Computers and Geotechnics 122 (2020) 103552

Contents lists available at ScienceDirect

Computers and Geotechnics

journal homepage: www.elsevier.com/locate/compgeo

Research Paper

Predicting the cyclic behaviour of suction anchors based on a stiffness T

degradation model for soft clays
Xinglei Chenga,b, Piguang Wangb, , Na Lia, Zhongxian Liua, Yadong Zhoua
⁎

a
Key Laboratory of Soft Soil Engineering Character and Engineering Environment of Tianjin, Tianjin Chengjian University, Tianjin 300384, China
b
Beijing University of Technology, Beijing 100124, China

ARTICLE INFO ABSTRACT

Keywords: The behaviour of suction anchors subjected to combined average and cyclic loads in soft clays is an essential
Stiffness degradation consideration in their design. A new numerical computation method was developed based on a stiffness de-
Soft clays gradation model for soft clay that was proposed by performing a series of undrained cyclic triaxial tests and
Suction anchors embedded in the ABAQUS software package by encoding the USDFLD subroutine and defining relevant field
Cyclic loads
variables. The numerical method predicts the cyclic behaviour of suction anchors by coupling the elastic-per-
Numerical simulation
fectly plastic model with the Mohr-Coulomb yield criterion and the proposed stiffness degradation model that
can reflect the stiffness degradation and the accumulation of plastic deformation of soils around the anchor
during cyclic loading. The numerical method was verified by a comparison with the model test results of suction
anchors subjected to combined average and cyclic loads in soft clays. The proposed method can predict the cyclic
deformation as well as the bearing capacity and capture the nonlinearity, hysteresis and cyclic accumulation
characteristics of the load-displacement responses of suction anchors in soft clays.

1. Introduction Sukumaran et al. [5] determined the capacity of a suction anchor

subjected to lateral loads in soft clays under undrained conditions using
Suction anchors have been widely used in taut mooring systems for a linear elastic model with the Von Mises strength criterion. Cao et al.
floating facilities, such as floating production storage and offloading [6] simulated the behaviour of suction anchors subjected to vertical
(FPSO) units. In many cases, the suction anchor needs to be installed in loading in normally consolidated clays based on the modified Cam-Clay
soft clay foundations, which are subjected to average loads caused by model of porous soil materials. A computational procedure was pro-
the floatage of the superstructure as well as cyclic loads induced from posed by Maniar [7] to simulate the installation process and the load-
various environmental factors such as winds, waves and flows in the displacement response of anchors subjected to axial and inclined loads,
marine environment [1–4]. The strength and stiffness of the soil around and a bounding surface model was used to describe the nonlinear be-
the suction anchor will be degraded under long-term cyclic loading, haviour of the clayey soil. Monajemi and Razak [8] investigated the
which will induce the cumulative displacement of the anchor along the failure mechanism and ultimate capacity of an anchor under combined
mooring direction. The failure of suction anchors occurs when the cu- V-H-M loading by conducting nonlinear analysis using a simplified
mulative displacement reaches a certain failure standard. Therefore, elastic-perfectly plastic model for saturated clays. Ahn et al. [9] esti-
proposing an appropriate method to analyse the cyclic deformation mated the holding capacities of optimally loaded suction caisson an-
process and evaluate the cumulative displacement is of great sig- chors embedded in cohesive soils with a linear strength distribution by
nificance for the design and safe service of suction anchor foundations the total stress analysis based on the Von Mises model of soil. Kim et al.
in soft clays. [10] investigated the performance of suction anchor groups subjected
The finite element method has been widely used for the study of to horizontal pull-out loading by performing centrifuge model tests and
suction anchor behaviours in soils. Some studies have also been re- numerical simulations, and the soil was modelled using the Mohr-
ported on the finite element analysis of suction anchors under static or Coulomb failure criterion with the non-associated flow rule. Wang and
cyclic loads. Many attentions have been focused on the simulation of Guo et al. proposed a novel two-dimensional static model for a mooring
the behaviours of suction anchors under static loads. The selection of a cable during pretensioning and a three-dimensional quasi-static model
soil constitutive model is very important for finite element analysis. for a cable in service [11]. In addition, Guo et al. performed model tests

⁎
Corresponding author.
E-mail address: [email protected] (P. Wang).

https://doi.org/10.1016/j.compgeo.2020.103552
Received 3 November 2019; Received in revised form 14 March 2020; Accepted 15 March 2020
Available online 20 March 2020
0266-352X/ © 2020 Elsevier Ltd. All rights reserved.
X. Cheng, et al. Computers and Geotechnics 122 (2020) 103552

with different loading angles and cyclic parameters to investigate the

failure mode and capacity of suction caissons under inclined short-term
static and one-way cyclic loadings [12].
A series of model tests were performed by Chen and Randolph [1]
and Zhu et al. [13] to investigate the responses of suction anchors
subjected to cyclic loading. However, research on numerical simula-
tions of the cyclic behaviours of suction anchors is limited. The main
difficulty is the lack of constitutive models that can not only reasonably
describe the nonlinear cyclic stress-strain responses of soft clays but
also be easily implemented in 3D nonlinear finite element analysis
codes. These models, such as the elastic-perfectly plastic model, mod-
ified Cam-Clay model, and Mohr-Coulomb model encoded into com-
mercial finite element software, are incapable of calculating the cyclic
responses of soil. Therefore, researchers have performed numerical si-
mulations of the cyclic behaviours of suction anchors using advanced
constitutive models. Gelagoti et al. [14] proposed a simplified equiva- Fig. 1. Definition of the secant shear modulus and cyclic degradation.
lent linear iterative approach to account for the nonlinear lateral stiff-
ness and bearing capacity of suction caissons by a simplified kinematic characteristic of clay is described by the decay of the secant shear
hardening model in the context of Von Mises associative plasticity. modulus with the increase in the number of cycles. As shown in Fig. 1,
Kourkoulis et al. [15] simulated the behaviour of suction caissons the secant shear modulus of the unloading curve is defined as the slope
subjected to wave and earthquake loading using a simple kinematic of a straight line connecting the two ends of the unloading curve, the
hardening model. Zhang et al. [16] predicted the cyclic behaviour of longitudinal coordinate is the axial deviatoric stress represented as R ,
suction caissons using a finite element method based on a thermo- and the transverse coordinate is the axial strain represented as y . The
dynamics-based constitutive model. Cheng et al. [17,18] simulated the secant shear modulus is represented by Eq. (1).
cyclic behaviours of suction anchors by developing an elastoplastic
N N
bounding surface model and encoding it into ABAQUS software. R,max R,min
GSN =
Although the advanced constitutive model with different kinematic N
y,max
N
y,min (1)
hardening rules can describe the cyclic responses of soil to some extent,
the complexity of the model leads to inefficient numerical calculation, where N represents the number of stress cycles, is the secant shear
GSN
modulus for the Nth cycle, RN,max and RN,min are the maximum and
especially for long-term cyclic loads. The cyclic loading can lead to the
degradation of stiffness for soft clays, and the stress-strain curve shows minimum deviatoric stresses for the Nth cycle, respectively, and yN,max
significant nonlinear and hysteresis properties. In this paper, the stiff- and yN,min are the corresponding axial strain, as shown in Fig. 1.
ness degradation model for soft clays was proposed by performing a Referring to the definition of the degradation index by Idriss et al.
series of cyclic triaxial tests, and then the model was embedded in the [19] based on the results of cyclic triaxial tests under the strain-con-
ABAQUS software package by encoding the USDFLD subroutine and trolled mode, the expression for the cyclic degradation index under a
defining relevant field variables. The behaviours of suction anchors constant amplitude stress-controlled mode is defined as follows:
subjected to combined average and cyclic loads in soft clays were si- N
R,max
N
R,min
mulated by coupling the elastic-perfectly plastic model with the Mohr- GN N
y,max
N
y,min
1
y,max
1
y,min
Coulomb yield criterion and the proposed stiffness degradation model. = S1 = 1 1
= N N
GS R,max R,min y,max y,min
This numerical method has higher computational efficiency than pre- 1
y,max
1
y,min (2)
vious methods using advanced constitutive models.
where is the cyclic degradation index. RN,max R,min is equal to
N

2. Stiffness degradation model of soft clays

1
R,max
1
R,min for constant amplitude stress-controlled cyclic tests.
The secant shear modulus for each cycle number can be determined
Many studies have revealed that the stiffness degrades and the by the tested stress-strain curve according to the definition of Eq. (1),
strength decreases for soft clay under cyclic loading [19–22]. The de- and then the cyclic degradation index for each cycle number can be
gree of degradation is related to the number of cycles, the level of initial determined based on Eq. (2). The functional relationship between the
static stress and the level of cyclic stress, and it is affected by many cyclic degradation index and the cycle number N can be determined.
factors, such as the overconsolidation ratio, principal stress direction,
strain rate and vibration frequency [23–27]. The results of soil dynamic 2.2. Cyclic degradation model
tests show that the cyclic degradation of soft clays means that the shear
modulus of the stress-strain hysteresis curve decreases gradually; Idriss et al. [19] defined the concept of the degradation index and
meanwhile, the cyclic deformation increases gradually as the number of proposed the exponential relationship of the degradation index versus
cycles increases. The plastic strain of soil accumulates when there is an the cycle number based on the results of cyclic triaxial tests the under
initial static deviate stress. strain-controlled mode but did not give the definition of the degrada-
tion index under the stress-controlled mode or consider the effect of the
2.1. Cyclic degradation index initial deviatoric stress on the exponential relationship. In this paper,
the relationship between the degradation index and the number of
The test results for soft clay subjected to combined static and cyclic cycles for soft clays subjected to the combination of initial static de-
deviatoric stresses show that the soil stiffness degradation is not sig- viatoric stress and cyclic deviatoric stress is studied. The undrained
nificant with no cyclic stress reversal; otherwise, significant stiffness cyclic triaxial tests on soft clay under the joint actions of different static
degradation will occur. The greater the cyclic stress is relative to the and cyclic deviatoric stresses are performed under the stress-controlled
static deviatoric stress, the more obvious the stress reversal is, and the mode. In these tests, an axial static deviatoric stress was first applied to
more significant the stiffness degradation is [20,21]. In this paper, the the soil sample, and then a 0.1 Hz sinusoidal constant-amplitude cyclic
secant shear modulus of the unloading curve is used to describe the deviatoric stress was exerted under undrained conditions. The stress
stiffness of soft clay under cyclic loads, and the stiffness degradation level of the soil elements in the general stress states can be represented

2
X. Cheng, et al. Computers and Geotechnics 122 (2020) 103552

1.0 τ8,cy / τ8, f =0.781

0.9
τ8,cy / τ8, f =0.667
0.8
τ8,cy / τ8, f =0.483
0.7
τ8,cy / τ8, f =0.322
0.6
0.5

ζ
0.4
0.3
0.2
0.1 τ8,a / τ8, f =0
0.0
0 50 100 150 200
Fig. 2. Typical stress-strain curve of soil under the joint actions of static and
N
cyclic deviatoric stress.
1.0
Table 1 0.9
Different combinations of static and cyclic stress levels for the cyclic triaxial 0.8
tests.
0.7
8, a 8, f 8, cy 8, f 8, a 8, f 8, cy 8, f 8, a 8, f 8, cy 8, f 0.6
0 0.781 0.3 0.549 0.5 0.494 0.5

ζ
0 0.667 0.3 0.412 0.5 0.423 0.4 τ8,cy / τ8, f =0.549
0 0.483 0.3 0.323 0.5 0.358 τ8,cy / τ8, f =0.412
0 0.322 0.3 0.263 0.5 0.310
0.3
0.2 τ8,cy / τ8, f =0.323
0.1 τ8,a / τ8, f =0.3 τ8,cy / τ8, f =0.263
using the octahedral shear stress 8 . Here, the octahedral static shear 0.0
stress 8, a is defined as the octahedral shear stress of the soil element 0 50 100 150 200 250
after static loading. The octahedral cyclic shear stress 8, cy is defined as N
one half of the difference between the maximum and minimum octa-
hedral shear stress of the soil element under cyclic loading, as shown in 1.0
Fig. 2. In addition, the octahedral peak shear stress 8, f is defined as the 0.9
octahedral shear stress when the soil sample reaches the deformation 0.8
failure criteria of 10% axial strain. The normalized octahedral static 0.7
and cyclic shear stresses were defined as 8, a 8, f and 8, cy 8, f , respec-
0.6
tively, which represent the static deviatoric stress level and cyclic stress
level that were applied to the sample during the experiment. For each
0.5
ζ

0.4 τ8,cy / τ8, f =0.494

8, a 8, f , different 8, cy 8, f values were selected for the tests, as shown in
Table 1. 0.3 τ8,cy / τ8, f =0.423
According to the definition of the degradation index in Section 2.1 0.2 τ8,cy / τ8, f =0.358
above, the relationships between the degradation index and the number 0.1 τ8,a / τ8, f =0.5 τ8,cy / τ8, f =0.310
of cycles under different stress levels can be determined based on cyclic 0.0
triaxial test results, as shown by the discrete points in Fig. 3. It is found 0 10 20 30 40 50 60 70
that Eq. (3) proposed by Idriss et al. [19] can be used to express the N
relationship between the degradation index and the number of cycles.
Fig. 3. Cyclic degradation index ς versus cycle number N.
=N t
(3)

where t is the cyclic degradation parameter and reflects the rate of 0.35
cyclic degradation as the cycle number increases. The value of t is
0.30
closely related to the static and cyclic stress levels of soils.
To reflect the change rule of the degradation parameter t with the 0.25
stress level, the cyclic stress parameter S is defined as Eq. (4) or (5).
0.20
8,cy (2 8, f 8,a )
t

S= 0.15
2 8,f
2
(4)
τ8,a / τ8, f =0
Or 0.10
τ8,a / τ8, f =0.3
8,cy
0.05 τ8,a / τ8, f =0.5
8,a
S= 1 0.5
8,f 8,f (5) 0.00
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
t is associated with different combinations of static and cyclic deviatoric S
stresses and can be determined by fitting the test results in Fig. 3 using
Eq. (3), and then t against different values of S can be plotted as the Fig. 4. Cyclic degradation parameter t versus cyclic stress parameter S.
discrete points shown in Fig. 4. It can be seen that t increases mono-
tonically with S under different 8, a 8, f , and the two basically follow a
linear relationship, which can be expressed in Eq. (6) as follows.

3
X. Cheng, et al. Computers and Geotechnics 122 (2020) 103552

t = 0.3353S (6) Su/kPa

In Eq. (4) or Eq. (5), 8, a 8, f 1, 8, cy 8, f 1, and 0 3 6 9 12 15
0
( 8, a + 8, cy ) 8, f 1. Hence, it can be considered that 0 S 1. S = 0
when 8, cy 8, f = 0 , and S = 1 when 8, cy 8, f = 0 and 8, cy 8, f = 1. Eq. (4)
10
or Eq. (5) also show that for a given 8, a , the larger 8, cy is, the larger S is, S1
and then the larger t correspondingly is through Eq. (6), which in- 20 S2
dicates that when the static deviatoric stress is constant, the larger the S3
cyclic deviatoric stress is, the more significant the cyclic degradation S4
30

Depth/cm
S5
will be. In addition, the equations also show that for a given 8, cy , the S6
smaller 8, a is, the larger S is, and then the larger t is, which means that 40
when cyclic deviatoric stress is constant, the smaller the static devia-
toric stress is, the more significant the cyclic degradation is. It can also 50
be seen from the test results in Fig. 3 that the cyclic degradation is the
60
most significant when 8, a 8, f is 0. The cyclic degradation gradually
becomes insignificant with increasing 8, a 8, f . Based on the above Eqs.
70
(3)–(6), the degradation index can be expressed as a function of the
static and cyclic deviatoric stress levels and the number of cycles, as
80
shown in Eq. (7).
Fig. 5. Shear strength of the soft clay along the depth.
=N
0.3353 (
8,cy
8,f
1 0.5
8,a
8,f ) (7)
Several LVDT displacement sensors were used to measure the dis-
The curves of the cyclic degradation index versus the number of
placement of the anchor along multiple directions, and the load sensor
cycles can be plotted based on Eq. (7) after determining the values of
was used to measure the force along the loading direction.
8, a 8, f and 8, cy 8, f , as shown by the solid line in Fig. 3. It can be seen
that the solid line generally agrees with the experimental data, which
indicates that Eq. (7) can predict the change in the degradation index 3.2. Model test steps
with the number of cycles. The secant shear modulus for the Nth cycle
GSN can be determined by Eq. (8) which is deduced from Eq. (2) after A static tensional load (average load) with a magnitude of Fa was
obtaining the value of . In this research, to easily apply the degrada- first applied to the anchor using the grading loading mode. When the
tion relationship to the numerical calculation, the elastic modulus of displacement of the anchor was relatively stable under the static load,
the soil is used to represent the stiffness of the soil, and the degradation sinusoidal cyclic loads with a peak value of Fcy and a frequency of
relationship for the secant shear modulus is used to represent the de- 0.1 Hz were applied to the anchor in the load control mode. The static
gradation of the elastic modulus of the soil. Therefore, we can obtain a and cyclic loads resulted in static and cyclic cumulative displacements
more applicable degradation relationship, as shown by Eq. (9). of the anchor. It is considered that the failure of the anchor occurred
when the sum of the static and cyclic cumulative displacements along
GSN = GS1 (8)
the loading direction at the loading point reached the failure criterion
of 0.6 times the anchor plate width (static loading and cyclic loading
EN = E1 (9)
follow the same displacement failure criterion; therefore, the failure
where N represents the number of stress cycles, EN is the elastic mod- criterion is determined by first performing static monotonic tests, and
ulus for the Nth cycle, and E1 is the elastic modulus for the first cycle or the determining method can be seen in reference [18]). Four groups of
initial elastic modulus. model tests were carried out. The test programs are shown in Table 2.
The number of load cycles corresponding to the failure of the anchors is
3. Model tests on suction anchors subjected to cyclic loads in soft defined as the number of load cycles to reach failure, denoted as Nf . The
clays cyclic bearing capacity corresponding to Nf is determined based on
Fa + Fcy . Ff is the ultimate bearing capacity of the anchor under
3.1. Model test apparatus monotonic static loading. Fa Ff is the normalised average load that
specifies the magnitude of the average load. Fcy Ff is the normalised
cyclic load that specifies the magnitude of the cyclic load. (Fa + Fcy ) Ff
In the following section, numerical simulations for the suction an-
is the normalised cyclic bearing capacity that specifies the magnitude of
chor model tests will be performed based on the above stiffness de-
the cyclic bearing capacity. The test results for the cyclic responses of
gradation model of soft clays. Here, the model tests of the suction an-
the suction anchors will be compared with the numerical simulation
chor are first introduced. Model tests were conducted in a test tank with
results in the following sections.
a length of 1.5 m, width of 1 m and height of 1.2 m. The soft clays for
the model tests are prepared using the vacuum preloading method and
clay slurry collected from Bohai Bay Beach of Tianjin, China. The unit 4. Numerical simulation for cyclic degradation behaviours of
weight, plastic limit, liquid limit, plastic index and sensitivity of the suction anchors in soft clays
clay are 17.5 kN/m3, 27.01, 44.44, 17.43 and 4.0, respectively. The
vane shear strength of the soft clay Su is basically uniform along the 4.1. Geometries and meshes
depth of the stratum with an average value of 6 kPa, as shown in Fig. 5.
The model anchor was made of stainless steel with a diameter of The 3D finite element model was established using the ABAQUS
0.152 m, a height of 0.456 m and a wall thickness of 0.002 m. The software package to simulate the above model tests on suction anchors.
loading system used in the model tests consists of a loading frame, an Considering the symmetries of the geometries and loading conditions
oriented plate with pulleys, and a multifunctional electric servo control during the tests, only half of the foundation was meshed to improve the
loading device, as shown in Fig. 6. Cyclic loading tests on the suction calculation efficiency, as shown in Fig. 7. The finite element results
anchor were carried out using the loading system in the load control dependent on the mesh density and the incremental step size. It is
mode. The upper pulley in the oriented plate was fixed, and the loading shown that more accurate calculation results can be obtained with
direction was changed by regulating the location of the pulley below. higher mesh density and smaller incremental step size, while

4
X. Cheng, et al. Computers and Geotechnics 122 (2020) 103552

LVDT 1 Loading actuator

Upper leading pulley Displacement sensor
Loading
actuator
Guide plate
Loading
frame
Load sensor
LVDT 2 LVDT 3

Guide plate Below leading pulley

Load sensor
Suction
anchor
Displacement gauge

Suction anchor
Below leading pulley

(a) Real loading system (b) Schematic diagram of the loading system
Fig. 6. Loading system of the model tests.

calculation efficiency is lower. After many trial calculations, we found degradation index of 1. As the degradation index decreases, the mod-
that the displacement of anchors increases as the mesh density increases ulus degrades more significantly, the value of the field variables in-
and the incremental step size decreases, and the calculation results will creases, and the elastic modulus decreases gradually. The degradation
have an acceptable accuracy when the number of elements is more than index is divided into several intervals. The increment of each interval is
10,000 and the incremental step size is not greater than 0.01 (100 in- 0.1. The corresponding elastic modulus is the product of the initial
cremental steps per loading cycle). Hence, the anchors and the soils elastic modulus and the average value of the field variable for each
were all simulated using 14,320 8-node linear brick elements and the interval. The function of the field variable is to establish the corre-
incremental step size is set as 0.01 in this research. Normal horizontal sponding relationship between the elastic modulus and degradation
constraints were applied to the vertical boundaries, and fixed con- coefficient. In addition, the degradation index is closely related to the
straints were applied to the bottom boundary. The top boundary was initial static stress level, the cyclic stress level and the cyclic number.
fully free. The anchor was treated as a rigid body during the calcula- The main function of the USDFLD subroutine is to determine the cyclic
tions. The model test results showed that the soil plug within the anchor degradation index according to Eq. (7) based on the static and cyclic
remained in close contact with the inner wall of the anchor during stress levels and the cyclic number, and then the corresponding de-
loading due to a large negative pressure; therefore, the tie-contact graded elastic modulus can be determined.
condition was set between the soil plug and the inner wall. Besides, The main steps of simulating the cyclic behaviours of suction an-
relative displacements and the separation could occur between the chors in soft clays based on the stiffness degradation model are as fol-
outer wall of the anchor and the stratum, so the contact conditions of lows:
the tangential slip and normal separation were set between the outer
wall and the stratum. The tangential slip conditions were set according (1) The first step is the geostatic analysis step. The purpose of this step
to the Coulomb friction model with a friction coefficient of 0.26. is to impose self-weight loads on the soil so that the soil domain has
an initial geostress field; meanwhile, the initial displacement field is
4.2. Numerical implementation of the stiffness degradation model virtually zero.
(2) The second step is to apply a static tensile load (average load) to the
The elastic-perfectly plastic model with the Mohr-Coulomb yield suction anchor. At the end of this step, six stress components of any
criterion was used to analyse the deformation of the soil. The cohesion soil element in the whole soil domain are output to calculate the
force was set as 6 kPa. The Poisson’s ratio was = 0.49 and the friction corresponding octahedral shear stress of the soil element, which is
and dilation angles of = = 0 were set to simulate the undrained defined as the initial octahedral static shear stress 8, a , and then the
condition of soft clays (the Mohr-Coulomb yield criterion reduces to the normalized octahedral static shear stress 8, a 8, f can be obtained for
Tresca yield criterion in the case of = 0 ). However, the model cannot any soil element in the whole soil domain at the end of static
describe the stiffness degradation and cumulative deformation char- loading.
acteristics of soils under cyclic loads. Therefore, the cyclic stress-strain (3) The third step is to apply a tension sinusoidal cyclic load to the
responses of soil were described by combining the above stiffness de- suction anchor. The numerical calculation in this step is performed
gradation model with the ideal elastic-plastic model. by calling the USDFLD subroutine. Six stress components of the soil
The stiffness degradation model was implemented in ABAQUS by element at the end of any time increment step in each cycle are
encoding the USDFLD subroutine. The elastic modulus and the corre- output, and the corresponding octahedral stress is calculated by
sponding field variables and cyclic degradation index are shown in calling the subroutine. The peak value of the octahedral stress 8,max
Table 3. The elastic modulus varies with the field variables that are in each cycle is tracked, judged and then output. The octahedral
determined by the cyclic degradation index . The initial maximum cyclic shear stress 8, cy can be determined by the formula of
elastic modulus E1 is 1.8 MPa, corresponding to a field variable of 0 and 8,max 8,a , and then the normalized octahedral cyclic shear stress

Table 2
Programs for the constant amplitude cyclic loading tests on suction anchors.
Test number Ff/kN Fa/Ff Fcy/Ff (Fa + Fcy)/Ff Nf (test) Nf (simulation) (Nf Nf ) Nf \%

T1 2.10 0.5 0.45 0.95 52 47 9.6

T2 2.00 0.5 0.38 0.88 195 218 11.8
T3 1.95 0.5 0.33 0.83 268 286 6.7
T4 2.07 0.5 0.24 0.74 No failure No failure \

5
X. Cheng, et al. Computers and Geotechnics 122 (2020) 103552

0.152m

0.456m
Load
0.8m

0.456m

1m

Fig. 7. 3D finite element model of the suction anchor in soft clays.

Table 3 degradation responses of soils can be calculated by combining the

Elastic modulus and field variables. ideal elastic-plastic soil model with the stiffness degradation model.
Elastic modulus (MPa) Field variables Degradation index
The main calculation process is shown in Fig. 9.

1.80 0 1
1.71 1 0.9 < ≤ 1 5. Comparisons of the predictions and model test results
1.53 2 0.8 < ≤ 0.9
1.35 3 0.7 < ≤ 0.8
5.1. Cyclic degradation behaviours
1.17 4 0.6 < ≤ 0.7
0.99 5 0.5 < ≤ 0.6
0.81 6
Fig. 10 shows the contours of the cyclic degradation index as the
0.4 < ≤ 0.5
0.63 7 0.3 < ≤ 0.4
cycle number increases (N = 5, 15, 30, 45). It can be seen that the
0.45 8 0.2 < ≤ 0.3 range of degradation in the passive soil domain of the anchor wall and
0.27 9 0.1 < ≤ 0.2 the soil domain of the anchor bottom expands and the degradation
0.09 10 0.0 < ≤ 0.1 degree increases as the number of cycles increases. The degradation
degree for the same cycle number is different in different soil domains
around the anchor. The degradation indexes are different for each soil
8000 element in the whole soil domain around the anchor due to different
Element K
7000 Element J values of 8, a 8, f and 8, cy 8, f for the same cycle number N; meanwhile,
Element E
6000 the degraded elastic modulus is also different. Therefore, the degraded
elastic modulus for each soil element corresponding to different cycle
5000
numbers is determined according to the static and cyclic stress levels of
4000 each element. Degradation mainly occurs in the passive soil domain of
τ8

3000 the anchor wall and the soil domain at the anchor bottom. The de-
2000 gradation in the active soil domain of the anchor wall is relatively in-
significant. The most significant degradation positions are the soil do-
1000
mains near the lower part of the anchor wall and near 1 times the
0 anchor diameter from the anchor bottom.
0 1 2 3 4 5 6 7 8 9 10 Fig. 11 shows the degradation index versus the cycle number N for
Time/s several typical soil elements around the anchor. The selected several
types of soil elements are shown in Fig. 11(a). Fig. 11(b) shows the
Fig. 8. Octahedral stress-time history curves for three typical soil elements
degradation rules of Element A in the active soil domain and Element E
(elements can be seen in Fig. 11(a)).
in the passive soil domain. It can be seen that the two soil elements
experience a certain stiffness degradation as the cycle number in-
8, cy 8, f can be determined for any soil element in the whole soil creases; however, the degradation of Element E is more significant than
domain during cyclic loading. The magnitude of the octahedral that of Element A. This is because the soil elements in the passive soil
stress varies periodically with time, as shown in Fig. 8 (indicated by domain are constantly and forcefully squeezed by the anchor wall
only 10 cycles), and the period of the stress cycle is virtually the during cyclic loading, while the soil elements in the active soil domain
same as the period of the sinusoidal cyclic loads, so the number of only undergo slight lateral unloading. Fig. 11(c) shows the degradation
stress cycles N can be determined according to the number of si- rules of Elements B ~ H along the stratum depth in the passive soil
nusoidal loads applied to anchors. The modulus degradation index domain. The stiffness degradation of soft clays becomes more obvious
corresponding to N cycles can be determined after obtaining the with increasing depth along the anchor wall. The elastic modulus near
values of 8, a 8, f , 8, cy 8, f and N, and then the corresponding de- the anchor top degrades to approximately 90% of the initial maximum
graded elastic modulus can be determined. Therefore, the cyclic elastic modulus, while the elastic modulus near the anchor bottom

6
X. Cheng, et al. Computers and Geotechnics 122 (2020) 103552

Geostatic analysis Define field variables

Static monotonic tension load analysis Output τ 8,cy / τ8,f and N

Output the value of τ 8,a / τ8,f for each soil Calculate ζ

element in the whole soil domain at the end of
static loading
Calculate EN

Perform cyclic load analysis by calling the

USDFLD subroutine based on the stiffness Calculate soil deformation using EN
degradation model based on the ideal elastic-plastic model

Fig. 9. Diagram of the calculation program.

0.8m

0.8m

1.0m 1.0m
(a) N=5 (b) N=15
0.8m

0.8m

1.0m 1.0m
(c) N=30 (d) N=45

0.00 0.10 0.21 0.31 0.41 0.52 0.62 0.73 0.83 0.93 1.00
Fig. 10. Contours of the cyclic degradation index with increasing cycle number.

7
X. Cheng, et al. Computers and Geotechnics 122 (2020) 103552

Fig. 11. Modulus degradation rules for different soil elements around the anchor as the number of cycles increases.

degrades to approximately 35% of the initial elastic modulus. Fig. 11(d) 5.2. Cyclic deformation process
shows the degradation rules of Elements E ~ L along the horizontal
direction in the passive soil domain. The soil element with the most The typical test result and prediction for the anchor at failure are
significant degradation is not Element E nearest to the anchor wall but shown in Fig. 12(a) and (b), respectively. The translational motion of
Element J. The degradation degree of Element L farthest from the an- the anchor occurs, and the vertical cumulative displacement is greater
chor wall is also relatively small. Therefore, the degradation degree of than the horizontal cumulative displacement in Fig. 12(b). The se-
the soil element along the horizontal direction is not inversely pro- paration occurs between the outer wall of the anchor and the soils in
portional to the distance from the anchor wall but is most significant at the active regions; meanwhile, the soils are compressed and uplifted in
a certain distance from the anchor wall, and the distance should be the passive regions. These predictions are consistent with the test re-
related to the direction of the loading and movement of the anchor. sults.

Heave
Gap

Heave
Gap

(a) Model test (b) Numerical simulation

Fig. 12. Comparison between the numerical simulation and model test results at anchor failure.

8
X. Cheng, et al. Computers and Geotechnics 122 (2020) 103552

100 100
Predicted result Fcy F f =0.45 Fcy / Ff =0.38
90 90 Predicted result
Test result Test result
80 80
Displacement/mm

Displacement/mm
70 70
60 60
50 50
40 40
30 30
20 20
10 10
0 0
0 100 200 300 400 500 600 0 500 1000 1500 2000 2500
Time/s Time/s
100 50
Predicted result Fcy F f =0.33 Predicted result Fcy Ff =0.24
90 45 Test result
Test result
80 40

Displacement/mm
Displacement/mm

70 35
60 30
50 25
40 20
30 15
20 10
10 5
0 0
0 500 1000 1500 2000 2500 3000 3500 0 500 1000 1500 2000 2500 3000 3500 4000
Time/s Time/s
Fig. 13. Comparison between the numerical simulation and model test results for displacement-time curves.

100 (7) is relatively higher than the true value of the T2 test, which in-
Prediction for T1
90 Fcy F f =0.45
Fcy F f =0.38 dicates that the stiffness of the soil is overestimated as the number of
Average displacement/mm

Test for T1
80 Prediction for T2 cycles increases during cyclic loading. This is possibly because the clay
Test for T2
70
Prediction for T3
in the T2 test is relatively softer than the clay in the other tests. In
60 Fcy F f =0.33
Test for T3 addition, there are some other reasons for the deviation. One reason is
50 Prediction for T4
that the strengths of the soft clays along the depth in different locations
Test for T4
40 of the model tank show certain differences; however, the strength of the
30 soft clay used in determining the model parameters is constant. Another
20 Fcy F f =0.24 reason might be that cracking of the soils in passive regions cannot be
10 captured using the FEM; however, it has an effect on the deformation of
0 the caisson. Actually, due to the strong nonlinearity of the deformation
0 50 100 150 200 250 300 350 400 450
of the soft clays and the soil-anchor contact under cyclic loading, it is
Cycle number
still challenging to predict the cyclic deformation process of the anchor
Fig. 14. Comparison between the numerical simulation and model test results perfectly under three-dimensional stress conditions. Although accu-
for cyclic cumulative displacement. rately predicting the test results is still a challenge, the consistency of
the experimental results in the overall trend proves the feasibility of the
Fig. 13 shows a comparison of the predicted and measured dis- numerical method to a certain extent.
placement-time curves along the mooring direction at the mooring Fig. 15 shows a comparison of the load-displacement curves of the
point. The predicted results show that the displacements increase ra- anchor between the predictions and test results. It can be seen from the
pidly at the beginning of cyclic loading, and then the increasing rate figure that the finite element method based on the stiffness degradation
gradually decreases as the number of load cycles increases. The dis- model can describe the nonlinear, hysteretic and displacement accu-
placement in a single cycle changes less compared to the cumulative mulation characteristics of the suction anchor in soft clays under cyclic
displacement along the tension direction during cyclic loading. Ex- loads, and the predictions are basically consistent with the test results.
cessive cumulative displacement is the main cause of anchor failure. It should be pointed out that the ideal elastic-plastic constitutive model
The predictions are generally in agreement with test results. Fig. 14 itself cannot reasonably describe the cyclic stress-strain response of soft
shows a comparison of the cumulative displacement between the pre- clays; however, the introduction of the stiffness degradation model
dictions and the model test results under different load levels. The causes the elastic modulus of soils to decrease and the plastic de-
predictions indicate that the higher the cyclic load level is, the faster the formation of soils occurs as the cycle number increases. In addition,
increase in the cyclic cumulative displacement, and the smaller the ABAQUS automatically linearly interpolates the elastic modulus set in
number of load cycles required to reach the failure displacement. The Table 3 during the calculation for each time-increment step in each
predictions are consistent with the overall trend of the test results, al- cycle, so the elastic modulus used in each time-incremental step will be
though there are some deviations in the magnitude. Especially for the different. As a result, the nonlinear and hysteretic characteristics of soil
T2 test, the numerical method underestimates the real displacement during each loading and unloading cycle can be captured.
value of the anchor, so the predicted displacement is smaller than the
experimental displacement, which results in predictions for the T2 and 5.3. Cyclic bearing capacity
T3 tests crossing over each other. The underestimation of the dis-
placement means that the cyclic degradation index determined by Eq. The cyclic bearing capacity can be predicted based on the failure

9
X. Cheng, et al. Computers and Geotechnics 122 (2020) 103552

1.2 1.2
1.0 1.0
0.8 0.8
0.6 0.6
0.4 0.4
0.2
Load/kN

0.2

Load/kN
0.0 0.0
-0.2 -0.2
-0.4 -0.4
-0.6 -0.6
-0.8 Fcy Ff =0.45 -0.8
-1.0 -1.0 Fcy Ff =0.45
-1.2 -1.2
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
Displacement/mm Displacement/mm
(a) T1 model test (b) T1 prediction

1.0 1.0
0.8 0.8
0.6 0.6
0.4 0.4
0.2
Load/kN

0.2

Load/kN
0.0 0.0
-0.2 -0.2
-0.4 -0.4
-0.6 -0.6
-0.8 Fcy Ff =0.38
-0.8 Fcy Ff =0.38
-1.0 -1.0
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
Displacement/mm Displacement/mm
(c) T2 model test (d) T2 prediction
Fig. 15. Comparison between the numerical simulation and test results for the load-displacement curves.

1.2 Establishing the stiffness degradation model and its numerical im-
plementation are crucial for the numerical method. The stiffness de-
1.0
gradation model is proposed first referring to the definition of the de-
gradation index by Idriss et al. [19]. by performing a series of
(Fa+Fcy )/ Ff

0.8
undrained cyclic triaxial tests on soft clays under the joint actions of
0.6 different static and cyclic deviatoric stresses. Subsequently, the pro-
Pridiction posed stiffness degradation model for soft clays is embedded in the
0.4
Test ABAQUS software package by encoding the USDFLD subroutine and
0.2 Fitting line defining relevant field variables. The function of the stiffness de-
gradation model is to establish the relationship between the elastic
0.0
10 100 1000 modulus of the soil elements and the initial static stress level, cyclic
Nf stress level and stress cycle number. When applied to the analysis of the
suction anchor, it can reflect the degradation of the stiffness and the
Fig. 16. Comparison between the numerical simulation and test results for the
accumulation of the plastic deformation of soils around the anchor
cyclic bearing capacity.
during cyclic loading.
The new numerical computation method predicts the cyclic beha-
criterion of 0.6 times the anchor plate width after obtaining the dis- viour of suction anchors by coupling the elastic-perfectly plastic model
placement-time curves. The predicted and measured number of load with the Mohr-Coulomb yield criterion and the proposed stiffness de-
cycles to failure is denoted as Nf and Nf , respectively. The specific gradation model for soft clays. The numerical method can simulate the
values of both are shown in Table 2. The maximum relative error be- cyclic deformation process of suction anchors in soft clays to some
tween the predictions and test results for the number of load cycles to extent and capture the nonlinearity, hysteresis and cyclic accumulation
failure is less than 15%, as shown in Table 2. The normalised cyclic characteristics of the load-displacement responses of anchors. The
bearing capacity versus the number of load cycles to failure is shown in cyclic bearing capacity of anchors can also be predicted well using the
Fig. 16. They can be fitted using the same straight line in semiloga- method based on an appropriate displacement failure criterion.
rithmic coordinates based on Origin 9.0 software, and the coefficient of As a simplification, only the applicability of the new numerical
determination R2 is 0.91, which indicates that the predictions generally method for harmonic loads was discussed in this paper. However, cyclic
agree with the test results, so the cyclic bearing capacity can also be loads such as wave loads or earthquake loads are more irregular vari-
predicted well using the finite element method. able-amplitude cyclic loads than constant-amplitude harmonic loads in
the marine environment. “The number of load cycles” was introduced
6. Conclusions into the stiffness degradation model for soft clays in the paper.
However, the number of load cycles is only applicable to constant-
The main contribution of the paper is to present a new numerical amplitude harmonic loads but not to irregular cyclic loads. Ordinarily,
computation method that can predict the cyclic behaviour of suction the laboratory tests are run with one constant cyclic shear stress am-
anchors based on a stiffness degradation model for soft clays. plitude throughout each test. Therefore, the concept of an “equivalent

10
X. Cheng, et al. Computers and Geotechnics 122 (2020) 103552

number of load cycles” [3,20] should be introduced into the research to [4] Ravichandran V, Maji VB, Gandhi SR. Field Testing of suction anchors for mooring
consider the effects of the cyclic loading history for irregular variable- applications. Indian Geotech J 2015;45(3):267–77.
[5] Sukumaran B, McCarron WO, Jeanjean P, Abouseeda H. Effcient finite element
amplitude cyclic loads. According to the concept, the equivalent techniques for limit analysis of suction anchors under lateral loads. Comput Geotech
number of load cycles is the number of cycles at the loading level of a 1999;24:89–107.
constant-amplitude load packet that would produce the same perma- [6] Cao J, Phillips R, Popescu R. Numerical analysis of the behavior of suction anchors
in clay. Int J Offshore Polar 2003;13(2):154–9.
nent deformation (or permanent pore pressure) as that produced by [7] Maniar DR. A computational procedure for simulation of suetion anchor behavior
irregular cyclic variable-amplitude cyclic loads. The applicability of under axial and inclined loads. Austin: University of Texas at Austin; 2004.
irregular cyclic variable-amplitude loading will be further studied and [8] Monajemi H, Razak HA. Finite element modeling of suction anchors under com-
bined loading. Mar Struct 2009;22(4):660–9.
discussed in future research. [9] Ahn J, Lee H, Kim YT. Holding capacity of suction caisson anchors embedded in
The cyclic behaviours of the reduced-scale suction anchor in soft cohesive soils based on finite element analysis. Int J Numer Anal Met
clays are simulated well using the proposed numerical computation 2014;38(15):1541–55.
[10] Kim S, Choo YW, Kim JH, Kim DS, Kwon O. Pullout resistance of group suction
method in this study. However, due to some limitations of the reduced-
anchors in parallel array installed in silty sand subjected to horizontal loading-
scale model tests, such as incorrect stress levels, the applicability of the centrifuge and numerical modeling. Ocean Eng 2015;107:85–96.
numerical method to real offshore conditions should be further vali- [11] Wang L, Guo Z, Yuan F. Quasi-static three-dimensional analysis of suction anchor
dated by simulating prototype tests or centrifuge model tests of suction mooring system. Ocean Eng 2010;37(13):1127–38.
[12] Guo Z, Jeng DS, Guo W, Wang LZ. Failure mode and capacity of suction caisson
anchors in future work. under inclined short-term static and one-way cyclic loadings. Mar Georesour Geotec
2018;36(1):52–63.
CRediT authorship contribution statement [13] Zhu FY, O'Loughlin CD, Bienen B, Cassidy MJ, Morgan N. The response of suction
caissons to long-term lateral cyclic loading in single-layer and layered seabeds.
Géotechnique 2017;68(8):729–41.
Xinglei Cheng: Conceptualization, Methodology, Software, Writing [14] Gelagoti F, Georgiou I, Kourkoulis R, Gazetas G. Nonlinear lateral stiffness and
- original draft. Piguang Wang: Investigation, Validation, Software. Na bearing capacity of suction caissons for offshore wind-turbines. Ocean Eng
2018;170:445–65.
Li: Data curation, Formal analysis. Zhongxian Liu: Supervision. [15] Kourkoulis RS, Lekkakis PC, Gelagoti FM, Kaynia AM. Suction caisson foundations
Yadong Zhou: Writing - review & editing. for offshore wind turbines subjected to wave and earthquake loading: effect of soil-
foundation interface. Géotechnique 2014;64(3):171.
[16] Zhang Z, Cheng X. Predicting the cyclic behaviour of suction caisson foundations
Declaration of Competing Interest using the finite element method. Ships Offshore Struct 2017;12(7):900–9.
[17] Cheng X, Wang J, Wang Z. Incremental elastoplastic FEM for simulating the de-
formation process of suction caissons subjected to cyclic loads in soft clays. Appl
The authors declare that they have no known competing financial
Ocean Res 2016;59:274–85.
interests or personal relationships that could have appeared to influ- [18] Cheng X, Yang A, Li G. Model tests and finite element analysis for the cyclic de-
ence the work reported in this paper. formation process of suction anchors in soft clays. Ocean Eng 2018;151:329–41.
[19] Idriss IM, Dobry R, Singh RD. Nonlinear behavior of soft clays during cyclic loading.
J Geotech Geoenviron 1978;104(12):1427–47.
Acknowledgements [20] Anderson KH, Pool JH, Brown SF, Pool JH. Cyclic and static laboratory tests on
Drammen clay. J Geotech Eng Div ASCE 1980;106(5):499–529.
Financial support from the National Natural Science Foundation of [21] Vucetic M, Dobry R. Degradation of marine clays under cyclic loading. J Geotech
Eng 1988;114(2):133–49.
China (Grant No. 51878434, 51678014) and China Postdoctoral [22] Huang MS, Shuai LI. Degradation of stiffness and strength of offshore saturated soft
Science Foundation funded project (Grant No. 2019M650411) and clay under long-term cyclic loading. Chinese J Geotech Eng 2010;32(10):1491–8.
Beijing Postdoctoral Research Foundation (Grant No. ZZ2019-101) is [23] Guo L, Wang J, Cai Y, Liu H, Gao Y, Sun H. Undrained deformation behavior of
saturated soft clay under long-term cyclic loading. Soil Dyn Earthq Eng
gratefully acknowledged. 2013;50:28–37.
[24] Li LL, Dan HB, Wang LZ. Undrained behavior of natural marine clay under cyclic
References loading. Ocean Eng 2011;38(16):1792–805.
[25] Hyodo M, Yamamoto Y, Sugiyama M. Undrained cyclic shear behaviour of normally
consolidated clay subjected to initial static shear stress. Soils Found
[1] Chen W, Randolph MF. Uplift capacity of suction anchors under sustained and cyclic 1994;34(4):1–11.
loading in soft clay. J Geotech Geoenviron 2007;133(11):1352–63. [26] Zhou J, Gong X. Strain degradation of saturated clay under cyclic loading. Can
[2] Anderson KH. Bearing capacity under cyclic loading-offshore, along the coast, and Geotech J 2001;38(1):208–12.
on land. Can Geotech J 2009;46(5):513–35. [27] Yasuhara K, Hirao K, Hyde AF. Effects of cyclic loading on undrained strength and
[3] Andersen KH, Lauritzsen R. Bearing capacity for foundations with cyclic loads. J compressibility of clay. Soils Found 1992;32(1):100–16.
Geotech Geoenviron 1988;114(5):540–55.

11