Jmse 09 00235 v2

Journal of
Marine Science
and Engineering
Article
Effect of Horizontal Multidirectional Cyclic Loading on Piles
in Sand: A Numerical Analysis
Orianne Jenck * , Armita Obaei, Fabrice Emeriault and Christophe Dano
Grenoble Alpes University, CNRS, Grenoble INP, 3SR, 38000 Grenoble, France; armita.obaei@gmail.com (A.O.);
fabrice.emeriault@3sr-grenoble.fr (F.E.); christophe.dano@3sr-grenoble.fr (C.D.)
* Correspondence: orianne.jenck@3sr-grenoble.fr
Abstract: Foundations of offshore and nearshore wind energy production systems are subjected to
multidirectional and cyclic loads, due to the combined action of wind and waves and in the particular
case of mutualized anchor foundations for floating wind turbines, to the phase shift between the loads
generated in the adjacent anchored turbines. This article presents a three-dimensional numerical
model developed with FLAC3D to analyse the impact of the change in direction of the horizontal
load during the cycles. The typical case of a 1.7 m diameter and 10 m-long pile founded in a
dense homogeneous sand is considered. A specific procedure has been implemented to apply force-
controlled cycles with a change in lateral load direction. The results are compared to mono-directional
lateral cyclic loads with the same average and cyclic forces. The results of the parametric study
highlight the effect of the average value and amplitude of the cyclic loading on the accumulation
of pile head horizontal displacements during the cycles. When a multidirectional cyclic loading is
applied, it also leads to an accumulation of the deviated horizontal displacements, and the resulting
accumulated horizontal displacements are larger than for a mono-directional cyclic loading of the

same amplitude.
Citation: Jenck, O.; Obaei, A.;
Emeriault, F.; Dano, C. Effect of
Keywords: numerical modelling; pile; lateral cyclic loading; multidirectional loading; offshore
Horizontal Multidirectional Cyclic
wind turbine
Loading on Piles in Sand: A
Numerical Analysis. J. Mar. Sci. Eng.
2021, 9, 235. https://doi.org/
10.3390/jmse9020235
1. Introduction
Academic Editor: Luke J. Prendergast Wind power offers an interesting source of energy, leading to the current development
of onshore, nearshore and offshore wind farms. Compared to onshore, nearshore and
Received: 26 January 2021 offshore wind turbines benefit from a number of advantages such as limited visual impact,
Accepted: 18 February 2021 higher and more constant wind speed, and the possibility of using large capacity turbines,
Published: 23 February 2021 resulting in higher power generation. There are several foundation concepts for offshore
wind turbines, which specifically depend on the loading, water depth and soil conditions:
Publisher’s Note: MDPI stays neutral gravity-base; monopole, tripile, tripod, jacket or anchor for floating structures [1,2]. In
with regard to jurisdictional claims in the case of floating wind turbines, the structures are connected to the foundation system
published maps and institutional affil- through several anchor lines and the cyclic load then acts in multiple directions, due to
iations. the combined action of wind and waves. In the particular case of mutualized anchor
foundations for floating wind turbines, the foundations are also subjected to the phase shift
between the loads generated in the adjacent anchored turbines. This study thus focuses
on the case of piles as mutualized anchors for floating wind turbines, thus submitted to
Copyright: © 2021 by the authors. multidirectional lateral cyclic loads.
Licensee MDPI, Basel, Switzerland. When a pile is subjected to lateral loads, one of the essential features is known to be
This article is an open access article the pile bending stiffness compared to the surrounding soil stiffness and strength. Short
distributed under the terms and rigid piles are mainly submitted to displacement and rotation, whereas for flexible piles,
conditions of the Creative Commons the soil–pile behaviour is affected by the flexibility of the pile (Figure 1).
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
J. Mar. Sci. Eng. 2021, 9, 235. https://doi.org/10.3390/jmse9020235 https://www.mdpi.com/journal/jmse

J.J.Mar.
Mar.Sci.
Sci.Eng. 2021,9,9,235
Eng.2021, x FOR PEER REVIEW 22 of
of2223
Figure1.1.Rigid
Figure Rigidvs.
vs.flexible
flexiblepile
pilesubmitted
submittedto
tolateral
lateralload.
load.
Poulos
Poulosand
andHull
Hull(1989)
(1989)[3]
[3]defined
definedthetherigidity
rigidityparameter
parameterRR(Equation
(Equation(1)),
(1)),depending
depending
on
onthe
thepile
pileYoung modulusEEp pand
Youngmodulus andquadratic momentIpIp,, and
quadraticmoment and on
on the
thesoil “stiffness”EEs .s.
soil“stiffness”
They
Theysuggested
suggestedthat
thataapile
pilewith
withaalength
lengthless
lessthan
than1.48
1.48RRbehaves
behavesrigidly,
rigidly,and
andififthe
thelength
length
exceeds
exceeds4.44
4.44R,
R,the
thepile
pilebehaves
behavesflexibly.
flexibly.However,
However,one oneshould
shouldnote
notethat
thatthe
thechoice
choiceofofaa
homogeneous
homogeneoussoil stiffnessEEs sis
soilstiffness is still
still problematic
problematicin inpractice:
practice:
.
E p𝐸I p 𝐼 0.25

R =𝑅 (1)
(1)
Es 𝐸
The piles under lateral load are traditionally designed using the p–y curve concept
The piles under lateral load are traditionally designed using the p–y curve concept [4,5],
[4,5], which formulates the soil response as uncoupled non-linear springs (Winkler ap-
which formulates the soil response as uncoupled non-linear springs (Winkler approach),
proach), relating the soil pressure p acting against the pile shaft and the corresponding
relating the soil pressure p acting against the pile shaft and the corresponding lateral
lateral displacement y of the pile. However, the concept has been developed based on
displacement y of the pile. However, the concept has been developed based on semi-
semi-empirical
empirical relations
relations for flexible
for flexible pilesmonotonic
piles under under monotonic
loading loading
and has and
somehas some limita-
limitations for
tions for large diameter piles, as particularly highlighted by numerical modelling
large diameter piles, as particularly highlighted by numerical modelling approaches [6,7]. ap-
proaches
Indeed, [6,7].aIndeed,
when when rigidly,
pile behaves a pile behaves
a passiverigidly,
wedge a passive wedge of
of soil beneath thesoil beneath
point of zerothe
point of zero deflection will develop, an aspect which is not considered
deflection will develop, an aspect which is not considered in the current methodology. in the current
methodology.
Furthermore, design methods of these offshore piles and their response to cyclic
lateralFurthermore,
loading are an design methods
important of these
issue in theoffshore piles and
geotechnical their response
engineering to cyclic
practice and haslat-
eral loading are an important issue in the geotechnical engineering practice
been addressed by several studies around the world, mainly focusing on mono-directional and has been
addressed by several studies around the world, mainly focusing on mono-directional
loading.
loading.
Among others, experimental testing studies in centrifuges have been reported in [8–10].
Most ofAmong others,
the studies areexperimental
synthetized intesting studies Project
the SOLCYP in centrifuges have been reported
recommendations in [8–
[11]. All these
10]. Most of the studies are synthetized in the SOLCYP Project recommendations
studies have highlighted the lateral displacement accumulation during the cycles. Based [11]. All
these studies have highlighted the lateral displacement accumulation
on field-scale piles, Lin and Liao (1999) [12] proposed a logarithmic law as during the cycles.
Based on field-scale piles, Lin and Liao (1999) [12] proposed a logarithmic law as
yN /y1 − 1 = α lnN (N ≥ 1) (2)
yN/y1 – 1 = α lnN (N ≥ 1) (2)
where
whereyy1 1and
and yyNN are
are the
the pile
pilehead
headlateral
lateraldisplacement
displacementatatfirstfirstloading
loadingandandatatNth
Nthcycle,
cycle,
respectively,
respectively,andandααisisaaparameter
parameterwhich
whichthusthusrepresents
representsthe thedisplacement
displacementaccumulation
accumulation
due
dueto tothe
thecycles.
cycles.This
Thisparameter
parameterααhighly
highlydepends
dependson onthe
theloading
loadingtype,
type,cyclic
cyclicamplitude,
amplitude,
pile installation, soil properties, pile embedment length and pile–soil relative
pile installation, soil properties, pile embedment length and pile–soil relative stiffness. stiffness.
For
For
instance, Verdure et al. (2003) [13] presented a linear relationship of α value with the with
instance, Verdure et al. (2003) [13] presented a linear relationship of α value cyclic
the cyclic amplitude,
amplitude, based on based on the centrifuge
the centrifuge tests of pile
tests of a flexible a flexible
in a drypile in a dry
dense sanddense
undersand
one-
under one-way cyclic loading. They proposed α = 0.18
way cyclic loading. They proposed α = 0.18 ΔH/Hmax where ΔHmax ∆H/H where ∆H is the cyclic
is the cyclic amplitude and
amplitude and Hmax is the maximum loading. They also mentioned that the load cycles
Hmax is the maximum loading. They also mentioned that the load cycles have much more
have much more effect on the pile displacement when the maximum applied load is closer
effect on the pile displacement when the maximum applied load is closer to the ultimate
to the ultimate lateral static resistance of the pile. Concerning the effect of pile–soil relative
lateral static resistance of the pile. Concerning the effect of pile–soil relative stiffness, the
stiffness, the effect of cycles is more important for flexible piles than for rigid piles [14]. For
effect of cycles is more important for flexible piles than for rigid piles [14]. For rigid piles,
rigid piles, the value of Ep Ip has no more influence [11].
the value of Ep Ip has no more influence [11].
Based on the expression proposed by Rosquoët et al. (2007) [15] obtained from
Based on the expression proposed by Rosquoët et al. (2007) [15] obtained from exper-
experimental results on flexible piles in sand, the SOLCYP recommendations [11] propose
imental results on flexible piles in sand, the SOLCYP recommendations [11] propose the
J. Mar. Sci. Eng. 2021, 9, 235 3 of 22
the following logarithmic law for the case of intermediate or rigid piles in sand, thus also
accounting for the pile bending stiffness:
yN /y1 − 1 = 0.102/CR × ln(N) × (Hc /Hmax )0.35

(3)
N≥1
where CR is a rigidity coefficient defined as
CR = [Ep Ip /(Ep Ip )fl ]1/5 (4)
(Ep Ip )fl corresponds to the limit value of the pile bending stiffness Ep Ip for flexible
piles and should be first assessed by analysing the effect of the pile bending stiffness on the
pile head displacement under monotonic horizontal loading. (Ep Ip )fl corresponds to the
value below which the bending stiffness has a large impact on the pile head displacement.
For flexible piles, CR = 1. Hmax and Hc are, respectively, the maximum lateral load and
half of the amplitude of the cyclic loading (see Equations (5) and (6)).
An alternative or additional approach to experimental and semi-empirical investiga-
tion consists of implementing numerical models. In this approach, special attention should
be paid to the modelling of the soil (and pile–soil interface), by using relevant constitutive
models, even under monotonic loading [16]. A few numerical approaches under lateral
cyclic loading have been proposed. For instance, Achmus et al. (2009) [17] implemented
an approach termed “degradation stiffness model” to account for long-term cyclic lateral
loading of piles. Model parameters were assessed from cyclic triaxial tests results with up
to 104 cycles, and numerical results were compared to test results of pile in sand, showing a
good agreement. The numerical parametric study thus permitted to develop design charts.
Nevertheless, most of these studies under lateral cyclic loading focus on the case of
mono-directional loading. The behaviour of the piles under lateral loading that varies in
direction and amplitude has been little studied to date, and even less under cyclic loading
conditions. Su and Li (2013) [18] investigated the effect of multidirectional lateral loading
using a finite-element modelling approach, by applying an imposed trajectory at pile
head (with no unloading). They obtained that the lateral resistance of the pile is lower
under multidirectional than under mono-directional loading. In this latter, the directions
of the force increment vector and lateral displacement increment vectors are generally
non-coaxial. Lovera et al. (2020) [19] extended the framework of p–y curve method to
multidirectional loading, and they obtained a misalignment between load direction and
total displacement when the loading direction changes.
Concerning the effect of multidirectional lateral loading under cyclic conditions, a few
experimental studies were conducted. Mayoral et al. (2016) [20] focused on the case of a
pile in soft clay under seismic loading using an experimental modelling approach. They
highlighted the influence of the cyclic loading path on the pile response and they noted
that the concept of p–y curve could not simply be applied in two orthogonal directions.
Rudolph et al. (2014) [21] performed centrifuge experiments of a large-diameter pile in sand,
with up to 13,000 cycles and a variation of loading direction of 0◦ , 30◦ , 60◦ or 90◦ . They
highlighted the significant increase in lateral displacement accumulation, even under load
levels representative of serviceability conditions, which makes the current guidelines—as
in the American Petroleum Institute (API) recommendations [4]—non-conservative.
Experimental studies are complex to perform, expensive and time consuming, even
more under complex loading paths, with cycles and loading direction changes, so that
complete parametric studies under such complex and multiple conditions are difficult to
conduct. Numerical studies thus can help understanding the behaviour under complex
conditions and highlight governing features. Moreover, numerical studies allow accessing
information difficult to obtain experimentally, such as the bending moment in the pile.
However, no numerical study of pile subjected to lateral multidirectional cyclic loading
has been reported in the literature to date.
J. Mar. Sci. Eng. 2021, 9, x FOR PEER REVIEW 4 of 23
However, no numerical study of pile subjected to lateral multidirectional cyclic loading

J. Mar. Sci. Eng. 2021, 9, 235 has been reported in the literature to date. 4 of 22
The objective of this work was to develop a numerical approach to acquire

knowledge concerning the system behaviour and governing parameters, in the case of
multi-directional
The objectiveand cyclic
of this work lateral
was toloading
developasa numerical
encountered for mutualized
approach anchoring
to acquire knowledge
piles. The aim was to be able to estimate the accumulated pile displacements,
concerning the system behaviour and governing parameters, in the case of multi-directionalin particular
atand
thecyclic
mudline, and the accumulated pile rotation and/or bending, in such
lateral loading as encountered for mutualized anchoring piles. The aim was complex load- to
ing
be situations.
able to estimate the accumulated pile displacements, in particular at the mudline, and
To achieve these
the accumulated pile objectives, a conceptual
rotation and/or bending, 3Dinnumerical
such complexmodel usingsituations.
loading the commercial
finite difference
To achieve these objectives, a conceptual 3D numerical model using thesurrounding
software Flac3D [22] was developed, simulating the pile, the commercial
ground mass andsoftware
finite difference their interface. Specific
Flac3D [22] was loading procedures
developed, arethe
simulating implemented to apply
pile, the surrounding
mono-directional
ground mass and their interface. Specific loading procedures are implemented to load-
and then multi-directional cyclic loadings. The multi-directional apply
ing consists of varying
mono-directional and thenthe multi-directional
direction of the lateral load during
cyclic loadings. Thethe unloading–reloading
multi-directional loading
process.
consists of varying the direction of the lateral load during the unloading–reloading process.
2.2.Materials
Materialsand
andMethods
Methodsfor
forthe
theNumerical
NumericalModelling
Modelling
2.1. Model Specifications
2.1. Model Specifications
InInthe
thecase
caseofof mutualized
mutualized anchor
anchor piles
piles for for floating
floating windwind turbines,
turbines, each each pilecon-
pile was was
connected to several turbines (Figure 2) and each turbine was connected
nected to several turbines (Figure 2) and each turbine was connected to several piles. Into several piles. In
the case depicted in Figure 2, the horizontal load applied to a pile head
the case depicted in Figure 2, the horizontal load applied to a pile head (H) thus results (H) thus results
fromthe
from thetension
tensionapplied
appliedbyby both
both lines
lines (H
(H11 and
and H H22).).HH1 1and
andHH2 2both
bothvary
varyinintime,
time,more
moreor or
lessindependently
less independentlyfrom fromthe
theother,
other,thus
thusresulting
resultingin inaamultidirectional
multidirectionalcyclic
cyclicvertical
vertical(not
(not
investigated here) and horizontal loading: H = f(ω, t), where ω is the loading
investigated here) and horizontal loading: H = f(ω, t), where ω is the loading direction (or direction (or
azimuth) and t, time.
azimuth) and t, time.
Figure
Figure2.2.Concept
Conceptof
ofmutualized
mutualized pile foundationsfor
pile foundations forfloating
floatingwind
windturbines
turbines(conceptual
(conceptual orienta-
orientations).
tions).
The reference case considered in this study is taken from a relevant case proposed by
TheEnergies
France referenceMarines
case considered in this
[23], in order to study
proceed is taken from
with the a relevant case
development proposed
of the by
numerical
France
model.Energies
Figure Marines
3a depicts[23], in order
this relevant to proceed
case in with
termstheofdevelopment
the horizontal of loading
the numerical
on an
model.
anchoredFigure
pile 3a depicts
(named this
A1), relevant
induced bycase
two in terms oflines
anchored the (M1-L4
horizontal
andloading
M2-L1) on an an-
connected
chored pile (named A1), induced by two anchored lines (M1-L4 and M2-L1) connected
to two floating turbines (M1 and M2). In this case, the resulting horizontal force (H) at pile to
two
headfloating
varies turbines
between(M1 Hmax and M2).MN
= 1.77 In this
downcase,
to the
Hminresulting horizontal
= 1.25 MN, with aforce
change (H)of
at force
pile
head
direction 30◦ in a horizontal
variesofbetween Hmax = 1.77 plane
MN down (azimuthto Hchange).
min = 1.25Figure
MN, with a change
3b depicts theof force di-
considered
rection
loadingofpath
30° in
H aashorizontal
a functionplane
of the(azimuth
azimuth change).
ω adopted Figure 3bnumerical
in the depicts the considered
study. In this
loading path study,
exploratory H as aunloading
function of andthereloading
azimuth follow
ω adopted in the
the same numerical
path. The case study.
of theInmono-
this
exploratory
directional study,
loadingunloading
of the sameandloading
reloading follow the
amplitude (Hsame path.1.25
between Theand
case1.77
of the
MN,mono-
thus
with ω = constant,
directional loading ofequal
the to 0◦ on
same Figureamplitude
loading 3b) is first(Hconsidered
between in order
1.25 andto1.77
(i) proceed
MN, thus to
a numerical
with model
ω = constant, validation
equal to 0° onunder
Figurea 3b)more “conventional”
is first considered in situation;
order to(ii)
(i) highlight
proceed tothe a
impact of loading azimuth change on the pile response afterwards.
J. Mar. Sci. Eng. 2021, 9, 235 5 of 22

numerical model validation under a more “conventional” situation; (ii) highlight the im-
pact of loading azimuth change on the pile response afterwards.
(a)
(b)
Figure
Figure 3. 3.Multidirectional
Multidirectionalhorizontal
horizontal loading
loading path
pathapplied
appliedonona apile head.
pile (a) (a)
head. Relevant loading
Relevant casecase
loading applied on pile
applied onanchor
pile
A1 by 2 anchor lines M2-L1 and M1-L4, adapted from [23]. (b) Reference case of the loading path for the current
anchor A1 by 2 anchor lines M2-L1 and M1-L4, adapted from [23]. (b) Reference case of the loading path for the current numerical
study.
numerical study.
AsAsanan initialnumerical
initial numericalparametric
parametricstudy,
study, the
the impact
impact ofof the
the average
average horizontal
horizontal force
force
and amplitude of mono-directional cycling loading (i.e., ω
◦ = 0°) is investigated before-
and amplitude of mono-directional cycling loading (i.e., ω = 0 ) is investigated beforehand.
hand.
The average lateral load Hm and half-amplitude of cyclic loading Hc are defined as
The average lateral load Hm and half-amplitude of cyclic loading Hc are defined as
Hmax + Hmin
Hm = 𝐻 𝐻 (5)
𝐻 2 (5)
2
Hc =
Hmax𝐻 − Hmin𝐻 (6)(6)
𝐻 2 2
The reference case corresponds thereby to Hm = 1.51 MN and Hc = 0.26 MN. The
Hc /Hmax ratio (cf. Equation (3)) is equal to 0.147. Additionally, the ultimate lateral loading
J. Mar. Sci. Eng. 2021, 9, 235 6 of 22

The reference case corresponds thereby to Hm = 1.51 MN and Hc = 0.26 MN. The
Hc/Hmax ratio (cf. Equation (3)) is equal to 0.147. Additionally, the ultimate lateral loading
for the case study is numerically determined to equal Hlim = 4.3 MN (see Section 3.1) in
for the to
order case study is
compute thenumerically determined
average force ratio (Hm/H ) and H
tolimequal lim = amplitude
cyclic 4.3 MN (see Section
ratio (Hc/H3.1)
lim).
in
order The
to compute the average force ratio (H
numerical parametric study conductedm /H ) and cyclic amplitude ratio (H c
lim is summarized in Table 1 and Figure /Hlim ). 4.
The numerical parametric study conducted is summarized in Table 1 and Figure 4.
The reference case is “case 1”. Cases with a larger cyclic amplitude than the reference case
The reference case is “case 1”. Cases with a larger cyclic amplitude than the reference case
were mainly investigated (cases 3–7), whereas case 2 is for a smaller amplitude. The effect
were mainly investigated (cases 3–7), whereas case 2 is for a smaller amplitude. The effect
of the average force value for a fixed amplitude can be explored by comparing cases 3, 4
of the average force value for a fixed amplitude can be explored by comparing cases 3, 4
and 5, and cases 6 and 7. Figure 4 also highlights that only “one-way” cycles are applied
and 5, and cases 6 and 7. Figure 4 also highlights that only “one-way” cycles are applied
(no load reversal).
(no load reversal).
Table 1. Parametric study under mono-directional cyclic loading.
Table 1. Parametric study under mono-directional cyclic loading.
Avg. force ratio
Avg. Force Ratio Hc (MN)
Cyclic amplitude ratio
Cyclic Amplitude
Case
Case H
Hmax (MN)
(MN) HHmin (MN)
(MN) HmH(MN)
(MN) H (MN) HcH /Hmax
/Hcmax
max min m HmH/H lim
m /Hlim
c HcH
Ratio /Hc /H
lim
lim
11 1.77
1.77 1.25
1.25 1.51
1.51 0.35
0.35 0.26
0.26 0.06
0.06 0.147
0.147
22 1.77
1.77 1.50
1.50 1.64
1.64 0.38
0.38 0.14
0.14 0.03
0.03 0.08
0.08
33 1.77
1.77 1.00
1.00 1.39
1.39 0.32
0.32 0.39
0.39 0.09
0.09 0.220
0.220
4 2.02 1.25 1.64 0.38 0.39 0.09 0.193
4 2.02 1.25 1.64 0.38 0.39 0.09 0.193
5 2.27 1.50 1.89 0.44 0.39 0.09 0.172
65 2.27
1.77 1.50
0.64 1.89
1.21 0.44
0.28 0.39
0.57 0.09
0.13 0.172
0.322
76 1.77
2.57 0.64
1.44 1.21
2.01 0.28
0.47 0.57
0.57 0.13
0.13 0.322
0.222
7 2.57 1.44 2.01 0.47 0.57 0.13 0.222
Figure4.4.Parametric
Figure Parametricstudy
studyunder
undermono-directional
mono-directionalcyclic
cyclicloading:
loading:cases
cases1–7.
1–7.
Theperiod
The period of of loading
loading cycles is around
around 10 10 s.s. InInthe
thenumerical
numericalmodel, model,nono dynamic
dynamic ef-
fect is thus considered—i.e.,
effect is thus considered—i.e., cyclic loading is applied under quasi-static conditions,
applied under quasi-static conditions, and and
calculationwill
calculation willbe
beperformed
performedunder underdrained
drainedconditions,
conditions,all allthethemore
morethat
thataasand
sandstratum
stratum
isisenvisaged.
envisaged.
Thecase
The caseofof1.7
1.7m-diameter
m-diameter(D (D==1.71.7m) m)and and10 10m-long
m-longpile pile(L(L==1010m)m)founded
foundedininaa
densehomogeneous
dense homogeneoussand sandmass
masswaswasexamined.
examined. The The pile
pile isis aahollow
hollow pile
pilemade
madeof ofsteel,
steel,
with
withaawall
wallthickness
thicknessofof35 35mmmm(internal diameterDDintint = 1.63 m). With Epp ==200
(internaldiameter 200GPa
GPaand anda
apile
pilequadratic momentIpI=p =
quadraticmoment × –D 4
(D int–D 4 ) = 0.0635 4
m flexural
, the flexural pile rigidity
π/64
π/64 × (D 4 4 ) int
= 0.0635 m , the
4 pile rigidity is Ep ×isIp
× Ip ×10
E=p 1.27 = 1.27
10 N.m². 10 N·m2 . The soil “stiffness” E is taken equal to 50 MPa. The value of
× 10The soil “stiffness” Es is taken sequal to 50 MPa. The value of R of equa-
Rtion
of Equation
(1) is then(1)
R =is4.0then
andRthe
= 4.0
pileand the pile
length is L =length
10 m =is2.5 L R.
= 10 The mpile
= 2.5 R. The
length pilebetween
being length
being between
1.48 and 4.44 R,1.48
thisandpile4.44 R, this pile
is therefore is therefore
supposed supposed
to have a medium to have a medium
relative relative
flexibility [3].
flexibility [3].
Concerning the soil mass, a dense and homogenous layer of sand is considered. The
main Concerning
geotechnical thecharacteristics
soil mass, a dense
that and
will homogenous
be consideredlayer of sand the
to calibrate is considered. The
soil constitutive
main geotechnical characteristics that will be considered to calibrate the soil constitutive
model for the numerical analysis are its submerged unit weight γ’ = 10.2 kN/m3 ; the
J. Mar. Sci. Eng. 2021, 9, 235 7 of 22
model for the numerical analysis are its submerged unit weight γ’ = 10.2 kN/m3; the max-
imum friction angle φ = 37°; the◦ cohesion c = 0 kPa; the dilation angle Ψ = φ – 30 = 7; and
maximum friction angle ϕ = 37 ; the cohesion c = 0 kPa; the dilation angle Ψ = ϕ − 30 = 7;
a deformation modulus Es = 50 MPa.
and a deformation modulus Es = 50 MPa.
2.2.
2.2. Numerical
Numerical Model
Model
The
The issue
issue addressed
addressedin inthis
thisstudy
studyisisaatypical
typicalsoil–structure
soil–structureinteraction
interactionproblem.
problem. It is
It
handled here by implementing a numerical method termed
is handled here by implementing a numerical method termed a “direct method”, since a “direct method”, since the
soil mass
the soil and and
mass the structure
the structure elements are part
elements of the
are part ofsame
the same model, here here
model, in a continuum.
in a continuum. The
numerical model is developed with the commercial
The numerical model is developed with the commercial finite difference program finite difference program FLAC3D
(Fast Lagrangian AnalysisAnalysis of Continua) v. 5.01 [22]. The code uses an explicit explicit numerical
numerical
scheme
scheme thatthatsolves
solvesthe thedynamic
dynamicequationsequations of ofmotion,
motion, even for for
even static problems,
static in conjunc-
problems, in con-
tion with with
junction an incremental
an incremental constitutive law over
constitutive lawa over
smallatime small step,
time at step,
discrete points inpoints
at discrete space
(grid-points). This method
in space (grid-points). Thisismethod
particularly well adapted
is particularly wellfor analysing
adapted the non-linear
for analysing the non-be-
haviour of soils and soil–structure interactions. Because no
linear behaviour of soils and soil–structure interactions. Because no matrices are formed, matrices are formed, large
three-dimensional
large three-dimensional calculations
calculations can can
be made
be made without
without excessive
excessivememorymemoryrequirements.
requirements.
Moreover, the internal programming language (FISH) and command-driven mode permit
applying the the desired
desiredloading
loadingpath. path.This
Thisisisa apowerful
powerful feature
feature when when dealing
dealing with a complex
with a com-
multidirectional
plex multidirectional loading process,
loading are addressed
process, are addressed in thisinstudy.
this study.
The numerical
numerical model modelininaacontinuum
continuumconsistsconsists of of
a soil mass
a soil masswith boundary
with boundaryconditions
condi-
tions in which a pile made of zone elements is embedded (Figure 5). For the case of mono-
in which a pile made of zone elements is embedded (Figure 5). For the case mono-
directional horizontal loading of the pile, only half of the soil mass needs to be simulated
due to symmetry
symmetry condition
condition(loading
(loadingHHisisonly onlyapplied
appliedininthe thex xdirection
direction defined
definedin in
Figure
Figure 5)
and
5) andin in
order
ordertotooptimize
optimizethe thecalculation
calculationduration.
duration.On Onthe thecontrary,
contrary,the the multi-directional
multi-directional
loading case
loading case requires
requires the the implementation
implementation of of aa full
full model.
model. The The boundary
boundary conditions
conditions consist
consist
of node fixity in the x direction for the vertical planes perpendicular to the x direction, in
of node fixity in the x direction for the vertical planes perpendicular to the x direction, in
the y
the y direction
direction forfor the
the vertical
vertical planes
planes perpendicular
perpendicular to to the
the y y direction
direction (also(also valid
valid for
for the
the
symmetry plane
symmetry plane ininthethecase
caseofofthe thehalf
halfmodel)
model)and and inin
allall
directions
directions forfor
thethehorizontal
horizontalbottom
bot-
plane.
tom The The
plane. global size size
global of theof model
the modeland and
the size of the
the size of zones
the zones (mesh refinement)
(mesh were
refinement) the
were
subject of a preliminary parametric study performed by Obaei
the subject of a preliminary parametric study performed by Obaei (2020) [24]. The opti- (2020) [24]. The optimized
modelsmodels
mized are depicted
are depictedon Figure 5. The
on Figure mesh
5. The meshis finer
is finernearnearthethepile,
pile,and
andthethesize
size of
of the
the
zones regularly increases when radially moving away from
zones regularly increases when radially moving away from the pile axis. The model width the pile axis. The model width
is equal
is equal to to 30
30 D,
D, which
which was was sufficient
sufficient to to prevent
prevent any any boundary
boundary effect effect under
under pile
pile lateral
lateral
loading. The model for mono-directional loading (half model) contains 11,472(10,464
loading. The model for mono-directional loading (half model) contains 11,472 zones zones
for the soil and 1008 for the pile) and 12,956 grid-points; the
(10,464 for the soil and 1008 for the pile) and 12,956 grid-points; the model for multi-di- model for multi-directional
loading (full model) contains 22,944 zones (20,928 for the soil and 2016 for the pile) and
rectional loading (full model) contains 22,944 zones (20,928 for the soil and 2,016 for the
24,847 grid-points. The total height of the model could be optimized as only horizontal
pile) and 24,847 grid-points. The total height of the model could be optimized as only
loading is applied on the pile head, and it is here equal to 17.85 m, i.e., 7.85 m below the
horizontal loading is applied on the pile head, and it is here equal to 17.85 m, i.e., 7.85 m
10 m pile toe.
below the 10 m pile toe.
Figure
Figure 5.5. Numerical
Numericalmodels
modelsfor
forthe
themono-directional
mono-directionallateral loading
lateral (left)
loading andand
(left) multi-directional
multi-directional
lateral loading (right).
lateral loading (right).
J.J.Mar.
Mar.Sci.
Sci.Eng. 2021,9,9,235
of22
23
The pile
The pile is actually
actually aahollow
hollowcylinder
cylinderbut but inin
thethe
present
present study it isitsimulated
study is simulated as a solid
as a
solid cylinder
cylinder withwith an equivalent
an equivalent bending
bending stiffness
stiffness (Ep (E × for
× pIp), Ip ),modelling
for modelling simplification
simplification rea-
reasons.
sons. The The pile
pile is ismade
madeofofcylindrical
cylindricalzone
zone elements.
elements. The The quadratic moment for for aasolid
solid
cylinder
cylinder ofof diameter
diameter D D being Ip_eq π/64 ×
p_eq ==π/64×D D4 , equivalent
4, the the equivalent YoungYoungmodulusmodulus for theforsolid
the
solid pile, is equal to
pile, is equal to Ep_eq = 31E = 31 GPa. The 10 m-long pile is made of 20
GPa. The 10 m-long pile is made of 20 slices of vertical zones plus
p_eq slices of vertical
zones plus an additional
an additional 0.1 m-thick0.1 m-thickabove
element, element,the above the soilinsurface,
soil surface, order to inapply
order theto apply the
pile head
pile head as
loading, loading,
depicted as on
depicted
Figureon Figure 6. Moreover,
6. Moreover, in order toinfacilitate
order tothe facilitate
analysis theinanalysis
terms of
in terms of
bending bendinggenerated
moments moments in generated in the pile
the pile during the during the lateral
lateral loading, loading,beam
a vertical a vertical
made
beam
of 20 structural elements is generated in the pile axis (Figure 6). This beam has low Eb has
made of 20 structural elements is generated in the pile axis (Figure 6). This beam and
low Eb and Ib characteristics
Ib characteristics (compared(compared
to Ep and Ito Ep and Iin
p values), p values),
order not in order not the
to affect to affect
pile the pile
bending
bending
stiffness,stiffness,
but has but has the
the same same deflection
deflection as the
as the pile. pile. According
According to the Bernoulli
to the Bernoulli beam
beam theory,
theory, the pile bending moment M is then equal to M ×
the pile bending moment Mp is thenp equal to Mb × (Ep × Ibp)/(Eb p× Ib), where (E × Ip )/(E × I
b Mb is ), where
b the bendingMb
ismoment
the bending moment
directly directly
recorded in therecorded
numerical in the numerical
model using the model beamusing the beam
element. element.
Interfaces are
Interfaces are generated between the pile and the soil
generated between the pile and the soil (at the pile shaft and toe). (at the pile shaft and toe).
Figure6.6.Numerical
Figure Numericalmodelling
modellingof
ofthe
thepile.
pile.
2.3.
2.3.Constitutive
ConstitutiveModels
Models
When
Whendealing
dealingwithwitha asoil–structure
soil–structure interaction
interaction problem
problem using numerical
using numerical modelling
modellingin
aincontinuum,
a continuum, the the
choices of material
choices of materialandand
interface constitutive
interface models
constitutive and and
models the calibration
the calibra-
of their
tion parameter
of their are essential.
parameter are essential.
As
As already discussed,the
already discussed, thepile–soil
pile–soilinteraction
interactionbehaviour
behaviourisishighly
highlyinfluenced
influencedby bythe
the
rigidity of the pile, characterized by (E × I ) value. The pile is simulated
rigidity of the pile, characterized by (Epp × Ip)pvalue. The pile is simulated by volume ele- by volume
element
ment with withananelastic
elasticbehaviour
behaviourand andan anequivalent
equivalent Young
Young modulus
modulus equal
equal toto3131GPa.
GPa.The
The
Poisson’s ratio is taken equal to
Poisson’s ratio is taken equal to 0.3. 0.3.
Concerning
Concerning the the modelling
modelling of ofthe
thesoil
soilbehaviour,
behaviour,to toovercome
overcomethe thelimitations
limitationsofofthethe
classical elastic–perfectly plastic model with Mohr–Coulomb
classical elastic–perfectly plastic model with Mohr–Coulomb failure criteria, failure criteria, a friction hard-
a friction
ening elastoplastic
hardening model,model,
elastoplastic with shear-induced
with shear-induced volumetric changeschanges
volumetric and stress-dependent
and stress-de-
stiffness,
pendent stiffness, was implemented is this study. The Flac3D built-inmodel
was implemented is this study. The Flac3D built-in “CHsoil” “CHsoil”wasmodel
used [25].
was
This model permits to account for a higher soil stiffness during unloading–reloading
used [25]. This model permits to account for a higher soil stiffness during unloading–re- pro-
cess and process
loading for a soiland
stiffness
for a dependent
soil stiffnessondependent
the initial onstress
thestate.
initialIndeed, the initial
stress state. stress
Indeed, the
state
initial stress state significantly increases from the head of the pile, at the mudline, 10
significantly increases from the head of the pile, at the mudline, to the toe at m
to the
depth, actually leading to variation in soil stiffness along the pile. To calibrate the soil
toe at 10 m depth, actually leading to variation in soil stiffness along the pile. To calibrate
model elastic stiffness parameters, an average value of a deformation modulus equal to
the soil model elastic stiffness parameters, an average value of a deformation modulus
50 MPa and Poisson’s ratio equal to 0.2 (i.e., a bulk modulus Kref = 27.8 MPa and shear
equal to 50 MPa and Poisson’s ratio equal to 0.2 (i.e., a bulk modulus Kref = 27.8 MPa and
modulus Gref = 20.8 MPa) are targeted at a mid-depth of 5 m, where the initial mean
shear modulus Gref = 20.8 MPa) are targeted at a mid-depth of 5 m, where the initial mean
effective pressure is equal to p’m = 30 kPa, also set as the reference pressure pref . In CHsoil
effective pressure is equal to p’m = 30 kPa, also set as the reference pressure pref. In CHsoil
model, the elastic bulk and shear modulus are then computed using Equations (7) and (8)
model, the elastic bulk and shear modulus are then computed using Equations (7) and (8)
according to the initial mean effective pressure p’m , which increases with depth due to soil
according to the initial mean effective pressure p’m, which increases with depth due to soil
J. J.Mar.
Mar.Sci.
Sci.Eng. 2021,9,9,235
of2223
weight. With
weight. With m
m == n = 0.5,
0.5, at
at aa depth
depth10 10m,
m,the
theelastic
elasticdeformation
deformationmodulus
modulusisisequal
equaltoto
70
MPa:
70 MPa: !m
p0 m
e
K = Kre f . pre f 𝑝′ (7)
𝐾 𝐾 . 𝑝 pre f (7)
𝑝
!n
p0 m𝑝′
G e 𝐺= Gre𝐺
f . pre. 𝑝
f (8)
(8)
pre𝑝f
The
TheCHsoil
CHsoilmodel
modelparameters
parametersused usedininthis
thisstudy
studyarearegiven
givenin inTable
Table22andandthethecorre-
corre-
sponding behaviour obtained on triaxial compression tests at 30 and 50 kPa
sponding behaviour obtained on triaxial compression tests at 30 and 50 kPa of confine- of confinement
are
mentdepicted on Figure
are depicted 7, with7,anwith
on Figure unloading–reloading loop. loop.
an unloading–reloading
However,
However, this model is not able to account for specificfeatures
this model is not able to account for specific featuresofofsand
sandbehaviour
behaviour
under
under a high number of load cycles (strain accumulation due to fatigue, etc.), asthere
a high number of load cycles (strain accumulation due to fatigue, etc.), as thereisis
only
only an isotropic shear hardening mechanism. Indeed, as this numerical study firstaims
an isotropic shear hardening mechanism. Indeed, as this numerical study first aims
atatdeveloping
developingaacomplex
complexloading
loadingprocedure
procedure(multi-directional
(multi-directionalcyclic
cycliclateral
lateralloading),
loading),thethe
efforts
effortswere
werenot
notconcentrated
concentrated onon
implementing
implementing a constitutive
a constitutive model
modelwith a high
with degree
a high of
degree
complexity, and this aspect is part of the numerical work perspectives.
of complexity, and this aspect is part of the numerical work perspectives.
Figure 7. Triaxial numerical test results with CHsoil.

Figure 7. Triaxial numerical test results with CHsoil.
Table 2. Parameters of the CHsoil model.
Parameter Significance value

pref Reference pressure 30 kPa
Eref Deformation modulus number 1 680 1
J. Mar. Sci. Eng. 2021, 9, 235 10 of 22
Table 2. Parameters of the CHsoil model.
Parameter Significance Value

pref Reference pressure 30 kPa
Eref Deformation modulus number 1680 1
υ Poisson’s ratio 0.2
m Bulk modulus exponent 0.5
n Shear modulus exponent 0.5
ϕf Maximum friction angle 37◦
Rf Failure ratio 0.9
ϕcv Constant volume friction angle 30◦
c Cohesion 0 kPa
Ψf Maximum dilation angle 7◦
1 The deformation modulus for the reference pressure is then equal to Eref · pref = 50.4 MPa.
Interfaces are placed between the soil and the pile by using the standard interface
feature from the software. The interfaces are characterized by normal and shear stiffness
and sliding properties. The interface normal stiffness kn and shear stiffness ks are computed
equal to 1.2 GPa/m, according to the adjacent zone size and stiffness, in order to optimize
the computational time [22], so they have no real physical meaning. The interfaces have
no cohesion and a friction angle equal to 2/3 of the soil ultimate friction angle, i.e., equal
to 2/3 × 37◦ = 24.7◦ , which is a conventional friction angle for a relatively rough soil–
steel interface.
2.4. General Modelling Procedure

A lateral pile loading simulation follows the subsequent steps:
(1) Mesh generation for the ground mass, the pile and the interfaces. The pile is first
generated separately above the soil mass, and is then moved down into its final position,
in order to correctly create the contact with the interfaces [22]. The pile is thus numerically
wished-in-place and the possible pile installation effects are therefore not represented.
Moreover, this procedure would correspond to the case of a close-ended rather than an
open-ended pile.
(2) Stepping until reaching static equilibrium under gravity and soil self-weight. The
zones occupied by the future pile are first considered as soil. The calculation is performed
under drained conditions, so the submerged unit weight γ’ = 10.2 kN/m3 is assigned to
the zones and an earth coefficient K0 = 0.5 is used to initialize the horizontal effective stress.
Static equilibrium is considered reached when the unbalanced force ratio is below 10−6 .
This ratio is defined as the largest ratio between the maximum grid-point unbalanced force
to the average applied force amongst all of the model grid-points.
(3) The pile zones are assigned pile properties and a new static equilibrium is reached.
(4) The displacements are initialized (at this stage, they are very small and have no
physical meaning).
(5) Lateral loading is applied using the procedure described below.
(6) All the calculations are performed in small strain conditions.
2.5. Loading Procedure

Due to the formulation of the program Flac3D, a loading can be defined by applying
a velocity to model grid-points. This procedure is especially appropriate as the loading
(and loading direction) varies during the process. The applied velocity then corresponds
to an increment in displacement (in the same direction) at each time-step (of duration
∆t) of the stepping process. In the case of a quasi-static study, the velocity has only a
numerical signification, and should be sufficiently small in order to keep the model in a
quasi-static equilibrium (i.e., with a negligible unbalanced force ratio). The choice of the
applied velocity is discussed later. Moreover, numerical damping is introduced in the
computational process to reach a quasi-static equilibrium state (combined damping is used
in this study).
J. Mar. Sci. Eng. 2021, 9, 235 11 of 22
To apply a horizontal force at the head of the pile, a horizontal velocity is therefore
applied to all pile head grid-points (on the 0.1 m thick pile slice above the mudline).
Nevertheless, the process is governed in terms of force value, by implementing adequate
servo files that compute the current resulting force at pile head during the process and then
adapts the required velocity accordingly (in particular, its direction). The current force at
the pile head is computed as the sum of the unbalanced forces of all the grid-points of the
pile head where the velocity is applied (in the x direction to obtain the force in x direction
Hx and in the y direction to compute Hy , where applicable, i.e., for the multi-directional
loading case).
Two types of loading procedures are implemented: for mono-directional lateral load-
ing (only in the x direction) and multi-directional lateral loading (in x and y directions).
2.5.1. Loading Procedure for Mono-Directional Monotonic Loading and Cyclic Loading
A velocity is applied on the pile head only in x- direction, and due to geometrical
and material symmetry of the model, it induces a pile head horizontal force in the x
direction. The impact of the numerical value of the velocity on the model response was
the subject of a parametric study [24] and a value equal to 5 · 10−8 (m/step) permitted to
maintain an acceptable quasi-static equilibrium during the loading process (i.e., negligible
maximum unbalanced force), with a reasonable computational time. Horizontal loading is
accomplished with a positive velocity in the x direction (pile head displacement in positive
x direction, see axes in Figures 5 and 6), until the desired Hmax value; then, unloading is
performed by applying a velocity in the opposite direction (-x) until the desired Hmin value.
The velocity was set to 1 · 10−8 (m/step). Reloading and unloading loops are applied in the
same manner. Twenty cycles were applied for each calculation (under mono-directional
loading, system behaviour was considered unaffected after around 20 cycles, see Section 3).
Under lateral monotonic loading, with a PC (CPU 1.9 GHz–RAM 16 Go), the computa-
tion duration was approximately 2 h to reach H = 1.77 MN and 9 more hours were needed
to reach a pile head displacement equal to 0.15 D (H = 3.46 MN), on the half model. The
computation time is directly proportional to the applied velocity value. Concerning the
cyclic loading, achieving 20 loops of unloading–reloading took about 20 h for case 1 and
48 h for case 7 (as the computation time is directly correlated to the pile head displacement,
for a given applied velocity).
2.5.2. Loading Procedure for Multi-Directional Cyclic Loading

Concerning a multi-directional horizontal loading process as described in Section 2.1,
a specific procedure has been developed in the frame of this work, in order to apply a given
loading path, but controlled in terms of displacements. Indeed, the complexity lies in the
fact that force vectors and displacement vectors increments are not co-linear.
The first loading is applied using the mono-directional procedure (but on the full
model), until Hmax value is reached at pile head.
Concerning the unloading, the resulting horizontal value H is considered varying
linearly according to the azimuth ω (Figure 3) from Hmax down to Hmin , for ω = 0 to
ωmax = 30◦ . The horizontal force H has two components: Hx = H cosω and Hy = H sinω,
as shown on Figures 8 and 9a) depicts the imposed velocity for unloading (xvel negative
to decrease Hx and yvel positive to increase Hy ). The absolute value of the norm of the
velocity vector is kept constant (vel = 5 × 10−8 m/steps as a cruise value, after a ramp
of increase) but the orientation (angle β) of the velocity vector is adjusted during the
calculation, in order to follow the imposed path in terms of force H = f(ω). The initial value
for β is arbitrarily picked at 45◦ and the x and y component of the velocity (xvel and yvel)
are re-calculated based on current β value at each step (xvel = vel × cos β, yvel = vel ×
sin β). On Figure 10, the angle δ is defined as the azimuth of the resulting horizontal force
at the current step (step n), whereas the targeted azimuth is ω (computed according to
the current H value and given function H = f(ω)). The difference is ∆ω = ω − δ. For the
next step (n + 1), the orientation of velocity is then adjusted to correct the orientation of the
and given function H = f(ω)). The difference is Δω = ω – δ. For the next step (n+1), the
J. Mar. Sci. Eng. 2021, 9, 235
orientation of velocity is then adjusted to correct the orientation of the force. The12new of 22
value of β isfunction
and given computed H =withf(ω)). βn+1The = βdifference
n + Δβ, with is Δβ
Δω==Δω. ω –During the next
δ. For the whole process,
step (n+1),the the
and given function H = f(ω)). The difference is Δω = ω – δ. For the next step (n+1), the
adjusting
orientation angle Δβ = Δω
of velocity is actually
then adjusted oscillates aroundthe
to correct zero. The unloading
orientation process
of the force. Thestops
new
orientation of velocity is then adjusted to correct the orientation of the force. The new
when
valuethe of βhorizontal
is computed force H reaches
with βn+1 = βH n + . with Δβ = Δω. During the whole process, the
Δβ,
min
value
force. of
Theβ is computed
new value ofwith is n+1 = βn + Δβ, with Δβ = Δω. During the whole process, the
βcomputed with + ∆β, ∆β ∆ω.
The same concept is β
adjusting angle Δβ = Δω actually oscillatesreloading,
implemented for β n+1 = βzero.
around nby with
inverting
The the=sign
unloading ofDuring the
xvel stops
process and
adjusting
whole anglethe
process, Δβadjusting
= Δω actuallyangle ∆βoscillates
= ∆ω aroundoscillates
actually zero. The unloading
around zero. process
The stops
unloading
yvel (Figure 9b) and with Δβ
when the horizontal force H reaches Hmin. = −Δω. The norm of velocity vector is also equal to 5 × 10 −8
when
process thestops
horizontal
when force
the H reaches
horizontal HminH. reaches H .
force
m/stepsTheassamea cruise value,isafter
concept a ramp of for
implemented increase. The
reloading, minreloading
by invertingprocess stopsofwhen
the sign xvel the
and
The same conceptisisimplemented
implementedfor forreloading,
reloading, byby inverting the sign of xvel and−8
yvel The
horizontal same
(Figure force
9b)concept
Handreaches
with Δβ Hmax = ;−Δω.
then The a new inverting
unloading–reloading
norm of velocity vectorthe loop
is sign of xvel
isequal
also toand
achieved. yvel
5 ×This
10
yvel (Figure
(Figure 9b) 9b) and with
∆βfor =Δβ = −Δω.
−∆ω. The norm of velocity vectoris isalso
alsoequal
equaltoto5 5××10 10−−88
cyclic
m/steps as aand
process with
is applied
cruise value, 40
after loops
a rampThe norm
(compared ofto
of increase. velocity
the
The vector
mono-directional
reloading process loading, 20 addi-
stops when the
m/steps
m/steps asasa acruise
cruise value,
value, after
after a ramp
a ramp of of
increase.
increase. TheThe
reloading
reloadingprocess stops
process when
stops the
when
tional cycles had to be applied to confirm the trend observed
horizontal force H reaches Hmax; then a new unloading–reloading loop is achieved. This after approximately the ten
horizontal
the horizontalforceforce
H reaches
H Section).
reaches HmaxH ; then athen
newa unloading–reloading
max ;(compared newtounloading–reloading loop is achieved. This
first cycles
cyclic (seeisResults
process applied for 40 loops the mono-directional loop is achieved.
loading, 20 addi-
cyclic
This process
cyclic is
processapplied for 40 loops (compared to the mono-directional loading, 20 addi-
The
tional cycles had toisbeapplied
computation ofapplied
the first forto2040 loops (compared
cycles
confirm took to
6.5the
aboutobserved
the trend daysmono-directional
with approximately
after the same PC. loading, 20
the ten
tional cyclescycles
additional had to hadbeto
applied
be applied to confirm
to confirm the trend observed
the trend afterafter
observed approximately
approximately the ten
the
first cycles (see Results Section).
first cycles
ten first (see (see
cycles Results Section).
Results Section).
The computation of the first 20 cycles took about 6.5 days with the same PC.
The
The computation
computation of of thethe first
first 20 20 cycles took about 6.5 days with the same PC.
Figure 8. Force applied on the pile head.
Figure 8. Force applied on the pile head.

Figure
Figure 8.
8. Force
Force applied
applied on
on the
the pile
pile head.
head.
Figure 9. Nodal velocity applied on the pile head for multi-directional (a) unloading; and (b) re-
loading.
Figure 9. Nodal velocity applied on the pile head for multi-directional (a) unloading; and (b) reloading.
loading.
loading.
Figure 10. Numerical procedure for multi-directional unloading.

3. Results
3.1. Monotonic
Figure Lateral
10. Numerical Loading for multi-directional unloading.
procedure
The preliminary calculation consists in the mono-directional and monotonic lateral
loading of the pile, to assess the conventional lateral limit load (Hlim ). Indeed, the effect
of the cyclic loading is supposed to be more important when the maximum applied force
3. Results
3.1. Monotonic Lateral Loading
J. Mar. Sci. Eng. 2021, 9, 235 13 of 22
The preliminary calculation consists in the mono-directional and monotonic lateral
loading of the pile, to assess the conventional lateral limit load (Hlim). Indeed, the effect of
the cyclic loading is supposed to be more important when the maximum applied force
HHmax
max becomes
becomes closer
closer to
to the
the limit
limit load
load H lim.. The
Hlim The ratio
ratio H
Hmax
max /H
/H should
should
limlim thenthen necessarily
necessarily in-
intervene in the cyclic degradation of the behaviour. Several procedures
tervene in the cyclic degradation of the behaviour. Several procedures exist to determine exist to determine
HHlim value: the
lim value: theconvention
conventionchosen chosenhere
hereis is
thethe horizontal
horizontal forceforce
thatthat corresponds
corresponds to a to a
hori-
horizontal head displacement equal to 25 % of the pile diameter D
zontal head displacement equal to 25 % of the pile diameter D (with D = 1.7 m). Figure 11 (with D = 1.7 m).
Figure
shows11 shows
Hlim = 4.3HMN
lim = 4.3extrapolating
(by MN (by extrapolating the numerical
the numerical results that results
werethat were stopped
stopped for a dis-
for a displacement
placement equal toequal
0.15 Dto =0.15 D =m.0.255 m.
0.255
Figure 11. Force–displacement curve for the monotonic lateral loading of the pile.
Figure 11. Force–displacement curve for the monotonic lateral loading of the pile.
3.2.
3.2.Mono-Directional
Mono-DirectionalLateral
LateralCyclic
CyclicLoading
Loading
Figure
Figure1212presents
presentsthe thehorizontal
horizontalforceforcedeveloped
developedatatthe
thepile
pilehead
head(H(H= =HH x )x)according
according
totothe
thepile
pilehead
headdisplacement,
displacement,for forthe
thereference
referencecasecase(case
(case1)1)ofofmono-directional
mono-directionalcyclic cyclic
loading.
loading. Twenty
Twentycycles
cyclesof ofunload
unloadand andreload
reloadwere
wereapplied.
applied.This
Thisfigure
figureshows
showsthat thatthe
the
irreversible
irreversiblehorizontal
horizontaldisplacement
displacementincreases
increasesafter
aftereach
eachcycle,
cycle,with
withaadecreasing
decreasingrate rateall
all
along
alongthe
thecycles.
cycles.The
Thepile
piledeflection
deflectionisisdepicted
depictedininFigure
Figure13a.
13a.The
Thepile
pileisismainly
mainlysubjected
subjected
totorotation
rotationaround
aroundaafixedfixedpoint
pointlocated
locatedatataadepth
depthofof88mm(with
(withan anangle
angleofofrotation
rotationofof
around 0.3 ◦ when the horizontal force is equal to 1.77 MN), but also to bending, leading
around 0.3° when the horizontal force is equal to 1.77 MN), but also to bending, leading
totoaabending
bendingmoment
momentininthe thepile
pile(Figure
(Figure13b).
13b).The
Thebending
bendingmoment
momentisismaximum
maximum(Mf (Mfmax
max))
atataa depth
depth around 4–4.5 m
around 4–4.5 mandandonlyonlyincreases
increasesbyby around
around 1%1%fromfromthethe
firstfirst loading
loading at Hatmax
H=max
1.77=MN
1.77(Mf
MN (Mf 4.40 MN.m)
max_0 = max_0
= 4.40 toMN ·m)
the 20to the 20th
th cycle cycle= 4.44
(Mfmax_20 (Mfmax_20
MN.m). = 4.44 MN·deflection
The pile m). The
pile deflection
curve curve
also shows thatalso shows
pile that pile displacement
displacement accumulationaccumulation
mainly occursmainly occurs
at the pile at the
head, and
J. Mar. Sci. Eng. 2021, 9, x FOR PEERpile head,
REVIEW and the pile toe has a negligible backward displacement accumulation along
the pile toe has a negligible backward displacement accumulation along the cycles. of 23 14
the cycles.
Figure12.
Figure 12. Force–displacement
Force–displacement curve
curve for
formono-directional
mono-directionallateral
lateralcyclic loading
cyclic (reference
loading case,
(reference case,
case 1).
case 1).
J. Mar. Sci. Eng. 2021, 9, 235 14 of 22
Figure
Figure 12.
12. Force–displacement
Force–displacement curve
curve for
for mono-directional
mono-directional lateral
lateral cyclic
cyclic loading
loading (reference
(reference case,
case,
case 1).
case 1).
Figure 13.
Figure13.
Figure Pile
13.Pile horizontal
Pilehorizontal displacement
horizontaldisplacement and
displacementand bending
andbending moment
bendingmoment (CH1
moment(CH1 case).
(CH1case).
case).
The
The increase
increaseinin
Theincrease horizontal
inhorizontal
horizontalpile pile head
pilehead deflection
headdeflection
deflection (Δy
(∆y
(Δy =y ==Ny
yNN−––yy
y111)))normalized
normalized
normalizedby by the
bythe first
thefirst
first
loading
loading displacement
displacement
loading displacement at H atatH H (y
max (y drawn
1) is drawn according
according to
to the
the cycle
cycle number
number
max (y11) is drawn according to the cycle number N in the natural
max NN in
in the
the natural
natural
logarithm
logarithm
logarithm(lnN) (lnN)
(lnN)inin Figure
inFigure
Figure14.14. From
14.From
Fromthe the law
thelaw proposed
proposedbyby
lawproposed byLinLinand
Lin and
and Liao Liao
Liao (1999)
(1999) (equ
(equ 1):
(Equation
(1999) 1): y N/y1
y(1)):
N/y1
y––N11/y==1α −lnN;
1 = the
α parameter
lnN; the α can
parameter be
α obtained.
can be The
obtained. numerical
The results
numerical
α lnN; the parameter α can be obtained. The numerical results approximately verify approximately
results approximately verify
the
the logarithmic
verify law
the logarithmic
logarithmic law and lawan
and and
an α
α anvalue
value equal
α value to
to 0.008
equalequal was
was obtained
to 0.008
0.008 was obtained
obtained for
for the
forreference
the the reference
reference case.
case.
Figure
case.
Figure 14
14 also
Figure also14shows the
the head
also shows
shows the displacement
head head displacement
displacement results obtained
resultsresults
obtained for
for case
obtained casefor 77 (case
case with
(case 7 (case
with the larg-
thewith
larg-
est
the average
largest force
average and cyclic
force and amplitude),
cyclic leading
amplitude), to α
leading= 0.017.
to α
est average force and cyclic amplitude), leading to α = 0.017. The values of the parameter The
= 0.017.values of
The the
valuesparameter
of the
α obtained
parameter α for all
obtained the investigated
for all the cases are
investigated summarized
cases are in Figure
summarized
α obtained for all the investigated cases are summarized in Figure 15. This figure high- 15.
in This
Figure figure
15. high-
This
figure
lightshighlights
lights that
that the that the
the higher
higher thehigher
the cyclicthe
cyclic cyclic amplitude,
amplitude,
amplitude, the
the largerthethe
larger larger
the α the α value;
α value;
value; however,
however,however,
for for a
for aa given
given
given
cyclic cyclic amplitude,
amplitude, the the
average average
force force
has
cyclic amplitude, the average force has no impact on α. has
no no
impactimpact
on α.on α.
Figure 14.
Figure14. Accumulation
Accumulationofof
14.Accumulation pile
ofpile head
pilehead displacement
headdisplacement
displacement during
during the
the cycles
cycles and
and linear
linear approximation
approximation
Figure
with the law yN/y1 – 1 = α lnN.
during the cycles and linear approximation with
with the law yN/y1 – 1 = α lnN.
the law yN /y1 − 1 = α lnN.
For all the cases investigated, the pile is subjected to a rotation around a fixed point,
all along the cycles located at a depth of 8 m, as depicted in Figure 16 for cases 1 and 7.
Figure 17 depicts the relative increase in the maximum bending moment obtained in the
pile for Hmax during the 20 cycles. The increase is limited for all the cases (0.2–2.4%), but it
is higher for a larger cyclic amplitude, whereas the average force has no impact.
J. Mar. Sci. Eng. 2021, 9, 235 15 of 22
Figure 15. Parameter α from the law yN/y1 – 1 = α lnN.
isFigure
higher15. Parameter
for a largerααcyclic
fromthe
thelaw
lawyyN/y
amplitude, /y1whereas
– 1 = α lnN.
the average force has no impact.
Figure 15. Parameter from N 1 − 1 = α lnN.
is higher for a larger cyclic amplitude, whereas the average force has no impact.
Figure
Figure 16.
16. Pile
Pile horizontal
horizontal displacement
displacement after
after 20
20 cycles
cycles for
for case
case 11 (reference)
(reference) and
and case
case 77 (displacement
(displacement of
of the
the nodes
nodes located
located
along the pile axis).
along the pile axis).
3.3. Multi-Directional Lateral Cyclic Loading

The curve shown on Figure 18 (horizontal force in y direction Hy vs. horizontal force
in x direction Hx )20iscycles
Figure 16. Pile horizontal displacement after a numerical
for case result takenand
1 (reference) from the7 simulation
case under
(displacement of themulti-directional
nodes located
along the pile axis). cycling loading. It demonstrates that the implemented procedure, described in Section 2.5.2,
successfully permits applying the targeted loading path. This figure also reminds that the
same unloading and reloading path is trailed along the cycles.
J. Mar. Sci. Eng. 2021, 9, 235 16 of 22
Figure 17. Increase in the maximum bending moment (Mfmax) in the pile from the first loading
(Mfmax_1 for Hmax) to the 20th loading (increase is defined as ΔMfmax_N/Mfmax_1).

in x direction Hx) is a numerical result taken from the simulation under multi-directional
cycling loading. It demonstrates that the implemented procedure, described in Section
Figure17.
17. Increase
Increase in
in the
the maximum
maximum bending
bending moment
moment(M (Mfmax) in the pile from the first loading
Figure
2.5.2, successfully permits applying the targeted fmax ) in
loading the pile
path. Thisfrom
figurethe also
first loading
reminds
(Mfmax_1 for Hmax) to the 20th loading (increase is defined as ΔMfmax_N/Mfmax_1).
(Mf for H ) to the 20th loading (increase is defined as ∆Mf
that the same unloading and reloading path is trailed along the cycles.
max_1 max max_N /Mf max_1 ).

in x direction Hx) is a numerical result taken from the simulation under multi-directional
cycling loading. It demonstrates that the implemented procedure, described in Section
2.5.2, successfully permits applying the targeted loading path. This figure also reminds
that the same unloading and reloading path is trailed along the cycles.
Figure18.
Figure 18.Loading
Loadingpath
pathfor
forthe
themulti-directional
multi-directionalcyclic
cyclicloading
loadingsimulation.
simulation.
Figure1919depicts
Figure depictsthetheevolution
evolutionofofthe thepile
pilehead
headdisplacement
displacementininthe thex xdirection
directionduring
during
thelateral
the lateralmultidirectional
multidirectional cyclic
cyclic loading
loading between
between H =H1.25= 1.25
andand1.771.77
MN,MN, whereas
whereas FigureFigure
20
is a plan view of the pile head displacement. The latter highlights that the pile head isis
20 is a plan view of the pile head displacement. The latter highlights that the pile head
subjectedtotoa ahorizontal
subjected horizontaldisplacement
accumulationininboth boththethex xand
andy ydirections
directionsduring
during
theloading
the loadingcycles.
cycles.Thereby,
Thereby,residual
residualpilepiledisplacements
displacementsoccuroccurininthethedeviated
deviateddirection
direction(y(y
Figure 18. Loading path for the multi-directional cyclic loading simulation.
direction),whereas
direction), whereasthe thehorizontal
horizontalforce
forceonlyonlygoes
goesback
backtotoananx xcomponent
componentatateach eachcycle
cycle
(azimuthωω==00°).
(azimuth ◦ ). The
Thedeviation
deviationangle
angleisisdefined Figure2020asasΩ.Ω.The
definedininFigure Thefinal
finaldeviation
deviation
Figure 19 depicts the evolution of the pile head displacement in the x direction during
angleΩ,
angle Ω,for
forH H=H Hxxafter
after2020cycles
cycles(final
(final position
position of of
thethe
pilepile compared
compared to xthe
to the x axis)
axis) is
is equal
the lateral multidirectional
◦ cyclic loading between H = 1.25 and 1.77 MN, whereas Figure
◦ after
equal
to 4.3°to(Ω
4.3
L20).(Ω
The
L20 ). The
value ofvalue
this of this
angle isangle
Ω L = is
2.3°Ω =
after
L 2.3
the first the
cycle first cycle
(Figure 21),(Figure
it 21),
increases
20 is a plan view of the pile head displacement. The latter highlights that the pile head is
it increases during the first ten cycles to reach a nearly constant value at the end of each
subjected to a horizontal displacement accumulation in both the x and y directions during
cycle. Figure 20 also shows this angle at the end of each unloading stages (ΩU ), i.e., for
the loading cycles. Thereby, residual pile displacements occur in the deviated direction (y
a loading azimuth ω equal to 30◦ . This angle is equal to 7.9◦ for the first unloading and
direction),
also rapidly whereas
(after 5–6the horizontal
cycles) tends to force only goes
a constant angleback to an x 9component
of around ◦ . The angle atdifference
each cycle
between unloading and reloading ∆Ω = ΩU –ΩL is constant after about 10 cycles. deviation
(azimuth ω = 0°). The deviation angle is defined in Figure 20 as Ω. The final
angle Ω, for H = Hx after 20 cycles (final position of the pile compared to the x axis) is equal
to 4.3° (ΩL20). The value of this angle is ΩL = 2.3° after the first cycle (Figure 21), it increases
during the first ten cycles to reach a nearly constant value at the end of each cycle. Figure
during the first ten cycles to reach a nearly constant value at the end of each cycle. Figure
20 also shows this angle at the end of each unloading stages (ΩU), i.e., for a loading azi-
20 also shows this angle at the end of each unloading stages (ΩU), i.e., for a loading azi-
muth ω equal to 30°. This angle is equal to 7.9° for the first unloading and also rapidly
J. Mar. Sci. Eng. 2021, 9, 235 muth ω equal to 30°. This angle is equal to 7.9° for the first unloading and also rapidly
17 of 22
(after 5–6 cycles) tends to a constant angle of around 9°. The angle difference between
(after 5–6 cycles) tends to a constant angle of around 9°. The angle difference between
unloading and reloading ΔΩ = ΩU – ΩL is constant after about 10 cycles.
unloading and reloading ΔΩ = ΩU – ΩL is constant after about 10 cycles.
Figure 19. Horizontal applied force (multidirectional) vs. the pile head horizontal displacement in
Figure19.
Figure 19.Horizontal
Horizontalapplied
appliedforce
force(multidirectional)
(multidirectional)vs.
vs.the
thepile
pilehead
headhorizontal
horizontal displacement
displacement in
in
the x direction.
the x direction.
the x direction.
Figure 20. Pile

Figure20. Pile head
head horizontal
displacementduring
duringthe
themulti-directional
multi-directionalcyclic
cyclicloading.
loading.
Figure 20. Pile head horizontal displacement during the multi-directional cyclic loading.
Figure 22 presents the horizontal pile head displacement accumulation (∆y/y1 =
yN /y1 − 1) during the loading cycles (with cycle number N in natural logarithmic scale), for
the multi-directional loading and also for the mono-directional loading of same amplitude
(lateral force H between 1.25 and 1.77 MN). In the case of the multi-directional loading,
the pile head displacement (termed yN ) has both x and y horizontal components, whereas
only a horizontal displacement in x direction occurs for the mono-directional loading.
However, the initial displacement (termed y1 ) under the first monotonic loading equal to
1.77 MN is the same in both calculations (y1 = 52 mm). This figure reveals that the pile
head displacement accumulation is much larger when cycles are applied with a change
J. Mar. Sci. Eng. 2021, 9, 235 18 of 22
J. Mar. Sci. Eng. 2021, 9, x FOR PEERin loading

REVIEW direction. After 20 cycles, a displacement increase of 11% was reached for the23
18 of
multi-directional loading, against an increase of only 2% for the mono-directional loading.
Figure21.
Figure 21.Deviation
Deviationangle
angleofofthe
thepile
pilehead
headhorizontal
horizontaldisplacement.
displacement.
Figure
This 22 presents
figure the horizontal
also highlights that thepile head displacement
accumulation accumulation
of pile head (Δy/y
displacement 1 = ynot
does N/y1
– 1) during
follow the loading
the logarithmic cycles
law (with (2))
(Equation cycle numberwhen
anymore N in anatural logarithmiccyclic
multi-directional scale), for the
loading
ismulti-directional
applied, contrarilyloading
to the and also for made
observation the mono-directional loading loading.
for the mono-directional of same However,
amplitude
(lateral force H between 1.25 and 1.77 MN). In the case of the multi-directional
a linear trend was observed again for a larger amount of cycles (a total of 40 cycles loading,
were
the piletohead
J. Mar. Sci. Eng. 2021, 9, x FOR PEERapplied
REVIEW displacement
confirm the trend, (termed yN)the
as well for hasmono-directional
both x and y horizontal
loading),components,
with a slopewhereas
19 of 23
similar
toonly
the aone
horizontal
obtained displacement in x direction
under mono-directional occurs
loading (α =for the mono-directional loading.
0.008).
However, the initial displacement (termed y1) under the first monotonic loading equal to
1.77 MN is the same in both calculations (y1 = 52 mm). This figure reveals that the pile
head displacement accumulation is much larger when cycles are applied with a change in
loading direction. After 20 cycles, a displacement increase of 11% was reached for the
multi-directional loading, against an increase of only 2% for the mono-directional loading.
This figure also highlights that the accumulation of pile head displacement does not
follow the logarithmic law (Equation (2)) anymore when a multi-directional cyclic loading
is applied, contrarily to the observation made for the mono-directional loading. However,
a linear trend was observed again for a larger amount of cycles (a total of 40 cycles were
applied to confirm the trend, as well for the mono-directional loading), with a slope sim-
ilar to the one obtained under mono-directional loading (α = 0.008).
Figure22.
Figure 22.Pile
Pilehorizontal
accumulationduring
duringcyclic
cyclicloading.
loading.
4.4.Discussion
Discussion
The
Thenumerical
numericalresults
resultsobtained
obtainedononthe
themono-directional
mono-directionallateral
lateralcyclic
cyclicloading
loadingare
areinin
qualitative good agreement with the predictive logarithmic law for the additional
qualitative good agreement with the predictive logarithmic law for the additional pilepile head
head displacement (Equations (2) and (3)). The parameter α from the Lin and Lao (1999)
[12] law: yN/y1 – 1 = α lnN could be obtained and varied between 0.004 and 0.018 mainly
according to the cyclic amplitude Hc for the investigated cases, but, in all likelihood, not
according to Hm.
J. Mar. Sci. Eng. 2021, 9, 235 19 of 22
displacement (Equations (2) and (3)). The parameter α from the Lin and Lao (1999) [12]
law: yN /y1 − 1 = α lnN could be obtained and varied between 0.004 and 0.018 mainly
according to the cyclic amplitude Hc for the investigated cases, but, in all likelihood, not
according to Hm .
The logarithmic law proposed in the SolCyP recommendation [11] would lead to the
expression of the parameter α (from Equations (2) and (3)):
α = 0.102/ CR × (Hc /Hmax )0.35 (9)
Thereby, Figure 23 depicts the evolution of α obtained from the numerical results,
according to (Hc /Hmax )0.35 , that should lead to a linear relation with a slope 0.102/CR if
Equation (9) is satisfied, with the assumption that the rigidity coefficient CR (defined in
Equation (4)) is independent on the loading. The trend of α increasing with (Hc /Hmax )0.35
is observed, but the linear relation does not accurately apply. A value of 0.102/CR = 0.02
(that satisfies the numerical cases 3, 4 and 5), would lead to CR = 5.1. However, the rule
given by [3] (Equation (1)) with a soil stiffness Es = 50 MPa, leads to a flexible pile for Ep Ip
below 1290 MN·m2 (=(Ep Ip )fl ), then to CR = 1.6. According to Equation (9), the parameter
J. Mar. Sci. Eng. 2021, 9, x FOR PEERαREVIEW
would vary between 0.026 for case 2 and 0.043 for case 6. These values are thus 2–620 of 23
times
higher than the values obtained with the current numerical modelling.
Figure23.
23.Parameter
Parameterααobtained
obtainedon
onthe
thenumerical
numericalmodel
modelaccording
accordingtoto(H
(H/H
c/Hmax)0.35.
0.35 .
Figure c max )
Thetrends
The trendsgiven
givenbybythe
theproposed
proposednumerical
numericalmodel modelare arethus
thusqualitatively
qualitativelyacceptable,
acceptable,
butititseems
but seemsthatthatititunderestimates
underestimatesthe thepile
pilehead
headdisplacement
accumulationunder underlateral
lateral
mono-directionalcyclic
mono-directional cyclicloading.
loading. This
Thisisisprobably
probablydue duetotothetheimplemented
implementedconstitutive
constitutive
modelfor
model forthe
thesoil
soilthat
thatisisnot
notable
abletototake
takeaccurately
accuratelyinto intoaccount
accountthe thestrain
strainaccumulation
accumulation
ininthe
thesoil
soil under
under repeated loading,
loading, andandthis
thisisisa amajor
major numerical
numerical work
work perspective
perspective to this
to
current
this study.
current In fact,
study. the elastoplastic
In fact, the elastoplasticmodel implemented
model implemented in thisinstudy has only
this study hasan iso-
only
antropic hardening
isotropic shearshear
hardening mechanism, with an
mechanism, elastic
with response
an elastic on an unload–reload
response on an unload–reload loop on
loop on a test
a triaxial triaxial
pathtest path7),
(Figure (Figure
thereby 7),initially
therebynot initially
able tonot
takeable to take effect
ratcheting ratcheting effect
into account
into
[26].account
However, [26].in However, in this soil–structure
this soil–structure interaction
interaction system, system,
the stress paththe
is stress path is
more complex
more
than complex
on a purethan on a pure axisymmetric
axisymmetric triaxial
triaxial test path, test path,
leading leading
to stress to stress redistribu-
redistribution in the sur-
tion
rounding soil, to additional plastic zones at following cycles and thus cycles
in the surrounding soil, to additional plastic zones at following and thus
to additional to
defor-
additional
mations. deformations.
AsAsthis
thisnumerical
numericalmodel modelaccurately
accuratelyrepresents
representsthe thequalitative
qualitativepile–soil
pile–soilinteraction
interaction
behaviour
behaviourunder under mono-directional
mono-directional lateral cyclic cyclicload,
load,ititisisimplemented
implementedtoto explore
explore thethe
be-
behaviour under multi-directional lateral cyclic loading. This latter
haviour under multi-directional lateral cyclic loading. This latter requires the implemen- requires the imple-
mentation
tation of aofspecific
a specific loadingprocedure.
loading procedure.This Thiswaswas successfully
successfully done done andand the
thenumerical
numerical
model conducted to highlight interesting and encouraging aspects
model conducted to highlight interesting and encouraging aspects of the specific of the specific behaviour
behav-
iour under a lateral load that varies in its direction during the cyclic process. The main
observations are:
— As expected, the pile head displacement suffers a displacement accumulation in both
horizontal directions during the cycles, leading to a misalignment of the loading and
pile displacement directions;
J. Mar. Sci. Eng. 2021, 9, 235 20 of 22
under a lateral load that varies in its direction during the cyclic process. The main observa-
tions are:
— As expected, the pile head displacement suffers a displacement accumulation in both
horizontal directions during the cycles, leading to a misalignment of the loading and
pile displacement directions;
— The pile head displacements are greatly increased compared to the mono-directional
cyclic loading with the same average and cyclic amplitude values. New plastic
zones can develop throughout the cyclic loading process as additional deviated
displacements go along;
— The pile head displacement does not follow a logarithmic law anymore. The rate
of displacement accumulation decreases after about ten cycles, to reach a rate of
accumulation similar to the one obtained on the mono-directional loading;
Nonetheless, the effect of the constitutive model on these numerical findings still
needs to be highlighted.
In this study, a limited number of cycles of load were applied, as they were all
simulated one after the other (cycle-by-cycle simulation). To investigate the behaviour
under higher cycle numbers (hundreds, thousands or more), particularly an appropriate
constitutive model should be used, but for cycle-by-cycle simulation, as this would lead to
a numerical drift of the result along a high number of cycles (accumulation of numerical
inaccuracies), as well as a high computational time demand. An alternative approach could
be by using a Miner type law [27] or a strain accumulation model [28], that accounts for
the accumulation of strains as a function of the number of cycles, relative density and load
characteristics. Nevertheless, the applicability to multi-directional cyclic loading still needs
to be demonstrated and calibrated for this type of model. The “cycle-by-cycle” numerical
approach thus still has a useful future in order to highlight the mechanisms taking place in
the system under such a complex loading path.
Moreover, this numerical approach could be extended to a parametric study on the
pile bending stiffness (stiffer or more flexible piles compared to surrounding soil), to other
loading conditions in terms of average and cyclic amplitude, but also by varying the load
history, as it seems to be of high impact on the pile behaviour [26].
5. Conclusions
A numerical procedure has been implemented in Flac3D using the program built-in
“FISH” language, to apply a multi-directional cyclic lateral loading at pile head. The
specific feature lays in the fact that a displacement is applied on the pile head (through a
velocity due to the program formulation), that is thus controlled and adjusted based on pile
forces conditions to meet the loading direction requirement. The effect of multi-directional
loading compared to mono-directional loading has been highlighted, leading to larger pile
displacements.
The main numerical work perspective is to perform the simulations with a more
complex constitutive model, for instance with kinematic hardening shear mechanism or
multi-surface, to account for ratcheting and soil stiffness evolution.
Furthermore, the azimuth amplitude and the loading path followed during the cycles
of the multi-directional loading certainly influence the results. The proposed numerical
model and developed procedures could be easily implemented to perform simulations
under additional and more complex multi-directional loading scenarios. Simulations
combining the effects of lateral and vertical cyclic loading could also be performed, to get
even closer to the reality of the loading conditions of mutualized anchor foundations for
floating wind turbines.
J. Mar. Sci. Eng. 2021, 9, 235 21 of 22
Author Contributions: Bibliographic review, A.O. and O.J.; conceptualization, F.E., C.D. and O.J.;
general methodology, O.J. and F.E.; numerical methodology, O.J. and F.E.; numerical development,
O.J.; preliminary numerical parametric study and result analysis, A.O. under the supervision of F.E.
and O.J.; final numerical investigation and data curation, O.J.; writing—original draft preparation,
O.J.; writing—review and editing, F.E., C.D. and O.J.; validation, F.E. and C.D. All authors have read
and agreed to the published version of the manuscript.
Funding: This research received no external funding.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: The data presented in this study are available on request from the
corresponding author.
Acknowledgments: The authors thank the reviewers for their relevant comments. The characteristics
of the numerical model (loading cases and pile geometry) come from discussions with G. Damblans
and R. Isorna from France Energies Marines. The authors thank them for these data that permitted to
perform a study in a relevant frame, with practical applications.
Conflicts of Interest: The authors declare no conflict of interest.
References
1. Abadie, C.N. Cyclic Lateral Loading of Monopile Foundations in Cohesionless Soils. Ph.D. Thesis, St Catherine’s College
University of Oxford, Oxford, UK, 2015.
2. Doherty, P.; Gavin, K. Laterally loaded monopole design for offshore wind farms. Proc. Inst. Civ. Eng. Energy 2012, 165, 7–17.
3. Poulos, H.; Hull, T. The role of analytical geomechanics in foundation engineering. In Foundation Engineering: Current Principles
and Practices; ASCE: Reston, VA, USA, 1989; Volume 2, pp. 1578–1606.
4. API. Recommended Practice for Planning, Designing and Constructing Fixed Offshore Platforms—Working Stress Design (API RP 2A
WSD); American Petroleum Institute (API): Washington, DC, USA, 2000.
5. DNV. Offshore Standard DNV-OS-J101, Design of Offshore Wind Turbine Structures; Det Norske Veritas: Høvik, Norway, 2014.
6. Augustesen, A.H.; Brødbæk, K.T.; Møller, M.; Sørensen, S.P.H.; Ibsen, L.B.; Pedersen, T.S.; Andersen, L. Numerical Modelling
of Large-Diameter Steel Piles at Horns Rev. In Proceedings of the Twelfth International Conference on Civil, Structural and
Environmental Engineering Computing 2009, Funchal, Madeira, Portugal, 1–4 September 2009. Paper 239.
7. Yang, M.; Ge, B.; Li, W. Force on the laterally loaded monopile in sandy soil. Eur. J. Environ. Civ. Eng. 2018. [CrossRef]
8. LeBlanc, C.; Houlsby, G.T.; Byrne, B.W. Response of stiff piles in sand to long-term cyclic lateral loading. Géotechnique 2010, 60,
79–90. [CrossRef]
9. Roesen, H.R.; Ibsen, L.B.; Andersen, L.V. Experimental testing of monopiles in sand subjected to one-way long-term cyclic lateral
loading. In Proceedings of the Int. Conference on Soil Mechanics and Geotechnical Engineering (ICSMGE), Paris, France, 2–6
September 2013; pp. 2391–2394.
10. El Haffar, I. Physical Modeling and Study of the Behaviour of Deep Foundations of Offshore Wind Turbines in Sand. Ph.D. Thesis,
École Centrale de Nantes, Nantes, France, 2020.
11. Puech, A.; Garnier, J. Design of Piles under Cyclic Loading, SOLCYP Recommendations; ISTE Editions: London, UK, 2017.
12. Lin, S.-S.; Liao, J.-C. Permanent strains of piles in sand due to cyclic lateral loads. J. Geotech. Geoenviron. Eng. 1999, 125, 798–802.
[CrossRef]
13. Verdure, L.; Garnier, J.; Levacher, D. Lateral cyclic loading of single piles in sand. Int. J. Phys. Model. Geotech. 2003, 3, 17–28.
[CrossRef]
14. Peralta, V.P. Investigations on the Behaviour of Large Diameter Piles under Long-Term Lateral Cyclic Loading in Cohesionless
Soil. Ph.D. Thesis, University of Hannover, Hannover, Germany, 2010.
15. Rosquoët, F.; Thorel, L.; Garnier, J.; Canepa, Y. Lateral cyclic loading of sand installed piles. Soils Found. 2007, 47, 821–832.
[CrossRef]
16. Wolf, T.K.; Rasmussen, K.L.; Hansen, M.; Roesen, H.R.; Ibsen, L.B. Assessment of p-y Curves from Numerical Methods for a Non-slender
Monopile in Cohesionless Soil; Technical Memorandum No. 024; Department of Civil Engineering, Aalborg University: Aalborg,
Denmark, 2013; 17p.
17. Achmus, M.; Kuo, Y.S.; Abdel-Rahman, K. Behaviour of monopile foundations under cyclic lateral load. Comput. Geotech. 2009,
36, 725–735. [CrossRef]
18. Su, D.; Li, J.H. Three-dimensional finite element study of a single pile response to multidirectional lateral loadings incorporating
the simplified state-dependent dilatancy model. Comput. Geotech. 2013, 50, 129–142. [CrossRef]
19. Lovera, A.; Ghabezloo, S.; Sulem, J.; Randoplh, M.; Kham, M.; Palix, E. Pile response to multi-directional lateral loading using
P–y curves approach. Géotechnique 2020. [CrossRef]
20. Mayoral, J.M.; Pestana, J.M.; Seed, R.B. Multi-directional cyclic p–y curves for soft clay. Ocean Eng. 2016, 115, 1–18. [CrossRef]
J. Mar. Sci. Eng. 2021, 9, 235 22 of 22
21. Rudolph, C.; Bienen, B.; Grabe, J. Effect of variation of the loading direction on the displacement accumulation of large-diameter
piles under cyclic lateral loading in sand. Can. Geotech. J. 2014, 51, 1–11. [CrossRef]
22. Itasca Consulting Group Inc. Flac3D User’s Guide, v.5.01; Itasca Consulting group Inc.: Minneapolis, MN, USA, November 2013.
23. France Energies Marines. Actions R&D sur la mutualisation des ancrages pour l’éolien flottant. Presented at CFMS Mirror Group
TC 209, Marne-La-Vallée, France, 26 September 2019.
24. Obaei, A. Numerical Modelling of a New Type of Foundation System for Offshore Applications. Master’s Thesis, Université
Grenoble Alpes, Grenoble, France, 2020.
25. Detournay, C.; Wang, Z.; Han, Y. Modeling of Nevada Sand Behaviour Using CHSOIL. In Proceedings of the 2nd FLAC/DEM
Symposium - Continuum and Distinct Element Numerical Modeling in Geomechanics, Melbourne, Australia, 14–16 February
2011; Sainsbury, D., Hart, R., Detournay, C., Nelson, M., Eds.; pp. 8–13.
26. Abadie, C.N.; Byrne, B.W.; Houlsby, G.T. Rigid pile response to cyclic lateral loading: Laboratory tests. Géotechnique 2019, 69,
863–876. [CrossRef]
27. Lesny, K.; Hinz, P. Investigation of monopile behaviour under cyclic lateral loading. In Proceedings of the 6th International
Conference on Offshore Site Investigation and Geotechnics, London, UK, 26–28 November 2007; pp. 383–390.
28. Page, A.M.; Klinkvort, R.T.; Bayton, S.; Zhang, Y.; Jostad, H.P. A procedure for predicting the permanent rotation of monopiles in
sand supporting offshore wind turbines. Mar. Struct. 2021, 75. [CrossRef]

Jmse 09 00235 v2

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Jmse 09 00235 v2

Uploaded by

Copyright:

Available Formats

Journal of

J. Mar. Sci. Eng. 2021, 9, 235. https://doi.org/10.3390/jmse9020235 https://www.mdpi.com/journal/jmse

yN /y1 − 1 = 0.102/CR × ln(N) × (Hc /Hmax )0.35

where CR is a rigidity coefficient defined as

CR = [Ep Ip /(Ep Ip )fl ]1/5 (4)

However, no numerical study of pile subjected to lateral multidirectional cyclic loading

The objective of this work was to develop a numerical approach to acquire

J. Mar. Sci. Eng. 2021, 9, 235 5 of 22

J. Mar. Sci. Eng. 2021, 9, 235 6 of 22

Figure 7. Triaxial numerical test results with CHsoil.

Table 2. Parameters of the CHsoil model.

Parameter Significance value

Table 2. Parameters of the CHsoil model.

Parameter Significance Value

2.4. General Modelling Procedure

2.5. Loading Procedure

2.5.2. Loading Procedure for Multi-Directional Cyclic Loading

Figure 8. Force applied on the pile head.

Figure 8. Force applied on the pile head.

Figure 10. Numerical procedure for multi-directional unloading.

Figure 15. Parameter α from the law yN/y1 – 1 = α lnN.

3.3. Multi-Directional Lateral Cyclic Loading

3.3. Multi-Directional Lateral Cyclic Loading

3.3. Multi-Directional Lateral Cyclic Loading

Figure 20. Pile

J. Mar. Sci. Eng. 2021, 9, x FOR PEERin loading

α = 0.102/ CR × (Hc /Hmax )0.35 (9)

You might also like