LukePrendergast PhDThesis FinalSubmission PDF

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/327749574
Monitoring of scour around structures using changes in natural frequency of

vibration
Thesis · March 2015
CITATIONS READS
0 247
1 author:
Luke J. Prendergast
University of Nottingham
75 PUBLICATIONS 1,216 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
COST TU1406 Quality specifications for roadway bridges, standardization at a European level (BridgeSpec) View project
PhD project 2011-2015: Monitoring of scour around structures using changes in natural frequency of vibration View project
All content following this page was uploaded by Luke J. Prendergast on 26 August 2020.
The user has requested enhancement of the downloaded file.

Scoil na hInnealtóireachta Comhshaoil, Structúr agus Sibhialta
SCHOOL OF CIVIL, STRUCTURAL & ENVIRONMENTAL ENGINEERING
An Coláiste Ollscoile, Baile Átha Cliath
UNIVERSITY COLLEGE DUBLIN
Ph.D. THESIS
MONITORING OF SCOUR AROUND STRUCTURES USING CHANGES IN

NATURAL FREQUENCY OF VIBRATION
Luke J. Prendergast
Bachelor of Engineering BE (Hons) MIEI
University College Cork, Ireland
This thesis is submitted to University College Dublin for the degree of Doctor of Philosophy
in the College of Engineering and Architecture
March 2015
Head of School: Supervisor:

Dr. Mark G. Richardson Dr. Kenneth Gavin
“If you want to find the secrets of the Universe, think in terms of energy, frequency and
vibration”
- Nikola Tesla
ii
In memory of Tiernan Byrne
R.I.P.
iii
The secret listener
Standing strong, reaching over

Murky darkness, swirling, wailing,
Traversing quickly, unaware,
The bridges, creaking, changing, failing.
Depths below, muddy, secret,

Eroding, drowning, until none,
The swirling wishing soils away,
The bridges know, the day may come.
The listener rests upon the bridge,

Eavesdropping, waiting, for the sound,
A sign of change, foundations done,
The bridge, creeping into the ground.
The listener hears, it cries alarm,

To let the passers know the ruin,
The damage done, the weakened pass,
Closing the pathway prior to doom.
iv
Table of Contents
List of Tables ............................................................................................................................ x
List of Figures ......................................................................................................................... xi
Declaration ........................................................................................................................ xviii
Acknowledgements ............................................................................................................... xix
Executive Summary .............................................................................................................. xxi
Chapter 1 Introduction ....................................................................................................... 1
1.1 Background ............................................................................................................. 1
1.2 Research objectives ................................................................................................. 2
1.3 Outline of thesis ...................................................................................................... 3
Chapter 2 A review of bridge scour monitoring techniques ............................................ 7
2.1 Introduction to scour ............................................................................................... 8
2.2 Scour monitoring using depth-measuring instrumentation ................................... 12
2.2.1 Single-use devices ...................................................................................... 12
2.2.2 Pulse or radar devices ................................................................................ 13
2.2.3 Fibre-Bragg Grating (FBG) ....................................................................... 15
2.2.4 Driven or buried rod devices ..................................................................... 15
2.2.5 Sound wave devices ................................................................................... 19
2.2.6 Electrical conductivity devices .................................................................. 20
2.3 Scour monitoring using changes in structural dynamic properties ....................... 21
2.4 Conclusions ........................................................................................................... 34
Chapter 3 A comparison of initial stiffness formulations for small-strain soil-pile

dynamic Winkler modelling ................................................................................................. 35
3.1 Introduction ........................................................................................................... 36
3.1.1 Dynamic soil-structure interaction (DSSI) ................................................ 36
3.1.2 Winkler modelling approach ..................................................................... 38
3.1.3 Synopsis of chapter .................................................................................... 40
v
3.2 Background to subgrade reaction theories ............................................................ 42
3.3 Structural modelling ............................................................................................. 45
3.4 Sensitivity study .................................................................................................... 48
3.4.1 Background to analysis & soil stiffness ..................................................... 48
3.4.2 Dynamic response ...................................................................................... 51
3.5 Field experiment ................................................................................................... 56
3.5.1 Test procedure and site investigation ......................................................... 56
3.5.2 Experimental frequency response and damping ratios .............................. 59
3.6 Comparison of experimental results with numerical analysis .............................. 65
3.7 Conclusions ........................................................................................................... 69
Chapter 4 An investigation of the changes in the natural frequency of a pile

affected by scour .................................................................................................................... 71
4.1 Introduction ........................................................................................................... 72
4.2 Methodology ......................................................................................................... 72
4.2.1 Laboratory model ....................................................................................... 72
4.2.2 Field test ..................................................................................................... 77
4.3 Theoretical analysis .............................................................................................. 83
4.3.1 Structural model ......................................................................................... 83
4.3.2 Geotechnical stiffness determination ......................................................... 86
4.3.2.1 Method 1 – small-strain stiffness (G0) ................................................... 86
4.3.2.2 Method 2 – API design code .................................................................. 87
4.4 Results of laboratory test, field trial and numerical simulations .......................... 90
4.4.1 Numerical model of laboratory pile ........................................................... 90
4.4.2 Numerical model of field pile .................................................................... 93
4.5 Conclusions ........................................................................................................... 96
Chapter 5 Determining the presence of scour around bridge foundations using

vehicle-induced vibrations: A feasibility study ................................................................... 99
5.1 Introduction ......................................................................................................... 100

vi
5.2 Development of a Vehicle-Bridge-Soil Interaction (VBSI) model .................... 101
5.2.1 Bridge structure to be modelled and modelling philosophy .................... 102
5.2.2 Vehicle model .......................................................................................... 104
5.2.3 Vehicle-Bridge Interaction (VBI) ............................................................ 106
5.2.4 Dynamic Soil-Structure Interaction ......................................................... 107
5.2.4.1 Generating a theoretical qc profile........................................................ 108
5.2.4.2 Calculating G0 profiles ......................................................................... 109
5.2.4.3 Calculating spring stiffness coefficients (ksi) ....................................... 109
5.3 Analysis & results ............................................................................................... 110
5.3.1 Eigen frequencies and mode shapes ........................................................ 111
5.3.2 Response of structure to moving half-car vehicle model......................... 112
5.3.2.1 Simulation of noise free pier accelerations due to the passage of a

vehicle .............................................................................................................. 112
5.3.2.2 The effect of noise on determining the frequency of the pier vibrations ...
.............................................................................................................. 114
5.3.2.3 Identifying the presence of scour by analysing pier acceleration signals ..

.............................................................................................................. 116
5.4 Conclusions ......................................................................................................... 120
Chapter 6 An investigation into the effect of scour on the natural frequency of an

offshore wind turbine .......................................................................................................... 122
6.1 Introduction ......................................................................................................... 123
6.2 Scour around offshore wind turbine foundations................................................ 123
6.3 Scale-model scour test ........................................................................................ 127
6.4 Numerical modelling of scaled-model test ......................................................... 133
6.4.1 Structural model ....................................................................................... 133
6.4.2 Sand spring stiffness ................................................................................ 134
6.5 Results of experimental and numerical modelling of scaled-model test ............ 136
6.6 Design example – wind turbine model ............................................................... 137
vii
6.7 Conclusions ......................................................................................................... 146
Chapter 7 Conclusions .................................................................................................... 148
7.1 Summary ............................................................................................................. 148
7.2 Recommendations for further research ............................................................... 149
References ......................................................................................................................... 151
Appendix A List of journal and conference publications ............................... 160
A.1 Journal publications ............................................................................. 160
A.2 Conference publications ....................................................................... 160
Appendix B Dynamic soil-structure interaction modelling using stiffness

derived from in-situ cone penetration tests ....................................................................... 162
B.1 Introduction .......................................................................................... 163
B.2 Test site ................................................................................................ 164
B.3 Field installation & vibration test......................................................... 166
B.4 Soil-structure dynamic interaction model ............................................ 167
B.5 Conclusions .......................................................................................... 170
Appendix C Experimental data summary ....................................................... 172
C.1 Synopsis ............................................................................................... 172
C.2 Blessington data – pile scour ............................................................... 172
C.3 Blessington data – monopile scour ...................................................... 179
Appendix D An experimental investigation into the effect of water added

mass on the first natural frequency of cantilever beams ................................................. 185
D.1 Synopsis ............................................................................................... 185
D.2 Introduction .......................................................................................... 185
D.3 Experimental layout ............................................................................. 186
D.4 Analysis and results ............................................................................. 188
D.5 Discussion and conclusions ................................................................. 191
Appendix E Long-term monitoring of the frequency response of an

offshore METMAST structure ........................................................................................... 192
viii
E.1 Synopsis ............................................................................................... 192
E.2 Project description ............................................................................... 192
E.3 METMAST twisted jacket structure .................................................... 192
E.4 Instrumentation scheme ....................................................................... 194
E.5 Frequency analysis ............................................................................... 195
E.6 Summary .............................................................................................. 198
Appendix F Formulation of modelling elements ............................................ 200
F.1 Elemental beam stiffness matrix derivation ......................................... 200
F.2 Spring stiffness matrix derivation ........................................................ 206
F.3 Spring-beam element formulation ....................................................... 209
Appendix G MATLAB Code – Dynamic modelling of spring-beam systems212
G.1 Synopsis ............................................................................................... 212
G.2 MATLAB Code ................................................................................... 212
Appendix H Images of research ........................................................................ 223
ix
List of Tables
Table 1-1 Journal publications - chapters .............................................................................. 3

Table 1-2 Appendices ............................................................................................................ 5
Table 2-1 Modal analysis of bridge spans before and after retrofitting .............................. 23
Table 3-1 Comparison of coefficients of subgrade reaction models ................................... 44
Table 3-2 Range of pile geometries considered in the sensitivity study (grey cells) .......... 49
Table 3-3 Comparison of design cases for loose and dense sand profiles .......................... 53
Table 3-4 Experimental results from pile testing ................................................................ 65
Table 3-5 Comparison of experimental results and numerical predictions ......................... 68
Table 4-1 Test cantilever properties .................................................................................... 76
Table 4-2 Experimental results ............................................................................................ 77
Table 5-1 Parameters of vehicle model ............................................................................. 106
Table 5-2 Eigenvalue analysis of the scour effect ............................................................. 111
Table 6-1 Offshore wind turbine properties ...................................................................... 138
Table B-1 Site properties ................................................................................................... 164
Table B-2 Frequency values from test ............................................................................... 167
Table B-3 Experimental & numerical frequencies ............................................................ 169
Table D-1 Structural properties of the test cantilevers ...................................................... 186
Table D-2 Results of natural frequency analysis ............................................................... 190
Table E-1 Accelerometer layout ........................................................................................ 194
Table E-2 Frequency analysis............................................................................................ 198
x
List of Figures
Figure 2-1 Schematic of the scour process ............................................................................ 9

Figure 2-2 Failure of bridges due to scour. (a) Sava Bridge, Zagreb (b) Malahide Viaduct,
Dublin. Both bridges failed in 2009 .................................................................................... 10
Figure 2-3 Single-use devices for scour measurement. (a) Tethered Buried Switch (Briaud
et al. 2011), (b) Float-Out devices (NCHRP 2009) ............................................................. 13
Figure 2-4 Schematic of a typical TDR system (Yu 2009) ................................................. 14
Figure 2-5 Typical GPR profile (Anderson et al. 2007) ...................................................... 15
Figure 2-6 Schematic of Magnetic Sliding Collar (NCHRP 2009) ..................................... 16
Figure 2-7 Schematic of vibration-based scour sensor (Zarafshan et al. 2012) .................. 18
Figure 2-8 Scour monitoring instrumentation ..................................................................... 21
Figure 2-9 Reduction in stiffness caused by scour .............................................................. 22
Figure 2-10 Retrofitting Pier P2. (a) Before retrofit, (b) After retrofit (Foti & Sabia 2011)
............................................................................................................................................. 24
Figure 2-11 Dynamic response of piers before retrofitting for traffic on the upstream side
of the carriageway; the results for three different records are shown (Foti & Sabia 2011) 25
Figure 2-12 Model-scale bridge at Haynes Coastal Engineering Laboratory (Briaud et al.
2011) .................................................................................................................................... 27
Figure 2-13 Laboratory results shallow foundation. (a) variation in first, second and third
natural frequency in flow direction for shallow bridge foundation (Hz); (b) Change in the
ratio of RMS acceleration values for three different directional combinations (g/g) (Briaud
et al. 2011) ........................................................................................................................... 28
Figure 2-14 Relationship between scour level and natural frequencies (a) first, (b) third and
(c) fifth horizontally displaced mode shape [converted from inches] (Elsaid 2012) .......... 31
Figure 2-15 Kao-Ping Hsi bridge (Chen et al. 2014) .......................................................... 32
Figure 3-1 Frequency bands for typical offshore wind turbines (Hz) ................................. 37
Figure 3-2 Variation of coefficient of subgrade reaction with diameter and soil elastic
modulus. (a) Loose sand with E0 = 50 MPa; (b) Medium dense sand with E0 = 100 MPa;
(c) Dense sand with E0 = 150 MPa ...................................................................................... 45
Figure 3-3 Embedded pile numerical schematic ................................................................. 48
xi
Figure 3-4 Synthetic soil profiles. (a) Loose and dense CPT qc profiles (MPa); (b) Loose
and dense Young’s modulus (E0) profiles (MPa) ................................................................ 51
Figure 3-5 Comparison of predicted frequency responses for each subgrade reaction theory
for loose sand. (a) Frequency response by each theory for L=10m and varied pile diameter
(Hz); (b) Frequency response by each theory for D=1m and varied penetration depth (Hz)
............................................................................................................................................. 53
Figure 3-6 Dynamic response example - design case 1; (a) Acceleration responses
predicted by each model (g); (b) Frequency spectrum of acceleration signals in (a) (Hz) . 55
Figure 3-7 First mode shapes predicted by implementation of different subgrade reaction
theories ................................................................................................................................ 56
Figure 3-8 Experimental layout.(a) Experimental schematic for Pile 1 & 2 (b) Numerical
schematic for Pile 1 (c) Numerical schematic for Pile 2 (all dimensions are in mm) ......... 57
Figure 3-9 Layout of experiment – Pile 2............................................................................ 57
Figure 3-10 MASW test results from Blessington test site. (a) Shear wave velocity
measurements (m s-1), (b) Young’s modulus (E0) profile (MPa) ........................................ 59
Figure 3-11 Example of analysis for Pile 1. (a) Ambient acceleration signals (200 second
segment shown) (g), (b) Frequency Domain Decomposition (FDD) of ambient signals
shown in (a) (Hz), (c) Forced acceleration response from top accelerometer (g), (d) FRFs
of forced accelerations responses (Hz) ................................................................................ 61
Figure 3-12 Measurement of damping ratio. (a) Unfiltered and filtered accelerations (g);
(b) Fitting exponential decay function................................................................................. 64
Figure 3-13 Stiffness profiles from each subgrade theory for the given pile dimensions for
Pile 1(N m-1) ........................................................................................................................ 66
Figure 3-14 Comparison of experimental and numerical results for Pile 1 impact test 1. (a)
Time-domain acceleration signals from the experiment and each of the five numerical
models (g), (b) Frequency response functions of experiment and each of the five numerical
models (Hz) ......................................................................................................................... 67
Figure 4-1 Laboratory model geometry............................................................................... 73
Figure 4-2 Frequency change with scour. (a) Acceleration response at scour levels A and E
(m s-2), (b) Frequency content of signals shown in (a) (Hz), (c) Change in frequency with
scour (Hz) ............................................................................................................................ 75
Figure 4-3 Section properties used in experimental vibration test. (a) stiff section, (b) stiff-
flexible section, (c) flexible section ..................................................................................... 76
xii
Figure 4-4 Blessington site properties. (a) Cone tip resistance qc (MPa), (b) Shear wave
profile (m s-1), (c) G0 profiles (MPa) ................................................................................... 78
Figure 4-5 Installed pile schematic (all dimensions are in mm) ......................................... 79
Figure 4-6 Isolating frequency zone of interest. (a) Acceleration signal at scour level -4 (m
s-2), (b) Frequency content of signal shown in (a) (Hz), (c) First natural frequency of pile
(Hz) ...................................................................................................................................... 80
Figure 4-7 Change in frequency content between two scour depths. (a) Scour level -6
acceleration response (m s-2), (b) Frequency spectra for scour levels -4 and -6 (Hz) ......... 82
Figure 4-8 Frequency change with scour depth (Hz) .......................................................... 82
Figure 4-9 Scour test in progress at advanced stage ............................................................ 83
Figure 4-10 Numerical model to simulate laboratory test. (a) Geometry at start of scour
test, (b) Numerical model corresponding to dimensions shown in (a) (all dimensions are in
mm)...................................................................................................................................... 85
Figure 4-11 Numerical model to simulate field test. (a) Geometry at start of scour test,
(b) Numerical model corresponding to dimensions shown in (a) (all dimensions are in mm)
............................................................................................................................................. 85
Figure 4-12 Winkler approach and definition of p-y curves (Augustesen et al. 2009) ....... 88
Figure 4-13 Spring stiffness profiles for Blessington site (N m-1) ...................................... 89
Figure 4-14 Theoretical and experimental results for laboratory test. (a) Experimental and
numerical acceleration signals for scour level A (m s-2), (b) Experimental and numerical
acceleration signals for scour level E (m s-2), (c) Frequency content of signals (Hz), (d)
Frequency change with scour depth (Hz) ............................................................................ 91
Figure 4-15 Stiffness profiles for the laboratory pile used in the numerical model ............ 92
theoretical acceleration at scour level A (m s-2), (b) Experimental and theoretical
acceleration at scour level E (m s-2), (c) Frequency change with scour depth (Hz) ............ 93
Figure 4-17 Theoretical and experimental results for field test. (a) Impulse force applied to
top of pile (N), (b) Unfiltered acceleration signal at scour level -2 (m s-2), (c) Filtered
signal compared with numerical model at scour level -2 .................................................... 94
Figure 4-18 Frequency change with scour showing experimental response, numerical
response and analytical upper-bound equivalent cantilever (Hz) ........................................ 96
Figure 5-1 Schematic of model. (a) un-scoured, (b) post scour ........................................ 101
Figure 5-2 Bridge layout with all dimensions shown in mm. (a) elevation, (b) section A-A
........................................................................................................................................... 103
xiii
Figure 5-3 Schematic of the Vehicle-Bridge-Soil Interaction (VBSI) model ................... 105
Figure 5-4 Postulated soil parameters for a loose, medium-dense and dense sand. (a) CPT
qc profiles (MPa), (b) G0 profiles (MPa), (c) E0 profiles (MPa) and (d) spring constants (ks)
for pier foundation piles (N m-1) for the analysis .............................................................. 110
Figure 5-5 Fundamental mode shapes in loose sand – global sway. (a) Zero scour, (b) Full
scour .................................................................................................................................. 112
Figure 5-6 Road profile on the bridge ............................................................................... 113
Figure 5-7 Results for vehicle crossing bridge for zero scour level and loose sand profile.
(a) Axle contact forces (N), (b) Lateral acceleration response at top of pier (m s-2)......... 114
Figure 5-8 Sensitivity of frequency content to noise. (a) Signal from bridge pier with SNR
= 20, (b) Signal with SNR = 10, (c) Signal with SNR = 5, (d) Frequency content of signals
shown in Figs. 8(a)-(c) (Hz) .............................................................................................. 116
Figure 5-9 Bridge response due to passing vehicle and subsequent free vibration. (a)
Acceleration response from bridge pier for loose, medium-dense and dense sand profiles
with 40 seconds of free vibration (m s-2), (b) Acceleration response from bridge pier for
loose, medium-dense and dense sand profiles with 7.5 seconds of free vibration (m s-2), (c)
Frequency response of signals shown in (a) (Hz).............................................................. 117
Figure 5-10 Effect of 10 m of scour on the pier acceleration response for loose sand
profile. (a) Acceleration response (laterally) at top of bridge pier for zero and 10 m scour
due to passage of vehicle, including 40 seconds of free vibration (m s-2), (b) Acceleration
response of bridge pier with 7.5 seconds of free vibration (m s-2), (c) Frequency content of
signals shown in (a) (Hz) ................................................................................................... 119
Figure 5-11 Frequency change with scour for all three soil stiffness profiles .................. 120
Figure 6-1 Excitation frequency ranges for typical offshore wind turbines ...................... 127
Figure 6-2 Blessington site investigation. (a) Cone tip resistance profile (MPa), (b) Shear
wave velocity profile (m s-1).............................................................................................. 128
Figure 6-3 Test layout schematic (all dimensions in metres) ............................................ 129
Figure 6-4 Experimental results for scour depth -6. (a) Input force time history (N), (b)
Output accelerations for scour depth -6 (g), (c) Frequency spectrum from FDD (Hz), (d)
Isolated first natural frequency for scour depth -6 and scour depth -7 (Hz) ..................... 130
Figure 6-5 Frequency change with dimensionless scour depth ......................................... 131
Figure 6-6 Schematic of numerical model of soil-structure system .................................. 134
Figure 6-7 Profile of shear modulus (G0) developed using vs measurements and CPT qc
data (MPa) ......................................................................................................................... 136
xiv
Figure 6-8 Results of numerical and experimental investigation - frequency change with
scour (Hz) .......................................................................................................................... 137
Figure 6-9 Wind turbine model and numerical schematic ................................................ 140
Figure 6-10 Synthetic site profiles. (a) Generated qc profiles (MPa), (b) Generated G0
profiles for loose, medium dense and very dense sand deposits (MPa) ............................ 141
Figure 6-11 Effect of scour on the system frequency of an offshore wind turbine. (a) Effect
of scour with loose, medium dense and very dense sand profiles (Hz), (b) Relative effect of
scour using each stiffness profile....................................................................................... 143
Figure 6-12 Change in mode shape in loose sand stiffness profile over design scour ratio of
1.3 for full structure ........................................................................................................... 146
Figure B-1 Ten cone penetration tests for Blessington site with min & max envelopes .. 165
Figure B-2 Installed pile .................................................................................................... 166
Figure B-3 Numerical model schematic (all dimensions are in mm) ................................ 169
Figure B-4 Numerical stiffness profiles ............................................................................ 170
Figure C-1 Scour levels for pile (all dimensions are in mm) ............................................ 173
Figure C-2 Overall change in natural frequency with scour (Hz) ..................................... 173
Figure C-3 0 m scour depth - pile ...................................................................................... 174
Figure C-4 0.2 m scour depth - pile ................................................................................... 174
Figure C-5 0.4 m scour depth - pile .................................................................................. 174
Figure C-10 2.1 m scour depth - pile ................................................................................. 176
Figure C-19 Scour levels for monopile (all dimensions are in mm) ................................. 180
Figure C-20 0 m scour depth - monopile........................................................................... 180
xv
Figure C-21 0.2 m scour - monopile ................................................................................. 181
Figure C-25 1 m scour - monopile .................................................................................... 182
Figure D-1 Experimental layout. (a) Cross-section of stiff beam section; (b) Cross-section
of stiff-flexible beam section; (c) Cross-section of flexible beam section ........................ 187
Figure D-2 Experimental arrangement. (a) Flexible section under wet test; (b) Stiff-flexible
Section under wet test ........................................................................................................ 187
Figure D-3 Comparison between dry and wet vibration tests for flexible section. (a) Dry
test acceleration signal (g), (b) Wet test acceleration signal (g), (c) Frequency content of
dry and wet accelerations signals shown in (a) and (b) (Hz)............................................. 189
Figure D-4 Percentage frequency change (%) vs. flexural rigidity of each section (N m2)
........................................................................................................................................... 191
Figure E-1 METMAST structure being transported to sea ............................................... 193
Figure E-2 Accelerometer layout on the twisted jacket structure ..................................... 194
Figure E-3 Acceleration signals measured on two different dates (m s-2)......................... 195
Figure E-4 Frequency response February 10th 2013.......................................................... 197
Figure E-5 Global dominant frequency response of structure (measured using FDD) (Hz)
........................................................................................................................................... 197
Figure E-6 Frequency response in North-South & East-West directions (measured using
FDD) (Hz) ......................................................................................................................... 198
Figure F-1 Derivation of beam stiffness matrix ................................................................ 200
Figure F-2 Internal shear forces and bending moments on beam element ........................ 203
Figure F-3 Derivation of spring stiffness matrix ............................................................... 206
Figure F-4 Beam element stiffness matrix summary formulation ..................................... 209
Figure F-5 Spring element stiffness matrix formulation ................................................... 210
Figure F-6 Spring-beam combined element ...................................................................... 211
Figure H-1 Laboratory investigation into scour effect on natural frequency .................... 223
Figure H-2 Fitting accelerometers for pile scour test ........................................................ 223
xvi
Figure H-3 Pile scour test at advanced stage ..................................................................... 224
Figure H-4 Close-up of fitted accelerometers ................................................................... 224
Figure H-5 Scour depth measurement ............................................................................... 225
Figure H-6 Pilot monopile scour vibration test ................................................................. 226
Figure H-7 Vibration test for subgrade reaction comparison ............................................ 226
xvii
Declaration
The author hereby declares that this thesis has not been previously used, in whole or in
part, to obtain any degree in this, or any other university or institution. Except in cases
where reference is given in the text, the contents of this thesis are entirely the author’s own
work.
The author hereby permits the loan of this thesis by the library in University College
Dublin, upon request, for academic purposes.
_______________
Luke J. Prendergast
xviii
Acknowledgements
I would like to acknowledge everyone who made this thesis possible, both personal and
academic. I am very grateful.
My supervisor, Dr Kenneth Gavin, for his encouragement, critique and support throughout
the duration of this project. This Ph.D. would not have been possible without his help and I
will always be grateful for the faith and courage given to me to keep going.
Cormac Reale, a special thanks for everything over the last eight years in Cork and Dublin.
I am forever grateful for every chat, trip, pint or rock-climb, every hike, walk, cup of
coffee or just hanging around. This journey would have been significantly more difficult
without the constant encouragement and friendly banter.
Jennifer McAree, for putting up with all of it and never complaining about the late nights
and early mornings, thank you.
To everyone who helped out with the research, in particular Dr David Hester. I would be at
a complete loss without his help and encouraging chats. Dr David Igoe and Dr Paul
Doherty, for the help with site work, modelling and everything else. I really appreciate it.
To the members of the Geotechnical Research Group at UCD, Lisa Kirwan, Sogol Fallah,
Weichao Li, Carl Brangan, Gerry Murphy, Tim Hennessy, Yeganeh Attari, John Barrett,
thanks for all the help with everything.
To the technical staff in the UCD laboratories; in particular Mr Derek Holmes, for his help
and support with the laboratory and field studies throughout the duration of the project.
To the staff at Newstead, in particular Dr Mike Long, Dr John O’Sullivan, Prof. Eugene
O’Brien, Dr Mark Richardson, Dr Arturo Gonzalez, Dr Amanda Gibney, thanks for the
help with the administrative aspects over the course of the work.
xix
To Mr Andrew Griffiths, a note of gratitude for the daily inevitable distraction, it was
always a welcome break. Remember me when the brewing production goes global.
To Austin Devin, for his help with Frequency Domain Decomposition analysis.
To the patrons of the Emigrant’s Rest in Clonmel, for teaching me that the acquisition of
knowledge never slows and that I should strive for the highest of heights.
I wish to acknowledge the financial support obtained from the Higher Education Authority
(HEA) through the programme for research at higher level institutions, Cycle 5 (PRTLI),
co-funded by the European Regional Development Fund (ERDF). I also wish to
acknowledge partial financing of the work by the European Union Framework 7 project
SMART RAIL (Project No. 285683). I also wish to acknowledge Dr Edward Casey for his
tireless work in ensuring the programme ran smoothly.
Last but certainly not least, to my parents Phil and Ray Prendergast, for their unyielding
support through every aspect of my life. It goes without saying that I would never have
come close to achieving this goal without their love and support. I hope this makes you
proud. To my brother Alan, this has been my Olympics, I hope you make yours.
xx
Executive Summary
Bridges comprise a large portion of our infrastructural asset stock. Many bridges in
operation today worldwide are either reaching or have exceeded their original design life.
The maintenance and inspection of these structures is paramount to their continued safe
operation. Bridges located over waterways can have their foundations exposed by erosion
of the foundation soil, a phenomenon known as scour. Due to changes in environmental
conditions as a result of climate change, flooding and scour are imposing higher demands
on an already aging bridge network. The inspection and monitoring of scour around bridge
foundations is becoming an area of increasing importance as scour is the number one cause
of failure in bridges located over waterways.
Traditional scour monitoring typically involves diving inspections whereby divers inspect
the condition of bridge foundations and measure the depth of scour using crude depth
measuring instrumentation. This practice can be both dangerous and misleading as diving
inspections typically cannot be carried out during flooding when the risk of scour is
highest and scour depths may reduce as flood waters subside and wash soil back into the
hole. The misleading aspect of this practice is that in-filled scour holes may have inferior
mechanical properties to the pre-existing foundation soil thus compromising structural
integrity. A range of scour-measuring instruments have been developed that aim to
remotely monitor the occurrence of scour. These instruments typically measure the
changing depth of a scour hole near a foundation element of interest (such as a bridge pier)
and can either be used for continuous monitoring or discretely as part of a maintenance
procedure. The main drawback with instruments of this nature is that they typically cannot
give any information about the distress experienced by a structure as a result of scour.
Recognising that scour causes a loss in foundation stiffness leads onto the idea that it can
be detected using changes in the modal properties of a structure. Several authors have
investigated the potential of using changes in structural dynamic properties to infer the
presence or extent of scour affecting a structure. This thesis aims to build on existing
investigations and assess the potential of detecting the presence of scour by monitoring the
natural frequency response of the structure. This type of monitoring removes the need for
xxi
complex instrumentation to be installed into the riverbed and may potentially only require
the installation of one accelerometer above the waterline, thus reducing unnecessary
maintenance and expense. The first part of the thesis is concerned with validating a
numerical model capable of estimating the natural frequency of a soil-pile system with a
view to extending this model to include a full superstructure. The performance of this
model in estimating the change in natural frequency with scour is assessed against a full-
scale field investigation of a single pile subjected to experimentally induced scour. This
validated model is then extended for two case applications in this thesis. In the first
instance, the model is extended to include a full bridge model incorporating Vehicle-
Bridge Interaction (VBI) effects to assess the potential of detecting scour around a bridge
foundation by analysing the acceleration response induced by a passing vehicle. The
bridge’s acceleration response is analysed for its frequency content, the sensitivity of
which is assessed as a possible scour indicator. The validated soil-pile dynamic interaction
model is also extended in the latter part of the thesis to include a full-scale offshore wind
turbine model in order to assess the potential changes in natural frequency that could arise
due to scour around a monopile foundation. A range of soil densities are considered for the
foundation models of both the bridge case study and the offshore wind turbine case study.
The results from this thesis are encouraging that accelerometers placed on the structure of
interest may be capable of detecting scour around the foundation due to the changes in the
natural frequency of the structure that can occur as a result of the reduction in foundation
stiffness. In the absence of further experimental verification, however, no firm conclusions
can be given on the efficacy of the approach on real structures. It is recommended to
perform a full-scale validation as part of future work.
xxii
Chapter 1 Introduction
1.1 Background
Scour is the term given to describe the excavation and removal of material from around
foundation systems of structures located in harsh hydraulic conditions. There are three
main terms used to classify scour, namely general scour, contraction scour and local scour.
General scour includes the natural processes of aggradation and degradation of streambeds
that may occur as a result of changing hydraulic parameters governing channel form such
as changes in flow rate or sediment supply (Forde et al. 1999). This predominately pertains
to the natural evolution of the waterway and is associated with the progression of scour
erosion and deposition in the absence of obstacles to the flow (Federico et al. 2003).
Contraction scour occurs due to sudden changes in channel geometry introducing a
constriction to flow area such as at the location of a hydraulic structure such as a bridge
pier or abutments. The decrease in flow area can lead to an increase in flow velocity and
an associated increase in bed shear stresses. If the shear stresses imposed by the flow on
the bed increase above the threshold shear stress of the bed material, the sediments can
mobilise enabling scour to occur (Briaud et al. 1999). Local scour is the term given to
describe erosion that occurs around hydraulic structures such as bridge piers and
abutments. When the flow meets an obstacle such as a bridge pier, downward flow is
induced at the upstream end leading to localised erosion around the structure (Hamill
1999). The amalgamation of the three forms of scour described can lead to significant
losses in soil from around foundation elements which can prove detrimental to the stability
and safety of structures in operation.
The detection and monitoring of scour has gained significant traction in recent years due to
an increase in the failure of structures due to this phenomenon. Climate change trends
indicate an increase in frequency and magnitude of flood events affecting an already aging
infrastructure network. Traditionally, scour is monitored using diving inspections where
divers explore the condition of foundation elements using crude instrumentation. This
practice is still in use worldwide but has many disadvantages, most notably that it is often
far too dangerous to undertake a diving inspection during times of flooding, when the risk
of scour is highest. Also, due to the re-filling of scour holes as flood waters subside, the
1
Chapter 1 - Introduction
actual extent of scour may be missed by a visual inspection. A range of different methods
and instruments have been developed that aim to monitor the progression of scour depths
around vulnerable foundation components of structures, in particular bridges. While many
of these methods and instruments are quite adept at monitoring scour, they tend not to give
any indication of the distress on the structure experienced due to the scour event. An
emerging area of Structural Health Monitoring (SHM) considers the response of the
structure under investigation as a means to detect the effects of scour. Different methods
have been put forward that aim to use the dynamic response of a structure of interest, such
as a bridge, to detect the existence and extent of scour around critical foundation elements.
1.2 Research objectives
Recent advances in scour monitoring have begun to look at the response of the structure
under investigation to give an indication of the existence and severity of scour affecting
the structure. These methods are based on the fact that scour causes a loss in foundation
stiffness, and any loss in stiffness to a structural system should manifest itself as a change
in structural dynamic properties. In relation to bridges, these methods typically involve
using accelerometers mounted on the structure of interest and using characteristics of the
signals obtained to assess the condition of the foundation in terms of scour degradation. In
relation to traditional SHM, the observation that changes in structural properties cause
changes in vibration frequencies was the impetus for using modal methods for damage
identification and health monitoring (Doebling & Farrar 1996). This thesis aims to build
on these methods and investigate a potentially reliable method to detect scour using
changes in natural frequency of the structure under consideration. The scope of this thesis
relates to bridges founded in riverine environments and offshore structures such as wind
turbines, since both of these types of structure can be adversely affected by the scour
phenomenon. The research presents a suite of experimental and numerical investigations
of the effect of scour on the natural frequency of bridges and offshore wind turbines
founded on piles. The efficacy of using changes in natural frequency of vibration as a
reliable scour indicator is investigated for both types of structure.
2
1.3 Outline of thesis
This thesis is presented in the format of a thesis by publications. The thesis comprises five
published and draft journal papers in Chapters 2-6. A number of appendices are also
included that provide additional supportive materials by the author in the form of
conference publications and additional information complimenting the work in the
chapters as well as modelling methods and experimental data summaries. The material
presented in each chapter is solely the work of the author under guidance. The status of
each paper (either published, accepted or in review) is given at the start of each chapter
and the contribution by the author is also presented. It should be noted that since each
chapter is in effect a stand-alone entity, there will be some repetition between chapters,
particularly when discussing modelling methods or site information. Appropriate efforts
have been made to avoid unnecessary repetition; however for ease of readability, some
repetition is necessary. Table 1-1 gives an overview of the content of each chapter in this
thesis and Table 1-2 describes the content of the appendices.
Table 1-1 Journal publications - chapters

Chapter Contribution
2 ‘A review of bridge scour monitoring This chapter provides a comprehensive overview of
techniques’ existing instrumentation used to detect and monitor
scour around bridge foundations. This chapter also
reviews previous works that use dynamic
measurements from structures to detect and monitor
scour. The purpose of this chapter is to establish the
state-of-the-art in terms of scour monitoring and to
outline the course of research presented in
subsequent chapters of this thesis.
3 ‘A comparison of initial stiffness formulations The purpose of this chapter is to establish a method
for small-strain soil-pile dynamic Winkler of modelling pile-soil dynamic interaction for the
modelling’ purpose of scour assessment and evaluation using
changes in natural frequency. Different formulations
used to numerically model soil-pile dynamic
interaction at small strains are investigated. A
numerical parametric study is presented and a field
validation is undertaken to ascertain which of a
number of models provides the best fit to
experimental data. The model developed in this
3
chapter is used in subsequent chapters of the thesis.
4 ‘An investigation of the changes in the natural This chapter builds on the model developed in
frequency of a pile affected by scour’ Chapter 3 and investigates the effect of scour on the
dynamic response of a single pile. This chapter
investigates a method to detect and monitor scour by
observing how the natural frequency of a single pile
changes with scour. Both laboratory and full-scale
field experiments were undertaken to validate the
numerical model.
5 ‘Determining the presence of scour around This chapter presents a numerical investigation into
bridge foundations using vehicle-induced the effect of scour on the natural frequency of a full-
vibrations: A feasibility study’ scale bridge structure and aims to ascertain if scour
has a noticeable effect on the bridge’s frequency
response. The potential for changes in natural
frequency to be used as a scour monitoring tool is
investigated. The pile scour numerical model
developed in Chapter 4 is adopted for modelling the
pile foundation components in this full numerical
study. The purpose of this chapter is to ascertain if
measurable changes in natural frequency occur due
to scour and moreover, is it possible to detect these
frequencies from structural accelerations arising due
to a passing vehicle. The bridge is loaded with a
realistic vehicle loading model and Vehicle-Bridge
Interaction effects are incorporated into this model to
make the simulated acceleration signals as realistic
as possible.
6 ‘An investigation into the effect of scour on This chapter investigates the effect of scour on the
the natural frequency of an offshore wind natural frequency of a typical 3.6 MW offshore wind
turbine’ turbine. These structures are very sensitive to
stiffness changes in the soil since resonance from the
rotating blade system and sea waves can potentially
occur which could be very detrimental to the design
life of these structures in operation. The numerical
model developed in Chapter 4 is validated against a
field investigation of the effect of scour on the
frequency of a scale-model monopile, which is in
4
keeping with rigid monopiles used in the offshore

wind industry. A numerical investigation of the
effect of scour on the natural frequency of a full
wind turbine structural model is then undertaken for
three postulated soil stiffness profiles. The purpose
of the chapter is to ascertain if scour could have a
detrimental effect on the design life of offshore wind
turbines founded on monopiles.
The appendices in Table 1-2 comprise supplementary information to compliment portions

of the thesis that have been shortened due to the submission by papers method. Also
included is information on the modelling philosophy adopted and a summary of some of
the data obtained throughout the course of the research.
Table 1-2 Appendices

Appendix Description
A ‘List of journal and conference publications’ This appendix contains a list of the published journal
and conference papers arising during the research
period of the doctoral studies.
B ‘Dynamic soil-structure interaction This is a conference publication that highlights a

modelling using stiffness derived from in- method developed whereby soil springs can be
situ Cone Penetration Tests’ created from a correlation to Cone Penetration Test
(CPT) data measured at a particular site. This paper
is supplementary information to the methods
presented in Chapter 4 of this thesis.
C ‘Experimental data summary’ This appendix provides some supplementary plots of

the experimental data obtained during the course of
the research. The data presented arises from the
experimental portion of the research in Chapter 4
and Chapter 6.
D ‘An experimental investigation into the This appendix expands on a portion of the research
effect of water added mass on the first from Chapter 4 investigating the effect of water
natural frequency of cantilever beams’ added mass on the natural frequency of structures
founded in hydraulic environments. The purpose of
this appendix is to highlight that the change in
frequency of a structure due to scour should be
5
significantly higher than any changes induced due to

the presence of water and that water added mass can
generally be ignored in soil-structure interaction
problems for stiff structures.
E ‘Long-term monitoring of the frequency This appendix presents the results of an analysis
response of an offshore METMAST undertaken by the candidate into monitoring the
structure’ frequency response of an offshore METMAST to
establish if any discernible changes in frequency
could be measured over the course of 15 months.
This appendix serves as a case study for this thesis
and all the analysis was undertaken by the candidate
through the development of algorithms to deal with
the large datasets.
F ‘Formulation of modelling elements’ This appendix contains the mathematical derivations

of the stiffness matrices for the modelling elements
used widely in this thesis.
G ‘MATLAB Code – Dynamic modelling of This appendix contains example MATLAB code of a
spring-beam systems’ pile model using spring-beam elements. The raw
code used to develop the model, apply an impulse
force, and calculate the resulting displacement,
acceleration and frequency responses of the pile
structure is presented.
H ‘Images of research’ Some images from the experimental component of

the research are included in this brief appendix.
6
Chapter 2 A review of bridge scour monitoring techniques
Authors:
Luke J. Prendergast
Kenneth Gavin
Paper Status:
Published in Journal of Rock Mechanics and Geotechnical Engineering 6 (2) 2014 138-
149
Note to Reader:
This review paper was researched and written by the candidate under the supervision of
Dr Kenneth Gavin. It has been amended slightly from the published version to include
more recent research.
7
Chapter 2 – A review of bridge scour monitoring techniques
2.1 Introduction to scour
Scour of foundations is the number one cause of bridge collapse in the United States
(Melville and Coleman 2000; Briaud et al. 2001 and 2005). During the last 30 years 600
bridges have failed due to scour problems (Shirole & Holt 1991; Briaud et al. 1999)
causing major operating disruption and financial losses (De Falco & Mele 2002). In the
United States the average cost for flood damage repair of highways is estimated at $50
million per year (Lagasse et al. 1995). During a single flood event in the upper Mississippi
and lower Missouri river basins which occurred in 1993, at least 22 of the 28 bridges that
failed were due to scour. The associated repair costs were more than $8,000,000
(Kamojjala et al. 1994).
Scour can be defined as the excavation and removal of material from the bed and banks of
streams as a result of the erosive action of flowing water (Hamill 1999). Scour occurs in
three main forms; namely general scour, contraction scour and local scour. General scour
occurs naturally in river channels and includes the aggradation and degradation of the river
bed that may occur as a result of changes in the hydraulic parameters governing the
channel form such as changes in the flow rate or changes in the quantity of sediment in the
channel (Forde et al. 1999). It relates to the evolution of the waterway and is associated
with the progression of scour and filling, in the absence of obstacles (Federico et al. 2003).
Contraction scour occurs as a result of the reduction in the channel’s cross-sectional area
that arises due to the construction of structures such as bridge piers and abutments. It
manifests itself as an increase in flow velocity and resulting bed shear stresses, caused by a
reduction in the channel’s cross-sectional area at the location of a bridge. The increased
shear stresses can overcome the channel bed’s threshold shear stress and mobilize the
sediments (Briaud et al. 1999). Finally, Local scour occurs around individual bridge piers
and abutments. Downward flow is induced at the upstream end of bridge piers leading to
very localized erosion in the direct vicinity of the structure (Hamill 1999), see Figure 2-1.
Horseshoe vortices develop due to the separation of the flow at the edge of the scour hole
upstream of the pier and result in pushing the down-flow inside the scour hole closer to the
pier. Horseshoe vortices are a result of initial scouring and not the primary cause of scour.
Furthermore, separation of the flow at the sides of the pier result in wake vortices
(Heidarpour et al. 2010). Local scour depends on the balance between streambed erosion
and sediment deposition. Clear-water scour is the term given to describe the situation when
8
no sediments are delivered by the river whereas live-bed scour describes the situation
where an interaction exists between sediment transport and the scour process (Brandimarte
et al. 2006). The presence of live-bed conditions leads to smaller ultimate scour depths
than in clear-water conditions.
Bridge Pier
Downflow Wake Vortices
Scour Hole
Horseshoe Vortices
Figure 2-1 Schematic of the scour process
Scour poses obvious problems to the stability of bridge structures. Current practice dictates
that the depth of scour is determined by the addition of the individual scour depths caused
by the aforementioned mechanisms (general, contraction and local scour) (Briaud et al.
2005). This is the most critical design case. The scour hole generated has the effect of
reducing the stiffness of foundation systems and can cause bridge piers to fail without
warning. Notable bridge failures in Europe due to scour include the failure of the Sava
bridge in Zagreb, Croatia and the collapse of the Malahide viaduct in Dublin, Ireland (see
Figure 2-2).
9
(a)
(b)
Figure 2-2 Failure of bridges due to scour. (a) Sava Bridge, Zagreb (b) Malahide Viaduct,
Dublin. Both bridges failed in 2009
Scour can be combatted in a number of ways. At the bridge design stage, it is possible to
allow for scour mitigation by providing both hydraulic and structural countermeasures
(NCHRP 2009). Hydraulic countermeasures involve the prevention of rapid flow
expansion or contraction caused by sudden induced changes in flow direction that would
10
occur due to blunt pier faces obstructing the flow. These sudden flow changes can lead to
the creation of the vortices responsible for the occurrence of scour. They can be prevented
by maintaining larger bridge openings at the design stage and also by streamlining pier
geometries (May et al. 2002). It is imperative to maintain clear openings by removing
debris such as fallen trees and other objects that can often become lodged in bridge
openings, obstructing the flow. However, it is noteworthy that maintaining large bridge
openings and streamlined pier faces can often be a futile exercise as natural changes in
channel deposition and erosion upstream of a bridge can often change the angle of flow
relative to the alignment of a bridge and cause these hydraulic problems to occur.
Structural countermeasures can be implemented at the design stage by ensuring spread
footings are located below maximum design scour depths, or as remediation by adding
rock-armour and rip-rap to the base of piers and abutments. This countermeasure is limited
by uncertainties in predicted design scour depth obtained using formulas such as the
Colorado State University (CSU) formula (Bolduc et al. 2008) formulated in the Hydraulic
Engineering Circular (HEC-18) design code (Arneson et al. 2012). It can also only be
implemented on new structures, since many existing structures have unknown foundation
depths. More information on the uncertainties in bridge scour depth estimation is available
in Johnson & Dock (1998).
A more effective and economically viable method of combatting scour is to monitor its
evolution over time and implement remediation works as they are required (Briaud et al.
2011). The most widespread monitoring scheme in place as part of any national bridge
asset management framework is to undertake visual inspections. These types of
inspections are commonplace in engineering and are used to detect structural anomalies
such as cracking and other damage (Sohn et al. 2004). With regard to scour, visual
inspections involve the use of divers to inspect the condition of foundation elements and to
measure the depth of scour using basic instrumentation (Avent & Alawady 2005). Two
particular disadvantages associated with this inspection method include the fact that
inspections cannot be carried out during times of flooding, when the risk of scour is
highest, and the maximum depth of scour may not be recorded as scour holes tend to fill in
as flood waters subside (Foti & Sabia 2011; Lin et al. 2010). The fact that scour holes tend
to refill can be dangerous and misleading as the true extent of the scour problem may be
missed in the inspection. A more effective alternative is to use fixed or discrete scour
depth recording instrumentation. A number of instruments have been developed that can
11
monitor the depth of scour around bridge piers and abutments. Some of these sensing
instruments are discussed in section 2.2.
2.2 Scour monitoring using depth-measuring instrumentation
Given the importance of the problem of scour, a range of instrumentation has been
developed to monitor scour-hole development. These instruments can be broadly
categorized as follows: single-use devices (section 2.2.1), pulse or radar devices (section
2.2.2), Fibre-Bragg Grating devices (section 2.2.3), driven or buried rod systems (section
2.2.4), sound-wave devices (section 2.2.5), and electrical conductivity devices (section
2.2.6). These are detailed separately in the following sub-sections.
2.2.1 Single-use devices
Single-use devices consist of float-out devices and tethered buried switches (NCHRP
2009; Briaud et al. 2011) that can detect scour at their location of installation (see Figure
2-3). These devices are installed vertically in the riverbed, near a pier or abutment of scour
interest, and work on the principle that when the depth of scour reaches the installation
depth of the device, they simply float out of the soil. When the device changes from a
vertical orientation to a horizontal one (upon floating out) an electrical switch triggers,
which can indicate to a data acquisition system that the device is no longer in the ground
and that the scour depth has reached its elevation. Tethered Buried Switches typically have
three status indicators; “in position”, “floated out” and “not operational” (Briaud et al.
2011). Although these are very simple mechanical devices, they have a number of distinct
disadvantages. They require expensive installation, have only a single use before they must
be re-installed and can only indicate that the scour depth has reached the position of the
device. As a result, they give no further information on maximum scour depth reached.
Tethered Buried Switches must also be directly hard-wired to a data acquisition system
and as such are susceptible to debris damage. Float-out devices have a finite amount of
stored power, which means they have to be replaced eventually as part of normal
maintenance procedures. However, Float-out devices and Tethered Buried Switches are
otherwise reliable pieces of equipment, due to their inherent simplicity, and can adequately
indicate when the depth of scour has reached their installation depth.
12
(a) (b)
Figure 2-3 Single-use devices for scour measurement. (a) Tethered Buried Switch (Briaud
et al. 2011), (b) Float-Out devices (NCHRP 2009)
2.2.2 Pulse or radar devices
Pulse or radar devices utilize radar signals or electromagnetic pulses to determine changes
in the material properties that occur when a signal is propagated through a changing
physical medium (Forde et al. 1999). This typically occurs at a water-sediment interface
and thus this type of device can detect a depth of scour at a particular location. Time-
Domain Reflectometry (TDR) is one method that uses changes in the dielectric
permittivity constants between materials to determine a depth of scour at a particular
location (Yu 2009). This method was originally developed by electrical engineers who
were interested in detecting discontinuities along power and communication lines
(Yankielun & Zabilansky 1999). Measuring probes are installed into the soil at a location
of scour interest and a fast-rising step impulse is sent down a tube which determines the
interface between the water and the soil, and hence the depth of scour. They operate based
on the principle that when the propagating wave reaches an area where the dielectric
permittivity changes (e.g. the water-sediment interface), a portion of the energy is reflected
back to the receiver. They can therefore be used to observe the variation of scour depth
with time (Elsaid 2012). The method requires long probes to be installed into the riverbed
at the location of interest, which can prove expensive and time consuming. The
measurement accuracy is affected by varying temperature in the channel, with relative
errors of the order of 5% being reported in some studies (Fisher et al. 2013). However,
monitoring the channel temperature in addition to the TDR waveform can mitigate this
effect. A schematic showing a typical TDR arrangement is presented in Figure 2-4.
13
Figure 2-4 Schematic of a typical TDR system (Yu 2009)
A second method that uses pulse or radar technology to detect scour is Ground-Penetrating
Radar (GPR) (Forde et al. 1999; Anderson et al. 2007). This method uses radar pulses to
determine the water-sediment interface and hence the depth of scour. The device works
using similar principles to the TDR arrangement described previously but does not require
probes to be installed into the stratum. Instead, a floating GPR transmitter is pulled along
the water surface thus obtaining a geophysical profile of the riverbed as it passes (see
Figure 2-5 for example GPR profile). It operates by sending out high frequency
electromagnetic waves which are partially reflected as they pass through different media,
thus building up a geophysical map of the subterranean lithology (Anderson et al. 2007). A
disadvantage of this method is that it requires manual operation and cannot be used during
times of heavy-flood flow when scour is often at its highest risk of occurring. Furthermore,
it can only give information on the depth of scour present at a certain location at its time of
deployment and as a result cannot be used as a continuous monitoring solution. However,
the method is easy to implement and can provide very accurate information about the sub-
surface ground conditions.
14
Figure 2-5 Typical GPR profile (Anderson et al. 2007)
2.2.3 Fibre-Bragg Grating (FBG)
Fibre-Bragg Grating sensors are a form of Piezo-electric device (Sohn et al. 2004). These
types of sensor operate based on the concept of measuring strain along embedded
cantilever rods to generate electrical signals, which can indicate the progression of scour
along the rod. It has been found that the shift of the Bragg wavelength has a linear
relationship to the applied strain in the axial direction (Lin et al. 2006). An embedded rod
that becomes partially exposed due to scour will become subject to hydrodynamic forces
from the flow of water that induce bending in the exposed rod. This bending allows the
strain sensors to detect that the rod is free. If a number of strain gauges are positioned
along a rod, the progression of scour may be monitored. These devices perform
particularly well at monitoring the change in scour depth with time at their installed
location and are relatively cheap to fabricate. The resolution, however, depends on the
spacing of the sensor array along the rod and it can also be highly sensitive to vibrations of
the support structure used, with vibrations occurring due to flowing water or traffic
excitation from the bridge. Some reviews of the efficacy of the approach report that there
is little difference between buried and exposed sensors as a result of this phenomenon
(May et al. 2002).
2.2.4 Driven or buried rod devices
Driven or buried rod systems include such systems as the Magnetic Sliding Collar (MSC),
the “Scubamouse”, the Wallingford “Tell-Tail” Device and Mercury Tip Switches. These
instruments work on the principle of a manual or automated gravity-based physical probe
that rests on the streambed and moves downward as scour develops. The gravity sensor is
15
usually positioned around a buried or driven rod system in the streambed. It must be
sufficiently large to prevent penetrating into the streambed while in a static state. A remote
sensing element is typically used to detect the change in depth of the gravity sensor. In the
case of a Magnetic Sliding Collar, the location of the collar relative to its original position
is determined by the closure of magnetic switches along the structurally rigid rod. As the
streambed erodes, the sensor moves along the rod and its magnetic nature causes these
switches to close. The data can be manually or automatically read. In the case of automatic
reading, flexible cables can be attached to the system, which convey magnetic switch
closures back to a data acquisition unit. This device provides a relatively straightforward
method to monitor scour depth progression; however there are a number of disadvantages.
In the manual readout mode, the sensor requires infrastructure in the form of metal tubing
that can be susceptible to damage from debris impacts, particularly during times of heavy
flood flow. Scour depths can only be detected in the direct vicinity of the device so a
number of devices may be required to capture the true (global) effect of scour. The device
uses a gravity sensor and as such, it remains at the lowest depth of scour after each flood
event (NCHRP 2009). This means it may have to be reset which can be costly and time-
consuming and provides no information on scour hole re-filling. A schematic of an
installed MSC is shown in Figure 2-6.
Datalogger
Bridge Deck
Bridge Pier
Flow
Direction Magnetic Sliding Collar
Scour Hole
Figure 2-6 Schematic of Magnetic Sliding Collar (NCHRP 2009)
16
The “Scubamouse” works in a very similar way to the Magnetic Sliding Collar, except in
this case the location of the collar is determined by sliding a radioactive sensing element
into the supporting steel tube, which locates the collar. The Wallingford “Tell-Tail”
Device (De Falco & Mele 2002) consists of a set of omni-directional motion sensors
mounted on “tails” and connected to a rod that can be buried in the ground at a range of
depths. This device works on the principle that the motion sensors detect bed movements
as scour reaches the depth of embedment of the sensor. It can be connected to a data
acquisition system and thus provides a relatively straightforward mechanism to detect
scour.
Mercury tip switches can be used on a driven or buried rod to detect scour. This works on
the principle that when a steel pipe is augured into the ground, switches located along the
shaft fold up against the pipe, thus closing the circuit. If the streambed material is eroded
away, the switches open thus breaking the circuit. Due to its simplicity, this device
provides a robust way to detect scour local to the sensor installation area. Decreasing the
sensor switch array spacing allows the accuracy to be improved. Disadvantages associated
with this instrument include the fact that mercury is used in the sensor, which can prove
hazardous to the environment if released (NCHRP 2009). Also, the sensor cannot indicate
that scour hole re-filling has occurred because the switches cannot close again once
opened. It can, however, provide a good estimate of the maximum scour depth developed.
Zarafshan et al. (2012) developed a driven rod system that uses changes in the natural
frequency of the rod to detect the presence of scour (see Figure 2-7 for schematic). The rod
is fitted with a strain-sensor that can detect dynamic strains along the rod in real time. The
system works on the principle that the natural frequency of a cantilevered rod is inversely
proportional to its length. As scour occurs, the rod becomes exposed and its cantilevered
length increases. The strain sensing element on the rod measures the dynamic strain, which
is converted to the frequency domain using spectral analysis tools such as the Fast Fourier
Transform (FFT). The change in frequency is correlated to a depth of scour by developing
a Winkler-based numerical model of the soil-rod system and choosing an appropriate
coefficient of subgrade reaction (Dutta & Roy 2002). The system allows for self-
calibration, where the natural frequency measured at installation is used to choose an
appropriate subgrade modulus for the system. This allows the scour depth to be correlated
directly to the observed change in frequency caused by scour. The system works
17
reasonably well at the real time monitoring of scour but is limited in that it can only detect
scour local to the sensor and it requires installation into the streambed, which can be
labour intensive and expensive. Fisher et al. (2013) describe the development of a similar
device that also uses a vibration-based method to detect scour using a partially buried pipe
installed upstream of a river bridge pier or abutment. The device consists of a series of
sensors located along the pipe. These sensors are each equipped with a flexible disk chosen
such that it will respond (by vibrating) to the dynamic pressure present as a result of the
highly turbulent flow. These are referred to as vibration-based turbulent pressure (VTP)
sensors. The device works on the principle that the vibrations of the VTP sensors can be
measured by an accelerometer in the time-domain. The energy content is determined as the
mean squared signal from each sensor and is proportional to the vibrational energy of each
VTP sensor. By monitoring the energy content of an array of these sensors along the pipe,
it can be shown that the energy content of the sensors exposed to the flow can be one to
two orders of magnitude greater than that of sensors located in the sediment thus allowing
for scour to be directly measured. This device has been shown to be unaffected by turbid
flow conditions which have been noted to affect other instrument types and it also works
very well even in cases where the flow is misaligned to the order of 90° to the sensors’
orientation (Fisher et al. 2013). The device provides a simple mechanism to detect scour
and has been shown to be reliable and robust in harsh hydraulic environments.
Water level
Vibrating rod
Fibre optic cable

Scour hole
Figure 2-7 Schematic of vibration-based scour sensor (Zarafshan et al. 2012)
18
2.2.5 Sound wave devices
A number of devices have been developed that use sound waves to monitor the
progression of scour holes. They work on the same principle as devices that use
electromagnetic waves, in that waves are reflected from materials of different densities
thus establishing the location of the water-sediment interface. One example, Sonic
fathometers (Nassif et al. 2002; Fisher et al. 2013) can be mounted on bridge piers just
below the level of the water stage. They then build up continuous profiles of the streambed
by emitting sonic pulses from a pulse generator, which travel through the medium to the
water-sediment interface. In doing so, the device can detect and monitor the depth of scour
over time. The device can, however, only be operated within certain depth tolerances and
is susceptible to interference from entrained air present in highly turbulent flow. In
addition, where the bed topography is variable, this type of sensor only measures the
shallowest depth. Therefore, the beam width at the bed with respect to the scour hole may
significantly affect the accuracy of the scour depth measurements (Fisher et al. 2013).
Reflection Seismic Profilers also utilize sound waves to monitor and detect scour
(Anderson et al. 2007). This type of device typically employs a coupled acoustic source
transducer and receiver transducer that are placed immediately below the water surface. As
the system is towed manually across the water surface, the source transducer produces
short period pulsed acoustic signals at regular time or distance intervals. The high
frequency seismic pulse propagates through the water column and into the subterranean
sediments below. This device can build up profiles of the streambed as some of the
acoustic energy is reflected back to the receiver when the water-sediment interface is
encountered. By combining the signals from multiple locations and using estimated
seismic interval velocities, the time-depth profile can be converted into a depth profile.
Some disadvantages of the system include; noise with variable streambeds leading to the
crossing-over of signals, both the source and receiver need to be submerged meaning that
data cannot be obtained continuously over sand bars and the device also requires
significant manual input, which may make it unsuitable as a viable monitoring regime in a
lot of cases. If used correctly, however, it can provide a very accurate map of the channel
sub-features.
19
Echo Sounders work in a very similar manner to Reflection Seismic Profilers and can be
used to determine scour hole depths (Anderson et al. 2007). The only major difference is
that they emit higher frequency acoustic source pulses and due to the rapid attenuation of
the high frequency pulsed acoustic energy, relatively little signal is transmitted into or
reflected from within the sub-bottom sediment. A time-depth profile is generated by
plotting traces from adjacent source and receiver locations. Using estimated seismic
interval velocities, these time plots may be converted to depth plots. The only disadvantage
of this system over Reflection Seismic Profilers is that no information about previously
filled in scour holes can be obtained since the high frequency waves cannot penetrate into
the sub-bottom strata.
2.2.6 Electrical conductivity devices
Devices of this type use the differences in the electrical conductivity of various media to
determine the location of the water-sediment interface. They work on the principle of
measuring an electrical current between two probes. If the material between the probes
changes the ability for a current to be drawn also changes. This phenomenon can be used
to indicate the presence and depth of scour. An example of a device that uses this
technology is an Electrical Conductivity Probe (Anderson et al. 2007).
A schematic showing the deployment of some of the instrumentation described in section

2.2 is shown in Figure 2-8. Most of the instrumentation described has the disadvantage of
either requiring underwater installation, which can be costly, or can only be used discretely
as part of routine inspections. These are notable disadvantages. Monitoring the dynamic
response of the structure itself to the occurrence of scour has gained significant interest in
recent times and is described in more detail in section 2.3.
20
Datalogger
Bridge Deck (Cross-Section)
Ground
Bridge Pier Penetrating
Radar
Flow Direction
TDR / Sonic Fathometer
Electrical
Conductivity Magnetic Sliding
Probes Collar
Scour Hole
Float-Out Device
Figure 2-8 Scour monitoring instrumentation
2.3 Scour monitoring using changes in structural dynamic properties
To date scour monitoring has mostly used underwater instrumentation that measures the
progression of scour depths with time. Limited research has been undertaken to consider
the effect that scour has on the response of the bridge structure itself. Some of the
instruments developed to measure the response of a bridge structure to scour include
Datalogger
Tiltmeters and Accelerometers.Bridge
Tiltmeters, also known as Inclinometers, measure the
Deck (Cross-Section)
relative rotation of a structural element and as such can be used to detect differential
settlement, which can occur as a result of the scour process. The only major disadvantage
of the device is that it does not give a direct indication of scour depth. Devices capable of
measuring structural distress directly are more likely to be successful in allowing engineers
Accelerometer
Bridge Pier
to implement the necessary repair schemes on critical structures prior to the occurrence of
failure. Flow Direction
Accelerometers allow for the measurement of the structural Scour

response
Depth particularly to a
change in boundary conditions. The soil-structure interaction process is complex during
21
Scour Hole
scour (Foti & Sabia 2011), however, removal of material from under (or around) the
foundation during scour will cause increased stress and consequently reduced stiffness in
the remaining soil (see Figure 2-9). Since the frequency of vibration of the structure
depends on the system stiffness, observing changes in vibration frequencies is a potential
method for damage identification and health monitoring. The natural frequency of the
bridge pier-foundation system can be determined from accelerometers placed on the
structure using spectral analysis tools such as Fast Fourier Transforms (FFTs) or
Frequency Domain Decomposition (FDD) (Brincker et al. 2001) among many other
methods. A number of authors have investigated the feasibility of using dynamic
measurements to detect the presence of scour. These methods usually comprise the use of
accelerometers (or other motion sensors) to detect modal properties such as natural
frequency in concert with the development of reference numerical models.
V V
Bridge Pier Bridge Pier
No Scour Scour Hole
E1 E2 E1 = Stiffness before scour
Stress E2= Stiffness after scour

E2 < E1
Strain
Figure 2-9 Reduction in stiffness caused by scour
Foti and Sabia (2011) describe a full-scale investigation that was undertaken on a bridge in
Northern Italy which had been adversely affected by scour during a flood in 2000. The
bridge, located in Turin, contained five spans each 30 m long, and was supported by four
22
large concrete piers. The original foundation system of each of the four piers comprised a
mat of 24 piles with diameters of 600 mm and lengths of 15 m. In the 1980s as part of a
scour protection regime, a series of 55 piles with diameters of 400 mm and lengths of 8 m
were placed around the perimeter of each foundation mat. Scour affected the foundations
continuously over time. When originally constructed, the riverbed was at the same level as
the top of the foundation mat. As time progressed, the top 2 m to 3 m of the piles became
exposed (with Pier P2 on the bridge developing a 6 m deep scour hole) leaving this portion
of the foundation without lateral restraint. In response to this, a continuous monitoring
regime was implemented which involved the collection of topographical measurements to
detect the rotations of the piers as well as the installation of a permanent inclinometer on
the critical pier. The worst affected pier continued to settle consistently and in March
2004, was noted to have settlements of 16 mm and 44 mm on either side of the pier. The
movement of the other piers was very limited in contrast.
A replacement scheme for the critical pier was agreed and a dynamic survey was
undertaken before and after the replacement of the pier by analyzing traffic-induced
vibrations (see Figure 2-10). The survey involved using a method that defines a
mathematical model capable of linking the theoretical response of a dynamic system to the
time-history measurements from the system. A vector extension of the Auto-Regressive
Moving Average (ARMA) technique (Juang 1994) was used to estimate the modal
response using signals acquired from a number of accelerometers distributed along the
bridge spans. For the bridge spans, a total of six modes were identified, with modes 1 and
3 showing the most significant differences before and after the retrofit. Some of this data is
summarized in Table 2-1.
Table 2-1 Modal analysis of bridge spans before and after retrofitting
Span 1 2 3 4 5
Mode 1 Before (Hz) 4.7 4.5 4.7 4.7 4.9
Mode 1 After (Hz) 4.7 4.4 4.4 4.7 4.7
In the survey performed before rehabilitation, the second span, which was supported by
Pier P2, exhibited the lowest frequency and had an anomalous mode shape when compared
to the other spans. In the retrofitted case, the natural frequencies of spans 1, 4 and 5 were
identical whilst the frequencies of spans 2 and 3 were lower. This was because the
rehabilitation scheme for Pier P2 involved placing a beam spanning between the new
23
structural supports, on which the original bridge spans would now rest. This new
construction resulted in a lower structural stiffness than the original mass pier construction.
(a)
(b)
Figure 2-10 Retrofitting Pier P2. (a) Before retrofit, (b) After retrofit (Foti & Sabia 2011)
The damaged pier itself was analyzed by observing the asymmetric dynamic response of
the foundation system using accelerometers placed along a line parallel to the direction of
flow. The asymmetry is expected due to the rotation of the foundation caused by uneven
support conditions. Modal analysis is less useful with the expected rigid body motion of
the mass pier. The covariance matrix of the measured acceleration signals was used to
detect the presence of scour, but cannot be used to quantify its extent due to the effect of
external load intensity on the magnitude of the covariance. However, it was expected that
the variation in covariance along the element would give an indication of the presence of
scour. The retrofitted pier was also tested. A simplified numerical model of the bridge was
developed to act as a sensitivity analysis in determining the parameters of interest for scour
monitoring and to assess the effects of load position and severity of scour on the dynamic
response of the bridge structure. A modal analysis conducted in the numerical model
24
before and after scouring showed only moderate differences in natural frequencies and
mode shapes and as such, indicated that experimental accuracy would not be sufficient to
detect variations in soil-structure interaction induced by scour. This was mainly because
the pier behaved as a rigid body. However, a load applied on the same side of the model as
the scour hole did have a different effect than a load applied on the opposite side (This
load models the effect of traffic moving in different directions along the bridge deck). This
observation was confirmed by the experimental measurements with Pier P2 showing a
marked asymmetry when compared to the other piers. The results of this analysis are
shown in Figure 2-11, which presents a plot of the diagonal terms of the covariance matrix
obtained for each pier prior to retrofitting. Three repetitions are shown on each plot. Pier
P2 shows both larger magnitude variance values than other piers as well as significant
asymmetry. Pier P3 also showed high variance magnitudes. This was the other central pier
and as such was more likely to have been affected by scour erosion.
variance [(m s -2 )2 ]
Pier Rec 1 Pier Rec 1

0.02 P1 0.02
Rec 2 P2 Rec 2
Rec 3 Rec 3
0.01 0.01
0 0
0 2 4 6 8 10 12 0 2 4 6 8 10 12
Receiver Position (m) Receiver Position (m)
Pier Rec 1 Pier Rec 1

0.02 0.02 P4
P3 Rec 2 Rec 2
Rec 3 Rec 3
0.01 0.01
0 0
0 2 4 6 8 10 12 0 2 4 6 8 10 12
Receiver Position (m) Receiver Position (m)
Figure 2-11 Dynamic response of piers before retrofitting for traffic on the upstream side
of the carriageway; the results for three different records are shown (Foti & Sabia 2011)
The research concluded that the analyses presented can be used to highlight the difference
in behaviour of a single pier compared to others, or to monitor the evolution of scour if
repeated over time. The purpose of the research was not to determine the actual extent of
scour, however, as the methods developed are not suited to this purpose. Detecting the
25
actual depth of scour would be a useful addition to the analysis developed by Foti & Sabia
(2011).
Briaud et al. (2011) present a laboratory and field investigation undertaken to study the
effectiveness of various instruments, including accelerometers, at monitoring scour. A
model-scale bridge was constructed in a large hydraulic flume at Haynes Coastal
Engineering Laboratory, Texas A&M University (see Figure 2-12). The model bridge
comprised a single central pier (column) and two deck slabs fabricated from concrete. The
pier column had a length of 4 m and a diameter of 0.45 m and each deck slab was 2.03 m
in length by 0.53 m in width by 0.1 m in thickness. Two foundations types were tested,
shallow and deep, as these are the most common foundations used on bridges. Modelling
the shallow foundation involved embedding the column to a depth of 0.3 m into the sand at
the base of the flume. Modelling the deep foundation involved adding eight 0.3 m long
bars to the base of the column to model the presence of piles and embedding the column a
total of 0.45 m into the sand base (0.3 m piles + 0.15 m concrete column). The research
involved the construction of the bridge in the flume with both foundation types and
instrumenting them with various scour monitoring devices. Devices tested included float
out devices, sonar sensors, water stage measurement devices, accelerometers and
tiltmeters. The study involved increasing the velocity of water in the flume in a controlled
manner to induce scour gradually. The acceleration response of the structure was obtained
by impacting various pre-determined points on the deck with a rubber hammer to simulate
excitation that would occur due to traffic loading. This acceleration was picked up by
accelerometers placed on the bridge pier. The accelerations in three directions were
analyzed; namely the flow direction, the traffic direction and the vertical direction. Two
different methods were used to analyze the effect of scour on the acceleration responses
measured. Firstly, Fast Fourier Transforms (FFTs) were used to obtain the frequency
content of the measured acceleration responses. In order to study the variation in natural
frequency with changing water velocity and scour conditions, the acceleration time
histories were divided up into time intervals before being transformed into the frequency
domain. Secondly, the ratio of the Root Mean Square (RMS) values of the acceleration
measured in two different directions (traffic/vertical, traffic/flow or flow/vertical) was
investigated as a parameter to measure scour. This is shown in Eq. (2.1) for the flow and
traffic directions, for example.
26
a x21  a x22  ...  a xn

2
ax n
 (Eq.2.1)
ay a y1  a y 2  ...  a yn
2 2 2
where a x is the RMS value of the measured acceleration in the flow direction and a y is the
RMS value of the measured acceleration in the traffic direction.
Figure 2-12 Model-scale bridge at Haynes Coastal Engineering Laboratory (Briaud et al.
2011)
The frequency response in the direction of flow showed the highest sensitivity to the
occurrence of scour in both the shallow and deep foundation laboratory trials. The results
from the shallow foundation trial are shown in Figure 2-13(a). This figure charts the
change in frequency for the first, second and third mode of vibration of the structure
against time. After 4.5 hours in the hydraulic experiment, a notable drop in frequency was
observed in all three modes corresponding to the depth of scour reaching the bottom of the
installed column. The higher vibration modes (2 and 3) exhibited relatively larger drops in
natural frequency at this point than the fundamental mode. At the same time in the
experiment, the ratio of the RMS values of acceleration in the flow/traffic direction and
27
traffic/vertical direction showed a significant change in relative magnitude, as determined

using Eq. (2.1). This is shown by the sudden increase in these RMS ratios at the 4.5 hour
mark in Figure 2-13(b).
(a)
60
1st Natural Frequency
Frequency (Hz)
2nd Natural Frequency

40
3rd Natural Frequency
20
0
0 1 2 3 4 5 6 7
Time (Hours)
(b)
20
Flow Direction / Traffic Direction
15 Traffic Direction / Vertical Direction
Ratio (g/g)
Flow Direction / Vertical Direction

10
0
0 1 2 3 4 5 6 7
Time (Hours)
Figure 2-13 Laboratory results shallow foundation. (a) variation in first, second and third
natural frequency in flow direction for shallow bridge foundation (Hz); (b) Change in the
ratio of RMS acceleration values for three different directional combinations (g/g) (Briaud
et al. 2011)
Similar results were obtained for the deep foundation trial. The study concluded that
accelerometers placed on bridge piers showed significant potential at detecting the
presence of scour at laboratory scale for both the shallow and deep foundation types. The
sensitive parameters identified were changes to the natural frequency and ratio of RMS
values of acceleration. This was particularly true for the case of vibration in the flow
direction, which showed that the methods seemed promising.
The laboratory trials were compared to reference numerical models generated using the
commercial software LS-DYNA in order to quantify critical conditions required for bridge
failure so that suitable warning criteria can be specified. The ratio of RMS acceleration
values and natural frequency are investigated in these analyses. Different scour depths are
modeled in the analyses and the effect on the structure is obtained. Eigenvalue analyses are
conducted such that the numerical and experimental natural frequencies can be compared
28
and transient dynamic analyses are conducted such that the ratio of RMS acceleration
values for the numerical and experimental trials can be compared. The results of the
analyses for the shallow experiment reported good agreement between both parameters
being obtained. Due to the sudden failure of the deep experiment during the experimental
scour trial, a full numerical comparison was not possible, thus results could only partially
be compared. The laboratory trial concluded that accelerometers along with tiltmeters and
sonar can be used to predict bridge failure. The FFT approach as well as the RMS
approach was shown to be effective to analyze the accelerometer data as they showed
significant changes when the scour depth reached the bottom of the column.
Following the success of the laboratory trial, a full-scale field investigation on two
different bridges was conducted (Briaud et al. 2011). One of the bridges tested was the
US59 Bridge, which is 111 m long with three spans. This bridge had been subject to
remediation works over its life span because of scour problems. The field-scale
deployment for scour monitoring involved testing the various instruments from the
laboratory investigation such as accelerometers, tiltmeters, float-out devices, water-stage
sensors and tethered buried switches. The scour monitoring regime was implemented and
initially showed promise. However, the accelerometers did not give satisfactory results as
in the laboratory trials. Low excitation due to the traffic loading, low signal to noise ratios,
the high energy required to transmit accelerometer data and harsh weather conditions
which led to failures of the solar re-charge system for the data storage were some of the
reasons given for the lack of success at field deployment. Due to the success in the
laboratory and the fact that many of these issues from the field application can be
overcome with better instrumentation, it was concluded overall that using accelerometers
to detect and monitor bridge scour showed potential. However, it would require much
more research and resources to conclusively achieve results.
Elsaid (2012) published a study which involved assessing the effect of scour on the
supports of a model-scale bridge. The research initially considered a finite-element
numerical model that was used to assess the effect of scour on various aspects of the
bridge’s dynamic response. The numerical model was created in SAP2000 v12.0.2 with
the bridge supports modeled as piles that extended from the base to the deck and were
fully-fixed at the base. Scour was modeled as an increase in the effective length of these
piles. Eigenvalue analyses were performed to extract the dynamic characteristics of the
29
structure. It was concluded from these analyses that vertical mode shapes of the structure
were insensitive to scour due to the limited effect of the pile’s axial stiffness on the
dynamic response of the superstructure (Elsaid & Seracino 2014). However, the
horizontally displaced mode shapes showed significant sensitivity to the progression of
scour, due to the reduction in the flexural stiffness of the bridge piles with increasing
effective length. This is in agreement with the laboratory study reported by Briaud et al.
(2011). Other damage indicators, namely; mode shape curvature, flexibility-based
deflection and curvature were also assessed in the finite-element environment to
investigate their applicability to detecting scour. Results from the numerical analyses
indicated that these methods not only showed promise at detecting scour but also showed
that it may be possible to quantify the extent of scour.
Following on from the numerical investigation, an experimental regime was undertaken to

assess the applicability of the methods on a real structural frame (Elsaid & Seracino 2014;
Elsaid 2012). The test regime involved impacting the structure in the vertical and
horizontal directions and measuring its dynamic response using accelerometers placed on
the web of steel girders. Frequency Response Functions (FRFs) were developed to obtain
the natural frequency for each case, which show the ratio of the output response to the
input stimulus. In this case, the input stimulus is the time-history of forces over the strike
duration and the output response is the acceleration readings from the accelerometers.
Three different scour scenarios were modeled; symmetrical scour, unsymmetrical scour
and braced piles (representing the case of scour that exposes the pier’s foundation or a
ground beam connecting the two piles). As in the finite-element investigation, scour was
modeled simply as increasing the effective length of piles, which are fixed at the base to
the laboratory floor. For the unsymmetrical case, the shorter pile was fixed to a concrete
block, in order to keep the steel frame level. The experimental investigation verified the
numerical prediction in that vertically displaced mode shapes were deemed insensitive to
scour with little differences being recorded in the FRFs before and after scour. For the
horizontal impact assessment, the first, third and fifth mode shapes showed the most
sensitivity to scour with a decrease in natural frequency being obtained as scour depth
increased (see Figure 2-14). The first and third mode shapes showed a decrease in
frequency and the fifth mode shape showed an initial increase in frequency. The second
and fourth mode shapes were deemed insensitive to scour based on the geometry of the
system under investigation.
30
(a) (b)
20 38
Mode 1 Mode 3
Frequency (Hz)
Frequency (Hz)
18 36
16 34
14 32
12 30
0 200 400 600 0 200 400 600
Scour (mm) Scour (mm)
(c)
65
Mode 5
Frequency (Hz)
64
63
62
61
0 200 400 600
Scour (mm)
Figure 2-14 Relationship between scour level and natural frequencies (a) first, (b) third and
(c) fifth horizontally displaced mode shape [converted from inches] (Elsaid 2012)
In addition to the horizontally displaced mode shapes showing sensitivity to the

progression of scour, the three damage indicators (mode shape curvature, flexibility-based
deflection and curvature) were also investigated to study their effectiveness at detecting
scour. Results indicated that the three damage indicators were able to identify the exact
location of scour for the symmetrical scour cases and the scour zone for the unsymmetrical
and braced scour cases. The change in flexibility-based deflection was capable of
determining the extent of scour as well as its location. Although the laboratory
investigation showed promise, the scour modelling was greatly simplified in that no
treatment of soil-structure interaction was considered and this would have a significant
effect on the dynamic response of the structural supports.
Chen et al. (2014) developed a foundation scour evaluation method using ambient
vibration measurements taken at different locations of a cable-stayed bridge. Using these
measurements, they performed a successful case study to estimate the scour depth around
one of the bridge piers. The bridge in question is the Kao-Ping-Hsi Cable-Stay Bridge,
which spans the Kao-Ping-Hsi River in southern Taiwan (see Figure 2-15). It
31
accommodates six traffic lanes, has a total length of 510 m and comprises two spans. In
the first instance, they performed a full modal analysis of the Kao-Ping-Hsi cable-stayed
bridge structure using velocity sensors distributed at various locations on the bridge, by
performing Fourier analysis on the velocity signals. From these measurements, they
determined 10 modal frequencies of interest covering flexural, axial and torsional bridge
modes for various bridge elements.
Figure 2-15 Kao-Ping Hsi bridge (Chen et al. 2014)
A Finite-Element (FE) numerical model of the bridge was created using space-frame
elements in SAP 2000. In the first version, a simplification was introduced whereby Pier
P2, the pier of scour interest, was initially neglected from the model (its presence was
represented by a boundary condition). In the first run of the model with assumed boundary
conditions, the discrepancy between measured and predicted frequency values was
generally less than ± 3% for most of the modes with some modes exceeding this (the first
vertical flexural mode had a 12% error). The boundary conditions of the model were tuned
based on the disparity between predicted modal frequencies and those measured on the
structure, since an adjustment of the geometric or material properties could not account for
the error due to the non-uniform effect on different modes. Tuning the boundary
conditions was based on a more comprehensive study of the actual structural boundary
conditions, and included the addition of linear and rotational springs where appropriate.
Once the boundary conditions had been updated to allow the predicted frequencies to
match the measured data more coherently, a new FE model incorporating Pier P2 was
developed. This new model incorporating the Pier P2 matched the response of the previous
tuned model very well since in the previous model the pier had been effectively treated as
a boundary. However, the new model contained additional flexural modes local to the pier
32
which were obviously omitted from the previous model. Using the known soil deposit
height at Pylon P1, it was possible to estimate optimal soil stiffness for the foundation.
This was achieved by observing the vibration modes most affected by scour of the pylon
(first horizontal flexural mode and second torsional mode) and tuning the soil stiffness
until the errors between the predicted and measured modes of interested were minimised.
The sensitivity to scour of the horizontal flexural mode is similar to the findings of Elsaid
& Seracino (2014), however the sensitivity of the torsional mode is novel to this research.
It was noted that the deposit height at Pylon P1 must be known a priori as a lower deposit
height with higher soil stiffness estimate or a higher deposit height with lower soil stiffness
estimate will lead to the same optimised model, therefore to mitigate potential
inaccuracies, the deposit height should be known. To evaluate the scour depth at Pier P2,
the scour depth was varied in the model to assess the sensitive frequencies. The first and
second vertical flexural modes of the Pier P2 proved most sensitive to scour of the pier
with an obvious decrease in frequency with increasing scour depth. The actual measured
frequencies of Pier P2 can be obtained from the ambient measurements taken at the top of
the pier and compared to the numerical model for varying scour depths. The average
relative error between the actual measured frequencies and the numerical model
predictions with several scour levels was obtained, with the minimum error indicating the
most likely scour depth around the real structure. The predicted scour depth of 3.5 m was
verified with good agreement using a plumb line.
The results of this paper suggest that the scour condition at Pier P2 can be further
monitored solely by a sensor placed at the top of the pier. It is noteworthy that although
this paper has provided evidence of a successful application of a vibration-based scour
monitoring framework, the bridge chosen had several favourable attributes. Firstly it has
very few piers and a single pylon which reduces significantly the unknowns. The deposit
height at the pylon P1 was known which allowed for optimised soil stiffness to be
estimated. The long-span nature of the bridge meant that more of the modes fall in the low-
frequency range and are more easily identified from ambient vibrations. These attributes
aside, the research is still a significant step towards long-term scour monitoring using
vibration measurements.
33
2.4 Conclusions
Traditional scour monitoring instrumentation often requires expensive installation and

maintenance and can also be susceptible to debris damage during flooding. Often, the
interpretation of data from these instruments can also be time consuming and difficult.
There is much on-going research into the use of the structural dynamic response to detect
and measure the depth of scour around structures. The research is paving the way for low-
maintenance non-intrusive structural health monitoring to detect and monitor scour
development. The advantages of dynamic measurements over traditional scour depth
measuring instrumentation are the ease of installation above the waterline and the low
maintenance required. Frequency shifts offer potential to detect the loss of stiffness
associated with scour and as a result, to detect the effect that this loss of stiffness has on
the structure of interest. This aspect is often missed by instruments installed in the stream
bed, as the global effect of scour may not be observed unless a high density of instruments
are used around scour critical areas. There is still significant room to improve dynamic
measurement systems. Some of the difficulties associated with these methods include
issues such as the high volume of acceleration data required to obtain sensible information
(Briaud et al. 2011), the high power requirements for data acquisition systems as well as
other effects. The issue of environmental effects on the measured natural frequency is also
a factor worth considering (Sohn 2007). In time, improvements in technology will mitigate
some of these issues and more reliable dynamic based scour monitoring instrumentation
may be developed.
From the literature survey undertaken in this chapter, several key questions arise. Is it
possible to accurately correlate a natural frequency value obtained from a structure to a
depth of scour around the foundation? How much does scour affect the natural frequency
of a typical bridge? Can acceleration signals arising on a bridge structure due to a passing
vehicle be effectively used to detect the natural frequency for scour evaluation? It is
recognised also that scour can pose problems for offshore wind turbines located in harsh
hydraulic environments. The question arises whether scour can change the natural
frequency of an offshore wind turbine to such a degree to cause issues with potential
resonance from rotating blades? These issues are investigated in subsequent chapters of
this thesis.
34
Chapter 3 A comparison of initial stiffness formulations for small-strain soil-pile
dynamic Winkler modelling
Authors:
Luke J. Prendergast
Kenneth Gavin
Paper Status:
Under Review with Soil Dynamics and Earthquake Engineering
Note to Reader:
The analysis and modelling within this paper was undertaken by the candidate under the
supervision of Dr Kenneth Gavin. The modelling methods were originally aided by Dr
David Hester and modified for the current purpose by the candidate.
35
Chapter 3 – A comparison of initial stiffness formulations for small-strain soil-pile
3.1 Introduction
This thesis aims to investigate the potential viability of using changes in natural frequency
of a structure to detect and monitor scour around foundation piles. One way to link the
observed change in natural frequency of a structure to a depth of scour around the
foundation is to develop accurate numerical models of the structure and foundation system
so that a direct comparison can be made between a frequency value and the scour
condition of the foundation. This chapter is concerned with developing an accurate soil-
pile numerical model that uses site specific geotechnical data to model the dynamic
behaviour of the foundation system. The model developed in this chapter forms the basis
for all subsequent numerical modelling in this thesis. An overview of dynamic soil-
structure interaction is presented in section 3.1.1, background to the Winkler modelling
philosophy is presented in section 3.1.2 and the methodology employed to develop a viable
numerical framework in this chapter is discussed in section 3.1.3.
3.1.1 Dynamic soil-structure interaction (DSSI)
Dynamic Soil-Structure Interaction (DSSI) is a vital aspect of the design of many

structures subjected to variable external excitation as part of their in service operation. The
response of soil-pile systems to lateral loading is an area of growing research interest. The
term ‘dynamic’ covers a broad spectrum of structural schemes ranging from large-strain
cyclic loading to small-strain system vibrations. The response of a soil-pile system is
heavily dependent on the nature and magnitude of the loading and a variety of modelling
approaches exist that aim to predict the response of these systems under various load
schemes. In particular, DSSI is an integral part in the design of offshore wind turbines,
which experience periodic excitation from a combination of environmental loading (wind
and wave action) and structural effects. The rotor spinning at a given rotational velocity
creates an excitation force with a frequency termed the 1P frequency. For a standard,
three-bladed, wind turbine, the blades passing the tower induce a second excitation force,
the frequency of which is termed the 3P frequency. Waves typically affect wind turbines
with excitation frequencies lower than the 1P band (see Figure 3-1). Flexible monopiles
are often designed in such a manner as to ensure that the global system has a natural
frequency between the 1P and 3P range and it is critical that a designer can accurately
predict the system’s natural frequency and avoid resonance (Doebling & Farrar 1996;
36
Tempel & Molenaar 2002). However, recent field measurements suggest that the soil
stiffness values recommended in offshore design codes (Det Norske Veritas 2007; API
2007) may result in significant errors in estimating the structure’s natural frequency.
Wave Spectrum
Rotor Spinning Blade Passing
Power Spectral Density
1P 3P
Soft - Soft - Stiff -

Soft Stiff Stiff
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
Frequency (Hz)
Figure 3-1 Frequency bands for typical offshore wind turbines (Hz)
Understanding a soil’s dynamic stiffness is also very important with regard to Structural
Health Monitoring (SHM). Recent advances in SHM use changes in the modal properties
of structures in order to infer some form of damage (Doebling & Farrar 1996). In the case
of bridges, most of the work to date has focused on monitoring the superstructure. More
recent research has begun to focus on using these damage detection methodologies on sub-
structural elements to detect the presence of scour, see Prendergast et al. (2013); Ju (2013);
Foti & Sabia (2011). In these cases, the analyses can be quite sensitive to the soil stiffness
assumed in the design where the dynamic oscillations typically remain in the small-strain
region.
DSSI is also very important in the field of earthquake engineering where propagating
ground motion waves can generate high stresses in a pile foundation. The stiffness contrast
between a pile and the surrounding soil tends to modify the transmitted excitation from
seismic shear waves leading to an effect known as kinematic interaction. Coupled with this
phenomenon, the dynamic response of a superstructure to a seismic excitation leads to
additional deformations in the pile foundation, an effect known as inertial interaction
(Kampitsis et al. 2013). It is very important to be able to accurately model the various
components of a soil-pile dynamic system so that the detrimental effects of external
37
actions may be mitigated by design. There are a variety of methods available to model the
dynamic behaviour of soil-pile systems. One such approach, termed the Winkler method,
is discussed in section 3.1.2.
3.1.2 Winkler modelling approach
In this chapter an approach, termed the Winkler model, commonly employed by structural
engineers for both static and dynamic soil-structure interaction problems is considered.
The model considers the soil as a system of discrete, mutually independent, closely-
spaced, springs (Winkler 1867; Dutta & Roy 2002). The pressure-deflection relationship at
any point of the foundation element can be generally represented by the equation shown in
Eq. (3.1), (in the absence of energy loss or inertial contributions – static analogue).
px, t   k wx,t  (Eq.3.1)
where px, t  is the applied pressure (N m-2) at a given unit of time, wx,t  is the
deflection (m) at a given time, and k is the coefficient of subgrade reaction (N m-3) (Note:
this expression is simplified for the purpose of illustration since a dynamic system will
typically include a significant inertial contribution). The key uncertainty with using a
Winkler model for dynamic applications lies with the specification of the parameters
required to model the behaviour of the soil under dynamic motion. The issues arise due to
the non-linear and inelastic nature of soil. These parameters include, among others; the
initial (elastic) stiffness, load-displacement response curves, cyclic degradation and
hardening parameters, unload-reload stiffness parameters and radiation and hysteretic
damping coefficients (Allotey & El Naggar 2008). A number of authors have developed
dynamic beam on non-linear Winkler foundation (BNWF) models for the purposes of
modelling the soil-structure response under large-strain dynamic loading. This topic has
received much interest in recent times from researchers working in the area of earthquake
engineering (Dezi et al. 2012; Kampitsis et al. 2013; Boulanger et al. 1999). Both Nogami
et al. (1992) and Allotey & El Naggar (2008) present a broad overview of the development
of general non-linear soil-pile interaction models for dynamic applications. In Allotey & El
Naggar (2008), a comprehensive discussion is given on the various soil-structure
interaction response features and how they can be modelled in a BNWF model. In
38
particular, the paper highlights the inefficiencies of static non-linear models in accounting
for cycle-by-cycle soil-structure interaction effects and kinematic interaction effects for
seismic applications, hence the need for the improved dynamic model. A generic cyclic
normal force-displacement scheme (cyclic p-y) incorporating backbone curves, unload-
reload curves, cyclic degradation and radiation damping as well as other modelling aspects
is discussed. Backbone curves are analogous to monotonic loading curves (static p-y), see
API (2007) and represent the non-linear load-displacement response of the system due to
the first application of the load (virgin loading). These can be represented by either a non-
linear or a multi-linear curve. Unload-reload curves represent the soil-structure behaviour
when the load is removed and re-applied (as per a cyclic load regime). The purpose of
modelling this aspect is such that the previous maximum force (stress) applied to the soil is
memorized by the model. Some coupled BNWF models are capable of directly modelling
cyclic degradation, however most other models require the specification of parameters that
are a function of dissipated hysteretic energy or cumulative displacement ductility.
Generally, cyclic degradation can be modelled by specifying stiffness or strength
degradation factors to be applied to unload-reload curves. The rate of degradation for
variable amplitude loading will depend on the number of load cycles. Radiation damping,
caused by the propagation of waves away from the foundation, can be modelled using a
linear or non-linear dashpot attached in parallel with a Winkler spring. There are a range of
methods available to specify damping constants for use with this model.
Kampitsis et al. (2013) describe the development of an advanced dynamic BNWF model,
developed based on Timoshenko beam theory, to investigate its accuracy in terms of
modelling kinematic and inertial interaction of a soil-pile-structure system for seismic
applications. The model encompasses the effects of geometrical non-linearity, rotary
inertia and shear deformation. A case study of a pile-column-bridge deck founded in two
cohesive soil layers and subjected to earthquake excitation is investigated. The efficacy of
the proposed model is investigated against a simplified beam finite-element (FE) model
and a fully 3-D continuum FE scheme. The spring configuration in the model consists of a
non-linear p-y spring connected in series with an elastic spring-damper element. The near
field plastification of the soil is accounted for by the non-linear spring and the far field
confining stiffness (viscoelastic characteristics) is incorporated by the elastic spring-
damper, known as a Kelvin-Voigt element, see Love (1927). The model is shown to be
39
capable of producing accurate results with a fraction of the computational time required by
the full 3-D FE model. Boulanger et al. (1999) evaluates the performance of a dynamic
BNWF model against the results of a series of dynamic centrifuge model tests, where two
piled supported structures are founded in a soil profile comprising soft clay overlying sand
and subjected to nine earthquake shaking events. A parametric study is undertaken to
assess the sensitivity of the analysis results to the chosen dynamic p-y parameters and site
response calculations. The backbone curves for the p-y analysis in the clay were based on
Matlock’s recommendation (Matlock 1970) for soft clay and the backbone curves for the
sand layer were based on recommendations from the American Petroleum Institute (API)
(API 2007). However, the initial stiffness component for the sand p-y curve was estimated
using the elastic theory of Vesic (Vesic 1961), with the small-strain shear modulus (G0)
adopted in the site response analyses. The dynamic response of the free-field soil and the
dynamic p-y analyses were undertaken separately. The results of the sensitivity studies
suggest that there is a greater uncertainty associated with the site response calculations
than with the dynamic p-y analysis when predicting the superstructure response for the
conditions tested in this study. However, by varying the dynamic p-y parameters by
different amounts, the peak superstructure displacement varied in the range of -26% to
+24% but the change was only greater than ±10% for 6 of the 32 parametric cases. The
method, therefore, would still seem to be quite sensitive to the magnitude of the
parameters chosen to model the non-linear dynamic response.
3.1.3 Synopsis of chapter
As mentioned above, one of the primary issues facing researchers and designers is that
beam on non-linear Winkler foundation models for dynamic applications typically require
the specification of many parameters and the results obtained in many cases can be quite
sensitive to the parameters chosen. Undertaking additional site investigations or laboratory
testing is often required to accurately calibrate the models developed so as to ensure that
the results obtained are realistic. In this chapter, the specification of an initial stiffness
parameter and its effect on the resulting dynamic response at very low strain levels is
investigated. For small-strain applications, a unique value of stiffness, either the small
strain shear (G0) or Young’s (E0) moduli can be used to model the elastic soil response.
This parameter is also one of the key inputs required to model the initial stiffness response
when large-strain non-linear dynamic response is modelled, see Boulanger et al. (1999).
40
The primary issue considered in this chapter is the specification of an initial stiffness
parameter that accurately reflects the in-situ small-strain stiffness of the combined soil-pile
system. This initial stiffness parameter, known as the coefficient of subgrade reaction in a
static analogue, can be a difficult parameter to specify in that it typically varies with
loading scheme, geometry of the foundation and the type of subgrade material.
This chapter presents a comparison of the performance of five different subgrade reaction
models used in the small-strain dynamic modelling of a soil-pile system. These models are
discussed in section 3.2. A numerical model programmed in MATLAB considers the pile
and surrounding soil as a beam supported by linear-elastic springs, as per the Winkler
approximation for small-strain applications. In the first instance, a range of synthetic soil
stiffness profiles and pile geometries are generated and a sensitivity analysis is performed
to predict the natural frequencies and mode shapes of the system and assess their
sensitivity to changes in the pile geometry and soil properties. Secondly, a field study is
performed in which two 340 mm diameter steel pipe piles were driven to various
slenderness ratios (depth, L divided by diameter, D) and their natural frequency and
damping ratios are measured. The natural frequency is measured by inputting a lateral
excitation to the pile head using a calibrated modal hammer and measuring the resulting
acceleration from accelerometers placed along the pile shaft. These input force and output
acceleration signals are used to develop Frequency Response Functions (FRFs) in order to
estimate their frequency content. Damping is measured using a curve-fitting technique to
estimate the damping ratio of the pile-soil system from the acceleration time-domain decay
resulting from the application of the impulse force. An estimate of the small-strain shear
modulus (G0) for the site was obtained using shear wave velocity data from a site
investigation. This small-strain G0 value was used to estimate coefficients of subgrade
reaction for the pile-soil systems using a range of theories discussed in section 3.2. A
direct modelling of the experimental impact test for each pile was undertaken and the
ability of each subgrade reaction model to predict the pile response at low strain levels is
assessed and discussed in the final section of the chapter.
The model developed in this chapter forms the basis for the numerical modelling of the
effect of scour on the frequency response of structures investigated in subsequent chapters
of this thesis.
41
3.2 Background to subgrade reaction theories
There have been a number of methods proposed to model soil-structure interaction using
Winkler springs. These models were typically developed for the application to static
problems; however, their efficacy at modelling the small-strain (elastic) dynamic
behaviour is of interest in this chapter. Some methods account for the flexural rigidity of
the foundation to which the spring is attached whereas other methods specify spring
constants without consideration for foundation geometry or flexibility. A brief review of
key assumptions in these models is discussed herein.
Biot (1937) presented a solution for the problem of an infinite beam with a concentrated
load, resting on a 3D elastic soil continuum. He found a correlation between the continuum
elastic theory and the Winkler model by equating the maximum moments in the infinite
beam and developed an empirical equation for the coefficient of subgrade reaction, ks
(Daloglu & Vallabhan 2000) as shown in Eq. (3.2).
0.108
0.95E0  E0 D 4 
ks    (Eq.3.2)
D(1  v s2 )  (1  v s2 ) EI 
where E0 is the small-strain Young’s modulus of the soil (N m-2), D is the width of the
foundation element (m), EI is the flexural rigidity of the foundation element (N m2), and vs
is the Poisson’s ratio of the foundation soil. Using a similar approach, Vesic (Vesic 1961)
derived an equation for ks by matching the maximum displacements of the infinite beam
(Daloglu & Vallabhan 2000) as shown in Eq. (3.3).
1 / 12
0.65E0  E0 D 4 
ks    (Eq.3.3)
D(1  v s2 )  EI 
which is of similar form to Eq. (3.2). A modified version of the equation shown in Eq.
(3.3) was used by Ashford & Juirnarongrit (2003) for modelling the dynamic response of
piles at low strains to investigate the effect of pile diameter on the initial modulus of
subgrade reaction. They substituted the constant value 0.65 in Eq. (3.3) for 1.0 since this
was deemed to provide the closest agreement between Vesic’s lower-bound estimation and
42
an upper-bound version provided by Bowles (1988) which indicated that the modulus of
subgrade reaction should be doubled to account for soil presence on both sides of the pile.
The value of 1.0 was found to provide the closest agreement between both estimations,
since in reality for a pile under lateral loading gapping may occur, and therefore soil may
lose contact with the foundation. Since piles are commonly modelled as beams supported
by springs for both static and dynamic applications, both the aforementioned formulations
for coefficient of subgrade reaction are suitable for comparison in this chapter.
Comparable expressions have been developed to compute the coefficient of horizontal

subgrade reaction for buried circular conduits (e.g. pipelines) (Okeagu & Abdel‐Sayed
1984; Sadrekarimi & Akbarzad 2009). Since these are similar to piles, the approaches are
considered in this study. Meyerhof & Baike (Okeagu & Abdel‐Sayed 1984; Sadrekarimi &
Akbarzad 2009) proposed an equation in which the coefficient of subgrade reaction for a
circular cross-section is expressed as a function of the soil elastic modulus (E0), the
Poisson’s ratio and the element width (D). The soil is assumed to be isotropic,
homogeneous and to have a linear stress-strain relationship, (a viable assumption for
small-strain dynamic loading). This is shown in Eq. (3.4).
E0
ks  (Eq.3.4)
D(1  vs2 )
Alternative formulations were developed by Klopple & Glock (Sadrekarimi & Akbarzad
2009; Elachachi et al. 2004; Okeagu & Abdel‐Sayed 1984), see Eq. (3.5) and Selvadurai
(Elachachi et al. 2004; Sadrekarimi & Akbarzad 2009), see Eq. (3.6).
2 E0
ks  (Eq.3.5)
D(1  vs )
0.65 E0
ks  (Eq.3.6)
D (1  vs2 )
The variations evident in Eq. (3.4) to Eq. (3.6) arise from the adoption of different
assumptions in their formulation. As a result for identical soil conditions and pile
43
geometries, different estimates of soil stiffness are obtained. In order to highlight the
variation of ks values obtained using the five methods considered, a sensitivity study was
performed in the first instance in which the coefficient of subgrade reaction was calculated
for a range of pile diameters and constant soil densities. For the purpose of this preliminary
study, Young’s moduli (E0) values of 50 MPa, 100 MPa and 150 MPa are used to
approximate the stiffness of loose, medium dense and dense sand deposits at small strains.
The difficulty faced by a designer in choosing stiffness parameters is evident in Table 3-1,
which presents values of ks predicted using the five models for a 0.75 m diameter pile
installed in soil with a constant elastic modulus (E0) of 50 MPa. The predicted ks values
ranged from 30.2 MN m-3 to 121.2 MN m-3. The two methods most widely used for
foundation analysis, Biot’s and Vesic’s produced ks values which varied by ≈ 27%. The
effect of increasing the pile diameter from 0.25 m to 2 m for the three soil densities
considered is illustrated in Figure 3-2, which reveals:
(i) As expected the ks value increased as the soil density increased. For a given
sand density ks reduced as the pile diameter increased.
(ii) The Klopple and Glock model in Eq. (3.5) predicted the highest values of ks,
whilst the Vesic model in Eq. (3.3) gave the lowest.
(iii) Whilst significant differences were evident between the ks values predicted
using the different models, the relative difference depended on the sand density
and pile geometry. For a pile with D = 0.5 m in loose sand the Vesic model
predicted ks values approximately 24% of those predicted using the Klopple
and Glock model. For the same pile geometry in dense sand, predicted ks values
using the Vesic model were approximately 26.6% of those using the Klopple
and Glock model.
Table 3-1 Comparison of coefficients of subgrade reaction models

Pile properties Soil properties Coef. Subgrade Reaction ks (MN m-3)
D (m) E (GPa) I (m4) vs E0 (MPa) Biot Vesic M&B K&G Selvadurai
0.75 200 0.00677 0.1 50 39.6 30.2 67.3 121.2 43.8
44
(a)
500 Biot
Vesic
400 Loose sand E0 = 50 MPa M&B
300 K&G
200 Selvadurai
100
0
0.25 0.5 0.75 1 1.25 1.5 1.75 2
Coeff. Subgrade Reaction (MN m )
-3
D (m)
(b)
1000
Medium dense sand E0 = 100 MPa
500
0
0.25 0.5 0.75 1 1.25 1.5 1.75 2
D (m)
(c)
1500
Dense sand E0 = 150 MPa
1000
500
0
0.25 0.5 0.75 1 1.25 1.5 1.75 2
D (m)
Figure 3-2 Variation of coefficient of subgrade reaction with diameter and soil elastic
modulus. (a) Loose sand with E0 = 50 MPa; (b) Medium dense sand with E0 = 100 MPa;
(c) Dense sand with E0 = 150 MPa
3.3 Structural modelling
In order to perform the sensitivity analyses of the different frequency responses predicted
by each formulation a numerical finite-element model of the pile and soil was created in
the MATLAB programming environment. Pile-soil systems can be modelled using beam
and spring elements, whereby the beam elements model the pile structural behaviour and
the spring elements model the soil behaviour. p-y springs can be used to model lateral soil
behaviour whereas t-z springs can be used to model the vertical soil reaction along the
pile. As the current analysis solely pertains to lateral pile response, only the elements
relevant to lateral motion are modelled. The pile is therefore modelled using standard four
45
degree of freedom (4-DOF) Euler-Bernoulli beam elements, each containing two nodes.
Each node has a translational and rotational degree of freedom and cannot deform axially.
The mass and stiffness matrices of this type of element are available in Kwon & Bang
(2000). The influence of the soil on the dynamic system was modelled using a linear p-y
spring, attached to one node of each beam element (see Figure 3-3). This combined
element has five degrees of freedom (5-DOF), with the lateral translation governed by a
combination of the beam element flexural rigidity and the soil-spring lateral stiffness (see
the insert in Figure 3-3) (Dutta & Roy 2002). It is noteworthy that the exclusion of vertical
degrees of freedom (and t-z springs) from the model will have no effect on the modelled
lateral response, due to the 1D nature of the system (i.e. mathematically, these are
mutually exclusive mechanisms). A system of elemental stiffness and mass matrices were
assembled together in line with the procedure outlined in Kwon & Bang (2000) in order to
create global stiffness and mass matrices capable of representing the entire pile-soil
system. The dynamic response of this discretized finite-element model can be obtained by
solving the second-order matrix differential equation shown in Eq. (3.7).
MG x CG x  K G x  F (Eq.3.7)
where M G , CG  and K G  are the (n x n) global mass, damping and stiffness matrices
for the pile model respectively. The vector x describes the displacement of each degree
of freedom for a given time step. Similarly, the vectors x  and x describe the velocity
and acceleration of each degree of freedom for a given time step. F is a vector defining
the external forces acting on each degree of freedom for a given time step.
A time-domain dynamic analysis can be undertaken by solving the second-order

differential equation shown in Eq. (3.7) using numerical integration. There are a number of
integration methods available that can be broadly segregated into implicit and explicit
methods. Implicit integration methods such as Wilson-theta and Newmark methods, see
Tedesco et al. (1999) can provide an unconditionally stable solution, through careful
choice of the integration parameters. The integration scheme employed for modelling in
this analysis is the Wilson-Theta method as described in Dukkipati (2009). A value of  =
1.4 was chosen as this is known to provide unconditional stability in the model (Tedesco et
46
al. 1999). It is noteworthy that the Wilson-theta method may underestimate the
contribution from higher order modes of vibration in the transient response, therefore if a
more comprehensive analysis is desired (i.e. observing higher modes), the Newmark
method may be more appropriate. However, for the purpose of obtaining the first natural
frequency response for the present analysis, either method works well. Integrating Eq.
(3.7) using numerical integration allows for obtaining numerical displacements, velocities
and accelerations in the time-domain from any degree of freedom in the model.
Damping is modelled as Rayleigh damping, where the damping matrix [CG] is formulated
as a linear combination of the mass and stiffness matrices for the combined soil-pile
system. This is shown in Eq. (3.8).
CG    MG   K G  (Eq.3.8)
where   212 / 1  2  and   2 / 1  2  . 1 and  2 are the first and second
circular frequencies of the system and 1   2   is the damping ratio adopted. Damping
can be modelled by specifying a damping ratio and the first and second circular
frequencies can be obtained by performing an eigenvalue analysis in the model. This is
achieved by specifying a system matrix as shown in Eq. (3.9).
Z  MG 1 K G  (Eq.3.9)
where [Z] is the system matrix, obtained from the matrix division of the global stiffness
and mass matrices. The eigenvalues and eigenvectors of this matrix represent the system
undamped natural frequencies and mode shapes and are obtained using MATLAB’s in-
built ‘eig’ function. The degrees of freedom corresponding to laterally displaced mode
shapes are extracted from each model such that a comparison of mode shapes can be
undertaken for each formulation of subgrade reaction. A schematic of the numerical model
is shown in Figure 3-3 with an insert showing the degrees of freedom for the combined
pile-soil elements.
47
Stick-up
D
E, I, , A
k S,N
Embedded Pile, L
5
2 1
3
k S,2
k S,1 x
Figure 3-3 Embedded pile numerical schematic
3.4 Sensitivity study
3.4.1 Background to analysis & soil stiffness
In order to conduct a sensitivity study investigating the different frequency responses

predicted by the implementation of different subgrade reaction (initial) stiffnesses, it was
first necessary to develop synthetic soil profiles for comparison purposes. The profiles
considered in this analysis represent a loose sand deposit and a dense sand deposit and they
vary with effective stress (depth). Open-ended, tubular steel piles embedded L = 10, 20, 30
and 40 m into the sand profiles are considered as the pile foundation element, each with a
1 m stick-up above ground level (see Figure 3-3). The pile diameter (D) was varied
between 0.25 m and 10 m using the increments shown in Table 3-2, such that L/D ratios
ranging from 4 to 40 were represented. The pile thickness is maintained constant at 50 mm
for each case. The sensitivity design cases considered are summarized in Table 3-2. The
grey cells represent design cases considered.
48
Table 3-2 Range of pile geometries considered in the sensitivity study (grey cells)
D (m)
0.25 0.5 1 2 5 10
10
L (m) 20
30
40
Soil profiles with depth-dependent stiffness were created. In order to achieve this, a
Young’s modulus value that varied with depth was hypothesized based on a number of
assumptions. In the first instance, the American Petroleum Institute design code (API
2007), which describes sand behaviour based on relative density (Dr) values, was used to
classify soil stiffness. Relative density values of 30% and 80% were considered to
approximate the conditions required to classify sand profiles as “loose” and “dense”
(Fugro 2011). These Dr values were considered constant over the numerical depth of
embedment of the pile. Synthetic Cone Penetration Test (CPT) tip resistance qc profiles
were generated based on these relative density values using a relation developed by Lunne
& Christopherson (1983) and shown in Eq. (3.10).
1  qc 
Dr  ln  0.7  (Eq.3.10a)
2.91 [60 'v  ] 
which can be re-arranged in order to obtain a synthetic cone tip resistance parameter (qc)
profile as follows:
qc  60 'v  exp 2.91Dr 

0.7
(Eq.3.10b)
where qc is the cone tip resistance (kPa) and  'v is the vertical effective stress (kPa)
For the loose and dense sand cases, values of bulk unit weight (γ) were taken as 18 kN m-3
and 20 kN m-3 respectively and the analysis assumes saturated soil conditions, with γw =
10 kN m-3. These qc profiles were converted to equivalent profiles of shear modulus (G0)
using the expression shown in Eq. (3.11) (Jardine et al. 2005; Baldi et al. 1989).
49

G 0  qc A  B  C 2 
1
(Eq.3.11)
where A = 0.0203; B = 0.00125; C = 1.216E-6 and   qc Pa 'v 0 

0.5
. Pa = 100 kPa,  'v 0 =
vertical effective stress (kPa). The resulting G0/qc values for the soil profiles assumed
ranged from 13-24 for the loose sand to 4-9 for the dense sand profiles. The Young’s
modulus (E0) profiles were obtained from the G0 profiles using Eq. (3.12), with a small-
strain Poisson ratio ν = 0.1. The synthetic qc and E0 profiles that were derived are shown in
Figure 3-4.
E0  2G0 1  v  (Eq.3.12)
These E0 profiles were used with each of the subgrade reaction formulations described in
section 3.2 to generate modulus of subgrade reaction profiles for use in the dynamic
analysis of each pile design case shown in Table 3-2. The first step involved converting the
coefficient of subgrade reaction (ks) values to a profile of the modulus of subgrade reaction
(K), by multiplying by the pile diameter. The next step involved creating individual spring
stiffness moduli by multiplying the modulus of subgrade reaction (K) at a given depth by
the spacing between subsequent springs at that depth (see Figure 3-3). Spring spacing was
maintained constant at 0.25 m for each design case considered.
50
(a) (b)
0 0
Loose CPT qc Loose E0
-2 Dense CPT qc -2 Dense E0
-4 -4
-6 -6
-8 -8
Depth (m)
Depth (m)
-10 -10
-12 -12
-14 -14
-16 -16
-18 -18
-20 -20
0 10 20 30 0 100 200 300
CPT qc (MPa) E0 (MPa)
Figure 3-4 Synthetic soil profiles. (a) Loose and dense CPT qc profiles (MPa); (b) Loose
and dense Young’s modulus (E0) profiles (MPa)
3.4.2 Dynamic response
The un-damped natural frequencies predicted by the numerical models for each subgrade
reaction formulation are summarized in Table 3-3. Both the loose sand case, see Table 3-
3(a), and the dense sand case, see Table 3-3(b), are presented, with the coefficient of
variation (COV) for each L and D also calculated. The coefficient of variation is defined as
the sample standard deviation (σ) divided by the sample mean (μ). The un-damped natural
frequencies were obtained by performing an eigenvalue analysis in each numerical model.

The percentage difference between Vesic’s model and Biot’s model, which are most
widely used for foundation design was also calculated, see Tables 3-3(a) and 3-3(b). Biot’s
model is seen to provide a higher estimate of the natural frequency with the difference
becoming larger as the L/D ratio of the pile decreased. For L/D ratios typical of offshore
51
piles (L/D ≈ 5), the difference between the natural frequency predicted using these two
subgrade models were in the range 13 to 15.5% for loose sand and 13.9% to 15.6% for
dense sand. A comparison of frequencies predicted using all the models is presented
graphically in Figure 3-5 for the loose sand case, with Figure 3-5(a) illustrating the effect
of varying the pile diameter for a pile of fixed length, L = 10 m. Figure 3-5(b) considers
the effect of varying the pile length for a pile with a fixed diameter, D = 1 m. The data
reveal that:
(i) When the pile length was fixed at 10 m, see Figure 3-5(a), all models predicted
that the frequency decreased when the pile diameter increased above 0.5 m.
(ii) When the pile diameter was fixed at 1 m, see Figure 3-5(b), the natural
frequency of the piles was not very sensitive to the pile length in the range
considered, and was relatively constant for L > 20 m or L/D > 20.
(iii) It is noteworthy that the formulations put forward by Biot and Vesic for the
spring stiffness are dependent on the pile structural properties as well as the soil
input stiffness, whereas the remaining methods are independent of the pile
properties. This can lead to changes in the order of predicted responses
depending on the magnitude of the pile structural properties under
investigation.
52
(a) BIOT (b) BIOT

35 30
VESIC VESIC
M&B M&B
28
K&G K&G
30
SELVADURAI SELVADURAI
26
25
Freq (Hz)
Freq (Hz)
24
Loose Sand L = 10 m 22 Loose Sand D = 1 m

20
20
15
18
10 16
0.25 0.5 1 2 10 20 30 40
Diameter (m) Length (m)
Figure 3-5 Comparison of predicted frequency responses for each subgrade reaction theory
for loose sand. (a) Frequency response by each theory for L=10m and varied pile diameter
(Hz); (b) Frequency response by each theory for D=1m and varied penetration depth (Hz)
Table 3-3(a) Comparison of design cases for loose sand profile

Loose sand
L D BIOT VESIC M&B K&G SELV Mean: Stand. Coef. % Diff Biot &
[m] [m] [Hz] [Hz] [Hz] [Hz] [Hz] Dev: Var: Vesic:
10 0.25 19.80 18.41 24.39 28.86 21.43 22.576 4.158 0.184 7.288
10 0.5 20.02 18.31 24.86 30.38 21.38 22.990 4.783 0.208 8.952
10 1 18.08 16.01 22.64 28.71 18.81 20.850 5.007 0.240 12.184
10 2 13.80 12.00 17.20 22.80 13.90 15.940 4.269 0.268 13.953
20 0.5 20.03 18.32 24.87 30.39 21.38 22.997 4.781 0.208 8.944
20 1 18.92 17.04 23.15 28.89 19.62 21.524 4.673 0.217 10.450
20 2 17.02 15.01 20.33 25.77 16.98 19.021 4.230 0.222 12.530
20 5 12.40 10.60 14.40 19.10 11.70 13.640 3.352 0.246 15.652
30 1 18.92 17.05 23.15 28.89 19.63 21.527 4.672 0.217 10.442
30 2 17.14 15.23 20.38 25.78 17.13 19.134 4.153 0.217 11.814
30 5 13.98 12.02 16.00 20.78 13.11 15.179 3.453 0.227 15.082
40 1 18.92 17.05 23.15 28.89 19.63 21.527 4.672 0.217 10.442
40 2 17.15 15.23 20.39 25.78 17.13 19.137 4.152 0.217 11.813
40 5 14.45 12.58 16.38 20.95 13.63 15.597 3.304 0.212 13.862
40 10 11.66 9.86 12.81 16.90 10.40 12.324 2.802 0.227 16.731
Max 0.268 16.731
Min 0.184 7.288
53
Table 3-3(b) Comparison of design cases for dense sand profile

Dense sand
L D BIOT VESIC M&B K&G SELV Mean: Stand. Coef. % Diff Biot &
[m] [m] [Hz] [Hz] [Hz] [Hz] [Hz] Dev: Var: Vesic:
10 0.25 24.24 22.51 29.09 34.14 25.72 27.138 4.600 0.169 7.369
10 0.5 25.12 22.91 30.47 37.03 26.28 28.363 5.569 0.196 9.185
10 1 23.62 20.95 28.57 35.76 24.00 26.578 5.816 0.219 11.955
10 2 18.70 16.10 22.50 29.80 18.30 21.080 5.391 0.256 14.943
20 0.5 25.13 22.92 30.47 37.03 26.29 28.368 5.566 0.196 9.179
20 1 24.04 21.58 28.77 35.80 24.43 26.922 5.597 0.208 10.794
20 2 21.87 19.30 25.48 32.20 21.39 24.050 5.070 0.211 12.479
20 5 16.60 14.10 18.70 24.80 15.20 17.880 4.233 0.237 16.287
30 1 24.04 21.58 28.77 35.80 24.43 26.923 5.596 0.208 10.788

30 2 21.90 19.37 25.50 32.21 21.44 24.083 5.049 0.210 12.237
30 5 18.27 15.71 20.35 26.21 16.79 19.466 4.153 0.213 15.050
40 1 24.04 21.58 28.77 35.80 24.43 26.923 5.596 0.208 10.788

40 2 21.90 19.38 25.50 32.21 21.44 24.085 5.048 0.210 12.229
40 5 18.53 16.10 20.55 26.26 17.14 19.716 4.021 0.204 14.024
40 10 15.37 12.97 16.45 21.56 13.42 15.954 3.438 0.216 16.932
Max 0.256 16.932
Min 0.169 7.369
A summary of all the data from the sensitivity analyses is set out in Table 3-3(a) and Table
3-3(b). Overall, the coefficient of variation between the frequencies predicted by each
model varied from 18.4% to 26.8% for the loose sand case, and 16.9% to 25.6% for the
dense sand case. This highlights the sensitivity of the calculated frequency response to the
choice of subgrade model when the soil and pile properties are maintained constant. The
percentage difference between Biot’s model and Vesic’s model varied from 7.3% to 16.7%
for the loose sand case and 7.4% to 16.9% for the dense sand case. This large variation
would have a significant effect on the frequency response predicted at the various limit
states considered in design, since this initial stiffness parameter governs the small-strain
behaviour for non-linear modelling.
In order to highlight the effect of the variation in the predicted natural frequency, a time-
domain dynamic response of a pile is undertaken by solving the second order differential
equation shown in Eq. (3.7) using the Wilson-theta integration scheme. The pile with a
diameter of 0.25 m and embedded depth of 10 m in loose sand is considered. An un-
damped time-domain acceleration signal is generated as a result of an excitation input, in
this case a short duration impulse force of magnitude equal to 500 N applied to the
transverse degree of freedom at the pile head. The analyses were repeated using the soil
54
stiffness derived for the five subgrade models considered. The results are shown in the
time domain in Figure 3-6(a). A signal of length equal to ten seconds is generated for each
subgrade model however, for clarity only the first 0.5 seconds are shown. From this plot, it
is evident that each signal is oscillating at a different frequency and amplitude. The
responses in Figure 3-6(a) were analysed with a fast Fourier transform to obtain the
frequency response spectra, shown in Figure 3-6(b). The peak of each plot represents the
dominant frequency and should closely approximate the value obtained in the eigenvalue
analyses in Table 3-3 allowing for some signal resolution errors. The significant effect of
input stiffness on the natural frequency obtained is illustrated clearly in Figure 3-6(b) by
the disparity between the frequency peaks for constant soil and pile properties.
(a) Biot
0.5 Vesic
Acceleration (g)
M&B
K&G
0 Selv
-0.5
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Time (s)
(b) Biot
2000 Vesic
28.81 Hz
19.78 Hz 24.35 Hz M&B
1500 21.42 Hz
Magnitude
K&G
1000 18.37 Hz Selv
500
0
15 20 25 30 35
Frequency (Hz)
Figure 3-6 Dynamic response example - design case 1; (a) Acceleration responses
predicted by each model (g); (b) Frequency spectrum of acceleration signals in (a) (Hz)
The disparity in predicted frequency can be further highlighted by observing the

differences in the first mode shapes extracted from the eigenvalue analyses in the models.
These mode shapes are plotted for the same design case in Figure 3-7, where 0 m
represents the pile base, and a 1 m stick-up above ground level is considered. The mode
shapes correspond to the first mode of vibration and are plotted relative to distance from
the pile tip. As can be seen, a difference is observed which matches the response in Figure
3-6. The difference between the mode shapes shown in Figure 3-7 arises directly as a
55
result of the difference in the spring stiffness formulation from each subgrade reaction
method.
12
Ground Line
10
Dist along pile (m)
6
Biot
4 Vesic
M&B
2 K&G
Selv
0
-0.05 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
Modal Displacement
Figure 3-7 First mode shapes predicted by implementation of different subgrade reaction
theories
3.5 Field experiment
3.5.1 Test procedure and site investigation
In order to examine the effect of pile slenderness (L/D ratio) on the measured and
predicted pile response, a field test was undertaken. Two 340 mm diameter steel open-
ended piles were driven to an embedment depth of 7 m (L/D = 21) in dense, over-
consolidated sand. The sand around Pile 1 was then excavated (see Figure 3-9 for photo of
experiment) until the pile embedment depth was 4.5 m (L/D = 13). A similar excavation
process was undertaken for Pile 2 until the pile embedment depth was 3.1 m (L/D = 9). An
experiment was performed in which three accelerometers were fitted near the pile head of
each pile (see Figure 3-8) and a modal hammer was used to perform a small-strain
vibration test. The natural frequencies obtained in the experiment are compared in this
section to numerical predictions obtained using the modelling procedure outlined in
section 3.3.
56
200
100
Original Level
500 500
26 Beam Elements
Accelerometer 1
Accelerometer 2
40 Beam Elements
Pile
Accelerometer 3
Pile 1 Level
Blessington Sand
7000
46 Sprung-Beam Elements
Pile 2 Level
Blessington Sand
32 Sprung-Beam Elements
4500
R1
6 70
3100
5
R1
Section A-A
A A
(a) (b) (c)
Figure 3-8 Experimental layout.(a) Experimental schematic for Pile 1 & 2 (b) Numerical
schematic for Pile 1 (c) Numerical schematic for Pile 2 (all dimensions are in mm)
Figure 3-9 Layout of experiment – Pile 2
57
The field investigation was undertaken at the University College Dublin (UCD) dense sand
test bed located near Blessington, southwest of Dublin, Ireland. This test site has been
developed over the past eleven years and has been used for a number of model, prototype
and full scale foundation experiments (Gavin & Lehane 2007; Gavin & O’Kelly 2007;
Igoe et al. 2011; Gavin & Tolooiyan 2012). A full description of the geotechnical
properties of the site can be obtained in Gavin & Lehane (2007); Gavin et al. (2009);
Doherty et al. (2012) and Prendergast et al. (2013). The site is comprised of very dense,
fine sand with a relative density between 90% and 100%, as determined from sand
replacement tests. The sand has a bulk density of 2.10 Mg m-3 and a unit weight of 19.8 kN
m-3. Due to its angularity it has a relatively high constant volume friction angle (  'cv ).
Triaxial compression tests undertaken on reconstituted samples of Blessington sand
reported in Tolooiyan & Gavin (2011) reveal that the sand has a constant volume friction
angle of 37 and a peak friction angle (  ' p ) which varies from 54 near the ground surface
to 42 at 5 m below ground level (bgl). Ring shear tests reported in Doherty et al. (2012)
reveal a soil-on-soil average residual friction angle of 36, which broadly agrees with the
values reported above by Tolooiyan & Gavin (2011). The specific gravity of particles is
2.69. The maximum and minimum void ratios are 0.73 and 0.37 respectively. The
equilibrium water depth is approximately 13 m below ground level (bgl). The sand is
partially saturated, with the degree of saturation above the water table being between 63%
and 75%. The natural water content is relatively uniform at 10% – 12% above the water
table and was found not to vary seasonally. Particle size distribution analyses performed
on samples taken from depths ranging from 0.7 m to 2 m below ground level (bgl)
indicated that the mean particle size, D50, varied between 0.1 mm and 0.15 mm. The well-
graded angular sand has a fines content (percentage of clay or silt particles) of between 5%
and 10%. Samples typically had less than 10% coarse-grained particles (> 0.6 mm).The
shear wave velocity profile for the site is shown in Figure 3-10(a) and the derived Young’s
modulus (E0) profile for the site is shown in Figure 3-10(b). The shear wave velocity
profile (vs) was obtained using the Multi-channel Analysis of Surface Waves (MASW)
method, see Donohue et al. (2004). The E0 profile is obtained by first converting the shear
wave velocity measurements to the small-strain shear modulus (G0) using the relation
shown in Eq. (3.13).
58
G0  vs
2
(Eq.3.13)
where  is the soil density (kg m-3). Once the shear modulus profile is obtained, the
Young’s modulus (E0) profile can be derived using the relation shown in Eq. (3.12). This
derived E0 profile is used with each of the subgrade reaction formulations discussed in
section 3.2 to ascertain which formulation gives the closest approximation of the system
natural frequency at low strain for the given experimental geometries considered.
0 0
-1 -1
-2 -2
-3 -3
Depth (m)
Depth (m)
-4 -4
-5 -5
-6 -6
-7 -7
-8 -8
150 200 250 300 100 150 200 250 300 350
-1 E0 (MPa)
Vs (m s )
Figure 3-10 MASW test results from Blessington test site. (a) Shear wave velocity
measurements (m s-1), (b) Young’s modulus (E0) profile (MPa)
3.5.2 Experimental frequency response and damping ratios
The accelerometers placed on the pile were hard-wired into a Campbell Scientific
CR9000x Datalogger and programmed to scan at 1000 Hz. This scan rate was more than
adequate to ensure an accurate estimation of the first natural frequency of the system under
59
test. In fact, a much lower scanning frequency can be used to estimate the first natural
frequency of the system; however the high scan rate was chosen to ensure highly accurate
data was obtained.
Both ambient and impact vibration testing were undertaken on each pile to obtain an
estimate of the system natural frequency at low strains. The ambient testing regime
involved switching on the acquisition system to begin scanning at 1000 Hz and monitoring
the system response for a period of five minutes. The input excitation for this testing
regime comes from environmental influences such as wind and other sources. The ambient
acceleration signals were analysed using two methods. The first method involved passing
each individual time-domain acceleration signal from each accelerometer through a fast
Fourier transform algorithm implemented in MATLAB. The second method involved
analysing all three time-domain accelerometer signals simultaneously using Frequency
Domain Decomposition (FDD) (Brincker et al. 2001). The FDD procedure is an
improvement on the classical Fourier transform approach as it is less sensitive to frequency
resolution problems and allows for easier estimation of closely spaced modes. The FDD
process involves taking the Singular Value Decomposition (SVD) of the spectral matrix
and decomposing it into a set of auto-spectral density functions that each correspond to a
single degree of freedom (SDOF) system (Brincker et al. 2000). The method is most
effective when the input excitation is broad-banded (white noise) and the structure is
lightly damped. Due to the relatively large free length of each pile above the new ground
line, this method is ideal for the analysis of the signals. The input ambient excitation is
assumed broad-banded.
The impact testing regime involved applying an impulse to each pile using a calibrated
modal hammer and measuring the output acceleration response. The hammer used was a
086D50 model manufactured by PCB Piezotronics. This is a large sledgehammer, which
was calibrated to excite low frequency resonances (fundamental mode) by fitting a soft
impact tip to the heavy impact head (mass = 5.5 kg). A number of impulses were applied
during each testing phase to ensure repeatability of the results. The impact test signals
were analysed by developing Frequency Response Functions (FRFs), which show the ratio
of the Fourier transform of the measured output v(t) to the Fourier transform of the input
stimulus u(t) (Elsaid & Seracino 2014; Dezi et al. 2012). This is shown in Eq. (3.14). In
60
this instance, the input stimulus is the time history of forces over the impact duration and
the output response is the measured acceleration from each accelerometer. A separate FRF
is developed for each force-acceleration pair and the results are averaged.
V  
H    (Eq.3.14)
U  
where V(ω) is the Fourier transform of the system output and U(ω) is the Fourier
transform of the system input. An example of an ambient and forced acceleration signal
and the resulting frequency response plots for Pile 1 is shown in Figure 3-11.
(a) -6 (b)
x 10
2
0.02
Acc 1 SV1
0.01 Acc 2 1.5 SV2
Amplitude
Acc 3
Acc (g)
0 1
-0.01 0.5
-0.02
0
0 50 100 150 200 10 20 30 40 50
Time (s) Frequency (Hz)
(c) (d)
2 0.01
Acc 1 Acc 1
0.008
1 Acc 2
FRF (g / N)
Acc 3
Acc (g)
0.006
0
0.004
-1
0.002
-2 0
0 0.2 0.4 0.6 0.8 1 0 10 20 30 40 50
Figure 3-11 Example of analysis for Pile 1. (a) Ambient acceleration signals (200 second
segment shown) (g), (b) Frequency Domain Decomposition (FDD) of ambient signals
shown in (a) (Hz), (c) Forced acceleration response from top accelerometer (g), (d) FRFs
of forced accelerations responses (Hz)
Figure 3-11(a) shows a 200 second long sample of the ambient acceleration signal
obtained from the vibration testing of Pile 1. In this figure, it is clear that the signal
resolution is somewhat poor from the sensors. In terms of frequency detection, this is not a
61
major issue as exemplified in Figure 3-11(b) however it may pose issues when attempting
to measure a damping ratio from the signals. Figure 3-11(b) shows the frequency response
spectrum obtained using the FDD algorithm on all three of the signals shown in Figure 3-
11(a). Both Singular Value 1 (SV1) and Singular Value 2 (SV2) are shown on the plot as
obtained from the singular value decomposition of the spectral matrix. Figure 3-11(c)
shows the acceleration response measured as a result of applying the impact force from the
modal hammer. The response shown is from accelerometer 1 (the other two accelerations
are omitted for clarity) which is at the top of the pile. The impulse force was applied at a
distance of 1 m below the pile head. Figure 3-11(d) shows the FRFs developed for the
impulse applied to the pile and each of the acceleration responses measured. The results
from each accelerometer were quite consistent. The slightly lower natural frequencies
measured using the impact response testing relative to the ambient response testing may be
as a result of the strain-level dependence of soil stiffness, i.e. the act of impacting the pile
with the modal hammer may be imparting larger strains than in the ambient case resulting
in a lower mobilised soil stiffness response. The lower frequencies in the impact response
testing may also be due to signal resolution issues with the shorter signals, which can be
resolved by using longer signals in the analysis. The slight difference in frequency (≈
4.6%) is evident between Figures 3-11(b) and 3-11(d) and is also displayed in Table 3-4.
Damping was measured from the decay in the acceleration signals resulting from the
impact testing. The purpose of measuring the damping ratio for the experimental signals is
to allow a direct modelling of the experimental tests in the numerical environment so that
the predicted frequencies from the numerical model can be compared directly to the
experimental results (i.e. a comparison of damped natural frequencies from the experiment
and the model can be made). The measured damping ratios can be modelled in the
numerical model using the Rayleigh damping approach discussed in section 3.3. The
damping ratio for each of the impact vibration tests was measured from the exponential
time-domain decay as seen in the signal shown in Figure 3-12. There are a number of ways
to estimate damping such as logarithmic decrement method (Ashford & Juirnarongrit
2003; Gutenbrunner et al. 2007; Dezi et al. 2012; Chopra 1981), half-power bandwidth
method (Tsai et al. 2011; Sohn et al. 2004) and curve-fitting (Gutenbrunner et al. 2007), as
well as numerous others. In this chapter, damping was estimated using a curve fitting
technique, whereby an exponential decay function is fit to the oscillation peaks of the time-
62
domain signal in a least-squares sense (Gutenbrunner et al. 2007). By way of a check, the
damping ratios were also measured using the logarithmic decrement technique averaged
over the number of peaks used for the curve-fitting method. For the exponential curve
fitting method, the equation takes the form shown in Eq. (3.15).
ut   u0 exp   nt   u0 exp t  (Eq.3.15)
where u0 and  are the curve-fitting parameters for the exponential function,  is the
damping ratio and  n is the un-damped circular frequency (rad s-1).  can be estimated
using the relation shown in Eq. (3.16).
 
  (Eq.3.16)
 n 2f
where f is the natural cyclic frequency (Hz) of the signal. As with all multi-degree of
freedom systems, the time-domain acceleration response will contain contributions from
numerous modes as well as from noise (Dezi et al. 2012). In the first instance, the signal is
filtered using a low-pass Butterworth filter to isolate the portion of the signal pertaining to
the first natural frequency in bending, which removes the influence of higher modes and
noise associated with the impact testing from the signal. Once filtered, an exponential
decay function is fitted to the peaks in the time domain using a least squares algorithm by
specifying the number of peaks to fit. In this case, a total of 50 peaks were used to obtain
an estimate of the damping ratio for the Pile 1 testing and 30 peaks for the Pile 2 testing
(due to the lower frequency and hence lower number of peaks available for Pile 2). The
first peak is ignored in each signal as this typically arises as a result of the application of
the impulse force to the pile.
The filtering and curve-fitting process is shown in Figure 3-12. Figure 3-12(a) shows the
original and filtered accelerations signals from the top accelerometer of Pile 2 due to the
application of an impulse from the modal hammer. From this figure, it can be seen that the
unfiltered signal in grey contains a significant amount of high frequency vibrations, which
arise as a result of the application of the impulse force to the pile inducing localized cross-
sectional deformation in the vicinity of the impact. Similar examples of this phenomenon
63
can be found in Dezi et al. (2012) and Prendergast et al. (2013). Filtering the signal has the
effect of removing the high frequency oscillations at the beginning of the unfiltered signal
and exposing the signal pertaining to the first natural (bending) frequency of vibration. The
filtered signal is shown in heavy black. Figure 3-12(b) illustrates the exponential curve
fitting to the filtered signal to estimate the damping ratio. As mentioned, the first peak is
ignored as this is generally understood to be as a result of the application of the impulse
force (Dezi et al. 2012).
(a) (b)
1.5 Unfiltered Acc 1.5 Filtered Acc
Filtered Acc Exponential decay
1 1
u(t) = u0 e-t
0.5 0.5
Acceleration (g)
Acceleration (g)
0 0
-0.5 -0.5
-1 -1
-1.5 -1.5
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Time (s) Time (s)
Figure 3-12 Measurement of damping ratio. (a) Unfiltered and filtered accelerations (g);
(b) Fitting exponential decay function
The results of the experimental analysis of both piles are presented in Table 3-4. The
damping ratios shown are the average values taken over a number of impacts for the top
accelerometer located near the head of each pile. No damping information is shown for the
ambient tests as the presence of noise in the signals made the estimation of damping using
the FDD approach very difficult.
64
Table 3-4 Experimental results from pile testing

Pile 1 - L/D = 13
Method: Frequency ±Standard deviation Damping ratio ±Standard
(Hz): deviation (%)
Ambient (FDD) 21 -
Forced (FRF) 20.06±0.098 1.8 ±0.076
Pile 2 - L/D = 9
Method: Frequency ±Standard deviation Damping ratio ±Standard
(Hz): deviation (%)
Ambient (FDD) 12.49 -
Forced (FRF) 12.14±0.122 1.26 ±0.1
3.6 Comparison of experimental results with numerical analysis
In order to compare the experimental results with numerical predictions obtained using the
five subgrade reaction stiffness formulations, a direct modelling of the forced vibration test
was undertaken for both Pile 1 and Pile 2. The reason behind comparing the forced
vibration tests to numerical models instead of comparing the ambient tests is simply
because the input excitation for the ambient test is unknown hence modelling this would
require modelling a random excitation input. Since a number of impact tests were
undertaken on each pile, to facilitate a direct comparison, only the first impact test is
modelled for each pile (however, it is noteworthy that the results are similar for each case,
see Table 3-4). The input force measured from the modal hammer for each pile test was
modelled as an input to the numerical model and is applied to the lateral degree of freedom
closest to the point of application of the force on the real system. For Pile 1, the force is
applied 1 m below the pile head and for Pile 2 the force is applied 2 m below the pile head.
The damping ratio measured using the curve-fitting technique described above was
modelled in the numerical model using the Rayleigh damping formulation discussed in
section 3.3.
The stiffness values derived for Pile 1 using the five subgrade reaction models are shown
in Figure 3-13, with the individual spring constants highlighted using markers. The
stiffness profile is plotted relative to the original ground level (pre-excavation) hence why
the depth begins at -2.5 m below ground level. From this figure, the disparity in predicted
stiffness using each of the five methods is evident once again by the spread of values.
65
-2.5
-3
BIOT
-3.5 VESIC
M&B
-4 K&G
Depth (m BGL)
SELVADURAI
-4.5
-5
-5.5
-6
-6.5
-7
0 1 2 3 4 5 6
-1 7
Stiffness (N m ) x 10
Figure 3-13 Stiffness profiles from each subgrade theory for the given pile dimensions for
Pile 1(N m-1)
Both the experimental results and numerical predictions for the impact test on Pile 1 are
shown in Figure 3-14. Each of the five subgrade models is shown. Figure 3-14(a) shows
the experimental and predicted time-domain acceleration signals for each of the subgrade
stiffness formulations for the first impact test on Pile 1. These are arranged in a column,
with each row showing the experimental results and the relevant numerical prediction. The
signals from Figure 3-14(a) were each analysed along with the force input to generate
Frequency Response Functions (FRF) to obtain the frequency content as presented in
Figure 3-14(b), see Eq. (3.14). This figure shows that the frequencies predicted by each
model employing the different subgrade reaction stiffnesses each overestimate the
measured system response to varying degrees. The percentage difference between the
experimental results and those predicted by each numerical model is presented in Table 3-
5.
66
(a) (b)
2 0.02 20.26 Hz
EXPT EXPT
BIOT BIOT
0 0.01
24.9 Hz
-2 0
0 0.5 1 1.5 2 15 20 25 30 35 40
2 0.02
EXPT 20.26 Hz EXPT
VESIC VESIC
0 0.01
23.93 Hz
-2 0
0 0.5 1 1.5 2 15 20 25 30 35 40
Acceleration (g)
2 0.02 20.26 Hz
EXPT EXPT
FRF (g N-1 )
M&B M&B
0 0.01
26.12 Hz
-2 0
0 0.5 1 1.5 2 15 20 25 30 35 40
2 0.02
EXPT 20.26 Hz EXPT
K&G K&G
0 0.01
27.83 Hz
-2 0
0 0.5 1 1.5 2 15 20 25 30 35 40
2 0.02
EXPT 20.26 Hz EXPT
SELVADURAI SELVADURAI
0 0.01
24.66 Hz
-2 0
0 0.5 1 1.5 2 15 20 25 30 35 40
Figure 3-14 Comparison of experimental and numerical results for Pile 1 impact test 1. (a)
Time-domain acceleration signals from the experiment and each of the five numerical
models (g), (b) Frequency response functions of experiment and each of the five numerical
models (Hz)
67
Table 3-5 Comparison of experimental results and numerical predictions

Pile 1 – L/D = 13
Method Numerical frequency Impact experimental (Hz) % Difference
(Hz)
BIOT 24.9 20.26 20.5
VESIC 23.93 20.26 16.6
M&B 26.12 20.26 25.3
K&G 27.83 20.26 31.5
SELVADURAI 24.66 20.26 19.6
Pile 2 – L/D = 9
Method Numerical frequency Impact experimental (Hz) % Difference
(Hz)
BIOT 13.18 12.21 7.6
VESIC 12.70 12.21 3.9
M&B 13.43 12.21 9.5
K&G 14.16 12.21 14.8
SELVADURAI 12.94 12.21 5.8
As indicated in Figure 3-14 and Table 3-5, all models overestimate the frequency of both
piles, with the largest over-prediction occurring for the pile with the highest L/D ratio. The
numerical models also slightly over-estimate the magnitude of the acceleration in the time-
domain, as can be seen in Figure 3-14(a). For Pile 2 (with a L/D ratio of 9), the models
predicted a reasonable estimate of the frequency with over-estimates in the range 3.9% to
14.8%, whereas for Pile 1 (with a L/D ratio of 13) the over-estimation ranged from 16.6%
to 31.5%.
One viable reason for the over prediction of each numerical model relative to the
experimental results may lie with the effect of pile installation (driving) on the in-situ G0,
an effect which was ignored in this study. While the experimental arrangement of inducing
very small strains into the soil mass by impacting the pile with the modal hammer may
justify the use of G0, the effect of pile installation inducing large strains in the shear zone
around the pile may change significantly the operational shear modulus (G) of the sand.
The difficulty with choosing an appropriate operational G value is that this parameter
typically decreases with strain and increases with stress. The large strains that arise during
the installation of the pile may reduce the operational G value in the vicinity of the pile.
However, this effect may be counterbalanced by an increase in the far-field confining
68
stiffness, an effect which arises due to the high stresses and over-consolidation that occurs
as the pile tip passes during driving. In addition to these phenomena, aging has been shown
to increase the stiffness characteristics of a soil mass, which may offset the effects of
installation. As a result, the G0 profile was chosen for the analysis in this chapter due to the
difficulty in estimating an operational G value, which may have given improved results.
It is noteworthy that there may also exist some errors in the estimation of the G0 profile for
the test site from the multi-channel analysis of surface waves method. Any errors in the
estimation of the G0 profile will have knock-on effects for the comparison of the
experimental and numerical predictions, particularly in the case of Pile 2 where the L/D
ratio was obtained by reducing the depth of soil surrounding the pile to a greater degree
than for Pile 1. Moreover, the G0 profile adopted is assumed to be isotropic (same in both
vertical and horizontal directions), therefore potential anisotropic behaviour in the
Blessington sand is not accounted for. Errors in the assumed G0 profile may contribute for
some of the disparities evident between the experimental results and numerical model
predictions. It is also noteworthy that contribution of the mass of soil in the dynamic
modelling was not accounted for, which may have given improved predictions.
Under the assumptions made in this chapter, for both piles, Vesic’s model provided the
closest approximation to the experimental measurements. Therefore, for the purpose of
further numerical modelling undertaken in this thesis, the Vesic model will be used to
model the dynamic soil-pile interaction.
3.7 Conclusions
The purpose of the analysis in this chapter was to develop an accurate soil-pile numerical
model that could be calibrated from site specific geotechnical data. The idea behind this is
that if an accurate model could be developed, then an observed natural frequency value
from an actual structure could be correlated to a depth of scour around the foundation. This
chapter focussed on variations in the formulation of the initial (small-strain) stiffness of
pile-soil systems. A range of subgrade reaction theories are available with which to
estimate the initial stiffness of piles founded in sand. In this chapter, a sensitivity analysis
was undertaken to establish the coefficient of variation between the frequency responses
predicted using different stiffness formulations for a range of pile diameters and depths of
69
embedment. The primary focus of the analysis was to establish the percentage difference
between models proposed by Vesic and Biot, which were derived specifically for the case
of infinite beams resting on elastic foundations. This analysis was undertaken by
developing a numerical FE model using MATLAB, employing a Winkler spring-beam
philosophy. Furthermore, a field investigation to obtain the frequency and damping
response of two piles with different L/D ratios was undertaken at a dense sand test bed site.
The pile response was compared to numerical models employing the site specific stiffness
obtained from a geophysical investigation of the site and utilizing the different subgrade
reaction models to develop spring stiffness coefficients. In the models, the force from the
modal hammer and the damping ratios measured in the experiments were inputted into the
numerical model so that a direct comparison of damped natural frequencies could be
undertaken for each stiffness formulation. It was found that the Vesic model provided the
closest approximation to the experimental responses for the given geometries and soil
conditions considered in the experiments.
The investigation undertaken in this chapter has resulted in a viable numerical model
(namely, the Vesic model) capable of modelling the dynamic behaviour of soil-pile
systems at small-strains to a reasonable degree of accuracy. Subsequent chapters will build
on this model. Chapter 4 assesses the potential for this model to be used in estimating the
change in the natural frequency of a single pile affected by scour.
70
Chapter 4 An investigation of the changes in the natural frequency of a pile
affected by scour
Authors:
Luke J. Prendergast
David Hester
Kenneth Gavin
John O’Sullivan
Paper Status:
Published in Journal of Sound and Vibration 332 (2013) 6685-6702
Note to Reader:
The experiments and subsequent analysis were undertaken by the candidate. Dr David
Hester aided the candidate in the development of the numerical model. The works
presented were undertaken under the supervision of Dr Kenneth Gavin. Dr John
O’Sullivan provided guidance on hydraulic countermeasures for scour design.
71
Chapter 4 - An investigation of the changes in the natural frequency of a pile affected by
scour
4.1 Introduction
Scour can be defined as the excavation and removal of material from the bed and banks of
streams as a result of the erosive action of flowing water (Hamill 1999). In relation to
bridges local scour is focussed around piers and abutments and occurs as a result of the
increased velocities and associated vortices as water accelerates around these obstructions
(Hamill 1999). Bed material is removed in the immediate vicinity of the structure, which
can reduce the foundation stiffness and lead to catastrophic structural collapse (Bolduc et
al. 2008). A full discussion of the scour phenomenon is provided in section 2.1.
From the literature survey conducted in section 2.3, there is a general agreement that scour
may be detected using changes in the natural frequency of a structure. In this chapter, the
effect of scour on a piled foundation system is closely investigated through scaled
experimental tests in the laboratory and full-scale field tests. For these experiments, the top
of the pile is excited using an impulse force and the resulting acceleration response is
recorded. This is undertaken at a number of scour depths. The natural frequency for a
given scour depth can be determined using a Fourier transform applied to the acceleration
signals. The effect of scour can be observed through the decrease in natural frequency as
the depth of scour increases. The soil-pile numerical modelling framework developed in
Chapter 3 is applied (and enhanced) in this chapter to test its ability to track the changes in
the natural frequency of a single pile with increasing scour. Section 4.2 details the
laboratory and field testing and section 4.3 details the numerical modelling and theoretical
framework.
4.2 Methodology
4.2.1 Laboratory model
A laboratory scale model of a pile foundation structure was constructed in order to assess
the plausibility of measuring changes in natural frequency due to progressive scour. The
pile was installed in a sand matrix, which was housed in a steel mould. The pile was left to
protrude approximately half its length above the sand (see Figure 4-1). The pile is a 0.1 m
x 0.1 m steel square hollow section with a Young’s modulus (E) value of 2 x 108 kN m-2.
72
scour
Accelerometer
Scour Depth = 200mm

Laboratory Pile
768 mm
100 mm
1260 mm
A A
d = 50 mm
Scour Level A Initial Level
292 mm d d d d
100 mm
8 mm
492 mm
Scour Level E
Final Level
Section A-A
Mould
Blessington Sand
1000 mm
Figure 4-1 Laboratory model geometry
The mould used was a 1 m x 1 m x 1 m steel box. It was chosen as its high mass and
stiffness provided a rigid structural framework in which to conduct dynamic testing. The
mould was incrementally filled with Blessington sand (Co. Wicklow, Ireland). The sand
was compacted in increments of 100 mm. The sand had a bulk density of 2.03 Mg m -3 and
a specific gravity value of 2.69. A sieve analysis was undertaken on the sand to establish
its grading. Once the steel box had been filled to an initial level of 300 mm, the pile was
placed in the centre. Sand was then added in further 100 mm increments until a final fill
level of approximately 800 mm had been reached. The sand was incrementally compacted
to a high value of relative density (close to 100%).
A uniaxial accelerometer was installed at the top of the box section as shown in Figure 4-1.
The accelerometer is a capacitive spring-mass system with integrated sensor electronics
fabricated by Sensor’s UK. A full list of its technical specifications is available online
(Seika.de 2012). The accelerometer was programmed into a Campbell Scientific CR9000x
Datalogger using accompanying Loggernet software. The software was programmed to
take data samples at a scanning frequency of 1000 Hz. This was adequate in order to
receive a relatively full waveform such that effective post-processing may be undertaken.
73
scour
The system was excited using an impulse force. For the purpose of consistency, the force
was applied using a swinging arc mechanism, where a known weight was allowed to swing
through a fixed arc. The force was applied in the plane of the accelerometer. Once the
impulse force was applied, the system responded transiently (W.J. Staszewski 1998). It
was the transient response that underwent post-processing in MATLAB in order to assess
the relevant dynamic characteristics such as natural frequency and damping ratio.
The scour process was modelled as the incremental removal of sand from around the pile.
The experiment commenced at level A (see Figure 4-1), which corresponded to the initial
ground level. Sand was removed in 50 mm increments until a total of 200 mm of sand had
been removed (level E). At each level, the impulse force was applied and the acceleration
response recorded. Figure 4-2(a) shows the acceleration response for scour levels A and E
respectively. It can be seen from the figure that the period between successive oscillations
is larger for scour level E than for scour level A. The natural frequency at each scour level
was obtained using a Fourier transform. Figure 4-2(b) shows the frequency content of the
acceleration signals shown in Figure 4-2(a). In Figure 4-2(b), it can be seen that there has
been a clear reduction in frequency between scour levels A and E. The reason for the peak
at scour level E being of a lower magnitude than scour level A is due to inconsistencies
with the applied force magnitude using the swinging arc mechanism. The magnitude of the
peak in the frequency domain is related to the acceleration response magnitude in the time
domain, which explains this result. Fourier transforms were also carried out on the
acceleration signals recorded at all other scour levels between A and E. Figure 4-2(c)
shows the frequency observed at each scour level. There is a clear reduction in frequency
as the depth of scour increases.
74
scour
(a) (b)
50 3000
Acceleration (m s-2 )
Scour Level A Scour Level A

25 Scour Level E Scour Level E
Magnitude
2000
0
1000
-25
-50 0
0 0.1 0.2 0.3 0 20 40 60 80 100
(c)
70
Frequency (Hz)
60
Level A
50
Level E
40
30
0 20 40 60 80 100 120 140 160 180 200
Scour depth (mm)
Figure 4-2 Frequency change with scour. (a) Acceleration response at scour levels A and E
(m s-2), (b) Frequency content of signals shown in (a) (Hz), (c) Change in frequency with
scour (Hz)
The detection of scour using changes in natural frequency of a pile will, in reality, take
place under water. It is important to assess the effect that this water will have on the
natural frequency of the pile, as this may have an effect on the ability of the proposed
approach to identify the occurrence of scour. The sensitivity of the proposed technique to
the presence of water was investigated via laboratory experimentation. Three different
cantilever structures were fabricated with increasing flexural rigidity (EI). Each section
had a length of 1.26 m and a width of 0.1 m. The geometric properties of each are detailed
in Table 4-1 and shown graphically in Figure 4-3 (Figure 4-3 also shows the direction of
vibration of the sections). The sections were welded to baseplates such that they could be
fixed in place at the base of a (1 m3) tank that was subsequently filled with water. A
vibration testing regime was implemented whereby all three sections were displaced then
released and the resulting acceleration recorded. These acceleration signals were
subsequently analysed using a Fourier transform to determine the natural frequency of the
system. The sections were tested in air and in water and the natural frequency of the
section in both mediums was obtained. To ensure that the results obtained experimentally
75
scour
were sensible, they were compared to the analytically calculated frequency, obtained using
the well-known expression shown in Eq. (4.1).
f1 
1
1.8752 EI 4 (Eq.4.1)
2 AL
where f1 is the first natural frequency of the cantilever (Hz), E is the Young’s modulus (N
m-2), I is the moment of inertia of the cross-section (m4). The results of the investigation
are shown in Table 4-2.
Table 4-1 Test cantilever properties

Section: Length Width Depth Thickness Flexural Stiffness Index
(m) (m) (m) (m) Rigidity (EI) (EI/L3)
(N m2) (N m-1)
Stiff Section 1.26 0.1 0.1 0.008 8.37E+05 418362
Stiff-Flexible 1.26 0.1 0.05 0.006 1.28E+05 63915

Section
Flexible 1.26 0.1 - 0.0065 4.58E+02 229
Section
100 mm
6.5 mm
100 mm
100 mm
100 mm
50 mm
8 mm 6 mm Direction of
vibration
(a) (b) (c)
Figure 4-3 Section properties used in experimental vibration test. (a) stiff section, (b) stiff-
flexible section, (c) flexible section
76
scour
Table 4-2 Experimental results

Section: Analytical Avg. Frequency Avg. Frequency Percentage change in
Cantilever measured in Air measured in Water natural frequency
Frequency (Hz) (Hz) ± Standard (Hz) ± Standard dev. between air and water
dev. (%)
Stiff 67.1 62.26 ± 0.33 62.08 ± 0.15 0.3

Section
Stiff- 35.0 28.81 ± 0.17 27.08 ± 0.07 6
Flexible
Section
Flexible 3.3 3.052 ± 0.0 2.81 ± 0.0 8
Section
Table 4-2 displays the values of the measured change in frequency between all three
structures vibrating in air and in water. The difference between the analytical frequency
and the actual frequency measured in air arises due to the difficulty in fabricating a fully
rigid baseplate connection. It can be seen that the first natural frequency of the flexible
structure is much more affected by the presence of water than the first natural frequency of
the stiff structure, with changes of 8% and 0.3% respectively. It must be noted, however,
that the acceleration signals for the stiff section had to undergo band pass filtering in the
region of the analytical first natural frequency as it was initially difficult to obtain a clear
frequency peak from the signals. This difficulty arose due to the high stiffness of the ‘stiff’
system in bending leading to poor signal clarity (high noise fraction) and the difficulty in
securing the base beam from lateral swaying, which clouded the frequency spectra
adversely. Bridge piles could have a stiffness index (EI/L3) within the ranges tested and as
a result may have their natural frequency altered by the presence of water. The effect of
water added mass is ignored in this study but could potentially be an issue. Further details
of this experiment can be found in Appendix D.
4.2.2 Field test
A full scale field test was undertaken at the University College Dublin (UCD) dense sand
test site which is located in Blessington approximately 25 km southwest of Dublin City. A
full description of the site can be found in section 3.5.1. In addition to the site
characteristics described in section 3.5.1, ten Cone Penetration Tests (CPTs) were also
performed at the site. The average CPT tip resistance (qc) profile is shown in Figure 4-4(a),
along with the maximum and minimum measured qc envelopes. The values are relatively
consistent revealing a uniform sand deposit where qc increased from ≈ 10 MPa at ground
77
scour
level to ≈ 17 MPa at 2 m below ground level, thereafter increasing gradually with depth.
The shear wave profile measured using the Multi-Channel Analysis of Surface Waves
(MASW) method (Donohue et al. 2004) for the site is shown in Figure 4-4(b). The small-
strain shear modulus (G0) profiles derived from the profiles in Figures 4-4(a) and 4-4(b)
are shown in Figure 4-4(c). Given the bulk density of the sand is ≈ 2 Mg m-3, it is
straightforward to derive the small-strain shear modulus from the vs measurements using
Eq. (3.13). The variation of G0/qc with qc1 for a range of sands was examined by Lunne et
al. (1997) and Schnaid et al. (2004). For an aged, over-consolidated material at the stress
levels present at the site, a G0/qc ratio in the range 5 to 8 is expected. A value of 6 is
adopted here.
(a) (b) (c)

0 0 0
-1 -1 -1
Max qc
-2 -2 -2
-3 -3 -3
Depth (m)
Depth (m)
Depth (m)
Min qc
-4 -4 -4
Average qc
-5 -5 -5
-6 -6 -6
Go from vs
vs profile Go from CPT
-7 -7 -7
0 20 40 150 200 250 300 0 50 100 150
Cone tip resistance Shear wave Shear modulus
qc (MPa) velocity (m s -1 ) (G0 ) (MPa)
Figure 4-4 Blessington site properties. (a) Cone tip resistance qc (MPa), (b) Shear wave
profile (m s-1), (c) G0 profiles (MPa)
An open-ended steel pile was installed in the sand stratum, as illustrated in Figure 4-5. The
steel pile has a Young’s modulus (E) value of 2 x 108 kN m-2. The pile length is 8.76 m
78
scour
and, at the start of the test, the top of the pile protruded a distance of 2.26 m above the
sand. After driving, the level of sand inside the pile was plugged at approximately 2 m bgl.
Four accelerometers were fitted along the initial exposed length (see Figure 4-5). All were
hard-wired into a Campbell Scientific CR9000x datalogger. The program RTDAQ was
engaged to record samples at a scanning frequency of 1000 Hz. This scanning frequency
was sufficiently high to allow for the detection of numerous vibration modes.
Accelerometer 1
Accelerometer 2
500 500
Accelerometer 3
2260
1000
R1
Accelerometer 4
57
70
Initial Level R1
500
Scour Level -1
Blessington Sand
Scour Level -2
Pile Scour Level -3
Scour Level -4 Section A-A

8760
Scour Level -5
Scour Level -6
6500
Scour Level -7
Scour Level -8
A A
Scour Level -9
Scour Level -10
Scour Level -11
Scour Level -12
Base Level
Figure 4-5 Installed pile schematic (all dimensions are in mm)
Figure 4-5 shows the experimental arrangement for the full scale field test. There were
thirteen scour levels tested; initial ground level and twelve scour depths. Each scour level
79
scour
was separated by 0.5 m. An excavator was used to remove the sand from around the pile
for each test. At each level, a lateral impulse force was applied to the top of the pile using a
modal hammer and the transient acceleration response was recorded by the accelerometers.
Figure 4-6 gives an example of pile acceleration signals observed during the field test.
(Note: the acceleration signal shown in Figure 4-6 is for scour level -4).
(a) (b)
15000
50
Scour Level -4 Scour Level -4
Acceleration (m s )
-2
Analytical Cantilever 2.26m
Magnitude
10000
Length - Frequency
0 = 63.8 Hz
5000
-50
0
0 0.5 1 1.5 2 0 100 200 300 400 500
TIme (s) Frequency (Hz)
(c)
200
Scour Level -4
150
Magnitude
100
50
0
0 10 20 30 40 50 60
Frequency (Hz)
Figure 4-6 Isolating frequency zone of interest. (a) Acceleration signal at scour level -4 (m
s-2), (b) Frequency content of signal shown in (a) (Hz), (c) First natural frequency of pile
(Hz)
Figure 4-6(a) shows the time-domain acceleration response obtained from accelerometer 1
as a result of applying a lateral impulse force to the top of the pile while testing at scour
level -4 (see Figure 4-5 for test schematic). A Fourier transform was applied to this signal
to obtain the frequency content. The frequency content of the signal in Figure 4-6(a) is
displayed in Figure 4-6(b). Two large peaks at 86 Hz and 322 Hz dominate the frequency
spectrum. When the top of the pile is subjected to an impulse force from the modal
hammer, it can be seen that in addition to creating global bending in the pile (which is the
phenomenon of interest in this analysis), there is also some localised three-dimensional
distortion of the pile in the vicinity of the zone that experienced the impact. This distortion
results in a localised vibration that is picked up by the accelerometers. The frequency of
80
scour
these localised vibrations is 86 Hz and 322 Hz. Irrespective of the scour level being tested,
the same two frequencies dominated the frequency spectrum, i.e. because they are local
effects they are independent of scour depth. To further verify that the frequencies at 86 Hz
and 322 Hz were due to local distortion/vibration of the pile, a 1.2 m long free-free section
of pile was subjected to impact testing using the same modal hammer as was used in the
field test. When the resulting acceleration signals were analysed, the frequency spectrum
was again dominated by peaks at 86 Hz and 322 Hz. One way to reduce the interference of
localised vibration peaks from distorting the frequency plot is to discard the first second or
so of acceleration data immediately after the application of the impulse force. This has the
result of removing a portion of the impact related vibration from the frequency spectrum.
In the present analysis, one is interested in detecting the frequency peak corresponding to
the first natural frequency of the pile and observing how this peak changes with scour. By
treating the initial exposed length of pile as a rigid cantilever, one obtains an upper-bound
value for the natural frequency of the pile analytically using the expression shown in Eq.
(4.1). The initial exposed length is equal to 2.26 m, which gives a first natural frequency of
63.8 Hz. Therefore, for all scour depths, the natural frequency of the pile will be less than
63.8 Hz. By focussing on the frequency spectrum below 63 Hz, one can isolate the pile’s
first natural frequency. Figure 4-6(c) shows a magnified view of the zone from the origin
to 63 Hz, of Figure 4-6(b), and one can see a clear peak at 12.2 Hz. This corresponds to the
first natural frequency of the pile in bending, at scour level -4.
Once the frequency zone of interest has been defined by the upper-bound cantilever model,
one can observe the change in the frequency spectra for different levels of scour. Figure 4-
7(a) shows a time-history acceleration response obtained from accelerometer 1 due to the
applied impulse force, while testing at scour level -6 (see Figure 4-5 for schematic). Figure
4-7(b) displays the frequency content of the signal in Figure 4-7(a) up to a maximum limit
of 63.8 Hz. Figure 4-7(b) also displays the frequency content of the acceleration signal
when the scour level was -4. It can be seen that a reduction in natural frequency is
observed with increased depth of scour.
81
scour
(a) (b)
100 300
Scour Level -6 Scour Level -6
250 Scour Level -4
50
200
Magnitude
0 150
100
-50
50
-100 0
0 0.5 1 1.5 2 10 20 30 40 50 60
Figure 4-7 Change in frequency content between two scour depths. (a) Scour level -6
acceleration response (m s-2), (b) Frequency spectra for scour levels -4 and -6 (Hz)
Figure 4-8 shows the change in first natural frequency with increasing depth of scour. A
gradual reduction in frequency is observed as scour depth increases. For completeness, the
first natural frequency of an equivalent length rigid cantilever with the same geometric
properties is also displayed in this figure. The cantilever results define the upper-bound
threshold expected in the analysis. Our experimental results are in line with expectations,
since all observed frequencies lie below the upper bound values. The photograph in Figure
4-9 shows the experimental arrangement of the field test with scour testing at an advanced
stage.
-1
-2
Scour depth (m)
-3
-4
-5
-6 Experimental frequency
Fixed cantilever frequency
-7
0 5 10 15 20 25 30 35 40
Frequency (Hz)
Figure 4-8 Frequency change with scour depth (Hz)
82
scour
Figure 4-9 Scour test in progress at advanced stage
4.3 Theoretical analysis
It can be seen from the experimental response in Figure 4-8 that a decrease in frequency
may be observed as scour depth increases. However, it is the depth of scour that is often
the parameter of interest, rather than the frequency value. If the soil-structure interaction
could be modelled accurately, then it would be possible to obtain an indication of the
severity of the scour problem by observing the natural frequency of the structure. To this
end, a finite element model was developed in MATLAB to numerically simulate the
laboratory and field tests. The structural model as well as the methods used to determine
accurate geotechnical stiffness are described in the next section.
4.3.1 Structural model
In the numerical simulations, the pile is modelled as being supported by a series of

horizontal springs; see Figures 4-10(b) and 4-11(b). Referred to as the Winkler hypothesis
(Dutta & Roy 2002), this method is known to yield reasonably good performance and is
relatively straightforward to implement. The scour process is simulated by progressively
removing the springs from the model commencing with those nearest the top. The
exposed/embedded lengths at the commencement of the scour test for the laboratory and
83
scour
field piles are given in Figures 4-10(a) and 4-11(a) respectively. The numerical models
used to simulate the laboratory and field piles are shown in Figures 4-10(b) and 4-11(b)
respectively.
The free end of the pile was modelled using standard four degree of freedom beam
elements. The elemental stiffness and mass matrices for this kind of element are given in
Kwon & Bang (2000). The embedded end of the pile was modelled using five degree of
freedom spring-beam elements. An increased effective area was used to model the cross-
section of the pile containing plugged soil to account for the extra mass in the dynamic
motion. The global stiffness and mass matrices for the structure were assembled in line
with the procedure set out in Kwon & Bang (2000). The response of the discretised finite
element model to a given impulse force is described by the dynamic equation of motion
shown in Eq. (3.7). The impulse force in the model is applied to the transverse degree of
freedom at the top of the pile.
Damping was modelled as Rayleigh damping, where the damping matrix is defined as a
linear combination of the mass and stiffness matrices for the global system as shown in Eq.
(3.8). The damping ratio used in the model ( 1   2   ) was the same as that calculated
from the experimental data. The damping ratio was obtained using a relatively
straightforward exponential curve fitting process (Gutenbrunner et al. 2007), see section
3.5.2.
The dynamic response of the system is calculated by solving the second order differential
equation shown in Eq. (3.7). The equation is solved using numerical integration. The
vectors x, x  and x are determined for each time step. The integration scheme used
was the Wilson-Theta method as given in Dukkipati (2009). A value of  = 1.4 was
chosen for the integration scheme as this provides unconditional stability in the model
(Tedesco et al. 1999). An analysis to calculate the system eigenvalues, which correspond
to the natural frequencies and the eigenvectors, which correspond to the mode shapes, was
also carried out, see Eq. (3.9). In order to check the numerical model and ensure its correct
operation, the static displacements, natural frequencies and mode shapes predicted by the
model were checked against those predicted by the commercial finite element package
PATRAN. Good agreement between both models was observed.
84
scour
23 Beam Elements
(each 31.5mm long)
768 mm
Sand - Air Interface
492 mm 17 Sprung-Beam
Elements
(each 31.5mm long)
(a) (b)
Figure 4-10 Numerical model to simulate laboratory test. (a) Geometry at start of scour
test, (b) Numerical model corresponding to dimensions shown in (a) (all dimensions are in
mm)
10 Beam
2260 mm
Elements
(each 219 mm
long)

30 Sprung-Beam Elements (each 219 mm long)
Blessington
Sand
6500 mm
(a) (b)
Figure 4-11 Numerical model to simulate field test. (a) Geometry at start of scour test,
(b) Numerical model corresponding to dimensions shown in (a) (all dimensions are in mm)
85
scour
4.3.2 Geotechnical stiffness determination
For the numerical model to correctly simulate the observed behaviour of the laboratory or
field piles, it is crucial that the stiffness assigned to the springs accurately reflects the
lateral stiffness of the sand around the pile. In this study, two methods are considered to
model lateral spring stiffness values.
The first method stems from the Vesic model developed in Chapter 3, Eq. (3.3), which
uses the small strain stiffness values measured directly or indirectly at the site. Direct
measurements of the small-strain shear modulus (G0) were derived from shear wave
velocity measurements made using the MASW technique. For sites where MASW
measurements are not available the use of indirect measures of small-strain stiffness were
considered, in this instance linking G0 and CPT qc. The rigidity index (ratio of small-strain
shear modulus to strength) is a useful index in terms of soil classification. For a given
deposit G0/qc increases with sand age and cementation (Schnaid et al. 2004). The variation
of G0/qc with qc1 for a range of sands was examined by Lunne et al. (1997); Schnaid et al.
(2004); Fahey et al. (2003) and others. For an aged, over-consolidated material at the stress
levels present at the test site, a rigidity index in the range 5 to 8 is expected. In the present
analysis, a value of 6 was assumed.
The second method utilises spring stiffness proposed in a widely employed design code for
laterally loaded piles, the American Petroleum Institute (API) method (API 2007). The
laboratory test was modelled using the API method only as no shear wave or CPT data was
available for this test. The results of both methods are presented in section 4.4. Individual
spring stiffness values calculated in this study are referred to as ksi.
4.3.2.1 Method 1 – small-strain stiffness (G0)
The stiffness of soils is strain level dependent (Atkinson 2000). Since dynamic loading
imparts very small strains in a soil mass, the small-strain shear modulus (G0) is a
particularly useful parameter for dynamic analyses. G0 is controlled by the in-situ soil state
(density and stress level) and also by the effects of ageing and cementation. Therefore, it is
relatively difficult to predict. However, G0 values can be measured using reliable and
relatively low cost in-situ geophysical techniques such as the MASW method. It is also
86
scour
possible to correlate a G0 value from in-situ measured CPT qc values (Schnaid et al. 2004).
The shear modulus (G0), which varies with depth, can be converted to equivalent lateral
spring stiffness (ksi) values at discrete locations along the pile shaft (see Figure 4-4(c) for
G0 profile derived from both the MASW method and CPT data). The first stage involves
converting the G0 profile to a Young’s modulus (E0) profile, using an appropriate small-
strain Poisson’s ratio () as shown in Eq. (3.12). The E0 profile may be converted to a
modulus of subgrade reaction (K) profile. The modulus of subgrade reaction (K) for the
sand-pile interface is derived using Eq. (4.2), which provides a relationship between the
modulus of subgrade reaction (of the sand) and the material properties of the pile in the
elastic continuum (Ashford & Juirnarongrit 2003). Eq. (4.2) is a modified version of the
Vesic equation presented in Eq. (3.3) (Vesic 1961). In this version, the presence of soil
contact on both sides of the pile is more explicitly accounted for by altering the coefficient
from 0.65 to 1.0 (Ashford & Juirnarongrit 2003).
1 / 12
1.0 E0  E0 D 4 
K   (Eq.4.2)
1 v2  E p I p 
where D = pile diameter (m); Ep = Young’s modulus of steel (N m-2); Ip = moment of

inertia of pile cross-section (m4). The individual spring stiffness values (ksi) may be
calculated by multiplying the modulus of subgrade reaction (K) at any particular depth by
the soil spring spacing (z), see Figure 4-11(b).
4.3.2.2 Method 2 – API design code
In geotechnical engineering practice, the analysis of laterally loaded piles is routinely

based on a Winkler model in which the pile-soil interaction problem is equivalent to a
beam supported by a series of uncoupled springs. The principle of using soil springs to
represent the soil reaction is illustrated in Figure 4-12. The lateral force (p) against lateral
displacement (y) response to a set of external loads (vertical (V), horizontal (H) and
moment (M)) can be represented by a linear or non-linear curve. The curve describes the
soil reaction (p) at a given depth as a function of the lateral movement (y). The spring
stiffness (Epy) is defined as the secant modulus of the p-y curve (see Figure 4-12). The
87
scour
approach is formulated in design methods including the American Petroleum Institute, API
(2007) code (API 2007).
V V
M M
H H
p
EI
E1 E1 y
E2
E2 y
E3
E3 y
Figure 4-12 Winkler approach and definition of p-y curves (Augustesen et al. 2009)
The application of the Winker approach for laterally loaded piles was first suggested by
Reese & Matlock (1956). In Murchison & O’Neill (1984), a database of fourteen lateral
load tests was compiled and a hyperbolic model to predict the pile response was proposed,
which has been incorporated into the current API design method as shown in Eq. (4.3).
 kx 
p  Apu tanh y  (Eq.4.3)
 Apu 
where pu = ultimate lateral resistance at depth ‘x’ below the surface (N m-1), k = initial
modulus of subgrade reaction (N m-3), A = empirical factor accounting for static or cyclic
loading conditions, y = lateral deflection (m). Values of k which depend on the soil density
(or friction angle) are given in the API design code, although k is assumed to be constant
for relative densities above 80%, or for friction angles above 39 (API 2007).
Eq. (4.3) allows for the existence of non-linear spring stiffness over a relatively large range
of displacements up to failure, since the stiffness is defined as the secant modulus of the
non-linear p-y curve. However, in the present analysis, the application of an impulse force
88
scour
to dynamically excite the pile mobilises only very small strains in the sand mass. By
differentiating Eq. (4.3) with respect to ‘y’, the initial (small-strain) stiffness in the p-y
curve can be determined (Kallehave et al. 2012; Augustesen et al. 2012). This is shown in
Eq. (4.4).
kx
dp Apu
y 0  Apu y 0  kx (Eq.4.4)
dy  kxy 
cosh 2  
 Apu 
Discrete spring stiffness values (ksi) along the pile shaft are obtained by multiplying the
small-strain stiffness modulus term in Eq. (4.4) (kx) by the spacing between springs, at
each depth (z).
The stiffness profiles derived using the small-strain shear modulus (G0) from the MASW
and CPT testing as well as that derived from the API method are shown in Figure 4-13.
0
Shear wave stiffness
-1 CPT stiffness
API stiffness
-2
Depth (m)
-3
-4
-5
-6
-7
0 2 4 6 8 10 12
Spring stiffness (N m-1 ) x 10
7
Figure 4-13 Spring stiffness profiles for Blessington site (N m-1)
89
scour
4.4 Results of laboratory test, field trial and numerical simulations
4.4.1 Numerical model of laboratory pile
When modelling the laboratory pile numerically, the spring stiffness values used in the
model were calculated using the API method, as no G0 data was available for the small
scale experiment. Figures 4-14(a) and 4-14(b) present the experimental and numerical
acceleration signals for scour levels A and E respectively, (Figure 4-1 shows a diagram of
scour levels). The solid line in Figure 4-14(a) shows the acceleration response at scour
level A when the laboratory pile was subject to an impact force, the dashed line shows the
acceleration response predicted by the numerical model. Figure 4-14(b) presents similar
data for scour level E. It can be seen in Figures 4-14(a) and 4-14(b) that the acceleration
signal predicted by the numerical model does not match well with the acceleration signal
observed experimentally. Figure 4-14(c) shows the frequency content of the signals shown
in Figures 4-14(a) and 4-14(b). For clarity, the magnitudes of the numerical and
experimental signals are plotted on the y-axes to the left and right of the figure
respectively. It can be seen in Figure 4-14(c) that the frequency of the signals generated
numerically are well below the frequencies observed experimentally. However, the
numerical signals do show a reduction in frequency between scour levels A and E. The
solid and dashed lines in Figure 4-14(d) respectively show the change in frequency with
increasing scour depth for the experimental test and the numerical simulation. It is evident
from Figure 4-14(d) that the numerical model is not correctly simulating the behaviour of
the laboratory pile. Firstly, the frequencies observed in the numerical signals are far lower
than those determined experimentally. Secondly, although the numerical model identifies
some reduction in frequency with increasing scour depth, the fall off is much less
pronounced than observed in experimental signals.
90
scour
40 40
Acceleration (m s-2 ) Exp acc Level A Exp acc Level E
Num acc Level A Num acc Level E
20 20
0 0
-20 -20
-40
0 0.1 0.2 0.3 0.4 0.1 0.2 0.3 0.4
4 Time (s) Time (s)
x 10
2 70
Experimental magnitude
Num L-A Num frequency (Hz)
Numerical magnitude
2000 60
Num L-E Exp frequency (Hz)
Frequency (Hz)
Exp L-A 50
1500 Level A
1 Exp L-E 40
1000
30 Level E
500 20
0 0 10
0 20 40 60 80 100 0 50 100 150 200 250
Frequency (Hz) Scour depth (mm)
numerical acceleration signals for scour level A (m s-2), (b) Experimental and numerical
acceleration signals for scour level E (m s-2), (c) Frequency content of signals (Hz), (d)
Frequency change with scour depth (Hz)
The error in the numerical model is due to the fact that the stiffness of the springs used in
the model are too low, and as a result the numerical model is underestimating the natural
frequency of the system. The API method is regarded as quite conservative and can
underestimate in-situ stiffness, particularly in the case of stiff piles (LeBlanc et al. 2010;
Doherty & Gavin 2012). So, some underestimation of sand stiffness is to be expected.
However, in the view of the authors, one particular aspect of the laboratory experiment is
causing the API method to substantially underestimate the stiffness of the sand around the
laboratory pile. The in-situ stiffness of soil depends on the sand density and mean stress
level, where mean stress is the sum of the vertical stress and twice the horizontal stress
over three (Fahey 1992). The sand used in the model was compacted into the mould in lifts
of 100 mm, leading to the generation of high lateral stresses and high in-situ density. As a
result of this compaction mechanism, the mean stress profile in the sample would be
relatively uniform, with perhaps some surface effects. As such, a uniform stiffness profile
for the model may be more appropriate. In order to improve the accuracy of the numerical
model, the spring stiffness values were increased from the API values to a uniform
stiffness of 5x106 N m-1 (see Figure 4-15).
91
scour
0
API Stiffness
Uniform Stiffness
-0.1
-0.2
Depth (m)
-0.3
-0.4
-0.5
0 1 2 3 4 5 6
-1 6
Stiffness (N m ) x 10
Figure 4-15 Stiffness profiles for the laboratory pile used in the numerical model
The solid line in Figure 4-15 shows the spring stiffness values prescribed by the API
approach. These were the spring stiffness constants used in the model that generated the
numerical acceleration signals shown in Figures 4-14(a) and 4-14(b). The dashed line in
Figure 4-15 shows the spring stiffness used in the model that generated the numerical
acceleration signals shown in Figures 4-16(a) and 4-16(b). In Figure 4-16(a) and 4-16(b), it
can be seen that there is good agreement between the signals generated numerically and
those observed experimentally for two different scour levels.
The solid and dashed lines in Figure 4-16(c) respectively show the change in frequency
with increasing scour depth for the experimental test and the revised numerical simulation.
Unlike Figure 4-14(d), this time the frequency changes predicted by the numerical model
provide a good match with those observed experimentally. The plots in Figure 4-16
demonstrate that if accurate values of spring stiffness are used, it is feasible to numerically
model the change in frequency with increasing scour depth. Therefore, a numerical model
could be used to estimate scour depths on a pile, i.e. one can match the frequency observed
in the field with the scour depth corresponding to this frequency in the numerical model.
92
scour
(a) (b)
Acceleration (m s ) 50 40
Exp Acc L-A Exp Acc L-E
-2
25 Num Acc L-A 20 Num Acc L-E
0 0
-25 -20
-50 -40
0 0.05 0.1 0.15 0.2 0.25 0.3 0 0.05 0.1 0.15 0.2 0.25 0.3
Time (s) Time (s)
(c)
70
Experimental frequency (Hz)
60
Frequency (Hz)
Numerical frequency (Hz)

Level A
50
Level E
40
30
20
0 50 100 150 200 250
Scour depth (mm)
theoretical acceleration at scour level A (m s-2), (b) Experimental and theoretical
acceleration at scour level E (m s-2), (c) Frequency change with scour depth (Hz)
4.4.2 Numerical model of field pile
The structural model that was developed in this study is a 1-D spring-beam model that is
designed to detect changes in first natural frequency of the pile as the depth of scour
increases. The reason behind choosing a 1-D model was to show that, for the purposes of
tracking changes in the first natural frequency, a simple numerical model was sufficient to
model the sand-pile interaction.
93
scour
(a) (b)
20 40
Impulse force (N) Unfiltered expt acc Scour Level -2
15
20
Force (N)
10
0
5
-20
0
-5 -40
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
Time (s) Time (s)
(c)
1
Numerical acceleration Scour Level -2
0.5 Filtered experimental acceleration Scour Level -2
a / amax
-0.5
-1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (s)
Figure 4-17 Theoretical and experimental results for field test. (a) Impulse force applied to
top of pile (N), (b) Unfiltered acceleration signal at scour level -2 (m s-2), (c) Filtered
signal compared with numerical model at scour level -2
Figure 4-17(a) shows the impulse force applied to the top of the pile while testing at scour
level -2, measured using the modal hammer. Figure 4-17(b) shows the experimental
acceleration response from accelerometer 1, due to the impulse force in Figure 4-17(a).
This acceleration response contains a significant amount of high frequency vibration (>
63.8 Hz), which is due to local effects in the pile generated during the testing regime (see
section 4.2.2). The authors did not aim to get the model to match the signal shown in
Figure 4-17(b) because much of the high frequency oscillations (> 63 Hz) evident in the
signal are local effects and are therefore not relevant to the global issue of trying to
identify scour. Incidentally, if one wanted to get the numerical results to match the signal
shown in Figure 4-17(b) it would be necessary to use a numerically intensive 3-D finite
element model to simulate the distortion of the pile in the vicinity of the zone that
experienced the impact. Also, omitting the first second or so of acceleration data from the
frequency analysis can aid in the reduction of these localised peaks. The authors show that
if a low pass filter is applied to the acceleration signal in Figure 4-17(b), the solid plot in
Figure 4-17(c) results. The application of the low pass filter has the effect of exposing the
portion of the acceleration signal pertaining to the transverse bending of the pile. When the
94
scour
impulse force in Figure 4-17(a) was applied in the numerical model, the dashed plot in
Figure 4-17(c) results. The unsteady response in the initial portion of both the
experimental and numerical signals in Figure 4-17(c), up to a value of approximately 0.1
seconds, is due to the duration of application of the impulse load. A similar study of this
phenomenon is available in Roy & Ganesan (1995). The numerical results presented in
Figure 4-17(c) were simulated using the spring stiffness values derived from G0 obtained
from shear wave velocity measurements, (see Figure 4-13 for exact spring stiffness
values). The plots in Figure 4-17(c) suggest that the numerical model is simulating very
accurately the frequency response associated with the transverse bending of the pile.
Therefore, the use of G0 values to approximate spring stiffness is quite good for dynamic
applications.
Figure 4-18 charts the drop in frequency with increasing scour depth and it shows that the
frequencies predicted by the numerical models using G0 derived stiffness, (namely the
shear wave velocity and CPT profiles) agreed well with those observed experimentally
across the range of scour depths tested. The numerical model developed using stiffness
derived from the API design code showed a tendency to underestimate the frequency
response for low depths of scour and overestimate slightly at higher depths of scour. It is
also noteworthy that for a pile with a different geometry (longer penetration depth), the
API method would overestimate the frequency response at higher scour depths
significantly due to its linearly increasing stiffness profile with depth. Overall it was found
that the experimentally measured first natural frequency drops from an initial value of 33
Hz for the case of zero scour to approximately 2 Hz for a scour depth of 6 m.
As a checking tool, at each scour level the frequency of an equivalent cantilever is

calculated analytically and the results are plotted using ‘x’ data markers in Figure 4-18. As
expected, for all scour depths, the frequencies observed experimentally are lower than the
frequency of an equivalent cantilever, i.e. the cantilever results provide an upper-bound
value in the analysis. Figure 4-18 shows that the models using stiffness derived from the
small-strain shear modulus (G0), obtained from in-situ MASW and CPT testing could be
used to provide a relatively accurate estimate of scour depth for a given observed
frequency.
95
scour
-1
-2
Scour depth (m)
-3
-4
Experimental frequency
-5 Numerical (shear wave) frequency
Numerical (CPT) frequency
-6 Numerical (API) frequency
Fixed cantilever frequency
-7
0 5 10 15 20 25 30 35 40
Frequency (Hz)
Figure 4-18 Frequency change with scour showing experimental response, numerical
response and analytical upper-bound equivalent cantilever (Hz)
4.5 Conclusions
This chapter has shown that it is possible to detect the presence of scour by monitoring the
natural frequency of a pile, and moreover that the depth of scour can be estimated from the
observed frequency.
Initially, a laboratory experiment to simulate pile scour was undertaken. In this test, a
model pile was installed in a sand matrix in a steel mould and was instrumented with an
accelerometer. The top layer of sand was progressively removed to simulate scour. At each
scour level, the pile was subjected to transverse impact loading and the resulting
acceleration signal was recorded. Fourier transforms were used to calculate the dominant
frequency at each scour level and it was found that there was a clear reduction in
frequency with increasing scour depth.
Subsequently, a similar approach was used to test a pile in the field. Unlike the laboratory
acceleration signals, the field acceleration signals contained many frequencies and initially
were more difficult to interpret. However, by regarding the initial exposed length of the
pile as the effective length of an equivalent cantilever, it was possible to calculate an upper
bound frequency. Therefore, tracking the change in frequency with increasing scour was
simply a matter of examining the frequency spectra between 0 Hz and the upper bound
96
scour
value. Similar to the laboratory tests, it was found that the natural frequency was inversely
proportional to the scour depth.
In reality, being able to track the change in frequency of a bridge pile is of limited use to
the structure owner especially if that frequency cannot be correlated to a scour depth.
However, if the observed frequency could be used to estimate a scour depth, this would
support more informed decision making. To that end, a spring-beam finite element model
of the sand-pile system was developed using the procedures outlined in Chapter 3 in order
to establish whether a correlation between observed frequency and scour depth was
possible. Obviously, for the model to be effective it is crucial that the spring stiffness
values used in the model accurately reflect the stiffness at the sand-pile interface. Two
recognised geotechnical methods were used to calculate the spring stiffness; the first
method stemmed from the study undertaken in Chapter 3 and used values derived from in-
situ G0 data obtained from the MASW method and a correlation to the CPT qc value. The
second method used stiffness derived from the API design code. The results from both
methods were fed into the development of separate numerical models. It was found that the
frequencies predicted by the numerical models using G0 derived stiffness (for a given
scour depth) agreed well with the frequencies observed experimentally. The API method,
however, did underestimate the frequency response for low depths of scour. The response
predicted by the API method would also be very sensitive to the depth of pile penetration
with longer penetration depths resulting in an over-estimation of stiffness at depth. It is
therefore not recommended to use the API method to derive stiffness profiles for the
purpose of scour depth estimation. The analyses showed that by using appropriate
geotechnical techniques and numerical modelling approaches, the sand-pile system could
be modelled accurately. Once the system is modelled correctly, it is possible to correlate
observed frequency with scour depth and thereby provide decision support to the structure
owner.
The novelty in this chapter is the development and validation of a method that uses the
actual structural response to indicate the severity of the scour problem, with possible
excitation by means of traffic loading on a bridge structure, for example. Any distress felt
by the structure as a result of scour will be indicated in its dynamic response and picked up
by accelerometers as a change in frequency. Other sensors that are placed near bridge
97
scour
foundations may not necessarily capture the scour effect on the structure as only the scour
hole local to the sensor will be captured. Although the sensor may be placed near the
problem area, the global effect of scour on the structure may still be missed. In the present
study, the global effect of scour is captured as a change in the response of the pile
foundation itself, which is a novel approach. The applications of the methods developed
herein extend beyond scour detection. Any stiffness degradation in the foundation due to
excessive cyclic loading may also be captured as a change in dynamic characteristics by
accelerometers placed on the structure, an effect which may be beneficial in the offshore
wind industry. This is due to the fact that the stiffness of soil is strain-level dependant and
any excessive loading may reduce the foundation stiffness, which could be detrimental to
the carrying capacity of the foundation elements. Accelerometers allow for direct
measurement of structural distress.
The study presented in this chapter deals with the development of the concepts required to
estimate the depth of scour based on an observed natural frequency of a single pile. Further
research is required to apply these concepts to the monitoring of full-scale real structures,
due to the complexity of the interactions at this scale. The current study serves as a
preliminary investigation in this regard. The validated pile-soil numerical model developed
in this chapter is applied to the case of modelling a full scale integral bridge in Chapter 5
and a full offshore wind turbine in Chapter 6.
98
Chapter 5 Determining the presence of scour around bridge foundations using
vehicle-induced vibrations: A feasibility study
Authors:
Luke J. Prendergast
David Hester
Kenneth Gavin
Paper Status:
Under Review with Engineering Structures
Note to Reader:
The analysis and modelling within this paper was undertaken by the candidate. The
modelling was closely aided by Dr David Hester and the vehicle model used was modified
from previous work by Dr Hester. The analysis was undertaken under the guidance and
supervision of Dr Kenneth Gavin.
99
Chapter 5 – Determining the presence of scour around bridge foundations using vehicle-
induced vibrations: A feasibility study
5.1 Introduction
This chapter builds on work developed in Chapter 4 whereby a numerical soil-structure

interaction model was developed to describe the change in natural frequency of a pile
foundation subjected to scour. The model was shown to be capable of tracking the change
in the natural frequency of a single pile affected by scour using input parameters which
included the structural properties of the pile and the small strain stiffness of the soil
(obtained from site investigation data). Experimental validation of the numerical model
was undertaken both on a laboratory model scale and full-scale field test on a 8.76 m long
pile embedded in dense sand. The pile geometry was typical of those used to support road
and rail bridges. This validated numerical model is represented by the pile/spring system
shown boxed in Figure 5-1(a).
The work described in Chapter 4, see Prendergast et al. (2013), was validated for the case
of a stand-alone pile foundation with forced vibration being imposed through the use of a
modal hammer. In reality, pile foundations are used to provide vertical and/or lateral
support for a structure (in this case a bridge) the presence of which will have a significant
effect on the natural frequency response of the pile-soil interaction problem. In this study,
the previous model from Chapter 4 is extended to consider the effect of a superstructure.
The structure considered is an integral bridge, comprised of two abutments, and a central
pier supported on piled foundations. The purpose of extending the model to include a
bridge superstructure is to ascertain the degree to which the structure’s natural frequency
changes due to scour of the foundation and moreover to investigate if it is practicable to
detect these changes by analysing the acceleration signals caused by traffic loading (i.e.
when a truck crosses the bridge). To make the simulated acceleration signals as realistic as
possible, interaction effects between the vehicle and the bridge are considered. In this
work, the change in natural frequency due to scour around the central pier foundation is
modelled, see Figure 5-1(b). The possibility of detecting these changes by analysing the
acceleration response signals from vehicular loading is considered. Details of the model
are given in section 5.2.
100
Vehicle
Road Surface
Embankment
Water Surface
Riverbed
Previously
validated
pile model
(a)
Vehicle
Road Surface
Embankment
Water Surface
Scour Hole
(b)
Figure 5-1 Schematic of model. (a) un-scoured, (b) post scour
5.2 Development of a Vehicle-Bridge-Soil Interaction (VBSI) model
A numerical model of a bridge incorporating both soil-structure interaction and vehicle-

bridge interaction was developed to investigate if measurable changes in natural frequency
due to scour could be obtained for a range of soil states. The model was built in the
101
MATLAB programming environment and the various components are discussed in the
following sub-sections, namely the model of the bridge structure (section 5.2.1), the
vehicle model (section 5.2.2), Vehicle-Bridge Interaction (VBI) (section 5.2.3) and Soil-
Structure Interaction (SSI) (section 5.2.4).
5.2.1 Bridge structure to be modelled and modelling philosophy
The bridge modelled is a two-span integral bridge. For this type of bridge the abutment is
formed using a series of vertical concrete columns and reinforced earth. The columns
support the deck and the reinforced earth retains the embankment fill (see Figure 5-2). The
bridge is not intended to represent any particular real-life structure. However, the
properties were chosen to be representative of bridges of this type. The bridge deck is
comprised of nine U10 concrete bridge beams (Concast 2014) supporting a 200 mm deep
concrete deck slab, which is typical of this type of bridge. The abutment consists of nine
columns supporting the bridge deck, each column is 500 mm in diameter and the columns
are at 1900 mm centres. This type of bridge does not have a conventional expansion joint
so the thermal movements of the deck have to be accommodated by lateral movements of
the abutment columns. To facilitate this movement, the abutment columns are cast in
vertical sleeves so that there is a gap of 50 to 100 mm on all sides, i.e. the reinforced earth
provides no lateral restraint to the columns. These abutment columns are therefore
assumed free to move laterally. Two piers support the bridge at the centre and have plan
dimensions of 1375 mm x 2625 mm. These piers are large stiff elements and they provide
lateral restraint to the bridge deck.
The abutment columns each rest on a pilecap, under which ten, 15 m long concrete bored
piles are used as the foundation system (see Figure 5-2). The pier columns each rest on a
pilecap supported by four piles. The scour action is assumed to be uniform along the
transverse length of a given support, so for modelling purposes, the structure shown in
Figure 5-2 is idealised as the 2D frame shown in Figure 5-3. (Note: scour is assumed to be
equal on both sides of the pier). The properties of each of the elements of the model in
Figure 5-3 are calculated by summing the properties of the individual components shown
in Figure 5-2. For example, the moment of inertia of the left abutment column shown in
Figure 5-3 is calculated by summing the moment of inertia of the nine abutment columns
shown in Figure 5-2. Similarly the stiffness of the two leaves of the pier shown in Section
102
A-A of Figure 5-2 is attributed to the central pier element of Figure 5-3. When
apportioning stiffness to the pile elements shown in Figure 5-3, a similar philosophy was
adopted. Details on calculating the spring stiffness coefficients for the soil are given in
section 5.2.4.
(a)
25000 25000
Bridge Deck
1600
Bank Seat Pier Cross-Head
500 1375 500
6000
Bridge Pier Abutment

Columns
A A
1800 600 2500 750 1800 600

15000
Abutment Piles Pier Foundation
3500 (b)
2625
3250
1900
Reinforced Earth
4500
3800
17100
Piles
Piles
Abutment Columns
Pier Column
Figure 5-2 Bridge layout with all dimensions shown in mm. (a) elevation, (b) section A-A
The elements used in the numerical bridge model are 6 degree-of-freedom (6-DOF) Euler-
Bernoulli frame elements (Kwon & Bang 2000). Each frame element has two nodes and
each node has an axial, transverse and a rotational degree of freedom as shown in the insert
in Figure 5-3. The lateral stiffness of the soil surrounding the piles is modelled using 2-
DOF spring elements connected to the pile nodes. No loss in bearing capacity of the piles
(due to scour) is considered in this model. This is because it is assumed that in service, any
103
loss of skin friction due to scour will be compensated by increased end bearing resistance
and therefore loss of bearing resistance is not considered. To represent this assumption in
the numerical model, the bottom end of the pile is modelled as being fully restrained in the
vertical direction. The global mass and stiffness matrices for the model are assembled
together according to the procedure outlined in Kwon & Bang (2000), taking appropriate
care to apportion the stiffness correctly at beam-column junctions. The dynamic response
of the bridge model is obtained by solving the second order matrix differential equation
shown in Eq. (3.7), where [MG], [CG] and [KG] are the (n x n) global mass, damping and
stiffness matrices for the full bridge model respectively. The vector x describes the
displacement of each degree of freedom of the bridge model for a given time step.
Similarly, the vectors x  and x describe the velocity and acceleration of each degree of
freedom for a given time step. F is a vector defining the external forces acting on each
degree of freedom for a given time step. Eq. (3.7) is solved using numerical integration.
The Wilson-Theta integration scheme was used in this analysis as given in Dukkipati
(2009). A value of  = 1.4 was chosen for the scheme as this provides unconditional
stability in the model (Tedesco et al. 1999).
Damping was considered in the model as Rayleigh damping, where the damping matrix
[CG] is defined as a linear combination of the mass and stiffness matrices for the global
bridge system as shown in Eq. (3.8). A damping ratio of 2% was assumed for all
simulations in this study.
Mode shapes and natural frequencies were extracted from the model by performing an
eigenvalue analysis on the system, see Eq. (3.9). In order to verify that the model was
operating correctly, the static displacements, mode shapes and natural frequencies
predicted by the model were verified against a commercially available finite-element
package. Good agreement was observed between the model and the commercial software.
5.2.2 Vehicle model
The vehicle model, which is used in this work is similar to a model described in Hester &
González (2012) and González & Hester (2013). The vehicle model has four degrees of
freedom, namely a vertical displacement for each of the two axles (y1 and y2), the body
104
bounce (yb) and body pitch (φp) (see Figure 5-3). The body has mass mb and has rotational
moment of inertia Ip (for pitch). The body is supported on a suspension/axle assembly. The
mass of the wheel/axle assembly is mw. The suspension has a stiffness Ks and a damping
coefficient Cs. Finally, the tyre is modelled as a spring with stiffness Kt. Table 5-1 provides
the parameters of the vehicle (Cantero et al. 2011; El Madany 1988). Using the properties
given in Table 5-1, stiffness [Kv], mass [Mv] and damping [Cv] matrices for the vehicle
can be populated. As the vehicle modelled has 4 DOFs, each of the vehicle matrices [Kv],
[Mv] and [Cv] are 4 x 4 matrices. By performing an eigenvalue analysis on the vehicle
system matrix, the natural frequencies of the vehicle for bounce, pitch, and front and rear
axle hops were determined to be 1.43 Hz, 2.07 Hz, 8.860 Hz and 10.22 Hz respectively.
S
S2 S1
Mb ,I p o
yb
p
K s2 C s2 Ks1 C s1
mw2 mw1
y2 K t2 y1 K t1 E,I, ,A
4
6 5
3 2
1
k S,N k S,N k S,N
k S,2 k S,2 k S,2

k S,1 k S,1 k S,1
x
Figure 5-3 Schematic of the Vehicle-Bridge-Soil Interaction (VBSI) model
105
Table 5-1 Parameters of vehicle model

Dimensional data
Dimensions (m) Wheel base (S) 5.5

Dist from centre of mass to front axle (S1) 3.66
Dist from centre of mass to rear axle (S2) 1.84
Mass and inertia
Mass (kg) Front wheel/axle mass (mw1) 700

Rear wheel/axle mass (mw2) 1,100
Sprung body mass (mb) 13,300
Inertia (kg m2) Pitch moment of inertia of truck (Ip) 41,008
Suspension
Spring stiffness (kN m-1) Front axle (Ks1) 400

Rear axle (Ks2) 1,000
-1
Damping (kN s m ) Front axle (Cs1) 10
Rear axle (Cs2) 10
-1
Tyre stiffness (kN m ) Front axle (Kt1) 1,750
Rear axle (Kt2) 3,500
5.2.3 Vehicle-Bridge Interaction (VBI)
Modelling the dynamic behaviour of a vehicle-bridge interaction system is complex as

there are two sub-systems, namely the moving vehicle and the bridge/substructure. These
two systems interact with each other via the contact forces that exist between the vehicle
wheels and the bridge surface, therefore mathematically the problem is coupled and time
dependant (Yang et al. 2004). It is necessary to solve both subsystems while ensuring
compatibility at the contact points (González n.d.). In this study, an iterative approach was
employed to implement the VBI model (Green & Cebon 1997; Yang & Fonder 1996).
In the first iteration, the contact forces applied to the finite-element model of the bridge are
determined using a ‘stationary’ road profile, i.e. the fact that the axle displacements will
change due to the bridge deflection is initially ignored. The contact forces between the
axles of the vehicle and the road profile are obtained by solving the equation of motion of
the vehicle, shown in Eq. (5.1) (Fryba 1999) for each time step.
106
M v xv  Cv x v  K v x v   Fv  (Eq.5.1)
where M v  , C v  and K v  are the 4 x 4 mass, damping and stiffness matrices for the
vehicle. The vector x v  describes the displacement of each of the 4 vehicle degrees of
freedom for a given time step in the analysis. The vectors x v  and x v  describe the
velocity and acceleration of the 4 vehicle degrees of freedom at each time step in the
analysis. The vector Fv  decribes the external forces acting on each of the 4 vehicle
degrees of freedom for a given time step in the analysis. Once obtained, these contact
forces are applied to the FE model of the bridge, i.e. it is possible to populate the force
vector {F} in Eq. (3.7) and then the Wilson-Theta integration scheme is used to calculate
the displacement of each of the bridge DOFs at each time step in the analysis. Using the
hermitian shape functions (interpolating functions), it is possible to calculate the vertical
bridge displacement under each vehicle axle for each time step in the analysis. This allows
a new road profile to be defined, i.e. the displacement of the bridge is added to the original
road irregularities. By repeating this procedure using the updated bridge road profile, a
new set of axle contact forces can be obtained. These calculations are iterated until a preset
convergence tolerance is achieved. Convergence is deemed to have occurred when the
difference in displacements between successive iterations is sufficiently small. Guidance
on convergence thresholds are given in Green & Cebon (1997).
The road profile used in the model is an array of numbers that defines the height of the
road irregularities at each time step in the analysis. An artificial road surface topography
corresponding to a specific ISO roughness classification for use in time-domain vehicle
vibration simulations can be generated as per Cebon (1999). In this study, only ISO Class
‘A’ (“very good”) profiles, typical of well-maintained highways, are used (see Figure 5-6).
5.2.4 Dynamic Soil-Structure Interaction
Soil-structure interaction is incorporated into the model by means of the Winkler method.
The soil is modelled as a system of discrete, mutually independent and closely-spaced
springs (Winkler 1867; Dutta & Roy 2002). The method for developing spring stiffness
values is described in section 4.3.2.1 whereby spring stiffness values were derived using
107
the small-strain shear modulus (G0) profile from the experimental site (Blessington,
Ireland). G0 data was derived using two methods, (i) A correlation to Cone Penetration
Test (CPT) tip resistance (qc) profiles and (ii) shear wave velocity measurements obtained
from the Multi-Channel Analysis of Surface Waves (MASW), see Donohue et al. (2004).
The resulting G0 profiles obtained from both methods matched well.
In this analysis, rather than modelling data from just one test site the response of a bridge
supported by piles founded in a range of sand densities, from loose to dense was
considered. To facilitate this modelling, three different synthetic CPT qc profiles
(corresponding to loose, medium-dense and dense sand) were generated. These qc profiles
were then used to calculate small-strain shear modulus (G0) profiles. From these G0
profiles, the spring stiffness coefficients were calculated using the method described in
section 4.3.2.1. The modelling process is summarised below.
5.2.4.1 Generating a theoretical qc profile
The method for developing a theoretical CPT qc profile is presented in detail in section
3.4.1, however for convenience some details are reproduced in this section. The strength
and stiffness of sand depends on the in-situ stress level and relative density (Dr). Design
codes such as the American Petroleum Institute (API) design code (API 2007) classify
relative density values of 30%, 50% and 80% to correspond to loose, medium-dense, and
dense sand deposits respectively. Assuming that Dr values in a given deposit are constant
with depth, synthetic CPT qc profiles were generated using Eq. (3.10). The vertical
effective stress (  'v ) for the three soil profiles; loose, medium-dense and dense sand are
calculated using bulk unit weights of 18 kN m-3, 19 kN m-3 and 20 kN m-3 respectively.
Given that the application considered in this chapter pertains to river bridges, the soil was
assumed to be saturated and effective unit weights were used in the calculations with the
bulk unit weight of water assumed to be 10 kN m-3. The qc profiles generated using Eq.
(3.10) for loose, medium-dense and dense sands are shown in Figure 5-4(a). The synthetic
qc profiles derived using this procedure, although idealised, do conform in accordance with
the values expected for the given soil conditions (Fugro 2011).
108
5.2.4.2 Calculating G0 profiles
The theoretical CPT qc profiles that were generated in section 5.2.4.1 can be converted to
profiles of small-strain shear modulus (G0) using Eq. (3.11). The results of this calculation
for loose, medium-dense and dense sand are shown in Figure 5-4(b). The ratio G0/qc
(known as the rigidity index) predicted varies from 15 to 22 for the loose sand profile and
from 5 to 8 for the dense sand profile, which is in the expected range for these soil
densities (Lunne et al. 1997).
5.2.4.3 Calculating spring stiffness coefficients (ksi)
The method for deriving individual spring moduli from the theoretical G0 profiles is
discussed in detail in section 4.3.2.1, however once again for convenience some details are
reproduced in this section. The first step in the procedure for deriving stiffness coefficients
(ksi) for the Winkler springs is to convert the G0 profiles to Young’s modulus (E0) profiles
using Eq. (3.12), see Figure 5-4(c). A Poisson’s ratio of 0.1 is adopted for the small strain
problem considered.
The E0 profiles may be converted to profiles of the modulus of subgrade reaction (K) for
the sand-pile interface using Eq. (5.2) (Prendergast et al. 2013; Ashford & Juirnarongrit
2003). This expression is similar to Eq. (4.2) but uses effective pile properties for the
group.
1 / 12
1.0 E0  E0 Deff 4 
K   (Eq.5.2)
1 v2  E p I p ,eff 
where Deff is the effective diameter of the pile group (m); Ep is the Young’s modulus of the
pile material (N m-2); Ip,eff is the effective moment of inertia of the pile group (m4). The
effective diameter of the pile group is obtained by summing the diameters of the individual
piles in the group and the effective moment of inertia of the pile group is obtained in the
same way. For example, in the bridge being modelled there are 10 no. 0.6 m diameter piles
in each abutment, each with a moment of inertia of 0.00636 m4. This results in Deff = 6 m
and Ip,eff = 0.0636 m4. The reason grouped properties are used in the spring stiffness
calculation is that for the purpose of modelling, it is assumed that the pile groups act in
109
block. Note: the units of K are N m-2 and Eq. (5.2) essentially provides a relationship
between the modulus of subgrade reaction of the sand and the material properties of the
pile in the elastic continuum, see Ashford & Juirnarongrit (2003). Individual spring
coefficients (ks1, ks2 etc., shown in Figure 5-3) may be obtained by multiplying the K value
at a given depth by the spacing between the springs at this depth. In this study springs were
spaced at 0.5 m centres and the stiffness of the individual springs for the different soil
stiffness profiles (loose, medium-dense and dense) are shown in Figure 5-4(d).
(a) (b)
0 loose 0 loose
medium-dense medium-dense
dense dense
-5 -5
Depth (m)
-10 Depth (m) -10
-15 -15
0 5 10 15 20 25 0 20 40 60 80 100 120
CPT qc (MPa) G0 (MPa)
(c) (d)
loose loose
0 0
medium-dense medium-dense
dense dense
-5 -5
Depth (m)
Depth (m)
-10 -10
-15 -15
0 50 100 150 200 250 0 0.5 1 1.5 2
E0 (MPa) -1 8
Spring stiffness ks (N m ) x 10
Figure 5-4 Postulated soil parameters for a loose, medium-dense and dense sand. (a) CPT
qc profiles (MPa), (b) G0 profiles (MPa), (c) E0 profiles (MPa) and (d) spring constants (ks)
for pier foundation piles (N m-1) for the analysis
5.3 Analysis & results
In the analyses performed using the model described above, a moving vehicle excites the
bridge. The vehicle crosses the bridge at a constant highway speed of 80 km hr-1 (22.22 m
s-1) and the horizontal acceleration from the top of the pier and abutments is recorded and
analysed. The effect of soil stiffness on the frequency changes with scour was examined
110
for the three soil stiffness profiles. To aid in choosing appropriate locations to place
accelerometers on the structure and to ascertain a baseline for the expected change in
natural frequency due to scour, an eigenvalue modal analysis was conducted in the first
instance.
5.3.1 Eigen frequencies and mode shapes
An eigenvalue analysis was conducted in the model to extract the fundamental frequency
of (lateral) vibration for different depths of scour. A maximum scour depth of 10 m was
considered in the model and the difference in frequency between zero scour and this
maximum value is shown in Table 5-2. The results indicate that a scour depth of 10 m
produced a change in fundamental frequency of ≈ 40 % for the three soil stiffness profiles
considered.
Table 5-2 Eigenvalue analysis of the scour effect

Scour depth (m) Frequency (loose sand) Frequency (medium Frequency (dense sand)
(Hz) dense sand) (Hz) (Hz)
0m 1.5643 1.6481 1.7357
10 m (full) 0.9386 0.9772 1.017
% Difference -39.99% -40.708% -41.4%
Once the expected shift in frequency due to scour was established, the next step was to
determine the optimum points on the structure to record accelerations to give the best
opportunity to capture the first mode of vibration of the integral bridge. By plotting the
first mode shape of the structure for zero scour and full (10 m) scour, it is possible to
obtain a pictorial view of the element locations showing the highest modal displacements
for the fundamental vibration mode. Figure 5-5 shows that the first mode shape for both
zero scour and maximum scour (10 m) is a global sway mode. The data shown in Figure 5-
5 pertains to the analysis performed in loose sand. However, the shape was the same for all
three soil stiffness profiles considered. From the figure, it can be seen that the maximum
modal amplitude occurred at the top of the bridge pier and this location was chosen as the
point to measure acceleration as it assists in identifying the frequency when using signal
processing and also aids with signal to noise ratio (SNR) issues.
111
(a) (b)
5 5
Zero scour mode 1 Full scour mode 1
0 0
-5 -5
Y-coordinate
Y-coordinate
Ground line 10 m
-10 -10 scour
-15 -15
Ground line
-20 -20
0 10 20 30 40 50 0 10 20 30 40 50
X-coordinate X-coordinate
Figure 5-5 Fundamental mode shapes in loose sand – global sway. (a) Zero scour, (b) Full
scour
5.3.2 Response of structure to moving half-car vehicle model
5.3.2.1 Simulation of noise free pier accelerations due to the passage of a vehicle
From the eigenvalue analysis in the previous section, it was observed that significant
reductions in natural frequency occurred due to scour of the central pile foundation system.
However, the fact that frequency changes will occur is of little use if the relevant mode is
not excited in the structure. The most practical way to excite a rail/highway bridge is to use
ambient traffic (Farrar et al. 1999). Therefore, in this section the aim is to ascertain if it is
possible to detect these frequency changes by analysing the bridge acceleration response to
a moving sprung vehicle. In this analysis, accelerations generated at a lateral degree of
freedom near the top of the bridge pier are analysed using a fast Fourier transform to
obtain the frequency content. The vehicle modelled is a 15 tonne two-axle truck, and to
make the model as realistic as possible interaction between the vehicle and the bridge is
allowed for. The bridge is excited by the sudden arrival of the vehicle on the bridge deck,
which effectively acts as an impact load.
112
The vehicle is a four-degree-of-freedom system that moves along the bridge deck. The
vehicle is excited by the presence of a road profile which causes the body to pitch and
bounce and this in turn means that the forces that the vehicle applies to the bridge are not
constant. The vehicle commences movement at an approach distance of 100 m from the
start of the bridge so that the initial vehicle motion conditions (axle displacements and
body displacement/pitch) when the vehicle meets the bridge are more realistic. The road
profile used in this study is a Class ‘A’ profile, see Cebon (1999), and the part of the road
profile on the bridge is reproduced in Figure 5-6 for the bridge section. (Note: a Class ‘A’
profile corresponds to a well maintained road surface).
-3
x 10
6
2
Height (m)
-2
-4
Class 'A' road profile
-6
0 5 10 15 20 25 30 35 40 45 50
Distance (m)
Figure 5-6 Road profile on the bridge
The vertical forces generated by the vehicle moving over the road profile are shown in
Figure 5-7(a) for a vehicle speed of 80 km hr-1, the loose sand profile and the case of zero
scour. In Figure 5-7(a) it can be seen that the rear axle is significantly heavier than the
front axle, which is typical of a fully loaded 2 axle truck.
Figure 5-7(b) shows the lateral acceleration response of the top of the pier when the truck
crosses the bridge (for the loose sand profile). The large peaks in acceleration at 0 seconds
and 2.5 seconds correspond to the vehicle entering and leaving the bridge. After the
vehicle leaves the bridge there is a logarithmic decay in the acceleration signal over the
113
following 27.5 seconds. This is to be expected as a damping ratio of 2% is used in the

simulations.
4
x 10 (a)
10
Force (N)
5
Axle 1
Axle 2
0
0 0.5 1 1.5 2 2.5
Time (s)
(b)
Acceleration (m s )
Pier acceleration
-2
0.01
-0.01
0 5 10 15 20 25 30
Time (s)
Figure 5-7 Results for vehicle crossing bridge for zero scour level and loose sand profile.
(a) Axle contact forces (N), (b) Lateral acceleration response at top of pier (m s-2)
5.3.2.2 The effect of noise on determining the frequency of the pier vibrations
Real data will contain noise, so in this study, noise was added to the simulated signal. In
order to check if the (scour detection) method was sensitive to the level of noise in the
signal, signals with different levels of noise are analysed. The method used to add noise is
based on the signal-to-noise ratio (SNR), given in Eq. (5.3) (Lyons 2011).
Signal Power
SNR  10 log10 (Eq.5.3)
Noise Power
where SNR is the ratio of the strength of a signal carrying information equating to that of
unwanted interference. Eq. (5.3) is rearranged to give Eq. (5.4).
114
Signal Power
 N  Noise Power  (Eq.5.4)
 SNRlog e 10 
exp  
 10 
where  N is the noise variance. Using Eq. (5.4), noisy signals with different signal-to-
noise ratios were added to the original clean signal. This process is shown in Eq. (5.5).
Sig NOISE   N rand   Sig CLEAN (Eq.5.5)
In this study, three noise levels were examined, namely SNRs of 20, 10 and 5. Figures 5-
8(a-c) show the result of adding noise to the signal shown in Figure 5-7(b). Figure 5-8(d)
shows the frequency content of the signals in Figure 5-8(a-c). It can be seen in the figure
that for all levels of noise the frequency plot is practically identical which proves that the
method will not be particularly sensitive to noise. For the purpose of completeness, the
figure has an insert which shows a zoomed in view of the frequency peak. In the insert it
can be seen that that there are small differences in the magnitude of the frequency peak for
the different levels of noise. However, in relative terms these differences are insignificant.
Since noise does not impede the ability of the method to detect the frequency accurately,
all analyses from this point will contain a SNR = 20 as it is easier for the reader to interpret
the time domain plots in Figures 5-9(a) and 5-10(a) for lower values of noise.
115
(a) (b)
Acceleration (m s )
-2
0.01 SNR = 20 0.01 SNR = 10
0 0
-0.01 -0.01
0 10 20 30 0 10 20 30
Time (s) -3
Time (s)
(c) x 10 (d) SNR = 20
6
0.01 SNR = 5 SNR = 10
Magnitude
4 SNR = 5
0
2
-0.01 1.56 1.58
0
0 10 20 30 1 1.2 1.4 1.6 1.8 2
Figure 5-8 Sensitivity of frequency content to noise. (a) Signal from bridge pier with SNR
= 20, (b) Signal with SNR = 10, (c) Signal with SNR = 5, (d) Frequency content of signals
shown in Figs. 8(a)-(c) (Hz)
5.3.2.3 Identifying the presence of scour by analysing pier acceleration signals
Figure 5-9(a) shows the lateral acceleration signal measured at the top of the bridge pier
due to the passing vehicle for the three soil stiffness profiles considered for the case of
zero scour. The three signals in Figure 5-9(a) are difficult to distinguish so Figure 5-9(b)
shows only the first 10 seconds of data. Figure 5-9(c) shows the frequency content of the
signals shown in Figure 5-9(a). From this figure, it is clear that it is possible to detect the
first natural frequency of the bridge (which is lateral sway) for each of the soil stiffness
profiles modelled. The difference in magnitude between each frequency peak is due to the
relative stiffness of the soil impeding the lateral sway motion. The loose sand profile
allows more movement than the dense sand profile (due to the difference in spring
stiffness), hence a higher peak was observed for the loose sand.
116
(a)
loose sand
0.01 medium-dense sand
Acc (m s )
-2
dense sand
0
-0.01
0 5 10 15 20 25 30 35 40
Time (s)
(b)
loose sand
0.01 medium-dense sand
Acc (m s )
dense sand
-2
-0.01
0 1 2 3 4 5 6 7 8 9 10
Time (s)
-3 (c)
x 10 loose sand
6
medium-dense sand
dense sand
Magnitude
0
1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2
Frequency (Hz)
Figure 5-9 Bridge response due to passing vehicle and subsequent free vibration. (a)
Acceleration response from bridge pier for loose, medium-dense and dense sand profiles
with 40 seconds of free vibration (m s-2), (b) Acceleration response from bridge pier for
loose, medium-dense and dense sand profiles with 7.5 seconds of free vibration (m s-2), (c)
Frequency response of signals shown in (a) (Hz)
Figure 5-9 demonstrates that the natural frequency of mode 1 can be accurately determined
by analysing the lateral acceleration response of the pier with a Fourier transform. The
next step is to induce scour in the analysis and observe the change in frequency. The
analysis involved running the vehicle over the bridge to generate an acceleration signal at
the top of the bridge pier and adding noise. The signal was then analysed with a fast
Fourier transform to determine the frequency content of the signal.
Scour was induced around the central piled foundation by removing springs iteratively
from the model, this corresponding to an increase in scour depth and a loss of associated
117
soil stiffness. A spring is removed and the vehicle is re-run across the bridge to generate a
new acceleration signal, which is analysed for its frequency content. The process is
repeated until a maximum depth of scour of 10 m was induced around the central pile
foundation.
An example of the analysis is shown in Figure 5-10. The solid and dashed plots in Figure
5-10(a) shows the acceleration signals generated at the top of the bridge pier for the case of
zero scour and the maximum 10 m scour depth respectively (for a loose sand profile). For
ease of visualising the signals, Figure 5-10(b) shows just the first 10 seconds of the pier
acceleration response. On the left hand side of this plot, a total of four impulses in the
acceleration signals (between t = 0 and t = 2.5 s) are visible. This corresponds to the front
and rear axles of the vehicle entering and leaving the bridge. The front axle enters the
bridge at t = 0 s and the rear axle leaves the bridge at t = 2.5 s. Figure 5-10(c) shows the
frequency content of the signals shown in Figure 5-10(a). It can be seen in Figure 5-10(c)
that the natural frequency for zero scour was 1.556 Hz. It can also be seen in Figure 5-
10(c) that the natural frequency at the maximum scour depth of 10 m was 0.9308 Hz.
Therefore, a significant and measureable reduction in natural frequency was observed.
118
(a) zero scour

0.01 10m scour
Acc (m s )
-2
-0.01
0 5 10 15 20 25 30 35 40
Forced (b) Time (s) Free Vibration
Front axle enters bridge zero scour

0.01
Front axle leaves bridge 10m scour
Acc (m s )
-2
0
Rear axle leaves bridge
-0.01
Rear axle enters bridge
0 1 2 3 4 5 6 7 8 9 10
Time (s)
(c)
0.01 zero scour
f 10m scour
Magnitude
0.005
0
0 0.5 1 1.5 2 2.5 3
Frequency (Hz)
Figure 5-10 Effect of 10 m of scour on the pier acceleration response for loose sand
profile. (a) Acceleration response (laterally) at top of bridge pier for zero and 10 m scour
due to passage of vehicle, including 40 seconds of free vibration (m s-2), (b) Acceleration
response of bridge pier with 7.5 seconds of free vibration (m s-2), (c) Frequency content of
signals shown in (a) (Hz)
By repeating the analysis for scour depths ranging from 0.5 m to 10 m, the natural
frequency for each scour depth was determined. The variation in natural frequency with
scour depth for the ‘loose sand’ is shown by the solid plot with circular data markers in
Figure 5-11. Figure 5-11 also shows the change in the natural frequency plotted against the
depth of scour for the ‘medium-dense sand’ and ‘dense sand’ stiffness profiles. It is clear
from this figure that for the three soil stiffness profiles simulated, it was possible to detect
a change in the natural frequency of the bridge due to scour using vehicle induced
vibrations. It is worth noting that the method was not particularly sensitive to soil stiffness
(loose, medium-dense or dense) i.e. for all soil densities considered, there is a clear
reduction in natural frequency with increasing scour. Not surprisingly, the magnitude of
119
the frequency for a given scour depth varies with the soil stiffness. However, the variation
with soil stiffness is significantly less than the variation with scour depth. This basically
implies that the increase in effective length resulting from scour had a much larger effect
on the frequency response of the structure than changes in the stiffness of the soil
supporting the foundation.
-2
Scour depth (m)
-4
-6
Loose sand
-8
Medium dense sand
Dense sand
-10
0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8
Frequency (Hz)
Figure 5-11 Frequency change with scour for all three soil stiffness profiles
5.4 Conclusions
A field-validated numerical model was developed in Chapter 4 which is capable of

tracking the change in the natural frequency of a single pile affected by scour. This model
was extended in this chapter to consider the case of a full bridge subjected to traffic
loading. A novel Vehicle-Bridge-Soil Interaction (VBSI) model was developed to explore
the potential frequency changes due to scour of an integral bridge structure for a range of
soil stiffnesses typically encountered in the field.
In the first instance, it was necessary to establish how scour affects the natural frequency
of the bridge and if the changes in frequency would be sufficiently large to warrant further
exploration of this method as a potential scour monitoring tool. A numerical modal study
was conducted to address this question. The aim of this study was to assess the magnitude
of frequency changes that can be expected for a typical bridge structure subjected to scour
of the central piles. From this study, the expected magnitude of the frequency shift was
120
established and deemed sufficiently large (≈ 40%) to warrant an investigation into the
feasibility of detecting scour by analysing the bridge’s response to a moving vehicle. The
VBSI model was used in the first instance to generate realistic acceleration signals from
the structure due to a vehicle passing at typical highway speeds (80 km hr-1). The
generated lateral acceleration from the top of the bridge pier was then analysed for its
frequency content. Results indicate that for all three soil stiffness profiles modelled (loose,
medium-dense and dense sand) the response signals generated from vehicular loading are
sufficient to allow the changes in natural frequency caused by scour to be detected.
Moreover, the shape of the scour depth vs frequency plot was the same for all three soil
stiffness profiles which shows that the method is not particularly sensitive to the
magnitude of soil stiffness.
The method developed in this chapter shows promise in terms of use as part of an
infrastructure management framework incorporating real-time low maintenance scour
monitoring. The advantage of the method is that it does not require complex underwater
installations and negates the requirement for dangerous diving inspections to monitor
scour. The results indicate that accelerometers fixed to the structure above the waterline
may possibly be used as a continuous scour monitoring solution. Real-time analysis of
signals from a structure of interest could be monitored for frequency changes or signals
could be analysed before and after major flood events to attempt to detect losses of
stiffness caused by scour. Whilst this appears promising, a full-scale application of the
method on a real bridge is recommended as future work.
121
Chapter 6 An investigation into the effect of scour on the natural frequency of an
offshore wind turbine
Authors:
Luke J. Prendergast
Kenneth Gavin
Paul Doherty
Paper Status:
Published in Ocean Engineering 101 (2015) 1-11
Note to Reader:
The experimental analysis and numerical modelling within this paper was undertaken by
the candidate under the supervision of Dr Kenneth Gavin. The model properties were
derived from discussions with Dr Paul Doherty.
122
Chapter 6 –An investigation into the effect of scour on the natural frequency of an offshore
wind turbine
6.1 Introduction
This chapter presents scale model tests and simple finite element numerical modelling to
examine the effects of scour on the natural frequency of a wind turbine supported by a
monopile embedded in typical offshore sand deposits. In the first part of the chapter, an
experiment is performed on a scale model monopile where scour is artificially induced and
the change in the natural frequency is measured. The results of this experiment are used to
calibrate a numerical model of the soil-pile system developed using simple finite-element
methods, similar to the methods presented in section 4.3.2.1. The purpose behind
performing these tests is to examine the effect of pile rigidity on the performance of the
numerical model, as the pile tested in Chapter 4 was quite slender in comparison to piles
used in the offshore industry. In the latter part of this chapter, the numerical model is
extended to include a full-scale offshore wind turbine founded on a monopile using
realistic structural properties corresponding to a 3.6 MW turbine. The numerical model
which is developed in MATLAB is used to consider the relative effects of different soil
stiffness profiles and scour severity on the system’s natural frequency. The potential for
accelerometers to be used as a method of monitoring scour around offshore wind turbines
is inferred from the results obtained. Section 6.2 provides information on scour in the
marine environment.
6.2 Scour around offshore wind turbine foundations
The development of offshore wind energy converters is viewed as one of the most cost
effective and politically acceptable short-term means of reducing society's reliance on
fossil fuels. Since 1991, the majority (over 75%) of offshore wind turbines have been
founded on large diameter steel tubular monopiles (EWEA 2012). These piles with
diameters (Dpile) in the range 4 m to 6 m are typically driven to penetrations (Lembed) of
between 20 - 30 m (Doherty & Gavin 2012). The resulting piles are relatively stiff, with
low slenderness (Lembed/Dpile) ratios. Monopile design is usually performed using spring-
beam (p-y) models which are described in the American Petroleum Institute (API 2007)
and Det Norske Veritas (Det Norske Veritas 2007) design codes. The design should
consider all limit states including the ultimate limit state (ULS). However, because of strict
rotational tolerance specifications, serviceability requirements tend to govern design.
123
wind turbine
Wind turbines are dynamically sensitive structures with the primary excitation forces
arising from the rotor spinning at a given rotational velocity, termed the 1P frequency and
the blade passing frequency, termed NbP, where Nb is the number of blades on the turbine.
LeBlanc et al. (2010) note that typical ranges for the 1P frequency are 0.17 to 0.33 Hz,
whilst for a standard three bladed turbine; 3P is in the range 0.5 to 1 Hz (see Figure 6-1).
Wind and wave frequencies are typically below the 1P frequency, although wave
frequencies are highly variable and can span a relatively wide frequency spectrum, see
Tempel & Molenaar (2002). During the design phase for the turbine support structure, it is
important to minimise the potential dynamic amplification of the load and avoid resonance
by ensuring that the natural system frequency of the turbine and its support system does
not coincide with any of the excitation frequencies discussed above. This is achieved by
tailoring the stiffness of the structural system at the foundation-superstructure interface
through careful design of a sub-structure system. In terms of the foundation, a designer can
choose a very stiff structure (termed stiff-stiff) such as a jacket, leading to a natural system
frequency above the 3P range. Alternatively, a structure with a system frequency between
the 1P and 3P values, termed a soft-stiff structure can be created by using a monopile
foundation. It is possible to design a flexible structure, with a system frequency below 1P,
termed a soft-soft structure. However, these soft-soft design options are susceptible to
resonance from sea waves due to the alternating stresses induced on the structure by the
period of passing waves.
It is widely recognised that offshore piles installed in non-cohesive deposits can be

affected by local scour. Local scour occurs around piles when the near bed shear stresses
exceed the critical shear stress at which sediment starts to move and occurs as a result of
the installed pile changing the local waterflow characteristics (Sørensen and Ibsen 2013).
The DNV design codes recommend that a local scour hole depth of 1.3 pile diameters
(1.3Dpile) should be considered for both the ultimate and serviceability limit state design
cases in the case of a monopile subjected to current only-induced scour (Det Norske
Veritas 2007). In the marine environment however, it must be recognised that these
structures can be jointly subjected to currents, tides and waves which makes the problem
more complex than that of scour around structures in rivers (which are generally under
steady current conditions) (Negro et al. 2014). It is also noteworthy that natural variation
in seabed level over a large area can give rise to a further global scour effect, which could
124
wind turbine
have an effect over an entire wind farm site leading to more severe scour depths. Under the
action of waves, the expression for the equilibrium scour depth (Seq) around a tubular
monopile as recommended by the DNV is shown in Eq. (6.1) (Det Norske Veritas 2011;
Sørensen and Ibsen 2013). This expression was derived by Sumer et al. (1992) based on
several small-scale tests.
Seq  1.3Dpile1  exp  0.03KC  6, KC  6 (Eq.6.1)
where KC is the Keulegan-Carpenter number, a dimensionless parameter that is a function

of the maximum horizontal particle velocity at still water level (umax), the intrinsic period
of the waves (Ti) and the monopile diameter (Dpile). The equilibrium scour depth increases
as KC values increase and tends toward the current-only equilibrium scour depth for large
KC values. Sumer & Fredsøe (2001) proposed a modification to Eq. (6.1) for the case of a
pile subjected to combined steady current and waves. Their modification introduces a
dimensionless parameter that describes whether currents or waves are dominating. In
general, the design scour depth will be large when currents are dominating and small when
waves are dominating (Sørensen and Ibsen 2013).
Uncertainties in the magnitude and variability of the various parameters governing the
scour process can lead to difficulties in the accurate estimation of a design scour depth for
a given foundation design. Moreover, Matutano et al. (2013) have compiled a list of
different formulae used in the prediction of maximum scour depths which have been
derived assuming different flow conditions (current-only, wave-only or combined current
and waves). They have highlighted that different standards and recommendations used in
the offshore industry suggest different formulations for scour depth estimation, leading to
natural variability in design scour depths. In a detailed study, they compared the maximum
design scour depths obtained by the implementation of various design formulae with the
actual measured scour depths for various wind farm sites around Europe. This analysis
revealed that the maximum observed scour depth was less than the estimated value in all
but two of the cases considered, which further highlights the uncertainties present in
accurate scour depth estimation. Furthermore, Negro et al. (2014) acknowledges the
requirement that for effective scour protection design, it is necessary to include sediment
properties, geotechnical characteristics of the site, environmental parameters for wave
125
wind turbine
loads and turbine foundation specifications in order to accurately predict the maximum
scour that could occur in the absence of such protection. They also call into question the
DNV’s (Det Norske Veritas 2013) recommendation for scour characterization around
offshore wind turbines under combined current and waves, deeming it to be inaccurate.
In the absence of adequate scour protection, scour has the effect in design of increasing the
cantilever length (or effective water depth) by between the order of 5.2 and 7.8 m for the
piles sizes currently used in practice, under the assumption of current-only induced scour.
Scour has the effect of reducing the ultimate capacity of a foundation and reducing the
system stiffness (Prendergast & Gavin 2014). This could, in certain circumstances, lead to
the occurrence of resonant vibrations in the structure if the structure’s natural system
frequency aligns with any of the excitation frequencies from the rotating blades or the
passing sea waves. Sørensen and Ibsen (2013) present a desk study investigation on the
potential effects of scour on an existing offshore wind farm at Horns Rev 1. Using
frequencies obtained from estimates of soil stiffness derived from p-y curves given in the
API code, they found that the first natural frequency of a structure reduced by 5% when
the scour depth reached 1.3Dpile. To counteract the effects of scour, mitigation measures
which typically comprise placing rock armour at the sea bed, have been undertaken at a
number of sites. These can have the effect of stiffening the foundation response and hence,
increasing the system’s natural frequency. For a soft-stiff structure, such measures have
the potential to cause dynamic load effects with excitation frequencies above the
structure’s natural frequency.
Recent efforts to measure the natural frequency of offshore structures have revealed that
the stiffness of heavily over-consolidated soils may be underestimated when using design
values recommended in current design codes (API 2007; Det Norske Veritas 2007). For
heavy offshore oil and gas platforms, under-estimating the soil stiffness does not have
major implications for safety. However when considering light, dynamically sensitive
offshore wind turbines it is crucial to accurately predict the structure’s system frequency.
Even more importantly, it is crucial to demonstrate how changes arising during the lifetime
of the structure due to soil stiffness degradation as a result of cyclic loading, or from scour
may affect the dynamic response.
126
wind turbine
Soft-Soft Soft-Stiff Stiff-Stiff
Amplitude
0.07 - 0.14 0.17 - 0.33 0.5 - 1.0 Frequency (Hz)

Wave Range 1P Range 3P Range
Figure 6-1 Excitation frequency ranges for typical offshore wind turbines
6.3 Scale-model scour test
A scale-model experiment was performed by driving a 340 mm diameter open-ended steel

pile to a final penetration depth of 2.2 m in very dense sand at the University College
Dublin (UCD) test site at Blessington, County Wicklow, Ireland. The pile diameter (Dpile)
of this experimental monopile is scaled in the range ≈ 1:12 to 1:18 of those currently used
in industry as offshore wind turbine foundations. The embedded depth (Lembed) is scaled in
the range ≈ 1:7 to 1:14 of typical embedment depths for offshore piles (Sørensen and Ibsen
2013). The ratio of embedded depth to pile diameter (Lembed/Dpile) was of the order of 6.5,
which is similar to values used in the offshore wind sector.
The site characteristics at the UCD test site are described in detail in section 3.5.1. For the
present study, we are interested in the shear wave velocity measurements and Cone
Penetration Test (CPT) data, therefore this data is reproduced in this section for
convenience. Ten CPT tip resistance qc profiles were measured at the site with the average
and maximum and minimum envelopes as shown in Figure 6-2(a), which reveal qc values
that range from 10 to 20 MPa over a depth of 8 m bgl. The shear wave velocity profile
measured using the Multi-Channel Analysis of Surface Waves (MASW) method (Donohue
et al. 2004), is shown in Figure 6-2(b). The sand density and bedding conditions at the site
are typical of those found in offshore environments.
127
wind turbine
Cone tip resistance qc (MPa) Shear wave velocity (m s -1 )

0 10 20 30 40 50 150 200 250 300
0 0
Shear wave
Max qc velocity
-1 -1
-2 -2
-3 -3
Depth (m)
Depth (m)
-4 -4
-5 -5
Min qc
-6 -6
-7 -7
Average qc
-8 -8
(a) (b)
Figure 6-2 Blessington site investigation. (a) Cone tip resistance profile (MPa), (b) Shear
wave velocity profile (m s-1)
An experimental scour test was performed whereby sand was removed in increments from
around the monopile and the natural frequency of the system was obtained at each scour
depth. The test pile was fitted with accelerometers (see Figure 6-3) which were wired into
a Campbell Scientific CR9000X datalogger and monitored at a scanning frequency of
1000 Hz. Three accelerometers were fitted before the start of the test, whilst a fourth (the
lowest) was added when the scour test had reached a depth of 0.8 m bgl. A modal hammer
was used to produce an initial lateral excitation at the top of the pile. The hammer was
calibrated to excite low frequency resonances in the structure (heavy tip mass coupled with
a soft impact tip) so the fundamental mode of vibration could be more easily obtained.
Scour was then induced in stages by removing 0.2 m deep soil layers. The impact test was
repeated a minimum of five times at each scour depth, with the modal hammer hitting the
same location on the pile during each test. In total, nine scour depths were considered, the
final scour depth being 1.8 m or 5.3Dpile bgl.
128
wind turbine
0.075 m
Accelerometer
0.3 0.3 0.3

Reference accelerometer
0.8 m
R0
0.2 m
.15
Initial Level
6m
Blessington
Scour depth -1
.17m
sand Scour depth -2 R0
Scour depth -3
3m
Scour depth -4
1.8 m
Scour depth -5
2.2 m
Scour depth -6 Section A-A

Monopile Scour depth -7
Scour depth -8
Scour depth -9
A A
Figure 6-3 Test layout schematic (all dimensions in metres)
The frequency response of the monopile structure at each scour depth was obtained using
two methods. In the first approach, Frequency Response Functions (FRFs) were developed
which show the ratio of the measured output to the input stimulus. In this instance, the
input stimulus is the time history of the forces applied over the impact duration and the
output response is the acceleration measured by each accelerometer. Individual response
functions were obtained from each input-output pair to ensure that the frequency obtained
by each was broadly in agreement. The second approach used to obtain the frequency
response of the structure involved using Frequency Domain Decomposition (FDD), a tool
which utilises the signals from all the accelerometers simultaneously to obtain an estimate
of the modal properties of the system (Brincker et al. 2001). This procedure is an
improvement on the classical Fourier transform approach as it is less sensitive to frequency
resolution problems and allows for easier estimation of close modes and damping. In
overview, the FDD process involves taking the Singular Value Decomposition (SVD) of
the spectral matrix and decomposing it into a set of auto spectral density functions that
each correspond to a single degree of freedom (SDOF) system (Brincker et al. 2000). It is
generally most effective when the type of input excitation is broad-banded (white noise).
129
wind turbine
In this case, due to the application of an impulse to excite the system, the decomposition
into SDOF systems was more approximate (it is exact for white noise excitation and light
damping) but the results are still more accurate than those obtained using classical
approaches (Brincker et al. 2000). The FDD results are used in this analysis as these results
were generally much clearer than the FRFs over the range of scour depths tested.
(a) (b)
15000 10
Acc 1
Acceleration (g)
5
10000
Force (N)
Peak impulse force ~ 12,000 N

0
5000
-5
0 -10
0 0.05 0.1 0.15 0.2 0.25 0.3 0 0.05 0.1 0.15 0.2 0.25 0.3
Time (s) -3 Time (s)
(c) x 10 (d)
0.06 f
Singular Value (SV)
Singular Value (SV)
SV1 scour D-6 SV1 scour depth -6

1 SV1 scour depth -7
0.04
0.5
0.02
0 0
0 100 200 300 400 500 10 20 30 40 50 60 70 80
Frequency (Hz) Frequency (Hz)
Figure 6-4 Experimental results for scour depth -6. (a) Input force time history (N), (b)
Output accelerations for scour depth -6 (g), (c) Frequency spectrum from FDD (Hz), (d)
Isolated first natural frequency for scour depth -6 and scour depth -7 (Hz)
Typical results for scour depth -6 are shown in Figure 6-4. Figure 6-4(a) shows the input
force time history, measured using the modal hammer. Figure 6-4(b) shows the
acceleration response measured by the top accelerometer arising due to the application of
the load shown in Figure 6-4(a). In total there are four acceleration signals, however, three
are omitted from the figure for clarity. The four acceleration signals are used in the FDD
analysis to obtain an estimate of the frequency response of the structure, for the given
scour depth. Figure 6-4(c) shows the dominant frequency content obtained from the
analysis of the four acceleration signals. A dominant peak is seen at approximately 305
Hz, which is too high to be the first natural frequency of the system. In order to isolate the
frequency corresponding to the first natural frequency of the structural system, it is
necessary to compare the system response to an analytical model. In this case, it is best to
obtain an upper-bound frequency threshold by comparing the system response to that of a
130
wind turbine
cantilever, with a free length (L) equal to the portion of the pile above the ground line (at a
given scour depth). The first natural frequency of a cantilever is obtained using Eq. (4.1).
In Figure 6-4(d) the frequency content of the spectrum shown in Figure 6-4(c) is displayed
up to a value of 80 Hz (the theoretical first natural frequency for a cantilever with a free
length corresponding to scour depth -6 is 81.5 Hz). From this plot, it can be seen that the
measured first natural frequency is 21.3 Hz, for scour depth -6. The change in first natural
frequency between scour depth -6 and scour depth -7 is also shown in this plot. The
dominant high frequency peak observed in Figure 6-4(c) arises due to localised vibration
in the annular cross-section as a result of the application of the impulse load. Similar
studies of this phenomenon are reported in Dezi et al. (2012) and Prendergast et al. (2013).
This phenomenon can be mitigated somewhat by omitting a portion of the acceleration
signal directly after the application of the impact force from the frequency analysis.
-1
-2
S/Dpile
-3
-4
-5 +/-  freq,expt
Experimental results
-6
0 10 20 30 40 50 60
Frequency (Hz)
Figure 6-5 Frequency change with dimensionless scour depth
The change in first natural frequency obtained as the scour depth around the pile increased
is shown in Figure 6-5 for all scour depths tested. The depth of scour is presented in terms
of the dimensionless scour depth to pile diameter ratio (S/Dpile). At each scour depth, a
number of impacts (minimum of five) were undertaken to ensure repeatability of the
results. It is noteworthy that at each scour depth, the frequency was observed to reduce
very slightly over the number of impacts, most probably arising as a result of the strain-
dependency of soil stiffness with the stiffness reducing due to the repeated application of
131
wind turbine
the heavy modal hammer. As a result, the average values for each scour depth are reported
with the standard deviation (σfreq,expt) shown by the error bars on the plot at each scour
depth. It is also noteworthy that for the upper two scour depths, the determination of
distinct natural frequency peaks was difficult. This problem arose due to the difficulty in
exciting a clear bending vibration mode in the pile due to the high bending stiffness of the
pile system when it was close to fully embedded. The application of the hammer to the pile
for these two scour depths had the effect of predominately exciting the annular vibration,
as discussed above, which clouded the frequency spectra adversely. Selection of the most-
likely frequency values for the upper two scour depths from the response spectra was aided
by following the trend of frequency change with scour starting with the lowest results (see
Figure 6-4(d) for example of lower scour depth results), which was predominately linear.
Therefore, there is some uncertainty in the frequency results presented for the top two
depths. From the overall results presented in Figure 6-5, however, it is evident that scour
has a noticeable effect on the frequency response of this scaled monopile foundation with a
reduction in measured frequency of between ≈ 57.6 Hz for a S/Dpile of zero to ≈ 7.4 Hz for
a S/Dpile of 5.3.
The scale-model test described above was performed in dry sand, i.e. no hydrostatic effects
(due to water depth at a turbine location) or excess pore pressures which could develop
during cyclic loading, were considered explicitly. In terms of accounting for hydrostatic
water pressure in the offshore environment, the actual water depth has no effect on the soil
response, i.e. if the saturated unit weight of the soil is constant, then the effective stress at
any depth below sea bed level is independent of the water depth. With respect to the
potential development of excess pore water pressures during cyclic loading, its effect
would be to gradually reduce the effective stress of the soil. The soil strength and stiffness
are controlled by the soil’s effective stress; therefore any reduction in effective stress will
cause a reduction in soil stiffness. Since the natural frequency is directly related to the soil
stiffness, in addition to reductions caused by scour, the natural frequency of the structure
could reduce as a result of cyclic pore pressure build-up. The development of excess pore
pressures will be controlled by the loading frequency, soil permeability, and pile diameter
amongst other factors.
132
wind turbine
6.4 Numerical modelling of scaled-model test
A one-dimensional (1-D) finite element numerical model of the soil-structure system was
developed in MATLAB to model the effect of scour on the structural vibration response of
the experimental monopile. The purpose of this numerical study is to replicate the
experimental measurements in the numerical environment such that further sensitivity
studies may be performed using realistic properties for a full-scale wind turbine structure.
6.4.1 Structural model
The soil-structure model consists of Euler-Bernoulli beam elements supported by lateral

linear springs, and is created using the same procedure as that described in section 3.3. The
model properties were obtained from the experimental geometries and material properties
discussed in section 6.3. The exposed end of the monopile was modelled using standard
four degree of freedom (4-DOF) beam elements, with elemental stiffness and mass
matrices recommended by Kwon & Bang (2000). The embedded portion of the pile was
modelled using modified 5-DOF spring-beam elements. The extra degree of freedom was
required to model the lateral soil spring. This spring-beam type model, known as the
Winkler hypothesis has been shown to provide a reasonable representation of the complex
soil-structure interaction problem (Dutta & Roy 2002). The elements are assembled
together into global mass and stiffness matrices. The undamped system natural frequencies
and mode shapes are obtained from the model by solving the Eigen problem shown in Eq.
(6.2) (Tedesco et al. 1999).
Z  IA  0 (Eq.6.2)
where I  is the Identity matrix and Z  I is a square matrix of order n and is called
the characteristic matrix of Z .  is the eigenvalue and A is the associated
eigenvector. Z is the system matrix formed in Eq. (3.9). The eigenvalues and
eigenvectors of Z can be obtained as the roots of the characteristic equation, obtained by
expansion of Eq. (6.3).
Z  I  0 (Eq.6.3)
133
wind turbine
The system matrix Z was specified in MATLAB for the soil-structure system and the
eigenvalues and eigenvectors (natural frequencies and mode shapes) were obtained using
in-built functions available in the MATLAB programming environment.
Scour was modelled as the removal of springs from the numerical model. An eigenvalue
analysis was undertaken for each spring removal phase and a profile of frequency against
scour was generated. The numerical model of the soil-structure system is shown
schematically in Figure 6-6.
0.8 m
7 Beam Elements
kS,N
R0
Blessington m
.17
.15
20 Spring-Beam Elements
sand
R0
6
m
2.2 m
Section A-A
kS,2
A A kS,1
Figure 6-6 Schematic of numerical model of soil-structure system
6.4.2 Sand spring stiffness
In the numerical modelling of this problem, the largest uncertainty with respect to input
parameters relates to the spring stiffness coefficients that are used to model the in-situ soil
stiffness. In spite of the slight deviation in observed frequency at each scour depth due to
the repeated application of the impulse load, it can generally be assumed that the dynamic
loading applied to establish the natural frequency of the structure induced very small
displacements. Therefore, the small-strain stiffness of the soil can be used as a fair
approximation to model soil displacements. Shear wave velocity (vs) measurements can be
134
wind turbine
used to determine the small-strain shear modulus (G0) of the soil if the soil density  is
known using Eq. (3.13). For sites where vs data is not available correlations with CPT qc
values are commonly adopted in practice (Schnaid et al. 2004). The ratio G0/qc, known as
the rigidity index is known to depend on soil density, current stress level and loading
history, degree of cementation and age of the deposit amongst other factors (Fahey et al.
2003; Lunne et al. 1997). For the aged, over-consolidated sand at the Blessington test site
where the experiment was performed, a G0/qc ratio in the range 5 to 8 would be expected
(see Figure 6-2 for site properties). A ratio of 6, which was adopted in the analysis
described herein, was seen to provide a reasonable estimate of the small-strain shear
modulus inferred from the MASW measurements of shear wave velocity at the site (see
Figure 6-7).
The derived G0 profiles were used to form discrete spring stiffness coefficients (ksi) for use
in the dynamic analysis in the 1-D finite element model. This procedure is discussed in
detail in section 4.3.2.1, and is summarised herein. In the first instance, the shear modulus
profile was converted to an equivalent Young’s modulus (E0) profile using an appropriate
small-strain Poisson’s ratio ( = 0.1) using Eq. (3.12). The material properties of the soil
and pile are coupled using Eq. (4.2). This expression allows a profile of the modulus of
subgrade reaction (K) to be determined for the solution to the problem relating to a beam
on an elastic foundation (Ashford & Juirnarongrit 2003). The G0 profiles derived from
both the shear wave velocity and CPT measurements used in the analysis are shown in
Figure 6-7 over the depth of embedment of the experimental monopile.
A total of 20 springs spaced at 0.111 m intervals were used to model the initial un-scoured
state as shown in Figure 6-6. Individual spring coefficients (ksi) are calculated by
multiplying the K value at a given spring depth by the spacing between subsequent springs
(z) at that depth. Scour is modelled by sequentially removing springs from the model,
corresponding to a loss in soil support, and performing an eigenvalue analysis at each
depth to obtain the natural frequencies of the system (see section 6.4.1).
135
wind turbine
0
G0 from shear wave
G0 from CPT qc
-0.5
Depth (m)
-1
-1.5
-2
0 20 40 60 80 100 120
Shear modulus (G0 ) (MPa)
Figure 6-7 Profile of shear modulus (G0) developed using vs measurements and CPT qc
data (MPa)
6.5 Results of experimental and numerical modelling of scaled-model test
The frequency values for a range of scour depths were obtained as the eigenvalues of the
system matrix after removal of each spring used to simulate scour (see Figure 6-6). The
results of the numerical analyses for all scour depths using G0 profiles derived from both
shear wave velocity measurements and CPT qc values are compared to the measured
experimental response in Figure 6-8. This figure shows the change in natural frequency of
the monopile due to progressive scour with the numerical predictions also shown, plotted
against the dimensionless S/Dpile.
136
wind turbine
-1
-2
S/Dpile
-3
-4 +/-  freq,expt
Experimental results
-5 Shear wave numerical model
CPT numerical model
-6
0 10 20 30 40 50 60 70 80
Frequency (Hz)
Figure 6-8 Results of numerical and experimental investigation - frequency change with
scour (Hz)
From the results shown in Figure 6-8, it is clear that the model provides a good match to
the measured experimental response using both the shear wave velocity and CPT data to
estimate the soil spring stiffness. It is worth noting that the experimental response at the
uppermost scour depths deviates away from the numerical predictions. As was stated in
section 6.3, there was a degree of uncertainty associated with obtaining a clear frequency
peak for these depths. This uncertainty arose due to the difficulty in inducing a clear
bending vibration mode with the impulse hammer as a result of the high system stiffness
when the pile was fully embedded. Overall however, the results are satisfactory and
moreover, they indicate that once a reasonable estimate of G0 can be made, the simple 1-D
model presented can be used to quantify the effect of scour on the natural frequency of a
scaled monopile system.
In practice, the presence of a structure, e.g. a wind turbine tower, nacelle etc., will have a
significant influence on the frequency response of the system. In section 6.6, a turbine-
foundation numerical model is developed to explore this interaction.
6.6 Design example – wind turbine model
A numerical model of a full-scale offshore wind turbine was developed using MATLAB
by employing the same spring-beam modelling philosophy as outlined in section 6.4.1. In
addition to considering scale effects, the purpose of this model was to include the
137
wind turbine
structural mass, realistic loads and to investigate the effect of sand density or initial
stiffness on the change in frequency induced by scour. The wind turbine model consisted
of a nacelle supported on an annular tower, which was founded on a monopile embedded
in sand. The turbine was founded on a 6 m diameter monopile with a wall thickness of 90
mm embedded 30 m into sand (see Figure 6-9). Details of the model properties are
outlined in Table 6-1.
Table 6-1 Offshore wind turbine properties

Tower properties: Value: Monopile properties: Value:
Tower length (m) 70 Monopile length (m) 75
Material Steel Embedded length (m) 30
Density (kg m-3) 7850 Material Steel
Young’s modulus (MPa) 210,000 Density (kg m-3) 7850
Tower diameter (m) 5 – 3.5 Young’s modulus (MPa) 200,000
Tower wall thickness (m) 0.045 Monopile diameter (m) 6
X-sectional area (Ait) (m2) 0.7005-0.4884 Monopile wall thickness (m) 0.09
Moment of inertia (Iit) 2.15-0.7289 X-sectional area (Am) (m2) 1.67

4
(m )
Nacelle mass (MTop) (kg) 230,000 Moment of inertia (Im) (m4) 7.2974
Nacelle rotational inertia 2.21 x 107 Mass of power unit at 27,000

transverse direction (J) interface level (MTransition)
(kg)
(kg m2)
Nacelle rotational inertia 3.5 x 107

fore-aft direction (J) (kg
m2)
The tower was modelled using 140 standard Euler-Bernoulli 4-DOF beam elements
(Ntower), each element being 0.5 m in length giving a total tower height of 70 m. The
monopile was modelled using a total of 150 elements (Npile), giving a total monopile length
of 75 m. Of these 150 monopile elements, the embedded portion was modelled using
modified 5-DOF (Nsprings) elements to account for the extra DOF associated with the soil
spring. The remaining Npile – Nsprings elements were modelled using standard 4-DOF beam
138
wind turbine
elements, similar to the tower, to model the exposed portion of the monopile. The
embedded portion of the monopile was initially modelled using a total of 60 Nsprings
elements, to model the un-scoured state, giving an initial embedded depth of 30 m. A
water depth of 30 m was included in the model. This water depth is in the approximate
range of depths where monopile foundations are typically used (Sørensen and Ibsen 2013).
The effect of the water depth is to increase the cantilevered length of the monopile. No
fluid-structure interaction was modelled between the monopile and the surrounding water,
although it is recognised that the presence of water may alter the dynamic response. The
nacelle was modelled as a lumped mass at the top node of the tower (MTop) with a
rotational inertia (J) as specified in Table 6-1 and includes the contribution from the mass
of the rotor and the blades. The discretized modelling approach can give a very good
approximation of the natural frequencies of the system and removes the need for more
computationally intensive finite element analysis of the continuous system (Murtagh et al.
2004). The gyroscopic effect of blade rotation on the frequency of the turbine was not
modelled. The transition piece between the monopile and the tower was modelled as a
lumped mass applied to the top node of the monopile (MTransition). A fixed connection was
considered between the turbine tower and the monopile. In total, 290 elements were used
to define the model from the base of the monopile to the lumped mass at the tower top,
each 0.5 m in length.
In order to check for model compliance, the model was fixed at the tower base and the
natural frequencies obtained for a range of tower diameters and thicknesses were
compared to the following approximation for a beam with a mass at its tip shown in Eq.
(6.4), (Tempel & Molenaar 2002).
1
 3.04 EI 2
f1   2 
 4 M  0.227 L free L free 
3
(Eq.6.4)
where μ = tower mass per metre (kg m-1), M = tip mass (kg) and Lfree is the free
cantilevered length (m). Excellent agreement between the clamped numerical model and
the approximations from Eq. (6.4) were observed, with deviations typically being of the
order of 1%.
139
wind turbine
Nacelle & blades
MTop, J
ANt, I Nt
Ntower = 140
Tower
Tapered
70 m
m
R2.91
R3 m
Monopile A1t, I 1t MTransition
15 m
Section A-A
Nsprings = 60
Mean sea level
Npile = 150
30 m
A A
Am, I m
75 m
Ks,Ns
Mud line
30 m
Ks,1
Figure 6-9 Wind turbine model and numerical schematic
The procedure for developing soil stiffness profiles for the analysis of the offshore wind
turbine is detailed in section 3.4.1, with a brief synopsis provided herein. To consider the
range of sands typically encountered in the offshore environment, uniform soil deposits 30
m deep with relative density (Dr) values of 30%, 50% and 80% were postulated, to
approximate loose, medium-dense and dense sand conditions. Synthetic CPT profiles (see
Figure 6-10) were derived using Eq. (3.10). The vertical effective stress was calculated by
assuming bulk unit weights of 16, 18 and 20 kN m-3, corresponding to loose, medium
dense and very dense sand and a bulk unit weight of water (γw) of 10 kN m-3. G0 profiles
140
wind turbine
were derived from the CPT qc values using Eq. (3.11). The resultant G0/qc values for the
soil profiles assumed ranged from 14-23 for loose sand and from 4-8 for dense sand
profiles. These G0 profiles were converted to profiles of modulus of subgrade reaction (K)
and then to individual spring moduli (ksi) using the procedure outlined in section 4.3.2.1.
(a) (b)
0 0
Loose qc Loose G0
Medium dense qc Medium dense G0
-5 Very dense qc -5 Very dense G0
-10 -10
Depth (m)
Depth (m)
-15 -15
-20 -20
-25 -25
-30 -30
0 10 20 30 40 0 50 100 150 200
Cone tip resistance qc (MPa) G0 profile (MPa)
Figure 6-10 Synthetic site profiles. (a) Generated qc profiles (MPa), (b) Generated G0
profiles for loose, medium dense and very dense sand deposits (MPa)
In order to investigate the potential changes in natural frequency of the turbine structure
arising due to scour, a reasonable estimate of a design scour depth is required in the first
instance. In this particular example, the current-only design value of 1.3Dpile is considered.
Since scour depth is generally small when waves are dominating and large when currents
are dominating (Sørensen and Ibsen 2013), the current-only design value is a reasonable
datum for potential extreme scour affecting the offshore turbine. The results of the
analyses on the full turbine model are shown in Figure 6-11(a), with scour being induced
up to and beyond the current-only design scour depth of 1.3Dpile. This figure shows the
141
wind turbine
change in natural frequency of the full turbine structure plotted against the dimensionless
scour depth ratio (S/Dpile) for all three soil stiffness estimates. For a monopile with no
scour, the natural frequency of the full system varied from 0.2913 Hz in loose sand, to
0.3056 Hz in very dense sand. This can be compared to a 1P frequency of 0.22 Hz for a
typical 3.6 MW turbine and a 3P frequency of 0.66 Hz. The structure is safely within the
soft-stiff range. As scour proceeded beyond 1.3Dpile to 1.66Dpile, or an absolute scour depth
of 10 m, the natural frequency of the structure reduced to 0.2548 Hz for the loose sand and
0.2771 Hz for very dense sand, i.e. safely above the 1P frequency level. The rate at which
the frequency reduced with scour for different soil densities is considered in Figure 6-
11(b). This figure shows that the loose sand model exhibited relatively greater reductions
in natural frequency than the medium-dense or very dense sand models with a maximum
reduction of ≈ 12.5% at S/Dpile = 1.66 (S = 10 m) or ≈ 8.5% at the design scour depth ratio
of 1.3 (S = 7.8 m). This is larger than the differences reported from studies. Sørensen and
Ibsen (2013) reported relative reductions of approximately 5% for scour over 1.3Dpile.
However, in their study, the authors adopted the API recommended stiffness profiles for
sand which increase linearly with depth. The G0 profiles adopted in the present study, see
Figure 6-10(b), are arguably more realistic for offshore soil conditions showing stiffness’s
for a given sand density which vary non-linearly with the mean stress level. The result of
this is that the soil stiffness near the sea bed level is relatively higher, and as a result, the
loss of material due to scour has a more significant effect on the frequency response of the
turbine using the soil profiles adopted in this study.
142
wind turbine
0
Loose sand freq
Design scour depth (1.3Dpile) = 7.8 m
Med dense sand freq
-0.5
Very dense sand freq
S / Dpile
1P frequency
-1
-1.5
0.21 0.22 0.23 0.24 0.25 0.26 0.27 0.28 0.29 0.3 0.31
Frequency (Hz)
0
Loose sand F
Design scour depth (1.3Dpile) = 7.8 m
-0.5 Med dense sand F
Very dense sand F
S / Dpile
-1
-1.5
0.86 0.88 0.9 0.92 0.94 0.96 0.98 1

F = F/FNo scour
Figure 6-11 Effect of scour on the system frequency of an offshore wind turbine. (a) Effect
of scour with loose, medium dense and very dense sand profiles (Hz), (b) Relative effect of
scour using each stiffness profile
In the analysis described above, continuous scour was assumed to occur over the full
design depth and beyond without any backfilling. In reality, scour will be time-dependant,
with the action of currents and waves creating oscillations in the equilibrium scour depth
and will typically be less than the current-only design value presented above with the
action of waves generally leading to a lower local scour depth. By assuming constant
removal of material, the system frequency can be observed to drop significantly as the
support due to the soil is removed. It is noteworthy that when backfilling occurs, the
stiffness of the in-fill material may be similar to that of the removed material. One
experiment undertaken by Sørensen et al. (2010) measured the relative density of
backfilled material to be 65-80%, which corresponds to dense sand. However, the relative
density was only determined for one backfill experiment so further research would be
required to make a definitive conclusion in this regard. Also, the fact that there will be a
lag in time between the scouring and backfilling process means that for a given period of
time, the natural frequency of the structure may be significantly reduced. Waves
compacting sediment near the mudline can lead to high relative density in sub-surface
sands, therefore it can be expected that backfilled material may become more dense with
143
wind turbine
time (Augustesen et al. 2009). This means that the natural frequency of the wind turbine
structure could vary considerably as scour and backfilling occurs. It is also noteworthy
that, in many cases, the actual scour depth occurring around a turbine foundation at a wind
farm site may be less than the design value so there is still some uncertainty present in the
prediction of an equilibrium scour depth (Matutano et al. 2013).
The change in mode shape exhibited by the full wind turbine structure for the loose sand
case is shown in Figure 6-12. The mode shapes were derived from the eigenvectors of the
system as discussed in section 6.4.1. Figure 6-12(a) shows the first mode shape for the
structure founded in the loose sand deposit prior to scouring. Figure 6-12(b) shows the un-
scoured mode shape in black with dashed line style and the scoured mode shape in grey
with solid line style. The scour depth shown corresponds to the current-only design scour
depth of 1.3Dpile, which in this case is 7.8 m absolute scour. The results for the medium
dense and very dense sand are not shown but they exhibit a similar mode shape for the
change in soil support level. This figure provides a visualisation of the effect of scour on
the natural frequency response of the wind turbine structure.
On a real full-scale monopile foundation in the offshore environment, the scour process
takes place under water therefore it is important to acknowledge the potential effect that
the presence of water around the pile could have on the natural frequency of the system. In
the case of using accelerometers to detect and monitor scour via changes in frequency, it is
imperative to establish whether the presence of water would have any significant effect on
the perceived stiffness of the system. A laboratory investigation performed by Prendergast
et al. (2013) aimed to highlight the changes in measured natural frequency of vibration
between cantilever structures of varying cross-sectional stiffness vibrating in air and
immersed in water (see Appendix D). Three different cantilever structures, with stiffness
index (EI/Lfree3) values ranging from 229 N m-1 to 418362 N m-1 were tested. The research
concluded that the stiff structure (highest EI/Lfree3 value) seemed to exhibit relatively little
change in measured natural frequency between air and water compared to the flexible
structure (lowest EI/Lfree3), which showed more sensitivity to the presence of water. With
respect to the current analysis, by taking a conservative Lfree value of 75 m equal to the full
length of the numerical monopile, this gives a EI/Lfree3 value of 3.46E+6 N m-1, which is an
order of magnitude higher than the stiffest experimental section tested by Prendergast et al.
144
wind turbine
(2013). In reality, due to the partial embedment of the pile, the Lfree value will be lower
leading to an even higher EI/Lfree3 value than that reported here. Therefore, it is assumed
that the water-added mass effect will be negligible for stiff monopiles. It must be noted,
however, that there is still some uncertainty with regard to the water-added mass effect on
structures and this is an area of ongoing research (Sedlar et al. 2011). Due to the difficulty
in quantifying water-added mass and the numerical complexity involved (Ju 2013), its
effect on the natural frequency of vibration has been ignored in the present study.
The dynamic movements estimated from the analysis presented above have a total range of
0.3056 Hz in the un-scoured dense sand deposit to 0.2548 Hz in the scoured loose sand
deposit, which indicates a relatively slow cyclic loading rate at the soil-pile interface.
Since the analysis presented here pertains to sand deposits, drained conditions are assumed
to govern the soil-structure interaction (Lehane 1992; Chow 1997; Gavin et al. 2013).
Therefore, the potential for the build-up of pore pressures in the soil as a result of cyclic
loading is ignored in the present study. However, it is noteworthy that this is an area of
ongoing uncertainty and may not be a true representation of the actual complex soil-
structure interaction (LeBlanc et al. 2010). It should be noted that the potential build-up of
pore pressures would lead to a gradual reduction in the soil’s effective stress, and
moreover since the soil strength and stiffness are controlled by this effective stress, a
reduction in soil stiffness could occur. This, in tandem with the scour effect, could further
lower the natural frequency of the system. Further studies are required to quantify potential
pore water effects that have not been considered here.
145
wind turbine
150 150
Loose sand S/Dpile = 0 Loose sand S/Dpile = 0
Loose sand S/Dpile ~ 1.3
Distance from pile tip (m)
Distance from pile tip (m)

100 100
Mean sea level Mean sea level
50 50
Mudline Mudline
 scour
0 0
-0.05 0 0.05 0.1 0.15 -0.05 0 0.05 0.1 0.15
Modal displacement (-) Modal displacement (-)
Figure 6-12 Change in mode shape in loose sand stiffness profile over design scour ratio of
1.3 for full structure
6.7 Conclusions
An experimental programme to examine the effect of scour on a scale-model monopile

was undertaken at a dense sand test site. Scour was found to have a noticeable effect on the
natural frequency response of this foundation structure with a measured reduction in first
natural frequency from ≈ 57.59 Hz to 7.42 Hz for 1.8 m or 5.3Dpile of (extreme) scour.
This extreme scour was performed for the purpose of validating numerical finite-element
models of the soil-pile system capable of tracking the change in the natural frequency of
the system due to scour. These simple 1-D spring-beam numerical models were developed
in MATLAB where the soil spring stiffness was modelled using small-strain stiffness (G0)
data obtained from in-situ tests such as MASW and CPT tests.
A numerical model of a full offshore wind turbine and monopile foundation was developed
using the same spring-beam modelling philosophy as that used to model the experiment.
The purpose of this model was to investigate the effect of scour on the natural system
frequency of a full-scale structure for a range of soil densities typically encountered in the
offshore environment. The nacelle and rotor system was modelled as a lumped mass at the
146
wind turbine
top of the tower. Typical tower and monopile geometric and material properties were used
in the numerical model which corresponded approximately to a 3.6 MW offshore turbine.
Three different soil stiffness profiles were developed corresponding to loose, medium
dense and very dense sand. The effect of scour on the natural frequency of the system was
obtained by removing springs sequentially from the model and performing an eigenvalue
analysis for each removal stage. The loose sand stiffness profile demonstrated the largest
relative change in frequency response with an overall reduction in natural frequency of the
order of 8.5% for a design scour depth of 1.3Dpile compared to approximately 6.5% for
very dense sand, suggesting that resonance effects are most likely in loose deposits.
Limitations in the numerical model include the fact that the gyroscopic effects of blade
rotation have not been encompassed, which would have some effect on the dynamics of
the system. Also, fluid-structure interaction effects around the exposed monopile and the
potential build-up of pore pressures in the foundation soil due to the cyclic loading have
not been accounted for in the numerical study. However, the model is still a reasonable
approximation for the analysis of offshore turbines and moreover, the results suggest that
the frequency response framework adopted provides a viable means of real-time remote
monitoring of scour.
147
Chapter 7 Conclusions
7.1 Summary
This thesis is presented in the format of individual publishable chapters, each in effect a
standalone entity. The individual conclusions of each chapter are presented at the end of
the relevant chapter. This chapter summarises the conclusions in terms of the overall thesis
aims.
The literature survey conducted in Chapter 2 provides information on various types of

scour monitoring sensors that have been developed to date. More importantly, however,
this chapter introduces the concept of monitoring/detecting scour using changes in the
modal properties of a structure. Several works are presented that aim to use various
structural dynamic characteristics to detect the presence of scour around bridge
foundations. Some of these works focus on merely detecting the presence of scour whereas
others focus on measuring the depth of a scour hole using changes in modal information.
Several key questions come to light from this research. (1) Is it possible to correlate a
depth of scour to a change in the natural frequency of a structure? (2) How sensitive are
the natural frequencies of an integral bridge to foundation scour? (3) Can changes in
frequency be detected using the acceleration signals generated on a bridge due to a passing
vehicle? (4) In relation to the offshore environment, what effect does scour have on the
natural frequency of an offshore wind turbine and is resonance a potential issue?
To address these questions, it is recognised that there is a requirement to develop an

accurate foundation model that can track the changes in the natural frequency of a
structure with scour. This model should allow for correlating a frequency value measured
on a structure of interest to a depth of scour around the foundation. Since soil is highly
heterogeneous, it is important that this model be capable of encapsulating the stiffness
response of a given soil stratum. Therefore, allowing for the model to be calibrated using
measurable geotechnical data is an important aspect. Chapters 3 and 4 of this thesis are
concerned with developing and validating a numerical pile-soil dynamic interaction model
capable of tracking the changes in the natural frequency of a single pile with scour. This
148
Chapter 7 –Conclusions
model incorporates soil stiffness using a Winkler spring-beam philosophy whereby the
spring stiffnesses are specified from small-strain stiffness parameters for the soil, which
can be obtained from correlations to shear wave velocity measurements and cone
penetration test data. The model is validated against laboratory and field tests.
The pile-soil dynamic interaction model developed in Chapters 3 and 4 is used as the
foundation model for a two-span integral bridge developed to explore how sensitive the
bridge natural frequency is to scour (Chapter 5). This model utilises the same modelling
philosophy as used to develop the foundation model. In order to make the model as
realistic as possible, it is loaded by a two-axle truck and incorporates vehicle-bridge
interaction effects, whereby the axle loads applied to the bridge vary according to the
roughness of the road surface on the bridge. This model is used to explore if changes in
natural frequency due to foundation scour can be detected by analysing the bridge
acceleration signals due to the passage of this truck model. The main findings of this
chapter are that the first natural frequency of the integral bridge model changes by ≈ 40%
for 10 m of (extreme) scour affecting the central pier foundation and it is possible to track
these frequency changes by analysing the lateral pier top acceleration due to a passing
truck.
Finally, the pile-soil dynamic interaction model developed in Chapters 3 and 4 is applied
to the case of modelling a full 3.6 MW offshore wind turbine founded on a monopile in
Chapter 6. The purpose of this chapter is to evaluate how scour affects the natural
frequency of an offshore wind turbine and to ascertain if resonance between the structure’s
natural frequency and the excitation frequencies of the rotating blades could become an
issue. The main finding of this chapter is that offshore wind turbines founded in loose sand
deposits experience relatively larger reductions in natural frequency than those founded in
dense sand deposits. However, even for extreme scour affecting the foundation, the
structure remains safely above the rotating blade nominal excitation frequency for a
standard 3.6 MW turbine.
7.2 Recommendations for further research
This thesis has explored the feasibility of using changes in natural frequency as an
indicator of scour around bridge foundations and offshore wind turbines. All of the
149
Chapter 7 –Conclusions
structures investigated have piled foundations. A range of experimental trials and

numerical modelling was undertaken as part of the research and several areas of potential
future research have come to the fore from these trials. Some of these are discussed herein.
The application of the method to a full-scale real bridge would be a significant benefit to
forwarding the research in terms of creating a viable and usable method to detect and
monitor scour using changes in natural frequency. Although it would be prohibitively
costly to scour a real bridge for investigative purposes, a bridge that is due to undergo
significant foundation retrofit may be useful as a pilot project to test the theories put
forward in this thesis. Some collaboration between potential researchers and infrastructure
managers would be very useful in this regard.
Since a lot of bridges and offshore structures alike have shallow foundations, further
investigation into the potential for the methods developed in this thesis to be used in this
context would be very useful. It would be necessary to observe what effect scour would
have on the natural frequency of a structure employing a shallow foundation and to
ascertain if it is possible to detect scour using frequency shifts as put forward in this thesis.
The soil modelling in this thesis remained in the small-strain region with the linear-elastic
simplification being employed. The development of a non-linear large strain dynamic
modelling framework would be a very useful addition to the works put forward in this
thesis. In particular, it would allow a more realistic treatment of the offshore monopile
case, where the dynamic movements may frequently enter the large-strain region. The
development of a non-linear model requires the specification of many inter-related
parameters therefore there is potential for significant further research in this area.
The research undertaken in this thesis used massless soil springs and concentrated on the
first mode of vibration in each case. Extending the work to consider higher order modes as
well as the contribution of the mass of soil in the analyses would be of interest for further
research.
150
References
Allotey, N. & El Naggar, M.H., 2008. Generalized dynamic Winkler model for nonlinear
soil–structure interaction analysis. Canadian Geotechnical Journal, 45(4), pp.560–
573.
Anderson, N.L., Ismael, A.M. & Thitimakorn, T., 2007. Ground-Penetrating Radar : A
Tool for Monitoring Bridge Scour. Environmental & Engineering Geoscience,
XIII(1), pp.1–10.
API, 2007. RP2A: Recommended practice for planning, designing and constructing
offshore platforms - Working stress design, Washington, DC.
Arneson, L.A., Zevenbergen, L.W., Lagasse, P.F., Clopper, P.E., 2012. HEC-18
Evaluating Scour at Bridges.
Ashford, S.A. & Juirnarongrit, T., 2003. Evaluation of Pile Diameter Effect on Initial
Modulus of Subgrade Reaction. Geotechnical and Geoenvironmental Engineering,
129(3), pp.234–242.
Atkinson, J., 2000. Non-linear Soil Stiffness in Routine Design. Géotechnique, 50(5),
pp.487–508.
Augustesen, A.H., Brødbæk, K.T., Møller, M., Sørensen, S.P.H., Ibsen, L.B., Pedersen,
T.S., Andersen, L., 2009. Numerical Modelling of Large-Diameter Steel Piles at
Horns Rev. In Proceedings of the Twelfth International Conference on Civil,
Structural and Environmental Engineering Computing. pp. 1–14.
Augustesen, A.H., Møller, M., Brødbæk, K.T., Sørensen, S.P.H., 2012. Review of laterally
loaded monopiles employed as the foundation for offshore wind turbines.
Avent, R.R. & Alawady, M., 2005. Bridge Scour and Substructure Deterioration : Case
Study. Journal Of Bridge Engineering, 10(3), pp.247–254.
Badiane, D., Gasser, A. & Blond, E., 2012. Vibrating beam in viscous fluid for viscosity
sensing - Application to an industrial vibrating viscometer. In Proceedings of the 12th
Pan-American Congress of Applied Mechanics. Port of Spain, Trinidad.
Baldi, G., Bellotti, R., Ghionna, V.N., Jamiolkowski, M., Lo Presti, D.C.F., 1989.
Modulus of Sands from CPTs and DMTs. In Proceedings of the 12th ICSMFE Vol. 1.
pp. 165–170.
Biot, M.A., 1937. Bending of an infinite beam on an elastic foundation. Journal of Applied
Mechanics, 59, pp.A1–A7.
151
References
Bolduc, L.C., Gardoni, P. & Briaud, J.L., 2008. Probability of Exceedance Estimates for
Scour Depth around Bridge Piers. Journal of Geotechnical and Geoenvironmental
Engineering, 134(2), pp.175–185.
Boulanger, R.W., Curras, C.J., Kutter, B.L., Wilson, D.W., Abghari, A., 1999. Seismic
soil-pile-structure interaction experiments and analysis. Journal of Geotechnical and
Geoenvironmental Engineering, 125(9), pp.750–759.
Bowles, J.E., 1988. Foundation analysis and design 4th ed., New York, 1004: McGraw-
Hill.
Brandimarte, L., Montanari, A., Briaud, J.L., D’Odorico, P., 2006. Stochastic Flow
Analysis for Predicting River Scour of Cohesive Soils. Journal of Hydraulic
Engineering, 132(5), pp.493-500.
Briaud, J.L., Chen, H.C., Ting, F.C.K., Cao, Y., Han, S.W., Kwak, K.W., 2001. Erosion
Function Apparatus for Scour Rate Predictions. Journal of Geotechnical and
Geoenvironmental Engineering, pp.105–113.
Briaud, J.L., Hurlebaus, S., Chang, K., Yao, C., Sharma, H., Yu, O., Darby, C., Hunt, B.E.,
Price, G.R., 2011. Realtime monitoring of bridge scour using remote monitoring
technology, Austin, TX.
Briaud, J.L., Chen, H., Li, Y., Nurtjahyo, P., Wang, J., 2005. SRICOS-EFA Method for
Contraction Scour in Fine-Grained Soils. Journal of Geotechnical and
Geoenvironmental Engineering, 131(10), pp.1283–1295.
Briaud, J.L., Ting, F. & Chen, H.C., 1999. SRICOS: Prediction of Scour Rate in Cohesive
Soils at Bridge Piers. Journal of Geotechnical and Geoenvironmental Engineering,
(April), pp.237–246.
Brincker, R., Zhang, L. & Andersen, P., 2000. Modal Identification from Ambient
Responses using Frequency Domain Decomposition. In Proceedings of 18th
International Modal Analysis Conference.
Brincker, R., Zhang, L. & Andersen, P., 2001. Modal identification of output-only systems
using frequency domain decomposition. Smart Materials and Structures, 10(3),
pp.441–445.
Cantero, D., Gonzalez, A. & O’Brien, E.J., 2011. Comparison of bridge dynamic
amplification due to articulated 5-axle trucks and large cranes. Baltic Journal of Road
and Bridge Engineering, 6(1), pp.39–47.
Cebon, D., 1999. Handbook of Vehicle-Road Interaction, Netherlands: Swets & Zeitlinger.
Chen, C.C., Wu, W.H., Shih, F., Wang, S.W., 2014. Scour evaluation for foundation of a
cable-stayed bridge based on ambient vibration measurements of superstructure. NDT
& E International, 66, pp.16–27.
152
References
Chopra, A.K., 1981. Dynamics of Structures. A Primer, Earthquake Engineering Research

Institute.
Chow, F.C., 1997. Investigations into displacement pile behaviour for offshore
foundations. University of London, Imperial College.
Concast, 2014. Concast Precast Group. Civil Engineering Solutions, p.7. Available at:
http://www.concastprecast.co.uk/images/uploads/brochures/Concast_Civil.pdf
[Accessed May 1, 2014].
Daloglu, A.T. & Vallabhan, C.V.G., 2000. Values of k for Slab on Winkler Foundation.
Journal of Geotechnical and Geoenvironmental Engineering, 126(5), pp.463–471.
Det Norske Veritas, 2007. DNV Offshore Standard DNV-OS-J101 Design of Offshore
Wind Turbine Structures.
Dezi, F., Gara, F. & Roia, D., 2012. Dynamic response of a near-shore pile to lateral
impact load. Soil Dynamics and Earthquake Engineering, 40, pp.34–47.
Doebling, S. & Farrar, C., 1996. Damage identification and health monitoring of
structural and mechanical systems from changes in their vibration characteristics: a
literature review.
Doherty, P., Kirwan, L., Gavin, K., Igoe, D., Tyrrell, S., Ward, D., O'Kelly, B., 2012. Soil
Properties at the UCD Geotechnical Research Site at Blessington. In Proceedings of
the Bridge and Concrete Research in Ireland Conference. Dublin, Ireland.
Doherty, P. & Gavin, K., 2012. Laterally loaded monopile design for offshore wind farms.
Proceedings of the ICE - Energy, 165(1), pp.7–17.
Donohue, S., Long, M., Gavin, K., O'Connor, P., 2004. Shear Wave Stiffness of Irish
Glacial Till. In International Conference of Site Characterisation I. Porto, Portugal,
pp. 459–466.
Dukkipati, R.V., 2009. Matlab for Mechanical Engineers, New Age Science.
Dutta, S.C. & Roy, R., 2002. A critical review on idealization and modeling for interaction
among soil–foundation–structure system. Computers & Structures, 80(20-21),
pp.1579–1594.
Elachachi, S.M., Breysse, D. & Houy, L., 2004. Longitudinal variability of soils and
structural response of sewer networks. Computers and Geotechnics, 31(8), pp.625–
641.
153
References
Elsaid, A., 2012. Vibration Based Damage Detection of Scour in Coastal Bridges. North
Carolina State University.
Elsaid, A. & Seracino, R., 2014. Rapid assessment of foundation scour using the dynamic
features of bridge superstructure. Construction and Building Materials, 50, pp.42–49.
Ergin, A. & Uğurlu, B., 2003. Linear vibration analysis of cantilever plates partially
submerged in fluid. Journal of Fluids and Structures, 17(7), pp.927–939.
Esmailzadeh, M., Lakis, A., Thomas, M., Marcouiller, L., 2008. Three-dimensional
modeling of curved structures containing and/or submerged in fluid. Finite Elements
in Analysis and Design, 44(6-7), pp.334–345.
EWEA, 2012. The European offshore wind industry key 2011 trends and statistics.
Fahey, M., 1992. Shear modulus of cohesionless soil: variation with stress and strain level.
Canadian Geotechnical Journal.
Fahey, M., Lehane, B.M. & Stewart, D., 2003. Soil Stiffness for shallow foundation design
in Perth CBD. Australian Geotechnics Journal, 38(3), pp.61–90.
De Falco, F. & Mele, R., 2002. The monitoring of bridges for scour by sonar and
sedimetri. NDT&E International, 35, pp.117–123.
Farrar, C.R., Duffey, T.A., Cornwell, P.J., Doebling, S.W., 1999. Excitation methods for
bridge structures. In Proceedings of the 17th International Modal Analysis
Conference Kissimmee. Kissimmee, FL.
Federico, F., Silvagni, G. & Volpi, F., 2003. Scour Vulnerability of River Bridge Piers.
Journal of Geotechnical and Geoenvironmental Engineering, 129(10), pp.890–900.
Fisher, M., Chowdhury, M.N., Khan, A., Atamturktur, S., 2013. An evaluation of scour
measurement devices. Flow Measurement and Instrumentation, 33, pp.55–67.
Forde, M.C., McCann, D.M., Clark, M.R., Broughton, K.J., Fenning, P.J., Brown, A.,
1999. Radar measurement of bridge scour. NDT&E International, 32, pp.481–492.
Foti, S. & Sabia, D., 2011. Influence of Foundation Scour on the Dynamic Response of an
Existing Bridge. Journal Of Bridge Engineering, 16(2), pp.295–304.
Fryba, L., 1999. Vibration of solids and structures under moving loads, London: Thomas
Telford.
Fugro, 2011. Guide for Estimating Soil Type and Characteristics using Cone Penetration
Testing. Cone Penetration Tests. Available at: http://www.fes.co.uk/downloads/CPT-
general.pdf [Accessed May 12, 2014].
Gavin, K., Adekunte, A. & O’Kelly, B., 2009. A field investigation of vertical footing
response on sand. Proceedings of the ICE, Geotechnical Engineering, 162(GE5),
pp.257–267.
154
References
Gavin, K.G., Kirwan, L. & Igoe, D., 2013. The effect of ageing on the axial capacity of
piles in sand. Proceedings of the ICE - Geotechnical Engineering, 166(GE2), pp.122–
130.
Gavin, K.G. & Lehane, B.M., 2007. Base Load-Displacement Response of Piles in Sand.
Canadian Geotechnical Journal, 44(9), pp.1053–1063.
Gavin, K.G. & O’Kelly, B.C., 2007. Effect of friction fatigue on pile capacity in dense
sand. Journal of Geotechnical and Geoenvironmental Engineering, 133(1), pp.63–71.
Gavin, K.G. & Tolooiyan, A., 2012. An Investigation of correlation factors linking footing
resistance on sand with cone penetration results. Computers and Geotechnics,
46(November), pp.84–92.
González, A., Vehicle-Bridge Dynamic Interaction Using Finite Element Modelling. In

Finite-Element Analysis. pp. 637–662.
González, A. & Hester, D., 2013. An investigation into the acceleration response of a
damaged beam-type structure to a moving force. Journal of Sound and Vibration,
332(13), pp.3201–3217.
Green, F. & Cebon, D., 1997. Dynamic interaction between heavy vehicles and highway
bridges. Computers and Structures, 62(2), pp.253–264.
Gutenbrunner, G., Savov, K. & Wenzel, H., 2007. Sensitivity Studies on Damping
Estimation. In Proceedings of the Second International Conference on Experimental
Vibration Analysis for Civil Engineering Structures (EVACES). Porto, Portugal.
Hamill, L., 1999. Bridge Hydraulics, London: E.& F.N. Spon.
Heidarpour, M., Afzalimehr, H. & Izadinia, E., 2010. Reduction of local scour around
bridge pier groups using collars. International Journal of Sediment Research, 25(4),
pp.411–422.
Hester, D. & González, A., 2012. A wavelet-based damage detection algorithm based on
bridge acceleration response to a vehicle. Mechanical Systems and Signal Processing,
28, pp.145–166.
Igoe, D., Gavin, K. & O’Kelly, B., 2011. The shaft capacity of pipe piles in sand. Journal
of Geotechnical and Geoenvironmental Engineering, 137(10), pp.903–912.
Jardine, R.J., Chow, F.C., Overy, R.F., Standing, J., 2005. ICP Design Methods for Driven
Piles in Sands and Clays, London.
Johnson, P.A. & Dock, D.A., 1998. Probabilistic Bridge Scour Estimates. Journal of
Hydraulic Engineering, 124(7), p.750.
Ju, S.H., 2013. Determination of scoured bridge natural frequencies with soil–structure
interaction. Soil Dynamics and Earthquake Engineering, 55, pp.247–254.
155
References
Juang, J.N., 1994. Applied system identification, Englewood Cliifs, NJ: Prentice-Hall.
Kallehave, D., Thilsted, C.L. & Liingaard, M.A., 2012. Modification of the API P-Y
Formulation of Initial Stiffness of Sand. In Proceedings of the 7th International
Conference on Offshore Site Investigations and Geotechnics. London, UK, pp. 465–
472.
Kamojjala, S., Gattu, N.P., Parola, A.C., Hagerty, D.J., 1994. Analysis of 1993 Upper
Mississippi flood highway infrastructure damage. In Proceedings of the 1st
International Conference of Water Resource Engineering. New York, NY: ASCE, pp.
1061–1065.
Kampitsis, A.E., Sapountzakis, E.J., Giannakos, S.K., Gerolymos, N.A., 2013. Seismic
soil–pile–structure kinematic and inertial interaction—A new beam approach. Soil
Dynamics and Earthquake Engineering, 55, pp.211–224.
Kwon, Y.W. & Bang, H., 2000. The Finite Element Method using MATLAB, Boca Raton,
FL: CRC Press, Inc.
Lagasse, P.F., Schall, J.D., Johnson, F., Richardson, E.V., Chang, F., 1995. Stream
stability at highway structures, Washington, DC.
LeBlanc, C., Houlsby, G.T. & Byrne, B.W., 2010. Response of stiff piles in sand to long-
term cyclic lateral loading. Géotechnique, 60(2), pp.79–90.
Lehane, B.M., 1992. Experimental investigations of displacement pile behaviour using

instrumented field piles. University of London, Imperial College.
Lin, Y.B., Lai, J.S., Chang, K.C., Li, L.S., 2006. Flood scour monitoring system using
fiber Bragg grating sensors. Smart Materials and Structures, 15(6), pp.1950–1959.
Lin, Y.B., Lai, J.S., Chang, K.C., Chang, W.Y., Lee, F.Z., 2010. Using mems sensors in
the bridge scour monitoring system. Journal of the Chinese Institute of Engineers,
33(1), pp.25–35.
Love, A.E.H., 1927. The mathemathical theory of elasticity, University Press, Cambridge.
Lunne, T. & Christopherson, H.P., 1983. Interpretation of Cone Penetrometer Data for
Offshore Sands. In Offshore Technology Conference OTC4464. Houston, Texas.
Lunne, T., Robertson, P.K. & Powell, J.J.M., 1997. Cone Penetration Testing in
Geotechnical Practice, Blackie Academic and Professional.
Lyons, R., 2011. Understanding digital signal processing 3rd Edition., Boston, MA:
Prentice Hall.
El Madany, M., 1988. Design and optimization of truck suspensions using covariance
analysis. Computers & structures, 28(2), pp.241-246.
156
References
Matlock, H., 1970. Correlations for design of laterally loaded piles in soft clay. In Offshore
Technology in Civil Engineering. pp. 577 – 594.
Matutano, C., Negro, V., López-Gutiérrez, J.S., Esteban, M.D., 2013. Scour prediction and
scour protections in offshore wind farms. Renewable Energy, 57, pp.358–365.
May, R.W.P., Ackers, J.C. & Kirby, A.M., 2002. Manual on scour at bridges and other
hydraulic structures, London.
Melville, B.W. & Coleman, S.E., 2000. Bridge scour, Highlands Ranch, CO: Water
Resources Publications.
Murchison, J.M. & O’Neill, M.W., 1984. Evaluation of p-y relationships in cohesionless
soils. In Analysis and Design of Pile Foundations. Proceedings of a Symposium in
conjunction with the ASCE National Convention. pp. 174–191.
Murtagh, P.J., Basu, B. & Broderick, B.M., 2004. Simple models for natural frequencies
and mode shapes of towers supporting utilities. Computers & Structures, 82(20-21),
pp.1745–1750.
Nassif, H., Ertekin, A.O. & Davis, J., 2002. Evaluation of Bridge Scour Monitoring
Methods, Trenton, NJ.
NCHRP, 2009. Monitoring Scour Critical Bridges - A Synthesis of Highway Practice,

Washington, DC.
Negro, V., López-Gutiérrez, J.S., Esteban, M.D., Matutano, C., 2014. Uncertainties in the
design of support structures and foundations for offshore wind turbines. Renewable
Energy, 63, pp.125–132.
Nogami, T., Otani, J., Konagai, K., Chen, H., 1992. Nonlinear Soil-Pile Interaction Model
for Dynamic Lateral Motion. Journal of Geotechnical Engineering, 118(1), pp.89–
106.
Okeagu, B. & Abdel‐Sayed, G., 1984. Coefficients of Soil Reaction for Buried Flexible
Conduits. Journal of Geotechnical Engineering, 110(7), pp.908–922.
Peder Hyldal Sørensen, S. & Bo Ibsen, L., 2013. Assessment of foundation design for
offshore monopiles unprotected against scour. Ocean Engineering, 63, pp.17–25.
Prendergast, L.J., Hester, D., Gavin, K., O'Sullivan, J.J., 2013. An investigation of the
changes in the natural frequency of a pile affected by scour. Journal of Sound and
Vibration, 332(25), pp.6685–6702.
Prendergast, L.J. & Gavin, K., 2014. A review of bridge scour monitoring techniques.
Journal of Rock Mechanics and Geotechnical Engineering, 6(2), pp.138–149.
Reese, L.C. & Matlock, H., 1956. Non-dimensional Solutions for Laterally Loaded Piles
with Soil Modulus Assumed Proportional to Depth. In Proceedings of the 8th
157
References
International Conference on Soil Mechanics and Foundation Engineering. Austin,

TX, pp. 1–41.
Roy, P.K. & Ganesan, N., 1995. Transient response of a cantilever beam subjected to an
Impulse Load. Journal of Sound and Vibration, 183(5), pp.873–880.
Sadrekarimi, J. & Akbarzad, M., 2009. Comparative Study of Methods of Determination

of Coefficient of Subgrade Reaction. Electronic Journal of Geotechnical
Engineering, 14.
Schnaid, F., Lehane, B.M. & Fahey, M., 2004. In situ test characterisation of unusual
geomaterials. In Proceedings of the International Conference of Site
Characterisation. Porto, Portugal, pp. 49–73.
Sedlar, D., Lozina, Ž. & Vučina, D., 2011. Experimental investigation of the added mass
of the cantilever beam partially submerged in water. Tehnički vjesnik, 18(4), pp.589–
594.
Seika.de, 2012. SEIKA Mikrosystemtechnik GmbH. BDK3 Accelerometer Specifications.

Available at: http://www.seika.de/english/index.htm [Accessed August 22, 2012].
Shabani, R., Hatami, H., Golzar, F. G., Tariverdilo, S., Rezazadeh, G., 2013. Coupled
vibration of a cantilever micro-beam submerged in a bounded incompressible fluid
domain. Acta Mechanica, 224(4), pp.841–850.
Shirole, A.M. & Holt, R.C., 1991. Planning for a comprehensive bridge safety assurance
program. In Transport Research Record. Washington, DC: Transport Research
Board, pp. 137–142.
Sohn, H., Farrar, C.R., Hemez, F., Shunk, D., Stinemates, D., Nadler, B., Czarmecki, J.,
2004. A Review of Structural Health Monitoring Literature : 1996 – 2001.
Sohn, H., 2007. Effects of environmental and operational variability on structural health
monitoring. Philosophical transactions. Series A, Mathematical, physical, and
engineering sciences, 365(1851), pp.539–560.
Sørensen, S.P.H., Ibsen, L.B. & Frigaard, P., 2010. Experimental evaluation of backfill in
scour holes around offshore monopiles. In Proceedings of the second international
symposium on frontiers in offshore geotechnics. Perth, Western Australia, pp. 617–
622.
Sumer, B.M. & Fredsøe, J., 2001. Scour around pile in combined waves and current.
Journal of Hydraulic Engineering, 127, pp.403–411.
Sumer, B.M., Fredsøe, J. & Christiansen, N., 1992. Scour Around Vertical Pile in Waves.
Journal of waterway, port, coastal and ocean engineering, 118(1), pp.15–31.
Tedesco, J.W., McDougal, W.G. & Allen Ross, C., 1999. Structural Dynamics: Theory
and Applications.
158
References
Tempel, J. Van Der & Molenaar, D., 2002. Wind Turbine Structural Dynamics – A
Review of the Principles for Modern Power Generation , Onshore and Offshore. Wind
Engineering, 26(4), pp.211–220.
Tolooiyan, A. & Gavin, K., 2011. Modelling the Cone Penetration Test in Sand Using
Cavity Expansion and Arbitrary Lagrangian Eulerian Finite Element Methods.
Computers and Geotechnics, 38(4), pp.482–490.
Tsai, P., Feng, Z. & Lin, S., 2011. A wavelet based method for estimating the damping
ratio in statnamic pile load tests. Computers and Geotechnics, 38(2), pp.205–216.
Vesic, A.B., 1961. Bending of beams resting on isotropic elastic solid. Journal of Soil
Mechanics and Foundation Engineering, 87, pp.35–53.
W.J. Staszewski, 1998. Identification of non-linear systems using multi-scale ridges and
skeletons of the wavelet transform. Journal of Sound and Vibration, 214(4), pp.639–
658.
Winkler, E., 1867. Theory of elasticity and strength, Dominicus Prague.
Yang, F. & Fonder, G., 1996. An iterative solution method for dynamic response of
bridge–vehicles systems. Earthquake engineering & structural dynamics, 25, pp.195–
215.
Yang, Y., Yau, J. & Wu, Y., 2004. Vehicle-bridge interaction dynamics.
Yankielun, N. & Zabilansky, L., 1999. Laboratory Investigation of Time-Domain

Reflectometry System for Monitoring Bridge Scour. Journal of Hydraulic
Engineering, 125(12), pp.1279–1284.
Yu, X., 2009. Time Domain Reflectometry Automatic Bridge Scour Measurement System:
Principles and Potentials. Structural Health Monitoring, 8(6), pp.463–476.
Zarafshan, A., Iranmanesh, A. & Ansari, F., 2012. Vibration-Based Method and Sensor for
Monitoring of Bridge Scour. Journal Of Bridge Engineering, 17(6), pp.829–838.
159
Appendix A List of journal and conference publications
The information contained within this section is correct at the time of thesis submission.
A.1 Journal publications
An investigation into the effect of scour on the natural frequency of an offshore wind
turbine
Luke J. Prendergast, Kenneth Gavin, Paul Doherty
Ocean Engineering 101 2015, 1-11
A review of bridge scour monitoring techniques

Luke J. Prendergast, Kenneth Gavin
Journal of Rock Mechanics and Geotechnical Engineering 6 (2) 2014, 138-149
An investigation of the changes in the natural frequency of a pile affected by scour

Luke J. Prendergast, David Hester, Kenneth Gavin, John O’Sullivan
Journal of Sound and Vibration 332 (25) 2013, 6685-6702
A.2 Conference publications
Non-intrusive bridge scour analysis technique using laboratory test apparatus

Proceedings of the bridge and concrete research in Ireland conference, Dublin, Ireland,
2012
Luke J. Prendergast, Kenneth Gavin, John O’Sullivan
Monitoring of bridge scour using changes in natural frequency of vibration - A field

investigation
Proceedings of the International Young Geotechnical Engineer’s Conference, Paris,
France, 2013
Luke J. Prendergast
The effect of variations in soil stiffness on the dynamic response of an offshore wind
turbine
Proceedings of the European Wind Energy Association Offshore Conference, Frankfurt,
Germany, 2013
Luke J. Prendergast, Paul Doherty, Kenneth Gavin
Design tools available for monopile engineering

Proceedings of the European Wind Energy Association Conference, Barcelona, Spain,
2014
Paul Doherty, Luke J. Prendergast, Gerry Murphy, Kenneth Gavin
160
Appendix A – List of Journal and Conference Publications
Monitoring of scour critical bridges using changes in the natural frequency of vibration of
foundation piles – a field investigation
Proceedings of the Transport Research Arena, Paris, France, 2014
Bridge scour monitoring using accelerometers placed on bridge piers – a numerical

investigation
Proceedings of the 3rd International Conference on Road and Rail Infrastructure, Split,
Croatia, 2014
Dynamic soil-structure interaction modelling using stiffness derived from in-situ cone
penetration tests
Proceedings of the 3rd International Symposium on Cone Penetration Testing, Las Vegas,
USA, 2014
Luke J. Prendergast, Kenneth Gavin, David Igoe
The effect of scour on the dynamic response of an offshore wind turbine

Proceedings of the civil engineering research in Ireland conference, Belfast, UK, 2014
Luke J. Prendergast, Kenneth Gavin, Paul Doherty
161
Appendix B Dynamic soil-structure interaction modelling using stiffness derived
from in-situ cone penetration tests
Authors:
Luke J. Prendergast
Kenneth Gavin
David Igoe
Paper Status:
Published in the Proceedings of the 3rd International Symposium on Cone Penetration
Testing, Las Vegas, USA, 2014
Note to Reader:
The work presented in this conference paper was undertaken by the candidate under the
supervision of Dr Kenneth Gavin with Dr David Igoe assisting in the experimental phase.
162
Appendix B – Dynamic soil-structure interaction modelling using stiffness derived from
in-situ cone penetration tests
B.1 Introduction
Dynamic soil-structure interaction is a pivotal aspect of the design process for large scale
wind turbine design. These dynamically sensitive structures require accurate soil stiffness
assessments in order to ensure that the design frequency matches the actual operational
frequency when the structures are constructed. The reason for this is that wind turbines,
unlike most large scale civil engineering structures, are subjected to periodic excitation as
part of their operation. This excitation arises due to the rotor spinning at a given rotational
velocity which induces a gyroscopic effect on the structure that is known as the 1P
frequency. In addition to this, the effect of a standard turbine having three blades induces a
further excitation due to the blades passing the tower. This is known as the 3P frequency.
The system stiffness must be such that the natural frequency of the wind turbine does not
lie within the rotor frequency excitation bands, as this may induce resonance which could
lower the design life significantly. There are three design options available. The first
involves designing the system to have a frequency lower than the 1P band, known as the
soft-soft option (Tempel & Molenaar 2002). This is difficult in offshore conditions due to
the presence of waves inducing low frequency resonances in the structure lower than the
1P band, and therefore it is often more suited to onshore turbine design. The second option
involves designing the system so that its natural frequency resides between the 1P and 3P
bands. This is known as the soft-stiff design option and is the most common. The third
option involves designing a very stiff structure that has a system frequency above the 3P
band. This is not so common due to the amount of material required to ensure a high
stiffness. In order to combat our growing energy needs, more and more wind-farms are
being constructed in deeper waters off our coasts. Over 75% of wind turbines have
monopiled foundations (Doherty & Gavin 2012). The design of these monopiles is often
undertaken using American Petroleum Institute (API) design codes from the offshore
engineering industry (API 2007). The natural frequency of a wind turbine is a function of
the material properties used in its construction, and is highly affected by the stiffness of the
soil surrounding the monopile. As such, the accurate assessment of this stiffness in terms
of modeling for the design process is very important in ensuring that the system frequency
can be reliably estimated.
The API method has been shown to be conservative, particularly in the case of stiff piles
(LeBlanc et al. 2010; Doherty & Gavin 2012). The dynamic stiffness is normally taken as
163
the initial slope of the soil reaction – displacement (p-y) curve derived from the API
method (API 2007). This is used in the modelling process by idealizing the pile as a beam
supported by linear springs, known as the Winkler hypothesis (Dutta & Roy 2002). This
idealization is known to yield reasonably good performance and is relatively
straightforward to implement as part of a modelling regime. The stiffness values obtained
from the API method are based on a number of simplified soil properties for the site and
typically only require knowledge of the angle of internal friction (  ' ) and the relative
density (Dr). The stiffness profile with depth is linearly increasing for strata with uniform
 ' and Dr. In reality, the stiffness of the soil for dynamic applications will not be as simple
as those proposed by the API method. Soil is highly heterogeneous from point to point.
The use of site investigations gives us more insight into the variable properties of soil at
different locations in a stratum.
In this paper, a method to obtain more accurate soil stiffness estimations based on Cone
Penetration Test (CPT) data is described and an experimental validation is detailed.
Stiffness profiles are derived from CPT data by linking the small-strain shear modulus (G0)
to the CPT tip resistance (qc). Known as the rigidity index, the variation of G0/qc with qc,1
for a range of sands was investigated by Lunne et al. (1997); Schnaid et al. (2004). The G0
profile that is obtained from this correlation is used to form discrete spring stiffness values
for use in a numerical model of a pile embedded in sand. The research paves the way for
more accurate assessments of soil stiffness at installation locations for wind turbine
foundations.
B.2 Test site
The test bed is a dense sand quarry located in Blessington, south-west of Dublin city,
Ireland. The site conditions comprise of very dense, fine sand, with a relative density
between 90 – 100%. The geotechnical properties of the site are outlined in Table B-1.
Table B-1 Site properties

Property
Soil Type Dense Sand
Fine Content (%) 5% - 10%
Coarse-grained particles (%) < 10%
164
Dr (%) 90% - 100%

Sand D50 (mm) 0.1 mm – 0.15 mm
Degree of saturation (%) 63% - 75%
Equilibrium WT (m BGL) 13 m BGL
-3
Bulk Density (Mg m ) 2.10
Unit Weight (kN m-3) 19.8
’v (°) 37°
’p (°) 54° - 40°
Specific gravity 2.69
emax 0.73
emin 0.37
Natural water content (%) 10% - 12%
Ten CPT tests were performed at the site. The values were relatively consistent, which
reveals a uniform sand deposit where qc increased from ≈ 10 MPa at ground level to ≈ 17
MPa at 2 m below ground level, and increases gradually with depth thereafter. The CPT
profiles are shown in Figure B-1.
0 qc 1
Max qc qc 2
-1
qc 3
-2 qc 4
-3 qc 5
qc 6
-4 Avg qc
Depth (m)
qc 7
-5
qc 8
-6 qc 9
Min qc
qc 10
-7
Min qc Envelope
-8
Max qc Envelope
-9 Avg qc
-10
0 5 10 15 20 25 30 35 40 45
Cone Tip Resistance qc (MPa)
Figure B-1 Ten cone penetration tests for Blessington site with min & max envelopes
165
The average CPT profile from Figure B-1 is used to derive spring stiffness values for a
numerical model of an installed pile at the test location.
B.3 Field installation & vibration test
An open-ended steel pile was driven into the stratum at the test site. The pile had a length
of 8.76 m, a diameter of 0.34 m and an annular thickness of 13 mm. The Young’s modulus
is taken as 2 x 1011 N m-2 and the cross-sectional moment of inertia is 1.91 x 10-4 m4. The
pile was embedded 6.5 m when the vibration test was undertaken. An image of the
embedded pile is shown in Figure B-2.
Figure B-2 Installed pile
The vibration test involved impacting the pile head with a modal hammer that was
calibrated to excite low frequency resonances in the pile. This was achieved by using a
heavy tip mass with a soft impact head on the modal hammer. The resulting lateral
166
vibration was picked up by three accelerometers embedded in the shaft of the pile along
the exposed length. The vibration was transformed from the time-domain acceleration
signal to the frequency domain by applying a Fourier transform in MATLAB. Five impact
tests were performed. The results of the vibration test are shown in Table B-2.
Table B-2 Frequency values from test

st
1 Frequency (Hz) Standard deviation (Hz)
31.09 0.748
B.4 Soil-structure dynamic interaction model
A numerical model is developed that incorporates dynamic soil-structure interaction by

modeling the pile-soil system as a beam on an elastic foundation, commonly referred to as
the Winkler hypothesis (Dutta & Roy 2002). The pile is modeled using Euler-Bernoulli
beam elements, while the soil is modeled by attaching a linear spring to one of each of the
embedded beam nodes. The mass and stiffness matrices for the beam-type elements can be
found in (Tedesco et al. 1999). The numerical model is capable of calculating the natural
frequencies of the soil-structure system, by obtaining the eigenvalues of the system matrix
as defined in the following Eq. (B.1).
Z  M1 K  (Eq.B.1)
where [Z] is the system matrix, [M] is the global mass matrix, and [K] is the global
stiffness matrix for the combined soil-structure model. The eigenvalues were obtained by
using MATLAB’s in-built ‘eig’ function and sorting the entries in descending order, to
obtain the fundamental frequency.
The stiffness component of the soil is obtained by manipulating the CPT qc profile and
discretizing it into individual spring moduli. The first step involves converting the qc
profile into a small-strain shear modulus (G0) profile using the rigidity index, a correlation
between G0 and qc which was undertaken for a range of sands (Lunne et al. 1997; Schnaid
et al. 2004). Dense sand has a rigidity index which varies between 5 and 8. For the purpose
of our analysis, a value of 6 was chosen for each of the springs. Variation of the rigidity
index with qc,1 was not considered in the analysis. The next step involves converting the G0
167
profile to a Young’s modulus (E0) profile using the well-known relation shown in Eq.
(B.2).
E0  2G0 1  v  (Eq.B.2)
where v is the small-strain Poisson ratio. A value of 0.1 was chosen for this parameter.
Once completed, the modulus of subgrade reaction (K) profile can be obtained using a
formula that couples the material properties of the soil and the pile in the elastic continuum
problem. This is available in Ashford & Juirnarongrit (2003) and is shown in Eq. (B.3).
1 / 12
1.0 E0  E0 D 4 
K   (Eq.B.3)
1 v2  E p I p 
where D is the pile diameter (m), Ep is the Young’s modulus of the pile (N m-2) and Ip is
the moment of inertia of the pile cross-section (m4). Individual spring constants are
obtained by multiplying the K value at a given spring depth by the spacing between
subsequent springs. The model is used to perform an eigenvalue analysis and obtain the
natural frequencies corresponding to those of the soil structure system.
Our numerical model contains 30 springs, each spaced at 0.219 m giving a total depth of
embedment of 6.57 m closing approximating the depth of embedment of our field pile.
This is shown in Figure B-3. The average CPT qc resistance from Figure B-1 was used to
obtain spring constants. The results of the eigenvalue analyses are shown in Table B-3.
The stiffness profiles adopted in this paper are shown in Figure B-4 for the CPT and API
springs.
168
10 Beam
Elements
2260
(each 219 mm
long)
30 Sprung-Beam Elements (each 219 mm long)

Blessington
Sand
6500
(a) (b)
Figure B-3 Numerical model schematic (all dimensions are in mm)
Table B-3 Experimental & numerical frequencies

Method: Avg. Experimental Frequency Numerical Frequency % Difference
(Hz) (Hz)
CPT Springs 31.09 28.00 9.93%
API Springs 31.09 22.50 27.6%
169
0
CPT Model Stiffness
-1 API Model Stiffness
-2
Depth (m)
-3
-4
-5
-6
-7
0 2 4 6 8 10 12
-1 7
Spring Stiffness (N m ) x 10
Figure B-4 Numerical stiffness profiles
The results from the eigenvalue analysis show that the CPT generated springs are more
accurate than those developed using the traditional API approach, where the spring
stiffness was obtained as the initial slope of the p-y formula available in API (2007). The
percentage difference lowers from 27.6% to 9.93%. It is worth noting that the spectral
resolution of the experimental signals will lead to some errors in the estimation of the
natural frequency due to problems with signal clarity and signal length. This may account
for some of the 9.93% error noted in Table B-3. Other than this, the CPT based stiffness
derivation is much more accurate than that of the API design approach, as it uses actual
variable in situ soil properties as part of the stiffness derivation. A suggested improvement
may be to use a higher rigidity index value in the transformation from qc to G0.
B.5 Conclusions
An experimental investigation into the natural frequency of a pile embedded in dense sand
was undertaken at a test site in Blessington, Ireland. A numerical model was developed
that models the pile embedded in sand as a beam supported by linear springs. The stiffness
moduli of the springs were obtained using a correlation between the CPT qc profile and the
small strain shear modulus (G0) for the site. Known as the rigidity index, this allows for
the G0 value to be estimated based on in-situ CPT measurements. The G0 profile was
converted to a modulus of subgrade reaction profile and thence to individual spring
170
constants for use in the numerical model. The natural frequencies obtained from the
experiment and the numerical model were compared. For comparison, spring stiffness
values were also generated using the API design code. It was shown that the discrepancy
between the numerical model employing CPT springs and the experiment was 9.93%
whereas the discrepancy between the experiment and the numerical model employing API
springs was 27.6%. Some of the error in the experiment will be due to spectral resolution
when transforming the time domain signals to the frequency domain and arises due to
issues with signal length and noise presence.
While the very small strains induced by dynamically exciting the pile with the modal
hammer may justify the use of the small strain shear modulus, G0, the effect of pile
installation on the in-situ G0 is ignored in this study. Complications arise in estimating an
operative shear modulus, G, after pile installation, as for a given sand, G increases with
stress and reduces with strain. The large strains imposed during pile installation may result
in a significantly reduced operative G value in the shear zone surrounding the pile when
compared with G0. However, this may be counterbalanced by an increase in the far-field
confining stiffness, due to the high stresses and over-consolidation which occurs as the pile
tip passes. In addition, ageing has been shown to result in increased stiffness
characteristics of the soil which may reduce the effects of pile installation. Due to the
difficulty in quantifying these effects accurately, the in-situ G0 value was used and was
found to provide a good match with the experimental data.
Overall, the CPT numerical model showed greater agreement with the experiment than the
API model. This research paves the way for more accurate soil-structure dynamic
interaction modeling, which is particularly pertinent to the design of offshore wind
turbines and other structures that are dynamically sensitive.
171
Appendix C Experimental data summary
C.1 Synopsis
The purpose of this appendix is to display a summary of the data obtained during the
experimental phase of the doctoral studies undertaken by the candidate. Examples of
typical data are displayed in lieu of producing all of the data gathered. The data and the
relevant chapter of this thesis / publication are also cited below. Some observations on the
data are also given.
C.2 Blessington data – pile scour
The plots contained within this section were used in Chapter 4 for the experimental section
of the paper “An investigation of the changes in the natural frequency of a pile affected by
scour”.
172
Appendix C – Experimental data summary
Accelerometer 1
Accelerometer 2
500 500
Accelerometer 3
2260
1000
R1
Accelerometer 4
57
70
Initial Level R1
500
Scour Level -1
Blessington Sand
Scour Level -2
Pile Scour Level -3
Scour Level -4 Section A-A

8760
Scour Level -5
Scour Level -6
6500
Scour Level -7
Scour Level -8
A A
Scour Level -9
Scour Level -10
Scour Level -11
Scour Level -12
Base Level
Figure C-1 Scour levels for pile (all dimensions are in mm)
-1
-2
Scour Depth (m)
-3
-4
-5
-6 Experimental Frequency
-7
0 5 10 15 20 25 30 35 40
Frequency (Hz)
Figure C-2 Overall change in natural frequency with scour (Hz)
173
2
0m BGL
X: 31.25
Y: 1.845
Magnitude 1.5
0.5
0
5 10 15 20 25 30 35 40 45 50
Frequency (Hz)
Figure C-3 0 m scour depth - pile
40 X: 26.86
Y: 34.88
0.2m BGL
30
Magnitude
20
10
0
5 10 15 20 25 30 35 40 45 50
Frequency (Hz)
Figure C-4 0.2 m scour depth - pile
35 X: 24.66
Y: 30.18
0.4m BGL
30
25
Magnitude
20
15
10
0
5 10 15 20 25 30 35 40 45 50
Frequency (Hz)
174
20
X: 22.95
Y: 16.08 0.6m BGL
Magnitude 15
10
0
5 10 15 20 25 30 35 40 45 50
Frequency (Hz)
10
X: 22.22 0.8m BGL
Y: 7.856
8
Magnitude
0
5 10 15 20 25 30 35 40 45 50
Frequency (Hz)
50 X: 18.43
Y: 43.02
1.1m BGL
40
Magnitude
30
20
10
0
5 10 15 20 25 30 35 40 45 50
Frequency (Hz)
175
X: 15.14
150 Y: 137.7
1.6m BGL
100
Magnitude
50
0
5 10 15 20 25 30 35 40 45 50
Frequency (Hz)
200
X: 12.21
Y: 161.5 2.1m BGL
150
Magnitude
100
50
0
5 10 15 20 25 30 35 40 45 50
Frequency (Hz)
X: 10.5
300 Y: 280.9
2.6m BGL
250
200
Magnitude
150
100
50
0
5 10 15 20 25 30 35 40 45 50
Frequency (Hz)
176
X: 8.545
Y: 189.4
200 3.1m BGL
150
Magnitude
100
50
0
5 10 15 20 25 30 35 40 45 50
Frequency (Hz)
X: 6.714
1400 Y: 1289
3.6m BGL
1200
1000
Magnitude
800
600
400
200
0
5 10 15 20 25 30 35 40 45 50
Frequency (Hz)
700
X: 5.371 4.1m BGL
600 Y: 529.4
500
Magnitude
400
300
200
100
0
0 5 10 15 20 25 30 35 40 45 50
Frequency (Hz)
177
1000
4.6m BGL
Magnitude 800
600
X: 4.639
400
Y: 288.8
200
0
0 5 10 15 20 25 30 35 40 45 50
Frequency (Hz)
1200
5.1m BGL
1000
800
Magnitude
600
X: 4.15
400 Y: 274.5
200
0
0 5 10 15 20 25 30 35 40 45 50
Frequency (Hz)
60
5.6m BGL
50
40
Magnitude
30
X: 3.662
Y: 19.16
20
10
0
0 5 10 15 20 25 30 35 40 45 50
Frequency (Hz)
178
50
6.1m BGL
40
Magnitude
30
X: 2.197
20
Y: 14.38
10
0
0 5 10 15 20 25 30 35 40 45 50
Frequency (Hz)
The data obtained for the frequency response of the pile for all but the last two levels are
quite good and show a clear peak for the first natural frequency of the system. The last two
scour levels are more difficult to interpret and the interpretation was aided by observing
the trend for the change in natural frequency for all the scour levels above. The reason for
the poor data quality for the lower depths, in the opinion of the author, is that the pile at
this stage was quite unstable with only a small penetration depth into the soil (≈ 400 mm
for the last depth tested). This meant that the pile was easily perturbed and the soil elastic
response was possibly compromised meaning the frequency was quite difficult to obtain
for the last two levels. In short, the pile acted like a mechanism for the lowest scour levels
tested. Overall, however, the data show a clear reduction in the measured natural
frequency with increasing severity of scour.
C.3 Blessington data – monopile scour
The plots contained within this section were used in Chapter 6 for the experimental section
of the paper “An investigation into the effect of scour on the natural frequency of an
offshore wind turbine”.
179
75
Accelerometer
300 300 300

Reference Accelerometer
800
200
R1
Initial Level
56
Scour Depth -1
Blessington 70
Sand Scour Depth -2 R1
Scour Depth -3
3000
Scour Depth -4
1800
Scour Depth -5
2200
Scour Depth -6 Section A-A

Monopile Scour Depth -7
Scour Depth -8
Scour Depth -9
A A
Figure C-19 Scour levels for monopile (all dimensions are in mm)
-6
x 10 X: 57.2
3.5 Y: 3.075e-006
0.0m BGL scour - SV1
3
SV2
2.5
Amplitude
2
1.5
0.5
0
10 20 30 40 50 60 70 80
Frequency (Hz)
Figure C-20 0 m scour depth - monopile
180
-5
x 10
1.2
X: 51
0.2m BGL scour - SV1 Y: 9.63e-006
1 SV2
0.8
Amplitude
0.6
0.4
0.2
0
10 20 30 40 50 60 70
Frequency (Hz)
Figure C-21 0.2 m scour - monopile

-6 X: 47.94
x 10 Y: 4.695e-006
5
4 SV2
Amplitude
0
10 15 20 25 30 35 40 45 50 55 60
Frequency (Hz)

-4 X: 42.7
x 10
1.2 Y: 0.0001114
1 SV2
0.8
Amplitude
0.6
0.4
0.2
0
10 15 20 25 30 35 40 45 50 55 60
Frequency (Hz)
181
-5
x 10 X: 33.93
7 Y: 6.337e-005
6
SV2
5
Amplitude
4
3
0
10 15 20 25 30 35 40 45 50
Frequency (Hz)

-4
x 10 X: 28.53
3.5
Y: 0.000307
3
SV2
2.5
Amplitude
1.5
0.5
0
0 5 10 15 20 25 30 35 40 45 50
Frequency (Hz)
Figure C-25 1 m scour - monopile
-4
x 10
X: 22.64
6 Y: 0.0005301
5
SV2
Amplitude
0
0 5 10 15 20 25 30 35 40 45 50
Frequency (Hz)
182
-3
x 10 X: 17.48
1.2 Y: 0.001079
1 SV2
0.8
Amplitude
0.6
0.4
0.2
0
0 5 10 15 20 25 30 35 40 45 50
Frequency (Hz)
-3
x 10 X: 12.93
Y: 0.001196
1.2 1.6m BGL scour - SV1
SV2
1
Amplitude
0.8
0.6
0.4
0.2
0
0 5 10 15 20 25 30 35 40 45 50
Frequency (Hz)

-3 X: 7.992
x 10 Y: 0.001313
1.4
1.2
SV2
1
Amplitude
0.8
0.6
0.4
0.2
0
0 5 10 15 20 25 30 35 40 45 50
Frequency (Hz)
183
Overall, the data are quite good for most of the scour levels tested. The data was difficult
to interpret for the upper scour levels for the pilot-scale monopile. The reason for this is
that the pile-soil system was very stiff for the upper scour levels, which made it very
difficult to induce a clear bending vibration mode in the pile structure when applying the
modal hammer. The application of the modal hammer for the upper scour depths had the
effect of predominately exciting a cross-sectional annular vibration mode, which
dominated the frequency spectrum making it difficult to clearly identify the vibration
frequency corresponding to bending in the first mode. This means that the frequency
clarity is quite poor for the uppermost scour levels. However, by following the trend of the
change in natural frequency with scour commencing at the most-scoured state, it is
relatively straightforward to observe that the frequency steadily reduces with scour (since
the data for the lower levels is very clear and shows this trend). Since the purpose of this
experiment was to validate a numerical model of the small-strain dynamic pile-soil system,
the data suffices in this regard.
184
Appendix D An experimental investigation into the effect of water added mass on
the first natural frequency of cantilever beams
D.1 Synopsis
The use of vibration based methods to assess the condition of structures such as bridge
piers and offshore platforms is an area of growing interest. This appendix investigates the
effect of water added (virtual) mass on the first natural frequencies of cantilever structures.
The aim of this analysis is to show that scour will have a much greater effect on the natural
frequency of the structures of interest in this thesis than the effects of virtual added mass
from water. Three test cantilever sections of equal length and width were constructed with
differing cross-sectional dimensions, and thus varying flexural properties. The sections
were fixed in place at the base of a water tank and a vibration testing regime was
implemented to obtain the natural frequencies firstly in air then in a fixed depth of water.
Fourier transforms were applied to the time domain acceleration signals obtained from an
accelerometer fixed at the free end of each section, in order to obtain the natural frequency
for each test. The results indicated that the dynamic response of flexible structures is much
more affected by the presence of water than stiff structures.
D.2 Introduction
The effect of water on the natural frequency of structures has been investigated by a
number of authors (Sedlar et al. 2011; Ergin & Uğurlu 2003; Shabani et al. 2013; Badiane
et al. 2012; Esmailzadeh et al. 2008). In the first of these works (Sedlar et al. 2011), the
effect of the virtual added mass of water was investigated by submerging one element of a
discretised test cantilever in two water tanks with different dimensions and studying the
effect of the volume of water on the frequency response compared to the value measured
in air. Finite element models were used to compare the dry test and the submerged test
results and to quantify the amount of local added mass required to obtain the observed
change in frequency. It was found that the amount of added mass depended on the size of
the tank in which the section was tested. In the second work (Ergin & Uğurlu 2003), the
wet natural frequencies and associated mode shapes of two different, partially submerged
cantilever plates were calculated by an approach based on the boundary-integral equation.
185
Appendix D – An experimental investigation into the effect of water added mass on the
first natural frequency of cantilever beams
The frequencies calculated were compared to experimental data. The numerical

frequencies were obtained for a given set of depth ratios, which is the ratio of the
submerged depth to the total cantilever plate length. The wet natural frequencies were
observed to decrease with increased depth of submergence, with the lowest frequencies
occurring for the case of fully submerged cantilever plates. In the third of these works
(Shabani et al. 2013), the vibration of a cantilever micro-beam submerged in a bounded
frictionless fluid cavity was investigated numerically. The movement pattern of the fluid
for each structural mode was depicted. It was observed that the added (virtual) masses of
lower modes were greater than those of higher modes, but that the presence of fluid
changed the higher modes more effectively with fluid density playing an important role.
In the present discussion, the effect of the flexural rigidity on the first natural frequency
response of a structure in air and water is investigated. It is postulated that the effect of
water added mass will be higher for flexible sections than for stiff sections.
D.3 Experimental layout
An experimental investigation into the effect of water on the natural frequency of vibrating
cantilever sections was undertaken in the civil engineering laboratories at University
College Dublin. Three cantilever sections were fabricated using steel box section or steel
plate. Each section was 1.26 m long and 0.1 m wide. The flexural rigidity EI (product of
the Young’s modulus, E and the moment of inertia, I) were varied by changing the wall
thickness and cross-sectional area, see Figure D-1.The resulting EI values varied from 8.37
x 105 N m2 to 4.58 x 102 N m2, see Table D-1. The sections were thus considered to be
stiff, stiff-flexible and flexible in the order of reducing flexural rigidity.
Table D-1 Structural properties of the test cantilevers

Section Length (m) Width (m) Depth (m) Thickness Flexural Stiffness
(m) Rigidity (EI) Index
(N m2) (EI/L3)
(N m-1)
Stiff 1.26 0.1 0.1 0.008 8.37E+05 418362
Stiff- 1.26 0.1 0.05 0.006 1.28E+05 63915
Flexible
Flexible 1.26 0.1 - 0.0065 4.58E+02 229
In order to investigate the effect of water on the first natural frequency of vibration of
these beams, a 1 m3 water tank was constructed. Each test section was fixed in place to the
186
base of this tank by attaching the base-plate of each beam to the top flange of a heavy H-
section beam located at the bottom of the tank, see Figure D-1. The vibration tests were
undertaken along the longitudinal axis of this heavy H-section. A diagram of the test set-up
is shown in Figure D-1 with a photograph of the test arrangement displayed in Figure D-2.
Accelerometer
Test section
X X
50 mm
6.5 mm
100 mm
Water level
100 mm
1260 mm
100 mm
100 mm
8 mm 6 mm
Water tank
1000 mm
(a) Stiff section (b) Stiff-flexible (c) Flexible

section section
220 mm
Section X-X
Baseplate
H-Section
1000 mm
Figure D-1 Experimental layout. (a) Cross-section of stiff beam section; (b) Cross-section
of stiff-flexible beam section; (c) Cross-section of flexible beam section
(a) (b)
Figure D-2 Experimental arrangement. (a) Flexible section under wet test; (b) Stiff-flexible
Section under wet test
187
The vibration testing regime involved displacing the top of the cantilever section then
releasing it and measuring the dynamic response with an accelerometer attached to the top
of the section. A number of trials were undertaken on each section and the results were
averaged. Each section was first tested in air, and then subsequently tested in a depth of 1
m of water. The natural frequency of each section was obtained by applying a Fourier
transform to the time-domain acceleration signals obtained from both the air vibration and
water vibration tests.
D.4 Analysis and results
The effect of water on the acceleration signals obtained from the vibration test performed
on the section with the lowest flexural rigidity (i.e. the flexible section) is shown in Figure
D-3. All tests were performed a minimum of 12 times and the results are reported in terms
of the average natural frequency and standard deviation in Table D-2. It is clear that the
presence of water caused a reduction of the natural frequency, from 3.502 Hz in the dry
test, to 2.808 Hz in the test with water. An interesting feature of the wet test is the effect of
reflections, manifesting itself as an increase in the magnitude of the acceleration signal at
discrete points in Figure D-3(b).
188
Dry Test 0.4 Wet Test

0.5
0.2
Acceleration (g)
Acceleration (g)
0
0
-0.2
-0.4
-0.5
-0.6
-0.8
0 5 10 15 20 0 5 10 15 20
(a) Time (s) (b) Time (s)
2000 50
Dry Freq
Wet Freq
0 0
0 1 2 3 4 5
(c) Frequency (Hz)
Figure D-3 Comparison between dry and wet vibration tests for flexible section. (a) Dry
test acceleration signal (g), (b) Wet test acceleration signal (g), (c) Frequency content of
dry and wet accelerations signals shown in (a) and (b) (Hz)
The addition of water has resulted in more prominent damping in the signals obtained from
the wet vibration test also. The damping ratios for the dry and wet tests were measured as
ξ= 0.267% and ξ= 2.08% respectively, measured using a relatively straightforward
exponential curve-fitting process to the time-domain acceleration signals (Gutenbrunner et
al. 2007). This is in line with expectations that the presence of water would lead to a higher
value of damping. The same vibration tests were undertaken on the ‘Stiff-Flexible’ and the
‘Stiff’ cantilever sections also.
The measured differences between the frequency of vibration in air and that in water for all
three sections are presented in Table D-2. For comparison purposes, the frequency
measured in air is compared to the theoretical first natural frequency of vibration of a
cantilever, obtained from Eq. (D.1).
189
f1 
1
1.8752 EI 4 (Eq.D.1)
2 AL
where E is the Young’s modulus (N m-2); I is the moment of inertia of the cross-section
(m4); ρ is the density of steel used to fabricate the beam (kg m-3); A is the cross-sectional
area of the beam (m2); L is the length of the beam (m). It is worth noting that the results for
the stiff section were quite difficult to interpret and required signal filtering within the
expected band width of the first natural frequency. The reasons for the difficulty arise due
to the high value of damping present for the stiff section which made the signal clarity
quite low as well as the fact that some lateral rocking of the base beam H-section was also
detected in the frequency spectrum, which clouded the frequency results somewhat. The
filtering process allowed for easier estimation of the first natural frequency in bending.
Table D-2 Results of natural frequency analysis

Section Analytical Avg. Frequency in Avg. Frequency in Percentage
Cantilever Air ± STD. Dev. Water ± STD. change in
Frequency (Hz) (Hz) Dev. (Hz) Frequency
between Air and
Water (%)
Flexible 3.3 3.052 ± 0.0 2.81 ± 0.0 8
Stiff-Flexible 35.0 28.81 ± 0.17 27.08 ± 0.07 6
Stiff 67.1 62.26 ± 0.33 62.08 ± 0.15 0.3
From the results in Table D-2, it is evident that the effect of water on the natural frequency
of vibration is dependent on the flexural rigidity of the section tested. The stiff section
showed a much smaller change in measured frequency than the flexible section with
changes of 0.3% and 8% respectively. This is shown more clearly in Figure D-4.
190
9
Flexible Section 4.58E+02 N m2
8
7
Stiff-Flexible Section 1.28E+05 N m 2
Percentage Change (%)
4
Stiff Section 8.37E+05 N m2
3
0
0 1 2 3 4 5 6 7 8 9
Flexural Rigidity (EI) (N m2 ) 5
x 10
Figure D-4 Percentage frequency change (%) vs. flexural rigidity of each section (N m2)
D.5 Discussion and conclusions
This appendix considers the effect of water on the natural frequency of cantilever
structures. Three sections with different flexural rigidities were tested in air and water. It
was shown that the natural frequency of a flexible structure (with low flexural rigidity)
was sensitive to the presence of water added (virtual) mass, with a reduction of natural
frequency of 8% occurring whilst damping increased significantly. In contrast the presence
of water seemed to have little effect on the vibration response of the stiffest structure
tested. The findings of this research may be beneficial in areas such as the offshore
engineering industry, where dynamically sensitive wind turbine structures are currently
being developed. The effects of water added mass will depend on the system stiffness of
the foundation system adopted. Future testing of this problem should consider a range of
different structural stiffness values as the values tested within this analysis may be too
varied for a comprehensive conclusion to be given. One recommendation would be to re-
test the stiffest section with a more rigid base beam to avoid lateral rocking motion from
affecting the frequency spectra in the analysis. The results of the analysis suggest that the
effects of water added mass may be much less significant than the effect of scour around
structures that are relatively stiff, however it is acknowledged that the presence of water
will change the natural frequency of a system to some degree.
191
Appendix E Long-term monitoring of the frequency response of an offshore
METMAST structure
E.1 Synopsis
The purpose of this appendix is to highlight work undertaken by the candidate into the
long-term monitoring of the frequency response of an offshore METMAST structure
subjected to environmental loading in the HORNSEA Zone, off the coast of the UK. The
work contained within this appendix is part of a larger research project so only the relevant
information is presented. The data presented was analysed solely by the candidate.
E.2 Project description
In 2009, permission to develop an offshore wind farm in the North Sea was awarded to the
SMARTWIND consortium, a 50/50 joint venture between Mainstream Renewable Power
and Siemens Project Ventures, as part of The Crown Estate’s Round 3 site distribution
process. The HORNSEA Zone is 4735 square kilometres in size and is located between 34
km-190 km off the East coast of England adjacent to the River Humber, 200 km south of
Newcastle and 75 km north of The Wash. The water depth ranges from 24-70 m across the
zone with the majority of the area falling between depths of 30-40 m. The target electrical
generating capacity is approximately 4 GW, which will provide enough electricity to meet
4% of all electricity demand in the UK. The purpose of the project was to monitor the
structural response of a novel twisted jacket foundation structure subjected to the wind and
wave regimes of the HORNSEA zone.
E.3 METMAST twisted jacket structure
The twisted jacket structure was designed and patented by Keystone Engineering in the
USA. The primary steelwork for the structure consists of five main components, (i) a large
diameter central pile often referred to as a caisson that is driven vertically into the seabed,
(ii) a guide structure that is lifted and lowered over the central caisson, and (iii) three
raking piles driven through the guide structure at an angle of 1H:4V prior to grouting the
piles into the guide sleeves. The main features of the structure can be seen in Figure E-1.
192
Appendix E – Long-term monitoring of the frequency response of an offshore METMAST
structure
Keystone Engineering have experience of designing and installing two similar structures in
the Gulf of Mexico where the foundations have demonstrated their robust nature as they
survived extreme storm loading generated as Hurricane Katrina passed directly overhead.
The key advantages of the twisted jacket include the relatively lightweight construction
(using approximately 20% less primary steel than a traditional square jacket structure). The
structural frequency can also be changed by minor variations in the geometry, which can
provide efficient support for a wide variety of turbine models that are sensitive to different
resonance frequencies. The 'twisted jacket' uses less steel than a conventional jacket and is
easier to fabricate, using fewer components. Installation costs will also be lower because
more units can be fitted onto an installation barge due to its small footprint which occupies
less deck space than comparable foundations. Furthermore, no piling template is required,
which can simplify the installation process and may lead to quicker offshore operations.
Figure E-1 METMAST structure being transported to sea
193
structure
E.4 Instrumentation scheme
A number of accelerometers were installed on key structural components of the

METMAST, and all measure accelerations in both the North-South (N-S) and East-West
(E-W) directions. Table E-1 shows the layout and location of the accelerometers used in
this study and is divided by direction. Figure E-2 shows the approximate location of these
accelerometers on the installed structure.
Table E-1 Accelerometer layout

Accelerometer Location
AIM_SO_1N Central vertical member of guide structure
AIM_SO_1E Central vertical member of guide structure
AIM_SA_1N On (A) Sleeve of Guide Structure
AIM_SA_1E On (A) Sleeve of Guide Structure
The accelerometers shown in Table E-1 are located just below deck level on the structure.
They are the 4575 model fabricated by Bruel and Kjaer. The accelerometers have a
measurement range of ± 20 m s-2 and are programmed to take data samples at a rate of 4
Hz. They record the acceleration signals for a 30 minute period corresponding to the
maximum excitation encountered each day. This acceleration is logged each day in
separate data files and can be analysed separately to extract frequency information.
Figure E-2 Accelerometer layout on the twisted jacket structure
194
structure
An example of an acceleration signal measured on two different dates is shown in Figure

E-3. The dates correspond to times when different extreme maximum significant wave
heights were experienced by the structure during its operation. It is noteworthy that since
the acceleration is measured over the 30 minutes corresponding to the most active
excitation, the Hs values shown may only have occurred once during this period and were
measured from a nearby wave buoy.
0.1
20/6/2012 - Hs Max = 4.72 m
28/5/2012 - Hs Max = 0.35 m
0.05
-0.05
-0.1
0 200 400 600 800 1000 1200 1400 1600 1800
Time (s)
Figure E-3 Acceleration signals measured on two different dates (m s-2)
E.5 Frequency analysis
The frequency response of the structure was monitored between 20th October 2011 and
16th February 2013. The data acquisition system continuously samples acceleration data at
a rate of 4 Hz. The acceleration data corresponding to the most active 30 minutes of the
day in terms of input excitation from the environmental conditions is then committed to
memory. The acceleration data is analysed for its frequency content using an output-only
modal identification technique known as Frequency Domain Decomposition (FDD)
(Brincker et al. 2001). This modal identification technique is ideal for circumstances where
the input excitation is unknown as is the case for this METMAST structure. The procedure
utilises the acceleration signals from all the accelerometers simultaneously in order to
obtain an estimate of the frequency response. The FDD procedure is an improvement on
the classical Fourier transform approach as it is less sensitive to frequency resolution
problems and allows for easier estimation of closely spaced modes as well as damping.
195
structure
The FDD process involves taking the Singular Value Decomposition (SVD) of the spectral
matrix and decomposing it into a set of auto-spectral density functions that each
correspond to a single degree of freedom (SDOF) system (Brincker et al. 2000). The
method is most effective when the input excitation is broad-banded (white noise) with light
damping, making it ideal for the structure under consideration.
An algorithm was developed by the candidate to analyse the daily acceleration signals
from the four accelerometers on the structure, two orientated in the North-South (N-S)
direction and two orientated in the East-West (E-W) direction. As the sampling rate is
quite low, only the first natural frequency (global sway) of the structure is measurable in
this study. The algorithm reads in the acceleration signals for each date from all the
accelerometers and creates a system matrix comprising the individual acceleration vectors.
Linear baseline correction is applied to each acceleration vector in order to remove any
drift bias and cosine tapers are applied to the start and end of the acceleration data in order
to reduce spectral leakage. A bandwidth filter is also applied to the raw acceleration data
with a band width of 0.5 Hz to 1.5 Hz. This helps to remove noise not related to the first
natural frequency of the structure.
The acceleration vectors are then analysed using the FDD algorithm where the frequency
content is extracted. The frequency value at the maximum amplitude is automatically
extracted and logged as the dominant frequency in the spectrum for each date. A typical
example of the structural frequency response for an arbitrary date (February 10th, 2013) is
shown in terms of the N-S and E-W directions in Figure E-4.
196
structure
0.05
North Frequency (Hz)
0.04 East Frequency (Hz)
Magnitude
0.03
0.02
0.01
0
0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4
Frequency (Hz)
Figure E-4 Frequency response February 10th 2013
The frequency response of the structure obtained from the four accelerometers is shown
logged against each date in Figure E-5. A comparison with the design frequency is also
shown in this plot (The design frequency was established using a numerical model
developed by Keystone Engineering and was not undertaken by the candidate). Figure E-6
shows the frequency content obtained separately in the N-S direction and the E-W
direction.
1.19 Measured Frequency Data

Design Frequency
1.17
1.15
Frequency (Hz)
1.13
1.11
1.09
1.07
1.05
Oct 11 Jan 12 Apr 12 Jul 12 Oct 12 Jan 13
Figure E-5 Global dominant frequency response of structure (measured using FDD) (Hz)
197
structure
1.19 North Frequency Data

East Frequency Data
1.17 Design Frequency
1.15
Frequency (Hz)
1.13
1.11
1.09
1.07
1.05
Oct 11 Jan 12 Apr 12 Jul 12 Oct 12 Jan 13
Figure E-6 Frequency response in North-South & East-West directions (measured using
FDD) (Hz)
The frequency response of the structure is higher than that predicted in the original design
(which was undertaken by Keystone Engineering). The results of the frequency analysis
are shown in Table E-2.
Table E-2 Frequency analysis

Parameter: Average: Standard Deviation:
Design Frequency (Hz) 1.08 0.00
Measured Overall Frequency (Hz) 1.13 0.01
Frequency N-S (Hz) 1.126 0.007
Frequency E-W (Hz) 1.135 0.0076
The average difference between the measured and design frequency is 4.64 ± 0.92 %. This
difference most probably arises due to minor differences in the operational stiffness of the
soil stratum compared with the modelled stiffness of the soil stratum in which the structure
is installed.
E.6 Summary
A novel offshore twisted jacket structure was installed at the HORNSEA Zone in the UK
sector of the North Sea. This structure was monitored for the first 15 months using
accelerometers mounted on the guide sleeves of the jacket structure. This appendix
describes the main findings resulting from the analysis of the resulting data.
198
structure
(i) The structure was observed to have a higher natural frequency than designed,
suggesting that the structure has behaved stiffer than expected. This is possibly
due to the in-situ soil acting in a stiffer manner than considered in design.
(ii) The trend of frequency with date seems to drop slightly towards the later dates
of the analysis. However, the drop is slight and the daily changes in measured
frequency are much higher in comparison.
(iii) No correlation with scour depths were possible as a scour sensor installed on
the piles failed early in the lifetime of the structure.
199
Appendix F Formulation of modelling elements
F.1 Elemental beam stiffness matrix derivation
L
M1 M2
V1 V2
y,v
x,u
r(x) = dv(x)
r1 dx r2
v(x)
v1
v2
Figure F-1 Derivation of beam stiffness matrix
The beam element shown in Figure F-1 contains four degrees of freedom, namely two
nodal transverse displacements, v1 and v2, and two nodal rotations, r1 and r2. The
transverse displacement of a beam element is assumed to follow the shape of a cubic
polynomial v(x) as shown in Eq. (F.1).
vx   a0  a1 x  a2 x 2  a3 x3 (Eq.F.1)
A polynomial in x of degree 3 is assumed since there are four degrees of freedom hence a
minimum of four parameters are required to describe the displacement field. The rotation
r(x) is obtained by differentiating the expression shown in Eq. (F.1). This is shown in Eq.
(F.2).
200
Appendix F – Formulation of modelling elements
d v x 
 x    a1  2a2 x  3a3 x 2 (Eq.F.2)
dx
where r(x) is written as θ(x). By substituting the boundary conditions of the beam element
into Eq. (F.1), one can obtain an expression relating the boundary conditions to the
unknown coefficients ai .This is shown in Eq. (F.3).
v1  vx  x 0  a0 (Eq.F.3a)
1   x  x0  a1 (Eq.F.3b)
v2  vx  x  L  a0  a1L  a2 L2  a3 L3 (Eq.F.3c)
 2   x  x L  a1  2a2 L  3a3 L2 (Eq.F.3d)
One wishes to express the system in terms of nodal forces and nodal displacements. This
can be expressed as follows in Eq. (F.4).
P  K d  (Eq.F.4)
where P  V1 M1 V2 M 2  is the nodal force vector, d   v1 1

T
v2  2  is the
T
nodal displacement vector and K  is the stiffness matrix of the beam element. From Eq.
(F.3), one can re-write the nodal displacement vector d  in terms of the coefficients of
the displacement function v(x). This is shown in Eq. (F.5).
 v1   a0  1 0 0 0 a0 
    0 0 
 
 1  a1   1 0  a1 
d       3
   (Eq.F.5)
v2  a0  a1 L  a2 L  a3 L   1
2
L L2 L3 a2 

  1 2 L 3L2 
 2 
 
 a1  2a2 L  3a3 L  a3 
2
 0 
The nodal displacement vector d  has now been expressed in terms of the coefficients of
the transverse displacement function as d   [A]a. One now solves this equation for a
201
as a  [A]1d . This is shown in Eq. (F.6). The matrix manipulation with relation to
inverting the matrix [A] is omitted.
 1 0 0 0   v1 
a0    v1   
a   0 1 0 0    1 
 1  3 2 3 1 1   3 2 3 1 
    2        2 v1  1  2 v2   2 

(Eq.F.6)
a 2   L L L2 L  v2  L L L L
a3   2 1 2 1    2 1 2 1 
 3  2   3 v1  2 1  3 v2  2  2 
 L3 L2 L L2  L L L L 
One has by now obtained expressions for the coefficients of the transverse displacement
function v(x) in terms of the boundary conditions of the beam element. One can now
substitute these expressions back into Eq. (F.1). By doing this and grouping together the
terms in respect of the boundary conditions, the following expression in Eq. (F.7) is
obtained.
 3x 2 2 x 2   2 x 2 x3   3x 2 2 x 3   x 2 x3 
vx   1  2  3 v1   x   2 1   2  3 v2     2  2 (Eq.F.7)
 L L   L L   L L   L L 
Eq. (F.7) can be re-written in the following format in Eq. (F.8).
 v1 
 
 
vx   N1 x  N 2 x  N 3 x  N 4 x  1  (Eq.F.8)
 v2 

 2 

where N(x) are the interpolation shape functions used to describe the displacement at any
point along the beam element in terms of the nodal displacements / rotations. These are
summarised in Eq. (F.9).
202
 3x 2 2 x 2 
1  2  3 
 N1  x   
L L 
  2x 2
x3 
 
 N 2  x  
x
 L L2 
 N x    (Eq.F.9)
 3x 2 2 x 3 
 3
  3
 N  x   L2 L 
 4 
 x 2 x3 
   2 
 L L 
At this point, it is necessary to prescribe some beam theory in order to develop expressions
for the stiffness matrix of the beam elements. One wants to obtain expressions for the
external loads in terms of the nodal displacements. Figure F-2 shows the internal forces
experienced by a beam element when subjected to the global nodal loads / reactions {Vi,
Mi}.
L
M1 M2
V1 Mb1 Mb2 V2
Vb1 Vb2
Figure F-2 Internal shear forces and bending moments on beam element
Shear force is defined as the rate of change of bending moment. Since V1 is in the same
direction as Vb1, one can write Eq. (F.10).
d M b1
V1  (Eq.F.10)
dx
However, it is known (from beam theory) that M b1   x  . Hence, one can write in Eq.
I
y
(F.11):
I d  x 
V1   (Eq.F.11)
y dx
203
One can write  x   E x  and  x   

y
, where  is the radius of curvature.

Therefore, Eq. (F.11) can be written in the form shown in Eq. (F.12).
d 1
V1  EI  
dx   
(Eq.F.12)
d 2 v x 
1
From beam theory, one can write  , therefore Eq. (F.12) can be re-written as
 dx 2
Eq. (F.13).
d 3 v x 
V1  EI (Eq.F.13)
dx 3
One now moves to the derivation of an expression for M1. M1 has the opposite sign to Mb1.
Therefore, one can write Eq. (F.14).
 yI 1
M 1  M b1   x   E x   E     EI  
I I
(Eq.F.14)
y y  y 
From Eq. (F.12), one can re-write Eq. (F.14) as Eq. (F.15).
d 2 v x 
M 1   EI (Eq.F.15)
dx 2
One now moves to find an expression for V2. V2 is in the opposite direction to Vb2 so one
can write Eq. (F.16).
d M b2
V2   (Eq.F.16)
dx
Following the derivation from Eq. (F.10) to Eq. (F.13), one can write Eq. (F.17).
204
d 3 v x 
V2   EI (Eq.F.17)
dx 3
One now moves to the derivation of an expression for M2. M2 has the same sign as Mb2.
Therefore, one can write Eq. (F.18).
 yI 1 d 2 v x 
  x    E x    E    EI    EI
I I
M 2  M b2 (Eq.F.18)
y y  y  dx 2
One can now summarize the nodal force vector P as follows in Eq. (F.19).
 d 3v  x  
 EI 3 x 0 
dx
 V1   d 2 v x  

 M   EI
P   1    
x 0
dx 2
 (Eq.F.19)
 V2   EI d v x  x  L 
3
M 2   dx 3 
 d v x 
2

 EI xL 
 dx 2

Using Eq. (F.8), one can re-write Eq. (F.19) in terms of the derived interpolation shape
functions. This is shown in Eq. (F.20).
 d3 
 EI 3 N1v1  N 2 2  N 3v3  N 4 4  x 0 
dx
 V1   d2

M    EI 2  N1v1  N 2 2  N 3v3  N 4 4  x 0 
 1  
P      dx
3  (Eq.F.20)
 V2   EI d  N1v1  N 2 2  N 3v3  N 4 4  x  L 
M 2   dx 3 
 d2 
 EI 2 N1v1  N 2 2  N 3v3  N 4 4  x  L 
 dx 
From Eq. (F.9), one has expressions in x to describe each Ni. By differentiating these
expressions appropriately and substituting the resulting expressions into Eq. (F.20), one
obtains an expression for P. Eq. (F.21) shows the resulting expression obtained in matrix
205
form, where the nodal displacements have been extracted from the expressions. This
equation is now in the form P  K d  .
 V1   12 6 L  12 6 L   v1 
M   6 L 4 L2  6 L 2 L2   
 1  EI   1
  3   (Eq.F.21)
 V2  L  12  6 L 12  6 L  v2 
  2 
M 2   6L 2L  6L 4L   2 
2
 
Therefore the stiffness matrix for a four-degree-of-freedom beam can be expressed in Eq.
(F.22).
 12 6 L  12 6 L 
 6 L 4 L2  6 L 2 L2 
K   EI3   (Eq.F.22)
L  12  6 L 12  6 L 
 2 
 6L 2L  6L 4L 
2
F.2 Spring stiffness matrix derivation
u1 u2
u(x)
Figure F-3 Derivation of spring stiffness matrix
The spring element shown in Figure F-3 contains two degrees of freedom, namely two
linear nodal displacements, u1 and u2. The linear displacement of the spring can be
represented by a polynomial of the first degree u(x) as shown in Eq. (F.23).
ux   a0 a1x (Eq.F.23)
206
By substituting the boundary conditions of the spring element into Eq. (F.23), one can
obtain an expression relating the boundary conditions to the unknown coefficients ai as
shown in Eq. (F.24).
u1  ux  x 0  a0 (Eq.F.23a)
u2  ux  x L  a0  a1L (Eq.F.23b)
One wishes to express the system in terms of nodal forces and displacements, as in the
form shown in Eq. (F.4). Firstly, one writes the nodal displacement vector d  in terms of
the unknown coefficients of the displacement function. This is shown in Eq. (F.24).
u1   a0  1 0 a0 
d          (Eq.F.24)
u 2  a0  a1 L 1 L  a1 
By now, one has expressed the nodal displacements in terms of the coefficients of the
displacement function as d   [A]a. One can now solve this equation for a as
a  [A]1d . This is shown in Eq. (F.25).
a0  1  L 0 u1   1 0  u 
       1 1  1  (Eq.F.25)
 1
a det  A    1 1 u 2   L  u
L  2 
One has now obtained expressions for the coefficients ai in terms of the nodal boundary
conditions. These are summarised in Eq. (F.26).
a0  u1 (Eq.F.26a)
1 1
a1   u1   u2 (Eq.F.26b)
 L  L
One can now substitute these expressions into the original displacement function u(x) from
Eq. (F.23) and grouping together the displacement terms one obtains the following
formulation in Eq. (F.27).
207
 x x
u x   u1 1    u2   (Eq.F.27)
 L L
Eq. (F.27) can be written as follows in Eq. (F.28)
u 
u x   N1 x  N 2 x  1  (Eq.F.28)
 u2 
where N(x) are the interpolating shape functions used to describe the displacement at any
point along the element.
For a linear spring element, the force-displacement (P-d) relationship is linear and can be
described by f  k , where  is the net spring deformation u2  u1 (See Figure F-3) and k
is the stiffness of the system. One can describe the force acting one each end of the spring
as f1 and f 2 . For equilibrium, f1  f 2  0 , therefore, f1   f 2 . So, one can write Eq.
(F.29).
f  k  k u2  u1  (Eq.F.29)
One can write Eq. (F.29) in terms of the nodal forces f1 and f 2 .as shown in Eq.(F.30).
f1  k u2  u1  (Eq.F.30a)
f 2  k u2  u1  (Eq.F.30b)
which can be re-written in matrix form as Eq. (F.31).
 f1   k  k  u1 
     
k  u2 
(Eq.F.31)
 f 2   k
The stiffness matrix for a linear spring can thus be described as Eq. (F.32).
208
k  k
K     (Eq.F.32)
 k k 
F.3 Spring-beam element formulation
This section describes briefly how to formulate a stiffness matrix for a combined beam and
spring system, which is used widely as part of the numerical modelling within this thesis.
In order to formulate the stiffness matrix for the four-degree-of-freedom (4-DOF) beam
shown in Eq. (F.22) a pictorial view of the unit displacements applied to a beam element is
shown in Figure F-4.
F1
B
1
F2 F3
A EI
F4
1
F6
F5
F8 F7
Figure F-4 Beam element stiffness matrix summary formulation
The stiffness formulations arising from the unit displacements in Figure F-3 are shown in
Eq. (F.32). These correspond to the four degrees of movement, two transverse
displacements and two rotations.
6 EI
F1  F 2  (Eq.F.32a)
l2
12 EI
F 3  F 4  (Eq.F.32b)
l3
209
4 EI
F5  (Eq.F.32c)
l
2 EI
F6  (Eq.F.32d)
l
6 EI
F 7  F 8  (Eq.F.32e)
l2
These can be represented in matrix form by Eq. (F.22).
Figure F-5 shows a two-degree-of-freedom (2-DOF) linear spring element undergoing unit
displacements applied to each end.
F1 F2
ks
1
F1 F2
ks
A B
Figure F-5 Spring element stiffness matrix formulation
Using the relation F  Kx where ‘x’ is the unit displacement, one can write Eq. (F.33).
F1   F 2  ks (Eq.F.33)
In matrix form, one can write F  Kd as Eq. (F.34).
FA   k s  k s  d A 
   
k s  d B 
(Eq.F.34)
FB   k s
One can combine the 4-DOF beam element and the 2-DOF spring element to create a
combined 5-DOF element as shown in Figure F-6. The stiffness matrix for this combined
element is shown in Eq. (F.35).
210
3 2 5 4
EI
ks
Figure F-6 Spring-beam combined element
 ks  ks 0 0 0 
 12 EI 6 EI 12 EI 6 EI 
 k s ks  3  3 
 L L2 L L2 
 6 EI 4 EI 6 EI 2 EI 
 2
K    0 L2 L L L  (Eq.F.35)
 12 EI 6 EI 12 EI 6 EI 
 0  3  2  2 
 L L L3 L
6 EI 2 EI 6 EI 4 EI 
 0  2 
 L2 L L L 
211
Appendix G MATLAB Code – Dynamic modelling of spring-beam systems
G.1 Synopsis
This appendix displays an example of the MATLAB code used to model the dynamic
behaviour of a spring-beam model subjected to an impulse load. The model properties are
arbitrarily chosen for the purpose of this example. A 22 m long model pile embedded 20 m
into soil is modelled for the purpose of this example. The pile is subjected to a lateral
impact load of 500 N applied to the top of the pile. The discretised displacement,
acceleration and frequency of the system is calculated for a duration of 1 second after the
load has been applied.
G.2 MATLAB Code

%Spring Beam Numerical Model
%Luke J. Prendergast
%Model Notes:
% - Input dynamic impulse load
% - Calculate resulting displacement, acceleration and frequency response
%------------%
%---INPUTS---%
%------------%
clear all;clc; close all;
%---Pile Inputs---%
D_Out = 1; %Outer diameter [m]
Pile_Thickness = 0.05; %Pile thickness [m]
D_In = D_Out - (2*Pile_Thickness); %Inner diameter [m]
A = (pi*(D_Out^2)/4)-(pi*(D_In^2)/4); %Area of pile [sq.m]

I = (pi*(D_Out^4)/64)-(pi*(D_In^4)/64); %Moment of inertia [m^4]
E = 2e11; %Young's modulus [N/sq.m]

rho = 7850; %Density of steel [kg/m^3]
L = 1; %Length of the elements [m]
N = 22; %Total number of elements
Ns = 20; %No. spring elements
Nb = N - Ns; %No. standard beam elements
%---Spring Inputs---%
ks_val = 100000000; %Default stiffness value [N/m]
for i = 1:Ns
ks(i) = ks_val; %Create matrix of stiffness
end
212
Appendix G – MATLAB Code – Dynamic modelling of spring-beam systems
%----------------%
%---END INPUTS---%
%----------------%
%-------------------------%
%---BOUNDARY CONDITIONS---%
%-------------------------%
restrain = [1:1:Ns];
bc = (restrain.*3)-(2*(ones(1,Ns))); %Restrain spring nodes
%-----------------------------%
%---END BOUNDARY CONDITIONS---%
%-----------------------------%
%-----------------------------------------%
%---CALCULATE SYSTEM DEGREES OF FREEDOM---%
%-----------------------------------------%
NEL = N; %Number elements
nNELs = 3; %Number of nodes per spring beam element

nDOFs = 5; %Number of dofs per spring beam element
nNEL = 2; %Number of nodes per beam element

nDOF = 4; %Number of dofs per beam element
tDOF = ((Ns*3)+2) + (Nb*2); %Total number of dofs
%---------------------------------------------%
%---END CALCULATE SYSTEM DEGREES OF FREEDOM---%
%---------------------------------------------%
%-------------------------------%
%---DYNAMIC SIMULATION INPUTS---%
%-------------------------------%
tk = 0.001; %Time interval you wish to use in computer model [s]

Ax1 = -500; %Load to be applied to the beam in newtons [N]
xi = 0; %The damping ratio you want to use
th = 1.4; %The value of thetha to use in wilson
theta method (thetha > 1.4)
delta = 0.5; %Newmark-beta parameter
alpha = 0.25; %Newmark-beta parameter
T = 1; %Time duration of dynamic simulation[s]
time = [0:tk:T]; %Time Vector
Adof = tDOF-1; %The degree of freedom to plot at the end of the
simulation (End of pile - transverse direction)
Ldof = tDOF-1; %Degree of freedom at which the impulse load is
applied
modeplot = 1; %Mode number to be plotted at end of simulation
%-----------------------------------%
%---END DYNAMIC SIMULATION INPUTS---%
%-----------------------------------%
%--------------------------%
%---INITIALISE VARIABLES---%
%--------------------------%
Kg = zeros(tDOF,tDOF); %Initialise global stiffness matrix
Mg = zeros(tDOF,tDOF); %Initialise global mass matrix
213
indexS = zeros(5,1); %Initialisation of index vector for spring beam

elements
index = zeros(4,1); %Initialisation of index vector for standard
beam elements
%------------------------------%
%---END INITIALISE VARIABLES---%
%------------------------------%
%--------------------------------%
%---COMPUTE ELEMENTAL MATRICES---%
%--------------------------------%
a = (12*E*I)/(L^3); b = (6*E*I)/(L^2); c = (4*E*I)/(L); d = (2*E*I)/(L);
%Elementary stiffness matrix for the spring beam element

Kes = [ks_val -ks_val 0 0 0
-ks_val a+ks_val b -a b
0 b c -b d
0 -a -b a -b
0 b d -b c];
%Elementary stiffness matrix for the standard beam element

Ke = [ a b -a b
b c -b d
-a -b a -b
b d -b c];
%Define elementary mass matrices for spring beam element and standard
%standard beam elements
e = (rho*A*L)/420;
Mes = zeros(size(Kes)); %Massless Springs

ms = 0.001*156*e; %Default Spring Mass (nominal)
Mes = [ms -ms 0 0 0

-ms 156*e+ms 22*L*e 54*e -13*L*e
0 22*L*e 4*e*(L^2) 13*L*e -3*e*(L^2)
0 54*e 13*L*e 156*e -22*L*e
0 -13*L*e -3*e*(L^2) -22*L*e 4*e*(L^2)];
Me = [ 156*e 22*L*e 54*e -13*L*e

22*L*e 4*e*(L^2) 13*L*e -3*e*(L^2)
54*e 13*L*e 156*e -22*L*e
-13*L*e -3*e*(L^2) -22*L*e 4*e*(L^2)];
%------------------------------------%
%---END COMPUTE ELEMENTAL MATRICES---%
%------------------------------------%
%------------------------------------------------------------%
%---ASSEMBLE GLOBAL MASS AND STIFFNESS MATRICES FOR SYSTEM---%
%------------------------------------------------------------%
%First we include the contribution of the first '(Ns-1)' spring beam

%elements into the global stiffness matrix
for iel=1:(Ns-1)
Kes = [ks(iel) -ks(iel) 0 0 0

-ks(iel) a+ks(iel) b -a b
0 b c -b d
0 -a -b a -b
214
0 b d -b c];
hanger = iel*3;
index = [hanger-2, hanger-1, hanger, hanger+2, hanger+3];
edof = length(index);
for i=1:edof
ii=index(i);
for j=1:edof
jj=index(j);
Kg(ii,jj)= Kg(ii,jj) + Kes(i,j);
Mg(ii,jj)= Mg(ii,jj) + Mes(i,j);
end
end
end
%next we include the contribution of the 'Ns' sprung beam

%element into the global stiffness matrix
iel = Ns;
Kes = [ks(iel) -ks(iel) 0 0 0

-ks(iel) a+ks(iel) b -a b
0 b c -b d
0 -a -b a -b
0 b d -b c];
hanger = iel*3;
index = [hanger-2, hanger-1, hanger, hanger+1, hanger+2];
for i=1:edof
ii=index(i);
for j=1:edof
jj=index(j);
Kg(ii,jj)= Kg(ii,jj) + Kes(i,j);
Mg(ii,jj)= Mg(ii,jj) + Mes(i,j);
end
end
%Finally we need to include the contribution of the standard beam

%elements to the global stiffness matrix
datum = index(5);
for iel = 1:Nb

hanger = datum+(iel*2);
index = [ hanger-3, hanger-2 ,hanger-1, hanger];
for i=1:edof
ii=index(i);
for j=1:edof
jj=index(j);
Kg(ii,jj)= Kg(ii,jj) + Ke(i,j);
Mg(ii,jj)= Mg(ii,jj) + Me(i,j);
end
end
end
215
%----------------------------------------------------------------%
%---END ASSEMBLE GLOBAL MASS AND STIFFNESS MATRICES FOR SYSTEM---%
%----------------------------------------------------------------%
%--------------------------------------------------------------%
%---APPLY BOUNDARY CONDITIONS TO STIFFNESS AND MASS MATRICES---%
%--------------------------------------------------------------%
for i = 1:length(bc)
Kg(bc(i),:) = 0;
Kg(:,bc(i)) = 0;
Kg(bc(i),bc(i)) = 1;
end
Mg(bc(i),:) = 0;
Mg(:,bc(i)) = 0;
Mg(bc(i),bc(i)) = 1;
end
%------------------------------------------------------------------%
%---END APPLY BOUNDARY CONDITIONS TO STIFFNESS AND MASS MATRICES---%
%------------------------------------------------------------------%
%----------------------------------%
%---EIGENVALUES AND EIGENVECTORS---%
%----------------------------------%
%Define the system matrix [D] as follows

D = (inv(Mg))*Kg; %Natural frequencies relate to eigenvalues of system
matrix
[eig_vect,lambda] = eig(D); %By default, this gives you a vector of

eigenvalues starting with largest and the rest are in descending order.
%Because we are generally interested in the lowest frequencies, we need
to sort the eigenvalues (ie the entries of lambda) so that we are working
%with the lowest natural frequencies
%the eigenvector matrix is also sorted accordingly
diag_lambda = diag(lambda); %Assorts Eigenvalues into single vector
diag_lambda = diag_lambda(1:tDOF-length(bc)); %Ignore entries

corresponding to boundary conditions
eig_vect = eig_vect(:,(1:tDOF-length(bc))); %Ignore entries corresponding

to boundary conditions
%Sort and index entries in array

[lambda_sort,ind_array] = sort(diag_lambda); %Tracks entries for ease of
manipulation and querying of array for first natural frequency and
associated mode shapes
%Initialise the arrays that will store the frequency information

freqcalc = zeros(1,8); %This array gives the frequencies calculated from
the stiffness and mass matrices, 8 are chosen for default
216
for i = 1:8
freqcalc(i) = ((sqrt(lambda_sort(i)))/(2*pi));%Natural frequencies (Hz)
end
%Get coordinates for each node

transDOF_spring = 3*Ns; %Total DOF Spring Elements (when connected
together)
i = 2:3:transDOF_spring; %Isolates Transverse DOFs (spring elements)
transDOF_beam = 2*Nb; %Total DOF beam elements (when connected together)

j = (transDOF_spring+1):2:tDOF; %Counts from end of spring elements to
end of beam elements and isolates transverse DOFs
ij = [i,j]; %vector of transverse DOFs of both spring and beam elements

(Full)
BeamMode = zeros(1,length(ij));
for i=1:length(BeamMode)
BeamMode(i) = eig_vect(ij(i),ind_array(modeplot));
end
BeamX = zeros(1,(N+1));%X-coordinate
BeamYPlot = zeros(1,N+1);
BeamX = BeamMode; %BeamMode only contains lateral translations, no axial
for i =1:(N+1)
BeamYPlot(i) = ((i-1)*L);%Nodal locations
end
freqcalc %print frequencies to screen
%--------------------------------------%
%---END EIGENVALUES AND EIGENVECTORS---%
%--------------------------------------%
%------------------------------%
%---CALCULATE DAMPING MATRIX---%
%------------------------------%
%NOTES:
%Assume Rayleigh Damping
%It is possible to calculate a damping matrix as a linear combination of
%the mass and stiffness matrices i.e. Cg = aa[Mg] + bb[Kg], where
%aa = (2*xi*w1*w2)/(w1+w2) bb = (2*xi)/(w1 + w2)
%W1 = 1st Natural Frequency
%W2 = 2nd Natural Frequency
%w1 and w2 can be calculated from the mass and stiffness matrices
%Once you have Mg and Kg you can calculate the (undamped) natural
%frequencies of the beam
217
w1 = freqcalc(1).*(2*pi);%Use natural circular frequency instead of

cyclic for damping
w2 = freqcalc(2).*(2*pi);%Use natural circular frequency instead of
cyclic for damping
aa = (2*xi*w1*w2)/(w1+w2);
bb = (2*xi)/(w1+w2);
Cg = (aa*Mg)+(bb*Kg);
%Apply the boundary conditions
Cg(bc(i),:) = 0;
Cg(:,bc(i)) = 0;
Cg(bc(i),bc(i)) = 1;
end
%----------------------------------%
%---END CALCULATE DAMPING MATRIX---%
%----------------------------------%
%------------------------%
%---DYNAMIC SIMULATION---%
%------------------------%
n = length(time); %No. of elements in time vector
Fextnew = zeros(tDOF,n); %Initialise new force matrix
Fextnew(Ldof,1:10) = Ax1; %Apply impulse load over a number of time

increments to the loaded dof
%Apply Boundary Conditions to Force Vector
Fextnew(bc(i),:) = 0;
end
%Output the Dynamic Simulation to a function file 'Wilson-Theta Solver'

[displace, VEL, acc] = wilson_theta_solver_2(tk, tDOF, Mg, Cg, Kg, th, n,
Fextnew);
%Output the Dynamic Simulation to a function file 'Newmark-Beta Solver'

[displace_nb, VEL_nb, acc_nb] = newmark_beta_solver(tk, 0, tDOF, Mg, Cg,
Kg, delta,alpha, n, Fextnew);
%----------------------------%
%---END DYNAMIC SIMULATION---%
%----------------------------%
218
%----------------------------------------------%
%---CALCULATE FREQUENCY CONTENT FROM SIGNALS---%
%----------------------------------------------%
fs = 1/tk; %sampling frequency
[f_1,m_wave_1, psd_1] = My_FFT(displace(Adof,:),fs);%displacement

[f_2,m_wave_2, psd_2] = My_FFT(acc(Adof,:),fs);%acceleration
[f_1nb,m_wave_1nb, psd_1nb] = My_FFT(displace_nb(Adof,:),fs);
%--------------------------------------------------%
%---END CALCULATE FREQUENCY CONTENT FROM SIGNALS---%
%--------------------------------------------------%
%-----------%
%---PLOTS---%
%-----------%
%FIGURES
figure;plot(displace(Adof,:),'b');hold on;plot(displace_nb(Adof,:),'r');
figure;plot(acc(Adof,:),'b');hold on;plot(acc_nb(Adof,:),'r');
figure;plot(f_1,m_wave_1,'b');hold on;plot(f_1nb,m_wave_1nb,'r');xlim([0
40])
figure;plot(BeamX,BeamYPlot,'b');title('Mode Shape');xlabel('Modal disp
[-]');ylabel('Distance along structure [m]')
%---------------%
%---END PLOTS---%
%---------------%
%------------------------%
%---EXTERNAL FUNCTIONS---%
%------------------------%
%---Wilson-Theta Method---%
function [displace, VEL, acc] = wilson_theta_solver_2(tk, tDOF, Mg, Cg,
Kg, th, n, Fextnew)
acc=zeros(tDOF,n); %this array will store the acceleration information
VEL=zeros(tDOF,n); %this array will store the velocity information
displace=zeros(tDOF,n); %this array will store the displacement

information
%In this method, the displacement at time t+(theta*tk) is calculated

%and using this, we calculate acceleration, displacement and velocity at
%time t+tk, therefore we need one more array 'dispth' which defines the
%displacement of the beam-column at time step t+(theta*tk)
dispth = zeros(tDOF, n);
%We now need to input the initial displacements and velocities. In this
%case, the beam is un-displaced and at rest at the start of the motion
%therefore disp(:,1)=0 and VEL(:,1)=0, which are already set to zero so
we do not need to do anything
%Next we calculate the acceleration at time t = 0
219
%THIS IS A STEP-BY-STEP INTEGRATION METHOD

%The following equation holds true for accelerations
acc(:,1) = inv(Mg)*(Fextnew(:,1)-(Cg*VEL(:,1)) - (Kg*displace(:,1)));

%All DOFs, Time Step '1'
%We now define the integration constants

a0= 6 / ((th*tk)^2);
a1= 3/(th*tk);
a2= 2*a1;
a3= (th*tk)/2;
a4= a0/th;
a5= -a2/th;
a6= 1-(3/th);
a7= tk/2;
a8= (tk^2)/6;
%We now define the effective stiffness matrix Ke

Ke = zeros(tDOF,tDOF); %Initialise effective stiffness matrix
Ke = Kg + (a0*Mg) + (a1*Cg);
%We now define the effective force vector

Fe = zeros(tDOF,n); %initialise force vector
%To calculate the displacement of the beam, we need to go through the

%following procedures for each of the 'n' timesteps
inKe = inv(Ke); %Inverse of effective stiffness matrix
for i = 2:n
%Procedure 1 - Need to calculate the effective force vector at time
%[t+(th*tk)]
Fe(:,i)= Fextnew(:,i-1) + (th*( Fextnew(:,i)- Fextnew(:,i-1) ) ) + Mg*(

a0*displace(:,i-1) + a2*VEL(:,i-1) + 2*acc(:,i-1) )+Cg*( a1*displace(:,i-
1) +2*VEL(:,i-1) + a3*acc(:,i-1) );
%Procedure 2 - Solve for the displacements at time [t + (th*tk)]

dispth(:,i) = (inKe)*(Fe(:,i));
%Procedure 3 - Calculate the displacements, velocities and accelerations

at time [t + tk]
acc(:,i)= ( a4*( dispth(:,i) - displace(:,i-1) ) ) + a5*VEL(:,i-1)+

a6*acc(:,i-1) ; %ACCELERATION
VEL(:,i)= VEL(:,i-1) + a7*( acc(:,i) + acc(:,i-1) ) ; %VELOCITY
displace(:,i)= displace(:,i-1) + tk*( VEL(:,i-1) ) + a8*( acc(:,i) +( 2*(

acc(:,i-1) ) ) ) ; %DISPLACEMENT
end
end
220
%---Newmark-Beta Method---%
function [displace, VEL, acc] = newmark_beta_solver(tk, vel, tDOF, Mg,
Cg, Kg, delta, alpha, n, Fextnew)
%The next thing we need to do is to define the acceleration, velocity and

%displacement vectors. Since these vectors are different for each time
step, we need to define them as tDOF x n arrays (where n refers to the
number of timesteps to be used)
acc=zeros(tDOF,n); %this array will store the acceleration information
VEL=zeros(tDOF,n); %this array will store the velocity information
displace=zeros(tDOF,n); %this array will store the displacement

information
%We will use the newmark-beta method to solve for the displacements of
the system.
%We now need to input the initial displacements and velocities. In this
%case, the system is un-displaced and at rest at the start of the motion
%therefore disp(:,1)=0 and VEL(:,1)=0, which are already set to zero so
we do not need to do anything
%Next we calculate the acceleration at time t = 0

%We use the newmark-beta method to undertake this.
%THIS IS A STEP-BY-STEP INTEGRATION METHOD

%The following equation holds true for accelerations
%Fextnew is magnitude of force at each time step - apportioned to each

%degree of freedom on the beam
acc(:,1) = inv(Mg)*((Fextnew(:,1))-((Cg*VEL(:,1))) -
((Kg*displace(:,1)))); %All DOFs, time step '1' - Initial Acceleration
%Integration Constants - Newmark Beta

a0 = 1 / (alpha*(tk^2));
a1 = delta / (alpha*tk);
a2 = 1 / (alpha*tk);
a3 = (1 / (2*alpha)) - 1;
a4 = (delta / alpha) - 1;
a5 = (tk / 2)*((delta / alpha) - 2);
a6 = tk*(1 - delta);
a7 = delta*tk;
%We now define the effective stiffness matrix Ke
Ke = zeros(tDOF,tDOF); %Initialise effective stiffness matrix
Ke = Kg + (a0*Mg) + (a1*Cg);
%We now define the effective force vector

Fe = zeros(tDOF,n); %initialise effective force vector
%To calculate the displacement of the beam, we need to go through the

%following procedures for each of the 'n' timesteps
inKe = inv(Ke); %Inverse of effective stiffness matrix
221
for i = 2:n
%Procedure 1 - Need to calculate the effective force vector at time

%[t+tk)]
Fe(:,i)= Fextnew(:,i) + Mg*( a0*displace(:,i-1) + a2*VEL(:,i-1) +
a3*acc(:,i-1)) + Cg*( a1*displace(:,i-1) + a4*VEL(:,i-1) + a5*acc(:,i-1)
);
%Procedure 2 - Solve for the displacements at time [t + tk]

displace(:,i) = (inKe)*(Fe(:,i));
%Procedure 3 - Calculate the velocities and accelerations at

%time [t + tk]
acc(:,i)= a0*((displace(:,i) - displace(:,i-1))) - (a2*(VEL(:,i-1))) -

(a3*(acc(:,i-1))) ; %ACCELERATION
VEL(:,i)= VEL(:,i-1) + (a6*acc(:,i-1)) + (a7*acc(:,i)) ; %VELOCITY
end
end
%----------------------------%
%---END EXTERNAL FUNCTIONS---%
%----------------------------%
222
Appendix H Images of research
Figure H-1 Laboratory investigation into scour effect on natural frequency
Figure H-2 Fitting accelerometers for pile scour test
223
Appendix H – Images of research
Figure H-3 Pile scour test at advanced stage
Figure H-4 Close-up of fitted accelerometers
224
Figure H-5 Scour depth measurement
225
Figure H-6 Pilot monopile scour vibration test
Figure H-7 Vibration test for subgrade reaction comparison
Sin É
226
View publication stats

LukePrendergast PhDThesis FinalSubmission PDF

Uploaded by

Copyright:

Available Formats

You might also like

LukePrendergast PhDThesis FinalSubmission PDF

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

LukePrendergast PhDThesis FinalSubmission PDF

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Monitoring of scour around structures using changes in natural frequency of

Thesis · March 2015

The user has requested enhancement of the downloaded file.

MONITORING OF SCOUR AROUND STRUCTURES USING CHANGES IN

Bachelor of Engineering BE (Hons) MIEI

University College Cork, Ireland

Head of School: Supervisor:

Standing strong, reaching over

Depths below, muddy, secret,

The listener rests upon the bridge,

The listener hears, it cries alarm,

List of Tables ............................................................................................................................ x

List of Figures ......................................................................................................................... xi

Declaration ........................................................................................................................ xviii

Acknowledgements ............................................................................................................... xix

Executive Summary .............................................................................................................. xxi

Chapter 1 Introduction ....................................................................................................... 1

1.1 Background ............................................................................................................. 1

1.2 Research objectives ................................................................................................. 2

1.3 Outline of thesis ...................................................................................................... 3

Chapter 2 A review of bridge scour monitoring techniques ............................................ 7

2.1 Introduction to scour ............................................................................................... 8

2.2 Scour monitoring using depth-measuring instrumentation ................................... 12

2.2.1 Single-use devices ...................................................................................... 12

2.2.2 Pulse or radar devices ................................................................................ 13

2.2.3 Fibre-Bragg Grating (FBG) ....................................................................... 15

2.2.4 Driven or buried rod devices ..................................................................... 15

2.2.5 Sound wave devices ................................................................................... 19

2.2.6 Electrical conductivity devices .................................................................. 20

2.3 Scour monitoring using changes in structural dynamic properties ....................... 21

2.4 Conclusions ........................................................................................................... 34

Chapter 3 A comparison of initial stiffness formulations for small-strain soil-pile

3.1 Introduction ........................................................................................................... 36

3.1.1 Dynamic soil-structure interaction (DSSI) ................................................ 36

3.1.2 Winkler modelling approach ..................................................................... 38

3.1.3 Synopsis of chapter .................................................................................... 40

3.3 Structural modelling ............................................................................................. 45

3.4 Sensitivity study .................................................................................................... 48

3.4.1 Background to analysis & soil stiffness ..................................................... 48

3.4.2 Dynamic response ...................................................................................... 51

3.5 Field experiment ................................................................................................... 56

3.5.1 Test procedure and site investigation ......................................................... 56

3.5.2 Experimental frequency response and damping ratios .............................. 59

3.6 Comparison of experimental results with numerical analysis .............................. 65

3.7 Conclusions ........................................................................................................... 69

Chapter 4 An investigation of the changes in the natural frequency of a pile

4.1 Introduction ........................................................................................................... 72

4.2 Methodology ......................................................................................................... 72

4.2.1 Laboratory model ....................................................................................... 72

4.2.2 Field test ..................................................................................................... 77

4.3 Theoretical analysis .............................................................................................. 83

4.3.1 Structural model ......................................................................................... 83

4.3.2 Geotechnical stiffness determination ......................................................... 86

4.3.2.1 Method 1 – small-strain stiffness (G0) ................................................... 86

4.3.2.2 Method 2 – API design code .................................................................. 87

4.4.1 Numerical model of laboratory pile ........................................................... 90