lournal of Fluids and Structures (1988) 2, 425-435

AMPLIFICATION OF PRESSURE FLUCTUATIONS

D U E TO F L U I D - S T R U C T U R E INTERACTION
G. D. C. KUIKEN
Delft University of Technology, Laboratory for Aero- and Hydrodynamics,
2628 AL Delft, The Netherlands
(Received 17 November 1986 and in revised form 13 January 1988)

For a single blocked pipeline subjected to an oscillating flow, the fluid-structure

interaction is discussed taking into account not only the pressure and flow conditions at the
entry and exit of the tube, but also boundary conditions for the radial and longitudinal tube
motions at the entry and exit. In a frequency analysis the pressure amplification at the exit
of the tube is compared with the pressure at the entry of the tube. It is shown that
fluid-structure interaction--typified by allowing the tube to move freely in axial direction
at the exit--should be taken into account in order to obtain a realistic analysis of the
frequency response of a pipeline.

1. INTRODUCTION
AMPLIFICATIONOF PRESSUREFLUCTUATIONS has to be examined if there is an exciter
present somewhere in the system. Exciters can for example be (centrifugal) pumps,
oscillating valves or the interaction of guide vanes and impeller vanes in tur-
bomachinery. In the interaction with the pipeline network an excitation may happen if
frequencies are generated near the eigenfrequencies of the network system. Frequency
analysis of a pipeline network is therefore important to avoid the undesired
consequences of pressure amplification such as excessive vibrations or inaccuracy in the
measurement of a volume flow at some point in the pipeline network.
The frequency analysis of a pipeline network is very complicated. Not only do the
time-dependent Navier-Stokes equations for compressible, viscous and heat-
conducting fluids have to be considered, but also the equations for (prestressed)
circular cylindrical shells have to be taken into account [1]. Simplifications of the
governing equations are obtained by investigation of the order of magnitude of the
various contributions in the equations and by comparison with numerical results using
the original set of equations [2, 3]. The low frequency approximation and non-
prestressed shell equations for isotropic tubes are quite adequate for showing that
boundary conditions for displacements of the tube at the exit are important for the
frequency analysis of pipeline systems. A simplified but general low frequency solution
of the fluid equations will be given. Using these solutions the dispersion equation will
be derived for an elastic tube filled with a compressible Newtonian fluid. In the low
frequency approximation this equation has two different solutions, representing two
different wave propagation modes. Both modes are taken into account in this paper in
a frequency analysis of a single pipeline of finite length to discuss the interaction of the
two modes, whereas in the classical approach of the frequency analysis of pipeline
systems only one mode is considered. It is noted that having obtained the solutions for
a single pipeline element a 4 x 4 matrix equation can be constructed in contrast with
0889-9746/88/050425 + 11 $03.00 (~) 1988 Academic Press Limited
426 o.D.C. KUIKEN

the classical 2 x 2 matrix equation, which serves as the fundamental element for
building an overall matrix equation for the whole network.
In Section 2, the basic equations of the fluid and the shell are given. In the next
section the dispersion equation is derived and the low frequency solutions for the
pressure, the flow, the radial and axial displacements of the tube are specified. In
Section 4 of this paper, the fluid-structure interaction is discussed for a single blocked
pipeline subjected to an oscillating flow. In a frequency analysis, the pressure
amplification at the exit of the tube is compared with the pressure at the entrance of
the tube. The influence of fluid-structure interaction is discussed by taking into
account axial motions of the tube at the tube exit.

2. THE BASIC EQUATIONS FOR THE WAVE PROPAGATION

IN FLUID-FILLED TUBES
The pulsatile flow of a Newtonian compressible, viscous and heat-conducting fluid is
determined by the linearized equations of continuity, of motion and the equation of
energy, supplemented by a constitutive equation for the density. Denoting the pressure
by p, the axial and radial components of the velocity vector v by vx and Vr, the fluid
density by p, the viscosity by ~/, the axial, radial and time coordinates by x, r, and t
respectively, these equations reduce in the low frequency approximation to the
following equations of motion in the axial and the radial direction:

"
Dv~
Dt Dp
3x t- ~_r [~r~rJ,
Dv~'~ (1)

av.~ = _3,0 + r/ 1 5 (FDVr~_7 , (2)

P Dt Dr r ~ \ Dr~
and to the equation of continuity;
1 3,o ~v~ l a { ( v 3T)}
pc 2 3t ~-~-x + -r- Dr r o~+ Prr 09 -~r =0, (3)

where G is the isentropic velocity of sound, v the kinematic viscosity ~I/P, otp the
coefficient of cubic thermal expansion, Pr the Prandtl number defined by rlcp/~., cp
denoting the specific heat at constant pressure and & the thermal conductivity. The
Prandtl number indicates the relative importance of heat transfer with respect to
viscosity. Its magnitude is O(1) for gases, and viscosity and heat transfer have to be
considered simultaneously for gases. On the other hand, for non-conducting fluids
having large Pr, the thermal contribution to the wave propagation velocity can be
neglected almost always [3], as will be assumed in this paper.
A general solution of the equations (1)-(3) is obtained by Fourier or Laplace
analysis. Assuming that v and p vary harmonically in the axial direction x and the time
t, only one term of the Fourier decomposition of the velocities v,, Vr and the pressure p
has to be considered. Hence, a term of the solution of the fluid equations (1)-(3) in the
frequency-domain is assumed to be of the form:
x
[v,,Vr, p]=[f),(r), f)r(r),l~(r)]exp{io9(t-c) }, (4)

where o9 = 2zrf, where f is the frequency of the pulsation, is a real constant denoting
the circular frequency and c is the complex propagation velocity. The flow is assumed
 AMPLIFICATIONOF PRESSUREFLUCTUATIONS 427
to be fully developed, which means that the amplitudes indicated by.the hat (^) are
independent of x. Substitution of (4) into the fluid equations, ordinary Bessel
differential equations of zeroth- and first order result. For axisymmetric flows varying
harmonically in x and t the general solution [3], satisfying the condition that the
amplitudes are finite for r -- 0, becomes

f~x(r) = ; c ° ko{l-AFo(i3/2OCR)}, (5)

vr(r) - ~Co~ifl{ (k2- fl2os)Fl(flflOSR) - gFl(i3/2OCR)}, (6)

/)(r) =Po, (7)

where A is a constant of integration and F0(~y), Fl(~y) are the zeroth- and first-order
members of the functions F,,(~y) defined by
2nn! J~ (~y)
F.(¢y) = ~%(~)
(8)

In equation (8) Jn(~y) denotes the nth-order Bessel function of the first kind. In the
low-frequency approximation the pressure wave reduces to a plane pressure wave (7).
The various parameters in equations (6) and (7) are defined by
Co (2zf] m, = 2zfR Co
ko = --,
C
ol = R
\V/
fl
CO
, flos = - - ,
Cs
(9)

where R is a scale for r, for which in the discussion of pipeline systems the radius of a
tube can be taken, and Cois a reference velocity, which depends on the fluid properties
as well as on the tube properties.
The dynamic and kinematic boundary conditions describe the interaction between
the fluid and the tube. The dynamic boundary conditions are the equations of motion
for a circular cylindrical shell loaded by the forces exerted on the wall by the fluid at
the internal surface of the tube. For the wave propagation in straight pipes the
membrane shell equations are sufficient, since the bending deformation is much smaller
than the extensional deformation. Denoting by ux and Ur the axial and radial
components of the wall displacement vector u, the membrane equations in the x- and
r-directions for an elastic isotropic tube loaded by the pressure and shear stresses
exerted on the wall by the fluid at r = R are:

pwh ~-~ - 1 -- ~ 2 ~x + 3x z ] - ~ / [ - ~ - r + -3X

- J r=R ' (10)

~2U r Eh 1 5u~] 3vr]

pwh-~ 1=~2 ( ~ ur q-/2 ~ X ] + [ p -- 2r/-~--r]r=g , (11)

where p,,, E, t*, and h denote the density, Young's modulus, Poisson's ratio and the
thickness of the wall, respectively.
With the linearization of the equations of motion correspond linear boundary
conditions at the wall. Since terms of O(h/R) are neglected, this wall may be situated
at r = R. The linearized kinematic boundary conditions follow easily using the
adherence condition, i.e., that the fluid layer next to a solid surface sticks to that
surface. The adherence condition is firmly established. Hence, the kinematic boundary
428 G . D . C . KUIKEN

conditions are given by

~ux ~Ur
Vx r = R = ~ - r=R and l}rr=R=-~ r=R" (12)

3. THE DISPERSION EQUATION FOR AN ELASTIC PIPE

The linearized kinematic and dynamic boundary conditions can be satisfied only for
definite values of the complex wave velocity vector. These complex values are
determined by the solution of the dispersion equation, yielding the propagation
velocities [1-5]. The dispersion equation has in general an infinite number of solutions,
but in the low frequency approximation only the two fundamental modes propagate.
These modes are usually named after the early investigators Young [6] and Lamb [7],
dating back to 1808 and 1898, respectively. The two kinematic boundary conditions
(12) for the radial and axial velocity at the wall and the two dynamic boundary
conditions (10) and (11) for the motion of the tube wall in the radial and axial direction
yield a set of four homogeneous equations in the unknown amplitudes. These
equations can only be solved if the determinant of the coefficients of the unknown
amplitudes is zero. For an isotropic elastic tube the following fourth-order algebraic
dispersion equation in the inverse dimensionless wave propagation velocity ko is
obtained:

k4(1 - F)(1 - gE)B' + kZo[{E#B' - ½(B' + K')}F - 2B'

+-~(B'+K')(1-F) + F-2-~ {l+½flZs(B'+K')}=O, (13)

where the accented variables are dimensionless elastic and mass properties of the pipe:

B'=l_l~ 2 1 + ~ , 0 Rflz' (14)

where K and Pt denote the bulk modulus of the fluid and the density of the tube Wall,
respectively. The reference velocity Co in ko = co/c becomes equal to the "waterham-
mer" wave speed,

~/ Kip (15)
Co= 2RK"
1 + E--h-

The function F in equation (13) is defined by

2/1(o:) (16)
F - od0(o:) "
For large values of o: the asymptotic expansion of F is

F=~ 1 2o:il/2 ~- O(o:-2) as o:---~% (17)

with i the imaginary unit. For an inviscid fluid F---~0, and the dispersion equation
reduces to those derived by Lin and Morgan [8], and to the long wavelength
approximation of Korteweg's [9] dispersion equation.
 AMPLIFICATION OF PRESSURE FLUCTUATIONS 429

In the long wavelength approximation the radial inertia of the pipe is neglected.
Neglecting the radial inertia means that in the dispersion equation (13) it is assumed
that K' <<B', hence, ( B ' + K') reduces to B'. To satisfy this condition it follows that
neglecting K' with respect to B' in (13) is justified only if the wavelength of the tube
sound wave is much larger than the radius of the tube. This illustrates and specifies the
notion of the long wavelength approximation. A numerical solution of the inviscid
equations--using the method of characteristics--has been obtained by Walker and
Phillips [10]. From their numerical results, they also conclude that radial inertia does
not play a major role in the dispersion of the pressure pulse for low frequencies.
The dispersion equation is a quadratic equation in k 2. Its solution yields two modes,
each propagating upstream and downstream. The mode with the lowest propagation
velocity is the Young or "waterhammer" mode and the other mode is the Lamb or
"precursor" mode. Having determined the waterhammer and the precursor wave
solutions of the dispersion equation--indexed by Y and L--the pressure, the.
volume-flux ~ , the displacements in radial direction and the velocity of the wall in axial
direction yield:
p(x) = p ( R ) e i ~ [ - D 1 sinh Fyx + D2 cosh Frx - D3 sinh FLX + D4 cosh FLX], (18)
~rR2
• = - - p ( R ) e i ° ~ [ k o y ( 1 - A y F ) { D I cosh Fyx - l)2 sinh Fyx}
pCo
+ koL(1 - A L F ) { D 3 cosh FLX -- 04 sinh FLX}], (19)
R
ur(x) = 2--~c2p (R)ei°"[{(1 - ArF)k2or - fi2os}{ - D I sinh Fyx + D: cosh Frx}

+ {(1 - ALF)k2L -- fl2os}{-D3 sinh FLX + D4 cosh FLX}], (20)

~ux(x) = ~ p ( R )eiO,,[koy(1 _ Ay){DI cosh Frx - D2 sinh Fyx}

5t pco
+ koL(1 -- AL){D3 cosh FLx - D4 sinh FLX}], (21)
where F is the propagation constant defined by
iwko i~o
F - -- - , (22)
CO C

the constant of integration A--introduced in equations (5) and (6)--becomes

{
1 - 2/.t - ~1 -
\ 1 + Eh /2RK]-~oo
A- (23)
1 - 2/2 - (1 - F)
and D1-D4 are constants of solution determined by the axial boundary conditions.

3.1. AXIAL BOUNDARY CONDITIONS

In the classical theory of pipeline networks the mean values of the pressure and the
volume flux at two cross sections of the pipeline are related to each other using only
the Young or "waterhammer" mode. The axial boundary conditions at two cross
sections determines the values of solution constants D1 and/)2. With these values, a
2 × 2 transfer matrix is built to obtain the fundamental pipeline element to calculate
the frequency analysis of a complicated network. In this classical approach nothing is
430 ~. D. C. KUIKEN

stated about the motion of the pipe itself. Fluid-structure interaction-the influence of
the motion of the pipeline on the dynamic behaviour of the pipeline--is therefore
ignored in the classical approach. Hence, since in fluid-structure interaction both wall
displacements are taken into account as well, a 4 × 4 transfer matrix will result.
To illustrate the influence of the fluid-structure interaction a frequency analysis of
the pressure will be given for a fluid-filled blocked tube for various axial boundary
conditions. It is assumed that at the entrance x = 0 the tube is driven by an oscillating
flow produced by a piston operating with the speed Vp at the frequency f , and t h a t - - a t
the entrance--the tube is fixed axially. At the closed end x = L, two different boundary
conditions might be considered. One consistent with the classical approach, the other
with fluid-structure interaction. The classical boundary condition for a blocked tube is
that there is no flow at the tube end. This means that the axial motion at the tube end
is assumed to be zero-i.e., the tube is fixed axially. In many practical situations tubes
are not fixed; e.g., when a piping system has several bends or is connected to other
flexible pipes. In these cases there is an interaction between the pipe axial motion and
the motion of the internal fluid. For the single blocked pipeline the piston that drives
the fluid causes a motion of the tube end, which has an influence on the fluid motion
itself, and vice versa. The boundary condition consistent with this fluid-structure
interaction is that the tube is free to move axially at the exit.
In the classical case the explicit boundary conditions are:
¢ -----zR2Vpe i°t for x = 0,
(24)
¢P=0 for x = L ,
which are sufficient to determine the constants D1 and D2 associated with the Young
mode. Using equation (24) and realizing that the Lamb mode is not considered in the
classical case, the constants DI-D4 in (18)-(21) are:

D1 = Vppco 1
p(R) koy(1 - A y F ) ' (25)
D2 = Da coth FyL, (26)
/93 = 0, (27)
D 4 = 0. (28)

Taking into account the Lamb mode, but still assuming that the tube is rigidly
anchored at the end, the boundary conditions for the two-mode solution are
= ~R2Vpe i°~t for x = 0,
Ux=0 for x = 0 ,
(29)
0=0 for x = L,
~u__2x___
0 for x = L.
9t
The four constants D1-D4 become in this case:

D1 = Vppco i -A L
(30)
p ( R ) koy(Ay - AL)(1 -- F ) '
D 2 = D 1 coth FyL, (31)
 AMPLIFICATION OF PRESSURE FLUCTUATIONS 431

VpPCo 1 -- Ay
D3-- ~(R) koa(Ay- Ac)(1 - F ) ' (32)

/)4 = D3 coth FLL. (33)

For fluid-structure interaction the boundary conditions at x = L are completely
different from those in the previous two cases. One boundary condition follows from
the consideration that the tube is closed by a rigid plate. With this plate inserted at the
exit, no radial displacements Ur are possible at the tube end. For the blocked tube with
an end that is free to move axially the other boundary condition follows from the
adherence condition--the fluid sticks to the closing plate at the tube exit. This gives the
boundary condition that the flow is determined by the axial velocity of the plate that
blocks the tube at its exit. The boundary conditions for the investigation of
fluid-structure interaction are therefore
= ~R2Vpei~ for x = 0,
ux=0 for x = 0 ,

= zrR2 9ux for x = L, (34)

3t
Ur=O for x = L .
Since the constants D1 and D3 are determined by the boundary conditions at x = 0,
they are already given by (30) and (32). By applying equation (34) the constants D2
and D4 become for the free-end condition
D1
D2=
1 - coth FyL tanh FLL/C

X [coth FyL tanh FLL 1 - 1/At 1 ](35)

C 1 1/AcsinhFgL coshFcL '
D3
D4-
1 - C coth FLL tanh FvL
[ 1--I/AL . I ]
x cothFLL-CtanhFrL-_-Z~rcosh~flfsinhFLL , (36)

where the coefficient C stands for

(1 - ALF)k2oL - fiZo,korAr (37)
C = (1 - ArF)k2or - flos
2 koLAL

4. AMPLIFICATION OF PRESSURE FLUCTUATIONS

To discuss the influence of the various boundary conditions stated in the previous
section, figures will be given of the absolute value of the ratio of the pressure at the
end of the tube to the pressure at the entrance for a particular tube filled with water.
The properties of the Newtonian fluid and the elastic tube are listed in Table 1. The
elastic properties are for a plastic material, which was used in an experimental test
section in the Delft Hydraulics Laboratory. Possible viscoelastic behaviour has been
neglected. In view of the dimensionless quantities involved, one might also think of an
432 G. D. C. KUIKEN

TABLE 1
Fluid and tube properties and dimensions

Fluid Tube

p = 1000 k g / m 3 p, = 1100 k g / m 3
y = 10 -6 m 2 / s E = 19 x 108 N / m 2
K = 21 × 1 0 8 N / m 2 /t = 0-35
R = 13-5 m m
h = 3.175 mm
L = 12.0 m

elastic water-filled zinc tube with h = 0.443 mm, p, = 7876 kg/m 3, E = 13.6 x 109 N / m 2,
and all other properties the same as above.

4.1. O N E M O D E - - C L A S S I C A L CASE

Using equations (25)-(28), the ratio of the pressure at the end of the tube to the
pressure at the entrance becomes

p (x = L) = 1 (38)
p (x = 0) cosh ryL"
This result is well known. Figure 1 shows for the above specified fluid-filled tube a
logarithmic plot of the absolute values of (38) for the frequency range 1 H z - < f - <
1000 Hz. Figure 1 represents the classical result of a frequency analysis o f a pipeline
and is reproduced in the following figures to serve as a reference.
In the figure many resonances are noticed. Resonances are defined as an oscillatory
condition that leads to an extraordinary pressure amplification at some point of a fluid
line network. For a block inviscid fluid-filled tube having no mass, resonances occur if
the length of the tube is an odd integer multiple of a quarter of the wavelength of
Young's mode; hence, for frequencies at odd-integer multiples of Co/4L. The

IO3 -~B . . . .
F6

102~ 2
~ I
i!
LO i I
i
-.j
L-
./
i
i
10 - I _

10 - 2 _J
3 4 5 6 789

I I0 10z 103
Frequency (Hz)

Figure 1. O n e - m o d e frequency response of the pressure ratio.

 AMPLIFICATION OF PRESSURE FLUCTUATIONS 433
"waterhammer" wave speed Co of the elastic tube is 449 m/s, yielding resonances in
odd-integer multiples of 9.26 Hz. The resonances in Figure 1 occur at odd-integer
multiples of 10.8 Hz. This somewhat higher value is due to the combined effects of the
mass of the tube, the viscosity of the fluid and Poisson's ratio.

4.2. T w o MODES--CLAsSICAL BOUNDARY CONDITIONS

For a rigidly anchored blocked pipe driven by an oscillating flow the solution is given
by (18)-(21) with (30)-(33). The ratio of the pressure at the end of the tube to the
pressure at the entrance is given by
p (x = L) _ 1 [1 -- kor(1 - A~.) sinh FyL/koL(1 -- Am) sinh FLL ]
p (x=0) cosh F .L Li - J" (39)
This result shows that the full solution is more complicated, but also that the
resonances are close to those obtained from the classical one-mode solution. Figure 2
is a plot of the numerical calculations.
The figure shows that the classical prediction of the resonances differs not very much
from the resonances obtained by using the Young and the Lamb mode solution
together. It is noticed that there is a slight change of the first two resonances and that
the Lamb mode produces extra resonances for frequencies at integer multiples of
25.4 Hz. At 76 Hz the classical theory predicts a resonance, whereas in the interaction
of the Young mode with the Lamb mode this resonance is cancelled. In the low
frequency range for frequencies less then 7 Hz the pressure ratio is 20% smaller than
predicted by the classical approach. In spite of these differences, it is concluded that for
design purposes the classical theory can be used for rigidly anchored tubes to obtain an
estimate of the frequency response of a pipeline network.

4.3. Two MODES---FLUID--STRUCIVREINTERACTION

The frequency response changes drastically if fluid-structure interaction is taken into
account. The explicit formula for the pressure ratio becomes equation (18) divided by

'°3~!
F
I0 2

i)
IO

~l
ii

i0-1.

10-2_
, ,2 ~ 14 ,5 ,~7,~,~, ~ ? ~ ,~ ,6,7,8,9 ~ 3, ~ ~_Y_~
I I0 102 I0 3
Frequency (Hz)

Figure 2. Two-mode frequency response of the pressure ratio for a rigidly anchored tube.
434 G . D . C . KUIKEN

103 -

102 -

II
I0-

--4

i0 -r

10-2I
2 3 4 ~ 6 789 2 ~s 4 ~ e 789 ~ s 4 ~ 6 7~9

I I0 102 103
Frequency (Hz)

Figure 3. Two-mode frequency response of the pressure ratio for a tube with a free moving end.

(/)2 +/)4) with the constants D1 to/)4 given by (30), (32), (35) and (36). The result is
quite different from equation (38) and (39) indeed, as the numerical results show
(Figure 3).
The differences are remarkable. Resonances occur only in the low frequency range.
There is a strong resonance at 5.99 Hz, much earlier than the occurrence of the first
resonance at 10.8'Hz for a rigidly anchored tube. The second resonance happens to
appear just between the first and the second one of the rigidly anchored tube. These
results emphasize that neglect of the fluid-structure interaction might be disastrous.
The free motion of the tube end prevents the higher frequencies from being amplified,
while such amplifications are noticed for tubes that are rigidly anchored.

5. C O N C L U D I N G REMARKS
In practice it is economically not advisable to anchor all pipes rigidly in a pipeline
network. From the results discussed in Section 4 is evident that fluid-structure
interaction should be considered in a frequency analysis of the network, although the
more complicated 4 × 4 transfer matrices have to be used. Fluid-structure interaction
diminishes the sound generation in piping systems, since it attenuates most high
frequency resonances occurring in rigidly anchored systems. On the other hand, it can
damage the pipeline system. Particularly the strong amplifications in the low frequency
range give rise to high loadings on the connecting pipes. Designers of pipeline systems
should be aware that the first resonance in non-rigidly anchored systems occurs at
much lower frequencies than for rigidly anchored systems.

REFERENCES
1. G. D. C. KUIKEN1984 Wave propagation in a thin-walled liquid-filled initially stressed tube.
Journal of Fluid Mechanics 141, 289-308.
2. G. D. C. KU~I~EN1984 Approximate dispersion equations for thin-walled liquid-filled tubes.
Applied Scientific Research 41, 37-53.
 AMPLIFICATION OF PRESSURE FLUCTUATIONS 435

3. G. D. C. KUIKEN 1984 Wave propagation in compliant tubes containing a heat-conducting

viscous fluid. Thesis, Werktuigbouwkunde, WTHD 165, Delft University of Technology,
Delft.
4. G. D. C. KUIKEN 1984 Wave propagation in fluid lines. Applied Scientific Research 41,
69-91.
5. S. I. RUBINOV and J. B. KELLER 1978 Wave propagation in a viscoelastic tube containing a
viscous fluid. Journal of Fluid Mechanics 88, 181-203.
6. T. YOUN6 1808 Hydraulic investigations, subservient to an intended Croonian lecture on the
motion of blood. Philosophical Transactions of the Royal Society (London) 98, 164-186.
7. H. LAMB 1898 On the velocity of sound in a tube, as affected by the elasticity of the walls.
Memoirs and Proceedings--Manchester Literary and Philosophical Society 42, 1-16.
8. T. C. LIN and G. W. MORGAN 1956 Wave propagation through fluid contained in a
cylindrical, elastic shell. Journal of the Acoustical Society of America 28, 1165-1176.
9. D. J. KORTEWEG 1878 Uber die Fortpflanzungesgeschwindigkeit des Schalles in elastischen
R6chren. Annalen der Physik und Chemie, Neue Folge 5, 525-542.
10. J. S. WALKER and J. W. PHILLIPS 1977 Pulse propagation in fluid-filled tubes. Journal of
Applied Mechanics 44, 31-35.