Soil Dynamics and Earthquake Engineering 138 (2020) 106223

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering

journal homepage: http://www.elsevier.com/locate/soildyn

Impact of the installation on the long-term cyclic behaviour of piles in sand:

A numerical study
P. Staubach a, *, J. Machaček b, M.C. Moscoso c, T. Wichtmann d
a
Chair of Geotechnics, Bauhaus University Weimar, Germany
b
Chair of Soil Mechanics, Foundation Engineering and Environmental Geotechnics, Ruhr University Bochum, Germany
c
Chair of Steel Construction, Bauhaus University Weimar, Germany
d
Professor and Head of the Chair of Soil Mechanics, Foundation Engineering and Environmental Geotechnics, Ruhr University Bochum, Germany

A R T I C L E I N F O A B S T R A C T

Keywords: The installation process changes the state of the soil in the vicinity of a pile and therefore influences its load-
Pile driving displacement behaviour during the operational phase following the installation. The installation process itself
High-cyclic loading and its effect on the pile behaviour under a subsequent high-cyclic loading are numerically investigated in this
High-cycle accumulation model
work. Two different installation techniques, namely impact driving and jacking, are investigated using a coupled
Coupled Eulerian-Lagrangian
Partially drained
Eulerian-Lagrangian approach in combination with a hypoplastic material model. Either ideally drained or
partially drained conditions are assumed. Afterwards the pile response to a high-cyclic loading with up to five
million load cycles is studied using a high-cycle accumulation (HCA) model. Compared to a wished-in-place
(WIP) pile, the simulations accounting for the installation process showed considerable less permanent pile
displacements in case of initially medium dense sand. However, the opposite was observed in case of initially
dense sand. Current practice assuming a WIP state as the starting point in numerical simulations seems thus to be
conservative for initially medium dense sand but leads to an underestimation of permanent deformations for
initially dense sand.

1. Introduction intergranular strain extension of Niemunis & Herle [4]. In order to take
into account the generation and dissipation of excess pore water pres­
The realistic and reliable prediction of the long-term deformations of sures in the soil during the installation process, the coupled
pile foundations subjected to high-cyclic loading is an important field of Eulerian-Lagrangian method is extended such that the consolidation
current research in geotechnical engineering, in particular with respect process is incorporated. The installation of the pile by either impact
to offshore wind turbine foundations. So far, some numerical studies on driving or jacking is simulated. The change of the soil state (density,
the long-term behaviour of piles have been performed (e.g. Ref. [1,2]). effective stress) during the installation process obtained from the sim­
However, the influence of the installation process has not been taken ulations is qualitatively compared to observations from pile driving
into account. Therefore, its influence on the load-displacement behav­ experiments documented in the literature.
iour of the pile has not been quantified yet. Furthermore, model tests, After the simulation of the installation, the response of the pile under
which are often used for the validation of numerical models, usually high-cyclic loading is investigated in a subsequent implicit (in terms of
include the installation of the piles into the soil. The corresponding time integration) calculation, using the high-cycle accumulation (HCA)
numerical simulations, however, start from wished-in-place (WIP) model of Niemunis et al. [5]. As shown in Fig. 1, the soil state at the end
conditions. of the simulation of the installation process is imported into a new model
The present work aims to investigate the influence of the installation in order to perform the calculation with the HCA model under
process on the long-term behaviour of piles under high-cyclic loading. quasi-static conditions. This is necessary because the CEL model for the
To account for the large deformations during the installation process, a installation process is restricted to an explicit time integration scheme
coupled Eulerian-Lagrangian (CEL) approach is used in combination used for dynamic processes. Two different cyclic loading conditions are
with the hypoplastic material model of von Wolffersdorff [3] with the considered: a predominantly horizontal and a purely vertical loading of

* Corresponding author.
E-mail address: [email protected] (P. Staubach).

https://doi.org/10.1016/j.soildyn.2020.106223
Received 18 October 2019; Received in revised form 5 May 2020; Accepted 5 May 2020
Available online 21 August 2020
0267-7261/© 2020 Elsevier Ltd. All rights reserved.
P. Staubach et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106223

The simulations with the HCA model are conceptually divided into a
low-cycle (also denoted as “implicit”) and a high-cycle (also denoted as
“explicit”) mode. In the low-cycle mode a conventional calculation is
performed using a common constitutive model (hypoplasticity with
intergranular strain is used in the present study). The strain amplitude
εampl , which is an important input parameter of the HCA model, is
determined in each element from based on the strain path recorded
during an individual load cycle. As depicted in Fig. 1, this can either be
the second cycle or a so-called update cycle. Based on the recorded six-
dimensional strain path (six independent components of the strain
tensor for 3D cases), the strain amplitude is defined as the euclidean
norm of six spans obtained by consecutive projection (degeneration) on
(hyper-) planes (for details see Ref. [8]). Recently, this definition of a
strain amplitude for multidimensional cycles has been confirmed by
laboratory tests with up to four stress components that have been
cyclically varied at the same time [14]. For a one-dimensional cyclic
loading path, the classical definition εampl = (εmax − εmin )/2 holds with
εmax/min denoting maximum/minimum principal strains.
In the high-cycle mode the accumulation trend is predicted by the
HCA model for a given number of loading cycles, using the evaluated
strain amplitude as input. This calculation strategy is illustrated in the
scheme on the right-hand side of Fig. 1. During the high-cycle phases the
strain amplitude is assumed constant. The field of the strain amplitude
can be updated in an update cycle, which is simulated in the low-cycle
mode again. For many applications the assumption of a constant strain
amplitude throughout the whole simulation is justified. Update cycles in
regular intervals may be necessary if the high-cyclic loading evokes
Fig. 1. Sequence of calculation steps: After the installation of the pile the high- large changes in pore water pressure and thus changes the stiffness of the
cyclic loading is simulated by means of the HCA model using a combination of soil significantly. The importance of update cycles under such conditions
low-cycle and high-cycle calculation strategies. is demonstrated in Ref. [2].
During the high-cycle phase only the changes of the average values of
the pile. The influence of the installation on the short-term and stress, strain and other state variables for a given increment ΔN of the
long-term (number of cycles N ≳ 104 ) displacements of the pile is studied number of cycles are calculated. The rate of accumulated strain (rate
and explained with changes of the stress state and the density distri­ denotes here the derivative of a variable with respect to the number of
bution. In addition, the influence of the installation technique (impact cycles N) during this high-cycle mode is calculated as the product of the
driving vs. jacking) on the long-term behaviour of the pile foundation is intensity of accumulation ε̇acc (scalar variable) and the direction of
analysed. accumulation m (unit tensor). The intensity of accumulation is again
The following notation is used: Vectors and second order tensors are obtained as the product of five factors (fi ). These factors account for the
written with bold font, e.g. u and σ . The divergence is div(a) = ai,i for i = influences of the strain amplitude (fampl ), the number of previous loading
1, 2, 3 and the gradient grad(a) = ai,j for i, j = 1, 2, 3. For the vector cycles weighted by their amplitude (f˙ ), the void ratio (fe ), the mean
N
product (⋅), a⋅b = ai bi holds. The trace of a tensor is tr(ε) = εi,i for i = effective stress (fp ) and the stress obliquity (fY ). An influence of a change
1, 2, 3. The first and the second time derivative of the displacement u are of the loading direction (fπ ) is irrelevant for the present simulations with
u̇ and ü, respectively. a constant direction of the cycles and therefore neglected herein.
The rate of accumulated strain is thus calculated from
2. Utilized constitutive models
ε̇acc = fampl ​ f˙N ​ fe ​ fp ​ fY ​ m. (1)
2.1. Hypoplastic model
For a detailed definition of the factors fi and the direction of accumu­
The hypoplastic model proposed by von Wolffersdorff [3] with the lation m the interested reader is referred to Ref. [15].
intergranular strain extension by Niemunis & Herle [4] is applied in the The changes of average effective stress Δσ = σ̇ ΔN or strain Δε = ε̇ΔN
simulations of the installation process. The same constitutive model is caused by an increment ΔN of the number of cycles are calculated from
also used for the first two cycles and the update cycles in the subsequent the basic equation of the HCA model:
calculation with the HCA model as explained in Section 2.2. If the mean σ̇ = E : (ε̇ − ε̇acc − ε̇p ), (2)
effective stress approaches zero, the material response becomes unde­
fined in this constitutive model. As a remedy, a correction for small wherein E is a barotropic elastic stiffness and ε̇p is a plastic strain rate.
mean effective stresses (p < 0.01 kPa) followed by another correction of Depending on the boundary conditions, Eq. (2) predicts either a change
stress states located outside the Matsuoka-Nakai failure locus (see Sec­ of (average) effective stress, or a change of (accumulated) strain, or both
tion 2.2) is applied as proposed in Ref. [6] and also used in Ref. [2,7]. under high-cyclic loading.
The plastic strain rate ε̇p is only needed if the stress path reaches the
M − N failure locus during the high-cyclic loading. This failure locus is
2.2. High-cycle accumulation model
given by
The high-cycle accumulation (HCA) model of Niemunis et al. [8] is I1 I2
F= − − 9 − 8tan2 (ϕ) ≤ 0, (3)
applicable to soils loaded by a large number of cycles (N ≳ 100). The I3
constitutive description is based on numerous laboratory tests on sand
samples subjected to high-cyclic loading [9–13]. where I1 , I2 , I3 are the principal invariants of the effective stress tensor

2
P. Staubach et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106223

and ϕ is the friction angle. Table 2

The HCA model has been validated by the successful numerical Parameters of the HCA model for Karlsruhe fine sand [15].
reproduction of measured data from model tests on monopiles and Campl Ce Cp CY CN1 [10− 4 ] CN2 CN3 [10− 5 ]
shallow foundations [16–18], a large-scale field test on a gravity foun­
dation [19] and real structures [20]. An extensive parametric study on 1.33 0.60 0.23 1.68 2.95 0.41 1.90

the long-term behaviour of an offshore monopile foundation using the

HCA model can be found in Refs. [2]. 3.1. Coupled Eulerian-Lagrangian analysis
The described calculation procedure with low-cycle and high-cycle
phases has several advantage: 1) there are no limitations with respect In a Eulerian analysis the mesh is fixed in space and the material
to the maximum number of cycles that can be considered in the simu­ moves through it. Opposite to a Lagrangian analysis, the elements are
lations; 2) the computational effort is much less than in calculations with not necessarily completely occupied by material. In most numerical
conventional constitutive models where every cycle is calculated with codes using Eulerian elements, an increment is divided into two steps. In
many increments; 3) the accumulated error resulting from inaccuracies the first step the solution is progressed in time. In the second step the
of the constitutive model and from the residuals allowed in the inte­ solution is then mapped back to the Eulerian mesh, which is called the
gration scheme is much less, because a significantly lower number of remap or advection step. The strategy of solving the problem in two
increments is involved in the simulations. Furthermore, with the HCA separate steps is often referred to as operator ​ split since the underlying
model a special constitutive description capturing the various de­ differential equations are solved term-by-term (see e.g. Refs. [24–26]).
pendencies of the strain accumulation rate is used. It should be stressed This procedure is the same for the arbitrary Lagrangian-Eulerian (ALE)
that even more sophisticated “conventional” constitutive models like approach. However, in ALE an additional step is performed in which the
hypoplasticity with intergranular strain or Sanisand [21] predict an Eulerian grid is re-meshed.
almost linear increase of the accumulated strain with increasing number A method in which both Lagrangian and Eulerian elements are pre­
of cycles, which contradicts the experimental observations. This deficit sent and interact with each other is referred to as coupled Eulerian-
will inevitably lead to a strong overestimation of the permanent de­ Lagrangian (CEL). In the commercial finite element code Abaqus,
formations if large numbers of cycles are simulated. Thus, simulations which has been used for the present study, the CEL method is imple­
with these constitutive models should be restricted to low numbers of mented using an explicit time integration scheme. Recently, the CEL
cycles (N ≤ 10). However, it should also be mentioned that new con­ method has gained increasing interest in geotechnics due to its easy
stititive models are currently under development which could at least accessibility through commercial finite element codes such as Abaqus.
overcome the latter problem. Successful applications of the CEL approach to geotechnical problems
For instance, Liu et al. [22] extended a multi-surface plasticity model have been reported e.g. in Refs. [27–31]. However, these simulations
by a memory surface allowing for the successful reproduction of drained were restricted to either perfectly drained or locally undrained
triaxial tests with up to 104 cycles in simulations where every individual conditions.
cycle was calculated. Recently, Houlsby et al. [23] presented a ther­ It is worth mentioning that apart from using Eulerian-based analysis
modynamically consistent model based on the framework of hyper­ approaches, pile driving has recently also been successfully simulated
plasticity which is able to describe cumulative effects due to a large using the material point method (MPM) (see e.g. Ref. [32]). Simulations
number of cycles using “accelerated” ratcheting. In this approach only a of impact driving incorporating fluid flow have been successfully
few cycles are calculated conventionally with the hyperplastic model, benchmarked with measurements reported in Refs. [33].
and the resulting permanent deformations are increased in dependence
of the maximum number of cycles to be considered in the high-cycle 3.2. Incorporation of the consolidation process
problem. No calculation of a strain amplitude is necessary in such sim­
ulations which may be advantageous for complicated loading paths To account for the change in pore water pressure and the resulting
during individual cycles. consolidation, the CEL method is extended by the pore water pressure as
an additional field variable. This is done based on an idea reported in
2.3. Constitutive model parameters Ref. [34], where a coupled temperature-displacement analysis is per­
formed with Abaqus, but the temperature is replaced by the excess pore
The installation and the high-cyclic loading are simulated using the water pressure. In the following, this approach is elaborated in detail
constitutive model parameters of Karlsruhe fine sand. The parameters since in the present work an extension of the method used in Ref. [34] is
for the hypoplastic model with intergranular strain are given in Table 1, introduced.
while those for the HCA model are summarized in Table 2. These pa­ To introduce the pore water pressure pw as an additional unknown,
rameters have been determined from laboratory tests performed on the mass balance of the pore water is chosen as additional field equation
Karlsruhe fine sand as reported in Refs. [15]. 1 w
ϕw w w w s
w ṗ + ϕ div(u̇ ) + (1 − ϕ ) ​ div(u̇ ) = 0, (4)
K
3. Simulation of pile installation
w
where ϕw is the volume fraction of the pore water, K is the bulk
Due to the large deformations involved in the installation of the pile, modulus of the water, u̇w is the velocity of the water and u̇s is the velocity
a coupled Eulerian-Lagrangian approach is used in this work. To take of the solid. The resulting element formulation is known as the u-p
into account partially drained conditions, an extension incorporating formulation [35] (discretisation of displacement (u) and pore water
the capability to simulate the build-up of excess pore water and its pressure (p)).
dissipation by consolidation is proposed. Since the velocity of the pore water u̇w is unknown, the generalized

Table 1
Parameters of hypoplasticity with intergranular strain for Karlsruhe fine sand [15].
ϕc ei0 ec0 ed0 hs n α β R mR mT βR χ

[− ] [− ] [− ] [− ] [MPa] [− ] [− ] [− ] [− ] [− ] [− ] [− ] [− ]
33.1◦ 1.212 1.054 0.677 4000 0.27 0.14 2.5 10− 4 2.4 1.2 0.1 6.0

3
P. Staubach et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106223

Darcy law is used in addition phase is calculated based on the time derivative of the volumetric strain
Perm
rate, using an explicit mid-point rule given by
K
ww = [ − grad(pw ) + ρw (b − üs )], (5)
ηw tr(ε̇)t+Δt − tr(ε̇)t div(üs )t
div(üs )t+Δt = + , (13)
2Δt 2
with the relative velocity between water and soil ww = ϕw (u̇w − u̇s ), the
w w
permeability of the solid KPerm = k γwη (with the hydraulic conductivity where ()t+Δt corresponds to the current and ()t to the previous
increment.
kw and the dead weight of water γw ), the dynamic viscosity of the water
The verification of this approach is given in Appendix A where the
ηw , the density of the water ρw and the gravity b. Note that the perme­
wave propagation in a two-phase medium is simulated under consid­
ability is assumed to be isotropic.
eration of the pore fluid flow and compared to an analytical approach.
Note that in Eq. (5) the acceleration of the pore water üw has been
The major difference to Refs. [34] is the incorporation of the accel­
replaced by the acceleration of the solid üs . This is necessary since the
eration of the solid in Eq. (11) which has been neglected in the previous
acceleration of the water is unknown. This simplification is justified if
study. Therefore, the presented approach enables to simulate problems
the wave frequency and the hydraulic conductivity are not too large. As
with high accelerations as faced during fast jacking, impact driving or
demonstrated in Ref. [36], this approximation is appropriate even for
vibratory driving.
large frequencies (> 30 Hz) in combination with large hydraulic con­
ductivities (> 1⋅10− 3 m/s) and can be applied for instance also for the
simulation of vibratory driving. In the present case it is not expected that 3.3. Specifications of the model and sequence of steps
waves with considerable magnitudes and frequencies higher than 30 Hz
are caused by the pile driving process. Therefore, the assumption üw = üs The model used for the simulation of the installation including the
is justified. finite element mesh is shown in Fig. 2. Considering the symmetry only
Substituting Eq. (5) in Eq. (4) results in one quarter of the problem is modelled. The red area represents the
{ Perm } region initially filled with material while the blue area is initially empty
1
ϕw w ṗw + div
K
[ − grad(p w
) + ρ w
(b − üs
)] + div(u̇s ) = 0. (6) but could be filled if the material moves into the elements during the
K ηw simulation. The pile has an outer diameter of 2 m, a wall thickness of 9
Eq. (6) is discretised with respect to the pore water pressure in the cm and is driven into the soil up to a depth of 10 m. The pile is assumed
finite element framework. to be rigid. The considered soil area has a diameter and a height of 100
To incorporate Eq. (6) into the explicit framework offered by Abaqus, m. The model size is chosen such that all mechanical waves resulting
the similarity between the energy balance used for the simulation of from the installation process are damped out before they approach the
thermal processes and Eq. (6) is utilized. The energy balance is given by boundaries. The size of the elements increases towards the boundaries of
the model in order to achieve a “damping” effect by damping the waves
ρcθ̇ + λ ​ div[ − grad(θ)] = − ṁT . (7) when travelling towards the borders as no transparent boundary con­
ditions were used. The applicability of this approach was verified by
Therein, ρ is the density (the saturated soil density in the present case), c examining the acceleration magnitudes during the driving process
the specific heat, θ the temperature, λ the thermal conductivity and ṁT which were found negligible for distances to the pile greater than
the internal heat production. Introducing the excess pore water pressure approximately 30 m. A more detailed analysis of this aspect is included
Δpw , Eq. (6) can be rewritten as in Appendix B. Due to the huge calculation time, especially in case of the
Δṗw K Perm K Perm partially drained conditions, the number of elements was kept at a
ϕw w + div[ − grad(Δpw )] − div(üs ) + div(u̇s ) = 0. (8) minimum.
K η w ηw
The pile is either jacked with a constant penetration rate of 0.3 m/s
This conversion is possible since the hydrostatic pore water pressure or driven by impact. In case of impact driving a static load of 2000 kN
does not change with time and ρw b − grad(pw ) = − grad(Δpw ) holds. due to the self weight of the pile and the weight of the ram as well as an
Setting the temperature equal to the excess pore water pressure θ = impact force of 860 kN are applied as point loads at the head of the pile
Δpw , the following matching terms can be identified from Eq. (8) and (all values of the loads are given for a full model). The blow rate of the
Eq. (7) ram is assumed to be 38 blows per minute. These specifications are
Δṗw associated with the hydro-hammer IHC S-900 [37]. The loading history
cρθ̇ = ϕw w , (9) for the impact driving is shown in Fig. 3.
K

K Perm
λ div[ − grad(θ)] = div[ − grad(Δpw )], (10)
ηw

K Perm
ṁT = div(u̇s ) − div(üs ). (11)
ηw
The constants c and λ are given following Eq. (9) and Eq. (10):

ϕw K Perm
c= w and λ = w . (12)
K ρ η
To perform an explicit calculation with Abaqus under consideration
of excess pore water pressure and seepage flow, a dynamic thermal-
displacement analysis is conducted. The material constants for the
specific heat and the thermal conductivity are defined as given by Eq.
(12). The boundary conditions of the temperature are set according to
free water surfaces (zero excess pore water pressure). The material heat
production is defined according to Eq. (11) within the VUMAT interface
provided by Abaqus. The divergence of the acceleration of the solid
Fig. 2. Finite element model for the installation process.

4
P. Staubach et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106223

The contact between pile and soil is modelled using a surface-to- where jacking tended to lead to greater soil plugging compared to
surface discretisation. In case of partially filled Eulerian elements a impact driving. Also, all simulations predict a densification alongside
reconstruction of the surface is performed in order to correctly enforce the inner of the pile shaft, while a significant decrease of relative density
the non-penetration condition. A Mohr-Coulomb friction model with a is observed outside of the pile shaft. This is most prone in case of impact
friction coefficient of μ = 0.25 for the ideally drained case is used. Un­ driving and the initially dense case. As will be shown in the following,
fortunately, Abaqus provides no possibility to define the frictional the stress alongside the outer shaft is significantly reduced (approxi­
behaviour by a user subroutine when using coupled Eulerian-Lagrangian mately to 10 kPa). Using the empirical relation for the critical void ratio
elements. This, however, leads to an overestimation of shear stresses in proposed by Bauer [39].
the simulations considering the build-up of excess pore water pressure as
the calculation of frictional forces is based on total stresses (and not ec = ec0 ⋅e(− 3p/hs )n
, (14)
effective stresses). To counterbalance this, the friction coefficient was
where ec0 is the critical void ratio at a mean effective stress of p = 0 kPa
reduced by a factor of 2 to μ = 0.125 in the simulations with partially
and hs and n are the material constants given in Table 1,
drained conditions (since the specific weight of the water-saturated soil
ec (p = 10 kPa) ≈ 1.0 is calculated. Hence, the critical void ratio
is 20 kN/m3 and the specific buoyant weight is 10 kN/m3 the difference
belonging to the stress state alongside the outer shaft is comparably
between effective and total stress corresponds to a factor 2). Obviously,
large (ec ≈ 1.0, whereas emax = ec0 = 1.054). Thus, the soil state at the
this assumption leads to completely wrong friction forces if significant
outer pile shaft (e ≈ 0.8) is dense after the passing of the pile tip and
changes in the pore pressure occur, e.g. in case of a full liquefaction of
behaves dilatant during shearing. Due to the shearing while driving, the
the soil.
void ratio tends towards the critical void ratio at the outside, while the
For the simulations incorporating excess pore water pressure, the
tendency towards soil plugging inside of the pile leads to reduced
phreatic level is assumed to be located 2 m below the ground level. The
w shearing alongside the shaft and therefore less dilatancy. This effect is
bulk modulus of the pore water is assumed to be K = 30 MPa which is amplified by the higher stress state inside the pile which results in a
approximately equivalent to a degree of saturation of S = 0.995 (using a smaller ec .
porosity of n = 0.5 and a constant bulk modulus of air of 100 kPa). The Fig. 5 presents the spatial distribution of mean effective stress during
hydraulic conductivity is varied as described in Section 3.5. the impact driving at two different times. A sharp increase in effective
For the soil two different initial relative densities are considered: stress is observed below the pile tip. This increase spreads over a large
Dr0 = 40% and Dr0 = 80%. The density of the saturated soil is assumed area and reaches to a depth below the tip of approximately two times the
to be ρ = 2 g/cm3 for both cases. During the calculation, the density is radius of the pile. The distribution of the mean effective stress along the
assumed to remain constant. The initial value of the lateral stress coef­
ficient is K0 = 0.5.
In the first step of the calculation the gravity (g = 10 m/s2) is applied
to the soil, without generating any deformations. In case of jacking, in
the second step the pile is pushed into the soil with a prescribed velocity
until an embedment depth of 10 m is reached. In case of impact driving
the self weight of the pile and the ram is applied in the second step,
leading to an initial penetration of the pile into the soil. After 2 s, the
impact load is brought up on the pile in the third step. Again, this third
step lasts until the pile is driven 10 m into the soil.

3.4. Results for ideally drained conditions

In this section, the results of simulations assuming ideally drained

conditions are presented.
The fields of the relative density after the installation of the pile are
given in Fig. 4 for the initially medium dense and for the initially dense
sand. The upper figures depict the results for jacking while the lower
ones those for impact driving. In all cases a strong densification at the
pile tip is observed. This is slightly more pronounced in case of jacking
which is in accordance with experimental results presented in Ref. [38],

Fig. 4. Field of relative density after the installation of the pile for the initial
medium dense state (left) and initial dense state (right) using jacking and
Fig. 3. Force history in case of impact driving. impact driving of the pile.

5
P. Staubach et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106223

black horizontal line shows a fast decrease with increasing distance from
the pile tip which is schematically drawn in the graph below the contour
plots. With increasing pile driving, the effective stress alongside the
shaft is strongly reduced as it is evident in the effective mean stress
distribution at the later time instance t = 35 s and the corresponding
scheme below the field plot which sketches the course of the effective
mean stress along the black horizontal line. Similar stress changes were
also observed in experiments of pile installation processes described e.g.
in Ref. [40–42]. When the pile passes the soil, the strongly densified soil
around the (outer) pile tip collapses as the load from the pile is removed
which leads to a sudden reduction of the stress acting on the pile shaft.
Fig. 6 depicts the fields of mean effective stress after the simulation of
the installation by jacking for the two studied initial densities. Again, a
considerable increase in effective stress is observed at the pile tip. The
region of high effective stress reaches far below the pile tip, which is
Fig. 6. Field of mean effective stress at the end of the pile installation
even more pronounced in case of the initially medium dense sand. In­
via jacking.
dependent of the initial relative density, a great reduction of effective
stress is observed alongside the outer pile shaft. This stress reduction is
caused by the same mechanism as explained earlier for the impact
driving. Behind the zone of stress reduction, a zone of stress increase is
observed. Therefore, the stress increase caused by the compaction of the
soil below and along the side of the pile tip spreads over a greater dis­
tance than the effect of the collapse of the densified zone when the tip
passes. This is also visible in the schematic drawings at the base of in the
Fig. 5. Compared to the impact driven pile, the reduction of effective
stress inside the jacked pile is less pronounced which results from the
increased tendency of the jacked pile to form a soil plug.
To investigate the influence of the friction along the pile-soil inter­
face, a comparison of the fields of relative density after the installation of
the pile via impact driving is shown in Fig. 7 for the friction coefficients
μ = 0.25 and μ = 0.5, respectively. Clearly, the higher friction leads to
an increased tendency towards soil plugging and to greater dilatancy
alongside the outer shaft. Both effects are in accordance with theoretical
considerations.
In order to study if the same tendencies are observed when the build-
up of excess pore water pressure caused by the installation process is
considered, partially drained simulations are discussed in the next
section.

Fig. 7. Field of relative density after an installation via impact driving for two
different friction coefficients.

3.5. Results for partially drained conditions

The spatial distribution of excess pore water pressure during impact

driving of the pile is given in Fig. 8. The left part shows the distribution
at a time where the impact load is not active whereas the right part
depicts the excess pore water pressure directly after the re-activation of
the impact load. In addition, two different hydraulic conductivities kw =
10− 3 m/s and kw = 10− 4 m/s are compared. As expected, the installation
process leads to a build-up of excess pore water pressure concentrated
inside the pile and below the pile tip which is more pronounced in case
of the lower hydraulic conductivity.
Fig. 9 compares the distributions of relative density during impact
driving starting from an initially medium dense state (Dr0 = 40%) for
the different hydraulic conditions. Despite the relative large hydraulic
conductivity, the partially drained simulations with both kw values
predict considerably less densification in the vicinity of the pile tip and
no tendencies towards soil plugging. This can be traced back to the
development of excess pore water pressure which prevents the soil to
compact below the pile tip.
It has to be noted that the distributions shown in Fig. 9 do not belong
Fig. 5. Field of mean effective stress during impact driving of the pile. The
to a state where the consolidation process is finished. Therefore, it is to
sketched distributions below show the change of mean effective stress with
be expected that the distribution of relative density will slightly change
respect to the distance from the pile for the depths marked by the black hori­
with time after the installation.
zontal line in the field plots above (schematic graphs based on [40]).

6
P. Staubach et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106223

It can be concluded that already partially drained conditions with

comparably large hydraulic conductivities of kw = 10− 3 m/s to kw =
10− 4 m/s lead to great differences in the soil state obtained during and
after the installation of the pile compared to the ideally drained case.
Since no tendency towards soil plugging is observed for the case with
smaller hydraulic conductivity, this driven pile may be expected to have
considerable less loading capacity compared to a pile driven into a very
permeable sandy soil which is in accordance with previous in­
vestigations [43].

4. Response of the pile to high-cyclic loading

After the simulation of the installation of the pile, the fields of stress
and void ratio are imported into a new finite element model in order to
perform the subsequent simulation of the high-cyclic loading using an
implicit Lagrangian calculation. Note that the intergranular strain (in­
ternal tensorial variable of the hypoplastic model) has not been trans­
ferred as the alteration caused by the installation process proved to
negligibly influence the results of the cyclic loading analysis. Only the
simulations assuming ideally drained conditions with a friction coeffi­
cient of μ = 0.25 are considered in the following. The transfer of the
state variables is performed by the use of a python script where the state
variables of the integration points of the CEL model are transferred to
the closest (in terms of euclidean norm) integration points of the
Lagrangian model.
The model for the high-cyclic loading is shown in Fig. 10. Since a
constant direction of the cyclic loading is assumed, only one half of the
problem is modelled considering the symmetry. The modelled soil re­
gion has a height of 35 m and a diameter of 50 m. The pile is already
embedded and has the same specifications as the one in the simulation of
the installation process. However, the pile is deformable now (except of
an extension of the pile head by a short rigid body which is used to apply
the concentrated loading without distortion of the pile). The deforma­
tion of the soil caused by the installation is not considered as the in­
fluence is judged negligible in comparison to the influence of the
changes in effective stress and void ratio. In addition to the simulations
incorporating the effects of the installation process, calculations of a
wished-in-place (WIP) pile are performed assuming lithostatic and ho­
Fig. 8. Field of excess pore water pressure during impact driving. The left mogeneous initial fields of effective stress and void ratio, respectively.
figures shows the distribution while the impact load is not active. The right Those latter simulations give results for the case that the installation
figures shows the distribution directly after the application of the impact load. process is not considered.
The upper fields are obtained for a hydraulic conductivity kw = 10− 3 m/s while After the initialisation of the initial state in the first step, a vertical
the lower ones correspond to kw = 10− 4 m/s. force of 1000 kN is applied on the top of the pile in step No. 2 to consider
the self-weight of the structure (value is given for a full model). Two
cycles with a sinusoidal loading of the pile follow in steps Nos. 3 and 4.

Fig. 9. Comparison of the spatial distributions of relative density during

(approximately 8 m of driving) impact driving of the pile for an initially me­
dium dense case (Dr0 = 40%) and different hydraulic conditions.
Fig. 10. Finite element model for the high-cyclic loading.

7
P. Staubach et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106223

The loading is a combination of horizontal force and moment further The spatial distribution of the strain amplitude calculated based on
described in Section 4.1 or a purely vertical one, see Section 4.2. These the strain path recorded during the second cycle (phase 4–5) is given in
two steps are calculated with the hypoplastic model (low-cycle mode). Fig. 12. Therein, the calculations with and without consideration of the
In step No. 4, which is a repetition of step No. 3 in terms of loading, the installation process are compared. Since the fields for the impact driven
strain path is recorded in each element. Based on the recorded path, the and jacked pile are similar, only the impact driven pile is discussed in the
spatial field of the strain amplitude is calculated in step No. 5. In the following. For the initially medium dense soil, in case of the simulation
sixth step, five million load cycles are simulated using the HCA model including the installation the area with large strain amplitudes
(high-cycle mode). (εampl > 10− 3 ) is smaller than in case of the WIP pile. Furthermore, large
The simulation is performed with ideally drained conditions since values of the strain amplitude are observed at the pile tip in case of the
the high-cyclic loading is assumed to be an operational load which does WIP calculation which are considerably smaller for the impact driven
not lead to an accumulation of excess pore water pressure. However, pile. Since larger strain amplitudes result in larger long-term de­
simulations considering pore water pressure build-up and consolidation formations, the WIP pile is expected to show greater permanent de­
processes are generally possible with the HCA model. formations in the initially medium dense case. For the initially dense
The simulations have been performed without the use of update case, the fields of strain amplitude are similar for both the impact driven
cycles and thus a constant field of the strain amplitude is used and the WIP pile. Therefore, the influence of the strain amplitude on the
throughout the calculation phase with high-cyclic loading. In pre­ long-term behaviour is expected to be comparable for both types of piles.
liminary investigations, simulations incorporating update cycles were It is important to note that the information about the strain amplitude
performed to evaluate if the assumption of a constant strain amplitude is alone is not sufficient to conclude which simulation will lead to larger
justified. A simulation incorporating update cycles is presented in Ap­ permanent deformations as, according to Eq. (1), the strain amplitude is
pendix C proving that update cycles do not alter the predicted long-term only one of five influencing factors.
deformations considerably in the present case. Fig. 13 presents the horizontal displacement of the pile head as a
function of the loading cycles for the initially medium dense and dense
4.1. Combined horizontal and moment loading state. The diagrams show the deformation starting with the second cycle
(phase 4–5). The dashed line represents the displacement curve of the
For the case of a combined horizontal and moment loading of the pile WIP pile shifted to the initial displacement of the installed piles to make
an average value of the horizontal force of Hav = 180 kN, an amplitude the differences in the long-term behaviour of both types of piles more
of the horizontal force of Hampl = 180 kN, an average moment of Mav = apparent. For the initially medium dense case the difference between the
400 kNm and an amplitude of the moment of Mampl = 1200 kNm are WIP and the installed piles becomes larger during the high-cyclic
considered (all values are given for the full model). loading and reaches its maximum at the end of the calculation
In Fig. 11 the horizontal displacement of the pile head during the (Δ ≈ 0.1 m). Fig. 14 compares the rate of displacement accumulation (=
application of the mean values of loading (phase 2–3), the first cycle accumulation per cycle) for the WIP and the impact driven pile for the
(phase 3–4) and the second cycle (phase 4–5) is shown. Independent of initially medium dense case. While the difference in the accumulation
the initial density the WIP simulation predicts considerably greater rates tends to increase slightly until 105 cycles it almost vanishes af­
displacements than the simulations considering the installation of the terwards. Hence, the accumulation rate of the WIP pile tends towards
pile. The differences occur mainly during the application of the mean the rate of the installed pile. The influence of the installation tends to
values of loading and during the first quarter of the first cycle. The vanish with increasing number of cycles. This is caused by the change in
second cycle shows a displacement amplitude being comparable to the stress and density during the accumulation phase which tend to the same
one in the first cycle for both installed and WIP piles. Furthermore, the final distribution for both WIP and installed piles as can be seen in
increase of permanent displacement is small in the second cycle. Fig. 15. Here, the spatial distribution of relative density is given after
Compared to the initially dense case, the simulations with initially five million loading cycles for the simulation of an impact driven and a
medium dense sand reveal a larger influence of the installation on the WIP pile. In both cases a large area has densified due to the cyclic
permanent displacement. An influence of the different installation loading. Apart from the soil inside the pile, the installation effects on the
techniques is only slightly present: for the initially medium dense case density distribution are hardly visible at that time. Therefore, the effect
the simulation of the impact driven pile gives a larger permanent of the differences in density caused by the installation vanishes with
displacement and cyclic amplitude compared to the jacked pile. In the increasing number of cycles which leads to the aforementioned
initially dense case the opposite tendency is observed.

Fig. 11. Horizontal displacement of the pile head during the application of the Fig. 12. Field of the strain amplitude for the impact driven and WIP pile for
mean values of loading and the first two cycles for different initial states. different initial densities.

8
P. Staubach et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106223

equalization of accumulation rates of the WIP and the installed piles (the
factor fe of Eq. (1) becomes similar). Analogous to the fields of relative
density, the fields of effective stress will eventually equalize with
increasing number of cycles. In that case also, similar to the factor fe , the
factor fp of Eq. (1) becomes equal for the installed and the WIP piles.
Similar to the initially medium dense case, also the WIP calculation
starting from an initially dense case predicts considerable larger dis­
placements after the second cycle compared to the installed piles.
However, the relative difference is less than in the medium dense case.
During the subsequent high-cyclic loading, the accumulation rate in the
simulation with the WIP pile is significantly lower compared to the
simulations including the installation process as can be seen in Fig. 16.
After approximately 105 loading cycles the displacement curves from
the simulations accounting for the installation pass the respective curves
from the WIP simulation. Therefore, the permanent displacements of the
installed piles after five million loading cycles exceed those of the WIP
piles. This difference is, however, small compared to the differences
observed for the initially medium dense case.
Fig. 13. Horizontal displacement of the pile head as a function of the number Only a small difference in the displacement for the two different
of loading cycles for different initial states. In case of WIP-shifted the
installation techniques is found. This is in accordance with the similar­
displacement of the WIP simulation after 100 cycles was adjusted to the cor­
ities in the fields of effective stress and void ratio described in Section
responding values of the simulations incorporating the installation.
3.4. At very large numbers of cycles the impact driven pile shows slightly
more permanent displacement compared to the jacked pile. These dif­
ferences are more pronounced in the dense case.
Considering the fields of effective stress and relative density given in
Figs. 4 and 5 it is not clear, in particular for the dense case, why the
simulations incorporating the installation process show lower pile de­
formations since the effective stress and the density are decreased
alongside the shaft by the installation process. These unexpected results
seem to be related to a sharp increase in the stiffness of the soil at the tip
of the pile as can be seen from Fig. 17. In that figure, the distribution of
contact stress between soil and pile at the time of the maximum loading
during the first cycle is given. Compared to the WIP calculation the
Fig. 14. Rate of permanent horizontal pile head displacement in dependence of simulation with jacking of the pile shows greater contact stresses near
the number of cycles for the WIP and the impact driven pile starting from an the tip of the pile which decrease rapidly with increasing distance from
initially medium dense state. the tip. In case of the WIP calculation the maximum contact stress is
approximately 3 times smaller but spreads over a larger pile length. The
increase in stiffness in case of the simulations including the installation
process results from the increase in effective stress around the tip of the
pile as depicted in Fig. 17 (and an increased relative density as can be
seen from Fig. 4). In case of the simulations incorporating the installa­
tion and starting from dense conditions the pile initially acts similar to a
clamped beam.
The pile deflection curves after the second cycle (dashed) and after
five million cycles are given in Fig. 18 for the different initial configu­
rations. Independent of the initial density and the number of cycles the
point of rotation (zero horizontal displacement) is lower in case of the
installed piles. This is due to the stiffer soil near the pile tip as explained
above. For initially medium dense conditions the simulation with the
WIP pile predicts considerably larger displacements after five million
loading cycles compared to the installed piles whereas for initially dense
conditions only small differences between the simulations with or

Fig. 15. Field of relative density after five million loading cycles for the impact Fig. 16. Rate of permanent horizontal pile head displacement in dependence of
driven and the WIP pile. the number of cycles for the WIP and the impact driven pile starting from an
initially dense state.

9
P. Staubach et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106223

Fig. 17. Distribution of normal contact stress between soil and pile at the
maximum load during the first loading cycle and distribution of effective hor­ Fig. 19. Vertical displacement of the pile head versus number of loading cycles
izontal stress after the installation by jacking for the initially dense case. for the different initial configurations. In case of WIP-shifted the displacement of
the WIP simulation after 100 cycles was adjusted to the corresponding values of
the simulations incorporating the installation.

dependency on the installation process is reasonable since the installa­

tion leads to an increase in both effective stress and density below the
pile tip. Both leads to a stiffer pile response whereas for the horizontal
loading scenario the effect of the higher stiffness at the pile tip is
somewhat counterbalanced by the decrease in lateral effective stress and
density alongside the outer pile shaft.

5. Summary and conclusions

The impact driving and the jacking of a pile were successfully

simulated under ideally drained and partially drained conditions using a
coupled Eulerian-Lagrangian approach in combination with a hypo­
plastic constitutive model. The predicted change in state variables such
as the effective stress and the relative density during driving were in
good accordance with experimental observations. The comparison of the
Fig. 18. Pile deflection curves after the second cycle (dashed) and after 5
spatial distribution of the effective stress and density revealed only
million loading cycles for different initial conditions. minor differences between jacking and impact driving. For hydraulic
conductivities typical for uniformly graded sands the consideration of
excess pore water pressures and consolidation lead to considerable
without considering the installation are observed.
changes in the fields of effective stress and relative density. A reduced
tendency towards soil plugging and a lower change in relative density in
4.2. Vertical loading the vicinity of the pile compared to the ideally drained simulations were
observed.
In addition to the horizontal loading scenario, the behaviour under a The high-cyclic loading of the pile was simulated using the high-
purely vertical loading is studied. The pile is loaded with an average cycle-accumulation (HCA) model. The results revealed a considerable
value of Vav = 200 kN and an amplitude of the vertical force of Vampl = difference between simulations assuming a wished-in-place (WIP) pile
600 kN (all values are given for a full model). and those incorporating the installation process. Independent of the type
The vertical displacement of the pile for the different initial states is of cyclic loading (horizontal or vertical) the assumption of a WIP state
given in Fig. 19. Again, the plots start with the second cycle in step No. 4. lead to an overestimation of permanent pile deformations during the
Similar to the horizontal loading scenario, the WIP simulations show first about 104 cycles compared to the simulations considering instal­
considerably larger permanent displacements during the first cycles lation effects. Depending on the initial density and the loading scenario,
compared to the simulations incorporating the installation. For the the differences in the permanent deformations between the installed and
vertical loading this difference is even greater than for the horizontal the WIP piles became either larger (all vertical loading scenarios inde­
loading, which was expected as the installation leads to a sharp increase pendent of density and the horizontal loading case at medium initial
in stress and density below the pile tip. Most design methods applied in density) or smaller with increasing number of cycles (horizontal loading
practice assume a fully plugged state open-profile piles and thus do take with initially dense conditions). With increasing number of cycles the
into account this effect. However, it has to be stressed here that the influence of the installation (in terms of the accumulation rate of pile
occurrence of plugging largely depends, among other influencing fac­ head displacement) decreased as the spatial distribution of effective
tors, on the ability of the soil to drain during the installation as was stress and density equalized as the result of the accumulation of defor­
demonstrated in Section 3.5. mation in the soil around the pile. For the horizontally loaded pile the
Independent of the initial density, the influence of the installation influence by the installation was found less in the initially dense state
increases with increasing number of cycles. This is opposite to the hor­ compared to the initially medium dense case. The vertically loaded pile
izontal loading scenario where for the WIP pile in the dense case showed a larger influence of the installation compared to the horizon­
eventually less accumulated deformation was obtained. The increased tally loaded pile due to the formation of a soil plug while driving which

10
P. Staubach et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106223

lead to a considerably stiffer response to the vertical cyclic loading. CRediT authorship contribution statement
Moreover, the increased effective stress and the decreased relative
density around the pile tip caused by soil plugging also affected the pile P. Staubach: Software, Formal analysis, Data curation, Visualiza­
response to the horizontal cyclic loading case. Amongst others the point tion, Validation, Writing - original draft, Conceptualization, Writing -
of rotation was found to be located lower for the installed piles than for review & editing. J. Machaček: Visualization, Writing - original draft,
the WIP pile. Writing - review & editing. M.C. Moscoso: Software, Data curation. T.
In general, the current study suggests that WIP simulations over­ Wichtmann: Writing - review & editing, Conceptualization.
estimate the permanent deformations under cyclic loading significantly
in case of initially medium dense soil. Therefore, current practice
assuming WIP conditions leads to an overestimation of pile displace­ Declaration of competing interest
ment and is conservative if the soil is initially medium dense. However,
the opposite is true in case of initially dense soil. The authors declare that there is no conflict of interest.

Appendix A. Comparison of numerical and semi-analytical results for 1D wave propagation

The following example aims to assess the performance of the proposed fully coupled explicit approach by comparison with an analytical solution
for the wave propagation in a poro-elastic medium (note that a similar study was also performed in Ref. [44], however, without the consideration of
the acceleration of the solid as explained earlier). Consider a one-dimensional column with a length of l = 10 m, as depicted in Fig. A20. The lateral
and bottom boundaries are assumed to be rigid, frictionless and impermeable. At the top, the stress Δσ = 1 kPa and the pore-water pressure pw = 0 kPa
are prescribed (thus, drainage along the top surface is allowed).

Fig. A.20. Schematic illustration of the numerical model of the one-dimensional soil column and important parameters for the simulations.

It is assumed that the time history of the loading is a Heaviside step function. The simulation results are compared to a semi-analytical approach,
which accounts for the relative acceleration of the pore-water phase. A detailed derivation of the semi-analytical solution as well as a comparison to
different finite element formulations are given by the authors in Refs. [36,45]. The parameters used for the numerical as well as for the semi-analytical
approach are listed in Fig. A.20. For the analytical solution and the explicit approach, the time increments are set to Δt = 10− 5 s and Δt = 10− 6 s,
respectively. The soil column was once discretised using 100 elements and once using 1000 elements. For the explicit calculations a time scaling factor,
which can be used to change the maximum stable time increment of an explicit calculation, of 0.1 is used (see Ref. [46]).
The development of excess pore water pressure at the bottom of the column and the settlement-time history recorded at the top of the column are
illustrated in Figure A.21. For both the predicted excess pore-water pressure Δpw as well as the vertical displacements u2 the semi-analytical solution
matches the explicit numerical solution well. The displacement of the top of the column follows a triangular cyclic pattern. With ongoing consolidation
process (drainage along the top surface is allowed), the absolute value of the mean displacement increases, while the amplitude is reduced. The time
needed by the wave to pass through the column increases gradually, as visible in the Δpw -time history in Fig. A.21.

11
P. Staubach et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106223

Fig. A.21. Comparison of numerical and semi-analytical results for the one-dimensional wave propagation problem. Top: Development of excess pore water pressure
(Δpw ) recorded at the bottom of the column. Bottom: Settlement-time history u2 (t) recorded at the top of the column.
Although the accordance with the semi-analytical solution is generally good, strong oscillations of the excess pore water pressure Δpw are observed
in the numerical solution in case of a discretisation with 100 elements. However, these oscillations are absent in case of the displacements u2 . The
oscillations can be noticeably reduced by refining the spatial discretisation, as evident from the results obtained with 1000 elements. The simulations
of the pile installation presented in the previous sections do not show strong oscillations since the reduced bulk modulus of the pore water leads to less
oscillations. Note in addition that implicit elements based on the u-p approximation exhibit similar oscillations (see Ref. [36]).

Appendix B. Field of acceleration during impact driving

Figure B.22 shows the field of acceleration during impact driving of the pile. It can be concluded that the waves generated by this dynamic process
do not reach to the border of the model as they are damped out before. Therefore, the use of absorbing boundary conditions seems unnecessary in the
considered cases.

12
P. Staubach et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106223

Fig. B.22. Acceleration magnitude during impact driving.

Appendix C. Comparison of simulations with and without update cycles

To prove that the assumption of a constant strain amplitude throughout the high-cyclic loading is valid in the present case, an additional FE
calculation was performed. In that simulation two update cycles were incorporated after 103 and 105 cycles. The simulation considered the case of
impact driving, an initially medium dense state and the horizontal loading case. The curve of the horizontal pile head displacements obtained from the
simulation without update cycles are compared in Fig. C.23. The simulation with update cycles shows a slightly lower accumulation rate as a result of
the updated strain amplitude. However, the relative difference to the simulation without update cycles is negligible, thus confirming that omitting the
update cycles in the present study is justified.

Fig. C.23. Horizontal displacement of the pile head as a function of the number of cycles for simulations with either nor or two update cycles.

References [8] Niemunis A, Wichtmann T, Petryna Y, Triantafyllidis T. Stochastic modelling of

settlements due to cyclic loading for soil-structure interaction. In: Proceedings of
session 14 on structural damage and lifetime assessment of the ICOSSAR05
[1] Achmus M, Kuo Y-S, Abdel-Rahman K. Behavior of monopile foundations under
conference in rome; 2005.
cyclic lateral load. Comput Geotech 2009;36(5):725–35. https://doi.org/10.1016/
[9] Wichtmann T. Explicit accumulation model for non-cohesive soils under cyclic
j.compgeo.2008.12.003.
loading publications of the institute of soil mechanics and foundation engineering,
[2] Staubach P., Wichtmann T. Long-term deformations of monopile foundations for
Ruhr-University Bochum. Ph.D. thesis. 2005.
offshore wind turbines studied with a high-cycle accumulation model, Comput
[10] Wichtmann T, Niemunis A, Triantafyllidis T. Strain accumulation in sand due to
Geotech 124. https://doi.org/10.1016/j.compgeo.2020.103553.
cyclic loading: drained triaxial tests. Soil Dynam Earthq Eng 2005:967–79. https://
[3] von Wolffersdorff PA. A hypoplastic relation for granular materials with a
doi.org/10.1016/j.soildyn.2005.02.022.
predefined limit state surface. Mech Cohesive-Frict Mater 1996;1(3):251–71.
[11] Wichtmann T, Niemunis A, Triantafyllidis T. Experimental evidence of a unique
[4] Niemunis A, Herle I. Hypoplastic model for cohesionless soils with elastic strain
flow rule of non-cohesive soils under high-cyclic loading. Acta Geotechnica 2006;1:
range. Mech Cohesive-Frict Mater 1997;2(4):279–99.
59–73. https://doi.org/10.1007/s11440-006-0006-8.
[5] Niemunis A, Wichtmann T, Triantafyllidis T. A high-cycle accumulation model for
[12] Wichtmann T, Niemunis A, Triantafyllidis T. On the influence of the polarization
sand. Comput Geotech 2005:245. https://doi.org/10.1016/j.
and the shape of the strain loop on strain accumulation in sand under high-cycle
compgeo.2005.03.002.
loading. Soil Dynam Earthq Eng 2007;14. https://doi.org/10.1016/j.
[6] Osinov VA, Chrisopoulos S, Grandas-Tavera C. Vibration-induced stress changes in
soildyn.2006.05.002.
saturated soil: a high-cycle problem. In: Holistic simulation of geotechnical
[13] Wichtmann T, Niemunis A, Triantafyllidis T. Strain accumulation in sand due to
installation processes: benchmarks and simulations, 80; 2016. p. 69–84.
cyclic loading: drained cyclic tests with triaxial extension. Soil Dynam Earthq Eng
[7] Machacek J, Triantafyllidis T, Staubach P. Fully coupled simulation of an opencast
2007;27:42–8. https://doi.org/10.1016/j.soildyn.2006.04.001.
mine subjected to earthquake loading. Soil Dynam Earthq Eng 2018;115:853–67.
[14] Wichtmann T, Knittel L. Behaviour of granular soils under uni- and
https://doi.org/10.1016/j.soildyn.2018.09.016.
multidimensional drained high-cyclic loading. In: Triantafyllidis T, editor. Recent

13
P. Staubach et al. Soil Dynamics and Earthquake Engineering 138 (2020) 106223

developments of soil mechanics and geotechnics in theory and practice. Cham: [31] Bakroon M, Daryaei R, Aubram D, Rackwitz F. Multi-material arbitrary
Springer International Publishing; 2020. p. 136–65. Lagrangian-eulerian and coupled eulerian-Lagrangian methods for large
[15] Wichtmann T. Soil behaviour under cyclic loading - experimental observations, deformation geotechnical problems, numerical methods in geotechnical
constitutive description and applications, publications of the Institute of soil engineering IX. In: Proceedings of the 9th European conference on numerical
mechanics and rock mechanics. Habilitation thesis. 2016. p. 181. methods in geotechnical engineering; 2018. p. 673–81. NUMGE 2018.
[16] Solf O. Zum mechanischen Verhalten von zyklisch belasteten Offshore- [32] Phuong N, van Tol A, Elkadi A, Rohe A. Modelling of pile installation using the
Gründungen. Ph.D. thesis. Heft: Veröffentlichungen des Institutes für material point method (mpm). In: Hicks Brinkgreve, Rohe, editors. Numerical
Bodenmechanik und Felsmechanik am Karlsruher Institut für Technologie; 2012. methods in geotechnical engineering; 2014. https://doi.org/10.1201/b17017-50.
p. 176. [33] Galavi V, Martinelli M, Elkadi A, Ghasemi P, Thijssen R. Numerical simulation of
[17] Zachert H. Zur Gebrauchstauglichkeit von Gründungen für Offshore- impact driven offshore monopiles using the material point method. 2019. https://
Windenergieanlagen. Ph.D. thesis. Heft: Veröffentlichungen des Institutes für doi.org/10.32075/17ECSMGE-2019-0758.
Bodenmechanik und Felsmechanik am Karlsruher Institut für Technologie; 2015. [34] Hamann T, Qiu G, Grabe J. Application of a Coupled Eulerian–Lagrangian
p. 180. approach on pile installation problems under partially drained conditions. Comput
[18] T. Wichtmann, J. Machacek, H. Zachert, H. Günther, Validierung eines Geotech 2015;63:279–90. https://doi.org/10.1016/j.compgeo.2014.10.006.
hochzyklischen Akkumulationsmodells anhand von Modellversuchen und [35] Zienkiewicz OC, Shiomi T. Dynamic behaviour of saturated porous media; the
Messungen an realen Bauwerken, Bautechnik 96. doi:10.1002/bate.201800056. generalized biot formulation and its numerical solution. Int J Numer Anal Methods
[19] Zachert H, Wichtmann T, Triantafyllidis T, Hartwig U. Simulation of a full-scale GeoMech 1984;8(1):71–96.
test on a gravity base foundation for offshore wind turbines using a high cycle [36] Staubach P, Machaček J. Influence of relative acceleration in saturated sand:
accumulation model. 2015. p. 819–24. https://doi.org/10.1201/b18442-115. analytical approach and simulation of vibratory pile driving tests. Comput Geotech
[20] Machacek J, Wichtmann T, Zachert H, Triantafyllidis T. Long-term settlements of a 2019;112:173–84. https://doi.org/10.1016/j.compgeo.2019.03.027.
ship lock: measurements vs. FE-prediction using a high cycle accumulation model. [37] Ihc s-900. https://nam03.safelinks.protection.outlook.com/?url=http%3A%2F%
Comput Geotech 2018;97:222–32. https://doi.org/10.1016/j. 2Fgreenglobalgroup3g.com%2Fimages%2Fsampledata%2Fparks%2Flandscape%
compgeo.2018.01.009. 2FIHCHydrohammerOffshore.pdf&data=02%7C01%7Cs.natarajan.2%40elsevier.
[21] Dafalias YF, Manzari MT. Simple plasticity sand model accounting for fabric com%7Cb05a68bc6cb541e5569f08d81cb5b96d%7C9274ee3f94254109a27f9fb
change effects. J Eng Mech 2004;130(6):622–34. 15c10675d%7C0%7C0%7C637290914535756262&sdata=w%2Ba%2B2
[22] Liu HY, Abell JA, Diambra A, Pisanò F. Modelling the cyclic ratcheting of sands czpyy6FHLacWeDXidDMCv%2Fc658XEnT9bYvyu2k%3D&reserved=0.
through memory-enhanced bounding surface plasticity. Geotechnique 2019;69(9): [38] De Nicola A, Randolph MF. The plugging behaviour of driven and jacked piles in
783–800. https://doi.org/10.1680/jgeot.17.P.307. sand. Geotechnique 1997;47(4):841–56. https://doi.org/10.1680/
[23] Houlsby G, Abadie C, Beuckelaers W, Byrne B. A model for nonlinear hysteretic and geot.1997.47.4.841.
ratcheting behaviour. Int J Solid Struct 2017;120:67–80. https://doi.org/10.1016/ [39] Bauer E. Calibration of a comprehensive constitutive equation for granular
j.ijsolstr.2017.04.031. materials. Soils Found 1996;36:13–26.
[24] Benson DJ. Computational methods in Lagrangian and Eulerian hydrocodes. [40] White DJ, Bolton MD. Displacement and strain paths during plane-strain model
Comput Methods Appl Mech Eng 1992;99(2):235–394. pile installation in sand. Geotechnique 2004;54(6):375–97. https://doi.org/
[25] Benson DJ, Okazawa S. Contact in a multi-material Eulerian finite element 10.1680/geot.2004.54.6.375.
formulation. Comput Methods Appl Mech Eng 2004;193(39):4277–98. the [41] Yang Z, Jardine R, Zhu B, Tsuha C, Foray P. Sand grain crushing and interface
Arbitrary Lagrangian-Eulerian Formulation. shearing during displacement pile installation in sand. Geotechnique 2010;60(6):
[26] Waltz J. Operator splitting and time accuracy in Lagrange plus remap solution 469–82. https://doi.org/10.1680/geot.2010.60.6.469.
methods. J Comput Phys 2013;253:247–58. https://doi.org/10.1016/j. [42] Gourvenec S, Randolph M. Offshore. Geotech. Eng. 2011. https://doi.org/
jcp.2013.07.016. 10.1201/9781315272474.
[27] Qiu G, Henke S, Grabe J. Application of a coupled eulerian-Lagrangian approach [43] Randolph MF, Leong EC, Houlsby GT. One-dimensional analysis of soil plugs in
on geomechanical problems involving large deformation. Comput Geotech 2011; pipe piles. Geotechnique 1991;41(4):587–98. https://doi.org/10.1680/
38:30–9. geot.1991.41.4.587.
[28] Andresen L, Khoa HX. Ldfe analysis of installation effects for offshore anchors and [44] Hamann T. Zur Modellierung wassergesättigter Böden unter dynamischer
foundations. In: Hicks, editor. Installation effects in geotechnical engineering; Belastung und großen Bodenverformungen am Beispiel der Pfahleinbringung,
2013. p. 162–8. Veröffentlichungen des Instituts für Geotechnik und Baubetrieb der Technischen
[29] Wang D, Bienen B, Nazem M, Tian Y, Zheng J, Pucker T, et al. Large deformation Universität Hamburg. Ph.D. thesis. 2015.
finite element analyses in geotechnical engineering. Comput Geotech 2015;65: [45] Machaček J. Contributions to the numerical modeling of saturated and unsaturated
104–14. https://doi.org/10.1016/j.compgeo.2014.12.005. soils, veröffentlichung des instituts für bodenmechanik und felsmechanik am
[30] Heins E, Grabe J. Class-a-prediction of lateral pile deformation with respect to karlsruher institut für technologie (KIT). Ph.D. thesis. 2019.
vibratory and impact pile driving. Comput Geotech 2017;86:108–19. https://doi. [46] Corp DSS. Abaqus 2017. http://www.3ds.com/products-services/simulia/abaqus;
org/10.1016/j.compgeo.2017.01.007. 2017.

14