Time scales and FSI in unsteady liquid-filled pipe flow

Arris S Tijsseling Alan E Vardy

Department of Mathematics and Computer Science Civil Engineering Division
Eindhoven University of Technology University of Dundee
P.O. Box 513, 5600 MB Eindhoven Dundee DD1 4HN
The Netherlands United Kingdom

ABSTRACT

Time scales are used to classify unsteady flow in liquid-filled pipes. Seven types of pipe flow
are distinguished, with time scales ranging from infinity to zero. Nothing happens when the
time scale tends to infinity: the flow is steady. Everything seems to happen when the time
scale tends to zero: all (coupled) modes of vibration of liquid and pipe are excited. The
complexity of mathematical models describing the phenomena increases as time scales
become smaller.

An attempt is made to provide and rationalise characteristic time scales, length scales and
natural frequencies. Governing equations for the fluid and structure are given in dimensional
form (but eventually should be scaled to find characteristic non-dimensional groups).
Illustrating examples are given for accelerating flow, water hammer and FSI in a single pipe.
One aim of the present investigation is to find the time scale for which 1D-FSI is of
importance.

Keywords: rigid column, water hammer, fluid-structure interaction (FSI), pipe stress

NOMENCLATURE
P fluid pressure, Pa
Scalars PDE partial differential equation
c fluid wave speed, m/s r radial coordinate, m
cs solid wave speed, m/s t time, s
D inner diameter of pipe, m u& z axial pipe velocity, m/s
E Young modulus of pipe wall material, Pa V cross-sectional average of axial fluid
e pipe wall thickness, m velocity, m/s
f Fanning or skin friction coefficient; v axial fluid velocity, m/s
frequency, Hz w radial fluid velocity, m/s
FSI fluid-structure interaction z axial coordinate, m
Im modified Bessel function of the first kind of γ angle between pipe and horizontal, rad
order m ∆P (linear) pressure rise within time ∆t, Pa
J0 Bessel function of the first kind of order 0 ∆t duration of excitation, s
J1 Bessel function of the first kind of order 1 ∆V (linear) velocity rise within time ∆t, m/s
K fluid bulk modulus, Pa λ Darcy-Weisbach friction coefficient:
L pipe length, m λ =4f

1
λn nth zero of J0 int_bdys internal boundaries
µ Poisson ratio r reversal
ν kinematic viscosity, m2/s rc rigid column
ρ fluid mass density, kg/m3 s structure, solid, pipe
ρs solid mass density, kg/m3 sh “steel hammer”
σhoop hoop stress in pipe wall, Pa su start up
σz axial stress in pipe wall, Pa wh water hammer
φ circumferential coordinate z axial direction
0 initial steady state
Subscripts ∞ final steady state
ext_bdys external boundaries

1 INTRODUCTION

Background
The authors have recently participated in a project aimed at the definition of practical
guidelines for dealing with fluid-structure interaction in industrial pipe systems. See
www.win.tue.nl/fsi. In such a project, it is wise to fall back on basic fluid principles (1). One
of these principles is to consider the time scale of the problem at hand. For unsteady flow in
pipes, this leads to a natural set of regimes. One of these regimes is water hammer with FSI;
this is the regime where pipe motion starts to influence the unsteady flow. The main objective
of this paper is to identify this particular time scale within a ranking of time scales associated
with flows in pipes that are not fully restrained (i.e. are able to move in response to internal
fluid forces). Previous efforts in this direction are described by Lavooij & Tijsseling (2), Uffer
(3) and Hamilton & Taylor (4).

Scope
We consider cylindrical pipes of circular cross-section with thin linearly-elastic walls and
filled with weakly-compressible (elastic) liquid. The flow velocities are small (e.g. less than
10 m/s for water in a pipeline), the absolute pressures are above vapour pressure, and pressure
variations caused by turbulence and vortex shedding are disregarded, as are surface and
electro-magnetic effects. Both one-way (liquid → pipe) and two-way (liquid ↔ pipe) FSI are
considered.

Outline
Section 2 classifies unsteady flows according to time scales (defined in Section 3) and gives
relevant governing equations. The FSI mechanisms in each type of flow are explained and the
equations needed in a pipe stress analysis are given. Section 3 lists and discusses the
characteristic time scales and frequencies. Section 4 illustrates and Section 5 concludes.
Appendix A gives a summary of closed-form solutions in rigid-column theory.
The Nomenclature defines the symbols; these are not declared in the text.

2 CLASSIFICATION OF FLOW

The purpose of this section is to categorise the types of flow that might be encountered in
industrial systems and to demonstrate a natural progression from each type to the next. The
corresponding flow-induced stresses and FSI mechanisms are described.

2.1 No-flow

2
Fluid
The no-flow situation gives the hydrostatic pressure in the pipe system. For a single pipe,
gravity defines the pressure according to
∂P
= ρ g sin γ (1)
∂z
Pipes
A similar relation holds for the axial tensile stress in an empty pipe:
∂σ z
− = ρs g sin γ (2)
∂z
However, a static stress analysis is more complicated than a hydrostatic pressure analysis;
non-vertical pipes sag under their own weight, causing bending moments and shear forces.
The support conditions and the flexibility of bends are crucial in a static stress analysis and
the largest load is usually due to thermal expansion. The static analysis of the coupled
compression, shear, bending and torsion of pipes and resulting anchor forces can be
important, but are not considered herein. Account is taken, however, of hydrostatic pressure
loads.

FSI
In liquid-filled pipes, the hydrostatic pressure at any axial location is counterbalanced by a
hoop stress in the pipe wall satisfying the equilibrium condition:
1D
σ hoop = P (3)
2 e
which is probably the oldest FSI formula. This hoop stress induces hoop strain and it also
induces axial stress and/or strain as a consequence of Poisson contraction. Thus, there is a link
between the fluid pressure and the axial stress/strain in the pipe wall. This is a simple form of
so-called Poisson coupling (so named because the amplitude of the interaction is proportional
to the Poisson contraction ratio µ).

A simple form of junction coupling can also exist, namely at an unrestrained closed end
where the pressure causes an axial stress in the pipe wall satisfying the equilibrium condition:
1D
σz = P (4)
4 e
The description of these examples of Poisson coupling and junction coupling as "simple" does
not imply any difference in principle from "non-simple" cases. It is merely a convenient way
of distinguishing them from cases where the coupling causes pipe movements.

2.2 Steady flow

Fluid
In steady flows, shear forces exist between the fluid and the pipe in addition to pressure
forces. The flow is driven by a piezometric pressure gradient and is resisted by skin friction at
the pipe wall. The resulting pressure gradient is in equilibrium with gravitational and friction
forces, satisfying
∂P V V
= ρ g sin γ − λ ρ 0 0 (5)
∂z 2D
Pipes
The pipes experience skin friction according to

3
 ∂σ z V V
− = ρ s g sin γ + λ ρ 0 0 (6)
∂z 8e
FSI
Skin friction acts on the whole of the interface between the liquid and the solid as a
distributed axial force. The effect is known as friction coupling. At pipe junctions such as
bends and branches directional changes of momentum and pressure changes due to local
losses cause local axial forces in the pipe wall proportional to ρ V0 V0 . These are further
examples of junction coupling. In a steady flow, however, they do not cause any movement of
the pipe (although they sustain static deflection of the pipe).

2.3 Quasi-steady flow

Fluid
In quasi-steady flows, conditions vary in time, but do so sufficiently slowly for the equations
of flow at any instant to be assumed identical to those in a truly steady flow with the same
instantaneous mean velocity. At a typical instant, the assumed equation of flow is
∂P V V
= ρ g sin γ − λ ρ (7)
∂z 2D
Pipes
The corresponding equation for the pipe is
∂σ z V V
− = ρ s g sin γ + λ ρ (8)
∂z 8e
Equations (7) and (8) differ from equations (5) and (6) only in the replacement of the steady
velocity V0 by a time-varying velocity V(t).

FSI
The pipes experience variations in skin friction, pressure gradient, momentum changes and
local losses. These progress through a succession of quasi-steady states. They may cause the
pipe to deflect, but the movements are too slow to have a significant influence on the fluid
flows. The pressures and wall stresses closely satisfy equilibrium relationships.

2.4 Rigid column

Fluid
When the fluid accelerates too rapidly for inertial forces to be neglected, the unsteady flow
ceases to resemble a succession of quasi-steady states. Nevertheless, if the accelerations are
sufficiently smooth, the velocities (and pressure gradients) may closely approximate to quasi-
steady behaviour, in the sense that the velocity along any particular pipe may be regarded as
uniform. In this condition, known as rigid-column behaviour, the governing equation of flow
is:
∂V ∂ P V V
ρ + = ρ g sin γ − λ ρ (9A)
∂t ∂ z 2D
Pipes
For this case, the axial motion of a single pipe satisfies:
∂ u& ∂σ z ( V − u& z ) V − u& z
ρs z − = ρs g sin γ + λ ρ (10)
∂t ∂z 8e

4
In practice, supports will usually prevent rigid-body pipe motion so that u& z ≈ 0.

FSI
In common with quasi-steady flows, pipe movements are assumed to have negligible
influence on the fluid flow. The pressures and wall stresses at any location again closely
satisfy equilibrium relationships. The actual values of these parameters vary in time, but the
equations relating them are the same as those for a simple steady flow. That is, equilibrium is
assumed to prevail locally.

Analytical solutions
Closed-form solutions of Eq (9A) can be obtained for a small number of special cases. These
include rigid columns driven by steps and ramps in pressure or velocity; see Appendix A. The
solutions give mathematical insight into the combined effects of friction and inertia. They
have been used herein to check numerical results.

2.5 Water hammer and 1D - FSI

Fluid
When sufficiently rapid accelerations occur, fluid compressibility effects come into play and
it ceases to be acceptable to neglect non-uniformities in velocity along any particular pipe. It
becomes necessary to take explicit account of pressure waves that propagate at the speed of
sound c in the liquid. The velocity V(z,t) becomes dependent on the axial position z, and a
continuity equation (11) must be considered in addition to the equation of motion (9). That is:
∂V ∂ P ( V − u& z ) V − u& z
ρ + = ρ g sin γ − λ ρ (9B)
∂t ∂ z 2D
∂V 1 ∂ P 2 µ ∂σ z
ρ + 2 = (11)
∂ z c ∂t E ∂t
Classical water hammer theory (5) neglects convective terms, takes u& z = 0 in Eq (9B),
σ z = 0 in Eq (11), and assumes c to depend on the support conditions of the pipes.

Pipes
Similarly, axial-stress waves will propagate at the speed of sound cs in the pipe wall. A stress
versus rate-of-strain relationship (12) must be used in addition to the equation of motion (10):
∂ u& ∂σ z ( V − u& z ) V − u& z
ρs z − = ρs g sin γ + λ ρ (10)
∂t ∂z 8e
∂ u& 1 ∂σ z µR ∂P
ρs z − 2 = − ρs (12)
∂ z cs ∂ t E e ∂t
FSI
In this flow state (i.e. water hammer in a movable pipe), the motion of the fluid is influenced
by pipe movements. This is a fundamental difference from all of the above flow states in
which equilibrium was assumed to prevail locally between fluid and solid forces.

Consider Poisson coupling, for example. The pipe wall expands radially when the internal
pressure increases. This effect, modelled quasi-statically through Eq (3), causes the dynamic
coupling terms on the right sides of the Eqs (11) and (12). The last terms in the Eqs (9B) and
(10), modelling friction coupling, depend on the relative velocity V − u& z . Junction coupling at
dead ends, elbows and branches, etc is often modelled as above, Eq (4), assuming local

5
equilibrium. More generally, however, account may need to be taken of the inertia of
additional masses at these locations. Irrespective of additional inertia terms, additional
equations (namely continuity) are needed to allow for changes in pipe length caused by the
movement of junctions.

To summarise, FSI in water hammer involves two-way coupling. Pipe vibrations affect the
fluid motion and vice versa. Moreover, this is a fundamentally interactive process, thus
explaining why an uncoupled analysis (where fluid force histories are used as input data in a
structural dynamics code for the pipes, without coupling back) may give erroneous results.

The FSI four-equation model Eqs (9B)-(12) is valid for the axial vibration of straight pipes
(1D systems). For the in-plane vibration of planar pipe systems (2D systems), four first-order
PDEs describing in-plane lateral motion need to be added. Eight lateral and two torsional
first-order PDEs need to be added for the description of general non-planar pipe
configurations (3D systems). These additional equations are presented by Tijsseling (6) and
Wiggert &Tijsseling (7).

2.6 2D - FSI

Fluid and Pipes

In 2D models the fluid-pipe system is assumed axisymmetric. The variables, like v(z,r,t) and
w(z,r,t), are functions of the radial distance from the central axis of the pipe. The simplest
model couples potential flow equations for the fluid to membrane equations for the pipes. The
effects of axisymmetric bending stiffness, shear deformation, and rotatory inertia can usually
be neglected, because these can be seen only in the vicinities of very steep wave fronts and of
anchors. Skalak (8), Lin & Morgan (9, 10), Thorley (11), Bürmann (12, 13) and De Jong (14,
Chapter 2) investigated and developed 2D-FSI models.

FSI
FSI is modelled through pressure, shear, radial stick and axial no-slip conditions at the fluid-
solid interface at r = D/2. Frequency-domain analysis shows that infinitely many coupled
radial-axial modes of vibration exist. Only two modes have a finite phase-velocity at low
frequencies; these correspond to water hammer and “steel hammer” in the 1D-FSI model.

2D-FSI aims at accurate modelling acoustic phenomena. Frequencies in the audible range
lead to applications in noise reduction, and in haemodynamics, where the large flexibility of
the vessels, the medical importance of shear stresses and the accurate calculation of flow
patterns are essential (15-18).

2.7 3D - FSI

3-D FSI is the most general case. The variables, like v(z,r,φ,t), are functions of all cylindrical
coordinates. Circumferential (lobar) modes of pipe vibration are of importance. In principle,
3D Navier-Stokes equations (acoustic, without convective terms) for the fluid and full shell
equations for the structure are needed to describe high-frequency behaviour. At high
frequencies the dynamic behaviour becomes more localised, remote boundary conditions are
less important locally, and reflections can often be ignored. Rayleigh surface waves may
appear in the solid. In practice, a statistical approach is commonly the best option (19). 3D-
FSI might be of significance for highly accurate modelling of curved pipes and for ultrasonic
phenomena (as used in wall-thickness measurement and flaw detection).

6
3 TIME SCALES

System response time scales

Mathematically, a multi-scale (or dimensional) analysis of the governing equations will reveal
the relative importance of individual terms as functions of the chosen time scales. Physically,
one speaks of response (relaxation, retardation) times in the transition from one physical state
to another, or of eigenperiods when this transfer induces vibration (or when vibration is
maintained by a driving force). In reality, in observing phenomena, the sampling rate and
duration of the observation is important. For example, if the sampling rate is large enough,
transient events will not be recorded.

Excitation time scales

The time scale of the excitation causing unsteadiness is of the utmost importance. Typical
excitation time scales might include, for example, (i) the time period of a repetitive input or
(ii) the duration of an impact force.

System response versus excitation

If the time scale of excitation is (much) larger than the time scales of system response not
much will happen; the situation will be quasi-steady. To have significant unsteadiness, the
time scale of the excitation must be smaller than the time scales of the system. The latter
determine the dynamic response of the system.

Amplitudes
It is not only the time scale that is important, but also the magnitude of excitation and
response. It is up to the analyst, what he/she thinks is significant. It is noted that events at
smaller time scales (higher frequencies) typically have smaller amplitudes (this follows from
an energy consideration) and more damping.

Definition
The time scales defined below concern system response times and eigenperiods. These have
to be compared with the time scale of the excitation.

No-flow
In the no-flow case, V ≡ 0 and the representative time scale is t = ∞. This case is an
asymptotic extreme of steady flow, valid as V0 → 0.

Steady flow
In steady flow, V is constant and the representative time scale is again t = ∞ (there are no
changes). This case is an asymptotic extreme of quasi-steady flow, valid for t → ∞ and V∞ =
V.

Quasi-steady flow
In quasi-steady flow, V varies slowly and the representative time scale is large, but t < ∞. All
changes occur sufficiently slowly for inertia effects to be neglected. This case is an
asymptotic extreme of rigid column theory; it is valid for t >> 2D/(λV).

Rigid column
Rigid column theory is valid for time scales of the order of t = 2D/(λV). Inertia significantly
impedes changes in velocity, but the elasticity of the liquid can be ignored. This case is the

7
asymptotic extreme of water hammer; it is valid for ∆t >> 4L/c, where L is a representative
pipe length and ∆t is the time scale of the excitation.

Water hammer and 1D-FSI

Representative time scales for water hammer are of the order of t = Lext_bdys/c, where the
length scale Lext_bdys represents the distance between external boundaries. These time scales
apply whether or not pipe motions are sufficient to merit the use of FSI analyses. If they are,
however, smaller time scales may exist because movements of internal boundaries can lead to
significant wave reflections that are absent when the pipe does not move. Such pipe motion
may occur when, for example, the pipe system has sufficient flexibility and the wave fronts
are sufficiently steep to create large unbalanced pressure forces. This consideration leads to
time scales of the order of t = Lint_bdys/c, where the length scale Lint_bdys represents the distance
between interaction points (junction coupling), for example movable elbows.

Although FSI is due to axial motion, the excited pipe system prefers to vibrate in its natural
modes, which have axial, lateral and torsional components. Therefore, it is convenient to
characterise wave propagation phenomena in terms of natural frequencies. The fundamental
frequencies for water hammer are c/(4L) and c/(2L) for open-closed and open-open (or closed-
closed) systems, respectively. The fundamental frequencies for axial-pipe vibration are cs/(4L)
and cs/(2L) for fixed-free and fixed-fixed (or free-free) straight pipes, respectively. The
natural frequencies of lateral and torsional pipe vibration can be found in textbooks on
structural dynamics. The natural frequencies of pipe systems strongly depend on the support
conditions. They provide a fingerprint of the system’s dynamic behaviour and they may
indicate the importance of FSI (20).

The 1-D FSI case may be regarded as an asymptotic extreme of 2D-FSI when radial changes
occur sufficiently slowly for cross-sectional values to be used. It uses only the two low-
frequency modes of the infinitely many modes existing in 2D-FSI models.

2D-FSI
The characteristic time scale t is of the order of D/c.

The ring frequency

c
f ring = s (13)
πD
is the natural frequency of a radially vibrating hoop (the hoop remains circular at all times).
This frequency is high: the ratio hoop/axial, fring/fsh, is about L/D. This is the reason that radial
inertia of the pipe wall is neglected in Eq (3) in the 1D-FSI model of Section 2.5 (21, pp. 97-
99, pp. 121-135).

3D-FSI
The characteristic time scale t is of the order of e/c.

The ovalizing frequency is:

2e
f oval ≈ f ring (14)
ρ D
D +5
ρs e
This is the cut-on frequency of the first lobar mode (14, pp. 18-23). This frequency is low
with respect to the ring frequency and very low with respect to the defined time scale for 3D-

8
FSI (the interaction between lobar and axial-radial modes becomes significant at high
frequencies). Although the influence of lobar modes on axial vibration is small at low
frequencies, the phenomenon can sometimes be of importance (22).

4 EXAMPLES

To illustrate the relative importance of friction, inertia, elasticity and FSI (Poisson coupling)
at different time scales, and to gain insight into the phenomena, calculations have been carried
out for accelerating liquid flow in a straight pipe. The dimensional test system is a steel pipe
filled with water. Turbulent flow is considered in a 200 m long pipe of 0.8 m inner diameter
and 8 mm wall thickness. Laminar flow is considered in a 20 m long tube of 0.8 mm inner
diameter. The water has mass density 1000 kg/m3, bulk modulus 2.1 GPa and kinematic
viscosity 1 mm2/s. The steel pipe has mass density 7900 kg/m3, modulus of elasticity 210 GPa
and Poisson ratio 0.3. The Darcy-Weisbach friction factor in turbulent flow is λ = 0.02. The
pressure and stress wave speeds are taken c = 1000 m/s and cs = 5000 m/s. The system is
excited by a linearly increasing pressure (or velocity) at one of its ends; the pressure remains
zero at the other end. The pressure (or velocity) rise is ∆P (or ∆V) within time ∆t, after which
the pressure (or velocity) remains constant; see Figure 1. The time interval ∆t is the important
time scale of excitation, which is to be compared with the other time scales listed in the
Tables 1 and 2.

Table 1 Representative time scales and frequencies in turbulent flow

[Numerical values provided for the particular case of water (V∞=2 m/s, c=1000 m/s) in a
steel pipe (L=200 m, D=0.8 m, cs=5000 m/s)]

Typical time scales Frequencies Eq

ρ V∞ L
start-up tsu = = 40 s (5, Eq 5-7) (15)
∆P
rigid 2D
trc = = 40 s (16A)
column λ V∞
water 2L 4L c c
0.4 s = ≤ twh ≤ = 0.8 s 1.25 Hz = ≤ f wh ≤ = 2.5 Hz (17)
hammer c c 4L 2L
steel 2L 4L c c
0.08 s = ≤ tsh ≤ = 0.16 s 6.25 Hz = s ≤ fsh ≤ s = 12.5 Hz (18)
hammer cs cs 4L 2L
c
ringing f ring = s = 2150 Hz (13)
πD
2e
f oval ≈ f ring = 18 Hz
ovalizing ρ D (14)
D +5
ρs e

9
 Table 2 Representative time scales in laminar flow
[Numerical values provided for the particular case of water (V∞=2 m/s) in a steel pipe
(L=20 m, D=0.0008 m)]

Typical time scales Eq

ρ V∞ L
start-up tsu = = 0.02 s (5, Eq 5-7) (15)
∆P
D2
rigid column trc = = 0.02 s (16B)
32ν
D2
2D t 2D = = 0.16 s (19)
4ν

Analytical solutions used in the calculations are given in the Appendices A and B.

4.1 Starting turbulent flow

Herein, turbulent flow means that skin friction is proportional to the square of the average
axial velocity. 2D effects in turbulent flow are not considered. A linear pressure ramp ∆P =
λ ( L / D) 1 2 ρ V∞2 = 0.1 bar is applied to accelerate the liquid column from rest to a final
steady-state velocity of V∞ = 2 m/s. Figure 2 shows the velocity history when the time of
excitation ∆t = 3600 seconds. The excitation is so slow that after, say, fifteen times trc = 600
seconds the situation is quasi-steady. Because the acceleration starts from zero flow, there is
no frictional resistance initially. Figure 3 shows the result for ∆t = 60 seconds, which is of the
order of trc = 40 seconds. The quasi-steady solution fails to describe the transition from the
initial steady state V0 = 0 m/s to the final steady state V∞ = 2 m/s. Figure 4 shows the result for
∆t = 1 second, which is much smaller than trc = 40 seconds so that the excitation is practically
a step load for the rigid column. The water-hammer solution, which is also drawn in the
figure, is not distinguishable from the rigid-column solution at this plotting scale. Figure 5
shows results for ∆t = 0.1 seconds, which is of the order of twh = 0.4 seconds. The water-
hammer effect is clearly visible at this time scale. The effect is still small because ∆P – equal
to the final steady-state friction loss – is small, so that the corresponding velocity jumps
∆ V = ∆ P / ( ρ c ) = 0.01 m/s are small too.

Water hammer can become spectacular when sudden, large velocity changes exist. In
Figure 6, a velocity change ∆V = 2 m/s, again within ∆t = 0.1 seconds, causes a water-
hammer pressure ∆ P = ρ c ∆ V = 20 bar. Figure 7 shows the pressure variation caused by the
same velocity jump of ∆V = 2 m/s, but now linearly applied over a time period of ∆t = 30
seconds. Rigid column theory gives an initial pressure rise of ∆ P = ∆ V . ( ρ L /(2 ∆t ) ) = 1/15
bar at the pipe midpoint, which also is the initial amplitude of benign water hammer at the
midpoint. The figure nicely confirms that rigid column theory gives the time-average of the
water-hammer oscillation. Figure 8 shows the equivalent of Figure 5 for ∆t = 0.01 seconds,
which is smaller than the time scale for FSI in a single straight pipe: tsh = 0.16 seconds. The
pipe is structurally fixed at the excited end and stress-free at the other end. The FSI effect,
visible now, is small initially because the interaction consists of Poisson coupling only. In
general, junction coupling is a stronger effect.

10
The frequencies in Table 1 confirm that the ring frequency, characterising radial pipe
vibration, is much higher than the water-hammer frequency and therefore of negligible
importance in 1D-FSI. The frequency of ovalizing vibration has the same order of magnitude
as the “steel-hammer” frequency, but there is no significant oval-axial interaction mechanism
at these low frequencies.

1.5 2

P ( t) 1.5

velocity (m/s)
1
∆P
 V ( t)  1

 ∆V  0.5
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4
0
t 0 600 1200 1800 2400 3000 3600
∆t time (s)

Fig 2 Starting flow driven by 0.1 bar pressure

Fig 1 Pressure (or velocity) ramp excitation at rise in 3600 seconds time; solid line = rigid
upstream end of pipe. column; dashed line = quasi steady; dotted line =
frictionless rigid column.
2 2

1.5 1.5
velocity (m/s)

velocity (m/s)

1 1

0.5 0.5

0 0
0 20 40 60 80 100 120 0 10 20 30 40 50 60 70 80
time (s) time (s)

Fig 3 Starting flow driven by 0.1 bar pressure Fig 4 Starting flow driven by 0.1 bar pressure
rise in 60 seconds time; solid line = rigid column; rise in 1 second time; solid line = rigid column ≈
dashed line = quasi steady; dotted line = dashed line = water hammer; dotted line =
frictionless rigid column. frictionless rigid column.

1.2 .10
4
0.06
(a) (b)
9000

0.04
velocity (m/s)

pressure (Pa)

6000

3000
0.02

0 3000
0 0.4 0.8 1.2 0 0.4 0.8 1.2
time (s) time (s)

Fig 5 Starting flow driven by 0.1 bar pressure rise in 0.1 second time: (a) velocity at
midpoint, (b) pressure at midpoint; solid line = rigid column; dashed line = water hammer.

11
 3 .10
6
5
(a) (b)
2 .10
6
4

1 .10
6
3

velocity (m/s)

pressure (Pa)
2 0

1 .10
6
1

2 .10
6
0

3 .10
6
1
0 0.8 1.6 2.4 0 0.8 1.6 2.4
time (s) time (s)

Fig 6 Starting flow driven by 2 m/s velocity rise in 0.1 second time: (a) velocity at midpoint,
(b) pressure at midpoint; solid line = rigid column; dashed line = water hammer.

4
2 . 10

4
1 . 10
pressure (Pa)

4
1 . 10
0 10 20 30 40 50 60
time (s)

Fig 7 Starting flow driven by 2 m/s velocity rise in 30 seconds time; average = rigid column;
oscillation = water hammer at midpoint.

1.2 .10
4
0.06
(a) (b)
9000

0.04
pressure (Pa)
velocity (s)

6000

3000
0.02

0 3000
0 0.4 0.8 1.2 0 0.4 0.8 1.2
time (s) time (s)

Fig 8 Starting flow driven by 0.1 bar pressure rise in 0.01 second time: (a) velocity at
midpoint, (b) pressure at midpoint; solid line = water hammer + 1D-FSI; dotted line = water
hammer.

4.2 Starting laminar flow

Herein, laminar flow means that skin friction is proportional to the mean velocity. A pressure
step (with ∆t = 0 s) of ∆P = 20 bar is applied to accelerate the liquid column from rest to a
final steady-state velocity of V∞ = 2 m/s. In principle, any flow is initially laminar when
starting from (theoretical) rest. Steady, quasi-steady, and rigid-column, laminar flow assume
paraboloid (Hagen-Poiseuille) velocity profiles in the pipe cross-section. For fast events, with
∆t = 0 s as the extreme, this is not always the case. Figure 9 compares 1D results with 2D
results by Szymanski (23) that include the effect of non-paraboloid velocity profiles. The 1D
model predicts a too fast response to the pressure step. The profile is practically paraboloid
after 0.75 t 2D .

12
 2

1.5

mean velocity (m/s)

1

0.5

0
0 0.02 0.04 0.06 0.08 0.1
time (s)

Fig 9 Starting laminar flow driven by 20 bar instantaneous pressure rise; solid line = rigid
column (1D); dotted line = frictionless rigid column (1D); upper dashed line = 2D-result,
cross-sectional average of velocity; lower dashed line = 2D-result, half the maximum velocity
(2D-results by Szymanski (23)).

5 CONCLUSIONS

Conditions of no-flow, steady flow, quasi-steady flow, rigid-column flow and water hammer
have been described with special reference to (i) coupling between the fluid and the structure
and (ii) time scales typifying the key characteristics of the flow.

Three types of coupling exist in all types of flow, namely junction coupling, Poisson coupling
and friction coupling. The first two of these exist even in no-flow conditions.

Notwithstanding the existence of coupling, accurate predictions of pressures and stresses can
be obtained from separate (uncoupled) structural and fluid analyses in all types of flow with
time scales larger than the water-hammer time scale.

The characteristic time scales of no-flow and steady flow are infinite and those for quasi-
steady flow are very large. Smaller characteristic time scales apply for rigid column flow,
relating primarily to the balance between inertia and frictional resistance. The smallest time
scales for 1-D phenomena, namely for water hammer and FSI, are independent of friction
(just like no-flow!). They relate to the time taken for waves to travel between adjacent outer
boundaries and adjacent inner boundaries respectively.

The paper is a first attempt to define time scales for problems possibly involving FSI. The
illustrating examples have been kept simple for clarity. Further work should illustrate piping
systems with elbows and branches, steady oscillatory flow, resonance behaviour, non-
dimensional parameters, 2D-FSI, etc.

ACKNOWLEDGEMENT

The Surge-Net project (http://www.surge-net.info) is supported by funding under the

European Commission’s Fifth Framework ‘Growth’ Programme via Thematic Network
“Surge-Net” contract reference: G1RT-CT-2002-05069. The authors of this paper are solely
responsible for the content and it does not represent the opinion of the Commission. The
Commission is not responsible for any use that might be made of data therein.

13
REFERENCES

(1) Vardy AE (1990). Fluid Principles. Maidenhead, UK: McGraw-Hill.

(2) Lavooij CSW, Tijsseling AS (1991). Fluid-structure interaction in liquid-filled piping systems. Journal of
Fluids and Structures 5, 573-595.
(3) Uffer RA (1993). Water hammer conservatisms. ASME - PVP, Vol. 253, Fluid-structure interaction, transient
thermal-hydraulics, and structural mechanics, 179-184.
(4) Hamilton M, Taylor G (1996). Pressure surge – Criteria for acceptability. In Proceedings of the 7th
International Conference on Pressure Surges and Fluid Transients in Pipelines and Open Channels
(Editor A Boldy), BHR Group, Harrogate, UK; Bury St Edmunds and London, UK: Mechanical
Engineering Publications Limited, 343-362.
(5) Wylie EB, Streeter VL (1993). Fluid Transients in Systems. Englewood Cliffs, USA: Prentice Hall.
(6) Tijsseling AS (1996). Fluid-structure interaction in liquid-filled pipe systems: a review. Journal of Fluids
and Structures 10, 109-146.
(7) Wiggert DC, Tijsseling AS (2001). Fluid transients and fluid-structure interaction in flexible liquid-filled
piping. ASME Applied Mechanics Reviews 54, 455-481.
(8) Skalak R (1956). An extension of the theory of waterhammer. Transactions of the ASME 78, 105-116.
(9) Lin TC, Morgan GW (1956a). A study of axisymmetric vibrations of cylindrical shells as affected by rotatory
inertia and transverse shear. ASME Journal of Applied Mechanics 23, 255-261.
(10) Lin TC, Morgan GW (1956b). Wave propagation through fluid contained in a cylindrical, elastic shell.
Journal of the Acoustical Society of America 28, 1165-1176.
(11) Thorley ARD (1969). Pressure transients in hydraulic pipelines. ASME Journal of Basic Engineering 91,
453-461.
(12) Bürmann W (1980). Längsbewegung frei verlegter Rohrleitungen durch Druckstöße. (Longitudinal motion of
pipelines laid in the open due to water hammer.) 3R international 19, 84-91 (in German).
(13) Bürmann W (1983). Beanspruchung der Rohrwandung infolge von Druckstößen. (Strain on pipe walls
resulting from water hammer.) 3R international 22, 426-431 (in German).
(14) Jong CAF de (1994). Analysis of pulsations and vibrations in fluid-filled pipe systems. PhD Thesis,
Eindhoven University of Technology, Department of Mechanical Engineering, Eindhoven, The Netherlands,
ISBN 90-386-0074-7.
(15) Womersley JR (1957). An elastic tube theory of pulse transmission and oscillatory flow in mammalian
arteries. WADC Technical Report TR 56-614, Wright Air Development Center, Air Research and
Development Command, United States Air Force, Wright-Patterson Air Force Base, Ohio, USA, pp. 1-115
+ Appendix.
(16) Cox RH (1969). Comparison of linearized wave propagation models for arterial blood flow analysis.
Journal of Biomechanics 2, 251-265.
(17) Stecki JS, Davis DC (1986). Fluid transmission lines - distributed parameter models. Part 1: a review of the
state of the art. Part 2: comparison of models. Proceedings of the Institution of Mechanical Engineers, Part A,
Power and Process Engineering 200, 215-236.
(18) Rutten MCM (1998). Fluid-solid interaction in large arteries. PhD Thesis, Eindhoven University of
Technology, Department of Mechanical Engineering, Eindhoven, The Netherlands, ISBN 90-386-0790-3.
(19) Fahy FJ (1994). Statistical energy analysis: a critical overview. Philosophical Transactions of the Royal
Society of London, A 346, 431-447.
(20) Moussou P. Fluid structure interactions in conservative linear systems: a kinematic variational method.
Submitted (in 2003) for publication in Journal of Fluids and Structures.
(21) Schwarz W (1978). Druckstoßberechnung unter Berücksichtigung der Radial- und Längsverschiebungen der
Rohrwandung. (Waterhammer calculations taking into account the radial and longitudinal displacements of the
pipe wall.) PhD Thesis, Universität Stuttgart, Institut für Wasserbau, Mitteilungen, Heft 43, Stuttgart,
Germany, ISSN 0343-1150 (in German).
(22) Caillaud S, Lambert C, Devos J-P, Lafon P (2002). Aeroacoustical coupling and its structural effects on a
PWR steam line. Part 2. Vibroacoustical analysis of pipe shell deformations. In Proceedings of the
5th International Symposium on Fluid-Structure Interactions, Aeroelasticity, Flow-Induced Vibration and
Noise (Editor MP Païdoussis), New Orleans, USA, ASME - AMD, Vol. 253, Paper IMECE2002-33363, pp.
1-8 (CD-ROM, Vol. 3) or ISBN 0-7918-3659-2 (printed book).
(23) Szymanski P (1932). Quelques solutions exactes des équations de l’hydrodynamique du fluide visqueux
dans le cas d’un tube cylindrique. (Some exact solutions of the hydrodynamic equations for viscous fluid in
the case of a cylindrical tube.) Journal de Mathématiques Pures et Appliquées 11(9), 67-107 (in French).

14
15
16