Professional Documents
Culture Documents
Gas Dynamic
Gas Dynamic
net/publication/307206748
CITATIONS READS
0 3,732
3 authors:
D. M. J. Smeulders
Eindhoven University of Technology
136 PUBLICATIONS 1,209 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
1) Micro-porous plates (TANGO) 2) Pneumatic sound source (LAUM) 3) Vortex nozzle interaction (VKI) View project
All content following this page was uploaded by Avraham Hirschberg on 29 August 2016.
26 August 2016
Contents
1 Introduction 2
2 Basic Equations 7
2.1 Conservation Laws . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Constitutive equations . . . . . . . . . . . . . . . . . . . . . . 8
2.3 Conservation laws in differential form . . . . . . . . . . . . . 9
2.4 Thermodynamic relations . . . . . . . . . . . . . . . . . . . . 9
2.5 Simplification of the laws of conservation;
Euler equations . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.6 The Bernoulli equation . . . . . . . . . . . . . . . . . . . . . . 18
2.7 Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.8 Answers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
i
3.6 Answers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
1
Chapter 1
Introduction
The most spectacular phenomena in gas dynamics are related to the ratio
of the flow velocity V and the speed of sound c. The speed of sound c is
the velocity of propagation of a small pressure perturbation relative to the
gas. This corresponds to the velocity of the propagation of information in
the flow. The ratio M = V /c is called the Mach number. A flow in which
M < 1 is called subsonic. A flow with M > 1 is called supersonic. A
subsonic flow approaching an obstacle will be warned by acoustic waves and
will smoothly flow around the obstacle. A supersonic flow will not receive
warnings, because the acoustic waves are washed away by the flow. The flow
will collide on the obstacle forming a shock wave. In figure 1.1 we illustrate
the difference between a subsonic and a supersonic flow. As we will see later
compressibility effects become dominant for M ≥ 1.
The phenomenon corresponds to a fundamental change in the type of the
differential equations describing the flow. Subsonic flow are described by
so called elliptical differential equations. Supersonic flows are described by
hyperbolical differential equations. The same mathematics and therefore
physics is found in many other phenomena. For example the flow of shallow
water and car traffic. In both cases when the flow travel faster than the
information one can observe discontinuities. In shallow water, this is called
a water jump (see figure 1.2). In car traffic this can be an accident due
to fog. In gas flows the existence of such discontinuities was controversial
until Ernst Mach obtained a photograph of such shock waves (see figure 1.3).
2
Figure 1.1: Spherical waves generated by point sources in a) a subsonic flow
M < 1 and in b) a supersonic flow M > 1. In the subsonic flow the waves
travel upstream and can warn the flow for the presence of an obstacle. In
the supersonic flow the waves are convected away and only reach a lim-
ited conical region of space (Mach cone with opening angle 2 arcsin (c/V ) )
downstream of the origin of the perturbation.
ν ∼ cλ̄ (1.1)
where λ̄ is the mean free path of molecules. This is the average distance
which molecules travel between two collisions. The speed of sound is used
here as an estimate for the average thermal random velocity of the molecules,
which takes care of the viscous transport of momentum within the gas.
Hence we see that [Landau (1987)]:
L
Re ∼ M (1.2)
λ̄
For compressible flows M = O(1) so that we conclude that if we use a con-
tinuum hypothesis L/λ̄ >> 1 the Reynolds number of the flow is very large
Re >> 1. Considering low Reynolds numbers at high Mach numbers in-
volves rarefied gas effects (Knudsen effects) such as the failure of the no-slip
boundary condition at the walls. Also the assumption of local thermody-
namic equilibrium will fail. The velocity distribution of molecules is not
necessarily described by one single temperature T . In these lecture notes we
ignore such effects, which are essential in vacuum and space technology. We
focus on large Reynolds number flows, for which the fluid can be described
as a continuum with locally defined thermodynamic state variables such as
the pressure p, temperature T and the density ρ.1
1
We always consider absolute pressures and absolute temperatures.
3
Figure 1.2: Shallow water in a kitchen sink displaying a circular water jump.
Within the circle the water flow is supercritical. The flow velocity is larger
than the propagation speed of small surface waves. We observe Mach cones
on the water surface. The jump is an abrupt transition to subcritical flow.
4
Figure 1.3: (a) Visualization of Mach waves around a bullet in supersonic
flight obtained by Ernst Mach. The density changes are made visible by
the refraction of light, due to the large density change (schlieren method)
[Settles (2001)]. For an observer moving with the bullet the approaching
flow is supersonic. (b) Bullet shot by a rifle. Schlieren photograph by Gary
Settles.
The understanding of gas dynamics has been developed mainly in the nine-
teenth century by: Poisson (1808) non-linear plane waves, Riemann (1860)
method of characteristics, Rankine (1870) and Hugoniot (1887), Mach (1887)
shock-waves, Prandtl and Meyer (1908) 2-D steady gas dynamics. The sig-
nificance of gas dynamics has dramatically increased in the twentieth century
with the large scale use of gas turbines for electricity production, combustion
car engines and aircraft engines. In the second half of the twentieth century
the space exploration and military applications have become important (su-
personic aircraft, rockets, scramjets). Also turbocompressors are currently
used in modern cars. Supersonic flows are also found in torches used for
welding and plasma surface treatment. Also gas dynamics and shock waves
are essential in astronomy to understand the evolution of stars and galaxies.
As indicated above many non-linear wave phenomena are described by the
same mathematical tools as gas dynamic flows, such as: tsunami waves, car
traffic waves, compositional waves in chromatography, water breakthrough
in secondary oil production, kidney stone destruction by ultrasound waves.
The formation of shock waves explains also simple daily phenomena such as
the whip crack, the sound of hand clapping, the sound of thunder, the tongue
clap and the brassy sound of trombones and trumpets [Hirschberg (1996)].
In these lecture notes we will limit ourselves to a very modest study of a few
basic flow phenomena.
Chapter 2 describes the basic equations of gas flows starting from the mass,
momentum and energy integral balance equations. The differential form of
these equations and constitutive equations are then discussed. Thermody-
namics is refreshed, which is needed to obtain relations between basic state
variables such as pressure, density, temperature, speed of sound, enthalpy
5
and entropy. We then focus on the thermal properties of ideal gases and de-
fine perfect gases. Simplified equations for frictionless flows are introduced.
In particular we discuss the equation of Bernoulli for steady frictionless flows.
At the end of each chapter we provide questions with detailed answers. Ba-
sic questions can be used to study the theory. Additional question marked
with a star are useful for students looking for a deeper understanding of the
theory. At the end of the lecture notes a collection of old exam questions is
provided.
In the list of references at the end of these lecture notes we give some refer-
ences to lecture notes and textbooks. The student can find is these excellent
books complementary information or alternative explanations.
These lecture notes are a translation of the original lecture notes of M.E.H.
van Dongen (published in 2008), with some modifications and the addition
of exercises.
6
Chapter 2
Basic Equations
n
F
v
dA
B
7
forces. This leads to two terms on the right hand side of the equation,
describing momentum production:
" " " "
d
ρv dτ + ρv(v · n) dA = ρg dτ + t dA, (2.2)
dt B F B F
with ρg a body force (e.g., gravity) and t a surface force (e.g., caused by
friction or pressure).
For the same volume B we can also consider the energy of the fluid per unit
mass, which comprises kinetic energy (1/2V 2 ) and internal energy (e) per
unit of mass. This energy can flow through F and can be changed by the
work of body forces, surface force, by transport and production of heat:
" # $ " # $
d 1 2 1 2
V + e ρ dτ + V + e ρ(v · n) dA =
dt B 2 F 2
" " " "
−(q · n) dA + ρ(g · v) dτ + (t · v) dA + φE dτ (2.3)
F B F B
with q the heat flow density and V 2 = |v|2 . The rate of energy production
per unit volume φE includes any process not accounted for in the definition
of e. This could be chemical or nuclear reactions, or electrical heating.
Equation (3.39) is a generalization of the first law of thermodynamics to
open systems.
t = n · S ; S = −pI + S ′ , (2.4)
with p the mechanical pressure, I the unit tensor and S ′ the deviatoric stress
tensor. For a Newtonian fluid it holds that:
S ′ = 2ηD + η ′ (∇ · v)I
1
= 2η(D − (∇ · v)I) + κ(∇ · v)I. (2.5)
3
For such a fluid, the stress tensor S ′ depends linearly on the deformation
% &
∂vi ∂vj
tensor D. This deformation tensor is defined as [D]ij = 12 ∂x j
+ ∂xi with
i = {1, 2, 3} and vi the components of the velocity v. η is the dynamic
viscosity, η ′ is the second viscosity coefficient and κ is the so-called bulk
viscosity. They are related by:
8
2
κ = η ′ + η. (2.6)
3
The use of the bulk viscosity instead of the second viscosity coefficient is
preferred, because of its direct relation to the compressibility of the fluid,
and the fact that it is always positive and vanishes for noble gasses (He, Ne,
Ar, Xe, Kr).
The heat flow density q satisfies Fourier’s Law:
q = −λ∇T, (2.7)
in which λ is the heat conduction coefficient and the temperature T will be
defined in section 2.4 where we discuss the thermal equation of state. This
linear relationship between q and ∇T is consistent with Newton’s approxi-
mation (2.5) for the stress tensor.
9
with the additional variable s the specific entropy (entropy per unit mass).
Indeed, from the fundamental thermodynamical law
1
de = T ds − pd (Gibbs’ relation) (2.12)
ρ
and the differential form of (2.11), we find that:
# $ ( ) # $
∂e ∂e 1
de = ds + d , (2.13)
∂s ρ ∂ ρ1 ρ
s
so that we immediately find the thermodynamic definitions of T and p:
# $
∂e
T ≡ , (2.14)
∂s ρ
and ( )
∂e
p≡− . (2.15)
∂ ρ1
s
The set of equations (2.8) - (2.10), (2.11), complemented with (2.14) and
(2.15) now contains 8 equations with 8 variables.2
We will now discuss a number of important terms and definitions we will
encounter during the study of gas dynamics.
Enthalpy
The enthalpy is defined as:
p
h≡e+ . (2.16)
ρ
Substitution into (2.12) gives that:
1
dh = T ds + dp. (2.17)
ρ
Specific heat
The specific heats cv and cp at constant volume and constant pressure,
respectively, are given by3 :
# $ # $ # $ # $
δQ ∂e δQ ∂h
cv = = ; cp = = . (2.18)
δT ρ ∂T ρ δT p ∂T p
the so-called ”local” thermodynamic equilibrium assumption, which is consistent with the
choice of Newton’s approximation (2.5) for the viscous stress and Fourier’s law (2.7) for
the heat flux, assuming that fluxes are proportional to the gradients in the flow.
2
As explained by [Thompson (1972)] the hydrodynamic definition (2.15) of the pressure
is the only useful definition of pressure.
3
We use δQ/δT to stress the fact that Q is not a state variable
10
where δQ is the amount of heat added to the system and δT the correspond-
ing increase in temperature. We have following the first law of thermody-
namics δQ = de + pd(1/ρ).
The Poisson ratio γ is defined by: γ ≡ cp /cv . The specific heats cv and cp
correspond to the amount of heat one should supply to increase the temper-
ature of one kilogram of gas by one degree, at respectively constant volume
and constant pressure.
Speed of sound
The thermodynamic property c is defined as:
# $
2 ∂p
c = , (2.19)
∂ρ s
Later is will become clear that the property c is the propagation speed of
sound waves (speed of sound).
Ideal gas
At low densities, gases are found to obey closely the Boyle-Gay Lussac equa-
tion of state of the form:
p = ρRT (2.21)
where R is the specific gas constant: R = R0 /M ; with R0 the universal gas
constant (8, 314kJ kmol−1 K−1 ) and M the molar mass (kg kmol−1 ) of the
gas. An ideal gas is defined as a gas which obeys the above equation of
state. By substituting (2.21) into (2.19) it follows directly that
c2 = γRT. (2.22)
For an ideal gas, it can be shown that the internal energy and enthalpy are
functions of temperature only: e = e(T ) and h = h(T ), so that de = cv dT
and dh = cp dT with cv (T ) and cp (T ). By definition of h (2.16), the difference
between the specific heats is:
# $
p
dh − de = (cp − cv )dT = d = RdT
ρ
hence:
cp − cv = R (2.23)
11
Perfect gas
A calorically perfect gas is an ideal gas with constant specific heat. As the
specific heat is independent of temperature, we find:
p 1 c2
e = cv T = = ;
ργ−1 γ(γ − 1)
p γ c2
h = cp T = =
ργ−1 γ−1
and using the ideal gas law (2.21), the isentropic gas relation becomes:
# $ # $γ # $γ/(γ−1) # $(2γ)/(γ−1)
p ρ T c
= = = . (2.26)
p0 ρ0 T0 c0
Internal energy
The specific internal energy of a molecule is determined by the number of
degrees of freedom available to store energy:
1
e = cv T = f RT , f = ntr + nrot + 2nvibr + nel (2.27)
2
with:
ntr the number of translational degrees of freedom, each contributing
1
2 RT to the internal energy;
12
nrot the rotational degrees of freedom, contributing 12 RT each when the
temperature is above the characteristic rotation temperature, θr ;
nvibr the vibrational degrees of freedom, contributing RT each when the
temperature is above the characteristic vibration temperature, θv . More
complex molecules have a θv for every vibrational mode;
nel takes into account the electron excitation energy, when the temper-
ature reaches θel .
Figure 2.2 shows how the specific heat cv depends on temperature for hy-
drogen.4 Note the temperature ranges in which cv is constant. This is where
the gas behaves ’perfectly’. In between are regions where the gas gradually
excites new degrees of freedom.
13
With γ = cp /cv , we have
f +2
γ= (2.28)
f
for Poisson’s ratio. At room temperature, linear molecules like O2 , N2 , CO,
CO2 have 2 rotational degrees of freedom while the vibrational mode is
frozen, leading to γ = 7/5.
14
Figure 2.3: cv /R vs. T for hydrogen. ([Owczarek (1964)])
Relaxation time
A system does not instantaneously achieve a new equilibrium state. It takes
a certain time for the system to adapt to changes in pressure, temperature,
etc; especially when these changes occur very fast. The lagging of internal
processes such as dissociation, ionization, evaporation, chemical reaction
and transfer of energy between molecular modes (translation, rotation, vi-
bration) is called a relaxation process. It is possible to define a relaxation
time, which is a measure of the time required for a system τ to reach equilib-
rium after a disturbance. When the time scale of the process one considers
is small, compared to τ or of the order of τ , the equilibrium equation of
state is no longer valid.
Table 2.2 shows for some common gases the relaxation times for the different
molecular processes ([Thompson (1972)]).
The difference in relaxation time for translation and rotation of air molecules
largely determine the volume viscosity of air. This explains the dissipa-
15
Gas mode T [K] τ [s]
N2 translation 300 1.6 · 10−10
N2 rotation 300 1.2 · 10−9
N2 vibration 3000 3.0 · 10−5
H2 O vibration 486 3.7 · 10−8
O2 dissociation 3100 2.3 · 10−5
De D ρ1 1
+p = − ∇ · q + (viscous dissipation) (2.29)
Dt Dt ρ * +, -
>0
16
to obtain:
Ds q 1
ρ + ∇ · = − 2 (q · ∇T ) + ρ(viscous dissipation), (2.33)
Dt T * T +, -
>0
or simply:
ds q
ρ +∇· ≥0 (2.34)
dt T
This relation can be seen as the differential formulation of the second law of
thermodynamics. By using conservation of mass and Gauss’ law, we obtain
the integral version of (2.34):
" " "
d 1
ρs dτ + ρs(v · n) dA + (q · n) dA ≥ 0. (2.35)
dt B F F T
17
Euler equations:
∂ρ
+ ∇ · ρv = 0, (2.38)
∂t
Dv
ρ = −∇p, (2.39)
Dt
Ds
ρT = 0, (2.40)
Dt
where we neglected gravity as well. The last equation, conservation of en-
ergy, follows from the relation between internal energy and enthalpy, (2.16):
De D ρ1 Dh Dp Ds
ρ + pρ =ρ − = ρT
Dt Dt Dt Dt Dt
18
b Bernoulli, in case of a stationary, homentropic (uniform entropy), rotation-
free flow: ∇s = 0 and ∇ × v = 0. Then:
1 2
V + h = h0 (homo-energetic, valid in the entire flow field) (2.43)
2
c Bernoulli, in case of an instationary, homentropic, rotation-free flow:
∇s = 0 and with ∇ × v = 0. Because ∇ × ⃗v = 0 there exists a scalar
potential φ such that v = ∇φ. Substitution into equation (2.41) gives:
∂φ 1 2
+ V + h = C(t) (entire flow field) (2.44)
∂t 2
d Bernoulli in differential form. From (2.43) it follows that:
1
V dV + dh = V dV + dp = 0. (2.45)
ρ
Stationary Bernoulli equation for a perfect gas
For a perfect gas, by making use of h = cp T ,
1 2
V + cp T = cp T0 ; (2.46)
2
and c2 = γRT, R = cp − cv and γ = cp /cv (2.22 - 2.23), we obtain:
T0 γ−1 2
=1+ M , (2.47)
T 2
where M is the Mach number: M = Vc . Starting with the same Bernoulli
equation, it is also possible by using (2.26) to derive expressions containing p,
ρ and c. The quantities T0 , p0 , ρ0 , c0 are the “total” temperature, -pressure,
-density and speed of sound. They are the magnitudes of temperature,
pressure, etc. that are reached when the fluid is decelerated to V = 0 by
means of an hypothetical isentropic, and quasi-steady process.7 They can be
different from the real values Ts , ps , ρs , cs in a stagnation point. In general
the flow around a stagnation point is not isentropic nor frictionless.
In practice a supersonic flow originates from a settling chamber. The state
in this reservoir where the velocity is negligible is called the reservoir state
pc , ρc and Tc (with the index c stands for settling “chamber”).
From (2.47) we derive expressions relating the density, pressure, speed of
sound by using the isentropic relations for a perfect gas (2.26). Thanks to
this we have now for given total conditions T0 , ρ0 , ... expressed T, ρ, ... as a
function of the Mach number M .
From (2.46) it follows that for a given T0 , there exists a maximum achievable
velocity: . .
Vmax = 2cp T0 = 2h0 (2.48)
which can be reached in a steady flow.
7
a quasi-steady isentropic process is by definition frictionless
19
The critical state
The temperature, pressure and speed of sound at which the Mach number
is equal to 1 are referred to as the critical temperature T ∗ , critical pressure
p∗ and critical speed of sound c∗ . This is a so-called sonic condition, where
the velocity is equal to the speed of sound. Starting with (2.47) and taking
M = 1, we obtain for a steady isentropic flow:
T∗ 2
= . (2.49)
T0 γ+1
Conditions for p∗ , ρ∗ and c∗ can also be found using (2.49) in combination
with (2.26).
c
M=1
c0 subsonic
transsonic
c* sup
ers
on
ic
V* Vmax V
We observe from figure 2.4 that for low subsonic flows (M ≤ 0.2) the speed
of sound is almost constant hence the flow is almost isothermal. Also the
density changes remain small so that the flow is in good approximation
“incompressible”. For very large Mach number (M ≥ 5), in √ the so called
hypersonic region, the flow velocity is very close to Vmax = 2h0 . Changes
in the Mach number M = V /c are dominated by the variation of the speed
of sound and associated with a strong density variation. Hypersonic flows
are associated with spectacular variations in temperature across the flow
field (between the free flow and a stagnation point for example).
20
For some applications an alternative Mach number M ∗ is used instead of
M . This Mach number is defined as:
V
M∗ = (2.50)
c∗
where c∗ = V ∗ is the critical flow speed, which is reached when the flow
is isentropically and steadily accelerated/decelerated to M = 1. As M ∗ is
proportional to V it is in the finite range:
0 ≤ M < ∞. (2.52)
2.7 Questions
Questions with a star (*) are more advanced and not considered as basic
study material.
2. How is the unit vector ⃗n determined when applying the integral equa-
tions (eq. 2.1, 2.2 and 2.3)?
3. Why is the bulk viscosity more convenient than the second viscosity?
4. What are the main additional limitations to the validity of the dif-
ferential form of the conservation equations (eq. 2.8, 2.9 and 2.10)
compared to the integral formulation (eq. 2.1, 2.2 and 2.3)?
21
* 10. Show that: # $ / # $ 0
∂e 1 ∂p
= 2 p−T
∂ρ T ρ ∂T ρ
and verify that the internal energy of an ideal gas is independent of
its density.
cp − cv = R
13. Starting from the Gibbs’ relation, show that for an ideal gas:
γp
c2 =
ρ
15. Starting from Gibbs’ relation show that for a perfect gas:
# $γ # $
p ρ s − s0
= exp
po ρ0 cv
18. How many degrees of freedom has a molecule with N atoms when
we assume that electronic states are not excited. How many of these
degrees of freedom are vibrational?
* 20. Starting from the energy equation (2.10) derive the entropy equation
(2.40).
22
21. Using the integral energy equation (2.3) applied to the adiabatic flow
through a segment of a pipe one can show that the total enthalpy is
conserved:
V2 V2
h0 = h1 + 1 = h2 + 2
2 2
What are the differences between this equation and the equation of
Bernoulli:
V2
h0 = h + ?
2
22. Derive equation:
T0 /T = 1 + (γ − 1)M 2 /2
Specify the assumptions you need.
27. What are the differences between the reservoir pressure, the total pres-
sure and the stagnation pressure in a flow?
28. An aircraft flies at a typical Mach number M = 0.8. Estimate the tem-
perature difference between the stagnation point and the surrounding
air. Same question for the Blackbird with M = 3, a scramjet M = 7
and Apollo during re-entry in the atmosphere M = 24. Assume air to
be a perfect gas. Will this assumption be realistic. Provide arguments.
2.8 Answers
1. The control volume B is chosen by the user. Usually it is chosen for
its convenience.
23
4. The flow is described by continuous fields that allow differentiation.
The integral formulations allow discontinuous fields as found in shocks
or contact surfaces (which we will discuss later).
T ds = de + pd(1/ρ) ⇒ de = T ds − pd(1/ρ)
and: # $ # $
∂f ∂f
= −s ; = (p/ρ2 )
∂T ρ ∂ρ T
Furthermore:
( # $ ) # # $ $
∂ ∂f ∂ ∂f
=
∂ρ ∂T ρ ∂T ∂ρ T ρ
T
implying: # $ # $
∂s ∂(p/ρ2 )
=−
∂ρ T ∂T ρ
24
10.
p
T ds = de + pd(1/ρ) ⇒ de = T ds − pd(1/ρ) = T ds + dρ
ρ2
# $ # $
∂s ∂s
ds = dρ + dT
∂ρ T ∂T ρ
⇒
1 # $ 2 # $
∂s p ∂s
de = T + dρ + T dT
∂ρ T ρ2 ∂T ρ
From question 1.7 we know that:
# $ # $ # $
∂(p/ρ2 ) 1 ∂p ∂s
= 2 =−
∂T ρ ρ ∂T ρ ∂ρ T
Hence: # $ / # $ 0
∂e 1 ∂p
= 2 p−T
∂ρ T ρ ∂T ρ
12. # $
1
dQ = de + pd
ρ
# $ # $
δQ ∂e
cv ≡ =
δT ρ ∂T ρ
1
dQ = dh − dp
ρ
p
h≡e+
ρ
# $ # $ # $ # $
δQ ∂h ∂e ∂(p/ρ)
cp ≡ = = +
δT p ∂T p ∂T ρ ∂T ρ
25
13.
T ds = de + pd(1/ρ)
Isentropic and ideal gas:
# $
∂e p dρ dT R dρ
0= dT − 2 dρ = cv dT − RT ⇒ =
∂T ρ ρ ρ T cv ρ
and
1 2 # $
dp dT dρ cp − cv dρ cp dρ ∂p cp p
= + = 1+ = ⇒ =
p T ρ cv ρ cv ρ ∂ρ s cv ρ
26
18. In 3-D space we need 3 coordinates to determine the position of each
atom. Hence the total number of degrees of freedom of the molecule
is 3N . The number of vibrational degrees of freedom is:
ntr = 3
20.
De D(1/ρ)
ρ + pρ = ∇ · (λ∇T ) + η ′ (∇ · ⃗v )2 + 2η(D : D)
Dt Dt
Reversible process, Gibbs’ relation:
T ds = de − pd(1/ρ)
Yields:
Ds
ρT = ∇ · (λ∇T ) + η ′ (∇ · ⃗v )2 + 2η(D : D)
Dt
No friction nor heat transfer:
Ds
ρT =0
Dt
27
account so that the flow is not isentropic. Due to the no-slip condition
at the wall the friction forces do not carry any work. The flow is steady
but not uniform. The velocity term V 2 /2 is the average of this term
over the pipe cross section (not a local value).
V12 V2
h0 = h1 + = h2 + 2
2 2
For a perfect gas we have:
h = cp T
For an ideal gas we have furthermore:
R = cp − cv
And:
c2 = γRT
Hence:
V2
cp T0 = cp T +
2
# $2
T0 V2 γR V cp c p − cv 2
=1+ =1+ =1+ M
T 2cp T 2cp c cv 2cp
T0 γ−1 γ+1
∗
=1+ =
T 2 2
We have furthermore:
# $γ # $γ
p ρ p T0
= =
p0 ρ0 p0 T
# $ γ # $ γ
p∗ T∗ γ−1 2 γ−1
= =
p0 T0 γ+1
24. At M = ∞ the gas has reached the absolute zero (T = 0K), which
is impossible following the third law of thermodynamics. This means
that all the random kinetic energy (translation, rotation and vibra-
tion) of the molecules in the reservoir has been transformed into a
macroscopic translation kinetic energy. The density is vanishing small
so that we have a kind of “vacuum” in which there is no collision
between molecules.
28
25. For a perfect gas:
V2
cp T0 = cp T +
2
T0 γ−1 2
=1+ M
T 2
# $ γ ( ) γ
γ−1
p T γ−1 1
= =
p0 T0 1 + γ−1
2 M
2
Taylor expansion:
# $ γ
p0 γ−1 2 γ−1
= 1+ M
p 2
# $# $
γ−1 γ 1 γ γ γ−1 2 2
≈1+ M2 + −1 M ...
2 γ−1 2γ−1 γ−1 2
# $
γ 2 1 2
= 1 + M 1 + M + ...
2 4
# $
2 V2 V2 γ−1 2
M = = 1+ M
γRT γRT0 2
# # $ $
p0 − p γ V2 γ−1 1 V2
= 1+ + + ...
p 2 γRT 2 4 γRT0
# $# $
p0 − p ρ M2
1 2
= 1+ + ...
2 ρ0 V
ρ0 4
# $ −1
ρ γ − 1 2 γ−1 M2
= 1+ M =1− + ...
ρ0 2 2
# $
p0 − p M2
1 2
= 1− + ...
2 ρ0 V
4
26. We look for the critical velocity as a function of the reservoir condi-
tions.
. .
V ∗ = c∗ = c0 (c∗ /c0 ) = c0 T ∗ /T0 = 2γRT0 /(γ + 1)
29
which are reached in a stagnation point. As the flow near a stagna-
tion point is not necessarily isentropic and reversible, the stagnation
pressure deviates in general from the total pressure and temperature.
In the particular case of a Pitot tube placed in a supersonic flow, the
stagnation pressure will be lower than the total pressure upstream of
the tube, because of dissipation on a shock wave in front of the tube
(see chapter 3). Furthermore the heat conduction along the tube and
radiation heat transfer will result into a temperature at the stagna-
tion point which is different from the total temperature upstream of
the tube.
28.
γ = 1.4
γ−1 2
(T0 /T ) = 1 + M = 1 + 0.2M 2
2
T = 300K
M = 0.8 ⇒ (T0 /T ) = 1.128 ⇒ T0 − T = 38.4K
M = 3 ⇒ (T0 /T ) = 2.8 ⇒ T0 − T = 540K
M = 7 ⇒ (T0 /T ) = 10.8 ⇒ T0 − T = 2940K
M = 24 ⇒ (T0 /T ) = 116.2 ⇒ T0 − T = 34560K
Note that the temperature in the air varies strongly depending on the
altitude from 250K (11 km) up to 450K (20 km).
Obviously an aircraft flying at M = 3 cannot be build from aluminium.
One uses titanium.
At M = 7 the theory fails (perfect gas) because air starts dissociating.
A classical rocket has to carry for a given mass of hydrogen eight
time the mass in oxygen. A scram-jet uses the oxygen from the air
for its propulsion. This allows to reach the critical escape velocity
(from the earth gravitation, about M = 7) with a much lighter engine
than a classical rocket. Until most scram-jet experiments failed due
to thermal problems.
At M = 24 the theory fails (perfect gas) because the gas is dissociated
and ionized. Above a temperature of the order of 12 000 K radiation
heat transfer becomes very important. In general there are strong
deviations from local thermodynamic equilibrium. The electrons will
have another temperature than atoms and ions.
30
Chapter 3
Stationary Quasi-1D
Gas Dynamics
Consider the situation in figure 3.1 above, a so-called Laval tube (or nozzle):
a perfect gas flows steadily from a reservoir with pressure p0 and entropy s0
through a converging channel followed by a diverging channel after passing
a throat of section Ak . We assume there is no friction (Re >> 1). The
diameter is assumed to change slowly with x, so that it holds that:
L dA
<< 1
A dx
where L is the nozzle length. This allows us to regard this flow as quasi-
one-dimensional : the flow is in good approximation parallel to the walls
31
and only a function of x: ⃗v = (V, 0, 0) with V = V (x). We assume that this
steady frictionless isentropic flow emerges from a uniform reservoir. Hence
all the flow properties are uniform in a cross section: p = p(x), T = T (x),
and ρ = ρ(x). From the conservation of mass (2.1) we have:
d
(ρV A) = 0,
dx
Bernoulli’s equation (2.45) can be written in differential form:
dV 1 dp
V + = 0,
dx ρ dx
and from the definition of the speed of sound c (2.19) we have for this
isentropic flow:
dp dρ
= c2 .
dx dx
It follows after some algebra that:
1 dV 1 1 dA
=− 2
. (3.1)
V dx 1 − M A dx
Concerning M , there are three possible scenarios:
In this last case, in order to determine the acceleration dV /dx at the throat,
we look at the second derivative of A(x). To find an expression for dV /dx
when dA/dx = 0 and M = 1 we use de l’Hôpital’s rule:
32
3 d
4 1 dA 5 33 3
1 d2 A 3
1 dV 33 − dx A dx 3 A dx2 3
= d 3 = 3 .
V dx 3throat dx (1 − M 2) 3 2M dM 3
dx throat
throat
Rearranging terms and using c = V /M and M = 1 we obtain:
2 dV dM 1 d2 A
= (3.2)
c dx dx A dx2
2
Because dV dM d A
dx and dx always have the same sign, dx2 will always be positive.
Indeed, the Mach number can pass through unity only at an area minimum
or throat. Hence we need a Laval nozzle to reach supersonic flows by a
steady expansion from a high pressure reservoir.
Next we look at the mass flux through the nozzle. This will provide us with
a less mathematical explanation for the fact that M = 1 can only be reached
at the throat of the Laval nozzle.
ρV
ρ∗ V ∗ 1
0
0 1 2 3 4
M
Figure 3.2: Normalized mass flux density as a function of the Mach number.
33
of the mass flux, the maximum mass flow ρ∗ V ∗ Ak through the nozzle oc-
curring when M = 1 at the throat is reached. We then say that the nozzle
is “choked” or “critical”.
Using the isentropic gas relations in combination with the equation of Bernoulli
we can express the ratios p/p0 , T /T0 , ρ/ρ0 ...as functions of the Mach num-
ber M (2.47 and 2.26). Hence given a nozzle geometry and the reservoir
conditions we can determine the state of the gas along the nozzle.
An alternative for the Mach number M is the ratio of the velocity and
the critical velocity M ∗ = V /V ∗ = V /c∗ (2.50). The advantage of this
normalized
. velocity is that it has a finite range 0 ≤ M ∗ < Vmax /c∗ =
(γ + 1)/(γ − 1) while for M = 1 we have M ∗ = 1. In terms of the
normalized velocity M ∗ we find:
# $1/(γ−1)
ρV γ + 1 γ − 1 ∗2
= M∗ − M , (3.4)
ρ∗ V ∗ 2 2
where M ∗ ≡ V /c∗ . Because c∗ is a constant along a stream line, the right
hand side is now a function of V only.
This formula is plotted in figure 3.3. From this graph we see that, as the
velocity increases along a stream line, the mass flux increases as long as
the flow remains subsonic. For low Mach numbers, this curve rises as a
straight line, corresponding to an incompressible flow. In the supersonic
range however, the mass flux density diminishes for increasing velocity and
vanishes when V → Vmax .
The flux has its maximum value ρ∗ V ∗ when M ∗ = 1 (at the critical velocity).
34
ρV
leb
ρ*V*
ssi
pre
1.0
om
inc
0.0
0.0 1.0 2.0 3.0 4.0
M*=V/c*
a flow would follow the sharp angle wall at the exit of the pipe and would
expand radially in all directions. Such a flow is never observed in practice.
In reality, due to friction losses, the flow will not be able to follow the
sharp edges at the outlet. It separates tangentially from the wall, as shown
in figure 3.3. A shear layer is formed between the main jet flow and the
surrounding stagnant gas. Across this shear layer, pressure and velocity
must be continuous.
a) b)
pe pe
pa
pa
Figure 3.4: (a) Potential flow, (b) Separated discharge in a subsonic flow
(from [Thompson (1972)]).
35
The discharge of fluid into a reservoir may either be subsonic or super-
sonic:
Figure 3.5 shows the pressure distribution p(x)/p0 of the flow in the nozzle
and just outside the nozzle, for different values of pa /p0 . In case a, where
1
Let us assume a subsonic flow. If one assumes that pk > pa the gas will expand as it
leaves the nozzle to adapt to the surrounding pressure pa . Hence the streamlines should
diverge following the continuity equation applied to a steady flow ⃗v · ∇ρ = −ρ∇ · ⃗v , which
demonstrates that ∇ · ⃗v > 0 when ⃗v · ∇ρ < 0. Following our analysis (1/V )(dV /dx) =
(1/A)(dA/dx)/(M 2 − 1) (equation 3.1). As sign[(1/A)(dA/dx)] = sign[∇ · ⃗v ] we conclude
that the flow velocity should decrease. Following Bernoulli a decrease in velocity implies an
increase in pressure. Hence we come to a contradiction. We assumed that the pressure in
the jet was decreasing in order to adjust itself to atmospheric pressure and we come to the
conclusion that the pressure is increasing. Obviously the same conclusion can be deduced
for a steady subsonic flow with pk < pa . Hence pk = pa should prevail. This boundary
condition is actually a dynamic boundary condition. When a pressure disturbance arrives
at the outlet it is reflected as an inverse pressure perturbation travelling as a sound wave
to the reservoir and adjusting the mass flow to match the boundary condition pk = pa .
2
These waves will meet at the center of the flow. In first approximation they add up,
resulting into an over expansion or compression on the jet center line. Further down-
stream these waves reflect on the shear layer, resulting into an inversion of the wave sign.
An expansion (compression) wave reflects as a compression (expansion). The process is
then repeated, resulting into an oscillating pressure along the jet center-line. A further
discussion of this will be provided later because this flow is not one-dimensional and due
to non-linearity shock waves can be formed.
36
Figure 3.5: Pressure distributions in a converging nozzle for differ-
ent ratios pa /p0 of outlet pressure pa to reservoir pressure p0 (from
[Thompson (1972)]).
37
Figure 3.6: Pressure distributions in a Laval nozzle for different ratios pa /p0
of outlet pressure pa to reservoir pressure p0 (from [Thompson (1972)]).
Laval nozzle, i.e., a converging- diverging nozzle with throat cross section Ak
and outlet (exit) cross section Ae . Figure 3.6 shows the geometry and cor-
responding pressure distribution p(x)/p0 inside the nozzle. Again, lowering
pa /p0 by increasing p0 will increase the mass flow through the throat, until
the critical condition is reached Φmax (equation 3.5) . Consider the cases a
through g, where we vary the ratio of ambient to reservoir pressures pa /p0 .
At a, pa = p0 and there is no flow. In case b, the flow is entirely subsonic,
and the velocity changes with the cross-sectional area A according to:
ρe Ve Ae = ρ(x)V (x)A(x) < Φmax
In case c, the sonic condition pk /p0 = p∗ /p0 is reached at the throat and
the nozzle is chocked: ρ(x)V (x)A(x) = Φmax . In the diverging section, the
flow is subsonic in order to let the exit pressure pe match with the ambient
pressure pe = pa . There exists exactly one other pressure (f ), for which the
diverging flow leads to pressure matching at the exit, without shock for a
supersonic flow. The flow is supersonic downstream of the throat. In this
case the jet flow is supersonic and fully expanded. Further lowering of the
ambient pressure (g) does not affect the flow inside the nozzle. The flow
38
within the nozzle is fully determined by the reservoir conditions and the
fact that the throat is critical. Lowering of the ambient pressure implies,
that pe > pa . The jet flow outside the nozzle is under-expanded now.
Figure 3.8: Schlieren photograph of the flow in a Laval nozzle with a slightly
under-expanded jet at the outlet (L. Prandtl).
We now discuss the flow for the pressure range pa /p0 in between cases c
and f. In this case the differential equation we have derived (equation 3.1)
has no solution. Nature finds a solution by violating our assumptions of a
continuous and isentropic solution. In the diverging section a discontinuous
pressure change occurs: a shock wave. We will now simply assume that it is
39
a narrow region, which can be described as a discontinuity with negligible
volume 3 .
Figure 3.9: Normal shock in divergent part of Laval nozzle (G. Settles).
Please note the forks due to the interaction of the shock with the viscous
boundary layers at the wall.
Choking does not only occurs in converging and Laval nozzles. Due to
friction forces, volume forces, combustion or heat transfer it can also occur
3
In practice, depending on the geometry of the nozzle and the interaction with viscous
boundary layers, this can be a complicated structure of interacting oblique shock waves
and expansion wave.
40
in a straight pipe. We will discuss this at the end of this chapter.
We restrict our discussion now to normal shock, which are in a (y, z) plane
normal to the flow direction ⃗v = (u, 0, 0). In the present discussion we
simply assume that the shock is infinitely thin4 . In other words its volume
is negligible.
Figure 3.10: Control volume used to derive the shock relations from the
integral mass, momentum and energy balances.
Applying the integral conservation laws over such a shock while we assume
a uniform flow upstream and downstream of the shock we find the so called
shock relations. Note that because the flow is uniform upstream and down-
stream of the shock the friction forces and heat flux vanish (the flow outside
4
A “strong” shock has a thickness of the order of a few times the mean free path λ̄. For
steady shocks the thickness of the shock is not relevant. In that case the steady integral
conservation laws hold anyhow!
41
the shock is isentropic). We choose a thin control volume delimited by planes
parallel to the shock, which it encloses. The mass conservation law yields:
ρu = ρ̂û (3.6)
where ρ is the density in front of the shock and u the corresponding flow
velocity. The density behind the shock is ρ̂ and û is the corresponding
velocity. In other words the mass flux φm = ρu is conserved over a normal
shock φm = φ̂m . The momentum equation yields:
p + ρu2 = p̂ + ρ̂û2 (3.7)
where p and p̂ are the pressures respectively in front and behind the shock.
The energy equation yields:
u2 û2
h+ = ĥ + (3.8)
2 2
where h and ĥ are the enthalpy respectively upstream and downstream of
the shock. Obviously the total enthalpy is conserved across a shock h0 = ĥ0 .
p̂ µ ρ̂ρ − 1
= (3.12)
p µ − ρ̂ρ
42
or:
ρ̂ 1 + µ p̂p
= (3.13)
ρ µ + p̂p
where we define µ by:
γ+1
µ= . (3.14)
γ−1
It is interesting to compare the Rankine-Hugoniot
% &γ adiabate with the isen-
p̂ ρ̂
tropic perfect gas relation: p = ρ . A first essential difference is that
the Rankine-Hugoniot depends on two parameters (p, ρ). Along a Rankine
Hugoniot curve in the (p, 1/ρ) plane there is only one point corresponding
to the initial state (in front of the shock). All the other points correspond to
the state behind a shock (p̂, 1/ρ̂). Starting a new Rankine-Hugoniot curve
from this final state yields a different curve than the one starting at (p, 1/ρ).
Along the isentropic equation of state, any point can be used as initial state,
because the curve depends only on one parameter, the entropy s. This
entropy along a Rankine-Hugoniot adiabate changes following the equation
of state: # $
ŝ − s p̂ ρ̂
= ln − γ ln . (3.15)
cv p ρ
From this equation in combination with the Rankine-Hugoniot relation (3.13)
we conclude that the entropy increases over a compression shock p̂ > p
for a perfect gas. It decreases for an expansion shock, which implies that
an expansion shock of a perfect gas does not exist. For real gasses how-
ever expansion shocks can exist depending on the sign of (∂ 2 p/∂(1/ρ)2 )s
[Landau (1987)], [Thompson (1972)]].
p̂ γ 2 − γ 2 γ 3 − 3γ 2 + 2γ 3
= (1 + ϵ)γ = 1 + γϵ + ϵ + ϵ + ... (3.16)
p 2 6
and the Rankine-Hugoniot equation:
p̂ µ(1 + ϵ) − 1 γ 2 − γ 2 γ 3−2 γ 2 + γ 3
= = 1 + γϵ + ϵ + ϵ + ... (3.17)
p µ − (1 + ϵ) 2 4
The two expressions differ only for terms of order ϵ3 or higher. Hence the
entropy production of a weak shock scales as ϵ3 :
ŝ − s p̂ ρ̂ γ3 − γ 3
= ln − ln ( )γ = ϵ + ... (3.18)
cv p ρ 12
43
Rankine−Hugoniot
Isentroop
Impulswet
10
p̂
p
γ = 1.4
0
0 γ−1 0.5 1
γ+1
ρ
ρ̂
As shown in Figure 3.12, the straight line joining the initial state (p, 1/ρ)
to the end %state (p̂, 2
& 1/ρ̂) has following the momentum law the slope φm =
1 1
−(p̂ − p)/ ρ̂ − ρ . The slope of the Rankine-Hugoniot adiabate at the
initial point is: (∂p/∂(1/ρ))RH = −ρ2 (∂p/∂ρ)s = −(ρc)2 . For a weak shock
we have furthermore (∂ p̂/∂(1/ρ̂))RH = −ρ̂2 (∂p/∂ρ)RH ≃ −(ρ̂ĉ)2 . Because
for a perfect gas we have (∂ 2 p/∂(1/ρ)2 )s > 0 we see that at the initial point
−(φ2m /(∂ p̂/∂(1/ρ̂)RH ) = (u/c)2 = M 2 > 1 hence the flow is supersonic
ahead of the shock M > 1. At the end point of the Rankine-Hugoniot
we have (φ2m /(∂p/∂ρ)RH ≃ (û/ĉ)2 = M̂ 2 < 1 so that the flow is subsonic
44
behind the shock M̂ < 1. This result appears to be much more general as
Figure 3.12: For a weak shock the Rankine-Hugoniot adiabate can be ap-
proximated by an isentrope. The Rayleigh line derived from the momentum
balance joints the initial point to the state behind the shock. The slope
of the Rayleigh line is proportional to the square of the mass flux density
φ2m = (ρu)2 = (ρ̂û)2 . The tangents at the initial point and the end point
have a slope proportional to the square of the speed of sound, (ρc)2 and
(ρ̂ĉ)2 respectively. We clearly see that M = u/c > 1 while M̂ = û/ĉ < 1.
45
3.2.3 Mach number dependance
The Mach number M in front of the shock is a convenient parameter to
relate the state of the gas in front it to the state behind it. The momentum
equation (3.9) can be written as:
# $
p̂ φ2 ρ
−1= m 1− . (3.19)
p ρp ρ̂
46
From these equations we can see that in the limit M → 1 the shock strength
ϵ = (ρ̂ − ρ)/ρ → 0 vanishes. In fact this very weak normal shock is an
acoustic wave travelling with the velocity u = c with respect to the fluid.
This can be best visualized by changing of coordinate system. Until now,
the shock had a fixed position. Now we choose a coordinate system in which
the fluid in front of the shock is stagnant. This is illustrated in figure 3.13.
In figure 3.14 we show the shock relations for the range 1 ≤ M ≤ 5 for a
perfect gas with γ = 1.4. We see that limM →∞ ρ̂ρ = (γ + 1)/(γ − 1) and
.
limM →∞ M̂ = (γ − 1)/(2γ)
2
10
γ = 1.4
p̂
p
1
10
ρ̂
ρ
T̂
T
0 M̂1
10
M̂
−1
10
1 2 3 4 5
M
Figure 3.14: Shock relations for a perfect gas γ = 1.4 as functions of the
Mach number.
47
3.2.4 Mach number from pressure measurements
We now consider a Pitot tube as show in figures 3.15, 3.16 and 3.17.
Figure 3.15: Schlieren photograph of the flow around the Pitot tube (D.
Rouwenhorst and H. Stobbe, UTwente)
we use the relation ps /p ≃ (p0 /p)(p̂0 /p0 ) to determine the Mach number.
48
Figure 3.16: Detail of the shock wave in front of a Pitot tube (D. Rouwen-
horst and H. Stobbe, UTwente)
Figure 3.17: A Pitot tube measures the stagnation pressure ps . For su-
personic flows M > 1 a normal shock is formed in front of the tube. The
stagnation pressure is approximatively equal to the total pressure behind
the normal shock ps = p̂0 .
ps
p0
0.5
p
✙
✟ ps
p ✯
✟
p0
0
0 1 2 3 4
M
Figure 3.18: From the measurement of the stagnation point pressure ps ≃ p̂0
by means of a Pitot tube and the static pressure p by means of a static
pressure tube or a pressure hole in a wall, we can calculate the Mach number
M.
50
If we consider a continuous source of infinitesimal perturbations at ⃗x0 in a
uniform steady supersonic flow, the boundary between the Mach cone and
the unperturbed region is a M ach wave. In contrast with a normal shock
of vanishing strength a compression Mach wave does not induce a transition
from a supersonic to a subsonic flow. The flow behind this very weak oblique
shock remains supersonic. We will see later that for finite shock strength
an oblique shock can either be a strong shock with M̂ < 1 (subsonic flow
behind the shock) or a weak shock with M̂ > 1 (supersonic flow behind the
shock).
Figure 3.20: Definition of approach angle β and flow turning angle θ for a
plane oblique shock.
The flow ⃗v2 = (û, v̂, 0) behind the shock is also uniform, making an angle θ
with respect to the approach flow: cos θ = (⃗v1 · ⃗v2 )/(|⃗v1 ||⃗v2 |).
Using the integral conservation law for mass over a thin controle volume
enclosing the shock, we have:
ρu = ρ̂û = φm (3.27)
φm (v − v̂) = 0 (3.28)
51
and from the x-momentum:
u2 + v 2 û2 + v̂ 2
h+ = ĥ + . (3.30)
2 2
This implies that v = v̂ and that the shock relations for an oblique shock are
identical to those for a normal shock observed in a reference frame moving
with the velocity (0, v, 0). Such a Galilean transformation does not affect
the flow physics. Hence we can use all the results obtained for a normal
shock by simply replacing the Mach number by the normal Mach number
Mn = u/c = M sin β. Obviously a shock wave can only be found if Mn >
1. The Mach number behind the shock M̂ is related to the normal Mach
number M̂n by the relation: M̂n = M̂ tan(β − θ).
Figure 3.21: Relationship between the approaching angle β and the flow
turning angle θ for a planar shock at fixed approach Mach number M (after
[Shapiro (1953)]).
52
Figure 3.22: Weak shock attached to a wedge in a supersonic flow (D.
Rouwenhorst and H. Stobbe, UTwente)
û tan(β − θ)
= . (3.31)
u tan β
From the normal shock relations for a perfect gas we have:
û 2γ 1 − Mn2
=1+ . (3.32)
u γ + 1 Mn2
Eliminating û/u we find a relationship between β, θ and Mn = M sin β:
tan(β − θ) 2γ 1 − Mn2
=1+ . (3.33)
tan β γ + 1 Mn2
For a fixed Mach number we obtain a curve in the (β, θ) plane. Such curves
are shown in figure 3.21. We observe that for a given Mach number there is
a maximum flow turning angle θmax . For θ > θmax there is no plane shock
solution. A detached shock will be formed ahead of the wedge (see figure
3.24). This is the reason why supersonic aircraft have a sharp nose and
wings with sharp leading edges.
For any turning angle θ < θmax we have two solutions for β. They both
correspond to an attached plane shock. One of the solutions corresponds to
a subsonic flow behind the shock M̂ < 1, which we call a strong shock. The
other solution is a weak shock solution M̂ > 1. Of course the weak shock
solution is preferred, as it has less dissipation than the strong shock.
53
Figure 3.23: For large apex angles a detached shock will be formed in front
of a wedge placed in a supersonic flow.
54
conservation implies that the mass flux is conserved along the pipe:
d[ρV ] = 0 (3.34)
or:
ρ(x)V (x) = φm (3.35)
where φm is a constant determined by the boundary conditions. The mo-
mentum balance implies that:
where τ is the wall shear stress (viscous force per unit area of the pipe wall
on the fluid). The perimeter of the pipe section is Pp = 2πa and the cross
sectional area is Ap = πa2 . The integral energy balance applied to the pipe
section yields: 1 2
V2
Ap φm d (h + ) = Pp qdx (3.37)
2
where q is the heat flux from the wall to the flow. Using the conservation of
mass flux ρV = φm we rewrite the momentum equation as:
dp 2τ
+ V dV = dx (3.38)
ρ ρa
and energy balance equation as:
2q
dh + V dV = dx. (3.39)
φm a
Eliminating the flow velocity V by substraction the momentum equation
3.38 from the energy equation 3.39 and using the fundamental Gibbs’ equa-
tion yields an equation for the entropy:
1 2
dp 2q 2τ
T ds = dh − = − dx. (3.40)
ρ φm a ρa
Heat flow into the pipe and friction do correspond to entropy production.
Note that the sign of τ is negative because the wall friction opposes the flow.
Hence friction increases the entropy of the flow.
55
We can furthermore describe the enthalpy as a function of p and ρ, which
implies: # $ # $
∂h ∂h
dh = dp + dρ. (3.42)
∂p ρ ∂ρ p
Substitution into equation 3.40 we have:
# $ # $
∂h 1 ∂h 2q 2τ
[ − ]dp + dρ = [ − ]dx. (3.43)
∂p ρ ρ ∂ρ p φm a ρa
The two equations 3.41 and 3.43 form a set of linear equations for dp and
dρ. The determinant D of this set of equations is given by:
( 1
)
hρ 2ρ
− hp
D= 1−V (3.44)
ρ hρ
hρ
D= (1 − M 2 ). (3.46)
ρ
When D ̸= 0 the system of equations yields a solution so we find expressions
for dp/dx and dρ/dx. For M = 1 the determinant vanishes (D = 0), so that
we have no solutions for dp/dx and dρ/dx. They become infinitely large.
This situation corresponds to choking induced by friction and heat transfer.
56
-the mass flow decreased for given inlet pressure.
As in the case of a Laval nozzle the behaviour for supersonic flows M > 1
is opposite to that found for subsonic flows. This is due to the fact that
D > 0. Instead of accelerating the flow will now decelerate, contradicting
our intuition. The Mach number decreases approaching M = 1. As choking
is approached (the pipe is too long) a shock wave will form and the flow will
continue as subsonic flow behind the shock. This subsonic flow accelerates
up to M = 1. If the pipe is too long, the shock will move up to the pipe
inlet causing the mass flow to change.
This complex behaviour has been studied for an adiabatic pipe flow of a per-
fect gas by [Fanno (1904)]. He obtained an analytical solution [[Shapiro (1953)],
[Thompson (1972)], [Owczarek (1964)], [Landau (1987)]].
57
We recognize the isothermal
. Mach number MT = V /cT with the isothermal
speed of sound cT = p/ρ. Hence we have:
1 dV cf MT2
= . (3.53)
V dx a(1 − MT2 )
As for friction in an adiabatic flow (Fanno flow) we observe choking when
the flow is accelerated from a subsonic velocity MT < 0 up . to MT = 1.
The fact that this involves the isothermal
. speed of sound c T = p/ρ rather
than the adiabatic speed of sound c = γp/ρ is logical because the flow is
isothermal. Hence acoustic waves transmitting information to the reservoir
to regulate the mass flow travel at the speed cT rather than c.5
By integrating between position x1 with MT = MT,1 and x2 with MT = MT,2
we find: / # $0
x2 − x1 1 1 1 MT,1
a
=
cf 2MT,1 2 − 2M 2 + ln M (3.54)
T,2 T,2
58
Figure 3.25: Laval nozzle downstream of the combustion chamber of the
solid propellant motor of the Ariane V rocket.
3.5 Questions
Questions with a star (*) are more advanced and not considered as basic
study material.
1. Following the quasi-one dimensional theory for steady flows the accel-
eration dV /dx along a pipe is related to the change in cross-section
area dA/dx by:
1 dV 1 dA
= 2
.
V dx (1 − M )A dx
where M = V /c is the Mach number. Is this theory also valid for a
liquid flow?
59
2. Following equation 1 there is a singularity of the factor (M 2 − 1)−1 at
M = 1. What does this implies for the actual flow in a Laval nozzle?
5. Make a sketch of the pressure profile p(x) along a Laval nozzle for
various values of the parameter pa /p0 .
7. How much does the mass flow Φ of a perfect gas through a choked
Laval nozzle increase if we double the reservoir pressure p0 at fixed
T0 ?
8. How much does the mass flow Φ of a perfect gas through a choked Laval
nozzle increase if we double the reservoir temperature T0 at fixed p0 ?
10. A blow down supersonic wind tunnel consists of an air reservoir with
T0 = 298 K and p0 = 60 bar. Such tunnels are commonly used up to
M = 4. Why are they not used at M = 24?
* 12. Due to the non-uniformity of the flow at the throat discussed in the
previous exercise, the critical line M = 1 will be a curve. Make a
sketch of this curve in a Laval nozzle.
60
13. What are the conditions of validity of the integral balances for mass,
momentum and energy taken over a normal shock wave?
ρu = ρ̂û
p + ρu2 = p̂ + ρ̂û2
u2 û2
h+ = ĥ +
2 2
Did we neglect friction and heat transfer to derive these equations?
14. Why are the shock relations for mass, momentum and energy discussed
in the previous exercise also valid for a moving shock (speed us )? Are
they also valid for a time dependent shock velocity us (t)?
16. What is the maximum of the compression ρ̂/ρ that can be reached
by a single shock compression? What is the maximum of compression
that can be reached by two successive shock compressions?
18. Determine the ratio T̂0 /T0 of the total temperatures across a normal
shock in a perfect gas.
20. Given the initial point (p, ρ1 ) and the mass flow φm through a shock,
draw the line along which one can find all the possible end states (p̂, ρ̂1 )
behind a normal shock as imposed by the momentum balance equation
(so called Rayleigh line). What is the physical meaning of the slope
dp̂/d(1/ρ̂) of the Rayleigh line?
61
21. The increase in entropy across a shock can be calculated for a perfect
gas by combining the equation of state:
ŝ − s p̂ ρ̂
= ln( ) − γ ln( )
cv p ρ
with the Rankine-Hugoniot. Compare, for ρ̂/ρ = 2 and γ = 1.4, the
value of (ŝ − s)/cv calculated in this way to the result obtained by
means of the weak shock approximation:
ŝ − s γ3 − γ 3
= ϵ
cv 12
What happens for ρ̂ = 1.2.
22. Which additional information do we need in order to determine the
Mach number M from the Pitot tube pressure measurement ps ? Can
we determine the Mach number from the ratio ps /p0 of Pitot tube
pressure and reservoir pressure?
* 23. We place a Pitot tube in the diverging section of a Laval nozzle. Make
a sketch of the variation of the Pitot pressure ps as a function of the
variation in atmospheric pressure pa for fixed reservoir pressure p0 .
24. Derive the expression for the angle α of Mach waves with the flow
velocity direction as a function of the Mach number M . What occurs
when M < 1?
25. What is the difference between a weak and a strong planar oblique
shock for given flow turning angle θ?
26. Estimate the Mach numbers of these two flows, shown in figure 3.26:
(a) Flow in supersonic wind tunnel HST of NLR (A. Elsenaar [Elsenaar (2012)],
50 jaar HST)
(b) Flow in supersonic wind tunnel (Spaceex, Space Launcher, DNW)
62
a b
Figure 3.26: (a) Supersonic HST wind tunnel of NLR (Amsterdam); (b)
Supersonic flow around a space launcher, DNW.
3.6 Answers
1. The theory is valid for a steady frictionless quasi-one dimensional fluid
flow. We assumed local thermodynamic equilibrium. No other as-
sumptions where made concerning the type of fluid. It can be a gas
or a liquid. Hence the theory is valid for a liquid.
63
Figure 3.27: Comparison of a strong single shock compression from (p, 1/ρ)
to (p̂/ρ̂) with a double shock compression with an initial shock from (p, 1/ρ)
to (p̂1 , 1/ρ̂1 ) followed by a second shock from (p̂1 , 1/ρ̂1 ) to (p̂2 , 1/ρ̂2 ) in the
(p, 1/ρ) diagram.
9.
T0 γ−1 2
=1+ M
T 2
p0 ρ0 p0 T
= ( ) γ = ( )γ ( )γ
p ρ p T0
T0 γ−1 2 p0 γ−1
=1=+ M =( ) γ
T 2 p
64
6 # $
2 p0 γ−1
M= ( ) γ −1 = 3.24
γ−1 p
T = 1166 K
Ak ρ V ρ0 ρ c 0 c
= ∗ ∗ = ∗ M
Ae ρ c ρ ρ0 c ∗ c 0
/ 0 γ+1
Ak (γ + 1)/2 2(γ−1)
=M = 0.05
Ae 1 + γ−1
2 M
2
11. The curvature of streamlines following the wall involves a pressure gra-
dient which should compensate the centrifugal acceleration. Hence the
pressure on the center line is larger than at the wall. The pressure dif-
ference is of the order pcenter −pwall ≃ [ρ∗ (c∗ )2 /rk ]ak = (ak /rk )γp0 (p∗ /p0 ) =
γ
2
(ak /rk )γp0 ( γ+1 ) γ−1 .
12. The critical condition is reached near the wall earlier than on the center
line. Hence the critical line is convex seen from downstream.
65
(and only one) initial point and depends on two parameters (the coor-
dinates of this initial point). Furthermore some processes are forbidden
because they would reduce the entropy. This is the case for expansion
shocks for ideal gasses. The Rankine-Hugoniot describes irreversible
processes (with increase in entropy).
16. For γ = 1.4 the maximum compression with a single shock compression
is ρ̂max /ρ = (γ + 1)/(γ − 1) = 6. For two successive shocks we have
the total temperature is conserved ρ̂max /ρ = [(γ + 1)/(γ − 1)]2 = 36.
17. See Figure 3.27.
18. From the energy balance across a shock we have: h0 = ĥ0 . For a
perfect gas h = cp T . Hence we have T0 = T̂0 over a shock in a perfect
gas.
19. An expansion shock is possible if (∂ 2 p/∂(1/ρ)2 )s < 0. This can occur
near the critical region of a fluid.
dp
20. The slope is d(1/ρ) = −φ2m . As this is a constant imposed by the
upstream boundary conditions, the Rayleigh line is a straight line in
the (p, ρ1 ) plane.
21. For ρ̂/ρ = 2:
p̂ µ ρ̂ρ − 1
= = 2.75
p µ − ρ̂ρ
# $
ŝ − s
= 0.041
cv exact
# $
ŝ − s
= 0.112
cv weak−shock
While the shock is almost isentropic, the prediction of the entropy is
not accurate. For ρ̂/ρ = 1.2:
p̂ µ ρ̂ρ − 1
= = 1.29
p µ − ρ̂ρ
# $
ŝ − s
= 6.8 × 10−4
cv exact
# $
ŝ − s
= 9.010−4
cv weak−shock
The shock is almost perfectly isentropic and the predicted entropy
increase is reasonably accurate.
66
22. From a measurement of the stagnation pressure ps (Pitot tube) and the
static pressure p (wall pressure) we can determine the Mach number
because ps /p is a function of M . For M >> 1 we can also determine
the Mach number from ps /p0 . For M ≤ 1 we have ps /p0 = 1 so we
cannot determine the Mach number from a measurement of pS /p0 .
1 dV 1 dMT cf MT2
= =
V dx MT dx a(1 − MT2 )
yields: / 0
a 1 1 MT,1
x2 − x 1 = 2 − 2 + ln
cf 2MT,1 2MT,2 MT,2
For MT,2 = 1 we have:
/ 0
a 1 1
x∗ = x2 − x1 = 2 − + ln MT,1
cf 2MT,1 2
√
R = R̄/Mmol = 520Jkg −1 K −1 , cT = RT = 380m/s, MT,1 =
V1 /cT = 0.0263
x∗ = 180 km. In practice a compressor is placed each 100 km. The
pressure is not allowed to decrease below 40 bar. The flow velocity
can be higher than 10 m/s.
67
sound. Consequently choking will occurs in an adiabatic process after
a longer distance than in an isothermal process.
This can also be understood by the fact that in an isothermal flow
there is a heat flux into the pipe, which by itself promotes choking
already. As explained in 3.4.4 heat addition to the flow can cause
choking by itself (without friction).
68
Chapter 4
Instationary 1D compressible
flows
4.1 Introduction
Disturbances in pressure and density of a compressible medium give rise to
travelling pressure waves. In case the amplitudes are small the waves are
called acoustic waves or sound waves; their propagation velocity is the speed
of sound and a simple linear relation between pressure amplitude and veloc-
ity amplitude exists. In case the amplitudes are not small, the propagation
velocity in a disturbance becomes dependent on the amplitude. Wavefronts
will show a change in shape. In some cases wave fronts will become steeper
which eventually gives rise to the appearance of shock waves. This non-linear
wave propagation is subject of this chapter. Analytic solutions of non-linear
wave problems will be found with the so called method of characteristics.
The 1D gasdynamics equations are applied to, amongst others, the descrip-
tion of the so called shock tube and the supersonic wind tunnel based on
the shock tube principle. Finally, we will pay attention to the striking anal-
ogy between non-linear wave propagation in compressible media and wave
phenomena which occur on a freeway, in traffic flow.
69
∂u ∂u 1 ∂p
+u + =0, (4.2)
∂t ∂x ρ ∂x
# $
2 ∂p dp
p = p(ρ) ; c = = ; s = s0 . (4.3)
∂ρ s dρ
We eliminate dρ with dρ = dp/c2 , and in addition we introduce the gasdy-
namic variable Γ:
dp
dΓ = . (4.4)
ρc
With dρ = ρc dΓ and dp = ρcdΓ (4.1-4.2) become:
∂u ∂u ∂Γ
+u +c =0, (4.5)
∂t ∂x ∂x
∂Γ ∂u ∂Γ
+c +u =0. (4.6)
∂t ∂x ∂x
Adding up, respectively subtracting gives:
# $
∂ ∂
+ (u + c) (u + Γ) = 0 , (4.7)
∂t ∂x
# $
∂ ∂
+ (u − c) (u − Γ) = 0 . (4.8)
∂t ∂x
These are the non-linear characteristic equations. In case ρc ≈ ρ0 c0 and
u << c0 , c ≈ c0 , then Γ ≈ p′ /ρ0 c0 and we find the linear equations. The
variable Γ plays an important role in gasdynamics:
" "
1 c
Γ= dp = dρ ; s = s0 . (4.9)
ρc ρ
γ−1
For a calorically perfect gas we find with c ∝ ρ 2 ; γ = cp /cv :
"
2 dc 2
Γ= c = c. (4.10)
γ−1 c γ−1
In that case the characteristic equations are:
# $
dx 2c
along =u±c: d u± =0. (4.11)
dt γ−1
70
Figure 4.1: C ± characteristics. Γ± characteristics.
2
For a perfect gas these are straight lines with slopes du/dc = − γ−1 for Γ+
2
en du/dc = γ−1 for Γ− .
Solving a gasdynamic problem by means of the method of characteristics
means:
finding the trajectories of the C + and C − characteristics in the x − t plane.
When these trajectories are known , also u, c can be found in any point by
applying the Riemann invariants. Suppose that the solution (u, c) is known
at line segment AB in the x − t plane of figure 4.2. We divide the line
segment in a number of elements with nodes (1 t/m 6 in figure 4.2.), and
calculate the characteristic directions u + c en u − c in those points. These
C + and C − characteristics cross (7 t/m 11). Applying the Γ± relations gives
the new values of u and c etc. It is clear that line segment AB affects only
a limited area in the x − t plane, terminating in point D.
71
Waves).
Figure 4.3: Left travelling simple wave with its representation in the x − t
plane and its representation in the c − u plane.
We will first treat the left travelling simple wave (of the C − type). Assume,
we generate a disturbance in a long tube for x > x0 . At t = 0, x < x0 , a
uniform flow exists with velocity u0 and speed of sound c0 . The disturbance
propagates in the negative x-direction.
Property I
2c 2c0
All C + characteristics emerge from the uniform region. So u+ γ−1 = u0 + γ−1
for all (x, t) bounded by the uniform area. So a unique relation exists be-
tween u and c in the whole field. This means that the whole x−t plane (con-
nected to the uniform region) is represented by one single Γ− characteristic
in the c − u plane corresponding to state (u0 , c0 ).
Property II
For each single wave travelling to the left, the C − characteristics are lines
72
of constant state. The C − characteristics are straight lines! The proof is
2
simple, du − γ−1 dc = 0 along the C − characteristics. Moreover, because of
2
Property I: du + γ−1 dc = 0, such that du = 0, dc = 0. This implies that in
−
the u − c plane, Γ characteristics reduce to points.
Note
If we consider a single wave travelling to the right, now the C + characteristics
2c 2c0
are lines of constant state. Also u − γ−1 = u0 − γ−1 for all (x, t) and the
+
C characteristics are straight lines in this case.
Note
The indication “to the left” and “to the right” should be seen with regard to
the gas particles. The whole treatment remains valid in case of a supersonic
flow (u > c > 0), in which both the u + c waves and the u − c waves travel
to the right with regard to the laboratory coordinate system.
73
Figure 4.4: Expansion wave caused by a accelerating piston.
u2
ξ= , τ = u/a , (4.17)
2a
and express c in terms of u by using (4.13):
γ−1
c = c0 + u. (4.18)
2
Substitution of (4.17) and (4.18) in (4.16) gives:
# $
u2 γ+1
x− − c0 + u (t − u/a) = 0 , (4.19)
2a 2
a quadratic equation for u that can be solved for arbitrary x, t, i.e. within
the range −V < u < 0. Outside this range, the velocity is either zero or −V .
b. Expansion fan
A special case is found when the piston velocity is changed step-wise from
0 to -V . An expansion fan is formed as shown in figure 4.5.
74
Figure 4.5: An abruptly accelerating piston causes an expansion wave. Dot-
ted line: particle trajectory.
The solution follows from (4.19) by taking the limiting case for a → −∞,
τ0 → 0, such that a − τ0 = −V :
2 %x &
u= − c0 ; −V < u < 0 . (4.20)
γ+1 t
The spatial extent of the expansion fan follows from (4.20) by substitution
of u = 0, respectively −V :
γ+1 x
c0 − V < < c0 . (4.21)
2 t
The value of the piston velocity V is in principle unlimited. The particle
2c0
velocity u however is bounded. From (4.13) it follows that |u| ≤ γ−1 , because
2c0
c ≥ 0. When in this example u = − γ−1 , then c = 0, p = 0, ρ = 0, which
means that a vacuum state is found. Expansion fans are bounded:
2 x
− c0 < < c 0 . (4.22)
γ−1 t
75
Figure 4.6: Expansion in vacuum; dotted line: particle trajectory.
We now consider the case of accelerating piston (figure 4.7), which now ac-
celerates from rest into the positive x-direction with a constant acceleration
a > 0 until it attains the final velocity V . The solution is identical to the
solution of case a. However, the straight C + characteristics are not divergent
but convergent, which has important consequences.
Since u increases, also c and u + c increase. The straight C + characteristics
will approach each other so that the compression wave becomes abrupter
and steeper. Finally a shock wave is formed, for which the condition of
isentropy is no longer valid. With solution (4.19) we can estimate where
and when the shock wave is formed. Equation (4.19) is of the form:
F (x, t; u) = 0 (4.25)
76
Figure 4.7: An accelerating piston causes a compression wave.
77
Figure 4.8: Abruptly accelerating piston causes a shock wave.
78
Because it concerns a C + simple wave, c must also satisfy:
γ−1
c = c0 + F (ξ) . (4.31)
2
The equation for the straight C + characteristic again gives the solution of
the problem:
# $
γ+1
x = ξ + (u + c)(t − t0 ) = ξ + c0 + F (ξ) (t − t0 ) . (4.32)
2
∂u ∂u 8 ∂x F ′ (ξ)
= = γ+1 ′ . (4.33)
∂x ∂ξ ∂ξ 1+ 2 F (ξ)(t − t0 )
ρ dc
1+ > 0, s = s0 . (4.36)
c dρ
The left hand side is called the fundamental derivative or the non-linearity
coefficient by P.A. Thompson.
79
When (4.36) is not satisfied, expansion waves will become steeper. This can
be seen in gas-vapour mixtures with many internal degrees of freedom, as has
been demonstrated by P.A. Thompson and co-workers. See e.g. S.C. Gulen,
P.A. Thompson, H.J. Cho in “Adiabatic Waves in Liquid-Vapor Systems”,
G. Meier, P.A. Thompson, eds, Springer 1989 or the extensive contribution
of A. Kluwick in the Handbook of Shock Waves (Academic Press, 2001).
side is filled with the driver gas with a pressure p4 higher than the pressure
p1 in the test gas. The diaphragm is removed at t = 0 and the driver gas
expands due to which the test gas endures shock compression. The wave
80
phenomena are shown in the x − t plane of figure 4.10. We recognize the
expansion fan in the driver gas and the shock wave in the test gas. Both
gases are separated by a contact plane over which velocities and pressures
must be continuous. In the x − t plane we distinguish several uniform re-
gions: the initial states of the test gas (1) and the driver gas (4), the test
gas, compressed and accelerated by the shock wave (2) and the expanded
driver gas (3). The transition (4) → (3) passes a simple expansion wave and
is isentropic:
γ4 − 1
c3 = c4 − u3 , (4.37)
2
p3 c3 2γ4
= ( ) γ4 −1 (s3 = s4 ) . (4.38)
p4 c4
The transition (3) → (2), a contact plane, is characterized by:
u3 = u2 ; p3 = p2 . (4.39)
The transition (1) → (2) is a shock transition. Therefore, applying the shock
relations of a perfect gas yields:
2 M2 − 1
u2 = c1 , (4.40)
γ1 + 1 M
p2 2γ1 γ1 − 1
= M2 − , (4.41)
p1 γ1 + 1 γ1 + 1
with M : shock Mach number (M = Us /c1 ).
Equations (4.37) to (4.41) can be combined to:
p4 p4 p2 γ4 − 1 c1 M 2 − 1 −2γ 4 2γ1 γ1 − 1
= = {1 − } γ4 −1 { M2 − }. (4.42)
p1 p3 p1 γ1 + 1 c 4 M γ1 + 1 γ1 + 1
So, this provides us the relation between the pressure ratio over the di-
aphragm p4 /p1 and the shock Mach number for an arbitrary combination of
perfect gases. The maximum shock Mach number is attained for p4 /p1 → ∞;
this corresponds to:
γ4 − 1 c 1 M 2 − 1 γ 1 + 1 c4
1− → 0 , Mmax ≈ . (4.43)
γ1 + 1 c 4 M γ 4 − 1 c1
The combination air/air leads to Mmax ∼ 6.2. The combination H2 as
driver gas and Ar as test gas is very effective: Mmax ≈ 27. For a perfect gas
this would yield a temperature jump (absolute temperature) of a factor 228!
In reality this cannot be realized, but the example gives an impression of
the large temperature (and pressure) ranges attainable by means of a shock
tube.
Light driver gases with many internal degrees of freedom have a large maxi-
mum expansion velocity Vmax = 2c0 /(γ −1), and therefore are very effective.
81
Sometimes the driver gas is heated in order to produce stronger shock waves,
since c ∼ T 1/2 . This can be done by burning a mixture of hydrogen and
oxygen, possibly diluted with helium in the so called “combustion-driven
shock tube”.
Shock reflection
The shock will reflect at the end wall of a closed shock tube. A stagnant
gas of relative high temperature and pressure appears: region (5) in figure
4.10. We have to use the Rankine Hugoniot relations of a shock wave to
determine p5 and T5 . We restrict ourselves to a perfect gas. In that case
the shock relations are completely symmetrical. We will use this property.
Assume velocity u and pressure p in front of the shock wave. The velocity
and pressure behind the shock wave are û and p̂; the speed of sound before
and after shock wave passage are c and ĉ. The velocity jump is described
by (see Appendix A):
# $
u − û 2 1 u − Us
= M− ; M= (4.44)
c γ+1 M c
in which Us is the shock velocity. All velocities have been defined here with
respect to the same coordinate system. We shall now apply (4.45) to the
transition (1) → (2) and use the notation M+ = (u2 − Us )/c2 :
# $
u2 2 1
= M+ − ,
c2 γ+1 M+
u2 −Ur
and we apply (4.44) to transition (2) → (5) with notation M− = c2 .
82
With Ur < 0 this gives:
# $
u2 2 1
= M− − .
c2 γ+1 M−
p̂ 2γ γ−1
= M2 − ; (4.46)
p γ+1 γ+1
Because of symmetry the following relation must also hold:
p 2γ γ−1
= M̂ 2 − . (4.47)
p̂ γ+1 γ+1
p5 − p1 1 + γ−1
γ+1
= 1 + p1 γ−1 . (4.48)
p2 − p1 p2 + γ+1
p5 −p1
For weak shocks p1 ≃ p2 : p2 −p1 = 2, the acoustic result. For strong shocks
p5 −p1 γ+1
p1 /p2 → 0 and we find p2 −p1 = 2 + γ−1 . For air (γ =1, 4) the pressure
jump at the end wall of the tube is at most 8 times the pressure jump over
the first shock wave! For the temperature we can deduce:
9 γ+1 p5 :
T5 p5 γ−1 + p2
= . (4.49)
T2 p2 1 + γ+1 p5
γ−1 p2
Example:
83
4.7 Shock tube applications
With a shock tube, an almost ideal and known step-wise change in the state
of a gas can be realized. The state jump satisfies elementary conservation
laws such that a simple measurement of the shock velocity at known initial
conditions is sufficient to calculate the velocity, temperature, pressure, den-
sity of the gas behind the shock.
A large range of temperatures and pressures can be realized with a shock
tube. Shock tubes are applied for studying properties of gases and plasma’s
from cryogenic temperatures up to temperatures of 3 104 K. When focusing
the shock waves, even much higher temperatures and pressures can be ob-
tained.
We will give some examples of fields of application:
b. Aerodynamics/hydrodynamics.
The shock tube can be used as a wind tunnel because the shock wave also
causes a velocity jump in the gas. The flow Mach number with respect to the
laboratory frame is dependent on the shock Mach number as shown in figure
4.12 (Appendix A). The maximum Mach number is 1.8 for air (γ = 1, 4).
The shock tube is especially useful as a transonic wind tunnel. Because in
principle high pressures can be applied, also high Reynolds numbers can be
realized.
A so called shock tunnel is obtained by connecting a shock tube to a vacuum
vessel through a Laval nozzle, as sketched in figure 4.13.
After a shock reflection at the side of the Laval nozzle a high enthalpy stag-
nation area is formed which feeds the Laval nozzle, the supersonic wind
tunnel. With a normal shock tube, the stagnation enthalpy is insufficiently
high to simulate the conditions of spacecrafts re-entering the atmosphere.
To realize this anyway, the “free-piston shock tunnel”is used, also called
the “Stalker-tunnel”, named after the inventor R.J. Stalker, AIAA J, 5, 12,
84
Figure 4.12: Relation between shock Mach number and the flow Mach number
induced by the shock wave, γ = 1.4.
Figure 4.13: Shock tunnel. HD: high pressure section, D1,2 : diaphragms,
LD: low pressure section, V : vacuum vessel, L: Laval nozzle.
85
2160-2165, 1967. An impressive shock tunnel is used by DLR in Göttingen
to study the high enthalphy flows associated with the re-entry of space ve-
hicles in the tunnel.
Figure 4.14: Principle of the “free-piston shock tunnel”, R.J. Stalker, AIAA
J, 5, 12, 2160-2165, 1967.
The driver gas, mostly helium or hydrogen is heated by use of a piston which
is driven by compressed air. The driver gas has both high temperature and
pressure and is able to cause strong shock waves.
Shock tubes are of course used to study the consequences of shock wave
interactions with obstacles. In the Netherlands this is done at the Prins
Maurits Laboratory, TNO, Rijswijk. The laboratory also has a so called
“blast-wave simulator”, with which the consequences of an explosion can be
studied.
The step-wise velocity jump, induced by a shock wave gives excellent op-
portunities to study the time dependent flow resistance of small spheres,
accelerated by this shock induced flow. Shock tubes are used to study prop-
erties of dusty gases, of water bubble mixtures, of cavitating bubbles. A
good insight in the diversity of shock tube applications is given by the pro-
ceedings of the bi-annual Shock Wave Symposia. See for example:
Ball, G.J., Hillier, R., Roberts, G.T., Shock Waves, Proceedings of the 22th
International Symposium on Shock Waves, Imperial College, London, 1999,
86
or the journal “Shock Waves”. Also the Handbook of Shock Waves (Aca-
demic Press, 2001, three volumes) offers aan extensive overview of the rich
variety of shock wave applications, and so does the Shock Wave Science and
Technology Reference Library (2007).
87
The boundary characteristic I (see Fig.) applies to (4.52) and is just part of
2c0
the expansion fan such that, with β0 = γ−1 :
# $
γ+1 γ−3
x = (u − c)t = − α− β0 t . (4.54)
4 4
∂t γ+1 t
+ =0. (4.55)
∂α 2(γ − 1) α + β0
2c0
The solution must satisfy: t = t0 = L/c0 , α = α0 = γ−1 and is:
# $ 1 γ+1
β0 + α 0 2 γ−1
t(α, β0 ) = t0 . (4.56)
β0 + α
The solution must satisfy (4.56) and is given by Shaw (Courant and Friedrichs,
p. 196):
# $ # $
α0 + β0 λ α0 β0 + αβ
t(α, β) = t0 Pλ−1 ;
α+β αβ0 + βα0
γ+1
Pλ : Légendre polynomials; λ = 2(γ−1)
For γ = 1, 4, λ = 3 ; P2 (x) = 12 (3x2 − 1).
88
A second shock will always overtake a first shock in the same direction.
So, shock B in figure 4.16 will travel faster than shock A. This is because
state (2) is subsonic with respect to shock (a) and supersonic with respect
to shock B. Thus uA − u2 ≤ c2 and uB − u2 ≥ c2 such that uB ≥ uA .
A new shock will be generated: shock E. A contact discontinuity D and
an expansion fan C will leave from the interaction point. The entropy in
(4) is influenced by two shock compressions, the entropy in (5) just by one
shock compression. Because of the non-linear relation between the entropy
jump and the shock strength: s4 ̸= s5 . This explains the formation of a
contact discontinuity separating two gaseous states with the same pressure,
but different entropy, temperature and density. That an expansion wave
must be generated follows from a representation of the different states in
a p − u diagram. The shock polar A, E in such a p − u plane is found by
eliminating the Mach number from the shock relations (Appendix A) for p̂
and û. This shock polar gives all possible states that can be attained from
state (1) by means of a shock wave transition. From state (2) a new shock
polar can be drawn. The gas arrives then in state (3). From here the gas
has to expand in order to arrive again at the first shock polar A, E. This is
only possible by means of an expansion wave C.
89
Figure 4.17: Relation between flux q and density ρ at a freeway.
90
This results in:
# # $$1/2
1 1
V = 2a −L and q = ρV = {2aρ(1−Lρ)}1/2 , ρm = . (4.66)
ρ 2L
Assume a = 7m/s2 , L = 5m. This gives Vm = (2aL)1/2 ∼ 8, 5m/s.
Shock waves
The shock wave conservation law for a road is simple and follows immediately
from the integral mass conservation law.
Suppose the velocity and density ρ, V jump to the values ρ̂, V̂ by a shock
with velocity Vs . Mass conservation demands:
ρ(Vs − V ) = ρ̂(Vs − V̂ ) , (4.67)
91
Figure 4.19: Development of a density disturbance into a shock wave.
such that:
q̂ − q
Vs = . (4.68)
ρ̂ − ρ
This corresponds with the connecting line AB in figure 4.20. Remember
again that q̂ and ρ̂ are not independent and that the expression above in
fact gives the relation between Vs and ρ̂ for given ρ.
We treat again a traffic light which alternately turns green and red. We
denote the ‘green’ and ‘red’ times with tg and tr . A suddenly red turning
traffic light causes a shock wave with velocity:
−q
Vs = . (4.69)
ρj − ρ
q(tr + tg ) < tg qm ,
92
Figure 4.21: Traffic wave at a traffic light. Drawn are “particle trajectories’,
shock waves - - - , and the characteristics of the expansion fan.
93
so that:
tr q
tg ≥ ,
qm − q
in which it is used again that at the traffic light the relation q = qm holds.
If this condition is not satisfied the following picture will appear: (see figure
4.22)
See also: S.A. Smulders, Control of freeway traffic flow, Proefschrift UT,
1989.
94
4.10 Questions
1. Starting from the equations of Euler:
1 Dρ
= −∇ · ⃗v
ρ Dt
D⃗v
ρ = −∇p
Dt
Ds
=0
Dt
Write these equations for a one-dimensional unsteady flow. Which are
the conditions of validity of these equations?
10. What is a self-similar solution of the wave equation? What are the
conditions for the occurrence of such a solution? Give two examples
of simple waves for one-dimensional unsteady frictionless flows?
95
Figure 4.23: A physically impossible to realize (x, t) diagram.
13. Show that for γ = 3 the characteristics C ± are straight lines in a (x, t)
diagram.
16. What is the maximum velocity a perfect gas (with γ = 5/3) can
reach when starting from a uniform stagnant state (p0 , c0 , u0 = 0) in a
one dimensional unsteady frictionless expansion? Compare this to the
maximum velocity that can be reached in a quasi-steady frictionless
expansion. Which of the two velocities is largest?
96
(a) The fluid state a the left side of the valve x = −ϵ is determined
by a boundary condition at the valve and by the message carried by a
family of characteristics. Which characteristics are doing the job? (b)
What is the message carried by these characteristics to the wall of the
valve? (c) What is the boundary condition, which prevails for t > 0
at x = −ϵ? (d) Calculate the magnitude of the perturbations p′ and
u′ for t > 0 at x = −ϵ. (e) Same questions for the other side of the
valve. (f) Make a drawing of the waves in an (x, t) diagram. Indicate
in the diagram the uniform regions. (g) Make a sketch of the pressure
perturbation p′ as a function of x for a given time t > 0. Specify
the position of the front of the waves and their amplitudes. Consider
a water flow with p0 = 1 bar, ρ0 = 1000 kg/m3 , c0 = 1500 m/s and
u0 = 1 m/s. Determine p′ and u′ on both sides of the valve. Determine
the force acting on the valve for t > 0 when the pipe has a cross section
of A = 10−2 m2 . Is a linear approximation accurate?
97
Figure 4.24: Pressures measured within the mouthpiece of a trombone at
piano (p), mezzo forte (mf ) and fortissimo (ff ) level.
21. The tunnel entrance for high speed trains M = 0.3 is a long converg-
ing tube (see figure 4.25). In some cases the walls of the horn are
perforated. Why is this useful?
98
Figure 4.25: Entrance (right) and exit (left) of a tunnel for high speed trains
M = 0.3.
99
p′ = 0 for |x| ≥ L.
Furthermore we have u′ = 0 and s′ = 0. Assume that the perturbation
is so small that a linear approximation can be used in order to deter-
mine the wave pattern for t > 0. Make a sketch of the (x, t) diagram
that you expect. Indicate in this diagram the uniform regions, the
simple wave regions and the complex wave regions. Calculate the flow
p′ and u′ in each region and specify the boundaries of the regions.
25. Consider a pipe of length 2L and cross section A that is divided in two
equally long sections by a fixed piston of mass m at x = 0. The section
left (x < 0) is filled at pressure p0 with a perfect gas γ = 3 with the
speed of sound c0 . The section right (x > 0) is filled with the same
gas at pressure p = p0 /2 with speed of sound c0 . At t = 0 the piston
is released. Assuming a frictionless movement of the piston, write the
equation of motion for the piston in terms of the pressures on each
sides. Derive an expression for the pressures on each side of the piston
before waves reflecting from the end walls reach the piston. Assume
that L is much shorter than the shock wave formation distance xs .
Show that for short times we have:
m du p0 u
= [1 − 9 ].
A dt 2 c0
Estimate the shock formation length xs . Make a sketch of the wave
pattern before the reflections from the walls reach the piston.
100
27. Consider a reservoir with stagnant water height h0 for x > 0. The
water level is h = 0 for x < 0 for t < 0. The dam suddenly collapses
at t = 0. Using a shallow water approximation calculate h(x, t) and
u(x, t) for t > 0.
4.11 Answers
1.
∂ρ ∂ρ ∂u
+u +ρ =0
∂t ∂x ∂x
1 2
∂u ∂u ∂p
ρ +u + =0
∂t ∂x ∂x
∂s ∂s
+u =0
∂t ∂x
These equations are valid for a one dimensional unsteady frictionless
isentropic flow.
101
Addition and substraction of the mass and momentum equation yields:
1 2 1 2
∂u ∂u ∂u 1 ∂p ∂p ∂p
+u ±c ± +u ±c =0
∂t ∂x ∂x ρc ∂t ∂x ∂x
or: 1 2 1 2
∂u ∂u 1 ∂p ∂p
+ (u ± c) ± + (u ± c) = 0.
∂t ∂x ρc ∂t ∂x
As by definition: "
dp dp
=d
ρc ρc
we can write this equation in the form:
1 2 1 2
∂u ∂u ∂Γ ∂Γ
+ (u ± c) ± + (u ± c) = 0.
∂t ∂x ∂t ∂x
or: 1 2
∂(u ± Γ) ∂(u ± Γ)
+ (u ± c) = 0.
∂t ∂x
with: "
dp
Γ= .
ρc
102
Figure 4.26: C + characteristic with a negative slope due to supersonic flow
in negative x-direction.
6. The path of the piston travels back in the time, which is impossible.
Also some of the C + travel back in the time providing information
form the future to the present. Please note that if the piston travels
back in time there will be two simultaneous positions (x) for a given
time (t).
7. A particle path is the line in the (x, t) plane described by the differ-
ential equation dx/dt = u.
8. The flow parameters (p, c, u) are uniform in the region. This implies
that the characteristics are straight lines because the slopes dx/dt =
(u + c) and dx/dt = u − c are constant. A constant slope corresponds
to a straight line. (See figure 4.27)
103
9. A simple wave is a region that is reached by waves emanating from a
uniform region. As a consequence the other family of characteristics
are straight lines. See figure 4.28 Demonstration: Let us assume that
10. When there are no independent time nor length scales in a flow the
solution should depend on a dimensionless combination of parameters
involving both time t and space coordinate x. In one dimensional fric-
tionless isentropic flows the only dimensionless combination we can
obtain is x/(c0 t) where c0 is a characteristic speed of sound. Such a
self-similar solution is generated when a uniform region is abruptly
perturbed. There are three types of waves generated by such pertu-
bations (see figure 4.29):
-the expansion fan
-the shock wave
-the contact discontinuity.
104
The expansion fan is a simple wave with (straight) characteristics
Figure 4.29: Self-similar solutions of the 1-D wave problem. Two pressure
waves (shock wave or expansion fan) and a contact discontinuity are formed.
emanating from a point in the (x, t) diagram. The self similar shock
is due to a sudden compression and should have a constant velocity
(straight line in the (x, t) diagram). The contact line is a particle path
separating two regions of fluids which can have different entropies. It
is again a straight line when we consider a self similar solution. In the
Rieman problem two uniform states (p1 , c1 , u1 ) and (p2 , c2 , u2 ) prevail
each at one side of a contact line at t = 0. Three waves are emanating
from this contact point. In general this will be two pressure waves and
a contact discontinuity.
11. Linearization is allowed when u′ /c0 << 1 and p′ /(ρ0 c20 ) << 1. In
linear approximation we have:
# $
∂ ∂
+ (u0 ± c0 ) (u′ ± Γ′ ) = 0
∂t ∂x
By definition:
dp
dΓ =
ρc
hence for small perturbations of a uniform region:
p′
Γ′ =
ρ0 c 0
105
where we assume dp = c2 dρ because the flow is isentropic.
Furthermore we have for an isentropic flow of an ideal gas with dp =
(γp/ρ)dρ and c2 = γRT :
1 2 1 2
dρ 1 dp 1 dρ dT 1 dρ dc
= = + = +2
ρ γ p γ ρ T γ ρ c
This yields:
γ − 1 dρ dc
= .
2 ρ c
So that: "
2dc
Γ= .
γ−1
As for a perfect gas γ is constant we find:
2c
Γ=
γ−1
106
2c 2c0
16. Along a C + characteristic we have u + γ−1 = u0 + γ−1 . Expanding
2 c0
to vacuum if u0 = 0 we find c = 0 so that umax = γ−1 = 3 c0 for
an unsteady expansion. For a steady frictionless
√ expansion
. we have
u2
from Bernoulli h + 2 = h0 so that umax = 2 h0 = 2 cp T0 =
. √ √
c0 2/(γ − 1) = c0 3. This is a factor 3 lower than the maximum
velocity achievable in an unsteady expansion.
17. The C + characteristics emerging from the uniform region x < 0 and
t < −x/c0 reach the left side of the closed valve. They carry the
message:
u + Γ = u0 + Γ 0 .
In linear approximation we can linearize the equations. The perturba-
tion are defined by: u′ = u − u0 and p′ = p − p0 . We have furthermore
Γ′ = p′ /(ρ0 c0 ). See figure 4.30. This implies:
p′
u′ + =0
ρ0 c 0
Figure 4.30: Linear theory of response of water flow in pipe to the abrupt
closure of a valve.
from the valve meet the C + emerging from the original uniform region,
107
which implies:
p′
u′ − = −2u0
ρ0 c 0
And:
p′
u′ + = 0.
ρ0 c 0
This implies that we find a new uniform region for x < 0 and t >
−x/c0 for which: u′ = −u0 and p′ = ρ0 c0 u0 as at the valve. At
the right side of the valve x > 0 and t > x/c0 we find: u′ = −u0
and p′ = −ρ0 c0 u0 . For a water flow with u0 = 1 m/s we find p′ =
15 bar at the left of the valve and p′ = −15 bar at the right. As
p′ /(ρ0 c20 ) = O(10−3 ) << 1 a linear approximation is certainly valid at
the right hand side. The negative pressures found at the right hand
side p = p0 + p′ = −14 bar imply that cavitation can occur. In such a
case the linear approximation fails because ρ′ /ρ0 = O(1). The largest
pressure difference we can have accros the valve is of the order of
30 bar. With a cross section of 10−2 m2 this implies a force of 3 kN
on the valve. If cavitation occurs this force is reduced to 1.6 kN . If
you do not understand this explanation, please read sections 4.2 and
4.3.
18. In linear approximation (valid for (p0 − p1 )/p0 << 1) we find at the
wall the message:
p′
u′ + = 0.
ρ0 c 0
Carried by the C + characteristics emanating from the uniform region
x < −c0 t. At the wall we have the boundary condition:
p′ = p1 − p0 .
p′ p0 − p1
u′ = − = .
ρ0 c 0 ρ0 c 0
The intersection of the C + and C − in the region x < 0 and t > −x0 /c0
yields:
p′
u′ + = 0.
ρ0 c 0
For the C + from the original uniform region and:
p′ p0 − p1
u′ − =2 .
ρ0 c 0 ρ0 c 0
108
For the C − coming from the wall. As a consequence we have a uniform
region x < 0 and t > −x/c0 with:
p′ = p1 − p0
and:
p0 − p1
u′ = .
ρ0 c 0
At a point (x, t) within the simple wave region we have a perturbation
! "
% & p0 −p1 t− cx
′ ′ ′ x ′
(p , u ) given by p (x, t) = p1 t − c0 − p0 and u (x, t) = ρ0 c0
0
.
For large pressure differences p0 − p1 we consider a perfect gas. In that
Hence we have:
# $ γ−1
c p 2 γ
= .
c0 p0
Consequently at x = 0 for t > 0 we have:
# $ γ−1
p1 2 γ
c = c0 .
p0
109
And: / # $ γ−1 0
2 c0 p1 2 γ
u= 1− .
γ−1 p0
A general solution of the simple wave region can be obtained by follow-
ing C − characteristics with slope dxdt = u(0, tw ) − c(0, tw ) along which
u = u(0, tw ) and c = c(0, tw ) where tw is the time at which the C −
leaves the wall at x = 0. If p1 is contant the state found at the porous
wall can be extended to a uniform region t > (u − c)t left of the wall.
The original uniform region is delimited by t < −x/c0 . Between these
two uniform regions, we a a simple wave region corresponding to an
expansion fan. In the expansion fan the C − characteristics are straight
lines emanating from the origin, so that:
x
= u − c.
t
Combining this equation for given (x, t) with the message carried by
the C + emanating from the original uniform region:
2c 2 c0
u+ =
γ−1 γ−1
yields:
2 c0 γ − 1 %x&
c= −
γ+1 γ+1 t
And:
x 2 x
u=c+ = (c0 + ).
t γ+1 t
19. As there are no independent time nor length scales in this problem we
expect a self similar solution, which can consist of two pressure waves
(expansion fan or shock wave) and a contact discontinuity. We do not
have a complex wave region. As we have an expansion to vacuum,
the compression wave (shock) and the contact discontinuity vanish.
We have only an expansion fan, which is a simple wave. As we have a
uniform region for x > 0 the characteristics travelling from the uniform
region into the simple wave region are C − and they carry the message
(see figure 4.32):
2c 2 c0
u− =− .
γ−1 γ−1
With γ = 5/3 we have 2/(γ − 1) = 3:
u − 3 c = −3 c0 .
The other family characteristics in the simple wave region are the C +
emerging from the origin (x, t) = (0, 0). As these characteristics are
110
straight lines with the slope dx/dt = u + c we have:
x
= u + c.
t
From these equations we find for the expansion fan:
3 %x &
u= − c0
4 t
and:
1 %x &
c= + 3 c0 .
4 t
The fan is delimited on the right by the line x = c0 t on which u = 0
(limit of uniform stagnant region). On the left it is limited by the
line x = −3 c0 t corresponding to the maximum expansion (c = 0 and
umax = −3 c0 ). The pressure follows from:
# $5
c
p = p0 .
c0
20. We assume that the wave in the trombone is a simple wave propagating
in a uniform region (p0 , c0 , u0 = 0). This assumption is reasonable
because p−p 0
p0 < 1. We assume air to be a perfect gas with γ = 7/5.
−
The C characteristics emerging from the uniform region carry the
message:
2c 2 c0
u− =− .
γ−1 γ−1
Assuming an isentropic flow, we have:
# $ # $ γ−1
c p 2γ
= .
c0 p0
111
Figure 4.33: Pressure measured at the end of the slide of a trombone. We
clearly observe a shock wave at fortissimo level.
21. A gradual inlet of the tunnel increases the rise time of the pressure
in the tunnel (decrease ∂p/∂t), which increases the shock formation
distance. A good design should avoid shock wave formation, which is
an environmental nuissance (as the sonic boom of aircrafts).
112
22. For M0 > 0 we have an expansion of the gas in the pipe. As there
are no independent time or length scales we expect an expansion fan
as simple wave in the region −c0 t ≤ x ≤ 0. At x = 0 we have the
message carried by the C + emerging from the uniform region:
u + 5c = 5c0 .
u + 5c = 5c0 .
u − 5c = u0 − 5c0 .
113
At the wall we have u = 0 so that c = c0 − u0 /5. The limiting C +
characteristics of the new uniform region has the slope dx/dt = u+c =
c0 − u0 /5 and passes through the origin: x = (c0 − u0 /5)t. Between
this characteristics and the limit of the old uniform region x = c0 t we
have an expansion fan with straight characteristics:
x
= u + c.
t
In this fan region we have:
1% x &
c= 5c0 + − u0
6 t
and: # $7
c
p = p0 .
c0
On the other side of the valve x = −ϵ < 0 we have a sudden compres-
sion involving a shock wave. We cannot use the method of character-
istics. As we expect a self similar solution, the shock should have a
time independent velocity us . We expect furthermore a uniform region
between the shock and the valve −us t < x < 0. Using the continuity
of mass flux over a shock ρw = ρ̂ŵ we find a shock relation for the
velocities w and ŵ in the reference frame moving with the shock. The
valve the flow velocity vanishes u = 0 for t > 0 . We find using the
shock relations in a framework moving with the speed us of the shock
(w = u0 − us and ŵ = −us ):
24. We have four lines delimiting the various wave regions: Two C + char-
acteristics:
-Line I defined by x = L + (u0 + c0 )t.
114
-Line II defined by x = (u0 + c0 )t.
Two C − characteristics:
-Line III defined by x = (u0 − c0 )t.
-Line IV defined by x = L + (u0 − c0 )t.
Right from line I (x > L + (u0 + c0 )t) we have the original uniform
region. Left from line III (x < (u0 − c0 )t) we have the original uniform
region. Below lines II and IV (t < L/(2c0 )) we have a uniform pertur-
bation p′ = Deltap and u′ = 0. Above lines II and IV (t > L/(2c0 ))
we have a uniform region p′ = 0 and u′ = 0. The two regions left
are simple waves. In the simple wave above line IV and between lines
I and II we have a perturbation p′ = ∆p/2 and u′ = ∆p/(2ρc0 ). In
the simple wave above line II and between lines III and IV we have a
perturbation p′ = ∆p/2 and u′ = −∆p/(2ρ0 c0 ).
At le right side of the piston the C − from the uniform region x > c0 t
bring the message:
u − c = −c0 .
At the left side of the piston the C + from the uniform region x < −c0 t
bring the message:
u + c = c0 .
The equation of motion of the piston is:
1 2
dup − + up 3 1 up 3
m = A[p − p ] = Ap0 (1 − ) − (1 + ) .
dt c0 2 c0
115
and:
∂p
0=− − ρg.
∂z
Integration of the z-component yields:
p − patm = ρg(h − z)
where z = 0 is the bottom of the flow and h the height of the free
surface. Applying a integral mass conservation on a slab between x
and x + ∆x we have:
∂h∆x
= u(x)h(x) − u(x + ∆x)h(x + ∆x).
∂t
Dividing by ∆x and taking the limit for ∆x → 0 we find:
∂h ∂u h
=− .
∂t ∂x
Applying the momentum equation for z = 0 we find:
∂u ∂u ∂h
+u = −g
∂t ∂x ∂x
√
Using the notation c = gh we can write these equations as:
A
∂u ∂u g ∂h
+u + c =0
∂t ∂x h ∂x
and: A 1 2
g ∂h ∂h ∂u
+u +c = 0.
h ∂t ∂x ∂x
Addition and substraction yields:
# $1 " A 2
∂ ∂ g
+ (u ± c) u± dh .
∂t ∂x h
Hence: " A
g
Γ= dh = 2 c.
h
√
27. We have an expansion fan for x < t g h0 . The maximum velocity√
−
√ C message u − 2c = −2c0 = −2 gh0 ,
of the water is given by the
which√implies umax = −2√ gh0 . Hence the expansion fan is delimited
by t gh0 > x > −2 t gh0 . Within the expansion fan we have:
x
=u+c
t
and:
u − 2c = −2c0 .
116
Hence:
. 1 ?x @ 1 ?x . @
c= g h= + 2c0 = + 2 g h0 .
3 t 3 t
or:
1 ?x . @2
h= + 2 g h0 .
9 t
and:
x 2 ?x . @
u= −c= − g h0 .
t 3 t
Appendix A
p̂ 2γ 2
A1 : p =1+ γ+1 (M − 1)
M 2 −1 û
A4 : 1 − M̂ 2 = 2γ ; M̂ = ĉ
1+ γ+1 (M 2 −1)
7
u−û 2 M 2 −1 T
A5 : M̂1 ≡ ĉ = γ+1 M T̂
A6 : ρu = ρ̂û,
A7 : p + ρu2 = p̂ + ρ̂û2 ,
1 2
A8 : 2u + h = 12 û2 + ĥ,
combined with the following expression for the enthalpy of a calorically per-
fect gas:
γ p γ γ p̂ γ
h = γ−1 ρ = γ−1 RT , ĥ = γ−1 ρ̂ = γ−1 RT̂ .
The solution of the problem is very much simplified by the condition that
the medium in the tube is initially at rest. The disturbance generated at
x = 0 propagates in the +x-direction. There are no reflections or other
disturbances which travel in the −x-direction. This is an example of a so
called simple wave, a concept that will appear to be very important in the
analytical description of non-linear waves. In this case we have a so-called
right-running simple wave. Such a simple wave has two important proper-
ties:
117
Bibliography
[Courant (1948)] R. Courant and K.O. Friedrichs, Supersonic flow and shock
waves, Interscience Pub., NY, 1948.
[Courant (1985)] Courant, R., Friedrichs, K.O., Supersonic Flow and Shock
Waves, Springer, 1985. (Original edition: 1948)
[Jaumotte 1971] Jaumotte, A.L., (ed), Chocs et Ondes de Choc, I, II, Mas-
son, 1971.
[Kundu (1990)] P.K. Kundu, Fluid Mechanics, Academic Press, San Diego,
1990.
118
[Landau (1987)] L.D. Landau and E.M. Lifshitz, Fluid Mechanics, Course
of Theoretical Physics Vol. 6, Second Edition, Pergamon Press, Oxford,
1987.
[Lighthill (1955)] M.J. Lighthill and G.B. Whitham, On kinematic waves .2.
A theory of traffic flow on long crowded roads, Proc. Roy. Soc. London,
Series A-345 (1955) pp. 317-345.
[Van Dongen (2007)] Van Dongen, M.E.H. (ed.), Shock Wave Science and
Technology Library, Vol. 1. Multiphase Flows. Springer, 2007.
119
[Zel’dovic (2002)] Ya.B. Zel’dovic and Yu.P. Raiser, Physics of shock Waves
and High-Temperature Hydrodynamic Phenomena, Dover Pub., NY,
2002.
120