Você está na página 1de 141

CENTRIFUGE MODELLING OF MONOPILE FOUNDATION FOR OFFSHORE

WIND TURBINES

Maria Fernanda Wamser Barra

Dissertação de Mestrado apresentada ao Programa


de Pós-graduação em Engenharia Civil, COPPE,
da Universidade Federal do Rio de Janeiro, como
parte dos requisitos necessários à obtenção do
título de Mestre em Engenharia Civil.

Orientadores: Marcio de Souza Soares de Almeida


Marcos Massao Futai

Rio de Janeiro
Março de 2020
CENTRIFUGE MODELLING OF MONOPILE FOUNDATION FOR OFFSHORE
WIND TURBINES

Maria Fernanda Wamser Barra

DISSERTAÇÃO SUBMETIDA AO CORPO DOCENTE DO INSTITUTO ALBERTO


LUIZ COIMBRA DE PÓS-GRADUAÇÃO E PESQUISA DE ENGENHARIA DA
UNIVERSIDADE FEDERAL DO RIO DE JANEIRO COMO PARTE DOS
REQUISITOS NECESSÁRIOS PARA A OBTENÇÃO DO GRAU DE MESTRE EM
CIÊNCIAS EM ENGENHARIA CIVIL.

Orientadores: Marcio de Souza Soares de Almeida


Marcos Massao Futai

Aprovada por: Prof. Marcio de Souza Soares de Almeida


Prof. Marcos Massao Futai
Prof.ª Maria Cascão Ferreira de Almeida
Dr. Rasmus Tofte Klinkvort
Dr. Samuel Felipe Mollepaza Tarazona

RIO DE JANEIRO, RJ - BRASIL


MARÇO DE 2020
Barra, Maria Fernanda Wamser
Centrifuge modelling of monopile foundation for
offshore wind turbines / Maria Fernanda Wamser Barra. -
Rio de Janeiro: UFRJ/COPPE, 2020.
XVIII, 122 p.: il.; 29,7 cm.
Orientadores: Marcio de Souza Soares de Almeida
Marcos Massao Futai
Dissertação (mestrado) - UFRJ/ COPPE/ Programa de
Engenharia Civil, 2020.
Referências Bibliográficas: p. 114-123.
1. Centrifuge modelling. 2. Offshore wind turbines. 3.
Lateral loads. 4. Monopiles I. Almeida, Márcio de Souza
Soares de et al. II. Universidade Federal do Rio de Janeiro,
COPPE, Programa de Engenharia Civil. III. Título.

iii
To my parents, Carlos and Cristina.
To my friend and love, Humberto.

iv
ACKNOWLEDGEMENTS

To my dear parents, Cristina and Carlos, for their amazing efforts that allowed me to
reach this far, I will always be grateful for your love and for inspiring me by the example.
To my sister Ana Luiza, for her friendship and companionship all the time.

To my beloved fiancé Humberto, that always encouraged me and stood by my side


during this journey, for never let me give up and for all the assistance in my research.
Your support was essential at this moment.

To my advisors, Marcio Almeida and Massao Futai, for expanding my horizons over
the last two years and for allowing me to develop this research on such an important topic,
by sharing their knowledge and providing me with the necessary support every step of
the way. For that, I am sincerely grateful and will take their teachings for my entire career.

To the members of the board, Professor Maria Cascão, Rasmus Klinkvort and Samuel
Tarazona for their availability and thoughtfulness on taking part in the evaluation of my
dissertation.

To the “Geofriends” for sharing with me all the experiences over this course. Your
support, talks and laughs made this challenging time easier to take.

To Silvia Polido, for your support since the beginning of the research, for the wise
advice and talks, I learned a lot with you. To Samuel Tarazona, for sharing your
knowledge about centrifuge modelling and assisting me on the preliminary tests and data
analysis.

To the technicians at the Geotechnical Lab of COPPE, Luizão, Carlinhos, Mauro, Gil,
Cid e Gilberto, for all your assistance and overtime on behalf of my research. To Tatiane
Benites for her availability and goodwill every time I needed it.

To the laboratory interns, Iago for your amazing efforts to help me, and Matheus,
Camila and Rodolfo for all your assistance over the test campaign, I would not have done
without your support. To Lucas Chinem, for all his help, especially with Matlab. To
Guilherme Geraldino for the assistance in the elaboration of the theoretical p-y curves

Finally, I thank all the other technicians, staff and students of LabGeo, for the support
over this period at COPPE/UFRJ.

v
Resumo da Dissertação apresentada à COPPE/UFRJ como parte dos requisitos
necessários para a obtenção do grau de Mestre em Ciências (M.Sc.)

MODELAGEM CENTRÍFUGA DE FUNDAÇÕES EM MONOPILES PARA


TORRES EÓLICAS OFFSHORE

Maria Fernanda Wamser Barra


Março/2020

Orientadores: Marcio de Souza Soares de Almeida


Marcos Massao Futai
Programa: Engenharia Civil

No Brasil, apesar de ainda não existirem parques eólicos offshore, a possibilidade


de implantação desses parques já é uma realidade. Assim, apresenta-se a necessidade de
compreender o comportamento de fundações para essas estruturas. Neste trabalho são
abordados os aspectos envolvidos na modelagem centrífuga de fundações do tipo
monopiles submetidos a cargas laterais de ondas e ventos. O principal objetivo é avaliar
o comportamento do solo frente a esses carregamentos. Foram realizados ensaios a 100g
em areias quartzosa e carbonatada, simulando um protótipo de 1,84 m de diâmetro externo
e 9,23 m de comprimento enterrado. Foram realizados ensaios monotônicos, envolvidos
em um projeto de benchmark, cujos resultados se mostraram consistentes e com
comportamento dentro do encontrado na literatura. Também foram feitos ensaios cíclicos
com 150 e 4800 ciclos, que mostraram a influência do número de ciclos na rigidez e no
mecanismo de ruptura do solo. Além disso, a partir de strain gauges colados ao longo do
modelo, perfis de momento fletor ao longo da profundidade permitiram obter a reação do
solo e o deslocamento sofrido pelo monopile ao longo da profundidade. Foi observado
que, quanto maior o percentual de carbonato de cálcio na amostra, maior a capacidade de
carga e rigidez frente ao carregamento lateral.

vi
Abstract of Dissertation presented to COPPE/UFRJ as a partial fulfillment of the
requirements for the degree of Master of Science (M.Sc.)

CENTRIFUGE MODELLING OF MONOPILE FOUNDATION FOR OFFSHORE


WIND TURBINES

Maria Fernanda Wamser Barra


March/2020

Advisors: Marcio de Souza Soares de Almeida


Marcos Massao Futai

Department: Civil Engineering

In Brazil, although there are no offshore wind farms yet, the possibility of
implementing these parks is already a reality. Thus, there is a need to understand the
behaviour of foundations for these structures. In this research, the aspects involved in the
centrifuge modelling of monopiles foundations subjected to lateral loads due to the action
of waves and winds are analysed. The main objective is to evaluate the behaviour of the
soil subjected to these loads. Tests were carried out at a centrifuge acceleration of 100g
on silica and carbonated sands, simulating a prototype of 1.84 m in external diameter and
9.23 m in embedded length. Monotonic tests were carried out, involved in a benchmark
project, and the results were shown to be consistent with behaviour observed in the
literature. Cyclic tests were also carried out with 150 and 4800 cycles, which showed the
influence of the number of cycles on the stiffness and failure mechanism of the soil. In
addition, using strain gauges along the model, bending moment profiles along the depth
were obtained, which made it possible to obtain the soil reaction and the displacement
profiles of the monopile along with the depth. It was observed that the higher the
percentage of calcium carbonate in the sample, the greater the carrying capacity and
stiffness in relation to lateral loading.

vii
TABLE OF CONTENTS

LIST OF FIGURES ......................................................................................................... xi

LIST OF TABLES ......................................................................................................... xv

LIST OF SYMBOLS ..................................................................................................... xvi

1. INTRODUCTION .................................................................................................. 1

1.1 GENERAL CONSIDERATIONS ..................................................................... 1

1.2 OBJECTIVES .................................................................................................... 3

1.3 THESIS LAYOUT ............................................................................................ 4

2. LITERATURE REVIEW ...................................................................................... 5

2.1 FOUNDATIONS FOR OFFSHORE WIND TURBINES ................................ 5

2.1.1 LOADS ACTING ON OFFSHORE WIND TURBINES ............................... 9

2.2 LATERALLY LOADED PILES.......................................................................... 13

2.2.1 P-Y CURVES ................................................................................................ 14

2.2.1.1 P-Y CURVES FOR PILES .................................................................... 14

2.2.1.2 CONSIDERATIONS FOR USING P-Y CURVES FOR MONOPILES21

2.2.1.3 PISA DESIGN MODEL ......................................................................... 25

2.2.2 ULTIMATE LIMIT STATE METHODS ..................................................... 30

2.3 PHYSICAL MODELLING IN CENTRIFUGE .............................................. 34

2.3.1 CENTRIFUGE MODELLING LIMITATIONS ........................................... 36

2.3.1.1 VARIATION IN GRAVITY FIELD ..................................................... 36

2.3.1.2 GRAIN SIZE EFFECTS ........................................................................ 38

2.3.2 PHYSICAL MODELLING OF MONOPILES ............................................. 39

2.4 FINAL CONSIDERATIONS ............................................................................... 41

3 MATERIALS AND METHODS ......................................................................... 43

3.3 MATERIALS AND TESTING EQUIPMENT ............................................... 43

3.1.1 SAND OF SÃO FRANCISCO BEACH (QZ) .............................................. 43

viii
3.1.2 CARBONATED SANDS (CA80 and CA50) ............................................... 43

3.1.3 BEAM CENTRIFUGE .................................................................................. 45

3.1.4 CYLINDRICAL CRADLE ........................................................................... 48

3.1.5 MONOPILE MODEL ................................................................................... 50

3.2 INSTRUMENTATION USED IN THE CENTRIFUGE TESTS ........................ 53

3.2.1 CALIBRATION OF STRAIN GAUGES FOR BENDING MOMENT ....... 56

3.3 MODEL PREPARATION AND CONFIGURATION ........................................ 61

3.3.1 PLUVIATION METHOD ............................................................................. 61

3.3.2 TEST PREPARATION ................................................................................. 62

3.4 TESTING PROGRAM AND SEQUENCE ......................................................... 66

3.4 1 CENTRIFUGE TESTS PERFORMED ........................................................ 66

3.5 FINAL CONSIDERATIONS ............................................................................... 70

4 RESULTS OF THE CENTRIFUGE TESTS ..................................................... 71

4.1 BCH TESTS .................................................................................................... 71

4.2 MF5 TESTS ..................................................................................................... 75

4.2.1 RESULTS FOR LOAD AND HORIZONTAL DISPLACEMENT OVER


TIME 75

4.2.2 LOAD-DISPLACEMENT CURVES ...................................................... 77

4.4.3 COMPARISON OF BEHAVIOUR BEFORE AND AFTER CYCLES ...... 80

4.3 MF6 TESTS ..................................................................................................... 83

4.3.1 RESULTS FOR LOAD AND HORIZONTAL DISPLACEMENT OVER


TIME 83

4.3.2 COMPARISON OF BEHAVIOUR BEFORE AND AFTER CYCLES . 86

4.4 CHANGE OF SECANT STIFFNESS WITH NUMBER OF CYCLES - MF6


TESTS ........................................................................................................................ 88

4.5 COMPARISON BETWEEN MF5 AND MF6 TESTS ........................................ 92

4.6 BENDING MOMENT PROFILES ALONG DEPTH ......................................... 95

4.6.1 PRELIMINARY CONSIDERATIONS ........................................................ 95

ix
4.6.2 RESULTS OF BENDING MOMENT FOR INITIAL MONOTONIC
LOADING .............................................................................................................. 95

4.7 INTERPRETATION OF BENDING MOMENT RESULTS ALONG DEPTH


98

4.7.1 PRELIMINARY CONSIDERATIONS ........................................................ 98

4.7.2 SOIL REACTION AND DISPLACEMENT CURVES ........................ 101

4.8 THEORETICAL P-Y CURVES .................................................................... 105

4.9 RESULTS ANALYSIS AND DISCUSSION ............................................... 109

5 CONCLUSIONS AND SUGGESTIONS FOR FUTURE RESEARCH ....... 111

5.1 INTRODUCTION .............................................................................................. 111

5.2 CONCLUSIONS ................................................................................................ 112

5.3 SUGGESTIONS FOR FUTURE RESEARCH.................................................. 113

REFERENCES ........................................................................................................... 114

x
LIST OF FIGURES

Figure 1 – Electricity production in Brazil in 2018 (BRASIL, 2019) .............................. 1


Figure 2 – The increasing size of offshore wind turbines, 1985-2016 (IRENA, 2016). .. 5
Figure 3 – Estimated costs of an offshore wind farm (BLANCO, 2009) ......................... 6
Figure 4 – Types of foundations for offshore wind turbines (WORLD STEEL, 2017) .. 7
Figure 5 – Transition piece for monopiles (Adapted from IXWIND, 2020) ................... 8
Figure 6 – Typical loads acting on an offshore wind turbine (a) (BYRNE and
HOULSBY, 2015) and an oil platform (b) (BYRNE and HOULSBY, 2003) ............... 10
Figure 7 - External loads acting on an offshore wind turbine, together with their typical
waveforms. (NIKITAS et al., 2016) ............................................................................... 12
Figure 8 – Frequency of loads acting on an offshore wind turbine (Adapted from LE
BLANC, 2009). .............................................................................................................. 13
Figure 9 – Pile subjected to lateral load (VELLOSO and LOPES, 2010)...................... 15
Figure 10 – Soil pressure around a pile (KIRKWOOD, 2015) ...................................... 15
Figure 11 – Soil pressure represented by Winkler springs (ROCSCIENCE, 2018) ...... 17
Figure 12 – p-y curve for sand by the Method of Reese et al., 1974 (ROCSCIENCE,
2018) ............................................................................................................................... 18
Figure 13 – Coefficients of soil resistance versus depth (REESE and VAN IMPE, 2014)
........................................................................................................................................ 19
Figure 14 – p-y curve for sand by the API Method (ROCSCIENCE, 2018) ................. 19
Figure 15 – Values for C1, C2 and C3 for sands in API (ISENHOWER and WANG,
2016) ............................................................................................................................... 20
Figure 16 – Values for 𝑘𝑖𝑛𝑖 for sands in API (MEYER and REESE, 1979) ................. 21
Figure 17 – Difference in the behaviour of a rigid pile (a) and flexible pile (b)
(BRODBAEK et al., 2009)............................................................................................. 22
Figure 18 - Kr plotted against L/D ratio for full-scale offshore wind farms (OWF) in the
United Kingdom, and piles used for the development of p-y curves. (ABADIE et al.,
2019) ............................................................................................................................... 23
Figure 19 - Geometry and parameters considered for the design problem (BURD et al.,
2019a) ............................................................................................................................. 25
Figure 20 - PISA design model: (a) idealisation of the soil reaction components acting
on the pile (b) 1D finite-element implementation of the model (BURD et al., 2019a) . 26

xi
Figure 21 – Conic function adopted for the parametric soil reaction curves (a) conic
form (b) bilinear form (BURD et al., 2019a) ................................................................. 28
Figure 22 – Performance of the 1D model, for Dr = 75 (a) pile L/D = 2 and (b) pile L/D
= 6 (BURD et al., 2019a) ............................................................................................... 29
Figure 23 – Comparison between computed responses determined from the 1D model
and equivalent 3D finite-element analyses for Dr = 85% (a) ultimate response (b) small
displacement response and (c) bending moment profile with depth- Solid line represent
3D finite-element results and dashed line the 1D model (BURD et al., 2019a) ............ 29
Figure 24 - Failure at small depths (REESE and VAN IMPE, 2014) ............................ 31
Figure 25 - Failure at great depths (REESE and VAN IMPE, 2014) ............................. 31
Figure 26 – Soil reaction at failure: Broms Method (1964) ........................................... 32
Figure 27 - Stress distribution around a laterally loaded pile (ZHANG et al., 2005) .... 33
Figure 28 - Correspondence between the inertial stress in a model and the gravitational
stress in the prototype (TAYLOR, 1995). ...................................................................... 35
Figure 29 - Prototype and centrifuge model of a horizontal soil layer (MADABHUSHI,
2014) ............................................................................................................................... 37
Figure 30 - Difference in stress distribution with a depth between model and prototype
(MADABHUSHI, 2014) ................................................................................................ 37
Figure 31 - Load-displacement curves for different installation methods (RANDOLPH
et al., 2001) ..................................................................................................................... 40
Figure 32 – Parameters to analyse the boundary effects ................................................ 41
Figure 33 – Beam Centrifuge of LM2C .......................................................................... 46
Figure 34 - Bi-directional actuator (BROADBENT Inc, 2011) ..................................... 47
Figure 35 - Plain strain cradle (BROADBENT Inc, 2011) ............................................ 47
Figure 36 - General arrangement of the beam centrifuge (BROADBENT Inc, 2011) .. 48
Figure 37 – Cylindrical box dimensions (in mm) .......................................................... 49
Figure 38 – Cylindrical box used in the tests ................................................................. 49
Figure 39 – Model dimensions (in mm) ......................................................................... 51
Figure 40 – Kr plotted against L/D ratio for the model used in the present study, full-
scale offshore wind farms (OWF) in the United Kingdom, and piles used for the
development of p-y curves. (AFTER ABADIE et al.., 2019) ........................................ 52
Figure 41 – Servo actuator with displacement control ................................................... 53
Figure 42 - Load cell 1250 N.......................................................................................... 54
Figure 43 – Fixation between the actuator and load cell system .................................... 54

xii
Figure 44 – Model instrumented with strain-gauges ...................................................... 55
Figure 45 - Laser for displacement measurement .......................................................... 55
Figure 46 - Distribution of strain gauges throughout the model .................................... 56
Figure 47 – Calibration with both sides of the model facing up: side I (a) and side II (b)
........................................................................................................................................ 57
Figure 48 – Calibration scheme ...................................................................................... 58
Figure 49 - Calibration loading and unloading steps...................................................... 59
Figure 50 – Calibration curve with loading and unloading ............................................ 60
Figure 51 – Pluviator used in sample preparation .......................................................... 62
Figure 52 – Weighting of the container with sand ......................................................... 63
Figure 53 - Test preparation ........................................................................................... 64
Figure 54 – Simplified test layout (a) and superior view (b) (dimensions in mm) ........ 65
Figure 55 – Type of movement in the cyclic tests .......................................................... 68
Figure 56 – Load-displacement curve (a) and normalized curve (b) for QZ sand ......... 71
Figure 57 - BCH result validation with the model proposed by Abadie (2015) ............ 72
Figure 58 – Load-displacement curve (a) and normalized curve (b) Results of the
benchmark (KLINKVORT, 2020) ................................................................................. 73
Figure 59 – Load-displacement curve (a) and normalized curve (b): Benchmark tests
performed on small diameter models (KLINKVORT, 2020) ........................................ 74
Figure 60 – Load over time: test MF5_QZ .................................................................... 75
Figure 61 – Displacement over time: Test MF5_QZ (at model) .................................... 76
Figure 62 – Displacement over time: detail of type of movement (at model)................ 76
Figure 63 – Normalized load-displacement curve: Test MF5_QZ................................. 77
Figure 64 – Normalized load-displacement curves: Tests MF5_CA80_1 and
MF5_CA80_2 ................................................................................................................. 78
Figure 65 – Load-displacement curves: Tests MF5_CA50_1 and MF5_CA50_2 ......... 79
Figure 66 –MF5 tests: Comparison of normalized load-displacement curves for QZ,
CA50, and CA80 sands .................................................................................................. 80
Figure 67 – Load-displacement curves: Tests MF5 – monotonic loading before (a) and
after (b) 150 cycles for sands (prototype scale) .............................................................. 81
Figure 68 - MF6 tests: Horizontal load over time during the for QZ sand (a) and CA80
sand (b) ........................................................................................................................... 84
Figure 69 - MF6_QZ: Load over time (model scale) ..................................................... 85
Figure 70 - MF6_CA80: Load over time (model scale) ................................................. 86

xiii
Figure 71 - MF6 tests: monotonic loading before (a) and after (b) 4800 cycles
(prototype scale) ............................................................................................................. 87
Figure 72 – Secant stiffness Kn for cycle n .................................................................... 88
Figure 73 – Load-displacement over 4800 cycles (prototype scale) for QZ sand (a) and
CA80 (b) ......................................................................................................................... 89
Figure 74 - MF6_QZ test: Load displacement for cycles 1-50 (a), 100-200 (b), 1000-
2000 (c), 3800-4800 (d) .................................................................................................. 90
Figure 75 - MF6_CA80 test: Load displacement for cycles 1-50 (a), 100-200 (b), 1000-
2000 (c), 3800-4800 (d) .................................................................................................. 91
Figure 76 – Change in secant stiffness with the number of cycles ................................ 92
Figure 77 – Load-displacement curves: before and after 150 cycles and before and after
4800 cycles, for QZ sand (a) and CA80 sand (b) ........................................................... 94
Figure 78 - MF5_QZ test: bending moment profile ....................................................... 96
Figure 79 - MF5_CA50_1 test: bending moment profile ............................................... 97
Figure 80 - MF5_CA80_1 test: bending moment profile ............................................... 98
Figure 81 –MF5_QZ test: Soil pressure, bending moment profile and pile displacement
...................................................................................................................................... 102
Figure 82 – MF5_CA50_1 test: Soil pressure, bending moment profile and pile
displacement ................................................................................................................. 103
Figure 83 - MF5_CA80_1 test: Soil pressure, bending moment profile and pile
displacement ................................................................................................................. 104
Figure 84 - p-y curve for QZ sand - z = 2 m (a) and z = 4 m (b) ................................. 106
Figure 85 - p-y curve for CA50 sand - z = 2 m (a) and z = 4 m (b) ............................. 107
Figure 86 - p-y curve for CA80 sand - z = 2 m (a) and z = 4 m (b) ............................. 108

xiv
LIST OF TABLES

Table 1 – Typical loads on offshore wind turbines ........................................................ 11


Table 2 – Dimensionless forms for the soil reaction curves (BURD et al., 2019a) ....... 27
Table 3 – Centrifuge scaling laws .................................................................................. 35
Table 4 - Physical indexes of the soils studied ............................................................... 45
Table 5 - Dimensions for model and prototype at N = 96 .............................................. 50
Table 6 – Parameters of scaling effects of the present study.......................................... 51
Table 7 – Soil-Pile relative stiffness ............................................................................... 52
Table 8 – Loads used in the calibration .......................................................................... 59
Table 9 – Strain-gauges calibration constants ................................................................ 60
Table 10 – Preliminary test campaign ............................................................................ 66
Table 11 – Final test campaign ....................................................................................... 67
Table 12 - Friction angle and average particle diameter for each sand in the benchmark
test (KLINKVORT, 2020) .............................................................................................. 75
Table 13 – MF5 tests: ultimate load before 150 cycles and comparison with Broms
(1964) ............................................................................................................................. 82
Table 14 - MF5 test: comparison of the ultimate load before and after 150 cycles ....... 83
Table 15 – Ultimate load before and after 4800 cycles .................................................. 86
Table 16 – Change in secant stiffness with the number of cycles .................................. 89
Table 17 – Comparison of ultimate load (MN) for tests MF5 and MF6 ........................ 93
Table 18 – Research parameters for centrifuge modelling of laterally loaded piles -
prototype dimensions (After Haiderali, 2015) .............................................................. 100

xv
LIST OF SYMBOLS

A Empirical factor for ultimate soil resistance

d Internal pile diameter

𝑑50 Average grain size

D External pile diameter

𝐷𝑟 Relative density of sample

e Void index

ex Load eccentricity

𝑒𝑚á𝑥 Maximum void index

𝑒𝑚í𝑛 Minimum void index

E Young’s modulus

𝐸𝑝𝑦 Soil reaction modulus based on p-y curves

𝐸𝑆𝐿 Soil Young’s modulus

Ep Pile Young’s modulus

𝐸𝑝 𝐼𝑝 Pile bending stiffness

f1 Natural frequency

g Gravity acceleration

𝐺𝑚á𝑥 Maximum shear modulus of the soil

Gs Grain unit weight

H Soil layer thickness

Ip Moment of inertia of the pile

k Permeability

xvi
𝑘𝑖𝑛𝑖 Initial stiffness

𝐾𝑝 Rankine’s earth pressure coefficient

𝐾𝑟 Soil-pile relative stiffness

𝐿 Embedded length of the pile

lb Distance between bottom of the box and model tip

le Free length

lt Pile thickness

Lt Pile total length

M Bending moment

𝑀𝑐 Mass of the cylindrical box

𝑀𝑠+𝑐 Mass of cylindrical box + soil

MR Modulus stiffness

N Scale factor

p Soil pressure

𝑝𝑢 Ultimate soil resistance

P Lateral load applied

Ra Roughness factor

T Time

V Axial load at pile head

𝑉𝑐 Cylindrical box volume

y Pile lateral displacement

z Depth

xvii
Greek symbols

𝛾 Soil unit weight

𝛾𝑚á𝑥 Maximum unit weight

𝛾𝑚í𝑛 Minimum unit weight

γ' Effective unit weight

γw Unit weight of water

𝛿 Angle of friction of the soil-pile interface

θ Angular speed

𝜈 Poisson’s ratio of the soil

𝜌 Soil density

𝜑 Friction angle of the soil

φ′ Effective friction angle of the soil

xviii
1. INTRODUCTION

1.1 GENERAL CONSIDERATIONS

In Brazil, hydroelectric energy production corresponded to about 80% of the total


generated in the country in 2018, as shown in Figure 1 (BRASIL, 2019). Despite being a
source of energy considered to be renewable, as it uses a resource that is replaced by
nature, a hydroelectric plant causes several environmental impacts in its place of
installation, therefore, it is not a sustainable energy source.

Figure 1 – Electricity production in Brazil in 2018 (BRASIL, 2019)

Thus, renewable energy sources that are sustainable in their method of installation
and operation are becoming more attractive. In Brazil, onshore wind energy has already
reached an installed capacity of 15 GW (ABEEÓLICA, 2018), and the forecast is that by
2025, 22.1 GW will be reached (CANAL ENERGIA, 2019).

Within the perspective of growth of wind energy in the energy matrix, the
implementation of offshore parks - which do not yet exist in Brazil - is already part of the
discussions of the sector, such as EPE and ABEEólica.

1
Despite having high costs of installation and operation, offshore wind can present
several advantages, such as greater generation capacity due to winds with higher speeds
and more constant in the coastal region, in addition to having greater availability of
unexplored areas and consequently fewer restrictions on the use of the area and distance
from the ground. With this, it is possible to use larger turbines with a greater generation
capacity in the offshore environment, which generally improves the performance
indicators of these parks (EPE, 2020).

In Europe, the installed capacity by the end of 2018 was 18.5 GW, distributed among
105 wind farms in 11 countries with more than 4000 turbines. 85% of this total installed
capacity is concentrated in only 3 countries: the United Kingdom, Germany, and
Denmark (WIND EUROPE, 2019).

The offshore wind capacity on the Brazilian coast is estimated between 697 GW
(EPE, 2020) to 1,780 GW (ORTIZ AND KAMPEL, 2011). In terms of comparison, the
onshore wind generation capacity is 143 GW (AMARANTE, 2001), which means that
the offshore wind potential is about 12 times greater than the onshore potential, which
can leverage the development of offshore wind towers in Brazil.

Despite the great capacity to be explored, there are still no offshore wind farms in
Brazil, but there is a forecast of a pilot project to be put into operation by Petrobras in
2022 (PETROBRAS, 2018), in addition to other projects with greater capacity currently
in process of licensing. In the region of the Brazilian coast, where these projects will be
developed, the occurrence of carbonated material is common, which presents a high
degree of cementation and usually brings difficulties in the execution of offshore works
(YEUNG and CARTER, 1989; GOMES, 2020). This is a topic that has concerned the
Brazilian offshore industry over the last four decades (ALMEIDA et al., 1987).

There are several types of foundation for offshore wind turbines, and the choice is
made based on the depth of installation and the geological conditions of the site. Among
the existing ones, the monopile has been the most used in Europe, corresponding to 75%
of the foundations (WIND EUROPE, 2019). The design methodologies used for
monopiles follow those widely used by the oil and gas industry. However, the mechanism
of load transfer that occurs in monopiles is different from that in offshore piles, as
monopiles have much larger diameters. Thus, the adequacy of methods such as those
proposed by API (2011) and DNV (2016) has been widely questioned. This dissertation

2
intends to contribute to this gap observed in the literature by conducting studies on
physical models in a geotechnical centrifuge of laterally loaded monopiles on silica and
carbonated sands found on the northeast coast of the country, in order to obtain insights
related to the response of this kind of soils subjected to lateral loads.

1.2 OBJECTIVES

The need for an in-depth study concerning the type of monopile foundation was
identified, seeking to understand the mechanisms of horizontal load transfer and adequacy
of the commonly used design methods. In addition to the need to study the behaviour of
the carbonated material subjected to lateral loads from the action of winds and waves.
Thus, the objective of this work is to study the behaviour of monopiles subjected to
monotonic and cyclic loads through physical modelling in the geotechnical centrifuge, in
silica and carbonate sand.

Also, this work is part of a Benchmark project, a study network with several
institutions around the world, such as Cambridge University, Norwegian Geotechnical
Institute, and IFFSTAR. Modelling the same prototype in different geotechnical
centrifuges, this Benchmark project aims to minimize the uncertainties identified in the
physical modelling of monopiles. The Benchmark project is detailed in Klinkvort et al..
(2018).

The specific objectives are:

1. Obtaining “Load-Displacement”, “Bending moment -depth” and “Soil Pressure


(p)-displacement (y)” curves for the tests.
2. Compare the results obtained with the p-y curves proposed by the main design
codes.
3. Compare the results for three types of dense sand: silica sand, carbonated sand with
50% CaCO3 and carbonated sand with 80% CaCO3.

3
1.3 THESIS LAYOUT

This work is divided into 5 chapters, including this one of Introduction.

Chapter 2 presents a literature review on the subject, addressing the main types of
foundation for offshore wind turbines and the loads acting on these structures, the main
aspects involved in the design of laterally loaded piles, and the differences between the
design of monopiles, in addition to a description of the physical modelling in the
centrifuge.

Chapter 3 presents the methodology used to prepare the samples used in the physical
modelling tests, the characterization of the materials, equipment and instrumentation
used.

Chapter 4 presents the results of lateral load tests and the discussion of the results.

And in Chapter 5, conclusions and suggestions for future research are presented.

4
2. LITERATURE REVIEW

In this chapter, a literature review is made on the subject studied, addressing the main
types of offshore wind towers foundations and the loads to which they are subjected.
Besides, it deals with the design methodology for foundations subjected to lateral loading,
detailing the ultimate limit state method and the p-y curves, widely used in the practice
of the oil and gas industry. And finally, it provides details about the physical modelling
in the geotechnical centrifuge and its use in the study of monopile foundations.

2.1 FOUNDATIONS FOR OFFSHORE WIND TURBINES

The world's first offshore wind farm was built in Vindeby, Denmark, in 1991, and
had 11 turbines with a capacity of 450 kW each. Vindeby was decommissioned in 2017,
but since its development, offshore wind energy has become a worldwide industry.
Today, a single offshore wind turbine has more than twice the production capacity of the
entire Vindeby complex (4C Offshore, 2016). Figure 2 shows the evolution of the
capacity and, consequently, of the size of wind turbines.

Figure 2 – The increasing size of offshore wind turbines, 1985-2016 (IRENA, 2016).

5
The largest wind farm in the world, Hornsea One, is already partially operating, with
50 of its 174 turbines in operation. When completed, it will have the generation capacity
to supply up to 1 million houses in the United Kingdom (CNN, 2019).

As the capacity and size of offshore turbines grow, there is a need for more robust
foundations to support these structures. The foundation is responsible for keeping the
structure stable so that it can operate safely and efficiently. The choice of the type of
foundation for offshore wind turbines depends on some variables and can vary greatly
according to the location of the installation.

The costs with the foundation can have significant values within the total cost of an
offshore wind turbine (BYRNE and HOULSBY, 2003; BLANCO, 2009). Blanco (2008)
indicates that the costs with foundations can reach 19%, based on reports of wind farms
in the United Kingdom (Figure 3), but these values may vary according to the project and
country. This shows the need to search for more efficient projects, seeking to reduce total
costs with the foundation.

Figure 3 – Estimated costs of an offshore wind farm (BLANCO, 2009)

The foundations for offshore wind turbines can be of different types, the most used
are gravity, suction, monopile, tripod, jacket and floating (Figure 4).

6
Figure 4 – Types of foundations for offshore wind turbines (WORLD STEEL, 2017)

Gravity-base foundations are made of reinforced concrete and are stabilized by their
self-weight. They are generally used for water depths up to 25 m and do not penetrate the
soil. In terms of cost, they can be advantageous for the materials used, but they are limited
when it comes to larger water depths, which ends up making the foundation bigger and
heavier and, therefore, increasing the cost and difficulty in transportation (SALEEM,
2011).

The suction bucket foundations work by applying suction to the foundation, where
the hydrostatic pressure and the weight of the structure cause it to penetrate the soil. They
are relatively quick to install, as the penetration into the soil is not great, but they can
suffer a high effect of scour around the foundation, as they have a large diameter.

The tripod foundations have three piles or three suction-type foundations and can be
installed in water depths of up to 50 m. They have good lateral stability, as they work
with a combination of tension and compression. Transport and installation are more
complex, which ends up reducing the use of this type of foundation for offshore wind
turbines.

Jacket foundations are widely used in the oil and gas industry, and their use for wind
turbines has benefited greatly from this experience. It is a truss structure, with great

7
stability, which can be installed in large water depths. However, the manufacture of
jackets can be complex, as they have several components welded together, which
increases the cost and time required.

Floating-type foundations can be used for turbines installed at large water depths. The
turbine and the foundation can be mounted onshore, and the entire structure is towed to
the operating site. Stability, however, is the main concern, and several studies have been
carried out to make this type of foundation feasible (SALEEM, 2011).

Monopiles are cylindrical structures, usually made of steel, and are installed on the
seabed using a hydraulic hammer or by vibration. The diameter (D) is up to 8 m, and the
relationship between diameter and embedded length (L) is sometimes as low as 3 (L/D =
3). A transition piece (Figure 5), usually made of steel, makes the connection between the
foundation and the turbine, being used as access for maintenance of the tower.

Figure 5 – Transition piece for monopiles (Adapted from IXWIND, 2020)

Negro et al. (2017) present a database with information on 30 offshore wind farms in
Europe, with monopile foundation. It includes the main characteristics such as the water
depth, the diameter of the monopile, length, and weight of the structure. From this
database, equations are proposed so that, once the diameter of the monopile is known,

8
length and weight can be estimated, which will then be verified with detailed geotechnical
and structural design.

As they are simple to manufacture, and transport and installation are quick when
compared to other types of foundation, the monopile is the type of foundation most used
worldwide in offshore wind farms. In Europe, in 2018, approximately 75% of foundations
were of the monopile type (WIND EUROPE, 2019). In the projects to be developed in
Brazil, however, the adequacy of this type of foundation is still questioned due to the
geological characteristics of the Brazilian coast, particularly in the northeast of the
country.

2.1.1 LOADS ACTING ON OFFSHORE WIND TURBINES

Despite being structures installed in an offshore environment, the loads acting on wind
turbines differ greatly from the loads operating in the typical structures of the oil and gas
industry. While in offshore structures the vertical load resulting from the structure's self-
weight is the most expressive, in wind turbines what governs are the lateral loads arising
from winds, waves, currents and rotation of the blades that generate bending moments of
great magnitude.

In 2018, the average rated capacity of newly installed turbines in Europe was 6.8 MW
(WIND EUROPE, 2019). But the values for loads acting on the turbines presented on the
literature are for turbines with a capacity of 3.5 - 5 MW (considered as small turbines).
Considering a wind turbine with the dimensions shown in Figure 6 (a), Byrne and
Houlsby (2015) differentiate the loads acting on the offshore wind towers (Figure 6a) and
on the platforms (Figure 6b). For an offshore platform, the vertical load is around 200
MN, due to its self-weight, and the horizontal load due to waves reaches 25 MN. While
a wind turbine, between 90-110 m high, installed at a depth of 20 to 50 m, is subject to
the following loads:

1. vertical load relative to self-weight, which can vary from 6 - 10 MN;


2. wind load acting on the turbine axis: 1-2 MN;
3. combination of current and waves acting at 10 m above the seabed for 10 seconds:
3 ± 6 MN.

9
(a)

(b)

Figure 6 – Typical loads acting on an offshore wind turbine (a) (BYRNE and HOULSBY, 2015) and an
oil platform (b) (BYRNE and HOULSBY, 2003)

10
The loading values suggested by Byrne and Houlsby (2015) are based on typical
values of wind and wave in the United Kingdom. Other authors also propose typical loads,
but for turbines with different capacities and installed in different locations. Lesny and
Wiemann (2005), when detailing aspects of offshore wind energy in Germany, provide
load values for a 5 MW turbine, with typical Baltic Sea wind and wave values. Le Blanc
et al. (2010) suggest loads for a 2 MW turbine. The values proposed by these authors are
shown in Table 1.

Table 1 – Typical loads on offshore wind turbines

Lesny and Le Blanc et al.


Byrne and
Wiemann (2010) (Ultimate
Houlsby (2015)
(2005) Limit State)

Turbine Capacity [MW] 5 2 3.5 – 5

Vertical Load [MN] 35 5 6 – 10

Horizontal Load [MN] 6 4.6 4-8

Bending Moment [MN.m] 280 95 90 - 180

Loads acting on wind turbines are predominantly cyclic, acting throughout the
lifetime of the structure, with unique characteristics in terms of magnitude, frequency and
number of applied cycles (ARANY et al., 2016; NIKITAS et al., 2016). Figure 7
illustrates these loads. It is important to know these characteristics since the structure
must be designed to avoid the resonance effect.

Wind intensity varies widely in space and time, with greater speed at greater heights.
Wind turbines start operating in winds with a speed of 3-4 m/s and stop their operation at
25 m/s. It is represented through the Kaimal or von-Karman spectrum, which includes the
complex characteristics of turbulence due to gusts of wind (TEMPEL, 2006), and will not
be detailed in this work. Le Blanc (2009) indicates that the frequency of wind loading is
not greater than 0.1 Hz.

11
The waves are also described through spectra, the most known being those of Pierson-
Moskowitz and JONSWAP. The frequency for this loading is in the range of 0.05-0.5 Hz,
and extreme events occur at 0.07-0.14 Hz (LE BLANC, 2009).

When in operation, the turbines are also subjected to the loads generated by the
rotation of the rotor and the blades, called 1P and 3P, respectively. The 1P load is caused
by rotor mass and aerodynamic imbalances and the frequency is equal to the rotor
frequency, between 0.17-0.33 Hz. The 3P load (for a 3-bladed turbine and 2P for a 2-
bladed turbine), is caused by the rotation effect of the blades, and its frequency is between
0.5-1 Hz.

Figure 7 - External loads acting on an offshore wind turbine, together with their typical waveforms.
(NIKITAS et al., 2016)

Spectral densities of wind and waves specific to each location can be obtained through
available measurements, meteorological databases or numerical models. These
frequencies, together with the frequencies 1P and 3P are presented in Figure 8. The
structure must be designed so that its first natural frequency does not coincide with the
wave, wind, 1P and 3P frequencies. Ideally, the natural frequency (f1) should be between
the frequencies 1P and 3P, where the structure is classified as soft-stiff. If f1 is below the

12
1P frequency, it would be a very flexible structure, more economic because it uses less
material, but is more susceptible to fatigue effects. If f1 is greater than 3P, its
classification would be stiff-stiff, being a very heavy structure and economically
unfeasible.

soft-soft

soft-stiff

stiff-stiff

Figure 8 – Frequency of loads acting on an offshore wind turbine (Adapted from LE BLANC, 2009).

According to DNV (2016), the design of foundations must cover four loading
conditions: ultimate limit state (ULS), which corresponds to the maximum resistance;
fatigue limit state (FLS), which corresponds to failure due to the effect of dynamic loads;
accidental limit state (ALS) which involves the maximum load capacity for accidental
loads or post-accidental integrity for damaged structures; and service limit state (SLS)
that corresponds to the tolerance criteria applicable to normal use.

In general, the design of the structures is given predominantly by SLS, FLS and ULS,
due to the maximum values of rotation and deflection allowed for the foundation in the
service state, which is specified by the project.

2.2 LATERALLY LOADED PILES

The foundations are designed to withstand vertical and horizontal loads and
bending moment. Horizontal loads can be from inclined loads, ground pressure, winds,
waves, earthquakes, among others. (FAN and LONG, 2005).

13
Velloso and Lopes (2010) mention three important aspects in the design of
structures subjected to lateral loads: the capability of the soil to resist the stresses
transmitted by the foundation (stability), if the displacements and rotation at the top of
the pile are compatible with the structure, and the structural design of the element, where
internal efforts must be foreseen.

Fan and Long (2005) classified the methods to analyse laterally loaded piles into
5 categories: 1) Limit state method; 2) Reaction coefficient method; 3) p-y curves; 4)
Theory of Elasticity; 5) Finite element method. The Limit State Method analyzes the
condition of the ultimate soil resistance and some methods for the calculation will be
presented in item 2.2.2. In the reaction coefficient method, the soil is modelled as linear
springs by the Winkler hypothesis. The p-y curves model the soil as non-linear springs
and were constantly used in practice. Its principles and limitations will be discussed in
item 2.2.1. The Theory of Elasticity considers the continuity of the soil for calculating the
ultimate load. The Finite Element Method (that is becoming the new standard in practice
today) considers the continuity and non-linearity of the soil, in addition to the soil-
structure interface through computational methods and constitutive models of the
materials involved.

2.2.1 P-Y CURVES


2.2.1.1 P-Y CURVES FOR PILES

The reaction of the soil to the lateral loading is a complex problem for foundations
subjected to this type of loading. Assuming a vertical pile subjected to a horizontal load
P applied above the soil surface (Figure 9), there is the horizontal displacement y
corresponding to the soil reaction.

14
Figure 9 – Pile subjected to lateral load (VELLOSO and LOPES, 2010)

While the pile is not subjected to loading, the stress distribution is uniform
throughout the section (Figure 10a). When the lateral load is applied, the stresses start to
be distributed asymmetrically, being higher in front of the pile and lower in the back
(Figure 10b). This non-uniform stress distribution is represented by the reaction of the
soil p acting along the diameter (Figure 10c).

Figure 10 – Soil pressure around a pile (KIRKWOOD, 2015)

The differential equation presented in Equation 1, proposed by Hetenyi (1946),


provides the deflection of the pile as a function of depth, and from equilibrium equations
and free body diagram, it is possible to find equations that describe the shear force,
bending moment and slope:

15
𝑑4 𝑦 𝑑2 𝑦
𝐸𝑝 𝐼𝑝 𝑑𝑧 4 + 𝑉 𝑑𝑧 2 + 𝐸𝑝𝑦 𝑦 = 0 1

where:

y – lateral deflection of the pile

𝐸𝑝 𝐼𝑝 – pile bending stiffness

𝑉 – axial load at the top of the pile

𝐸𝑝𝑦 – soil pressure module based on p-y curves

z – depth

Using a system of non-linear and independent springs to represent soil stiffness


(Figure 11), it is possible to perform load-deflection analysis and solve Equation 1. The
method must correctly represent the mobilization of soil resistance against lateral loading.
Besides, it is an analysis of the soil-pile system and not just a property of the soil.

16
Figure 11 – Soil pressure represented by Winkler springs (ROCSCIENCE, 2018)

The p-y curve methodology has been successfully adopted for pile foundations of
structures in the oil and gas industry. There are recommendations for obtaining p-y curves
for different types of soil (clay, sand, rock and stratified soils), that were proposed based
on field tests of instrumented piles. All models provide the initial stiffness (or initial slope
of the curve), the ultimate capacity (asymptote to the p-y curve), and the shape of the
curve between the points of the initial response and the ultimate capacity (DASGUPTA,
2017). In this work, two models for p-y curves in sand will be presented: Reese et al.
(1974) and API (2011). There is also the DNV (2016), similar to the API (2011), the
difference being that DNV (2016) recommends validating the method through numerical
modelling for diameters greater than 1 m.

Below is a list of variables used in the equations:

D = pile diameter (m)

z = depth below the soil surface (m)

𝛾′ = effective unit weight (kN/m3)

17
𝜑 = friction angle of soil

Reese et al. (1974) initially proposed a method, to which others made their due
contributions. The proposed curve has the shape shown in Figure 12.

Figure 12 – p-y curve for sand by the Method of Reese et al., 1974 (ROCSCIENCE, 2018)

The values of 𝑝𝑢 and 𝑝𝑚 are calculated using the smallest value of 𝑝𝑠 (Equation 2
and 3), multiplied by the coefficients A and B obtained from Figure 13.

𝑝𝑢 = ̅̅̅
𝐴𝑠 𝑝𝑠 ; 𝑝𝑚 = 𝐵𝑠 𝑝𝑠

0𝐾 𝑡𝑎𝑛 𝜑 sin 𝛽 tan 𝛽


𝑝𝑠 = 𝛾𝑧 [tan(𝛽−𝜑) + tan(𝛽−𝜑) (𝐷 + 𝑧 tan β tan 𝛼) + 𝐾0 𝑧 tan 𝛽 (tan 𝛽 sin 𝛽 −
cos 𝛼

tan 𝛼) − 𝐾𝑎 𝐷] 2

𝑝𝑠 = 𝐾𝑎 𝐷𝛾𝑧 (𝑡𝑎𝑛8 𝛽 − 1) + 𝐾0 𝐷𝛾𝑧 tan 𝜑 𝑡𝑎𝑛4 𝛽 3

where

𝜑 𝜑 𝜑
𝛼= ; 𝛽 = 45° + ; 𝐾0 = 0.4; 𝐾𝑎 = 𝑡𝑎𝑛2 (45° + )
2 2 2

18
Figure 13 – Coefficients of soil resistance versus depth (REESE and VAN IMPE, 2014)

The method proposed by the API (2011) uses a hyperbolic tangent function to
compute the curve. The main difference concerning the method of Reese et al. (1974) is
the initial reaction module and the shape of the curve, shown in Figure 14 (DASGUPTA,
2017).

Figure 14 – p-y curve for sand by the API Method (ROCSCIENCE, 2018)

19
The p-y curve is given by Equation 4:

𝑘 𝑧
𝑝 = 𝐴 𝑝𝑢 tanh ( 𝐴𝑖𝑛𝑖𝑝 𝑦) 4
𝑢

where

𝑧
A = factor for static load (3 − 0.8 𝐷) or cyclic (0.9)

𝑘𝑖𝑛𝑖 = initial reaction module (abacus in Figure 16)

The value of 𝑝𝑢 is given by the lowest value between 𝑝𝑢𝑠 , for small depths
(Equation 5), and 𝑝𝑢𝑑 , to great depths (Equation 6). The values of 𝐶1 , 𝐶2 and 𝐶3 are
provided by the abacus in Figure 15.

𝑝𝑢𝑠 = (𝐶1 𝑧 + 𝐶2 𝐷) 𝛾 ′ 𝑧 5

𝑝𝑢𝑑 = 𝐶3 𝐷𝛾 ′ 𝑧 6

Figure 15 – Values for C1, C2 and C3 for sands in API (ISENHOWER and WANG, 2016)

20
Figure 16 – Values for 𝑘𝑖𝑛𝑖 for sands in API (MEYER and REESE, 1979)

2.2.1.2 CONSIDERATIONS FOR USING P-Y CURVES FOR MONOPILES

The semi-empirical methodology used in the p-y curves was developed from tests
on flexible piles, with a diameter of up to 2 m (KLINKVORT, 2012), and has also been
used in the design of monopiles. However, the adequacy of this use is questioned by
several authors (ACHMUS et al., 2005; BRODBAEK et al., 2009; KLINKVORT, 2012,
ABADIE, 2015; BAYTON and BLACK, 2016), mainly due to the difference between
the mechanisms of load transfer and failure, because of differences in stiffness of
monopiles and flexible piles. Figure 17 shows the difference between rigid pile behaviour
(a) and flexible pile behaviour (b).

DNV (2016), when discussing the application of these curves, highlights that for
piles with a diameter greater than 1.0 m, it is recommended for the design to be validated
with numerical analysis.

21
Poulos (1982) establishes the following expression (Equation 7) to determine soil-
pile relative stiffness (𝐾𝑟 ), and indicates the limits to which the foundation has rigid or
flexible behaviour:

𝐸𝑝 𝐼𝑝 > 0.208 rigid


𝐾𝑟 = 𝐸 4
{ 7
𝑆𝐿 𝐿 < 0.0025 flexible

where

𝐸𝑝 𝐼𝑝 = pile bending stiffness

𝐿 = pile length

𝐸𝑆𝐿 = Young's modulus of the soil, given by Equation 8:

𝐸𝑆𝐿 = 2(1 + 𝜈)𝐺𝑚á𝑥 8

where 𝐺𝑚á𝑥 is the maximum shear modulus of the soil and 𝜈 is the Poisson ratio of the
soil.

(a) (b)

Figure 17 – Difference in the behaviour of a rigid pile (a) and flexible pile (b) (BRODBAEK et al., 2009)

Figure 18 presents a graph with the relative stiffness versus the ratio between the
length (L) and the diameter (D) of a pile for a variety of projects relevant to UK offshore
wind farms, presented by Abadie et al. (2019). Three data sets are highlighted: (a)
monopiles on sand; (b) monopiles on clay; and (c) piles used in the 1960s and 1970s for
the development of p-y methods (for example, API, 2010; DNV, 2016).

22
Figure 18 - Kr plotted against L/D ratio for full-scale offshore wind farms (OWF) in the United Kingdom,
and piles used for the development of p-y curves. (ABADIE et al., 2019)

Considering the typical dimensions of a monopile in dense sand, the classification


of relative stiffness is closer to the condition of a rigid pile, that still may present some
type of bending (KIRKWOOD, 2015).

Also, as mentioned in item 2.1.1, there are significant differences between the
loads imposed on the piles of offshore platforms and the wind turbines. Thus, cyclic
lateral loads are not considered in the methodology suggested by DNV (2016) and API
(2011) (ABADIE, 2015). Despite proposing a load reduction factor A, which
differentiates static load from the cyclic load, according to Equation 4, the API method
computes the same soil reaction value regardless of the number of cycles, that is, it ignores
the change in soil stiffness with an increasing number of cycles, despite being a relevant
factor, as exposed by Lombardi et al. (2013) and Nikitas et al. (2016).

Many authors have carried out researches to evaluate the transfer of this
methodology to the design of offshore wind tower foundations. Klinkvort (2012), after

23
centrifuge tests on rigid piles, suggests changes in the methodology used for API (2011)
and DNV (2014) p-y curves. The author proposes the use of a hyperbolic formulation
(Equation 9) for the curves, developed by Kondner (1963) and used by Kim et al. (2004).

𝑦
𝑝= 1 𝑦 9
+
𝑘𝑖𝑛𝑖 𝑧 𝐴 𝑝𝑢

where

𝑘𝑖𝑛𝑖 = initial stiffness of the p-y curve.

The author suggests a new empirical formulation for 𝑘𝑖𝑛𝑖 (Equation 10).

𝑘𝑖𝑛𝑖 = 100 𝐾𝑝 𝛾 ′ 10

where

𝐾𝑝 = Rankine’s passive earth pressure coefficient

Another suggestion of Klinkvort (2012) is the change in empirical value A,


applied to the ultimate soil resistance in the API method (Equation 10). This change is
suggested to represent the difference in the failure mechanism between rigid and flexible
piles.

1,1 𝑧
𝐴 = 0,9 + {1 + tanh (9 − 3 )} 11
2 𝑏

Byrne et al. (2017) also describe a new analysis method for lateral loading on
monopiles, the result of a joint industry project, called PISA (Pile Soil Analysis). The
PISA project, based on an extensive theoretical study and numerical models validated
from field tests, proposes a broader approach to accurately detail the behaviour of laterally
loaded piles. By doing that, it eliminates the conservative aspect of the already known
methodologies, proposing a more efficient design. The PISA model will be described in
details in item 2.2.1.3.

24
2.2.1.3 PISA DESIGN MODEL

Burd et al (2019a) describe one-dimensional (1D) computational method used to


analyse and design monopiles in sand, submitted to monotonic loading.

The model is calibrated from a 3D numerical modelling, with sand parameters


from Dunkirk (France). The proposed model is applicable for sand with relative density
between 45% ≤ Dr ≤ 90%. The monopile is subjected to a lateral load H, applied at a
height h of the seabed (Figure 19). In this item, the symbols used by Burd et al. (2019)
will be adopted, which differs slightly from that used in the rest of this dissertation.

Figure 19 - Geometry and parameters considered for the design problem (BURD et al., 2019a)

The four reaction components are: distributed load p, distributed moment m,


lateral force Hb and moment Mb at the base of the pile (Figure 20b). The monopile is
represented by Timoshenko beam theory, which allows shear strains to be incorporated
into the analysis. Soil reactions are applied using a generalized approach to the Winkler
hypothesis, where the force and moment of reaction are related only to the displacement

25
(or rotation) of the pile. The reaction curves, therefore, depend on the magnitude of the
pile displacement. In the PISA model, when the values of m, Hb and Mb are equal to zero,
is reduced to the traditional p-y approach. However, these values increase significantly
as the L/D ratio value decreases, which is the case for monopiles.

Figure 20 - PISA design model: (a) idealisation of the soil reaction components acting on the pile (b) 1D
finite-element implementation of the model (BURD et al., 2019a)

The model used for the reaction curves is based on dimensionless parameters for
load, displacement, moment and force at the base of the pile (Table 2). Using an algebraic
function, the model realistically computes soil reactions for both small and large
displacements. The function used is represented by Equation 12, and its format is shown
in Figure 21.

26
Table 2 – Dimensionless forms for the soil reaction curves (BURD et al., 2019a)

Normalised variable Dimensionless form


𝑝
Distributed lateral load 𝑝̅ ′
𝜎𝑣𝑖 𝐷
𝑣 𝐺0
Lateral displacement 𝑣̅ ′
𝜎𝑣𝑖 𝐷
𝑚
Distributed moment 𝑚
̅
|𝑝| 𝐷
𝜓 𝐺0
Pile cross-section rotation 𝜓̅ ′
𝜎𝑣𝑖
𝐻𝑏
̅̅̅̅𝑏
Base horizontal load 𝐻 ′
𝜎𝑣𝑖 𝐷2
𝑀𝑏
Base moment ̅̅̅̅
𝑀𝑏 ′
𝜎𝑣𝑖 𝐷3

𝑦̅ 𝑥̅ 2 𝑦̅ 𝑥̅ 𝑘 𝑦̅
−𝑛 ( ̅ − ) + (1 − 𝑛) ( ̅ − ) ( ̅ − 1) = 0 12
𝑦𝑢 𝑥̅ 𝑢 𝑦𝑢 𝑦̅𝑢 𝑦𝑢

where 𝑥̅ signifies a normalised displacement or rotation variable and 𝑦̅ signifies the


corresponding normalised soil reaction component, formulated in term of the
dimensionless forms in Table 2. The normalised soil reactions can be determined
explicitly from the normalised displacements by Equation 13

2𝑐
𝑦̅ = ̅̅̅
𝑦𝑢 ; 𝑥̅ ≤ ̅̅̅𝑦
𝑥𝑢 ̅ = ̅̅̅;
𝑦𝑢 𝑥̅ > ̅̅̅
𝑥𝑢 13
−𝑏+ √𝑏2 −4𝑎𝑐

where

𝑎 = 1 − 2𝑛 14

𝑥̅ 𝑥̅ 𝑘
𝑏 = 2𝑛 ̅̅̅̅ − (1 − 𝑛) (1 + 𝑦̅ ) 15
𝑥 𝑢 𝑢

𝑥̅ 𝑘 𝑥̅ 2
𝑐= (1 − 𝑛) − 𝑛 16
𝑦̅𝑢 𝑥̅ 𝑢 2

27
The parameter k represents the initial slope, ̅̅̅
𝑦𝑢 is the ultimate value of the
normalised soil reaction and ̅̅̅
𝑥𝑢 is the normalised displacement (or rotation) at which this
ultimate value of soil reaction is reached. The parameter n (0 ≤ 𝑛 ≤ 1) determines the
shape of the curve, for extreme values n = 0 and n = 1, the function reduces to the bilinear
form illustrated in Figure 21b.

Figure 21 – Conic function adopted for the parametric soil reaction curves (a) conic form (b) bilinear
form (BURD et al., 2019a)

The calibration of the 1D model is done from finite element numerical modelling,
using the bounding surface model for the constitutive model, from the theory of critical
states. The soil parameters for the model were determined by Taborda et al. (2014) for
the Dunkirk sand. The parameters were determined only for monotonic loading and not
for cyclic loading. The calibration was done for piles with diameters of 5, 7.5 and 10
meters, varying L/D, h/D and thickness, configuring 11 different piles. The ICFED
software was used for the numerical modelling, with a total of 38 analyses. The numerical
model does not consider stress and state changes due to the installation of the monopile
but considers it as “wished in place”.

A comparison between the 1D model and the 3D numerical modelling shows a


good approximation between these two, as shown in Figure 22. In this figure, the case P
indicates an analysis where only the distributed lateral load terms are included, and the

28
case MP indicates the analysis where only the distributed lateral load and distributed
moment terms are included.

Figure 22 – Performance of the 1D model, for Dr = 75 (a) pile L/D = 2 and (b) pile L/D = 6 (BURD et al.,
2019a)

A process of calibration and optimization of the model is then performed,


described in detail by the authors.

Design examples for application of the method are presented and Figure 23 shows
the results for one pile, with sand at relative density of 85%, where the 1D and 3D models
are practically coincident.

Figure 23 – Comparison between computed responses determined from the 1D model and equivalent 3D
finite-element analyses for Dr = 85% (a) ultimate response (b) small displacement response and (c)
bending moment profile with depth- Solid line represent 3D finite-element results and dashed line the 1D
model (BURD et al., 2019a)

29
Considering the results obtained, the authors conclude that the proposed PISA
model provides a fast methodology for the design of monopiles. The authors presented
the method developed for sand and is complemented by Byrne et al. (2020) for clays. The
predictions made from the model were practically coincident with the equivalent finite
element modelling performed.

Also included in the PISA project, Zdravkovic et al. (2019) presents the ground
characterisation for Cowden (stiff glacial clay) and Dunkirk (dense sand), the sites
adopted for the field tests of the project. And Burd et al. (2019b) describe the
experimental set-up for the field tests, such as the procedures and the interpretation of
data.

It is out of the scope of this work to assess the PISA project.

2.2.2 ULTIMATE LIMIT STATE METHODS

In addition to the limitations explained in item 2.2.1.1, another point is the


determination of the ultimate soil resistance (𝑝𝑢 ) by the API method. As shown by
Equations 5 and 6, the method differentiates the type of failure mechanism for small and
large depths, based on the propositions of Reese et al. (1974).

For small depths, the failure mechanism would be that of a wedge forming ahead
of the pile (Figure 24). At great depths, a two-dimensional failure mechanism is proposed,
where the soil flows around the pile rather than forming a wedge, as indicated in Figure
25.

30
Figure 24 - Failure at small depths (REESE and VAN IMPE, 2014)

Figure 25 - Failure at great depths (REESE and VAN IMPE, 2014)

According to this proposition, for monopiles, the soil failure is predicted by a


wedge mechanism for all depths. However, this is not possible, as the point of rotation
restricts zero ground displacement. Therefore, a soil wedge cannot develop below the
depth of the rotation point (KIRKWOOD, 2015).

Some theoretical and empirical formulations are presented here to calculate the
ultimate resistance of the soil.

The method of Broms (1964) is based on analyzes made by the author on clays in
drained condition and on sands, for different failure mechanisms (for short or long piles,
31
with free or impeded tops). For a short pile with a free top, subjected to a lateral load at
an eccentricity ex (Figure 26), which is the case of monopiles, the author suggests the
expression presented in Equation 17.

𝑝𝑢 = 3𝐷𝛾 ′ 𝑧𝐾𝑝 17

Figure 26 – Soil reaction at failure: Broms Method (1964)

Zhang et al. (2005) propose a simple method to calculate the ultimate lateral
resistance (Equation 18), composed of the stresses acting on the front and sides of the pile
(Figure 27). As it is an analysis that considers the ultimate resistance of the soil (and not
the pile), it can be applied for both rigid and flexible piles.

𝑝𝑢 = (0,8 𝑝𝑧 + 𝜏𝑧 ) 𝐷 18

32
where

𝑝𝑧 = 𝐾𝑝2 𝛾′𝑧 19

𝜏𝑧 = 𝐾ℎ 𝛾 ′ 𝑧 𝑡𝑎𝑛 𝛿 20

and

𝐾ℎ is the lateral earth pressure coefficient

𝛿 is the friction angle of the soil-pile interface

Figure 27 - Stress distribution around a laterally loaded pile (ZHANG et al., 2005)

Other methodologies to calculate the ultimate lateral load can be found in


Meyerhof (1981), Hansen (1961) and Prassad and Chari (1999) and will not be detailed
in this work.

33
2.3 PHYSICAL MODELLING IN CENTRIFUGE

Physical modelling is a tool used in several fields of engineering, that allows to study
on a reduced or increased scale various physical phenomena. There are, for example,
hydraulic models, in experimental tanks and channels, and aerodynamic models of
airplanes and automobiles.

According to Lobo Carneiro (1996), “if two physical processes are similar, it is
possible to predict the behaviour of one of them when the behaviour of the other is known.
In experiments using models, the two similar physical processes are the prototype and its
model; in this case, the model is used because it is easier to test in the laboratory than to
test the prototype directly”.

It is important, therefore, to ensure physical similarity between the processes that


occur in model and prototype, to obtain representative models, whose results can be used
to predict the prototype behaviour. This can be achieved through the so-called similarity
conditions, which are determined through equations that control the phenomenon and by
the boundary conditions (SARAMAGO, 2002).

Among the scaled models, 1g and centrifuge models are important tools that can assist
in solving various engineering problems. In Geotechnical Engineering they are quite
efficient for situations where soil conditions are considered difficult, when the soil's
constitutive models are not well defined or when the predicted loads are unusual or very
extreme. Centrifuge modelling, specifically, has the purpose to get similar soil stress in
the model as in prototype.

The centrifuge modelling consists on testing a model scaled to 1/N of a prototype in


the increased gravitational field, generated by the geotechnical centrifuge. Gravity is
increased by the same N factor relative to Earth's gravity - referred to as 1 g
(MADABHUSHI, 2014).

This soil sample subjected to centrifuge acceleration has a stress-free surface and a
stress level that increases with depth, at a rate related to soil density and with the created
acceleration field (TAYLOR, 1995). If the same type of soil is used for the model and the
prototype, and the model is correctly designed, then at a depth hm, the model is subjected
to the same stress level as the prototype at a depth hp, where hp = N.hm, where N is the

34
scale factor (Figure 28). This is the basic principle of centrifuge models and the
demonstration of this expression can be found in Taylor (1995).

Figure 28 - Correspondence between the inertial stress in a model and the gravitational stress in the
prototype (TAYLOR, 1995).

The scaling laws are correspondences that relate the behaviour of the centrifuge
model with that of the prototype. They are necessary to guarantee the similarity of stresses
between model and prototype, as mentioned above. These relationships can be deduced
through dimensional analysis (MADABHUSHI, 2014; NG, 2014), and Table 3 presents
the main scaling laws between prototype and model.

Table 3 – Centrifuge scaling laws

Parameter Model / Prototype Ratio Unit


Length 1/N m
Area 1/N2 m2
Volume 1/N3 m3
Mass 1/N3 Nm-1s2
Stress 1 Nm-2
Strain 1 -
Force 1/N2 N
Bending Moment 1/N3 N.m
Frequency N s-1

35
2.3.1 CENTRIFUGE MODELLING LIMITATIONS

Despite its advantages, centrifuge modelling has some limitations that must be
well understood in order not to interfere with the analysis that will be made from the
results obtained in the physical tests (MADABHUSHI, 2014). Among these limitations,
two will be discussed below, which are: variation in the gravity field and scale effects on
the soil.

2.3.1.1 VARIATION IN GRAVITY FIELD

One of these limitations is related to the variation in the gravitational field created
by the centrifuge. In general, the Earth's gravitational field is considered constant with
depth. That is, the variation in the acceleration of gravity with depth is ignored when
calculating the effective stresses in the field. In centrifuge modelling, however, the
models are tested on a reduced scale and a small change in the rotation radius changes the
gravitational field, as shown in Equation 21.

𝑁 𝑔 = 𝑟 𝜃̇ 2 21

where

N is the scale factor, g is the acceleration of gravity, r is the radius from the center of
rotation and θ is the angular speed.

The vertical stress at a depth z in the model is given by Equation 22

𝐻
(𝜎𝑣 )𝑚𝑜𝑑𝑒𝑙 = 𝜌 𝜃̇ 2 𝐻 (𝑟𝑡 + ) 22
2

where

𝑟𝑡 is the distance from the center of rotation to the top of the model, as shown in Figure
29, H is the thickness of the soil layer and 𝜌 is the density of the soil.

36
Figure 29 - Prototype and centrifuge model of a horizontal soil layer (MADABHUSHI, 2014)

The vertical stress on the model and the prototype will be equal at the equivalent point
𝑟𝑒 . The angular speed θ must be calculated by Equation 23 to guarantee same stresses at
that point. Figure 30 shows a comparison between the stress distribution along the depth
for the prototype and model.

2𝐻
𝑟𝑒 = 𝑟𝑡 + 23
3

Figure 30 - Difference in stress distribution with a depth between model and prototype (MADABHUSHI,
2014)

37
2.3.1.2 GRAIN SIZE EFFECTS

In centrifuge models, the prototype dimensions are scaled by an N factor.


However, the particle size is not being scaled down by this same factor. And this is a
point that is commonly criticized in centrifuge modelling.

Madabhushi (2014) shows that, if the soil particles were subjected to the same
scale factor, this would lead to the use of a type of soil in the model that would be
completely different from the soil considered in the prototype.

For example, to model a prototype in fine sand, with an average particle diameter
of 0.4 mm, at a centrifuge acceleration of 100 g, in the model the particles should have a
diameter of 0.004 mm if the N factor was also used for the soil. However, this would lead
to the use of fine clay, which mineralogical constitution and behaviour in terms of strength
are very different from that of sand.

Thus, to simulate the same behaviour in the prototype and model, the same soil
must be used to maintain the stress-strain relationship. However, the continuity of the soil
must be guaranteed for all types of situations analysed. Each geotechnical problem
requires different theories to guarantee this continuity.

Garnier et al. (2007), using data from centrifuge tests from different authors,
brings together the main similarity relationships for different models. For piles subjected
𝐷
to lateral loading, no significant grain size effect was observed for a ratio > 45
𝑑50
𝐷
(NUNEZ et al., 1988) and > 60 (REMAUD, 1999) (where D is the diameter of the
𝑑50

pile and 𝑑50 is the average particle diameter).

Klinkvort et al. (2012), however, points out that these relationships are based on
tests made on long and flexible piles, and should be used with caution in the centrifuge
modelling of short and rigid piles, as is the case with monopiles.

More details on the principles of physical modelling in centrifuge can be found in


Taylor (2005) and Madabhushi (2014).

38
2.3.2 PHYSICAL MODELLING OF MONOPILES

Tests on full-scale monopiles are difficult to perform for technical and economic
reasons. Over the past few years, several authors have studied the effect of horizontal
loads on monopiles through physical modelling, both in centrifuge and 1g.

Le Blanc et al. (2010) and Abadie et al. (2019), performed cyclic tests at 1g and
present a model to predict the accumulation of displacement. Klinkvort (2012), Kirkwood
(2015) and Choo and Kim (2016) also performed cyclic tests but using a geotechnical
centrifuge. Futai et al. (2018) analysed the dynamic behaviour of piles and monopiles
also from centrifuge tests.

Kirkwood (2015), Zhu et al. (2016), Choo and Kim (2016) and El Haffar (2018)
compared the p-y curves proposed by the design codes with those obtained from
centrifuge tests.

Klinkvort et al. (2012) discuss three important aspects of the centrifuge modelling of
monopiles from a series of tests on different models. These aspects were: scale effects of
soil grains, non-linearity of the gravity field (discussed in item 2.3.1.1), and installation
effects.

The authors conclude that the relationship between the diameter of the model and the
𝐷
average diameter of the particles (𝑑 ) must be greater than 88 for a rigid pile, to avoid
50

scale effects. This value is much higher than those indicated in item 2.3.1.2.

Regarding the installation effects, Randolph et al. (2001) conducted a study in a


centrifuge to understand these effects on the response of laterally loaded piles. Four
installation methods were used: placing the pile in the centrifuge strongbox before adding
the soil, jacking the pile at 1g, and jacking or driving the pile at 160g. The authors
conclude that the ultimate lateral capacity of the piles differs depending on the type of
installation, as shown in Figure 31 where three installation methods are compared. Most
of the literature studies are carried out using the 1g installation method (e.g., LEHANE
et al., 2014; ZHU et al., 2016; FUTAI et al., 2018).

39
Figure 31 - Load-displacement curves for different installation methods (RANDOLPH et al., 2001)

Klinkvort et al. (2018) present the main parameters to be considered in the physical
modelling of monopiles installed in sand, highlighting those in which further studies are
needed. According to the authors, dimensional analysis of the monopile response shows
that the lateral resistance of these structures is a function of the parameters presented in
Equation 24, which must be constant between model/prototype.

P 𝑦 L Le Ip E G M kT lt D R
3 = f (D , D , D , D , γ' pD , γ' 0D , γ RD2 ,φ' , , , a) 24
γ 'D w d50 d50 d50

where

P = applied load; y = pile head displacement; D = pile diameter; γ' = effective unit weight;
L = pile length; Le = load eccentricity; Ip = pile moment of inertia; Ep = pile stiffness;
MR = modulus stiffness; k = permeability; t = time; γw = effective unit weight of water;
φ' = effective angle of friction; lt = thickness of the pile wall; d50 = average sand grain
size; R a = roughness factor.

Boundary effects can occur if the container where the test is performed is too small
for the problem in question. The distance S between the model and the cradle wall (Figure
32) must be such as to avoid this effect. There is still no consensus on what this value
should be for modelling of monopiles. In Zhu et al. (2016), this relationship is greater

40
than 7. However, in Sayles et al. (2018) and Bayton et al. (2018), this ratio is 4.5 and 5.5,
respectively.

The distance lb between the bottom of the container and the tip of the model must
also be considered (KLINKVORT et al., 2018). Despite this, there is not much
information in the literature on this point. In Zhu et al. (2016), this ratio is greater than 4
(lb/D > 4).

Figure 32 – Parameters to analyse the boundary effects

2.4 FINAL CONSIDERATIONS

The foundations for offshore wind turbines are subjected mainly to horizontal
loads due to waves and winds, which are more expressive than vertical loads. Thus, based
on what was exposed throughout this chapter, it is possible to conclude that the design of
piles subjected to horizontal loads involves several aspects, such as the failure mechanism
and classification regarding the element stiffness. Also, when it comes to monopile
structures, the methods usually used for design are not valid due to the difference in
stiffness, since they have larger diameters.
41
Thus, the physical modelling of monopiles is an important tool to assist in
understanding the behaviour of these structures concerning the lateral loads to which they
are subjected. Several aspects must be considered when performing tests in a geotechnical
centrifuge: variations in the gravity field, scale effects, and boundary effects.

Research done by several authors were presented addressing the centrifuge


modelling of monopiles, demonstrating the advances of research in this area and possible
aspects yet to be better understood.

In this research, lateral load tests done in monopile using geotechnical centrifuge
will be presented. It involves monotonic and cyclic tests, to evaluate the behaviour of the
soil when subjected to these loads.

42
3 MATERIALS AND METHODS

This chapter presents the main characteristics of the sands used in the study (silica
sand, carbonated sand with 50% CaCO3 and carbonated sand with 80% CaCO3). Gomes
(2020) performed the characterization of these materials and describes the procedures
used to obtain the physical indexes, as well as provides parameters for dynamic
characterization from resonant column tests and bender elements. It also describes the
process for obtaining artificially carbonated material representative of the Brazilian
offshore region, presenting the chemical characterization of this material. Following, the
description of the geotechnical centrifuge used in physical modelling is presented, as well
as the equipment used for monotonic tests, and the final test campaign with a description
of the tests performed.

3.3 MATERIALS AND TESTING EQUIPMENT

3.1.1 SAND OF SÃO FRANCISCO BEACH (QZ)

One of the soils studied is silica sand, already used in previous studies
(OLIVEIRA FILHO, 1987; GUIMARÃES, 2014; TARAZONA, 2015). It is a material
from the São Francisco beach (Niterói, RJ). To obtain an adequate relationship for the
development of the physical model and to avoid scale effects, the sand was sieved through
a No. 60 mesh sieve (0.250 mm opening) and subsequently washed in a No. 200 sieve
(0.074 mm opening), to use the particles retained between these meshes, thus obtaining a
fine and uniform material.

The physical indexes of this material, obtained in the present research, are
presented in Table 4.

3.1.2 CARBONATED SANDS (CA80 and CA50)

Carbonated sands present important differences mainly in physical properties


when compared to silica sand, in addition to different chemical and mineralogical

43
composition. This material is found in abundance on the Brazilian coast, which ends up
causing difficulties for offshore exploration works.

Petrobras was contacted in order to obtain samples of offshore carbonated sands


to use in the present study, but this contact proved to be unfruitful. As mentioned
previously, natural carbonated sands have quite complex characteristics (WATSON et
al., 2019), including the presence of internal voids that make them unsuitable for use in
centrifugal modelling. Considering all the questions above, the alternative found was to
use artificial carbonated sands in this research. However, there are important differences
between natural and artificial carbonated sands, such as the natural structure of the soil
and the size of the grains.

There are two main requirements in choosing the sand soil to be used in soil-
structure interaction studies in a centrifuge. The first requirement is that the sand be
homogeneous, a requirement of the pluviation technique used in the preparation of the
models. A second requirement is to present a ratio between the average grain diameter
(d50) and the diameter of the monopile.

To find a material with a significant percentage of CaCO3 that could be used for
the physical modelling tests at the Multi-User Centrifuge Modelling Laboratory (LM2C
- COPPE/UFRJ), the super-fine aragonite, with 80% of CaCO3, was chosen to be mixed
with silica sand, to obtain different percentages of CaCO3. Gomes (2020) details this
mixing process.

The material was also sieved between sieves No. 60 and No. 200, to obtain a fine
material that met the specifications of the physical model. However, as it is a carbonated
material that could be modified in contact with water, it was not washed as the QZ sand
was.

The physical indexes of the superfine aragonite (CA80) with 80% of CaCO3 and
the carbonated sand with 50% of CaCO3 (CA50) are shown in Table 4. As expected, it is
observed that the higher the percentage of CaCO3, the higher the void index values. The
same occurs for the value of the gravity weight of the grains (Gs). Also, the friction angle
of carbonated sands is greater than that of silica sand, and consequently, they are the
materials with greater resistance.

44
The values of maximum shear modulus (Gmáx) presented refer to the values found
by Gomes (2020) from resonant column tests. The soil Young’s modulus was calculated
from Equation 8, assuming a value for Poisson’s ratio of 0.3.

Table 4 - Physical indexes of the soils studied

Parameter Symbol Unit QZ CA50 CA80


Maximum void ratio 𝑒𝑚á𝑥 - 0.915 1.113 1.247
Minimum void ratio 𝑒𝑚í𝑛 - 0.602 0.762 0.862
Maximum unit weight 𝛾𝑚á𝑥 kN/m³ 16.47 15.64 15.13
Minimum unit weight 𝛾𝑚í𝑛 kN/m³ 13.78 13.04 12.54
Specific weight for Dr = 80% 𝛾 kN/m³ 15.55 14.76 14.25
Grain unit weight Gs - 2.638 2.756 2.817
Average particle diameter 𝑑50 mm 0.18 0.18 0.18
Non-uniformity coefficient - - 1.90 2.11 1.90
Friction angle (at 100 kPa) φ degrees 40.6 46.10 48.11
Maximum shear modulus (σo=100 kPa) Gmáx MPa 98.40 85.22 83.89
Soil Young’s modulus (σo=100 kPa) 𝐸𝑆𝐿 MPa 255.84 221.57 218.11

3.1.3 BEAM CENTRIFUGE

The Multi-user Centrifuge Modelling Laboratory (LM2C), located at


COPPE/UFRJ (www.lm2c.coppe.ufrj.br), has two geotechnical centrifuges. A drum
centrifuge, which has already been used in several kinds of research and theses (e.g.
OLIVEIRA et al., 2010; OLIVEIRA et al., 2017), and the beam centrifuge, in use since
2011 (e.g. ALMEIDA et al., 2014, GUIMARÃES et al., 2015). The tests of this research
were carried out in the beam centrifuge. The studies conducted at LM2C from 1996 to
2015 are summarized at Almeida et al. (2016). Since then, other researches were carried
out, such as Hotta et al. (2019).

The beam centrifuge has a diameter of 1.2 m and a maximum acceleration capacity
of 300 times the acceleration of gravity, reaching up to 638 rpm. Figure 33 shows the
beam centrifuge.

45
The main components of the centrifuge are:

1) Data acquisition system: allows to obtain test data through an on-board


computer. There are three different frequencies for the data acquisition
system: 1 Hz, 5 Hz, and 10 Hz. In this research, a rate of 10 Hz was used.
2) Beam rotor: makes the connection between the cradles where the sample is
placed and the counterweight cradles.
3) Bidirectional actuator (Figure 34): allows remote control of horizontal and
vertical movement through two axes. It lays on the box shown in Figure 31.
4) Swinging cradle (Figure 35): made of high-tensile aluminium, consisting of a
bottom, two end plates, two side plates and a pair of rails across the top of the
cradle.

Figure 33 – Beam Centrifuge of LM2C

46
X Axis
Z Axis

Figure 34 - Bi-directional actuator (BROADBENT Inc, 2011)

Figure 35 - Plain strain cradle (BROADBENT Inc, 2011)

Figure 36 shows the entire system assembled and the principle of operation, and
the main components are also highlighted.

47
Figure 36 - General arrangement of the beam centrifuge (BROADBENT Inc, 2011)

3.1.4 CYLINDRICAL CRADLE

A laterally loaded pile is not an axisymmetric problem and should be analysed in


a 3D approach. To correctly consider all aspects involved in this type of problem, the
plain strain cradle shown in Figure 31 had to be changed to a cylindrical type box.

The acrylic plates were removed, and it was manufactured in high-strength


aluminium a cylindrical tube, shown in Figure 37, whose dimensions are: internal
diameter of 220 mm and a height of 140 mm.

48
Figure 37 – Cylindrical box dimensions (in mm)

Figure 38 shows the cradle assembled, attached to the swinging cradle, used in the
tests. The height of the cylindrical container allows for it to be attached to the cradle
without disturbing the sample.

Figure 38 – Cylindrical box used in the tests

49
3.1.5 MONOPILE MODEL

As discussed in item 2.1, the external diameter of monopiles varies between 5 to


7 m and can reach up to 8 meters. However, the centrifuge has limits of operation due to
weight balance issues, limiting the maximum angular speed. An angular speed θ equal to
39.48 rad/s was adopted. Using Equation 21, considering that the rotation radius of the
centrifuge is 0.57 m, the scale factor N = 96 was calculated.

Besides, the choice of the prototype was made to meet with the Benchmark
project, as discussed by Klinkvort et al. (2018). The flexural rigidity of the prototype
should be approximately 20 GNm2, which led to the choice of aluminium for the material
of the model. Commercially available aluminium tubes have limitations in diameter and
thickness, which led to a model with an external diameter of 19.16 mm and a thickness
of 1.29 mm (Figure 39). Thus, the prototype adopted was 1.84 m in external diameter,
with embedded length L = 5D and free length Le = 4D. The prototype and model
dimensions are shown in Table 5.

Table 5 - Dimensions for model and prototype at N = 96

Parameter Symbol Prototype Model


External diameter D 1.84 m 19.16 mm
Internal diameter d 1.59 m 16.58 mm
Thickness lt 0.12 m 1.29 mm
Free length Le 7.57 m 78.85 mm
Load eccentricity ex 6.72 m 70.00 mm
Embedded length L 9.23 m 96.15 mm
Total length Lt 15.95m 175.00 mm
Young Modulus of aluminium E (GPa) 70.00 70.00
Moment of Inertia I 0.25 m4 2905.90 mm4
Pile stiffness EpIp 17.42 GNm² 0.2034 kNm²

50
Figure 39 – Model dimensions (in mm)

𝐷
With these dimensions for the model, a relationship = 106 was obtained,
𝑑50

higher than the value recommended by Klinkvort et al. (2012). As for the boundary effect,
the ratio S/D = 5.24 is also within the limits recommended by the literature. Table 6
presents the parameters considered in the present study of the physical modelling of
monopiles.

Table 6 – Parameters of scaling effects of the present study

D/d50 106.44
lt/D 0.07
L/D 5.02
Le/D 4.12
S/D 5.24
lb/D 2.56
EIprot/N4 0.205

51
With these prototype dimensions, the relative stiffness of soil-pile Kr was
calculated for each sand, from Equation 7 (see Chapter 2), and the values are shown in
Table 7. In Figure 40, the values of Kr and L/D for this research are shown, together with
values for monopiles in the United Kingdom, the model by Abadie et al. (2019) and the
models used in the development of the p-y curves.

Table 7 – Soil-Pile relative stiffness

Sand 𝑬𝑺𝑳 (MPa) 𝑲𝒓 (Equation 7)


QZ 255.84 0.00938151
CA50 221.57 0.01083244
CA80 218.11 0.01100418

Rigid

Model used for the


present study

Flexible

Figure 40 – Kr plotted against L/D ratio for the model used in the present study, full-scale offshore wind
farms (OWF) in the United Kingdom, and piles used for the development of p-y curves. (AFTER
ABADIE et al.., 2019)

52
3.2 INSTRUMENTATION USED IN THE CENTRIFUGE TESTS

The original bi-directional servo-actuator, shown in Figure 34, was modified to meet
this research. The adaptation involved removing the motor from the z-axis, leaving only
the motor on the x-axis, as shown in Figure 41, for the application of horizontal load. The
equipment has displacement control, the maximum force of ± 1.5 kN, and the maximum
linear speed of 6.9 mm/s (BROADBENT Inc, 2011).

Sensor for displacement


measurement

Figure 41 – Servo actuator with displacement control

A miniature load cell with a maximum capacity of 1250 N reads the applied load
(Figure 42). The load cell is coupled to the actuator through a hinge that prevents the
generation of bending moment at the point of application of the load (Figure 43). The
model is installed at 1g with the fixation system already attached. Then, the actuator is
positioned, and the connection is made to the model.

53
Figure 42 - Load cell 1250 N

Fixation piece

Monopile model
Hinge

Load cell

Figure 43 – Fixation between the actuator and load cell system

The bending moment measurement on the model is done using 5 pairs of strain
gauges, glued on ½ Wheatstone bridge, on diametrically opposite faces of the model, as
shown in Figure 44. To prevent the strain gauges from being damaged during the test, a
thin layer of epoxy is applied for protection. The change in the diameter of the pile was
very small considering it was a thin layer of epoxy. A cable connects the strain gauges to
the connectors of the data acquisition system, and during the test it remains in vertical
position, as shown in Figure 43. The voltage difference between the strain gauges on each
side, for the same depth, is recorded by the data acquisition system.
54
Fixation piece

Hinge

Strain gauges

Figure 44 – Model instrumented with strain-gauges

A laser sensor installed at ground level reads the horizontal displacement of the
model throughout the test (Figure 45). Another measurement of horizontal displacement
at the top of the pile is made by a magnetic sensor coupled to the actuator. The results of
laser reading showed noise in the electrical signal, so a Fourier transform filter was
applied to the displacement data to eliminate this noise.

Figure 45 - Laser for displacement measurement

55
3.2.1 CALIBRATION OF STRAIN GAUGES FOR BENDING MOMENT

The calibration of the instrumented model is necessary to determine the


coefficient between the electrical response of the strain gauges and the bending moment
generated by the application of lateral load. Figure 46 shows the distribution of strain
gauges throughout the model.

Side I Side II

Figure 46 - Distribution of strain gauges throughout the model

Calibration consists of submitting the model to a load with known boundary


conditions. The tube is clamped at one side, and known loads are applied at time intervals,
and the difference in electrical response between the strain gauges glued on opposite sides
is recorded by the data acquisition system. The process is done with the two sides of the
tube facing upwards, that is, the strain gauges of the side I upwards (Figure 47a) and then
with side II facing upwards (Figure 47b).

56
System used for
loading

(a)

(b)

Figure 47 – Calibration with both sides of the model facing up: side I (a) and side II (b)

It is necessary to calculate the maximum load to be applied during calibration so


that the aluminium model does not suffer plastic deformations. The aluminium tube is
made with an alloy of type 6063 T5, whose yield limit 𝜎𝑒 is equal to 107 MPa.

57
The maximum stress to be applied is given by:

𝑀(𝑧) 𝐷
𝜎𝑥 = 25
𝐼 2

with

𝑀(𝑧) = 𝑃. 𝑎 26

where P is the applied load and a is the distance from the crimp point to the load
application point (Figure 48).

Figure 48 – Calibration scheme

The maximum load that can be applied is given by Equation 27. However, the
load used for calibration was equal to 27% of 𝑃𝑚á𝑥 , or approximately 50 N.

𝐼
𝐷/2
𝑃𝑚á𝑥 = . 𝜎𝑒 = 185,5 𝑁 27
𝑎

Thus, 5 kg is applied in 5 steps of 1 kg at every 2 minutes. The masses are then


removed, also at every 2 minutes. The applied loads are shown in Table 8. Figure 49
shows an example of the loading and unloading steps for strain gauge number 3, for
calibration with side I facing up.

58
SG3_I_PGA_100x
0,35
0,3
Electrical Response (V) 0,25
0,2
0,15
0,1
0,05
0
-0,05
0 200 400 600 800 1000 1200 1400 1600
Time (s)

Figure 49 - Calibration loading and unloading steps

Table 8 – Loads used in the calibration

Step 1 2 3 4 5

Mass (g) 987.10 987.30 994.20 990.90 1008.20

Accumulated load 9.66 19.33 29.06 38.76 48.76

% of the elastic limit


5.21 10.42 15.66 20.89 26.21
of the tube

To calculate the bending moment, the distance between the center point of the part
where the load is applied and the center of each strain gauge is measured. Thus, the mass
applied at each stage is multiplied by the acceleration of gravity to obtain the applied load,
and the applied load is multiplied by the distance to obtain the bending moment at each
point in the model. The electrical response recorded for each pair of strain gauges is then
related to the corresponding bending moment at that point. The points are then plotted on
a straight line, and the slope is determined. Figure 50 shows this calibration curve, for
strain gauge 3 on side I. The curve does not pass by the origin, because at the beginning
of the test there is a reading value due to the mass of the system used for loading.

59
This procedure is done before each test for each of the strain gauges. The constant
used in the tests is the mean value of the constants obtained for the calibration of sides I
and II, as shown in Table 9.

Bending_Moment_N.m_SG_03_I_PGA_100x
7
y = -19,274x + 0,5483
R² = 0,9996
6
Bending Moment (N.m)

0
0 -0,05 -0,1 -0,15 -0,2 -0,25 -0,3 -0,35
Electrical Response (V)

Figure 50 – Calibration curve with loading and unloading

Table 9 – Strain-gauges calibration constants

Coefficients Table (N.m/V)


Strain Gauge Base I Base II Mean Value
1 26.43 22.47 24.45
2 19.12 18.36 18.74
3 19.27 18.76 19.02
4 19.29 18.25 18.77
5 19.06 18.96 19.01

60
3.3 MODEL PREPARATION AND CONFIGURATION

3.3.1 PLUVIATION METHOD

When preparing the centrifuge model, it is important to obtain a homogeneous


sample. For this research, the pluviation method proposed by Oliveira Filho (1987) was
adopted, based on the method of Miura and Toki (1982), and used by Gomes (2020) in
the preparation of samples for triaxial tests, bender elements, and resonant column.

The method consists on placing the sand in a nozzle with an opening of


approximately 0.14 cm, and the material passes through a sequence of 7 sieves with an
opening of 0.425 mm (No. 40 mesh), as shown in Figure 51. Thus, the sand is distributed
over an area and falls directly into the cylindrical container used in the tests.

The relative density of the sample (𝐷𝑟 ) is calculated as determined by the NBR
12051-1991. From equations 28 to 30, where 𝑀𝑠+𝑐 is the mass of the container + soil, 𝑀𝑐
is the mass of the container and 𝑉𝑐 the volume of the container, determined from a known
mass of distilled water at 25ºC. 𝛾 is the unit weight of the material, which must be equal
to the unit weight for the density of 80%.

𝑒𝑚á𝑥 − 𝑒
𝐷𝑟 = 𝑒 28
𝑚á𝑥 − 𝑒𝑚í𝑛

𝐺𝑠
𝑒= −1 29
𝛾

𝑀𝑠+𝑐 − 𝑀𝑐
𝛾= 30
𝑉𝑐

61
Nozzle with
opening of
0,14 cm

Set of 7
sieves

Container
used in the
test

Figure 51 – Pluviator used in sample preparation

3.3.2 TEST PREPARATION

All tests were carried out on dry sand, although the conditions of monopiles on
the offshore environment are on saturated soil. This is to represent an effective stress
approach. Also, this was chosen due to the difficulties on performing centrifuge tests on
saturated materials, where the viscosity of the material used to saturate the sample must
also be taken into account using the scaling laws, a topic that will not be discussed in
details in this work. Other authors also performed lateral tests on monopiles using dry
sand (e.g. KLINKVORT et al., 2012; CHOO and KIM, 2016; ABADIE et al., 2019).

The test preparation procedure then consists of: sample preparation (soil),
pluviation, weighing and checking the relative density (Figure 52), fixing the cylindrical
tube in the basket that goes into the centrifuge (Figure 53a), placing the actuator in the

62
box (Figure 53b), installation of the model at 1g and fixation of the actuator with the
model and checking the instrumentation and measurements (Figure 53c). Then, the
prepared set is placed inside the centrifuge (Figure 53d), where the instrumentation is
connected to the data acquisition system, and then the test, which will be described in
item 3.4, proceeds.

Figure 52 – Weighting of the container with sand

63
(a) (b)

(c) (d)

Figure 53 - Test preparation

64
Figure 54 shows an outline of the test and instrumentation.

(a)

(b)

Figure 54 – Simplified test layout (a) and superior view (b) (dimensions in mm)

65
3.4 TESTING PROGRAM AND SEQUENCE

3.4 1 CENTRIFUGE TESTS PERFORMED

During the development of this research, 7 different groups of centrifuge tests


were carried out. The MF1-MF4 type tests were preliminary tests, to adjust the
equipment, and develop the test procedure and to check the repeatability of the results.
These tests were carried out only for QZ sand and are summarized in Table 10. The results
of these tests will not be shown here.

Table 10 – Preliminary test campaign

Loading rate Displacement at the load


Test Type of actuation
(mm/s) application point

2 mm, 4 mm, 6 mm, 8 mm,


10 monotonic loads
MF1 0.37 10 mm, 12 mm, 14 mm, 16
(back and forth)
mm, 18 mm, 20 mm

11 mm (2 loadings) and 35
MF2 3 monotonic loads 0.32
mm (last loading)

MF3 1 monotonic load 0.32 11 mm

Monotonic + 50 cycles + 30 mm (monotonic) + 3,5


MF4 0.32
Monotonic mm (cycles)

The final tests campaign, with a total of 11 tests, involved tests of types BCH,
MF5, and MF6. Table 11 shows these tests, the relative density obtained for each one, the
type of actuation performed, loading rate, displacement applied at the point of application
of the load, and the number of cycles.

66
Table 11 – Final test campaign

Displacement at
Relative
Type of Loading rate the load Number
Test Sand density
actuation (mm/s) application of cycles
(%)
point
BCH_QZ_1 80
QZ Monotonic 0.32 11 mm -
BCH_QZ_2 80
MF5_QZ_1 QZ 79
MF5_CA50_1 47 Monotonic 30 mm
CA50
MF5_CA50_2 44 + cyclic + 0.32 (Monotonic) 150
MF5_CA80_1 80 Monotonic 3.5 mm (cyclic)
CA80
MF5_CA80_2 81
MF6_QZ_1 79 30 mm
QZ Monotonic
MF6_QZ_2 80 (Monotonic)
+ cyclic + 1.02 4800
MF6_CA80_1 79 3.5 mm (cyclic)
CA80 Monotonic
MF6_CA80_2 78

Considering the preliminary tests presented (Table 10) and the final tests (Table
11), 15 tests were carried out in this research. The results of these final tests are presented
in Chapter 4.

The BCH type test was carried out for the benchmark project. It consists on the
application of a monotonic loading, with displacement at the ground level of
approximately 0.25 times the diameter of the model, which corresponds to 11 mm at the
level of the actuator. The loading rate used was 0.32 mm/s. These tests were carried out
only for QZ sand with a mean relative density of 80%.

The MF5 test group consists of a monotonic loading of 30 mm at the actuator


level, followed by 150 cycles with a displacement of 3.5 mm, and another monotonic
loading at the end of the 150 cycles. The loading rate is 0.32 mm/s. These tests were
carried out on the sands QZ, CA50, and CA80. For the sands QZ and CA80 the relative
density obtained varied from 78% to 81%, with a mean value of 80%. As for the tests on

67
CA50 sand, due to an error during the pluviation process, the relative density obtained
was 44% and 47%, with a mean value of 46%.

The MF6 type test group also consisted of a monotonic loading of 30 mm at the
actuator level, followed by 4800 cycles with a displacement of 3.5 mm, and another
monotonic loading at the end of the cycles. The loading rate used was 1.02 mm/s. Due to
a problem in the centrifuge control system, this test was performed only on the QZ and
CA80 sands, and there was no time available to perform the test on the CA50 sand. The
relative density obtained for the QZ and CA80 sands was in the range of 78% to 80%,
with a mean value of 79%.

Figure 55 presents a diagram that shows the type of movement made in the cycles:
the model is displaced in one direction, returns to its original position and is displaced
again in the same direction.

Figure 55 – Type of movement in the cyclic tests

Little information is available on the appropriate definition for the criteria of pile
failure when subjected to lateral loading. This is because there is no clear failure
behaviour in the load-displacement curve, and the piles, especially the rigid and short,
accumulate displacement with increasing of the load. The criterion of failure is defined
by the project, but it is often considered as a displacement of 0.1D (10% of the external
diameter of the pile) for the limit state of use, which in the case of the present study would
be 0.1 x 19.16 mm = 1.9 mm, or 2o rotation on the pile head. In Rosquoet et al. (2007)
failure is defined by the intersection between the tangent at the origin and the tangent

68
corresponding to large displacements in the load-displacement curve (double tangent
method). Another criterion is the point where the curve shows an increase in displacement
without any increase in load.

The first and last monotonic loading for the MF5 and MF6 type test groups sought
the condition of soil failure. Then, a displacement of 30 mm at the load application point
was imposed, that is, 1.6D would be enough to achieve this condition. Besides, the choice
of maximum horizontal displacement of 30 mm was made not to damage the load cell,
which cannot be subjected to bending, as it only resists axial forces.

The ultimate load is an important information to define the load amplitude for
cyclic tests with force control. In the case of the present research, the purpose of these
loads was to evaluate the effect of soil degradation after a certain number of cycles. That
is, even if the cycles are applied to a sample that has already been subjected to a failure
condition, it was sought to verify whether the ultimate load would change after the
application of the cycles.

In controlled load tests, Solf et al. (2010) highlight a possible effect of self-healing
of the soil after the application of a large horizontal load, followed by low-intensity
cycles. Based on physical and numerical models, the authors conclude that this self-
healing effect was stronger for greater cyclic force amplitudes, greater magnitudes of the
initial static load and higher soil densities. This happens only if compaction happens to
the sand. If the sand dilates, then this self-healing effect will not happen.

The loading rate used was limited by the capacity of the actuator and so as not to
damage the load cell, which restricted the maximum frequency of cyclic loading. For tests
of type MF5, the frequency was 0.04 Hz and for tests of type MF6, it was equal to 0.14
Hz. Transformed for prototype scale are: 4.17 × 10−4 Hz for the MF5 test and
1.46 × 10−3 Hz for the MF6 test. These values are lower than the values of frequency of
waves and winds to which the monopile is subjected: up to 0.1 Hz for the wind of 0.05-
0.14 Hz for waves.

Thus, the cyclic performance was of the quasi-static type, seeking only to impose
a condition of degradation on the soil, and not to represent loading on the field. The
amplitude of the displacement for the cycles simulated the maximum displacement for
the service limit state (rotation of 0.5º at pile head).

69
3.5 FINAL CONSIDERATIONS

In this chapter, the main physical characteristics and resistance parameters of the
materials used for the physical modelling tests were presented: silica sand, carbonated
sand with 50% of CaCO3 and carbonated sand with 80% of CaCO3. The pluviation
method used for sample preparation was also described.

The beam centrifuge, in which the tests were carried out, was also presented. And
so was the actuator used to apply the horizontal load, and all the instrumentation used.

An aluminium model with a diameter of 19.16 mm was adopted, simulating a


prototype with a diameter of 1.84 m, for N = 96. This model was instrumented with strain
gauges calibrated to bending moment, and then obtain p-y curves with experimental data.

Finally, the preliminary and final test campaigns was detailed. The final tests
involved three groups of tests: benchmark, one with 150 cycles and another with 4800
cycles.

70
4 RESULTS OF THE CENTRIFUGE TESTS

This chapter presents the results of the tests described in item 3.4. It also presents
the bending moment profiles along with the depth, and the process performed to obtain
the p-y curves from these results. Finally, the results are compared and discussed.

In general, the results of tests for QZ sand are plotted in blue, those for CA50 sand
in green and those for CA80 sand in orange.

4.1 BCH TESTS

Figure 56(a) shows the load-displacement curve, that is, the horizontal load
applied by the displacement measured at ground level for the BCH_QZ_1 and
BCH_QZ_2 tests, and Figure 56(b) shows the normalized curve. The displacement on the
soil surface is normalized by the diameter of the model, and the load is normalized by the
factor NγD3. The curves obtained for the two tests are practically coincident, which shows
the repeatability of the test and the reliability of the test preparation.

350 35

300 30

250 25
P/(NγD3)

200 20
Load (N)

150 15

100 10

50 5

0 0
0 1 2 3 4 5 6 7 0 0,1 0,2 0,3
Displacement (mm) y0/D
BCH_QZ_1 BCH_QZ_2 BCH_QZ_1 BCH_QZ_2
(a) (b)

Figure 56 – Load-displacement curve (a) and normalized curve (b) for QZ sand

71
Abadie (2015), when validating the results of physical modelling from data of
other authors, proposes Equation 31, which adjusts the results of monotonic tests by a
power law. The database analysed by the author involves different values of eccentricity,
the geometry of the model, soil parameters and test procedures (1g and Ng).

𝑀 𝑦 0.31
= 2.04 (𝐷) 31
𝑀𝑟

where M is the bending moment applied, Mr is the moment referring to a displacement


of 0.1D and y is the displacement at ground level.

In Figure 57, the data for the BCH_QZ_1 and BCH_QZ_2 tests are plotted
together with the power-law by Abadie (2015). The bending moment is calculated by
multiplying the load by its eccentricity. The model used does not fit the test data, as it
predicts a more rigid behaviour than that obtained experimentally.

Figure 57 - BCH result validation with the model proposed by Abadie (2015)

72
Within the Benchmark study, the results of the BCH test are compared with the
results of the other study groups. Figure 58 presents the results of the tests performed by
the other groups together with the results of the tests in the present study, identified by
COPPE_1 and COPPE_2. All groups modelled the same prototype, but with different
diameter and values of N. All tests were performed on silica sand with different physical
indexes and resistance parameters. From the normalized curve (Figure 58b), it is possible
to observe that COPPE tests show a slightly more rigid behaviour than most of the other
tests, except the DTU Delft and University of Nottingham (UoN) tests.

Figure 58 – Load-displacement curve (a) and normalized curve (b) Results of the benchmark
(KLINKVORT, 2020)

73
Figure 59 shows the results of the two groups that performed tests on models with
higher values of N, therefore smaller diameters, and lower measured horizontal loads,
which were the cases of the COPPE and TU Delft tests, highlighted in Figure 59. As
shown in Figure 59, all curves show a similar behaviour at the beginning of loading, but
the two COPPE tests reach higher horizontal load values.

Figure 59 – Load-displacement curve (a) and normalized curve (b): Benchmark tests performed on small
diameter models (KLINKVORT, 2020)

Table 12 shows, for the sand used by each study group, the average particle diameter
(d50) and the friction angle (at 100 kPa). The friction angle for all sands is in the range of
40º to 41º, except the KAIST group, for which the friction angle was equal to 43.5º. The
average particle diameter varies from 0.111 mm to 0.224 mm, all fine sand.

74
Table 12 - Friction angle and average particle diameter for each sand in the benchmark test
(KLINKVORT, 2020)

Group Friction angle at 100 kPa (degrees) d50 (mm)


CEIGR 40.6 0.160
COFS 41.0 0.168
COPPE 40.6 0.180
CUED 40.7 0.224
DTU 40.5 0.194
IFSTTAR 40.1 0.184
KAIST 43.5 0.188
TU_Delft 40.6 0.111
UoN 40.6 -

4.2 MF5 TESTS

4.2.1 RESULTS FOR LOAD AND HORIZONTAL


DISPLACEMENT OVER TIME

Figure 60 shows the load curve over time (in the model) for the MF5 test
performed on QZ sand. The first monotonic loading can be observed at the beginning of
the test, followed by 150 cycles, and at the end another monotonic loading.

150 cycles

2nd monotonic
1st monotonic
loading
loading

Figure 60 – Load over time: test MF5_QZ

75
Figure 61 shows the displacement measured over time, for measurements of the
laser and the actuator displacement sensor, which are positioned at different heights
according to Figure 47a (Chapter 3). The 150 cycles are made with displacement in only
one direction, that is, the model is displaced in one direction, returns to the original
position and is displaced again in the same direction (see Figure 55), as shown in the
detail of Figure 62.

Detailed in Figure 62

Figure 61 – Displacement over time: Test MF5_QZ (at model)

Figure 62 – Displacement over time: detail of type of movement (at model)

76
4.2.2 LOAD-DISPLACEMENT CURVES

To provide an overview of the test (monotonic + cyclic + monotonic), Figure 63


shows the normalized load-displacement curve, for the horizontal load applied and
displacement measured at the soil surface by the laser. The first monotonic loading before
the cycles (in blue), the 150 cycles (highlighted in the ellipse), and the monotonic loading
after the cycles (in dashed line) can be observed.

1st monotonic
loading

150 cycles
2nd monotonic
loading

Figure 63 – Normalized load-displacement curve: Test MF5_QZ

Figure 64 shows the normalized load-displacement curves for the CA80 sand,
MF5_CA80_1 and MF5_CA80_2 tests. The dashed lines indicate the monotonic loading
after the cycles. The two tests have similar curves, which indicates the repeatability of the
tests.

77
Figure 64 – Normalized load-displacement curves: Tests MF5_CA80_1 and MF5_CA80_2

Figure 65 shows the normalized load-displacement curves for the tests with sand
CA50, MF_CA50_1, and MF5_CA50_2. The dashed lines indicate the monotonic
loading after the cycles. It is also possible to observe that the curves of these tests are
close, assuring the repeatability of the tests. Unlike the other tests, due to a problem with
the pluviation procedure, for the CA50 sand, the relative density of 80% was not reached.

78
Figure 65 – Load-displacement curves: Tests MF5_CA50_1 and MF5_CA50_2

The tests MF5_QZ, MF5_CA50_1 and MF5_CA80_1 are presented together in


Figure 66, which shows the normalized load-displacement curve for the three different
sands. The dashed lines represent the monotonic loading after the cycles. For comparison,
only one of the tests for CA50 and CA80 was chosen. It is observed that, for monotonic
loads, the CA80 sand presented the highest load-displacement ratio. This is in line with
what is expected for this material, which has the highest friction angle among the sands
under study.

The CA50 sand, on the other hand, had the lowest load-displacement ratio, as it
was tested with a lower relative density than the other two sands. It is known that a
variation in the relative density influences the response of the soil. In the case of lateral
loading, denser samples tend to be more rigid, and the failure mechanism in dense and
loose samples is different.

79
Figure 66 –MF5 tests: Comparison of normalized load-displacement curves for QZ, CA50, and CA80
sands

4.4.3 COMPARISON OF BEHAVIOUR BEFORE AND AFTER CYCLES

To assess the failure load, the two monotonic loadings were selected before
(Figure 67a) and after the cycles (Figure 67b). The double tangent method was used as a
failure criterion, and the failure load for each sand, before and after the cycles, are shown
in Table 13.

From Figure 67, it is possible to observe that the sands QZ and CA50 reached
failure. As for CA80 sand, even with an increase of displacement, there is still an increase
in load, so the material has not yet reached its ultimate capacity. This is in line with
expectations, as it is the most resistant material among the three soils studied, so the
failure load should be greater.

80
Before cycles
6

4
Load (MN)

0
0 0,2 0,4 0,6 0,8 1 1,2 1,4 1,6 1,8
Displacement (m)
QZ CA80 CA50

(a)

After cycles
6

4
Load (MN)

0
0 0,2 0,4 0,6 0,8 1 1,2 1,4 1,6 1,8
Displacement (m)
QZ CA80 CA50

(b)

Figure 67 – Load-displacement curves: Tests MF5 – monotonic loading before (a) and after (b) 150
cycles for sands (prototype scale)

From Figure 67 (a), it is observed that the CA50 sand had low stiffness from the
beginning of the test. Considering that in this test the relative density was lower than that
for the sands QZ and CA80, this is within the expected. If the test had been carried out at

81
a relative density of 80%, it would be expected that the CA50 sand would present an
intermediate behaviour between QZ and CA80, as it is a material with an intermediate
friction angle between the other two sands. Therefore, it was expected that its behaviour
on the load-displacement curve would also be intermediate between the other two
materials.

As it is a simple and widely used method (ROSQUOET, 2007; EL HAFFAR,


2018), the ultimate load was calculated using the Broms Method (1964) for QZ and CA80
sands for comparison with the values found experimentally, which are presented in Table
13. For QZ sand, the experimental failure load found was lower than that calculated by
Broms (1964), with a difference of about 15%. This may be due to the installation method
adopted in the tests, in which the model was installed at 1g. As discussed in item 2.3.3,
the installation method influences the ultimate soil resistance. In the installation at 1g, it
is not simulated the same level stress as in the prototype, which reduces the capacity of
the soil. This is in line with that presented by Dyson and Randolph (2001) and El Haffar
(2018), who compared the ultimate load for installation at 1g and Ng.

For the CA50 sand, the experimental ultimate load was approximately 58% less
than that calculated by Broms. This can be explained by the installation method, but it is
mainly related to the relative density obtained for this test. For CA80 sand, no failure was
observed, but the load for the greatest displacement, of approximately 5.60 MN, was
about 11% greater than the value calculated by the Broms Method. A preliminary
conclusion is that for the sands QZ and CA80, the experimental ultimate loads are
relatively close to the values calculated using Broms’ formulation.

Table 13 – MF5 tests: ultimate load before 150 cycles and comparison with Broms (1964)

QZ CA50 CA80
Experimental ultimate load (MN) 3.20 1.80 5.60
Broms ultimate load (MN) 3.75 4.35 4.96
Difference (%) 14.67 58.62 11.43
Mean relative density (%) 79 46 80

82
The ultimate loads before and after the cycles are presented and compared in Table
14. For the QZ sand, the ultimate load was the same, that is, the number of cycles applied
did not influence the soil resistance. For the CA50 sand, there was an increase of 44.4%
in the last load after the cycles. This increase was observed only for this sample and may
be due to an effect of self-healing of the soil, as discussed by Solf et al. (2010), and for a
densification effect of the sample, which at the beginning of the test was at loose state.
As for CA80 sand, as the failure load was not identified, it is not possible to discuss the
influence of the number of cycles on the ultimate load. Based on the results of the QZ and
CA80 sands, it can be concluded that the ultimate loads before and after the 150 cycles
had small variations.

Table 14 - MF5 test: comparison of the ultimate load before and after 150 cycles

QZ CA50 CA80
Before cycles (MN) 3.20 1.80 5.60
After cycles (MN) 3.20 2.60 5.50
Difference (%) 0.00 44.44 1.79
Mean relative density (%) 79 46 80

4.3 MF6 TESTS

4.3.1 RESULTS FOR LOAD AND HORIZONTAL


DISPLACEMENT OVER TIME

In the MF6 test, the number of cycles was greater than the number of cycles for
the MF5 test, to assess the effect of the number of cycles on the ultimate load. In addition,
the loading rate in this test was also greater than that of the MF5 test. Due to problems
with the control system of the centrifuge, there was no time available to carry out the MF6
test on CA50 sand. Then, only the results of the QZ and CA80 sands will be presented.

Figure 68(a) shows, for QZ sand, the horizontal load over time, and the first and
second monotonic loading can be observed. Figure 68(b) also shows, for CA80 sand, the
horizontal load over time. It is possible to observe, for the two sands, that the load
83
measured along with the cyclic loading gradually increases initially and reaches a
constant value in the final third of the cycles.

1st monotonic 2nd monotonic

loading loading

(a)

1st monotonic 2nd monotonic


loading loading

(b)

Figure 68 - MF6 tests: Horizontal load over time during the for QZ sand (a) and CA80 sand (b)

84
Figure 69 shows the load over time for the QZ sand test, detailing the beginning
and end of the cycles. It is possible to observe that, over the 4800 cycles, the load increases
for the same displacement, indicating possible densification of the soil.

Figure 69 - MF6_QZ: Load over time (model scale)

The differences between graphs in Figures 69 and 60 are due to the differences in
loading rate (frequency) and the greater number of cycles applied in the MF6 test. In
addition, as this is a controlled displacement test, the loading amplitude is the same, but
the measured load changes.

Figure 70 shows the load over time for the test on CA80 sand, detailing the
beginning and end of the cycles. In this test, in the first monotonic loading, the data
acquisition rate used was 1 Hz, which compromised the initial readings. From the end of
monotonic loading, the rate of 10 Hz was adopted.

85
Figure 70 - MF6_CA80: Load over time (model scale)

4.3.2 COMPARISON OF BEHAVIOUR BEFORE AND


AFTER CYCLES

The same analysis as the one done for the MF5 tests was done for MF6 tests,
which is to compare the failure loads before and after the cycles, as shown in Figure 71(a),
before the cycles, and Figure 71(b) after the cycles. The values obtained are shown in
Table 15. For QZ sand, there was an increase of 31.25% in the ultimate load after the
cycles, and for CA80 sand an increase of 8.93%. It can be concluded that the number of
cycles to which the model was subjected had a significant influence on the ultimate load,
mainly for the QZ sand.

Table 15 – Ultimate load before and after 4800 cycles

QZ CA80
Before 4800 cycles (MN) 3.20 5.60
After 4800 cycles (MN) 4.20 6.10
Difference (%) 31.25 8.93
Mean relative density (%) 80 79

86
(a)

(b)

Figure 71 - MF6 tests: monotonic loading before (a) and after (b) 4800 cycles (prototype scale)

87
4.4 CHANGE OF SECANT STIFFNESS WITH NUMBER OF CYCLES - MF6
TESTS

A model to calculate the accumulation of displacement over cyclic loading is


proposed by Leblanc et al. (2010) and used by Abadie (2019). Klinkvort (2012) also
proposes a model for calculating the accumulated displacement for cyclic tests. However,
the parameters used in these models are for cyclic tests performed with control of force,
which is not the case for the tests performed in this study, which were displacement
controlled.

An analysis of the change in secant stiffness was made over the 4800 cycles. The
secant stiffness for each cycle (Kn) is determined by the slope of the line between the
beginning and end of each cycle, as shown in Figure 72.

Figure 72 – Secant stiffness Kn for cycle n

Figure 73(a) shows the load-displacement curve of the 4800 cycles for the QZ
sand test, and Figure 73(b) for the CA80 sand.

88
(a) (b)

Figure 73 – Load-displacement over 4800 cycles (prototype scale) for QZ sand (a) and CA80 (b)

Figure 74 shows 4 sections of the cyclic loading of QZ sand: cycles 1-50, cycles
100-200, cycles 1000-2000 and cycles 3800-4800. Figure 75 shows the same
representation for CA80 sand.

An analysis of the change in the stiffness was made for cycles of number: 1, 50,
100, 500, 1000, 2000, 3000, 4000 and 4800, which are shown in Table 16.

Table 16 – Change in secant stiffness with the number of cycles

K (MN/m) QZ CA80
K1 0.69 1.10
K50 1.03 1.38
K100 1.07 1.61
K500 2.17 1.91
K1000 1.66 2.20
K2000 2.87 2.18
K3000 4.30 2.44
K4000 3.60 2.44
K4800 4.12 2.05

89
Figure 74 - MF6_QZ test: Load displacement for cycles 1-50 (a), 100-200 (b), 1000-2000 (c), 3800-4800
(d)

90
Figure 75 - MF6_CA80 test: Load displacement for cycles 1-50 (a), 100-200 (b), 1000-2000 (c), 3800-
4800 (d)

In Figure 76, the secant stiffness as a function of the number of cycles is plotted.
The analysis of Pearson's correlation coefficient for a logarithmic approach, as used by
Klinkvort (2012), provides for the QZ and CA80 sands a very high correlation.

91
5,00

4,50

4,00

y = ln(1,9980 + 0,0102x)
3,50 R² = 0,88546
Secant stiffness(Kn)

3,00

2,50

2,00
y = 0,159ln(x) + 0,9695
R² = 0,8732
1,50

1,00

0,50

0,00
0 1000 2000 3000 4000 5000 6000
Number of cycles
QZ CA80

Figure 76 – Change in secant stiffness with the number of cycles

It can be seen from figure 76 that the initial stiffness of CA80 sand is slightly
higher than that of QZ sand. However, during cyclic loading, the stiffness of QZ sand
increases and becomes higher than the CA80 sand.

4.5 COMPARISON BETWEEN MF5 AND MF6 TESTS

Table 17 shows a comparison between tests MF5 and MF6. It can be observed
that there was no difference between the values of ultimate loads for tests MF5 and MF6
before the cycles, for the sands QZ and CA80, indicating the repeatability of tests. But,
with an increase in the number of cycles, the effect of compaction in the sands increases.
For the QZ and CA80 sands, the ultimate load after 4800 cycles was greater than the one
after 150 cycles. For CA80 sand, the ultimate load of 6.10 MN was reached after cyclic

92
load in the MF6 test, and for the MF5 test, the ultimate load was not reached. As for the
QZ sand, the ultimate load had not changed after 150 cycles (MF5 test), and after 4800
cycles (MF6 test), there was an increase of 31.25%. That is, the ultimate load after 150
cycles remained the same, while after 4800 cycles there was a significant increase, mainly
for QZ sand, indicating the influence of the number of cycles on the behaviour of the soil.

Table 17 – Comparison of ultimate load (MN) for tests MF5 and MF6

Before cycles After cycles


Sand 150 4800 Difference 150 4800 Difference
cycles cycles (%) cycles cycles (%)
QZ 3.20 3.20 0.00 3.20 4.20 23.81
CA80 5.60 5.60 0.00 5.50 6.10 9.84

Figure 77(a) shows, for the QZ sand, the load-displacement curves before and
after 150 cycles, and before and after 4800 cycles. And Figure 77(b) shows, for CA80
sand, the load-displacement curves before and after 150 cycles and before and after 4800
cycles. In all cases, the dashed lines correspond to the condition after the cycles, making
it possible to observe the increase in stiffness with the increase in the number of cycles
and change in the failure mechanism. In the first monotonic loading (both before 150
cycles and before 4800 cycles), the failure occurs for large displacements in the QZ sand
and for the CA80 sand it does not reach failure. After 150 cycles, this behaviour does not
change significantly, only for QZ sand there is a small increase in the stiffness at the
beginning of loading. After 4800 cycles, there is a great change in stiffness, and the failure
occurs for higher load values but with lower values of displacement.

93
(a)

(b)

Figure 77 – Load-displacement curves: before and after 150 cycles and before and after 4800 cycles, for
QZ sand (a) and CA80 sand (b)

94
4.6 BENDING MOMENT PROFILES ALONG DEPTH

4.6.1 PRELIMINARY CONSIDERATIONS

The first monotonic loading of the MF5 type tests was analysed to present the
results of bending moments. From the readings of the 5 strain gauges, the values of
bending moment along the monopile depth are obtained. The bending moment value at
the soil surface is determined by multiplying the load measured by its eccentricity, and
the bending moment at the tip of the pile is assumed to be zero, with a total of 7 bending
moment points along the depth. The objective of determining the bending moment profile
is to obtain the experimental p-y curves, which will be discussed next.

Assuming a bending moment at the tip of the pile equal to zero is a way to increase
the number of points available so that the data can be interpolated, which will be discussed
in item 4.5. However, it is known that for large diameter piles, the value of the bending
moment at the tip is not necessarily zero, since there is some rotation at pile tip, and
therefore there is bending moment at that point. Haiderali (2015), when comparing
numerical modelling results with those of an instrumented model with strain gauges,
discusses the effect of assuming or not this value as zero for piles with a diameter of 0.61
m, 3.8 m, and 7.5 m, and concludes that the error is greater for the 7.5 m piles. In the
present study, it was assumed this point with zero bending moment.

4.6.2 RESULTS OF BENDING MOMENT FOR INITIAL MONOTONIC LOADING

Figure 78 shows bending moment profiles measured through strain gauges


calibrated for bending moment, distributed along with the depth as shown in Figure 45.
Bending moment profiles are presented for the MF5_QZ test, for horizontal loads of 1
MN, 1.5 MN, 2 MN, 2.5 MN, and 3 MN, referring to the initial monotonic loading. For
this test, as highlighted in item 4.2, the failure load was equal to 3.2 MN. It is observed
that, as the applied load approaches the failure load, the bending moment at the point
closest to the tip increases (see indicative arrow in Figure 74), which ends up deviating
the moment profile in relation to the smaller applied forces. It is observed that the

95
maximum bending moment during the entire loading of 1 to 3 MN occurred at a depth of
about 2.70 m (1.5D).

Bending moment (MN.m)


0 5 10 15 20 25
0

4
Depth (m)

10

1 MN 1.5 MN 2 MN 2.5 MN 3 MN

Figure 78 - MF5_QZ test: bending moment profile

Figure 79 shows bending moment profiles for CA50 sand, for test MF5_CA50_1.
For this test, the failure load was equal to 1.8 MN, so the bending moment profile up to 2
MN is shown. As in the MF5_QZ test, when the applied load was closer and, in this case,
greater than the failure load, the value of the bending moment at the point closest to the
tip increased, causing the deviation in the curve (see indicative arrow in Figure 79). The
point of maximum bending moment, in this case, was observed at a depth of about 2.8 m,
that is, 1.52D.

96
Bending moment (MN.m)
0 5 10 15 20
0

4
Depth (m)

10

1 MN 1.5 MN 2 MN

Figure 79 - MF5_CA50_1 test: bending moment profile

Figure 80 presents bending moment profiles for the CA80 sand, from the
MF5_CA80_1 test. In this test, no soil failure was observed and, for the load values
analysed, there are no deviations in the bending moment profiles. In the case of the
MF5_CA_80_1 test, the maximum bending moment value was observed at a depth of
about 1.6 m, that is, 1.15D.

97
Bending moment (MN.m)
0 5 10 15 20 25
0

4
Depth (m)

10

1 MN 1.5 MN 2 MN 2.5 MN 3 MN

Figure 80 - MF5_CA80_1 test: bending moment profile

4.7 INTERPRETATION OF BENDING MOMENT RESULTS ALONG DEPTH

4.7.1 PRELIMINARY CONSIDERATIONS

From the bending moment distribution along with depth, it is possible to obtain
the reaction of the soil (p) through Equation 32 and the displacement (y) through Equation
33.

98
𝑑2 𝑀
𝑝= 32
𝑑𝑧 2

𝑀
𝑦 = ∬ 𝐸𝐼 𝑑𝑧 33

For these operations to be performed, it is desirable to represent the bending


moment as a mathematical function. Thus, the bending moment readings obtained from
strain gauges must be approximated by an interpolation function. This approach can be
done in different ways and is a topic discussed by several authors (KING, 1994; YANG
and LIANG, 2006; LAU et al., 2014; KIRKWOOD, 2015; HAIDERALI and
MADABHUSHI, 2016). Klinkvort (2012) uses a 6th-degree polynomial, Choo and Kim
(2016) adopt an approach using a modified 5th-degree polynomial function. While Lau
(2015) and El Haffar (2018) use cubic spline interpolation.

The interpolation function must pass through the measurement data, but it must
also satisfy the boundary conditions for the derivation and integration processes. The first
derivative of the moment is the shear force, and the second derivative is the reaction of
the soil (p). At ground level, the shear force is equal to the lateral force applied. The
integration constants are obtained through the two displacement measures. The first
constant is the rotation in the pile head, and the second is the displacement measured at
the same point.

One problem when determining the p-y curves by experimental methods is


precisely to find the interpolation function that satisfies all these conditions. The double
derivation can amplify possible measurement errors, resulting in wrong values for p. As
for double integration, this is not a problem, and can even minimize these errors (YANG
and LIANG, 2006).

Table 18, adapted from Haiderali (2015), presents the characteristics of lateral
loading tests in sand done by 7 authors, made in centrifuge with strain gauges, indicating
the number of sensors used and the type of interpolation used for the bending moment
data. It can be inferred from the results in Table 17 that ideally, the number of pairs of
strain gauges along the depth of the pile should be greater than or equal to 6
(HAIDERALI, 2015).

99
Klinkvort (2012) used instrumented models with 5 and 10 pairs of strain gauges,
and reports that the interpretation of results for the model with 5 pairs, as adopted in the
present study, is not possible by the small number of sensors, and that only the curve
referring to one depth can be deduced from this model. In this research, 5 pairs of strain
gauges were adopted, that is, a number lower than that adopted in the literature, due to
the reduced dimensions of the monopile model used.

Table 18 – Research parameters for centrifuge modelling of laterally loaded piles - prototype dimensions
(After Haiderali, 2015)

Number of
External pile Pile length
Author pairs of strain Interpolation method used
diameter (m) (m)
gauges
4th to 7th-degree polynomial,
King (1994) 0.76 8.4 6
cubic and quartic spline
Dyson and
Randolph 2.08 41.6 12 3rd degree polynomial
(2001)
Kong and
Zhang 0.63 12 8 5th-degree polynomial
(2006)
Klinkvort
3 18 5-10 6th degree polynomial
(2012)
Kirkwood 6th-degree polynomial and cubic
3.81 20 6
(2015) spline
Choo and
6.0 31-42.8 7 Modified 5th-degree polynomial
Kim (2016)
El Haffar
1.8 20 17 Cubic spline
(2018)

100
4.7.2 SOIL REACTION AND DISPLACEMENT CURVES

A code using Python programming language was developed to perform the 5th
and 6th-degree cubic and polynomial spline adjustments. In the same code, double
integration and double derivation were made to determine the displacement and reaction
profiles of the soil, respectively. In addition, soil reaction points and displacements for
each depth were obtained to plot the p-y curves. From the derivation and integration
processes, it was observed that the results obtained were not compatible with the
boundary conditions of the problem. This is because the number of strain gauges are lower
than what would be desirable, as previously mentioned. So, for the results of this research,
this code was not used.

The results obtained for the bending moment in the tests were used to illustrate in
a simplified way the process of obtaining the reaction curves and displacement of the soil.
In this context, a 6th-degree polynomial was chosen to represent the bending moment
profile along with the depth, and from this polynomial, equations 32 and 33 were
manually applied to obtain the displacement and reaction profile of the soil along with
the depth. Manual interactions were necessary (CHOO and KIM, 2016) to obtain the
results presented here, to match the results of double integration with the horizontal
displacement measures.

Figure 81 shows the soil pressure profile (p) along with the depth, bending
moment and displacement (y) for the QZ sand, for the loads of 1, 1.5, 2 and 2.5 MN. It is
observed that, for the load of 3 MN, the soil pressure profile does not present the same
behaviour as for the other loads, which confirms how the derivation process is affected
by possible errors in the bending moment measurements. The displacement profile shows
that as the load increases, the deflection of the pile increases, which occurs under the same
point of rotation.

101
Figure 81 –MF5_QZ test: Soil pressure, bending moment profile and pile displacement

Figure 82 shows the soil pressure profiles, bending moment and displacement
along with the depth, on the CA50 sand, for the loads of 1, 1.5 and 2 MN. In the same
way as for QZ sand, for CA50 sand it is possible to observe a deviation in the soil pressure
profile, at greater depths, for the loads of 1.5 and 2 MN. The displacement profile also
shows the deflection of the monopile increasing with the increase in force around the
same point.

102
Figure 82 – MF5_CA50_1 test: Soil pressure, bending moment profile and pile displacement

Figure 83 shows the soil pressure profiles, bending moment and displacement
along with the depth, in the CA80 sand for the loads of 1, 1.5, 2, 2.5 and 3 MN. It is
observed, as for the other sands, the uncertainty in the soil pressure profile, obtained from
the double derivation process, both on the soil surface and at greater depths. For this sand,
it is also observed that, unlike the QZ and CA50 sands, the depth of the point of rotation
of the monopile varied during loading.

103
Figure 83 - MF5_CA80_1 test: Soil pressure, bending moment profile and pile displacement

From the results of tests with a greater number of pairs of strain gauges, the
experimental p-y curves can be obtained by taking the values for p and y at each depth.
These values are then plotted and can be compared with the p-y curves recommended in
the literature. However, in the case of the present study, due to the various uncertainties
involved in the interpolation process of bending moment values and the double derivation
and integration processes, it was not possible to have reliability in the experimental results
of soil reaction and displacement of the monopile to get a consistent comparison with
theoretical curves.

104
4.8 THEORETICAL P-Y CURVES

The equations proposed by the methods recommended by API (2010), API (2011)
and DNV (2016), in addition to the methodology proposed by Klinkvort (2012) for large
diameter piles were calculated for each of the sands studied. As described above, the aim
was to compare these theoretical curves with those obtained experimentally and to discuss
the results obtained. As it was not possible to plot the experimental results, this item
presents only the theoretical curves for the three sands studied here, QZ, CA50, and
CA80.

Figures 84, 85 and 86 show these curves for the four methods mentioned for the
sands QZ, CA50, and CA80, respectively, for the depths of 2.0 m and 4.0 m. It is observed
that the curves of API (2010), API (2011) and DNV (2016) methods are very close and
differ from Klinkvort (2012). As can be seen, the API and DNV methods overestimate
lateral stiffness compared to the equation proposed by Klinkvort (2012). This reinforces
the point previously discussed that the API and DNV methods are not suitable for the
design of monopiles, as they predict different behaviour from what happens in these
structures.

105
(a)

(b)

Figure 84 - p-y curve for QZ sand - z = 2 m (a) and z = 4 m (b)

106
(a)

(b)

Figure 85 - p-y curve for CA50 sand - z = 2 m (a) and z = 4 m (b)

107
(a)

(b)

Figure 86 - p-y curve for CA80 sand - z = 2 m (a) and z = 4 m (b)

108
4.9 RESULTS ANALYSIS AND DISCUSSION

From the results presented above, the following discussion can be made.

In the monotonic test done for the benchmark project, the results of the present
study fit well with the results from the other groups, showing the modelling of the models,
where the same prototype is tested at different stress levels and the response is consistent.

In the tests carried out taking the soil up to failure, or the closest possible to this
condition, some points were verified: the ultimate experimental load for the sands QZ and
CA50 was lower than the theoretical estimate calculated by the Broms Method (1964).
This is possibly related to the installation of the model at 1g. For the CA80 sand, the
ultimate experimental load was higher than the theoretical load, showing a condition of
greater resistance for this sand. It should be noted that the Broms Method (1964) was not
developed based on the behaviour of carbonated sands, considering only parameters such
as friction angle and specific weight of the soil, and the geometry of the pile.

As for cyclic loading, it was observed that the increase in the number of cycles,
from 150 to 4800, increased the ultimate capacity of the soil. The secant stiffness of the
soil also increased during cyclic loading. In terms of design, monopiles are considered to
be subjected to up to 107 cycles over their lifetime, so it is important to consider and
understand well this effect of change in the stiffness of the soil with the increase in the
number of cycles.

The ultimate load of the soil before and after 150 cycles did not change much for
the QZ and CA80 sands. For CA50 sand, the ultimate load increased by 44.44% after the
cycles, which can be due to the densification of the material that was in loose state. The
ultimate load before and after 4800 cycles had a difference for the QZ and CA80 sands,
in which this test was carried out. For QZ sand, there was an increase of 31.25% on the
ultimate load, and for CA80 sand, an increase of 8.93%. The effect of self-healing of the
soil is an aspect that may have occurred and that would explain this increase in resistance
in the final load after cycling.

It is also important to comment on the behaviour of each of the sands. It was


observed that, with the increase in the percentage of CaCO3, there was an increase in soil
resistance. At the level of stresses to which the materials were subjected, it was not

109
observed particle crushing, nor significant difference in the behaviour of carbonated
sands, in addition to what was already expected due to the greater friction angle.

CA50 sand presented a less rigid behaviour in the load-displacement curve than
the other sands, but this is due to the relative density obtained in the tests with this
material. CA80 sand, with 80% CaCO3, showed the most rigid behaviour among the three
materials, and QZ sand, which does not have calcium carbonate, had a less rigid behaviour
than CA80 sand.

In the analysis of secant stiffness during the tests carried out with 4800 cycles, it
was found that the CA80 sand presented greater initial stiffness, which increased over the
cycles. The stiffness of QZ sand, which was lower at the beginning of the cycles,
increased in a greater proportion and became higher than that of CA80 sand.

The process to obtain the experimental p-y curves from the bending moment
profiles along with the depth, however, proved to be complex in the sense of adjusting
the experimental points to a mathematical function that meets the boundary conditions of
the problem. In addition, uncertainties in the displacement measurements, which exist
due to the reduced size of the model, can cause errors in the integration, since the
constants used for the integration are obtained from the displacement measurements.
Also, the double derivation process is greatly affected by the type of adjustment made,
and small errors in the measurement of strain gauges. Besides, the number of pairs of
strain gauges used, and the spacing between them must be correctly considered to meet
the process for obtaining the experimental p-y curves.

110
5 CONCLUSIONS AND SUGGESTIONS FOR FUTURE RESEARCH

5.1 INTRODUCTION

The present work involved the development of a new research project at LM2C,
at COPPE/UFRJ, within the study of monopiles foundations for offshore wind turbines
and it was the first work of the group on renewable energies. It included an extensive and
detailed literature review to understand the subject, to bring this topic of study to the
reality of the beam centrifuge. The entire conception of the tests was performed, as well
as the preparation of the equipment and instrumentation.

Lateral loading tests were carried out on an aluminium model, at a centrifuge


acceleration N = 96g, simulating the prototype of a monopile with 1.84 m of external
diameter and 9.23 m of embedded length in dry sand. A total of 15 tests were performed,
of which 4 were preliminary tests, and 11 final tests.

From the preliminary tests, the equipment, the pluviation, the installation system,
and the fixation between pile and actuator were tested. From these tests, changes were
made until reaching the current system used for the final tests. For this system, the
repeatability and consistency of the tests were verified.

Throughout the test campaign, some limitations were identified regarding the
methodology and equipment used. First, the pluviation process was done manually, which
could bring some uncertainties to the process, it would be ideal to use an automatic
pluviator. Another important aspect is regarding the process of installation of the
monopile, that was also done manually. In addition, the position of the laser used to
measure displacement was possibly too distant from the tube, and that could cause noise
on the readings and some uncertainties to the measurements. It is important to mention
these limitations here, so that future works (specially at LM2C Laboratory) can overcome
these issues.

111
5.2 CONCLUSIONS

On the final testing campaign, three different types of tests were carried out. One
of them was the BCH test, which consisted of a monotonic loading, done in silica sand
with a relative density of 80%. This test is part of a benchmark project, in which other
study groups are involved, modelling the same prototype. When comparing the results of
the present study with the results of these other groups, the consistency of the soil
behaviour was verified, with COPPE’s sand showing a slightly more rigid response.

Another type of test performed consisted of a monotonic loading until failure,


followed by 150 cycles with displacement within the service condition, followed by
another monotonic loading at the end of the test. Three types of sand were used: silica
sand (QZ), carbonated sand with 50% CaCO3 (CA50) and carbonated sand with 80%
CaCO3 (CA80). CA50 sand was tested with a mean relative density of 46%, which did
not allow including this material in the comparison with the QZ and CA80 sands, both
with a mean relative density of 79% and 80%, respectively. For these two materials, it
was observed that the higher the percentage of calcium carbonate, the greater the strength
and stiffness of the material. It is relevant to mention that, as only two contents of calcium
carbonate were tested (under a same relative density), the experimental evidence is
limited, and more tests are recommended to confirm this conclusion. The values of the
experimental failure load found for the QZ and CA80 sands were close to the values found
by the Broms Method. Regarding the failure load before and after 150 cycles, little
difference was observed in the values, which shows little influence of this number of
cycles.

The last type of test was performed only on sands QZ and CA80, with mean
relative densities of 80% and 79%, respectively. It consisted of a monotonic loading until
failure, followed by 4800 cycles with displacement within the service condition, followed
again by a monotonic loading until failure. For this test, an increase in the ultimate load
was observed after the cycles, which was more significant for the QZ sand. There was
also an increase in soil stiffness during cyclic loading. Comparing the two sands, QZ had
lower initial secant stiffness than the CA80 sand, but over the cycles, the increase was
greater for QZ than for CA80. This increase in stiffness over cycles can be approximated
by a logarithmic function.

112
Comparing the ultimate loads obtained in the tests done with 150 and 4800 cycles,
it was observed that, with the increase in the number of cycles, there is an increase in the
failure load after cycling. The failure mechanism also changes with the increase in the
number of cycles, changing from a failure with lower forces at greater displacements, to
a failure with greater force, but at lower displacements.

As for the experimental p-y curves, due to limitations of the instrumentation used,
it was not possible to obtain them. However, a code using Python programming language
was developed that allows obtaining these curves from bending moment results with a
greater number of strain gauges.

Given the context of offshore wind turbines in Brazil, this research contributes on
understanding how the number of cycles influences on the behaviour of the foundation.
As for the response of carbonated sands subjected to lateral loads, no significant
difference was found, but it is necessary to mention the limitations on the use of the
carbonated samples, as it was an artificial material.

5.3 SUGGESTIONS FOR FUTURE RESEARCH

1) Installation of the model in flight, to simulate the same stress level as in


prototype, obtaining more relevant results in terms of design.
2) Cyclic tests with frequencies corresponding to the wind and wave frequencies
to which the monopiles are subjected in the field.
3) Cyclic tests with force control to analyse the change in the stiffness and
accumulation of displacement according to the model by Leblanc et al. (2010).
4) Carry out tests on samples with intermediate levels of calcium carbonate, to
evaluate, more detailed, the influence of this parameter on the response of the
soil to lateral loads, comparing with the dynamic parameters of these
materials.
5) Conduct tests with an instrumented model with a greater number of pairs of
strain gauges to obtain experimental p-y curves from the code developed in
Python.

113
REFERENCES

ABADIE, C. N., 2015. Cyclic Lateral Loading of Monopile Foundations in


Cohesionless Soils, PhD Thesis. University of Oxford, Oxford.

ABADIE, C. N., BYRNE, B. W., HOULSBY, G. T., 2019. “Rigid pile response to cyclic
lateral loading: Laboratory tests”, Geotechnique, v. 69, n. 10, pp. 863–876, 2019.

ABEEÓLICA, 2018. Boletim anual de geração eólica 2018. Available from:


<http://abeeolica.org.br/wp-content/uploads/2019/05/Boletim-Anual_2018.pdf>.
Accessed on: 27/02/2020.

ALMEIDA, M. S. S., OLIVEIRA, W. L., MEDEIROS, C. J., PORTO, E.C., 1987.


“Laboratory Tests and Design Paraments for Offshore Piles in Campos Basin Calcareous
Sands”. In: 6th International on Offshore Engineering, 1987. Laboratory Tests and
Design Paraments for Offshore Piles in Campos Basin Calcareous Sands. pp. 215-225.

ALMEIDA, MARCIO S. S., ALMEIDA, M. C. F., TREJO, P. C., et al., 2014. “The
geotechnical beam centrifuge at COPPE centrifuge laboratory”. In: 8 th international
Conference of Physical Modelling in Geotechnics. Perth.

ALMEIDA, M. S. S., ALMEIDA, M. C. F., OLIVEIRA, J.R.M.S., 2016. “Twenty years


of Centrifuge Modeling at the Federal University of Rio de Janeiro”. In: 3 European
Conference on Physical Modelling in Geotechnics - Eurofuge 2016. Nantes.

AMARANTE, O. A. C. do., BROWER, M., ZACK, J., SÁ, A. L.de, 2001. Atlas do
Potencial Eólico Brasileiro. Rio de Janeiro, Ministério de Minas e
Energia/Eletrobrás/CEPEL.

API (American Petroleum Institute), 2011. Geotechnical and foundation design


considerations. Washington, DC.

ARANY, L., BHATTACHARYA, S., MACDONALD, J. H. G., et al, 2016. “Closed


form solution of Eigen frequency of monopile supported offshore wind turbines in deeper
waters incorporating stiffness of substructure and SSI”, Soil Dynamics and Earthquake
Engineering, v. 83, n. (January), pp. 18–32.

114
BAYTON, S. M., BLACK, J. A., 2016. “The effect of soil density on offshore wind
turbine monopile”, Proceedings of the 3rd European Conference on Physical
Modelling (Eurofuge 2016), v. 1. pp. 245–251.

BAYTON, S. M., BLACK, J. A., KLINKVORT, R. T., 2018. “Centrifuge modelling of


long term cyclic lateral loading on monopiles”. Physical Modelling in Geotechnics, v.
1, pp. 689-694.

BLANCO, M. I., 2009. “The economics of wind energy”, Renewable and Sustainable
Energy Reviews, v. 13, n. 6–7, pp. 1372–1382.

Brasil, 2019. Boletim Mensal de Monitoramento Elétrico. Ministério de Minas e Energia.


Available from:
<http://www.mme.gov.br/documents/239673/909085/1.Boletim+de+Monitoramento+d
o+Sistema+El%C3%A9trico+-+Janeiro+-+2019.pdf/584d1db8-282e-92f9-52b2-
396ef8dc73c5>. Accessed on: 27/02/2020.

BRODBAEK, K. T., MOLLER, M., SORENSEN, S. P. H., AUGUSTESEN, A. H., 2009.


Review of p-y relationships in cohesionless soil. Aalborg: Department of Civil
Engineering, Aalborg University. DCE Technical reports, No. 57. Disponível em <
https://vbn.aau.dk/en/publications/review-of-emp-y-emrelationships-in-cohesionless-
soil>. Accessed on: 10/03/2020.

BROADBENT Inc., 2011. Operating Manual for GT6/0.75 Geotechnical Beam


Centrifuge. H83113 for COPPE/UFRJ, Brazil.

BURD, H. J., TABORDA, D. M. G., ZDRAVKOVIC, L., ABADIE, C. N., BYRNE,


B.W., HOULSBY, G. T., GAVIN, K. G., et al., 2019a. “PISA design model for
monopiles for offshore wind turbines: application to a marine sand”. Géotechnique.
https://doi.org/10.1680/jgeot.18.P.277

BURD, H. J., BEUCKELAERS, W. J. A. P., BYRNE, B. W., GAVIN, K. G., HOULSBY,


G. T., IGOE, D. J. P., JARDINE, R. J., MARTIN, C. M., MCADAM, R. A., MUIR
WOOD, A., POTTS, D. M., SKOV,GRETLUND, J., TABORDA, D. M. G. &
ZDRAVKOVIC´, L., 2019b. “New data analysis methods for instrumented medium scale
monopile field tests”. Géotechnique, https://doi.org/10.1680/jgeot.18.PISA.002.

115
BYRNE, B. W., HOULSBY, G. T., 2003. “Foundations for offshore wind turbines”,
Philosophical Transactions of the Royal Society A: Mathematical, Physical and
Engineering Sciences, v. 361, n. 1813, pp. 2909–2930.

BYRNE, B. W., HOULSBY, G. T., 2015. “Helical piles: An innovative foundation


design option for offshore wind turbines”, Philosophical Transactions of the Royal
Society A: Mathematical, Physical and Engineering Sciences, v. 373, n. 2035.

BYRNE, B. W., MCADAM, R., BURD, H. J., et al., 2015. “New design methods for
large diameter piles under lateral loading for offshore wind applications”, Frontiers in
Offshore Geotechnics III - Proceedings of the 3rd International Symposium on
Frontiers in Offshore Geotechnics, ISFOG 2015, n. (June), pp. 705–710.

BYRNE, B.W., HOULSBY, G. T., BURD, H. J., GAVIN, K. G., IGOE, D. J. P.,
JARDINE, R. J., MARTIN, C. M., MCADAM, R. A., POTTS, D. M., TABORDA, D.
M. G. & ZDRAVKOVIC´, L., 2020. “PISA design model for monopiles for offshore
wind turbines: application to a stiff glacial clay till”. Géotechnique,
https://doi.org/10.1680/jgeot.18.P.255.

BYRNE, B. W., MCADAM, R. A., BURD, H. J. et al., 2017. “PISA: New design
methods for offshore wind turbine monopiles”. In: Proceedings of the 8th International
Conference. pp. 142-161. London.

CANAL ENERGIA, 2019. “ABEEólica: os próximos 10 anos serão virtuosos, porém


desafiadores”. Available from:
<https://www.canalenergia.com.br/noticias/53119824/abeeolica-os-proximos-10-anos-
serao-virtuosos-porem-desafiadores>. Accessed on: 27/02/2020.

CHOO, Y. W., KIM, D., 2016. “Experimental Development of the p-y Relationship for
Large-Diameter Offshore Monopiles in Sands: Centrifuge Tests”, Journal of
Geotechnical and Geoenvironmental Engineering, v. 142, n. 1, pp. 1–12.

CNN, 2019. “The world's largest offshore wind farm is nearly complete. It can power 1
million homes”. Available from: <https://edition.cnn.com/2019/09/25/business/worlds-
largest-wind-farm/index.html>. Accessed on: 27/02/2020.

DASGUPTA, U. S., 2017. Design of Laterally Loaded Monopile Foundations in Sand


for Offshore Wind Turbines, Master thesis. The University of Texas, Austin.

116
DNV (Det Norske Veritas), 2016. Support structures for wind turbines. DNVGL-ST-
0126, Hłvik, Norway.

DYSON, G. J., RANDOLPH, M. F., 2001. “Monotonic lateral loading of piles in


calcareous sand”, Journal of Geotechnical and Geoenvironmental Engineering, v.
127, n. 4, pp. 346–352.

EL HAFFAR, I., 2018. Physical modeling and study of the behavior of deep
foundations of offshore wind turbine. DSc Thesis. L’École Centrale de Nantes, Nantes.

EPE, 2020 Roadmap Eólica Offshore Brasil. Rio de Janeiro. Disponível em


<http://www.epe.gov.br/sites-pt/publicacoes-dados-
abertos/publicacoes/PublicacoesArquivos/publicacao-
456/Roadmap_Eolica_Offshore_EPE_versao_R1.pdf>. Accessed on: 27/02/2020.

FAN, C. C., LONG, J. H., 2005. “Assessment of existing methods for predicting soil
response of laterally loaded piles in sand”, Computers and Geotechnics, v. 32, n. 4, pp.
274–289.

FUTAI, M. M., DONG, J., HAIGH, S. K., et al, 2018. “Dynamic response of monopiles
in sand using centrifuge modelling”, Soil Dynamics and Earthquake Engineering, v.
115, pp. 90–103.

GARNIER, J., GAUDIN, C., SPRINGMAN, S. M., CULLIGAN, P. G. et al, 2007.


“Catalogue of scaling laws and similitude questions in geotechnical centrifuge
modelling”. International Journal of Physical Modelling in Geotechnics. v. 3. pp 1-
23.

GOMES, N. F., 2020. Estudo de parâmetros geotécnicos estáticos e dinâmicos de


areias quartzosa e carbonatadas. Dissertação de Mestrado, COPPE/UFRJ.

GUIMARÃES, M. P. P, 2014. Modelagem centrífuga da movimentação lateral de


dutos em areia. Dissertação de mestrado. Coppe/UFRJ, Rio de Janeiro, Brasil.

GUIMARAES, M. P. P., LUKIANTCHUKI, J. A., OLIVEIRA, J. R. M. S., et al., 2015.


“Centrifuge Modelling of Lateral Soil-Pipe Interaction”. In: XV PANAMERICAN
CONFERENCE ON SOIL MECHANICS AND GEOTECHINICAL
ENGINEERING. BUENOS AIRES.

117
HAIDERALI, A. E., MADABHUSHI, G. P., 2016. Evaluation of Curve Fitting
Techniques in Deriving p–y Curves for Laterally Loaded Piles. Geotechnical and
Geological Engineering. v. 14. pp. 1453-1473.

HAIDERALI, A. E., 2015. Numerical modelling of monopiles for offshore wind


farms. Doctor of Philosophy Thesis. University of Cambridge, Cambridge.

HETENYI, M., 1946. Beams on elastic foundations: Theory with applications in the field
of civil and mechanical engineering, University of Michigan Press.

HOTTA, M., ALMEIDA, M. S. S., PELISSARO, D. T., OLIVEIRA, J. R. M. S., et al.,


2019. “Centrifuge tests for evaluation of submarine-mudflow hydroplaning and turbidity
currents”. International Journal of Physical Modelling in Geotechnics, v. 20, p. 1-39.

IRENA, 2016. International Renewable Energy Agency. Wind Power: Technology


brief. Available from: <https://irena.org/publications/2016/Mar/Wind-Power>.
Accessed on: 27/02/2020.

IX WIND, 2020. Sif Offshore Foundations. Available from:


<https://www.ixwind.com/services/sifoffshorefoundations/>. Accessed on: 27/02/2020.

KING, G. J. W., 1994. “The interpretation of data from tests on laterally loaded piles”. In
Centrifuge International Conference. pp. 515-520. Rotterdam.

KIRKWOOD, P. B., 2015. Cyclic lateral loading of monopile foundations in sand.


DSc Thesis. Churchill College, University of Cambridge, Cambridge.

KLINKVORT, R. T., BLACK, J. A., BAYTON, S. M., et al., 2018. “A review of


modelling effects in centrifuge monopile testing in sand”, Physical Modelling in
Geotechnics, v. 1, pp. 719–724.

KLINKVORT, R. T., 2012. Centrifuge modelling of drained lateral pile-soil response.


PhD Thesis. Technical University of Denmark, Denmark.

KLINKVORT, R. T., LETH, C. T., HEDEDAL, O., 2012. “Centrifuge modelling of


monopiles in dense sand at The Technical University of Denmark”, 2nd European
Conference on Physical Modelling in Geotechnics.

KLINKVORT, R. T., 2020. Personal communication.

118
KONG, L. G., ZHANG, L. M., 2006. Rate controlled lateral-load pile tests using a robotic
manipulator in centrifuge. Geotechnical Testing Journal. v. 30(3). pp. 192-201.

LAU, B. H., LAM, S. Y., HAIGH, S. K., et al., 2014. “Centrifuge testing of monopile in
clay under monotonic loads”, Physical Modelling in Geotechnics - Proceedings of the
8th International Conference on Physical Modelling in Geotechnics 2014, ICPMG
2014, v. 2, n. June 2015, pp. 689–695.

LAU, B. H., 2015. Cyclic behaviour of monopile foundations for offshore wind farms,
Doctor of Philosophy Dissertation, University of Cambridge, Cambridge.

LEBLANC, C. B., 2009. Design of Offshore Wind Turbine Support Structures:


Selected topics in the field of geotechnical engineering. Doctor of Philosophy Thesis.
Aalborg University, Denmark.

LEBLANC, C., HOULSBY, G. T., BYRNE, B. W., 2010. “Response of stiff piles in sand
to long-term cyclic lateral loading”, Geotechnique, v. 60, n. 2, pp. 79–90.

LEHANE, B. M., PEDRAM, B., DOHERTY, J. A., POWRIE, W., 2014. Improved
Performance of Monopiles When Combined with Footings for Tower Foundations in
Sand. Journal of Geotechnical and Geoenviromental Engineering.

LESNY, K., WIEMANN, J., 2005. “Design aspects of monopiles in German offshore
wind farms”, Frontiers in Offshore Geotechnics, ISFOG 2005 - Proceedings of the
1st International Symposium on Frontiers in Offshore Geotechnics, pp. 383–389.

LOBO CARNEIRO, F., 1996. Análise Dimensional e Teoria da Semelhança e dos


Modelos Físicos. Editora UFRJ, 2ª edição.

LOMBARDI, D., BHATTACHARYA, S., MUIR WOOD, D., 2013. “Dynamic soil-
structure interaction of monopile supported wind turbines in cohesive soil”, Soil
Dynamics and Earthquake Engineering, v. 49, pp. 165–180.

MADABHUSHI, G., 2014. Centrifuge Modelling for Civil Engineers. Taylor &
Francis Group. 1st. ed.

MEYER, B. J., REESE, L. C., 1979. Analysis of single piles under lateral loading.
Research report 244-1, Centre for Transportation Research, Bureau of Engineering
Research, University of Texas at Austin, Austin, TX.

119
MIURA, S., TOKI, S., 1982. A sample preparation method and its effect on static and
cyclic deformation-strength properties of sand. Soils and Foundations. v. 22. pp. 61-77.

NBR 12051/1991. ASSOCIAÇÃO BRASILEIRA DE NORMAS TÉCNICAS. Solo -


Determinação do índice de vazios mínimo de solos não-coesivos - Método de ensaio.

NEGRO, V., LÓPEZ-GUTIÉRREZ, J. S., ESTEBAN, M. D., et al., 2017. “Monopiles in


offshore wind: Preliminary estimate of main dimensions”, Ocean Engineering, v. 133,
n. December 2016, pp. 253–261.

NG, C. W. W., 2014. The state-of-the-art centrifuge modelling of geotechnical problems


at HKUST. Journal of Zhejiang University SCIENCE A. v 15. pp 1-21.

NIKITAS, G., VIMALAN, N. J., BHATTACHARYA, S., 2016. “An innovative cyclic
loading device to study long term performance of offshore wind turbines”, Soil Dynamics
and Earthquake Engineering, v. 82, pp. 154–160NUNEZ, I. L., 1988. Driving and
tension loading of piles in sand on a centrifuge. Proceedings International Conference
Centrifuge 88. pp 353-362.

OLIVEIRA FILHO, W., 1987. Considerações sobre ensaios Triaxiais em areias.


Dissertação de mestrado, Coppe/UFRJ, Rio de Janeiro, Brasil.

OLIVEIRA, J. R. M. S., ALMEIDA, M. C. F., ALMEIDA, M. S. S., BORGES, R.G.,


2010. “Physical modelling of Lateral Clay-Pipe Interaction”. Journal of Geotechnical
and Geoenvironmental Engineering, pp. 950-956.

OLIVEIRA, J. R. M. S., RAMMAH, K. I., TREJO, P. C., ALMEIDA, M. S. S.,


ALMEIDA, M. C. F., 2017. “Modelling of a pipeline subjected to soil mass movements”.
International Journal of Physical Modelling in Geotechnics, v. 17, pp. 246-256.

ORTIZ, G. P., KAMPEL, M., 2011. “Potencial de Energia Eólica Offshore na Margem d
Brasil”. In: V Simpósio Brasileiro de Oceanografia. Santos.

PETROBRAS, 2018. “Estamos desenvolvendo o primeiro projeto piloto de energia eólica


offshore do Brasil”. Available from: <https://petrobras.com.br/fatos-e-dados/estamos-
desenvolvendo-o-primeiro-projeto-piloto-de-energia-eolica-offshore-do-brasil.htm>.
Accessed on: 27/02/2020.

120
POULOS, H. G., 1982. Single pile response to cyclic lateral load. Journal of
Geotechnical and Geoenvironmental Engineering. v 108, No. 3, pp 355–375.

REESE, L. C., VAN IMPE, W. F., 2014. Single Piles and Pile Groups Under Lateral
Loading. Taylor & Francis Group. 2ª edição.

REMAUD, D., 1999. Pieux soux charges latérales: Etude expérimentale de l’effect
de group. Tese de Doutorado. Universidade de Nantes, Nantes. 328 p.

ROCSCIENCE, 2018. Laterally loaded piles. Available from: <


https://www.rocscience.com/help/rspile/pdf_files/theory/RSPile_-
_Laterally_Loaded_Pile_Theory.pdf>. Accessed on: 28/02/2020

ROSQUOET, F., THOREL, L., GARNIER, J., et al., 2007. “Lateral cyclic loading of
sand-installed piles”, Soils and Foundations, v. 47, n. 5, pp. 821–832.

ROSQUOET, F., THOREL, L., GARNIER, J., KHEMAKHEM, M., 2010. P-y curves on
model piles Uncertainty identification. Proceedings 7th International Conference on
Physical Modelling in Geotechnics. v. 2. pp. 997-1002.

SALEEM, Z., 2011. “Alternatives and modifications of Monopile foundation or its


installation technique for noise mitigation", Technical report. Delft University of
Technology and North Sea Foundation, p. 1–68, 2011.

SARAMAGO, R. P., 2002. Estudo da Influência da compactação no comportamento


de muros de solo reforçado com a utilização de modelos físicos. Tese de Doutorado.
Universidade Federal do Rio de Janeiro.

SAYLES, S., STONE, K. J. L., DIAKOUMI, M., RICHARDS, D. J., 2018. Centrifuge
model testing of fin piles in sand. Physical Modelling in Geotechnics, v. 1, pp. 743-747.

SOLF, O. KUDELLA, P., TRIANTAFYLLIDIS, TH., 2010. “Investigation of the self-


healing effect of monopile foundations”. In: ICPMG 2010 Physical Modelling in
Geotechnics. v. 2. pp 1003-1008. Zurich.

TABORDA, D. M. G., ZDRAVKOVIC´, L., POTTS, D. M., BURD, H. J., BYRNE, B.


W., GAVIN, K. G., HOULSBY, G. T., JARDINE, R. J., LIU, T., MARTIN, C. M. &
MCADAM, R. A., 2019. “Finite element modelling of laterally loaded piles in a dense
marine sand at Dunkirk”. Géotechnique, https://doi.org/10.1680/jgeot.18.PISA.006.

121
TARAZONA, S. F. M., 2015. Modelagem centrífuga da movimentação lateral e axial
de dutos em leito marinho arenoso. Dissertação de mestrado. PUC Rio. Rio de Janeiro,
Brasil.

TAYLOR R. N., 1995. Geotechnical Centrifuge Technology, Blackie Academic &


Professional, 1st Ed.

TEMPEL, J. V. D., 2006. Design of Support Structures for Offshore Wind Turbines.
PhD Thesis. TU Delft.

VELLOSO, D. de A., LOPES, F. R., 2010. Fundações – Critérios de projeto,


investigação do subsolo, fundações superficiais, fundações profundas. Oficina de
Textos. 1ª ed.

WATSON, P. BRANSBY, F., DELIMI, Z. L., ERBRICH, C., FINNIE, I., KIRSDANI,
H., MEECHAM, C., O’NEILL, M., RANDOLPH, M., RATTLEY, M., SILVA, M.,
STEVENS, B., THOMAS, S., WESTGATE, Z., 2019. “Foundation Design in Offshore
Carbonate Sediments – Building on Knowledge to Address Future Challenges”. In: XVI
Pan-American Conference on Soil Mechanics and Geotechnical Engineering (XVI
PCSMGE). Advance in Soil Mechanics and Geotechnical Engineering. v 7. pp. 240-274.

WIND EUROPE, 2019. Offshore Wind in Europe: Key trends and statistics 2018.
Available from: <https://windeurope.org/wp-content/uploads/files/about-
wind/statistics/WindEurope-Annual-Offshore-Statistics-2018.pdf>. Accessed on:
27/02/2020.

WORLD STEEL, 2012. Steel Solutions in the Green Economy – Wind Turbines.
Available from: < https://www.worldsteel.org/en/dam/jcr:41f65ea2-7447-4489-8ba7-
9c87e3975aab/Steel+solutions+in+the+green+economy:+Wind+turbines.pdf>.
Accessed on: 27/02/2020.

ZDRAVKOVIC´, L., TABORDA, D. M. G., POTTS, D. M., ABADIAS, D., BURD, H.


J., BYRNE, B. W., GAVIN, K. G., HOULSBY, G. T., JARDINE, R. J., MARTIN, C.
M., MCADAM, R. A. & USHEV, E., 2019. “Finite element modelling of laterally loaded
piles in a stiff glacial clay till at Cowden”. Géotechnique,
https://doi.org/10.1680/jgeot.18.PISA.005.

122
ZHU, B., LI, T., XIONG, G., et al., 2016. “Centrifuge model tests on laterally loaded
piles in sand”, International Journal of Physical Modelling in Geotechnics, v. 16, n.
4, pp. 160–172.

4C Offshore, 2016. “DONG to dismantle Vindeby”. Available from:


<https://www.4coffshore.com/news/dong-to-dismantle-vindeby-nid3279.html>.
Accessed on: 27/02/2020

123

Você também pode gostar