Soil Dynamics and Earthquake Engineering 136 (2020) 106212

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering

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

Response spectrum method for seismic analysis of monopile offshore

wind turbine
Mi Zhao , Zhidong Gao , Piguang Wang *, Xiuli Du
Key Laboratory of Urban Security and Disaster Engineering of Ministry of Education, Beijing University of Technology, Beijing 100124, China

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

Keywords: Monopile offshore wind turbines (MOWT) are inserted into sea water and sea bed soil. As the piles are usually
Monopile offshore wind turbine embedded into the soil as deep as about 40 m, dynamic pile-water and pile-soil interactions, as well as seismic
Pile-water interaction excitation non-uniformity, often occur under earthquake. In this paper, the response spectrum method (RSM) is
Pile-soil interaction
developed to calculate the maximum response of MOWT considering the pile-water and pile-soil interactions
Non-uniform seismic excitation
Response spectrum method
subjected to the non-uniform seismic excitation. The tower and pile are modeled by Timoshenko beam elements,
with both nacelle and rotor including hub and blades integrated as a lumped mass at the top of the tower. The
pile-water and pile-soil interactions are considered by added mass and soil spring at their interfaces, respectively.
The non-uniform seismic excitation is obtained by the site transfer function from the horizontal earthquake
motions or design response spectrum at the water-soil interface. The seismic analysis model is solved by using a
multiple-support RSM under design response spectrum. Seismic responses of a case of 5-MW MOWT on two types
of soil sites are analyzed. The results indicate that the pile-water and pile-soil interactions and seismic excitation
non-uniformity should be considered in analyzing such seismic problems of MOWT, while the RSM can be an
effective tool.

1. Introduction non-uniformity of seismic excitation acting on monopile, should be

considered in seismic analysis model of the monopile offshore wind
Over the past decades, focus on renewable and clean energy has turbines (MOWT).
strengthened the motivation to use wind energy. Offshore wind turbines Two types of models have been adopted to consider the pile-water
are playing an important role in producing electrical energy in many and pile-soil interactions. One is the three-dimensional solid model
countries and regions. An increasing number of offshore wind farms are [3]. The pile and parts of its surrounding soil are modeled by solid finite
being planned worldwide to satisfy the growing energy demand. The elements maybe with an absorbing boundary condition at model trun­
offshore wind turbines are subjected to environmental loads which in­ cation boundary. The pile-water interaction is modeled by added mass
cludes aerodynamic wind loads and ocean wave loads. Furthermore, it obtained using the method in Ref. [4]. It is inefficient use the solid model
may also be under threat from earthquake in an area of intense seismic in structural seismic design because of its high time costs in modeling
activities such as in the eastern coast of China and the western coast of and computation.
the USA [1]. It is therefore necessary to study the analysis model and The other is the substructure model. It is rather applicable in struc­
method for seismic design of offshore wind turbine. tural seismic design than the above mentioned solid model. In the sub­
At present, due to its simplicity in installation, cost saving and suc­ structure model, the pile is modeled by finite beam elements. The pile-
cessful application in offshore oil and gas infrastructures, monopile is soil interaction is simulated by soil springs including linear springs
the most widely used foundation for offshore wind turbine in shallow [5–8] and nonlinear springs [9–18] based on p-y curves. The pile-water
water depth. A commonly used monopile is a hollow steel member with interaction is simulated by the Morison equation [12,13,17] and added
an outer diameter of 3–6 m and a length of 22–40 m [2], which is mass [9,11,14,18]. While about seismic analysis, only very few refer­
inserted into the seabed soil under sea water. The interaction between ences [10,11,14,18] focus on MOWTs. The uniform seismic excitations
the monopile and its surrounding water and soil, as well as the are considered in Refs. [11,18] while the non-uniform ones in Refs. [10,

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

https://doi.org/10.1016/j.soildyn.2020.106212
Received 9 October 2019; Received in revised form 12 March 2020; Accepted 3 May 2020
Available online 16 June 2020
0267-7261/© 2020 Elsevier Ltd. All rights reserved.
M. Zhao et al. Soil Dynamics and Earthquake Engineering 136 (2020) 106212

Fig. 1. MOWT-water-soil system under non-uniform earthquake.

In this paper, the RSM is developed for seismic analysis of MOWT.

Table 1
The resting parts of this paper are organized as follows. The analysis
Parameters of the 5-MW MOWT [40].
model of MOWT including the pile-water and pile-soil interactions
Properties Values under the non-uniform seismic excitation is presented in Section 1. The
Height of tower 87.6 m multiple-support RSM is applied to solve the seismic analysis model of
Monopile length above water surface 10 m MOWT in Section 2. The numerical examples are given to indicate the
Depth of sea water 20 m
effectiveness of RSM in analyzing the seismic response of MOWT in
Monopile length buried in soil 45 m
Diameter of top and bottom of tower 3.87 m and 6 m Section 3. Conclusions follow in Section 4.
Thickness of top and bottom of tower 0.019 m and 0.027 m
Diameter and thickness of monopile 6 m and 0.06 m 2. Seismic analysis model of monopile offshore wind turbine
Density, young’s modulus and shear modulus of 8500 kg/m3, 210 GPa and 80.8 (MOWT)
tower GPa
Density, young’s modulus and shear modulus of 7850 kg/m3, 210 GPa and 80.8
monopile GPa Fig. 1 shows the MOWT-water-soil system under earthquake exci­
Mass of nacelle 240 ton tation, where z is the vertical coordinate with an origin at water-soil
Mass of rotor (blade þ hub) 110 ton interface. As shown in Fig. 1a, the MOWT consists of a rotor
(including hub and three blades), a nacelle, a tower and a monopile that
is inserted into seabed soil under sea water. As shown in Fig. 1b, the
14]. The time history method is adopted in all above seismic analyses.
tower and the pile are modeled by Timoshenko beam elements, with the
The Response Spectrum Method (RSM) [19] has been widely incor­
nacelle and rotor integrated as a lumped mass at the top of tower. The
porated into computing codes for seismic analysis and used in design of
pile-water and pile-soil interactions are considered by added mass and
aboveground buildings in many countries [20–23]. Recently, the RSM
soil spring at their interfaces, respectively. As shown in Fig. 1c, starting
has also been developed for seismic soil-structure interaction analysis
from the horizontal earthquake motions or design response spectrum at
for underground structures [24]. For seismic analysis of offshore wind
water-soil interface, the non-uniform seismic excitation is obtained by
turbine, the RSM have been suggested in the recent guidelines and
the site transfer function. Only the earthquake loading is considered in
standards [25–27]. However, it has not been demonstrated whether the
this paper. The detailed description about the seismic analysis model of
water-structure and the soil-structure interactions and the non-uniform
MOWT is given as follows.
seismic excitation should be considered or not. In such problems, only a
wind turbine on land has been studied by using the traditional RSM, but
2.1. MOWT-water-soil system
without considering soil-structure interaction and non-uniform seismic
excitation [28]. To the best of the authors’ knowledge, no RSM has yet
2.1.1. MOWT
been presented for seismic analysis of MOWT considering pile-water and
The tower and pile are modeled as the three-dimensional two-nodal
pile-soil interactions subjected to non-uniform seismic excitation.
Timoshenko beam elements that can consider structural shear and

Fig. 2. Dynamic shear modulus ratio and dynamic hysteretic damping ratio varying with shear strain, respectively denoted by solid and dotted lines.

2
M. Zhao et al. Soil Dynamics and Earthquake Engineering 136 (2020) 106212

2 38 9 2 38 9
Mr 0 0 < u€r = Crr Crw Crs < u_ r =
40 Mw 0 5 u€w þ 4 Cwr Cww Cws 5 u_ w
: ; : ;
0 0 Ms u€s Csr Csw Css u_ s
2 38 9
Krr Krw Krs < ur =
þ 4 Kwr Kww Kws 5 uw
: ;
Ksr Ksw Kss us
8 9
<0 =
¼ fw (3)
: ;
fs

where u is the structural absolute motion vector including the horizontal

displacement and rotation degrees of freedom; the dot over variable
denotes the derivative to time; M, C and K are the lumped mass,
damping and stiffness matrices, respectively; f is the force vector
denoting the action of the water and soil to the structure; and the sub­
scripts w and s correspond to the structural horizontal displacement in
water and that in soil, respectively, while subscript r for the rest struc­
tural degrees of freedom.

2.1.2. Pile-water interaction

The hydrodynamic force can be conveniently modeled as an inertia
force equal to the product of an equivalent mass of water and the ac­
celeration of structure. The equivalent mass of water is the so-called
hydrodynamic added mass. It is obtained by assuming the water
controlled by the Laplace’s equation with resting condition at water
surface and impermeable condition at water bottom. The hydrodynamic
Fig. 3. Testing model of a 5-WM MOWT above water surface. added mass that varies along z coordinate can be calculated using the
following equation [4].

16ρw hw a X
∞
ð 1Þj 1
K1 ðλaÞ
mw ðzÞ ¼ cosðλzÞ (4)
π j¼1 ð2j 1Þ2 K0 ðλaÞ þ K2 ðλaÞ

where ρw is the water mass density; hw is the water depth; λ ¼ (j-0.5)π /

hw; a is the outside radius of the pile; and Kn is the second kind of
modified Bessel function of n order.
The force fw denoting water action to structure in Equation (3) can
therefore be written as
fw ¼ M∞
wu€w (5)

Fig. 4. First four order natural frequencies and modal shapes of MOWT above where the hydrodynamic added mass matrix M∞ w is obtained from
water surface. Equation (4) by using finite element discretization and assembly.

bending deformations by the finite element method. A lumped mass is 2.1.3. Pile-soil interaction
structured at the top of tower to represent the quality of nacelle and Lumped parameter model [29] is widely applied in the pile-soil
rotor (including hub and blades). interaction analysis due to its simplicity and efficiency. The model
The cross section of the tapered tower changes along the length. For uses springs to describe the deformation properties of soil. Different
convenience in calculation, the tower cross section is treated to be types of soil springs mainly include linear and nonlinear springs. In this
uniform in geometry in each beam element, while the elements varies paper, a simple linear soil spring is used according to the Chinese Code
from top to bottom. For the i-th beam element, the equivalent inertia for Design of Ground Base and Foundation of Highway Bridges and
moment and area of the cross section are obtained, respectively, as Culverts (JTG D63-2007) [30]. One end of the soil spring is connected to
Z the pile element node, and the other end is subjected to the horizontal
1 zi earthquake motion. The stiffness of the i-th soil spring can be written as
Ii ¼ IðzÞdz (1)
li ziþ1
ki ¼ hi mjzj (6)
Z zi
1
Ai ¼ AðzÞdz (2) where hi is the computational height of pile elements connecting the soil
li
spring; jzj is the absolute value of z coordinate of soil spring location; m
ziþ1

where li is the element length; I(z) and A(z) are the inertia moment and is a given coefficient that is chosen according to the seismic code [30].
area of the cross section, respectively, which vary along z coordinate; The force fs denoting soil action to structure in Equation (3) can
and zi and ziþ1 are the two element node coordinates. therefore be written as
The dynamic finite element equation of the structure consisting of �
f s ¼ K∞ s us ug (7)
the tower (including the nacelle and rotor as a lumped mass) and pile
under earthquake can be expressed as where the soil spring matrix K∞ s is a diagonal matrix with ki from
Equation (6) as its elements; and ug is the horizontal displacement vector
of non-uniform multiple-support earthquake excitation, which will be

3
M. Zhao et al. Soil Dynamics and Earthquake Engineering 136 (2020) 106212

Fig. 5. Horizontal acceleration time histories and corresponding Fourier spectra.

given in Subsection 1.2. displacement as

Mu€d þ Cu_ d þ Kud ¼ MRu€g CRu_ g (12)

2.1.4. Interaction system
Substituting Equations (5) and (7) into Equation (3), and after some In engineering practice, the damping forces on the right-side of
manipulations, the dynamic equation of MOWT considering the pile- Equation (12) caused by seismic excitations might be neglected because
water and pile-soil interactions under the non-uniform earthquake they are generally much smaller than the corresponding inertia forces,
excitation could be obtained as which will also be verified in the Subsection 3.2.1 by numerical exam­
2 3
8 9 2 38 9 ples. Therefore, Equation (12) can be simplified as
Mr 0 0
6 < u€r >
7> = Crr Crw Crs > < u_ r >=
6 7 6 7 Mu€d þ Cu_ d þ Kud ¼ MRu€g (13)
60 Mw þ M∞ w 0 7 u€w þ 4 Cwr Cww Cws 5 u_ w þ
4 5>
: ; > >
: > ;
0 0 Ms u€s Csr Csw Css u_ s If the dynamic displacement is obtained from Equation (12) or (13),
2 3 8 9 8 9 (8) the absolute displacement can be given by Equation (10).
8 9 > > >0 >
Krr Krw Krs > ur = > >>0 >
> > > >
>
6 7< < = < =
6 7
6 Kwr Kww Kws 7 uw ¼ 0 ¼ 0 ug
4 5>
: > ; >> >
> >
> >
> 2.2. Non-uniform multiple-support seismic excitations
∞ u >
: ∞ >
; >
: ∞ >
;
Ksr Ksw Kss þ K s
s K ug
s K s
The non-uniform earthquake motions in soil site are required to solve
It should be noted that the structural displacement at water-soil Equation (12) or (13). For the seismic soil-structure interaction problem
interface is not given in Equation (8) for its brevity. This degree of on land, the earthquake motions at ground surface are usually given in
freedom requires to consider the pile-water and pile-soil interactions advance in engineering practices. Because the one-dimensional site
simultaneously as well as the earthquake excitation. response under horizontal earthquake excitation is not affected by the
Equation (8) can be rewritten in brief as existence of sea water, the earthquake motion on water-soil interface is
Mu€ þ Cu_ þ Ku ¼ K∞ ug (9) regarded the same as that on ground surface, as show in Fig. 1c.
Therefore, the non-uniform multiple-support earthquake motions can be
The absolute displacement is decomposed into pseudo-static and obtained from the given horizontal earthquake motion at water-soil
dynamic components as [31] interface by using the one-dimensional site transfer function in fre­
u ¼ us þ ud (10) quency domain.
b z ðωÞA0 ðωÞ
Az ðωÞ ¼ H (14)
where the superscripts s and d denote the two components. The pseudo-
static displacement can be written as where ω is the circular frequency; A0 ðωÞ and Az ðωÞ are the Fourier
s 1 ∞
u ¼ ðKÞ K ug ¼ Rug (11) spectra of the earthquake motions at the water-soil interface and the
location of coordinate z, respectively; and H
b z ðωÞ is the one-dimensional
where the superscript 1 denotes the matrix inversion; and site transfer function that can be written as [32,33].
R ¼ ðKÞ 1 K∞ is the influence matrix.
Substituting Equation (11) into Equation (10), and then the result b z ðωÞ ¼ sumðEN Þ
H (15)
2
into Equation (9), obtain the equation with respect to the dynamic

4
M. Zhao et al. Soil Dynamics and Earthquake Engineering 136 (2020) 106212

Fig. 6. Peak values of time histories of structural responses varying along z coordinate. Effect of damping force on structural response is studied.

! Q
Y
N N-th soil layers; denotes the continued multiplication; I is the unit
EN ¼ Tn I (16) vector; i is the imaginary unit; and ρn , Gn , ζn , cn and hn are the density,
shear modulus, hysteretic damping ratio, complex wave velocity and
n¼1

2 3 height of the n-th layer soil, respectively.

� � � �
6 1 þ ρn cn iωhn ρn cn iωhn
7 Besides, the design response spectra at different depths of soil are
e cn 1 e
required when using RSM to solve the MOWT-water-soil system sub­
cn
6 7
16 ρnþ1 cnþ1 ρnþ1 cnþ1 7
6
Tn ¼ 6 � � � � 7
7 (17) jected to non-uniform seismic excitation, as shown in Fig. 1c. Only the
26 ρn cn iωhn ρn cn iωhn 7
4 1 e cn 1þ e cn 5
design acceleration response spectrum at the ground surface is available
ρnþ1 cnþ1 ρnþ1 cnþ1
in the Chinese seismic design code [22], and it can be regarded as that at
the water-soil interface. A method [24] can be used to calculate the
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Gn ð1 þ 2ζn iÞ design response spectrum at different depths of soil from a given design
cn ¼ (18) response spectrum at the water-soil interface. Firstly, the power spectral
ρn
density function at the water-soil interface is derived from a given ac­
where sum() denotes the sum of all elements of a vector; N denotes the celeration response spectrum [34]. Secondly, the power spectral density

5
M. Zhao et al. Soil Dynamics and Earthquake Engineering 136 (2020) 106212

Fig. 7. Peak value of time history of inertia force of hydrodynamic added mass varying along z coordinate. Two sites and three earthquake records are used.

function at a location of soil is back-calculated by the 1D site transfer 3.2. Mode superposition
function [35] as shown in Equation (15). Thirdly, the design accelera­
tion response spectrum at the location is obtained based on the Substituting Equations (11) and (20) into Equation (10), obtains the
approximate relationship between the power spectral density function structural absolute displacement and internal force as
and the response spectrum.
X
L L X
X J
u ¼ us þ ud ¼ Rl ugl þ Xj γlj qlj (23)
3. Response spectrum method (RSM) l¼1 l¼1 j¼1

Equation (13) can be solved by the multi-support RSM. The details on where ugl is the l-th element in ug .
this method can be found in the references [36–39]. It is summarized in In Equation (23), the peak value of the earthquake motion ugl can be
brief as follows. denoted as Sl , and the peak values of qlj can be obtained from seismic
response spectra of Equation (22) and denoted as Slj . The peak value of
3.1. Mode decomposition structural absolute displacement can be obtained by the complete
quadratic combination (CQC) [36–39]. It can be written as
The generalized eigenvalue problem of the mass and stiffness vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
uX
matrices of Equation (13) can be written as u L X X
L L X L X J
u
u ρuk ul Rk *Rl Sk Sl þ 2 ρuk qlj Rk *Xj Sk γlj Slj
�
K ω2 M X ¼ 0 (19) u k¼1 l¼1
jujmax ¼ u
k¼1 l¼1 j¼1
(24)
u X L X L X J X J

By solving Equation (19), the first J natural frequencies ωj and modal

tþ ρ X *X γ S γ S qki qlj i j ki ki lj lj

shape vectors Xj for j ¼ 1, 2, …J could be obtained. The dynamic

k¼1 l¼1 i¼1 j¼1

displacement can be written in the following form

where | |max denotes to take the peak value of each element of a time
L X
X J history vector; and * denotes that two vectors element by element are
ud ¼ Xj γlj qlj (20) multiplied.
The corresponding coefficients ρuk ul , ρuk qlj and ρqki qlj in Equation (24)
l¼1 j¼1

where the subscript l ¼ 1, 2, …L denotes the numbers of multiple- can be expressed as

support earthquake motions; qlj are the modal displacements; and γlj 1
Z ∞
1
are the corresponding modal participation coefficients which can be ρu k u l ¼ G ðωÞdω
4 u€k u€l
(25–1)
σ uk σ ul ∞ ω
written as
Z ∞
1 1
XTj MRl ρuk qlj ¼ Hj ðωÞGu€k u€l ðωÞdω (25–2)
γ lj ¼ (21) σuk σqlj ∞ ω2
XTj MXj
Z ∞
1
where the superscript T denotes the transposition of a vector; and Rl is ρqki qlj ¼ H i ðωÞHj ðωÞGu€k u€l ðωÞdω (25–3)
σ qki σqlj ∞
the l-th column of the matrix R in Equation (13).
Substituting Equations (20) and (21) into Equation (13) and using with the root mean squares of ugk and qki , respectively, as
the orthogonality of the modal shapes, obtain the following dynamic Z ∞
single-degree-of-freedom equations for j ¼ 1, 2, …J and l ¼ 1, 2, …L as σ 2uk ¼
1
G ðωÞdω (26–1)
4 u€k u€k ω
∞
q€lj þ 2ζj ωj q_lj þ ω2j qlj ¼ u€gl (22)
Z ∞
σ 2qki ¼ 2
jHi ðωÞj Gu€k u€k ðωÞdω (26–2)
where u€gl is the l-th element in u€g ; and ζj is the modal damping ratio of ∞
the j-th mode.
where Hi and Hj are the frequency response functions of Equation (22)

6
M. Zhao et al. Soil Dynamics and Earthquake Engineering 136 (2020) 106212

Fig. 8. Peak values of time histories of structural responses varying along z coordinate. Effect of pile-water interaction on structural response is studied.

for the i-th and j-th modes, respectively; Hi is the conjugate function of obtained from Equation (27)–(2) by using the site transfer function of
Hi ; and Gu€k u€l is the real-valued power spectral density of acceleration Equation (15). In Equation (27)–(2), G0 is a scale factor; ωg and ζg are the
process of earthquake motion. The cross-power spectral density can be predominant frequency and damping ratio of site, respectively; and ωf
obtained from the power spectral densities as and ζf are the seismic energy parameters which reflect the changes of
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi seismic energy of low frequency [38].
Gu€k u€l ¼ Gu€k u€k Gu€l u€l (27–1)
4. Numerical examples
with the power spectral density for the earthquake motion at water-soil
interface as
Seismic responses of a modern NREL 5-MW MOWT on two types of
ω4g þ 4ζ2g ω2g ω2 ω4 soil sites are calculated using the MOWT-water-soil model and multiple-
Gu€k u€k ¼ G0 � �2 � �2 (27–2) support RSM in this section. The geometry and material parameters of
ω2 ω2g þ 4ζ2g ω2g ω2 ω2 ω2f þ 4ζ2f ω2f ω2 MOWT are obtained from Ref. [40] and listed in Table 1.
Two typical sites with homogeneous soil, namely Site 1 and Site 2,
for k ¼ 1 and the power spectral densities in soil for k ¼ 2, 3, …L are selected to study the influence of different site conditions on

7
M. Zhao et al. Soil Dynamics and Earthquake Engineering 136 (2020) 106212

Fig. 9. The peak value of time history of resisting force of soil spring varying along z coordinate. Two sites and three earthquake records are used.

structural response and calculation accuracy of RSM. A soil density of 4.2.1. Effect of seismic damping force
2020 kg/m3 and a shear wave velocity of 251 m/s are set for Site 1, and a Seismic responses of the 5-MW MOWT are computed to indicate the
soil density of 1870 kg/m3 and a shear wave velocity of 150 m/s for Site effect of the damping force. The pile-water and pile-soil interactions and
2. The bedrock depth for the two sites are 80 m, and its density and shear the non-uniform seismic excitation are considered in this study.
wave velocity are 2300 kg/m3 and 800 m/s, respectively. The two sites Fig. 6 shows the peak values of time histories of absolute horizontal
belong to Class II (the equivalent shear wave velocity is less than 500 m/ displacement, absolute horizontal acceleration and section bending
s and greater than 250 m/s, and the cover thickness of site is greater than moment of whole structure varying along z coordinate. In this figure, the
5 m) and IV (the equivalent shear wave velocity is less than 150 m/s, and structure is considered in above two site conditions subjected to above
the cover thickness of site is greater than 80 m) respectively according to three earthquake records. The results considering the damping force are
the Chinese seismic code in soil classification [22]. The soil nonlinearity compared with those without the damping force. The error of the result
is considered by the equivalent linearization model. The curves of dy­ without damping force relative to that with damping force is also given
namic shear modulus ratio and dynamic hysteretic damping ratio at the location where the maximum response occurs. It can be seen from
varying with shear strain are shown in Fig. 2. The coefficient of foun­ Fig. 6 that the displacement, acceleration and bending moment results
dation soil, m, for Site 1 and Site 2 are 120 MPa/m and 30 MPa/m obtained from considering and ignoring the damping force are very close
respectively according the seismic code [30]. for all of the cases of site conditions and earthquake records, and their
The modal damping ratio of MOWT system is set as 9% for all modes maximum response errors are less than 5.3%, 6.3% and 8.3%, respec­
according to Ref. [41]. Except that the mesh size at the bottom of tower tively. Therefore, for the sake of simplicity, the damping forces on the
is 2.6 m, the other parts of the tower and the pile are meshed with sizes right-side of Equation (12) caused by seismic excitations are neglected in
of 5 m, which satisfies the accuracy requirement. All analyses are ach­ the present study.
ieved by a self-compiled Matlab program.
4.2.2. Effect of pile-water interaction
4.1. Verification of the MOWT model Studying the effect of water on site or structure under earthquake is
very meaningful [41]. Seismic responses of the 5-MW MOWT are
The computational model of MOWT is verified using existing test computed to indicate the effect of the pile-water interaction. The
data. The model above water surface is established and fixed at water pile-soil interaction and non-uniform seismic excitation are considered
surface, as shown in Fig. 3. The natural frequencies and modal shapes in this study, and the damping force is neglected. The time history of
are presented in Fig. 4. The results indicate that the first two natural inertia force due to the hydrodynamic added mass is given in Equation
frequencies are close with the test data (0.312 Hz and 2.936 Hz) pro­ (5). Its peak value varying along z coordinate is shown in Fig. 7.
vided in Ref. [40]. Fig. 8 shows the same results as Fig. 6. The results considering the
pile-water interaction are compared with those without the pile-water
4.2. Seismic analysis of MOWT using time history method interaction. Ignoring pile-water interaction could be calculated by
setting the hydrodynamic added mass in Equation (8) as zero. The error
Seismic responses of the 5-MW MOWT on two types of sites are of the result ignoring pile-water interaction relative to that considering
calculated using the time history analysis in this section. The well- pile-water interaction is also given at the location where the maximum
known implicit Newmark’s time integration algorithm [19] is used to response occurs. It can be seen from Fig. 8 that the acceleration and
solve the dynamic finite element equation of structure. Its two param­ bending moment results obtained from considering and ignoring the
eters are 0.5 and 0.25. pile-water interaction are rather close, while the displacement results
Three different earthquake records are given at the water-soil are very different especially at the top of tower. The maximum response
interface of site. They are the Kobe record, the Friuli record and an error of displacement can be more than 31%, indicating the pile-water
artificial record. They are selected based on the different main frequency interaction should be considered in the seismic analysis of MOWT
range of their Fourier spectra. The peak ground acceleration is adjusted when the structural displacement response at the top of tower is
to 0.10 g and 0.12 g for Site 1 and Site 2, respectively. The horizontal concerned.
acceleration time histories with peak ground acceleration of 0.10 g and
the corresponding Fourier spectra are shown in Fig. 5.

8
M. Zhao et al. Soil Dynamics and Earthquake Engineering 136 (2020) 106212

Fig. 10. Peak values of time histories of structural responses varying along z coordinate. Effect of pile-soil interaction on structural response is studied.

9
M. Zhao et al. Soil Dynamics and Earthquake Engineering 136 (2020) 106212

Fig. 11. Peak values of time histories of structural responses varying along z coordinate. Effect of non-uniform seismic excitation on structural response is studied.

4.2.3. Effect of pile-soil interaction the non-uniform earthquake motions on the pile in soil. The error of the
Seismic responses of the 5-MW MOWT are computed to indicate the result using the constant soil spring parameter and ignoring the pile-soil
effect of the pile-soil interaction. The pile-water interaction and non- interaction relative to that using soil spring parameter of Equation (6) is
uniform seismic excitation are considered in this study, and the damp­ also given at the location where maximum response occurs. It can be
ing force is neglected. The time history of resisting force of linear soil seen from Fig. 10 that the displacement results obtained from using the
spring is given in Equation (7). Its peak value varying along z coordinate two different soil spring parameters and ignoring the pile-soil interac­
is shown in Fig. 9. tion are close, while the acceleration and bending moment results are
Fig. 10 shows the same results as Fig. 6. The soil spring parameter at very different. The maximum response errors of the acceleration and
the bottom of pile in Equation (6) is denoted as Ks. As a case for com­ bending moment obtained from using the two types of soil springs can be
parison, a constant soil spring parameter along z coordinate is chosen as more than 73% and 33%, respectively. Above results indicate the pile-
1/2 Ks The results considering the pile-soil interaction by the soil spring soil interaction should be reasonable considered in the seismic anal­
parameter of Equation (6) are compared with those using the soil spring ysis of MOWT when the structural acceleration and bending moment
parameter of constant 1/2Ks and with those ignoring the pile-soil responses are concerned.
interaction, respectively. Ignoring pile-soil interaction would enforce

10
M. Zhao et al. Soil Dynamics and Earthquake Engineering 136 (2020) 106212

this study.
At the water-soil interface, two different design response spectra as
shown in Fig. 12 are used for Site 1 and Site 2, respectively, according to
the Chinese seismic code [22]. The corresponding equation of the design
spectrum curve is also shown in Fig. 12a with the attenuation index of
descending segment and the adjustment factor of damping, respectively,
as
0:05 ζ 0:05 ζ
γ ¼ 1:0 þ and η ¼ 1:0 þ (28)
0:3 þ 6ζ 0:08 þ 1:6ζ

where Tg denotes the characteristic period; amax denotes the peak ground
acceleration; and ζ is the modal damping ratio. A group of artificial
earthquake motions can be generated to be compatible with the accel­
eration design response spectrum.
Fig. 12. Seismic acceleration response spectra of 9% modal damping ratio at
It is assumed that the structure is located in the region with a seismic
the water-soil interface for (a) Site 1 and (b) Site 2.
intensity of Degree 7. The corresponding peak ground acceleration is
0.10 g and 0.12 g for Site 1 and Site 2, respectively (i.e. probability of
Table 2 exceedance of 10% in 50 years). The characteristic periods are 0.40 s
Natural frequencies of MOWT for Site 1 and Site 2. and 0.75 s for the two sites. The modal damping ratio is 9%. Seven
artificial earthquake motions are generated by using the hybrid simu­
Mode order 1 2 3 4 5
lation method proposed by Zhao et al. [42]. The response spectra of the
Site 1 0.91 10.08 27.20 55.15 92.09 fourteen artificial earthquake motions are also given in Fig. 12.
Site 2 0.89 9.56 25.85 52.62 86.74

4.3.1. Choosing number of modes

4.2.4. Effect of seismic excitation non-uniformity Aim to high efficient calculation, only several of the main modes are
Seismic responses of the 5-MW MOWT are computed to indicate the used in RSM instead of running all the modes regularly, just as that in
effect of the non-uniform seismic excitation. The pile-water and pile-soil mode superposition method. The smallest number of modes for ensuring
interactions are considered in this study, and the damping force is the accuracy of the mode superposition method is studied here. A
neglected. contribution ratio of each mode to accuracy of the method is quantified
Fig. 11 shows the same results as Fig. 6. The results considering the by defining the following error.
non-uniform seismic excitation are compared with those considering the �
�juðtÞj
�
juðtÞjmax �
uniform seismic excitation. The uniform earthquake excitation is R¼ max
(29)
juðtÞjmax
applied by setting the earthquake motion at water-soil interface on each
node of pile in soil. The error of the result considering uniform seismic
where juðtÞjmax and juðtÞjmax are the peak values of time histories of
excitation relative to that considering non-uniform one is also given at
structural displacements obtained by using all modes and without the j-
the location where maximum response occurs. It can be seen from
th mode, respectively.
Fig. 11 that the displacement results obtained from considering non-
The first five natural frequencies of MOWT obtained from Equation
uniform and uniform seismic excitations are slightly different, while
(19) are listed in Table 2 for the two sites. The contribution ratios of the
the acceleration and bending moment results are very different. The
first five modes are shown in Fig. 13 at the top (z ¼ 117.6 m) and the
maximum response errors of the acceleration and bending moment are
middle (z ¼ 77.6 m) of tower. It is considered that the j-th mode can be
more than 57% and 99%, respectively, indicating that the non-uniform
neglected in the subsequent analysis if its contribution ratio is less than
seismic excitation should be considered in the seismic analysis of MOWT
1%. The level of 1% is shown as the dotted line in the figure. It can be
especially when the structural acceleration and bending moment re­
seen from Fig. 13 that the first mode has the biggest contribution ratio,
sponses are concerned.
and the second and third modes have relatively significant contribu­
tions. The other modes have the contribution ratios of less than 1%. Only
4.3. Seismic analysis of MOWT using RSM
the first three modes are therefore used in RSM in the following studies.

Seismic responses of the 5-MW MOWT on two types of sites are

4.3.2. Accuracy of the RSM under response spectra of earthquake motions
calculated using the RSM in this section. The pile-water and pile-soil
The accuracy of the RSM for seismic responses of the 5-MW MOWT is
interactions, and the non-uniform seismic excitation are considered in

Fig. 13. Contribution ratio of j-th mode on accuracy of mode superposition method for (a) Site 1 and (b) Site 2 under seven artificial earthquake motions.

11
M. Zhao et al. Soil Dynamics and Earthquake Engineering 136 (2020) 106212

Fig. 14. Peak values of structural displacements varying along z coordinate. Two sites and seven artificial earthquake records are used. Accuracy of RSM is studied by
comparing with time history analysis.

studied in this subsection. The fourteen artificial earthquake records for values of the structural responses of the 5-MW MOWT are analyzed
two sites shown in Fig. 12 are chosen for the RSM and time history using RSM under design response spectrum. Using the method in Sub­
analysis. The response spectrum under each earthquake record is section 1.2, the design response spectra in soil are obtained from the
computed by solving Equation (22). given one at the water-soil interface shown in Fig. 12.
Figs. 14–16 shows the peak values of absolute horizontal displace­ Fig. 17 shows the peak values of absolute horizontal displacement
ment, absolute horizontal acceleration and section bending moment of and section bending moment of the whole structure varying along z
the whole structure varying along z coordinate, respectively. In these coordinate, respectively. The results using RSM under design response
figures, the structure is considered in the above two site conditions spectra are compared with mean value of the solutions obtained from
subjected to above seven artificial earthquake records. The results are time history analysis under seven artificial earthquake motions shown in
compared with those obtained from the time history method. The error Fig. 12. The relative difference between the results obtained from RSM
of the result obtained from RSM relative to that obtained from time and time history method is given at the location where the maximum
history method is given at the location where the maximum response response occurs. It can be seen from Fig. 17 that the maximum response
occurs. It can be seen from the three figures that the 7-record mean value differences of the displacement and section bending moment are less
of response errors of the displacement, acceleration and section bending than 5.5% and 3.8% respectively for Site 1, and less than 13.9% and
moment results are 10.7%, 14.2% and 9.2% respectively for Site 1, and 6.8% respectively for Site 2. Above results indicate the multiple-support
are 7.3%, 16.4% and 17.8% respectively for Site 2. Above results indi­ RSM is a suitable tool to seismic analysis and design of MOWT consid­
cate the multiple-support RSM is a suitable tool to calculate the seismic ering the pile-water and pile-soil interactions as well as the seismic
response of MOWT considering the pile-water and pile-soil interactions excitation non-uniformity.
as well as the seismic excitation non-uniformity (see Fig. 16) (see
Fig. 15). 5. Conclusions

4.3.3. Accuracy of the RSM under design response spectra Seismic analysis of a 5-MW MOWT considering pile-water and pile-
The RSM has advantage of using design response spectrum. The peak soil interactions subjected to non-uniform earthquake excitation is

12
M. Zhao et al. Soil Dynamics and Earthquake Engineering 136 (2020) 106212

Fig. 15. Peak values of structural accelerations varying along z coordinate. Two sites and seven artificial earthquake records are used. Accuracy of RSM is studied by
comparing with time history analysis.

investigated in this paper. The time history analysis is used to study the response errors may be more than 73% and 57% for acceleration,
effect of seismic damping force, pile-water interaction, pile-soil inter­ and 33% and 99% for bending moment.
action and seismic excitation non-uniformity on structural displace­ (2) The effectiveness of the multiple-support RSM is discussed by
ment, acceleration and bending moment. Subsequently, the accuracy of using seven artificial earthquake motions and design response
the multiple-support RSM for seismic responses of a 5-MW MOWT spectrum. Compared with the results obtained from time history
considering the pile-water interaction, pile-soil interaction and seismic method, the 7-record mean values of the maximum response er­
excitation non-uniformity is studied. The conclusions based on the ob­ rors of the displacement, acceleration and section bending
tained results can be drawn as follows. moment results obtained using RSM are 10.7%, 14.2% and 9.2%
respectively for Site 1, and are 7.3%, 16.4% and 17.8% respec­
(1) The effects of several factors on structural responses are studied tively for Site 2. Compared with the mean results obtained from
by time history analysis under three earthquake motions. The time history method under seven artificial earthquake motions,
damping force has less effect on the structural displacement, ac­ the differences of the displacement and section bending moment
celeration and section bending moment, and the maximum results obtained using RSM under design response spectrum are
response errors are all less than 5.3%, 6.3% and 8.3%, respec­ less than 5.5%, and 3.8% respectively for Site 1, and are less than
tively. Therefore, the damping forces caused by seismic excita­ 13.9% and 6.8% respectively for Site 2. Above results indicate the
tions are neglected in the present study. The pile-water multiple-support RSM is a suitable tool to seismic analysis and
interaction has significant effect on the structural displacement design of MOWT considering the pile-water and pile-soil in­
but rather less effect on the structural acceleration and section teractions as well as the seismic excitation non-uniformity. Be­
bending moment. The maximum response errors of displacement sides, only the first three modes are used in RSM for high
may be more than 31%, while the maximum response errors of efficiency and accuracy.
acceleration and bending moment are less than 6% and 11%,
respectively. The pile-soil interaction and seismic excitation non- It should be noted that the above conclusions are drawn based on the
uniformity have less effect on the displacement but rather larger models and examples adopted in this paper. More reasonable and ac­
effect on the acceleration and bending moment. Their maximum curate pile-water and pile-soil interaction models need be further stud­
ied specially and deeply, which would be a more difficult research topic.

13
M. Zhao et al. Soil Dynamics and Earthquake Engineering 136 (2020) 106212

Fig. 16. Peak values of section bending moments varying along z coordinate. Two sites and seven artificial earthquake records are used. Accuracy of RSM is studied
by comparing with time history analysis.

Declaration of competing interest

The authors declare that they have no known competing financial

interests or personal relationships that could have appeared to influence
the work reported in this paper.

CRediT authorship contribution statement

Mi Zhao: Conceptualization, Methodology, Writing - review &

editing, Resources, Investigation. Zhidong Gao: Data curation, Writing -
original draft, Software. Piguang Wang: Investigation, Supervision,
Visualization. Xiuli Du: Methodology, Conceptualization, Supervision.

Acknowledgements

This work described in this paper is supported by National Natural

Science Foundation of China (51678015 and 51421005) and Ministry of
Fig. 17. Peak values of structural displacement and section bending moment
Education Innovation Team of China (IRT_17R03). Opinions and posi­
varying along z coordinate. The results using RSM under design response
tions expressed in this paper are those of the authors only and do not
spectra are compared with mean value of the solutions obtained from time
history analysis under seven artificial earthquake motions.
reflect those of sponsors.

14
M. Zhao et al. Soil Dynamics and Earthquake Engineering 136 (2020) 106212

References [22] Code for seismic design of urban rail transit structures (GB 50909-2014). Beijing:
China Planning Press; 2014 [in Chinese)].
[23] International ICC. Building code (IBC). Falls Church, VA, USA: International Code
[1] Zheng XY, Li H, Rong W, Li W. Joint earthquake and wave action on the monopile
Council; 2003.
wind turbine foundation: an experimental study. Mar Struct 2015;44:125–41.
[24] Zhao M, Gao ZD, Du XL, Wang JJ, Zhong ZL. Response spectrum method for
[2] Kuo YS, Achmus M, Abdel-Rahman K. Minimum embedded length of cyclic
seismic soil-structure interaction analysis of underground structure. Bull Earthq
horizontally loaded monopoles. J Geotech Geoenviron Eng-ASCE 2012;138:
Eng 2019;17(9):5339–63.
357–63.
[25] IEC. Wind turbines. Part 3: design requirements for offshore wind turbines.
[3] Corciulo S, Zanoli O, Pisan� o F. Transient response of offshore wind turbines on
Geneva, Switzerland: International Electrotechnical Commission; 2009. p. 1. IEC
monopiles in sand: role of cyclic hydro-mechanical soil behavior. Comput Geotech
61400-3 (ed.
2017;83:221–38.
[26] Guideline GL. For the certification of offshore wind turbines. Hamburg, Germany:
[4] Goyal A, Chopra A. Simplified evaluation of added hydrodynamic mass for intake
Germanischer Lloyd; 2012.
towers. J Eng Mech 1989;115(7):1393–412.
[27] DNV. Design of offshore wind turbine structures. DNV-OS-J101. Copenhagen,
[5] Zaaijer MB. Foundation modelling to assess dynamic behavior of offshore wind
Denmark: Det Norske Veritas; 2013.
turbine. Appl Ocean Res 2006;28:45–57.
[28] Stamatopoulos GN. Response of a wind turbine subjected to near-fault excitation
[6] Adhikari S, Bhattacharya S. Vibrations of wind-turbines considering soil-structure
and comparison with Greek aseismic code provisions. Soil Dynam Earthq Eng
interaction. Wind Struct 2011;14(2):85–112.
2013;46:77–84.
[7] Adhikari S, Bhattacharya S. Dynamic analysis of wind turbine towers on flexible
[29] Penzien J, Scneffey CF, Parmelee RA. Seismic analysis of bridges on long piles.
foundations. Shock Vib 2012;19:37–56.
J Eng Mech Div 1964;90(3):223–54.
[8] Bhattacharya S, Adhikari S. Experimental validation of soil-structure interaction of
[30] Code for design of ground base and foundation of highway bridges and culverts
offshore wind turbines. Soil Dynam Earthq Eng 2011;31:805–16.
(JTG D63-2007). Beijing: China Communication Press; 2007 [in Chinese)].
[9] Feyzollahzadeh M, Mahmoodi MJ, Yadavar-Nikravesh SM, Jamali J. Wind load
[31] Clough RW, Penzien J. Dynamics of structures. second ed. New York: McGraw-Hill
response of offshore wind turbine towers with fixed monopile platform. J Wind Eng
Inc; 1993.
Ind Aerod 2016;158:122–38.
[32] Liu GH, Lian JJ, Liang CL, Zhao M. An effective approach for simulating multi-
[10] Raffaele DR, Subhamoy B, Katsuichiro G. Seismic performance assessment of
support earthquake underground motions. Bull Earthq Eng 2017;15(11):4635–59.
monopile-supported offshore wind turbines using unscaled natural earthquake
[33] Schnabel PB, Lysmer J, Seed HB. SHAKE: a computer program for earthquake
records. Soil Dynam Earthq Eng 2018;109:154–72.
response analysis of horizontally layered sites. Berkeley: University of California;
[11] Mo RJ, Kang HG, Li M, Zhao XL. Seismic fragility analysis of monopile offshore
1972. Report No. UCB/EERC-72/12.
wind turbines under different operational conditions. Wenergies 2017;10. https://
[34] Gasparini DA, Vanmarcke EH. Simulated earthquake motions compatible with
doi.org/10.3390/en10071037.
prescribed response spectra. Department of Civil Engineering. Cambridge, MA:
[12] Bisoi S, Haldar S. Dynamic analysis of offshore wind turbine in clay considering
Massachusetts Institute of Technology; 1976. Research Report R76-4.
soil-monopile-tower interaction. Soil Dynam Earthq Eng 2014;63:19–35.
[35] Afra H, Pecker A. Calculation of free field response spectrum of a non-
[13] Bisoi S, Haldar S. Design of monopile supported offshore wind turbine in clay
homogeneous soil deposit from bed rock response spectrum. Soil Dynam Earthq
considering dynamic soil-structure-interaction. Soil Dynam Earthq Eng 2015;73:
Eng 2002;22(2):157–65.
103–17.
[36] Liu GH, Lian JJ, Liang C, Li G, Hu JJ. An improved complex multiple-support
[14] Dong HK, Sang GL, Il KL. Seismic fragility analysis of 5 MW offshore wind turbine.
response spectrum method for the non-classically damped linear system with
Renew Energy 2014;65:250–6.
coupled damping. Bull Earthq Eng 2016;14(1):161–84.
[15] Wystan Carswell1, , Arwade1 Sanjay Raja, DeGroot1 Don J, Lackner Matthew A.
[37] Yu RF, Zhou XY. Response spectrum analysis for non-classically damped linear
Soil-structure reliability of offshore wind turbine monopile foundations. Wind
system with multiple-support excitations. Bull Earthq Eng 2008;6(2):261–84.
Energy 2015;18:483–98.
[38] Der Kiureghian A, Neumnhofer A. A response spectrum method for multiple-
[16] Versteijlen WG, Metrikine AV, van Dalen KN. A method for identification of an
support seismic excitation. Report No. USB/EERC-91/08. College of Engineering,
effective Winkler foundation for largediameter offshore wind turbine support
University of California at Berkeley; 1991.
structures based on in-situ measured small-strain soil response and 3D modelling.
[39] Der Kiureghian A, Neumnhofer A. A response spectrum method for multiple-
Eng Struct 2016;124:221–36.
support seismic excitations. Earthq Eng Struct Dynam 1992;21:713–40.
[17] Zuo HR, Bi KM, Hao H. Dynamic analyses of operating offshore wind turbines
[40] Jonkman J, Butterfield S, Musial W, Scott G. Definition of a 5-MW reference wind
including soil-structure interaction. Eng Struct 2018;157:42–62.
turbine for offshore system development. Golden, CO: National Renewable Energy
[18] Wang PG, Zhao M, Du XL, Liu JB, Xu CS. Wind, wave and earthquake responses of
Laboratory; 2009 [R]. NREL/TP-500-38060.
offshore wind turbine on monopile foundation in clay. Soil Dynam Earthq Eng
[41] Zhao M, Zhang GL, Wang PG, et al. An accurate frequency-domain model for
2018;113:47–57.
seismic responses of breakwater-water-seabed-bedrock system. Ocean Eng 2020;
[19] Chopra AK. Dynamics of Structures: theory and applications to earthquake
197:106943.
engineering. fourth ed. Englewood Cliffs: Prentice Hall; 2011.
[42] Zhao FX, Zhang YS, Lv HS. Artifical gound motion compatible with specified
[20] BSL. The building standard law of Japan. Tokyo, Japan: The Ministry of
ground shaking peaks and target response spectrum. Earthq Eng Eng Vib 2006;5
Construction; 2000.
(1):41–8.
[21] Eurocode 8. Design of structures for earthquake resistance. European Committee
for Standardization; 2003.

15