Engineering Structures 162 (2018) 257–269

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Short communication

Structural failure simulation of onshore wind turbines impacted by strong T

winds
⁎
Jui-Sheng Choua, , Yu-Chen Oub, Kuan-Yu Lina, Zhi-Jia Wanga
a
Department of Civil and Construction Engineering, National Taiwan University of Science and Technology, Taipei, Taiwan
b
Department of Civil Engineering, National Taiwan University, Taipei, Taiwan

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

Keywords: The renewable energy industry is thriving in many countries against a global backdrop of growing environ-
Wind power mental awareness. In particular, in Taiwan, wind is emerging as a potential source. The Taiwanese government
Blade damage has implemented the “Thousand Wind Turbines” project as a collaboration with the private sector to construct a
Tower collapse wind power infrastructure. However, when Typhoon Soudelor hit Taiwan on August 7–8, 2015, the towers of
Structural behavior simulation
2.0 MW wind turbines close to Taichung harbor collapsed and falling blades damaged one of the collapsed
Root cause analysis
turbine towers. This incident of wind turbine collapse was the second in the nation and was more extensive than
System safety
Structural reinforcement the previous one. This study investigates the causes of this incident and the mechanical mechanisms of turbine
Wind turbine tower collapse and blade fracture to support risk prevention and hazard-resistant design of future wind turbines.
Relevant data are obtained to simulate wind turbine collapse. Next, mechanical analyses via finite element
method are carried out to identify mechanisms of failure and structural weakness planes, with the ultimate
purpose of summarizing potential causes of collapse. Improvement solutions on strong wind resistance of wind
turbine towers and their benefits are discussed. Finally, this study gives the recommendations on weaker blades
being a safety mechanism for the wind turbine tower, torque capacity of the pitch system as well as the required
strength of joint bolts to be installed in steel structural connections, so as to withstand severe storms.

1. Introduction
This study focuses on a wind farm in Taichung in central Taiwan,
whose Z72 wind turbines (2.0 MW) collapsed between August 7 and 8,
2015, as a result of Typhoon Soudelor. The maximum instantaneous
wind speed, measured by the supervisory control and data acquisitions
(SCADAs) in the nacelles of the turbines, exceeded 59.5 m/s. In the
wake of the typhoon, a maintenance team that was sent by the owner of
the wind turbines found that one wind turbine’s blades were fractured
and six towers of turbines had snapped at either the first or second of
the three sections from the ground.
In this study, data from in-situ inspections, technical information
from the manufacturer, and the results of a literature review are utilized
to develop a wind turbine model, simulate the crumbling of turbines,
and conduct forensic engineering assessments, to identify mechanisms
of the collapse of turbines and weak planes in the tower to determine
the causal factors of the incident and inform the construction, hazard-
resistant design [1], and risk prevention planning of future wind tur-
bines.
Fig. 1 depicts the geographic locations of the wind turbine units that
were damaged by Typhoon Soudelor. They were located in the north-
west of the Taichung port area, approximately 3 km south of the Dajia
River estuary.
A loss-making incident may occur as a result of multiple factors. All
of the inspection items that were covered by the incident investigation
were organized into a flow chart (Fig. 2). This investigation comprised
the following steps: (1) collecting data on wind turbine construction
and typhoon-induced turbine collapse and reviewing these data to
identify the causals of the turbine collapse incident and conduct in-situ
simulations; (2) conducing a finite element analysis (FEA) of forces on
the wind turbines in the ultimate limit state, identifying or excluding
potential causal factors; and (3) determining the causes of the incident
and proposing improvements to the turbines.
The rest of this paper is organized as follows. Section 2 reviews wind
turbine accidents and incidents worldwide. Section 3 investigates re-
levant documents including the proof strength, design strength, and
specifications of bolts, the manufacturer’s blueprints of the wind tur-
bines, the construction and maintenance of the turbines and possible
causes of the incident. Section 4 presents a structural analysis (which
consists of a series of force analyses of wind turbines under various

⁎
Corresponding author.
E-mail addresses: jschou@mail.ntust.edu.tw (J.-S. Chou), yuchenou@ntu.edu.tw (Y.-C. Ou).

https://doi.org/10.1016/j.engstruct.2018.02.006
Received 19 July 2017; Received in revised form 30 December 2017; Accepted 1 February 2018
0141-0296/ © 2018 Elsevier Ltd. All rights reserved.
J.-S. Chou et al.
Incident site
Satellite image (before turbines collapsed)
Satellite image (after turbines collapsed)
Fig. 1. Locations of damaged wind turbines (https://www.google.com.tw/maps/).
structural failure sequences and scenarios), and summarizes the results
of the investigation into the turbine collapse incident and its causal
factors. Section 5 provides suggestions concerning hazard-resistant
design and risk prevention planning.
2. Literature review
Reports on the investigations of relevant incident and forensic en-
gineering assessments of disaster-damaged wind turbines in Taiwan
[2–4] and other countries [5–8] were reviewed to identify the potential
causal factors of the turbine collapse incident.
258
Engineering Structures 162 (2018) 257–269
2.1. International case study
Table 1 presents the extent of damage to wind turbines that col-
lapsed during typhoons [3,4,6,8–10]. Previous cases have shown that
wind turbines with damaged blades can retain their respective towers
[9]. The FEA has been used to illustrate the failure mode of Z72 tower
structure, and identify the tower bolt junctions of a Z72 wind turbine as
its weak planes [4].
An investigation of the effect of rainfall and wind speed on wind
turbines revealed that rainfall has a negligible effect on them, ac-
counting for 1–3% of wind load [7]. The highest rainfall ever recorded
near the damaged wind turbines under investigation is 87.5 mm/hr at a
weather station in Wuqi District in Taichung; on the basis of [7], the
analysis in this study omits the effects of rainfall.
2.2. Statistics concerning wind turbine-related accidents and incidents
globally
A total of 1376 wind turbine-related accidents and incidents
worldwide, from 2006 to the end of March 2016, compiled by the
Caithness Windfarm Information Forum [11] were reviewed to analyze
the causal factors of the Port of Taichung turbine collapse incident and
the risk factors that may have contributed to turbine damage.
The majority of the identified cases were categorized as “Others”
(n = 306, 22% of all the cases), meaning that damage was unspecified
or negligible, investigation reports were not published, or causes were
not determined. Excluding such cases, the total number of wind tur-
bine-related accidents and incidents was 1070, of which 23% involved
“damaged blades” and 11% “damaged structures” (Fig. 3). Fig. 3 pre-
sents causal factors in the collapse of wind turbines and their respective
frequencies worldwide during 2006–2016.
A previous study [12] estimated the failure frequencies of the
components of onshore wind turbines and subsumed these components
under 16 subassemblies. Towers had the highest failure frequency
(0.151 times per year), followed by gearboxes (0.134 times per year)
and rotor blades (0.124 times per year). The average failure frequency
(caused by component fault or damage) of onshore wind turbines was
estimated to be 1.22 times per year. In the Port of Taichung incident,
the blade-pitch systems of the wind turbines were damaged, causing a
chain of effects that eventually induced the collapse of the towers.
3. Data collection and incident investigation
Data on Z72 wind turbines were obtained from the owner of the
turbines to ensure the reliability of upcoming analyses. Relevant data
were also derived from other credible sources to ensure comprehen-
siveness of the data and ensure the objectivity of the analysis of the
collapsed turbines.
3.1. Recorded operating data of wind turbines
Operating data were collected from July 1 to August 15, 2015, for
the damaged turbines in the Port of Taichung. The operating data of the
turbines before collapse, which were obtained through the SCADA
system [13,14] in each of the turbines, included wind speed, wind di-
rection, rotor speed, yaw angle, blade rake, and turbine condition.
Table 2 shows the conditions of the wind turbines before and during
collapse, which were inferred from the data on turbine operations, in-
situ inspections, and interviews with on-site engineers. Note that the
SCADA systems neither directly measured blade conditions nor re-
vealed irregular patterns prior to blade breakage [15].
3.2. In-situ wind speed during the typhoon
Table 3 presents the maximum instantaneous wind speed of each
collapsed turbine, which was estimated from recorded wind speed data.
J.-S. Chou et al. Engineering Structures 162 (2018) 257–269

Data collection and

analysis

Review the records of the

Review the blueprints Review structural construction, installation, Review other relevant
inspection, operation, and
of wind turbines calculation reports monitoring of wind documents
turbines

Structural models of
Analyze wind speed
Determination of turbine towers,
(force) and the
material properties blades, and pitch
external force
system

Analyze structural
stress-strain
relationship and
review the results
thereof

Overall assessment of
the root causes of the
incident of wind
turbine collapse

Fig. 2. Incident investigation procedure.

Table 1
Overview of damaged wind turbines in Japan, China, and Taiwan.

Year 2003 2003 2008 2008 2013 2015

Typhoon Maemi Dujuan Jangmi Jangmi Usagi Soudelor

Wind farm Miyakojima Island, Shanwei,China Changhua Coastal Industrial Taichung Harbor, Shanwei, China Taichung Harbor,
Japan Park, Taiwan Taiwan Taiwan
No. of damaged wind 3 8 5 0 7 2
turbines
No. of collapsed turbine 3 0 0 1 8 6
towers

The last record of operation before collapse covered the period from mass) portions that were bolted together. Fig. 4 depicts the structure of
3:00 AM to 5:30 AM on August 8, 2015. When Turbine H11, whose the wind turbine.
signal was disconnected at the time, was excluded, the maximum in-
stantaneous extreme wind speed of the other collapsed turbines was
3.4. Blade-pitch system
59.5 m/s and the minimum was 50.8 m/s. These results were used in
subsequent mechanical analyses.
The pitch system varies the windward angle of a blade to control the
amount of wind that is received by a turbine [16]. The blade-pitch
3.3. Blueprint of turbine structure
system in a Z72 turbine provides a braking torque of 32 N·m, which is
multiplied by a gear ratio to withstand torque at the roots of the blades
The manufacturer’s blueprint of the tower structure was obtained to
and to adjust their windward angles.
examine the geometric structures of collapsed turbines by developing a
numerical model of the tower structure.
The total height and gross mass of each turbine tower were 3.5. Verifying the strength of tower bolts
61.895 m and 101,357 kg, respectively. The tower comprised upper
(19.645 m in height and 19,372 kg in mass), middle (24.92 m in height The tensile strengths of the tower bolts were obtained from the
and 39,424 kg in mass), and lower (17.33 m in height and 39,139 kg in following sources.

259
J.-S. Chou et al. Engineering Structures 162 (2018) 257–269

Table 3
Maximum instantaneous wind speed measured for each collapsed wind turbine.

Turbine No. Maximum instantons wind speed Last record of operation before
collapse

H01 53.0 m/s 2015/8/8 03:04

H04 50.8 m/s 2015/8/8 05:33
H06 Detailed wind speed data were –
unavailable.
H11 33.2 m/s 2015/8/8 17:04
H12 51.2 m/s 2015/8/8 04:17
H14 59.5 m/s 2015/8/8 04:41
H18 59.0 m/s 2015/8/8 03:54

3.6. Maintenance records of wind turbines

Turbine maintenance records obtained from the turbine owner; they

Fig. 3. Statistics concerning wind turbine accidents and incidents worldwide during
2006–2016 (excluding cases categorized as “others”).
– Bolt supplier’s test reports.
– Turbine owner’s batch test reports, which are recognized by the
Taiwan Accreditation Foundation (TAF) and Societe Generale de
Surveillance (SGS) [17].
– Academic literature (undamaged bolts): a TAF-accredited laboratory
was used to conduct tensile strength tests on six sets of M30, 10.9-
grade bolts that were undamaged during Typhoon Jangmi. These
bolts were randomly selected from disassembled turbine towers at
the Port of Taichung [18].
– Academic literature (damaged bolts): a TAF-accredited laboratory
was used to conduct tensile strength tests on six sets of damaged
M30, 10.9-grade bolts, which had been sampled [4] from a col-
lapsed turbine tower in the Taichung Port.
Table 4 shows the summary of the tensile tests on bolts from the
specified sources. Comparing the results of the bolts from these sources
revealed the high completeness, reliability, and objectivity of the tur-
bine owner’s batch test reports, which were therefore subsequently
used in subsequent analyses.
included the torque specifications and annual maintenance of the bolts.
The 2015 torque specification record revealed variations in the bolt
torque among all connection layers of each turbine tower, which ac-
counted for in the subsequent mechanical analyses.
3.7. In-situ inspection
Turbine damage revealed by the first in-situ inspection was cate-
gorized into two types.
(1) Type 1: fractured towers in Turbines H1, H04, H06, H11, H14, and
H18 (Fig. 5). The tower of Turbine H1 fractured at the junction
between the upper and middle portions (42.3 m above the ground),
whereas the towers of the other turbines fractured at the junction
between the middle and lower portions (17 m above the ground).
These six turbines exhibited similar failure behaviors and failure
surfaces. In the cross-sectional view of the failure of the tower of
Turbine H04 (Fig. 5), some screw holes caved in, but the main steel
body was undamaged, indicating that the limited strength of the
junctions was responsible for the failure of the tower.
(2) Type 2: Straight turbine towers with fractured blades (Turbine

Table 2
Overview of damaged wind turbines in Port of Taichung during Typhoon Soudelor.

Turbine No. Estimated Status Rotor (Note 1) Pitch system (Note 2) Yaw system
collapse time

H01 2015/8/8 03:06 The pitch system failed, and two blades Over speed Failed Substantial change in the
were damaged. (41 rpm) (Blades rotated at 1.8°) rotor orientation toward
the wind
H04 2015/8/8 05:30 The pitch system failed, and the rotor Rotation Failed Incomplete recorded data
rotated. (11 rpm) (Blades rotated at 56°)
H06 2015/8/8 04:42 Data recording was disconnected at 04:42, Slight rotation Blades rotated constantly at 92°. The rotor rotated at a
and no instantaneous wind speed data (4 rpm) constant angle.
were recorded.
H11 2015/8/7 Signals were disconnected, and the nacelle Signal Signal disconnected Signal disconnected
15:14 detected smoke. disconnected
Signal
disconnected
H12 Uncollapsed Blades were fractured and sideswiped the Slight rotation Incomplete data The rotor rotated at a
tower tower when they crashed into the ground. (3 rpm) constant angle.
H14 2015/8/8 04:42 The pitch system failed. (16 rpm) Blades oscillated markedly. Substantial change in the
rotor orientation toward
the wind
H18 2015/8/8 03:59 The pitch system might have failed, but it (8 rpm) Subsequent signals were disconnected and the Substantial change in the
detected considerable change in wind pitch system might have been damaged. (The rotor rotation toward the
direction before the turbine collapsed. rotating angle of blades decreased from 91° to wind
86°) (100°/30 sect)

Note 1: Over speed occurs when the blade rotor rotated at > 26.5 rpm.
Note 2: 30° is the standup position of blades and 84° is their standby position.

260
J.-S. Chou et al. Engineering Structures 162 (2018) 257–269

(Fig. 6).

Notably, if a turbine blade is damaged by previous fractures, then its

ability to withstand direct wind is considerably reduced and so the wind
Blades force acts on the tower. Turbine towers with such damage may retain
their integrity owing to the strong-column (tower)-and-weak-beam
(blades) effect [19]. However, the tower of Turbine H12 was struck by
its falling blades.

4. Structural and mechanical analysis of the collapsed wind

Nacelle
turbines

This section presents a structural and mechanical analysis of the

collapsed wind turbines to elucidate forces on turbine blades and
towers under various conditions of wind speed, blade rotating angle,
19.6 m

Upper tower and number of fractured blades, to simulate the fracturing of the blades
and the collapse of the towers during Typhoon Soudelor, and to propose
measures against similar incidents in the future.

4.1. Estimating wind load

Wind load (wind pressure in the direction of the height of the wind
24.9 m

Middle tower turbine) was estimated using the equations in the Wind Resistance
Design Specifications and Commentary of Buildings [20]:
p = qGCp−qi (GCpi ) (1)

Vz z α
= ⎛ ⎞ 0 ≦ z ≦ zg
V10 ⎝ 10 ⎠ (2)

q (z ) = 0.06 K (z ) K zt [I V10 (C )]2 (3)

17.3 m

2α
Lower tower z
K (z ) = 2.774⎜⎛ ⎟⎞
⎝ zg ⎠ (4)

In these equations, the parameters were set according to the spe-

cification [20]. p is the wind pressure (kgf/m2) that the wind-resistant
system is designed to withstand. q is the external wind velocity pressure
Fig. 4. Illustration of the wind turbine structure. (kgf/m2), which was estimated using Eq. (3). G is the gust response
factor. Since wind speed is time-dependent, the gust response factor
Table 4 converts the dynamic wind pressure that is caused by the along-wind to
Strength of bolts of different grades. the equivalent wind pressure, accounting for the dynamic effects of
wind, and G was set at 1.78. Cp is the external wind pressure coefficient
Source Specs. Tensile Stipulated Yield Stipulated
(which was set to 0.8). qi is the internal wind velocity pressure (kgf/
strength tensile strength yield
(N/mm2) strength (N/mm2) strength m2), which was also estimated using Eq. (3). Cpi is the internal wind
(N/mm2) (N/mm2) pressure coefficient (which was set at −0.2). Vz is wind speed (m/s) at
height z. V10 is wind speed (m/s) at a height of 10 m (which was re-
(1) Manufacturer M30 1041 1040 941 940
placed in this study by the wind speed of the turbine nacelle and the
10.9
(1) Manufacturer M30 1291 1220 1277 1100
aforementioned height was changed from 10 m to 67 m, which is the
12.9 height of the nacelle). z is height above the ground (m). α is the vertical
(2) Batch test M30 1137 1040 1050 940 distribution law index (which was set to 0.15 in line with Ground
10.9 Condition C—flat and open ground or coastal area—at the incident site)
(2) Batch test M30 1283 1220 1207 1100
in relation to the 10-min average wind speed. zg is gradient height (m)
12.9
(3) Undamaged M30 1049 1040 904 940 (which was set to 300 m in line with Ground Condition C —flat and
bolts 10.9 open ground or coastal area—at the incident site). K(z) is the ground
(4) Damaged M30 997 1040 881 940 condition coefficient of wind velocity pressure, which was estimated
bolts 10.9
using Eq. (4). Kzt is the topographic coefficient (which was set to one
because the topography of the incident site is flat). I is the occupancy
coefficient and was set to one herein. Finally, V10(C) is wind speed (m/
H12). Wind turbines use a pitch brake system and blade locks s) at a height of 10 m. Eq. (2) was used to estimate wind speed at
(which together constitute a pitch system) to fix blades in the various heights.
standby (feathered) position to protect their surfaces from excessive The design pressure was estimated from wind speeds that were
wind. Gusts from Typhoon Soudelor caused the failure of both measured at a height of 67 m. On the basis of the minimum in-
mechanism in Turbine H12, despite the normal operation of the stantaneous wind speed of 50.8 m/s, which was measured at the six
power grid in the nacelle. Consequently, the blades of the turbine collapsed turbines (Table 3), the equations provided by the specifica-
were not fixed in the standby position and fractured in direct gusts tion [20] were used to estimate design pressure (Table 5).

261
J.-S. Chou et al. Engineering Structures 162 (2018) 257–269

(a) Tower collapse (b) Bolt junctions in the tower

(c) Damaged blades and nacelles (d) Crashed nacelle

Fig. 5. In-situ Type 1 turbine damage.

(a) Overview of the hub and blades (b) Blade damaged

(c) An entire blade (disassembled) (d) Dent in the blade

Fig. 6. In-situ Type 2 turbine damage.

262
J.-S. Chou et al. Engineering Structures 162 (2018) 257–269

Table 5 Table 6
Wind pressure at different tower heights. Parameters of glass fiber-reinforced plastic blades.

Tower height K(z) Vz q(z) p(z) Wind pressure Parameter Value (GPa)
(m) (m/s) (kgf/m2) (kgf/m2) (kgf/cm2)
EGFB,x 44.65
61.90 1.728 50.20 316.09 562.63 0.056263 EGFB,y 12.96
58.03 1.695 49.72 304.08 541.26 0.054126 EGFB,z 12.96
54.16 1.660 49.20 291.75 519.31 0.051931 PRGFB,x 0.50
50.29 1.623 48.66 279.06 496.73 0.049673 PRGFB,y 0.50
46.42 1.585 48.08 265.98 473.44 0.047344 PRGFB,z 0.50
42.55 1.544 47.46 252.45 449.35 0.044935 GGFB,x 4.32
38.68 1.500 46.78 238.41 424.37 0.042438 GGFB,y 4.32
34.82 1.458 46.05 223.81 398.38 0.039838 GGFB,z 4.32
30.94 1.403 45.24 208.54 371.10 0.037120 Density 0.00001715 kg/mm3
27.08 1.348 44.35 192.48 342.62 0.034262
23.21 1.287 43.33 175.48 312.35 0.031235
19.34 1.219 42.16 157.30 279.99 0.027999
15.47 1.140 40.77 137.58 244.90 0.024490
11.61 1.046 39.05 115.77 206.08 0.020608
7.74 0.926 36.75 90.77 161.57 0.016157
3.87 0.752 33.12 59.89 106.60 0.01066

4.2. Establishing a wind turbine model

ABAQUS FEA [21] was used to establish a wind turbine model that
comprised blades and a tower. The blade manufacturer refused to
provide the blueprint of the blades of the Z72 turbines because it in-
cludes confidential commercial information, so the blades of the model
were constructed from the cross-sectional scans of a V80 wind turbine
blade and scaled down to 34 m in proportion to the size of the Z72
turbine blade. The geometric dimensions of the components of the
tower were consistent with the blueprint of the Z72 wind turbine.

4.2.1. Turbine blade model

The V80 turbine blade, whose length was determined to be 38.5 m,
was modeled by scanning the blade structure into 12 cross-sections,
plotting coordinate points on the cross-sections, and extending the
cross-sections; the internal structure of the blade was thus re-
constructed.
The turbine blade model in this study was established by inputting
the V80 turbine blade model into ABAQUS FEA and scaling it down in
proportion to the length of the Z72 turbine blade (Fig. 7). The structural
analysis of the Z72 turbine blade, which is presented in Section 4.3, was
based on the scaled model and relevant material parameters.

4.2.2. Material parameters of blades
The material parameters of Z72 turbine blades, unfortunately, could
not be obtained directly. Hence, they were estimated using those of the
blade samples that were used in [3], which were acquired through
tensile, compressive, and flexural tests. Table 6 presents the parameters
of the glass fiber-reinforced plastic that were used in the Z72 turbine
models.
4.2.3. Turbine tower model
A turbine tower model was constructed using ABAQUS FEA in ac-
cordance with the Z72 manufacturer’s design of wind turbine; the
model incorporated the turbine blades and nacelle (Fig. 8). Following
Fig. 8. ABAQUS-generated full model of wind turbine.
the ABAQUS model, a sectional tower model was established using
Xtract and loaded with a wind load to conduct a mechanical analysis of
the bolts that connected the various portions of the turbine tower.
4.3. Mechanical analysis of blades
The torque at the roots of the rotor blades was analyzed to ascertain
whether the pitch brake system had sufficient torsional strength to fix
blades in their standby positions during gusts. The wind load on the

Fig. 7. 3-D wind turbine blade model.

263
J.-S. Chou et al. Engineering Structures 162 (2018) 257–269

blades was subsequently estimated to determine the gust-induced da- Based on the assumption that the blades were anchored in their
mage under the ultimate wind load. standby position with a wind inflow of 40 m/s, the maximum torque on
the blade root (during the normal operation of the pitch system), esti-
4.3.1. Torque capacity of blades mated using Bernoulli’s principle and Eq. (7), was 27,117 N·m, which is
Blades have significantly more exposure to wind in their standup well below that at a wind speed of 59.5 m/s. Therefore, the clamping
(nonfeathered) position than in their standby position. To reduce wind force of the pitch system must be increased to fix blades more securely
load on a turbine, the pitch brake system must have sufficient torsional in their standby position.
strength to anchor the blades in their standby position, enabling them
to withstand the torque that is induced by wind pressure and preventing 4.3.2. Failure analysis of blades
them from reverting to their standup position. The Tsai–Wu failure criterion is applied globally to determine
Based on the assumption that the blades were anchored in their whether fiber-reinforced plastic material has failed [22]. A score
standby position, the maximum torque on the blades during a typhoon greater than one indicates material failure. Eq. (8) expresses this cri-
was estimated as follows. The maximum instantaneous wind speed terion.
during Typhoon Soudelor was 59.5 m/s (Table 3) and wind blew di- F1 σ1 + F11 σ12 + F2 σ2 + F22 σ22 + 2F12 σ1 σ2 + F66 τ122 = 1
rectly toward the turbines; therefore, the maximum torque on the
blades was estimated by calculating their aerodynamic pressure of the In the above equation:
blades by applying Bernoulli’s principle, expressed by Eq. (5): 1 1 −1 1 1
F1 = + ,F11 = ,F2 = +
1 2 1 XT XC XT XC YT YC (8)
P1 + ρV1 = P2 + ρV22
2 2 (5) −1 1
F22 = ,F66 = 2 ,F12 = −0.5 F11 F22
where P1 and P2 denote pressure on both sides of a blade; V1 and V2 YT YC S
represent wind speed on both sides of the blade, and ρ denotes the where XT is the tensile strength in the direction of the parallel fibers, XC
density of air (1.29 kg/m3). is the compressive strength in the direction of the parallel fibers, YT is
The extreme force on both sides of a blade was considered (based on the tensile strength in the direction perpendicular to the fibers, YC is the
an assumption that pressure difference between them was maximal); V1 compressive strength in direction perpendicular to the fibers, and S is
represents the inflow wind speed and V2 was set to 0, thereby creating a the shear strength of the fibers.
maximizing pressure difference between both sides of the blade, which Section 4.3.1 notes that during Typhoon Soudelar, the blade-pitch
was given by Eq. (6): system could not keep the blades in their standby position, so in this
1 2 section, the FEA assumes that blades are in the standup position at a
ΔP = ρV1 maximum wind speed of 51.2 m/s (as measured by the SCADA in
2 (6)
Turbine H12, which retained its tower and had two fractured blades;
The maximum instantaneous wind speed (at 59.5 m/s) was sub- see Table 3).
stituted into Eq. (6), and the maximum pressure difference between This turbine blade model was divided into Regions A–J, each of
both sides of the blade (ΔP ) was estimated to be 0.023 kgf/cm2. which was subdivided into upper, middle, and lower elements (Fig. 9).
Wind pressure was assumed to be uniformly distributed across each Possible locations of failure of the blades in the standup position were
blade; the force area (A) of the blade was approximately 900,000 cm2 identified by applying the Tsai–Wu failure criterion, yielding results in
and the torsional eccentricity (e) of the blade was approximately 29 cm. Table 8.
Accordingly, the torque at the blade root was estimated using Eq. (7): When the blades reached the standup position during the typhoon,
T = ΔP × A × e (7) the Tsai–Wu failure indices for Region C (at one quarter of the length of
the blade from the root of the blade) and D (at one third of the length of
where T is rotor torque at the blade root, ΔP is the assumed maximum
the blade from the root of the blade) in the blade model exceeded one,
pressure difference between both sides of the blade, A is the force area
indicating that the blade fractured in these regions at a wind speed of
of the blade (900,000 cm2), and e is the torsional eccentricity of the
51.2 m/s. This finding was consistent with the actual failure (fractured
blade (29 cm).
blades) of Turbine H12 during Typhoon Soudelor.
On the basis of the estimate thus obtained, the external torque (T)
In contrast, when the blades of the turbine were assumed to be in
on blades at a maximum instantaneous wind speed of 59.5 m/s during
the standby position, even with a designed wind inflow of 70 m/s, the
Typhoon Soudelor was estimated to be 59,780 N·m. Table 7 lists the
Tsai–Wu failure indices for all regions of the blade model were less than
operational details of the turbines that were recorded by the SCADA
one, indicating no fracture of any blade.
during Typhoon Soudelor, and demonstrates that at a wind speed of
40 m/s, the pitch system operated normally; hence, the system was
4.4. Structural mechanics analyses of turbine towers
assumed to be capable of withstanding gusts of this speed.
Based on the results of the incident site investigation and recorded
Table 7
wind velocity data, finite element analysis was conducted on wind
Operating data recorded by nacelle.
turbines in the Port of Taichung during Typhoon Soudelor to examine
Turbine No. Wind speed (m/s) Rotational angle of the Inferred status of the force scenarios of the turbine towers.
blade (Note 1) the pitch system
4.4.1. Turbine force scenarios
H01 43 96–1° Failed
H04 50 92–52° Failed To assess the risk of turbine collapse and to propose improvements,
H06 Incomplete data Incomplete data – force analyses were performed on turbines under different blade con-
H11 Recording Recording disconnected – ditions (standby or standup) at various wind speeds (Fig. 10).
disconnected
H12 30 Rotated constantly at Normal
92°
4.4.2. Analytical procedure
H14 49 91–41° Failed
H18 40 92–86° Normal (1) Establishing a turbine model using ABAQUS FEA.

Note 1: 30° is the standup position of blades and 84° is their standby position.

264
J.-S. Chou et al. Engineering Structures 162 (2018) 257–269

Lower
Middle
Upper

Fig. 9. Simplified wind turbine blade model.

Table 8 the turbine base ring). This prediction was consistent with the findings
Results of applying Tsai–Wu failure criterion to blades (51.2 m/s). of the incident investigation.
Blade position Elemental position Tsai-Wu failure index
(3) Constructing tower section models for the weak planes and esti-
Upper 0.348 mating the bolt stress that corresponded to various bending mo-
A Middle 0.685 ments of the planes that were generated by wind loads during
Lower 0.375 Typhoon Soudelor, and comparing maximum stress with strength of
Upper 0.270 bolt material to evaluate tower safety.
B Middle 0.618
Lower 0.796
The prestress (σpre) of the bolts of all bolt types that were used in
Upper 1.019 each turbine tower were estimated using Eqs. (9) and (10), based on
C Middle 0.999
their corresponding designed torque values. Table 10 presents the es-
Lower 1.018
timated results.
Upper 1.006
D Middle 0.939 T=K×D×N (9)
Lower 0.922
σpre = N / A (10)
Upper 0.708
E Middle 0.708 where T is the design torque of the bolt; K is the torque coefficient
Lower 0.781
(0.127); D is the standard diameter of the bolt; N is the fastening tensile
Upper 0.704 force; σpre is the prestress of the bolt, and A is the effective cross-sec-
F Middle 0.648
tional area of the bolt.
Lower 0.679
Based on the assumption that planes remained planar during
Upper 0.522 bending and that the sectional strain varied linearly, a sectional model
G Middle 0.491
Lower 0.502
of the tower junctions was established using Xtract. Next, the bending
moments and axial force values of the weak planes were input into
Upper 0.377
Xtract to estimate the bolt stress. To determine Xtract settings, bolts
H Middle 0.345
Lower 0.347 were placed under pretensile stress, and the axial force of the weak
planes was applied to connecting section; subsequently, the curvature
Upper 0.218
I Middle 0.203 of that section was increased monotonically from 0 and the corre-
Lower 0.207 sponding sectional moments and bolt strains were analyzed. The max-
Upper 0.098
imum tensile stress of the bolt was interpolated for the moment that
J Middle 0.084 corresponded to each force scenario.
Lower 0.084 The allowable strength ( ρA ) of the bolts, which was the yield
strength of the bolts minus the prestress, was divided by the maximum
tensile stress ( ρS ) of the bolts to yield a safety factor ( ρA / ρS ); a higher
A turbine tower is composed of a wall body and bolts. The failure of safety factor corresponds to a safer tower. A safety factor of one in-
the wall body of the tower can be identified as tensile failure or stress- dicated that bolts yielded at the junctions, whereas a safety factor of
induced local buckling under compression, whereas that of tower bolts less than one suggested that the maximum tensile stress of the bolts
can be identified as tensile failure or shear failure. exceeded their allowable strength, revealing possible failure of the bolts
and, therefore, fracture and collapse of the tower.
(2) Identifying weak planes in the towers; estimating wind pressure
through the method that was described in 4.1, and inputting these
4.4.3. Analyzing force scenarios of turbines
estimates into turbine model to calculate bending moments of
Force scenarios of the turbines in Fig. 10 were examined using the
identified planes.
analytical procedure that was described in the preceding subsection.
Table 11 and Table 12 present the results thus obtained under the
In the standup position at a wind speed of 50.8 m/s, the tensile
following conditions; the minimum instantaneous wind speed was
stress and compressive stress on the tower body were found to be less
50.8 m/s; the maximum instantaneous wind speed was 59.5 m/s, and
than the yield strength of the steel, indicating no tensile or compressive
the blades were in the standup or unfeathered condition. In Force
failure of the tower body. Wind pressure was applied to the turbine
Scenario 9 (wind speed = 50.8 m/s), the maximum stress and allowable
model to estimate the maximum shear force at the sections of tower
strength of the bolts that connected the middle/lower to the upper/
bolts and compare it with the allowable shear strength of the bolt
middle portions of the tower were comparable, indicating that these
material. As Table 9 shows, the allowable shear strength of the tower
bolts approached the yielding state, so the towers might have fractured
bolts was higher than the maximum sectional shear force, so the bolts
at these two connection sites, and particularly in that of the middle and
could sufficiently resist the shear force at a wind speed of 50.8 m/s.
lower portions (17 m above the ground).
Based on the above, the potential weak planes of the tower structure
In Force Scenario 14 (wind speed = 59.5 m/s), the maximum stress
were predicted to be associated with the tensile failure of the bolts at
of the bolts that connected the turbine base ring to the upper portion of
the four junctions (which were respectively 0, 17, 42, and 62 m from
the tower, to the middle and lower portions of the tower, and to the

265
J.-S. Chou et al. Engineering Structures 162 (2018) 257–269

Force scenarios of
wind turbines
Shutdown wind Design wind
speed 25m/sec speed 70m/sec
standup position Normal
(Normal operation) operation
Pitch system
Failed Scenario 18 Scenario 19
Scenario 1 Maximum
instantaneous wind
speed Normal
59.5m/sec operation Failed

Unfeathered blades Number of blade fractured 1/2/3

Scenario 20 Scenario 21
Scenario 2

Scenario 5 Scenario 6 Scenario 7

Scenario 3 Scenario 4 Maximum instantaneous Maximum instantaneous

wind speed wind speed
59.5m/sec 59.5m/sec

Unfeathered blades Number of blade fractured 1/2/3 Unfeathered blades Number of blade fractured 1/2/3

Scenario 8 Scenario 9 Scenario 10 Scenario 11 Scenario 12 Scenario 13 Scenario 14 Scenario 15 Scenario 16 Scenario 17

Fig. 10. Force scenarios of wind turbines.

Table 9 Table 10
Allowable shear strength and maximum sectional shear force of tower bolts. Design torque and prestress of bolts at each junction of tower.

Bolt position Number of Type of Allowable shear Maximum sectional Bolt position Tower No. Type of bolt/Number Design Prestress
(m) bolts bolt strength (tf) shear force (tf) (m) of bolts torque (kgf/cm2)
(N·m)
62 72 M30 480.7 168.1
42 90 M30 600.8 194.4 62 H02, H03 M30/72 1300 6199.9
17 128 M30 854.5 224.7 H01, H04–18 1170 5579.9
0 142 M36 948.0 237.5
42 H02, H03 M30/90 1800 8584.5
H01, H04–18 1350 6438.4

upper and middle portions of the tower, exceeded their allowable 17 H04–18 M30/128 1900 9061.4
H02, H03 1800 8584.5
strength, revealing that these bolts had approached their yielding state.
H01 1500 7153.8
Notably, bolts that connected the middle and lower portions of the
0 H02, H03 M36/142 3300 9106.0
tower exhibited the smallest ratio of allowable strength to maximum
H01, H04, H08 2750 7588.3
tensile stress. Therefore, this connection site is associated with a H05–07, 2338 6451.5
structural weakness of the tower and was likely to have been the site of H09–18
failure. This finding is consistent with the actual failure condition of the
turbine.
Table 13 presents the results of analyses in the 21 force scenarios of of the towers. In Force Scenario 9, turbines H01, H04, H06, H11, H14,
the turbines. Force Scenarios 8 and 9 (wind speed = 50.8 m/s) and 13 and H18 might have fractured at the middle/lower (17 m above the
and 15 (wind speed = 59.5 m/s) exhibited possible structural failures ground) and upper/middle (42 m above the ground) connection sites.

266
J.-S. Chou et al. Engineering Structures 162 (2018) 257–269

Table 11 buildings in Taiwan [23], the spacing between extra (newly added) and
Maximum tensile stress and allowable bolt strength in Force Scenario 9. existing bolts, S (the distance from the center of an extra bolt hole to
ρA
that of an existing one), must be three times the bolt diameter; and the
Bolt Tower Bolt Maximum Allowable Turbines
position No. strength tensile strength ρA ρS
with clearance between extra bolts and the edges of steel plates, Lc, must be
(m) (grade) stress ρS (kgf/cm2) damaged one and a half times the bolt diameter. Furthermore, extra bolts should
(kgf/cm2) tower be symmetrically inserted into sections to give towers sufficient re-
sistance in all directions.
62 H02, H03 12.9 0 6103.9 >3 –
H01, 10.9 0 5123.5 >3
Table 14 presents the number of extra bolts (proposed in this study)
H04–18 in relation to the safety factors for the structural sections of towers at a
maximum instantaneous wind speed of 59.5 m/s and a minimum of
42 H02, H03 12.9 2329.0 3719.3 1.60 H01
H01, 10.9 4210.7 4265 1.01 H01 50.8 m/s. This table presents the number of bolts recommended to be
H04–18 added to tower sections on the basis of different levels of wind speed
17 H04–18 12.9 3279.3 3242.4 0.99 H04 during Typhoon Soudelor. The safety factors in the table were obtained
H02, H03 12.9 3719.6 3719.3 1.00 H06 by dividing allowable bolt strength by the maximum tensile stress of a
H01 12.9 4915.6 5150 1.05 H11 bolt.
H14
H18
5. Conclusions and recommendations
0 H02, H03 12.9 1203.2 3197.8 2.66 –
H01, 10.9 2160.3 3512.6 1.63
Based on the results in this investigation, this section draws con-
H04, H08
H05–07, 10.9 3324.5 4649.4 1.40 clusions and offers suggestions concerning (1) the blade-pitch system,
H09–18 (2) blade force, and (3) forces on bolts in tower.

5.1. Conclusions
Table 12
Maximum tensile stress and allowable bolt strength in Force Scenario 14.
(1) Blade-pitch system evaluation
Bolt Tower Bolt Maximum Allowable ρA Turbines
position No. strength tensile strength ρA ρS
with If turbine blades had been in the standby position when the wind
(m) (grade) stress ρS (kgf/cm2) damaged was at its maximum wind speed of 59.5 m/s during Typhoon Soudelor,
(kgf/cm2) tower then the external torque on the pitch systems was determined to be
59,780 N·m, whereas the actual torsional strength of the motors of the
62 H02, H03 12.9 0 6103.9 >3 –
H01, 10.9 0 5123.5 >3 systems was merely 27,117 N·m at a wind speed of 40 m/s.
H04–18 Consequently, the pitch systems had been damaged and the blades were
42 H02, H03 12.9 6292.9 3719.3 0.59 H01 not in the standby position, increasing the exposure of the turbines to
H01, 10.9 8405.0 4265 0.51 H01 wind, elevating the wind force on the tower bolts.
H04–18

17 H04–18 12.9 7736.6 3242.4 0.41 H04 (2) Blade force analysis
H02, H03 12.9 8270.6 3719.3 0.44 H06
H01 12.9 9609.9 5150 0.53 H11 At the maximum instantaneous wind speed of 51.2 m/s (as mea-
H14
sured by Turbine H12) and the blades in the standup position, a blade
H18
force analysis that applied the Tsai–Wu failure criterion revealed that
0 H02, H03 12.9 4270.9 3197.8 0.75 –
the force in the region one third of the length of the blade from the rotor
H01, 10.9 5884.7 3512.6 0.60
H04, H08 joint reached the critical value (Tsai–Wu failure index: 0.99–1.02).
H05–07, 10.9 6766.9 4649.4 0.69 Moreover, if the variability of blade material was accounted for, the
H09–18 blades were also likely to have been damaged at a minimum in-
stantaneous wind speed of 50.8 m/s (measured by all six collapsed
turbines).
Force Scenario 15 revealed how Turbine H01 collapsed, even though
only one of its blades was fractured. (3) Analysis of forces on bolts in tower
If the pitch system fails, then turbine towers are likely to collapse at (a) Tower bolts differ in specified torque, so the effects of bolt
a wind speed of 50.8 m/s. If the system operates normally, then the prestress on bolt strength must be considered. In this study, in
towers are likely to remain undamaged even at a wind speed of 70 m/s. the tower bolt force analysis, prestress was deducted from bolt
Thus, whether the pitch system operates normally has profound im- strength (which was obtained from technical documents from
plications for the safety of turbine towers. These towers may continue the turbine owner) to provide a yardstick for assessing bolt
to stand when their blades are fractured, reducing their wind exposure safety, with a minimum instantaneous wind speed of 50.8 m/s
to the wind; this condition suggests the significance of the strong- and the blades in the standup position. The results showed
column (tower)-and-weak-beam (blades) principle for wind turbine possible failures of bolts at 17 m and 42 m above the ground
tower design. (particularly those at 17 m), consistent with the actual failures
of the six collapsed turbines. (Turbines H04, H06, H11, H14,
4.5. Suggestions for reinforcing tower structures and H18 fractured 17 m above the ground, whereas Turbine
H01 fractured 42 m above the ground.)
Anchoring blades in the standby position to prevent typhoon-in- (b) With a maximum instantaneous wind speed of 59.5 m/s and
duced damage to junction bolts may be the most effective long-term blades under the standup condition, tower bolt force analysis
means of strengthening tower structures. To handle extreme weather, indicated that the bolts might have been damaged at the tower
bolts of higher strength (of Grade 12.9, for example) can be installed in base and 17 m and 42 m above the ground. Moreover, if only
flanges (Fig. 11). one blade fractured, then junction bolts might have snapped at
According to the construction specifications of steel structures for 17 m and 42 m above the ground. If two blades fractured, then

267
J.-S. Chou et al. Engineering Structures 162 (2018) 257–269

Table 13
Results of analysis of force scenarios.

Scenario Condition Wind speed Pitch system No. of fractured Safety factor Safety factor Safety factor Safety factor
(m/s) blades (62 m) (42 m) (17 m) (0 m)

1 Shutdown wind speed 25 Normal 0 >3 >3 >3 >3

2 25 Normal 0 >3 >3 >3 >3

3 Maximum instantaneous wind speed measured during 59.5 Normal 0 >3 >3 >3 >3
4 Typhoon Soudelor 59.5 Normal 0 >3 >3 >3 >3
5 59.5 Normal 1 >3 >3 >3 >3
6 59.5 Normal 2 >3 >3 >3 >3
7 59.5 Normal 3 >3 >3 >3 >3

8 Minimum instantaneous wind speed measured during 50.8 Failed 0 >3 1.04 1.02 1.58
9 Typhoon Soudelor 50.8 Failed 0 >3 1.01 0.99 1.4
10 50.8 Failed 1 >3 2.06 2.37 1.4
11 50.8 Failed 2 >3 >3 >3 >3
12 50.8 Failed 3 >3 >3 >3 >3

13 Maximum instantaneous wind speed measured during 59.5 Failed 0 >3 0.52 0.42 0.61
14 Typhoon Soudelor 59.5 Failed 0 >3 0.51 0.41 0.6
15 59.5 Failed 1 >3 0.94 0.87 1.29
16 59.5 Failed 2 >3 2.52 >3 >3
17 59.5 Failed 3 >3 >3 >3 >3

18 Design wind speed 70 Normal 0 >3 >3 2.86 2.99

19 70 Normal 0 >3 >3 2.76 2.91
20 70 Failed 0 >3 0.3 0.23 0.34
21 70 Failed 0 >3 0.3 0.23 0.33

Table 14
Efficiency of increasing number of bolts.

Bolt Specs Wind Number of Percentage of Minimum

position speed extra bolts extra bolts safety factor
(m) (m/s) (%)

42 M30/12.9 59.5 8 8.9 0.68

16 17.8 0.79
24 26.7 0.86
32 35.6 1
40 44.4 1.2
48 53.3 1.43

17 M30/12.9 59.5 44 34.4 0.68

56 43.8 0.78
Fig. 11. Additional bolts installed in a flange of tower. 68 53.1 0.9
80 62.5 1.02
92 71.9 1.17
stress on the bolts would have fallen to the safety level, pre- 50.8 4 3.1 1.01
8 6.3 1.02
venting the tower from collapsing. This latter condition corre- 12 9.4 1.07
sponded with the actual failure of Turbine H12, which re- 16 12.5 1.23
mained standing with two fractured blades. 20 15.6 1.3
(c) With the manufacturer-specified design wind speed of 70 m/s 24 18.80 1.38
28 21.90 1.48
and blades in the standup position, tower bolt force analysis
revealed that the force on the bolts at the tower base and 17 m 0 M36/12.9 59.5 8 5.60 0.86
16 11.30 0.97
and 42 m above the ground greatly exceeded their allowable
24 16.90 1.12
strength. Even if 128 bolts (the maximum number of extra bolts 32 22.60 1.19
that is stipulated in the regulation [23]) are added to the 40 28.20 1.45
junction bolts of the tower 17 m above the ground, then the
force on that junction may still surpass the allowable strength of
bolts, so the junction may not be able to withstand winds that (2) Blade force
blow at up to 70 m/s.
The blade force analysis revealed that turbines are exposed to less
wind when their blades are destroyed by gusts, reducing gust-induced
5.2. Recommendations damage to the tower and nacelle. Accordingly, blades can be perceived
as improving the safety of turbine towers when appropriate blade
(1) Blade-pitch system strength is designed.

No data on the design strength of the pitch system and its blade
locks could be obtained from the manufacturer. If such data were ever
to be obtained, then the actual strength of the pitch system could be
estimated, and plans for improving the system could be formulated on
the basis of the external blade torque (59,780 N·m).
(3) Forces on bolts in tower
(a) To prevent tower collapse when the pitch system fails during a
typhoon, extra bolts can be used to improve strength of the
tower flanges. Section 4.5 presents the recommended numbers
of bolts to be added to towers to withstand severe storms.

268
J.-S. Chou et al.
(b) Specific bolt torque vary across towers, leading to differences in
bolt prestress and strength. Moreover, manufacturers’ manuals
and construction plans specify different torques for tower bolts,
making installation of the bolts difficult. Thus, torque can be
specified based on the strength and size of the bolts, minimizing
the variability of the bolt strength of turbines.
References
[1] Valamanesh V, Myers AT, Arwade SR. Multivariate analysis of extreme metocean
conditions for offshore wind turbines. Struct Saf 2015;55:60–9.
[2] Thies PR, Johanning L, Karikari-Boateng KA, Ng C, McKeever P. Component relia-
bility test approaches for marine renewable energy. Proc Inst Mech Eng Part O: J
Risk Reliab 2015;229:403–16.
[3] Chou J-S, Chiu C-K, Huang IK, Chi K-N. Failure analysis of wind turbine blade under
critical wind loads. Eng Fail Anal 2013;27:99–118.
[4] Chou J-S, Tu W-T. Failure analysis and risk management of a collapsed large wind
turbine tower. Eng Fail Anal 2011;18:295–313.
[5] Brown S. Forensic engineering: reduction of risk and improving technology (for all
things great and small). Eng Fail Anal 2007;14:1019–37.
[6] Chen X, Xu JZ. Structural failure analysis of wind turbines impacted by super ty-
phoon Usagi. Eng Fail Anal 2016;60:391–404.
[7] Wang Z, Zhao Y, Li F, Jiang J. Extreme dynamic responses of MW-level wind turbine
tower in the strong typhoon considering wind-rain loads. Math Probl Eng
2013;2013:1–13.
[8] Li Z-q, Chen S-j, Ma H, Feng T. Design defect of wind turbine operating in typhoon
activity zone. Eng Fail Anal 2013;27:165–72.
[9] Chen X, Li C, Xu J. Failure investigation on a coastal wind farm damaged by super
typhoon: a forensic engineering study. J Wind Eng Ind Aerodyn 2015;147:132–42.
[10] Ishihara T, Yamaguchi A, Takahara K, Mekaru T, Matsuura S. An analysis of da-
maged wind turbines by typhoon maemi in 2003. In: Proceedings of the sixth asia-
269
Engineering Structures 162 (2018) 257–269
pacific conference on wind engineering; 2005. p. 1413–28.
[11] Caithness Windfarm Information Forum. Wind turbine accident and incident
compilation. < http://www.caithnesswindfarms.co.uk/AccidentStatistics.htm > ;
2016.
[12] Shafiee M, Dinmohammadi F. An FMEA-based risk assessment approach for wind
turbine systems: a comparative study of onshore and offshore. Energies
2014;7:619–42.
[13] Jin X, Ju W, Zhang Z, Guo L, Yang X. System safety analysis of large wind turbines.
Renew Sustain Energy Rev 2016;56:1293–307.
[14] Lebranchu A, Charbonnier S, Berenguer C, Prevost F. Review and analysis of SCADA
data-based methods for health monitoring of wind turbines. In: 25th European
Safety and Reliability Conference, ESREL 2015, September 7, 2015–September 10,
2015. Zurich, Swaziland: CRC Press/Balkema; 2015. p. 2413–21.
[15] Wang L, Zhang Z, Xu J, Liu R. Wind turbine blade breakage monitoring with deep
autoencoders. IEEE Trans Smart Grid 2016:1–10.
[16] Li H, Yang C, Hu YG, Liao XL, Zeng Z, Zhe C. An improved reduced-order model of
an electric pitch drive system for wind turbine control system design and simula-
tion. Renew Energy 2016;93:188–200.
[17] SGS Taiwan Ltd., Material & Engineering Laboratory, Test report:HV-11-05101X.
SGS Taiwan Ltd.; 2011.
[18] Pliego Marugán A, García Márquez F, Pinar Pérez J. Optimal maintenance man-
agement of offshore wind farms. Energies 2016;9. 46, 1–20.
[19] Park R, Paulay T. Ductile reinforced concrete frames – some comments on the
special provisions for seismic design of ACI 318–71 and on capacity design. Bull NZ
Soc Earthquake Eng 1975;8:70–90.
[20] Wind Resistance Design Specifications and Commentary of Buildings. Taiwan:
Ministry of Interior; 2014.
[21] Abaqus Analysis User's Manual Version 6.12; 2012.
[22] Tsai Y-S, Chiu P-H, Lin H-J, Huang H-H. Composite laminate design of an offshore
wind turbine blade (In Chinese). National Taiwan University, Taipei, Taiwan; 2014.
[23] CAPA. Construction Specifications of Steel Structures for Buildings. Taiwan:
Ministry of the Interior; 2007.