same as the boundary surface
shape in the (x, y) plane. Solving (12.339) using separation of variables gives:
(12.340)
c1 cos k c2 sin k c3 exp k c4 exp k
Since the flow must have finite velocities as (this is true for both the compressible and
incompressible flow cases), we can conclude that c3 must be zero. The boundary condition at the wavy
wall surface:
2 AU
2
u ,0
cos
(12.341)
c4 k c1 cos k c2 sin k

indicates that c2 must be zero, k must be (2/), and c1c4 must be -AU. Thus:
2
2
AU cos
(12.342)
exp

Transforming back into the (x, y) plane using equation (12.338) gives:
AU
1
2 x
2 y
(12.343)

cos
1 Ma 2
exp
2
1 Ma
This is precisely the same relation derived in equation (12.336). The pressure coefficient
corresponding to the incompressible flow (i.e. the (, ) plane) is:
2
4 A 2
Cp
sin
0
U
(12.344)
Using the Prandtl-Glauert rule, the pressure coefficient for the compressible flow (i.e. the (x, y) plane)
is:
Cp
4 A
2 x
(12.345)
Cp
sin
2
2

1 Ma 1 Ma
Equation (12.345) is the same as equation (12.337), as expected.
C. Wassgren
647
Supersonic Flow
Now consider the case where Ma2 > 1 so that (1-Ma2) < 0 and Eqn. (12.332) becomes a hyperbolic PDE
with a form identical to that of the wave equation. Thus, well utilize dAlemberts solution to the wave
equation which has the form:
Recall that the wave equation has the form:
f L x y Ma 1 f R x y Ma 1
or
f L f R
where
2u
2
c2
2u
t
x 2
and has the (dAlembert) solution:
u x, t f x ct g x ct
(12.346)
where f and g are functions determined by the initial

and boundary conditions. For our case, the governing
equation is:
x y Ma 1
2
x y Ma 2 1
2
Ma
1
2 2
2
Note that the functions fL and fR are unknown at this point but will eventually be determined using the
boundary conditions. To verify that Eqn. (12.346) is indeed a general solution, substitute it back into Eqn.
(12.332) and simplify:
2
1 Ma x
2
2
y 2
where
d f L
2
2
2
d f L
d fL d fL
2
2
2
2
d x
d x
d
d
2
d 2 fL d 2 fL
d 2 f L d 2 f L
2
Ma
1
2
d 2 y
d 2 y
d 2
y
d
Of particular interest for this solution are the curves corresponding to fL=constant and fR=constant (along
these curves and, hence ux and uy, will remain constant). Note that the form of these functions will be
dictated by the boundary conditions. Thus, the value of the perturbation potential, , (as well as the
perturbation velocities) will be propagated from the boundary conditions into the rest of the flow along the
curves where fL and fR are constant. The shape of these curves can be determined from:
dy
1
x y Ma 2 1 constant
fL=constant
dx f L constant
Ma 2 1
fR=constant
x y Ma 2 1 constant
dy
dx
f R constant
1
Ma 2 1
Thus we see that the two curves are, in fact, lines with opposite slopes. Moreover, the slope of these lines
is equal to the slope of a Mach line:
Ma
1
1
sin
tan
1
Ma
Ma 2 1
(Ma-1)1/2
It is worthwhile to re-iterate these last two important points:
1. Information propagates along the curves where fL and fR are constant.
2. Curves along which fL and fR are constant correspond to Mach lines of opposite slope.
C. Wassgren
648
For the problem at hand (supersonic flow over a wavy wall), the perturbation potential should only include
the function fL since if we included fR in the solution, then information could propagate upstream along the
Mach line with negative slope an impossibility in supersonic flows.
fR=constant
fL=constant
Ma
x
Thus, the perturbation potential is:
f L
x y Ma 2 1
where
(12.347)
The form of fL will be dependent on the boundary conditions. Recall that the boundary condition at the
wall is given by Eqn. (12.335):
2 AU
dy
2 x
'
U
uy
cos
x ,0
dx surface

Substituting Eqn. (12.347) into the boundary condition gives:
x y Ma 2 1
df L
df L 2 AU
2 x
'
2
cos
Ma 1
uy
y 0
Ma 2 1
y y 0 d y y 0
dx

y
df L
2 AU
2 x
cos
dx Ma 2 1

Note that in the previous equations the fact that at y=0, dfL/d = dfL/dx, has been used. Integrating gives:
AU
2 x
sin
fL x
constant
2

Ma 1
Thus,
AU
f L
Ma 1
2
2
constant

sin
The resulting perturbation potential and corresponding perturbation velocities are:

AU
2
2
sin
x y Ma 1 constant
2
Ma 1
'
ux
'
uy
2 AU
Ma 1
2
2 AU
2
x y Ma 1
cos
(12.348)
2
x y Ma 1
cos
C. Wassgren
649
The pressure coefficient at the wall (Eqn. (12.334)) for this flow is:
Cp
2u x'
U
4 A
Ma 1
2
2 x
(12.349)
cos
Notes:
1. Along lines of x-y(Ma2-1)1/2=constant (recall that these are the Mach lines), the perturbation potential
and velocities are constant. Thus, the disturbances produced by the wall are felt equally along the
same Mach line. Recall that in the subsonic flow case, the disturbances die out as y.
Mach lines
streamlines
x
2.
In the subsonic case the drag on the wall is zero. For supersonic flow, however, the drag is not zero.
This is because the pressure coefficient is out of phase with the wall shape (shown below). This type
of drag is commonly referred to as supersonic wave drag.
pressure profile is asymmetric net
horizontal force is not zero
y, C
p
3.
In an actual flow, the drag coefficient on the wall as a function of Mach number look like:
CD
perturbation model
experimental data
Ma
1
The drag occurring in real subsonic flows is due solely to viscous effects. For real supersonic flows
around slender bodies, the wave drag is typically much larger than the viscous drag. Note that the
assumptions for perturbation analysis do not allow for the investigation near Ma1.
C. Wassgren
650
21. Thin Airfoils in Supersonic Flow
Recall that for steady, irrotational, supersonic, compressible flow with negligible body and surface forces
and small perturbations, the perturbation potential is given by:
f R x y Ma 2 1 f L x y Ma 2 1
where fL and fR are arbitrary functions that are dependent on the boundary conditions.
y
fL=constant
Ma
x
fR=constant
Since information cannot be propagated upstream in a supersonic flow, the perturbation potential for flow
over the top of the airfoil (y>0) should only include left-running Mach waves:
f L x y Ma 2 1
while under the bottom of the airfoil (y<0) the solution should contain only right-running Mach waves:
f R x y Ma 2 1
The boundary condition on the upper surface of the airfoil is:

dy
dx
u 'y
upper
upper
1
U y
upper
so
dy
dx
upper
f L'
1
U y
U
Ma 2
upper
dy
1 dx
Ma 2 1
f L'
(Note:
df L
.)
y d y
(12.350)
upper
The pressure coefficient on the upper surface is given by:

2 u x'
2
2 '
df L
(Note:
.)
fL
x d x
U
U x y 0 U
Substituting equation (12.350) into equation (12.351) gives:
2 U
dy
C p ,upper
U Ma 2 1 dx upper
C p ,upper
C p ,upper
y 0
2
Ma 2
dy
1 dx
C. Wassgren
(12.351)
(12.352)
upper
651
A similar approach can be taken to determine that:

2
dy
C p ,lower
Ma 2 1 dx lower
(12.353)
Notes:
1. The slope of the upper and lower surfaces relative to the incoming flow depends not only on the airfoil
shape, but also on the airfoils angle of attack, .
slope at point on surface
relative to flow is different
y, y
Ma
Ma
y
y
x, x
c
(a)
(b)
For example, the local slope of the surface relative to the incoming flow, (dy/dx), for case (a) is
different than the slope of the surface at the same point for case (b). In order to isolate the effects of
the airfoil shape and angle of attack, lets define the quantity, , as:
dy
(12.354)
dx
so that represents the airfoils surface slope relative to the airfoils chord line, i.e. with respect to the
(x, y) axes in the figures above. The slope of the surface relative to the incoming flow, , will be:
airfoil surface
tan
Note that and are slopes,
not angles.
dx
dx
dy
dy
dy
dx
dy
dx
(12.355)
Note that tan since were concerned only with small perturbations to the flow. The pressure
coefficients (equations (12.352) and (12.353)) may be written as:
2 U
C p ,U
(12.356)
Ma 2 1
C p, L
2 L
(12.357)
Ma 2 1
where the subscripts U and L refer to the upper and lower airfoil surfaces, respectively.
C. Wassgren
652
2.
The airfoils lift and drag can be determined by integrating the net pressure force acting over the entire
airfoil surface. The pressures acting on the upper and lower airfoil surfaces are:
pU p C p ,U
U 2 C p ,U
1
2
Ma 2
Ma 2 1
p Ma 2
(12.358)
Ma 2
pL p p
Ma 2 1
(12.359)
Resolving the pressure force acting on the upper surface on a small area element, ds (Note that the
airfoil distance into the page, also known as the span, is assumed to have unit depth.), into lift (L) and
drag (D) components gives:
dLU pU ds cos U pU dx pU dx (since the slopes are small, dx dx)
dx
dy
dDU pU ds sin U pU
dx pU U dx
dx
dy
pds
ds
Similarly, for the lower surface:

dLL pL ds cos L pL dx pL dx
dy
dx
dy
dDL pL ds sin L pL
dx pL L dx
dx
dx
dx
dy
dy
Note that the small angle approximation has been used in the expressions above. The net lift and drag
are determined by integrating over the entire airfoil surface.
x c
x 0
dLU dLL pU pL dx
0
p
L p
x c
x 0
Ma 2
Ma 2 1 0
c
Ma 2
Ma 2 1 0
L dx
U 2 dx
(12.360)
dDU dDL pU U pL L dx
p
D p
Ma 2
Ma 2
Ma 2
Ma 2
C. Wassgren
2
U
2
2
L dx
L2 2 U L 2 2 dx
(12.361)
653
y
Note, however, that:
c
dx
dy
dx dy 0
dx
so that Eqns. (12.360) and (12.361) become:

Ma 2
L p
2 c
Ma 2 1
Recall that this is the chord

length so that:
y x 0 y x c 0
U2 L2 dx 2 2 c
Ma 2 1 0
Written in terms of the lift and drag coefficients (based on the chord length):
L
4
CL
1 p Ma 2 c
Ma 2 1
2
Ma 2
D p
CD
3.
D
1
2
p Ma 2 c
c Ma 2 1 0
U2 L2 dx
4 2
Ma 2 1
(12.362)
(12.363)
(12.364)
(12.365)
The lift coefficient is directly proportional to the angle of attack for thin supersonic airfoils at small
angles of attack. Note that there is also a component to the drag coefficient that depends only on the
angle of attack. This component is often referred to as the wave drag due to lift:
4 2
CD ,wave drag
(12.366)
due to lift
Ma 2 1
The other part of the drag coefficient depends only the shape of the airfoil and is known as the wave
drag due to thickness since the thicker the airfoil, the larger the integral term (the slopes will be
larger):
CD , wave drag
due to thickness
Ma 2
2
U
L2 dx
(12.367)
These previous results suggest that in order to minimize the drag acting on a supersonic airfoil, the
airfoil should be as thin as possible. Furthermore, the lift acting on the airfoil will be unaffected by the
shape of the airfoil.
C. Wassgren
654
4.
Lets now consider subsonic flow around the airfoil shown below.
Ma < 1
For sufficiently small free stream Mach numbers, Ma, the flow over the entire airfoil surface will
remain subsonic and the corresponding drag coefficient for the airfoil will remain at a relatively small
value (Point A on the plot shown below. Note that the total drag will be due to skin friction and form
drag.).
CD
C
B
Macr Madd 1
Ma
As the free stream Mach number is increased, a critical free stream Mach number, Macr, will be
reached where a sonic Mach number will occur at the minimum pressure point on the airfoil surface.
Ma = Macr < 1
Ma = 1
At free stream Mach numbers slightly greater than Macr, a small region of supersonic flow will occur
near the minimum pressure point. The drag coefficient for these flow conditions (point B in the
diagram above) remains close to, but only slightly greater than, the drag coefficient for purely subsonic
flow.
(Ma > Macr) < 1
Ma < 1
Ma > 1
At another critical Mach number, known as the drag divergence Mach number, Madd (Macr < Madd <
1), the drag on the airfoil increases suddenly (point C in the CD vs. Ma plot shown above) due to the
formation of a terminating shock wave. The shock wave on the airfoil surface causes the boundary
layer to separate (due to the large adverse pressure gradient across the shock wave) resulting in a
significant increase in the form drag.
Ma < 1
(Ma > Madd) < 1
separated flow
Ma > 1
shock wave
This sudden increase in the drag at Madd is the origin of the concept of the sound barrier.
C. Wassgren
655
In order to minimize the drag acting on the airfoil when operating near sonic conditions, the drag
divergence Mach number should be pushed as close to sonic conditions as much as possible.
a. One approach to increasing Madd is to decrease the thickness of the airfoil. Recall that the largest
Mach numbers will occur in the vicinity of the minimum pressure region. If the minimum
pressure can be brought closer to the free stream pressure, then the corresponding local Mach
number will deviate less from the free stream Mach number (which is subsonic). Making the
airfoil thinner will result in less expansion of the flow and hence the Mach numbers over the
airfoil will be smaller. As a result, the drag divergence Mach number for the thin airfoil will occur
at a higher free stream Mach number than a thicker airfoil. Thinner airfoils also produce less drag
for supersonic conditions as discussed previously in Note 3.
b.
Another approach to reducing the drag is to use a supercritical airfoil (shown below) which is
designed to give Madd 1. The airfoil shape is designed to give mostly supersonic flow and
discourage the formation of shock waves.
c.
A third approach to delaying Madd is to use a swept-wing design. Consider flow over straight and
swept wings (inclined by an angle from the straight wing) that have identical airfoil crosssections. Note that only the component of the flow normal to the airfoil will be important in
determining the drag divergence Mach number since the flow tangential to the surface does not
see variations in the airfoil geometry. Hence, the effective free stream Mach number for the
swept-back wing is smaller than that for the straight wing (Maeff = Macos). As a result, the
negative effects associated with shock formation on the airfoil can be delayed until a higher free
stream Mach number is reached. The downsides of swept wings are that the wing area must be
increased to generate the same lift as a straight wing (since the lift decreases as a result of the
lower effective free stream Mach number), and the structural design of the wing is more complex.
Ma
Macos
Ma
Swept-wings are also advantageous in supersonic flight since the wing may be subject to a
subsonic effective Mach number and, as a result, the penalty of supersonic wave drag can be
avoided.
C. Wassgren
656
Example:
A symmetric strut of chord length, L, is placed at a small angle of attack, , in a supersonic flow of Mach
number, Ma.
L
x
Ma
t(x)
The geometry of the strut is such that the centerline is straight and the foil thickness, t(x), is given by:
t x
x x
4 1
tM
L
L
where tM is the maximum thickness of the foil which occurs at x/L=1/2. Assuming that the foil is slender
and that the angle of attack is small, find expressions for the lift and drag coefficients for this strut as
functions of Ma, , and tM/L.
C. Wassgren
657
22. Unsteady, 1D Compressible Flow
Applications that might be approximated as being, 1D and unsteady:

a. accelerating piston in a cylinder
b. projectile moving through a cylinder
c. shock tube
d. start-up and shut-down transients in a wind tunnel
Governing equations for 1D, unsteady flow (ignoring viscous forces, body forces, and area changes):

u
u
0
(12.368)
u 0
continuity:
t
x
x
t x
u
u
1 p
u
t
x
x
momentum:
(12.369)
Assume that the flow is isentropic:

p dp
p p
c2
x
d x
x
(12.370)
c2
Note that in the previous expression, the sound speed, c, has been used because the flow is isentropic.
To make our analysis here look similar in form to the analysis we used while investigating steady, 2D
flows, lets re-write the velocity in the following manner:
u
x
x
where the subscript x signifies a partial differentiation with respect to x. Using this notation and
substituting Eqn. (12.370) into Eqn. (12.369) and simplifying, Eqns. (12.368) and (12.369) become:
(12.371)
x
xx 0
t
x
c 2
(12.372)
xt xxx
0
x
Multiply Eqn. (12.372) with dx and integrate with respect to x (note that Eqn. (12.372) is a function of both
x and t):
2
c
t 1 2 x2 d f t
(12.373)
0
where f(t) is an unknown function of time. Taking the partial derivative of Eqn. (12.373) with respect to t
gives:
c 2
f t
(12.374)
tt xxt
t
We can substitute Eqns. (12.372) and (12.374) into Eqn. (12.371) to give an expression that does not
include the density:
f t tt xxt xxt x2xx c 2xx 0
x2 xx 2xxt tt f t
C. Wassgren
(12.375)
658
The unknown function of time may be eliminated by defining:

f t dt
x x u
t t f t
(12.376)
xx xx
tt tt f t
xt xt
so that Eqn. (12.375) becomes:
c 2 u 2 xx 2uxt tt 0
(12.377)
Note also that Eqns. (12.373) and (12.374) become:

2
c
t 1 2 x2 d 0
(12.378)
tt xxt
c 2
0
t
(12.379)
Now lets examine the nature of Eqn. (12.377) more closely. This 2nd order PDE in two-independent
variables (x and t) will always be hyperbolic since:
B 2 4 AC 4u 2 4 c 2 u 2 1 4c 2 0
where A c 2 u 2 ; B 2u; C 1 .
The slope of the characteristic curves for Eqn. (12.377) can be found using an approach similar to that used
when investigating 2D, steady supersonic flows. We can write the following system of equations:
c 2 u 2 2u 1 xx 0
dx du
dt
0 xt dx where
(12.380)
dx
d f
d
0
dx dt tt dt
Across a characteristic curve, the derivatives of x and t are indeterminate. In order for this to occur, the
determinant of the matrix on the LHS of the previous equation must be zero:
c 2 u 2 2u 1
dx
dt
0 c 2 u 2 dt 2 dx 2 2udxdt 0
dx dt
0
2
dx
dx
2
2
2u c u 0
dt
dt
dx
dt
1
u c and
(12.381)
dt
dx u c
Slope of the characteristic curves in the x-t plane. The + sign represent a right-running
characteristic and the - represents a left-running characteristic.
Since disturbances propagate along characteristic curves, we see that disturbances will propagate at the
speed of sound relative to the local flow velocity.
C. Wassgren
659
In order to ensure that the velocity of the flow will always remain finite, we must also have finite velocity
gradients. The velocity gradient can be found from the system of equations given in Eqn. (12.380) using
Kramers rule:
0 2u 1
0
dx dt
d
dx
dt
d dx dt dt 2udx dt
u
xx
(12.382)
2 t 2
2 x 2
x c u 2u 1 c u dt 2 dx 2 2udxdt
0
dx
dt
0
dx dt
Since along a characteristic curve the denominator of the previous expression is zero, the only way for the
velocity gradient to remain finite is for the numerator to also be zero. Hence, we have the following
relation:
dx dx dt dt 2udx dt 0
dx
du 2u dt
dt
The RHS of the previous equation is found with the aid of Eqn. (12.378):
d
d
udu c 2
dt x dx c 2
(12.383)
Substituting back into Eqn. (12.383) and utilizing Eqn. (12.381):

d
du u c 2u udu c 2
du d 1 dp
d 1 dp
p
(Note:
since c 2
.)
c
c2
s
c2
(12.384)
Conditions along a characteristic curve. The - corresponds to a right-running characteristic

while the + corresponds to a left-running characteristic.
Notes:
1. We could have also solved Eqn. (12.380) for finite xt or tt to arrive at Eqn. (12.384).
2.
The numerical algorithms given during our previous notes on the method of characteristics for 2D,
steady flow may also be applied here.
3.
For an ideal gas, Eqn. (12.384) can be written as:

du d
1 dT
2 dc
c
1 T 1 c
2
dc
(12.385)
1
Integrating along the characteristic curve gives:
2
along a R-running characteristic
1 c
2
u2 u1
(12.386)
c2 c1 or u
1
2 c along a L-running characteristic
1
Conditions along a characteristic curve in a perfect gas. The + corresponds to a rightrunning characteristic while the - corresponds to a left-running characteristic.
du
C. Wassgren
660
4.
We can also determine how properties change across a characteristic curve by considering that by
moving along a right-running characteristic we cross left-running characteristics and vice-versa.
du d 1 dp
(Note: compression dp > 0, expansion dp < 0.)

(12.387)
c
c2
du
2
dc
1
u2 u1
(12.388)
2
(12.389)
c2 c1
1
Conditions across a characteristic curve in a perfect gas. The + corresponds to a rightrunning characteristic while the - corresponds to a left-running characteristic.
5.
A sketch of characteristic curves on the t-x plane looks as follows:

t
left-running
characteristics
uninfluenced
by initial data
right-running
characteristics
influenced
by initial data
influenced
by initial data
uninfluenced
by initial data
6.
R-running
uninfluenced
by initial data
initial data line
Note: The orientation of left

and right running characteristics
are relative to the pathline
direction of fluid particles.
L-running
pathline
For an isentropic flow of a perfect gas we have:

1
p
c2 T2
2
c1 T1
p1
so that Eqn. (12.389) can be written as:
u u
2 c2
2 1
1
c
c
c
1
1
1
1
2
(12.390)
u u
2 p2
2 1
1
(12.391)

c1 c1 1 p1
Conditions across a characteristic curve in a perfect gas. The + corresponds to a rightrunning characteristic while the - corresponds to a left-running characteristic.
7.
The previous relations do not apply to a shock wave. Shock waves are non-isentropic processes and,
hence, the isentropic assumption used in deriving the equations is not valid. In addition, shock waves
travel at speeds greater than the sonic speed.
C. Wassgren
661
Simple Waves
Flows involving simple waves contain either left or right-traveling waves but not both. A simple wave
flow can be produced using an accelerating piston as shown in the figure below:
waves
undisturbed gas
Lets consider two cases of simple waves in a perfect gas using this piston geometry. First well consider a
piston moving to the left and then well consider a piston moving to the right.
Piston Moving to the Left
The movement of the piston can be shown on the t-x diagram. Note that well assume that the piston
velocity increases with time.
t
right-running
characteristics
particle path
piston path
1
u1 c1
region uninfluenced
by the piston
(dead zone)
1
u0 c0
At t=0, a sound wave leaves the piston and travels into the undisturbed fluid. This wave will travel at the
sound speed in the undisturbed fluid, c0, and thus is a straight line on the t-x plot. Across this right-running
characteristic, we have from Eqn. (12.388):
2
2
(12.392)
du
dc (or u
c along a left running characteristic)
1
1
Since du < 0 (the piston moves to the left so the fluid velocity should also move to the left), then dc < 0.
Thus, the next pressure pulse will travel at a slightly slower speed which corresponds to a larger slope on
the t-x plot. Each characteristic curve will, in fact, be a straight line since u and c are constant in the
regions between the pressure waves (refer to Eqn. (12.381)). Thus, the characteristic lines diverge and the
influence of the piston motion is stretched out as the waves propagate downstream.
Notes:
1. The motion of an individual fluid particle, known as a pathline, can be determined using:
dx
u particle
dt particle
C. Wassgren
662
(12.393)
2.
Now lets consider what happens if we accelerate the piston to the left from rest (u0 = 0) to a very large
speed. From Eqn. (12.389) (were crossing right-running characteristics) we have:
umax
2
2
2
c 1 u
u u0
c c0
1
2
1
c
c
c
1
0
0
0
0
Since the piston is moving to the left, u < 0. Thus, the largest speed we can have that results in a nonnegative speed of sound (c 0) is:
umax
2
(12.394)
c0
1
This is known as the escape speed. If the piston continues to accelerate, a vacuum (called the
cavitation zone) will form on the face of the piston as shown in the diagram below.
t
1
1
umax cmin
2c0
vacuum
(cavitation zone)
1/c0
piston path
4.
An impulsive withdrawal to the left results in an expansion fan as shown below:

t
1
u P cP
1
1
1
1
up
uP c0
u p c0
2
2
particle path
expansion fan
uniform flow
region
piston path
1/c0
1/uP
C. Wassgren
663
undisturbed
region
x
Crossing R-running waves:

uP u0 = 2/( 1) (cp c0)
where u0 = 0 so that
cP = c0 + ( 1) up
Piston Moving to the Right

Now lets consider the case where the piston moves to the right to produce compression waves (refer to the
figure shown below).
t
piston path
shock wave
1/c0
x
Across each compression wave (right-running characteristics as shown in the diagram above) we have from
Eqn. (12.388):
2
du
dc
1
where du > 0. Hence, dc > 0 and the characteristic curves become steeper (refer to Eqn. (12.381)).
Eventually these compression waves will intersect and the isentropic assumption breaks down since the
velocity gradients across the wave are no longer infinitesimal and viscous (irreversible) losses become
significant. The point of the first intersection is defined as the start of a shock wave (refer to the previous
figure).
Notes:
1. Consider the diagram shown below where the piston is accelerated in very small, discrete increments
of du each at every time step dt.
shock wave
t
piston path
complex system
of sliplines and
reflected waves
1/c0
dt
x
2.
An impulsive acceleration to the right will immediately form a shock wave as shown in the following
figure.
t
piston path
shock wave
particle path
x
The strength of the shock wave will be such that the fluid velocity behind the shock will equal the
piston velocity.
C. Wassgren
664
Interactions with Boundaries

Stationary Boundaries
Near a stationary wall the fluid velocity must equal zero. This implies that a wave must reflect in a similar
sense as shown in the diagram below. For example, consider a right-running compression wave (dp > 0
and, according to Eqn. (12.387), du > 0) impinging against a stationary wall. In order for the velocity to
remain zero at the wall we must have for the reflected wave (a left-running characteristic), du < 0. Hence,
according to Eqn. (12.387) we see that dp > 0 and thus we have another compression wave.
Note that a fluid particle follows a compression wave and moves away from an expansion wave.
particle path for

expansion wave
reflected
wave
particle path for

compression wave
stationary
u=0
boundary
Across:
R-running: du/c = 1/c2 (dp/)
L-running: -du/c = 1/c2 (dp/)
incident
wave
u=0
x
Free Surface Boundaries

Reflections from an open end are more complicated since we must consider four different cases: the flow
may be either inflow or outflow and it may be either subsonic or supersonic.
Subsonic Outflow/Inflow
For low speed subsonic outflow or inflow, it is reasonable to assume that the pressure at the end of the
duct is equal to the ambient pressure. As a result, waves will reflect in an unlike sense.
For example, consider a right-running compression wave (dp > 0, du > 0) impinging on an open end.
In order for the pressure to remain constant at the open end (and equal to the ambient pressure), we
must have dp < 0 which implies that the reflected wave is an expansion wave. Across the left-running
reflected wave we observe from Eqn. (12.387) that du > 0. A similar approach may be taken to
determine the conditions for an incident expansion wave.
t
t
u>0
reflected
wave (exp.)
free surface
boundary
(p = constant)
reflected
wave (comp.)
incident
wave (comp.)
u<0
free surface
boundary
(p = constant)
incident
wave (exp.)
u=0
particle path
C. Wassgren
u=0
x
particle path
665
For high speed subsonic outflow or inflow, the unsteady effects at the open end must be included
making the analysis much more difficult.
Supersonic Outflow/Inflow
Since for supersonic flow the flow velocity is larger than the propagation velocity of the reflected
waves, the reflected waves are unable to propagate from the open end of the duct. Hence there are no
reflected waves.
C. Wassgren
666
Example:
A diaphragm at the end of a 4 m long pipe containing air at a pressure of 200 kPa and a temperature of 30
C suddenly ruptures causing an expansion wave to propagate down the pipe. Find the velocity at which
the air is discharged from the pipe if the ambient air pressure is 103 kPa. Also find the velocity of the front
and the back of the wave and hence find the time taken for the front of the wave to reach the end of the
pipe.
C. Wassgren
667
23. Description of a Shock Tube
A shock tube is a device consisting, in its simplest form, of a long tube of constant area divided into two
sections by a diaphragm. One of the sections contains a high pressure gas (aka driver gas) while the other
contains a low pressure gas (aka driven gas). A sketch of the device is shown below.
high pressure
low pressure
diaphragm
When the diaphragm between the two sections is broken, either by mechanical means or by increasing the
pressure on the high pressure side and using a scored diaphragm designed to burst at a specified
presssure, a shock wave propagates into the low pressure section and an expansion wave propagates into
the high pressure section. Between the shock wave and the expansion wave is a region of uniform velocity.
expansion
wave
shock
wave
Shock tubes are used:

as an inexpensive, but short duration (usually on the order of milliseconds), wind tunnel,
to study of transient aerodynamic effects,
to study dynamic and thermal response,
to study relaxation effects and reaction rates, and
to generate high enthalpies for studying dissociation and ionization.
C. Wassgren
668
Analysis of the Flow in a Shock Tube
The velocity and pressure behind the shock wave must be equal to the velocity and pressure behind the
expansion wave as shown in the figure below. The temperatures (and densities and entropies) are not
necessarily equal however. If the temperature in the tube is initially uniform, then behind the expansion fan
the temperature will be lower than the initial temperature while behind the shock wave the temperature will
be higher than the initial temperature. The interface separating the two regions is called the contact
surface. The contact surface is the interface dividing the gases that were originally separated by the
diaphragm. Over time, diffusion will cause this interface to spread out.
contact shock
surface wave
expansion
wave
1
diaphragm
location
velocity
pressure
temperature
position
time
current time for drawing

shown above
position
C. Wassgren
669
We know the following about the flow:

1. The gas properties are assumed identical on either side of the diaphragm, i.e. 1 = 2 = 3 = 4 and R1 =
R2 = R3 = R4.
2. The velocities in regions 1 and 4 are zero, i.e. u1 = u4 = 0.
3. The pressures and temperatures in regions 1 and 4 are known, i.e. p1, p4, T1, and T4 are known.
4. The pressure and velocity across the contact surface are equal, i.e. p2 = p3 and u2 = u3.
The remainder of the flow field is most easily analyzed using an iterative procedure as described below.
1. Assume a value for p2 = p3.
2.
The value of u2 can be determined from the pressure ratio, p2/p1, and a relationship developed
previously in our notes concerning 1D, unsteady, and isentropic compressible flow:
1
2
u2
2 p2

1 (crossing L-running waves)
(12.395)
c1 1 p1
where c1 = (RT1)1/2.
3.
Use the normal shock relations to determine MaS and u3:

p3
2
1
Ma 2S
(determine MaS)
p4 1
1
u3 uS
uS
1 Ma 2S
u S
2
u3 uS 1 Ma S 2
velocities are given w/r/t the ground
where uS = MaS(RT4)1/2.
4.
5.
(12.396)
(12.397)
Is u3 = u2? If not, then go back to step 1. If so, then weve correctly determine the pressure and
velocity in regions 2 and 3.
Now that the pressure and velocity in regions 1-4 are known, we can also determine the temperature in
region 2 since we know the speed of sound there:
u
2 c2
2
1
c1 1 c1
T2
c22
R
or
1
T2 p2

T1 p1
We can also determine the temperature in region 3 using the normal shock relations:
2 Ma 2 1
T3
S
2 1 Ma 2S
2
2
T4
1
Ma
6.
The speed of the contact surface can be determined by noting that it has the same velocity as the gas in
regions 2 and 3, i.e. uCS = u2 = u3.
C. Wassgren
670
Example:
A shock tube containing air has a high pressure section at 300 kPa and a low pressure section at 30 kPa.
The temperature of the air is uniform at 15 C. The diaphragm separating the two sections is suddenly
ruptured. Find:
a. the velocity of the air between the shock wave and the expansion wave relative to the ground,
b. the speed of the shock wave relative to the ground, and
c. the speed of the front and back of the expansion fan relative to the ground.
d. Sketch the process on a t-x diagram.
C. Wassgren
671

2010 08 14 NotesOnFluidMechanicsAndGasDynamics Wassgren PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2010 08 14 NotesOnFluidMechanicsAndGasDynamics Wassgren PDF

Uploaded by

Copyright:

Available Formats

Notes on Fluid Mechanics and Gas Dynamics

Carl Wassgren, Ph.D.

Symbolic vs. Numeric Approach

Last Updated: 05 Sep 2008

Symbolic vs. Numeric Calculations

Assume this height is negligible.

Use Newtons 2nd law to determine the acceleration of the projectile:

(10 kg ) ( 9.81 m/s 2 ) = (10 kg ) y

y = ( 4.91 m/s ) t + ( 50 m/s ) t

Last Updated: 05 Sep 2008

Assume this height is negligible.

The time aloft is found by setting y = 0:

Last Updated: 05 Sep 2008

We can conclude from this exercise the following:

Last Updated: 05 Sep 2008

Dimensions and Units

A dimension is a qualitative description of the physical nature of some quantity.

A unit is a quantitative description of a dimension. A unit gives size to a dimension.

Last Updated: 05 Sep 2008

Taylor Series Expansion Approximation

We can use a Taylor series expansion for y about location x:

Last Updated: 05 Sep 2008

Last Updated: 05 Sep 2008

Single Sample Experiments

The least count for the ruler to

Last Updated: 05 Sep 2008

Multiple Sample Experiment

many identical cavities and thermometers

(Note: The variance is 2.)

Last Updated: 05 Sep 2008

95% of measurements fall between x 2

used in most engineering

Last Updated: 05 Sep 2008

( g/cc ) = 0.03 g/cc

Last Updated: 05 Sep 2008

(Note that an = is only strictly true as xn dxn.)

and the relative uncertainty is:

Last Updated: 05 Sep 2008

Last Updated: 05 Sep 2008

Last Updated: 05 Sep 2008

Last Updated: 05 Sep 2008

um = um2 ,C + um2 , A + um2 , p1 + um2 ,T1 + um2 ,p

Last Updated: 05 Sep 2008

Substitute into Eqn. (1.30).

um = uC2 + u A2 + 14 u 2p1 + 14 uT21 + 14 u2p

uC2 = 5.4 *103

The uncertainty in the p1 measurement contributes the most to the uncertainty in m .

Last Updated: 05 Sep 2008

Statistical vs. Continuum Approach

Last Updated: 05 Sep 2008

For air at standard conditions:

2.7 *1019 cm-3 and A 1*1015 cm 2 3.7 *105 cm

Last Updated: 05 Sep 2008

Re Re L (where ReL = VL/ and Re = V/)

To learn more about non-continuum flows, refer to the following texts:

Last Updated: 05 Sep 2008

specific gravity, SG {no units dimensionless}

specific weight, {F/L3} [N/m3, lbf/ft3]

specific volume, v {L3/M} [m3/kg, ft3/slug, ft3/lbm]

In thermodynamics, the specific volume is commonly used in place of density.

Last Updated: 05 Sep 2008

pressure, p {F/L2} [Pa=N/m2, psf=lbf/ft2, psi=lbf/in2]

unit normal vector

2.7 1019 cm-3 and A 11015 cm 2 3.7 *105 cm