Computers and Structures 85 (2007) 844–851

www.elsevier.com/locate/compstruc

Water hammer with fluid–structure interaction in thick-walled pipes

A.S. Tijsseling *

Department of Mathematics and Computer Science, Eindhoven University of Technology, P.O. Box 513, 5600 MB Eindhoven, The Netherlands

Received 1 June 2006; accepted 2 January 2007

Available online 26 February 2007

Abstract

A one-dimensional mathematical model is presented which describes the acoustic behaviour of thick-walled liquid-filled pipes. The
model is based on conventional water-hammer and beam theories. Fluid–structure interaction (FSI) is taken into account. The equations
governing straight pipes are derived by the cross-sectional integration of axisymmetric two-dimensional basic equations. The resulting
FSI four-equation model has small correction terms and factors accounting for the wall thickness. Exact solutions of this model show
that these corrections are important only for very thick pipes, with, say, a radius/thickness ratio smaller than 2.
Ó 2007 Elsevier Ltd. All rights reserved.

Keywords: Water hammer; Fluid transients; Fluid–structure interaction (FSI); Pipe flow; Pipe vibration; Wave speed

1. Introduction ages are taken to arrive at a convenient one-dimensional

formulation. The motivation for the research lies in the fact
Water hammer concerns the generation, propagation that many pipes in practice have relatively thick walls, for
and reflection of pressure waves in liquid-filled pipe sys- example, high-pressure pipes in chemical and nuclear
tems. These waves may have very steep fronts when gener- industries, ductile-iron pipes with concrete lining, and plas-
ated by, for example, the closure of lightly damped check tic pipes.
valves [1] or the collapse of liquid column separations [2]. The governing equations can be solved exactly in the
Steep-fronted waves are prone to excite the structural sys- time and frequency domains. The influence of wall thick-
tem and as a result make individual pipes move. Pipe ness is studied from figures and tables showing wave prop-
motion itself generates water hammer, thus invoking agation speeds, natural frequencies and pressure histories.
fluid–structure interaction. Predictions from thin- and thick-wall theories are com-
The numerical simulation of water-hammer events (with pared with experimental data obtained in a relatively thick
and without fluid–structure interaction) is usually based on pipe.
one-dimensional mathematical models assuming thin-
walled cylindrical pipes [3]. The ratio e/R of wall-thickness 2. Governing equations for the liquid
to pipe-radius is then small with respect to unity, circum-
ferential (hoop) stresses are uniform in the wall cross-sec- Liquid pipe-flow is described by the following two-
tion, and radial stresses are neglected. The present paper dimensional equations [4]:
considers water hammer with axial/radial vibration of Continuity equation (2D)
thick-walled pipes. The distributions of hoop and radial
1 op ovz 1 o
stresses in the wall are given as quasi-static functions of þ þ ðrvr Þ ¼ 0 ð1Þ
K ot oz r or
the internal and external pressures. Cross-sectional aver-
Equation of motion in axial direction (2D)
*
Tel.: +31 40 247 2755; fax: +31 40 244 2489. ovz op
qf þ ¼0 ð2Þ
E-mail address: [email protected] ot oz

0045-7949/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compstruc.2007.01.008
 A.S. Tijsseling / Computers and Structures 85 (2007) 844–851 845

Equation of motion in radial direction (2D)

ovr op
qf þ ¼0 ð3Þ
ot or
The pressure p, the axial velocity vz and the radial velocity
vr are functions of time t and of the cylindrical co-ordinates
z (axial) and r (radial). The circumferential co-ordinate (u)
is omitted because of axial symmetry. Although the liquid Fig. 1. Definition sketch, side view (z–r plane): stresses acting on pipe
density qf is taken constant, the liquid compressibility (elas- wall.
ticity) is present in the first term of the continuity equation
(1) in which K is the bulk modulus of the liquid. Gravity,
viscosity and convective terms [vz  cf , Eq. (43a),
dz=dt ¼ cf ] are ignored herein, so that a linear model is
obtained. For an extended nonlinear model the reader is
referred to [5].
To arrive at a one-dimensional formulation, in accor-
dance with classical water-hammer theory, the Eqs. (1)
and (2) are multiplied by 2pr, integrated with respect to r
from 0 to R, the inner radius of the pipe, and divided by Fig. 2. Definition sketch, cross-sectional view (r–u plane): normal stresses
Af ¼ pR2 . The result is: acting on element of pipe wall.
Continuity equation (1D)
1 oP oV 2 Equation of motion in axial direction (2D)
þ þ vr j ¼ 0 ð4Þ
K ot oz R r¼R ou_ z orz
qt ¼ ð8Þ
Equation of motion in axial direction (1D) ot oz
oV oP Equation of motion in radial direction (2D)
qf þ ¼0 ð5Þ
ot oz ou_ r 1 oðrrr Þ ru
qt ¼  ð9Þ
in which ot r or r
Z R Z R with u_ z and u_ r the axial and radial pipe velocities, rz and rr
1 1
V ¼ 2prvz dr and P¼ 2prp dr ð6Þ the axial and radial stresses, ru the hoop stress, and qt the
pR2 0 pR2 0 mass density of the wall material. The thickness of the pipe
are the cross-sectionally averaged axial velocity and pres- wall is denoted by e. Gravity, convective terms [u_ z  ct ,
sure, respectively. The first integral in Eq. (6) is the flow Eq. (43b), dz=dt ¼ ct ], shear forces, bending stiffness, rota-
rate (discharge), the second the axial force due to pressure. tory inertia and structural damping are neglected herein.
The radial equation of motion (3) is multiplied by 2pr2, To arrive at a one-dimensional formulation, the Eqs. (8)
integrated with respect to r from 0 to R, and divided by and (9) are multiplied by 2pr, integrated with respect  to r
2pR2. The result is: from R to R + e, and divided by At ¼ 2p R þ 12 e e:
Equation of motion in radial direction (1D) Equation of motion in axial direction (1D)
1 ovr ou_ z orz
q R j þ pjr¼R  P ¼ 0 ð7Þ qt ¼ ð10Þ
2 f ot r¼R ot oz
Equation of motion in radial direction (1D)
In deriving this equation it has been assumed that
rvr ¼ Rvr jr¼R , which is in accordance with the continuity ou_ r Rþe R 1
qt ¼ rr jr¼Rþe    rr jr¼R  ru
equation (1) if compressibility is disregarded ðK ¼ 1Þ ot 1
ðR þ 2 eÞe 1
R þ2e e R þ 12 e
and axial inflow (vz) is concentrated at the central axis by
ð11Þ
means of a point source (singularity). Walker and Phillips
[6] found the same equation by using asymptotic expan- in which
sions. If one assumes that Rvr ¼ rvr jr¼R , which corresponds Z Rþe
1
to a linear distribution of the radial liquid velocity, the u_ z ¼   2pru_ z dr;
factor 1/2 (representing the total liquid mass) becomes 2p R þ 12 e e R
Z Rþe
1/4 Ref. [7] or 1/6 Ref. [8]. 1
u_ r ¼   2pru_ r dr;
2p R þ 12 e e R
3. Governing equations for the pipe Z Rþe
1
rz ¼   2prrz dr and
2p R þ 12 e e R
Pipe motion is described by the following two-dimen- Z Rþe
sional equations [4]; see Figs. 1 and 2 for definition 1
ru ¼ ru dr ð12Þ
sketches: e R
846 A.S. Tijsseling / Computers and Structures 85 (2007) 844–851

are averaged values of u_ z ; u_ r ; rz and ru. A single bar denotes Axial stress–displacement relation (2D)
cross-sectional area averages, a double bar denotes line(ar)
ouz
averages. The first integral in Eq. (12) is the ‘‘flow rate’’ in rz ¼ E þ mru þ mrr ð18Þ
oz
axial pipe movement, the third is the axial force in the pipe
wall. The fourth integral is the hoop force in the pipe wall and, applying the transformation between Eqs. (16) and
per unit area of (axial) wall surface. (17).
The equations of motion relate axial velocities to axial Axial stress–velocity relation (1D)
stresses (Eq. (10)), and radial velocities to radial and hoop
stresses (Eq. (11)). Stress–velocity relations will complete orz ou_ z oru orr
¼E þm þm ð19Þ
the mathematical model. Stress–strain relations are pro- ot oz ot ot
vided by the generalised Hooke’s law. For a three-dimen-
in which
sional isotropic solid the normal strains ez, eu and er
Z Rþe
depend linearly on the normal stresses rz, ru and rr: 1
ru ¼   2prru dr and
Stress–strain relations (3D) 2p R þ 12 e e R
1 1 Z Rþe
ez ¼ ½rz  mðru þ rr Þ; eu ¼ ½ru  mðrz þ rr Þ; 1
E E rr ¼   2prrr dr ð20Þ
2p R þ 12 e e R
1
er ¼ ½rr  mðrz þ ru Þ ð13Þ
E
4. Liquid–pipe coupling
which is equivalent with
E The liquid and pipe equations are coupled by means of
rz ¼ ½ð1  mÞez þ mðeu þ er Þ or
ð1 þ mÞð1  2mÞ boundary conditions representing the contact between
E m liquid and pipe wall on the interface at r ¼ R. Outside
rz ¼ ðez þ meu Þ þ rr ;
1  m2 1m the pipe a constant pressure, Pout, is assumed to exist.
E m The interface conditions are
ru ¼ 2
ðeu þ mez Þ þ rr and
1m 1m
rr jr¼R ¼ pjr¼R and rr jr¼Rþe ¼ P out ð21Þ
rr ¼ Eer þ mðrz þ ru Þ; ð14Þ
u_ r jr¼R ¼ vr jr¼R and u_ r jr¼Rþe ¼ ðvr Þout ð22Þ
where E is Young’s modulus of elasticity and v is Poisson’s
where ðvr Þout is the radial velocity of the external fluid.
ratio.
Buried pipes are not considered herein.
The strain–displacement (e–u) relations are:
The dynamic conditions (21) give the fluid pressures
ouz ur our acting on the pipe wall. The kinematic conditions (22)
ez ¼ ; eu ¼ and er ¼ ð15Þ
oz r or prescribe the adherence of solid and fluid. Except for its
The axial stress rz is expressed in displacements after sub- constant pressure, Pout, the fluid outside the pipe is not
stitution of Eq. (15) in the first of Eq. (14): modelled, so that ðvr Þout is not known. For this reason,
Axial stress–displacement relation (2D) and because the liquid inside the pipe will not be modelled
two-dimensionally, the relations (22) do not provide suit-
 
E ouz 1 oðrur Þ able boundary conditions for pipe equation (17) [herein
rz ¼ ð1  mÞ þm ð16Þ replaced by Eq. (19)]. Nevertheless, Eq. (17) may be useful
ð1 þ mÞð1  2mÞ oz r or
in other applications. For example, with u_ r jr¼R ¼ 0 and
Differentiation with respect to t, multiplication by 2pr, inte- u_ r jr¼Rþe ¼ 0, Eq. (17) describes walls that are fixed in the
gration with respect
 to r from R to R + e, and division by radial direction [9, p. 64].
At ¼ 2p R þ 12 e e leads to: The conditions (21) are substituted in the equation of
Axial stress–velocity relation (1D) radial pipe motion (11). The condition (22), at r ¼ R, is
orz E substituted in the liquid Eqs. (4) and (7). After the substi-
¼ tutions and a rearrangement of terms, eight basic equations
ot ð1 þ mÞð1  2mÞ
" # remain for the eight variables P, V, rz, u_ z , ru, ur, rr and
ou_ z Rþe R pjr¼R . The eight equations stem from, successively, the rela-
 ð1  mÞ þm  u_ r j  m  u_ r j
oz R þ 12 e e r¼Rþe R þ 12 e e r¼R tions (5), (4), (10), (19), (7) and (11), and the second and
ð17Þ third equations in (13) and (15). In summary:
Liquid, axial motion
Eq. (16) is basic for a full 2D analysis, but Eq. (17) is not
oV 1 oP
appropriate for the present 1D investigation as will be ex- þ ¼0 ð23Þ
plained in Section 4. An alternative equation is found by ot qf oz
substituting the axial strain from Eq. (15) in the first of oV 1 oP 2
þ þ u_ r j ¼ 0 ð24Þ
Eq. (13), yielding oz K ot R r¼R
 A.S. Tijsseling / Computers and Structures 85 (2007) 844–851 847

Pipe, axial motion means that pipe cross-sections remain almost plane when
ou_ z 1 orz stretched axially. Then, the quantity rz jr¼R in Eq. (32) can
 ¼0 ð25Þ be replaced by rz . Note that the stress ru þ rr , caused by
ot qt oz
an internal pressure P, is a function of z and t only [see
ou_ z 1 orz m o Appendix A, Eqs. (A.1) and (A.2)], so that according to
 þ ðru þ rr Þ ¼ 0 ð26Þ
oz E ot E ot Eq. (18) the hoop and radial stresses do not introduce
Liquid, radial motion r-dependency in rz. Four coupled basic equations remain,
1 ou_ r Eqs. (23), (32), (25) and (26), for the four unknowns
pjr¼R ¼ P  qf R j ð27Þ P ; V ; rz and u_ z :
2 ot r¼R
Liquid, axial
Pipe, radial motion
oV 1 oP
ou_ r Rþe R 1 þ ¼0 ð36Þ
qt ¼   P out þ   pjr¼R  ru ot qf oz
ot 1
R þ 2e e 1
R þ2e e R þ 12 e    
oV 1 2 R 1 þ Re oP 2m orz
þ þ þ þm ¼ ð37Þ
ð28Þ oz K E e 2 þ Re ot E ot
r
ur ¼ ½ru  mðrz þ rr Þ ð29Þ Pipe, axial
E
our 1 ou_ z 1 orz
¼ ½rr  mðrz þ ru Þ ð30Þ  ¼0 ð38Þ
or E ot qt oz
Substitution of expression (27) in Eq. (28) gives: ou_ z 1 orz mR 1 oP
Equation of motion in radial direction  ¼ ð39Þ
oz E ot Ee 1 þ 12 e
ot
  R
1 ou_ r 1 ou_ r
qt R þ e e þ qf R2 j This is the thick-wall FSI four-equation model. If the usual
2 ot 2 ot r¼R thin-wall assumption is made, in which e/R is neglected
¼ RP  ðR þ eÞP out  eru ð31Þ with respect to unity, the four-equation model of Skalak
[10] remains.
Substitution of expression (29) in Eq. (24) gives:
Continuity equation 6. Wave propagation speeds
oV 1 oP 2 o
þ þ ½ru jr¼R  mrz jr¼R  mrr jr¼R  ¼ 0 ð32Þ
oz K ot E ot The four eigenvalues k of the hyperbolic system of
partial differential equations (36)–(39) represent the pro-
5. FSI four-equation model for thick-walled pipes pagation speeds – in two directions – of axial waves in
straight liquid-filled pipes. The magnitudes are obtained
For long wavelengths (long compared to the radius of from the following bi-quadratic dispersion relation,
the pipe), accelerations in radial direction are negligible, k4  c2 k2 þ c2f c2t ¼ 0 ð40Þ
so that the radial inertia terms in Eq. (31) can be left out
with
[8,10–16]. Local effects near sharp wave fronts [17] are also !
neglected in the long-wave approximation. A quasi-static 2 2
qf R 1
c ¼ 1 þ 2m c2f þ c2t ð41Þ
relation between the hoop stress and the internal pressure qt e 1 þ 12 e
R
is the result,
This leads to modified (because of FSI) squared wave
R Rþe speeds
ru ¼ P  P out ð33Þ
e e 1h 1=2
i
k21;2 ¼ c2  ðc4  4c2f c2t Þ ð42aÞ
Neglecting radial liquid inertia means that p equals P; the 2
pressure is uniform in each cross-section of the pipe. and
Eq. (33) is confirmed in Appendix A, Eq. (A.7), where 1h 1=2
i
the quasi-static stress distribution in a pressurised ring is k23;4 ¼ c2 þ ðc4  4c2f c2t Þ ð42bÞ
2
given. Furthermore, from the Eqs. (A.10) and (A.6), it fol- where k1 and k3 are positive, and k2 and k4 are negative.
lows that, The constants
o R 1 oP ( " ! !#)1
ðru þ rr Þ ¼ ð34Þ 2 1 2 R m2 1 þ Re
ot e 1 þ 12 Re ot c f ¼ qf þ 1 þ þm
  K E e 1 þ 12 Re 2 þ Re
o R 1 þ Re oP
ðru jr¼R Þ ¼ þ ð35Þ ð43aÞ
ot e 2 þ Re ot
and
Eq. (34) is substituted in Eq. (26), and Eq. (35), together
with rr jr¼R ¼ P , is substituted in Eq. (32). The r-depen- E
c2t ¼ ð43bÞ
dency of ez, and hence of rz, is considered to be small, which qt
848 A.S. Tijsseling / Computers and Structures 85 (2007) 844–851

are the squares of the classical pressure wave speed Some results are presented for the Dundee water-filled steel
[3,18,19] and the bar velocity [20, p. 493], respectively. pipe [33–35] with data given in Table 1, where the thick-
Taking the limit to zero of e/R, one obtains ness/radius ratio e=R ¼ 0:152. The test pipe is 4.5 m long
and has an inner radius of 26 mm. A five metre long solid
Ee c2t
k21;2 ¼ ! 0 and k23;4 ¼ ð44Þ steel rod hits the pipe axially at one of its closed ends. The
2qf Rð1  m2 Þ 1  m2
water in the pipe has a static pressure of typically 2 MPa to
which are the squares of the conventional pressure wave prevent the occurrence of cavitation. Pressures, strains and
speed in flexible tubes filled with incompressible liquid structural velocities were measured at different positions
and the plate velocity [9, p. 81], respectively. Taking the along the pipe.
limit of e/R to infinity, one obtains The wave speeds calculated from Eq. (42) are shown in
   1 Fig. 3 as a function of e/R. The pressure wave speed k1 in
2 1 2 Fig. 3(a) increases, from zero at e=R ¼ 0 to the tunnel
k1;2 ¼ qf þ ð1 þ mÞ and k23;4 ¼ c2t ð45Þ
K E velocity of 1452 m/s (Eq. (45)) at e=R ¼ 1, because the
which are the squares of the conventional pressure wave pipe hoop stiffness increases with e/R. The stress wave
speed in circular tunnels [3,18,19] and the bar velocity, speed k3 in Fig. 3(b) decreases, from the thin plate velocity
respectively. of 4793 m/s (Eq. (44)) at e=R ¼ 0 to the bar velocity of
It appears that the coupled pressure and stress waves 4587 m/s (Eq. (45)) at e=R ¼ 1, because the wall thickness
travel without dispersion. Gromeka [21], Lamb [22], Ska- increases. Thick-wall theory predicts lower pressure wave
lak [10] and Lin and Morgan [11,12] analysed this problem speeds, because of additional mass of the pipe wall, and
in the frequency domain, and Bürmann [23], Stuckenbruck lower stress wave speeds, because of a heavier wall. The dif-
et al. [24] and Leslie and Tijsseling [25] in the time domain. ferences between thick-wall theory (thick solid lines) and
Refs. [26,27] give extensive reviews of the subject. thin-wall theory (where e/R terms are neglected with
respect to unity) (thin broken lines) are less than 1.5% for
the pressure wave speed and less than 0.05% for the stress
7. Exact solutions
wave speed, over the whole e/R range. These small
differences return in the calculated natural frequencies
The FSI four-equation model (36)–(39) can be solved
listed in Table 2, where the lower pressure wave speeds pre-
exactly in the time [28,29] and frequency [30–32] domains.
dicted by thick-wall theory lead to lower frequencies, also
for the ‘‘structural’’ frequencies measured at 485 Hz
Table 1
Geometrical and material properties of Dundee test pipe, closed at both
[ k3 =ð2 LÞ] and 968 Hz [ k3 =L]. Table 2 lists calculated
ends and freely suspended results obtained without [Table 2a] and with [Table 2b] tak-
Steel pipe
ing into account the lumped masses closing the pipe. The
Length L = 4.51 m thick-wall results in the fourth column are furthest away
Inner radius R = 26.01 mm from the measured data in the first column, but the last col-
Wall thickness e = 3.945 mm umn of Table 2b shows the most systematic deviation from
Young’s modulus E = 168 Gpa the experiment in the sense that all calculated frequencies
Mass density qt = 7985 kg/m3
Poisson’s ratio m = 0.29
are 2–3% lower than the measured ones. In this respect
End mass at z = 0 m0 = 1.312 kg one could remark that the statically measured modulus of
End mass at z = L mL = 0.3258 kg elasticity of 168 ± 5 GPa [33, p. 28] is rather low for stain-
Wall area At = 694 mm2 less steel.
Water The time-domain results, obtained without including the
Bulk modulus K = 2.14 Gpa lumped end masses, and shown in Fig. 4, confirm the
Mass density qf = 999 kg/m3 observations above. Thick-wall theory predicts a slightly
Flow area Af = 2125 mm2
slower transient, which slowly departs from the thin-wall

1500 4675

4650
1400
lambda_3
lambda_1

4625

1300
4600

1200 4575
0 0.5 1 1.5 2 0 0.5 1 1.5 2
e/R e/R

Fig. 3. Wave speeds (in m/s) (Eq. (42)) as function of thickness/radius ratio for thick-wall theory (thick solid line), thin-wall theory (thin broken line) and
limit values (Eq. (45)) (dotted lines): (a) pressure wave speed k1, (b) axial stress wave speed k3.
 A.S. Tijsseling / Computers and Structures 85 (2007) 844–851 849

Table 2a (A.1), (A.4) and (A.5)] and rr jr¼Rþe ¼ P out into the second
Natural frequencies (in Hz) of axial vibration of closed water-filled steel of Eq. (13), and ignoring the (usually) unknown axial stres-
pipe with free closed ends
ses rz, one obtains:
Measurement Thin-wall Difference Thick-wall Difference !
[31] calculation (%) calculation (%) 1 R 1
without end without end eu ¼ ðP  P out Þ  ð1  mÞP out ð46Þ
E e 1 þ 12 Re
masses masses
173 172 0.6 170 1.7 where P and Pout are the internal and external pressure,
289 286 1.0 284 1.7 respectively. The thick-wall equation (46) had to be used
459 453 1.3 449 2.1
485 493 +1.6 491 +1.2
to accurately predict the hoop strains in the Dundee test
636 633 0.5 628 1.3 pipe [33, p. 44–47]. The relative pressure is derived from
750 741 1.2 735 2.0 the hoop strain by
918 907 1.2 899 2.1  
968 980 +1.2 976 +1.0 e 1 e  
P  P out ¼ E 1þ eu  eu;out ð47Þ
Calculations without end masses.
R 2R
where eu;out ¼ ð1  mÞP out =E is the hoop strain when
Table 2b P ¼ P out .
Natural frequencies (in Hz) of axial vibration of closed water-filled steel
pipe with free closed ends
8. Conclusion
Measurement Thin-wall Difference Thick-wall Difference
[31] calculation (%) calculation (%)
with end with end A rigorous derivation of one-dimensional equations
masses Masses describing fluid–structure interaction mechanisms in the
173 171 1.2 169 2.3 axial/radial vibration of liquid-filled pipes has been pre-
289 285 1.4 283 2.1 sented, thereby taking the thickness of the pipe wall into
459 453 1.3 449 2.2 account through the averaging of hoop and radial stresses.
485 471 2.9 470 3.1 FSI coupled wave speeds have been formulated and inves-
636 626 1.6 621 2.4
tigated. It has been shown, for one example concerning a
750 740 1.3 734 2.1
918 906 1.3 899 2.1 water-filled steel pipe, that the thin-wall assumption is valid
968 944 2.5 941 2.6 for long waves in fairly thick pipes. Liquid frictional and
Calculations with lumped end masses. structural damping effects are usually small and have been
neglected herein, but these could be included in the analysis
[4,5,31,36].
20

Appendix A. Two-dimensional stress distribution in a

dynamic pressure (bar)

10 pressurised ring

0
In the present investigation the radial inertia forces in
both liquid and pipe wall are neglected. For the liquid this
means that the pressure is uniform in each pipe cross-sec-
-10 tion, whereas for the pipe a quasi-static stress distribution
across the thickness of the pipe wall is assumed. This par-
-12 ticular stress distribution, given in [20, pp. 68–71] and
0 1 2 3 4 5 6 7 8 9 10
time (ms)
attributed to Lamé [37], is integrated here over the pipe
wall cross-section.
Fig. 4. Dynamic pressure in Dundee single pipe experiment (pressure Consider the two-dimensional axially symmetric stress
3.376 m away from the impact end). Thick solid line: thick-wall theory;
problem of a circular ring subjected to an internal pressure
thin broken line: thin-wall theory; thin solid line: experiment.
P and an external pressure Pout. The ring, shown in Fig. 5,
result, in phase and amplitude, as time increases. The inclu- has inner radius R and thickness e. The hoop and radial
sion of the end masses, of friction and damping mecha- stresses have the form
nisms [36], and of a non-instantaneous excitation, will A
ru ¼  þ2C ðA:1Þ
give more realistic computational results, but exact solu- r2
tions like those in Fig. 4 cannot be obtained. A
rr ¼ þ 2C ðA:2Þ
Wall thickness may be of importance in the determina- r2
tion of fluid pressures from (measured) hoop strains. The where the constants of integration, A and C, are deter-
hoop strain eu is calculated at the outer radius r ¼ R þ e, mined from the boundary conditions
because strain gauges are normally glued to the external
surface of the pipe. Substituting r/ jr¼Rþe [from Eqs. rr jr¼R ¼ P ðz; tÞ and rr jr¼Rþe ¼ P out ðA:3Þ
850 A.S. Tijsseling / Computers and Structures 85 (2007) 844–851

The averaged value of rr, defined in Eq. (20) and calculated

with (A.2), (A.4) and (A.5), is
Z Rþe
1
rr ¼   2prrr dr
2p R þ 12 e e R
 
ln 1 þ Re
¼   A þ 2C
R þ 12 e e
" 2     #
R2 ðR þ eÞ ln 1 þ Re  R þ 12 e e
¼  2  2 P
2e R þ 12 e
" 2    2   #
R2 ðR þ eÞ ln 1 þ Re  1 þ Re R þ 12 e e
þ  2 P out
2e2 R þ 1e 2
Fig. 5. Definition sketch of stress distribution in a pressurised ring.
ðA:9Þ
The sum of the averaged values of ru and rr, defined in Eq.
It follows that
(20), and calculated with (A.1), (A.2), (A.4) and (A.5), is
2 Z Rþe
R2 ðR þ eÞ ðP out  P Þ 1
A¼   ðA:4Þ ru þ rr ¼   2prðru þ rr Þ dr ¼ 4C
2 R þ 12 e e 2p R þ 12 e e R
 2
R2 P  ðR þ eÞ2 P out R 1 R 1 þ Re
2C ¼   ðA:5Þ ¼ P P out ðA:10Þ
2 R þ 12 e e e 1 þ 12 Re e 1 þ 12 Re
The stress solutions ru and rr are independent of Poisson’s For thin-walled pipes, e  R, the approximation
ratio v, so that they are valid for both (axial) plane stress lnð1 þ e=RÞ  e=R is made. Terms of the order of e/R,
and (axial) plane strain situations. The sum ru þ rr is equal and higher, are neglected with respect to unity. The aver-
to 4C and therefore independent of r. aged values of ru and rr, given by the expressions (A.7),
The value of ru at r ¼ R follows from (A.1), (A.4) and (A.8) and (A.9), then become
(A.5), R
ru ¼ ru ¼ ðP  P out Þ ðA:11Þ
A e
ru jr¼R ¼  þ 2C 3 1
R2 rr ¼  P  P out ðA:12Þ
4 4
R Rþe 1 þ Re
¼ P P out þ ðP  P out Þ ðA:6Þ
e e 2 þ Re
References
The averaged value of ru, defined in Eq. (12) and
[1] Koetzier H, Kruisbrink ACH, Lavooij CSW. Dynamic behaviour of
calculated with (A.1), (A.4) and (A.5), is large non-return valves. In: Papworth M, editor. Proceedings of the
Z 5th international conference on pressure surges. Cranfield,
1 Rþe A
ru ¼ ru dr ¼  þ 2C UK: BHRA; 1986. p. 237–43.
e R RðR þ eÞ [2] Bergant A, Simpson AR, Tijsseling AS. Water hammer with column
R Rþe separation: A historical review. J Fluid Struct 2006;22(2):135–71.
¼ P P out ðA:7Þ [3] Wylie EB, Streeter VL. Fluid transients in systems. Englewood Cliffs,
e e New Jersey, USA: Prentice Hall; 1993.
[4] Tijsseling AS. Fluid–structure interaction in case of waterhammer
The other averaged value of ru, defined in Eq. (20) and with cavitation. Ph.D. Thesis, Delft University of Technology,
calculated with (A.1), (A.4) and (A.5), is Faculty of Civil Engineering, Communications on Hydraulic and
Z Rþe Geotechnical Engineering, Report No. 93-6, ISSN 0169-6548, Delft,
1 The Netherlands, 1993. Available from: www.darenet.nl/en/page/
ru ¼   2prru dr
2p R þ 12 e e R language.view/search.page.
  [5] Zhang YL, Vairavamoorthy K. Analysis of transient flow in pipelines
ln 1 þ Re with fluid–structure interaction using method of lines. Int J Numer
¼   A þ 2C
Meth Eng 2005;63(10):1446–60.
R þ 12 e e
" 2     # [6] Walker JS, Phillips JW. Pulse propagation in fluid-filled tubes.
R2 ðR þ eÞ ln 1 þ Re þ R þ 12 e e ASME J Appl Mech 1977;44(March):31–5.
¼  2 P [7] Kellner A, Schönfelder C. The effect of fluid/structure-interaction on
2e2 R þ 12 e pressure pulse loads on pipes. 3R Int 1982;21(8):443–9 [in German].
" 2    2   # [8] Barez F, Goldsmith W, Sackman JL. Longitudinal waves in liquid-
R2 ðR þ eÞ ln 1 þ Re þ 1 þ Re R þ 12 e e filled tubes – I. Theory, II. Experiments. Int J Mech Sci
  2 P out
2e2 R þ 1e 1979;21(4):213–36.
2 [9] Kolsky H. Stress waves in solids. New York, USA: Dover Publica-
ðA:8Þ tions; 1963.
 A.S. Tijsseling / Computers and Structures 85 (2007) 844–851 851

[10] Skalak R. An extension of the theory of water hammer. Trans ASME [25] Leslie DJ, Tijsseling AS. Wave speeds and phase velocities in liquid-
1956;78(1):105–16. filled pipes. In: Půlpitel L, editor. Proceedings of the 9th International
[11] Lin TC, Morgan GW. A study of axisymmetric vibrations of Meeting of the IAHR Work Group on the Behaviour of Hydraulic
cylindrical shells as affected by rotatory inertia and transverse shear. Machinery under Steady Oscillatory Conditions, Paper E1, Brno,
ASME J Appl Mech 1956;23(June):255–61. Czech Republic, September 1999.
[12] Lin TC, Morgan GW. Wave propagation through fluid contained in [26] Tijsseling AS. Fluid–structure interaction in liquid-filled pipe systems:
a cylindrical, elastic shell. J Acoust Soc Am 1956;28(6):1165–76. a review. J Fluids Struct 1996;10(2):109–46.
[13] Meißner E. Influence of radial inertia of a pipe wall on non-stationary [27] Wiggert DC, Tijsseling AS. Fluid transients and fluid–structure
flow phenomena in pressurized conduits. Technische Universität interaction in flexible liquid-filled piping. ASME Appl Mech Rev
München, Hydraulik und Gewässerkunde, Mitteilungen, Heft 25, 43– 2001;54(5):455–81.
62, Munich, Germany, 1978 [in German]. [28] Li QS, Yang K, Zhang L. Analytical solution for fluid–structure
[14] Schwarz W. Waterhammer calculations taking into account the radial interaction in liquid-filled pipes subjected to impact-induced water
and longitudinal displacements of the pipe wall. Ph.D. Thesis, hammer. ASCE J Eng Mech 2003;129(12):1408–17.
Universität Stuttgart, Institut für Wasserbau, Mitteilungen, Heft 43, [29] Tijsseling AS. Exact solution of linear hyperbolic four-equation
Stuttgart, Germany, 1978 [in German]. systems in axial liquid–pipe vibration. J Fluid Struct 2003;18(2):
[15] Bürmann W. Longitudinal motion of pipelines laid in the open due to 179–96.
water hammer. 3R Int 1980;19(1–2):84–91 [in German]. [30] de Jong CAF. Analysis of pulsations and vibrations in fluid-filled
[16] Kuiken GDC. Amplification of pressure fluctuations due to fluid– pipe systems. Ph.D. Thesis, Eindhoven University of Technology,
structure interaction. J Fluids Struct 1988;2(5):425–35. Department of Mechanical Engineering, Eindhoven, The Nether-
[17] Tijsseling AS, Lambert MF, Simpson AR, Stephens ML, Vı́tkovský lands, 1994, Errata: April 2000. Available from: www.darenet.nl/en/
JP, Bergant A. Wave front dispersion due to fluid–structure interac- page/language.view/search.page.
tion in long liquid-filled pipelines. In: Kurokawa J, editor. Proceed- [31] Zhang L, Tijsseling AS, Vardy AE. FSI analysis of liquid-filled pipes.
ings of the 23rd IAHR symposium on hydraulic machinery and J Sound Vib 1999;224(1):69–99.
systems, Paper 143, Yokohama, Japan, October 2006. [32] Brown FT, Tentarelli SC. Dynamic behavior of complex fluid-filled
[18] Halliwell AR. Velocity of a water-hammer wave in an elastic pipe. tubing systems – Part 1: Tubing analysis. J Dyn Syst, Measure, Ctrl
ASCE J Hydraul Div 1963;89(4):1–21. Discussed by VL Streeter in 2001;123:71–7.
89(6): 295–6. [33] Fan D. Fluid–structure interactions in internal flows. Ph.D. Thesis,
[19] Thorley ARD. Fluid transients in pipeline systems: a guide to the The University of Dundee, Department of Civil Engineering,
control and suppression of fluid transients in liquids in closed conduits. Dundee, UK, 1989.
2nd ed. London: Professional Engineering Publishing; 2004. [34] Vardy AE, Fan D. Flexural waves in a closed tube. In: Thorley ARD,
[20] Timoshenko SP, Goodier JN. Theory of elasticity. 3rd ed. Singa- editor. Proceedings of the 6th international conference on pressure
pore: McGraw-Hill; 1970. surges. Cranfield, UK: BHRA; 1989. p. 43–57.
[21] Gromeka IS. On the velocity of propagation of wave-like motion of [35] Tijsseling AS, Vardy AE. Twenty years of FSI experiments in
fluids in elastic tubes. Physical–mathematical section of the Scien- Dundee. In: Bathe KJ, editor. Proceedings of the third MIT
tific Society of the Imperial University of Kazan, Kazan, Russia, conference on computational fluid and solid mechanics. Oxford,
1883: p. 1–19 [in Russian]. UK: Elsevier; 2005. p. 1014–7.
[22] Lamb H. On the velocity of sound in a tube, as affected by the [36] Leslie DJ, Tijsseling AS. A review of modelling damping mechanisms
elasticity of the walls. Memoirs of the Manchester Literary and in coupled liquid-pipe vibrations. In: Liou JCP, editor. Symposium
Philosophical Society, Manchester, UK 1898:42(9);1–16. S-290 Water Hammer, Proceedings of the 3rd ASME & JSME joint
[23] Bürmann W. Water hammer in coaxial pipe systems. ASCE J fluids engineering conference, ASME – FED 248, Paper FEDSM99-
Hydraul Div 1975;101(6):699–715. Errata in 101(12): 1554. 6878, New York, USA: ASME, 1999.
[24] Stuckenbruck S, Wiggert DC, Otwell RS. The influence of pipe [37] Lamé G. Lessons in the theory of elasticity. Paris: Gauthier-Villars;
motion on acoustic wave propagation. ASME J Fluids Eng 1852 [in French].
1985;107(4):518–22.