IMPACT LOADING AND

TRANSIENT RESPONSE OF PIPES

TRANSPORTING GAS OR LIQUID

Roslina Mohammad

A thesis submitted in fulfilment of the requirements

For the degree of Doctor of Philosophy

School of Mechanical Engineering

The University of Adelaide

July 2011
DISCLAIMER

This work contains no material that has been accepted for the award of any other degree
or diploma in any university or other tertiary to Roslina Mohammad and, to the best of
my knowledge and belief, contains no material previously published or written by
another person, except where due reference has been made in the text.

I give consent to this copy of my thesis when deposited in the university library, being
made available for loan and photocopying, subject to the provisions of the Copyright
Act 1968.

I also give permission for the digital version of my thesis to be made available on the
web, via the university’s digital research repository, the Library catalogue and also
through web search engines, unless permission has been granted by the university to
restrict access for a period of time.

Signed ……………………………………………….

Date …………………………………………………

iii
ACKNOWLEDGEMENTS

Firstly, I would like to express my sincere appreciation and my deepest sense of

gratitude to my principal supervisor, Associate Professor Dr Andrei Kotousov, for his
patient guidance, invaluable assistance, encouragement and excellent advice throughout
this study. I am also deeply grateful to my other supervisors, Dr John Codrington and
Dr Sook-Ying Ho, for their continuous guidance and important support throughout the
course of the thesis.

I wish to express my warm and sincere thanks for the expertise provided by the
electrical and mechanical workshop staff, especially to Mr Marc Simpson, Manager of
Thebarton Laboratories and Mr Eyad Hassan. I also wish to thank Dr Antoni Blazewicz
for his valuable advice throughout my experiment and Ms Karen Adams for being kind
enough to proofread my thesis. Their kind support and guidance have been of great
value in this study.

I warmly thank my research colleagues especially Stuart Wildy, Jade Wildy, Phuc
Nguyen, Luiz Bortolan Neto and Donghoon Chang for their encouragement and
motivation during my candidature. My thanks are also extended to all Malaysian friends
here in Adelaide and Malaysia. Thanks for your warm hospitality.

I am deeply and forever indebted to my beloved husband, Fikrul Khair Ahmad

Shahabuddin, my parent, and parent-in-law, my sisters and brothers for their love,
patience, support and encouragement throughout my entire life. Your kindness and
thoughtfulness during my study was greatly appreciated.

Finally, I would like to express my gratitude to my scholarship sponsor, the University

Technology of Malaysia, Razak School of Engineering and Advanced Technology,
UTM International Campus, Kuala Lumpur and the Ministry of Higher Learning,
Malaysia for financing of my study. My thanks are extended to the University of
Adelaide, for sponsoring my attendance at various conferences.

And to everyone else I neglected to mention, thank you very much!

iv
PUBLICATIONS

Journal Papers
1.*Mohammad R, Kotousov A, Codrington J, Blazewicz A. Effect of flowing medium
for a simply supported pipe subjected to impulse loading, special ACAM6 issue.
Australian Journal of Mechanical Engineering (AJME). 2011; 8(2): 1-10 (see Appendix
A).
2.*Mohammad R, Kotousov A, Codrington J. Analytical modelling of a pipe with
flowing medium subjected to an impulse load. International Journal of Impact
Engineering. 2011; 38(2–3):115–122 (see Appendix B).
3.Kotousov A, *Mohammad R. Analytical modelling of the transient dynamics of pipes
with flowing medium. Journal of Physics: Conference Series. 2009; 181 012082, 8pp.

Conference Papers

4.*Mohammad R, Kotousov A, Codrington J. Dynamic behaviour of transporting liquid

under impulse loading. 6th Australasian Congress on Applied Mechanics, (ACAM 6);
2010 Dec. 10–12; Perth (Australia).
5.Kotousov A, *Mohammad R. Analytical modelling of the transient dynamics of pipes
with flowing medium. 7th International Conference on Modern Practice in Stress and
Vibration Analysis (MPSVA); 2009 Sept. 8–10; Cambridge (UK).

Awards and Achievements

Awarded 2nd prize for the Postgraduate Student Best Paper Award at the 6th
Australasian Congress on Applied Mechanics in 2010 for the paper:
*Mohammad R, Kotousov A, Codrington, J. Dynamic behaviour of transporting liquid
under impulse loading. 6th Australasian Congress on Applied Mechanics, (ACAM 6);
2010 Dec. 10–12; Perth (Australia).

v
ABSTRACT

This thesis focuses on the investigation of the effect of flowing medium on the transient
response of a pipe due to dynamically applied loading. The topic is very important in
many industrial and military applications including offshore structures, oil and gas,
power stations, petrochemical and defence industries where critical pipe components
transporting a gas or liquid can be subjected to impact loading due to an accident. In
many previous studies, such effects were largely ignored, simplified or considered
negligible. The conducted study demonstrated that in many practically important cases,
the influence of flowing medium on transient response is not small and has to be taken
into consideration.

In the current work, the classical Bernoulli-Euler beam theory is adopted to describe the
dynamic behaviour of an elastic pipe and a governing equation of a slender pipe
transporting gas or liquid was derived. This governing equation incorporates the effects
of inertia, centrifugal and Coriolis forces due to the flowing medium. This equation can
be normalised to demonstrate that only two non-dimensional parameters govern the
static and dynamic responses of the system incorporating a pipe and flowing medium.
Therefore, these non-dimensional parameters can be utilised to investigate various
dynamic phenomena using reduced size or scale physical models. Such scale models
would be adequate if the values of these parameters were kept the same for the scale
model and the real system. This is expected to result in substantial benefits if the
experimental approach is adopted for the investigation of the problem under
consideration.

The main effort in this thesis is devoted to the development of an analytical procedure
utilising the perturbation method and numerical approach adopting a central finite-
difference scheme to analyse the dynamic response of the system due to impulsively
applied loadings. This is then followed by a validation study against previously
published data as well as between both approaches, analytical and numerical. Further, a
detailed investigation was carried out on the effects of the flowing medium on the
transient response. It revealed two principally different types of behaviour of a pipe
subjected to impulse loading: stable decay and unstable associated with so-called pipe
whip.

Special attention is given to the above phenomena, pipe and flow characteristics, which
cause growing unlimited displacements of the pipe regardless of how small the value of
the applied loading is. Experiments were also conducted to support the theoretical
results and were found to follow the theoretically predicted tendencies. The developed
theoretical methods provide a framework for analysis of many other dynamic problems
of pipes with flowing media subjected to arbitrary boundary and loading conditions.
Lastly, the overall conclusion of the conducted research was provided and future work
was identified for further investigations, which follows from the obtained results.
CONTENTS

DISCLAIMER......................................................................................................................iii
ACKNOWLEDGEMENTS.............................................................................................iv
PUBLICATIONS................................................................................................................... v
ABSTRACT.....................................................................................................................vi
CONTENTS...................................................................................................................viii
LIST OF FIGURES..........................................................................................................xi
LIST OF TABLES.........................................................................................................xiv
GLOSSARY....................................................................................................................xv
CHAPTER 1........................................................................................................................... 1
INTRODUCTION.................................................................................................................. 1
1.1 Background and Significance..................................................................................2
1.2 Aims and Objectives................................................................................................5
1.3 Outline of Thesis.............................................................................................6
CHAPTER 2........................................................................................................................... 9
BACKGROUND AND LITERATURE REVIEW...........................................................9
2.1 Introduction...........................................................................................................10
2.2 Bernoulli-Euler Beam Equation..................................................................................13
2.2.1 Review of Bernoulli-Euler Beam Theory and Its Modifications..........................13
2.2.2 Influence of Shear Deformations.........................................................................18
2.2.3 Mechanics of Non-linear Problems......................................................................20
2.3 Effect of Internal Flow on Dynamic Behaviour of Pipes............................................22
2.4 Numerical Techniques...........................................................................................26
2.5 Experiments Investigations of Pipes subjected to Impact Loading.......................27
2.6 Summary.................................................................................................................... 31
CHAPTER 3......................................................................................................................... 33
GOVERNING EQUATION AND ASYMPTOTIC APPROACH.................................33
3.1 Introduction...........................................................................................................34
 3.2 Simplified Classical Bernoulli-Euler Beam Theory..............................................34
3.3 Mathematical Modelling of Beam with Flowing Medium....................................35
3.4 Scaling Transformation of the Dynamic Governing Equation..............................36
3.5 Asymptotic Solution: Cantilever Beam.................................................................38
3.5.1 Solution of Non-homogeneous Classical Beam Equation..............................39
3.5.2 Perturbation Approach....................................................................................41
3.6 Asymptotic Solution: Simply Supported Beam..........................................................43
3.6.1 Solution of Non-homogeneous Classical Beam Equation....................................44
3.6.2 Perturbation Approach.........................................................................................45
3.7 Summary.................................................................................................................... 48
CHAPTER 4......................................................................................................................... 49
NUMERICAL APPROACH AND VALIDATION STUDY.........................................49
4.1 Introduction...........................................................................................................50
4.2 Central-Difference Numerical Scheme..................................................................52
4.3 Initial and Boundary Conditions...................................................................54
4.3.1 Initial Conditions.................................................................................................54
4.3.2 Boundary Conditions...........................................................................................55
4.4 Computational Flow...................................................................................................57
4.5 Validation Study.........................................................................................................59
4.5.1 Numerical Approach versus Exact Analytical Solution.......................................60
4.5.2 Numerical Solution versus Asymptotic Analytical Approach..............................62
4.6 Summary.................................................................................................................... 64
CHAPTER 5......................................................................................................................... 65
EFFECT OF THE INTERNAL FLOW ON DYNAMIC RESPONSE..........................65
5.1 Introduction................................................................................................................66
5.2 General Features of Dynamic Response................................................................67
5.3 Effect of Dimensionless Parameters and Boundary Conditions on Dynamic Response. .70
5.3.1 Encastre Pipe...................................................................................................72
5.3.2 Propped Cantilever Pipe..................................................................................74
5.3.3 Simply Supported Pipe...................................................................................75
5.4 Initiation of Instability (Pipe Whip)............................................................................77

5.5 Analytical Modelling Approach to Instability in Pipes.........................................80

 5.6 Discussion and Conclusion....................................................................................87
CHAPTER 6......................................................................................................................... 89
EXPERIMENTAL STUDY..................................................................................................89
6.1 Introduction...........................................................................................................90
6.2 Experimental Setup................................................................................................91
6.2.1 Flow Tank.......................................................................................................91
6.2.2 Data Acquisition.............................................................................................95
6.3 Experimental Procedure..............................................................................................98
6.3.1 Filling/Emptying the Tank System......................................................................98
6.3.2 Operating the System Pump (P1)..................................................................101
6.3.3 Experiments..................................................................................................102
6.4 Results and Discussions............................................................................................104
CHAPTER 7.......................................................................................................................109
SUMMARY AND CONCLUSION....................................................................................109
APPENDICES....................................................................................................................115
APPENDIX A....................................................................................................................116
APPENDIX B.....................................................................................................................126
APPENDIX C.....................................................................................................................134
fixed-free condition................................................................................................134
Fixed-fixed condition.............................................................................................136
fixed ss condition...................................................................................................138
ss ss condition.............................................................................................................140
REFERENCES...................................................................................................................143
LIST OF FIGURES

Figure 1.1: Typical aboveground pipeline in Australian outback.....................................2

Figure 2.1: The close proximity of piping system to mechanical and electrical
equipment in a power generation aggregate [13]............................................................11
Figure 2.2a: Accidental release of toxic vapour [13]......................................................11
Figure 2.2b: Aftermath of an accident [13].....................................................................12
Figure 2.3: Setup of drop-weight impact test [80]..........................................................28
Figure 2.4: Schematic arrangement for the impact test on pressurised pipes [4]............29
Figure 2.5: Schematic diagram of the apparatus used for experiments with rubber tubes
conveying water. [82]......................................................................................................30
Figure 3.1: Deformation of pipe with flowing medium..................................................35
Figure 3.2: Cantilever beam subjected to impulse loading.............................................39
Figure 3.3: Simply supported beam subjected to impulse loading..................................44
Figure 4.1: Finite-difference grid....................................................................................51
Figure 4.2: Computational flow......................................................................................59
Figure 4.3: Comparison of the exact analytical (3.16) and numerical solutions for ε=
β = 0.................................................................................................................................61
Figure 4.4: Comparison of the analytical and numerical solutions for ε =0.2 and β = 0.1,
analytical solution for ε = β = 0 is given for comparison only........................................62
Figure 4.5: Comparison of the analytical and numerical solutions for ε =0.2 and β = 0.1
(the numerical and perturbation graphs are not distinguishable in this case).................63
Figure 5.1: Example of stable response of a cantilever pipe with flowing medium (at
ç=0.5 and Q=10)..............................................................................................................68

Figure 5.2: Example of unstable response of a cantilever pipe with flowing medium (at ε
=0.5 and β = 60)..............................................................................................................69
Figure 5.3: Transient response of a cantilever pipe for various values of ç while
parameter Q is kept constant (Q = 10).......................................................................................... 70

xi
Figure 5.4: Transient response of a cantilever pipe for various values of Q while
dimensionless parameter ç is kept constant (ç = 0. 5).............................................................71

Figure 5.5: Transient response of an encastre pipe for various values of ç while
parameter Q is kept constant (Q = 10).......................................................................................... 72
Figure 5.6: Transient response of an encastre pipe for various values of Q while
dimensionless parameter ç is kept constant (ç = 0. 5).............................................................73

Figure 5.7: Transient response of a propped cantilever pipe for various values of ç while
parameter Q is kept constant (Q = 10).......................................................................................... 74

Figure 5.8: Transient response of a propped cantilever pipe for various values of Q
while dimensionless parameter ç is kept constant (ç = 0. 5)..................................................75

Figure 5.9: Transient response of a simply supported pipe for various values of ç while
parameter Q is kept constant (Q = 10)......................................................................................... 76

Figure 5.10: Transient response of a simply supported pipe for various values of Q
while dimensionless parameter ç is kept constant (ç = 0. 5)................................................77

Figure 5.11: Unstable behaviour (initiation of dynamic instability) of a propped

cantilever pipe at ç = 1 and Q = 5................................................................................................. 79
Figure 5.12: Unstable behaviour of a simply supported pipe characterised by
dimensionless parameters ç = 1 and Q = 10.............................................................................. 80
Figure 5.13: Dynamic instability.....................................................................................84
Figure 5.14: Critical parameters corresponding to the initiation of the dynamic
instability when one end of the pipe is pinned and another one is free...........................85
Figure 5.15: Critical parameters corresponding to the initiation of the dynamic
instability when one end of the pipe is clamped, and another one is free.......................85
Figure 5.16: Critical conditions for initiation of dynamic instability for cantilever.......86
Figure 6.1: The schematic diagram of impact experiments............................................90
Figure 6.2: The experimental setup.................................................................................91
Figure 6.3: The schematic diagram of the flow facility..................................................92
Figure 6.4: A 2 hp 240 V Onga pump.............................................................................92
Figure 6.5: The supply tank with the top removed, showing the flow conditioning
features............................................................................................................................93
Figures 6.6 (a), (b) and (c): Fixed-fixed PVC pipe section.............................................94
Figures 6.7 (a) and (b): A Solarton-Morbrey Crawley 121" series no KS012342-2
variable-area flow meter to measure the flow rate..........................................................95
Figures 6.8 (a), (b) and (c): Sensor laser to measure deflection......................................96
Figures 6.9 (a) and (b): Load cell is screwed onto the aluminium handle......................96
Figure 6.10: Load cell that is fitted to the handle and used to impact the pipe...............97
Figure 6.11: Data acquisition system—USB-1208FS.....................................................98
Figure 6.12: Filling two-tank facility..............................................................................99
Figure 6.13: Process of filling two-tank facility............................................................100
Figure 6.14: Valve D.....................................................................................................100
Figure 6.15: Pump P2 via Vent V3...............................................................................101
Figure 6.16: Process of emptying two-tank facility......................................................101
Figure 6.17: Pump controller.........................................................................................102
Figure 6.18: Position of Vent V2..................................................................................102
Figure 6.19: The test setup during use..........................................................................103
Figure 6.20: Load intensity as a function of time..........................................................104
Figure 6.21a: Transient response of pipe mid-section at V = 0 m/s..............................105
Figure 6.21b: Transient response of pipe mid-section at V=1.21 m/s..........................105
Figure 6.21c: Transient response of pipe at mid-section at V=2.22 m/s.......................106
Figure 6.21d: Transient response of pipe at mid-section at V=3.0 m/s.........................106
Figure 6.22: Transient response of pipe at mid-section at various speeds of the internal
flow................................................................................................................................107
LIST OF TABLES

Table 6.1: Different range of flow conditions...............................................................103

xiv
GLOSSARY

Nomenclature

hf density of the flow per unit length (kg/m)

hp pipe density per unit length (kg/m)
V velocity of the , gas or liquid or their mixture (m/s)
E Young’s modulus (Pa)
I second moment of inertia of the pipe (m4)
µ mass per unit length (kg/m)
v characteristic time (s)
L length of the pipe (m)
F¯ (x¯, t¯) driving force per unit length (N/m)
F(x, t) dimensionless (normalised) function of the position and time
W(x) dimensionless (normalised) displacement
t time (s)
y specific heat ratio of the gas
ð(x — u) Dirac delta function
s and þ two parameters
Yk a set of normalised eigenfunctions
Ak normalisation constants
x× position along the pipe
∆x spatial grid size
∆t temporal grid size
Ns The upper bounds of the grid point in the spatial direction
Nt The upper bounds of the grid point in the temporal direction
A global evolution matrix
I identity matrix
R applied loading

xv
CHAPTER 1

INTRODUCTION

1
1.1 Background and Significance

The purpose of the research presented in this thesis is to investigate the effect of the
flowing medium on the transient response of pipes transporting liquid or gas subjected
to impulse loading. This will provide a better understanding of the consequences of a
failure so that they can be minimised.

Long flexible pipes such as the one shown in Figure 1.1 are used extensively across
many industries to transport liquid or gas or their mixtures. The failure rate of gas or oil
pipelines differs significantly, with variances in design factors, construction conditions,
maintenance techniques, and the environmental situation.

Figure 1.1: Typical aboveground pipeline in Australian outback

The European Gas Pipeline Incident Data Group presents an overall failure rate of 0.575
per 1000 kilometres/year based on the experiences over 1.47 106 kilometres/years of on-
shore gas pipelines [1]. A survey of 185 accidents involving natural gas [2] showed that,
of the total, 131 were caused during transportation, by road, railway, ship or pipeline.
The analysis of these data clearly shows the relatively high frequency of accidents in
pipes: 127 of them occurred in piping systems. The most frequent causes of the
accidents were mechanical failure, impact failure, human error and external events.
Amongst the accidents arising from impact failure (39 accidents), the most frequent
specific cause was excavating machinery (21 accidents), followed by vehicles (5
accidents) and heavy objects (5 accidents). Other specific causes due to external events
were ground subsidence (4 accidents) cases and sabotage/vandalism (4 accidents) cases.

Failures from impact loading often caused by third party mechanical interference can be
immediate or may occur sometime later due to fatigue. This type of failure is likely to
have severe consequences. Historically, some of the most serious pipeline accidents
resulting in ruptures have been caused by such incidents. In general, failure of the
pipeline can lead to various outcomes; some can pose a significant threat of damage to
people and properties in the immediate vicinity of the failure location.

Prior to the recent release of AS 2885.1-2007, Australian pipeline standards mandated

burial for all cross-country pipelines carrying gas. The reason was not stated explicitly
but would appear to be related to the hazard of pressure in a fluid that could lead to
devastating consequences if an aboveground gas line were to rupture. The 2007 revision
of AS 2885.1 includes a new section (5.8.3) on ‘Pipeline with reduced cover or
aboveground’, which makes it clear that pipelines may be installed aboveground. The
reason for this revision was probably motivated by economical considerations, as
aboveground gas pipelines can offer significant benefits while posing relatively low
risks. Some gas producers are interested in application of aboveground gas lines so that
they could have this option available for production flowlines in remote areas. Hence,
the issues surrounding safety and dynamic behaviour of aboveground gas pipelines are
very important, specifically for Australia.

On the other end of the structural length scale, various sensor technologies have been
develop recently based on in micro-pipes carrying fluid. These technologies involve the
use of dynamic measurements. Therefore, there is now a growing interest in studying
the transient response of pipes with flowing media caused by dynamic loading.

The behaviour of pipes subjected to dynamic loading has been extensively studied in the
past, through analytical, numerical and experimental approaches [15-82]. Many
previous studies were focused on the development of theoretical techniques for
investigation of linear or non-linear dynamic response within simplified one-
dimensional models [17, 33, 35, 59]. However, not many investigations were focused
on the effect of the flowing medium on the dynamic response of pipes. The flowing
medium can produce substantial forces on the pipe walls and affect the dynamic
response of the whole structure, and at the same time, the pipe response, in turn, can
change the flow characteristics in a very complicated manner. Due to the complex
coupling of fluid-structure interactions, it is not surprising that in many previous
theoretical works that have focused on the impact loading of pipes, the internal flow
effects were simply ignored or only partially incorporated into the governing equations,
utilising oversimplified assumptions. For example, a number of published theoretical
one-dimensional models describe the effects of the transported liquid by means of an
attached mass model in which an additional mass is distributed along the pipe with the
linear density equal to the one of the liquid [35,44]. The distributed mass is then
introduced into the corresponding governing equations. Such simplification could be
adequate if the flow velocities and rates of loading are comparatively low; however, at
higher flow speeds or rates of loading, such theoretical models can produce large errors
in the quantitative assessment of the pipe transient behaviour when it is subjected to
dynamically applied loading. For the same reason, which is complexity of the
mechanisms governing the transient response, there were no numerical attempts or
experimental studies, to focus on the effect of the flowing medium on the transient
behaviour of pipes.

The possible large errors in theoretical predictions of pipe response to dynamic loading
are not the only reason behind the proposed study. Another major concern, if one
disregards the effect of the internal flow in pipes, is that many critical (from a safety
point of view) phenomena could be completely neglected, leading to incorrect design
decisions. For example, the existence of critical values of the flow was demonstrated,
such as flow density and speed, at which any small external disturbance leads to
growing unbounded deflections of the pipe. Such behaviour is linked to the dynamic
instability of pipes with flowing medium. This behaviour is seen familiarly in the chaos
caused by an unrestrained garden hose with running water under relatively high
pressure, phenomena called pipe whip. For example, in the case of a full bore failure of
an aboveground pipeline, the pipe whip can represent a potential threat to structures and
people in the vicinity of the pipeline. Therefore, when designing aboveground pipelines,
the possible instability mechanisms have to be avoided or some special measures have
to be undertaken to avoid or minimise the consequences of pipe failures.

The main objective of the thesis is the analytical, numerical and experimental
investigation of the effects of the flowing medium on the transient response of a circular
pipe carrying gas or liquid, which are subjected to impact loading. These effects can be
very important in modern high-pressure pipelines, which now, in accordance with the
Australian standards, are permitted to be constructed above ground. Indeed, the
potential energy density accumulated in such pipelines is comparable with the energy of
some explosive substances. The research will provide a better understanding of the
consequences of a failure so that they can be minimised. Further, an attempt is made to
provide a theoretical basis for the development of design criteria that avoid possible
unstable behaviour of pipes in the case of an accident.

1.2 Aims and Objectives

The overall objective of this research project is to investigate, using analytical,

numerical and experimental approaches, the effect of the flowing medium on the
transient response of a pipe transporting gas or liquid and subjected to impact loading.
In order to achieve this objective, the following specific aims were formulated for the
project:

 Aim 1: Derive simplified governing equations of a pipe model with flowing gas
or liquid subjected to external loading.

 Aim 2: Develop an analytical procedure for analysis of the flowing medium

effects on the transient response of a pipe, which utilises the perturbation
approach. This approach is expected to be accurate at relatively small influence
of the internal flow on the overall response.

 Aim 3: Develop a computational tool adopting the central finite-difference

scheme for determining the dynamic response of a pipe with flowing medium
subjected to arbitrary dynamic load.
  Aim 4: Validate the developed theoretical approaches against known analytical
and numerical solutions.

 Aim 5: Using analytical and numerical approaches, analyse the dynamic

response of a pipe transporting liquid or gas and estimate the internal flow
effects.

 Aim 6: Investigate critical parameters of the internal flow leading to the

initiation of pipe instability. Estimate the effect of support conditions and
develop simple criteria avoiding the initiation of dynamic instability.

 Aim 7: Design an experimental rig for studying the transient response of pipes
with flowing medium subjected to impact loading.

 Aim 8: Conduct experimental investigations on the effect of the internal flow on

the transient response of pipes transporting liquid.

1.3 Outline of Thesis

The remainder of this thesis is structured as follows. Chapter 2 provides an in-depth

critical literature review of the previous research related to the current research topic.
This includes a review of major previous studies of the transient response of pipes with
and without flowing media and pipes subjected to low speed impact. Due to the great
number of articles on the topic, this review is restricted to a summary of the current
state of the art in this area including main results and general analytical, numerical and
experimental approaches to analysis of such types of problems.

Chapter 3 introduces the governing equation of a long flexible pipe with flowing
medium that takes into account the forces generated by the internal flow due to inertia,
centrifugal and Coriolis accelerations. Due to the approximate nature of the
mathematical modelling, a number of assumptions and simplifications have to be
introduced. It is assumed, for example, that the flow velocity and density in the pipe are
not affected by the pipe transient response but the pipe itself is subjected to the
additional external forces as a result of the pipe lateral deflections that change the flow
direction. Small and, hence, linear deformations are also assumed in the pipe as well as
all assumptions utilised in the classical Bernoulli-Euler beam theory. Thus, there is no
consideration of plastic deformation, fracture or other phenomena that might accompany
the impulse loading of pipes. Further, a scaled transformation of variables is conducted
to demonstrate that only two dimensionless parameters affect the transient response.
These parameters represent dimensionless combinations of the elastic properties of the
pipe, pipe geometry and flow characteristics. In the case of a small influence of the
flowing medium on the overall mechanical response of the whole system, the governing
equation can be solved using an analytical approach, which is based on the standard
perturbation technique. Several examples demonstrating the effectiveness of this
approach for particular problems are considered in detail. The asymptotic approach
developed in this chapter can be generalised relatively easy for other boundary
conditions as well as for arbitrary applied loading.

Chapter 4 presents a numerical method using central finite-differences scheme. In this

chapter, the consideration is given to validate the earlier developed analytical approach
and the numerical method against published results and classical solutions. Further, as
expected, both the approaches, asymptotic and numerical, produce essentially the same
results in situations in which the internal flow has a small influence on the mechanical
response; for example, this happens at low flow speeds or large moments of inertia of
the pipe.

As mentioned above, only two dimensionless parameters fully control the transient
response of the system within the adopted assumptions and simplifications. In Chapter
5, a comprehensive investigation of the effect of the flowing medium on the dynamic
response of pipes is conducted. Special attention is paid to the conditions that initiate
dynamic instability associated with pipe whip phenomena. A simple criterion to avoid
such dangerous behaviour is proposed and can be used in the design of aboveground
pipelines.

In Chapter 6, a rig is described for experimental investigation of the effect of flow

parameters on the dynamic response of circular pipes. The rig essentially represents
encastre pipe with water flowing through it. A laser sensor was used to measure the
deflection at the mid-point of the pipe which was impacted by a load cell device. This
chapter also presents the results of the experimental investigations followed by a
discussion.

Finally, Chapter 7 presents the overall conclusions of the conducted research, along
with recommendations for future work.
CHAPTER 2

BACKGROUND AND LITERATURE

REVIEW
2.1 Introduction

The aim of this chapter is to provide a general overview of problems, methods and
solutions related to the dynamic behaviour of long flexible pipes transporting liquid or
gas. A substantial number of sources were reviewed and a brief summary will be
provided here. This literature review is comprised of several sections covering various
aspects of dynamic behaviour of long flexible pipes having circular cross-sectional area.
The present chapter is not intended to provide a comprehensive review of problems and
complex phenomena associated with dynamic behaviour of pipes, specifically, vibration
and vibration control of pipes conveying gas or liquid, curved pipes and pipes of non-
circular cross-section. Instead, it focuses on specific issues that are important for the
current study.

Pipes are widely utilised in many industries and across many applications [3] to
transport, process, transmit and store fluids and materials that, sometimes, can be
hazardous, volatile and flammable. For example, in oil and gas installations (whether
on-shore or offshore) circular pipes are commonly used for the conveyance of either
crude or products. An accident leading to the loss of the pipe integrity could pose a
serious threat to property, environment and people in the close vicinity of the accident
location [4]. Therefore, it is not surprising that there were many investigations
conducted in the past two decades that are related to the dynamic behaviour, safety and
integrity of pipelines and pipe systems [5].

The importance of safety and integrity issues associated with pipelines and pipe systems
was highlighted by a number of very serious accidents in the past [6–12]. Figure 2.1
illustrates an example of a pressurised piping system that is in close proximity to
mechanical, electrical and electronic equipment and instruments of a power generation
aggregate [13]. Mechanical damage to the pipe system could result in a disastrous
consequence and severe damage to the power generator and the personnel.
 Figure 2.1: The close proximity of piping system to mechanical and electrical
equipment in a power generation aggregate [13]

Another typical example is a petrochemical industrial plant, which often incorporates

thousands of pipes for various technological purposes, which can potentially fail due to
many reasons. The cost of such failures is often much greater than the cost of the initial
failed component. Consequences of a pipe failure on the petrochemical plant are shown
in Figures 2.2a and 2.2b. These figures display images of the aftermath of an accidental
toxic vapour cloud that was discharged following a failure of a pipe component. The
initial failure resulted in an explosion that was followed by a series of other explosions
that resulted in a US$1.5 billion loss and 23 deaths at the facility [13].

Figure 2.2a: Accidental release of toxic vapour [13]

 Figure 2.2b: Aftermath of an accident [13]

As part of the safety precaution measures, effective protection solutions against possible
impact loading have to be implemented to any potentially dangerous piping system or
pipeline. Subsequently, safety engineers or safety authorities have to demonstrate the
proper and adequate design strength not only against quasi-static loading but also in the
case of abnormal events such as that caused by impact loadings because this sort of
loading is rather frequently found in practice. It can be associated with accidents such as
collision with a moving object, struck by a falling object or acts of vandalism.

Numerous surveys carried out on high-pressure pipelines have confirmed that that the
impact loading is one of the main reasons of heavy accidents. It is interesting to note
that although the accidents due to impact loading are less frequent than, for example,
accidents due to mechanical failure, the most serious pipeline accidents [14] resulting in
high financial, property or human life losses have been caused by incidents associated
with impact. Amongst the accidents arising from impact failure, the most frequent were
machinery accidents, followed by vehicles and heavy objects. Other specific causes of
such accidents due to external events were ground subsidence and acts of
sabotage/vandalism.

The purpose of the current project is to gain a further understanding of dynamic

behaviour of pipes under impulse loading (impact-like loading). The study is focused on
the investigation of the effects of the internal flow on the dynamic response of pipes,
which was often ignored in many previous studies. Special attention is given to the
condition leading to dynamic instability of pipes. Such conditions can occur during an
accident resulting in a full bore fracture of a high-pressure pipeline with a relatively
small diameter. In this case, the pipe transporting gas or liquid can become unstable and
generate substantial damage due to chaotic movements, similar to what almost everyone
has experienced when watering a garden with an unrestrained hose. The next section
will provide a brief review of simplified one-dimensional theories of pipes based on
Bernoulli-Euler assumptions, which will be also utilised and further extended in the
current study.

2.2 Bernoulli-Euler Beam Equation

This section will begin with considerations of the classical Bernoulli-Euler beam theory,
which is the most popular theory for modelling the transient response of long flexible
structures. Many previous studies utilised this theory and many interesting results have
been obtained in the past. It was also demonstrated in many previous studies that for
sufficiently flexible structures it provides an excellent correlation with experimental
results. The Bernoulli-Euler beam theory will be further extended and utilised to model
effects of the internal flow on the transient response of pipes.

2.2.1 Review of Bernoulli-Euler Beam Theory and Its

Modifications

The Bernoulli-Euler beam theory can be formally derived from the classical linear
theory of elasticity using well-developed asymptotic techniques, energy or variational
approaches. In general, the theory provides a way to calculate the load-carrying capacity
and dynamic deflections of long and flexible beams subjected to various support
conditions. Bernoulli-Euler beam theory is also known as simplified beam theory or
classical beam theory or engineer’s beam theory. This theory can reasonably be applied
only for the laterally loaded beams without taking into account the sheer deformation in
the material in contrast to more complicated theories, such as Timoshenko beam theory,
which will be briefly discussed later in this chapter. The Bernoulli-Euler beam theory
relates to the case of small lateral deflections, which must be much smaller than the
characteristic size of the cross-sectional area. The out-of-plane (or transverse)
displacement W(x) of a statically loaded beam is described within the Euler-Bernoulli
beam theory by the following ordinary differential equation:

d2 &2W
[EI
dx2 ]=p (2.1)
&x2

where p is the distributed loading (force per unit length) acting in the same lateral or
transverse direction as W, E is the Young’s modulus of the beam’s material (in Pa), and
I is the area moment of inertia of the beam’s cross-section (in m 4). If E and I (or
product of these variables) do not vary with x along the length of the beam, then the
beam equation simplifies to the especially simple form:

&4W
EI =p (2.2)
&x4

which is well known from many engineering textbooks.

This formulation also assumes that the internal damping, externally imposed tension,
inertia and pressurisation effects are either absent or can be neglected in comparison
with the elastic restoring reaction of the beam under the applied external loading
[15,16,17].

The quasi-static theory can be generalised immediately to the case in which the inertia
force is not negligible. In this case, the transient deflections, W(x, t) of a long flexible
beam will be described by the following partial differential equation (PDE), which
incorporates the inertia term (the second term on the left side of the equation below) as
&4W(x, t) EI +µ
&x4
&2W(x, t) =F
&t2 (x, t) (2.3)

where W(x, t) is already a function of x and t. The first term represents the flexural
elastic restoring force and the second term is the lateral inertia force of the beam, where
µ is the mass per unit length of the beam (in kg/m). The right-hand side term F(x, t) , F
is a function of span location and time; it represents the external applied force per unit
length (in N/m). In the above equation x is the spatial coordinate along the length of the
beam, and t is time. The origin of the coordinate system and time can be selected
arbitrary and does not affect the solution of this equation.

Many studies utilised the Bernoulli-Euler beam theory to investigate the dynamic
response of beam due to transversely moving point mass [18]. These studies were
strongly motivated by the rail industry demands and other industrial applications. In this
case, the governing equation can be written in the following form:

&4W(x, t) &2W(x, t)
EI
&x4
+µ = Fð (x — u) (2.4)
&t2

where F is the intensity of the applied point force and ð(x — u) is the ordinary Dirac
delta function. The product of the intensity and Dirac function represents the moving
point mass at location x — u, which also changes with time. For example in [18], the
effect of the force travelling velocities, boundary conditions and the ratio of the moving
mass to the mass of beam on the dynamic and vibration response of the beam were
comprehensively studied.

The Bernoulli-Euler beam framework was subsequently utilised to develop more

complicated formulations for analysis of the transient behaviour of long flexible beams
[19] incorporating damping. An example of extension of the classical beam theory with
linear damping can be written as:
&4W(x, t) &2W(x, t) &W(x, t)
EI 4
+µ +c &t ÷ F (x, t) (2.5)
&x &t2
In this equation (2.5), the third term represents damping with c is a positive constant
characterising the damping effect; c in this equation in (kg/ms).

The Bernoulli-Euler beam theory was also extended to long flexible pipes having an
internal flow of gas or liquid. The linearised equation of motion for a horizontal pipe
conveying fluid in the case of negligible damping can be written as:

&4W 2 2
& W
2
& W & W ) 2 ( )
(M + N (2.6)
EI 4 + MU + 2MU + = F x, t
&x &x2 &x&t &t2

where M is the mass of fluid per unit length, flowing with a steady flow velocity U, N
is the mass of the pipe per unit length, W is the lateral deflection of the pipe, x and t are
the spatial coordinate and time, respectively; and the right-hand side term is F (x,
t) represents the external dynamic loading. The various terms in the above equation can
be identified, sequentially, as the flexural restoring force, a centrifugal force, a
Coriolis force, and the inertia force. This equation represents the simplest
governing equation taking into account the effect of the internal flow on the
structural response of the system (pipe–internal flow) [20]. Formally, this equation
can also be further extended
[21] to incorporate the viscous damping effects, for example, as follows:

&4W &2W &2W &2W &W

2
(M + N ) ( ) (2.7)
EI + MU + 2MU + +c = F x, t
4 &x 2 &x&t 2
&x &t &t

The equation (2.7) was used in [21] to investigate the transient vibrations of a fluid-
conveying pipe. The results from experimental and numerical simulations demonstrated
that Coriolis force may have a significant damping effect at moderate flow velocities
and the damping of the flow in a pipe can be effective and may have potential in
engineering applications as a vibration controller. It was also demonstrated that the
damping effect may be exploited to control the vibration of a light and resonant
structure, if the pipe carrying the fluid is used as a controller in a cantilever pipe.

Another simplified model was developed to incorporate the axial tensile forces. When
such forces are significant, the motion of a pipe can be described by the following
fourth order partial differential equation (PDE):

&4W &2W &2W &2W

2
EI — (To — MfU ) + 2MfU + (Mf + =0 (2.8)
&x4 M&x
p) 2 &x&t &t2

where To the effective axial force representing the combined effect of an initially
imposed force To, which will be negative when it is compressive. The effect of the fluid
flow produces a centrifugal force due to curvature of the span, and a Coriolis force due
to combined flow and rotation of the fluid elements. The centrifugal force is equivalent
to a compressive end load, and in vibration problems leads to the reduction of the
natural frequencies of the pipe. The Coriolis force causes an asymmetrical distortion of
the classical mode shape. The last equation is based on the assumptions that the
dynamic deflections are small so that only linear terms are significant. The other
assumptions are that mass densities are uniform, transverse shearing and damping are
negligible and the fluid velocity and pressure are constant along the pipe.

In [22], the effect of fluid velocity on buckling instability of a pipe with the cantilever-
type boundary conditions was investigated. The effect of the axial load P, was analysed
utilising the following governing equation:

&4W 2 &2W &2W &2W

( ) ( )
EI + MU + P + 2MU + M+N =0 (2.9)
&x 2 &x&t
&x 4 &t2

where, as previously, the first term describes the flexural elastic restoring force, the
second term the centrifugal force term together with the effect of axial compression, the
third tem takes into account the Coriolis force and the fourth—the inertia term. This
study revealed that flutter or divergence instability may occur, depending on the
combination of velocity and axial loading.

There was also an interest to study the stability of a tensioned clamped-pinned pipe
conveying fluid. [23]. The study has proven that for small fluid velocities, the pipe is
stable. The pipe loses stability by divergence at relatively high fluid velocities. The
equation of motion governing the lateral in-plane motion of the pipe without internal
and external fluid as a function of the axial distance, x and time, t, [23] it was taken in
the following form:

&4W &2W &2W ( ) ( )

EI +T + Nr &t = F x, t + x,
t , (2.10)
2 int F
&x 4 &x 2 est

where EI is the bending stiffness of the pipe, Nr is the mass of the pipe per unit length,
T is a prescribed axial tension, Fint (x, t) and Fest (x, t) are forces acting on the
pipe from inside and outside, respectively. The internal fluid flow is approximated as a
plug flow, as if it were an infinitely flexible rod travelling through the pipe, all points
of the fluid having a velocity U relative to the pipe. This was accepted as a
reasonable approximation for a fully developed turbulent flow profile [24].

2.2.2 Influence of Shear Deformations

The effects of shear deformation, which are neglected in the classical beam theory, have
been extensively investigated in literature. The Timoshenko’s beam theory, for
example, can be used to allow for the effect of transverse shear deformation and
constitutes an improvement over the standard beam. A numerical example for the
supported-clamped pipe was considered by Huang [25]. The influence of rotary inertia,
shear deformations, internal dissipation (the dissipation in this work was simulated
within Kelvin-Voigt model) on the dynamic response of pipes transporting fluid were
studied by Bratt [26]. Stack et al. [27] and Lin and Tsai [28] presented a finite element
(FE) approach for non-linear vibration analysis of Timoshenko pipes conveying fluid.
The results of the simulations were partially validated using data from experimental
studies available in the literature.

A vast majority of the previous studies on the investigation of the effect of shear
deformations on the transient response of pipes utilised the Timoshenko theory of
beams subjected to transverse loading and were conducted using computer-based
techniques. Considering an elastic Timoshenko beam of length L, uniform cross-section
area A, and moment of inertia I, subjected to specified boundary conditions at the left
and right ends of the beam and applying Hamilton’s principle, one can derive the
coupled Euler–Lagrange governing equations in the absence of body forces as

&2W(x, t) &ƒ(x, t) &2W(x, t)

kG
&x2 &x ) — q =0 (2.11a)
( &t2
—

& 2 ƒ(x, t) & 2 ƒ(x, t)

&W(x, t)
EI + kGA &x — (x, )) — =0 (2.11b)
( &x2 ƒ t qI &t2

where W (x, t) is the transverse deflection function and ƒ(x, t) the slope of the
deflection curve due to bending. Parameters E, G and q represent the Young’s
modulus, the shear modulus, and the material mass density, respectively; k
represents the Timoshenko shear coefficient that is introduced to account for the
geometry-dependent distribution of shear stress. One can rewrite the above system
of PDEs as a fourth order single PDE.

Chao et al. [29] developed a set of coupled-non-linear PDEs of motion for pipes with
flowing medium using the continuity and momentum equations of unsteady flow as well
as Timoshenko beam theory. However, the governing equations are too complicated to
allow any analytical solution or simple analysis of various effects. They will not be
given here.

Timoshenko beam theory was implemented to study the transient response and stability
of relatively pipes transporting fluid by Pramila et al. [30]. The hydro-elastic stability of
short simply supported pipes was studied by Tani and Doki [31] also within
Timoshenko beam theory. These researchers investigated the effects of shearing loads
and support conditions on the dynamic behaviour. It was demonstrated that an increase
in the shearing load reduces the critical velocity leading to pipe instability and raises the
number of the corresponding hoop waves. A change of the boundary conditions was
found to influence significantly the short pipe dynamics. The dynamics and stability
issues of short pipes were re-examined by means of Timoshenko beam theory for pipes
and a three-dimensional fluid-mechanical model for the fluid flow [32, 33]. These short
pipes were subjected to clamped-clamped or cantilever-type boundary conditions.

2.2.3 Mechanics of Non-linear Problems

The linear models were found incapable of describing dynamics of pipes with flowing
medium when the flow velocity approaches a critical value leading to static or dynamic
instability [34]. The non-linear dynamics of pipes transporting liquid or gas and
subjected to various boundary conditions has been investigated extensively in recent
years due to wide applications of the relevant problems in many industrial situations,
such as heat exchangers, nuclear reactor piping, power stations and pipelines.

The non-linear equations of pipes with flowing medium are normally derived using the
Hamilton’s principle or from Newton’s second law. An example of a non-linear
governing equation derived by Semler et al. [35] in a relatively simple form is given
below. It is based on the energy and Newtonian methods and can be written in the
following form:

&25 &45 &55 2

&25
&25 &5
+ +α +u + 2uƒµ +y
&v 2 &£ 4 &v&£ 4 &£ 2 &v&£ &£
&5 2
&25 &25 y &5
+ ( ) (u + 2uĵ + )
&£ 2 &£ 2 &v&£ 2 &£

& &4 5 &5 2

&5 &2 5 &3 5 &5 3
+ (1 + α ) ( 4 ( ) + 4 +( ))
&v &£ &£ &£ &£2 &£3 &£
 &u
— y) (1 — £) (1 3 &5 2
+ &v &25
+
( ) &£2
( ƒµ )2 &£

1£
&25 &2 5 2 &5 &35
— (ƒ ƒ (( )+ ) d£d£
&£2 £ 0 &v&£ &£ &£&v2

1
1 &u &5 2
&5 &25 &5 &25
+ƒ( + 2uƒµ + u2 ) d£)
ƒµ &v &£ &£ 2
2 ( ) &v&£ &£
&£
£

£ 2
+ƒ &25 ) + &5 &35 ) d£ = 0 (2.12)
(( 2
&v&£ &£ &£&v
0

where 5 = W⁄L, W is the lateral deflection of the pipe, L is the pipe length, £ = x⁄L is
the dimensionless spatial variable, u = VL⁄ƒM⁄EI is the flow parameter, V is the fluid
velocity, which can vary with time, EI is the pipe flexural rigidity,

v = tƒEI ⁄(M + N)⁄L2 is dimensionless time, α = cƒEI ⁄(M + N)⁄L, c is the

coefficient of Kelvin-Voigt damping of the pipe material, µ = M⁄(M + N) is the mass
ratio of the linear densities of the pipe and pipe with fluid, and y = (M + N)gL3 ⁄EI is
a gravitational parameter.

The first term in the above equation is the linear inertia, the second term is the linear
stiffness, and the third term is due to the Kevin-Voigt damping. All terms involving the
square the fluid flow speed u2 are due to the centrifugal force, while those having linear
term u are due to the Coriolis acceleration. All terms including y represent the effect of
gravity. The terms in the third line represent the non-linear stiffness due to non-linear
curvature; the fifth and the last lines represent non-linear inertial. Terms with time
derivative of the flow velocity take into account the unsteady mature of the internal
flow. A systematic comparison of the derived equation with previous studies was
conducted. However, no quantitative analysis of the governing equation or investigation
of the effect of various controlling parameters have been carried out.
The governing equations near the critical fluid velocity above which the lateral
deflections become self-excited and unbounded (within the linear theory) were derived
by Yoshizawa et al. [36]. The beating phenomena, as predicted by the theory, were
observed experimentally at slightly above the critical fluid velocity. Miles et al. [37]
demonstrated that as flow through the non-linear stiffened cantilever pipe increases,
phase-coupled harmonics of the power-spectral primary frequency grow. For parameters
close to those associated with chaotic movements, sub-harmonics become phase
coupled to the primary frequency. Cross-bi-spectra indicated that both sum and
difference interactions result in non-linear phase coupling and energy exchange between
vibration modes.

2.3 Effect of Internal Flow on Dynamic Behaviour of

Pipes

The effect of the internal flow on the dynamic behaviour of pipes conveying fluids is
essentially based on the interaction of inertia, centrifugal forces and Coriolis forces of
the fluid with the pipe structure [38]. The centrifugal and Coriolis forces occur from the
fluid flow, while the inertia force is a combination of the fluid and pipes inertias. As
mentioned above, the centrifugal force arises from the fluid flowing in a pipe that is
curved due to lateral deflection. If the element of the pipe rotates, this rotation
generates a gyroscopic or Coriolis force that is proportional to the fluid velocity. From
the previous studies, it was highlighted that the Coriolis force has a marked influence on
the dynamic pipe behaviour.

Most of the earlier contributions to the dynamics of pipes neglected most of the above
mentioned interactions effects. However, later, it was realised that at sufficiently high
internal pressure and flow speed in a relatively flexible pipe these effects are significant
and can totally change the dynamic or static behaviour of pipes [17,35,39–51]. For
example, for a sufficiently high flow velocity, the centrifugal force can overcome the
flexural restoring force and the pipe can buckle in its first mode, provided that gravity
and external damping are relatively small. This phenomena is also known as divergence
[52,53]. It was also recognised that when the fluid is stationary (the flow speed is
negligible or zero) [54], the pipe response as a classical beam to an initial disturbance.

If the centrifugal force in the pipe acts in much the same way as compressive forces in
the classical beam theory, the Coriolis force acts to restabilise the pipe after divergence
before flutter, or dynamic instability, finally destabilises the whole pipe. Once the pipe
becomes unstable, the non-linear effects associated with relatively large pipe
displacements become very important and the linear theory is then not applicable
beyond this critical flow velocity associated with the initiation of the dynamic instability
[55, 56].

The interesting findings are that the pipe systems can be categorised into two types:
gyroscopic conservative and non-conservative type system. Gyroscopic conservative
systems refer to the pipes not being allowed to move at the downstream end, and it was
concluded that such pipes cannot absorb fluid energy. Non-conservative systems take
place when the downstream end is allowed to move and fluid energy can be absorbed by
the pipe. In the subcritical flow range, the pipe systems of different types response
differently. In a gyroscopic conservative system, the Coriolis force does not contribute
to damping hence the damping values do not increase because of the increase of flow
velocity. In a non-conservative system, the Coriolis force will lead to the generation of
dissipation forces for certain modes of pipe oscillation.

Propagation of free harmonic waves in a periodically supported infinite pipe has been
studied in [57]. Due to the presence of the Coriolis force from the fluid flowing in a
curved path, it was found that the elastic waves travelling in positive and the negative
direction propagate at different speeds. As a result of this, no classical deflection modes
present in the dynamic response of the pipe.

In simplified beam theories, the centrifugal force acts in the same manner or equivalent
to a compressive end load [17]. This can be realised by comparing the governing
equation of a pipe taking into account the internal flow,
&4W &2W &2W &2W
2
( )
EI 4
+ MV
&x
 + 2MV + M+N =0 (2.13)
&x2 &x&t &t 2

with the classical governing equation describing dynamic behaviour of a pipe subjected
to a compressive load P at the pipe’s ends.

&4W &2W &2W

EI +P +N =0 (2.14)
&x 4 &x 2 &t 2

and x, t are the axial coordinate and time, respectively, EI is the flexural rigidity of the
pipe, M is the mass of fluid per unit length, flowing from the fixed end to the free one
with a steady flow velocity V, N is the mass of the pipe per unit length, and W (x, t) is
the lateral deflection of the pipe.

The first term in both equations is the flexural restoring force with ð2W⁄ðx2 ~ 1⁄R,
where R is the local radius of curvature. The second term in the former equation is
associated with the centrifugal forces as the fluid flows in curved portions of the pipe
[24]. The third term is recognised as being associated with the Coriolis acceleration, and
the last term represents the inertial effects. In this way, with increasing the flow speed,
V, the effective stiffness of the system is diminished, and for sufficiently large V, the
destabilising centrifugal force may overcome the restoring flexural force. The resulting
divergence is known as buckling or fire-hose instability.

For sufficiently small V, the dynamics of the system is dominated by the Coriolis force
2MV ð2W⁄ðxðt, and the system is subjected to the flow-induced damping. For
sufficiently large V, the centrifugal force MV2 ð2 W ⁄ ðx 2 , which may also be viewed as
a compressive follower force, overcomes the Coriolis damping effect, and the system
can lose stability from single-mode flutter [24].

If the pipe vibrates with periodic motions of period T, it was shown in [58,59,60] that
the work done by the fluid on the pipe is equal to:

T ðw 2 ðw ðw
∆W = —MV ƒ [( )+V( )( ) ] dt G 0 (2.15)
ðt L ðt L ðx L
0
where (ðw⁄ðt)L and (ðw⁄ðt)L are, respectively, the lateral velocity and slope of
the free end (for the case of a fixed-free pipe). For small magnitudes of the velocity
of the internal flow (V Σ 0), the first term dominates, and the work done is negative;
hence, the pipe loses energy to the flowing fluid, and free pipe motions are damped.
However, for the high enough velocities, V, the second term will dominate. If the
slope and velocity of the free end have opposite signs over a period, or

ðw
( ) ðw
ðt (ðx ) € 0, (2.16)
L L

then the work done may be positive, and energy flows from the fluid, to the pipe,
resulting in amplified oscillations. The fluid is a source of unbounded energy. The
opposite-sign characteristic of the free-end slope and velocity corresponds to the
dragging and lagging modal form of flutter, observed in experiments and commented in
many studies including [61,62,63]. With the situation of V € 0, for example an
aspirating system, exactly the opposite conclusions may be reached by consideration of
the same equation [24].

In [64] it was concluded that in the case of free motions, the pipe absorbs energy from
the fluid for all sufficiently small |V| and is therefore subject to flutter. For relatively
higher flow speeds |V|, the pipe loses energy to the fluid, and hence it is stabilised and
its motions are damped. The startling conclusion is that the system is unstable for
infinitesimally small |V| or if energy dissipation is taken into account, for very small
|V|.
2.4 Numerical Techniques

The effect of the flowing medium and stability conditions for circular pipes transporting
fluid or liquid has been studied with different numerical approaches, such as the FE
method and the spectral element approach. In Langthjem [65], an FE method was
developed for a cantilever pipe to investigate the critical conditions when pipe loses
stability and experience chaotic movements. The FE method was used by Liang and
Tang [66] to examine the dynamic response and stability issues of simply supported
pipe. Shizhong and et al. [67], Wang et al. [68], Cui and Tani [69] and Hongwu and
Junji [70], used FE to solve the solid-liquid coupling damping matrix and a symmetrical
solid-liquid coupling stiffness matrix. The first several natural frequencies of pipes were
determined.

Another popular method, a spectral element approach, is often used to predict the flow-
induced vibrations accurately. Lee and Oh [71] developed a spectral element model that
was formulated from the exact wave solution of the governing equations. It was claimed
that the spectral element approach provide much better accuracy in comparison with FE
for the uniform pipes regardless its length. Later, Lee and Park [72] derived coupled
governing equations for a pipe transporting liquid using Hamilton’s principle and the
classical governing equations of uncompressible fluid. The obtained non-linear
equations were liberalised about the steady-state solution using the frequency-domain
approach for the linearised equations.

Stein and Tobriner [73] presented a numerical solution to the governing equations of a
pipe conveying fluid and resting on an elastic foundation. An ideal gas and infinite pipe
approximations were utilised by Mizoguchi and Komori [74] to investigate the vibration
and stability of pipes. This study took into consideration the effect of the fluid boundary
layer. The frequencies of pipes were obtained for a wide range of the internal flow
velocities ranging from 10 m/s to Mach 3.
Wang and Bloom [75] formulated an analytical model to study the dynamics of
submerged and inclined concentric pipes of different length and subjected to different
boundary conditions. The discretised dynamic equations were solved numerically.

Lee and Chung [76] suggested a non-linear approach for a clamped-clamped pipe
transporting fluid using the classical Euler-Bernoulli beam theory described in the
previous section of this chapter and the non-linear Lagrange strain theory. Similar to the
previous study, the governing equations were discretised, linearised and solved
numerically. Yang and Jin [77] applied the Galerkin method with improved truncation
term and Maalawi and Ziada [78] presented a mathematical model for a pipeline
composed of uniform modules. As a case study, the developed model was applied to a
simply supported pipeline consisting of two and more modules.

The influence of stochastic uncertainties of flexible cantilever pipe material properties,

such as Young’s modulus, densities was investigated by Ganesan and Ramu [79]. The
statistical information was presented in terms of average values, variance and
autocorrelation functions. The stochastic equations were solved for the moments of
characteristic values and effect of the stochastic uncertainness on the dynamic behaviour
was investigated.

2.5 Experiments Investigations of Pipes subjected to

Impact Loading

Many experimental studies have been conducted in the past investigating various
aspects of pipe integrity and pipe behaviour under dynamic and impulse loading. Many
of these studies were driven by industry or military applications. The purpose of this
section is to provide a brief overview of key studies focusing on the development of test
equipment.
 A - FRAME
A
B - GUIDES

C - DROP WEIGHT

D - SPECIMEN BG
F
E - CLAMPING FIXTURE
C
F - PHOTO CELL
VIEW A
G - ACCELEROMETER

H - L.V.D.T
D
I - TABLE VIEW A
H

I
SPAN
E

Figure 2.3: Setup of drop-weight impact test [80]

In the laboratory environment, the simulation of impact loading ranges from low
velocity drop-weight tests to high velocity ballistic missile tests. For example, the
experiment conducted in [80] studied impact response of pipes by setting up a drop-
weight test to simulate low velocity impact. The test setup, shown in Figure 2.3,
included an instrumented drop weight and data acquisition system. The aims of the
experimentations were to relate the mechanical behaviour of the pipe to the loss of
kinetic energy of the impactor under low velocity loading conditions. It also aimed to
build a standardised test method for low velocity impact studies, since at that time, there
were no acceptable standard testing procedures for pipes [80].

A quasi-static behaviour of buried pipelines under localised radial (impact-like) loading

was undertaken in [81]. This study was motivated by numerous accidents during pipe
excavation. The pipelines were pressurised and radial impulse contact loads were
applied on the pipe. This study shows that the response of flat plates can approximate
the mechanics of local indentation of pressurised pipes when the indenter is of small in
comparison to the diameter of the pipeline. However, the results of this experimental
study are limited by the small range of considered geometries and loading condition.
In a number of experimental and theoretical papers, it was demonstrated that the
structure-fill medium interactions significantly influence the dynamic behaviour. An
experimental study carried out indicated that the critical perforation energy and
deformation of the wall of the pipes were significantly influenced by the presence of the
fluid (water) and fluid pressure [4]. Figure 2.4 depicts the experimental setup.

Drop hammer orbit Hammer mass

Support
Load cell
Indenter
Pressure sensor Tube

40 20 20 40

Figure 2.4: Schematic arrangement for the impact test on pressurised pipes [4]

The experiment reported that irrespective of pipe thickness, the limit perforation energy
of empty pipes was greater than that filled with water. When the thin wall tube was
filled with water and not pressurised, the limit perforation energy dropped by 20–30%,
but the energy drop in thick wall tubes are smaller. Figure 2.4 is referred and stated that
the ballistic limit of the tubes impacted by a hemispherical-nose indenter is greater than
that of the tubes impacted by the blunt and conical ones under the same impact
conditions. The thickness of the wall is also an important parameter in perforation of the
tube wall with the ballistic energy of the thick wall tube about three times that of the
thin-walled pipes. The fill liquid tends to retain the circular shape of the pipe walls and
localised the impact point. This change may lead to an increase of the perforation limit.

The experimental studies on pipes with flowing medium were conducted mainly with
rubber pipes conveying water or air and with very thin-walled metal pipes conveying
oil. The experiments were also extended to a cantilever pipes fitted with various nozzles
at the free end [82].
 Ceiling level

Supporting Strings

11 ft.
Water collecting device
Control valves

Flexible Pipe
Measuring Tank

Flow straightener

Pump

Drain pipe DrainWeighing Scales

Slip knots

Figure 2.5: Schematic diagram of the apparatus used for experiments with rubber
tubes conveying water. [82]

Figure 2.5 shows the apparatus used in experiments with rubber pipes conveying water.
The rubber tube was glued to a suitable length of rigid metal pipe that was clamped
securely in a horizontal position. The pipe was supported by a number of thin threads
hung from the ceiling and of sufficient length to ensure that at least for motions of small
amplitude, the tube effectively moved in horizontal plane. The centrifugal pump was
used in some cases to boost the available mains pressure. The water, after leaving the
flexible pipe, was collected in a tapered duct leading to a tank resting on weighing
scales. The apparatus for experiments with air flowing in rubber tubes was essentially
the same except that a volumetric flow meter was inserted in the air supply line. The air
pressure was always sufficiently low for compressibility effects to be neglected. The
experiments appear to prove that neglecting internal friction in the material of the tube
and the effect of the surrounding fluid, a universal stability curve is constructed
corresponding to conditions of neutral stability and hence separating stable and unstable
regimes. It was shown that damping, associated either with energy dissipation in the
tube material or with energy transfer to surround fluid, in certain cases may have
destabilised the system [82].
2.6 Summary

From a wide study of the literature, a few gaps in knowledge were identified:
1. There is a lack of analytical approaches and analytical solutions that can
explicitly provide the estimate of the effect of the internal flow on the dynamic
response of pipes and serve as a benchmark for numerical solutions. Meanwhile,
the speed of the flow represent a suitable parameter for the development
perturbation approaches to the investigation of the pipe dynamics.
2. Most of the previous studies were focused on vibration problems in pipes
including the calculation of natural frequencies and vibration control. Despite
the fact that the vibration and impact loading problems are closely related, it was
surprising that not much research has been conducted to investigate the dynamic
response of pipes to impulse loading, especially to understand how the internal
flow changes the maximum deflections of the pipe.
3. Despite the fact that many previous studies comprehensively investigated
problems of stability of pipes conveying fluid or gas, there were no engineering
criteria suggested to avoid unstable behaviour. With the recent release of the AS
2885.1-2007 standard, which clearly states that pipelines may now be installed
aboveground, such criteria are vital and must be established. In addition, the vast
majority of these studies were focused on a particular type of boundary
conditions, namely, cantilever-type boundary conditions.
4. A similar gap was found with the previous experimental studies. The focus of
these studies was on the investigation of the effect of the internal flow on the
integrity or fracture under impulse loading. Many studies were focusing on the
investigation of the critical flow speed, which causes the dynamic instability.
However, it seems there are no experimental studies on the investigation of the
effect of the flowing medium on the impact response of circular pipes conveying
fluid or gas.

The current study will partially address these gaps. In the next chapter, an analytical
perturbation approach will be developed and validated via a numerical method based
on central-difference scheme. It will also present an experimental rig and the
results from the experimental study. This method and validation study will be
described in Chapter 4. A detailed study on the effect of the flowing medium on the
transient response of pipes subjected to impulse loading will be presented in Chapter
5. A general conclusion will be provided in Chapter 7.
CHAPTER 3

GOVERNING EQUATION AND

ASYMPTOTIC APPROACH
3.1 Introduction

In this chapter, a simplified governing equation is formulated that describes the dynamic
response of a long flexible pipe carrying gas or liquid and subjected to transverse
impulse loading. This equation is developed from the simplified classical Bernoulli-
Euler beam theory. Further, a scaling transformation of the governing equations is
conducted to demonstrate that the dynamic response within the adopted assumptions
and simplifications is controlled by only two parameters. These parameters represent
dimensionless combinations of the elastic properties of the pipe, pipe geometry and flow
characteristics. In the case of a slight influence of the flowing medium on the overall
mechanical response of the whole system, the governing equation can be solved using
an original analytical approach that is based on the standard perturbation technique.
Several examples demonstrating the effectiveness of the developed in this thesis
approach for particular problems are considered in detail. The development of the
asymptotic approach presented in this chapter is quite general and can be generalised
relatively easily for other boundary conditions as well as for arbitrary applied loading.

3.2 Simplified Classical Bernoulli-Euler Beam Theory

This section describes a simplified mathematical model based on standard assumptions

for bending long flexible beams also adopted in [83,84,85] and many other papers. This
model will be utilised throughout this thesis (1) to analyse the transient response of a
pipe with flowing medium when the pipe is subjected to transverse impulse loading, and
(2) for investigation of the effects of parameters of the internal flow (density and
velocity) on the overall mechanical response. The classical Bernoulli-Euler beam theory
is adopted as a starting point of modelling and this theory is then further extended to
incorporate the contribution of flowing medium into the mechanical behaviour of the
system. Indeed, the lateral displacement of the pipe leads to the additional forces due to
flow inertia, centrifugal acceleration, and Coriolis force because the flow follows along
the curved deflected pipe. All these forces will be incorporated into the pipe governing
equation. The derived governing equation represents a fourth order non-homogeneous
PDE and will be solved using asymptotic and numerical approaches in Chapter 4. The
outcomes of the theoretical investigation are also validated by experiments presented in
Chapter 6.

3.3 Mathematical Modelling of Beam with Flowing

Medium

Consider a pipe element of length dx¯ as shown in Figure 3.1.

W¯

F¯(x¯, t¯)
hf V
hp

O dx¯ x¯

Figure 3.1: Deformation of pipe with flowing medium

Let hf and hp be the medium and pipe densities per unit length (say 1 meter). In the
case of the pipeline’s transverse motion, the element is subjected to the inertia force of
intensity

& 2 W¯
— (hf + hp) . (3.1)
&t¯2
As the flow rotates with angular speed & 2 W¯ ⁄(&x¯&t)¯ , the Coriolis force acting on the
element dx¯ is given by

& 2 W¯
—2 hfVdx¯ , (3.2)
&x¯&t¯
where V is the velocity of the medium. With the same sign, we can write the centrifugal
force acting on the element due to the flow that follows along the curved deflected pipe:

& 2 W¯
2
— 2
dx¯ . (3.3)
hfV&x¯

It is assumed for simplicity that the dynamics of the pipe do not affect significantly the
flow characteristics such as flow velocity and density. Thus, we can decouple the
structural and flow mechanisms of the problem. Taking also into account the elastic
response of the pipe element, dx¯, the governing dynamic equation can be written in the
following form:

&4 &2W &2 &2 2 ¯( ¯)

W¯ ¯ W¯ W¯
EI (hf + hp) + 2 hfV &x¯ 2 =
t F x¯, (3.4)
+ &x¯ 4 +
&t¯2 &x¯&t¯ h fV

where E represents Young’s modulus and I is the second moment of inertia of the
pipe. The right-hand side term F¯ (x¯, t¯) represents a driving force per unit length.
We assume that the length of the pipe is finite and 0 ≤ x¯ ≤ L and t¯ ≤ 0.

3.4 Scaling Transformation of the Dynamic

Governing Equation

It is always advantageous and often necessary to rewrite a governing equation in a

dimensionless form. To do so, the following scaling transformations are introduced:

x t W¯ (x, t)
x = , t = , W(x, t) = (3.5)
L v L
where v is a parameter that will be defined in equation (3.8a). Then, equation (3.4) can
be rewritten as

EI2& 4W (hf + hp)L Vhf & 2W V2hf & 2W

& W ¯
+ +2 +
L3 &x4 v2 &t2 v &x&t L &x2 = F(x, t) . (3.6)

It is straightforward to show the dimensional consistency of the equation. To proceed

with the solution of the governing equation, it is convenient to recast this equation into a
more compact form. Dividing both sides by EI⁄L3, we obtain

&4W (hf + hp)L4 VhfL3

V2hfL2 &2W
2 2
& W & W
+ +2 + = F(x, t) . (3.7)
&x4 EIv2 &t2 EIv &x&t EI &x2

Introduce characteristic time, v.

h + hf
2 p
v=L J . (3.8a)
EI

Now we define the following dimensionless parameters

VLhf
s= (3.8b)
JEI(hp + hf )

and

hp + h f
þ = VLJ (3.8c)
EI

and
L
( ) 3 ¯( )
F x, t . (3.8d)
Then equation (3.6) becomes
W,ssss + W,tt + s(2W,st + þW,ss) = F(x, t) . (3.9)

In the last equation, the following notation for the derivative: &W⁄&x ÷ W,s was
introduced. The similar rule is applied to the variable t as well as throughout higher
order derivatives. Observe also that following the scaling transformation (3.8) and (3.9),
the space variable x varies between 0 and 1. Finally, to solve the problem, the initial and
boundary conditions must be imposed.

In the derivation of the governing equation (3.9), the effect of axial forces on the
transverse movements was also omitted; this can be developed in the pipe as a result of
the viscosity of the flow or reactions in the pipe supports. Such forces can also
contribute to the dynamic response at large values of pipe deflections or slopes of the
deflection curve. However, in the following analysis, it is assumed that the deflections
are small and the effect of the axial forces (due to flow viscosity and support reactions)
on the transverse movements is negligible in comparison with other forces.

It is important to stress that, within the developed model, only two parameters, s and þ
fully control the transient response of the system. Therefore, the derived governing
equation (3.9) can be utilised to investigate various dynamic phenomena using reduced
size or scaled physical models. Such scaled models would be adequate if the values of
the governing parameters s and þ are kept the same for the scaled model and the
reference system. This can potentially result in substantial benefits if the experimental
approach is adopted for the investigation of the problem under consideration.

3.5 Asymptotic Solution: Cantilever Beam

A boundary-value problem for the governing equation (3.9) can be solved by means of a
standard perturbation approach utilising the smallness of parameter s. To demonstrate
the application of the perturbation approach for the analysis of the transient behaviour of
pipes, specific boundary conditions at the ends of the pipe x = 0 and x = 1 must be
prescribed. In this sub-section, consideration is given to the cantilever-type boundary
conditions, as shown in Figure 3.2, or

at x = 0: W = W,s = 0, (3.10a)

which correspond to zero displacement and zero slope at the fixed end (x = 0), and

at x = 1: W,ss = W,sss = 0, (3.10b)

which means zero resultant bending moment and shear force at the free end (x = 1).

x
0 1
F(t)
L

Figure 3.2: Cantilever beam subjected to impulse loading

3.5.1 Solution of Non-homogeneous Classical Beam Equation

The transient analysis of the derived equation (3.9) in the previous section is
commenced with a consideration of the classical non-homogeneous beam equation
subjected to the boundary conditions given in equations (3.10a) and (3.10b):

W,ssss + W,tt = F(x, t) . (3.11)

Assumption is made that the pipe was stationary at t € 0, meaning that the applied
driving load distribution

F(x, t) ÷ 0 at t € 0 (3.12)

Several solution techniques can be applied to analyse the beam’s transient response to
an impulse loading, including Laplace-Laplace transforms, the method of undetermined
coefficients, variation of parameters, and the Joint transform scheme [86] and [87]. An
exact analytical solution of equation (3.11) with boundary conditions (3.10) can be
obtained by the standard separation of variables technique and an application of Laplace
transform [87]. It can be written as

∞ 1 t
sin(m (t — v))
W(x, t) = Σ Yk(x) ƒ Yk(u) ƒ F(u, v) k dvdu (3.13)
mk
k=1 0 0

where mk = hk2 . In the previous equation, Yk are a set of normalised eigenfunctions

corresponding to the solution of the homogeneous equation, i.e. W,ssss + W,tt = 0,
subjected to the specified boundary conditions [87]:

Yk(x) = Ak[(cosh(hkx) — cos(hkx)) — ck(sinh(hkx) — sin(hkx))] (3.14)

where

cosh(hk) + cos(hk)
ck = sinh(hk) + sin(hk) (3.15)

with h1 = 1.8752, h2 = 4.6942, h3 = 7.8849, h4 = 10.9956, h5 = 14.1373 and hk

(k ≤ 6) = (2k — 1) n⁄2 [13,14].

The coefficients Ak in Eq. (3.14) are the normalisation constants that ensure the
eigenfunctions are not only orthogonal but also orthonormal. For k ≤ 12, Ak = 1,
while for k Σ 12, Ak departs from unity and must be determined (see for example
[87]).

Equation (3.13) provides the general analytical expression for the transient spatio-
temporal dynamics of a cantilever beam driven by arbitrary load distribution, F(x, t).
As an example, consider a spatio-temporal impulse driving force at time t = 0 with a
magnitude fa (in scaled variables). We use Dirac’s time-and-space delta function to
model this impulse. Then, F(x, t) takes the form

F(x, t) = fað(x — x×)ð(t) (3.16)

where x× represents the position along the pipe at which the driving impulse is applied.
Carrying out the space and time integrations and invoking the basic properties of
Dirac’s delta function, we obtain

œ
sin(m t)
×
W(x, t) = fa Σ Yk(x)Yk(x ) k . (3.17)
mk
k=1

The above solution is also well known in the literature and describes the mechanical
response of a cantilever beam subjected to a spatio-temporal impulse [88–93]. This
relationship (3.17) also describes the exact analytical solution of the corresponding
problem for a pipe with flowing medium (3.9) when parameters s as well as þ are
limited to zero. Such situations, for example, correspond to the case when the velocity
of the flowing medium is negligible, or when the pipe has a large bending stiffness (EI).

3.5.2 Perturbation Approach

The case of a cantilever pipe with flowing medium can be related to a practically
important situation in which, for example, the pipe is ruptured circumferentially due to
an accident such as fracture propagation and is subjected to a dynamic loading. In this
case, the simplified governing equation (3.9) was derived in the previous section and,
for convenience of the reader, is rewritten below:

W,ssss + W,tt + s(2W,st + þW,ss) = F(x, t) .

In contrast to the previously considered case (the classical Bernoulli-Euler beam

equation neglecting the internal flow considered earlier), an attempt at a standard
separation of variables approach to determine the dynamic response fails because it is
impossible to satisfy the boundary conditions with a nontrivial solution. However, as
was mentioned in Section 3.4.1, at fixed þ, (3.18) reduces to the classical Bernoulli-
Euler equation when s is limited to zero. Alternatively, the problem can be solved as a
regular perturbation problem developed by the candidate in [83] with, s as a small
parameter; that is, the solution has the form

W = W0 + sW1 + s 2 W2 + … s N WN + … . (3.18)

If the expansions (3.18) are substituted into (3.9) and like powers of s are equated, then
the following expressions for the transverse deflection of W, correct up to order s N , are
obtained:

W0 ,ssss + W0 ,tt = F(x, t)

W1 ,ssss + W1 ,tt = F0(x, t) F0(x, t) = —2W0 ,st — þW0 ,ss

W2 ,ssss + W2 ,tt = F1(x, t) F1(x, t) = —2W1 ,st — þW1 ,ss

(3.19)

...

WN ,ssss + WN ,tt = FN–1 (x, t) FN–1 (x, t) = —2WN–1 ,st — þWN–1 ,ss .

These equations are subjected to the specified boundary conditions such as (3.10). This
sequence of equations can be considered a recurrent system of differential equations that
can be solved through a step-by-step integration using the general solution for the
cantilever beam subjected to arbitrary load distribution (3.13). This solution can also be
written using the earlier considered solution for the classical Bernoulli-Euler beam
(3.13) as

∞ 1 t
sin(m (t — v))
WN (x, t) = Σ Yk (x) ƒ Yk (u) ƒ FN–1 (u, v) k
dvdu . (3.20)
mk
k=1 0 0

The above recurrent equations (3.19) and general solution for non-homogeneous beam
equation (3.20) provide an analytical tool for the analysis of the effect of the flowing
media on the transient response of pipes. Next, the different boundary conditions are
considered to demonstrate further the developed approach.

3.6 Asymptotic Solution: Simply Supported Beam

For this sub-section, consideration is given to the simply supported type of boundary
conditions, as illustrated in Figure 3.3, or

at x = 0: W = W,ss = 0, (3.21a)

and

at x = 1: W = W,ss = 0 . (3.21b)

These conditions correspond to zero displacement and bending moment enforced at

both ends of the pipe.
 W

x
0 1
F(t)

Figure 3.3: Simply supported beam subjected to impulse loading

3.6.1 Solution of Non-homogeneous Classical Beam Equation

Similar to sub-section 3.4, the driving load distribution F¯ (x, t) = 0 at t € 0 is

assumed. First, consideration is given to the solution of the non-homogeneous equations
for the classical Bernoulli-Euler beam corresponding to the prescribed boundary
conditions (simply supported pipe). This solution will be further utilised in the
asymptotic approach for a pipe with flowing medium. An analytical solution of equation
(3.11) with boundary conditions (3.21) can also be obtained by the separation of
variables technique and Laplace transform [86] and can be written as

∞ 1 t
sin(m (t — v))
W(x, t) = Σ Yk(x) ƒ Yk(u) ƒ F(u, v) k dvdu (3.22)
mk
k=1 0 0

where Yk are the set of normalised eigenfunctions corresponding to the solution of the
homogeneous equation W,ssss + W,tt = 0, subjected to the specified boundary
conditions:

Yk(x) = √2 sin hkx, hk = kn (3.23)

and
mk = h2k = (kn)2 . (3.24)

The coefficient √2 in eigenfunctions, equation (3.23), is a normalisation constant that

ensures that the eigenfunctions are not only orthogonal but also orthonormal. Further,
the perturbation approach is developed for the simply supported beam.

3.6.2 Perturbation Approach

Similar to the cantilever case considered previously, the problem can be solved as a
regular perturbation problem with  as a small parameter; that is, the solution has the
form

W = W0 + sW1 + s 2 W2 + … s N WN + … . (3.25)

If this expansion is substituted into (3.9) and like powers of s are equated, then
the similar recurrent expressions for the transverse deflection of W, correct up to order
s N , can be obtained:

W0 ,ssss + W0 ,tt = F(x, t)

W1 ,ssss + W1 ,tt = F0(x, t) F0(x, t) = —2W0 ,st — þW0 ,ss

W2 ,ssss + W2 ,tt = F1(x, t) F1(x, t) = —2W1 ,st — þW1 ,ss (3.26)

...

WN ,ssss + WN ,tt = FN–1 (x, t) FN–1 (x, t) = —2WN–1 ,st — þWN–1 ,ss .

Based on the similar approach as adopted in the case of cantilever pipe and omitting
computational details, the expansion series terms are

œ 1 t
sin(m (t — v))
WN (x, t) = 2 Σ sin(hk x) ƒ sin(hn u) ƒ FN–1 (u, v) k dvdu . (3.27)
mk
k=1 0 0
The above recurrent equations (3.26) and general solution for non-homogeneous beam,
equation (3.27) with hk as given by (3.23), provide an analytical procedure for the
analysis of the effect of the flowing media on the transient response of simply supported
pipes. Further, several examples are considered to demonstrate the developed approach.
For example, first consider a harmonical impulse driving force at time t = 0 with a
magnitude fn (in scaled variables); that is

F(x, t) = fn sin(hnx) ð(t) (3.28)

where n is a positive integer number and hn is defined by equation (3.24).

Then, in accordance with equations (3.19),

Wn (x, t) = sin(hn x) sin(mnt) (3.29)

mn
0 fn

which also describes the transient response of pipe with a stationary filled media. In
accordance with equation (3.24),

Fn0 (x, t) = —2fn hn cos(hn x) cos(mn t) + þ sin(hn x) sin(mn x) . (3.30)

Then, carrying out all integrations and after simplifications, the first-order correction of
the deflection function can be written as a sum of two terms:

Wn1(x, t) = WCor (x, t) + WCen(x, t) , (3.31)

n1 n1

where

∞
Cor k(1 — (—1)n+k) cos(mkt) — cos(mnt)
Wn1 = —4hn fn Σ sin(hk x) ( ) (3.32)
k=1
n(k + n)(k — n) (mn + mk)(mn — mk)

and
þfn
WCen(x, t) = sin(h x) (sin(m t) — cos(m t)m
t) . (3.33)
n1 n n n n
2mn2
 Cor
The first term, Wn1 (x, t), represents the contribution of Coriolis force and the second
Cen
term, Wn1 (x, t) is due to the centrifugal acceleration of the flowing media through the
Cor
curved pipe. It is interesting to note that the first term, Wn1 (x, t), is limited in time;
however, the second term, WCenn1(x, t), in the first-order correction is growing with time,
and is thus demonstrating a synergetic effect of the flowing media in the structural
response of the system. Both terms are continuous functions with zero values at t = 0.
This means that the effect of the flowing medium on the transient response is small
within relatively small time intervals from the start of the impulse loading. This effect is
characterised by the dimensionless parameter v (see equation (3.8)). In the case of an
arbitrary driving impulse force, the spatial component can be represented in Fourier
series as

F(x, t) = ð(t) Σ fn . (3.34)

sin(hnx)
n=1

Using the principle of superposition, the general solution can be found by the simple
summation of the particular solutions for each harmonic. For example, if the system is
subjected to a spatio-temporal impulse driving force, i.e.

F(x, t) = fa · ð(x×)ð(t) (3.35)

where fa is the amplitude of the driving impulse and x× is the point at where the force is
applied, then, the s0 - order solution for the pipe deflection is

∞
sin(m t)
W0(x, t) = 2fa Σ sin(hnx) sin(hnx×) n . (3.36)
mn
n=1

The first-order correction of this solution, which incorporates the flowing medium
effects, can be found from equation (3.31) as

W1(x, t) = Σ fnWn1(x, , (3.37)

t)
n=1
where Wn1 is given by Eq (3.31) and fn = 2fa sin(hnx×).

In the next chapter, a numerical approach is presented for the solution of equation (3.9)
in the case of arbitrary (not small) controlling parameters and the obtained analytical
solutions are used to validate the numerical approach.

3.7 Summary

In this chapter, the governing equation was derived for a long flexible pipe with flowing
medium based on the classical Bernoulli-Euler beam theory. After that, scale
transformations were conducted to demonstrate that only two parameters, (s and þ), that
govern the transient dynamics of the pipe. This finding can be utilised to investigate
various dynamic phenomena using reduced size or scaled physical models. Such scaled
models would be adequate if the values of the governing parameters s and þ are kept
the same for the scaled model and the reference system. Of course, the physical models
have to meet all other conditions corresponding to the adopted assumptions in
mathematical modelling. For example, the pipe has to be sufficiently long and flexible,
and the loading should not significantly affect the internal flow characteristics.

Further, an asymptotic approach was developed to investigate the transient response of

pipes subjected to impulse loading. The approach is based on the standard perturbation
technique. The solution represents a recurrent system of equations, each term of which
can be obtained by integration. The application of this approach was demonstrated to
obtain the first-order correction to the classical solution for the corresponding problem
of both a cantilever beam and a simply supported pipe transporting gas or liquid.
However, if higher order terms are required, the analytical procedure could become
quite cumbersome and time-consuming. Therefore, the derived asymptotic solutions
will be used mainly to validate important results obtained by a more comprehensive
numerical method to be developed in the next Chapter.
CHAPTER 4

NUMERICAL APPROACH AND

VALIDATION STUDY
4.1 Introduction

In Chapter 3, an analytical perturbation approach was developed to obtain solutions for

particular problems and analyse the governing equation describing dynamic response of
a circular pipe transporting gas or liquid. Due to the asymptotic nature of the developed
approach, it has obvious limitations. In particular, the perturbation approach is only
applicable to the situations in which the influence of the parameters of the flowing
media on the transient response of the pipe is relatively small. Such situations, for
example, take place when the velocity and density of the internal flow are low or when
the flexural rigidity of the pipe is relatively high. It seems, that other analytical
approaches, such as the Fourier transform approach or integral equation method, for
analysis of the governing equation, which represents a forth order PDE, are too
cumbersome and can lead to significant technical difficulties. Therefore, a more
simplified numerical approach is developed, which will be used in the next chapter,
Chapter 5 to investigate comprehensively the dynamic response of pipes subjected to
impulse loading as well as to study the effects of the flowing medium on the dynamic
response.

The numerical approach to be developed in this chapter is based on the central finite-
difference scheme. This numerical scheme including convergence and accuracy issues
is well investigated and widely described in the literature. It was previously applied to
study similar dynamic problems, see for example [87]. In this scheme, the
spatial- temporal domain of interest is partitioned with a rectangular uniform grid
of mesh size: ∆x × ∆t, where ∆x represents the spatial grid size and ∆t denotes
the temporal grid size. The upper bounds of the grid point in the spatial and temporal
directions will be denoted by Ns and Nt , respectively. Consequently, the spatial
domain is limited by points n = 0 and n = Ns ; and the time domain spreads from
N = 0 to N = Nt . The discredited two-dimensional spatial-temporal domain is
shown in Figure 4.1. The inclusion of additional spatial points at x € 0 and x
Σ 1 is often necessary to implement the specified boundary conditions, which
will be discussed later in this chapter.
The following notations are used: the transverse deflection of the beam at a grid point
(xn , tN ) will be denoted as W (xn , tN ) or, simply as Wn,N .

Tine ∆x

Nt

N Wn,N

2
∆t
1

-1 0 1 n Ns Ns + 2 Space

Figure 4.1: Finite-difference grid

The dynamic deformation behaviour of the pipe with flowing medium within the
adopted assumptions and simplifications is given by equation (3.9), which can be
written in a dimensionless form; and it is repeated below:

W,ssss + W,tt + s(2W,st + þW,ss) = F(x, t) ,

where the first term in this equation (W,ssss) represents the elastic response of the pipe,
the second term (W,tt) is a contribution from the inertia force of pipe material and
flowing medium; (2sW,st) is the contribution from the Coriolis force on the pipe
behaviour, (sþW,ss) represents the contribution from the centrifugal force of the flow
on pipe because the flow follows along the curved deflected pipe, and the right side of
this equation, F(x, t), is the applied impulsive force.

In the following the above dimensionless PDE will be discredited and solved using the
central finite-difference scheme for various boundary conditions specified at the ends of
the pipe, i.e. at points x = 0 and x = 1.
4.2 Central-Difference Numerical Scheme

To implement the central finite-difference scheme, we will first write down the discrete
representation of the derivates comprising equations (3.9) or W,ssss, W,tt, W,st and W,ss.
Using Taylor series expansion, we obtain the following representation of the partial
derivatives:
ðW (n, N)
W (n, N) ÷
,t
ðt

W(n, N + ∆t) — W(n, N — ∆t) 2

= 2∆t + 0(∆t )
W(n, N + 1) — W(n, N — 1) (4.1)
= 2∆t

and

ð 2 W(n, N)
W,tt (n, N) ÷
ðt 2

W (n, N + ∆t) — 2W (n, N) + W (n, N — ∆t)

= ∆t2

W(n, N + 1) — 2W (n, N) + W (n, N — 1)

= ∆t2 (4.2)

and

ð 2 W(n, N)
W,ss (n, N) ÷
ðx 2

W (n + ∆t, N) — W (n, N) — W (n, N) + W (n — ∆t, N)

= ∆x2

W (n + 1, N) — 2W (n, N) + W (n — 1,
=)
N
∆x2 . (4.3)

In a similar vein, the spatial derivative term is approximated as

W,st (n, N) ÷
ð2 (n, N)
W ðxðt

W(n + 1, N + 1) — W(n + 1, N — 1)
= 4∆t∆x —

W (n — 1, N + 1) — W (n — 1, N — 1)
4∆t∆x (4.4)

and

ð 4 W(n, N)
W,ssss (n, N) ÷
ðx 4

W (n + 2, N) — 4W (n + 1, N) + 6W (n, N)
= ∆x4 —

4W(n — 1, N) + 4W(n — 2, N)
∆x4 . (4.5)

Substituting the above equations into the governing equation (3.9) the final discrete
representation of the partial deferential equation can be written in the following form

Wn,N+1 s ∆t
2 ( ∆x ) [Wn–1,N+1 — Wn+1,N+1 ]
=
2
∆t
—( ) [Wn+2,N — 4Wn+1,N + 6Wn,N — 4Wn–1,N +Wn–2,N ]
∆x2

+[2Wn,N — W ] s ∆t ) [Wn+1,N– — Wn–1,N–1 ]

+ n,N–1 ( 1
2 ∆x
2
∆t
+þs ( ) [Wn+1,N — 2Wn,N + Wn–1,N ]
∆x
+ ∆t 2F(x, t) . (4.6)

The central-difference scheme under consideration has a truncation error of 0[∆x4 +

∆t2]. The sufficient condition for stability of this scheme is the condition that
∆t⁄∆x2 € 0.5, which indicates that the finite-difference method is stable if the error
goes to zero as the truncation error goes to zero. Further, when the both governing
parameter, s and þ, equal to zero, we can derive the classical dimensionless form of the
beam governing equation (3.11):

W,ssss + W,tt = F(x, t) .

The finite-difference representation of the last equation is

∆t 2
Wn,N+1 =— ) [Wn+2,N — 4Wn+1,N + 6Wn,N — 4Wn–1,N +Wn–2,N ]
∆x 2
(

+[2Wn,N — Wn,N–1 ] + ∆t 2 F(x, t)

(4.7)

which is a particular case of the derived equation (4.6).

4.3 Initial and Boundary Conditions

4.3.1 Initial Conditions

In the following sub-section, consideration is given to three types of boundary

conditions. These are fixed support, free end and simply supported boundary conditions.
The initial conditions (deflection and the velocity of the pipe) for all cases are assumed
to be zero:

W(x, 0) = 0 (4.8)

and
Wt(x, 0) = 0 . (4.9)

In the finite-differences representation, the initial conditions are reduced to the

following expressions:
Wn,1 = 0 (4.10)
and

Wn,2 = 0 . (4.11)

Next, consideration is given to the boundary conditions for all cases under
consideration.

4.3.2 Boundary Conditions

At the fixed end (which is assumed to be the left end), the deflection and slope have to
be zero. It means that for any time t

W(0, t) = 0 (4.12)

and

Ws(0, t) = 0 . (4.13)

In terms of finite-differences representation, these can be expressed as follows:

W0,N = 0 (4.14)

and

W–1,N = W1,N (4.15)

which is valid for any time step N.

The boundary conditions at the free end (without losing the generality, we will assume
that these conditions apply to the left end of the pipe or at x = 0) are

Wss(0, t) = Wsss(0, t) = 0 (4.16)

which means that the bending moment and shear force have to be zero. These
conditions can be expressed in terms of the finite differences as
W1,N — 2W0,N + W–1,N = 0 (4.17)

and
1 1
W2,N — W1,N + W–1,N — W–2,N = 0 . (4.18)
2 2

Finally, for simply supported boundary conditions, the deflection and bending moment
have to be zero, or

W(0, t) = 0 (4.19)

and

Wss(0, t) = 0 .(4.20)

The corresponding finite-differences representations of these conditions are

W0,N = 0 (4.21)

and
W1,N — 2W0,N + W–1,N = 0 (4.22)

One can see that many boundary conditions include virtual special points n = —1 and
n = —2 for the left end of the pipe. The similar points have to be introduced for the
right part of the part, or at n = Ns + 1 and n = Ns + 2 to model the appropriate
boundary condition. They are shown as dashed line in Figure 4.1.
4.4 Computational Flow

In this section, the matrix approach is introduced to obtain a computational numerical

solution. Once the central finite-difference scheme has been established, the evolution
equation (4.6) is set in a matrix form. The number of unknowns must match the number
of equations and for this reason, the initial and boundary conditions are merged with the
main numerical scheme of equation. The matrix forms of the equations are developed
for the three earlier considered boundary conditions: fixed end, free end and simply
supported end. The evolution equation together with the boundary conditions is
combined into the global evolution matrix A and it is then programmed in MATLAB
and solved numerically as follows:

AW=R (4.23)

(A–1A) W = A–1 R (4.24)

and

IW = A–1R (4.25)

where I is the identity matrix and R represents the applied loading F(x, t), which is a
function of the position and time. The complete MATLAB code is listed in Appendix C.

As mentioned above, for solution using the finite-difference scheme, the natural grid
(the natural domain boundaries are within the bold lines in Figure 4.1) must include
virtual grid points (on vertical dashed lines) to accommodate the appropriate boundary
conditions. By examining the boundary condition, one needs to extend the natural grid
of size Ns × Nt by introducing the virtual nodes. These grid nodes correspond to indices
n = —1 and n = —2 on the left-hand side of the spatial grid representation, and
to n = Ns + 1, Nt + 2 on the right-hand side.

To generate the first Wn,3 grid point, the central-difference scheme must be combined
with known initial and boundary conditions. The computational flow proceeds as
follows. First, the initial conditions require that these equations be evaluated, where the
range of the space index, n, extends from —1 or —2 to Ns + 1 or Ns + 2 (depending on
the boundary conditions). Then, the time increment evolution begins. For
each successive time step N, starting N = 3, the general formulation in equation
(4.9) is applied to compute Wn,N+1 . For this expression, the spatial component
index varies between 1 and Ns . Thereafter, the boundary equations specified by
the boundary conditions, equations (4.12) to (4.22), are applied. In order to complete
the list of data needed to compute numerical results from either analytical formula
or the finite- differences scheme, we need to specify the grid parameters. In this
study, the scaled values selected ∆x = 0.025 and ∆t = 0.00005. These results in
stability factor of 0.08, much lower than the minimum required ∆t/∆x 2 € 0.5. The
time step ∆t = 0.00005 was selected to insure convergence of the numerical results
with 5% error in the time interval0 € t € 1.

A shifted spatio-temporal delta impulse is modelled as a point load applied to a specific

grid node within a short period, normally ∆t = 0.025. A comparison of the finite-size
spatio-temporal load impulse with analytical results obtained for delta impulse, which
will be presented later in this chapter, demonstrated that the results are similar. The total
number of simulation time steps is arbitrary. However, as time progresses, numerical
error accumulate, resulting in a slight time lag due to the well-known limitations in
accuracy of the finite-differencing numerical schemes. The computational flow is shown
in Figure 4.2.
 Set values of pipe (dimension,
properties, time, etc)

Set dimensionless parameters

of the boundary-value problem

Set initial values and boundary

conditions

Yes N of steps No
achieved? Time loop

End

Form global evolution matrix

Calculate W in natural and

virtual nodes for the next step

Figure 4.2: Computational flow

4.5 Validation Study

Before conducting the study on the effect of flowing media parameters on the dynamic
response of a pipe in Chapter 5, it is first necessary to validate the developed numerical
and analytical approaches. This will be achieved through the comparison of the
numerical results with the known analytical solutions described in Chapter 3, as well as
by comparison of the numerical method results with the perturbation method, which is
possible at small values of s and þ, where the perturbation approach can be applied to
model the dynamic response of the pipe. Below, the consideration is given to the three
particular cases that demonstrate the accuracy of the developed approaches.

4.5.1 Numerical Approach versus Exact Analytical Solution

In this sub-section, the results of the calculations are compared using the developed
numerical approach with the exact analytical solution presented in Section 3.6. Recall
that the dynamic behaviour of the pipe is controlled by two dimensionless parameters s
and þ. In the case of a finite value of parameter þ and when s = 0, the governing
equation (3.9) can be reduced to the classical beam equation; and the dynamic
deflection of pipe (beam) can be obtained analytically. It can be written in the form of a
series as described in Chapter 3, equation (3.13) and given below:

∞ 1 t
sin(m (t — v))
W(x, t) = Σ Yk(x) ƒ Yk(u) ƒ F(u, v) k dvdu
mk
k=1 0 0

where mk = h2k , k number of modes in the series solution. In the previous equation, Yk
are a set of normalised eigenfunctions corresponding to the solution of the
homogeneous equation, i.e. W,ssss + W,tt = 0. For a cantilever

Yk(x) = Ak[(cosh(hkx) — cos(hkx)) — ck(sinh(hkx) — sin(hkx))] ,

cosh(hk) + cos(hk)
ck = sinh(hk) + sin(hk) ,

with h1 = 1.8752, h2 = 4.6942, h3 = 7.8849, h4 = 10.9956, h5 = 14.1373 and hk

(k ≤ 6) = (2k — 1) n⁄2.

The numerical calculations are conducted for a pipe with fixed left end (x = 0) and
while the right end (x = 1) is free. The corresponding boundary conditions are given by
equations (4.12) to (4.22). The results of the numerical and analytical calculations are
shown in Figure 4.3 for the displacement of the pipe’s free end (x = 1) for the total
of 2 and 10 first terms in the analytical series solutions (3.16). In this figure, the
pipe displacement W was normalised by the magnitude of the applied load. The
impulse loading of magnitude, fa, is applied at the half-way point along the pipe
length (x× = 0.5) for duration of ∆t = 0.025. The analytical solution based on first
10 term series expansion agrees very well with the results obtained from the
numerical approach with the spatial and time steps: ∆x = 0.02 (or Ns = 40) and ∆t
= 0.00005, respectively. In this case, the both graphs are virtually not
distinguishable, which indicates a very good accuracy of the numerical approach as
well as applicability of the finite-size spatio- temporal distribution adopted in the
numerical approach to model impulse loading.

W(1, t)
Normalised tip deflection

fa Numerical
Analytical (2 terms)
Analytical (10 terms)
0.4

0.2

-0.2

-0.4
0 0.2 0.4 0.6 0.8 t
Non-dimensional time

Figure 4.3: Comparison of the exact analytical (3.16) and numerical solutions for
ε=β=0

As time increases, the error in the numerical solution starts to build up due to the
numerical approximations and the finite spatial and temporal grid sizes. Reducing the
grid sizes will further improve the numerical results; however, this will be at the cost of
increased computational requirements.
4.5.2 Numerical Solution versus Asymptotic Analytical
Approach

For small values of the controlling parameters, the developed perturbation approach via
equations (3.18) to (3.20) is implemented. Figure 4.4 displays the first-order
perturbation correction (only one term was retained from the analytical asymptotic
solution) for the case of s = 0.2 and þ = 0.1, along with the numerical results.
Calculations were performed for a cantilever pipe the left end of which was fixed and
the right end is free. The results shown in Figure 4.4 are for the pipe free-end
displacements in the case of an impulse load applied as described previously at the pipe
mid-point. The numerical and analytical results again match well with any differences
being due to only using the first-order perturbation correction and, as before, the finite
approximations of the numerical approach.

W(1, t)
Normalised tip deflection

fa Numerical Method
Perturbation Method Analytical (ε = β = 0)

0.4

0.2

-0.2

-0.4
0 0.2 0.4 0.6 0.8 t
Non-dimensional time

Figure 4.4: Comparison of the analytical and numerical solutions for ε =0.2 and
β = 0.1, analytical solution for ε = β = 0 is given for comparison only
The following results describe the transient response of a simply supported pipe.
The impulse load of magnitude, fa, is applied, in this case, again at the half-way
point along the beam length (x× = 0.5). Figure 4.5 displays the first-order
perturbation correction calculations for the case of s = 0.2 and þ = 0.1, along with
the numerical results, for the displacements in the centre of the pipe. In this figure,
the pipe displacement W has been normalised by the magnitude of the applied load,
fa, and the following computational parameters: ∆x = 0.02 (or 40 subdivisions) and
∆t = 0.00005 is used.

As can be seen from Figure 4.5, the numerical and analytical results match very well.
Although both graphs are almost not distinguishable, as time increases, the error in the
numerical and analytical solutions starts to build up due to the numerical
approximations, and the asymptotic nature of the perturbation approach.
Normalised mid-point deflection

W(1, t)
Numerical Method
fa
Perturbation Method
0.4

0.2

-0.2

-0.4

0 0.2 0.4 0.6 0.8 t

Non-dimensional time

Figure 4.5: Comparison of the analytical and numerical solutions for ε =0.2 and β
= 0.1 (the numerical and perturbation graphs are not distinguishable in this case)
4.6 Summary

In this chapter, the central finite-difference scheme was developed for the analysis of
the governing equation describing the dynamic behaviour of pipes with flowing
medium. The variables in this approach, such as the time step and number of spatial grid
points, were selected from computational results testing the convergence and stability of
the numerical scheme. Further, a validation study was completed that demonstrated a
high accuracy and good convergence of the numerical approach. The numerical
approach was also used to validate the previously developed perturbation approach (see
Chapter 3). This demonstrated that both approaches are in agreement and can be
confidently applied. Due to technical difficulties with the development of analytical
solutions using perturbation approach as well as its limitations (the assumption that the
flow has a relatively small influence on the pipe response), the numerical approach
developed in this chapter will be a major tool for the investigation of the effect of
parameters of the flowing medium on the transient response of the pipe to be presented
in Chapter 5.
CHAPTER 5

EFFECT OF THE INTERNAL FLOW

ON DYNAMIC RESPONSE
5.1 Introduction

This chapter presents a systematic study of the effect of the flowing medium on
dynamic response of circular pipes subjected to impulse loading. The numerical
approach developed in the previous chapter is used to investigate the influence of the
boundary conditions, flow speed and its density on the dynamic behaviour of pipes. It
was demonstrated in Chapter 3 that within the adopted assumptions and simplifications,
the dynamic response of a pipe transporting gas or liquid is governed by two
dimensionless parameters

VLhf
s= (3.8b)
JEI(hp + hf )

and

hp + h f
þ = VLJ (3.8c)
EI

where hf and hp are the fluid and pipe densities per unit length; E represents Young’s
modulus of the pipe’s material, I is the second moment of inertia of the cross-sectional
area of the pipe, L is the characteristic length of the pipe and V is the speed of the
flowing gas or liquid or their mixture.

Therefore, the effect of the flowing medium on the dynamic response of pipes will be
expressed through these two parameters. An increase of the flow speed or its density
will lead to the increase of s and þ. While parameters s and þ are proportional to the
flow speed, the change of the flow density (linear density) is proportional to the product
of s and þ.

First, general features of the dynamic response of pipes will be investigated and
characterised. After that, a sensitivity study will be carried out to quantify the effect of
the above dimensionless parameters on the dynamic response of the pipe subjected to
impulse loading. Then, an important dynamic effect, a possible unstable behaviour of
the pipe, so-called pipe whip, will be investigated numerically and analytically. The
effect of the boundary (support) conditions will be studied leading to a simple
engineering criterion that can be used to avoid the unstable behaviour of a pipe during
accidents such as, for example, full bore rupture of a high-pressure pipeline. Further
examples will demonstrate that the developed criterion is relevant to the typical high-
pressure aboveground pipelines, which are now allowed to be built in remote areas.

5.2 General Features of Dynamic Response

The numerical simulations with various controlling parameters s and þ reveal two
principally different types of behaviour of a pipe subjected to impulse loading: stable
decay behaviour (a typical response is shown in Figure 5.1) and unstable behaviour (a
typical response is shown in Figure 5.2). As in the previous chapter, the impulse load of
magnitude fa was applied at the mid-point of the pipe. The figures plotted below
represent the normalised deflection of the tip mid-point of pipes as a function of the
non-dimensional time.
 W(1, t)
fa

Normalised tip deflection

0.4

-0.2
0 1 2 3 4 t
Non-dimensional time

Figure 5.1: Example of stable response of a cantilever pipe with flowing medium
(at ç=0.5 and Q=10)

The numerical results were obtained with a spatial grid size of ∆x = 0.02 and time
step, ∆t = 0.00005, this provides the stability factor of 0.125, which is well below the
critical factor, 0.5. From the validation study in the previous sections, it was
demonstrated that these parameters ensure stability of the numerical finite-difference
scheme as well as good accuracy.

The stable type behaviour (see Figure 5.1) is associated with a transient response of the
pipe having a magnitude that decays with time. The effect of the flowing medium in this
case is that it provides a damping on the transient pipe response.
 W(1, t)
fa

Normalised tip deflection

10

-5

-10

-15
0 0.2 0.4 0.6 0.8 t

Non-dimensional time

Figure 5.2: Example of unstable response of a cantilever pipe with flowing medium
(at ε =0.5 and β = 60)

The opposite takes place for the unstable type of dynamic behaviour. In this case, the
presence of the internal flow serves as a source of an additional energy, powering the
development of growing unbounded deflections, i.e. dynamic instability (see Figure
5.2). It was found numerically that the unstable behaviour is a property of the system
(pipe-flowing medium) and is independent of the location and amplitude of the applied
impulse loading regardless of the location of the applied force. It means that regardless
of the magnitude and location of the applied dynamic loading, the pipe will experience
growing and unlimited deflections. Of course, in practical situations, the amplitude of
the pipe deflections will be limited due to non-conservative energy consumption
mechanisms, such as friction or thermal dissipation.

Later in this chapter, the critical conditions of the dynamic instability will be
investigated in detail using numerical and analytical approaches. Further, the effect of
variation of the dimensionless controlling parameters on the dynamic response will be
systematically studied.
5.3 Effect of Dimensionless Parameters and Boundary
Conditions on Dynamic Response

Figures 5.3 and 5.4 show the transient beam deflection of the tip of the cantilever pipe
(fixed-free support condition) subjected to impulse loading for various combinations of
dimensionless parameters s. and þ.

W(1, t)
fa ε=0
Normalised tip deflection

ε = 0.5
ε=1
0.4 ε = 1.5

0.2

-0.2

0 0.2 0.4 0.6 0.8 t

Non-dimensional time

Figure 5.3: Transient response of a cantilever pipe for various values of ç while
parameter Q is kept constant (Q = 10)

Figure 5.3 depicts a transient response of a cantilever pipe for various values of s while
parameter þ is kept constant. From this figure, one can see that the magnitudes of the
normalised deflections are decreased with increase of s. Thus, this parameter can be
associated with damping properties of the system and the increase of this parameter
leads to a faster dissipation of the energy supplied by the impact loading.
Figure 5.4 shows a transient response of a cantilever pipe for various values of þ while
dimensionless parameter s (both parameters are defined in the beginning of this chapter
as well as in Chapter 3) is kept constant.

W(1, t)
fa
β=0
Normalised tip deflection

β = 10
β = 20
β = 30

0.4

-0.4
0 0.2 0.4 0.6 0.8 t

Non-dimensional time

Figure 5.4: Transient response of a cantilever pipe for various values of Q while
dimensionless parameter ç is kept constant (ç = 0. 5)

In the above figure, the parameter þ was set at 0, 10, 20 and 30 with s = 0.5. In contrast
to the previous numerical test, the magnitudes of the normalised deflections are
increased with time with the increase of the magnitude of þ. Thus, higher values of
parameter þ lead to decreasing the damping properties and, likely to lead to the
initiation of the dynamic instability of the system.

To confirm these general tendencies, a number of simulations will be presented for

different boundary conditions, which encastre pipe, propped cantilever pipe and simply
supported pipe.
5.3.1 Encastre Pipe

Figure 5.5 shows a transient response of an encastre pipe (fixed-fixed support

condition), for various values of s while parameter þ is kept constant.

Normalised mid-point deflection

W(1, t)
ε=0
fa ε = 0.5
ε=1
ε = 1.5
0.2

0.2

-0.4

0 0.2 0.4 0.6 0.8 t

Non-dimensional time

Figure 5.5: Transient response of an encastre pipe for various values of ç

while parameter Q is kept constant (Q = 10)
W(1, t)
Normalised mid-point deflection β=0
fa
β = 10
β = 20
β = 30
0.2

0.2

-0.4

0 0.2 0.4 0.6 0.8 t

Non-dimensional time

Figure 5.6: Transient response of an encastre pipe for various values of Q while
dimensionless parameter ç is kept constant (ç = 0. 5)

From Figures 5.5 and 5.6, one can see that the effect of flow damping when changing
dimensionless controlling parameters is significantly weaker in comparison with the
previous case (cantilever pipe). This is obviously connected with the boundary
conditions, which in the current case provide more resistance against any unstable
behaviour, similar to the classical theory of instability of long flexural beams. However,
the general tendencies essentially stay the same: an increase of parameter s leads to a
faster dissipation of the energy supplied by the impact loading and an increase of
parameter þ leads to a reduction of the dissipation due to the internal flow.
5.3.2 Propped Cantilever Pipe

Figure 5.7 shows a transient response of a propped cantilever pipe (fixed at the left end
and simply supported at the right condition) for various values of s while parameter þ is
kept constant. All simulations are limited by dimensionless time t = 1, after this time,
the dynamic behaviour is essentially the same; however, the magnitude of the
deflections steadily decays. One can see two major modes of the dynamic response: one
is associated with global mode of vibration of the pipe and the second mode is related to
the transient response of the pipe to impact loading.
Normalised mid-point deflection

W(1, t)
fa ε=0
ε = 0.5
ε=1
ε = 1.5
0.2

-0.2

-0.4

0 0.2 0.4 0.6 0.8 t

Non-dimensional time

Figure 5.7: Transient response of a propped cantilever pipe for various values of ç
while parameter Q is kept constant (Q = 10)

Figure 5.8 shows a transient response of a propped cantilever pipe for various values of
þ while dimensionless parameter s (both parameters are defined in the beginning of this
chapter and in Chapter 3) is kept constant.
 W(1, t) β=0
Normalised mid-point deflection
fa β = 10
β = 20
β = 30

-0.2

-0.4
0 0.2 0.4 0.6 0.8 t

Non-dimensional time

Figure 5.8: Transient response of a propped cantilever pipe for various values of Q
while dimensionless parameter ç is kept constant (ç = 0. 5)

5.3.3 Simply Supported Pipe

Figure 5.9 shows a transient response of a simply supported pipe for various values of s
while parameter þ is kept constant and Figure 5.10 shows a transient response of the
same type of boundary conditions for various values of þ while dimensionless
parameter s is kept constant.

Figure 5.9 also shows the transition from stable behaviour at s = 0 and 0.5 to unstable
behaviour, which takes place at s = 1. This is characterised by an unlimited deflection
of the pipe.
 Normalised mid-point deflection
ε=0
W(1, t) ε = 0.5
fa ε=1
0.4

-0.4

0.8

-1.2

0 0.4 0.8 1.2 1.6

Non-dimensional time

Figure 5.9: Transient response of a simply supported pipe for various values of ç
while parameter Q is kept constant (Q = 10)
 Normalised mid-point deflection
β=0
W(1, t) β=5
fa β = 10
0.4

0.2

-0.2

-0.4
0 0.4 t
0.8 1.2 1.6
Non-dimensional time

Figure 5.10: Transient response of a simply supported pipe for various values of Q
while dimensionless parameter ç is kept constant (ç = 0. 5)

5.4 Initiation of Instability (Pipe Whip)

The stability of pipes conveying a fluid was first recognised by Marcel Brillion in 1885,
[85,90], but it did not attract much attention from the research community until the
1950s, when RH Long, Jr began his pioneering investigations [91]. However, his
approach pertained to the pipes with relatively small velocities of the internal flow,
considerably below the threshold of instability. In fact, it seems he was unaware of an
instability threshold. Instead, he perceived and confirmed by experimentation that, in
contrast to simply supported pipes, forced motions of cantilevers are damped by internal
flow at relatively low flow velocities.

In 1963, Gregory and Paidoussis [91–93] became the first researchers to confirm both
theoretically and experimentally that at sufficiently high flow velocities, cantilevered
pipes are subject to oscillatory instability (flutter) behaviour. They also pointed out that
at zero flow velocity, the cantilever pipe system under consideration was lightly damped
with the pipe oscillated in its first dynamic mode. At a somewhat higher flow velocity,
the system became critically damped. At higher flow velocities, the system became less
heavily damped and the pipe oscillated in its second mode. With a further increase in
flow velocity, the pipe eventually became unstable spontaneously. Gregory and
Paidoussis [91–93] reported further that decreasing the flow velocity to the point where
instability first occurred did not avoid the oscillation in progress. These facts, along
with the reported results that show the onset of instability depended upon the amplitude
of the applied excitation displacement, vividly demonstrate the non-linear behaviour of
the cantilever system.

The conditions discussed above of the initiation of the dynamic instability for a
cantilever are also very important in the aerospace industry, when designing nozzles of
rocket engines. The nozzle of a rocket engine can be viewed as a cantilever with
variable flexibility along the length, and the high speed of the gas, which is a product of
burning reaction in the chamber of the rocket engine, passing through the nozzle can
cause the similar dynamic instability of the rocket engine [15,84].

A new impulse for more thorough investigations of possible unstable behaviour of pipes
was caused by the recent release of AS 2885.1-2007. Before this release, the Australian
pipeline standards mandated burial for all cross-country pipelines carrying gas. It was
realised that in the case of a full bore failure of an aboveground pipeline, the unstable
behaviour of the pipe can represent a potential threat to structures and people in the
vicinity of the pipeline. Therefore, when designing aboveground pipelines, the possible
instability mechanisms must be avoided at the design stage or, alternatively, special
measures have to be undertaken to avoid or minimise the consequences of pipe failures.
Such an engineering criterion will be formulated later in this chapter.

First, it will be demonstrated that the unstable behaviour of pipes transporting gas or
liquid is possible for other boundary conditions then those considered earlier in previous
studies (i.e. for cantilever pipe).
Figures 5.11 and 5.12 demonstrate the development of dynamic instability for a propped
cantilever pipe and simply supported pipe, respectively. Therefore, the unstable
behaviour is also possible in the case of other boundary conditions.

Normalised mid-point deflection

W(1, t)
fa

-0.2

-0.4

-0.6

-0.8
0 0.2 0.4 0.6 0.8 t

Non-dimensional time

Figure 5.11: Unstable behaviour (initiation of dynamic instability) of a propped

cantilever pipe at ç = 1 and Q = 5
 Normalised mid-point deflection
W(1, t)
fa
0

-0.2

-0.4

-0.6

-0.8

0 0.4 0.8 1.2 1.6 t

Non-dimensional time

Figure 5.12: Unstable behaviour of a simply supported pipe characterised by

dimensionless parameters ç = 1 and Q = 10

In the next section, an analytical approach to the instability of pipes transporting gas or
liquid will be presented and validated through numerical simulations. This analytical
approach will result in the formulation of an engineering criterion to avoid the possible
unstable behaviour of high-pressure pipelines.

5.5 Analytical Modelling Approach to Instability in

Pipes

In this section, consideration is given to conditions corresponding to the initiation of the

dynamic instability of the pipe. For the convenience of the reader, the homogeneous
governing equation of a pipe with the flowing medium is presented again below:
W,ssss + W,tt + s(2W,st + þW,ss) = 0. (3.9)

Further, assuming the solution in the form

W(x, t) = w(x)e(y+im)t , (5.1)

then the governing equation (5.1) can be rewritten as

w,ssss + (5 + im)2w + s(2(5 + im)w,s + þw,ss) = 0 . (5.2)

Mathematically, the dynamic instability takes place when the real part of exponent
index 5 + im of the assumed solution (5.1) takes positive values. In this case,
theoretically, the amplitude of oscillations can grow unbound with time. However,
in practice, the amplitude of oscillations is limited and controlled by the damping
properties of the pipeline structure, which always exist in real conditions.
Neglecting the damping, the problem is reduced to determination of the domain of
parameters s and þ variation, at which 5 > 0. Thus, at fixed s and þ = 0, we need to
determine minimum þ that satisfies equation (5.2) with an additional condition &
5⁄þ Σ 0 that corresponds to the transition from stable conditions to dynamic
instability characterised by the growing amplitude of oscillations with time.

Following [83, 94, 95] we represent the solution in the interval 0 ≤ x ≤ 1 in the
following form

w(x) = Σ Cnxn . (5.3)

n=0

From the boundary conditions at the left end (fixed end) follows that C0 = C1 = 0. The
remaining boundary conditions (free end) take the form

Σ Cnn(n — 1) = 0 (5.4a)
n=0

and
∞ Σ Cnn(n — 1)(n — 2) = 0
n=0

. (5.4b)

Now, we transform the complex form of equation (5.2) to the real-valued one.
Representing solution w as

w = w1 + iw2 (5.5)

and complex coefficients

Cn = An + iBn , (5.6)

such that

w1 = Σ A n x n (5.7a)
n=0

and

w2 = Σ Bnxn , (5.7b)
n=0

where An and Bn are real constants, which can be obtained by substituting equations
(5.5) to (5.7b) into equation (5.2) producing the following recurrent equations for the
coefficients An and Bn:
1
A= {—sþAn–2 (n — 2)(n — 3)
n
n(n — 1)(n — 2)(n —
3)

—2s5An–3(n — 3) + (m2 — 52)An–4

+2smBn–3(n — 3) + 25mBn–4} (5.8a)

and
1
B=
n n(n — 1)(n — 2)(n — {—sþBn–2 (n — 2)(n — 3)
3)

—2s5Bn–3(n — 3) + (m2 — 52)Bn–4

—2smAn–3(n — 3) — 25mAn–4} . (5.8b)

With these equations, the boundary conditions (5.4a) and (5.4b) can be rewritten as

aC2 + bC3 = 0 (5.9a)

and

cC2 + dC3 = 0 (5.9b)

where a, b, c, d are complex constants.

Specifying, for example, C1 = 1 (A1 = 1 and B2 = 0), and C3 = 0 (A3 = B3 = 0), from
equations (5.4a) and (5.4b) we have

a = Σ Ann(n — 1) + iBnn(n — 1) (5.10a)

n=0

and
∞

c = Σ Ann(n — 1)(n — 2) + iBnn(n — 1)(n — 2) . (5.10b)

n=0

Similar to the previous case, if we specify C2 = 0 and C3 = 1, i.e. we set A2 = B2 =

B3 = 0, A3 = 1, then we obtain

b = Σ Ann(n — 1) + iBnn(n — 1)
(5.11a)
n=0

and

d = Σ Ann(n — 1)(n — 2) + iBnn(n — 1)(n — 2) . (5.11b)

n=0

where An and Bn are given by recurrent equations (5.8a) and (5.8b).

The condition of existence of a nontrivial solution for the homogeneous system of linear
algebraic equations (34) is that the determinant formed from the coefficients must be
equal to zero or
a b
det [ ]=0 (5.12)
c d
from which the critical parameters of the initiation of the dynamic instability can be
obtained by means of a simple numerical procedure for determining the roots of a non-
linear algebraic equation.

The analytical approach assumes the cantilever-type boundary conditions. However, it

can be easily generalised for other boundary conditions. Below, a particular problem
relevant to pipeline industry will be considered. Namely, what would be conditions to
avoid the dynamic instability in the case of full bore rupture of an aboveground high-
pressure pipeline? As it is always difficult to model boundary conditions accurately for
real problems, two limiting cases will be considered: pinned and clamped supports at
the left end of the pipe, as shown in Figure 5.13.

x
0
L

Figure 5.13: Dynamic instability

As explained previously, the strategy is to determine minimum þ (at fixed s and þ =

0), which satisfies equation (5.2). An additional condition is & 5⁄þ Σ 0, which
corresponds to the transition from stable conditions to dynamic instability
characterised by the growing amplitude of oscillations with time. Results of the
analytical calculations for two different support conditions are shown in Figures
5.14 and 5.15, where the region of the unstable behaviour is located above the solid
line.
ƒsþ, m

12

8
ƒsþ
m

0
0 0.2 0.4 0.6 s ⁄þ

Figure 5.14: Critical parameters corresponding to the initiation of the dynamic

instability when one end of the pipe is pinned and another one is free

ƒsþ, m

24

16
m

ƒsþ
8

0
0 0.2 0.4
0.6 s⁄þ

Figure 5.15: Critical parameters corresponding to the initiation of the dynamic

instability when one end of the pipe is clamped, and another one is free
These results show that for gas pipelines (when parameter s⁄þ = hf⁄(hf + hp) is
small) the criterion to avoid the initiation of the pipe whip, regardless of the support
conditions, can be written as

hf
ƒsþ = VLJ €4 (5.13)
EI

The results of analytical calculations are also supported by numerical calculations using
the approach developed in the previous chapter (see Figure 5.16). From this figure, one
can see some differences between the analytical and numerical solutions, which are
likely caused by the error in the numerical calculation and a criterion of instability
implemented the numerical solution. The area below the curves corresponds to the case
when flowing medium generates the damping effect and in the area above the
theoretical curves, it powers the development of unstable and unbound deflections.

β
Numerical

120 Analytical

90
Unstable

60

30 þ = 15s
Stable

0
0 0.4 0.8 1.2 ε

Figure 5.16: Critical conditions for initiation of dynamic instability for cantilever
5.6 Discussion and Conclusion

A numerical method to analyse the effect of flowing medium on the elastic transient
response of a pipe transporting gas or liquid and subjected to dynamic loading was
systematically employed in this chapter. It was previously demonstrated above that the
dynamic response is governed by two non-dimensional parameters, namely s and þ,
incorporating mechanical properties of the pipe and characteristics of the flowing
medium. It was found that the first parameter (s) increases the damping properties of
the system and the increase of its value leads to the faster dissipation of the energy
supplied by the external loading (impact). The other non-dimensional parameter (þ)
acts in the opposite way. An increase of its value reduces the damping properties of the
system and can lead, eventually, to the initiation of dynamic instability.

As an example of the usefulness of the obtained results, the initiation of dynamic

instability investigated in this work has to be considered a major risk factor for high-
pressure aboveground pipelines as a full bore failure can provoke the development of
such behaviour. In this case, significant damage can be expected due to chaotic
movements of the pipe. The derived conditions of the initiation of instability can be
considered a conservative estimate, as the present mathematical model does not take
into account other important factors, for example, friction or gravity, which can
contribute significantly into the damping properties and energy absorption mechanisms.

To assess the relevance of the derived conditions of the initiation of instability to

practical problems, consider, for example, a full bore failure in a
hypothetical aboveground gas flowline of a diameter Di = 160 NN and wall thickness
t=4 NN, subjected to operating pressure P0 = 10 MPa. Let the gas density
at operating conditions be q0 = 100 kg⁄N3 and the specific heat ratio of the gas, y
= 1.3, which represent typical values for raw natural gas. The velocity of the gas
flow through the orifice can be roughly estimated using the simplified gas
dynamics model presented
[89] and this is found to be around 350 m/s. The gas flow density behind the
decompression wave is approximately 60 kg/m3. Further, we can find the ratio
 ßp +ßf
þ/s = = 15 . The line þ = 15s intersect the critical curve (see Figure 5.16) at
ßf
s = 1.25. From equation (5.13), the critical length of the pipe is found to be around 15
meters.

This example explicitly demonstrates that the pipe whip phenomena or dynamic
instability is relevant to the typical operating conditions normally occurring in high-
pressure aboveground pipelines of relatively small diameter; and to avoid this
potentially dangerous dynamic effect, the pipe has to be restrained against lateral
movements, in the case considered above, every 15 meters. The anchor spacing can be
increased by selecting pipes with large moment of inertia (lager diameter and wall
thickness) and by reduction of the operating pressure or gas density.

The equations derived here may also be useful in reducing potential damage from other
equipment working under high-pressure conditions. The criterion (equation 5.13) can be
applied, for example, when designing pipe attachments, exhausts or dispensers of
relatively small diameter or flexible connections.

It is important to mention that the derived model represents a simplest extension of the
Bernoulli-Euler beam theory, and, of course, cannot capture all mechanisms and effects
associated with pipe-flow interactions and more sophisticated models might be required
for more accurate assessments. The latter will be a subject of further investigations.
CHAPTER 6

EXPERIMENTAL STUDY
6.1 Introduction

An experimental study was conducted to gain a further understanding of how the

characteristic of the internal flow affects the transient response of circular pipes. The
scope of the experimental study was limited to the case of a long flexible pipe with
fixed-fixed support conditions or encastre pipe, as schematically shown in Figure 6.1.
This configuration was selected because of the simplicity of the rig design and an ability
to modify relatively easy the speed of the flow. The speed of the flow leads to the
change of the dimensionless controlling parameters s and þ introduced in Chapter 3. All
other parameters of the test rig including the pipe cross-section and span were kept
constant. The dynamic response of the pipe to impulse loading was measured using a
displacement sensor for a range of flow speeds. This will be discussed in detail in the
following sections. This chapter first outlines the setup and experimental approach and
then discusses the results derived from the experimental measurements.

x
0 1
F(t)

Figure 6.1: The schematic diagram of impact experiments

6.2 Experimental Setup
6.2.1 Flow Tank

Two flow tanks with the closed loop system have been designed and fabricated. The
experimental rig consists of a supply tank that delivers fluid, which is tap water, to an
open receiver tank through a 2 m long pipe section. The pipe section forms the working
area to be used in the experimental study. This pipe section was fixed from both ends.

The supply tank is made from 10 mm thick PVC plates and the receiver tank is made
from 10 mm clear acrylic plates to allow visualisation of the flow. Figure 6.2 shows a
photo of the experimental setup and Figure 6.3 is the schematic diagram of the flow
facility.

Figure 6.2: The experimental setup

 Figure 6.3: The schematic diagram of the flow facility

Flow circulation is achieved via a 2 hp 240 V Onga pump as shown in Figure 6.4. The
pump delivers fluid through the supply tank to the pipe working section. The fluid
(water) then flows into the receiver tank, out through a flow meter and back into the
supply tank (see Figure 6.3). The pump is operated by a variable frequency controller
that provides pump speeds from 0 to 25 Hertz.

Figure 6.4: A 2 hp 240 V Onga pump

The supply tank was designed to provide a low turbulence flow with uniform velocity
distribution across the cross-sectional area of the pipe. This was achieved using a
carefully designed flow distributor at the inlet, followed by perforated plates and flow
screens, discharging into a large settling chamber.

Perforated Screens

Inlet Contour

Flow Distributor

Figure 6.5: The supply tank with the top removed, showing the flow conditioning
features

The flow entered the pipe through a contoured inlet, designed according to guidelines
provided [96–101]. A 10 mm thick PVC plate is affixed to the top of the supply tank by
30 bolts. A rubber seal is used to prevent pressure leakage. Figure 6.5 is a photo of the
supply tank with the top removed, showing the flow conditioning features.

A 20 mm inside-diameter polyvinyl chloride (PVC) pipe was used (with relatively low
Young modulus E = 3.4 GPa, and the density, ρ = 1.38x106g/m3). The pipe length, L,
was 2 m with the ends clamped to simulate the fixed-fixed support conditions. Although
it should be noted that rubber-O-rings at the pipe connection to the tanks did allow some
movement. In general, it is very difficult to achieve the idealised boundary conditions in
practice. This represents a major source of discrepancies between analytical or
numerical studies and experiment. The test pipe is connected between the receiver tank
and supply tank. Figures 6.6 (a), (b) and (c) show the connection of the PVC pipe. The
material of the pipe (PVC) was selected to ensure sufficient flexibility of the pipe, as it
was demonstrated in Chapters 3 and 4 that the effect of the internal flow is negligible if
the pipe has high bending stiffness.
 Figures 6.6 (a), (b) and (c): Fixed-fixed PVC pipe section

"
The flow velocity of the fluid was measured using a Solarton-Morbrey Crawley 121
series no KS012342-2 variable-area flow meter (as shown in Figures 6.7 (a) and (b)).
This provided a value for the volumetric flow rate, which was converted to flow
velocity. The flow meter is driven by a variable frequency controller starting from 0 to
25 Hertz.
 "
Figures 6.7 (a) and (b): A Solarton-Morbrey Crawley 211 series no
KS012342-2 variable-area flow meter to measure the flow
rate.

6.2.2 Data Acquisition

To enable the investigation of the flowing medium the pipe response, it was necessary
to measure the displacement of the pipe after it had been loaded by an impact force.
This section discusses the transducers and data logging equipment used for the
experiments.

A laser sensor, Wenglor YP05MGV80, was used to measure the deflection at the mid-
point of the pipe. The reflex sensor has a working range of 43–53 mm (10 mm
measuring range) and a resolution of less than 20µN. The output of the sensor was a 0-
10 V signal, which is proportional to the measured lateral deflection of the pipe. The
sensor was placed on the pipe. This was chosen to be the mid-point where the maximum
deflections are expected. The use of this sensor is justified by a very easy procedure for
the alignment and measurements. The deflection-measuring laser setup is shown in
Figures 6.8 (a), (b) and (c).
 Figures 6.8 (a), (b) and (c): Sensor laser to measure deflection

The pipe was loaded by hitting it with a specially designed impact device. The device
consists of a load cell attached to an aluminium handle. The load cell used was a WMC
Miniature Sealed Stainless Steel load cell with a measuring capacity of ±50lbf. The load
cell utilised strain gauges in a full bridge, which provides for temperature compensation
and maximum sensitivity. Figures 6.9 (a) and (b) shows how the load cell is screwed
onto the aluminium handle. When the handle is to be used to impact the pipe, the cover
is taken off.

Figures 6.9 (a) and (b): Load cell is screwed onto the aluminium handle
Figure 6.10 shows the load cell that is fitted to the handle used to impact the pipe. The
pipe was impacted at the mid-point along its length.

Figure 6.10: Load cell that is fitted to the handle and used to impact the pipe

The load and deflection data was recorded using a Measurement Computing USB-
1208FS acquisition board. The data acquisition system, USB-1208FS, shown in Figure
6.11 was a low cost, USB-based module with four channels of 12-bit differential input.
The load cell was connected to a bridge amplifier before passing to the data acquisition
card. The laser sensor connected directly to the data acquisition card. All data was
recorded at a sampling frequency of 500 Hz for later processing.
 Figure 6.11: Data acquisition system—USB-1208FS

6.3 Experimental Procedure

6.3.1 Filling/Emptying the Tank System

Before conducting the program of experiments, the tanks have to be filled in according
to a special procedure, which was developed to meet the necessary electrical safety
precautions. To fill the tank, the drain line valves A, B and C are first closed as shown
in Figure 6.12. The supply tank vent V1 is opened to bleed out air from the receiver
tank. The main water supply tap then is opened so that the supply and receiver tanks are
filled with water. When both tanks are sufficiently full, the main water supply tap is
turned off.
 Vent V1

Valve C Valve B

Valve A

Figure 6.12: Filling two-tank facility

When water in the supply tank has displaced the air, vent V1 is closed. Vent V2 is then
opened to bleed out air from the pump. The filling time of the tanks takes approximately
20–30 minutes. The facility should be monitored during filling to ensure no water leaks
and overflows. The filling procedure is shown in Figure 6.13.
 Figure 6.13: Process of filling two-tank facility

To empty the system, the main water supply is closed. The drain line valves A, B and C
are opened. Vent V1 is opened and air is bled from drain pump P2 via bleed valve V3.
Then valve D is opened (see Figure 6.14) and the drain pump P2 is turned on (see
Figure 6.15). When the supply and receiver tanks are empty, pump P2 is switched off
and valve D is closed. It takes about 10 minutes for the tanks to be emptied of water.
The process of emptying two-tank facilities is shown in Figure 6.16.

Valve D

Figure 6.14: Valve D

 Vent V3

Figure 6.15: Pump P2 via Vent V3

Figure 6.16: Process of emptying two-tank facility

6.3.2 Operating the System Pump (P1)

The pump controller can be seen in Figure 6.17. Before operating the pump, air is bled
from the pump casing via vent V2 as shown in Figure 6.18. In operation, a pump is not
allowed to be void of water. It is important to ensure the system pump P1 is turned off
in the event of water overflowing the tanks. Pump P1 is turned off and valves A, B and
C are opened. Water supply to the facility should be turned off.
 Figure 6.17: Pump controller

Figure 6.18: Position of Vent V2

6.3.3 Experiments

The impact test was conducted when water has filled the supplier tank and water was
flowing in the pipe. The pump was operated at various frequencies so that various flow
rates are obtained. A photo of the test setup during experiments is shown in Figure 6.19.
 Figure 6.19: The test setup during use

The dynamic response of the pipe was then measured using a displacement sensor for a
range of flow conditions as shown in Table 6.1.

Table 6.1: Different range of flow conditions

No. Velocity (m/s) s þ

1 0.19 0.0653 0.0718

2 1.21 0.4137 0.4548

3 2.22 0.7620 0.8377

4 3.01 1.023 1.1250

As one can see, the selected design of the test rig provides a wide range of the
controlling parameters, which can be changed to investigate the internal flow effects.
However, both controlling parameters change simultaneously with the change of the
speed of the water flow and with the current experimental set up it was unachievable to
vary these parameters independently.
6.4 Results and Discussions

An impulse load was applied to the centre point of the pipe. Figure 6.20 shows a typical
diagram of load intensity normalised by its maximum value (f⁄fNas ) as a function of
dimensionless time, t, introduced in Chapter 3:

t
¯= t
v

where t¯ is the actual time in sec and

2
hp + hf
v=L J
EI

hf and hp be the medium and pipe densities per unit length, L is the pipe length, E is
Young modulus and I is the second moment of inertia of the pipe’s cross-section. v is
the characteristic time and can vary in a wide range, say from 1 for the laboratory tests
and up to 103 for high-pressure pipelines.

1
Normalised load

0.8

0.6

0.4

0.2

0
0 0.02 0.04 0.06 0.08 0.1
Non-dimensional time

Figure 6.20: Load intensity as a function of time

The next four figures show the transient response of the pipe, normalised deflections at
the mid-span, to the quasi-impulse loading.

Normalised mid-point deflection 0.8

0.6

0.4

0.2

0
-0.1 6E-16 0.1 0.2 0.3 0.4 0.5 0.6 0.7
-0.2

-0.4

-0.6

-0.8

Non-dimensional time

Figure 6.21a: Transient response of pipe mid-section at V = 0 m/s

Normalised mid-point deflection

0.8

0.6

0.4

0.2

0
-0.1 6E-16 0.1 0.2 0.3 0.4 0.5 0.6 0.7
-0.2

-0.4

-0.6

-0.8

Non-dimensional time
Figure 6.21b: Transient response of pipe mid-section at V=1.21 m/s
 0.8

Normalised mid-point deflection 0.6

0.4

0.2

0
-0.1 6E-16 0.1 0.2 0.3 0.4 0.5 0.6 0.7
-0.2

-0.4

-0.6

-0.8

Non-dimensional time

Figure 6.21c: Transient response of pipe at mid-section at V=2.22 m/s

Normalised mid-point deflection

0.8

0.6

0.4

0.2

0
-0.1 6E-16 0.1 0.2 0.3 0.4 0.5 0.6 0.7
-0.2

-0.4

-0.6

-0.8
Non-dimensional time
Figure 6.21d: Transient response of pipe at mid-section at V=3.0 m/s
All four sets of experiments were combined in a single diagram shown in Figure 6.22.
The first resonant frequency dominates in the dynamic response of the system. This is
caused by the internal damping, which always exists in real systems. The resonant
frequency of the pipe is almost not affected by the flow speed, at least at the first few
oscillations.

From the comparison, one can see a significant influence (up to 30% in terms of the
normalised deflections) of the parameters of the internal flow on the transient response
of the pipe. It can be clearly seen from this figure that the increase of the flow speed
leads to the increase of negative damping effect. This effect tends to increase or, at least,
to support the intensity of pipe deflections at higher speeds of the internal flow. As
discussed in the previous sections, the Coriolis force is responsible for the ‘negative’
damping effect.
Normalised mid-point deflection

2.22ms
0.8
1.21ms
0.6
3ms
0.4
0ms
0.2

0
-0.1 6E-16 0.1 0.2 0.3 0.4 0.5 0.6 0.7
-0.2

-0.4

-0.6

-0.8

Non-dimensional time

Figure 6.22: Transient response of pipe at mid-section at various speeds of the

internal flow
In the previous chapter, the numerical simulations predicted the similar effect for the
fixed-fixed boundary conditions as well as for other support conditions. Due to a very
high internal damping due to friction, gravity and non-ideal support conditions, it was
impossible to conduct a more detailed experimental investigation of the effects of the
internal flow on the transient response of pipes transporting liquid, In particular, it was
impossible to study higher harmonics of the transient response and the conditions
corresponding to the initiation of pipe instability predicted analytically and via
numerical modelling. This is something that was very difficult to predict at the earlier
stage of the rig design. Nevertheless, it was demonstrated that even with a high positive
damping the flow effects can be significant. In addition, the tendencies obtained from
the experiments (the effect of ‘negative’ damping) fully agree with the previous
analytical and numerical studies.

Further improvements and modifications of the test rig are required. This could involve
redesign of supports and use of a more powerful pump to achieve higher flow speeds.
Unfortunately, a comparison between experimental studies and analytical/numerical
results was not conducted as the scale physical system had completely different
properties to the idealised. However, this will be a subject of further work.
CHAPTER 7

SUMMARY AND CONCLUSION

The main objective of this project was to investigate effect of the flowing medium on
the transient response of pipes conveying fluid or gas and subjected to impulse (impact-
loading). To achieve this ambitious objective, analytical, numerical and experimental
studies were conducted. This chapter presents a brief summary of the work conducted
and the main results, and draws overall conclusions.

The selected topic is very important in many industrial and military applications such as
offshore structures, oil and gas, power stations, petrochemical and defence industries
where the critical pipe components transporting gas or liquid can be subjected to impact
loading due to an accident, for example. On the other end of the structural length scale,
in the last few years, micro-pipes carrying liquid or gas have been gaining popularity in
various sensor technologies, which involve some dynamic measurements. Therefore,
there is now a growing interest in studying the transient response of pipes with flowing
media caused by dynamic loading.

However, the main motivation behind the current study was the recent release of AS
2885.1-2007, which includes a new section (5.8.3) on ‘Pipeline with reduced cover or
aboveground’. It makes it clear that pipelines may be installed aboveground. The reason
for this revision was probably economical considerations because aboveground gas
pipelines can offer significant benefits while posing relatively low risks. Some gas
producers are interested in the application of aboveground gas lines so they can have
this option available for production flowlines in remote areas. Hence, the issues
surrounding safety and dynamic behaviour of aboveground high-pressure gas pipelines
are very important, specifically for Australia.

A careful literature review was conducted, which indicated a significant amount of

literature on the topic. The literature review presented in this thesis was limited to the
issues that represent an interest for the current project. It was found that the previous
studies mostly focused on vibration problems in pipes including the calculation of
natural frequencies and vibration control. Despite the fact that the vibration and impact
loading problems are closely related, it is surprising that not much research has been
conducted on the dynamic response of pipes subjected to impulse loading, especially to
understand how the internal flow changes the maximum deflections of the pipe.
Other gaps in the literature include a lack of analytical approaches and analytical
solutions that can explicitly provide an estimate of the effect of the internal flow on the
dynamic response of pipes and serve as a benchmark for numerical solutions. Driven
mainly by the power chemical industries, the previous experimental studies mostly
focused on vibration issues and vibration control of pipes conveying liquid; and not
much work has been completed to investigate the response of pipe to impact (except the
ballistic impact experiments on damage and perforation).

In Chapter 2, mathematical models describing a pipe conveying fluid were reviewed. It

was found that the most popular models for long flexible pipes are based on Bernoulli-
Euler beam theory. Many previous attempts to incorporate shear deformation using
Timoshenko model or non-linear behaviour led to a significant complication of the
governing equations, which make it virtually impossible to derive analytical solutions or
validate numerical simulations. Therefore, the present work is based on the most simple
yet reasonably accurate model utilising the classical beam theory.

In the third chapter, the governing equation for a long flexible pipe with flowing
medium based on the classical Bernoulli-Euler beam theory was presented and
analysed. After that, new scale transformations were conducted to demonstrate that only
two parameters govern the transient dynamics of the pipe. This fact can be utilised to
investigate various dynamic phenomena using reduced size or scaled physical models.
Such scaled models would be adequate if the values of the governing parameters are
kept the same for the scaled model and the reference system. Of course, the physical
models have to meet all other conditions corresponding to the adopted assumptions in
mathematical modelling. For example, the pipe has to be sufficiently long and flexible,
and the loading should not significantly affect the internal flow characteristics.

Further, an original asymptotic approach was developed to investigate the transient

response of pipes subjected to impulse loading. The approach is based on the standard
perturbation technique widely applied elsewhere. The solution represents a recurrent
system of equations, each term of which can be obtained by integration. The application
of this approach was used to obtain the first-order correction to the classical solution for
the corresponding problem of both a cantilever beam and a simply supported pipe
transporting gas or liquid. However, when higher order terms are required, the analytical
procedure may become quite cumbersome and time-consuming.

In the fourth chapter, a central finite-difference scheme was developed for the fourth
order governing equation and a validation study was completed. This demonstrated that
the both approaches are free from errors and can be confidently applied to investigate
the effect of parameters of the flowing medium on the transient response of pipes
subjected to impulse loading.

A systematic study of these effects was conducted in Chapter 5. The suggested method
of scaling of the governing equation proved very efficient, as there was significant
controversy surrounding the effect of the flow speed on the dynamic behaviour of pipes.
In some cases, it leads to negative damping, but in other cases, it leads to positive
damping. With the introduced dimensionless parameters, which combine elastic,
geometric properties of the pipe as well as properties of the internal flow, there was
observed two tendencies. One of the non-dimensionless parameters increases damping
properties of the system and the increase of its value leads to the faster dissipation of the
energy supplied by the external loading (impact). The other non-dimensional parameter
acts in an opposite way regardless of the boundary conditions and loading. An increase
of its value reduces the damping properties of the system and can lead, eventually, to the
initiation of dynamic instability.

The conditions of the initiation of dynamic instability were investigated using an

analytical approach as well as the numerical central-difference approach. As stated in
the literature review, the developed liberalised approaches are capable of determining
the critical conditions for the initiation of dynamic instability. However, they are not
capable of describing the dynamic behaviour near these conditions. As an example of
the usefulness of the obtained results, a simple engineering criterion avoiding the
dynamic instability of pipes conveying fluid or gas was suggested. The derived criterion
of the initiation of instability can be considered a conservative estimate as the present
mathematical model does not take into account other important factors, for example,
friction or gravity, which can contribute significantly into the damping properties and
energy absorption mechanisms. The critical conditions derived in this chapter may also
be useful in reducing potential damage by designing some measures from the pipe
equipment working under high-pressure conditions.

In the sixth chapter, an experimental study was conducted to gain a further

understanding of how the characteristics of the internal flow affect the transient
response of circular pipes. The scope of the experimental study was limited to the case
of a long flexible pipe with fixed-fixed support conditions. It was found the first
resonant frequency of the pipe dominated in the dynamic response of the system. This is
caused by the internal damping, which always exists in real systems. The resonant
frequency of the pipe is almost unaffected by the flow speed, at least at the first few
oscillations. From the comparison of the dynamic response of the pipe when all
parameters were fixed and only fluid speed was varied, a significant influence of the
parameters of the internal flow on the transient respond of the pipe was observed. It can
be clearly seen that the increase of the flow speed leads to the increase of negative
damping effect. As discussed in the literature review, the Coriolis force is responsible
for the negative damping effect.

Due to very high internal damping due to friction, gravity and non-ideal support
conditions, it was impossible to conduct a more detailed experimental investigation of
the effects of the internal flow on the transient respond of pipes transporting liquid. In
particular, it was impossible to study higher harmonics of the transient response and the
conditions corresponding to the initiation of pipe instability predicted analytically and
via numerical modelling. This is something that was very difficult to predict at the
earlier stage of the rig design. Nevertheless, it was demonstrated that even with a high
positive damping, the flow effects can be significant. In addition, the tendencies
obtained from the experiments (the effect of negative damping) fully agree with the
previous analytical and numerical studies summarised in the literature review. Further
improvements and modifications of the test rig are required and specified in the
conclusion section of Chapter 6. However, this will be a subject of further work.

Overall, the current work demonstrated the applicability of various approaches,

analytical, numerical and experimental, to the investigation of extremely complex
phenomena. An excellent correlation between the analytical asymptotic approach and
numerical approach was found. Unfortunately, a comparison between experimental
studies and analytical/numerical results was not conducted as the scale physical system
had completely different properties to the idealised. In particular, the test pipe had a
very high internal damping, which was difficult to model analytically. Nevertheless, the
result demonstrated a significant effect of the internal flow on the transient response of
pipes.

A number of journal and conferences papers resulted from this work and were published
in local and international journals and conference proceedings. One of the conference
contributions won a prize for best student work at ACAM 7. This also represents one of
main outcomes of the PhD project.

It is believed that the current theoretical and experimental work has made a notable
contribution to the field of research. It has allowed a better understanding of the
dynamic of pipes and the outcomes can be directly utilised to improve the safety of
pipelines or develop more advanced procedures for analysis of the dynamic behaviour
of pipes conveying gas or liquid. It is also believed that the current work can be further
extended and new useful results can be obtained as a result of further in-depth study.
The main area of the further research seems to be in the experimental field. Further
improvements and modifications of the test rig are required. This could involve redesign
of supports and use of a more powerful pump to achieve higher flow speeds and fully
validate the theoretical models and solution presented in this thesis. However, full-scale
experiments with high-pressure pipelines can be quite expensive and not feasible
without an industry support.
APPENDICES
 APPENDIX A

Mohammad, R., Kotousov, A., Codrington, J. and Blazewicz (2010)

Effect of flowing medium for a simply supported pipe subjected to impulse loading.
IN 6th Australasian Congress on Applied Mechanics, ACAM 6,
12-15 December 2010, Perth, Australia.

NOTE:
This publication is included on pages 116-125 in the
print copy of the thesis held in the University of Adelaide
Library.

Also available in:

Australian Journal of Mechanical
Engineering v. 8(2), pp. 133-42, 2011

116
APPENDIX B

Mohammad, R., Kotousov, A. and Codrington, J. (2011) Analytical modelling

of a pipe with flowing medium subjected to an impulse load.
International Journal of Impact Engineering, v. 38(2-3), pp. 115-122

NOTE: This publication is included on pages 126-133 in the print

copy of the thesis held in the University of Adelaide Library.

http://dx.doi.org/10.1016/j.ijimpeng.2010.09.00

6
APPENDIX C

fixed-free condition

clear all; clc

clc
mm=20001;
nn=51;
%V=250;
L=10;
do=0.2;
di=0.19;
I=3.142/64*(do^4-di^4);
densp=7850;
densf=1000;
pvol=3.142/4*(do^2-di^2)*L;
fvol=3.142/4*(di^2)*L;
E=200e9;
force=5;
eta = 1.22807;
%alpha = 0;

%Mass of pipe and fluid

mp=pvol*densp;
mf=fvol*densf;
twt=mp+mf;

miu=twt/L;
gamma=2*sqrt(E*I*miu)*eta/L^2;
alpha=0; %gamma*L^2/sqrt(E*I*miu);
delx = 1/(nn-1);
delt = 0.00005;
%f(44,mm)=0;

lambp=densp/L;
lambf=densf/L;

eps=0.5
%eps=V*L*mf/L/sqrt(E*I*miu);
%eps=V*L*mf/sqrt(E*I*miu);
beta=0
%beta=V*L*sqrt(miu/E/I);

for n = 1:nn+1
for m = 1:mm
f(n,m) = 0;
end
end
F = L^3*force/E/I;
f(ceil(nn/2),500)=F/delx/delt;

%eta = 1.22807;
134
A = eye(nn+3);
A(1,3) = -1;
A(nn+2,nn) = 1;
A(nn+2,nn+1) = -2;
A(nn+3,nn-1) = -0.5;
A(nn+3,nn) = 1;
A(nn+3,nn+1) = 0;
A(nn+3,nn+2) = -1;
A(nn+3,nn+3) = 0.5;
for i = 3:1:nn+1;
A(i,i-1) = -eps*delt/delx;
A(i,i) = 2+alpha*delt;
A(i,i+1) = eps*delt/delx;
end
Ainv = minv(A);

%Boundary and initial conditions

for n = 1:1:nn+3;
W(n,1) = 0;
W(n,2) = 0;
end

R(1:2,1)=0;
R(nn+2:nn+3,1)=0;
W(2,1) = 0;

for m = 2:1:mm;
W(2,m) = 0;

for n = 3:1:nn+1;
R(n,1) = -2*(delt^2/delx^4)*(W(n+2,m)-
4*W(n+1,m)+6*W(n,m)...
-4*W(n-1,m)+W(n-2,m))+2*(2*W(n,m)-W(n,m-1))...
+eps*(delt/delx)*(W(n+1,m-1)-W(n-1,m-1))...
-2*beta*eps*(delt^2/delx^2)*(W(n+1,m)-2*W(n,m)...
+W(n-1,m))+alpha*delt*(W(n,m-1))+2*delt^2*f(n,m);
end

Sol = transpose(Ainv*R);
for n = 1:1:nn+3;
W(n,m+1) = Sol(n);
end

end

%CSVWRITE('Data-Eps_10-Beta_11',W)

%figure()
m=1:1:mm;
plot(m,W(nn+1,m)/F)
n=2:1:nn+1;
%plot(n,W(n,1000)/F)
hold on;
xlabel('m','Fontweight','Bold');
ylabel('w','Fontweight','Bold');
title('Scaled deflection at n=10, Eps=0.35','Fontweight','Bold');

grid on;
hold on;

tip = max(W(nn+1,m))
mid = max(W(ceil(nn/2),m))

Fixed-fixed condition

clear all; clc

clc
mm=20001;
nn=51;
%V=250;
L=10;
do=0.2;
di=0.19;
I=3.142/64*(do^4-di^4);
densp=7850;
densf=1000;
pvol=3.142/4*(do^2-di^2)*L;
fvol=3.142/4*(di^2)*L;
E=200e9;
force=5;

eta = 1.22807;
%alpha = 0;

%Mass of pipe and fluid

mp=pvol*densp;
mf=fvol*densf;
twt=mp+mf;

miu=twt/L;
gamma=2*sqrt(E*I*miu)*eta/L^2;
alpha=0; %gamma*L^2/sqrt(E*I*miu);
delx = 1/(nn-1);
delt = 0.00005;
%f(44,mm)=0;

lambp=densp/L;
lambf=densf/L;

eps=0.5
%eps=V*L*mf/L/sqrt(E*I*miu);
%eps=V*L*mf/sqrt(E*I*miu);
beta=30
%beta=V*L*sqrt(miu/E/I);

for n = 1:nn+1
for m = 1:mm
f(n,m) = 0;
 end
end
F=L^3*force/E/I;
f(ceil(nn/2),500)=F/delx/delt;

A = eye(nn+1);
A(1,3) = -1;
A(nn+1,nn-1) = -1;

for i = 3:1:nn-1;
A(i,i-1) = -eps*delt/delx;
A(i,i) = 2+alpha*delt;
A(i,i+1) = eps*delt/delx;

end
Ainv = minv(A);

for n = 1:1:nn+1;
W(n,1) = 0;
W(n,2) = 0;
end

R(1:2,1)=0;
R(nn:nn+1,1)=0;
W(2,1) = 0;
W(nn,1)= 0

for m = 2:1:mm;
W(2,m) = 0;

%for n = 3:1:40;
for n = 3:1:nn-1;
R(n,1) = -2*(delt^2/delx^4)*(W(n+2,m)-
4*W(n+1,m)+6*W(n,m)...
-4*W(n-1,m)+W(n-2,m))+2*(2*W(n,m)-W(n,m-1))...
+eps*(delt/delx)*(W(n+1,m-1)-W(n-1,m-1))...
-2*beta*eps*(delt^2/delx^2)*(W(n+1,m)-2*W(n,m)...
+W(n-1,m))+alpha*delt*(W(n,m-1))+2*delt^2*f(n,m);
end

Sol = transpose(Ainv*R);
%for n = 1:1:42;
% W(n,m+1) = Sol(n);
%end

for n = 1:1:nn+1;
W(n,m+1) = Sol(n);
end
end
%figure()
m=1:1:mm;
plot(m,W(nn+1,m)/F)
n=2:1:nn+1;
%plot(n,W(n,1000)/F)
hold on;
xlabel('m','Fontweight','Bold');
ylabel('w','Fontweight','Bold');
title('Scaled deflection at n=10, Eps=0.35','Fontweight','Bold');

grid on;
hold on;

tip = max(W(nn+1,m))
mid = max(W(ceil(nn/2),m))

fixed ss condition

clear all; clc

clc
mm=20001;
nn=51;
%V=250;
L=10;
do=0.2;
di=0.19;
I=3.142/64*(do^4-di^4);
densp=7850;
densf=1000;
pvol=3.142/4*(do^2-di^2)*L;
fvol=3.142/4*(di^2)*L;
E=200e9;
force=5;
eta = 1.22807;
%alpha = 0;

mp=pvol*densp;
mf=fvol*densf;
twt=mp+mf;

miu=twt/L;
gamma=2*sqrt(E*I*miu)*eta/L^2;
alpha=0; %gamma*L^2/sqrt(E*I*miu);
delx = 1/(nn-1);
delt = 0.00005;
%f(44,mm)=0;

lambp=densp/L;
lambf=densf/L;

eps=1.5
%eps=V*L*mf/L/sqrt(E*I*miu);
%eps=V*L*mf/sqrt(E*I*miu);
beta=10
%beta=V*L*sqrt(miu/E/I);

%for n = 1:42
for n = 1:nn+1
for m = 1:mm
f(n,m) = 0;
end
end
%f(20,500)=L^3*force/E/I;

F=L^3*force/E/I;
f(ceil(nn/2),500)=F/delx/delt;

%eta = 1.22807;
%A = eye(42);
A = eye(nn+1);
A(1,3) = -1;
%A(42,40)=1
A(nn+1,nn-1) = 1;

%for i = 3:1:40;
for i = 3:1:nn-1;
A(i,i-1) = -eps*delt/delx;
A(i,i) = 2+alpha*delt;
A(i,i+1) = eps*delt/delx;
end
Ainv = minv(A);

%Boundary and initial conditions

%for n = 1:1:42;
for n = 1:1:nn+1;
W(n,1) = 0;
W(n,2) = 0;
end

R(1:2,1)=0;
%R(41:42,1)=0;
R(nn:nn+1,1)=0;
W(2,1) = 0;
%W(41,1) = 0
W(nn,1) = 0

for m = 2:1:mm;
W(2,m) = 0;

%for n = 3:1:40;
for n = 3:1:nn-1;
R(n,1) = -2*(delt^2/delx^4)*(W(n+2,m)-
4*W(n+1,m)+6*W(n,m)...
-4*W(n-1,m)+W(n-2,m))+2*(2*W(n,m)-W(n,m-1))...
+eps*(delt/delx)*(W(n+1,m-1)-W(n-1,m-1))...
-2*beta*eps*(delt^2/delx^2)*(W(n+1,m)-2*W(n,m)...
+W(n-1,m))+alpha*delt*(W(n,m-1))+2*delt^2*f(n,m);
 end

Sol = transpose(Ainv*R);
%for n = 1:1:42;
for n = 1:1:nn+1;
W(n,m+1) = Sol(n);
end

end

ss ss condition

clear all; clc

clc
mm=2000;
nn=51;
%V=250;
L=10;
do=0.2;
di=0.19;
I=3.142/64*(do^4-di^4);
densp=7850;
densf=1000;
pvol=3.142/4*(do^2-di^2)*L;
fvol=3.142/4*(di^2)*L;
E=200e9;
force=5;
eta = 1.22807;
%alpha = 0;

%Mass of pipe and fluid

mp=pvol*densp;
mf=fvol*densf;
twt=mp+mf;

miu=twt/L;
gamma=2*sqrt(E*I*miu)*eta/L^2;
alpha=0; %gamma*L^2/sqrt(E*I*miu);
delx = 1/(nn-1);
delt = 0.00005;
%f(44,mm)=0;

lambp=densp/L;
lambf=densf/L;

eps=0.5
%eps=V*L*mf/L/sqrt(E*I*miu);
%eps=V*L*mf/sqrt(E*I*miu);
beta=10
%beta=V*L*sqrt(miu/E/I);

for n = 1:nn+1
for m = 1:mm
f(n,m) = 0;
 end
end
F = L^3*force/E/I;
f(ceil(nn/2),500)=F/delx/delt;

%for n = 1:42
%for m = 1:mm
%f(n,m) = 0;
%end
%end
%f(20,500)=L^3*force/E/I;

%eta = 1.22807;

%A = eye(42);
A = eye(nn+1);
A(1,3) = 1;
%A(42,40) = 1;
A(nn+1,nn-1) = 1;

%for i = 3:1:40;
for i = 3:1:nn-1;
A(i,i-1) = -eps*delt/delx;
A(i,i) = 2+alpha*delt;
A(i,i+1) = eps*delt/delx;
end
Ainv = minv(A);

%Boundary and initial conditions

%for n = 1:1:42;
for n = 1:1:nn+1;
W(n,1) = 0;
W(n,2) = 0;
end

R(1:2,1)=0;
R(41:42,1)=0;
R(nn:nn+1,1)=0;
W(2,1) = 0;
W(nn,1)= 0
W(41,1)= 0

for m = 2:1:mm;
W(2,m) = 0;

%for n = 3:1:40;
for n = 3:1:nn-1;
R(n,1) = -2*(delt^2/delx^4)*(W(n+2,m)-
4*W(n+1,m)+6*W(n,m)...
-4*W(n-1,m)+W(n-2,m))+2*(2*W(n,m)-W(n,m-1))...
+eps*(delt/delx)*(W(n+1,m-1)-W(n-1,m-1))...
-2*beta*eps*(delt^2/delx^2)*(W(n+1,m)-2*W(n,m)...
+W(n-1,m))+alpha*delt*(W(n,m-1))+2*delt^2*f(n,m);
end
Sol = transpose(Ainv*R);
%for n = 1:1:42;
for n = 1:1:nn+1;
W(n,m+1) = Sol(n);
end

end

%figure()
m=1:1:mm;
plot(m,W(nn+1,m)/F)
n=2:1:nn+1;
%plot(n,W(n,1000)/F)
hold on;
xlabel('m','Fontweight','Bold');
ylabel('w','Fontweight','Bold');
title('Scaled deflection at n=10, Eps=0.35','Fontweight','Bold');

grid on;
hold on;

tip = max(W(nn+1,m))
mid = max(W(ceil(nn/2),m))
REFERENCES

[1] Morgan B. The importance of realistic representation of design features in the

risk assessment of high-pressure gas pipeline, 5 th International Conference and
Exhibition Pipeline Reliability, Houston, Texas; 1995, Sept.
[2] Montiel H, Vilchez JA, Amaldos J, Casal J. Historical analysis of accidents in
the transportation of natural gas. Journal of Hazardous Materials. 1996; 51:77–92.
[3] Wang Z, Zhang Z, Zhao F. Stability analysis of viscoelastic curved pipes
conveying fluid. Applied Mathematics and Mechanics. 2005; 6(26):807–813.
[4] Sallstrom JH. Fluid-conveying damped Rayleigh-Timoshenko beams in
transient transverse vibration studied by use of complex modal synthesis. Journal of
Fluids Structures. 1993; 7:551–563.
[5] Yigit F. Active control of flow-induced vibrations via feedback decoupling.
Journal of Vibration Control. 2008; 14(4):591–608.
[6] Fan CN, Chen WH. Vibration and stability of helical pipes conveying fluid.
ASME. 1987; 402(109):402–410.
[7] Ashley H, Haviland G. Bending vibration of a pipe line containing flowing fluid.
ASME Journal of Applied Mechanics. 1950; 72:229–232.
[8] Houser GW. Bending vibration of a pipe line containing flowing fluid. ASME
Journal of Applied Mechanics. 1952; 74:206–208.
[9] Diza M, Jain RK. Vibration analysis of pump discharge line. Journal of
Hydraulic Division, ASCE 96. 1970; (HY11):2279–2296.
[10] Chen SS, Rosenberg. Vibration and stability of tube exposed to pulsating
parallel flow. ANS TRANS. 1970; 13:335–336.
[11] Mulcahy TM, Wambsganss MW. Flow-induced vibration of nuclear reactor
components. The Shock and Vibration Digest. 1976; 8(7):318–332.
[12] Li T, DiMaggio OD. Vibration of a propellant line containing flowing fluid.
AIAA Fifth Annual Structures and Material Conference, AIAA, No. CP-B: 194–199.
[13] Brown S. Forensic engineering: reduction of risk and improving technology (for
all things great and small). Engineering Failure Analysis. 2007; 14:1019–1037.
[14] Sultan G, Hemp J. Modelling of the Coriolis mass flowmeter. Journal of Sound
and Vibration. 1989; 132(3):473–489.
[15] Feodos'ev VP. Vibrations and stability of a pipe when liquid flows through it.
Inzhenernyi Sbornik. 1951; 10:169–170.
[16] Housner GW. Bending vibrations of a pipe line containing flowing fluid. ASME
Journal of Applied Mechanics. 1952; 19:205–208.
[17] Paidoussis MP. 1992 Calvin rice lecture: Some curiosity-driven research in fluid
structure interactions and its current applications. Journal of Pressure Vessel
Technology. 1993; 115:2–14.
[18] Foda MA, Abduljabbar Z. A dynamic green function formulation for the
response of a beam structure to a moving mass. Journal of Sound and Vibration. 1998;
210(3):295–306.
[19] Huang Y, Zeng G, Wei F. A new matrix method for solving vibration and
stability of curved pipes conveying fluid. 2002; 251(2):215–225.
[20] Niordson FI. Vibrations of a cylindrical tube containing flowing fluid. Kungliga
Tekniska Hogskolans Handlingar (Stockholm); 1953. p. 73.
[21] Semercigil SE, Turan OF, Lu S. Employing fluid flow in a cantilever pipe for
vibration control. Journal of Sound and Vibration. 1997; 205(1):103–111.
[22] Guran A, Plaut RH. Stability boundaries for fluid-conveying pipes with flexible
support under axial load. Archive of Applied Mechanics. 1994; 64:417–422.
[23] Kuiper GL, Metrikine AV. On stability of a clamped-pinned pipe conveying
fluid. Heron. 2004; 49(3):211–232.
[24] Paidoussis MP. Aspirating pipes do not flutter at infinitesimally small flow.
Journal of Fluids and Structures. 1999; 13:419–425.
[25] Huang CC. Vibrations of pipes containing flowing fluids according to
Timoshenko theory. ASME J. Appl. Mech. 1974; 41(3):814–817.
[26] Bratt JF. On lateral vibration of fluid conveying pipes. Proceedings of the
ASME Energy-Sources Technology Conference, Petroleum Division Publication PD,
New Orleans, LA. 1994; 63:1–4.
[27] Stack CP, Garnett RB, Pawlas GE. Finite element for the vibration analysis of a
fluid-conveying Timoshenko beam. Proceedings of the 34th AIAA/ASME Structures,
Structural Dynamics and Materials Conference, La Jolla, CA. 1993; 4:2120–2129.
[28] Lin YH, Tsai YK. Nonlinear vibrations of Timoshenko pipes conveying fluid.
International of Journal Solids Structures. 1997; 34(23):2945–2956.
[29] Chao Y, Menglin Y, Baoren L. Coupled vibration of piping system conveying
unsteady flow. China Mech. Eng. 2008; 19(4):406–410.
[30] Pramila A, Laukkanen J. Dynamics and stability of short fluid-conveying
Timoshenko element pipes. Journal of Sound and Vibration. 1991; 144(3):421–425.
[31] Tani J, Doki M. Effects of shearing loads and in-plane boundary conditions on
the stability of thin tubes conveying fluid. ASME J. Appl. Mech. 1979; 46:779–783.
[32] Paidoussis MP, Laithier BE. Dynamics of Timoshenko beams conveying fluid.
J. Mech. Eng. Sciences. 1976; 18:210–220.
[33] Paidoussis MP, Luu TP, Laithier BE. Dynamics of finite-length tubular beams
conveying fluid. Journal of Sound Vibration. 1986; 106:311–331.
[34] Maalawi KY, Ziada MA. On the static instability of flexible pipes conveying
fluid. Journal of Fluids Structures. 2002; 16:685–690.
[35] Semler C, Li GX, Paidoussis MP. The non-linear equations of motion of pipes
conveying fluid. Journal of Sound and Vibration. 1994; 169(5):577–599.
[36] Yoshizawa M, Ueno K, Hasegawa E, Tsujiioka Y. Lateral vibration of a
cantilevered flexible pipe conveying fluid: a horizontal excitation at the upper end of the
vertical pipe. Trans. Jpn. Soc. Mech. Eng. 1988; Ser C 54(497):100–107.
[37] Miles WH, Pezeshki C, Elgar S. Bispectral analysis of a fluid elastic system: the
cantilevered pipe. Journal of Fluids Structures. 1992; 6:633–640.
[38] Ibrahim RA. Overview of mechanics of pipes conveying fluids—part I:
fundamental studies. Journal of Pressure Vessel Technology. 2010; 132:034001–32.
[39] Dzhupanov VA. Systematic review on the modes of a cantilevered pipe
conveying fluid and lying on a multiparametric resisting medium. Journal of Theory.
Appl. Mech. 1998; 28:34–53.
[40] Paidoussis MP. Flow-induced vibrations in nuclear reactors and heat
exchangers: practical experiences and state of knowledge. Practical experiences with
flow-induced vibrations. Berlin: Springer; 1980. pp. 1–81.
[41] Paidoussis MP. A review of flow-induced vibrations in reactors and reactor
components. Nuclear Engineering Design. 1983; 74:31–60.
[42] Paidoussis MP. Flow-induced instabilities of cylindrical structures. Applied
Mechanic Revision. 1987; 40:163–175.
[43] Paidoussis MP. The canonical problem of the fluid-conveying pipe and radiation
of the knowledge gained to other dynamics problems across applied mechanics. Journal
of Sound Vibration. 2008; 310:462–492.
[44] Paidoussis MP, Li GX. Pipes conveying fluid: a model dynamical problem.
Journal of Fluids Structures. 1993; 7:137–204.
[45] Tijsseling AG. Fluid-structure interaction in liquid filled pipe systems: a review.
Journal of Fluids Structures. 1996; 10:109–146.
[46] Huang Y, Zhou S, Xu J, Qian Q, Li L. Advances and trends of nonlinear
dynamics of pipes conveying fluid. Advanced Mechanical. 1998; 28(1):30–42.
[47] Weaver D, Fitzpatrick J. A Review of cross-flow induced vibrations in heat
exchanger tube arrays. Journal of Fluids Structures. 1988; 2:73–93.
[48] Wiggert DC. Fluid transients in flexible piping systems. Proceedings of the
Eighth International Association of Hydraulic Research IAHR Symposium on
Hydraulic Machinery and Cavitation, Valencia, Spain. 1996; 58–67.
[49] Wiggert DC, Tijsseling AS. Fluid transients and fluid-structure interaction in
flexible liquid-filled piping. Applied Mechanic Revision. 2001; 54(5):455–481.
[50] Xu J, Yang QB. Recent development on models and nonlinear dynamics of
pipes conveying fluid. Advanced Mechanic. 2004; 34(2):1–13.
[51] Chen SS. Flow-induced vibration of circular structures. Washington, DC:
Hemisphere; 1987.
[52] Paidoussis MP, Luu TP, Laithier BE. Dynamics of finite-length tubular beams
conveying fluid. Journal of Sound and Vibration. 1986; 106(2):311–331.
[53] Housner GW. Bending vibrations of a pipe when liquid flows through it. Journal
of Applied Mechanics. 1952; 19:205–208.
[54] Edelstein WS, Chen SS, Jendrzejczyk JA. A finite element computation of the
flow-induced oscillations in a cantilevered tube. Journal of Sound and Vibration. 1986;
107(1):121–129.
[55] Jendrzejczyk JA, Chen SS. Experiments on tubes conveying fluid. Thin-Walled
Structures. 1985; 3:109–134.
[56] Holmes PJ. Pipes supported at both ends cannot flutter. Trans ASME Journal
Applied Mechanic. 1978; 45:619–622.
[57] Singh K, Mallik AK. Wave propagation and vibration response of a periodically
supported pipe conveying fluid. Journal of Sound and Vibration. 1977; 54(1):55–66.
[58] Benjamin TB. Dynamics of a system of articulated pipes conveying fluid. I.
Theory. Proceedings of the Royal Society (London). 1961; 261(A):457–486.
[59] Paidoussis MP. Dynamics of tubular cantilevers conveying fluid. I. Mech. E.
Journal of Mechanical Engineering Science. 1970; 12:85–103.
[60] Paidoussis MP. Fluid–structure interactions: slender structures and axial flow.
London: Academic Press; 1998.
[61] Paidoussis MP. A note added to the discussion by Dupuis & Rousselet. Journal
of Fluids and Structures. 1991; 5:600.
[62] Benjamin TB. Dynamics of a system of articulated pipes conveying fluid. II.
Experiments. Proceedings of the Royal Society (London). 1961; 261(A):487–499.
[63] Gregory RW, Paidoussis MP. Unstable oscillation of tubular cantilevers
conveying fluid. II. Experiments. Proceedings of the Royal Society (London). 1966;
293(A):528–542.
[64] Paidoussis MP, Luu TP. Dynamics of a pipe aspirating fluid, such as might be
used in ocean mining. ASME Journal of Energy Resources Technology. 1985; 107:250–
255.
[65] Langthjem M. Finite element analysis and optimization of a fluid-conveying
pipe. Journal of Mechanics Based Design of Structures and Machines. 1995; 23(3):343–
376.
[66] Liang B, Tang JX. Analysis of dynamic characteristic and stability of pipe
conveying fluid by finite element method. Acta Mech. Solida Sinica. 1993; 14(2):167–
170.
[67] Shizhong W, Yulan L, Wenhu H. Research on solid-liquid coupling dynamics of
pipe conveying fluid. Applied Mathematics and Mechanics. 1998; 19(11):1065–1071.
[68] Wang SZ, Yu SS, Zho Y. Solid-liquid coupling characteristics of fluid-
conveying pipes. Journal Harbin Inst. Technology 2002; 34(2):241–244.
[69] Cui HW, Tani J. Effect of boundary conditions on the stability of a cantilever
pipe discharging and aspirating fluid. JSME Int. J. Ser. C: Dyn. Control, Robotics, Des.
Manuf. 1996; 39:20–24.
[70] Hongwu C, Junji T. Effect of boundary conditions on the stability of a cantilever
pipe discharging and aspirating fluid. JSME Int. J., Series. 1996; 39(C):20–24.
[71] Lee U, Oh H. The spectral element model for pipelines conveying internal
steady flow. Eng. Structures. 2003; 25:1045–1055.
[72] Lee U, Park J. Spectral element modeling and analysis of a pipeline conveying
internal unsteady fluid. Journal of Fluids Structures. 2006; 22:273–292.
[73] Stein RA, Tobriner WM. Vibrations of pipes containing flowing fluids. ASME
Journal Applied Mechanic. 1970; 37(6):906–916.
[74] Mizoguchi K, Komori S. Vibration and dynamic instability of a cylindrical shell
conveying a compressible fluid. Bull. JSME. 1978; 21:628–636.
[75] Wang X, Bloom F. Stability of concentric pipes conveying steady and pulsatile
fluids. Journal of Fluids Structures. 1999; 13(4):443–460.
[76] Lee SI, Chung J. New nonlinear modeling for vibration analysis of a straight
pipe conveying fluid. Journal of Sound Vibration. 2002; 254(2):313–325.
[77] Yang XD, Jin JD. Comparison of Galerkin method and complex mode method
in natural frequency analysis of tube conveying fluid. Zhendong yu Chongji/Journal
Vibration Shock. 2008; 27(3):80–86.
[78] Maalawi KY, Ziada MA. On the static instability of flexible pipes conveying
fluid. Journal of Fluids Structures. 2002; 16:685–690.
[79] Ganesan R, Ramu SA. Vibration and stability of fluid conveying pipes with
stochastic parameters. Struct. Eng. Mech. 1995; 3(4):313–324.
[80] Lifshitz JM, Gov F, Gandelsman M. Instrumented low-velocity impact of CFRP
beams. International Journal of Impact Engineering. 1995; 16(2):201–215.
[81] Daniel CB. Denting of pressurised pipelines under localised radial loading.
International Journal of Mechanical Sciences. 2004; 46:1783–1805.
[82] Kumar DR, Vidivelli B. Acrylic rubber latex in ferrocement for strengthening
reinforced concrete beams. American Journal of Engineering and Applied Sciences.
2010; 3(2):277–285.
[83] Kotousov A, Mohammad R. Analytical modeling of the transient dynamics of
pipe with flowing medium. J. of Physics: Conf. Series. 2009; 181:8.
[84] Feodosiev VI. Advanced stress and stability analysis. Berlin: Springer-Verlag;
2005.
[85] Paīdoussis MP, Issid NT. Dynamic stability of pipes conveying fluid. Journal of
Sound Vibration. 1974; 33:267–294.
[86] Kreyszig E. Advanced engineering mathematics. Hoboken: John Wiley; 2006.
[87] Barhen S. Analytic and numerical modeling of the transient dynamics of a
microcantilever sensor. Phys. Letters. 2008; 372(A): 947–957.
[88] Steidel R. An introduction to mechanical vibrations. John Wiley; 1989.
[89] Jo Y-D, Ahn BJ. Analysis of hazard area associated with hydrogen gas
transmission pipelines. Int. J. Hydrogen Energy. 2006; 31:2122–2130.
[90] Shilling R, Lou YK. An experimental study on the dynamic response of a
vertical cantilever pipe conveying fluid. Journal of Energy Resources Technology.
1980; 102:129–135.
[91] Long RH Jr. Experimental and theoretical study of transverse vibration of a tube
containing flowing fluid. ASME Journal of Applied Mechanics. 1955; 22:65–68.
[92] Gregory RW, Paidoussis MP. Unstable oscillation of tubular cantilevers
conveying fluids. I. Theory. Proceedings of the Royal Society (London). 1966b;
293(A):528–542.
[93] Gregory RW, Paidoussis MP. Unstable oscillation of tubular cantilevers
conveying fluid. II. Experiments. Proceedings of the Royal Society (London). 1966b;
293(A):528–542.
[94] Kotousov AG. Method of integral inequalities to impact dynamics problems. Int.
J. Eng. Science. 1999; 37:1737–1746.
[95] Kotousov AG. A method for obtaining approximate solutions for highly
dynamic problems. Int. J. Eng. Science. 2001; 39:477–489.
[96] Toophanpour-Rami M, Hassan ER, Kelso RM, Denier JP. Preliminary
Investigation of impulsively blocked pipe flow. 16th Australasian Fluid Mechanics
Conference Crown Plaza, Gold Coast, Australia; 2007. pp. 1–4.
[97] Smith CR, Seal CV, Praisner TJ, Sabatini DR. Hydrogen bubble visualisation in
flow visualisation. In: Lim TT, Smits A, editors. Journal of Imperial College Press;
2000. pp. 27–42.
[98] Vardi AE, Hwang KL. A characteristics model of transient friction in pipes.
Journal of Hydraulic Research. 1992; 29(5):669–684.
[99] Wang XJ, Lambert MF, Simpson AR. Detection and location of a partial
blockage in a pipeline using damping of fluid transients. Journal of Water Resources
and Planning Management. 2005; 131(3):244–492.
[100] Wallis A. Axial-flow fans and ducts. Krieger Publishing Company; 1991.
[101] Weinbaum S, Parker KH. The laminar decay of suddenly blocked channel and
pipe flows. Journal of Fluid Mechanics. 1975; 69(4):729–752.