Anda di halaman 1dari 139

HSE

Health & Safety


Executive
aTI 93614
EVALUATION OF VORTEX

SHEDDING FREQUENCY AND
DYNAMIC SPAN RESPONSE
Prepared by J P Kenny & Partners Ltd
for the Health and Safety Executive

-t-
,
-.......... ....
..... ...
-...
.... -
.... __ .. -- .,
Health and Safety Executive
CONTENTS
Page
SUMMARY ix
1 .
2.
3.
INTRODUCTION
VORTEX SHEDDING FROM SUBMARINE
SPANS - GENERAL THEORY
2.1 Basic vortex shedding mechanism
2.2 Vortex induced excitation forces
SPAN RESPONSE AND LOCK-IN
3.1
3.2
3.3
3.3.1
3.3.2
3.4
Response types
Lock-in
Physical mechanism of lock-in
Cross-flow lock-in
In-line lock-in
Response assessment basis
PIPELINE
3
3
6
11
11
12
16
16
18
19
4_ VORTEX SHEDDING - PARAMETRIC STUDY 21
21
22
22
22
29
37
40
44
45
45
45
45
46
47
48
53
4.1 Introduction
4.2 Strouhal number
4.2.1 Physical factors influencing Strouhal number
4.2.2 Reynolds number
4.2.3 Pipe surface roughness
4.2.4 . Bed proximity
4.2.5 Velocity gradient across pipe diameter
4.2.6 Approach flow turbulence intensity
4.2.7 Span length /diameter ratio
4.3 Approach flow velocity
4.3.1 Definition
4.3.2 Steady state current
4.3.3 Comparison of methods
4.3.4 Wave induced currents
4.4 Vortex shedding in oscillatory wave flow
4.5 Vortex shedding in combined wave and tidal flow
5. SPAN RESPONSE - PARAMETRIC STUDY 55
5.1 Definition of parameters 55
5.2 Stability parameter 61
5.2.1 Components of the stability parameter 61
5.2.2 Effective mass, m. 61
5.2.3 Damping 62
5.3 Span response in Wave and combined wave/current flows 67
6. RESPONSE ANALYSIS
6.1 Definition of response thresholds
6.2 Full scale test analysis
6.2.1 Available full scale test data
6.2.2 Theoretical considerations
6.2.3 Special full scale study - smooth pipe tests
6.2.4 Strouhal number - smooth span
6.2.5 Frequency response - smooth span
6.2.6 Amplitude response - smooth span
6.2.7 Special full scale study - rough pipe tests
6.2.8 Strouhal number - rough span
77
77
77
77
78
83
83
83
92
100
102
v
Page
6.2.9 Frequency response - rough span 105
6.2.10 Amplitude response - rough span 108
6.2.11 Polar gas projects 110
6.2.12 Full scale tests - Bruschi et al (ref 22) 116
6.3 Practical response characteristics 117
6.3.1 In-line response 117
6.3.2 Cross-flow response 118
6.4 Alternative response rnechanism for very small bed gaps 122
7_
EVALUATION TECHNIQUE 125
7.1 Basis of method 125
7.2 Input data 125
7.2.1 Environmental data 125
7.2.2 Pipeline data . 126
7.2.3 Span assessment parameters 127
7.3 Analysis method 127
7.3.1 In-line response 127
7.3.2 Cross-flow response 129
7.3.3 Amplitude of in-line response 129
7.3.4 Number of in-line response cycles 130
8. CONCLUSIONS 131
LIST OF TABLES
3.1 Surnmary of locked-in response types 14
6.1 Cross-flow response thresholds from polar gas tests 116
6.2 Cross-flow response thresholds from Bruschi et al
(ref. 22) G/D = constant", 4.5 116
7.1 Environmental data parameters 126
7.2 Pipeline data parameters 126
7.3 Span assessment parameters 127
LIST OF FIGURES
Figure' 2.1 Flow regirne development static circular cylinder 4
2.2 Reynolds number as a function of pipeline diameter 5
2.3 Shedding frequency relationship static circu lar cylinder 6
2.4 Pressure field associated with vortex shedding 8
2.5 Alternate shedding of vortices and relation to the lift force 9
3.1a Staggered alternate vortex shedding - in-line and
cross-flow response 11
3.1b Symmetrical vortex shedding - in-line response only 12
3.2 Locked-in vortex shedding (cross-flow) 13
3.3 Influence of response arnplitude on excitation forces 15
4.1 Cross section through typical pipeline span 22
4.2 Typical wake shapes for (a) subcritical and (b) supercritical 23
flow
4.3 Strouhal-Reynolds number relationship for circular 24
cylinders
4.4 Separation positions for various Reynolds number ranges 25
4.5 Experimental Strouhal number - Reynolds number data 26
4.6 Power spectral density for vortex shedding at various 27
Reynolds numbers
4.7 Turbulent transition and boundary separation points 31
4.8 S as a function of kiD and Re Buresti (ref 47) 32
4.9 S as a function of k,/D and Re Achenbach and Heineke 33
(ref 58)
4.10 S as a function of k,/D and Re summary diagram (ref 58) 34
vi
Contents
Page
Figure 4.11 Boundaries between flow regimes 35
4.12 Strouhal - Reynolds number relationship for typical pipe
roughness conditions 37
4.13 Strouhal number as function of GID' 39
4.14 Normalised Strouhal number as function of GD 40
4.15 Normalised Strouhal number as function of GD for varying 41
approach flow conditions
4.16 Strouhal number as function of ~ U/U bed proximity 42
effect included
4.17 Normalised Strouhal number as function of A U/U bed 43
proximity effect removed
4.18 Normalised Strouhal number, based on velocity at top 44
of cylinder, as function of G/D
4.19 Typical boundary layer velocity profiles 47
4.20 Vortex shedding modes for plane oscillatory flow 50
4.21 Relative frequency of vortex shedding as a function of K 52
and R.
5.1 Logarithmic decrement definition sketch 56
5.2 Cross-flow response amplitude as a function of k, 59
5.3 In-line response amplitude as a function of k, 60
5.4 Added mass coefficient 63
5.5 In-line oscillation - body motion drag force interaction 65
5.6 Steady drag coefficient as a function of Reynolds number 66
5.7 Rough cylinder response in harmonic flow 69
5.8 Variation of vortex shedding frequency during wave cycle 69
5.9 Cylinder response to wave intermediate depth conditions 71
5.10 Cylinder response to wave deep water conditions 72
5.11 Relative amplitude as a function of the response parameter 73
5.12 Relative amplitude as a function of the generalised damping 74
ratio
5.13 Modulated span response under natural sea waves 74
6.1 Summary of the three vortex induced response types 79
6.2 Influence of the three fundamental response types on 80
vortex shedding frequency
6.3 Influence of an increased Strouhal number upon response 82
characteristics
6.4 Strouhal number results from '20-inch uncoated steel pipe 84
6.5 Smooth span - typical wake and span response frequency 85
plots
6.6 Smooth span - frequency response plots for all tests 88
6.7 Wake frequency spectra during in-line response 91
6.8 Wake frequency spectrum during cross-flow response 92
6.9 Possible interaction mechanisms between span wake and 93
current meters
6.10 Smooth span - typical span response amplitude plots 95
6.11 Plots of span mid-point motion. 40m span, GID = 2.0 96
6.12 Span deflection due to steady drag action 98
6.13 Steady drag coefficient for smooth and rough test spans 99
6.14 Strouhal number results from 20-inch pipe with simulated 101
concrete surface
6.15 Wake frequency spectra for stationary 2D-inch diameter 103
span
6.16 Wake frequency spectra for stationary rough span at 104
10wG/D
6.17 Rough span-typical wake and span response frequency 106
plots
vii
Contents
Figure 6.18
6.19
6.20a
6.20b
6.21
6.22
6.23
6.24
6.25
6.26
6.27
6.28
7.1
7.2
7.3
Plots of span mid-point motion rough span G/D = 1 .5
Plots of span mid-point motion rough span G/D = 0.5
Rough span - typical span response amplitude plots
Rough span - typical span response amplitude plots
Rough span - amplitude response plots for various
GlDvalues
Polar gas study, span mid-point motion for smooth and
rough pipes
In-line response thresholds
Cross-flow response thresholds onset of motion
Cross-flow response thresholds termination of motion
Model cylinder response at low K.
Cross-flow motion threshold V
R2
Response at very small G/D
In-line motion threshold VRl
Design cross-flow motion threshold V
R2
Amplitude of in-line motion as a function of K,
REFERENCES
viii
Page
107
109
111
112
113
115
117
120
120
121
123
124
127
128
129
133
Contents
SUMMARY
Vortex shedding from submarine pipeline spans is responsible for inducing the basic
excitational forces which can lead to large amplitude vibrational span response. The
principles of vortex shedding are therefore addressed in this document in order to provide a
physical insight into the excitation process.
For a basic case of a static pipeline span, the vortex shedding frequency may he simply
related to the approach flow velocity and the pipeline diameter by means of the well
established non-dimensional Strouhal number. Span assessment techniques in current use
often assume or imply that this Strouhal number remains constant.
Dynamic response of a pipeline sPlill can occur when the frequency of energy input due to
the fluctuating vortex induced forces approaches the natural vibration frequency of the span.
This response can occur as one of the three fundamental types, two of which induce the
span to oscillate 'in-line' ie., parallel to the flow direction. The third, most serious, type is
the 'cross-flow' or transverse oscillation where the pipe oscillates perpendicular to the
incident flow direction. Once span response is initiated, an interactive coupling mechanism
can be established whereby the vortex shedding becomes actively controlled by the span
motion itself. This phenomenon, known as 'lock-in' enables the span resporise to override
the basic Strouhal number relationship over a significant range of approach velocities.
In view of this strong interaction between flow behaviour and structural response the
mechanism of vortex shedding cannot be divorced from the span behaviour. Consequently,
the treatment of the vortex-induced excitation mechanism is combined with that of the
corresponding span response phenomenon.
The evaluation of the possibility of response can be achieved by means of a convenient
calculation procedure which makes use of a non-dimensional parameter known as the
'reduced velocity' which effectively combines two calculation steps. This operation
implicitly incorporates the calculation of vortex shedding frequency and its comparison with
the natural vibration frequency of the span.
If the above procedure is to be of practical utility however, the following physical
phenomena must be examined and their effects quantified:
The vortex shedding frequency is dependent upon the Strouhal number, which can
vary in response to environmental influence.
The onset threshold of a given type of span response occurs for a particular ratio of
vortex shedding frequency to span natural frequency. This threshold frequency ratio
may also be subjected to a variation as a function of environmental influences and
span properties.
The reaction of the Strouhal number to environmental influences is therefore examined in
detail, making use of the results from the specially performed laboratory test programme.
Further information regarding Strouhal number variations is obtained from the results of
the specially performed programme of full scale testing work. The full scale test results
have been further utilised to detennine the thresholds of span response for real spans under
realistic turbulent approach flow conditions in close proximity to the seabed.
An essential pre-requisite for use of the reduced velocity method for span assessment, is the
correct definition of the approach flow velocity. For unidirectional flow, this parameter is
defined with a high degree of confidence as the uninfluenced or free stream mean current
component acting at right angles to the pipe axis, level with the top of the pipe. For the
more general cases of wave-induced or combined steady and wave-induced flow, this is more
difficult and existing experimental data is extremely limited. On the basis of the few
experimental studies which have been published, it is concluded that the maximum wave-
induced and the uni-directional current components acting perpendicularly to the pipe should
be algebraically added to define the approach flow velocity. This approach is essentially
conservative, but the limited evidence available suggests that a very real risk of span
response exists under these conditions.
ix
Contents
The experimental studies carried out specially for the present work, performed their intended
function of examining span response under uni-directional flow. They were not designed
however, to investigate the effects of wave-induced currents on span response. Vortex
shedding under wave induced and combined uni-directionallwave induced flows is therefore
an area where further research is needed to resolve the significant uncertain.ties still
outstanding.
To take account of the practical circumstances existing in real situations, two thresholds,
VR, and VR2 , are defined. The lower threshold V." defines the point below which no
significant span vibration of any kind is anticipated. The upper threshold, V R2, defines the
point above which the risk of severe and unacceptable vibration levels is present. In the
intennediate region between VRI and VR2 , the possibility of some vibration exists, with its
acceptability dependel!t upon the circumstances prevailing for any giyen span. situation.
Conventional analysis methods, taking account of factors such as fatigue, may be used to
determine whether or not vibration in this region i ~ acceptable under the circumstances
prevailing.
The response amplitude of a vibrating span is essentially determined by the balance between
the energy input to the span from the flow, and the energy dissipated by structural and
hydrodynamic damping. The practical quantification of this energy balance in terms of the
commonly used 'stability parameter' is discussed, and the use of the stability parameter for
practical span assessment purposes is addressed.
Finally, a procedure is presented for the practical application of the span assessment
concepts developed within the document.
x
Contents
1. INTRODUCTION
Pipeline freespans on the seabed can arise due to a number of reasons but in each case a
known span requires assessment to ascertain whether correction is necessary or whether the
span may be left untended until the next pipeline inspection whereupon further assessment
would take place.
In the overall context of span assessment J P KENNY were assigned by the Department of
Energy to a project to develop guidelines for the evaluation of pipeline spans. In
association with the project, two specially commissioned experimental research
programmes were carried out under subcontract. These comprised a labora'tory test study
carried out by University COllege London and a full scale field study carried out by
Hydraulics Research Limited, Wallingford. Both studies were performed under the
management of J P KENNY.
The purpose of this document is to provide detailed technical back-up to the guidelines on
two related aspects of pipeline span assessment, namely the determination of frequency of
vortex shedding from the pipeline span and the evaluation of dynamic response. A separate
background document deals with the structural aspects of spanning.
The process of vortex shedding from a submarine pipeline span induces fluctuating
hydrodynamic forces which can excite the span into a resonant state. In order to ascertain
the potential for a span to undergo dynamic response it is necessary to determine the vortex
shedding behaviour specific to the span in question. This background report presents the
theoretical approach and draws on extensive empirical infonnation in explaining the vortex
shedding phenomenon in the context of a span close to the natural seabed. Having thus
examined the basic vortex-induced excitation mechanism, the report then proceeds to
investigate the consequent dynamic response of a submarine pipeline span.
The approach taken in the document is founded on the fundamental relationship between
shedding frequency and the velocity of the incident flow on the pipe, and the manner in
which that relationship is modified by conditions local to the span, such as bed proximity,
seabed velocity gradients and flow characteristics. In addition, the report discusses the
manner in which vortex shedding can induce a response. These aspects have been combined
in one report since it is unrealistic to consider vortex shedding in isolation due to the strong
physical coupling mechanism between the structure and the flow which occurs during
dynamic response.
The report is arranged in the following manner:
Section 2 provides a description of the general theory of vortex shedding.
Section 3 gives a qualitative description of the span response mechanism and the
manner in which response and flow behaviour interact.
Section 4 is concerned with a detailed evaluation of the physical parameters
influencing vortex shedding.
Section 5 evaluates the physical parameters which determine the response of a span
to vortex induced excitation.
Section 6 utilises full scale field tests results and existing literature to determine
response thresholds at which vortex induced vibration of a pipeline span may begin.
Section 7 presents a summary technique for the determination of the important
reduced velocity parameter which is used to predict span response characteristics.
Section 8 sets out the conclusions of this part of the study.
Contents
2. VORTEX SHEDDING FROM SUBMARINE
PIPELINE SPANS - GENERAL THEORY
2.1 BASIC VORTEX SHEDDING MECHANISM
A Submarine pipeline span comprises a section of unsupported pipeline lying freely
suspended close to the ocean floor. In a natural ocean environment, such a span will
generally be subject to the effects of currents incident upon the suspended length of
pipeline. These currents will generally consist of a long-term steady component due to tidal
flows, storm surge currents and other similar phenomena, and a fluctuating component due
to wave induced effects.
The phenomenon of fluid flow past solid bodies is of practical relevance to many
engineering applications, and consequently forms the subject of a significant volume of
available research data. Submarine pipelines fall into the category of circular cylindrical
bodies, along with other common applications including chimney stacks, marine piles and
power transmission cables. The practical relevance of the circular cylindrical configuration,
along with its relative experimental simplicity, has made it one of the better documented
cases, allowing a reasonable description of the basic fluidlbody interactions to be developed.
A number of points must be appreciated however, when interpreting data for use in pipeline
engineering applications. A significant point is that most data is derived from laboratory
tests on models, since full scale data is generally technically difficult and very expensive to
obtain. The extrapolation of model test data (particularly quantitative results) to full scale
must therefore be performed with care. A further point is that model tests tend to be carried
out under idealised conditions, eg., in uniform low turbulence free stream flows from
influences such as fixed boundaries.
Full scale pipeline spans, however, tend to be exposed to the effects of seabed proximity
and associated turbulent boundary layer effects. Dynamic response effects can also
significantly modify the flow regime around the pipe. Nevertheless, an appreciation of the
basic established principles of flow around a rigid circular cylinder as described below is a
necessary stage in understanding the overall phenomenon of vortex induced pipeline
vibration.
In the absence of other environmental influences, the flow around a circular cylinder placed
in a free stream flow can be divided into a number of basic regimes dependent upon the flow
Reynolds number, Re. Reynolds number is defined below:
Re = UD/u Eqn 2.1
where:
U = approach flow velocity:
D = cylinder outside diameter;
u = kinematic viscosity of surrounding fluid.
Figure 2.1 is taken from Lienhard (ref. I) and illustrates the respective regimes of fluid flow
past circular cylinders (eg., pipelines) for increasing ranges of Reynolds number.
Photographic evidence of these basic flow regimes has been presented by a number of
researchers including Prandtl (ref. 2), Prandtl and Tietjens (ref. 3), Taneda (ref. 4) and
Clutter, Smith and Brazier (ref. 5). Examples of these latter results are conveniently
summarised by Batchelor (ref. 6).
These basic flow regimes apply specifically to the case of a static pipeline, and are subject
to modification if any motion of the span is allowed to take place.
The very low Reynolds number laminar regimes (Figure 2.1 a), b) and c) ) are generally of
little practical interest to subsea pipeline engineers. Realistic Reynolds numbers that are
typically encountered in the North Sea can be estimated using the following representative
data values.
3
Contents
Kinematic viscosity of seawater, '\.):::: 1.5 x 10--ti m2/s
Steady current velocity component, U, = 0.5 m/s
Wave induced velocity component, Uw ::: 1-2 ruls
01 _.
b)
cl
dl
el

I)

Figure 2.1
Re <5
REGIME OF UNSEPARATED FLOW
5 TO 15 Re < 40
A FIXED PAIR OF FOPPL
VORTICES IN WAKE
40 .. Re < 90 AND 90<Re < 150
TWO REGIMES IN WHICH
VORTEX STREET IS LAMINAR
150 '" Re < 300
TRANSITION RANGE TO
TURBULENCE IN VORTEX STREET
.Re 3 105
VORTEX STREET IS FULLY
TURBULENT
3.10
5
<Re < 3'5.10
6
LAMINAR BOUNDARY LAYER HAS
UNDERGONE TURBULENT
TRANSITION AND WAKE IS
NARROWER AND DISORGANISED
3 5 10
6
s;;. Re
RE-ESTABLISHMENT OF
TURBULENT VORTEX STREET
FROM LIENHARD (REF. I
Flow regime ,de.velopment static circular cylinder
4
Contents
Nominal pipeline diameter, D; 12-36 in
; 0.3-0.91 m
Combined current velocity, U, + U.; 1.5-2.5 mls
This representative data is used to construct Figure 2.2, which shows the variation of pipe
Reynolds number as a function of pipe diameter. Two curves are shown, corresponding to a
combined approach flow velocity of 1.5 mls and 2.5 mls respectively. From this
information, North Sea pipelines would be expected to fall into a region which includes
both the stable alternate vortex shedding regime (Figure 2.1 d) and the transition zone where
laminar flow in the boundary layer around the pipe becomes turbulent (Figure 2.1 e).
The alternate vortex shedding regime of Figure 2.1 d) is stable over a wide Reynolds
number range, and is therefore commonly encountered in practical situations and hence well
documented. The frequency of vortex shedding, f., is regular and defined as the frequency at
which complete vortex pairs (as opposed to single vortices) are shed. Each complete vortex
pair is comprised of one vortex from the top and one from the bottom of the cylinder. It is
important to note this defmition of f .. to avoid later confusion.
The frequency with which vortex pairs are shed from a point on a static pipeline span is
determined by the following simple and well established relationship, originally discovered
by Strouhal (ref. 7) in the 19th Century:
where:
SU
f.; D
f. = vortex shedding frequency;
U ; approach flow velocity;
D ; pipeline diameter:
S ; Strouhal number.
.K)'
18,--------------------,
17
NOTES: SU8CRITICAL AND CRITICAL FLOW REGIMES ARE DEFINED IN FIGURE 51
THE THRESHOLD OF SU8CRITICAL TO CRITICAL TRANSITION CAN VARY
IN RESPONSE TO CYLINDER SURFACE ROUGHNESS AND ENVIRONMENTAL
EFFECTS. THIS IS ADDRESSED IN DETAIL IN SECTiON FIVE
Figure 2.2
Reynolds number as a function of pipeline diameter
Eqn 2.2
5
Contents
Equation 2.2 provides a simple two-dimensional description of tbe vortex sbedding
frequency at a point on a static span. Experiments show that the phase of the vortex
sbedding cycle does not generally remain constant along the lengtb of a spanning cylinder.
This gives rise to the concept of the 'correlation lengtb' of the vortex sbedding process,
described later in this section. For engineering purposes however, the most conservative
approach is obtained by utilising the two-dimensional approach, in which all vortex induced
forces are assumed to act in synchronisation along the length of a span.
S, tbe Strouhal number is a non-dimensional pbysical parameter dependent upon a number
of variables which are discussed later in this document.
U, the approacb velocity, for a pipeline span in a natural seabed environment, requires
considerable care in its definition. The selection of a suitable design approach velocity is
addressed in Section 4.
D, the pipeline diameter is basically a fixed quantity althougb features such as marine
growth thickness may require consideration.
Equation 2.2 sbows tbat for a given pipeline diameter and Strouhal number, the vortex
shedding frequency of a static pipeline is linearly related to the approach flow velocity as
shown in Figure 2.3.
u
(S AND 0 ASSUMED CONSTANT)
Figure 2.3
Shedding frequency relationship static circular cylinder
In the critical Reynolds number region of Figure 2.1 e) the situation is less well defined,
since the wake can be subject to disorganisation, with consequent disruption of the regular
vortex shedding regime. The full implications of the critical flow region and its effect upon
span behaviour are discussed later in this document.
2.2 VORTEX INDUCED EXCITATION FORCES
The time varying velocity field associated with the growth and shedding of vortices from a
submerged body generates an accompanying time-varying pressure field. The consequent
action of this pressure field on the surface of the body,ipsum imposes time-varying forces
upon the body itself. For the typical case of regular shedding from a circular
cylinder (eg., a pipeline) the fluctuating pressure field is illustrated in Figure 2.4 (Drescher,
ref. II). It can be seen from this figure that the net pressure components acting both
parallel to and normal to the approach flow are subject to regular fluctuations at frequencies
related to the "/ortex shedding process. The force component parallel to the approach flow is
commonly tenned the in-line or drag force, and that Donnal to the flow, the cross-flow or
lift force.
6
Contents
By examining Figure 2.4, a number of inferences can be drawn regarding the relationship of
the fluctuating forces to the vortex shedding process.
In the case of the cross-flow (lift) force component, a vortex shed from the top of the pipe
induces pressure forces which are opposite in sign to those of a similar vortex shed from
the bottom of the pipe. Each complete cross-flow excitation cycle therefore comprises the
successive effects of a complementary pair of vortices, acting first in one given direction
and then in the vertically opposite direction. The relationship between the direction of the
lift force and the shedding of the vortices can be predicted from classical theory of
circulation and lift generation. The frequency of the cross-flow excitation force is therefore
equal to the vortex shedding frequency, f,. The relationship of the cross-flow (lift) force to
the vortex shedding process is illustrated in Figure 2.5, taken from Sarpkaya (ref. 12).
The behaviour of the in-line (drag) force component is different however. The in-line
component is due to the pressure differentials which are induced between the upstream and
downstream faces of the cylinder each time an individual vortex forms and is shed into the
wake. The pipeline therefore experiences a similar in-line excitation force irrespective of
whether a given vortex is shed from the top or bottom of a pipeline. The in-line excitation
frequency is therefore equal to the frequency of shedding of individual vortices from the
pipeline. In other words, the fluctuation frequency of the in-line force component is equal to
2fs. ie., twice the vortex pair shedding frequency.
Experimental evidence suggests that for the case of a long static cylinder. the vortex
shedding process does not generally remain 'in phase' or correlated with itself along the
cylinder length. King (ref. 8) describes the vortex shedding as taking place in cells, where
the length of each cell is termed the correlation length. For the Reynolds number ranges
relevant to pipeline situation, King (ref. 8 ) quotes the. following values for the correlation
length. Values are expressed as multiples of the cylinder diameter, D.
Reynolds Number Correlation Length Original Source
150<R,<lO' 2D-3D Gerlach and Dodge (ref. 9)
R,>IO'
R, = 2 x 10'
O.5D
1.56D
Gerlach and Dodge (ref. 9)
Humphreys (ref. 10)
Care should be taken in applying these quantitative results to full scale subsea pipeline
applications, but general indication is that over the Reynolds number ranges considered, the
correlation length is small. The consequence of this effect is that over a long length of
stationary pipeline, the vortex-induced forces tend in the mean to have a cancelling effect
upon each other. The net force experienced by the entire length of pipeline is thus
considerably reduced. As will be discussed later, the correlation length and hence the net
force acting upon the pipeline can be significantly modified if pipeline motion takes place.
7
Contents
-2
t = 0'903 sec
u ~
2
t=0935 sec
\
-2
t = 0968 IIC
u ~
-2 t =1,000 sec
-I
u ..
FROM DRESCHER (REF II
Figure 2.4
Pressure fleld associated with vortex aheddlng
8
Contents
C
L

1-0 0-5 00 0-5
70



72


t


VORTEX
SHEDDING


74 BEGINS




76


78

/VORTEX
SHEDDING

BEGINS

80
t


FROM SARPKAYA (REF. 12
Figure 2.5
Alternate shedding of vortices and relation to the lift force
9
Contents
3. SPAN RESPONSE AND LOCK-IN
3.1 RESPONSE TYPES
The fluctuating hydrodynamic forces associated with vortex sbedding may result in the
dynamic response of a pipeline span. This response, wbicb takes the form of large
amplitude oscillation or vibration of the span, is obviously undesirable from the point of
view of maintaining the structural integrity of the span and its concrete coating.
A further important characteristic of vortex induced span vibration is that the pipeline
vibration itself, once initiated can interact with the vortex shedding process to force the
vortex sbedding into synchronisation with the span vibration, and magnify the amplitude of
tbe osciIlations. 1his self-sustaining interaction process is referred to in current literature by
several descriptions, including lock-in, wake capture, locking on, self excitation, and
others. For all subsequent references in this text, the term 'lock-in' is adopted.
The response of a pipeline span is essentially governed by the interaction between the
natural vibration frequency of the span, fN. and the excitation frequency at which energy is
input to the span by the vortex shedding process. A number of response mechanisms are
possible, corresponding to various alternative interaction regimes. These possible types of
response are described below.
Figure 3.1 a) illustrates the stable regime of staggered alternate vortex shedding which
exists over a large section of the Reynolds number range. Associated with the development
and shedding of each vortex is a corresponding fluctuating pressure field, responsible for the
hydrodynamic excitation forces experienced by the pipeline. Tbe mechanism of these
vortex-induced excitation forces is discussed in greater detail in Section 2.2. To summarise,
two mechanisms of excitation can be identified in association with the alternate vortex
shedding regime. These are an in-line (drag) force, fluctuating at twice the vortex shedding
frequency, and a cross-flow (lift) force, fluctuating at the vortex shedding frequency itself.
CURRENT
FLOW
CROSS- FLOW ~
RESPONSE
DIRECTION
COMPLETE VORTE)( PAIR SHEDDING CYCLES_
WITH CROSS-FLOW RESPONSE

SINGLE VORTEX SHEDDING CYCLES-
ASSOCIATED WITH IN-LINE RESPONSE
Figure 3.18

Staggered alternate vortex shedding - In-line and cross-flow response
In the absence of external influences, dynamic span response win take place when the
excitation frequency closely corresponds to the natural frequency of tbe span thus allowing
resonance to occur. Following this reasoning, response in the in-line direction will take
place when f. " 1/, fN, and cross-flow response will occur when f." fN (where f. = vortex
shedding frequency and fN = natural frequency of span.) If tbe linear Strouhal number
relationship between approach flow velocity and vortex shedding frequency is assumed, then
it can be deduced that response in the in-line direction will occur at half the velocity of the
II
Contents
cross-flow response. Additionally, as discussed for example by King, (ref. 8) higher
harmonic response modes are possible at higher flow velocities. The fundamental response
mode however, is the dominating case for pipe span response, and higher harmonic
response is not addressed further in this document.
There is however, a further response mechanism, occurring at a lower flow velocity than
either of the respoIL"C types described above. 1llis additional mechanism also causes span
response-in the in-line direction-and is associated with the symmetrical vortex shedding
pattern illustrated in Figure 3.1 b). This shedding regime is a 'forced' phenomenon imposed
on the flow by the pipe motion itself, and should not be confused with the fixed vortex pair
regime formed behind stationary cylinders at very low Reynolds number (see Fignre 3.1 b).
The symmetrical vortex street produced by this shedding regime is inherently unstable both
in theory and in practice, and King (ref. 8) offers photographic evidence ,to show that
symmetrically shed vortices rapidly undergo a transition (involving the coalescence of
adjacent vortices) into the stable staggered vortex street configuration.
r- - - -_'VTH1S ARRANGEMENT OF
CU
RRENT 70-1 ~ -::-.. 1 VORTICES 1$ UNSTABLE
I CJ eJ I AND TENDS IN PRACTICE
FLOW.. I TO UNDERGO COALESCENCE
j
/ r.. I Gl I TO FORM THE STAGGERED
::.....;::<' 1....9 J I ALTERNATE REGIME
I I I DEPICTED IN PART 0)
IN-LINE L _____ .J ABOVE
RESPONSE
(N.B. THIS IS A FORCED EFFECT INDUCED BY THE CYliNDER MOTION" ITSELF)
Figure 3.1 b
Symmetrical vortex shedding - in-line response only
The physical basis of the symmetrical vortex shedding mechanism is discussed by Sarpkaya
(ref. 12). He attributes the development of symmetrical vortex pairs to conditions of low
relative displacement between the cylinder and the f l o w . ~ A low relative displacement does
not allow the full development of the vortices, or of asymmetry whilst they are still
attached to the cylinder. The vortices are thus periodically shed from the cylinder in
symmetrical pairs, with the associated periodic drag overshoot giving rise to the excitation
force. The necessary condition of low relative displacement is obtained when the vibration
amplitude/cylinder diameter ratio, aID, is relatively small aod the flow velocity, U, is
approximately equal to the maximum velocity of the cylinder, Um
Since this symmetrical vortex shedding regime is a forced phenomenon imposed by the
motion of the span itself, an obvious question to ask is why does the motion arise in the
first instance, before the regular excitation mechanism has been established. This question
can be answered by considering the hydrodynamic forces acting upon the span prior to the
establishment of any form of regular periodic response. In an idealised uniform flow, the
span will experience regular fluctuating forces at frequencies much lower than fN, due to
conventional Stronhal number defined vortex shedding, Additionally, in a real environment
the span will experience further hydrodynamic forces due to flow turbulence and possible
wave-induced effects. The net result is a randomiSed excitation of the pipe, which will
respond with small amplitude vibrations, variable in direction, but with a tendency to occnr
at or near fN The mechanism by which small amplitude pipe motions can lead to the
fOlmation of a locked in in-line vortex shedding mechanism is discussed in Section 3.2.
3.2 LOCK-IN
The three types of pipeline response described above, two in-line and one cross-flow, are
associated with the phenomenon known as lock-in. A comprehensive description of the
detailed physical process associated with lock-in is given by Sarpkaya (ref. 12). For the
purposes of the present discussion, a much simplified description of the phenomenon is
given as follows.
12
Contents
In Section 3.1, span response is simplistically described in terms of a resonance effect
between coincident excitation and natural frequencies. Although resonance is a simple
concept to deal with, it by no means fully describes the physical processes inherent in span
response. Once a large amplitude span response has been initiated, further physical
processes occur, leading to the complex interaction process of lock-in.
The lock-in phenomenon arises when the amplitude of the span oscillation becomes
sufficiently large to dominate the vortex shedding process. Basically, the motion of the
cylinder is able to enhance the vortex formation process, thereby reinforcing the
hydrodynamic mechanisms which induce span motion in the first instance. Thus once the
vibration has been initiated, the periodic motion of the span actively assists the formation
and shedding of vortices. The excitation mechanism is therefore pulled into synchronisation
with the span vibration process, in addition to undergoing amplification due to the
. constructive interactions between the span motion and the vortex shedding regime.
The enforced synchronisation between the vortex shedding process and the span fuotion has
the effect of over-riding the conventional stationary cylinder Strouhal number relationship.
The over-ride process is well illustrated by Figure 3.2, from King et al (ref. 65). This
example specifically relates to cross-flow lock-in, but can be qualitatively applied to the in-
line lock-in also. The figure shows a graph of dominant wake frequency against approach
flow velocity for a flexible circular cylinder. It is thus comparable with Figure 2.3, which
showed the conventional linear relationship between the vortex shedding frequency and
approach flow velocity for a simple stationary cylinder. For the purposes of this qualitative
description, the number axis scales are not of great relevance.
.'\.
40
~
" lit
"
w ~ ' "
'"
~ C
<!
30
.,." X
,.
&X'
"
<-'I'
>-
"
u
z
w
CYLINDER
"
20 0
FREQUENCY, fn yA xB w
"
xxX XXXXX
~
;X
....
(X z
SYNCHR()'.jISATON <!
z
10
X/
RANGE
's = fN
"
0
0
~
o 5 10 15 20
VELOCITY
FROM KING ef 01 (REF 65)
Figure 3.2
. Locked-In vortex shedding (cross-flow)
Initially, Figure 3.2 shows a simple linear relationship between the wake frequency and
approach flow velocity. This indicates that the wake is experiencing simple Strouhal
number dominated vortex shedding. As the velocity is increased further however, a point is
reached (point A) where the Strouhal number dominated vortex shedding frequency
nominally coincides with the natural vibration frequency of the cylinder. At this point
resonance occurs, and large amplitude oscillations of the cylinder are able to build up. From
this point, the wake frequency no longer follows the simple Strouhal number based linear
relationship with the flow velocity. It becomes synchronised with the natural frequency
vibration of the cylinder, and remains nominally equal to this frequency despite significant
further increase in approach flow velocity. In other words, the vortex shedding process has
become 'locked-in' to the natural vibration frequency of the cylinder.
Synchronised lock-in continues upeto point B, where the discrepancy between the
theoretical Strouhal number dependent wake frequency and the actual locked-in frequency is
considerable. In this example, point B represents the limit, beyond which the span motion
can no longer control the vortex shedding. A further increase in approach flow velocity
13
Contents
therefore results in the cessation of lock-in, indicated by an abrupt jump of the wake
frequency back to the value predicted by the linear Strouhal number relationship. Vortex
shedding subsequently continues to conform to the Strouhal number relationship for further
. increases in approach flow velocity.
An inherent consequence of the locked-in coupling between the span vibration and the
vortex shedding process is that the vortex shedding correlation length is increased. The
vortex shedding is no longer free to exhibit a random phase relationship along the length of
the span, but tends to be forced into phase with the span motion itself. Sarpkaya (ref. 12)
points out however, that the increase of the correlation length at lock-in can be less
pronounced if the approach flow is turbulent
A further point indicated by Sarpkaya (ref. 12) is that the transverse forces necessary to
excite a span to the typical oscillation amplitudes observed during lock-in, are considerably
greater that the forces generated by stationary cylinder vortex shedding. In other words, the
dynamic lock-in process is able to magnify the vortex induced forces exerted upon a span.
The improved organisation and coherence of the vortex shedding, associated with the
increase in correlation length are able to account for a limited proportion of the increase,
but some additional more powerful enhancement process is clearly involved.
This additional enhancement effect is due to an increase in vortex strength resulting from
the relative motion between the pipeline and the surrounding fluid. The physical causes of
this vortex strength enhancement are discussed in some detail in Section 3.3, for descriptive
purposes, however, the practical effects experienced by the pipeline are well illustrated by
Figure 3.3 (King, ref. 8). This figure shows the variation of the fluctuating force coefficient
with vibration amplitude for the three primary types of responses. For both of the in-line
response mechanisms a) and b) the fluctuating force coefficients which define the excitation
forces experienced by the cylinder are seen to increase linearly with increasing amplitude of
response. In practice, it is found that the maximum response amplitude in the in-line
direction is generally self-limiting to approximately 0.2 D (where D = cylinder diameter),
although exceptions to this rule are found to occur under certain circumstances discussed
later.
The cross-flow behaviour, c), is somewhat different. The cross-flow response amplitudes
show an order of magnitude increase over those in-line and the relationship between the
fluctuating force coefficient and response amplitude is somewhat more complex. Initially
the fluctuating force coefficient increases with response amplitude, as in the in-line cases.
Above an amplitude of approximately 0.5 D however, the fluctuating force co-efficient
begins to progressively decrease, ultimately decaying to zero at a limiting response
amplitude. This limiting amplitude depends upon the individual characteristics of the span,
but in practice appears to vary up to approximately 2.0 D. Factors determining the limiting
amplitude are discussed in detail by Sarpkaya (ref. 12) and are addressed in Section 5 of this
document.
To summarise, lock-in span response involving synchronisataion between the excitation
and response frequencies may result from any of three basic mechanisms as detailed in Table
3.1:
Table 3.1
Summary of locked-in response types
RESPONSE TYPE COMMENTS TYPICAL LIMITING
AMPLITUDE-
Syminetrical vortex
lstin-Line
shedding
0.2D
Alternate vortex shedding
2nd In-Line fs" 0.5 fN 0.2 D
Alternate vortex shedding
Cross-Flow
fs" fN
2.0D
* The suggested limiting amplitudu are for general information only, and may be subject
in practice to significant modification depending upon physical Circumstances.
14
Contents
020,---------------,
o
.....
o
w
o
:>
....
:::i 0-10
Q.
:l!

o
g
o
UJ
0::
o 01 02 ().3 04 05 06 07 OB
FLUCTUATING FORCE COEFFICIENT Cd

o
W
o
:>
.... 010
:::i
Q.
:l!


u
:>
o
UJ
0::
'"
20
.....
0
w
15 0
:>
....
:::i
o 0-1 0-2 03 04 0-5
FLUCTUATING FORCE COEFFICIENT Cd
"-
1-0
:l!

0
w
05 u
:>
0
UJ
0::
0 01 0-2
FLUCTUATING
0-3 04 0-5 0-6 07 0-8
FORCE COEF F ICIENT C
Figure 3.3
NOTE'
THE CIRCLED SYMBOLS
ARE THE MAXIMA
0) FIRST TYPE
IN-LINE RESPONSE
NOTE'
ONLY THE MAXIMA ARE
SHOWN
b) SECOND TYPE
IN -LINE RESPONSE
NOTE-
ONLY THE MAXIMA ARE
SHOWN
c) CROSS-FLOW
RESPONSE
FROM KING REF 8)
Influence of response amplitude on excitation forces
15
Contents
3.3 PHYSICAL MECHANISM OF LOCK-IN
3.3.1 Cross-Flow Lock-In
. Dynamic lockin of the cross flow type, where the pipeline span vibrates transversely to
the direction of the approach flow, is potentially the most serious form of vortex induced
span vibration. The phenomenon has received a significant amount of research attention,
including -hydrodynamic and aerodynamic experimentation and numerical computation
investigations, with the result that a reasonable understanding of the phenomenon has been
developed. From the point of view of quantitative analysis however, the problem remains
that the prediction of the strength and distribution of the vorticity field is difficult and
sensitive to the theoretical and experimental means employed. as pointed out for example
by Sarpkaya (ref. 12) and Davies (ref. 56).
The lock-in phenomenon can be more readily appreciated if the vortex formation and
shedding mechanism is examined. Roshko (ref. 34 and 35) showed the dependence of the
vortex shedding frequency upon the following wake parameters; dw = wake width, and U, =
flow velocity at the outer edge of the boundary layer at the points of flow separation. The
mean value of Us can also be defined in terms of the base pressure coefficient, Cpb as
follows:
where:
U,
=
k
Cpb
=
=
U
=
U, = kU
mean value of Us;
(I . C,.);
base pressure coefficient on stagnation side of pipeline;
C
p
= pressure coefficient at separation points;
undisturbed approach flow velocity.
Eqn 3.1
Roshko (ref. 35) shows that the total circulation shed by a body is related to U, and Davies
(ref. 56) elaborates on this reasoning to give the following expression:
where:
dT
dt
=
dI'
- = [/ U2 (t)
dt 2 s
Eqn 3.2
rate of production of circulation at each separation point.
It must be noted however, that in practice, only a fraction of the circulation generated at the
separation points survives into the fully developed wake region.
By integrating Equation 3.2 over an entire shedding cycle period, T, the circulation
generated during the cycle is given as follows:
where:
r
T
U,
=
=
r = [/2 U
2
,T Eqn 3.3
total circulation generated during shedding cycle at one separation point
ie., the circulation of an individual vortex;
period of shedding cycle;
is as defined for Equation 3.1.
The circulation, r, can be r e l a ~ to the Strouhal number by considering two further
conditions. Firstly, the longitudinal separation of the vortices, lv, is related to the vortex
shedding frequency and the velocity of the wake vortices relative to the pipeline, by the
following expression (see Roshko ref. 35, Sarpkaya ref. 12):
Eqn 3.4
where:
t; = vortex shedding frequency;
Iv ::: longitudinal spacing of vortices in wake;
16
Contents
U = undisturbed approach flow velocity;
Us ;;;; velocity of wake vortices relative to U, which gives U - Us = velocity
of wake vortices relative to pipeline.
The second condition was derived by Von Karman (ref. 57) as an essential condition for the
stability of the wake, and is given by the following relationship:
where:
r =
u, =
I, =
rlu,l, = 2 ...J2 Eqn 3.5
circulation of an individual vortex;
velocity of wake vortices relative to undisturbed approach velocity, us.
longitudinal separation of vortices in wake.
By using the latter results, and incorporating the Strouhal number relationship, s = fs DIU,
the following result is obtained (see Sarpkaya, ref. 12): .
S = _1_ _ rlUD
I,ID 2-n (I,ID)'
Eqn 3.6
The derivation of Equation 3.6 can be followed in more detail by studying the quoted
references. The important result however, is that the Strouhal number is a function of the
strength of the Circulation, which can itself be influenced by motion of the cylinder relative
to the flow.
The motion of the cylinder causes a shift in the separation points, and brings the cylinder
closer to the vortex which is growing at the particular phase of the cycle. As a consequence,
the circulation, r, of the shed vortices is increased. By examining equation 3.6, it can be
seen that an increase in the circulation, r, will lead to a decrease in Strouhal number, S,
and it is this relationship which forms the basis of the lock-in process.
If the Strouhal number was strictly constant, the dynamic response would be restricted to a
simple resonance range when fs, == fN, with the motion tending to die out again if the flow
velocity, and hence f .. was further increased. In reality however, any build-up of span
motion due to resonant interaction between f8 and fN tends to increase the circulation, r, and
consequently reduce the Strouhal number, S. A further increase in approach velocity, U, is
therefore accompanied by the tendency for S to reduce as a consequence of the increased
circulation generated by the cylinder motion. The vortex shedding frequency, f, therefore
tends to remain nominally equal to the natural frequency, fN, despite the increase in U, and
the constructive interaction between the forcing frequency (= f8) and the natural response
frequency (=fN) is able to continue. By this process, the forcing frequency remains
synchronised with the natural response frequency of the span for a significant range of
approach flow velocities, increasing from the point where the stationary cylinder vortex
sheddiug frequency first becomes nominally equal to f
N
.
In view of this ability of the span motion to control the Strouhal number, Sarpkaya (ref.
12) questions why the vortex shedding does not lock onto the cylinder frequency for all flow
velocities. The lower end of the lock-in range is in fact imposed because there is no
physical mechanism for reducing the circulation of the shed vortices and thus increasing the
Strouhal number above its stationary cylinder value. Lock-in cannot therefore take place
when the stationary cylinder f, is less than fN, although some response may be encountered
in practice as a result of the cylinder being forced to vibrate at the vortex shedding
frequency.
The upper limit of the lock-in range is reached when the lock-in process becomes unstable.
In order to achieve sufficient depression of Strouhal number at velocities considerably in
excess of that required for the initial coincidence of f, and fN, very high values of r are
required. The presence of such highly energetic vortices in close proximity to the cylinder
and to each other, represents an unstable situation where small perturbations may introduce
an abrupt breakdown in the lock-in mechanism (see Sarpkaya, ref. 12).
17
Contents
3.3.2 In-line Lock-in
The qualitative description in Section 3.1 describes two possible types of in-line vortex
excitation. The second of these in-line mechanisms, occurring when f, ;: 0.5 fN, simply
comprises the situation where the frequency of the drag fluctuation associated with the
alternate shedding of individual vortices from opposite sides of the pipe corresponds
nominally with the natural frequency of the span. This mechanism is therefore analogous to
cross flow excitation, in that it results from a nominal correspondence between fN and a
multiple of the Strouhal number based vortex shedding frequency. In this case, however, the
increase in circulation resulting from the relative motion of the cylinder, becomes unstable
at much lower vibration amplitudes, typically of the order of 0.2 D.
A different physical mechanism is responsible for the remaining type of in-line response,
occurring at very low reduced velocity values. The vortex shedding associated with this type
of response is controlled by the relative velocities of the approach flow and the vibrating
cylinder and is thus not directly dependent upon Strouhal number effects. As "mentioned in
Section 3.1, some initial vibration of the cylinder is a condition for the development of the
symmetrical vortex pairs associated with this type of response.
lf a point on a vibrating cylinder is considered, it will be seen that the instantaneous
velocity of the point in a given plane of motion varies periodically in accordance with a
relationship approximating to the following form:
where:
2m
U = UmCOS
T
U = instantaneous velocity of point;
Urn ;;;; maximum velocity of point;
time from beginning of cycle;
T = period of vibration.
Eqn 3.7
If the cylinder is situated in a flow, with an approach velocity U, then it can be deduced that
the velocity of the flow relative to the cylinder will vary from U + Urn when the cylinder is
moving at maximum velocity against the flow direction, to U . Urn when the cylinder is
moving at maximum velocity with the direction of flow. When U and Urn are nominally
equal, the relative velocity will therefore vary between 2U and zero.
These latter conditions, in which the velocity of the flow relative to the cylinder decreases
to zero every cycle, can be interpreted as a situation in which the cylinder is being
repeatedly started impulsively from rest. If the amplitude of the vibration is additionally
small, then the displacement of the cylinder relative to the flow during each cycle will also
be small. Under these conditions, the vortices which grow from the separation points on
each side of the cylinder do not have sufficient time to develop instability, before they are
shed. Shedding in fact takes place each time the cylinder reverses into the flow direction
(ref, 8), and it is the periodic drag overshoot resulting from the growth and shedding of the
vortices which gives rise to the in-line excitation force.
The conditions that Urn ;: U, and that the relative displacement of the cylinder is small, tend
to occur for reduced velocity values in the appropriate range 1 < V R < 2 and vibration
amplitudes up to approximately 0.2 D. If Y
R
is increased above the latter range, then U
becomes greater then Urn, and the relative velocity of the cylinder and the flow no longer
decays to zero during each cycle. The conventional alternate vortex shedding regime then
becomes established, with the possibility that the second type in-line or subsequently the
cross-flow response may take place.
18
Contents
3.4 RESPONSE ASSESSMENT BASIS
Large amplitude locked-in span vibrations can occur if the vortex shedding flow regime
around the pipe is able to input energy to the pipe span at a frequency synchronised with
natural vibration frequency, fN, of the span.
The three fundamental types of response where this synchronisation can occur are described
in Section 3.1. The respective vortex-induced excitation and span vibration effects are linked
by an interactive physical process, which brings into question the relevance of considering
either effect in isolation once the span response has become established. For the purposes
of predicting the threshold at which vibration commences however, it is reasonable to
assume that the excitation frequency is proportional to the vortex shedding frequency, f"
where the proportionality relationship is dependent upon the response mode.
For a given response ~ o d e , three basic cases defining alternative relationships between the
vortex induced excitation frequency (proportional to f,) and fN, are considered below. The
general term 'vortex induced excitation frequency' is used in order to cover span excitation
by any of the three fundamental response mechanisms.
i) Excitation frequency fN
In this situation the frequency of the vortex-induced excitation process tending to
dynamically excite the span is much lower than the span's own natural vibration frequency.
The interactions between these two dynamic effects are therefore generally random and
uncorrelated. The span vibrations set up under these conditions are therefore small in
amplitude and are not sufficiently energetic to control the vortex shedding process and
initiate lock-in.
ii) Excitation frequency", fN
In this case, the frequency of the vortex induced span excitation effect closely matches the
natural vibration frequency of the span. The two synchronised effects therefore tend to
mutually reinforce each other and lead to a build-up of resonant span oscillations. These
comparatively energetic vibrations initiate the onset of 'lock-in', whereby the vortex
shedding process becomes controlled by and therefore synchronised with the motion of the
span itself. Once established, lock-in vortex-induced span vibration becomes a self-
sustaining effect, capable of modifying the vortex shedding frequency to coincide closely
with the natural frequency of the span.
iii) Excitation frequency fN
In this case the vortex-induced excitation frequency exceeds the natural vibration frequency
to the extent that synchronisataion of the two effects becomes impossible and resonance and
lock-in are prevented. This is not however a satisfactory situation from the point of view of
pipeline span stability, since in order to reach this high non-resonant range, fs must at some
stage pass through the unacceptable lock-in response region. The possibility of higher
harmonic vibration would also exist in this case.
The assessment of vortex-induced span response is -therefore essentially based upon a
comparison between the vortex-induced span excitation frequency (proportional to fs) and
the span natural vibration frequency, fN.
Comparison of fs and fN could in theory be carried out directly. It has become accepted
practice however, to combine this comparison with the calculation of fs to. form a single
step, making use of a dimensionless parameter known as the 'reduced velocity'.
The reduced velocity parameter is defined by the following expression:
where:
V.
U
fN
D
=
=
reduced velocity;
approach flow velocity;
span natural frequency;
pipeline diameter.
Eqn 3.8
19
Contents
The reduced velocity is thus seen to be a very simply calculated parameter. The physical
relevance of the parameter can be better appreciated however, if U/D is expressed in terms
of f, and S, using the standard Strouhal number relationship of equation 2.2. This gives an
alternative definition ofV
R
expressed as:
Eqn 3.9
where:
V. = reduced velocity;
U = approach flow velocity;
fN = span natural frequency;
D = pipeline diameter;
S = Strouhal number;
f, = vortex shedding frequency.
V. can thus be seen to incorporate effectively the Strouhal number, and the frequency ratio
fJ fN
The ratio fJ fN is an effective means of performing the required comparison between f, and
fN V R can thus be used as an indicator of possible span response. and has become
commonly accepted for this purpose, eg., Dn V 1981 (ref. 16).
A point which must be appreciated, however. when VR is used for span response
assessment is that the Strouhal number is also effectively incorporated into the parameter.
,S is not constant, and can vary as a function of a number of parameters as described in
Section 4. A given value of V R does not therefore define uniquely the ratio of fs/fN as
required for span response assessment. Additionally, the exact f,/fN ratio at which
significant response motion commences is subject to variation dependent upon the
characteristics of the system. Forced vibration effects can induce significant motion well
before the excitation frequency exactly coincides with fN. Furthermore, interaction between
the in-line and cross-flow response components can further modify the response thresholds.
In practice therefore, when using the reduced velocity method of span response assessment,
the thresholds of response must be defined with careful reference to real span conditions.
This should include full scale testing.
20
Contents
4. VORTEX SHEDDING - PARAMETRIC STUDY
4.1 INTRODUCTION
The frequency of vortex shedding from a stationary pipeline span is given by the following
simple relationship:
f,= SU
0
Eqn4.l
where:
f,
=
vortex shedding frequency
S = Strouhal number
U = approach flow velocity
0 = total external diameter of pipeline
The relationship between this stationary cylinder vortex shedding frequency, f, and the
natural vibration frequency of the span, fN. essentially detennines the dynamic response
behaviour of the span. The physical parameters which influence fs are therefore discussed in
detail in this section.
Equation 4.1 shows that f, is dependent upon the following three variables; Strouhal
number, S, approach flow velocity, V, and pipe diameter, D. Of these, the pipeline
diameter is a simply determined quantity, although care must be taken to include the
thickness of all external coatings. The Strouhal number however, is influenced by a
number of physical parameters, which are considered in Section 4.2. The remaining
variable is the approach flow velocity, which must be carefully defined in order to avoid
ambiguity in the results, and is addressed in Section 4.3.
Information contained in Section 4.2 and 4.3 was obtained from the following sources:
a) Existing literature;
b) A special study comprising laboratory tests on a stationary circular cylinder located
close to a solid plane boundary and subject to a controlled range of approach flow
conditions.
Experiments were conducted using two nominal approach flow valves of 80 mmls and
160 mmls respectively. With the 25 mm diameter test cylinder used for the
experiments, these nominal velocities gave nominal Reynolds number values of 2 x
10
3
and 4 x 10
3
respectively.
Three separate approach flow boundary layer configurations were used for this test
programme. These comprised a simulated uniform approach flow, a moderately thick
turbulent boundary layer formed over a smooth boundary, and a thick turbulent
boundary layer formed over a rough boundary. For subsequent reference, and
identification of data points on diagrams, these three boundary layer configurations
were given identifying code prefixes as follows; U = uniform approach flow, S =
smooth bed boundary layer (moderately thick), R = rough bed boundary layer (thick).
As an example. the designation S80 refers to a test series carried out using the smooth
bed boundary layer configuration and the nominal 80 mmls flow rate.
The principal results from this work have been published in the form of a paper
presented at the 15th Offshore Technology Conference, 1983 (ref. 17). A revised
version of the latter paper was also published in the ASME 10urnal of Energy
Resources Technology, March 1984 (ref. IS). A full description of the work and all
results is contained in the final report prepared by University College London (ref. 74).
c) A special study comprising tests on a full scale pipe span.
These tests were carried out in the tidal estuary of the River Severn, where current
velocities up to the order of 1.5 mls regularly occur. The test span comprised a length
of 20-inch 00 steel pipe, with a wall thickness of 0.5-inch. Span lengths of 50 m and
40 m were tested, and both bare metal and simulated concrete surface conditions were
investigated. The tests are described in reference 62, and the results are also extensively
discussed in Section 6 of this document.
21
Contents
Both the above special studies were commissioned by the United Kingdom Department of
Energy, and carried out by subcontractors under the management of J.P. KENNY. For the
laboratory tests, the subcontractor was University College London, whilst Hydraulics
Research Limited, Wallingford, carried out the full scale tests.
4.2 STROUHAL NUMBER
4.2.1 Physical Factors Influencing Strouhal Number
Figure 4.1 shows a typical cross sectional view of a pipeline span suspended a small
distance above the bed and subject to the influence of a subsea current. Adjacent to the
seabed, a thick turbulent boundary layer will generally exist, displaying a typical velocity
profile of the form shown in the figure. The resulting velocity gractient causes a velocity
differential ~ u , between the upper and lower surfaces of the pipe.
For the typical situation illustrated in Figure 4.1 the factors which may influence the value
of the Strouhal nwnber are listed below:
Reynolds number;
Pipe roughness;
Bed proximity;
Velocity gradient across pipehne diameter;
Approach flow turbulence intensity;
Span length/pipe diameter ratio.
These effects are described in greater detail below.
TURBULENT
BOUNDARY
LAYER
VELOCITY
PROFILE
u
u
Figure 4.1
PIPEUNE
Cross section through typical pipeline span
4,2.2 Reynolds Number
Reynolds number, Re = UD/u
wbere:
U = approach flow velocity
D = pipe diameter
u kinematic viscosity of surrounding water
22
Contents
The discussion of vortex shedding in Section 2 explains that the physical configuration of
the flow regime around a smooth circular cylinder varies with the flow Reynolds number. It
is therefore reasonable to find that the Strouhal number also demonstrates Reynolds number
dependence, reflecting as it does the physical characteristics of the vortex shedding process.
In order to appreciate the influence of the Reynolds number fully upon the frequency of
vortex shedding, it is helpful to understand the actual physical mechanisms which give rise
to and control the vortex shedding phenomenon. The following discussion therefore gives
an outline of the relevant processes, and shows how experimentally observed variations in
Strouhal number can be explained in terms of well established physical principles.
An extensive accumulation of experimental and general observational evidence now exists
which strongly supports the hypothesis that the sole requirement for regular vortex
shedding to take place from cylindrical bodies of arbitrary axisymmetric cross-sectional
shape, is that the flow must separate from the cylinder surface along straight lines at fixed
locations. This conclusion along with the supporting evidence has been previously
discussed in some detail by Naumann et al (ref. 32) and by Berger and Wille (ref. 33). On
the basis of these fundamental requirements, Grass (ref. 52) develops a logical description of
the physics of vortex shedding. Many of Grass' arguments, linking in experimental
observations with established physical principles, are repeated below.
Vortex shedding results from the basic instability which exists between the two free shear
layers released into the downstream flow from the separation points. These free shear layers
roll-up and feed vorticy and circulation into large discrete vortices which form alternately on
opposite sides of the generating cylinder. At a certain stage in the growth cycle of the
individual vortex, it becomes sufficiently strong to draw the other free shear layer with its
opposite signed vorticity across the wake. This action cuts off further supply of vorticity to
the vortex which ceases to grow in strength and is subsequently shed into the downstream
flow. The process then repeats itself on the opposite side of the wake resulting in regular
alternate vortex shedding.
These observed physical features suggest that the vortex shedding process is largely
controlled by two primary factors. Firstly, the distance, d .. between the two free shear
layers at the point just downstream of the body where they first became approximately
parallel (see Figure 4.2). Secondly, the rate at which velocity and circulation is fed into the
shear layers which is completely detennined by the velocity, U" at the outer edge of the
boundary layer at the points of flow separation. U, and d. are simply the relevantvelocity
and length scales, respectively determining the vortex shedding process.
u
---
FROM ROSHKO (REF. 181
Figure 4.2
,
o
TYPE B (dw/d<l)
Typical wake shapes for (a) subcrltlcal and (b) supercritical flow
A significant volume of experimental evidence is available to support the above line of
reasoning. Roshko (refs. 34 and 35) performed experiments upon a variety of bluff body
shapes at subcritical Reynolds Number and postulated the concept of a 'universal' Strouhal
number defined in terms of U, and lid . He later carried out further experiments to expand
the concept to very high Reynolds numbers above the critical transition region (ref. 18).
Other work, which has further supported the concept that the vortex shedding frequency is
proportional to U, and inversely proportional to d., has been carried out by Bearman (ref.
36), Chen (ref. 37), Simmons (ref. 38), Griffin (ref. 39) and Buresti (ref. 40).
In view of the practical difficulties of predicting U, and d. for any given case, the
'universal' Strouhal number concept developed from the above work has so far been limited
in its practical application. The work has however served to validate the concept that the
23
Contents
proportional to the wake width dw. Thus when d. is less than the diameter d, which is
typical of flows at Reynolds numbers higher than the critical transition range, the vortex
shedding frequency is expected to be higher than the frequency under subcritical conditions
for which d./d is greater than unity (see Figure 4.2). Experimental measurements and
momentum theory also demonstrate that there is a close correlation between the drag
coefficient, CD. of a body and the wake width, dw . When the wake width dw is large in
relation to width, d, then CD is large and vice versa. Thus there is also an
approximate inverse relationship between the drag coefficient and the vortex shedding
frequency.
In the case of circular cylinders therefore, the Strouhal number and vortex shedding
frequency are directly related to the position and stability of flow separation points on the
cylinder surface which determines d. and CD. A discussion of the influence of Reynolds
number and surface roughness on vortex shedding frequency and Strouhal number is thus
essentially narrowed down to the effects these parameters have on the position of the
separation points and hence on the wake width and the drag coefficient.
The fundamental condition for regular vortex shedding, namely that the flow should separate
along straight lines at stable positions. is satisfied in the case of flow round both smooth
and rough surfaced circular cylinders over a wide range of practically relevant Reynolds
numbers. Regular vortex shedding duly takes place. However, there is a critical range of
Reynolds numbers associated with the transition from laminar to turbulent flow in the
boundary layer formed on the upstream surface of the cylinder, within which the flow
behaviour is relatively complex. This results in unstable separation points and irregular
vortex shedding in the resulting disrupted regime.
From the practical point of view, it is the critical Reynolds number range which gives rise
to the greatest nncertainties in the prediction of Strouhal number behaviour. A typical plot
showing Strouhal number as a function of Reynolds number is shown in Figure 4.3, which
was originally presented by Lienhard (ref. I) and subsequently reproduced by Blevins (ref.
41). The plot Clearly shows the region between approximately R. 2 x 10' and
R. 3.5 x IO'where effects associated with the critical transition process cause the
Strouhal number curve to broaden out to fonn an extensive envelope of possible values.
The physical mechanisms responsible for this behaviour are discussed below for the
particular case of smooth circular cylinders, from which most existing experimental data
has been obtained. The important and practically relevant case of rough surfaced cylinders is
addressed separately in Section 4.2.3

,,'\
/ \
04
03
"'
'"
'"
"
02
10,
"' >-
m
/ I
/ I
/ I
/ I
/ I I
/ \
"
/
Z:?Z<
REYNOLDS NUMBER (Re)
FROM LIENHARD (REF. I)
Figure 4.3
Strouhal-Reynolds number relationship for circular cylinders
In the case of flow round a smooth surfaced circular cylinder, thin boundary layers of
retarded fluid develop on the upstream face of the cylinder and grow in thickness round the
circumference. On encountering regions of adverse pressure gradient in which the static
24
Contents
pressure increases in the flow direction, the already retarded velocity of the fluid particles in
the boundary layer is rapidly reduced to zero as its kinetic energy is transfonned into
pressure energy to match Ibe surrounding pressure field. Flow reversal results adjacent to
!be cylinder surface and Ibe flow separates from Ibe body fonning a downstream wake. The
highly sheared fluid in the boundary layers leaves the cylinder surface at the separation
points and forms the two free shear layers responsible for the vortex shedding as discussed
above.
At lower Reynolds numbers in Ibe so called subcritical range, 3 x 10' < 2 x 10', the
upstream surface boundary layer is laminar. Early separation occurs at an angle of
approximately 80
Q
from the upstream stagnation point with stable separation points and a
relatively wide wake width, d,.., between the two cast off free shear layers. This situation is
illustrated in Figure 4.4a) . The vortex shedding is highly regular and also remains closely
constant with a value of approximately 0.2.
As the Reynolds number is further increased, the flow enters a transition region,
comprising a relatively wide zone defined by 2 x 10' < 3.5 x 10' within which Ibe flow
regime surrounding the cylinder passes through a complex, changing state of transition. It
is convenient to subdivide the transition zone into two regions. respectively tenned
'critical' and 'supercritical' in Ibe terminology of Roshko (ref. 18). These regions enclose
stages in the overall transition process, as is explained in the following text.
The critical region, which represents the first part of the overall transition zone commences
with the onset of instability and turbulence in Ibe separated free shear layers downstream of
the cylinder which affects the near cylinder pressure distribution. The angle at which
separation takes place increases slightly (Achenbach ref. 42) thus reducing the wake width
in Ibe approximate Reynolds number range 2 x 10' < Re < 3 x 10'.
As a Reynolds number just in excess of 3 x 10', Ibe flow undergoes a mucb more profound
instability in which the detached free shear layers re-attach themselves, in a much thickened
turbulent condition, along with tbe outer flow, back to Ibe cylinder surface. This results in
the fonnation of trapped laminar separation bubbles as illustrated in Figure 4.4 b). Tbe
turbulent boundary layer flow eventually re-separates at an angular position of
approximately 140' well round on !be downstream face of!be cylinder (Acbenbacb ref. 42).

0)
LAMINAR
SEPARATION
TUR9JLENT
SEPARATION
TURBULENT
REATIACHMENT
3 1( <.3 X 10
6
b)
Figure 4.4
c)
TURB.JLENT
SEPARATION
Separation positions for various Reynolds number ranges
Bearman (ref. 43) observed Ibat this process occurred in two stages. Firstly, the flow re-
attached itself on one side of the cylinder only, at a Reynolds number of 3.4 x 10' This
produced a single laminar separation bubble and an asymmetric flow pattern which
generated a mean transverse lift force. Under the carefully controlled conditions of
Bearman's tests, twin laminar separation bubbles first formed at a Reynolds number of
3.8 x 10'. This re-established a symmetrical mean flow pattern and Ibe mean lift force
disappeared
25
Contents
o -5
o
S
o -3
o -2
o -I
0
26
!
,
Bearman
r
~ Delcny S Saensen
"
"
~
,
,
I
I
)
,
I
.} ,
I ,
,
: Retf 8 Simmons
), ", ,
Ribner a Elkin
Re
DATA FROM BEARMAN (REf 36)
ROSHKO (REF. IB)
RIBNER a ETKIN (REF. 19)
RELF a SI MMONS (REF 20)
DELANEY a SORENSEN (REF_ 21)
Figure 4.5
Experimental Strouhal number - Reynolds number data
< f D O ~ c P 0
Roshko
I
Contents
...J
..
z
o
iii
z
"'
2;
is
z
o
z
10
U
R = 224 .10'
(lUb- crilical)
~ ~
I 0
I
I ~ )
I
I
/' R = 368.10'
I
0
I I
I (one bubble)
I
1I-___ J.
-- 012
~
HOT - WIRE POSITION
j ~
I
V
ltX
f0
\
R=4.16.IO'
...
(Iwo bubbles)
I'
R=3'55xlo"
1\
I !II
R = 7 2.
0
~
l/
0-18 0'26 0-34 0-42 0-50 0-58 0-66
STROUHAL NUMBER
Figure 4.6
Power spectral density for vortex shedding at various Reynolds numbers
27
Contents
The significant narrowing of the wake width, d., in the critical flow regime would be
expected to produce a substantial increase in the vortex shedding frequency. This is
confirmed by Bearman's (ref, 43) measurements (Figure 4.5) of regular vortex shedding at
frequencies giving the unusually high Strouhal number of 0.46 which persisted up to a
Reynolds number of 5.5 x 10' marking the end of the critical flow regime, where the drag
coefficient, Co is minimum. Bearman drew attention to the fact that this critical zone
vortex shedding is relatively weak, producing peak spectral energy values almost two orders
of magnitude less than the corresponding levels in the subcritical regime as indicated in
Figure 4.6. This suggests, as pointed out by Bearman, that any fluctuating pressures
induced on the cylinder surface by the critical zone shedding would also be very much
weaker and thus less likely to generate vortex induced resonance.
An additional factor which furrher diminishes the likely practical significance of the regular
high frequency but. weak vortex shedding observed by Bearman in the critical zone, is its
sensitivity to upstream disturbances. These can take the form of either surface irregularities
or approach flow turbulence which trigger the formation of turbulence wedges or spots in
the upstream boundary layer. This disrupts the separation bubbles producing gross three-
dimensionality in the flow and irregular vortex shedding characteristic of and prevailing
throughout the supercritical flow regime following the critical zone. In the case of a
hydrodynamically smooth pipeline situated in the highly turbulent near bed flow of a tidal
boundary layer it thus appears highly unlikely that the high frequency shedding observed
under idealised wind tunnel conditions could ever become established.
The supercritical flow regime begins to become apparent for Reynolds numbers greater than
approximately 5 x 10', where the separation bubbles become increasingly disrupted by the
upstream disturbances. This breaks the condition of straight line separation and the vortex
shedding becomes irregular covering a broader band of frequencies. Delany and Sorensen
(ref. 21) presented measurements of Strouhal number based on the broad band spectral peak.
The values suggested ranged between 0.4 and 0.45 over the supercritical Reynolds number
range between 6 x 10' and 1.6 x 10
6
as shown in Figure 4.5. Roshko, (ref. 18) also made
measurements in the supercritical region but failed to detect any predominant frequency
ascribable to vortex shedding. He speculated that the reason for this was that his monitoring
position was further downstream than that used by Delany and Sorensen. It is clear
however, that any irregular. vortex shedding which may exist in this regime is relatively
weak, again probably high sensitive to upstream disturbance, and hence unlikely to be of
practical significance in the context of pipeline span excitation.
As the Reynolds number is increased still furrher, the supercritical regime gives way to the
so called 'transcritical' regime, again using the tenninology of Roshko (ref. 18) (it should
be noted that other researchers have used different tenninologies to define the various flow
regimes). In the transcritical region, the boundary layer has undergone transition from a
laminar to a turbulent state on the upstream face of the cylinder. A stable situation exists in
which the turbulent boundary ultimately separates from the downstream face of the cylinder
at an angle in the region of 110 to 120 from the upstream stagnation point. These
separation angles are larger than for the equivalent subcritical case, leading to a reduced
wake width as shown in Figure 4.4 c).
Once again the conditions of stable straight line separation points is satisfied and regular
vortex shedding is observed. Roshko (ref. 18) recorded an increased vortex shedding Strouhal
frequency of 0.27 in this transcritical regime. which is consistent with the inverse
relationship between Strouhal number and wake width discussed above.
Roshko (ref. 18) suggested that the transcritical regime commenced at Reynolds number
3.5 x 10
6
, coinciding with the observed re-establishment of regular vortex shedding during
his experiments. Achenbach (ref. 42) defined the end of the preceding supercritical region as
the stage where separation bubbles vanish and transition to turbulence takes place in the
upstream attached boundary layer, suggested that this occurred at a Reynolds number
1.5 x 10"
When experimental data derived from smooth cylinder research work is examined in
conjunction with the actual physical flow processes taking place, the observed variations in
vortex shedding Strouhal number throughout the Reynolds number range can be logically
explained. The basic mechanism initiating changes in the flow regime around the cylinder
is the process of transition from laminar to turbulent flow. a fundamental fluid mechanics
phenomenon first investigated by Reynolds (ref. 44) in the 19th Century. A feature of
28
Contents
laminar to turbulent transition, revealed by Reynolds' early experiments, is its sensitivity
to external disturbing influences, which in a practical pipeline situation includes the effect
of pipe surface roughness. It is therefore not surprising to find that the Strouhal number
behaviour discussed above for the specific case of a smooth cylinder is subject to significant
modification if the cylinder possesses a realistic hydraulically rough surface. In view of the
importance of this aspect, it is addressed separately in Section 4.2.3.
4.2.3 Pipe Surface Roughness
The physical influence of a rough surface upon an adjacent fluid flow is basically quite
simple. The individual elements of the roughness physically obstruct the flow, leading to
the production of turbulence and generation of fonn drag. The size of roughness required for
these effects to become apparent is generally quite small, being typically of the order of a
fraction of a millimetre. Truly hydraulically smooth conditions are therefore confined to
quite finely finished surfaces such as glass, machined metal etc., and would not normally be
expected on a typical subsea pipeline.
The boundary layer flow disturbances induced by pipe surface roughness can impose a major
influence upon the process of laminar to turbulent transition in the flow around the pipe.
This influence is reflected in the vortex shedding behaviour and is consequently highly
relevant to the assessment of vortex induced span vibration.
An important stage in determining the influence of roughness is the quantitative definition
of the roughness itself. This process is complicated by the many possible forms which
roughness can take, with a consequently large number of variables being present. Material
characteristics, manufacturing and handling procedures and marine growth can all contribute
towards the production of pipeline surface irregularities of varying shape, size and
distribution pattern.
A simple form of roughness is comprised of discrete granular particles (eg., sand or
pebbles) evenly distributed over the surface. An obvious parameter for quantifying this type
of roughness is the mean grain diameter, k, and important early work by Nikuradse (ref. 45)
on the flow through sand-lined pipes developed a roughness classification based on the sand
grain diameter. Subsequent research showed however, that other factors such as distribution,
shape and grading of the particles, significantly affect the roughness influence.
Nevertheless, Nikuradse's results have been adopted as a standard, and other roughnesses are
classified in terms of an 'equivalent sand roughness', les, defined as the sand grain diameter
on Nikuradse's scale which would give the same physical flow characteristics.
In the case of flow past circular cylinders such as near bed pipelines the surface roughness is
normally specified by relative roughness parameter, kJD, where 0 is the pipe diameter. The
relative roughness and the Reynolds number represent the two dominant variables which
determine the general state of flow including the vortex shedding characteristics from
circular cylinders.
Most of the studies of rough cylinder flow carried out to date have been specifically
concerned with the mechanics of drag force generation. However, because of the common
dependency of both the drag and the vortex shedding frequency on the flow
separation point position and the wake width, these drag studies also provide valuable
information relevant to the vortex shedding process. The investigation reported by
Achenbach (ref. 46) is particularly useful in this respect since he studied the flow behaviour
round circular cylinders over a relative roughness range typical for seabed pipelines.
Achenbach, in common with most of the smooth cylinder studies reported in Part a) carried
out his measurements in a wind tunnel with a low approach flow turbulence intensity of
approximately 0.7 percent. Three different relative roughnesses were studied, namely,
kJD = 1.1 X 10', 4.5 X 10' and 9 x 10' across a wide range of Reynolds numbers defined
by 4 x 10'< Re < 3 x 10
6

Achenbach used a single test cylinder with three different grades of emery paper for surface
roughness. The equivalent sand roughness size k.. was obtained by direct measurement of
friction factors in a duct flow and matching these values to Nikuradse's (ref. 45) standard
results for sand lined pipes.
29
Contents
____________________ ,
xX x
: \
0-24
: I 'xX


s
0-
I I "'--_
I
I ,
I I
I
,
,
I
0-20
; ) J

0-181+---.----.-...-::-------.
0-4 0-6 O-S 10" 2 Re
o x 6 V
D=6lmm kID 1-77 2-974-10 6-S9
QJ x 10-
3
0-26,-------------------------,
0-24 rx
t \. x
s
, Xx
0-22

f I
,
0-20
0-1
10'R. 04 0-6 O-S
0
x
6 V
D=34mm kID 3-18 5-32 7-35 12-35
b)
.10-3
0-26 0-26,--------------,
0-24
S
022
0-20
,
I
, I
I I /
I I /
I f /
I I
))/0
P
I
I
0-18'+--.----,r--.-.---------.----'
08 10
5
06 2 R.
o x b. V
D=90mm k/D
3
'-20 2-01 2-78 4-67
c) 10-
FROM, SURESTI (REF. 47)
Figure 4.8
0-24
s
0-22
0-20
I
I
,
I
I
I
I
, '
, /
,.
;..c
,
/
I
I

0-8 10
5
D= 119mm
d)
o
kID 0'91
.,0-
3
2 3Re
X 6 V
I-52 2-10 3-53
S 8S a function of kiD and Re Buresti (ref 47)
32
Contents
5
5
5
o 301+--i'---ii--;---<;>.:S"--'"' ..... t--+-t--t--;
030
o 15
5
,,"Cr,
10
4
2
03
020
o 15
Re
If''''
k,u,
""'"
l>A,
i"''''''''->
5 ~ 5 2 5 ~ 6 2
Re
..m.
In
Re
VV
V V
v .
nO
10
3
2 5 10
4
2
Re
5
Figure 4.9
0) ks 10 = 075x 10-
3
FROM ACHENBACH AND
HEINECKE (REF 58)
S as a function of k.1D and Re Achenbach and Heinecke (ref 58)
33
Contents
OA
s
03
02
34
... . ..
2
... .,
/
,
I
I
r-'- ::::"
''"i-'"
....
.. . ..
---
/ ;
... ...
5 2
SMOOTH
"s ID = 075 x 10-
3
kS/D = 3-0 x 10-
3
ks/D = 9-0 x 10-
3
ks/D = 30-0 x10-
3
I
I
I
,
I
"'- '-.
L"
'-
=.:.. .. ..;.:;:
5
Re
FROM' ACHENBACH AND HEINECKE (REF 58)
Figure 4.10
~
Irt,c- -
.: ='.L..:._:::
-------
2 5
S as a function of k./D and Re summary diagram (ref 58)
Contents
kiD
2,----------------------------------,
8
6
4
2
10-3
000""'" V V
o
04
o SUBCR ITICAL
CRITICAL
o
V SUPERCRITICAL
'" TRANSCRITICAL
o 0
FROM: SUREST I (REF. 47)
Figure 4.11
Boundaries between flow regimes
2 4
35
Contents
A detailed discussion of flow behaviour around rough circular cylinders, with particular
reference to vortex shedding, is presented by Grass (ref. 52). Grass analyses in detail the
inter-relationship between the basic physical influence of roughness upon the flow regime
and the ultimate observed vortex shedding behaviour. In addition to the extensive
discussion, Grass provides a summary ofroughuess effects, which is reproduced below.
As in the case of smooth cylinder flow, four regimes, namely the subcritical,
critical, supercritical and transcritical can be identified categorising the flow past
rough surfaced cylinders. These flow regimes are essentially detenruned by the
position of the transition from laminar to turbulent flow in either the attached
laminar boundary layers formed on the upstream face of the cylinder- or in their
'continuation as the separated free shear layers.
The major effect of increasing roughness is to lower the Reynolds number at which
the various phases of the shear layer transition take place. This has the effect of
progressively lowering the values of the bounding Reynolds number defining the
four flow regimes as illustrated in Figure 4.11.
In the subcritical regime, the flow separates from the cylinder in a laminar condition
at stable separation points. This produces regular vortex shedding with a Strouhal
number of approximately 0.20.
In the critical flow regime the separated laminar free shear layer undergoes transition
to turbulence. The subsequent re-attachment of the flow to the cylinder surface
produces a very narrow final wake width and a characteristic minimum drag
coefficient. General instability in the location of the final flow separation points
eliminates vortex shedding in the critical regime. As indicated in figure 4.11, the
Reynolds number range defining the width of the critical regime narrows rapidly
with increasing relative roughness, kJD until it reduces to zero at a kJD of
approximately 10 x 10-
3
For relative roughnesses in excess of 10 x 10.
3
vortex
shedding is always present and shows a step increase in Strouhal frequency at the
critical Reynolds number.
In the supercritical flow regime; the transition point moves rapidly into the attached
boundary layer flow on the upstream face of the cylinder. This produces stable
turbulent flow separation points and re-establishes regular strong vortex shedding
which contrasts with the irregular weak shedding from smooth cylinders in this flow
regime. Surface roughness therefore stabilises the shedding in the supercritical
regime which considerably narrows the Reynolds number range within which vortex
shedding is eliminated compared with the corresponding smooth cylinder range.
For practical range of relative roughness considered between 0.75 x 10" and 30 x 10'
" the Strouhal number typically falls from a peak value of between 0.46 and 0.22 at
the beginning of the supercritical regime to a value between 0.2 and 0.21 on
entering the transcritical zone. The results of Buresti ( ref. 47) tend to be slightly
lower than those of Achenbach and Heinecke (ref. 58), but as a general rule, larger
Strouhal numbers tend to be associated with smaller roughness and vice versa.
In the transcritical regime, the flow characteristics including the vortex shedding
behaviour become largely independent of the Reynolds number. The Strouhal
number remains constant and ranges between 0.23 and 0.21 with the larger of the
two values once again associated with the smaller relative roughness value.
The presence of surface roughness appears to eliminate the weak; double frequency
vortex shedding recorded on smooth cylinders in the critical and supercritical flow
regimes.
Spanwise correlation measurements reported by Buresti (ref. 47) indicated that the
correlation length was of the order 30 in the subcritical regime, less than 20 in the
critical regime, 30 in the supercritical regime and 40 in the post critical regime.
Buresti points out that these measurements demonstrate that vortex shedding from
roughened cylinders in the supercritical and especially the transcritical flow regimes
is at least as strong as in the case of subcritical shedding.
In order to apply these results to a practical pipeline span situation, it is ~ e c e s s a r y to
estimate the expected relative roughuess of the pipeline. When dealing with a subject as
complex as surface roughness, it can be difficult to define a precise value, but it should be
36
Contents
possible to make a reasonable approximate estimate. The process is simplified io a certain
extent by the limited number of pipeline surface finishes in nonnal use. These can be
basically classified as anti-corrosion coatings and concrete coating, both of which are of
course subject to the modifying effect of marine growth.
A typical illustration of the effect of varying pipeline surface roughness is shown in Figure
4.12. This example is based upon the assumption of a medium sized pipe, 0.5 m
(approximately 20-inch) in diameter, with typical k. values of 0.75 mm and 6.0 mOl for
corrosion wrap and concrete coating surfaces respectively. These values give relative
rougbnesses of 1.5 x 10-
3
and 12 x 10.
3
respectively. Using the data of Buresti (ref. 47, see
Figure 4.8) the Strouhal number - Reynolds number relationship curves for two cases are
presented in Figure 4.12. For comparison, the smooth pipe case as summarised by Lienhard
(ref. I) is also presented.

SMOOTH PIPE ------..; \
(ks/O; ,.sl'O-31 // \ !
TYPICAL / I .
2 CORROSION I I :
(ksf
D
=I-2.l10- \OATING " I i.
TYPICAL / I. !
CONCRETE I \ -V:
COATING 1-.
I , ? i ! fO 2 :1}; , ....... .... w .......... _ .. ;
03
04

01
R.
ASSUMED: 0 = 0 5 m
ks = CORROSION WRAP = 0-75 mm
ks = CONCRETE = 6 mm
Figure 4.12
Strouhal-Reynolds number relationship for typical pipe roughness conditions
4.2.4 Bed Proximity
The inherent configuration of most pipeline spans generally results in suspended length of
pipeline lying in close proximity to the seabed. Consideration of an idealised situation of
flow around a circular cylinder adjacent to a solid boundary suggests that the presence of the
boundary will have an effect upon the frequency of vortex shedding form the cylinder. The
results of experimental investigations generally support this view.
In the case of a circular cylinder placed in an infinite unifonn t10w of inviscid
incompressible fluid, the flow pattern is completely symmetrical upstream and downstream
of the cylinder in the absence of boundary layer separation effects induced by viscosity. The
portion of the incident fluid flow which originally passed through the space occupied by the
cylinder is diverted round its sides. This is accompanied by increases in local velocity which
can be precisely calculated using potential flow theory. The flow velocity past the sides of
the cylinder initially reduces rapidly from its large maximum value of twice tbe undisturbed
approach velocity at the two extreme side points on the cylinder surface to a value only 4
percent in excess of the approach velocity at a distance of 2 diameters from the cylinder
surface. Approximately 80 percent of the approach flow blocked by the presence of the
cylinder therefore passes the sides 'of the cylinder within 2 diameters of its surface, rather
than being uuifonmly re-distributed across the entire width of the side flow.
37
Contents
A solid boundary introduced parallel to the approach flow close to one side of the cylinder
has two main effects. Firstly, it destroys the original symmetry of the flow field between
the two sides of the cylinder. This generates cross-flow lift forces even in the ideal inviscid
flow under discussion. Secondly the presence of the bed further increases the velocities close
to the sides of the the cylinder particularly on the bed side. These velocity increases are
caused by the re-distribution of the portion of the approach flow blocked by the cylinder
which originally flowed past the bed side of the cylinder outside the gap zone. For example,
according to tbe figures quoted above, if the gap between the cylinder base and the bed is 2
cylinder diameters then approximately 10 percent of the total blocked approach flow will be
re-distributed round the cylinder. Most of this re-distributed flow passes through the bed gap
producing significant velocity increases. The velocity field can again be calculated using
potential flow theory by employing 'image' cylinders to simulate the bed plane boundary
conditions, as described by Carpenter (ref. 23).
Such calculations indicate that as the bed gap to cylinder diameter ratio GID, decreases, so
the velocity increments close to the cylinder sides produced by the bed proximity, increase.
The viscous property of real fluids, causes the flow to separate from the pipeline cylinder
surface leading to the fonnation of a downstream wake and vortex shedding. However,
outside the thin surface boundary layers, potential theory still predicts the flow field on the
upstream half of the cylinder reasonably accurately. It is therefore to be expected that the
velocity enhancement resulting from the differential bed blockage effect described above in
the case of an ideal inviscid fluid flow, will also occur in practical flow situations. Since
the velocity enhancement close to the cylinder surface caused by the bed proximity is
qualitatively equivalent to an increase in the uniform approach velocity, it is therefore also
to be expected that the frequency of vortex shedding, which is proportional to the effective
approach velocity, and hence the Strouhal number, will progressively increase as the bed
gap ratio is reduced.
Despite a convincing theoretical basis, the actual experimental measurement of the
influence imposed upon the Strouhal number by the proximity of the seabed is not a
completely straightforward task. The measurement of the vortex shedding frequency tends to
include a degree of inevitable experimental scatter, which can mask the comparatively small
Strouhal number shifts being measured. Previously existing research relating to the subject
had revealed some apparent inconsistencies between different sets of results. In order to
clarify the situation, the special study described in Section 4.1 was carried out, involving
laboratory tests over a range of controlled flow conditions upon a circular cylinder situated
close to a solid plane boundary. The principal results from this test programme are
presented in references 15 and 17, and a full report of the programme is given in reference
87.
Previous experimental investigations were reported by Goktun (ref. 24), Haffen (ref. 25),
Bearman and Zdravkovich (ref. 26), Buresti and Lanci.otti (ref. 27) and Agrilli et al (ref. 28),
in which the effects of bed proximity on vortex shedding frequency and suppression were
studied for cylinders close to a plane boundary in a simulated uniform approach flow. The
approach flow used in all these tests was formed by the free stream region outside a very
thin wall boundary layer developed on the surface of the flat bed plates. Goktun found that
as the test cylinder was moved in towards the bed plate, the Strouhal number increases
slightly to a maximum value approximately 5 percent greater that the free stream value at
gap ratio, G/D = 0.5.
Haffen recorded a similar increase in the Strouhal number which reached a maximum at
G/D = 0.75 and then decreased rapidly for smaller G/D values. Bearman and Zdravkovich
concluded from their measurements that the Strouhal number remained constant as the test
cylinder approached the bed plate. Careful re-examination of their data, however, making
allowance for the inevitable scatter in this type of measurement, does appear to reveal an
upward trend in the Strouhal number. It again reaches a maximum approximately 5 percent
greater than the free stream value at a GID ratio in the region of 0.75 prior to reducing
again closer to the bed. Buresti and Lanciotti failed to detect any significant charges in
Strouhal number with reducing bed gap ratio. Angrilli et at measured a maximum increase
in Strouhal number of slightly over 10 percent at GID = 0.5.
Results from the special study (ref.'I7) carried out under uniform approach flow conditions
are shown in Figure 4.13. These measurements indicate that as the gap between' the bottom
of the pipe and the bed is decreased below two pipe diameters, the vortex shedding Strouhal
number increases slightly. It rises to a maximum value between 5 and 10 percent greater
38
Contents
than its free stream magnitude remote from bed influences, at a gap ratio G/D, of
approximately 0.75. For GID ratios less than 0.5, the Strouhal number decreases below its
free stream value. The latter behaviour is peculiar to simulated uniform approach flows
with very thin bed boundary layers which form stable separation zones upstream of the
cylinder for small bed gap ratios.
0'30
l:!.

t:;,0
t::.
2
0-20
tP
'"
0'10
"

TEST U 80
1::.
TEST UI60
0'00
000 050 100 1'50 200
GID
RESULTS FROM' SPECIAL STUDY (REF 17 )
Figure 4.13
Strouhalnumber as function of G/D
2-50
These detached separation regions severely distort the local flow patterns and reduce the
velocities past the sides of the cylinder which in tum reduces the vortex shedding frequency.
The present observations thus draw attention to the fact that considerable care needs to be
exercised in interpreting the results of earlier wind tunnel studies using uniform approach
flows which were susceptible to this somewhat artificial effect. Regular vortex shedding
was sharply suppressed for gap ratios less than approximately 0.3. This suppression is
considered to be strongly influenced by the apparent cancellation of vorticity which takes
place in the lower wake region when the pipeline is close to the bed and which is enhanced
by the fonnation of detached separation zones on the bed downstream.
The results of the special study (ref.17) taken in conjunction with the overall conclusions
from previously published work, appear to be reasonably conclusive. They show that at the
sub-critical Reynolds numbers investigated, the effects of bed proximity can act to increase
the Strouhal number defining the frequency of vortex shedding from a circular cylinder. A
convenient summary of,available experimental results is given in Figure 4.14.
39
Contents
0
'"
,
'"
150,--------------------------,
('00
Or
~
o-so
0

0
0'00 050 100 ISO
G/D
RESULTS FROM: SPECIAL STUDY (REF. 17)
BEARMAN AND ZORAVKOVICH (REF. 26)
ANGRILLI et 01 (REF. 28)
. Figure 4.14

~
I
SPECIAL STUOY- I
UNIFORM APPROACH FLOW:
ANGAILLI ET AL I
saz UPPER
I
saz LOWER
200 2-50
Normalised Strouhal number 8S function of GO
4.2.5 Velocity Gradient Across Pipe Diameter
The characteristic turbulent boundary layer velocity profIle fonned when a steady current
flows over the seabed is shown in Figure 4.1. The incident approach flow experienced by a
pipeline immersed in this thick seabed boundary layer-will thus be highly non-umfonn_
Due to the velocity gradient within the boundary layer, large differences in velocity may
occur between the top and the bottom of the pipeline. This leads to problems in defiriing an
appropriate approach velocity to correlate vortex shedding frequencies since the shedding
frequency is no longer a unique function of the incident centre line velocity but also depends
on the velocity gradient across the pipe diameter.
No experiments appear to have been carried out previously, concerning the influence on
non-umfonn boundary layer velocity gradients on vortex shedding frequencies from smooth
circular cylinders. However, Kiya et aI (ref. 29) have reported the results of experiments in
which the vortex shedding frequency was measured for a smooth circular cylinder placed in
all approach flow with a unifonn velocity gradient. These experiments, which were carried
out at sub-critical Reynolds numbers in the range 80-1000, showed that the shedding
frequency and Strouhal number increased with increasing velocity gradient. At a Reynolds
number of 10', the Strouhal number increased by approximately 15 percent of its uniform
flow value when the difference in approach velocity across the cylinder was 25 percent of
the centre line velocity. There was also evidence that vortex shedding was suppressed when
the velocity gradient across the cylinder exceeded a certain critical value which was
dependent on Reynolds number. These experiments thus demonstrated that boundary layer
velocity gradients in the approach flow could produce significant increases in vortex
shedding Strouhal number for pipeline spans close to the bed.
The special study (ref. 117) induded experiments to measure the vortex shedding Strouhal
number for a model pipeline span immersed in the non-unifonn approach flow fanned by
the turbulent boundary layer adjacent to a solid surface. Experiments were carried out over
both rough and smooth boundaries, which respectively produced considerably differing
velocity profile configurations. This allowed the velocity gradient effect to be assessed
40
Contents
separately from the bed proximity influence, and also allowed a wider range of velocity
values to be studied. In view of the scarcity of previous infonnation on this
subject, the results of the special experimental study are discussed in some detail below.
The results obtained from this study are summarised in Figure 4.15. The plotted trend
curves respectively summarise the data from the rough boundary tests (prefix R), which
provided a thick approach flow boundary layer, and the smooth boundary tests (prefix S),
which provided an approach flow boundary layer of moderate thickness. For comparison
purposes, results from tbe uniform approach flow tests (prefix U), which were discussed in
part c) above, are also included. As a general rule the highest velocity gradients across the
cylinder occurred for the rough boundary tests, moderate gradients occurred for the smooth
bed tests, and tbe gradient was zero for the uniform flow tests.
It can be seen that the measured Strouhal numbers are significantly influenced by the
approach flow boundary layer condition. The discussion below considers these influences in
detail.

100
------
--- '"--. ..
..".- --- ----.--

0-50
TESTS U80, UI60
TESTS S60,5160
TESTS RBO,RI60
0'00
0-00 "50 I-DO 1'50 2-00 250
G/O
Figure 4.15
Normalised Strouhal number 88 function of G/D for varying approach flow
conditions
When the rigid model pipeline cylinder was immersed in the realistic non-uniform approach
flow formed by the thick bed boundary layers with significant velocity gradients, the
measured vortex shedding Stroubal numbers again increased as the bed gap ratio GJD was
decreased below 2 and the local boundary layer velocity gradients correspondingly increased
The Strouhal number reached a maximuin value somewhat closer to the bed than in the case
of uniform approach flow, at a GJD value of approximately 0.5. It then reduced for smaller
GJD values but remained above the free stream value down to the vortex shedding
suppression point. In the case of the smooth bed approach boundary layer the Strouhal
number increased to a maximum between 10 and 15 percent greater than the free'stream
value at a GJD value of approximately 0.5. The velocity gradient parameter, L\,U/U where
L\,U is the difference between the approach velocities at the top and bottom of the pipeline
and U is the centre line velocity, ranged between 0.02 at GJD = 2 and 0.30 at GJD = 0, for
the smooth bed boundary layers. At GJD = 0.5, where the Strouhal Dumber peaked, L\,U/U
was approximately equal to 0.2. Vortex shedding was again sharply suppressed for G/D
values less than 0.3.
This general pattern of behaviour was reproduced in the case of thick approach flow
boundary layer formed on the rough bed. However, in this case, the velocity gradients were
considerably increased with L\,U/U values now ranging between 0.11 and 0.55 across the test
41
Contents
zone. This produced a maximum increase in measured Strouhal number of approximately
25 percent at OlD = 0.5 where the velocity gradient parameter AVIV = 0.35. In the rough
bed flow, regular vortex shedding was severely disrupted for OlD values less than 0.5. This
enhanced suppression of vortex shedding is probably due to a combination of several
factors. These include the likely existence of some critical value of AVIV for maintenance
of vortex shedding which may have been exceeded at small OlD values in the rough bed
test. Gross perturbations in the approach flow produced by the violent large scale turbulence
fluctuations close to the rough bed are likely to have had a major disruptive effect on
regular vortex formation and shedding. Flow visualisation observations also suggested that
vorticity cancellation in the lower wake zone was considerably increased in the rough bed
flow.
The results discussed above are re-plotted in Figure 4.16, which shows a graph of measured
Strouhal number plotted against the velocity gradient parameter, AVIV. These data points
were measured near the channel bed where t.,VIV and OlD are both significant. The graph is
therefore slightly ambiguous, in that the measured Strouhal number values incorporate the
effects of the bed proximity influence superimposed on the separate velocity gradient effect.
The superposition of these two physical effects leads to an apparently significant degree of
scatter in the data.
030,-----------------------------,
0-20
0'10
0'001
0-00 010
o
o 0
(I '"
020
o
6U/U
o
030
RESULTS FROM: SPECiAL STUDY (REF. 17)
Figur. 4.16
o
o
OR 160
DR 80
\; S 160
6 S 80
DAD 050
Strouhal number as function of ,6,u/U bed proximity effect Included
In order to isolate the effect of the velocity gradient alone, Figure 4.17 is included. In this
graph, the previously measured Strouhal number increment due to the bed proximity effect
has been subtracted from the plotted data values, which therefore reflect the influence of the
velocity gradient parameter alone. In spite of non-linear interaction effects, this procedure is
likely to be reasonably valid except for very low OlD values where the upstream separation
effects, peculiar to simulated uniform approach flow tests, produce excessive distortion as
discussed in par! c) above. This corrected data has been plotted along with the data envelope
of Kiya et al (ref. 29), who measured Strouhal numbers under the influence of uniform
approach flow velocity gradients alone. To allow realistic comparison between sets of data
measured under different experimental conditions, the corrected Strouhal number values (SI)
are normalised on the respective free stream Strouhal numbers (So) measured remote from
the influence of bed proximity and velocity gradient effects. Figure 4.17 therefore represents
a plot of SdSo as a function of AVIV.
As can be seen, the data points exhibit significantly reduced scatter and are remarkably well
summarised by and in excellent agreement with the data envelope lines defining the
measurements by Kiya et al (ref. 29).
42
Contents
'50,---------------------------,
100
050
o
o RI60
o R 80
'\7 $160
lJ. S 80
o
DATA ENVELOPE;
- -- -- KIVA at 01 1960

000 oro
L. U/U
RESULTS FROM' SPECIAL STUDY (REF (7)
KIYA el 01 (REF 29)
Figure 4.17
030 040 0-50
Normalised Strouhal number as function of 6.U/U bed proximity effect removed
These results in fact confirm the intuitive expectation that when a cylinder is placed in a
sheared flow with a differential in approach velocity across its diameter and remote from
boundary influences, the frequency of vortex shedding increases relative to the frequency in a
uniform approach flow with the same cylinder centre line velocity. The Strouhal number
defined using the centre line velocity also therefore increases. In view of the close constancy
of the Strouhal number at approximately 0.2 over a wide range of sub-critical Reynolds
numbers, the question arises as to whether by selecting a defining approach velocity at
some position across the cylinder diameter other than at the centre line position, the
constancy of the Strouhal number can be maintained. This concept has attractions in the
context of the present overall study objectives since in order to be effectively applied in
practice the proposed guidelines for assessing pipeline span stability need'to be not only
reliably founded but also simply stated and uncomplicated to use.
Examining the 'corrected' Strouhal number data in Figure 4.17, it is apparent that the
average increment in SI/S0 defined by the central point in the data scatter range, is of the
order half the corresponding magnitude of .1UIU. For example, the average increments in
S,ISo are of approximately 0.1 and 0.15 for DoUfU values of 0.2 and 0.3 respectively. This
suggests that by defining the Strouhal number relative to the velocity at the top of the
model pipeline cylinder, which has a value given approximately by U + !'J.UI2, the Strouhal
number magnitude would be roughly restored to its value in a uniform approach flow
without velocity gradients.
Following this line of reasoning, the Strouhal numbers were recalculated using velocities at
the height of the top of model pipeline cylinder. Strouhal numbers calculated using this
method were defined by the symbol S,. The resulting normalised values of S,/So, were
plotted and trend lines fitted as shown in Figure 4.1. As can be seen, the curves fitted to the
smooth and rough bed data are now much closer together indicating that a large proportion
of DoUfU increment in Strouhal number has been removed by this simple re-definition of
approach velocity which is equally unambiguous from the point of view of practical
application. The residual increments in the Strouhal numbers indicated by the curves
presented in Figure 4.18 are due in large part to the bed proximity increments as is
demonstrated by the close agreement with the curve summarising the uniform flow results.
43
Contents

",0
,
N
'"
1-00
- TESTS U80, U\60
---- TESTS 560,5160
--- TESTS R80,R160

0-00 0-50 100 1'50 200 2,50
GID
RESULTS FROM: SPECIAL STUDY (REF 17)
Figure 4.18
Normalised Strouhal number, based on velocity at top of cylinder. as function of
G/D
4.2.6 Approach Flow Turbulence Intensity
Turbulence in an approach flow boundary layer can impose both indirect and direct
influences upon the process of vortex shedding from a pipeline immersed in the layer. The
indirect influence results from the role played by the turbulence in a determining the
velocity distribution within the boundary layer, whereas the direct influence is imposed by
the physical effect of the turbulent velocity fluctuations upon the process of vortex
formation and shedding;
Turbulence in a boundary layer results in the random transport of fluid particles within the
layer itself. The momentum possessed by these particles is thus subject to continuous re-
distribution within the layer, and this rate of change of momentum generates sbear stresses
within the fluid. It is these shear stresses which determine the velocity distribution within
the fluid, ie., the shape of the velocity profile. The effect on vortex shedding of approach
flow velocity gradients due to the presence of turbulent seabed boundary layers is addressed
in detail in Section 4.2.5, and methods of determining the velocity distribution within these
boundary layers arc discussed in Section 4.3. Since tbe indirect effect of turbulence is thus
dealt with elsewbere in this document, it is not addressed further in this Section.
The direct effects of turbulence upon the processes of vortex formation and sbedding are not
well documented, particularly for the case of the very large scale turbulence whicb may be
present in a very thick ocean boundary layer. Surry (ref. 59) and Bruun and Davies (ref. 60)
bave carried out investigations of the effect of turbulence on the flow around circular
cylinders, but their work was limited to relatively small scale turbulence (longitudinal scale
= 0.36 - 4.4 D; ref. 59, 0.19 -0.55 D, ref. 60).
For an oceanic boundary layer with a thickness of say 20m, the longitudinal scale of the
turbulence, whicb in this case would be of the sarne order of size as the boundary layer
thickness, would be of the order of 40D for a 20-inch pipeline. Assuming a steady current
velocity of 0.5 mIs, a turbulent structure of this size would take approximately 40 seconds
to pass a pipeline span. Experimental field measurements by Gordon and Witting (ref. 61)
have demonstrated that such large scale turbulence structures do in fact exist in real
situations.
TIle full implications of approach flow turbulence remain under investigation.
44
Contents
4.2.7 Span Length/Diameter Ratio
The effect of the ratio of the span length, 1, to the cylinder diameter, 0, is mentioned by
King (ref. 8). He presents data showing that for small lID ratios, the Strouhal number may
be influenced by three-dimensional effects induced at the cylinder ends. Results from the
work of Gouda (ref. 30) suggests lID> 45 for freedom from end effects, whilst Bernard (ref.
31) defined the limit as lID> 27.
It is not clear how severe the end effects would be "for the case of continuous pipeline lifting
off from seabed to form a span. They would probably be less severe than for the free-ended
, cylinders considered in the experimental studies mentioned above. In any event, if a pipeline
span is long enough to present the risk of vortex induced vibration, then it will generally
exceed the typical limiting 110 values given above. The 110 ratio is not therefore
considered to be a significant parameter for practical span assessment purposes and is not
addressed further.
4.3 APPROACH FLOW VELOCITY
4.3.1 Definition
The design flow velocity, U. appropriate for the calculation of the pipeline span vortex
shedding frequency, f" is the approach flow velocity at a height above the seabed level with
the top of the pipeline.
U is comprised of two components Us and Uw
where:
Us steady state current
Uw ;;;; wave induced current
4.3.2 Steady State Current
Steady state submarine currents can result from tidal flows, wind-induced effects and other
similar phenomena. Long term extreme current values can be predicted by statistical
analysis of measured current records from current meters situated in the area of the pipeline
span.
In-situ current meters can measure the current velocity at a fixed reference height above the
seabed. For design purposes it is necessary to utilise this reference data to calculate the
current velocity at the height of the top of the pipeline. Two alternative methods of
perfonning the latter calculation are outlined below. .
a) One seventh power law approximation
The following expression is used to define the velocity profile:
Eqn4.2
where:
v, = known velocity at known height, y, above seabed
U = velocity at given height y above seabed
This method possesses fue advantage that the entire velocity profile can be determined from
a single velocity measurement at a known reference height above seabed.
Power law approximations of this type are essentially empirical in nature, with no
rigourous physical basis. The '/,th power law is however widely used in engineering
practice, and generally yields results which are slightly conservative. When compared with
the uncertainties which are generally inherent in environmental design velocity data, the
errors produced by this simple approximation can be oflow significance:
45
Contents
b) Logarithmic velocity profile law
The velocity at a given height above the seabed can be obtained using the following
expression:
where:
U,
=
yo =
U =
U=U
,
In (Y/Yo)
In (y'/Yo)
known velocity at known height y, above the seabed
rouglmess length scale (describes seabed roughness condition)
velocity at given height y above seabed
Eqn 4.3
Equation 4.3 is derived from accepted turbulent boundary layer theory. which is a complex
subject containing many areas where existing knowledge remains inadequate. No attempt is
therefore made in this document to address the subject in detail. A number of standard tests
are available however, including those by Bradshaw (ref. 53), Reynolds (ref. 54) and
Townsend (ref. 55). .
This method has the advantage of being based upon a more realistic physical reasoning
process than the power law approximation method. The method can only be used however,
if a realistic estimate of the roughness length scale, Yo, can be made from available seabed
data. It should also be noted that the law only strictly applies to a comparatively narrow
region (approximately 10-20 percent of total boundary layer thickness) adjacent to the
seabed. Outside this region it provides a reasonable approximation but is not exact.
4.3.3 Comparison of Methods.
The major difference between the two methods is that the logarithmic law is able to take
account of seabed roughness conditions, whereas the power law approximation ignores this
factor and gives a generalised profile. Comparisons between profiles generated by the two
methods show that for a moderate seabed roughness, defined by a moderate value of yo. the
differences are small (ie., less than 5 percent). As yo increases however, reflecting
increasingly severe seabed roughness conditions, then the logarithmic law profile begins to
predict significantly lower near bed velocities than those predicted by the power law
approximation.
This variation can be simply explained by considering the physical influence of seabed
roughness upon the flow. Roughness has the effect of retarding the flow in the near bed
region due to the form drag effect of the roughness elements projecting into the flow.
Increasingly severe seabed roughness will therefore generally impose a correspondingly
increased retarding effect upon the near bed flow. A given current flowing over a rough
seabed can therefore be expected to give rise to lower velocities in the near bed region than
the same current flowing over a smoother seabed. The energetic turbulence generated by the
effects of seabed roughness will also have the effect of transmitting the retarding influence
further out into the flow due to the action of turbulent shear stresses generated by the
interchange of momentum between adjacent layers of the flow.
Some practical implications of seabed roughness for the prediction of near bed velocities are
shown in Figure 4.19. Environmental current data is, typically defined in terms of an
extreme current value, VJ, at a given height y" above the seabed, as illustrated in the
figure. Also shown are typical velocity profiles, corresponding respectively to high and
moderate seabed roughness conditions. Both profiles give the same velocity. VI. at the
reference point situated at the known height, YI, above the seabed but exhibit significant
differences at other points.
The profile corresponding to the moderate seabed roughness condition can be reasonably
predicted by inputting the appropriate value of Yo, to the logarithmic law expression.
Comparisons have shown however, that the simpler Ihth power law approximation will
also provide a quite reasonable prediction, giving errors that are small compared to the
inherent uncertainties typically present in environmental design data.
46
Contents
Y
,
,
,
,
,
,
,
,
I
I
I
I
I
I
I
I
I
UI
Y,
_ ....
I
I
I
HIGH SEABED /
ROUGHNESS
",,""
",,""
",,""
////
,,/
Figure 4.19
POINT
MODERATE
SEABED
ROUGHNESS
Typical boundary layer velocity profiles
U
For the high seabed roughness condition, the profile can again be reasonably predicted by
using;'the logarithmic law with an appropriately high value of Yo. The lhth power law
however, has no means of accounting for roughness effects, MId will therefore provide the
same generalised profile based solely upon UI and YI. It can therefore be seen that for a
general point situated between the seabed and the reference point (ie., y<yl), the Ihth power
law will provide a conservative over-estimate of the current velocity, since the retarding
effect of high bed roughness is not taken into account. Above the reference point however
(ie. Y>YI), the Ihth power law will begin to under-predict the velocity and hence
progressively become under-conservative.
4.3.4 Wave Induced Currents
Close to the seabed, where pipeline spans are located, the elliptical water particle orbits
induced by surface wave action beconte extrentely flat. The resulting fluid flow thus
approximates to so-called plane oscillatory motion. Because the amplitude of the wave
induced motion at the bed is relatively small, the layer of retarded fluid in the associated
wave boundary layer is correspondingly thin. An estimate of the amplitude of the
fluctuating wave induced current velocity at the level of the top of the pipeline span, can
thus be obtained using a standard wave theory appropriate to the local water depth and wave
characteristics. No correction need normally be made for velocity attenuation in the thin
wave boundary layer.
Owing to the multi-directional nature of the majority of sea wave states and the fact that the
waves cover a spectrum of frequencies and heights, the wave induced velocity field round
pipeline spans will be highly complex in practice having a random two dimensional form.
As in the case of the steady current component it is nonnally assumed with some back-up
experimental evidence, that the wOIst and hence conservative case occurs when the selected
design wave induces oscillating currents perpendicular to the pipeline axis. A degree of
judgement and inevitable uncertainty arises concerning the choice of an appropriate design
wave for span assessment. If long tenn avoidance of any span vibration is the criterion then
47
Contents
the extreme design wave height for the locality in question should be used. If on the other
hand as is commonly the case, some estimate is required of the likelihood and extent of
span vibration over a relatively shorrinterval of time prior to remedial action being taken,
then a more modest wave such as the significant wave or a wave reflecting the short return
period and the seasonal characteristics is probably more appropriate.
In general, at a particular span site the prevailing bed currents will be a combination of
both steady state and wave induced currents. Unfortunately, the superposition of wave
induced flow and pressure fields into a steady tidal boundary layer flow. produces strong
non-linear interaction effects particularly on the near bed turbulence generation and
structure. This affects the boundary layer velocity profile and introduces deviations in the
combined unsteady flow velocities actually occurring in the relevant near bed zone,
compared with the combined velocities conventionally predicted by simple linear addition of
the two velocity components. .
Recent experimental laboratory studies such as that reported for example by Kemp and
Simmons (ref. 74), have begun to address the extremely complex problem. However,
fundamental difficulties associated with correctly modelling the relevant length and velocity
scales hamper direct practical application of laboratory test results. Theoretical models
proposed for example by Grant and Madsen (ref. 75) and by Christoffersen (ref. 72) are thus
of limited use for design prediction in the absence of good quality field data for calibration
purposes. There is thus a considerable need to carry out full scale field tests to obtain direct
measurements of combined wave and tidal induced velocities in the near bed zone. Given the
lack of such data and in view of the uncertainties in using laboratory test results and
available theoretical techniques which attempt to model the formidably complex fluid
interaction processes, it is doubtful that the conventional method of vectorially adding the
two velocity components can be significantly improved upon at the present time. This
conclusion is further supported when one considers the substantial uncertainties in
establishing relevant local environmental conditions.
Having accepted this method of estimating the unsteady current velocity it is normally
further assumed, when assessing fluid lift and drag loading for example, that the worst case
situation once again arises when the wave and tidal current directions coincide. The total
instantaneous velocity componentU(t) can then be expressed in the form U(t) ; U, + Uw
sin 21ttff where T is the wave period and U, is the steady current velocity at the height of
. the top of the pipeline. The maximum wave induced velocity, U., equals u. 21tff where a
is the amplitude of the wave induced fluid particle motion close to the bed. It must once
again be emphasised that the actual prevailing combined velocity field in the near bed
region is exceedingly complex and that the above treatment is a necessary but highly
idealised description. In the present context of vortex induced span vibration, two further
key questions arise. Firstly, how does the unsteady wave flow affect vortex shedding.
Secondly, how does a particular span respond to the resulting unsteady vortex shedding.
behaviour with its transitory pattern of force-excitation. An attempt is made to address these
very difficult questions in the following Sections 4.4 and 4.5 and also in Section 5. As
pointed out in these discussions, analysis at the present time must remain somewhat
speculative, primarily because of the current lack of relevant experimental data from
laboratory and in particular field tests. As envisioned at the outset therefore, adequate
treatment of these questions lies eventually beyond the scope of the present study ..
4.4 VORTEX SHEDDING IN OSCILLATORY WAVE FLOW
A brief review and analysis of information relating to vortex shedding in wave and
combined wave/steady current flow was carried out by Grass (ref. 77) as part of the
background information for the present investigation. The description set out in this section
and also Section 4.5 outlines the main findings from this ancillary study.
48
Contents
Sarpkaya (78) carried out a comprehensive series of tests using both smooth and rough
surfaced cylinders at practical scale Reynolds numbers in the type of plane oscillatory flow
produced by waves close to the bed. The results of these experiments clearly demonstrate
that many of the manifestations of the laminar to turbulent flow transition in the cylinder
boundary layer present in steady flow discussed in Section 4.2, are also observed in the
more complex oscillatory flow. This is panicularly reflected in the behaviour of the in-line
drag coefficient which exhibits the same sharp drop in magnitude in a critical Reynolds
number range linked to the boundary layer transition in the reversing flow. The transition
takes place at lower Reynolds numbers than in steady flow however, due to the highly
turbulent wake being washed back past the cylinder in the return half of the flow cycle.
Sarpkaya's measurements show that in addition to the flow Reynolds number, Re, defined
by UwD/u where U
w
is the maximum oscillatory velocity, the flow characteristics and the
associated drag, lift and vortex shedding behaviour are also highly dependent on the
Keulegan Carpenter number, K. This dimensionless variable defined by 21taiD, where a is
the distance amplitude of fluid panicle oscillatory motion, it is simply a measure of the
relative distance moved by the fluid back and forwards past the pipeline cylinder, diameter
D.
The characteristics of plane oscillatory flow are severely complicated by the fact that the
near cylinder flow field has to "re-build' itself twice per complete cycle as the velocity
reverses direction and increases in magnitude from two zero velocity points. This cyclically
repeating acceleration of the fluid motion from rest generates a rapid accumulation of
vorticity in the initial vortices formed downstream of the cylinder which thus grow more
rapidly and become somewhat stronger than the vortices fonned under equilibrium steady
flow conditions. The development of these relatively strong and usually asymmetric
vortices and their subsequent action in producing strongly asymmetric flow as they are
swept back past the cylinder in the return half of the flow cycle, is largely responsible for
the high values of drag and lift coefficient which are a characteristic feature of plane
oscillatory flow.
These effects a r ~ most pronounced for low Keulegan Carpenter numbers, K, less than
approximately 25. Because of the relatively small distances moved by the fluid panicles
past the cylinder between the flow reversal phases at these low K values, typically only one
or two vortices have time to form and be shed in each half cycle. In the idealised, exactly
repeating plane oscillatory flows used in laboratory experiments such as Sarpkaya's, it has
been found that for K values in the approximate range 5 to 15, the vortex generation,
shedding and pairing patterns lock-in to the primary flow oscillation cycle for considerable
~ r i o d s of time. During these locked-in phases a single vortex is shed from the cylinder in
each half cycle producing a transverse lift force which fluctuates at a frequency, f" equal to
twice the flow oscillation frequency, fw, ie., fjfw = f. = 2.
For Keulegan Carpenter numbers in the approximate range 15 to 25 a second mode of
locked-in vortex shedding, pairing and cyclical motion occurs. Two vortices are now shed
from the cylinder during each half cycle which pair and are swept back past alternating sides
of the cylinder in sequential half cycles. This generates a fluctuating transverse lift force
with a dominant frequency of three times the primary flow frequency, ie., f.= 3.
The detailed mechanics of these two locked-in modes of vortex shedding which dominate
plane oscillatory flow regimes around cylinders for K values less than 25 and give rise to
very high transverse lift coefficients have been studied by Grass et aI (ref. 79). A diagram
illustrating the formation and migration patterns of the dominant vortices corresponding to
the two locked-in shedding modes, based on flow visualisation observations made in the
latter investigation, is reproduced in Figure 4.20.
For Keulegan Carpenter numbers greater than approximately 25, there is an increasing
randomisation and decoupling of the interaction between the shed wake vortices and the
newly forming vortices in the return half flow cycle. This results in a rapid and progressive
decrease in the lift coefficient which eventually approaches the steady current value at large
K values greater than say 100. As the Keulegan Carpenter number increases, so the number
of vortices completely formed and shed in each half cycle of the flow increases until the
wake resembles a steady flow Karman vortex street at large K values.
49
Contents
V>
o
<

;;


::r

0.
0.

3
o ."
0.-
G'"
C

-.
o
..
.., ..
!"o


2-
Ii'
o
-<
-


I' . I
I
I
. 1
@ 0
1 bi

REVERSAL
II ..
I.

01
REVERSAL
II II
FLOW

2


Fl...OW
2
K 10
FLOW
3
<

8

oW
FLOW
8
8
REVERS.!.:....

4
o


REVERSAL
II ..
4.
._- -------- -------------_._---
DOMINANT LOCKED-IN MODES OF VORTEX SHEDDING
AND C.YCLICAL MOTION FOR
KEUCEGAN-CARPENTER NUMBER (K)
oj K = 10 Reynolds number Re = 2-61 x 10
4
b) Ko 20,Rp370,IO'
Relative surface roughness of cylinder k/D=1/50
FROM GRASS ET Al iREF (9)
Contents
The vortex shedding frequency and hence the number of complete vortices shed in each half
cycle also become progressively decoupled from and independent of the primary flow
oscillation frequency with increasing Keulegan Carpenter number. The number of vortices
completely formed and shed in each half flow cycle, N, is directly related to the frequency of
the fluctuating transverse lift force, f,. Isaacson and Maull (ref. 80) have suggested that f, is
given by the simple relationship f, = (N + 1) fw. This formula is clearly valid for the two
dominant vortex shedding modes discussed above for which N = 1 and 2 and the lift force
frequencies are 2fw and 3fw respectively. However, at higher K values where the number of
vortices shed can vary from one cycle to the next and the instantaneous shedding frequency
fy varies as the velocity varies through the cycle, this simple relationship can only be
expected to yield an approximate estimate of the overall average lift force frequency.
For Keulegan Carpenter numbers greater than 15 the number of vortices shed in each half
cycle increasingly varies as K increases. Nevertheless, in particular K ranges the total
distances the fluid particles move past the cylinder in a particular direction coincide with and
embrace the average distance required to completely form and shed a specific number of
vortices. Within these K ranges it is therefore found that the major contribution to the lift
force occurs at frequencies corresponding to the particular number of vortices naturally shed
in the half cycle. Spectral analysis of the lift force produced in plane oscillatory flow would
thus be expected to show peak contributions at exact multiples (harmonics) of the primary
flow frequency, fw. This is confirmed by the experimental results obtained by Maull and
Milliner (ref. 81). These observations do not however mean for the reasons outlined above,
that the vortex shedding is closely synchronised and coupled with the primary flow
oscillation except for the two dominant locked in shedding modes occurring at K values less
than 25. The time scales of the vortex shedding and the flow oscillation thus remain largely
independent.
The above effects are also well illustrated by Sarpkaya's measurements of the maximum
frequency of the lift force, fy, averaged for a representative sample of flow cycles for a range
of Reynolds numbers and Keulegan Carpenter numbers. The results plotted in the form of
fR = fJfw as a function of K and R, are reproduced in Figure 4.21 from 5arpkaya and
, Isaacson (ref. 78). The interpretation of this diagram is that a line defined for example by
fR = 4 means the transverse lift force generated by the vortex shedding does not in general
contain relative frequencies larger than fR;;;; 4 for K and R values in that region to the left of
the line. It will be noted that the fR values are all integers for the reasons discussed above.
Intermediate values of fR such as f. = 3, 5, 7 etc., have been excluded for simplicity.
If the peak value of f" used to derive the Figure 4.21 correlation, is directly equated to the
peak instantaneous frequency of vortex shedding at the point where the velocity reaches its
maximum value in the cycle, then a vortex shedding Strouhal number can be defined as
5 = tD/Uw = fRIK. Sarpkaya's test results for smooth cylinders show that 5 so defined
remains reasonably constant at 0.22 for fR larger than 3 as indicated in Figure 4.21. For
large transcritical values of Reynolds number, S tends to increase to a value of
approximately 0.3 which is reflected in the curve trends in Figure 4.21 and follows the
steady flow pattern of behaviour discussed in Section Two. The average value of S based on
all the vortices shed during a given cycle (and presumably Uw although this is not stated)
was found by Sarpkaya to be between 0.14 and 0.16.
Sarpkaya's results also showed that the maximum lift coeffiCients for surface roughened
cylinders showed very little dependence on Reynolds number. This can once again be taken
as evidence of the effect of roughness in contracting the width of critical Reynolds number
range and maintaining continuous regular vortex shedding. The Strouhal number S
remained essentially constant at a value of 0.22 over the entire Reynolds number Keulegan
Carpenter number ranges tested.
The important practiCal conclusion to be drawn from Sarpkaya's experimental results is that
for Keulegan Carpenter numbers greater than approximately 20 a good estimate of the
instantaneous vortex shedding frequency. fs, and hence the frequency of the transverse
excitation force of pipeline spans, can be obtained from the relationship f, = 5U(t)1D where
U(t) is the instantaneous velocity. The 5trouhal number, 5, can be taken as constant at
0.22 for the pipeline surface roughness conditions normally encountered in practice.
51
Contents
10"
Re
10
4
52
4
'R= 2
FRACTIONAL
VORTEX
SHEDDING
5 10
~ 1 5
\
\
\
\
10
\
\
8
\
,
\
6
\
\
\
\
,
\
,
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
I
\
I
\
I
I
I
I
I
I
I
I
I
50
KEULEGAN CARPENTER No.
K
\
\
\
\
\
100 200
FROM SARPKAYA AND ISAACSON (REF. 78 1
Figure 4.21
Relative frequency of vortex shedding as a function of K and R.
Contents
The evidence of 'quasi-steady' vortex shedding provided by Sarpkaya's results, where the
shedding frequency closely follows the instantaneous cyclical variations in the unsteady
oscillating current velocity, is further supported by the results presented by Bearman et al
(ref. 82). The latter authors propose a simple model for predicting the wave form and
magnitude of the transverse lift force as it varies through the the half cycles of plane
oscillatory flow. Quasi-steady vortex shedding is a primary assumption on which this
prediction model is based. It was found that the value of Strouhal number substituted to
obtain the best fit with experimental data obtained at suhcritical Reynolds numbers.
remained closely constant at 0.2 for K values in the range 30 to 55. This Strouhal number
constancy confirms that the vortex shedding frequency adjusts quite rapidly to the
fluctuating velocity in conformity with the underlying quasi-steady shedding assumption.
It must be emphasised that the above conclusion regarding quasi-steady vortex shedding is
based on laboratory tests under highly regular oscillatory flow conditions. The complexity
and general randorrmess of near bed wave induced flow fields in nature discussed in Section
4.3, will tend to disrupt and weaken the coherence of vortex shedding. However, there is no
good reason to believe that the vortex shedding frequency will not tend to follow the
instantaneous velocity even in these adverse circumstances.
4_5 VORTEX SHEDDING IN COMBINED WAVE AND TIDAL
FLOW
Prediction of vortex shedding frequencies from pipeline spans situated in combined wave
and tidal flows close to the seabed is fraught with further difficulties and attendant
uncertainties. In addition to the problems of defining an approximate approach flow
velocity discussed in Section 4.3 and also the effects of velocity gradients, Reynolds
number, and surface roughness, the vortex shedding frequency from pipeline spans in such a
combined unsteady flow field will also clearly depend on the ratio of the current to the wave
velocity, UJU. and the Keulegan Carpenter number, K; 21t aID. Very little experimental
work has been carried out in this type of flow particularly relating to vortex shedding
behaviour. Most of the investigations conducted to date have been concerned with the
measurement of in-line drag forces as for example in the case of the studies reported by
Mercier (ref. 83), Sarpkaya (ref. 84) and Verley and Moe (ref. 85). These and other generally
relevant investigations have been briefly reviewed by Sarpkaya and Isaacson (ref. 78). A
study of lift forces generated on a vertical cylinder subjected to combined wave and current
flow has been reported by Every (ref. 86). Whilst lift force measurements are more useful in
attempting to infer vortex shedding characteristics, the relevance of this particular
investigation in this present context is compromised by the depth varying orbital nature of
the wave induced velocity field round the vertical cylinders.
The amount of information relating to vortex shedding behaviour in combined flows to be
gleaned from these laboratory studies is thus extremely limited. Recourse must therefore be
made to physical arguments based on the observed shedding characteristics under the much
more extensively studied separate steady and plane oscillatory conditions discussed in
Sections 4.2 and 4.4 respectively.
Two very different flow regimes can exist round seabed pipeline spans under combined flow
conditions dependent on whether the velocity ratio UJU
w
is greater or less than unity. In
the case when the steady tidal current velocity, U
c
is greater than the maximum wave
induced velocity, Uw, in the bed region, which is very rare in practice except in very deep
water, the combined flow velocity, U(t) as defined in Section 4.3, never reverses and always
stays in the same direction. In these circumstances, in view of the evidence that the vortex
shedding frequency closely follows the instantaneous velocity in the much more disruptive
reversing plane oscillatory flow regime presented in Section Three, it appears reasonable to
assume that the 'quasi-steady' shedding assumption will continue to be valid in this less
severe flow environment.
Typical values of U, and U. are 0.5 mls and 1 to 2 mls respectively. It is thus much more
common for the ratio Dc /Urn to take a value less than unity. In these circumstances the
combined flow velocity reverses direction during a part of the wave cycle which inevitably
interrupts the vortex generation process. Probably the simplest and most physically
relevant way of viewing this complex situation is to consider the net distance and direction
moved by the fluid particles past the pipeline cylinder in the two halves of the wave cycle.
53
Contents
These distances are respectively UcT/22a and can be used to define two loosely 'equivalent'
Keulegan Carpenter numbers K, and K2 for the two halves of the wave cycle given by
K, = 2rr(U,T/4 + a)ID, K, = 2rr(U,T/4 - a)ID, Since as discussed in Section Two, the
number of vortices shed in a particular half cycle of oscillatory flow is dependent' on the
distance measured in cylinder diameters that the water particles travel past the cylinder as
expressed by the K values, it can be seen that the presence of the current will produce
differences in the vortex shedding in the two halves of the flow cycle as indicated by the
differences in Kl and K2.
The evidence presented in Section 4-4 suggests that in the case of plane oscillatory flow,
when the fluid particle movement in half flow cycle exceeds distances defined by K values
in excess of 20 then the vortex shedding frequency follows the variations in instantaneous
flow velocity reasonably closely. Given the dearth of relevant experimental data and in the
absence of contradicting physical reasons. it would again appear reasonable in the present
circumstances to assume that the same criteria for quasi-steady vortex shedding. namely Kl
and K, greater than 20, can be applied in complex combined flow fields for the two halves
of the flow cycle respectively.
For K, and K2 values less than 20 it would appear reasonable to assume. based on the plane
oscillatory flow results, that the transverse excitation force has a maximum frequency of
three times the wave frequency. It must be understood that whilst this three times factor is
likely to be conservative. it is highly speculative in the absence of experimental data.
Experimental evidence discussed by Sarpkaya and Isaacson (ref. 78) indicates that the
presence of very small steady currents significantly reduces the drag forces experienced by
cylinders in oscillatory flow for K values less than 20. Flow visualisation studies carried
out by Grass, suggest that this is due to the disruption of the dominant lock-in modes of
vortex shedding at low K values associated with the K" K2 bias produced by the steady
current. The high liff forces generated by these locked-in modes would therefore also be
significantly reduced.
In the case of quasi-steady vortex shedding, the instantaneous shedding frequency, fs. can be
estimated from the instantaneous combined velocity from f, = SU(t)lD. Once again in the
absence of direct experimental measurements of S in combined flow fields, it would appear
reasonable to substitute the values of S suggested by Sarpkaya, based on measurements in
plane oscillatory flow, quoted in Section 4.4. into the above relationship for f,.
In conclusion therefore. under the environmental design conditions normally relevant to
pipeline span assessment the values of effective Keulegan Carpenter number, K, and K,
relating to sequential half cycles of the design wave will frequently exceed 20. In these
circumstances the evidence presented above suggests that quasi-steady instantaneous vortex
shedding will take place from a pipeline span with the shedding frequency closely following
the fluctuating instantaneous combined flow velocity. When conditions are such that
insantaneous vortex shedding occurs in both halves of the wave cycle, the effect of the
steady mean tidal current velocity is to cause a shift in the range of shedding frequencies
between the two halves of the wave cycle.
Once again it must be understood that the above conclusions regarding vortex behaviour in
combined flow fields are based on limited observations from idealised laboratory tests.
Whilst it appears likely that this pattern of behaviour will be at least qualitatively followed
in its more complex natural sea environment, some uncertainty must remain in the absence
of relevant field data.
54
Contents
/
5. SPAN RESPONSE - PARAMETRIC STUDY
5.1 DEFINITION OF PARAMETERS
The response amplitude, a, of a submerged cylinder subject to vortex induced vibration may
he functionally. expressed as follows (ref. 13):
a = f (fN, m, D,Ii., U, p) Eqn 5.1
where:
a
=
vibration amplitude;
fN
=
natural vibration frequency of cylinder in still water;
m
=
mass per unit length (including added mass);
D
=
cylinder diameter;
Ii. =
logarithmic decrement of structural damping;
U
=
approach flow velocity;
p
=
mass density of water.
Note that the above expression assumes the absence of Reynolds number and three-
dimensional effects.
By dimensional analysis, the fOllowing groups are obtained (ref. 13):
aID = f, (UlfND, mJpD',li,) Eqn 5.2
where the symbols are as defined for Equation 5.1, and the groups are described as follows:
aID
=
.relative amplitude
UlfND reduced velocity
/pD'
=
mass ratio
Ii,
=
logarithmic decrement of structural damping
The latter two terms can be grouped together to form a term known as the stability
parameter, as discussed later in this section.
aID,
mJpD'
the relative amplitude, simply expresses the amplitude of the vibration in terms
of the cylinder diameter.
the reduced velocity, has been described in some detail in Section 3, and is shown
to be an effective means of comparing the excitation and response frequencies of
a system.
the mass ratio, defines the relative inertia of the cylinder compared to the fluid. A
low mass ratio, in which the inertia of the fluid is great compared to that of the
cylinder, will provide the most favourable conditions for the occurrence of flow
induced vibration. Conversely, if the mass ratio is high, the energy content of
the flow may be insufficient to cause vibration, despite other conditions being
favourable for its onset.
the logarithmic decrement of structural damping, is defined by the following
expression:
where:
'0-\
Ii, = log, (-)
'0
a" . "a" = amplitudes of successive cycles of free, natural
frequency, vibration of the cylinder.
Eqn5.3
Figure 5.1 graphically defines Equation 5.3, for a typical structure undergoing freely
decaying natural frequency vibration.
55
Contents
B, represents a simplified method of defining the structural damping of a system. It is
derived from the more rigourous approach given by the following expression.
where:
a..1 = I (1 _
a"
'n-I, 'n are as defined for Equation 5.3

energy dissipated per cycle
41t x total energy of system
is generally termed the 'damping factor' or 'damping ratio'.
Eqn 5.4
If however, is small, then (I - I, hence Equation 5.4 approximates to the
following form:
In this situation, 0, can thus be seen to be equal to
AMPLlTUOE I 0
+

TIME,f
Figure 5.1
Logarithmic decrement definition sketch
From a practical point of view, damping can perhaps be most usefully thought of as a
mechanism for the dissipation of energy. Vickery and Watkins (ref. 64) utilised this concept
to define the energy balance of a cylinder undergoing resonant cross-flow vibration. They
equated the energy absorbed by the cylinder from the flow to the energy dissipated by
structural damping, by utilising the following argument.
Assume that the flow-induced excitation force is sinusoidal and may be expressed in the
form:
P = Po sin 21t fNt Eqn5.5
56
Contents
where:
P
=
instantaneous excitation force per unit length, acting at a given section, at
time t;
Po
fN
=
=
max. excitation force per unit length, acting at the given section;
natural vibration frequency of cylinder.
Po can be expressed as follows:
where:
,
CL =
. fluctuating lift coefficient;
p =
water density;
U
=
flow velocity;
D
=
cylinder diameter.
Eqn 5.6
The energy absorbed from the excitational forces along the cylinder is now equated to that
lost through structural damping. Vickery and Watkins (ref. 64) developed the following
expression where the left hand term is the net energy absorbed by the cylinder from the flow
induced excitation forces, and the right hand term is the energy loss due to structural
damping. Hydrodynamic damping effects are incorporated effectively into the fluctuating lift
coefficient and hence are not explicitly shown. Further discussion of damping effects is
given in Section 5.2.3.
Eqn5.7
where:
x = axial distance along cylinder:
a = transverse displacement of cylinder at a given section;
L cylinder length;
0, = logarithmic decrement of structural damping;
m ;;;; mass per unit length at a given section.
All other symbols are as defined above.
From this, the following expression is deduced:
Eqn 5.8
where:
a. = amplitude at some significant section of the cylinder (say mid-span point);
IIlo = equivalent mass per unit length, which for a uniform structure such as a
pipeline is simply equal to m.
All other symbols are as defined above.
The groups LID and ala. respectively define the fixed dimensions and deflected shape of the
cylinder. The remaining groups are the Reynolds number, the reduced velocity, which is
extensively dealt with earlier in this document, and a new group defined as the 'stability
. parameter' , 1<,.
wbere:
k,
2m Os
pO'
Eqn 5.9
57
Contents
It can be seen that k, is essentially the product of the mass ratio and the logarithmic
decrement of structural damping. Vickery and Watkins (ref. 64) substantiated their
theoretical deductions of the importance of k., by performing experiments to show that the
relative amplitude of vibration is indeed a function of ks itself rather than of the individual
parameter 0, and m1pD'.
Vickery and Watkins' analysis (ref. 64) refers specifically to response in the
direction, and states that an equivalent result may be obtained for the in-line case by simply
substituting the fluctuating drag coefficient CD for the coefficient of fluctuating lift. King
(ref. 68) however, adopts a similar in principle, but more detailed approach, by
incorporating the variations in relati,:e flow velocity induced by in-line cylinder vibration.
Under in-line response, energy is only input to the cylinder during the half of the vibration
cycle in which the direction of cylinder vibration is the same as the direction of the fluid
flow. During the remaining half of the vibration cycle. in which the cylinder is moving
against the direction of flow, work must be done by the cylinder against the drag force, and
energy is consequently lost. King (ref. 68) includes this effect in the damping side of the
equation, and obtains a result similar to that of equation, 5.8, except that the in-line
response amplitude is found to possess an additional functional dependence on the steady
drag coefficient, CD. King (ref. 68) argues however, that for a cylinder free to oscillate, CD
remains nominally constant for a wide range of experimental Reynolds numbers and does
not therefore impose a significant influence upon the response amplitude.
The theoretical relevance of k, as a parameter determining the response amplitude of
oscillating cylinders, is extensively confirmed by experimentation. Figure 5.2 shows a
diagram presented by King (ref. 68) plotting relative cross-flow amplitude at the point of
maximum deflection against k.. The referenced data sources shown on the figure include
experiments carried out in both air and water, and the various results are seen to correlate
extremely well. This diagram is in fact quoted directly by some existing design rules and
guidelines (eg., Det norske Veritas, ref. 16 and CIRIA, ref. 66), together with a similar
diagram relating to in-line response which is shown in Figure 5.3.
In addition to the observed gocxl correlation between different laboratory test results provided
by the above relationships, it was shown by King (ref. 72) that reasonable quantitative
agreement with the results offull scale tests on marine piles (ref.67) can also be obtained.
Caution must be exercised however, when directly applying the results to full scale pipeline
spans, since full scale tests, including the special study (ref. 62) and the work of Bruschi et
al (ref. 22) have indicated differences in response behaviour. In particular, both the latter test
programs indicated the possibility of significantly increased in-line vibration amplitudes,
which is a non-conservative effect from the point of view of maintaining the structural
integrity of a pipeline span.
It can be concluded from the above brief discussion, that the stability parameter, k
s
, with its
inherent incorporation of the effects of damping, is a potentially important parameter for
the determination of span response. The stability parameter is therefore discussed in greater
detail below. along with additional physical parameters which may influence the behaviour
of submarine pipeline spans.
58
Contents
a (max)
D
2'5
20
15
~ ,
\
. \
\ x
: ~
\
10
x,
05
I
i
!
'x
, .
'"
,
i
I
I
I
I
,
,
I I i
I
!
,
I
I
I
I
i
i
i
,
I
,
,
I
I
i
,
.... I .
'1"',

: ......... _1 -. x
..
r + - - " ~
4---< ----
---- !-----!,-...;--
I
0 2 4 6 8 10 12
ks
:
2 m.O
s
,.
j>D2
OJ lOX KING (IN WATER) diD: 20-33 (REF68)
VICKERY AND WATKINS (IN WATER) LID: 15 (REF 64)
VICKERY AND WATKINS (IN AIR) LID: 142 (REF. 64)
+ FENG (IN AIR) L/D:9 (REF 69)
" HARTLEN(IN AIR) LID: 138 (REF. 70)
v SCRUTON (INAIR)L/D: 275(REF.71)
L = CYLINDER LENGTH
D : CYLINDER DIAMETER
d : WATER DEPTH
Figure 5.2
14 16
FROM KING (REF 68)
Cross-flow response amplitude as a function of ks
,
,
I
18
59
Contents
o 20 ,.----------,-----------,-----------r----------,
a (max)
D
FIRST INSTABILITY REGION
0- I 5
IN - LINE MOTION
o -I 0 -------+----------1--------------1
SECOND INSTABILITY
REGION
0-05
o 0-5 1-0 1-5 20
FROM DnV AND CIRIA (REFS_ 16 AND 66)
Figure 5.3
In line response amplitude as a function of k.
60
Contents
5.2 STABILITY PARAMETER
5.2.1 Components of the Stability Parameter
The stability parameter definition incorporates a number of individual component
paramett::rs as shown below:
where:
k, =
m, =
8, =
p =
D =
stability parameter;
,,_ 2 I140,
"'- po'
effective mass per unit length;
logarithmic decrement of structural damping;
density of surrounding fluid;
cylinder diameter.
Eqn 5.10
The bottom line of the defining equation generally presents few problems when applied to a
submarine pipeline span, since p, the density of the surrounding seawater and D, diameter
of the pipeline with its external coatings are known.
Of the remaining component parameters, the effective mass per unit length is generally
well defined for a fully submerged pipeline, although complicating circumstances may
arise, which are discussed in more detail in Section 5.2.2.
Finally, the damping component requires careful definition, and this is addressed in detail in
Section 5.2.3.
5.2 .2 Effective Mass, m.
In its most general form, the effective mass per unit length of a structure is defined by the
following equation (see King, ref. 8):
oJLm y' dx
=
oJd y' dx
Eqn 5.11
where:
II4 = effective mass per unit length;
= mass per unit length at a given section;
y = deflection of structure at distance x from origin;
L = total length of structure;
d = submerged length of structure.
This generalised definition allows the effects of variations in mass along the structure to be
incorporated, as well as allowing the analysis of surface piercing vertical structures. For a
typical submarine pipeline span, the mass per unit length will be uniform, and the span
will be totally submerged ie., d = L, giving m, simply equal to m. Circumstances may
arise however, where the presence of installations such as anodes, buckle arrestors, repair
sleeves etc., may lead to potentially significant variations in m along the span length, and
equation 5.11 allows such effects to be incorporated into the analysis.
At any given section of a pipeline, the mass per unit length itself comprises up to three
individual components of mass as follows:
m=t11m+m,+m. Eqn 5.12
61
Contents
where:
m
lllm
m.,
m.
=
=
=
total mass per unit length;
mass of pipeline material (steel + all coatings);
mass of pipeline contents (if any);
added mass.
The material and contents masses will generally present few problems. The added mass
however, may be open to some uncertainty, and is consequently discussed further below.
In general terms, the added mass is defined as follows:
where:
m.
Cm
lllm",
=
=
rna = em x Illdisp
added mass per unit length;
coefficient of added mass;
mass of fluid displaced by pipeline per unit length.
Eqn 5.13
The physical significance of added mass is that when a body accelerates through a fluid,
energy is needed to accelerate not only the mass of the body itself, but also the mass of a
certain amount of the fluid which is entrained by the accelerating body. For a circular
cylinder the added mass is conventionally taken as equal to the mass of fluid displaced by
the cylinder, ie., the added mass coefficient, Cm, is equal to unity. The presence of a solid
boundary near to the cylinder is known to increase the added mass, and for the case of a
circular cylinder of diameter D, situated at a distance G from a plane boundary, em is
commonly expressed as a function of GID, as shown in Figure 5.4 (see ref. 16).
As always, when dealing with full scale applications, care must be taken in applying
'conventional' results, which tend to be derived from laboratory scale testing.
5.2.3 Damping
Of all the individual component parameters which make up the stability parameter, the
damping is most difficult to quantify. Existing span assessment methods tend to
incorporate highly generalised assumptions regarding pipeline span damping, based on
existing knowledge relating generally to structural steelwork. In view of the many variables
and uncertainties involved, it is inevitable that generalisations and assumptions will at
present remain. The basic concepts involved however, are discussed in some detail below.
Damping is essentially a process of energy dissipation. For a structure oscillating in a
fluid, three basic mechanisms exist whereby energy dissipation takes place. These are
summarised as follows:
Material damping;
Structural damping;
Fluid damping.
Material damping is present in any real material. It results from the fact that no material is
in practice perfectly elastic, and consequently the energy required to deform a material is
never recovered fully when the material-elastically recovers. It is sometimes referred to as
hysteretic damping, since the stress-strain curve follows different paths during the respective
deformation and recovery phases, thus describing a hysteresis loop. The energy lost during
the complete oscillation cycle is then proportional to the area within the hysteresis loop.
62
Contents
3-0
I
I
I
I
2-29 - !
j
I-
2-0
I
i
I
I
I
i
,
I
I
1-0
I
i
i
:
I
I
,
I
I
I
,
, o
o 0-5 1-0
6
JG
7 / / / / / / 7 7 ~ T / 7
i
,
!
I
I
,
I
I- 5
G
D
2-0
Figure 5.4
Added mass coefficient
I
I
I
I
25 3-0
FROM DnV (REF_ 16)
35
63
Contents
In physical terms, the energy loss takes place at molecular level, as the molecular structure
of the material undergoes slight irreversible rearrangement in response to the deformation.
The 'lost' energy is actually converted to heat by the frictional action of molecules shearing
past each other during the rearrangement process. In practice, common structural materials
such as steel and concrete possess low damping factors as long as they remain within their
elastic limits. Excessive strain of the materials however, leading to irreversible effects such
as plastic flow in steel, results in a significant increase in material damping. It therefore
follows that materials such as lead, which undergo plastic flow under comparatively small
loads, possess high material daroping factors.
The second type of damping listed above is structural damping. In this case, the energy is
again dissipated as frictional heat, due to different structural elements working against each
other. In the case of structural steelwork, structural damping takes place due to scraping and
impacting at joints and at any other location where different structural members are in
contact and able to move relative to each other. For a subsea pipeline span, there is little
scope for this type of daroping within the continuous welded steel pipeline itself, however
friction against the seabed at the touch down points will provide a means for energy
dissipation.
The third type of damping is fluid damping, resulting from the work done by a structure
oscillating in a fluid to overcome the fluid drag forces. In water, the fluid (hydrodynamic)
drag forces may be quite significant, however, their incorporation into an analytical
approach requires careful interpretation. For the simple case of a body oscillating freely in a
stationary fluid, hydrodynamic damping is a comparatively simple concept, resulting in a
steady decay of the oscillation amplitude as the body does work to overcome the fluid drag
force. Simple experiments to quantify the logarithmic decrement may be performed by
releasing the body from an initial deflection and recording the rate of amplitude decay of the
resulting oscillations."
The problem becomes much less well defined however, in the more general case of a body
oscillating in a moving fluid. As described in earlier sections of this document, a fluid
flowing past a body is able to exert fluctuating forces upon the body, giving the potential
for the transfer of energy from the fluid to the body. This process is fluid excitation, which
is the reverse of fluid damping. In this case therefore, the oscillating body may
simultaneously experience the superimposed effects of fluid excitation due to the motion of
the fluid past the body and fluid damping due to the motion of the body through the fluid,
with the perceived behaviour of the body reflecting only the net energy transfer.
In view of this uncertainty, the energy balance argument formulated by Vickery and
Watkins (ref. 64) and discussed in Section 5.1, made no attempt to incorporate the
hydrodynamic daroping, but considered only the net hydrodynamic excitation effect, and the
structural (and material) damping effects at resonance. All hydrodynamic damping effects are
effectively reflected in the net value of the fluctuating lift coefficient, CL'.
Hydrodynamic damping cannot however be totally ignored, as demonstrated by King (ref.
68), who extended the energy balance argument of Vickery and Watkins (ref. 64) to describe
the varying relative motion of a cylinder oscillating in the in-line direction. This argument
is discussed in Section 5.1, and essentially divides the in-line response cycle into two
halves, depending upon whether the direction of body motion is with or against the mean
flow direction. The two halves of the in-line oscillation cycle and the relationship to the net
drag force are illustrated in Figure 5.5.
During the half of the oscillation cycle in which the body is moving with the flow, the net
drag force acts to constructively reinforce the cylinder motion. When the body reverses
however, and moves against the flow direction, the body must do work to overcome the net
drag force and consequently loses energy. In other words, it undergoes hydrodynamic
daroping, the strength of which is related to the net drag force. This latter effect is expressed
by King (ref. 68) as a partial functional dependence of the in-line oscillation amplitude
upon the steady drag coefficient, CD.
Figure 5.6 (taken for Roshko, ref. 73), shows CD as a function of Reynolds number Re,
and illustrates the significant variations in CD which occur in the critical Reynolds number
region between Re = 10
5
and Re.= 10
6
for a smooth stationary cylinder. This shows that the
potential at least exists for the energy loss due to fluid damping, and hence the in-line
oscillation amplitude, to demonstrate a Reynolds number dependence. It would be a
64
Contents
MEAN FLOW
/BODY MOTION
ENVELOPE
MEAN BODY
POSITION
MOTION

I
I
G
-----
NET DRAG
FORCE
BODY MOVING WITH MEAN FLOW - DRAG FORCE REINFORCES MOTION
MOTION

Gr-,
BODY MOVING AGAINST MEAN FLOW - DRAG FORCE OPPOSES MOTION
Figure 5.5
In-line oscillation body motion and drag force Interaction
65
Contents
3 6
\'
~ / S
,
!
I I
\
i
i
;- \
1\
I ........
-
I
I
i
I
,
I
. \ .. - I
I
I
I
\
,
:1
\:
I
I
, ~ .
,
I
I I.
I j
I
~ '
I : i I'
i i
).-
~
I '
:J
i i j
\
!
I
\
i
''':.:
!


,
2
5
2 4
3 liS
o
!
-t,
I
I
- - - ~
i i I
,
10
NOTE, RECIPROCAL OF STROUHAL NUMBER ALSO SHOWN
REF ROSKHO (REF 73)
Figure 5.6
Steady drag coefficient as a function of Reynolds number
66
Contents
considerable over-simplification however, to directly incorporate CD values from Figure 5.6
into a dynamic energy balance analysis, since the results shown are specifically for
stationary cylinders. As has been discussed above in the context of the lock-in
phenomenon, the cylinder motion itself alters the flow regime around the cylinder, and in
consequence; the pressure distribution around the cylinder and hence the resultant drag forces
may be significantly modified from their stationary values. The effects of cylinder motion
may therefore over-ride the potential Reynolds number effects for the oscillating cylinder
case. A further point to be considered, is that if CD does fall significantly even for an
oscillating cylinder in the critical Reynolds number region, then the reduced wake width and
confused flow responsible for the effect will also probably reduce the fluctuating drag forces
responsible for the excitation as well, and so the net energy balance of the cylinder may
undergo little change.
Experimental evidence relating to the above question is ambiguous. King (ref. 72) obtained
good quantitative agreement of the in-line response behaviour between tests on model piles
at subcritical Reynolds numbers, and the tests on full scale piles at Reynolds numbers
extending through the critical region, carried out by Wootton et al (ref. 67). Tests on full
scale pipeline spans in the critical Reynolds number region however, comprising the
Department of Energy sponsored special study (ref. 62) show significantly increased in-line
response amplitudes. There are however, a number of alternative potential causes for this
latter effect in addition to a reduction in hydrodynamic drag induced damping, and these are
discussed later.
For the case of a submarine pipeline span, there are a number of additional sources of
potential damping which do not fall conveniently into the three principal categories defined
at the beginning of this section.
Friction at the span touchdown points has already been mentioned as a potential
contribution to structural damping. The nature of span touchdown points however, with the
range of configurations possible, provides scope for additional means of energy dissipation.
Span motion may lead to deformation of the seabed soil at the touchdown point, in which
case the material damping of the soil would contribute to the overall damping. A further
possible effect is the displacement of seabed sediment at the touchdown point due to the
'pumping' action of a moving span. This would again dissipate energy and increase the
total damping.
For a concrete coated line, the concrete will contribute to the total damping. Reinforced
concrete tends inherently to possess minute cracks in the structure due to the inability of
concrete to withstand significant tensile strain. Friction at the internal crack surfaces
therefore contribute a form of structural damping, in addition to the material damping of the
concrete matrix itself.
5.3 SPAN RESPONSE IN WAVE AND COMBINED
WAVE/CURRENT FLOWS
The parametric description of span response outlined in Sections 5.1 and 5.2, assumes that
the frequency of the primary excitation force generated by the vortex shedding remains
steady or at least quasi-steady over a sufficiently long period for the span response
amplitude to build and reach a steady state condition. This requires either a steady constant
approach flow velocity or an approach flow in which the velocity changes very slowly, as
in the case of a tidal flow where the time scale of the cyclical velocity changes in several
orders of magnitude larger than typical span vibration periods.
In the case of a span subjected to wave induced fluid motion the above condition is
ostensibly clearly violated. The wave period and hence the period of the reversing
oscillations of the approach flow back and forwards past the span is often of the same order
of magnitude or at most one, order of magnitude greater than the natural period of span
vibration. Very little information is available relating to the number of span oscillations,
starting from zero displacement position, to build up to the maximum response amplitude
even under steady state approach flow conditions. This number clearly depends on the mass
ratio between the pipeline and the surrounding fluid. From available information, as quoted
for example by King (ref. 8) and Hallam et al (ref. 90), for a typical pipeline span immersed
in water, the required number of oscillations to build up to an equilibrium amplitude is
probably of the order of 10. If this build up process is to take place over half a wave period
67
Contents
before the flow reverses, then the wave period must be at least of the order of 2 x tOffN
where fN is the natural frequency of span vibration. For spans with a natural frequency lower
than say 1 second, which are quite common in practice, this condition requires a wave
period in excess of 20 seconds which is outside the normally encountered range,
The finite time and number of oscillations required to build up the span vibration amplitude
tu a significant level also impuses constmints un Keulegan Carpenter number necessary
to produce such a response. Arguments were presented in Sections 4.4 and 4.5 that vortex
shedding frequency tends to follow the instantaneous variations in velocity in an unsteady
flow field. Accepting this, it can be shown by'simple integration that the number of
vortices, N, shed from one side of the pipeline and hence the number of force cycles
occurring in a given half wave cycle can be estimated from N=S.Kht where S is the
Strouhal number and K is the Keulegan Carpenter number given by 21ta/D, where a is the
amplitude of the water particle motion past the span. Assuming that close to span
resonance at the stage when oscillations begin to build up, the average vortex shedding
frequency is approximately equal to the span natural vibration frequency, fN then the
number of span oscillations in a half wave cycle is approximately equal to N the number of
shed vortex pairs and force cycles. Assuming once again that full amplitude build up
requires at least 10 oscillations then 10. This implies that the Keulegan Carpenter
number Substituting S= 0.2, indicating that the Keulegan Carpenter
number of the wave induced oscillatory flow close to the bed must be at least of the order of
150 for span response to reach maximum amplitude over a half wave cycle.
These simple considerations alone suggest that the wave induced amplitude response of
many pipeline spans will be significantly less than the equilibrium vibration amplitudes
generated by steady tidal currents.
Closer examination of the problem however, based on physical reasoning, indicates that the
vibrational response of a pipeline span subjected to oscillatory wave flow is unlikely to be
confined to short bursts of build up and complete damping out within the duration of
discrete half wave cycles. In practice, because of the considerable stored energy, the span
oscillations will tend to continue into the return half wave cycle in which the current
reverses direction. The strong possibility exists therefore that given conductive flow
conditions for which vortex shedding can synchronise, and lock in to the residual span
vibration, then the oscillations could continue to increase in amplitude during the return
half cycle. This continuity of response could thus extend over a sequence of several suitably
matched waves in a typical natural random wave train, and the amplitude response could
build up further beyond the limited level possible within an individual wave cycle.
These speculations are supported by the findings of relatively recent laboratory tests carried
out by Sarpkaya (ref. 91 and 92). Two sets of experiments were reported. Firstly, using
spring mounted rough and smooth surfaced cylinders placed in the plane oscillatory flow
generated in a water tunnel at the Naval Postgraduate School (named hereafter the NPS
experiments and data). Secondly, the response of a vertically-cantilevered model pile
cylinder under the action of regular surface waves was investigated at the Shell
Development Company's Westhollow Research Center in Houston (referred to hereafter as
the WRC experiments and data).
In the NPS tests the cylinder motion was restricted to transverse cross flow direction by the
mounting system. It was found that the cylinders exhibited hydroelastic response in a
relatively well defined range of reduced velocity, V., bounded by 4 < U.,JfND < 12 where Urn
is the maximum cyclic velocity. The maximum amplitUde response, a..JD, as a function of
reduced velocity is reproduced in Figure 5.7 from reference (ref. 91). As can be seen in
Figure 5.7, the amplitude response exhibits a relatively sharp and narrow peak at a reduced
velocity of approximately 5.5. Sarpkaya's measurements demonstrated that the peak
coincided with perfect synchronisation when the vortex shedding frequency remained
remarkably constant and equal to the natural cylinder vibration frequency, fN, throughout the
sinusoidally varying reversing flow cycle and also throughout consecutive cycles. Changing
the reduced velocity ! about the peak value of 5.5, introduces fluctuations in vortex
shedding frequency, f" of the order of 50 percent of the natural frequency which in turn
coincides with the sharp fall off in amplitude response either side of the perfectly
synchronised peak. The vortex shedding remains synchronised however exhibiting a
frequency which fluctuates about a level close to the natural frequency. This behaviour is
illustrated in Figure 5.8 also reproduced from reference (ref. 91).
68
Contents
_2
v.
VARIOU$ UFT COEfF!CIENtS ANb RfLA1'we: DISPLACEMENT
vERSUS V
R
-=-OR A SAN!) - Ff{)UGHI=:NEO CYliNDER IN
HARMON!C FLOW (II./O:;! 1100::
fROM SARPKAYA {REF gil
,-4-
,- a-
0-'
Figure 5.7
Rough cylinder response in harmonic flow
__
o 0-' (n 03 0-4 0-5 0-6 0-1 08 O-g
liT
VA"rAtroN Of THE VORTex SHEDOING fflQUf:NCY WlTH
TIME cuttiNG A CYCLE FOR VARIOU$ VALUES OF VR
rROM $ARPKAl'A (REF. 911
Figure S.8
Variation of vort .. shedding frequency during wav" cycle
'-0
69
Contents
The response of the vertically cantilevered cylinder in regular (single frequency) waves
shows a similar but slightly more complex pattern of behaviour as illustrated in Figures
5.9 and 5.10 reproduced from Sarpkaya et al (ref. 92) and corresponding respectively to
intermediate and deep water wave conditions. The root mean square of the relative amplitude
response expressed by (alD)= in Figures 5.9 and 5.10 now exhibits two relatively sharply
defined peaks with the dominant first peak once again located at a V R value close to 5.5.
The second peak, located at a slightly higher V R value, is thought to be linked to the
unstable pattern of vortex shedding observed round fixed cylinders in harmonic flow over
the range of surface Keulegan Carpenter number, K, namely 10 to 12.4 used in the tests.
Interactions also take place between the in-line and cross-flow response of the cantilevered
cylinder. This further complicates the response pattern as also does the inherent three
dimensionality of the wave induced flow field near the surface and the resulting depth
variations in K which affect the vertical coherence of the shed vortices.
Perhaps the most important point about these observations in the context of the present
discussion, is that they demonstrate that highly unsteady reversing wave type motion can in
appropriate circumstances induce and sustain significant levels of hydroelastic vibrational
response similar to that by a steady tidal or other current. It must be borne in mind however
that both sets of experiments reported by Sarpkaya were carried out using almost perfectly
regular single frequency oscillatory flow. The results of these extremely useful but idealised
tests thus need very careful interpretation as far as their relevance to the response behaviour
of pipeline spans located in the highly irregular multi-frequency flow field induced by
natural wave conditions is concerned. These crucially important implications are further
discussed below.
Sarpkaya (ref. 91) suggested that by using a modified form of the stability parameter, K"
discussed in Section 5.1 and 5.2, and appearing in Figure 5.2, a much improved correlation
could be achieved for peak amplitude response data at perfect synchronisation between the
smooth and rough surfaced cylinder results obtained in the NPS experiments. This new
'response' parameter Rp defined by in which MIL is the cylinder mass/unit
length, t; is the damping coefficient defined in Section 5.1, also includes the rms lift co-
efficient for the identical cylinder clamped against oscillation in the same flow. Sarpkaya
(ref. 91) reasonably argues that incorporation of C
L
o
rms
is necessary since the response of a
given body to a given flow must depend to a greater or lesser extent on the starting
dynamical forcing conditions expressed by this lift coefficient. A plot of peak response
amplitude aID versus Rp reproduced from reference (ref. 92) is shown in Figure 5.11. As
can be seen the data exhibits very modest scatter for ihis type of experiment. Also the
correlation between the NPS and the WRC results obtained under very different test
conditions is good.
In steady flow past cylinders the dynamic response is frequently expressed as a function of
&- defined by &- = Mt;(8rr2S2)/pLD2 Sarpkaya et at (92) again propose a generalised form of
this parameter namely Dr/CL',m, = (8rr2S2)Rp where S is the Strouhal number defined by
f,DN. The NPS and WRC peak response amplitude data is shown plotted against this
parameter in Figure 5.12 reproduced from reference (ref. 92). Sarpkaya et al (ref. 91) also
used ArieL arms to correlate aID data from steady flow experiments at perfect
synchronisation. A curve was fitted to the data with the following equation:
I
aID = 1.72y/[1.91 + Eqn 5.14
in which y is the response mode shape factor which equals I for a spring mounted rigid
cylinder, 1.305 for the first mode of a cantilevered cylindrical pile. as used in the WRC
tests and 1.155 for a sinusoidal mode.
The above equation 5.14 for aiD is plotted in Figure 5.12 for comparison with the
oscillatory flow data. Surprisingly, and rather remarkably, the equation based on steady flow
data also gives a good prediction of the peak response under unsteady regular oscillatory
flow conditions. This is particularly true in the case of the NPS data. The results from the
WRC tests with the vertical cantilever in regular waves lies slightly below the predictions
of the equation as shown in Figure 5.12. Sarpkaya et al (ref. 92) suggest that in the case of
the WRC experiments, the lift coefficient is an integral average and does not account for
changes with depth or KeulegaiI Carpenter number. The correlation of the lift
force is also reduced by an unknown amount even in this simplest of wave flows.
70
Contents
h
~
E
{ :
!J
u
. ,
;
;
80-,-------------------------,-10
48
32
16
0
SERIES 400 RESPONSE
LIFT COEFF ----
,
II
II
II
I I
I I
I I
I I
/ ~ .. ;\
'--...
2 3 4 5 6 7 6 9 10
V
R
ROOT MEAN SQUARE LIFT COEFFICIENT AND
NORMALIZED TRANSVERSE RESPONSE AS A FUNCTION OF
THE REDUCED VELOCITY FOR WRC- SERIES 400
06
OA
02
0
~
E
'" "-
0
N
FROM SARPKAYA ET AL (REF. 92)
Figure 5.9
Cylinder response to wave intermediate depth conditions
71
Contents
72
SERIES 500
SA
I
I
I
RESPONSE
LIFT COEFf ----.
06
-4
... '
I \
't
\ ~ - ,
\ '" ~
\ "I
l....I ..
O i - - - _ , - - - - , - - L ~ - - - - , _ - - _ , - - - - , _ - - _ , - - - - ~ O
2 3 4 5 6 7 8 9 10
V
R
ROOT MEAN SQUARE LIFT COEFFICIENT AND
NORMALIZED TRANSVERSE RESPONSE AS A FUNCTION OF
THE REDUCED VELOCITY FOR WRC-SERIES 500
~
E
-
" ....
0
N
FROM SARPKAYA ET AL (REF. 92)
Figure 5.10
Cylinder response to wave deep water conditions
Contents
10
09
08
07
O-G
~ 0'5
0
Q-
0-3
0-2

0-1
0
0 05 10 15 2-0 4-0 45
R,
FROM SARPKAYA ET AL (REF. 92)
Figure 5.11
Relative amplitude as a function of the response parameter
The test results presented by Sarpkaya and his colleagues thus not only demonstrate that
regular oscillatory flow can sustain vortex induced response from one cycle to the next, hut
also that perfect synchronisation between vortex shedding and natural oscillation frequency
can be achieved pro-ducing closely similar aruplitude response to that generated by an
equivalent steady current.
These remarkable observations are clearly highly relevant in attempting to assess the effects
of natural sea waves on pipeline span response. Severe difficulties arise in addressing this
problem however because of the random nature of the near bed oscillatory flow conditions
induced by multi-directional sea waves which cover a wide spectrum of heights and
frequency for a prevailing sea state. At modest Keulegan Carpenter numbers under closely
regular single frequency flow oscillations it will take several complete flow cycles to build
up response to maximum amplitude at V R= 5.5 corresponding to perfect synchronisation.
The relatively narrow band response peaks illustrated in Figures 5.7, 5.9 and 5.10 indicate
that quite modest deviations in reduced velocity about this peak 5.5 value will result in a
significant reduction in the response aruplitude. Given the considerable irregularity of the
velocity field generated by an individual wave in nature, coupled with the fact that in
general consecutive waves are likely to have appreciable variations in height and frequency
even within so-called wave 'groups', it thus appears highly unlikely that the peak
aruplitudes indicated in Figure 5.13 will build-up under natural sea wave conditions.
The peak aruplitude curve corresponding to equation 5.13 plotted in'Figure 5.13 can
therefore be considered as representing an upper bound on the likely response induced by sea
waves particularly for Keulegan Carpenter numbers less than say 100. Nevertheless, random
groupings of suitably matched consecutive waves generating a band of vortex shedding
frequencies 'bracketing' the span natural frequency, can be expected to induce significant
response aruplitudes. This is likely to pio-duce a mo-dulated pattern of response build-up and
subsequent daruping as illustrated diagrarumatically in Figure 5.13.
73
Contents
2,---------------------------------------,
l'o 0-2
o
,
o
0-05
{
NPS DATA
.. .


2 5 10 20 50 100
FROM SARPKAYA ET AL (REF. 92)
Figure 5.12
Relative amplitude as a function of the generalised damping ratio
UPPER BOUND
STEADY FLOW
AND REGULAR
WAVE RESPONSE
AMPUTUDE. FROM
FIGURE 6,13
MODULATED SPAN RESPONSE UNDER
NATURAL SEA WAVES
Figure 5.13
Modulated span response under natural sea waves
r.m.s
Whilst even the qualitative form of this postulated response behaviour remains somewhat
speculative in the absence of any direct observational evidence, the above arguments
underline the fact that there is a strong probability of this or some very similar type of
response occurring under natural wave conditions. This poses something of a dilemma in
the context of span assessment. Firstly. there is no directly relevant data or reliable
theoretical model currently available for estimating the magnitude of the wave induced span
response amplitude illustrated in Figure 5.13 including any variations which may occur for
different wave characteristics. Such an estimate is essential in order to establish whether or
not the "resulting vibrations are potentially damaging to pipeline spans. Secondly. if span
response is significant there is no directly relevant data obtained under real or simulated
random wave conditions available for estimating the threshold critical V R values at which
the response commences. In Sarpkaya's idealised tests with regular harmonic flow a value
74
Contents
of V
R
;;;; 4 was estimated as the threshold point. However, natural sea waves cover a wide
spectrum of heights and frequency which in tum will produce a similarly broad spectrum of
vortex shedding frequencies and linked force-excitation. This is likely to reduce the threshold
V R value below 4 to an at present unknown level.
These large gaps in existing knowledge. which can only be filled with new information
from future research, clearly make it impossible at the present time to propose rationaIly
based criteria for assessing span vibrational stability specifically relevant to the action of
natural sea waves. In these difficult circumstances, it appears inevitable that assessment
must sensibly be based on a simple, easily understood and essentially conservative method.
It is therefore recommended that the same criteria for the onset of significant span vibration
derived from the present full scale field test results in steady flow are used as a basis for
assessing span response under wave action. Instead of the steady current velocity at the
height of the top of the pipeline span however, the .amplitude of the velocity oscillations
induced in the bed region by the relevant design wave is substituted in calculating the
reduced velocity, YR. The relevant design wave itself would not necessarily be an extreme
wave since the discussion above is founded on build-up of excitation from consecutive
waves rather than from an extreme event. A root mean square or as commonly used for
design purposes, a 'significant' description of the wave would appear a more sensible
approach.
As described in Section 4.3 and 4.5, superposition of waves on steady tidal or other currents
adds new dimensions of complexity when attempting to estimate relevant combined
approach flow velocities and vortex shedding patterns. This enhanced uncertainty inevitably
applies also to span response behaviour. Once again there is a complete lack of data and
information relevant to span vibration under a randomly fluctuating oscillatory velocity
component superposed on a steady current. Tsahalis (ref. 93) investigated the response of a
model pipeline span to combined waves and steady currents. He concluded that the
superposition of waves on a steady flow has a pronounced effect on the amplitude response
in both the in-line and cross-flow directions. However, because regular single frequency
waves were used in these tests, and also the ratio of wave induced to steady current
velocities was less than unity in the response zone, it is difficult to draw general
conclusions from the results of what is nevertheless a pioneering study. In view of the fact
that in nature, wave induced velocities are typicaIly up to 4 or 5 times the bed region tidal
velocities, it seems reasonable to assume that the likely induced span response will be
similar to the wave only. pattern postulated previously and illustrated in Figure 5.13.
For safety reasons, therefore, because of the almost complete pervading ignorance regarding
the actual.span response, both qualitative and quantitative, under these complex but
unfortunately all too typical natural flow conditions, it is recommended that assessment
criteria should be based on the known steady flow situation. An appropriate conservative
value of approach velocity is obtained by simple linear addition to the steady current
velocity at the level of the top of the pipeline span and the wave induced velocity amplitude
close to the bed as discussed in Section 4.3.
The discussion in this section is restricted to resonant span response induced by vortex
shedding. When assessing general span stability it is clearly crucially important to assess
the fatigue implications of the lower frequency forced oscillations generated by the
oscillatory drag forces resulting from the reversing wave currents past the span. Drag
deflection observations from the present full scale tests suggest that larger wave induced
forces could produce cyclically reversing in-line deflections approximately one order of
magnitude greater than typical in-line response amplitudes produced by vortex shedding.
The effect could therefore dominate the in-line response analysis under wave and combined
wave/current conditions.
It is clear from the above discussion that there is an urgent need for further research into
span response under the complex flow conditions induced by random natural sea waves in
combination with steady currents. Such a future test programme should combine both
controlled laboratory tests with a full scale field investigation for maximum effectiveness.
75
Contents
6. RESPONSE ANALYSIS
6.1 DEFINITION OF RESPONSE THRESHOLDS
The definition of a design threshold above which unacceptable pipeline span vibration will
take place, involves two stages of development. Firstly, the level of vibration which is
deemed to be unacceptable must be defined. Secondly, the physical conditions which may
give rise to this unacceptable vibration must be determined.
In practice, the simple definition of unacceptable vibration may not be clear cut, and may
depend upon the individual circumstances relating to any given span. Factors influencing
the level of vibration which may be tolerated include the following:
concrete coating fatigue considerations;
steel fatigue considerations;
any existing damage to pipeline or coating;
possibility of impact with seabed rocks or other nearby objects or installations.
In addition, the practicality of perfonning immediate rectification work on a newly
discovered span may be severely limited due for example to adverse weather conditions.
Under the latter circumstances, some vibration may be unavoidable, and it would be
necessary to assess the degree or risk involved given the prevailing conditions, in order to
determine the most appropriate course of action.
It has already been shown in Section Three, that the reduced velocity, V R, is a simple and
effective parameter for assessing the stability of a pipeline span (providing that it is
suitably modified to account for variations in Strouhal number). VR has been commonly
used for span assessment in the past, and it is appropriate to retain its use for the definition
of span response thresholds. In view of the many variables involved however, it is not
realistic to attempt to define a single threshold of response, since acceptability criteria may
vary for different pipeline spans. Instead, two thresholds, VRI and VR2 are defined. By
comparing the calculated V
R
for a given span with these thresholds, the span can be placed
in one of the three categories;,a); b) or c), as 'defined below.
a) VR<VR1 No significant vibration of the span occurs.
b) VR1 <VR<VR2
Some vibration may occur, with acceptability dependent upon the
individual circumstances.
c) VR>VR2 Severe vibration occurs, unacceptable under all circumstances.
In practical terms, category a) in general simply comprises the region of low reduced
velocity, where the approach flow velocity is inadequate to generate any significant
vibration. Category c) can normally be practically defined as the large amplitude vibration
occurring during cross flow lock in. Category b) includes the greatest area of uncertainty. It
will generally comprise the region of predominantly in-line vibration prior to the onset of
cross-flow lock-in. If a pipeline span is found to lie in this category. then further analysis
may be performed to determine whether or not the level of vibration is acceptable.
6.2 FULL SCALE TEST ANALYSIS
6.2.1 Available Full Scale Test Data
The testing of full scale pipeline spans is an expensive and technically demanding
procedure. The amount of available full scale test data is therefore very limited, in contrast
to the relative abundance of laboratory scale data relating to flow around circular cylinders.
Three principle sources of full scale test data are available however, and these are described
below.
The most comprehensive source of full scale data is the special study commissioned by the
United Kingdom Department of Energy specifically for present work. This study (ref. 62)
was performed by Hydraulics Research Limited, under the management of l.P.Kenny. The
77
Contents
tests were carried out at Sheperdine in the estuary of the River Severn, on a 20-inch
diameter test pipeline span. Span lengths of 50m and 40m were tested at varying heights
above the seabed, and both bare metal and simulated concrete pipe surface conditions were
examined. A full report on the test programme is provided by reference 62.
A further source of full scale test data is provided by the work carried out in association
with the Polar Gas Project (ref. 63). This work was also carried out at Sheperdine in the
Severn Estuary, by Hydraulics Research Station (now Hydraulics Research Limited). A 24-
inch diameter x13m length test span was studied, incorporating a flexible suspension
system to simulate the effect of a much longer span. Smooth and rough pipe surface
conditions were examined, and tests were carried out at varying heights above the seabed.
Additionally, a series of full scale tests were carried out by Bruschi et al in Venice Lagoon,
Italy. These tests are reported in reference 22, and were carried out on a 20-inch diameter x
67 m length test span. Both smooth and rough pipe surface conditions were again tested,
but the test span was in this case fixed at a single height above the seabed.
6.2.2 Theoretical Considerations
The interpretation of experimental data can lead to ambiguous conclusions if it is attempted
on a purely empirical basis. It is therefore appropriate to first examine the theoretical
approach to the problem, in order to facilitate the interpretation of the test results. The
following discussion thus applies the theoretical concepts discussed previously to the
particular problem of span response analysis.
As discussed in Section 3, three types of vortex-excited response are known to occur. These
are well illustrated by Figure 6.1, which is taken from King et al (ref. 65). This figure
shows a plot of response oscillation amplitude against reduced velocity. The three types of
excitation are also schematically illustrated. It should be noted that although numerical
values are shown on this figure, they apply to model tests and may not be appropriate for
full scale applications.
Examinations of Figure 6.1 shows the following sequence of events. As the reduced
velocity, V R, is increased from zero, there is initially no cylinder motion. The first in-line
response region is then encountered, and the oscillation amplitude first builds up in
response to the symmetrical pattern vortex shedding. before dying away again with further
increase in V R. The second in-line response then builds up, corresponding to the region
where the frequency of the conventional alternate vortex shedding region is nominally equal
to one half of the cylinder natural frequency, fN This response then dies out with increasing
VR and is replaced by the considerably more energetic cross-flow response, occurring when
the vortex shedding frequency is nominally equal to fN An additional effect shown by
Figure 6.1 is the dependence of the response amplitude un the damping, Ii.
Whilst Figure 6.1 illustrates the three response regions in terms of their effect on cylinder
response amplitude, Figure 6.2 a) and b) shows their effect on the vortex shedding frequency
measured in the cylinder wake, fw. In the following discussion, the symbol fw is used to
define the frequencies actually measured in the wake of a vibrating cylinder, whereas f, refers
to the stationary cylinder vortex shedding frequency defined by the Strouhal number
relationship. Figure 6.2 shows diagrams from King et al (ref. 66) and King et al (ref. 65).
The diagrams are also reproduced and discussed by King (ref. 8).
Figure 6.2 a) and b) shows plots of vortex shedding frequency, measured in the cylinder
wake, against approach flow velocity. The straight lines show the theoretical linear
relationship that f, would be expecred to follow if the cylinder was stationary and the
Strouhal number possessed a conventional sub-critical value of 0.2. Since the figure is for
explanatory purposes only, the numerical values on the plot axes are not of great relevance.
78
Contents
o
"-
o
~
~
...J
...J
g
o
....
o
w
o
::>
i-
a'
:;
<I
',O+---+---t---+--+----+-++----j
o 5 +---\-L==-
2 3 4 5 6 7
REDUCED VELOCITY V
R
NOTES
i) THE NUMERICAL VALUES SHOWN ON THE FIGURE RELATE
TO MODEL TESTS AND MAY NOT BE RELEVANT TO A FULL
SCALE SITUATION
ii) & = LOGARITHMIC DECREMENT OF DAMPING
FLOW PATTERNS
V
R
8::
'"
...
~
12 - 25
~
, ..
SYMMETRIC
8 ~
.,
28 - 3-4
V
V
ALTERNATE

--
(C>
35-8
\.., ~
ALTERNATE
FROM KING 0' 01 (REF. 65)
Figure 6.1
Summary of the three vortex Induced response types
79
Contents
'"
"

'"
;:
"

6 " <0 6
,0 :!!"
<0'
0
,
"'
0
5 a: 5
0
(J
'"
a:

,...
4 4
u
/x
z
"'
:>
3 3 0
x Bose bending moment
'"
a:
//r mode
IL
)(
2 2
"' f-
a:
0
/
>
N
J:
A 8,
.Y
C
0 0
5 10 15 20 25 30 35
VELOCITY in/s
0) FIRST AND SECOND IN - LINE RESPONSE REGIONS
FROM KING el 01 (REF. 65)
N
J:
"'
40
"


,... 30
u
z
'"
:>
0
Cylinder w
20
a: frequency If N
Ie
/'
x l( x x
t-
Z
'"
1-0
/'
Z


0
0
..!-
5 10
VELOCITY
b) CROSS - FLOW RESPONSE REGION
FROM KING .101 (REF_ 66)
Figure 6.2
15
8
20
Influence of the three fundamental response types on vortex shedding frequency
80
Contents
As the flow velocity is increased from zero (see Figure 6.2 a) no significant cylinder
motion occurs, and fw initially follows the straight line Strouhal number relationship. With
further increase in velocity, the cylinder enters the first in-line response region, and this is
accompanied by a slight deviation of fw from the linear Strouhal number relationship. As
discussed by Wootton at al (ref. 67), the details of the flow in this symmetrical vortex
shedding regime are subject to uncertainty, whilst King et al (ref. 66) state that the ratio of
fw to fN is subject to variation within the same region. There is therefore no total
domination of fw by the cylinder motion in this region. despite the significant response
amplitude.
With further increase in velocity, the first type of in-line response dies out, and fw collapses
back onto the Strouhal number relationship (fs) line. The second type of in-line response.
due to alternate' vortex shedding, then commences and it can be seen that the cylinder
motion totally dominates the vortex shedding for a significant velocity range. It is reported
in reference 66 that fw within this in-line locked-in range is equal to which is in
agreement with the theoretical discussion of Section 3. At the end of the lock-in range, fw
reverts to following the linear Strouhal number relationship.
Figure 6.2 b) shows the effect of cross-flow cylinder response on fw. This type of response,
which occurs at higher values of V R, again results in domination of fw by the cylinder
motion, and fw remains locked in to fN for a significant range of velocity values before
jumping back to follow the linear Strouhal number relationship at the end of the lock-in
range.
The behaviour discussed above is summarised schematically in Figure 6.3 a). fw is shown
on fN. and plotted against V R. with the experimentally measured characteristics
shown in Figure 6.2 reproduced in schematic form. The main characteristics are the.
symmetrical in-line shedding response region, the in-line lock-in zone due to alternate
shedding, where fw = 0.5 fN, and the major lock-in zone where fw = fN and the cross-flow
response takes place. Outside these response the vortex shedding frequency follows
the linear Strouhal number relationship, with the Strouhal number, S, in this case assumed
to be the typical sub-critical value of 0.2 commonly measured in laboratory experiments.
It was shown in Section 4 however, that S may be significantly greater than 0.2 in the
critical Reynolds number range where turbulent transition is taking place in the cylinder
boundary layer. The effect of such an increase in S is shown schematically in Figure 6.3 b).
An immediate consequence is that the slope of the Strouhal number relationship line is
increased, or in other words for a given approach flow velocity and hence V R, f, will be
higher. The result of this is that the in-line and cross-flow lock-in zones in the regions,
where fs = 0.5 fN and fs = fN respectively, occur at lower values of V R. A further result if S
is sufficiently high, is that an overlap may occur between the two lock-in regions, where
suitable conditions simultaneously exist for both in-line and cross-flow locked-in response.
This is also illustrated in Figure 6.3 b).
The fIrst in-line response region, where the excitation is due to symmetrically shed vortex
pairs, is dependent upon a particular relationship between the maximum oscillation velocity
of the cylinder and the approach flow velocity. The symmetrical vortex shedding range is
controlled therefore by a different physical mechanism to the Strouhal number dominated
alternate vortex shedding regime. The influence of a change in Strouhal number upon the
onset of this first in-line response region is therefore expected to be weak. It is thus
probable that the earlier onset of the second (Strouhal number dependent) in-line response
region may introduce a tendency for the two in-line response regions to also overlap.
In summary, the effects of an increase in Strouhal number above the widely assumed sub-
critical value of 0.2, are that an earlier onset of the 2nd in-line and the cross-flow response
will occur, and the three response regions will be forced into closer proximity, with
possible overlapping of the respective regions.
81
Contents
82
I-a ---------------
2nd IN-LINE
//,'0
0
",,"'=,
CROSS-FLOW
IN -LINE DIES OUT
BEFORE CROSS-FLOW
BEGINS
1st IN-LINE
(SYMMETRICAL SHEDDING)
0) STROUHAL NUMBER = 0 -2
S> 0-2 (FULL SCALE TESTS)
aVE RLA P ZaN E
POSSIBLE OVERLAP ZONE
1st IN-LINE
(SYMMETRICAL SHEDDING)
b) STROUHAL NUMBER GREATER THAN 0-2
Figure 6.3
Influence of an increased Strouhal number upon response characteristics
Contents
6.2.3 Special Full Scale Study - Smooth Pipe Tests
The smooth pipe tests were carried out using an uncoated steel test span. The term
'smooth' is slightly misleading, since the pipe surface did possess slight irregularities,
presumably resulting from the manufacturing process with an estimated roughness size of
0.5 mm. This gave a relative roughness, kID of 1 x 10". The flow velocity at the test site
varied up to a maximum value of approximately 1.5 mis, giving a maximum Reynolds
number of approximately 6 x 10
5
Most test data was in fact obtained in the range of
Reynolds numbers between approximately 10' and 6 x 10'.
6.2.4 Strouhal Number - Smooth Span
Figure 4.9 a) of section Four, shows a plot of Strouhal number as a function of Reynolds
number for a circular cylinder with a relative roughness of 0.75 x 10-
3
as measured by
Achenbach and Heinecke (ref. 58). This is sufficiently close to the test spari relative
roughness of 1 x 10-
3
for a reasonable comparison to be made. It can be seen that over the
Reynolds number range relevant to the full scale test programme, the Strouhal number is
subject to significant elevation above its sub-critical value of 0.2, due to the effects of
turbulent transition in the so-called critical Reynolds number region.
Figure 6.4 shows plots of Strouhal number against Reynolds number measured for the full
scale test span. The span was rigidly clamped for these measurements, to avoid any
dynamic modification of the vortex shedding ,Strouhal number. These results tend to
confirm the measurements of Achenbach and Heinecke, showing that a cylinder of this
relative roughness is subject to critical transition over the Reynolds number range of
approximately 10' - 6 x 10'. Strouhal numbers encountered in this region therefore tend to
be significantly in excess of the commonly assumed sub-critical value of 0.2. In view of
these generally high Strouhal numbers compared to most model tests, it would be expected
that an earlier onset of the second in-line and the cross-flow lock-in would be observed,
together with a possible overlap between the response zones. The reasons for this are
discussed in Section 6.2.2, and examination of data from the dynamic response tests shows
that the expected effects do in fact occur.
6 _ 2.5 Frequency Response - Smooth Span
Figure 6.5 shows typical wake frequency and span response frequency plots relating to free
span tests on the smooth pipe. Parts a) and c) of the figure shows data from a SOm span
length, and in parts b) and d), the data relates to a span length of 40m. The examples shown
form part of a wider test series and are simply selected to give a representative illustration
of the wake behaviour. The significance of the gap ratio, GID, will be discussed in more
detail later. Plots a) and b) show the dominant wake vortex shedding frequency, fw,
normalised on the span natural frequency, fN, plotted against the reduced velocity,
VR (=U/fND). Parts a) and b) of the figure are thus directly comparable with the theoretical
plots shown in Figure 6.3. In parts c) and d)" of the figure similar graphs are shown, but in
this case the span response frequency, f, is plotted in place of f .
Examination of Figure 6.5 a) and b) immediately shows the two lock-in regions, where the
vortex shedding becomes controlled by the cylinder motion. The first lock-in region
corresponds with the in-line type response where f. should theoretically equal 0.5 f
N
and a
broad plateau is apparent on the response graphs, nominally corresponding with this value.
The second lock-in region, corresponding to span motion of the cross-flow type, is expected
to control f. at or near the span natural frequency, and it can correspondingly be seen that a
further plateau on the graphs is apparent where f.lfN is equal to unity.
The theoretical arguments presented in Section 6.2.2, predicted that a relatively high
Strouhal number would lead to a degree of overlap between the lock-in regions. Figure 6.4
shows that the measured Strouhal numbers are typically appreciably higher than the normal
sub-critical value of approximately 0.2 (with the mean value lying in the region of 0.28),
and Figure 6.5 confIrms that the expected overlap does exist. The vortex shedding frequency
jumps abruptly from one lock-in zone to the next as suggested in Figure 6.3 b), without
the intermediate stage of following the Strouhal number relationship as shown in Figure
6.3 a).
83
Contents
00
.. 00 ,
"
" "
" ,
03
"
"1
"
'"
030
"
00
"
" "
0>'
!a 00
"
OJ> 0000

" 0
"
0 Sa!)
0
0
"
"
:r
"
'"
!.
""
z
" "" " 0
3
" " "
'"
I
s 0.21
020
02


s s
..

0
iii'
-:!!
'CD
o c
3

9!"
O l ~ OH 0'10
:;"
n
:r
0
,
n
0

;
a.
!. 01 I o I I 0
l
10' 2 3 4 5 6 7 8 9 10
6
10' 2 3 4 5 6 7 8 910
6
10' 2 3 4 5 6 7 8 9 10
6
." Re Re Re
-0

G/D 05 G/DIO G/D 20
5 025 s 031 5 028
G 50rn SPAN
,
40m SPAN
FROM FULL SCALE TEST STUDY (REF.62)
Contents
14J
5'0'3//5'0'28
IA
1'2
::J
' '\jl 6'>0 '0""
KEY PlAN'- olAND b)
i"
0 VELOCITY INCREASING

q:>
. + &I ..
I{)-j ,,./
r
,
VELOCITY DECREASING

++++8
<II
fw
08
1
/..r
I, 0'8
,
a IN
I
IN
0
9- 0'6
0 06
:T
0

."
m
04

.;:
02
1 /
O}SOmSPAN c) 50 m SPAN
02
."
n G/D = 15 G/O 15
!.

o r I I 0
m
0 2 3 4 5 6 7 WAKE FREQUENCY SPAN RESPONSE 0 2 3 4 5 6 7 ,..
.. ."
V
R
PLOTS FREQUENCY PLOTS VR
m
cC
c
... ;
I
1-4
I
4
0>
.".
50,28
m en

00

1'2
12j
.. KEY PLAN -c) AND d)
8


..,
IN CROSS
0
10
0
II NE FLOW
" .. "

1'0 ...,0
0 VELOCITY I NCREA$ING
"
..
-
G

VELOCITY DECREASING
;
Iw 08 08 ..,
I,
c
.. IN
IN
n
06
06
'<
."
9:
0'11
04

0'2
b) 40m SPAN d)40m SPAN
02
00
G/O 2'0 G/O 2'0
v,
I 0
2 3 4 5 6 7 0 2 3 4 5 6
V
R
FROM FULL SCALE TEST STUDY (REF.62) V
R
Contents
Further information can be derived by observing the differing behaviour exhibited during the
respective phases of increasing and decreasing flow velocity. During the increasing flow
phase, fw remains locked-in nominally at 0.5 fN for a considerable reduced velocity range
before jumping up to the second lock-in region in which fw nominally.equals fN. Upon
subsequent reduction of the velocity however, fw does not drop back at the original point of
increase, but remains locked in at the higher frequency for a significant interval before
finally dropping down at a lower V R value. The resulting graph exhibits the characteristics
of a hysteresis loop. It can be seen that the hysteresis behaviour is less pronounced for the
40m span than the 50m case in the example shown, and this feature was reproduced for
other tests at different GID values.
The hysteresis behaviour can be physically explained as follows. The two lock-in regions
represent stable behaviour regimes in which a balance is established between the energy
input to the vibrating span from the flow. and the energy dissipated through damping. In
each regime, the span motion is able to control the vortex shedding frequency through the
dependence of the latter on the vortex strength as explained in Section 3. Consider first, the
situation where predominantly in-line response is taking place, and fw has just become
locked-in at 0.5 fN. In this situation it is natural that the span response should stabilise in a
. regime where the energy input and the energy dissipation exactly balance. If the flow
velocity is increased, the system readjusts until the energy balance is re-established, with
the span motion continuing to control the vortex shedding frequency and ensuring
synchronisation between the excitation and response frequencies. Stability of the system is
thus maintained.
If the flow velocity increases further, the point is reached where the stationary cylinder fs.
ie., the vortex shedding frequency predicted by the Strouhal number relationship, becomes
equal to f
N
. At this point, conditions become favourable for resonance to occur between fs
and fN, giving the potential for cross-flow response and lock-in, in the case of which a new
stable energy balance would be established. In the absence of other influences, this does in
fact occur, as shown for example by Figure 6.2. In the present case however, the span is
already established in a stable response regime, and the results of Figure 6.5 suggest that it
is reluctant to break out of the existing stable behaviour pattern despite the sudden
appearance of the potential alternative response regime. Lines have been drawn on Figure
6.5 corresponding to the theoretical Strouhal number relationship for Strouhal numbers of
0.28 and 0.30, which are representative of the typical Strouhal number values measured for
the smooth span. Examination of these lines shows that correspondence between fs and fN
should take place at VR values between 3 and 4, as opposed to 5 when the Strouhal number
takes it sub-critical value of 0.2. It can be seen however, that the wake vortex shedding
frequency, fw, remains locked in at 0.5 fN well past the point where the stationary cylinder
shedding frequency, f" and fN should coincide.
As described in Section 3, lock-in of any type cannot continue for an indefinitely increasing
flow, since the very strong vortices necessary to depress the Strouhal number and maintain
lock-in, become prone to instability and corresponding sudden breakdown of the lock-in
mechanism. Ultimately therefore, the stability of the in-line lock-in mechanism becomes
insufficient to survive the disruptive effects of the transverse vortex induced pressure forces
tending to initiate cross-flow motion plus any random environmentally induced forces, and
the in-line lock-in abruptly ceases. The span must naturally find a new equilibrium regime
in which energy input is equal to energy dissipation, and this is established in the cross-
flow direction, where the new 19ck-in mechanism takes over. Once again. the span motion
is able to control the vortex shedding process and preserve sychronisation between
excitation and response frequencies.
If the flow velocity is subsequently reduced, a similar train of events occurs, in which the
span motion remains locked into the stable cross-flow r ~ s p o n s e regime, despite the presence
of the alternative in-line regime which is shown to also be possible at the same VR value.
For the case of decreasing flow velocity, the theoretical lower limit of lock-in is reached
when the locked in fw reaches correspondence with its-stati,onary span fs value. As explained
in detail in Section 3, lock-in at higher velocities is possible because of the inverse
relationship between vortex strength and Strouhal number. The increased vortex strength
induced by the span motion in these cases depresses the Strouhal number below its
stationary cylinder value and mruntains fw nominaijy equal to fN despite increases in flow
velocity. There is no mechanism however, whereby the vortex strength can be consistently
depressed below its stationary cylinder value in order to increase the Strouhal number and
86
Contents
allow lock-in to continue for lower velocities. This theoretical lower limit of lock-in is
found to be confined in this case, with cross-flow lock-in ceasing when fw reduces to its
stationary Strouhal value, f" ie., where the fwlfN = 1.0 line intersects the Strouhal number
relationship line. It is possible that large amplitude span response could continue below
this limit, but as forced rather than synchronised interaction. The disrupted vortex shedding
in the initial Reynolds number region appears too weak to induce this effect here, however.
In applying the above results to determine practical onset thresholds for cross-flow
vibration, questions arise as to where the threshold should be taken. Firstly, the hysteresis
behaviour creates ambiguity by showing that cross-flow lock-in may exist at velocities
lower than its initial onset threshold. Secondly, the onset threshold appears to be span
length dependent, with onset occurring earlier for the 40m than the 50m span (this
observation was confirmed by other tests also).
Consideration of the hysteresis behaviour rapidly leads to the conclusion that it must be the
lower limit of the hysteresis range which is taken as the effective cross-flow lock-in
threshold. Although in the case shown, the lock-in does not initiate itself at this lower
limit, the tests show that any existing locked-in response may sustain itself once this limit
is exceeded. The risk then exists that in a practical situation, external effects such as wave
action may disrupt the stability of the in-line locked in response and prematurely initiate
the cross-flow lock-in.
The relative stability of the two lock-in regimes appears to be sensitive to extraneous
influences, as shown by the span length dependence demonstrated by the upper hysteresis
limit (on set threshold). The earlier onset of cross-flow lock-in for the 40 m span suggests
that the net cross-flow effect is more powerful for this span, andlor the in-line effect is less
stable. In both cases however, the potential for cross-flow lock-in exists once the lower
limit of the hysteresis range is exceeded.
The frequency response of the span to the above lock-in mechanisms can be detefl1.lined
from parts c) and d) of Figure 6.5. It can be seen that throughout the changing regimes of
vortex shedding, the span simply responds at an approximately constant frequency,
nominally equal to the still water natural frequency, fN It is known from the stationary
span tests, that the vortex shedding is generally weak and incoherent within the critical
Reynolds number regime associated with the smooth span. The excitation forces generated
by this type of shedding are therefore also weak, and act in a randomised manner which is
incapable of forcing the' span !\way from the preferred natural frequency. The vortex
shedding strength is admittedly substantially increased during the lock-in phases, however,
these only occur when the vortex is able to interact constructively with and
become controlled by the span hence the span vibration is still not pulled away
from its natural frequency. The general' conclusion for the smooth span tests, is that due to
the weak vortex shedding, the is the dominant influence within the system.
Having determined the basic principles of locked in span response behaviour from the two
example cases shown in Figure 6.5, ii is necessary to examine the wider set of full scale
test data to determine the detail variations which exist at different values of bed gap ratio,
OlD. The full set of smooth pipe wake frequency response data for 40 m and 50 m spans,
and OlD values ranging from 2.0 to 0.0: is shown in Figure 6.6.
In general, for OlD values of 1.0 and greater, the behaviour patterns described above appear
to be substantiated by the additional test results. In some of the 50 m span tests, the tidal
velocity had not sufficiently reduced at the end of the test run to allow the response to drop
back from the cross-flow to the in-line behaviour. The lower limit of the hysteresis range
was not therefore explicitly indicated in these cases. The lack of control over environmental
test conditions is of course one of the technical difficulties associated with field test work.
For the tests at low OlD values, analysis of the results becomes more complicated, due to
the presence of a number of apparently anomalous points, indicating the presence of wake
vortex shedding close to the span natural frequency, but as much lower VR values than
those associated with cross-flow response. This effect is predominant at low OlD values,
but isolated points do occur for some of the higher OlD tests. At first sight, these points
appear to contradict the physical behaviour arguments presented above, but upon closer
examination can be reconciled with a realistic interpretation of the situation.
87
Contents
OlD = 2,0 OlD = 15 OlD = j-O
00
00
::J
I ~
:::J :: J
x

5'0-30\
030\
II
0'0\
"
0
z
'"
n
n
1'4..( // \
L 0,28 \.. 028
'"
'" I'
"
'"
'" J>
J>
5; 028
'"
'"
"
,.,
"
Z
Z
'" <II
'"
3 '0 '0 '0 <
<:
,. ,.
'" 0
,.
'"
r
a
r
0
'0 08 '0 08
fo
0'
0
n
:T
x n
"
"
-<
-<
...
00 00 N

~
0 o. 0.
..
O'l[
o , ~
0'
...
c
0

o i
i
o i
~ "TI ,
3

,
0 7
, ,

, ,
7
,

0 7
n -
,<U>
C
V. V. v
~
:

...
..
0 en
~


...
~
G/D = 20 G/D,; j5 G/D = 15
,., ,., ,.,
0' 5-030 030

!!1:
'0
,0
,.,
..
~ , .. , ..
\.. 0-28
'"
500-28
0
~
.,
x ,t
~
,.,
.... .-.:. "
,
~ ~ ) ( ..
c
r
r
'0
10 10

'"
U
f. f. r c
n
'0
08 In 0' In 0'
O-u
J> ,. -u
r ",,,,
'"
'" '" ...
08 00 N
'"

'" "''' ...
O'
o. o. 00
'"
3 3
...
c
0' 0' 0'
"'''' ~
-u-u
J> J>
:u
z z
'"
0 0 0
"
'"

, ,
2 3

, ,
7 2
,

0 7
~
V. V. v,
Contents
~
6
0
,
"
~
...
6
0
,
"
o
" o
,
"
a
~
6
~

a
~
a
~ ~
~
N
"

~
N
a


\".
I.

N a ~
"
!IE
N a

x
N a
~
a
~
6
~
a

6

"

a

a
"
a
N
6
a
N a
6
...
~
~


>
~
...
~
~

>

N


x

'I
0
0
a
~
0
a
,
"
~ ~

N
~
N
0

~
0
0
a ,
~
"
a
~

N

a
a
~
N
6
~
"
! l ~
~
a

"



6

a
o.
a
"
a
N
6
a
--.------_ ... -.---
o
Figure 6.6 continued
~ x ~
N
a

XX
x,,
N


a

~ ~
6 6

6
N
a
a
Smooth epan .. frequency response plots for all tests

~

-ff
~
N
...
~


>
o
N
89
Contents
G/D= 0-5
's
030
"6
'4
0-Z8
"
'0
1..'<
'0 OS
06
04
,
0'
0
4 6 7
V.
Figure 6.6 continued
Smooth span . frequency response plots for all tests
As discussed earlier, the first occurrence of in-line span response is associated with a wake
flow regime in which the vortices tend to be shed in symmetrical pairs, rather than
alternately. Furthermore, the tests of Wootton et al (ref. 67) carried out on full scale marine
piles, showed that the overall wake behaviour in this region does not necessarily fall into
the simplest and most obvious behaviour patterns. Wootton et al pointed out that pure
symmetrical vortex shedding alone could not explain the multiple-peaked wake spectra
which they measured, and they commented in some detail upon alternative mechanisms.
Amongst their suggested possibilities were combinations of interspersed alternate and
symmetrical shedding, the occurrence of alternate and symmetrical shedding in blocks of
each type, and a mechanism in which symmetrical vortex shedding is superimposed upon a
longer period general wake instability.
Some comment on the physical mechanisms responsible for the first in-line response range
is given by King et al (ref. 13) who performed tests on the vortex excitation of model piles
in water. With reference to the two types of in-line excitation mechanism, they state: "The
first of these was an apparently selective mechanism in which the pile responded at a fixed
frequency to a range of input frequencies which it only partly controlled". For the full scale
pipe span tests described here, a range of wake frequencies clearly existed, as illustrated for
example by the typical wake frequency spectra shown in Figure 6.7. These spectra were
measured at respective VR values of 1.4, where the response should be of the first in-line
type, and 2.09, where it would be expected that the first and second in-line response regions
may undergo some degree of overlap. It can be seen that two main spectral peaks occur,
nominally at 0.5 fN aod fN. In such cases where multiple spectral peaks occur, it is perhaps
misleading to imply the existence of a unique wake frequency by picking off the value of
the higher peak only. In contrast, Figure 6.8 shows a wake spectrum measured by the same
current meter within the same test, but during the subsequent dominant reglon of cross-flow
lock-in. In this case there is no ambiguity, with a single dominant spectral peak present at
the pipe response frequency (nominally equal to fN).
In laboratory scale experiments, it would be normal practice to resolve the uncertainty by
performing flow visualisation tests. Unfortunately this was not possible in the field
conditions of the Severn Estuary, so, as in the full scale marine pile tests of Wootton et al,
(ref. 67), some uncertainty remains as to the exact excitation mechanism prevalent during
the fIrst in-line response region. It appears reasonable to speculate along similar lines to
Wootton et al, that the excitation mechanism is some combination of symmetrical and
alternate vortex shedding, with the further possibility that the symmetrical vortex shedding
may be superimposed on a generalised wake instability whose frequency is not directly
related to that of the shedding itself.
90
Contents
Further possibilities which must be considered of course, are that some of the spectral
energy input is derived from sources other than pure vortex shedding behaviour. For
example, it is possible that some cross-wake interference is occurring, allowing the reading
from a current meter at one side of the wake to be corrupted by effects from the other side.
A further possibility, especially at low velocities, is that the motion of the pipe is simply
generating associated pulsed distnrbances within the flow, which are being picked up by the
current meters, and registering as a spectral energy input. These possible extraneous effects
are shown in Figure 6.9, along with the expected influences of alternate and symmetrical
shedding upon the wake current meters.
Contents
POWER
SPECTRAL
DENSITY
MEASURED SPAN
RESPONSE FREQUENCY
DURING LOCK _ IN
o f = = = ; = = = ~ ~ = 7 ~ ~ ~ - = ~ = = ~ ~
o 01 02 0-3 04 0-5 06
'.(H
Z
)
Figure 6.8
Wake frequency spectrum during cross-flow response
6.2.6 Amplitude Response - Smooth Span
The above discussion of the frequency response of the test span provides useful physical
insight into the behaviour patterns, and shows consistency with theoretical arguments.
From the practical engineering point of view however, it is generally the respollse
amplitude for a given set of flow conditions which is the most relevant parameter. The
measured response amplitudes from the full scale test programme are therefore discussed in
detail below.
Figure 6.10 illustrates two typical response amplitude plots, which are used as examples in
describing the span behaviour. These plots show results from the same test runs as the
example frequency response plots shown in Figure 6.5. The amplitude plots show root
mean square (RMS) response amplitude plotted against reduced velocity, YR. The two
examples are used to illustrate the general qualitative features of the response amplitude
behaviour, and identify a generalised behaviour pattern. The full set of test results are then
presented, and variations in behaviour which occur at various GID values are addressed in
more specific detail.
As discussed above, the smooth full scale pipe span lay in the critical Reynolds number
range, where the vortex shedding Strouhal number was subject to significant elevation
above its subcritical value of 0.2. This was shown to impose a significant influence upon
the frequency lock-in behaviour of the test span. Not surprisingly therefore, the amplitude
response behaviour also exhibits marked differences when compared with the 'conventional'
response amplitude plot shown in Figure 6.1.
As the reduced velocity, V R, is increased from zero, the test span shows no significant
response for VR values less than approximately 1.0. Above this value however, span
motion begins to become apparent, which is consistent with the conventional behaviour
shown in Figure 6.1. At this initial response threshold, the vibration is expected to be
predominantly in-line and associated with a symmetrical type of vortex shedding which is
dependent upon the relationship between the span motion and the flow velocity rather than
the Strouhal number (see Section 3.3.2). Providing therefore, that the required conditions
on the span vibration velocity and amplitude are met, the initial response threshold should
be largely independent of Reynolds number effects, and tltis appears to be substantiated by
the experimental results.
92
Contents
FLOW
o
o}
r
Figure 6.9
0) CONVENTIONAL
ALTERNATE VORTEX
SHEDDING
b) CONVENTIONAL
ALTERNATE VORTEX
SHEDDING WITH CROSS-
WAKE INTERFERENCE
c) INCOHERENT VORTEX
SHEDDING SPAN MOTION
INDUCES PULSED
DISTURBANCES IN FLOW,
WHICH ARE DETECTED
By WAKE CURRENT
METERS.
d) SYMMETRICAL VORTEX
SHEDDING.
KEY
~ CURRENT METER
. LOCATIONS
Possible Interaction mechanisms between span wake and current meters
93
Contents
The motion in this initial response region is generally assumed to take place in the in-line
plane. Indeed, some model experimental arrangements have been set up to incoIporate
physical restraints to constrain the motion to one plane only (see eg., King, ref. 68). The
full scale test span however, was free to vibrate in any plane, and it can be seen from
Figure 6.10 that the in-line response is accompanied by smaller amplitude, but still
significant, cross-flow component. The physical reason for this is immediately apparent
from Figure 6.11, which shows plots of the motion of the 40 m span mid point at various
VR values. Figure 6.11 a) shows the span motion in the initial response region (V R = 1.3),
and it can be seen that the motion takes the fonn of <1!l ellipse.
Previous research data, as summarised for example in Figure 6.1 (from ref. 65), indicates
that the first response region will die out again if V R is increased to between 2.0 and 2.5.
The second response region is then expected to commence at a V R value between 2.5 and
3.0. This second response region however, is induced by an alternate vortex shedding
mechanism dependent upon the Strouhal number, S, and it has been shown that S for the
smooth pipe tests is significantly higher than the commonly assumed subcritical value of
0.2. The onset of the second response region would therefore be expected to occur earlier
than shown in Figure 6.1. Theoretically, this type of response should commence when f, is
nominally equal to 0.5 fN, and taking S to be in the range 0.28 to 0.3, this gives an onset
value of V R in the range 1.7 to 1.8, which is near the expected peak of the first response
region. In this case therefore, the first and second response regions do not occur
independently, but merge into each other as a result of the influence of the increased
Strouhal number on the second response region.
For both examples shown in Figure 6.10, the in-line motion appears to reach peak
amplitude at a V R value of approximately 3.5, and is accompanied by a significant
component in the cross-flow direction also. Figure 6.11 b) shows the corresponding span
mid-point motion, and a complex motion pattern is apparent. The span 'appears to vibrate
in an elongated ellipse, which constantly varies in orientation to fOIDl a 'rosette' pattern
motion path. Upon consideration of the physical characteristics of the test arrangement, it
is not surprising that the span motion is subject to wide variation from the idealised single
plane in-line vibration. The span is a slender and hence relatively flexible structure, and is
situated in a natural boundary layer flow with large scale turbulence present. It is inevitable
therefore, that random excitation forces will be superimposed on the periodically varying
vortex induced in-line force fluctuations, and in view of its inherent flexibility the span will
respond with a random as well as a basic in-line vibration component. The superposition of
the latter effects leads to the type of response motion shown in Figure 6.11 b).
From the practical span assessment point of view, the most significant parameter is
generally the maximum amplitude of the span response. In this respect, the full scale tests
reveal a significant increase in in-line response amplitude over the commonly suggested
'limiting' value of 0.20. Figure 6.11 shows in fact, that amplitudes of up to the order of
0.45D occur within the predominantly in-line response region.
A possible explanation for this increased response amplitude could be that the energetic
boundary layer turbulence of the approach flow may be inputting energy at or near the span
natural frequency. Extensive examination of spectral data derived from measurements in the
approach flow however, revealed negligible energy input near the natural frequencies for
either 50m or 40m spans. Furthermore, the fact that the roughened span failed to display
increased in-line response amplitude, despite being exposed to the same approach flow,
suggests that some other cause unique to the smooth span is responsible for the effect.
A more probable cause, which would be expecteo to affect the smooth span only, is the
potential for the hydrodynamic damping losses to be reduced, as a result of the drop in the
steady drag coefficient, Co, which occurs within the critical flow region. The critical flow
phenomenon is discussed at some length in Section 4.2, within the context of its effect
upon the vortex shedding Strouhal number. A physical explanation is given in the latter
section, describing the critical transition to turbulence in the pipe surface boundary layer,
and its subsequent effect on flow separation points, wake width, drag coefficient and vortex
shedding behaviour. The connection between the drag coefficient and the hydrodynamic
damping of the in-line response is then specifically discussed in Section 5.2.
94
Contents
26
2'
22
20
18
E
1

(II
", 1
. \ \
3

0
!a
"
,.


" ,
:5 12
Q.
Ul
"
0
0-
10
!!. ...
Ul
co
.""
" C
" 8
,,-
a:
'\
, .
1
/V
'"
I-
:..
6
1 .,
0
1
I
I/Iit I
0
i
,
/

:

.r
i
3

-g,
2
[
5 6

0
3 4 2
"
0
0"
;;
REDUCED VELOCITY, V
R
0) 50 m SMOOTH SPAN G/D I 5
::::J VERTICAL (CROSS FLOW;
'"
A HORIZONTAL (IN-LINE) RESPONSE
."
"
HORlZONTAL (IN.lINE) RESPONSE (SECOND MODE)
7 8

E
u
26
2'
22
20
18
- 16
>-
z
w
:E 14 '
w I
u I .i
G I f
12 :Ji' I
o -"Ii
10 ,/ \ \
Ul I
" / I I
1 I
6 II / I I
'" It I
a: 8 1/"! i l
1 I
RESIDUAL RESPONSE I 4 I II' '. /
FROM NATURAL -t . l JI>
FREQUENCY TEST
2
2 3

5 6
REDUCED VELOCITY, VR
b) 40m SMOOTH SPAN GID 20
7 8
Contents
018,-------------------------,
016
014
E
z
012
El
0
I-
U
W
...J
0
"-
o 10 w
0
z 0
'"
El
"-
UJ
008 El
0
El
::!O 0
El
z
0
El
'" w
::!O
006 El
(!)
El
004 0 El
0
El
0
002
0 El
. ~
El
"'
0
0 02 OA 06 08 10 12 IA
_2
U (m Is)
o G/D = 10 ROUGH SPAN 40m
El G / D = 15 SMOOTH SPAN 40 m
FROM FULL SCALE TEST STUDY (REF 62)
Figure 6.12
Span deflection due to steady drag action
98
Contents
15
10
05
ACHENBACH a HEIN'ECKE (REF 58)
kID = 0'75x 10-3
-----L " ........
ACHENBACH a
"
HEINECKE (REF
kiD = 90 x 10-3 \.\ . .-: ____ _
\ " ......
SMOOTH TEST SPAN
kiD = ],0 II 10- 3
El
\
....... \
_ ..... -' \
El
El
ROUGH TEST SPAN
kID = 85 J. 10-3
SPAN ASSUMED SPAN MOTION
EFFECTIVELY ..... .. ASSUMED TO
STATIONARY INFLUENCE CD
__
10' 2 3 4 5 6 7 8 9 105 2 3. 5 6 7 8 9
10
6
Re
FROM FULL SCALE TEST STUDY (REF 62)
Figure 6.13
Steady drag coefficient for smooth and rough test spans
99
Contents
As shown by the frequency response results discussed above, an abrupt transition occurs in
the span behaviour as the response jumps from the predominantly in-line lock-in where
fw = 0.5 fN, to the predominantly cross-flow lock-in where fw = f
N
. This jump is reflected
in the amplitude response results, with an abrupt increase in cross-flow amplitude
accompanied simultaneously by a reduction in the in-line amplitude, as cross-flow lock-in
commences. Figure 6.11 c) shows the span motion immediately prior to the
commencement of cross-flow lock-in, and displays an interesting motion pattern
resembling a much distorted 'figure of 8'. This pattern results from the superposition of
two distinct components of response, where the nominally in-line component possesses
twice the frequency of the component in the nominally cross-flow plane. This is consistent
with the relationship between the drag and lift force fluctuations resulting from alternative
vortex shedding. It can also be deduced that at least one of the components of response must
be a forced effect, taking place at a frequency not equal to the natural frequency of the span.
This double frequency vibration may be part of the mechanism which breaks down the
stability of the predominantly in-line lock-in regime and allows the cross-flow regime to
take over. The brief occurrence of this double frequency motion is also shown in Figure
6.S d).
The cross-flow response is immediately apparent due to its greatly increased amplitude.
Maximum amplitude is of the order of 1.00, compared with the 'limiting' value of 2.00
suggested by previous research. The hysteresis behaviour apparent from the frequency
response plots is reflected by the amplitude response results also. Figure 6.11 d) shows a
span motion plot during the cross-flow lock-in phase (during the decreasing phase of the
hysteresis cycle), demonstrating a coherent response predominantly in the cross-flow plane.
The orientation of the response plane appears in this case to be more stable than for the
predominantly in-line response.
For the SOm span example, the subsequent decay in cross-flow amplitude with further
increasing V R is clearly shown. Ii was in fact observed during the tests, that the SOm span
displayed a tendency to begin vibrating in a second 'double bowstring' mode at V. values in
the region of 6 and greater.
The end of cross-flow lock-in during the period of decreasing flow is equally abrupt as its
initial onset, with the reduction in cross-flow amplitude being accompanied by a
. corresponding increase in the in-line component. For the 40m span example (Figure 6.10
b), the amplitudes return almost exactly to the paths they followed during the increasing
flow phase. There is insufficient data to determine whether the SOm span will stabilise to a
similar behaviour path however. The span motion following the cessation of cross-flow
lock-in is shown in Figure 6.11 e) where the pattern of generally elliptical motion with a
varying axis of orientation is visible, as previously observed during the increaSing flow
phase.
6.2.7 Special Full Scale Study - Rough Pipe Tests
For the rough pipe test programme, the surface finish of a typical concrete coated pipeline
was replicated by attaching moulded glass fibre roughness panels to the test span. The
mould from which these panels were produced was manufactured using an actual length of
concrete coated pipe as a pattern, and resulting mean roughness height was of the order of
4.Smm.
The relative roughness height kID, was therefore of the order of 8.S x 10'3, where the slight
increase in D due to the thickness of the panels themselves was taken into account. A small
amount of the rough span test data has already been presented for comparative purposes in
the context of the smooth span discussion above, however a more complete series of
results, together with a discussion of the rough span behaviour, is given below.
100
Contents
!I.!
a
c
::r
!!.
"
c
3
.,.
!l
iil
..
50
;;;
-
a s
3
'" .,.."
:;' cD'
n c
::r ;;;
'O _ ...
'O.
m ~
..
!.
;:
..
3'
c
;;
l>
Q,
n
0
"
n
iil
l>
..
c
~
iil'
n
"
0
0' o
02
01
00 0
<DE
o 00
o
00
o
o 00
o 0 0
o 1 I
10
5
2 3 4'" 5 6 7 8 9 10
6
Re
GID ' 0 5
SO269
s
0'
02
01
o
oC co 0 a:>O 0
CD 0 (Jl 0
000
o 1 , 1
10
5
2 3 4 5 6 7 B 9 10
6
Re
GID' I 0
5'0225
RELATIVE ROUGHNESS kiD' 8 5,,0-'
0<30
o 00
o OCDO<>O
o <00 aD
020 o o
s
010
01 .
10
5
2 3 4 5 6 7 8 9 1 ~ 6
Re
G/D 20
5 0215
FROM FULL SCALE TEST STUDY (REF.62)
Contents
6.2.8 Strouhal Number - Rough Span
Figure 6.14 shows a plot of measured Strouhal number as a function of Reynolds number
for the rough span tests. The full significance of these results is best observed by
comparing the plot with Figure 6.4, which shows the equivalent results. obtained for the
smooth' span. Whereas the smooth span results displayed considerable scatter, and gave
generally higher values than would normally be expected, the rough span displayed vortex
shedding behaviour of a far more predictable and consistent nature. At GID = 2.0, the mean
Strouhal number is 0.215. For the tests closer to the seabed, where the boundary proximity
effect becomes significant. the measured mean Strouhal number values are 0.225 and 0.269
respectively. The effects of bed proximity are discussed in greater detail below.
As a comparison, Figures 4.8 and 4.10 show experimentally measured Strouhal numbers
from Buresti (ref. 47) and Achenbach and Heinecke (ref. 58), for a range of relative
roughness (kID) values. Both the latter references tested kID values closely similar to the
full scale test value of 8.5 x 10-
3
, and a reasonable qualitative comparison may be obtained.
For a kID of 7.35 x 10' (ie .. slightly less than the full scale test span), Buresli found that
the post critical Strouhal number stabilised at approximately 0.22. Achenbach and Heinecke
found that for a k,1D of 9.0 x 10.
3
, the post critical Strouhal number stabilised at
approximately 0.25. For both the latter cases, the transition to post critical conditions was
complete at a Reynolds number in the region of 7 x 10'. It should be noted that unlike the
full scale results, which were obtained from a test span situated in a highly turbulent
natural seabed boundary layer, the results quoted in references 47 and 58 were both obtained
from laboratory measurements in low turbulence wind tunnels, and are therefore free of any
significant turbulence wind tunnels, and are therefore free of any significant turbulence or
bed proximity induced effects. It is also relevant to note that there are inherent uncertainties
associated with the quantification of surface roughness. It is therefore not certain that an
exact quantitative comparison between the above sets of results is valid. despite the
nominal similarity of the relative roughness.
The physical effect of the full scale pipe surface roughness in stabilising the vortex
shedding, can be appreciated by examining Figure 6.15, which shows examples of velocity
spectra measured in the wake of the stationary pipe at GID = 1.0, where the bed proximity
effect is moderate. For comparison purposes, equivalent spectra at a closely similar
Reynolds number for the smooth pipe test at GID = 1.0, are also shown. It is immediately
apparent from the spectra shown. that the roughened pipe wake spectral energy is
concentrated into a narrow frequency band, whereas that for a smooth span is much less
coherent and spread across a wide band of frequencies.
It can be seen from Figure 6.14 a) that for the relatively low GID value of 0.5, the
measured Strouhal number values begin to show increased scatter compared to those at
higher GID values. This effect can be better appreciated by examining the wake spectra for
the GID = 0.5 tests, shown in Figure 6.16, and comparing them with equivalent spectra for
a test at higher GID shown in Figure 6.15 a).
It is immediately obvious from a comparison of the two sets of rough pipe spectra, that the
vortex shedding nearthe bed at GID = 0.5. is considerably less coherent than the shedding
further out from the bed at GID = 1.0. This is consistent with the findings of special model
test study described in reference 87, which showed similar disrupted and broad band shedding
for a model pipeline span at a similar GID value in the highly turbulent boundary layer
flow over a rough surface. Close to the seabed therefore, the disruptive effect of the very
energetic boundary layer turbulence tends to counteract the stabilising effect of the pipe
surface roughness upon the vortex shedding. There is therefore reason to expect that the
dynamic behaviour of the span at low GID values may differ from the behaviour further
away from the bed, due to the differences in the characteristics of the vortex-induced
excitation mechanism.
102
Contents
,..
t:
,..
(/)
R
e
=3IOxI0
5

006
1
z
Re = 316 x 10
5
lLJ 02 Z
0
0-05
...J
"
HORIZONTAL ci 004
0:
I- VELOCITY a:
u
COMPONENT t; 003
w
01
0- W
::;
(/) 0-
m
If) 0,02
,..
a:
a:
CD
w
0-01
-
" ;; 0 0
.., 0-
0
0-
0
t::
0 01 0-2 0-3 0-0 05 06 0 01 0-2 03 0-' 05 06
CD
"
0
f s (H ZJ f s (H z)
...
co
..,
CD
0

il
-" o _"
co >-
In
I-
>-
ar CD
iii I-
=. CD 0-3
iii
o " 0
Z
"
w
.. '"
0
-<
...J
VERTICAL
..J 001
"
'"
a:
02
VELOCITY
"
I- a:
?
u
COMPONENT
I-
w U
"
0- W
0 (/) 0-
".
01
(/)
0. a:
a:
;;"
w
w
3
"
" CD
0 0
iD
0-
0
0-
0
0 01 02 03 00 05 0-6
0 01 02 03 00 a 5 0-6
co
..,
f s (H z) f5 (H z) ..
"
oj SPAN WITH SIMULATED CONCRETE ROUGHNESS bJ UNCOATED STEEL SPAN kiD 10- 3


0
w
FROM FULL SCALE TEST STUDY (REF 62)
Contents
>
t::
<II
Z
'"
0
.;.J
" 0:
....
<..>
'"
Q.
<Il
0:
'" ~
0
Q.
>
....
<Il
Z
'"
0
..J
"
0:
....
<..>
'"
Q.
<Il
0:
w
~
0
Q.
104
009
008
0:07
006
005
004
003
002
001
0
0 01 02 03 0-4 05 OS
fS {Hz}
HORIZONTAL VELOCITY COMPONENT
03
02
01
O + - - - - - ~ - - - - - r - - - - ~ ~ - - - - r _ - - - - ~ - - - - ~
o 01 02
VERTICAL VELOCITY
G/D = 05
kiD = 85 x 10-
3
'
03 04
fS (Hz)
COMPONENT
Re = 331x 10
5
05 OS
FROM FULL SCALE TEST STUDY (REF.S2)
Figure 6.16
Wake frequency spectra for stationary rough span at low G/D
Contents
6.2.9 Frequency Response - Rough Span
The results of the stationary span Strouhal number tests show that significant differences in
vortex shedding behaviour exist between the rough and smooth test spans. These differences
would be expected to be reflected in the dynamic behaviour of the rough span compared to
that of the smooth. A further characteristic displayed by the stationary rough span test
results, is a variation in vortex shedding behaviour and Strouhal number with decreasing
gap ratio, GID. As a result of this influence of GID on the vortex shedding mechanism, it
is to be expected that the vortex induced dynamic behaviour of the rough span will also
show a GID dependence.
Typical examples of the rough span wake frequency response are depicted in Figure 6.17 a)
and b), which shows dominant wake frequency versus reduced velocity plots for GID values
of a) 1.5 and b) 0.5. Part a) of this figure (G/D = 1.5) is representative of the behaviour in
the region of strong and coherent shedding remote from the immediate bed proximity
influence, whereas part b) (GID = 0.5) reflects the influence of the more disrupted vortex
shedding in the region of energetic approach flow boundary layer turbulence near the seabed.
Parts c) and d) of Figure 6.17 show similar response graphs, but with the span response
frequency, f" plotted instead of fw.
An increase in vortex shedding strength naturally increases the associated excitation forces,
which are thus able to impose a more dominant influence upon the response of the span.
The resulting span response is therefore more complex. Although a lock-in region occurs,
where the span response controls the vortex shedding, it is also observed that the vortex
shedding is powerful enough at certain velocities to force the response away from the
natural frequency.
If Figure 6.17 a) is examined, it can be seen that the initial wake behaviour is similar to
that of thesmooth span tests. At low YR. where the span motion is very limited, the wake
frequency follows the stationary Strouhal number relationship. The convention in-line lock-
in region then begins when the Strouhal frequency, f, becomes nominally equal to 0.5 fN,
and the vortex shedding remains controlled at this frequency by the span motion over a
further range of.increasing YR. For the smooth span tests, this in-line lock-in region tends
to end abruptly as cross-flow lock-in begins and fw jumps abruptly to nominally equal fN
This latter behaviour is not observed in Figure 6.17 a) however, since at the end of the
lock-in region (at VR == 2.8), fw simply begins to progressively increase.
The behaviour of the rough span is complex, as can be seen by examining the span
response plot in Figure 6.17 c). For a better appreciation of the physical details of the
phenomenon this should be examined in conjunction with the span motion plots in Figure
6.18. At V R values less than 1.0, the amplitude of the span motion is negligible. Above
this value however, the expected in-line response begins to build-up, with the span
vibrating at slightly below its natural frequency. This in-line response at VR = 1.8 is shown
in Figure 6.18 a). If this is compared with the approximately equivalent plot for the
smooth span shown in Figure 6.11 a), it can be seen that the rough span motion is
constrained more closely to the in-line plane than the more circular orbit of the smooth
span. The span response frequency, fr, is approximately 0.9 fN at this point.
The next stages in the response build-up differ greatly from those for the smooth span. At a
V
R
value approaching 3, the frequencies of the in-line and cross-flow response'components
abruptly diverge and take up relative ratio of 2: I. This is reflected in the 'figure of eight'
shape of the span motion shown in Figure 6.18 b) for a VR of 3.1. At this point, the in-
line f, is approximately equal to fN, hence the cross flow f, is half this value and equal to f
w
.
In other words, the cross-flow response is a forced vibration in which the pipe response is
dominated by the vortex shedding.
The above described behaviour is in marked contrast to that of the smooth span. Whereas in
the latter case, the span response frequency appeared to dominate and the vortex shedding
was subject to abrupt changes to remain in sychronisation, a totally different effect occurs
for the rough span with its considerably more powerful vortex shedding. In the rough span
case the vortex shedding has a more active influence, and it is the span response which is
subject to abrupt frequency jumps in order to remain in synchronisation. The vortex
shedding is not totally dominant however, since it is subject to synchronisation with the
span response during the in-line lock-in region ..
105
Contents
-0
141 1-4
'"
s = 022
0
00
0
1-2 1-2
dl
KEY PLAN - a) AND b)
~
::c
0 VELOCITY INCREASING
0 -_ '0
c
1-0 VELOCITY DECREASING 1-0
~ . ~
<D
".
~
"
Iw 0-8
I,
0-8
m
~
1 "
IN IN
o '\ J.
~
0-6 0-6
,"
"
'V + ~ + l
n
*,.
.. +
!!.
0-4 0-4

~
m
,..
CD !!
0:1L
01 ROUGH SPAN clROUGH SPAN
02
m<D
~ c G/DI-5 G/OI-5
Co ~
m
~
0
" ..
m :..
0 2 3 4 5 6 WAKE FREQUENCY SPAN RESPONSE 0 2 3 4 5 6 7
~ ...
ii
VR PLOTS FREQUENCY PLOTS VR
~
" 0
1-4 1-4
~
~ S = 0269
m
/
- 1-2
ii
12
.Q KEY PLAN -c) AND d)
c
IN CROSS m
~
1-0 LINE FLOW 1-0 n
'<
1
0 VELOCITY INCREASING
' ~ j # ' .
"
G + VELOCITY DECREASING J000
~
Iw 08 0-8
I,
IN
IN J
0
0
0-6 0-6
0
,+
/
OA.....Jllll
+
0-4J
' "
, ~
0-2
b) ROUGH SPAN d) ROUGH SPAN 0-2
G/D 0-5 G/O 0-5
ROUGHNESS kiD = 85x 10-
3
0
0 2 3 4 5 6 7
RELATIVE
0 2 3 4 5 6 7
VR
FROM FULL SCALE TEST STUOY (REF.62) VR
Contents
FROM FULL SCALE TEST STUDY (REF 62)
Figure 6.18
Plots of span mid-point motion rough span G/O = 1.5
107
Contents
For further increasing Y. values, where f. shows a steady increase, f, continues to be forced
by the vortex shedding. The cross-flow f, component is equal to f., and the in-line
component twice this value. From Figure 6.17 a) it can also be seen that fw is lower than
the Strouhal number predicted frequency, thus it can also be deduced that the span motion is
imposing some influence on the vortex shedding.
As VR reduced from its maximum value, fw exactly retraces its path from the increasing
flow phase. This contrasts with the smooth span f. behaviour, where a hysteresis effect is
observed. The rough span does show a hysteresis effect in the f, behaviour however, as seen
in Figure 6.17 c). The double frequency 'figure of eight' span response persists to a lower
Y. value than that required for its additional onset. This is probably because the energy
input required to maintain the latter response pattern is lower than that required to trigger it.
The frequency response behaviour described above, relates to the rough span at a GID of
1.5. Closer to the seabed however, where the span is exposed to additional physical effects
resulting from increased turbulence intensity and immediate bed proximity, the response
pattern is different. The differences in the response can be observed by examining Figure
6.17 b) and d), which relate to a test at a GID of 0.5. The span motion plots for the test are
shown in Figure 6.19.
Figure 6.17 a) shows wake response behaviour for the near bed rough span test. It can be
seen that the behaviour characteristics are markedly different from the GID = 1.5 rough span
results described above. The main difference is at the end of the in-line lock-in, f. displays
an abrupt jump to a higher frequency. This is qualitatively similar to the behaviour of the
smooth span, although a significant quantitative difference does exist. Whereas for the
smooth span, the jump occurs after the theoretical stationary span shedding frequency, f"
has become equal to fN the jump occurs at a lower frequency than this for the rough near-
bed span. The span response frequency, fr, following the jump, also takes place at a
frequency lower than f
N
These observations imply that the vortex shedding is sufficiently
more powerful than the smooth pipe shedding to influence the span response to some
extent, but lacks the ability to dominantly force-vibrate the span as for the higher Gill
tests. This is consistent with the inferences from the stationary span vortex shedding
measurements in Figures 6.4 and 6.14. The Strouhal number measurements (Figure 6.14)
for the rough span at GID ;;::;: 0.5 are less consistent than those at higher GID, but are still
considerably more coherent than the randomly scattered results from the smooth span
(Figure 6.4).
The physical behaviour of the near bed rough span can be clearly seen from the span
motion plots in Figure 6.19. It can be seen that the double frequency 'figure of eight'
response apparent in Figure 6.18, is totally absent. Additionally, the response at low Y.
values (up to 2.5) is extremely limited. The only significant response occurs in the cross-
flow phase, and this of an elliptical pattern similar to the cross-flow response of the
smooth span. As shown by figure 6.17 d), this response is at a frequency slightly below fN
The latter figure does also show a very erratic span frequency response at low V R values,
but in view of the almost negligible amplitude of the motion in this region, it is not
considered valid to draw any significant conclusions from the results. Such low amplitudes
of motion are likely to be close to the resolution limits of the instrumentation, and may
therefore lead t<;> unreliability of the measurements.
6.2.1 0 Amplitude Response - Rough Span
When compared to the smooth span, the stronger vortex shedding induced by the simulated
concrete roughness produces marked changes in the interactions between the excitation and
response mechanisms. These changes are reflected in the amplitude response behaviour of
the spans.
Figure 6.20 a) shows the amplitude response plot for a GID value of 1.5. The initial onset
of span response takes place at a V R value of approximately 1.0, which is consistent with
the smooth span behaviour.
108
Contents
b) VR = 2'5
e)VR= 2'3
FROM FULL SCALE TEST STUDY (REF. 62)
Figure 6.19
Plots of span mldpoint motion rough span G/D = 0.5
109
Contents
The build up of response then takes place predominantly in the in-line phase, as confirmed
by the initial low vertical response values in Figure 6.20 a), and the flat span motion
pattern shown in Figure 6.18 a). This contrasts with the smooth span which showed
significant cross-flow response also, in the so-called in-line response region. The maximum
amplitude of the in-line response is also lower for the rough span than the smooth, and
appears to remain below the 'limiting' amplitude of 0.2 D found in several previous
experimental studies. As mentioned earlier, the unexpectedly high in-line amplitudes found
with the smooth span are believed to be due to the very low drag coefficient which exists
under critical Reynolds number conditions. Figures 6.12 and 6.13 show however, that for
the rough span these low drag conditions do not exist, and the amplitude is therefore similar
to that found in sub-critical model tests.
The physical characteristics of the cross-flow response also differ from those found for the
smooth span. The first difference is that the build up of the response is more progressive,
in contrast to the abrupt jump in amplitude which generally occurs for the smooth span
tests. This behaviour is consistent with the increased strength of the vortex shedding, which
imposes the characteristics of a forced oscillation upon the span response. Conventional
forced vibration theory, as described for example by Timoshenko et al (ref. 88) and Stoker
(ref. 89), shows a similar progressive build up of the dynamic magnification factor, and
hence oscillation amplitude, as resonance conditions are approached.
The physical behaviour of the span during cross flow response is well illustrated by the
span motion plot shown in Figure 6.18 c), This behaviour differs greatly from the smooth
span behaviour, in that a very significant double frequency in-line response component is
associated with the cross-flow oscillation. This response pattern appears to indicate a large
drag overshoot at each vertical extreme of the span motion, and contrasts with the flat
elliptical cross-flow response pattern of the smooth span (see Figure 6.11 d).
A further difference from the smooth span behaviour is that the history-dependent hysteresis
effect is minimal. This is consistent with the presence of stronger vortex shedding, which
is able to actively force the span response, rather than being itself largely controlled by the
pre-existing span response characteristics.
The above discussion relates to the behaviour of the rough span at higher GID values (ie.,
G/D ~ 1.0). Nearer the seabed, where the increased turbulence intensity imposes a
disruptive effect upon the vortex shedding, the behaviour shows different characteristics.
This is demonstrated by the response of the rough span at a GID of 0.5, as illustrated in
Figure 6.20 b). Additionally, the pipe motion plots relating to this test are shown in
Figure 6.19, which gives an immediate qualitative impression of the behaviour difference
induced by the low GID value. It can be seen that the double frequency forced response is
absent, and the elliptical cross-flow motion pattern is qualitatively similar to that
demonstrated by the smooth span.
A further characteristic shown by Figure 6.19 and 6.20 b), is the lack of any significant in-
line response in the region prior to the onset of cross-flow motion. Significant span
motion is thus associated entirely with the latter type of response, although the inclined
axis of the span motion ellipse introduces an apparent in-line component.
The onset characteristics of the cros's-flow response are also different for low GID. Instead
of the more progressive build-up in amplitude demonstrated by the rough span at higher
GID values, the near bed span displays an abrupt annplitude jump, similar to the behaviour
of the smooth span. The former progressive amplitude build-up is attributed to a forced type
of response produced by strong vortex excitation. It is therefore reasonable to find that the
more disrupted near-bed shedding fails to give this forced response behaviour.
A complete set of rough span amplitude response plots for GID values between 2.0 and
0.25 is shown in Figure 6.21.
6.2.11 Polar Gas Project
This test series was carried out using a 13m length of 24-inch diameter pipe, mounted on
flexible suspension beams to simulate the flexibility of a much longer span. A
disadvantage of this test series was that the maximum response amplitude of the test span
was limited to small displacements by mechanical stops, in order to prolong the fatigue life
of the suspension beam system. This characteristic therefore limits the scope for direct
quantitative comparison with other work. J
110
Contents
26
24
22
18
~
u
16
I- ,
t
z
i
w
14
~
ft w
"
u
<!
12
,f
..J
0..
"
fJ)
It
0
10
fJ) If
~
II
'"
8
6
4
2
/
0
0 2 3 4 5 6 7 8
REDUCED VELOCITY
0) ROUGH SPAN G/D= '5
G VERTICAL (CROSS - FLOW) RESPONSE
& HORIZONTAL (IN-LINE) RESPONSE
Figure 6.20a
Rough span - typical span response amplitude plots
III
Contents
26
24
22
20
16
::;;
<.)
16
I-
z
LU
14 ::;;
LU
<.)
<I
..J 12
Il.
en
0
10
en
::;;
0:: B
6
4
2
0
0 2 3 456 7 6
REDUCED VELOCITY
b) ROUGH SPAN G/D: 05
FROM FULL SCALE TEST STUDY (REf 62)
Figure 6.20b
Rough span typical span response amplitude plots
112
Contents
G/0=2-0 G/0=2-0 G/D= I'D 6/0=1,0 G/O= 1-0
2t5 26 26 26 26 26
24 24 24 I 24 24 24'
22 12 22 22 22 P 22
2: 20 20 20 20 2'
18 18 18 18 18 18
16 16 I&- 16 16 16
14 f 14 I 14 "14 14 14
;12 f;,. ;12 ;12 f ;12
" It: 1I:!t.l. II: a 10 10 0 10 0 10 0 10 -:- 0 10
J:J c,.. ,
08 8 8 8 8 8
.. t .. 4 4" ,... i 4
cE 6 6 6 6 6:;1 6
,/U '
::::I 0 0 0 0 0
02345618 0 2: 45678 023456780 45678 0234567802345618
3 VR V
R
V
R
V R VA VR
'!!.

a.

.. 'II
. -
." CD G/O=0-75 G/O=0-75 G/D= 0'5 G/O=0-5 G/O= 0'25
o .

(1)02:6 26 26 26
; 24 24' 24 I 2'" 24
0.... 22 22 22 22 22
"
'" 20 20 20 20 20
18 18 18 18 18
< Hi 16 16 16 16
14 14- 14 14 14
o
.:: (1)12 (/)12 "'12 VlIZ "'12
(I) ::I :I ::I :I ::Ii
II:: a:: a:: II:: a:
C) 0 10 0 10 0 10 0 10 0 10
08 8 8 8
< 6 6;" 6 6 6
.. .. :1.1" ....... .. 4
.;J l ' d if
o 0 0 o-h--r!
o 2::3.. 6 7 8 0 4 5 6 8 0 2 4 5 6 7 e 0 2:"3" 5 6 7 8 0 2::3 4 6 1 8
VA V
R
V
R
VA V
R
W RELATIVE ROUGHNESS kiD = 85 .. 10'3, FROM FULL SCALE TEST STUDY (REF. 62)
Contents
As with the Department of Energy sponsored tests described above and in reference 62, both
smooth and rough pipe surface finishes were tested. The rough surface finish was obtained
by securing stones to sheets of flexible backing material with epoxy resin, and attaching
the resulting panels around the pipe. Two roughness sizes were used, giving relative
roughness values of kID of 2.5 x 10-
2
and 8.2 x 10-
3

Figure 6.22 shows examples from reference 63 of span motion plots obtained for both the
smooth span (Fig 6.22 a) and the kID = 8.2 X 10-
3
rough span (Figure 6.22 b). Since the
quantity of data presented in reference 63 is limited, directly comparable rough and smooth
span plots for identical G/D values are not available. A number of the essential features of
rough and smooth span behaviour can however be readily identified from this figure.
The smooth span in-line response behaviour in Figure 6.22 a) i) can be compared with the
similar plot in Figure 6.11 a). The almost circular elliptical motion trajectory is apparent
in both cases. Similarly, the cross-flow response behaviour is Figure 6.22 a) ii) displays
the same qualitative behaviour as that shown in Figure 6.11 d).
The rough span motion plots shown in Figure 6.22 b) exhibit totally different behaviour
from that of the smooth span, as was also found for the special study described above and in
reference 62. The in-line response (Figure 6.22 b) i) shows a much flatter trajectory, and
there is evidence of a double frequency 'figure of eight' shaped trajectory. The VR value for
this plot is not closely matched by any of the equivalent plots in Figure 6.18, but lies in
between those of Figure 6.18 a) and b). Not surprisingly therefore, the behaviour also
shows characteristics lying between those of the latter two plots.
The Polar Gas Study rough span cross-flow behaviour (Figure 6.22 b) ii), shows a crescent
shaped response trajectory, essentially similar to that shown in Figure 6.18 c). The in-line
component of the motion is of a significantly lower magnitude for the Polar Gas test, but
this may be attributable to the constraints imposed by the amplitude limiting stops.
Reference 63 reports that the cross flow response frequency coincides with the wake
frequency, and is somewhat lower than the natural frequency of the span. Examination of
Figure 6.17 a) and c) shows that similar behaviour was also observed at higher GID values
for the special full scale study (ref. 62). A further similarity at higher GID between the two
sets of rough span test results, is a more progressive onset of cross-flow response, which
contrasts with the abrupt increase in amplitude observed for the smooth span test.
As shown in Figure 6.17 b) and d), the behaviour of the rough span of reference 62 in the
near bed region (ie., low GID) is different from the behaviour of the same span further away
from the seabed. This effect is also reported in reference 63, which states that the behaviour
of the rough span at GID = 0.25 was observed to the generally similar to the general
behaviour of the smooth span. The special study (ref. 62) and the Polar Gas Study tests
therefore again show qualitative agreement on the respective behaviour of rough and smooth
pipeline spans.
In view of the very low in-line motion amplitudes, it is not realistic to attempt to
determine in-line response thresholds from thePolar Gas study results. Cross-flow response
thresholds are however given in Table 6.1 below, for both rough and smooth spans. The
criterion adopted to define the threshold is a RMS amplitude of O.ID. Since no coherent
hysteresis pattern could be determined, the results do not distinguish between onset and
termination of response.
114
Contents
--.-. DIRECTION OF CURRENT llME INCREASING --.-.
i) IN -LINE
VR 12
iil CROSS - FLOW
V
R
= 26
i) IN - LINE
V
R
= 23
ii) CROSS-FLOW
V
R
= 38
I
o o
0) SMOOTH SPAN G/D' 025
(
b) ROUGH SPAN GID = 10
KID = 82 x 10-
3
FROM POLA R GAS STUDY (REF 63)
Figure 6.22
\
\
,
)
Polar gas study, span mid-point motion for smooth and rough pipes
115
Contents
Table 6.1
Crossflow response thresholds from polar gas tests
SPAN G/D THRESHOLD
CHARACTERISTICS
VR
Smooth 0.25 2.3
Smooth 0.25 2.5
Smooth 1.0 2.8
Smooth 1.5 3.1
Rough 0.25 2.4
Rough 1.0 3.1
6.2.12 Full Scale Tests - Bruschi et al (ref. 22)
These tests were carried out in a tidal channel in the Venice Lagoon, Italy. The test span
was 67 m in length and 20 inches in diameter. For these tests; only a single G/D value
(nominally equal to 4.5), was employed. Both 'smooth' (kID = 0.5 x 1 O ~ 3 ) and rough
(kID = 3 x 10-
3
) surfaced pipe spans were tested. The roughness in this case was produced
by wrapping a net around the pipe, and it can be seen that the resulting kID was lower than
for either the Special Study or Polar Gas tests described above.
As in the latter tests, there was considerable evidence that the smooth span lay within the
critical Reynolds number range. The consequent differences between the behaviour of the
smooth and rough spans were therefore once again apparent.
Characteristics of the rough span behaviour include the progressive onset of vibration as
opposed to the abrupt lock-in of the smooth span, and double frequency figure of eight
shaped span motion trajectories instead of the ellipses demonstrated by the smooth span. A
further similarity with the other full scale programmes described above, is that the response
frequency of the rough span during cross-flow lock-in was found to be significantly lower
than the natural frequency of the span. The smooth span, on the other hand; always vibrated
close to the span natural frequency.
Bruschi et al give a table of threshold velocities for the onset and termination of cross-flow
response. This information is converted into non-dimensional reduced velocity form, and
reproduced in Table 6.2.
Table 6.2
Cross-flow response thresholds from Bruschi et al (ref. 22)
G/D = constant ::= 4.5
SPAN ONSET TERMINATION
CHARACTERISTICS
VR VR
ROUGH 2.94 NOT GIVEN
ROUGH 3.37 3.79
ROUGH 5.05 4.63
ROUGH 3.79 2.52
ROUGH 3.79 2.52
SMOOTH 3.79 3.37
SMOOTH 4.04 2.52
SMOOTH 3.20 2.52
116
Contents
6.3 PRACTICAL RESPONSE CHARACTERISTICS
6.3.1 In-Line Response
In Section 6.1, two reduced velocity threshold values, V RI and V R2, were defined. The first
of these three thresholds, V
RI
, represents the onset of the relatively low amplitude span
response region associated with fluctuating drag forces acting parallel to the approach flow
direction. Span motion in this region is conventionally termed in-line response, due to the
predominant direction of the oscillations, but full scale tests described above show that an
associated cross-flow motion component of similar magnitude may also be present.
Existing design rules (eg., Det norske Veri las, reference 16 and CIRIA, reference 66), show
that the possibility for in-line motion exists for VR values higher than 1.0. The exact VR
threshold is given as a function of the stability parameter, Ks. For the full scale test span
described above, K.. is very low (= 0.1), hence an onset threshold equal to 1.0 is indicated
by the latter references. As explained earlier, the initial onset of in-line response is a
function of the relative motion between the flow and the span. This threshold is not
therefore expected to be heavily influenced by changes in the Strouhal vortex shedding
regime. Additionally, there is no pre-existing span response which may influence the onset
threshold. There is therefore no reason in this case, to expect that the response of a full
scale span will differ significantly from existing observations.
Examination of the full scale test results from the special study (ref. 62) described above,
shows that VR = 1.0 is generally a reasonable threshold for the absolute onset of in-line
motion. however the amplitudes are very low.
Figure 6.23 shows the onset threshold for motion with RMS amplitude of 4 cm, which
gives a RMSID = 0.08, and if the in-line motion is taken to be approximately sinusoidal in
form, a
Except for the immediate near-bed tests, the threshold for build-up of in-line motion with
an amplitude of approximately O.lD is reasonably consistent at VR = 1.4. The onset
threshold predicted by references 16 and 66 therefore appears conservative in this case.
5
VR
4
3
,
0
0
0 ,
+
,
FROM FULL SCALE" TEST STUDY (RE F. 62)
Figure 6.23
!
In-line response thresholds
0

2{)
117
Contents
An area where the established assessment technique for in-line vibration may be non-
conservative, is the prediction of response amplitude. The special full scale study (ref. 62)
showed that the smooth span may respond at amplitudes considerably in excess of
normally assumed 'limiting' value of 0.20. This effect is attributed to the reduction in
hydrodynamic damping associated with the critical Reynolds number region. The
phenomenon is therefore something of a special case, -restricted to smooth (or nearly
smooth) pipes, and a very limited Reynolds number range. In view of the very limited
applicability of this phenomenon, it is not formally incorporated into the span assessment
technique given in Section 7.
Where critical flow conditions do not exist, ie., for the rough span tests, the special study
results (ref. 62) give maximum in-line amplitude in close agreement with those predicted
by existing techniques (see Figure 5.3). For the test span Ks of approximately 0.1, an in-
line amplitude in the region of O.ISO is indicated. The results shown in Figure 6.21 show
typical in-line amplitudes, before the onset of cross-flow motion, of approximately 6 cm
RMS. If it is assumed that the span displacement follows an approximately sinusoidal
oscillation pattern then the maximum amplitude;;; -,j 2 x the RMS amplitude (at the span
centre). A maximum aID of approximately 0.170 is therefore indicated, which agrees
closely with the predicted value of O.ISO.
The objectives of the special full scale study (ref. 62) did not include an analysis of the
effect of varying damping and only this single data point is available to check the
relationship shown in Figure 5.3. Apart from where the effects associated with the critical
Reynolds number region affect the behaviour, there is no reason why the established
relationship of Figure 5.3 should not apply. This relationship is therefore retained for cases
where the assessment procedure described in Section 7 requires the calculation of the in-line
motion amplitude.
6.3.2 Cross-Flow Response
The second of the two previously defined response thresholds V R2, represents the beginning
of the cross-flow response region. Whereas the earlier in-line response may be found
tolerable in some cases due to its small amplitude. the cross-flow motion can be an order of
magnitude more severe, and is therefore unacceptable under virtually all practical
circumstances.
Established references, such as Oet norske Veritas (ref. 16) and CIRIA (ref. 66.), give cross-
flow onset thresholds ranging between 5.0 for sub-critical conditions to 3.9 in the post-
critical region. The full scale test results discussed above however, indicate that these
results are non-conservative for full scale pipeline spans. The latter conclusion is based
upon the data shown in Figures 6.24 and 6.25 which summarise results from the special
study (ref. 62) with the ranges of the results obtained by Bruschi et al (ref. 22) also shown.
When assessing the cross-flow response thresholds from the special study results, some
criterion must be established to define the boundary of the response region. For the in-line
response, where the amplitude is building up from effectively zero, it is relatively simple to
define a significant amplitude level (=0.10), and take the threshold at the point where this
amplitude is first exceeded. This simple approach cannot in most cases be readily applied to
the cross-flow behaviour, however. Examination of Figure 6.10 for example, shows a
significant cross-flow component associated with the in-line motion, which influences the
behaviour up to relatively high amplitudes. The cross-flow response threshold is
determined therefore, by continuing the steeply ascending section of the cross-flow response
curve down to the V R axis and taking the intercept as the threshold.
The criteria used by Bruschi et al (ref. 22) to define the onset and termination thresholds are
not known. The ranges of results obtained however, are shown on Figure 6.:'_4 :md in
the form of scatter bands.
liS
Contents
The onset threshold graph shown in Figure 6.24 displays a significant degree of
experimental scatter. Due to the hysteresis effect described earlier however, the onset of
span motion does not uniquely define the threshold of possible span motion. As described
earlier, the threshold for the termination of response is considered to be the appropriate
criterion for practical span assessment. This termination threshold represents the V R value
above which span response can be sustained, thus any exceedence of this value introduces
the risk of continuing span response. Although the test results indicate that a span may not
spontaneously begin to respond until a higher V R is reached, there is the risk in a practical
situation that an external influence may initiate the response. If the termination V.
threshold is exceeded, then conditions will be suitable to allow this response to continue.
Figure 6.25 shows little systematic differences in termination thresholds between the rough
and the smooth spans. The effect of the generally lower Strouhal numbers for the rough
span is cancelled out by the tendency for forced vibration to occur at frequencies below fN.
All comments relating to cross-flow response thresholds therefore relate equally to smooth
and rough surfaced spans.
A question obviously arises at this point, as to why the criteria shown in references 16 and
66, which were derived from careful experimental studies, differ significantly from those
derived from the full scale span tests described above. Various contributory factors have
already been discussed, and these include the following:
increased Strouhal number in near bed region;
increased Strouhal number in critical Reynolds number region (smooth span);
slightly increased Reynolds number in post-critical Reynolds number region (rough
span).
These effects in themselves however, do not fully explain the relatively large observed
reduction in cross-flow response threshold. Further insight into the phenomenon however,
is provided by the comprehensive study carried out by King (ref. 68). King performed tests
on cantilever mounted cylinders, over a wide range of stability parameter values. Pipeline
spans in general possess low values of stability parameter, Ks, and some of King's (ref. 68)
low Ks results are shown in Figure 6.26. The amplitudes shown relate to the in-line
response, but King states that cross-flow motion is' also present. If Figure 6.26 a) is
compared with the in-line response behaviour for the rough full 'scale span (see Figure
6.26 a, dashed line on graph), the close similarity in behaviour can be observed. This
similarity occurs, despite the fact that the two sets of results apply respectively to
subcritical and post-critical Reynolds number ranges.
Unfortunately, King (ref.28) does not show the cross-flow response corresponding to Figure
6.26. He does however strongly emphasise that the in-line response is linked to the
simultaneous occurrence of cross-flow oscillations. This would be consistent with an
overall response of the type shown in Figure 6.20 a).
King (ref. 68) performed' ~ further experiment which indirectly confirmed the influence of
cross-flow response at these relatively low VR values. A further cylinder with low K.. was
tested, and found to give a similar response to that of Figure 6.26 a). This cylinder was
then fitted with restraining wires to confine, the motion to the in-line plane and eliminate
cross-flow response. The results are shown in Figure 6.26 b). When the interaction between
the two planes of motion is prevented, the response amplitude drops back again to zero as
V R is further increased. This latter behaviour is more consistent with the occurrence of
discrete separated response regions as shown for example in Figure 6.1.
For pipeline spans therefore, which generally demonstrate low K ~ values and flexibility in
all planes, the response cannot be simplistically subdivided into in-line and cross-flow
vibration regions. Instead, there appears to be a coupled interaction between the motion in
the two orthogonal planes, leading to an early threshold of large amplitude response. As
seen from the full scale test results presented above, the large amplitude response may
possess components in both the cross-flow and in-line directions. .
119
Contents
V.
120
5
3
2
o
5
4-
3
2
KEY:
ONSET
BRIJSCHI ET Al
$- 50m SMOOTH SPAN
(REF 221
G/O4!j
o - 40m SMOOTH SPAN CROSS' FLOW
THRESHOLO
,
- 40m ROUGH SPAN
SCATTER
BAND
$
8
8
>-
0
w
I
~
~
z
8
0
,
,
,
,
,
0
~
~
,
0
,
,
05 10
G/D
FROM FULL SCALE TEST STUDY (REF. 62)
Figure 6.24
Cross-flow response thresholds onset of motion
KEY:
TERMINATION
- 50m SMOOTH SPAN
40m SMOOTH SPAN
+ - 40m ROUGH SPAN

+
+

+
t


+
+

*



BRUSCHI ET', I
(REF 221
G/O '45
CROSS'FLOW I
THRESHOLD
SCATTER .
BAND
z
Q
~
z
"
<r
w
>-
o+-------,-------,-----__ ,-______ .-L-__ ~
05 15 20
G/D
FROM FULL SCALE TEST STUDY (REF. 62)
Figure 6.25
Cross-flow response thresholds termination of motion
Contents
a
0
a
o
0-4 ,











0-3
,






0-2







xx
: X
0-1
f
,
,.-
.
,

I
0 ,
r
2 3 4 5
VR
0-4 ,---------------';r----,
03
0-2
0-1
I
j
:

(/!
~ .
2 3 4 5
Figure 6.26
0) RESPONSE OF CYLINDER
WITH COMBINED CROSS-
FLOW AND IN- LINE
MOTION_ (IN -LINE
COMPONENT SHOWN)
RUN 32
x pvc CYLINDER 1I-2 FREE ENDED
K," 0-16
b) EFFECT OF CONFINING
MOTION TO A SINGLE
PLANE_ (IN -LINE)
RUN 32
x PVC CYLINDER 1I-2 FREE ENDED.
K.=O16
RUN 14
AL CYLINDER Jr-I WITH
RESTRAINING WIRES.
K. = 012
FROM KING (REF. 68)
Model cylinder response at low K.
121
Contents
A further point which can be discussed in the light of King's result (Figure 6.26) is the
effect of Reynolds number on the response threshold. It has previously been a suggested
practice to consider the response threshold to be a function of the approach flow Reynolds
number (see refs. 16 and 66). This approach reflects the variation in Strouhal number as a
function of Re, and is backed up by experimental results at the higher values of K.., where
the above interaction effect does not take place. At the low Ks values typical of pipeline
spans however, the coupling mechanism between in-line and cross-flow motion can, as
seen for both full scale tests and the model test results in Figure 6.26; initiate early cross-
flow lock-in. Under these circumstances, the use of a higher cross-flow lock-in threshold for
suhcritical Reynolds numbers is not necessarily -valid. The span assessment technique
described in the next section does not therefore incorporate any Reynolds number
dependence.
Finally, the objective of the present work is to define a design limit on the cross-flow
response threshold, VRI This limit is defined in Figure 6.27 by fitting a design curve to the
full scale test data from references 62, 63 and 22. [n view of the experimental scatter, the
design curve is not rigourously defined by the data points. The curve fit is selected however
to be a conservative representation of the data trend.
The area of the curve defined by 1.0 < G/D < 2.0 comprises a transition zone where the bed
proximity effect attenuates and gives way to free-stream conditions. The Polar Gas study
results (ref. 63) demonstrate a more rapid attenuation of the seabed proximity effect than the
special study results (ref. 62). Other cases may be less conservative. In view of these
possible variations, it is recommended that the near bed V R2 threshold of 2.4 is adopted for
G/D S; 2.0, and the higher free stream threshold of V R2 = 3.0 for G/D > 2.0.
6.4 ALTERNATIVE RESPONSE MECHANISM FOR VERY
SMALL BED GAPS
Prior to the present study, it was commonly assumed that since vortex shedding from
clamped cylinders was observed to be suppressed for small bed gaps with G/D less than
approximately 0.3, then vortex induced resonant vibration should also be absent.
Exploratory tests using a small scale dynamic model of the 50m full scale pipeline span
were carried out as part of the laboratory investigation described in references 15, 17 and 87.
The results indicated that large cross-flow vibration amplitudes of the same order of
magnitude as occur remote from the bed, persisted even down to G/D values nominally
equal to zero. This important observation was confirmed by the full scale test results for the
uncoated 50 m span with G/D = O. The actual G/D, though nominally zero in both these
particular laboratory and full scale tests, was in fact finite due to the effect of the mean lift
force raising the centre of neutrally buoyant span slightly to produce a small gap with a
G/D of the order of 0.1. Such small gap ratios are relatively common in real pipeline spans
and the observation that significant vibration can still occur is thus highly relevant in the
practical context of span assessment.
Two key observations from the present study suggest that the large amplitude response at
very small GID values results from an alternative fluid loading mechanism to conventional
alternate vortex shedding relevant further away from the bed. Firstly, the flow of
visualisation observations obtained in the laboratory experiments showed that the near bed
vibration was accompanied by vortex shedding from one side of the pipeline only, namely,
the side furthest from the bed. The vortices are shed in the upper part of the oscillation and
are exactly synchronised and apparently triggered by the span oscillation. Secondly, the
behaviour of the response frequency, fr, of the vibrating span is quite distinctly different
from the behaviour observed with spans for larger gap ratios. In the latter case, following
vortex lock-in, f, usually increases slightly as the reduced velocity V, is increased through
the resonant response range, eventually either approaching or passing through and slightly
exceeding the span natural frequency, fn For very small G/D values the model results
indicate a quite different pattern of behaviour with the response frequency ratio f, If, now
increasing rapidly from a common initial response value less than 1 through unity finally
approaching a value of 2. The response frequency is initially less than the Strouhal vortex
shedding frequency f, for a clamped pipe at the same flow velocity and finally exceeds the
corresponding Strouhal frequency at higher V, values. For small G/D values, the response
frequency apparently remains more closely linked with the increasing Strouhal frequency, fs,
122
Contents
N
W
V,
0
0


0"

3 !!
0'"
-" o CD
,
-!"
IN
; ....
0
J
0
ii
<

N
7,--------------------------------------------------------------------------------,
6 -
5
4 -
a
,

+
+
"

+
2'
..


+
2 -
0
I
10
SPECIAL STUDY (REF 621

SOm SMOOTH

40m SMOOTH
+
40m ROUGH
DESIGN
BRUSCHI ET AL
(REF: 22)

BAND "-..



J 130

L ..... .. ..-
*


I I
20 30
G/D
POLAR GAS STUDY (REF 631

SMOOTH
0 ROUGH
,
I
40
I I
50
Contents
as the flow velocity increases, than is the case for the more conventional locked-in response
with spans having larger bed gaps. These differing patterns of response frequency behaviour
are illustrated in Figure 6.28 which is based on the small scale model observations. The
model results are corroborated by the full scale test results from the 50 m span tests with
GID = 0 in which a large value of response frequency ratio f,lf, of 1.6 was recorded. This
compares with a typical maximum value of fJf, of approximately 1.2 for spans with larger
GID.
2 , - - - - - - - - - - - - - - - - - - - - - - - - - - - - - ~
fr
Tn
o ~ - - - - - - - - - - _ . - - - - ~ - - - - ~
5 10
EFFECT OF BED IMPACT AT VERY SMALL BEOGAP RATIOS, G/D, ON SPAN
RESPONSE RATIO fr/fn AS A FUNCTION OF REDUCED VELOCITY VR
Figure 6.28
Response at very small G/D
The cause of this peculiar behaviour is not fully understood. However, on the evidence
available, it appears highly likely that an alternative excitation mechanism is involved
which is closely linked to the impact of the pipeline span with the bed soon after initial
response when the amplitude has built up to a small level. Once the span makes contact
with the bed the underflow is cut off, and a large lift force is applied to the pipeline, the
impulse of which, coupled with impact, raises it off the bed again. The upward motion of
the pipeline span induces the formation and shedding of a vortex from the top of the span
which further perturbs circulation round the pipeline and in tum influences the lift force
cycle. The span returns to make contact with the bed again under the natural spring action,
. with the lift force removed and the cycle repeats. Excitation force and response cycles are
thus exactly phase locked and synchronised and occur with increasing frequency as the flow
velocity and lift force amplitude increase.
The rather peculiar 'half-cycle' oscillation of a pipeline span for very small GID values can
be approximately modelled as a so called 'impact oscillator'. This particular type of
oscillator, which also models the behaviour of a moored ship rebounding against jetty
fenders, represents a limiting case of bilinear oscillators described in a recent paper by
Thompson et al (ref. 94). In the perfect elastic rebound case the response frequency
approaches twice the full cycle natural frequency ie., fr ----t 2fn. This impact effect is
reflected in the observed response frequency behaviour of both the full scale and model span
test results discussed above. In spite of the fact that the real physical situation has
additional complicating features compared with the simple model, it appears that the key
elements are correctly simulated and further study is recommended.
Indications from the full scale field tests are that the initial signifjcant cross flow response
still takes place close to the natural frequency and at a Vr value similar to those recorded for
larger GID ratios with the span well clear of the bed. However, the additional damage hazard
resulting from repeated span impact with the bed makes it all the more important to avoid
even modest span response for spans with very small bed clearances.
124
Contents
7. EVALUATION TECHNIQUE
7.1 BASIS OF METHOD
The potential for a pipeline span to undergo vortex-induced vibration is assessed in terms of
the dimensionless "reduced velocity" parameter, YR- The assessment is performed by
calculating the V R value for the pipe span under assessment, and comparing the calculated
V R with two threshold limits, V RI and V R2. The signifIcance of these thresholds is defined
as follows:
VR < VRI No significant vortex-induced vibration takes place. The span is therefore
acceptable.
V
R
> V
R2
Large amplitude and potentially catastrophic vortex-induced vibration may
take place. The span is therefore unacceptable.
VR1 < VR < VR2 . In this region, vortex-induced vibration of relatively low amplitude
may take place. If the. pipeline operator can demonstrate that this lintited amplitude
vibration will not adversely affect the pipeline or any of its coatings, a span in this
region can be accepted.
7.2 INPUT DATA
7.2.1 Environmental Data
The environmental data parameters consist primarily of the steady and wave-induced currents
acting on the span. Suitable values of these parameters must be carefully selected to provide
an appropriate representation of conditions experienced by the span. To determine whether
any response will occur during the lifetime of the span. the design current parameters shall
represent the most severe combination of steady current and stonn sea state that the span is
expected to experience during its lifetime. Established industry methods are available for
expressing such extreme conditions in .terms of a. given return period or encounter
probability for the condition. In some cases, the analysis may show that a limited degree of
response may occur. In these cases, the operator may attempt to demonstrate that this
limited response will not adversely affect the integrity of the pipeline. Under these
circumstances, the environmental parameters shall be selected and applied in order to
accurately represent the long term cumulative effect of the vibration upon the pipeline span.
The selection of an extreme wave parameter presents a potential uncertainty, in view of the
fact that alternative definitions may be adopted to define a given sea state. Definitions
commonly used in the offshore pipeline industry include the significant wave (representing
the highest third of the waves within the sea state), and the maximum wave (statistically
predicted highest wave within the duration of the sea state). By deflOition, a maximum
wave is therefore a rare event. The mechanism described in Section 5.3 for sustained vortex-
induced span response under waves, relies on the presence of a succession of similarly sized
waves. Response induced only by the individually occurring maximum wave will therefore
be transient in nature, may not build up to maximum amplitUde, and will be sustained only
for the period of the wave. In some cases, for example where there is the possibility of
impact with an adjacent installation or solid obstruction, even this limited response may be
unacceptable. If it can be demonstrated however, that the limited response induced under
maximum wave conditions will not represent any risk to the pipeline, then a less stringent
design condition may be adopted.
The significant wave represents more closely the sustained characteristics of the sea state.
This representation therefore provides a more appropriate definition of the environmental
wave parameters which are likely to cause sustained response for the duration of the sea
state. Since it is this sustained response which represents the most serious risk of
endangering the integrity of the pipeline, the significant wave parameters will normally
provide an acceptable definition of the environmental wave condition. It must be
demonstrated however, that the possible momentary response induced by a maximum wave
occurrence will not represent any risk to the span.
125
Contents
The wave-induced current component may be calculated from the design wave parameters
using an appropriate wave theory.
The detailed environmental data requirements are defined in Table 7.1.
Table 7.1
Environmental data parameters
Uc Steady approach flow velocity component acting at right angles to the
pipe axis at the level of the top of the pipe (see Section 4.3.2 for method
of detennining this parameter).
Uw Maximum wave-induced current velocity acting at right angles to the pipe
axis at the level of the top of the pipe.
p Mass density of surrounding water.
7.2.2 Pipeline Data
In this section, it is assumed that the natural frequency of vibration of the pipe span is
known. The calculation procedure for this parameter is addressed in detail in reference 14,
(Guidelines for the assessment of submarine pipeline spans, Background Document number
two).
The pipeline data para.meters required for the span assessment are summarised in Table 7.2.
Table 7.2
Pipeline data parameters
fN Natural vibration frequency of pipe span. See Reference 14 for description
of calculation procedure.
D Outside diameter of pipeline including all coatings.
G Gap between seabed and underside of pipe.
0, Logarithmic decrement of structural damping.
Ille Effective mass,per unit length, pipeline steel mass + contents
mass + coatings mass + added mass.
n.b. This definition of m. applies to a completely submerged pipeline of
uniform mass per unit length. A generalised definition is given in
Equation 7.1.
A generalised definition of m. is given by the following equation. This allows the effects of
non-uniform mass per unit length, and/or free surface penetration to be incorporated
where:
126
oiL my' dx
m, = old y' dx
I11e = effective mass per unit length;
m = mass per unit length at a given section;
y deflection of structure at distance x from origin (mode shape);
L total length of stnicture;
d = submerged length of structure.
Eqn 7.1
Contents
7.2.3 Span Assessment Parameters
The derived data parameters given in Table 8.3 below are obtained by combining the basic
environmental and pipeline data parameters defined above.
Table 7.3
Span assessment parameters
SYMBOL OEFlNmON TERMINOLOGY
U Uc + Uw Approach flow velocity
VR
U
Reduced velocity
fNO
K.
2me Os
Stability parameter
p 0
2
GID GID Gap ratio
7.3 ANALYSIS METHOD
7.3.1 In-line Response
The beginning of the in-line response region is defined by the first reduced velocity
threshold. VRI VRI has a lower bound value of 1.0, but may increase as a function of the
stability parameter, ks This functional dependence is shown in Figure 7.1.
2 5
FIRST INST ABIL:TY REGION
_! SECOND REGION
I
I
I
I
!
I
/i
I
I
i
I
,
I ,
1
I
V
I 1
MOTIQr.j
I
I
I
;,
1
5
20
v"
I
.
10
,
I
o 05 1'0 1'5
K.
Figure 7.1
In-line motion threshold VR1
127
Contents
I I I I
.... OS> ..
N

>
Figure 7.2
Design cross-flow motion
128
o
'"
0
-'
0
z'"
_ ...

... '" 0>-
V
... z
>'2


.....
'ii:"
00""":'
u>-o

... z ..
00>",

..... -
'"",z
'" ... Q

"'-''''
I I
'"
.. N
N
threshold VR2
,

r
0
..
o
'"
o
N
-,
r
o
-
o
o
,
to
Contents
7.3.2 Cross-Flow Response
The beginning of the cross flow response region is defined by the second reduced velocity
threshold, VIV.. As shown by the results of both model and full scale tests, the proximity of
a solid boundary will affect the processes of vortex shedding and span response. For a
pipeline span, the effect of the proximity of the seabed to a pipeline span can be quantified
in terms of the gap ratio, G/D, defined in Table 7.3. VR2 is then expressed as a function of
G/D in Figure 7.2.
7.3.3 Amplitude of In-Line Response
In cases where V
R1
< VR < VR2 it may be possible for the operator to demonstrate that the
limited span response can be accepted for a specified period of time without endangering the
integrity of the pipeline. In these cases it will be necessary to take account of steel fatigue,
fatigue and risk of cracking and/or spalling of the concrete coating, possible impact with
Obstructions, and any other considerations relevant to the individual span site.
Any quantitative assessment of these risks require knowledge of the amplitude of the span
response. The amplitude of this (in-line) response is expressed as a function of the stability
parameter, 1(" in Figure 7.3. Established methods of fatigue and structural analysis may
then be used to determine whether or not the span is acceptable under this level of response.
020T"-----.,...-----,..-----,..-----,
"'--FIRST INSTABILITY REGION,
v, < 12 I
015+---*--,--------'------i------i
IN-LINE MOTION
0 1 0 t - - - - - ~ ~ ~ - - - - ~ - - - - - - - - - - _ 1
,
"'..---1-SECOND INSTABILITY REGION,
V
r
>22
005+-----+---""""':-+--...3 ..... ---------1
20
Figure 7.3
Amplitude of in-line motion as a function of Ks
129
Contents
7.3.4 Number of In:Line Response Cycles
Limited amplitude in-line response will occur for the following conditions:
V R2 > V R > V RI, where V R is derived from the extreme environmental conditions expected
during, the lifetime of the span. Obviously. these extreme conditions only exist for a small
percentage of the time. Under less severe environmental conditions, V R is _very likely to
decrease below YR1 at which point the span motion will cease completely. Span motion
will therefore be restricted to a discrete upper range of environmental conditions.
Span motion will not occur, if the combined steady and wave-induced currents give a
-reduced velocity less.than YR.l. The minimum wave induced current, which when combined
with the steady current will give span motion, can then be determined. By using an
appropriate wave theory, the environmental wave height which generates this minimum
wave induced current necessary for span motion. can also be determined. By referring this
minimum wave height parameter to statistical wave height exceedence data (normally used
to derive a design waveheight), the proportion of the span's lifetime during which vortex-
induced span response will occur, is obtained. The number of response cycles occurring
during this reduced proportion of the span's lifetime then represents the true fatigue loading
on the span due to vortex-induced vibration.
130
Contents
8 CONCLUSIONS
The existing fonn of the 'reduced velocity' approach for the assessment of a span condition
is not acceptable, since the characteristics of real span behaviour are not adequately
represented. The reduced velocity effectively incorporates both the Strouhal number and the
ratio of the vortex-induced excitation frequency to the span natural response frequency. The
Strouhal number is subject to variation under practical environmental conditions, which are
not accounted for under the existing approach. Furthermore, the ratio of excitation to
response frequency at which significant span motion will commence, is not a fixed quantity
and varies according to the span properties and environmental conditions. Again, the
existing analysis approach does not incorporate the effect of these variations.
The Strouhal number and thus shedding frequency is found to be affected by:
Reynolds number;
Pipe roughness;
Bed proximity;
Local velocity gradient;
Local turbulence.
The orientation of the span to the approach flow can also influence the vortex shedding
process. For the purposes of this document however, the approach flow is assumed to
consist of the component acting at right angles to the span axis.
The critical vortex-excitation to response frequency ratio at which significant span motion
will commence is found to be affected by:
Strength and coherence of vortex shedding;
Damping of span;
Degrees of freedom of motion of span.
In view of the number of variables affecting span response, and the inadequate state of
existing knowledge relating to some areas, it is considered impractical to define a purely
analytical approach to the complex span response phenomenon. The high quality data from
the programme of full scale testing was therefore adopted as the main basis for a realistic
definition of span assessment criteria. The theoretical development work described in this
document is used to relate the full scale test data to established physical principles and
existing research data.
A response assessment technique is presented, comprising a development of the existing
reduced velocity method. The approach flow velocity to be used in calculating the reduced
velocity is precisely defined to incorporate the effects of velocity gradients and wave-induced
currents. A new design threshold, defining the onset of vortex induced vibration in the
cross-flow direction, is developed from the results of full scale testing. Existing techniques
are retained for predicting the threshold and amplitude of the less energetic in-line motion.
131
Contents
REFERENCES
I. LIENHARD, JH
Synopsis of Lift, Drag and Vortex Frequency Data for Rigid Circular Cylinders
Washington State University, College of Engineering, Research Division Bulletin
300, 1966
2. PRANDTL, L
J. Roy. Aero. Soc. 31, 730, 1927
3. PRANDTL, Land TIETJENS OG
Applied Hydro- and Aeromechanics
McGraw-Hill (Translated from the German Edition, Springer, 1931), 1934
4. TANEDA,S
J. Phys. Soc. Japan 11,302, 1956.
5. CLUTTER, DW; SMITH, AMO and BRAZIER, JG
Douglas Aircraft Company Report Number ES29075,1959
6. BATCHELOR, GK
An Introduction to Fluid Dynamics
Cambridge University Press, 1967
7. STROUHAL, V
Uber eine besondere Art der Tonerregung
Ann. Phys. und Chemie. New Series Vol 5,1878, pp. 216-251
8. KING, R
A Review of Vortex Shedding Research and Its Application
Ocean Engineering Vol. 4, pp. 141-171, Pergamon Press, 1977
9. GERLACH, CR and DODGE, Fr
An engineering approach to tube flow-induced vibrations.
Proc. Conf. on Flow-Induced Vibration in Reactor System Components, pp. 205-
225, Argonne National Laboratory; '1970'
10. HUMPHREYS, JS
On a circular cylinder in a steady wind at transition Reynolds numbers
J. Fluid Mech 9, pp. 603-612,1960
II. DRESCHER, H
Messung der auf querongestromte Zylinder ausgeubten zeitlich verandetten Drucke
Z.F. Flugwiss 34, 17-21, 1956
12. SARPKA Y A, T
Vorlex Induced Oscillations. A selective Review
AS ME Journal App. Mech. Vol. 46, pp 241-258,1979
13. KlNG, R; PROSSER, MJ; and JOHNS, OJ
On Vortex excitation afmodel piles in water
J. Sound Vibration 29, pp 169-188, 1973
14. J.P. KENNY
Structural Static and Dynamic Analysis of Pipeline Spans
Background Document Two, Guidelines for the Assessment of Submarine Pipeline
Spans, 1984 (Now OT! 93 613)
IS. GRASS, AJ; RAVEN, PWJ; STUART, RJ; and BRAY, JA
The Influence of Boundary Layer Velocity Gradients and Bed Proximity an Vortex
Shedding from Free-Spanning Pipelines
ASME 1. Energy Resources Technology, Vol. 106, pp. 70-78, 1984
16. DET NORSKE VERITAS
Rules for Submarine Pipeline Systems, 1981
133
Contents
17. GRASS, AJ; RAVEN, PWJ; STUART, RJ and BRAY, JA
The Influence of Boundary Layer Velocity Gradients and Bed Proximity on Vortex
Shedding from Free Spanning Pipelines
Paper OTC 4455, 15th Offshore Tech. Conf., Houston. 1983
18. ROSHKO, A
Experiments on the Flow Past a Circular Cylinder at Very High Reynolds Numbers
J. Fluid Mech. 6 pp. 345-356, 1961
19. RIBNER,HSandETKIN,B
Noise research in Canada.
Proc. 1st Int. Congr. Aero. Sci., Madrid (publ. by Pergamon Press, London, 1959),
1958
20. RELF, EF and SIMMONS, LFG
The frequency of eddies generated by the motion of circular cylinders through afluid.
Aero. Res. Counc., Lond., Rep. and Mem. no. 917, 1924
21. DELANY, NK and SORENSEN, NE
Low-speed drag of cylinders of various shapes.
Nat. Adv. Comm. Aero. Wash., Tech. Note 3038, 1953
22. BRUSCHI, RM; BURESTI, G; CASTOLDI, A and MIGLIAVACCA, E
Vortex Shedding Oscillations for Submarine Pipelines: Comparison Between Full
Scale Experiments and Analytical Models.
Paper OTC 4232, 14th Offshore Technology Conference, Houston, 1982
23. CARPENTER, LH
On the motion of two cylinders in an ideal fluid.
J. Res. Nat. Bur. Standards, Vol. 61, No.2, Res. paper 2889,1958
24. GOKTUN, S
The drag and lift characteristics of a cylinder placed near a plane su!face.
M.Sc. Thesis US Naval Postgraduate School, Monterey, California, December,
1975
25. HAFFEN, B
Forces on a transverse circular cylinder in a steady uniform flow near a plane
boundary.
M.Sc. These. Oregon State University, June, 1975
26. BEARMAN, PW and ZDRA VKOVICH, MM
Flow around a circular cylinder near a plane boundary.
J. Fluid Mech. Vol. 89, Part I, pp. 33-47, 1978
27. BURESTI, G and LANCIOTTI, A
Vortex shedding from smooth and roughened cylinders in cross flow near a plane
su!face.
Aeronautical Quarterly, February, pp. 305-321, 1978
28. ANGRILLI, F; BERGAMESCHI, Sand COSSALTER, V
Investigation on Wall Induced Modifications in Vortex Shedding from a Circular
Cylinder
J. Fluid Eng., 1982
29. KIYA, M; TAMURA, H and ARIE, M
Vortex shedding from a circular cylinder in moderate Reynolds number shear flow.
J. Fluid Mech., Vol. 141, Part 4, pp. 721-735, 1980
30. GOUDA, BHL
Some measurements of the p h e n o ~ n a of vortex shedding and induced vibrations of
circular cylinders.
Technische Universitat Berlin Report DLR-FB 75-01,1975
31. BENARD, H
Formation of the centres of rotation behind a moving object
c.r. hebd Seane. Acad. Sci. Paris, 147, pp. 839-842 (in French), 1907
134
Contents
32. NAUMANN. A; MORESBACH, M and KRAMER, C
The conditions of separation and vortex fonnation past cylinders
AGARD Conf. Papers 4, pp. 539-574, 1966
33. BERGER, E and WILLE, R
Periodic flow phenomena
Ann. Rev. Fluid Mech., 4. pp. 313-340, 1972
34. ROSHKO, A
On the drag and shedding frequency of bluff cylinders.
Nat. Adv. Comm. Aero, Wash., Tech. Note 3169, 1954
35. ROSHKO, A
On the wake and drag of bluff bodies.
J. Aero. Sci. 22. pp. 124-132, 1955
36. BEARMAN, PW
On vortex street wakes.
J. Fluid Mech. 28, pp. 625-641,1967
37. CHENYN
Fluctuating lift forces a/the Karman vortex streets on single circular cylinders and
in tube bundles. Part A. The vortex street geometry of the single circular cylinder.
AS ME 1. of Eng. for Industry 94, pp. 603-612, 1972
38. SIMMONS, JEL
Similarities between two-dimensional and axisymmetric vortex wakes.
Aero. Quart. 26, pp. 15-20, 1977
39. GRIFFIN, OM
A universal Strouhal number for the 'locking-in' of vortex shedding to the 'vibrations
of bluff cylinders.
J. Fluid Mech. 85, pp. 591-606, 1978
40. BUREST!, G
On the evaluation of universal wake numbers for roughened circular cylinders in
crossjlow.
ASME Winter Ann. Meeting of Fluid Eng. Div. Nov. 15-20 Paper 81-W A/FE-23,
1981
41. BLEVINS, RD
Flow induced vibration
Van Nostrand Reinhold, 1977
42. ACHENBACH, E
Distribution of local pressure and skin/riction around a circular cylinder in cross-
flow up to R, = 5 x Iif'.
J. Fluid Mech. 34, Part 4, pp. 625-639, 1968
43. BEARMAN, PW
On vortex shedding from a circular cylinder in the critical Reynolds number regime.
J. Fluid Mech. 37, Part 3, pp. 577-585, 1969
44. REYNOLDS, 0
An experimental investigation of the circumstances which determine whether the
motion of water shall be direct or sinuous.
Phil. Trans. Roy. Soc., 1983
45. NIKURADSE, J
Stromungsgesetze in rauhen Rohnen.
Forsch. Arb. Ing. - Wes. Heft. 361, 1933
46. ACHENBACH, E
Influence of sUiface roughness on the cross-flow around a circular cylinder
J. Fluid Mech. 43, Part 2, pp. 321-335,1971
47. BUREST!, G
The effect of surface roughness on the flow regime around circular cylinders.
1. Wind Eng. and Indus. Aero., 8, pp. 105-114, 1981
135
Contents
48. MILLER, BL; MA YBREY, JF and SALTER, IJ
The drag of roughened cylinders at high Reynolds numbers.
NPL Report Mar. Sci. R. 132, April 1975
49. SZECHENYI, E
Supercritical Reynolds number simulation for two-dimensional flow over circular
cylinders.
J. Fluid Mech. 70, Part 3, pp. 529-542, 1975
50. GUVEN, 0; FARELL, C and PATEL, VC
Surface roughness effects on the mean flow past circular cylinders.
J. Fluid Mech. 98, Part 4, pp. 673-701,1980
51. ACHENBACH, E
1979. Vortex shedding from rough cylinders in cross-flow.
Paper presented at European Coloq. 119 London, July 1979.
5L GRASS, AJ
Effects of sUrface roughness on vortex shedding from seabed pipelines.
Report prepared for J.P. Kenny and Partners Ltd, 1983
53. BRADSHAWP
An Introduction to Turbulence and its Measurement
Pergamon Press,1971
54. REYNOLDS, AJ
Turbulent Flows in Engineering
John Wiley and Sons, 1974
55. TOWNSEND,AA
The Structure of Turbulent Shear Flow. 2nd Edition
Cambridge University Press, 1976
56. DAVIES, ME
A comparison of the wake structure of a stationary and oscillating bluff body, using
a conditional averaging technique.
J. Fluid Mech. Vol. 75, part 2, pp 209-231,1976
57. VON KARMAN Th V
On the mechanics of resistance experienced by a body moving through a fluid.
Cottinger Nach. 13. pp. 547-562 (In German), 1912
58. ACHENBACH, E and HEINECKE, E
On vortex shedding from smooth and rough cylinders in the range of Reynolds
numbers 6x10' to 5x10'. .
1. Fluid Mech. Vol. 109, pp 239-251,1981
59. SURRY, D
Some effects of intense turbulence on the aerodynamics of a circular cylinder at sub-
critical Reynolds number.
J. Fluid Mech., Vol. 52, Pt. 3, pp 543-563, 1972
60. BRUUN, HH; DAVIES, POAL
An experimental investigation o/the unsteady pressure/orces on a circular cylinder
in a turbulent cross flow
Vol. 40, pp 535-560, June 1975
61. GORDON, CM and WITTING, J
Turbulent structure in a benthic boundary layer. Bottom Turbulence, Ed. J.C.l.
Nihoul, Proceedings of the 8th International Liege Colloquim on Ocean
Hydrodynamics.
Elsevier Scientific Publishing, 1977
62. HYDRAULICS RESEARCH LTD
Vibration of Pipeline Spans
Report No EXI268 1984, Background Document Three, Guidelines for the
Assessment of Submarine Pipeline Spans 1984 (Now OTI 92 553)
136
Contents
63. HYDRAULICS RESEARCH STATION (now HYDRAULICS RESEARCH
LIMITED)
Polar Gas Project - Vibration of a pipeline span in tidal current.
Report No. EX777 and Report No. EX790 (addendum). 1977
64. VICKERY, BJ and WATKINS, RD
Flow induced vibrations of cylindrical structures.
Proe. 1st Australasian Conf. on Hydraulics and Fluid Mech. Univ. Western
Australia, 1962
65. KING, R; PROSSER, MJ and VERLEY, RLP
The suppression of structuraL vibrations induced by currents and waves.
BOSS '76, 1976
66. CONSTRUCTION INDUSTRY RESEARCH AND INFORMATION
ASSOCIATION UNDERWATER ENGINEERING GROUP
Dynamics of marine structures.
Report UR8, 1978
67. WOOTTON, LR; WARNER, MH; SAINSBURY, RN and COOPER, DH
Oscillation afpiles in marine structures.
Construction Industry Research and Information Association, Technical Note 40,
1972
68. KING, R
Vortex excited structural oscillations of a circular cylinder in flowing water.
PhD Thesis, Loughborough University, 1974
69. FENG, CC
The Measurement of Vortex Induced effects on flow past stationary and oscillating
circular and D-section cylinders.
M.A.Sc. Thesis. The University of British Columbia, 1968
70. HARTLEN, RT et al
Vortex excited oscillations of a circular cylinder.
U.T.I.A.S. Report UTME_TP-6809, 1968
71. SCRUTON,C
On the wind-excited oscillations of stacks, towers and masts.
Paper 16. Proc. conf. on wind effects on buildings and structures, Teddington, U.K.,
1963
72. KING, R
Vortex excited structural oscillations of a circular cylinder in steady currents.
Paper OTC 1948. Offshore Technology Conf. Houston, 1974
73. ROSHKO, A
Experiments on the flow past a circular cylinder at very high Reynolds number.
J. Fluid Mech. 10. pp 345-356, 1961
74. KEMP, PH and SIMONS, RR
The interaction between waves and a turbulent current; waves propagating with the
current.
J. Fluid Mech 116, pp 227-250, 1982
75. GRANT, WD and MADSEN, OS
Combined wave and current interaction with a rough bottom.
J. Geophys. Res. 84, pp 1797-1808, 1979
76. CHRISTOFFERSEN, JB
A simple turbulence model for a three-dimensional wave motion on a rough bed.
Int. Rep. No. I, Inst. Hydrodyn. and Hydr. Tech. Univ. Denmark, 1980
77. GRASS, AJ
Vortex shedding from seabed pipeline spans in unsteady flow produced by combined
wave and tidal currents.
Report prepared for J P Kenny and Partners Ltd, 1983
137
Contents
78. SARPKA YA, T and ISAACSON, M
Mechanics of wave forces on offshore structures
Van Nostrand Reinhold Chapter 3,1981
79. GRASS, AJ; KEMP, PH and STUART, RJ
Vortex induced velocity magnification and loading effects forcylinders in oscillatory
flow.
SERC London Centre for Marine Technology. Report Number FL 28 January, 1981
80. ISAACSON, M and MAULL, OJ
Transverse/orees on vertical cylinders in waves.
J. Waterways, harbors and Coastal Eng. Div. ASCE WWI Feb. pp. 49-59, 1976
81. MAULL, OJ and MILLINER, MG
Sinusoidal flow past a circular cylinder.
Coastal Engineering Volume 2, pp 149-168, 1978
82. BEARMAN, PW; GRAHAM, JMR; NAYLOR, P and OBASAJU, EP
The roles a/vortices in oscillatory flow about bluff cylinders.
Int. Symp. Hydrodynamics in Ocean Engineering Trondheim, Norway. August 1981
83. MERCIER, JA
Large amplitude oscillations of a circular cylinder in a low speed stream.
Ph.D. Dissertation Stevens Institute of Technology, 1973
84. SARPKA Y A, T
Uni-directional periodic flow about bluff bodies.
Naval Postgraduate School Report Number NPS-69SL77-51, Monterey, California,
1977
85. VERLEY, RLP and MOE, G
The forces on a cylinder oscillating in a current
River and Harbour Lab. The Norwegian Institute of Tech., Report Number
STF60A79061, 1979
86. EVERY, MF
Lift Forces acting on circular elements subjected to wave and currents
BHRA Cranfield Report OT-R-8054, September 1980
87. GRASS, AJ and STUART, RJ
Development of guidelines for the assessment of submarine pipeline spans:
Laboratory scale test programme peiformed by University College London
Final Report, 1983
88. TIMOSHENKO, S; YOUNG, DH and WEAVER, W
Vibration problems in engineering
John Wiley and Sons, 1974
89. STOKER, 11
Non linear Vibrations
Interscience Publishers, 1950
90. HALLAM, MG; HEAF, NJ and WOOTTON, LR
Dynamics of Marine Structures.
CIRIA Report UR8 (Second Edition), 1978
91. SARPKA Y A, T
Hydroelastic response of cylinders in harmonic flow.
Proc. Spring Meeting, The Royal Institution of Naval Architects, 1979
92. SARPKA Y A, T; RAJABI, F; ZEDAN, FR and FISCHER, FJ
Hydroelastic response of cylinders in harnwnic and wave flow.
Paper OTC 3992, Offshore Technology Conference, Houston, 1981
93. TSAHALIS, DT
138
Vortex-induced vibrations of a flexible cylinder near a plane boundary exposed to
steady and wave-induced currents.
Proc. 3rd OMAE Conf., New Orleans, Feb. 1984
Contents
94. THOMPSON, JMT; BOKAIAN, AR and GHAFFARI, R
Subharmonic and Chaotic motions of compliant offshore structures and articulated
11Woring towers.
J. Energy Resources Tech., ASME Vol. 106, pp 191-198, June 1984
139
Contents
" "
~ ~
HSE
BOOKS
MAIL ORDER
HSE priced and free
publications are
available from:
HSE Books
PO Box 1999
Sudbury
Suffolk CO I 0 6FS
TeL 01787 881165
Fa" 01787 3\3995
RETAIL
HSE priced publications
are available from
good booksellers.
HEALTH AND SAFETY ENQUIRIES
HSE InfoLine
Tel, 0541545500
Or write to;
HSE Information Centre
Broad Lane
Sheffield S3 7HQ
15.00 net
ISBN 0-7176-0641-4
9
Contents

Anda mungkin juga menyukai