Ocean Engineering 266 (2022) 112692

Contents lists available at ScienceDirect

Ocean Engineering
journal homepage: www.elsevier.com/locate/oceaneng

A new FounDyn module in OpenFAST to consider foundation dynamics of

monopile supported wind turbines using a site-specific soil
reaction framework
Lilin Wang a, b, *, Takeshi Ishihara b
a
Ocean College, Zhejiang University, Zhoushan, 316021, PR China
b
Department of Civil Engineering, School of Engineering, The University of Tokyo, 7-3-1, Hongo, Bunkyo-ku, Tokyo, Japan

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

Keywords: A FounDyn module is created in OpenFAST to consider foundation dynamics, which is an appealing supplement
Wind turbine to the current version of OpenFAST. The FounDyn module receives the motions from the SubDyn module and
Monopile sends the forces back to the SubDyn module. In FounDyn, the soil-monopile interaction is captured using a site-
FounDyn module
specific soil reaction framework. The soil reaction framework possesses the same configuration of the semi-
Site-specific soil reaction
Earthquake
analytical 1D model to consider effects of the large plie diameter and the small pile aspect ratio but uses new
site-specific soil reaction models. The soil reaction models are nonlinear and hysteretic, which match the desired
modulus reduction curve by identifying three parameters in a hyperbolic function and a linear function using
genetic algorithm (GA) and manual parameter tuning, and the desired damping curve by applying the Ishihara-
Yoshida rule that controls the unloading-reloading curves iteratively through three parameters. The FounDyn
module is verified by the well-confined OC3 project in terms of modal frequencies, tower top displacement and
shear force and moment at the mudline, and reasonable agreements are achieved between them. A series of
emergency shutdown analyses of the NREL 5 MW wind turbine are performed using OpenFAST plus FounDyn.
The results show that the misalignment of wind and earthquake affects the tower bending moments largely. The
earthquake excitation is found to be the design driving load, prevailing over the wind excitation for the design of
wind turbine supporting structures.

1. Introduction software FAST. Jung et al. (2015) developed a macro-spring model in

FAST to consider the soil-pile interaction, that is, a concentrated
Offshore wind turbines are complicated systems that consist of rotor nonlinear elastic soil spring was set at the mudline to achieve a
nacelle assembly (RNA), control system, tower, substructure, and simplified soil-pile interaction simulation. Similarly, Krathe and Kaynia
foundation, which suffer from environmental loads such as wind, wave, (2017) also developed a macro-spring model at the mudline in FAST to
earthquake, etc. The dynamic responses of offshore wind turbines are further consider the nonlinear hysteresis characteristics of soil-pile
coupled with the external environmental conditions, foundation dy­ interaction, so that the damping effect of the soil-pile interaction can
namics, and intelligent control algorithms, which means the interactions be reflected. In addition, Loken and Kaynia (2019) also proposed an
among them shall be not neglected. As a popular simulation tool of integrated model for the analyses of offshore wind turbines based on the
offshore wind turbines, FAST can model the RNA and tower well, and modification of FAST. In their model, the foundation that is originally
the environmental loads are well considered in it. However, a fixed fixed at the mudline was expanded to a certain depth below the mudline
foundation is used in FAST till this study was performed, which makes it to mimic the stiffness of soil-pile interaction. A decrease in modal fre­
unable to consider the soil-structure interaction accurately (Jonkman quencies of offshore wind turbines can be realized by the flexibility of
and Buhl, 2005). the extended cantilever beam. However, this method cannot reflect the
Several pieces of research have been conducted to achieve the inte­ real condition of the pile base. In summary, the soil-pile interaction was
grated analysis of offshore wind turbines by modifying the open-source considered somehow in the above-mentioned researches. However,

* Corresponding author. Ocean College, Zhejiang University, Zhoushan, 316021, PR China.

E-mail address: [email protected] (L. Wang).

https://doi.org/10.1016/j.oceaneng.2022.112692
Received 3 March 2022; Received in revised form 21 September 2022; Accepted 22 September 2022
Available online 17 October 2022
0029-8018/© 2022 Elsevier Ltd. All rights reserved.
L. Wang and T. Ishihara Ocean Engineering 266 (2022) 112692

these considerations of soil-pile interactions are too simplified to predict 2. A new FounDyn module in OpenFAST with a site-specific soil
the dynamic responses of monopiles accurately. reaction framework
Several pieces of research were performed to implement the foun­
dation dynamics and earthquake excitations in FAST. Prowell (2011) 2.1. A site-specific soil reaction framework for soil-monopile interaction
implemented an advanced seismic module in FAST V7. The module can
specify a platform motion time series as displacement, velocity, or ac­ Wang and Ishihara (2022) proposed a semi-analytical one-dimen­
celeration independently in the X, Y, and Z axes. The force required to sional (1D) model for the soil-structure interaction of pile foundations
achieve the desired motion is calculated at run time for each time step under the combined lateral, vertical, and torsional forces. The reason
and applied to the wind turbine platform in FAST using a damped why the model is semi-analytical is that a correction factor κ (see Eq.
oscillator model. However, their implementation did not directly sup­ (A1)) is introduced in the calculation of ultimate capacities of lateral and
port the consideration of base rocking, twisting, or soil-structure inter­ rotational soil reactions and fitted based on three-dimensional finite
action (SSI). Yang et al. (2020) implemented a seismic analysis element analyses. However, the soil reactions in the proposed model use
framework (SAF) in FAST V7. Although they stated the shortcomings of empirical models (PySimple1, TzSimple1, etc.) that were fitted to the
the API p-y curve, it was used to model the soil-pile interaction for the test results of certain soil conditions. The applicability of these soil re­
seismic analysis in their study. All previous researchers implemented actions to soil conditions other than those examined is uncertain (Zhang
their methods in FAST V7 using a user-specified subroutine UserPtfmLd. et al., 2021; Lai et al., 2021). To improve the applicability of the pro­
However, no pieces of research focus on the widely-used version of posed model, a site-specific soil reaction framework is formed using the
FAST, OpenFAST. same configuration of the semi-analytical 1D model, but with
Previous dynamic analyses of offshore wind turbines were performed site-specific soil reactions. Since a site-specific p-y model PySimple5 has
with fixed foundations, in which the soil-pile interactions are not been proposed by the author (Wang and Ishihara, 2022a,b), other
considered (Wei et al., 2013; Kim et al., 2016). Thus, it is needed to site-specific soil reactions (Sy,pb − vy,pb , mx − θx , mx,pb − θx,pb , tz − vz ,
review the influence of soil-pile interaction on the dynamic loadings of Qz,pb − vz,pb ) can be proposed using the same ideas. More specifically,
wind turbine supporting structures. Bush and Manuel (2009) studied the these site-specific soil reactions shall match the desired modulus
dynamic responses of an offshore wind turbine supported by the reduction curve by identifying three parameters in a hyperbolic function
monopile and the sandy seabed. Compared to the fixed foundation, the and a linear function using the genetic algorithm (GA) and manual
peak value and the cyclic amplitude of bending moment at the mudline parameter tuning algorithm, and the desired damping curve by applying
will increase by 15% and 20%, respectively if the soil-pile interaction is the Ishihara-Yoshida rule that controls the unloading-reloading curves
considered. Similar conclusions were reported by Krathe (2015). Jung iteratively through three parameters. It should be noted that the pro­
et al. (2015) modelled the soil-pile interaction using the macro-spring posal of site-specific soil reactions is not the focus of this paper, the
model and found that the soil-pile interaction affects not only the dy­ originality of this study exists that the author tends to implement this
namic responses but also the modal properties of offshore wind turbines. advanced soil reaction framework in the widely-used aero-elastic soft­
The natural frequency of the wind turbine will be overestimated by 14% ware OpenFAST to support the integrated analysis of monopile sup­
if the soil-pile interaction is not considered. Loken and Kaynia (2019) ported wind turbines. According to the author’s knowledge, this
also found that when the soil-pile interaction is considered, the meaningful research has not been investigated in previous studies.
maximum bending moment at the mudline increased by 1%, while the Fig. 1 illustrates the procedure to propose the configuration of the
cyclic amplitude of the bending moment increased by 7%. The peak site-specific soil reaction framework (same as the semi-analytical 1D
value relates to the ultimate bearing capacity of the wind turbine sup­ model). Fig. 1(a) presents the monopile with the external excitations.
porting structure, while the cyclic amplitude affects the fatigue life of Fig. 1(b) shows the stress distribution on the monopile from the soil.
the wind turbine supporting structure. A 7% increase in the cyclic Fig. 1(c) shows the fishbone-shaped model for the monopile, from which
amplitude of the bending moment would increase fatigue damage by the configuration of the site-specific soil reaction framework can be
28% since an offshore wind turbine is subjected to millions of cyclic derived as shown in Fig. 1(d). The site-specific soil reaction framework
loads during its 25-year service life. To summarize, if the fixed foun­ includes the lateral, rotational and vertical soil reactions along the pile
dation is used, the cyclic amplitude of the dynamic loading will be shaft and at the pile base. More details about Fig. 1a-d can be found in
underestimated, and the fatigue life of the monopile will be over­ Wang and Ishihara (2022a). The site-specific soil reactions in the pro­
estimated, which leads to an unsafe design. Since the above results are posed model are summarized in Table 1. MθSimple5 is selected to
based on simple soil-structure interaction models, the accuracy and demonstrate the idea of site-specific soil reactions since few pieces of
reliability need to be further evaluated. Regarding the seismic consid­ research focused on the rotational soil reaction.
eration in the design of wind turbines, many studies were conducted in Following the idea of PySimple5, MθSimple5 is proposed to consider
the past decades (Wang and Ishihara, 2020). However, most researchers the coupling between axial and lateral resistances. Following the Japa­
tend to use the finite element models instead of the aero-elastic models nese guideline for the seismic soil-pile interaction analysis, the modulus
since the earthquake excitation and foundation dynamics are not well reduction and damping curves shall be matched in the proposed soil
addressed in the wind turbine simulation software, such as OpenFAST. reaction curve (Ishihara, 2010). MθSimple5 adopts the linear (Me − θe )
Compared to the aero-elastic model, an obvious shortcoming of the and hyperbolic (Mp − θp ) components in series. The governing equations
finite element model is that it cannot consider the rotor dynamics and of MθSimple5 are expressed in Eqs. (1)–(7). The force and stiffness in the
the servo dynamics. linear component (Eqs. (1) and (2)), in the hyperbolic component (Eqs.
In this study, the integrated dynamic analyses of offshore wind tur­ (3) and (4)), and the overall spring (Eqs. (5) and (6)) are presented
bines are performed by creating a new FounDyn module in OpenFAST to herein briefly. Eq. (7) defines the yield function used in MθSimple5.
consider the foundation dynamics of offshore wind turbines. The outline Note that Eq. (3) shows that the unloading and reloading are achieved
of this paper is given as follows: a site-specific soil reaction framework by updating the reversal point (θp0 , m0 ) in the plastic component of
for monopiles is updated and a FounDyn module is created in OpenFAST MθSimple5, which is different from the widely-used Masing rule. Pa­
in Section 2. The FounDyn module is verified and demonstrated in rameters in MθSimple5 can be divided into backbone curve relevant
Section 3. The application of OpenFAST plus FounDyn to the integrated parameters (c, n, η), unloading-reloading relevant parameters (A, B, E),
dynamic analysis of a monopile-supported wind turbine is demonstrated soil properties relevant parameters (mult , hmax , θ50 ) and others (m0 , θp0 ,
in Section 4. Conclusions are given in Section 5. g
md0 , θ0 , θ+
0 , θ0 ).
−

2
L. Wang and T. Ishihara Ocean Engineering 266 (2022) 112692

Fig. 1. Monopile foundation and its dynamic semi-analytical model.

Table 1 where Ke is the elastic modulus, Kp is the plastic modulus, and K is the
Soil reaction models in the Semi-analytical 1D model for monopiles. combined modulus. f is the yield function; mα is the value of m at the
center of the elastic region (analogous to the backstress in the classical
Spring name Soil reaction component Soil reaction curves
plasticity theory), Cr ⋅mult is the yielding force and Cr is 0 in Pysimple5,
py − vy Lateral for pile shaft PySimple5 m0 is the value of m at the start of current plastic loading cycle and mult is
Sy,pb − vy,pb Lateral for pile base TzSimple5
the ultimate bearing capacity. θe is the elastic component of displace­
mx − θx Rotational for pile shaft MθSimple5
mx,pb − θx,pb Rotational for pile base MθSimple5 ment, θp is the plastic component of displacement, θp0 is the value of θ0 at
t z − vz Vertical for pile shaft TzSimple5 the start of current plastic loading cycle and θ50 is the displacement
Qz,pb − vz,pb Vertical for pile base QzSimple5 where m = 0.5mult . θ+ 0 is the memory term for the positive side of the
gap, θ−0 is the memory term for the negative side of the gap. The initial
M e = K e θe (1) values of θ+ 0 and θ0 are θ50 /100 and − θ50 /100, respectively. Stiffness
−

constants c, n, η define the shape of the backbone curve of the

mult MθSimple5.
Ke = η (2)
θ50 The backbone curve relevant parameters (c, n, η) can be identified by
( )n GA using the criterion that the backbone curve shall be capable of
mp = B ⋅ E ⋅ mult − (B⋅ E ⋅ mult − m0 )
A⋅E⋅c⋅θ50
(3) matching the desired modulus reduction curve. The fitness function of
A⋅E⋅c⋅θ50 + |θp − θp0 | GA, 1/RMSE, is defined as the ratio of root mean squared error in Eq. (8),
[( )n ] while the iteration will be stopped when Eq. (9) is satisfied or the gen­
Kp =
∂mp n⋅sign(ẏ)(B⋅E⋅mult − m0 )
=
A⋅E⋅c⋅θ50
(4) eration reaches 100. Note that the results of GA may be not stable and
∂θp |θp − θp0 | + A⋅E⋅c⋅θ50 |θp − θp0 | + A⋅E⋅c⋅θ50 manual parameter tuning algorithm can be used to determine the final
values of c, n, and η.
m = me = mp (5) √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
√ N ( )2
√1 ∑ Ki Gi
K = (1/K e + 1/K p )− 1
(6) RMSE = √ − (8)
N i=1 K0 G0
f = |m − mα | − (Cr ⋅ mult ) ​ (7)

3
L. Wang and T. Ishihara Ocean Engineering 266 (2022) 112692

( )n
RMSE ≤ 10− 3 or φj − φj− ≤ 10− 3
(9) ⎧ c⋅θ50
mult − (mult − m0i /b0 ) (Δθ > 0)
1
⎪
⎪
⎨ c⋅θ50 + (θ − θp0i /a0 )
where RMSE means the root mean square error. Gi is the modulus at the g(θ) = ( )n
⎪
ith points of the modulus reduction curve, while Ki is the corresponding ⎪
⎩ c⋅θ50
− mult + (mult + m0i /b0 ) ​ (Δθ < 0)
stiffness of the p-y curve. N is the number of points on the modulus c⋅θ50 − (θ− θp0i /a0 )
reduction curve used for identification, which could be 10–20 uniformly ⎛ ⎞n
( )
distributed points on the modulus reduction curve. G0 and K0 are the ⎧ m0i+1 /b0 ⎜ c⋅θ 50 ⎟
⎪ mult − mult − (θ− θp /a )⎠ ​ (Δθ > 0)
⎪ ⎝
initial shear modulus and stiffness, respectively. φj means the parameter ⎪
e0
⎪ c⋅θ 0i+1 0
⎪
⎨ 50 + e0
at the jth generation and φ could be c, n, or η. f(θ) = ⎛ ⎞n
The site-specific modulus reduction and damping curves can be ⎪
⎪
⎪ ( )
described by the well-organized Hardin-Drnevich model (Hardin and
⎪
⎪ m0i+1 /b0 ⎜ c⋅θ50 ⎟
⎩ − mult + mult + ⎝ (θ− θp /a )⎠ ​ (Δθ < 0)
Drnevich, 1972). One example of identified parameters is given in e0 c⋅θ50 − 0i+1 0
e0
Table 1 by fitting to G/G0 = 1/(1 +γ/γ0.5 ) with γ0.5 = 0.25% in the where h(θ) is the desired damping curve. D(θ) represents the
Hardin-Drnevich model. The backbone curve and corresponding damping corresponds to the p-y loops. ΔW(θ) is the damping energy and
modulus reduction curve using the parameters in Table 2 are portrayed
W(θ) is the equivalent elastic strain energy. The coordinates (θp0i , m0i )
in Fig. 2. It is observed that the proposed model can match well with the
and (θp0 i + 1, m0i+1 ) are the most recent two reversal points.
desired modulus reduction curve. The relationship between displace­
( ( p) )/ ( ( p) )
ment (θ) and strain (γ) proposed by Lai et al. (2021) is used to convert θ θ
displacement into strain (θ = 0.8γ) and guarantees that K/ K0 = 1/ abs D 0i − h(θp0i ) h(θp0i ) ≤ 1.0e− 3 or abs D 0i − h(θp0i )
a0 a0
(1 +θ/θ50 ) and h = hmax θ/(θ50 +θ) for the p-y curve corresponds to G/
≤ 1.0e− 3
(14)
G0 = 1/(1 +γ/γ0.5 ) and h = hmax γ/(γ0.5 +γ) for the Hardin-Drnevich
model. 2m − me − mp
The unloading-reloading relevant parameters (A, B, E) can be ≤ 1.0e− 12
(15)
mult
updated using the Ishihara-Yoshida rule (Ishihara et al., 1985) following
the criterion that the backbone curve shall be capable of matching the
damping curve. The Ishihara-Yoshida rule was proposed for ground 2.2. A new FounDyn module in OpenFAST
response analyses to solve the overestimation of soil damping when the
soil strain is large. The same idea is used in this study for p-y modeling. OpenFAST is a multi-physics, multi-fidelity tool for simulating the
A, B and E have different values for different loading conditions as coupled dynamic response of wind turbines, which couples computa­
shown in Eq. (10) and are obtained iteratively to match the damping tional modules for aerodynamics, hydrodynamics for offshore struc­
curve using Eqs. (11)–(13). It is noted that Eqs. (12) and (13) cannot be tures, control, and electrical system (servo) dynamics, and structural
solved in a closed form for a0 and e0 . They are solved numerically using dynamics to enable coupled nonlinear aero-hydro-servo-elastic simula­
the bisection method. The iteration will be stopped when Eq. (14) is tion in the time domain. However, the foundation is the least considered
satisfied. Eq. (5) is also examined, and the iteration will be stopped when in OpenFAST among the important components of wind turbines. Since
Eq. (15) is satisfied or the number of iterations reaches 20. Fig. 3 com­ more and more wind turbines are installed in seismically active regions,
pares the hysteresis loops and damping curves with and without the it is also necessary to supplement the foundation dynamics into Open­
Ishihara-Yoshida rule. It is noticed that the damping at large strains will FAST for the seismic analysis of onshore and offshore wind turbines. In
be significantly overestimated without using the Ishihara-Yoshida rule this study, a FounDyn module is created in OpenFAST to address the
to modify the hysteresis loop. foundation dynamics. As given in Fig. 4, the newly implemented
⎧ FounDyn module works similarly to the HydroDyn module when it is
⎨ skeleton ​ curve: ​ A = 1, B = 1, E = 1 applied to an offshore wind turbine. That is, the FounDyn module first
unloading: ​ A = a0 , ​ B = b0 , ​ E = 1 (10) receives the motions from the SubDyn module and then sends back the
⎩
reloading ​ & ​ reunloading: ​ A = a0 , ​ B = b0 , ​ E = e0 ​ forces and moments from the soil-structure interaction and seismic
( )n excitation to the SubDyn module at each time step. To achieve this,
SubDyn module needs to be modified to remove the base fixity to
c⋅θ50
p
c⋅θ50 +|θ0 |
− 1
b0 = ( )n (11) incorporate the dynamics of a foundation, i.e., monopile. The configu­
a0 ⋅c⋅θ50
p
a0 ⋅c⋅θ50 +|θ0 |
− 1 ration of OpenFAST with the FounDyn module is depicted in Fig. 5, in
( p) which the red color shows the new implementation of the FounDyn
θ0i module in OpenFAST. Note that the seismic submodule in the FounDyn
D = h(θp0i ) (12)
a0 module uses the same method as that provided by Prowell (2011). The
( ) FounDyn module includes the SSI submodule and the earthquake sub­
yp0i p0i /b0 module. The proposed semi-analytical 1D model is extended to the 3D
f = (13)
a0 e0 model and implemented in the SSI submodule. Currently, a similar
( ) module called SoilDyn is implemented using constant spring stiffnesses
Here, h(θ) = hmax θ
θ50 +θ , D(θ) = 1 ΔW(θ)
4π W(θ) , W(θ) = 12 θ⋅ g(θ), ΔW(θ) = and dashpot dampings to consider the soil-structure interaction for wind
∫θ and wave dynamics in OpenFAST, which is also absorbed as part of the
2 − θ g(θ0 )dθ0 ,
SSI submodule. Therefore, the FounDyn module can consider different
SSI models and loading conditions including wind, wave, and
earthquake.
Table 2
One example of identified parameters for the modulus reduction curve in the H-
3. Verification of the FounDyn module in OpenFAST
D model.
Soil condition Material constants MθSimple5
The FounDyn module is verified by the OC3 project. The Interna­
Clay or Sand c 0.5 tional Energy Association (IEA) performed the Offshore Code Compar­
n 0.7 ison Collaboration (OC3) project to investigate offshore wind
η 3.5
technology and deployment. In the OC3 project, the NREL 5 MW wind

4
L. Wang and T. Ishihara Ocean Engineering 266 (2022) 112692

Fig. 2. One example of MθSimple5 backbone curve with parameters in Table 1 and modulus reduction curve derived from MθSimple5.

Fig. 3. One example of MθSimple5 hysteresis loop and damping curve with and without Ishihara-Yoshida rule.

Fig. 4. Comparison of original and modified SubDyn modules.

5
L. Wang and T. Ishihara Ocean Engineering 266 (2022) 112692

Fig. 5. Module configuration of OpenFAST with FounDyn.

turbine with a monopile foundation is studied. The diameter and

thickness of monopile are 6m and 0.06m, respectively. The embedment
of monopile is 36m. The information of the wind turbine is summarized
in Table 3. The soil profile is shown in Fig. 6. For the OC3 project,
Jonkman et al. (2008) and Jonkman and Musial (2010) simulated the
wind turbine responses using the distributed linear springs to consider
the soil-structure interaction. The spring stiffness is obtained by using
the initial stiffness of API p-y curves. During their simulation, the wind
speed of hub height is 11.4 m/s and the turbulent wind is generated
using the Kaimal spectrum with a turbulent density of 14%. The wave is
generated using the Jonswap spectrum with an effective wave height of
6m and a wave period of 10s.
To verify the FounDyn module, the parameters of the proposed
model shall be consistent with those in Jonkman and Musial (2010).
More specifically, the stiffnesses of rotational soil reactions for pile shaft
and base shall be small enough to be negligible. Therefore, the proposed
model is reduced to the p-y model. The initial stiffness of the p-y model
shall be the same as that in Jonkman and Musial (2010). Note that there
is an option used in the FounDyn to determine whether nonlinear soil
reactions are considered. When the option is set as false, the initial
stiffness is used to linearize soil reactions. Table 4 shows comparisons of
simulation results between the FounDyn and the OC3 project (Jonkman Fig. 6. The soil-monopile system in the IEA OC3 project.
et al., 2008). It is noticed that the predicted frequencies by the FounDyn
agree well with those in the OC3 project. Table 5 illustrates the dynamic
responses of the wind turbine predicted by the FounDyn and the OC3 Table 4
project. The results show that the differences between the two methods Comparison of predicted modal frequencies by the OC3 project and FounDyn.
Modal frequencies FA 1st mode SS 1st mode FA 2nd mode SS 2nd mode

OC3 project 0.248 0.246 1.546 1.533

Table 3 FounDyn 0.245 0.240 1.530 1.494
Properties of the NREL 5 MW Baseline OWT with a monopile support structure.
Note:FA means the fore-aft direction，SS means the side-side direction.
Rating 5 MW

Rotor orientation, configuration Upwind, 3 blades are smaller than 5% in terms of the mean value, maximum value, and
Rotor diameter 126 m
standard deviation of the tower top displacement, the shear force at the
Hub-height 90 m
Tower top diameter, wall thickness 3.87, 0.019 m mudline, and the moment at the mudline. It means that the FounDyn
Tower base diameter, wall thickness 6.0, 0.027 m module can predict the soil-structure interaction of monopiles
Substructure diameter, wall thickness 6.0, 0.06 m accurately.
Cut-in, rated, cut-out wind speed 3, 11.4, 25 m/s
It is important to examine whether the proposed soil reaction model
Rated rotor speed 12.1 rpm
Rated tip speed 80 m/s
is implemented properly. To illustrate the nonlinearity of the proposed
Rotor mass 110,000 kg soil reaction model, a series of sinusoidal uniform wind flows are defined
Nacelle mass 240,000 kg in Fig. 7 and applied to the above NREL 5 MW wind turbine. The am­
Tower mass 346,460 kg plitudes of sinusoidal uniform wind flows are set to mimic the realistic
Mean sea level 20.0m
behavior of wind and the period is close to the natural period of the

6
L. Wang and T. Ishihara Ocean Engineering 266 (2022) 112692

Table 5
Comparison of predicted dynamic responses by the OC3 project and FounDyn.
Response TTD (m) Qmudline (MN) Mmudline (MNm)

Mean Maximum STD Mean Maximum STD Mean Maximum STD

OC3 project 0.509 0.931 0.113 0.409 4.053 1.099 57.502 124.800 20.198
FounDyn 0.510 0.925 0.111 0.407 4.041 1.085 57.050 122.621 19.848
Difference (%) 0.196 0.644 1.770 0.489 0.296 1.274 0.786 1.746 1.733

Fig. 7. Harmonic wind velocities.

turbine. During the simulation, the sinusoidal wind is the only excitation obtained with a). To demonstrate the combination of earthquake
input, and others from the HydroDyn FounDyn modules are eliminated. loading and operational loading, a series of emergency shutdown ana­
The corresponding hysteretic responses for the My − θy spring at 1 m lyses are performed with the parameters in Table 6. In Table 6, the wind
below the mudline are illustrated in Fig. 8. The hysteresis loops show flow is uniform and constant with a velocity of 11.4 m/s, while 15
that a nonlinear behavior is successfully implemented in the FounDyn earthquake waves are used to consider the phase characteristic and for
module. Since the area under the loop represents soil damping, the each earthquake wave. The simulations are demonstrated using the
larger loop means that the more soil damping exits. It is also observed NREL 5 MW wind turbine, whose information can be found in Section
that larger rotations and moments are observed for the cases with the 3.1. The targeted wind turbine is assumed to be supported by a monopile
mean velocity of 10 m/s than those with the mean velocity of 0 m/s. embedded in a uniform sand layer. The information of monopile and
supporting soil are given in Table 7 and Table 8, respectively. It is
4. Application of the FounDyn module for dynamic analyses of assumed that the nonlinearity of the supporting soil can be modelled
monopile-supported wind turbines under wind and earthquake using the Hardin-Drnevich model with parameters of γ0.5 = 0.1% and
hmax = 21%. In addition, the properties of the engineering bedrock are
4.1. Configuration of numerical simulations assumed to be Vs = 144 ​ m/s and γ = 2.0E + 04N/m3 , which are needed
for the soil dynamic analyses later.
The IEC 61400-1 guideline suggests that the earthquake loading shall As recommended in Ishihara (2010), the structural integrity and
be superposed with operational loading that shall be equal to the higher safety of wind turbine support structures are required against the level II
of a) loads during normal power production by averaging over the earthquake with a recurrence period of 500 years. This also fulfills the
lifetime; b) loads during the emergency shutdown for a wind speed requirement of the IEC 61400-1 guideline which states that the ground
selected so that the loads prior to the shutdown are equal to those acceleration corresponding to a 475-year recurrence period should be

7
L. Wang and T. Ishihara Ocean Engineering 266 (2022) 112692

Fig. 8. Hysteretic response for the My − θy spring at 1 m below the mudline.

8
L. Wang and T. Ishihara Ocean Engineering 266 (2022) 112692

Table 6 in SHAKE to obtain the seismic waves at different elevations. Fig. 10

Parameters for the emergency stop. illustrates the time histories of seismic waves at typical elevations
Item Value generated by a level II earthquake with the phase of El Centro NS. It is
noticed that the seismic wave becomes stronger along with the decrease
Wind condition Wind speed from 11.4 m/s to 0 m/s
Seismic condition 15 artificial Level II earthquake waves of elevation depth.
Misalignment between wind and 0 deg; 45 deg; 90 deg As portrayed in Fig. 11, three misalignment levels of wind and
earthquake earthquake (0 deg, 45 deg, and 90 deg) are investigated to illustrate the
Nacelle acceleration limit 300gal effects of the misalignment of wind and earthquake. Larger mis­
Pitch feathering rate 8 deg/s
alignments are not considered since they would reduce the dynamic
responses of the monopile-supported wind turbine. To model the soil-
monopile interaction properly, the spring interval of 1m is used for
Table 7 the depth range of (-20m, -25m) while the spring interval of 2m is used
Monopile parameters. for the depth range of (-25m, -50m). It is important to note the damping
Foundation type Diameter (m) Thickness (m) Embedment (m) used in this study. Each of the simulations has a duration of 121s and a
Monopile 6 0.06 36 time step of 0.0005s. The emergency stops are triggered when the ac­
celeration of the nacelle reaches 300 gals. Once it is triggered, the blades
are turned to pitch feathering with a rate of 8 deg/s.
Table 8
Uniform sand properties used in this section.
4.2. Results and discussion

Soil properties Gr (MPa) Br (MPa) cu (kPa) φ (deg) γ (kN/m3) Fig. 12 illustrates various time histories during the emergency stop
′

Sand 41.5 90.0 0.0 39.5 9.81 with the nacelle acceleration limit of 300 gals and the pitch feathering
Note: Gr , low-strain shear modulus; Br , low-strain bulk modulus; cu , cohesion; φ, rate of 8◦ per second for the 0-degree misalignment. More specifically,
soil friction angle; γ , effective unit weight.
′ Fig. 12(a) depicts the time histories of wind velocity and seismic ac­
celeration. A uniform wind with the rated velocity of 11.4 m/s is
adopted since the turbulent component is less important to the extreme
load compared to the mean component. Fig. 12(b) shows the time his­
considered for the seismic response evaluation of a wind turbine. For the tories of the resultant nacelle acceleration of X and Y directions. That is
level II earthquake, the response spectrum at the engineering bedrock is why the value is no less than 0 gal. It is noticed that the nacelle accel­
defined as follows: eration limit reaches 300 gals at 10s, from which the emergency stop
⎧
⎨ a0 (1 + 9.375T) (0 ≤ T ≤ 0.16) starts by feathering the blade pitch at the rate of 8◦ per second. As shown
Sa0 (T, 0.05) = 2.5a0 (0.16 < T < 0.64) (22) in Fig. 12(c), the blade pitch angle keeps 0◦ until 10s and increases
⎩
1.6a0 /T (T ≥ 0.64) linearly to 90◦ at the interval of 21.25s, at which the emergency stop is
finished and after which the blade is feathered and the pitch angle does
where Sa0 (T, 0.05) specifies the basic peak ground acceleration a0 at the not change any more. It is interesting to note that the rotor speed ex­
engineering bedrock along with the frequency characteristic of the periences a short increase before it decreases sharply and it takes more
ground motions. a0 is 3.2 m/s2 for the level II earthquake and 1.6 m/s2 time for the rotor speed reduces to zero than that for the pitch angle
for the level I earthquake; T is the natural period (s). In this study, a 120- reaches 90◦ . Fig. 12(d) portrays the corresponding time histories of the
s seismic time history is generated based on the target spectrum tower base moment in the X direction for the misalignments of 0-degree
described above, in which the phase characteristics of 4 typical real and 90-degree, while Fig. 12(e) presents those in the Y direction. For the
earthquake waves such as El Centro NS, Taft NS, Hachinohe EW, and 0-degree misalignment, it is found although the amplitude of seismic
JMA Kobe EW and 11 random phase properties are used. Fig. 9 shows wave decreases after the 50s, the moment for the X direction (My) in­
the response spectrum of the generated earthquake wave with a creases after the 50s. This is because after the emergency stop occurs, the
damping ratio of 5%. Since the generated seismic waves are specified at X-direction aerodynamic damping reduces to near zero. The loss of
the engineering bedrock, a series of site response analyses are performed aerodynamic damping causes the increase of moment. By comparing the
moments of the X and Y directions for the 90-degree misalignment, it is
known that the aerodynamic load (My) is much smaller than the seismic
load (Mx), which means that the earthquake excitation is the design
driving load, prevailing over the wind excitation for the design of wind
turbine support structures.

Fig. 9. Response spectrum of the generated level II seismic wave with the
damping ratio of 5%. Fig. 10. Time histories of displacements at different elevations.

9
L. Wang and T. Ishihara Ocean Engineering 266 (2022) 112692

Fig. 11. The misalignment configuration of wind and earthquake.

To obtain the distribution of seismic loading along the tower, the

maximum combined seismic and aerodynamic loads of 15 seismic waves
are averaged for each misalignment and presented in Fig. 13. It is found
that due to the nonlinearity of soil-structure interaction and the char­
acteristics of seismic waves, the shear force and bending moment are not
linearly distributed along the tower, which is different from the pre­
dictions using the fixed foundation. In addition, the misalignment affects
the combined seismic and aerodynamic loads largely. The 0-degree
misalignment outputs the maximum bending moment, the 45-degree
misalignment shows the middle value, while the 90-degree misalign­
ment yields the smallest one, from which it is concluded that the
bending moment of tower decreases along with the increase of the
misalignment of wind and earthquake.

5. Conclusion and future research subject

In this study, a new FounDyn module is created in OpenFAST to

consider foundation dynamics based on a site-specific soil reaction
framework. The site-specific soil reaction framework has the same
configuration of the semi-analytical 1D model proposed by Wang and
Ishihara, 2022a to consider effects of plie diameter and aspect ratio but Fig. 12. Illustrations of various time histories during the emergency stop.
uses new site-specific soil reaction models. The soil reaction models are
nonlinear and hysteretic, which match the desired modulus reduction by the OpenFAST plus FounDyn module, the integrated dynamic anal­
curve by identifying three parameters in a hyperbolic function and a ysis of monopile supported wind turbines can be conducted considering
linear function using genetic algorithm (GA) and manual parameter the aerodynamics, hydrodynamics, aeroelasticity, servo dynamics and
tuning algorithm, and the desired damping curve by applying the foundation dynamics, which means the FounDyn module is an appealing
Ishihara-Yoshida rule that controls the unloading-reloading curves supplement to the current version of OpenFAST.
iteratively through three parameters. The application of FounDyn module is demonstrated to perform
The FounDyn module is created in OpenFAST to consider the foun­ dynamic responses of a NREL 5 MW monopile supported wind turbine
dation dynamics using the proposed site-specific soil reaction frame­ under coupled wind and earthquake. Considering the fact that emer­
work. The FounDyn module receives the motions from the SubDyn gency shutdown is the least known operational scenario, a series of
module and sends the forces back to the SubDyn module. The FounDyn emergency shutdown analyses are performed in which the misalignment
module is verified by the well-confined OC3 project in terms of modal of wind and wave is considered. The results show that the misalignment
frequencies, tower top displacement and shear force and moment at the of wind and wave affects the bending moments of tower significantly.
mudline, and reasonable agreement is achieved between them. Driven The earthquake excitation is found to be the design driving load,

10
L. Wang and T. Ishihara Ocean Engineering 266 (2022) 112692

Fig. 13. Profiles of maximum combined loads for the emergency stop with various angles between wind and earthquake.

prevailing over the wind excitation for the design of wind turbine sup­ 4. Although the site-specific soil reaction framework suits for all types
porting structures. of pile foundations, i.e., monopile, jacket, pile group, etc., the current
It is believed that the FounDyn module is a meaningful improvement version of the FounDyn module considers the monopile foundation
for the coupled analyses of offshore wind turbines. However, it suffers to only. In the future, other pile foundations and other types of foun­
some limitations when it is applied to design offshore wind turbines. dations (i.e., gravity foundation, monopod, tripod, etc.) shall be
Some examples of future research subjects are as follows: considered in the FounDyn module.
5. Since the FounDyn module is evaluated by a limited number of nu­
1. During the implementation of the Ishihara-Yoshida rule to match the merical analyses in this study, it shall be verified or validated by
damping curve in the proposed model, the bisection method is used more numerical analyses, experiments or field tests before it is
to iterate the damping parameters A, B, E, which is unconditionally widely-used for the design of offshore wind turbines.
stable, but less efficient. It is necessary to implement other more
advanced algorithms (i.e., Ridders’ Method, Ridders, 1979) to CRediT authorship contribution statement
improve the simulation efficiency, especially when the proposed
model is coupled with the OpenFAST program. Lilin Wang: Conceptualization, Methodology, Software, Data cura­
2. The current soil reaction framework contains 8 types of soil reactions tion, Writing – original draft, Visualization, Investigation, Validation.
that are distributed along the whole pile shaft and base. The Takeshi Ishihara: Methodology, Supervision, Writing – review &
complicated model configuration reduces the stability and efficiency editing.
of simulations when it is coupled with the OpenFAST program, i.e.,
the time increment usually shall be small enough to obtain the stable Declaration of competing interest
results. This inconvenience can be addressed by the recently pro­
posed mechanism-based family of “p-y+M-theta” models (Wang The authors declare that they have no known competing financial
et al., 2020; Lai et al., 2021), which unify the predictions for in­ interests or personal relationships that could have appeared to influence
teractions between soil and piles covering a wide range of rigidity the work reported in this paper.
and geometry (i.e., flexible, semi-rigid and rigid piles) using a single
set of parameters. The physical soundness of the model lies in the Data availability
new understanding associated with three-zones of failure mecha­
nisms around a laterally loaded pile, which were firstly discovered by No data was used for the research described in the article.
Hong et al., 2017 based on ‘visualized’ half-pile centrifuge model
tests and numerical analyses using an advanced cyclic model. Acknowledgement
3. The current site-specific soil reaction framework is mainly proposed
for seismic analyses of wind turbines. Therefore, it can be further The authors express his deepest gratitude to Dr. Jason Jonkman and
improved for wind and wave analyses of wind turbines by incorpo­ Dr. Andy Platt in National Renewable Energy Laboratory (NREL) for
rating the cyclic effect into the site-specific soil reaction framework. sharing the method to implement a new module in OpenFAST.

Appendix

Appendix A. The input parameters in the proposed model

The simple way to determine the ultimate capacity and representative displacement for the proposed model can be found in Eq. (A1)-(A10). More
information can be found in Wang and Ishihara (2022a).
( )
π 1 π
pult (z) = + tan δ Dσ r0 + cD (A1)
4 3 4

11
L. Wang and T. Ishihara Ocean Engineering 266 (2022) 112692

pult (z)
y50 (z) = (A2)
K ​ py (z)

{ √̅̅̅̅̅̅̅4
√̅̅̅̅̅̅ ( ◦ ) 6.612⋅D− 0.469 for ​ sand ( )
Here, σr0 = κ(2c Kp + γ zKp ), Kp = tan 2 45 + φ2 , κ = , δ = 13 ∼ 23 φ, Kpy = 0.65E s 12 Es D
Ep Ip , Es = 2(1 + υs )Gs , Ip =
′
1.112
min(2.714⋅D , ​ 2.2) for ​ clay 1− v2
s

π
64 [D
4
− (D− 2t)4 ]
( ) π D2
Qult,pb = η cNc Fcs Fcd Fci + γ’ Df Nq Fqs Fqd Fqi + 0.5γ’ BNγ Fγs Fγd Fri (A3)
4
Qult,pb
z50,pb = (A4)
KQz,pb

[ ( )0.75 ]
Here, KQz,pb = 1−GLυ 0.73 + 1.54 DL
( )
1 Npb
Mult,pb = ⋅Npb 1 − D2 (A5)
2 Qult,pb

Mult,pb
θ50,pb = (A6)
KMx,pb

Gs D π 23 √̅̅̅̅
Here, KMx,pb = 3(1− υs ), A = 4D , D2 = A

(A7)
′
tult (z) = (k0 γ z tan δ + k1 c)πD

tult (z)
z50 (z) = (A8)
Ktz

( ( ) )
6.8 L − 1.71
Here, Ktz = π D Es π D

1 π
Mult (z) = cD2 + D2 tan δσr0 (A9)
2 8

Mult (z)
θ50 (z) = (A10)
KMx

( )− 1.71 2
Here, KMx = 0.85 DL Es L
( π )
2
Sult,pb = c D + Npb tan δ (A11)
4
Sult,pb
y50,pb = (A12)
KSy,pb

⎛ ⎞
∑L
4Gs D ⎜ K ∑L ⎟
Here, KSy,pb = (2− υs ), Npb = N0 − min⎝∑L z=0 tz
N0 , ​ z=0 tult (z)⎠, k0 = 0.4, k1 = 0.5,
z=0
Ktz +KQz, ​ pb

where c is cohesion of soil, γ is the effective unit weight of soil, φ is the friction angle of soil, Dr is the relative density of soil, Pa is the atmospheric
′

pressure. Nc , Nq and Nγ are the bearing capacity factors, Fcs , Fqs and Fγs are the shape factors, Fcd , Fqd and Fγd are the depth factors, Fci , Fqi and Fri are the
inclination factors, which can be found in Meyerhof (1963). η is the plugging coefficient. In the Chinese method (Chinese Department of Construction,
2008), η = 1 for closed-ended piles, η = 0.16z/D when z/D < 5, and η = 0.8 when z/D ≥ 5. In the FinnRA method (FinnRA, 2000), η = 0.8 for sandy
soil if z/D = 15, and the value decreases linearly with the z/D ratio. Gs and υs are the shear modulus and Poisson ratio of soil, respectively. t and Ep are
the thickness and Young’s modulus of monopile, respectively.

References Hong, Y., He, B., Wang, L.Z., Wang, Z., Ng, C.W.W., Mašín, D., 2017. Cyclic lateral
response and failure mechanisms of a semi-rigid pile in soft clay: centrifuge tests and
numerical modeling. Can. Geotech. J. 54, 806–824.
Bush, E., Manuel, L., 2009. Foundation Models for Offshore Wind Turbines, Aerospace
IEC 61400-1, 2019. Wind Turbines – Part 1: Design Requirements, Ed, p. 4.
Sciences Meetings. American Institute of Aeronautics and Astronautics. https://doi.
Ishihara, T., 2010. Guidelines for Design of Wind Turbine Support Structures and
org/10.2514/6.2009-1037.
Foundations (In Japanese).
Chinese Department of Construction, 2008. JGJ 94-2008: Technical code for building
Ishihara, K., Yoshida, N., Tsujino, S., 1985. Modelling of stress-strain relations of soils in
pile foundations. (In Chinese).
cyclic loading. In: International Conference on Numerical Methods in Geomechanics,
FinnRA, 2000. Steel Pipe Piles, Helsinki.
pp. 373–380.
Hardin, B.O., Drnevich, V.P., 1972. Shear modulus and damping in soils: design
Jonkman, J.M., Buhl, M.L., 2005. FAST User’s Guide. National Renewable Energy
equations and curves. J. Soil Mech. Found Div. 98 (7), 667–692.
Laboratory.

12
L. Wang and T. Ishihara Ocean Engineering 266 (2022) 112692

Jonkman, J.M., Musial, W., 2010. Offshore Code Comparison Collaboration (OC3) for Prowell, I., 2011. An Experimental and Numerical Study of Wind Turbine Seismic
IEA Task 23 Offshore Wind Technology and Deployment. Technical Report NREL/ Behavior. University of California, San Diego.
TP-5000–48191. Ridders, C.J.F., 1979. A new algorithm for computing a single root of a real continuous
Jonkman, J.M., Butterfield, S., Passon, P., Larsen, T., Camp, T., Nichols, J., Azcona, J., function. IEEE Trans. Circ. Syst. 979–980.
Martinez, A., 2008. Offshore code comparison collaboration within IEA wind annex Wang, L., Ishihara, T., 2020. A study of the effects of foundation uplift on the seismic
XXIII: phase II results regarding monopile foundation modeling. Technical report no. loading of wind turbine tower and shallow foundation using a new dynamic Winkler
NREL/CP-500- 47534. model. Eng. Struct. 219, 110745.
Jung, S., Kim, S.R., Patil, A., Hung, L.C., 2015. Effect of monopile foundation modeling Wang, L., Ishihara, T., 2022a. A semi-analytical one-dimensional model for offshore pile
on the structural response of a 5-MW offshore wind turbine tower[J]. Ocean Eng. foundations considering effects of pile diameter and aspect ratio. Ocean Eng. 250,
109, 479–488. 110874.
Kim, B., Jin, J., Bitkina, O., Kang, K., 2016. Ultimate load characteristics of NREL 5-MW Wang, L., Ishihara, T., 2022b. New p-y model for seismic loading prediction of pile
offshore wind turbines with different substructures[J]. Int. J. Energy Res. 40, foundations in non-liquefiable and liquefiable soils considering modulus reduction
639–650. and damping curves. Soils Found. 62 (2022), 101201.
Krathe, V.L., 2015. Aero-hydro Dynamic Analysis of Offshore Wind Turbine - Wang, L., Lai, Y., Hong, Y., Mašín, D., 2020. A unified lateral soil reaction model for
Implementation of Nonlinear Soil-Structure Interaction in Software fast[D], vol. monopiles in soft clay considering various length-to-diameter (L/D) ratios. Ocean
2015. Msc Thesis. Engineering 212, 107492.
Krathe, V.L., Kaynia, A.M., 2017. Implementation of a non-linear foundation model for Wei, S., Hyun-chul, P., Chin-hwa, C., Jae-ha, B., Youngchan, K., Chang-wan, K., 2013.
soil-structure interaction analysis of offshore wind turbines in FAST[J]. Wind Energy Load analysis and comparison of different jacket foundations[J]. Renew. Energy 54,
20 (4), 695–712. 201–210.
Lai, Y., Wang, L., Zhang, Y., Hong, Y., 2021. Site-specific soil reaction model for Yang, Y., Bashir, M., Li, C., Michailides, C., Wang, J., 2020. Mitigation of coupled wind-
monopiles in soft clay based on laboratory element stress-strain curves. Ocean Eng. wave-earthquake responses of a 10 MW fixed-bottom offshore wind turbine. Renew.
220, 108437. Energy 157, 1171–1184.
Loken, I.B., Kaynia, A.M., 2019. Effect of foundation type and modelling on dynamic Zhang, Y., Aamodt, K.K., Kaynia, A.M., 2021. Hysteretic damping model for laterally
response of offshore wind turbines[J]. Wind Energy 22 (12), 1667–1683. loaded piles. Mar. Struct. 76, 102896.
Meyerhof, G.G., 1963. Some recent research on the bearing capacity of foundations. Can.
Geotech. J. 1 (1), 16–26.

13