You are on page 1of 34

Bull Earthquake Eng (2018) 16:1007–1040

https://doi.org/10.1007/s10518-017-0191-x

ORIGINAL RESEARCH PAPER

On time-domain uncoupled analyses for offshore wind


turbines under seismic loads

F. Santangelo1 • G. Failla1 • F. Arena1 • C. Ruzzo1

Received: 29 September 2016 / Accepted: 14 July 2017 / Published online: 30 August 2017
Ó Springer Science+Business Media B.V. 2017

Abstract For seismic assessment of wind turbines in seismically-active areas, Interna-


tional Standards and Guidelines allow the combination of two uncoupled analyses under
environmental and earthquake loads, respectively. The separate earthquake response is
generally computed including an additional aerodynamic damping in the structural model.
Although some work has been done to estimate the effectiveness of uncoupled analyses for
land-based wind turbines, and determine appropriate levels of aerodynamic damping, to
date no similar studies have been carried out for offshore wind turbines. This paper
assesses the accuracy of different time-domain implementations of uncoupled analyses for
offshore wind turbines, and investigates pertinent levels of aerodynamic damping. The case
study is a 5-MW wind turbine, resting on a tripod in intermediate waters.

Keywords Offshore wind turbine  Seismic response  Uncoupled analyses  Aerodynamic


damping  Fully-coupled analysis

1 Introduction

Bottom-fixed, offshore horizontal-axis wind turbines (HAWTs) are typically installed in


water depths ranging from shallow (\30 m) to intermediate (30–60 m) ones. International
Standards and Guidelines, such as IEC 61400-3 (IEC 2009), GL 2012 (GL 2012) and
DNV-OS-J101 (DNV 2013), recommend a seismic assessment of bottom-fixed offshore
HAWTs when installed in seismically-active areas. This has become a subject of particular
interest, considering that there exist several sites at medium-to-high seismic risk, with high
wind resources and water depth up to 60 m (Failla and Arena 2015; Schwartz et al. 2010;
USGS 2008).

& G. Failla
giuseppe.failla@unirc.it
1
Department of Civil, Energy, Environmental and Materials Engineering (DICEAM), University of
Reggio Calabria, 89124 Reggio Calabria, Italy

123
1008 Bull Earthquake Eng (2018) 16:1007–1040

Recent studies have investigated the seismic response of offshore HAWTs mounted on
either monopiles (Haciefendioğlu 2012; Kim et al. 2014) or support structures including a
tripod or jacket (Alati et al. 2015). Simplified or full system models have been used: the
first involves the support structure only, with the rotor-nacelle assembly (RNA) modelled
as a lumped mass at the tower top (Haciefendioğlu 2012; Kim et al. 2014); the second
includes the support structure and whole turbine, i.e. rotor blades, nacelle, as well as
mechanical/electrical/control turbine components (Alati et al. 2015). Earthquakes striking
in the rotor parked state (Haciefendioğlu 2012; Kim et al. 2014; Alati et al. 2015) or
operating conditions (Alati et al. 2015) have been considered. In particular, Alati et al.
(2015) have run time-domain, fully-coupled simulations on full system models of the
5-MW baseline HAWT of the National Renewable Energy Laboratory (NREL) (Jonkman
et al. 2009), mounted on a tripod and a jacket in intermediate waters, for a large set of
earthquakes. A comprehensive review on seismic analysis of offshore HAWTs and related
issues may be found in a recent paper by Katsanos et al. (2016).
Time-domain, fully-coupled simulations capture the mutual dependence of aerody-
namic, hydrodynamic and seismic responses. Indeed, tower top oscillations due to wave
loads and earthquake ground motion affect rotor aerodynamics, in particular the relative
wind speed at the blades depending on which aerodynamic lift and drag forces are cal-
culated. On the other hand earthquake ground motion, and wind loads to a certain extent,
affect the support structure velocities and, consequently, the hydrodynamic loads. Com-
putational costs of fully-coupled simulations, however, are significant and may become
almost prohibitive when several analyses have to be implemented for different environ-
mental states and system parameters, as in the early stages of design. Another limitation is
that fully-coupled simulations require a dedicated software package, capable of handling
the inherent interactions between aerodynamic, hydrodynamic and seismic responses. For
these reasons, Standards and Guidelines allow uncoupled analyses, where the responses to
wind-wave loads and earthquake loads are computed separately, and then linearly com-
bined (IEC 2009; GL 2012; DNV 2013). In uncoupled analyses on offshore HAWTs, the
separate earthquake response may be computed by a response spectrum approach (IEC
2009; GL 2012; DNV 2013) or in time domain (GL 2012), in agreement with similar
prescriptions for land-based HAWTs (IEC 2005; GL 2010; ASCE/AWEA 2011).
It is now well understood that, for earthquake striking a wind turbine in operating
conditions, either offshore or land-based, fore-aft oscillations of the tower top in wind
direction, induced by earthquake loads, significantly affect the instantaneous thrust force
and produce indeed damping effects. These damping effects are intrinsically considered in
the fully-coupled response (Asareh and Prowell 2012; Witcher 2005), as fully-coupled
simulations capture the instantaneous mutual dependence between the responses to envi-
ronmental and earthquake loads. On the contrary, these damping effects must be properly
incorporated in the structural model when implementing uncoupled analyses, because the
latter compute separately the responses to environmental and earthquake loads. This is
generally made by adding a certain amount of damping in the fore-aft support structure
modes of the structural model, when computing the separate earthquake response only
(Asareh and Prowell 2012; Asareh and Volz 2013; Asareh et al. 2016; Failla 2014; Prowell
et al. 2009, 2014; Santangelo et al. 2016; Valamanesh and Myers 2014; Witcher 2005).
The additional damping is referred to as aerodynamic damping.
Much work has already been done to assess appropriate levels of aerodynamic damping
for uncoupled analyses on land-based HAWTs. According to ASCE/AWEA RP2011 and
several studies, a 4% value proves sufficiently accurate in various operating conditions of

123
Bull Earthquake Eng (2018) 16:1007–1040 1009

land-based HAWTs (ASCE/AWEA 2011; Asareh and Prowell 2012; Asareh and Volz
2013; Santangelo et al. 2016; Valamanesh and Myers 2014; Witcher 2005).
A 4% aerodynamic damping is implicitly suggested also by Standards and Guidelines
for uncoupled analyses on offshore HAWTs (IEC 2009; GL 2012; DNV 2013). Indeed, the
standard response-spectrum approach allowed in IEC 61400-3 (IEC 2009), GL 2012 (GL
2012) and DNV-OS-J101 (DNV 2013) relies, in general, on a 5% design spectrum.
Considering that structural steel damping ratio amounts generally to 1%, this corresponds
to assume a 4% aerodynamic damping when computing the separate earthquake response
(Asareh and Prowell 2012; Asareh and Volz 2013; ASCE/AWEA 2011; Santangelo et al.
2016; Valamanesh and Myers 2014; Witcher 2005).
To the best of authors’ knowledge, however, so far no detailed study has been carried
out to assess the accuracy of uncoupled analyses on offshore HAWTs when a 4% aero-
dynamic damping is assumed. As previously noted, aerodynamic damping is added to the
structural model in the uncoupled-analyses approach, in order to account for damping
effects resulting from the influence of earthquake loads on the instantaneous thrust force.
However, since not only earthquake loads but also wave loads may have an influence on
the instantaneous thrust force in offshore HAWTs, the question arises whether, and to what
extent, the 4% aerodynamic damping used for land-based HAWTs is accurate also for
offshore HAWTs. Another limitation of existing literature on offshore HAWTs is that no
comprehensive study has been carried out on the most appropriate implementation of time-
domain uncoupled analyses allowed by GL 2012 (GL 2012), for which no specific indi-
cations are available yet. A time-domain approach, however, is certainly appealing because
no assumption is required on modal combinations to be adopted as, instead, in a response-
spectrum approach.
The purpose of this paper is to investigate various implementations of time-domain
uncoupled analyses allowed by GL 2012 (GL 2012) for seismic assessment of offshore
HAWTs, and assess pertinent levels of aerodynamic damping. The case study is the NREL
5-MW baseline HAWT (Jonkman et al. 2009), mounted on a typical tripod support structure
in intermediate water depth. For earthquake striking in operating conditions, fully-coupled
and uncoupled analyses are run on a full model of the system implemented in GH-BLADED
(Bossanyi 2000). Considering various wind-wave states and earthquake records, fully-
coupled simulation is compared with two different methods of uncoupled analyses:
(i) Method 1—Demands are obtained as maxima values from a time history built by
linearly superposing separately-computed time histories of wind-wave response
and earthquake response, respectively. The superposition of time histories is made
from the time instant at which earthquake shaking is triggered in the fully-coupled
simulation.
(ii) Method 2—Demands are obtained by summing separate maxima from wind-wave
response and earthquake response.
In both methods, the separate earthquake response is computed by adding different
levels of aerodynamic damping in the first two fore-aft support structure modes, consis-
tently with previous studies on land-based HAWTs (Dı́az and Suárez 2014; Prowell et al.
2014).
The study will involve two steps. The first step will assess whether there exists an
aerodynamic damping value minimizing a total error in terms of shear-force and bending-
moment demands at the tower base, as computed by fully-coupled simulation and the two
methods of uncoupled analyses. Once it will be established that such aerodynamic damping
value cannot be obtained for all environmental states and earthquake records, in the second

123
1010 Bull Earthquake Eng (2018) 16:1007–1040

step errors in bending-moment and shear-force demands will be computed assuming a 4%


aerodynamic damping, showing range of applicability and accuracy of the two methods of
uncoupled analyses.
The paper is organized in four sections. The test structure is presented in Sect. 2. Time-
domain fully-coupled and uncoupled analyses are described in Sect. 3. Numerical results
are discussed in Sect. 4.

2 Test structure

The turbine is the NREL 5-MW three-bladed HAWT (Jonkman et al. 2009), generally
considered, in recent literature, as representative of modern multi-megawatt wind turbines
(Agarwal and Manuel 2011; Asareh et al. 2016; Nguyen and Manuel 2015). The support
structure is a steel centre-column tripod shown in Fig. 1, designed according to current
practice (Alati et al. 2015), with details on structural members given in Fig. 2. Steel
parameters for the support structure are: Young’s modulus = 210 GPa, Poisson coeffi-
cient = 0.3, Mass density = 7850 kg/m3. Rotor blades are made of glass-fiber composite
material; pertinent elastic properties are provided at various stations along the blade in the
reference dataset (Jonkman et al. 2009), and are not reported here for brevity. It is assumed
that the water depth is 50 m.
The full system is implemented in GH-BLADED (Bossanyi 2000), a software package
validated by Germanischer-Lloyd for analysis and certification of offshore HAWTs. Rotor
blades, nacelle, drive train, mechanical/electrical/control components are modelled as
described in the reference dataset (Jonkman et al. 2009), for a full representation of wind
turbine aerodynamics. Beam elements are used for blades and structural members of the
support structure. The structure is assumed to be fixed at the mudline (-50 m).

2.1 Modal analysis

For a first insight, Table 1 reports the modal frequencies obtained by GH-BLADED
(Bossanyi 2000), considering the rotor in a parked state (one blade upward, two blades
downward). Fore-aft and side-to-side directions correspond to x and y directions in Fig. 1,
respectively.
Frequencies in Table 1 agree with those in the literature (Prowell et al. 2010). Shapes of
first and second fore-aft support structure modes are reported in Fig. 3; those in side-to-
side direction are similar and are not reported for brevity.
In the time-domain simulations described in Sect. 3, modal damping ratios are assumed
in agreement with previous studies (Jonkman et al. 2009; Dı́az and Suárez 2014; Prowell
et al. 2014), and specifically:
(i) Fully-coupled simulations: 0.4775% for blade modes and 1% for support structure
modes.
(ii) Uncoupled analyses: 0.4775% for blade modes and 1% for support structure
modes, to compute the separate response to wind-wave loads; 0.4775% for blade
modes, while the 1% modal damping ratio of the first two fore-aft support structure
modes will be increased to include aerodynamic damping, when computing the
separate response to earthquake loads.
Fully-coupled as well as uncoupled analyses assume that all system components behave
elastically. Indeed, offshore HAWTs are generally designed to withstand operational/

123
Bull Earthquake Eng (2018) 16:1007–1040 1011

Fig. 1 Tripod support structure

extreme loads without undergoing plastic strains (GL 2012), as they could affect stability
and, ultimately, power production.

2.2 Wind, wave, earthquake loads

It is assumed that wind and waves act both in x direction, see Fig. 1. Samples are generated
in GH-BLADED (Bossanyi 2000) based on Kaimal and JONSWAP spectra, respectively.
For wind loads, the Kaimal spectrum is given as (IEC 2005):

123
1012 Bull Earthquake Eng (2018) 16:1007–1040

Fig. 2 Geometry of tripod support structure (dimensions of structural members in mm; heights and depths
in m)

4r2k Lk =V
Sk ð f Þ ¼ ð1Þ
ð1 þ 6fLk =V Þ5=3
where f is the frequency (Hz), V is the wind velocity at hub height, k is the index referring
to the velocity component (1 = x direction, 2 = y direction and 3 = z direction), rk is the
standard deviation and Lk is the integral scale parameter of each velocity component. In
GH-BLADED (Bossanyi 2000), the aerodynamic loads on the spinning rotor are generated

123
Bull Earthquake Eng (2018) 16:1007–1040 1013

Table 1 Natural frequencies


Mode description Freq. (Hz)

1st support structure side-to-side 0.309


1st support structure fore-aft 0.311
1st blade asymmetric flapwise yaw 0.645
1st blade asymmetric flapwise pitch 0.677
1st blade collective flap 0.710
1st blade asymmetric edgewise pitch 1.081
1st blade asymmetric edgewise yaw 1.097
2nd blade asymmetric flapwise yaw 1.749
2nd blade asymmetric flapwise pitch 1.848
2nd blade collective flap 1.996
2nd support structure fore-aft 2.206
2nd support structure side-to-side 2.277

Fig. 3 First and second fore-aft


support structure modes

from a dynamic wake model for the axial inflow, in conjunction with classical Blade-
Element-Momentum model for the tangential inflow (Manwell et al. 2010). Wind loads on
the tower are included.
For wave loads, the JONSWAP spectrum is (Hasselmann et al. 1973; Bossanyi 2000):
 5
f 4
2
SJS ð f Þ ¼ a2 Hs Tp e1:25ðf =fp Þ  cb ð2Þ
fp

123
1014 Bull Earthquake Eng (2018) 16:1007–1040

with Tp and Hs wave period and significant wave height, fp = 1/Tp, c, a2, b and r (IEC
2009):
8 pffiffiffiffiffi
>
<5 Tp Hs  3:6
pffiffiffiffi pffiffiffiffiffi
c ¼ eð5:751:15Tp = Hs Þ 3:6  Tp Hs  5:0 ð3Þ
>
:1 pffiffiffiffiffi
T H  5:0
p s

"  2 #
0:0624 0:5 f
a2 ¼ ; b ¼ exp  2 1 ;r
0:230 þ 0:0336c  0:185=ð1:9 þ cÞ r fp

0:07 f  fp
¼ ð4acÞ
0:09 f [ fp

In GH-BLADED (Bossanyi 2000), the hydrodynamic loads on the structural members


are computed based on Morison’s equation (Chakrabarti 1987; Bossanyi 2000), with drag
and inertia coefficients set according to DNV-OS-J101 recommendations (DNV 2013).
Earthquake ground motion is modelled as an acceleration at the base with two hori-
zontal components in x and y directions.
In GH-BLADED (Bossanyi 2000), the equations of motion are derived based on a
multi-body dynamics approach combined with a modal representation of the flexible
components, like blades and support structure. Large displacements of the rotor with
respect to the support structure are accounted for. The equations of motion are numerically
integrated in the time-domain by a variable-step Runge–Kutta method.

3 Fully-coupled and uncoupled analyses

In this study, comparisons between fully-coupled and uncoupled analyses are carried out
for earthquake striking in operating conditions, i.e. while the rotor is spinning. All analyses
are implemented in the time domain, using GH-BLADED (Bossanyi 2000) as explained
below.
Fully-coupled analyses are carried out by numerical integration of the equations of
motion, considering mutual interactions of aerodynamic, hydrodynamic and seismic
responses. That is, aerodynamic loads on the rotor blades are built taking into account
blade motion due to global rotor motion and blade flexibility, as induced by wind loads and
control system, wave loads, earthquake shaking at the base. This means that fully-coupled
simulations capture the influence of earthquake loads on the instantaneous thrust force, and
related damping effects are intrinsic in the fully-coupled response. Thus, no additional
aerodynamic damping must be incorporated in the structural model when implementing
fully-coupled simulations (Asareh and Prowell 2012; Witcher 2005). The ground motion
starts at t0 = 400 s into the simulation, to ensure that the earthquake occurs as the system
response has already attained a steady state and initial transient behaviour has completely
disappeared (Asareh et al. 2016; Prowell et al. 2010; Prowell 2011). After t0 = 400 s, the
simulation runs until the end of the earthquake record. Since the longest earthquake record
in this study lasts about 70 s (Chi–Chi Taiwan record, see Table 2), a total simulation
length equal to 470 s is chosen and used, for simplicity, for all earthquake records (Asareh
et al. 2016; Prowell et al. 2010). Stress-resultant demands are computed as maxima SRSS
bending moment and shear force (SRSS = square root of the sum of the squares), i.e.:

123
Bull Earthquake Eng (2018) 16:1007–1040 1015

Table 2 Earthquake records, wind velocities at the hub


Earthquake records Year Station PGA (m/s2)

Wind-wave and earthquake loading


Cape Mendocino (CM) 1992 Petrolia 6.12
Chi Chi Taiwan (CC) 1999 TCU102 2.17
Ducze—Turkey (D) 1999 Ducze 4.23
Erzican—Turkey (E) 1992 Erzican 4.43
Imperial Valley-06 (IV) 1979 E.C. #3 2.44
Kobe—Japan (K) 1995 KJMA 6.71
Loma Prieta (LP) 1989 F.C.—APEEL 1 2.75
N. Palm Springs (NP) 1986 N.P.S 6.30
Northridge (N) 1994 P.D.d. 3.43
Superstition Hills – 01 (S) 1987 I.V. W.L.A. 1.50

Wind velocities at hub, V Wind samples

11.4 m/s 5
15 m/s 5
20 m/s 5

Wave periods and significant height Wave samples

Hs = 5 m; Tp = 9.5 s 2
Hs = 6 m; Tp = 11 s 2

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2
Mr ¼ max ðMx ðtÞÞ2 þ My ðtÞ ; t  400 s ð5Þ

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2
Fr ¼ max ðFx ðtÞÞ2 þ Fy ðtÞ ; t  400 s ð6Þ

where subscript ‘‘r’’ means resultant, while Mx , My , Fx and Fy are bending moments and
shear forces in x and y directions (see Fig. 1).
Uncoupled analyses allowed by GL 2012 (GL 2012) are implemented as follows. Wind-
wave response and earthquake response are computed separately. For wind-wave excita-
tion the rotor is spinning, while for earthquake excitation the rotor is considered in a parked
state. Aerodynamic damping is included when computing the separate response to earth-
quake only. In particular, consistently with numerical evidence on the seismic response of
land-based HAWTs (Dı́az and Suárez 2014; Prowell et al. 2014), the additional aerody-
namic damping is included in the first two fore-aft support structure modes in Table 1 by
increasing the pertinent structural modal damping ratios, set equal to 1% in Sect. 2.1. For
instance, selecting a 4% aerodynamic damping means that a (1 ? 4%) = 5% modal
damping ratio will be considered for the first two fore-aft support structure modes. Bending
moment and shear force in x and y directions, obtained from the separate analyses, are
denoted as:
Wind-wave response: MxðWW Þ ðtÞ; MyðWW Þ ðtÞ; FxðWW Þ ðtÞ; FyðWW Þ ðtÞ

123
1016 Bull Earthquake Eng (2018) 16:1007–1040

Earthquake response: MxðEÞ ðtÞ; MyðEÞ ðtÞ; FxðEÞ ðtÞ; FyðEÞ ðtÞ

Then, two different methods of combination will be adopted, referred to as ‘‘Method 1’’
and ‘‘Method 2’’, as detailed in the following.

3.1 Method 1

The idea is that demands are computed from a combined response built by summing wind-
wave response and earthquake response from the time instant at which the earthquake
occurs in the fully-coupled simulation, i.e. t0 = 400 s. The simulation length for wind-
wave response is 470 s, i.e. identical to the length of the fully-coupled simulation, while
the simulation length for earthquake response is 70 s, used for all earthquake records
(70 s = duration of the longest earthquake record, i.e. Chi–Chi Taiwan, in Table 2). Thus,
the time histories of the separate wind-wave and earthquake responses are summed starting
from t0 = 400 s and, from the combined response after t0 = 400 s, stress-resultant
demands are computed as maxima SRSS bending moment and shear force:
(rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi )

2
2
ð1Þ ð1 Þ ð1 Þ
Mr ¼ max Mx ðtÞ þ My ðtÞ ; t  400 s ð7Þ

(rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi )

2
2
ð1 Þ ð1Þ
Frð1Þ ¼ max Fx ðtÞ þ Fy ðtÞ ; t  400 s ð8Þ

where

Mxð1Þ ðtÞ ¼ MxðWW Þ ðtÞ þ MxðEÞ ðtÞ; Myð1Þ ðtÞ ¼ MyðWW Þ ðtÞ þ MyðEÞ ðtÞ ð9a; bÞ

Fxð1Þ ðtÞ ¼ FxðWW Þ ðtÞ þ FxðEÞ ðtÞ; Fyð1Þ ðtÞ ¼ FyðWW Þ ðtÞ þ FyðEÞ ðtÞ ð10a; bÞ

Notice that Method 1 keeps track of the actual time instant at which the earthquake
occurs in the reference fully-coupled simulation, mirroring a recent time-domain uncou-
pled-analyses approach for seismic assessment of land-based HAWTs (Santangelo et al.
2016).
Superscript (1) in Eqs. (7)–(8) will distinguish results from Method 1.

3.2 Method 2

Demands are computed by linearly combining the two maxima of the separate wind-wave
response and earthquake response, respectively. In accordance with standard procedures to
simulate normal operating conditions of wind turbines, the separate wind-wave response is
computed from a 600-s steady-state simulation (Ishihara et al. 2012; Ernst and Seume
2012; Leu et al. 2014; Prowell 2011; Resor 2013), obtained from a 1000-s simulation
discarding the first 400 s to allow dissipation of transient behaviour (Prowell 2011). The
simulation length for earthquake response is again 70 s, used for all earthquake records.
Stress-resultant demands are computed as follows:


Mrð2Þ ¼ k M rðWW Þ þ M
 rðEÞ ð11Þ

123
Bull Earthquake Eng (2018) 16:1007–1040 1017



Frð2Þ ¼ k FrðWW Þ þ FrðEÞ ð12Þ

where M  rðWW Þ ; M
 rðEÞ and FrðWW Þ ; FrðEÞ denote the maxima SRSS bending moments and shear
forces due to separate wind-wave loads and earthquake loads, respectively, i.e.
n o n o
M rðWW Þ ¼ max MrðWW Þ ðtÞ; 400  t  1000 s ; M  rðEÞ ¼ max MrðEÞ ðtÞ; 0  t  70 s

ð13a; bÞ
n o n o
FrðWW Þ ¼ max FrðWW Þ ðtÞ; 400  t  1000 s ; FrðEÞ ¼ max FrðEÞ ðtÞ; 0  t  70 s
ð14a; bÞ
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi


ffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi



ð Þ 2 ð Þ 2 ð Þ 2 ðÞ 2
where MrðÞ ðtÞ ¼ Mx ðtÞ þ My ðtÞ and FrðÞ ðtÞ ¼ Fx ðtÞ þ Fy ðtÞ for both
wind-wave and earthquake responses (i.e. () = (E) or (WW)). Specifically, in Eqs. (11)–
(12) two possible combinations are considered, selecting k as:
k ¼ 1; k ¼ 0:75 ð15a; bÞ

The combination factor k ¼ 1 is recommended by GL 2012 (GL 2012) for uncoupled


analyses on offshore HAWTs, as well as IEC 61400-3 (IEC 2009) and IEC 61400-1 (IEC
2005) for uncoupled analyses on offshore and land-based HAWTs. On the other hand, the
combination factor k ¼ 0:75 is the one suggested by ASCE/AWEA RP2011 (ASCE/
AWEA 2011) for uncoupled analyses on land-based HAWTs and, in this study, is con-
sidered for comparison (Asareh and Prowell 2012).
It is worth noticing that the maxima earthquake responses (13b)–(14b) could alterna-
tively be computed by a response-spectrum approach. However, the relevant advantage of
a time-domain implementation is that, unlike in a response-spectrum approach, no choice
has to be made on the combination rule for modal maxima.
Superscript (2) in Eqs. (11)–(12) will distinguish results from Method 2.

4 Numerical results

Ten real earthquake records, three wind velocities at the hub and two sea states are
considered, as reported in Table 2.
Earthquake records in Table 2 feature different peak ground acceleration and quite
different frequency content. Horizontal components acting along x and y directions in
Fig. 1 coincide with first and second column data of the records as taken from Peer Ground
Motion Database (PEER 2013). Records are not scaled to represent a specific site hazard,
since the purpose here is to assess the accuracy of time-domain uncoupled analyses under a
variety of peak ground accelerations. In Table 2, they range from 1.50 to 6.71 m/s2.
Wind velocities in Table 2 are representative of potential operating conditions, within
the cut-in-cut-out wind velocity range of the NREL 5 MW HAWT [cut-in = 3 m/s and
cut-out = 25 m/s, rated speed = 11.4 m/s, see Jonkman et al. (2009)]. Medium turbulence
characteristics are assumed, with parameters in the Kaimal spectrum (1) set according to
IEC 61400-1 prescriptions for a normal turbulence model (IEC 2005). Wave parameters in
Table 2, to be used in the JONSWAP spectrum (1), are typical of ocean sites.

123
1018 Bull Earthquake Eng (2018) 16:1007–1040

For every earthquake record, 5 wind samples are generated for each wind velocity at the
hub and 2 wave samples for each sea state, totalling 10 9 (5 9 3) 9 (2 9 2) = 600
potential earthquake-wind-wave scenarios.
For the purposes of this study, the following errors in terms of stress-resultant demands
will be estimated at the tower base:
Mr  MrðkÞ
Mr error ¼ k ¼ 1; 2 ð16Þ
Mr

Fr  FrðkÞ
Fr error ¼ k ¼ 1; 2 ð17Þ
Fr

Mr  MrðkÞ Fr  FrðkÞ
Total error: þ
F
k ¼ 1; 2
ð18Þ
Mr r

where MrðkÞ and FrðkÞ , for k = 1, 2, are the maxima SRSS bending moment and shear force
at the tower base computed by Method 1 or Method 2, while Mr and Fr are the maxima
SRSS bending moment and shear force at the tower base obtained from the fully-coupled
simulation. In addition, SRSS bending moment and shear force from Method 1/Method 2
and fully-coupled simulation will be compared along the whole tower, as explained next.

4.1 Uncoupled analyses: Method 1

For some wind-wave realizations and earthquake records in Table 2, Fig. 4 shows errors
(16)–(18) obtained by Method 1, as aerodynamic damping in the separate earthquake
response varies within 0–8% at steps equal to 0.5%. In particular, the left column includes
the results for the Superstition Hills earthquake and all wind velocities (sample 3). The
results for all 600 earthquake-wind-wave scenarios in Table 2 are omitted for brevity.
The first observation is that ‘‘Mr error’’ (16) and ‘‘Fr error’’ (17) generally increase as
aerodynamic damping increases. This means that adding aerodynamic damping in the fore-
aft support structure modes reduces the earthquake responses MyðEÞ and FxðEÞ in wind
direction (fore-aft = x direction Fig. 1) and, consequently, the demands Mrð1Þ and Frð1Þ built
by Method 1, see Eqs. (7)–(8). In particular, notice that errors (16)–(17) become positive as
Mrð1Þ and Frð1Þ are smaller than demands Mr and Fr from fully-coupled simulation. Further
comments are:
(i) In some cases, an aerodynamic damping minimizing the total error (18) may be
found, typically when the ‘‘Mr error’’ (16) and/or the ‘‘Fr error’’ (17) start from
negative values for the lowest aerodynamic damping 0.5% and tend towards zero
with increasing aerodynamic damping. In these cases, Method 1 provides larger
demands than fully-coupled simulation for the lowest aerodynamic damping 0.5%
(negative errors (16)–(17)), and progressively approaches the fully-coupled
simulation as aerodynamic damping increases.
(ii) However, there may also be cases in which no minimum is found for the total error
(18), typically when both ‘‘Mr error’’ (16) and ‘‘Fr error’’ (17) start from positive
values for the lowest aerodynamic damping 0.5%, and increase as aerodynamic
damping increases. In these cases, demands from Method 1 are always smaller
than those from fully-coupled simulation, for any aerodynamic damping. This
result is not surprising, since a study on land-based HAWTs already showed that

123
Bull Earthquake Eng (2018) 16:1007–1040 1019

Fig. 4 Method 1—errors (16)–(18) for various potential aerodynamic damping values, under various
earthquake records and wind-wave realizations in Table 2

123
1020 Bull Earthquake Eng (2018) 16:1007–1040

the combination of separate wind and earthquake responses, the latter being
computed with inclusion of aerodynamic damping, may provide smaller bending-
moment demands with respect to fully-coupled simulations (Asareh and Prowell
2012). It can be explained considering that any aerodynamic-damping based
approach is indeed an approximate way to account for the inherent interaction
between wind-wave and earthquake responses.
At any rate, based on the results in Fig. 4, it must be concluded that an optimal
aerodynamic damping capable of minimizing the total error (18) cannot exist, for all
earthquake-wind-wave realizations in Table 2. Notice that the same conclusion was drawn
in a previous study, where Method 1 was applied for land-based HAWTs (Santangelo et al.
2016).
Once established that an optimal aerodynamic damping value, as defined above, cannot
exist, attention is focused on ‘‘Mr error’’ (16) and ‘‘Fr error’’ (17) at the tower base when, in
Method 1, the separate earthquake response is built with 4% aerodynamic damping. In this
case, Figs. 5, 6 and 7 show that errors (16)–(17) are generally below 15%, with a few
maxima values slightly below 30%, for all the considered earthquake records and wind-

(a)

(b)

Fig. 5 Method 1—errors (16)–(17) at the tower base for 4% aerodynamic damping, all earthquakes, wind
velocity V = 11.4 m/s and waves with a Hs = 5 m, Tp = 9.5 s; b Hs = 6 m, Tp = 11 s

123
Bull Earthquake Eng (2018) 16:1007–1040 1021

(a)

(b)

Fig. 6 Method 1—errors (16)–(17) at the tower base for 4% aerodynamic damping, all earthquakes, wind
velocity V = 15 m/s and waves with a Hs = 5 m, Tp = 9.5 s; b Hs = 6 m, Tp = 11 s

wave realizations in Table 2. The errors are within the range of alternative formulations of
uncoupled analyses in the literature, see the SRSS combination of separate wind and
earthquake responses computed by the IEC 61400-1 method for land-based HAWTs
(Prowell 2011), which provide errors around 23% depending on the pseudo spectral
acceleration. Errors (16)–(17) in Figs. 5, 6 and 7 may be negative or positive, i.e. Method 1
may provide conservative or non-conservative results with respect to fully-coupled sim-
ulation. This is consistent with previous results on land-based HAWTs, obtained by the
IEC 61400-1 method (Prowell 2011) or alternative formulations of uncoupled analyses
proposed by Asareh and Prowell (2012), which already showed that the combination of
uncoupled analyses may be non-conservative.
For a further insight into the results obtained with 4% aerodynamic damping, Figs. 8
and 9 show the mean of SRSS bending-moment and shear-force demands along the tower,
computed by fully-coupled simulation and Method 1, for all earthquake records, sea states,
wind velocities V = 11.4 m/s and V = 20 m/s in Table 2 (results for V = 15 m/s are
similar and omitted for brevity; the ordinate range in Figs. 8 and 9 start from the structural
node at the tower base, located at -11 m, i.e. at the intersection between the axis of the

123
1022 Bull Earthquake Eng (2018) 16:1007–1040

(a)

(b)

Fig. 7 Method 1—errors (16)–(17) at the tower base for 4% aerodynamic damping, all earthquakes, wind
velocity V = 20 m/s and waves with a Hs = 5 m, Tp = 9.5 s; b Hs = 6 m, Tp = 11 s

tower and the axes of the three bracings of the tripod, see Fig. 2). For each earthquake
record and wind-wave state, the mean is obtained by averaging the demands from all wind-
wave realizations. Figures 8 and 9 show that results from Method 1 and fully-coupled
simulation agree well along the whole tower, with errors within engineering margins
typically encountered in uncoupled analyses on land-based HAWTs, e.g. see bending-
moment errors found by Asareh and Prowell (2012). The maximum error at tower base is
about 23%, as reported in Table 3 (including, for completeness, also the results for
V = 15 m/s). Again, either negative or positive errors may be encountered along the tower
and at the tower base, i.e. Method 1 may provide conservative or non-conservative results
with respect to fully-coupled simulation.
Next, an important remark is that stress demands in system components are found well
below the elastic limit, consistently with general design of offshore HAWTs (GL 2012).
For the tower, this may be seen considering the geometry of tower cross sections in Fig. 2,
stress-resultant demands in Figs. 8 and 9, and including the contribution to normal stress
associated with the axial force (maximum = 6.5 MPa at the tower base through all time-

123
Bull Earthquake Eng (2018) 16:1007–1040 1023

(a)

(b)

Fig. 8 Method 1—mean demands for 4% aerodynamic damping, all earthquakes, wind velocity
V = 11.4 m/s and waves with a Hs = 5 m, Tp = 9.5 s; b Hs = 6 m, Tp = 11 s; left column shear force;
right column bending moment

123
1024 Bull Earthquake Eng (2018) 16:1007–1040

(a)

(b)

Fig. 9 Method 1—mean demands for 4% aerodynamic damping, all earthquakes, wind velocity V = 20 m/
s and waves with a Hs = 5 m, Tp = 9.5 s; b Hs = 6 m, Tp = 11 s; left column shear force; right column
bending moment

123
Bull Earthquake Eng (2018) 16:1007–1040 1025

Table 3 Method 1—errors in mean demands at tower base for 4% aerodynamic damping, all earthquakes,
wind velocities, and waves with (a) Hs = 5 m,Tp = 9.5 s; (b) Hs = 6 m,Tp = 11 s
Earthq. V = 11.4 m/s V = 15 m/s V = 20 m/s

Mean shear Mean bending Mean shear Mean bending Mean shear Mean bending
force error moment error force error moment error force error moment error
(%) (%) (%) (%) (%) (%)

(a)
CM 1.89 4.14 1.38 3.28 0.87 2.04
CC -7.13 -3.68 -6.99 -4.96 -6.43 -7.26
D 18.34 0.40 19.97 2.19 22.85 6.10
E 1.85 4.34 1.46 4.64 4.06 3.52
IV -2.00 7.37 0.39 7.38 2.05 6.99
K 10.30 2.97 10.58 2.86 10.37 4.06
LP 13.44 6.03 14.76 6.29 20.40 6.34
NP 6.63 6.71 6.63 6.71 6.27 5.93
N 10.86 15.94 12.81 15.42 14.34 17.40
S 3.59 -2.98 1.78 -5.73 3.13 -8.76
(b)
CM -2.17 1.85 -4.25 0.93 -6.08 -0.75
CC -8.42 -0.80 -8.11 -0.85 -5.66 9.53
D 15.39 1.45 15.16 4.16 15.92 7.06
E -1.61 4.22 -2.18 1.80 3.43 2.81
IV 0.48 9.06 1.06 8.62 3.60 8.40
K 14.68 3.73 14.60 3.57 14.91 5.49
LP 13.28 5.21 16.01 6.25 20.64 7.21
NP 3.36 5.04 2.21 4.69 2.09 4.59
N 8.99 13.13 8.58 13.64 8.75 14.30
S 0.71 0.71 0.71 -4.76 1.80 -7.68
Bold values indicate maxima errors for each sea state

domain simulations, essentially due to total weight only). For the blades, stress demands
found in all simulations fall within the range of those obtained for other typical operational
load cases (IEC 2009), in agreement with previous findings of the authors for the system
under study (Alati et al. 2015). Detailed numerical results for all system components and
simulations cannot be reported for brevity. It can be concluded, however, that elastic
material properties and structural modal damping ratios taken in Sect. 2 are correct.
As for computational advantages, it is noticed that the wind-wave response, once com-
puted for a given sample of wind velocity at the hub and sea state, applies for any earthquake
record. On the contrary, the fully-coupled simulation must be re-run whenever the earthquake
record changes. In view of the simulation length set in Sect. 3.1, this means that, for a given
sample of wind velocity/sea state, and the 10 earthquake records of this study:
(i) The fully-coupled simulation approach requires 10 wind-wave-earthquake simu-
lations (470 s each), totalling 10 9 470 = 4700 s simulation time.

123
1026 Bull Earthquake Eng (2018) 16:1007–1040

(ii) The uncoupled-analyses approach requires 1 wind-wave response simulation


(470 s) ? 10 earthquake response simulations (70 s each), totalling
470 ? 10 9 70 = 1170 s simulation time.
Computational advantages become particularly relevant when several responses are
compared for various potential earthquake realizations, as is typical in a design process.

4.2 Uncoupled analyses: Method 2

For a few wind-wave realizations and earthquake records in Table 2, Fig. 10 shows errors
(16)–(18) at the tower base, obtained by Method 2 as potential values of aerodynamic
damping in the separate earthquake response vary within the interval 0–8% at steps equal
to 0.5%. Results in left and right columns refer to k = 1.0 and k = 0.75 in Eq. (15a, b),
respectively.
In some respects, results mirror those in Fig. 4 by Method 1. Indeed, a minimum for the
total error (18) can be found only when ‘‘Mr error’’ (16) and/or ‘‘Fr error’’ (17) start from
negative values for the lowest aerodynamic damping 0.5%, and increase towards zero as
aerodynamic damping increases. However, no minimum for the total error (18) is found when
both errors (16)–(17) start from positive values and increase with aerodynamic damping.
In addition, it is observed that the total error (18) may have no minimum because both
errors (16)–(17) do not tend to zero but remain practically constant as aerodynamic
damping increases, see for instance the bottom-left subfigure of Fig. 10. To explain this
result, Fig. 11 shows the earthquake-induced bending moments MyðEÞ ðtÞ and MxðEÞ ðtÞ (see
Fig. 1), along with the corresponding SRSS bending moment
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

2
2
ð E Þ ð E Þ  rðEÞ is extracted to compute the
MrðEÞ ðtÞ ¼ Mx ðtÞ þ My ðtÞ , whose maximum M
‘‘Mr error’’ (16) in the bottom-left subfigure of Fig. 10 (see Eq. (13b) for M  rðEÞ and Eq. (11)
for Mrð2Þ used in Eq. (16)). Specifically, 0.5% and 8% aerodynamic damping values are
considered in Fig. 11. It is noticed that the maximum M  rðEÞ is not affected by aerodynamic
damping and occurs always at about t = 3.65 s, where MxðEÞ ðtÞ attains its peak value
(negative peak in Fig. 11b). As shown in Fig. 11, the reasons for this result are that:
(i) On one hand, increasing aerodynamic damping does reduce the bending moment in
wind direction, i.e. MyðEÞ ðtÞ, but does not affect the bending moment in direction
orthogonal to wind, i.e. MxðEÞ ðtÞ, which is always the same. This is consistent with
the fact that aerodynamic damping must be included only in the fore-aft support-
structure modes, i.e. the modes in wind direction (Dı́az and Suárez 2014; Prowell
et al. 2014). Indeed, in this manner, uncoupled analyses account for the fact that
fore-aft oscillations of the tower top in wind direction, induced by earthquake
loads, affect the instantaneous thrust force and produce damping effects (Asareh
and Prowell 2012; Asareh and Volz 2013; Asareh et al. 2016; Prowell et al.
2009, 2014; Valamanesh and Myers 2014; Witcher 2005).
(ii) On the other hand, in this case MxðEÞ ðtÞ is much larger than MyðEÞ ðtÞ and,
consequently, the reduction of MyðEÞ ðtÞ caused by aerodynamic damping has
 rðEÞ and, in turn, on the ‘‘Mr error’’ (16).
negligible effect on the maximum M
The fact that MxðEÞ ðtÞ largely prevails over MyðEÞ ðtÞ depends on the components of the
earthquake record. At any rate, this result obtained by Method 2 confirms the conclusions

123
Bull Earthquake Eng (2018) 16:1007–1040 1027

Fig. 10 Method 2—errors (16)–(18) for various potential aerodynamic damping values, under some wind-
wave realizations and earthquakes in Table 2; left column k = 1 in Eq. (15a, b); right column k = 0.75 in
Eq. (15a, b)

123
1028 Bull Earthquake Eng (2018) 16:1007–1040

(a)

(b)

(c)

Fig. 11 Earthquake-induced SRSS bending moment MrðEÞ ðtÞ and corresponding MxðEÞ ðtÞ, MyðEÞ ðtÞ used to
compute ‘‘Mr error’’ (16) in the bottom-left subfigure of Fig. 10

already drawn for Method 1 (see point (ii) in Sect. 4.1), i.e. that for some earthquake
records the uncoupled-analyses approach may only approximate the results from fully-
coupled simulations, and that no aerodynamic damping value capable of minimizing the
total error (18) exists for all earthquake records and wind-wave states.

123
Bull Earthquake Eng (2018) 16:1007–1040 1029

(a)

(b)

Fig. 12 Method 2—errors (16)–(17) at the tower base for 4% aerodynamic damping, all earthquakes, wind
velocity V = 11.4 m/s and waves with a Hs = 5 m, Tp = 9.5 s; b Hs = 6 m, Tp = 11 s

123
1030 Bull Earthquake Eng (2018) 16:1007–1040

(a)

(b)

Fig. 13 Method 2—errors (16)–(17) at the tower base for 4% aerodynamic damping, all earthquakes, wind
velocity V = 15 m/s and waves with a Hs = 5 m, Tp = 9.5 s; b Hs = 6 m, Tp = 11 s

123
Bull Earthquake Eng (2018) 16:1007–1040 1031

(a)

(b)

Fig. 14 Method 2—errors (16)–(17) at the tower base for 4% aerodynamic damping, all earthquakes, wind
velocity V = 20 m/s and waves with a Hs = 5 m, Tp = 9.5 s; b Hs = 6 m, Tp = 11 s

123
1032 Bull Earthquake Eng (2018) 16:1007–1040

(a)

(b)

Fig. 15 Method 2 with k = 1 in Eq. (15a, b)—mean demands for 4% aerodynamic damping, all
earthquakes, wind velocity V = 11.4 m/s and waves with a Hs = 5 m, Tp = 9.5 s; b Hs = 6 m, Tp = 11 s;
left column shear force; right column bending moment

123
Bull Earthquake Eng (2018) 16:1007–1040 1033

(a)

(b)

Fig. 16 Method 2 with k = 1 in Eq. (15a, b)—mean demands for 4% aerodynamic damping, all
earthquakes, wind velocity V = 20 m/s and waves with a Hs = 5 m, Tp = 9.5 s; b Hs = 6 m, Tp = 11 s;
left column shear force; right column bending moment

123
1034 Bull Earthquake Eng (2018) 16:1007–1040

(a)

(b)

Fig. 17 Method 2 with k = 0.75 in Eq. (15a, b)—mean demands for 4% aerodynamic damping, all
earthquakes, wind velocity V = 11.4 m/s and waves with a Hs = 5 m, Tp = 9.5 s; b Hs = 6 m, Tp = 11 s;
left column shear force; right column bending moment

123
Bull Earthquake Eng (2018) 16:1007–1040 1035

(a)

(b)

Fig. 18 Method 2 with k = 0.75 in Eq. (15a, b)—mean demands for 4% aerodynamic damping, all
earthquakes, wind velocity V = 20 m/s and waves with a Hs = 5 m, Tp = 9.5 s; b Hs = 6 m, Tp = 11 s;
left column shear force; right column bending moment

123
1036 Bull Earthquake Eng (2018) 16:1007–1040

Table 4 Method 2 with k = 1 in Eq. (15a, b)—errors in mean demands at tower base for 4% aerodynamic
damping, all earthquakes, wind velocities, and waves with: (a) Hs = 5 m, Tp = 9.5 s; (b) Hs = 6 m,
Tp = 11 s
Earthq. V = 11.4 m/s V = 15 m/s V = 20 m/s

Mean shear Mean bending Mean shear Mean bending Mean shear Mean bending
force error moment error force error moment error force error moment error
(%) (%) (%) (%) (%) (%)

(a)
CM -13.65 -26.11 -14.35 -26.92 -14.29 -25.91
CC -28.58 -33.87 -28.68 -35.61 -25.90 -40.81
D -15.84 -27.28 -3.26 -32.81 -2.94 -33.34
E -31.43 -13.79 -16.16 -27.47 -13.90 -29.51
IV -31.43 -13.79 -30.05 -14.91 -31.56 -19.88
K -4.42 -13.46 -4.76 -13.83 -4.87 -12.09
LP -11.45 -36.93 -11.03 -37.18 -9.57 -36.31
NP -7.35 -18.09 -8.33 -22.06 -7.67 -25.76
N -3.51 -17.80 -2.28 -18.91 -3.60 -22.49
S -44.50 -38.81 -42.98 -43.18 -46.29 255.19
(b)
CM -18.44 -28.98 -21.06 -30.04 -23.23 -29.60
CC -30.85 -29.89 -31.05 -29.81 -26.26 -18.01
D -6.67 -34.82 -8.73 -30.95 -8.73 -31.46
E -20.91 -27.05 -22.27 -30.56 -17.76 -29.73
IV -27.33 -13.83 -26.59 -14.67 -28.14 -18.79
K -2.50 -13.23 -2.64 -13.35 -3.26 -10.47
LP -17.68 -38.29 -15.02 -37.27 -11.00 -35.31
NP -11.41 -20.25 -12.66 -24.93 -13.33 -27.67
N -6.00 -21.42 -7.17 -21.75 -7.01 -27.69
S -51.58 -36.53 -47.22 -42.41 -50.58 253.52

Bold values indicate maxima errors for each sea state

Based on the conclusion above, errors (16)–(17) obtained for a 4% aerodynamic


damping are then reported in Figs. 12, 13 and 14. It is evident that Method 2 with k = 1 in
Eq. (15a, b) provides always negative errors (16)–(17), i.e. conservative with respect to
fully-coupled simulation, with maxima values up to 75%. Instead, Method 2 with k = 0.75
in Eq. (15a, b) provides maxima errors generally below 15% with a few maxima slightly
below 30%, but errors (16)–(17) may be either negative or positive, i.e. conservative or
non-conservative with respect to fully-coupled simulation. Interestingly, these results
confirm previous findings of uncoupled analyses on land-based HAWTs implemented by
Asareh and Prowell (2012), showing that rather conservative results are generally obtained
when separate maxima are summed [this corresponds to assume k = 1 in Eq. (15a, b)],
while results may be conservative or non-conservative when the sum of the separate
maxima is scaled by a 0.75 factor (Asareh and Prowell 2012).
Similar observations can be made from the mean of the SRSS bending-moment and
shear-force demands along the tower reported in Figs. 15, 16, 17 and 18, computed, for
each earthquake record and wind-wave state, by averaging the demands from the wind-

123
Bull Earthquake Eng (2018) 16:1007–1040 1037

Table 5 Method 2 with k = 0.75 in Eq. (15a, b)—errors in mean demands at tower base for 4% aero-
dynamic damping, all earthquakes, wind velocities, and waves with: (a) Hs = 5 m, Tp = 9.5 s;
(b) Hs = 6 m, Tp = 11 s
Earthq. V = 11.4 m/s V = 15 m/s V = 20 m/s

Mean shear Mean bending Mean shear Mean bending Mean shear Mean bending
force error moment error force error moment error force error moment error
(%) (%) (%) (%) (%) (%)

(a)
CM 14.76 5.42 14.24 4.81 14.29 5.57
CC 3.56 -0.40 3.49 -1.71 5.57 -5.61
D 13.12 4.54 22.55 0.39 22.79 -0.01
E 1.43 14.66 12.88 4.40 14.57 2.87
IV 1.43 14.66 2.46 13.82 1.33 10.09
K 21.68 14.91 21.43 14.63 21.34 15.93
LP 16.41 -2.70 16.73 -2.88 17.82 -2.23
NP 19.49 11.43 18.76 8.45 19.24 5.68
N 22.36 11.65 23.29 10.82 22.30 8.13
S -8.38 -4.11 -7.23 -7.38 -9.72 -16.39
(b)
CM 11.17 3.26 9.20 2.47 7.58 2.80
CC 1.86 2.58 1.71 2.64 5.31 11.50
D 20.00 -1.12 18.45 1.79 18.45 1.40
E 9.32 4.71 8.30 2.08 11.68 2.70
IV 4.50 14.62 5.06 14.00 3.90 10.91
K 23.13 15.08 23.02 14.99 22.55 17.15
LP 11.74 -3.72 13.74 -2.96 16.75 -1.48
NP 16.44 9.81 15.50 6.30 15.00 4.25
N 20.50 8.94 19.62 8.68 19.74 4.24
S -13.69 -2.39 -10.42 -6.80 -12.93 -15.14

Bold values indicate maxima errors for each sea state

wave realizations in Table 2 (again, results for V = 15 m/s are similar to those for
V = 11.4 m/s and V = 20 m/s and are not included). Errors along the tower in Fig. 15
through Fig. 18 fall within the range of those obtained by Asareh and Prowell (2012) for
land-based HAWTs, for both k = 1 and k = 0.75 in Eq. (15a, b), and can be considered
acceptable from an engineering point of view. Again, while results from Method 2 with
k = 1 in Eq. (15a, b) are generally conservative along the whole tower, those from Method
2 with k = 0.75 in Eq. (15a, b) may be conservative or non-conservative. In particular, the
maximum error at the tower base is about 55% for k = 1 in Eq. (15a, b) and about 23% for
k = 0.75 in Eq. (15a, b), as reported in Tables 4 and 5 (including also the results for
V = 15 m/s).
As in Sect. 4.1, it is worth noting that stress demands found in system components are
found well below the elastic limit in all simulations, confirming that elastic material
properties and damping ratios in Sect. 2 are correct. Comments on computational advan-
tages are similar to those on Method 1. In this case, in view of the simulation length set in

123
1038 Bull Earthquake Eng (2018) 16:1007–1040

Sect. 3.2, for a given sample of wind velocity/sea state, and the 10 earthquake records of
this study:
(i) The fully-coupled simulation approach requires 10 wind-wave-earthquake simu-
lations (470 s each), totalling 10 9 470 = 4700 s simulation time.
(ii) The uncoupled-analyses approach requires 1 wind-wave response simulation
(1000 s) ? 10 earthquake response simulations (70 s each), totalling
1000 ? 10 9 70 = 1700 s simulation time.

5 Concluding remarks

The present study has investigated time-domain implementations of uncoupled analyses


prescribed by GL 2012 (GL 2012) for seismic assessment of offshore HAWTs in operating
conditions. The NREL 5 MW baseline HAWT, mounted on a steel tripod, has been
considered as case study. Modelling the full system in GH-BLADED (Bossanyi 2000),
wind-wave and earthquake responses have been computed separately and linearly super-
posed, with the earthquake response computed upon including an additional aerodynamic
damping in the structural model. Two different methods of uncoupled analyses have been
considered. Method 1, where demands are computed from a combined response built by
summing wind-wave response and earthquake response from the time instant at which the
earthquake occurs in the fully-coupled simulation; Method 2, where demands are com-
puted by linearly combining the maxima from separate wind-wave response and earth-
quake response. Method 2 has been implemented considering two combination
coefficients: k = 1 as in GL 2012 (GL 2012) and k = 0.75 as in ASCE-AWEA RP2011
(ASCE/AWEA 2011), for comparison.
Numerical results have shown that no optimal aerodynamic damping minimizing the
error at the tower base exists. However, when the separate earthquake response is com-
puted using a 4% aerodynamic damping, Method 1 and Method 2 with k = 0.75 provide
reasonably accurate results, either conservative or non-conservative, while Method 2 with
k = 1 provides always rather conservative results. Errors in terms of maxima demands at
the tower base and mean demands along the tower mirror those found for land-based
HAWTs, showing that a 4% aerodynamic damping, generally adopted for land-based
HAWTs (Asareh and Prowell 2012; Asareh and Volz 2013; ASCE/AWEA 2011; San-
tangelo et al. 2016; Valamanesh and Myers 2014; Witcher 2005), may be a reasonable
value also for offshore HAWTs.
To the best of authors’ knowledge, this study is the first assessing the accuracy of
various time-domain implementations of uncoupled analyses allowed by GL 2012 (GL
2012) for seismic assessment of offshore HAWTs. An important conclusion to retain is that
Method 1 and Method 2 with k = 0.75 may provide non-conservative results: although
they have proved sufficiently close to the results from fully-coupled simulation, this should
clearly kept in mind and possibly drive the selection of appropriate safety factors.
Conclusions of the present study are based on 600 potential ‘‘wind-wave-earthquake’’
scenarios. Although more simulations could be run, possibly with more earthquake records
at various hazard levels, reported data may provide a sufficient insight on results to be
expected from time-domain uncoupled analyses on offshore HAWTs, at least for the
considered NREL 5-MW HAWT and tripod support structure. Results may be of interest,
considering that the NREL 5-MW HAWT is generally representative of modern multi-

123
Bull Earthquake Eng (2018) 16:1007–1040 1039

megawatt wind turbines and, also, that the tripod support structure is a typical design
solution for intermediate waters.

References
Agarwal P, Manuel L (2011) Incorporating irregular nonlinear waves in coupled simulation and reliability
studies of offshore wind turbines. Appl Ocean Res 33:215–227
Alati N, Failla G, Arena F (2015) Seismic analysis of offshore wind turbines on bottom-fixed support
structures. Philos Trans R Soc Lond Ser A Math Phys Eng Sci 373:20140086
American Society of Civil Engineers/American Wind Energy Association (ASCE/AWEA) (2011) Rec-
ommended practice for compliance of large land-based wind turbine support structures. ASCE/AWEA
RP2011, Reston/Washington, USA
Asareh MA, Prowell I (2012) A simplified approach for implicitly considering aerodynamics in the seismic
response of utility scale wind turbines. In: Proceedings of 53rd AIAA/ASME/ASCE/AHS/ASC
structures, structural dynamics and materials conference, Honolulu, 23–26 April 2012
Asareh MA, Volz JS (2013) Evaluation of aerodynamic and seismic coupling for wind turbines using finite
element approach. In: Proceedings of ASME 2013 international mechanical engineering congress and
exposition, vol 4B—Dyn, Vib Control, San Diego, California, 15–21 November 2013
Asareh MA, Schonberg W, Volz J (2016) Effects of seismic and aerodynamic load interaction on structural
dynamic response of multi-megawatt utility scale horizontal axis wind turbines. Renew Energy
86:49–58
Bossanyi EA (2000) Bladed for windows user manual. Garrad Hassan and Partners, Bristol
Chakrabarti SK (1987) Hydrodynamics of offshore structures. WIT Press, Southampton
Det Norske Veritas (DNV) (2013) Design of offshore wind turbine structures. DNV-OS-J101, Copenhagen
Dı́az O, Suárez LE (2014) Seismic analysis of wind turbines. Earthq Spectra 30(2):743–765
Ernst B, Seume JR (2012) Investigation of site-specific wind field parameters and their effect on loads of
offshore wind turbines. Energies 5:3835–3855
Failla G (2014) Seismic analysis of wind energy converters. In: Beer M, Kougioumtzoglou IA, Patelli E,
Siu-Kui AuI (eds) Encyclopedia of earthquake engineering. Springer, Berlin, pp 1–19
Failla G, Arena F (2015) New perspectives in offshore wind energy. Philos Trans R Soc Lond Ser A Math
Phys Eng Sci 373:20140228
Germanischer Lloyd (GL) (2010) Guideline for the certification of wind turbines. Hamburg
Germanischer Lloyd (GL) (2012) Guideline for the certification of offshore wind turbines. GL 2012,
Hamburg
Haciefendioğlu K (2012) Stochastic seismic response analysis of offshore wind turbine including fluid-
structure-soil-interaction. Struct Des Tall Spec Build 21:867–878
Hasselmann K, Barnett TP, Bouws E et al (1973) Measurements of wind wave growth and swell decay
during the joint North Sea wave project (JONSWAP). Deutsche Hydrografische Zeitschrift A8:1–95
International Electrotechnical Commission (IEC) (2005) Wind turbines—part 1: design requirements. IEC
61400-1, 3rd edn, Geneva
International Electrotechnical Commission (IEC) (2009) Wind turbines—part 3: design requirements for
offshore wind turbines. IEC 61400-3, 1st edn, Geneva
Ishihara T, Yamaguchi A, Sarwar MW (2012) A study of the normal turbulence model in IEC 61400-1.
Wind Eng 36(6):759–766
Jonkman J, Butterfield S, Musial W, Scott G (2009) Definition of a 5-MW reference wind turbine for
offshore system development. Report No. NREL/TP-500-38060, National Renewable Energy Labo-
ratory (NREL), Golden, Colorado, USA
Katsanos EI, Thöns S, Georgakis CT (2016) Wind turbines and seismic hazard: a state-of-the-art review.
Wind Energy. doi:10.1002/we.1968
Kim DH, Lee SG, Lee IK (2014) Seismic fragility analysis of 5 MW offshore wind turbine. Renew Energy
65:250–256
Leu TS, Yo JM, Tsai YT, Miau JJ, Wang TC (2014) Assessment of IEC 61400-1 normal turbulence model
for wind conditions in Taiwan west coast areas. In: Proceedings of 5th international symposium on
physics fluids (ISPF5), Int J Modern Phys 34:1460382
Manwell JF, McGowan JG, Rogers AL (2010) Wind energy explained: theory, design and application.
Wiley, Chichester

123
1040 Bull Earthquake Eng (2018) 16:1007–1040

Nguyen HH, Manuel L (2015) A Monte Carlo simulation study of wind turbine loadsin thunderstorm
downbursts. Wind Energy 18:925–940
Pacific Earthquake Engineering Research Center (PEER) (2013) Peer ground motion database. University of
California, Berkeley, USA. http://ngawest2.berkeley.edu/
Prowell I (2011) An experimental and numerical study of wind turbine seismic behavior. PhD dissertation,
University of California, San Diego, USA
Prowell I, Veletzos M, Elgamal A, Restrepo J (2009) Experimental and numerical seismic response of a
65 kW wind turbine. J Earthq Eng 13(8):1172–1190
Prowell I, Elgamal A, Uang C, Jonkman J (2010) Estimation of seismic load demand for a wind turbine in
the time domain. In: Proceedings of European wind energy conference exhibition (EWEC), Warsaw,
20–23 April 2010
Prowell I, Elgamal A, Uang C, Luco JE, Romanowitz H, Duggan E (2014) Shake table testing and numerical
simulation of a utility-scale wind turbine including operational effects. Wind Energy 17(7):997–1016
Resor BR (2013) Definition of a 5 MW/61.5 m wind turbine blade reference model. Report no. SAND2013-
2569, Wind Energy Technology Department Sandia National Laboratories, Albuquerque, New Mex-
ico, USA
Santangelo F, Failla G, Santini A, Arena F (2016) Time-domain uncoupled analyses for seismic assessment
of land-based wind turbines. Eng Struct 123:275–299
Schwartz M, Heimiller D, Haymes S, Musial W (2010) Assessment of offshore wind energy resources for
the United States. Report No. NREL/TP-500-45889, National Renewable Energy Laboratory (NREL),
Golden, Colorado, USA
U.S. Geological Survey (USGS) (2008) Hazard map (PGA, 2% in 50 years). http://earthquake.usgs.gov/
hazards/products
Valamanesh V, Myers A (2014) Aerodynamic damping and seismic response of horizontal axis wind turbine
towers. J Struct Eng 140(11):04014090
Witcher D (2005) Seismic analysis of wind turbines in the time domain. Wind Energy 8(1):81–91

123

You might also like