Fatigue Failure Accident of Wind Turbine Tower in

Taikoyama Wind Farm

Yin LIU1, Takeshi ISHIHARA2

1,2Department of Civil Engineering, School of Engineering, the University of Tokyo, Tokyo, Japan

Abstract

One of the wind turbine nacelles at Taikoyama wind farm collapsed due to the fatigue failure of high tension

bolts. Strain gauges and accelerometers were installed on the wind turbine to verify the aerodynamic model.

Furthermore a FEM model was built in order to find out the relationship between tower tube and high tension

bolts at the position of flange joint, where the fracture occurred. When the bolt’s pre-tension force decreases,

its stress range increases. Less the pretension force left, the larger the stress range will be. Hence when pre-

tension force is 0%, the fatigue life is left for only a few days. On the other hand when 17 bolts are damaged,

the turbine tube stress is three times larger than the stress when all the bolts are in good condition. Hence the

fatigue evaluation shows that the life time rapidly decreases to less than two months compared with that of the

normal life time which is 20 years.

Key Words: Fatigue failure, pre-tension force, high tension bolt, nacelle collapse.

1. Introduction

The Taikoyama wind farm is located at the top of

Taikoyama Mountain, Kyoto Prefecture, Japan,

which is surrounded by the Tango peninsular and

faces north to the Sea of Japan. The construction

cost is approximately 12.5 million dollars and it

reduces nearly 5900 tons of carbon dioxide every

year. The wind farm information is summarized in

Table 1.

In March 2013 the nacelle of No.3 wind turbine

collapsed[1] and the accident scene and schematic

diagram of the wind turbine is shown in Fig. 1.

1 Presenting and corresponding author, PhD candidate, E-mail: liuyin@bridge.t.u-tokyo.ac.jp

Table 1 Summary of Taikoyama wind farm

Name

Operating time

Manufacturer

Unit

Max power output

Taikoyama Wind Farm

15th, November, 2001

Lagerwey

6×750kW

4500kW

Performance

Cut-in wind speed

Rated wind speed

Cut-out wind speed

Resistant wind speed

3m/s

12m/s

25m/s

60m/s

Rotor

Diameter

Generation rotor speed

Number of blades

Hub height

50.5m

13~33rpm

3

50m

Tower Height

Material

46m

SM400 (steel)

Flange connection

high-tension bolts F10T M24

Nacelle Dimensions

Material

W5.6×L3.3×H6.5m

SS400, GFRE

Wind direction

control Control method Active yaw control

Rated power

output control Control method Pitch control

(a) Collapsed nacelle

(b) Fracture section (c) Vertical cross section

Fig. 1 Accident scene and schematic diagram

The detailed structure is shown in Fig. 2.

(a) Flange joint (b) Fracture section in detail

Fig. 2 Detail drawing of fracture section

The field investigation indicates that the wind

condition satisfied the construction requirement

based on the IEC 61400-1[2] including annual wind

speed, turbulence intensity and flow inclination angle.

By observing the fracture section of the tower tube,

we found that the material strength was strong

enough, but evidence of fatigue crack propagation

was detected at the inner surface of the tube.

Furthermore, 17 broken bolts were found during the

field investigation and fatigue cracks were also

detected. By comparing the two aspects, fracture is

considered to be preceded by a certain degree of

fatigue damage caused by the reduction of bolts pre-

tension force up to 30%~100%.

The wind turbine collapsed very early in 12 years,

where the expected life period was 20 years.

Moreover, the accident happened only three months

after the periodical inspection was carried out.

Additionally, there are more than 120 wind turbines

in service of the same type across Japan. Therefore,

it is necessary and urgent to understand the cause

of this accident, so that this kind of accident can be

prevented in the future.

This paper proceeds as follows: 1) Field

measurement; 2) Aerodynamic modelling and

verification; 3) Clarify the fracture section’s

aerodynamic characteristics; 4) Explain the

relationship between nominal stress, local stress and

bolt stress using FEM model; 5) Evaluate the fatigue

life of both high-tension bolt and tower tube, and

reveal the reason for the failure.

2. Field measurement

2.1 Wind condition investigation

All the data were measured from Feb. 2nd 2015 to

Feb. 28th 2015.

Fig. 3 Occurrence frequency Fig. 4 Average wind speed

Fig. 3 and Fig.4 indicate the occurrence frequency

and average wind speed respectively. The

occurrence frequency of dominate wind direction

WSW, W and WNW is 9%, 27% and 15%

respectively.

Since the SCADA data contains only maximum wind

speed and average wind speed in a time scale of 1

minute, we calculated the turbulence intensity

according to reference [3] in equation (1).

𝐼𝑝 =𝑈𝑚𝑎𝑥 𝑈𝑚𝑒𝑎𝑛−1⁄

𝑃, 𝑃 =

1

2𝑙𝑛

𝑇

𝑡 (1)

0%

10%

20%

30%N

NNE

NE

ENE

E

ESE

SE

SSES

SSW

SW

WSW

W

WNW

NW

NNW

02468

1012

NNNE

NE

ENE

E

ESE

SE

SSES

SSW

SW

WSW

W

WNW

NW

NNW

50m

45.94m

0m

Fracture section

Local structur

Flange

Welding 10mm below flange

Fracture section

The maximum wind speed Umax and average wind

speed Umean are derived from the 10min SCADA data,

the peak factor P is evaluated by a time scale T of

600 seconds and average time t of 1 second.

Consequently 1m/s bin average is calculated. Fig. 5

shows the field turbulence intensity.

Because of the insufficient high wind speed data

(>17m/s) during the measurement period, the high

wind speed turbulence intensity is extrapolated

assuming the normal turbulence intensity in

reference[2], and it is described as equation (2)

𝜎1 = 𝐼𝑟𝑒𝑓(0.75𝑉ℎ𝑢𝑏 + 𝑏), 𝑏 = 3.8 (2)

Iref is the expected value of hub-height turbulence

intensity at a 10 min average wind speed of 15m/s,

Vhub is the wind speed at hub height and 𝜎1 is hub-

height longitudinal wind velocity standard deviation.

As a result for aerodynamic simulation, a combined

turbulence intensity is used: measurement value for

low wind speed ( ≪ 17m/s ) and the extrapolated

value for high wind speed respectively (>17m/s).

0

0.1

0.2

0.3

0.4

0.5

0 5 10 15 20 25

MeasurementExtroplatedbin average

Tu

rbu

lence in

ten

sity

Wind speed (m/s)

Fig. 5 Turbulence intensity in the direction of WSW+W+WNW

For the turbulence spectrum, the Kaimal model is

used. The lateral and vertical turbulence intensity

component are considered as 0.8 𝜎1 and 0.5 𝜎1

according to reference [2].

2.2 Moment measurement

Strain gauges with sampling frequency of 20Hz were

installed in eight directions in order to get the

moment at the height of 12.6m above tower base. Fig.

6 shows the strain gauges installment.

The nacelle was forced to rotate one circle without

operating for the estimation of the strain gauges’

installment error, and the compensation value can be

calculated by the amplitude of the sin curve.

Fig. 6 Strain gauges installment

Fig. 7 Moment calculation schematic diagram

The measurement moment was calculated following

the method by Ishihara and Phuc[4]. According to Fig.

7, the East-West moment and South-North moment

were given in equation (3) and (4) respectively.

Where M and ε is the moment and strain at

corresponding direction, EI is the stiffness of tower tube

and D is the inner diameter.

𝑀𝐸𝑊 = 𝐸𝐼𝜀

𝐷= 𝐸𝐼

𝜀𝐸−𝜀𝑊

𝐷 (3)

𝑀𝑆𝑁 = 𝐸𝐼𝜀

𝐷= 𝐸𝐼

𝜀𝑆−𝜀𝑁

𝐷 (4)

The total moment is given in equation (5). If the

direction of total moment is opposite to the nacelle

direction, then the total moment will be positive,

otherwise it is negative.

𝑀𝑡𝑜𝑡𝑎𝑙 = √𝑀𝐸𝑊2 + 𝑀𝑆𝑁

2 (5)

The average bending moment, maximum bending

moment and standard deviation of bending moment

are plotted in Fig. 8.

ε𝐸

ε𝐸

ε𝐸

ε𝐸ε𝐸

Nacelle direction ε𝑊

ε𝑠

ε𝐸

ε𝑁

M𝐸𝑊

M𝑆𝑁

M𝑡𝑜𝑡𝑎𝑙

α θ

-1000

0

1000

2000

3000

4000

5000

0 5 10 15 20 25

measurement

bin average

Mo

men

t (k

Nm

)

Wind speed (m/s)

-1000

0

1000

2000

3000

4000

5000

0 5 10 15 20 25

measurement

bin average

Mo

men

t (k

Nm

)

Wind speed (m/s)

(a) Average moment (b) Maximum moment

-1000

0

1000

2000

3000

4000

5000

0 5 10 15 20 25

measurement

bin average

Mo

men

t (k

Nm

)

Wind speed (m/s) (c) Standard deviation of moment

Fig. 8 Comparison of measurement and bins average moment

3. Aerodynamic analysis and

fatigue life investigation

3.1 Aerodynamic modelling

Aerodynamic model is built to simulate the dynamic

performance by GL’s Bladed wind turbine modelling

tool[ 5 ]. The tower section refers to the real

engineering drawings. For commercial confidentiality,

the blade profile is not available from manufacturer.

As a result we selected airfoils from NREL’s airfoil

family, which are S818 for root section, S830 for

primary section and S831 for tip section[ 6 ], and

thickness/chord ratio, Reynolds number, lift

coefficient Cl and draft coefficient Cd were

determined.

For control method, some adjustment had been

applied. In case of the high turbulence intensity in the

mountainous area, the wind turbine encounter over

speed at times. Once it exceeds the maximum rotor

speed of 33 rpm, it stops suddenly and starts to

operate again when the rotor speed drops below the

maximum value which causes frequent downtime.

Hence the manufacturer modified the maximum rotor

speed and power output to decrease the downtime.

Since the details were commercial confidentiality, we

adjust rated power output and maximum rotor speed

according to the measurement data. Moreover a five

degrees pitch angle error is considered to eliminate

the error in pitch control. With the adjustment above

the power output, rotor speed and pitch angle are

now close to the measurement data as shown in Fig.

9.

0

100

200

300

400

500

600

700

800

0 5 10 15 20 25

measurement

simulation modified

simulation default

Po

wer

ou

tpu

t (k

W)

Wind speed (m/s)

0

5

10

15

20

25

30

35

0 5 10 15 20 25

measurementsimulation modifiedsimulation default

Ro

tor

spee

d (

rpm

)

Wind speed (m/s)

(a) Power output (b) Rotor speed

0

5

10

15

20

25

30

0 5 10 15 20 25

measurementsimulation modifedsimulation default

Pit

ch a

ngle

(deg

)

Wind speed (m/s)

(c) Pitch angle

Fig. 9 Comparison of power output, rotor speed and pitch angle

The proportional gain KQP and integral gain KQI for

torque control, and proportional gain KSP and integral

gain KSI for pitch control were calculated based on

Guidelines for Design of Wind Turbine Support

Structures and Foundations, JSCE[7 ],and optimal

mode gain Kopt was modified to validate the dynamic

simulation results with measurement results.

Some key parameters for Bladed modelling are

summarized in Table 2.

Table 2 Key parameters for Bladed modelling

Optimal mode

gain Kopt

Demanded generator

toque (Nm)

Rated Power

generation (kW)

Rotor speed

(rpm)

Error in Pitch

angle (degree)

Torque

control Pitch control

Default 22583.5 216450 750 33rpm 0 KQP=789139

KQI=516780

KSP=0.458180

KSI=0.847957

Modified 23340.2 231387 630 26rpm 5 KQP=461249

KQI=176551

KSP=0.492799

KSI=0.771005

A field test was carried out to measure the natural

frequency of the tower. The damping ratio of the 1st

order frequency was applied as 0.5% based on the

field inspection [1]. The natural frequency is shown

in Table 3, which is consistent with the aerodynamic

simulation result.

Table 3 Comparison of tower natural frequencies

Tower natural frequencies Measurement Simulation

1st order (fore-art) 0.515Hz 0.533

1st order (side-side) 0.518Hz 0.533

2nd order (fore-art) 3.838Hz 3.685

2nd order (side-side) 3.832Hz 3.578

Finally, Fig. 10 shows the measurement and

simulation results for moment at 12.6m above tower

base were in good agreement, and the aerodynamic

model is verified to be correct.

-1000

0

1000

2000

3000

4000

5000

6000

0 5 10 15 20 25

measurementsimulation modifiedsimulation default

Mom

en

t (k

Nm

)

Wind speed (m/s)

-1000

0

1000

2000

3000

4000

5000

6000

0 5 10 15 20 25

measurementsimulation modifiedsimulation default

Mom

ent

(kN

m)

Wind speed (m/s)

(a) Average moment (12.6m) (b) Std of moment (12.6m)

-1000

0

1000

2000

3000

4000

5000

6000

0 5 10 15 20 25

measurementsimulation modifiedsimulation default

Mom

ent

(kN

m)

Wind speed (m/s)

(c) Maximum moment (12.6m)

Fig. 10 Comparison of moment

3.2 Characteristics of fracture section

Fig. 11 (a) and Fig. 11 (b) show simulated axial force

N and bending moment M at the tower fracture

section (45.94m) at different wind steps respectively

according to simulation result.

Hence the nominal stress can be calculated from

equation (6), where A is the sectional area and Z is

the sectional resistance moment.

σ𝑛 =𝑁

𝐴−

𝑀

𝑍 (6)

-800

-700

-600

-500

-400

-300

0 5 10 15 20 25

MinAverageMax

Ax

ial

forc

e (

kN

)

Wind speed (m/s)

-1000

-800

-600

-400

-200

0

0 5 10 15 20 25

MinAverageMax

Mo

men

t (k

Nm

)

Wind speed (m/s)

(a) Axial force N (45.94m) (b) Bending moment M (45.94m)

-15

-10

-5

0

5

10

15

20

25

0 5 10 15 20 25

MaxAverageMin

No

min

al

stre

ss (

N/m

m2)

Wind speed (m/s)

c) Nominal stress (45.94m)

Fig. 11 Aerodynamic characteristics at the fracture section

As shown in Fig. 11 (c), the nominal stress σ𝑛

changes and varies with the increase of the wind

speed. The minimum stress turns into negative value

when the wind speed is above 18 m/s.

3.3 FEM modelling

The fracture section is very close to the top flange

welding position, and according to the field

investigation the fatigue failure propagated at the

inner surface of the tower tube, so the stress

concentration and spatial effect may influence the

local stress σ𝑙𝑜𝑐𝑎𝑙 significantly. A 3D FEM model is

Thrust force by wind

built to clarify the relationship between nominal

stress σ𝑛 ， local stress σ𝑙𝑜𝑐𝑎𝑙 and bolt get pre-

tension force before and after the bolts damaged.

The relationship of nacelle weight, thrust force and

top flange is illustrated in Fig. 12. The nacelle weighs

53.3t and it is rigidly connected to the yaw bearing.

The stress concentration factor of welding geometric

profile was proposed by Caccese[8]. The case for

Taikomaya wind turbine is as shown in Fig. 13. Solid

element is used for the modelling of yaw bearing, top

flange and bolts, and shell element is used for tower

tube modelling. Furthermore, contact element is

considered for the contact surface of yaw bearing

and top flange and the friction factor is 0.2. The bolts

are rigidly connected to the yaw bearing.

Fig. 12 Force applying position relationship

Fig. 13 FEM detail at top flange position

3.4 Investigation of the tower tube

fatigue life

As for the tower tube, Fig. 14 shows the cases when

17 bolts broken.

Fig. 14 Diagram of the damage area

Thrust force is considered in seven cases from 0kN

to 250kN to simulate different wind loading. Fig. 15

shows an example of the local stress σ𝑙𝑜𝑐𝑎𝑙 before

and after 17 bolts are damaged at wind speed of

16m/s.

Fig. 15 (a) implies that the cause of maximum tensile

stress happens at the inner tube because of the law

of lever, which is consistent with the observation of

fracture face. According to Fig. 15 (b), the local

stress is much larger when 17 bolts are broken.

(a) Bolts normal

(b) 17 Bolts broken

Fig. 15 Comparison of the local stress (16m/s)

The relationship between nominal stress and local

stress considering the welding stress concentration

[7] is now given as following respectively:

Bolts normal

𝜎𝑙𝑜𝑐𝑎𝑙 = −3.05 + 2.65𝜎𝑛 (7)

17 bolts broken

𝜎𝑙𝑜𝑐𝑎𝑙 = −10.6 + 6.35𝜎𝑛 + 0.16𝜎𝑛2 (8)

Equation (7) and (8) are plotted in Fig. 16. When 17

W

N

S

C

L

5

m

m

Contact

element

Yaw bearing

Top flange

Tower tube

Shell element

Fracture section

Yaw bearing

Flange

Tower tube

Welding

Local stress

Bolts

Yaw bearing

Flange

Tower tube

Welding

Local stress

Bolts

Nacelle opposite side

E (0°)

Edge of damaged

Bolts (53.6°)

Top flange

Hub Height (=GL+50.0m) Nacelle weight (53.3t)

Thrust force by wind

Nacelle center

of gravity

4000m

1525mm

Yaw bearing

Lee wind side (180°)

Rotor side (0°)

bolts are broken, the local stress is more than three

times larger than bolts at normal condition.

-100

-50

0

50

100

150

200

-20 -10 0 10 20

Bolts normal17 Bolts broken

Lo

cal

stre

ss

local

(N/m

m2)

Nominal stress n (N/mm2)

0 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0 6 0 0

1 7 b o l t s b r o k e n

B o l t s n o r m a l

- 5 0

0

50

100

150

200

Lo

cal str

ess

local (N

/mm

2)

time (s) Fig. 16 Local stress vs. Fig. 17 Time history of local

nominal stress stress (22m/s)

With a time period of 10 minutes, the time series

simulation result is available for each wind speed

combining aerodynamic model with equation (7) and

(8). When the wind speed is low, the tensile stress

predominates. However with increase in wind speed,

compressive stress occurs and the stress amplitude

increases. The case of wind speed at 22m/s is shown

in Fig.17.

With the time history of bolt pre-tension stress, we

can investigate its fatigue life. Rain flow counting

algorithm is used for fatigue analysis in order to

reduce the spectrum of varying stress into a set of

simple stress reversals. Goodman relation as shown

in equation (9) is used to quantify the interaction of

mean and alternating stresses.

𝜎𝑎 = 𝜎𝑤(1 − 𝜎𝑚/𝜎𝐵) (9)

𝜎𝑎 is the alternating stress from rain flow counting

result, 𝜎𝑚 is the mean stress, 𝜎𝑤 is the fatigue limit

for comple`tely reversed loading and 𝜎𝐵 is the

ultimate tensile strength of the material, which is

493Mpa for SM400 steel.

By using the fatigue limit for completely reversed

loading 𝜎𝑤, S-N curve based on GL wind 2005 with

a detail category of 71[ 9 ], and Miner’s rule, the

accumulative fatigue damage D in 10 minutes is

given in Equation (10), and failure is reached when

D equals to 1.

∑𝑛𝑖

𝑁𝑖

𝑘𝑖=1 = 𝐷 (10)

Frequency distribution of the wind speed is based on

Rayleigh distribution with a mean annual wind speed

of 8.5m/s.

The fatigue life of tower tube is shown in Fig. 18.

When the bolts are in normal condition the fatigue life

is 27.5 years, which is in agreement with the design

requirement. However, if 17 bolts are broken, the

fatigue life decreases dramatically to 0.09 years,

approximately one months. It is in accordance with

the time interval between the last periodical

inspection and the accident.

0

5

10

15

20

25

30

17 bolts brokenBolts normal

Tow

er

tube

fa

tig

ue life (

ye

ar)

27.5

0.09

17 bolts broken Bolts normal

Fig. 18 Tower tube fatigue life

3.5 Investigation of the high tension

bolts fatigue life

Based on the field investigation [1], six bolts at

nacelle’s opposite side were found to have reduction

pre-tension force reduced as shown in Fig. 19.

Fig. 19 Bolts pre-tension force decreasing

In order to recreate the real situation, blots pre-

tension force is set in six different cases which were

100%, 80%, 60%, 40%, 20% and 0% of the design

pre-tension force corresponding to 850kNm torque.

The relationship between the nominal stress and bolt

pre-tension stress is given as shown in Fig. 20. With

the nominal stress increasing, the gradient increases

as pre-tension decreases, and it is much more

obvious when the pre-tension force decreases. The

larger the gradient the larger the bolt stress range will

be, and the bolt’s fatigue load. Since the nominal

E W

N

S

Nacelle’s opposite side

stress ranges mainly between -5N/mm2 to 25 N/mm2

according to Fig. 11(c), the stress range may vary a

lot especially when the bolts pre-tension stress drops

to 0% as illustrated in Fig. 20.

-400

-200

0

200

400

600

800

-30 -20 -10 0 10 20 30 40 50

0%20%40%60%80%100%

Bo

lt p

re-t

en

sio

n str

ess (

N/m

m2)

Nominal stress

-5 25

0

50

100

150

200

250

300

0 100 200 300 400 500 600

Pre-tension 20%Pre-tension 0%

Blo

t p

re-t

en

sio

n s

tress (

N/m

m2)

Time (s)

Fig. 20 Nominal stress Vs. Fig. 21 Time history of bolt

bolt pre-tension stress pre-tension stress (14m/s)

Fig.21 shows one example of the time history of the

bolt pre-tension stress at the wind speed of 14m/s. It

is clear that when the pre-tension force drops the

stress range increases significantly.

The fatigue life investigation follows the rules

mentioned in Section 3.4. The ultimate tensile

strength of FT10 bolts is1000Mpa and the detail

category is 36.

The bolts fatigue life is shown in Fig. 22.

0

0.001

0.01

0.1

1

10

100

1000

020406080100

Bo

lts f

atig

ue

life

(Y

)

Bolt pre-tension force(%)

20 years

0.22 days

6.95 years

318.25 years

Fig. 22 Bolts fatigue life vs. bolt pre-tension percentage

As we can see that when the pre-tension force is

over 40%, the life time does not decrease. However

when the pre-tension force is below 40% the fatigue

life time drops dramatically as only a few days left,

when the pre-tension force is 0%.

4. Conclusions

This research is based on the collapse accident of

Taikoyama wind farm No.3 turbine. The field

measurement of tower model frequency, SCADA

data and strain gauge data were measured. At the

same time the aerodynamic model was built. In

addition, the tower top FEM model was built to

evaluate the high-tension bolts and tower tube

fatigue life.

The cause of the collapse of the wind turbine is

discussed and the following conclusions were drawn:

1) Due to high turbulence intensity at site, the control

of the wind turbine was modified by manufacturer.

Power output and maximum rotor speed were

adjusted according to measurement data, and a five

degree of pitch error was applied. With this control

method the simulation results show good agreement

with measurement results;

2) For the high tension bolts, by considering the

nonlinear phenomenon and stress concentration

closed to welding zone, when the pre-tension force

decreases, the stress range increases, especially

when pre-tension force is 0% it is 30 times larger. The

less the pre-tension force left, the larger its range is.

As a result, when the pre-tension force is below 40%

the fatigue life time drops drastically and it is only a

few days when the pre-tension force is 0%;

3) Similarly, the FEM model shows that with 17 bolts

broken the local stress at fracture section increases

more than three times compared with the case of

bolts at normal condition. This phenomenon

accelerated the fatigue initiation and propagation

and the fatigue life of the fracture section decreases

dramatically to 1/200 of its life time.

4) The reason for the Taikoyama wind farm accident

is now clearly understood in a detailed manner. It is

not the matter of design or material, but was due to

the fatigue failure caused by the reduction of high

tension bolts’ pre-tension force.

For the Taikoyama wind turbines’ high tension bolts,

according to the service manual the temporary

torqueing and final torqueing was applied. And at the

time of 500 hours after bolt changing, the re-

torqueing must be applied. However at the time of

periodical bolt changing operation, the re-torqueing

was not applied. The wind turbine is a rotating

machine system, in which the contact surface and

the bolt itself plasticity deforms accompany with the

wind turbine operation, and therefore the pre-tension

force reduces.

Moreover, according to the service manual, 5% of the

bolts should be inspected per year, which means

only three bolts were inspected. We should check at

least 16 bolts per year in order to cover the bolts in

all wind direction.

Besides, during the year from 2005 to2008, the

workers only conducted the method of counter mark

inspection to make sure the torque was enough.

It is a serious problem between manufacturer and

operator that expertise technique is not transferred

accurately and efficiently. Clear rules must be made

even after guarantee periods, or it may lead to

devastating accident.

Reference

[1] Kyoto fu, Report of the accident in Taikoyama wind farm No.3 wind turbine, Kyoto, 2013.

[2] International Electrotechnical Commission, (2005). IEC 61400-1, 3rd edition, Part 1: Design requirements. Geneva.

[3] Ishizaki, H.(1983) Wind profiles, turbulence intensities and gust factors for design in typhoon-prone regions. Journal of Wind engineering & Industrial Aerodynamics, 13: 55-66.

[4] T. Ishihara, P.V. phuc, Yozo Fujino. A Field Test and Full Dynamic Simulation on a Stall Regulated Wind Turbine. The sixth Asia-Pacific Conference on Wind Engineering, Seoul, September 2005: 599-612.

[5] Garrad Hassan Bladed, version 4.4, DNV-GL, 2013.

[6] Tony Burton, David Sharpe, Nick Jenkins. Wind Energy Handbook. John Wiley & Sons Ltd, Chichester, 2001.

[7] Japan Society of Civil Engineers, (2010). Guidelines for Design of Wind Turbine Support Structures and Foundations. Task Committee on Dynamic Analysis and Structural Design of Wind Turbine Committee of Structural Engineering, Tokyo.

[8] V. Caccese, P.A. Blomquist, K.A. Berube. Effect of weld geometric profile on fatigue life of cruciform welds mad by laser/GMAW processes. Marine Structures, 2006, 19: 1-22.

[9] Germanischer Lloyd WindEnergie GmbH (2005), Guideline for the Certification of Offshore Wind Turbines. Germanischer Lloyd WindEnergie, Hamburg.