Doboku Gakkai Ronbunshuu F Vol.63 No.2, 181-189, 2007.

A PROPOSED TYPHOON RESISTANT DESIGN OF

A WIND TURBINE TOWER IN THE PHILIPPINES

Lessandro Estelito GARCIANO1 and Takeshi KOIKE2

1Member of JSCE, Dept. of Civil Eng., Musashi Institute of Technology
(1-28-1, Tamazutsumi, Setagaya-ku, Tokyo 158-8557, Japan)
E-mail:g0475002@sc.musashi-tech.ac.jp
2Member of JSCE, Professor, Dept. of Civil Eng., Musashi Institute of Technology
(1-28-1, Tamazutsumi, Setagaya-ku, Tokyo 158-8557, Japan)
E-mail:tkoike@sc.musashi-tech.ac.jp

The Philippines is essentially a typhoon-prone area which poses a threat to the structural integrity of wind
turbines. Due to this scenario, the authors classified the country’s wind climate according to the IEC
classes. The generalized extreme value distribution was used to model the extreme wind speeds from each
of the fifty stations. Areas exceeding the IEC class I 50-year return wind speeds were classified as Class 1a.
Subsequently the probability of buckling failure of a wind turbine was analyzed. The analysis shows that
there is a need to increase the mean of the tower buckling resistance by a factor of 1.02 to 1.74 to mitigate
buckling failure due to typhoons.

Key Words : typhoon, wind turbine, generalized extreme value, buckling failure

1. INTRODUCTION wind speed Ve50 for a class I wind turbine.

As far as the Philippines is concerned it has not
The plan of constructing wind farms in the experienced wind turbine failures yet because the
Philippines is gaining momentum in recent years. wind industry is relatively new. As mentioned earlier
The need to lessen dependence on fossil fuel for there is only one commercial wind farm and the rest
energy and to utilize clean energy are some of the few are small-sized wind turbines. Nevertheless future
reasons for this development. The first commercial wind farm owners are faced with a potential but
wind farm in the country was commissioned in 2005. uncertain risk of buckling failure due to the strong
This 25 MW wind farm is located far north of the
country where the best wind resource can be found.
Other commercial wind farms are also in the works.
The country has good potential for wind power
development1) but is also prone to strong typhoons.
These extreme wind phenomena can undermine the
structural performance of a wind turbine, e.g.
buckling failure of the tower. Although modern wind
turbines are designed to the highest standards the
codes and guidelines from which the turbines were
designed where not intended to cover wind climates
where strong typhoons occur 2).
In 2003, Super Typhoon Maemi hit a wind farm in
Okinawa, Japan causing damage to blades, tower
(Fig. 1) and foundation. The observed maximum
instantaneous wind speed during the typhoon was
about 74 m/s3). This exceeded IEC’s4) 50-year design Fig. 1 Buckling Failure of a Wind Turbine Due to Typhoon

181
 Doboku Gakkai Ronbunshuu F Vol.63 No.2, 181-189, 2007. 5
possibility of typhoons exceeding the Ve50 during its
economic life. Due to this extreme wind phenomenon
and the potential risk it poses to wind turbines studies
have been done to improve the performance of wind
turbines considering typhoon loads; e.g., Okinawa
Electric Power Company Inc.3), Clausen et al 2), and
Garciano et al 5). Although the IEC defines external
wind conditions in the design of wind turbines it
contains no specific information for wind conditions
in areas with typhoons. As such wind turbines
designed according to the IEC standards does not
necessarily ensure satisfactory performance if
constructed in the Philippines.
A higher probability of failure can also be
prescribed for wind turbines with due consideration
to the economics of the wind farm. This approach
requires that both the designer’s technical and
owner’s financial criteria are simultaneously met. In
this paper the authors will focus on the technical
approach which is necessary to evaluate the risk with
the probability of system failure. Once this technical
development is established the financial approach
can be done using the abovementioned probability of
failure. The optimal solution based on both
approaches is developed and discussed in a parallel
study6), 7).
The authors therefore propose a solution from a
technical approach a typhoon resistant design of a
wind turbine tower in the Philippines. To realize this
proposal the following design parameters of a wind
turbine tower will be established: (1) the target
probability of buckling failure due to a typhoon load
which is assumed to follow a Poisson distribution; (2)
the buckling resistance distribution; and the (3)
typhoon load distribution. These proposed design
parameters are compatible with the target probability
of failure of existing wind turbines designed for
European wind conditions.
2. DESIGN PARAMETERS OF A WIND
TURBINE TOWER
(1) Target probability of buckling failure PFT
As mentioned earlier PFT is assumed to follow a
Poisson distribution as shown below.
PFT = 1 − exp( −ηT ) (1)
The variable η is the mean occurrence rate and T
is time in years. In this paper η is assumed equal to
3.2×10-4. This value is the recommended rate of
collapse of an entire turbine at tower foot8) which is
based on accident data from European countries.
Then if the plant life of the wind farm is 20 years,
182
PFT is equal to 6.4×10-3. From this value the
reliability index β is calculated using Equation (2)
and is equal to 2.4904.
β = −Φ −1 ( PFT ) (2)
(2) Critical buckling resistance
a) Buckling capacity
There are several failure modes that a wind
turbine might experience when strong typhoons
strike. Failures such as blade fracture, anchor bolt
pull-out or buckling of the tower have been observed
not only in Japan but in other countries as well.
However due to the cost impact of tower buckling
failure the authors were motivated to study on this
major failure mode.
Modern wind turbine towers are made of steel and
are tubular in shape. The tower can be considered as a
thin wall structure due to a larger diameter compared
to the thickness. Due to this circumstance the tower is
prone to buckling. A localized increase of stress9) is
also likely because of a geometric discontinuity at
the base of the tower where a doorway is located.
There are several models which can estimate the
buckling capacity of a thin-wall structure but in this
paper the model recommended by the International
Organization for Standardization10) (ISO) is used.
⎛ fy ⎞
Fyc = ⎜⎜1.047 − 0.274 ⎟⎟ f y (3)
⎝ Fxe ⎠
where Fyc is the characteristic local buckling strength
in stress units, fy is the yield stress of steel and Fxe is
the characteristic elastic local buckling strength in
stress units. The value of Fxe 11) is determined using
Equation (4).
Fxe = 0.75 n E (t / D) (4)
where n = 0.11, E is the Young’s modulus of
elasticity, t is the thickness and D is the diameter.
Multiplying Equation (3) with the plastic section
modulus of the tower gives the bending moment
capacity that includes local buckling effects as shown
below:
fy ⎛ D fy ⎞ 3
Ro = ⎜⎜1 − 0.592 ⎟⎟[ D − ( D − 2t ) 3 ] (5)
6 ⎝ t E⎠
If uncertainties12) in the model and statistics of the
variables are taken into account Equation (5)
becomes:
 Doboku Gakkai Ronbunshuu F Vol.63 No.2, 181-189, 2007. 5

fy ⎛ fyX y , ss ⎞ drag coefficient and c is the blade chord.

Rn = ⎜1 − 0.592 D ⎟× (6) From Equations (6) and (8), the buckling
6 ⎜⎝ t X E , ss E ⎟⎠ resistance is expressed in terms of the wind speed14)
[ D 3 − ( D − 2t ) 3 ] X y , ss X cr as shown below.

−1 −1
where Xy,ss model uncertainties related to yield stress 2 Rn X aero X exp X st−1 X str
−1

for scale effects resulting from differences between R n (v ) = (10)

CT ρA(1 + 2k p Icamp X dyn ) 0.5 h
test specimens and full-scale structures, XE,ss model
uncertainties related to modulus of elasticity for scale
effects resulting from differences between test c) Numerical simulation of R(v)
specimens and full-scale structures and Xcr accounts To obtain sample realizations of Rn(v) monte carlo
for the natural fluctuations of the test results to the simulation of Equation (10) was performed. The
mean of which Equation (5) has been fitted. values of the variables that were used in the
b) Wind load effect on tower simulation are found in Table 1.
It is assumed that the wind turbine is in a Based on these realizations the distribution Rn(v)
non-operational state, e.g., parked. In this case, the of can be estimated for D/t each ratio. The mean,
load on the tower is given as: standard deviation and coefficient of variation
(COV) are also determined (see Table 2). Note that
S o = 0.5CT ρv 2 A(1 + 2k p Ic amp ) h the distributions are also lognormally distributed.
0.5
(7)
However from these results only the COV values will
be used. The means and standard deviations are
where CT is the thrust coefficient, ρ is the air density, discarded and are derived separately using partial
v is the wind speed at hub height , A is the rotor disk safety factor equations as shown later.
area, kp is the peak factor, I is the turbulence
intensity, camp is the combined admittance and
dynamic amplification factor and h is the distance Table 1 Variables Used in the Buckling Resistance Analysis
from the center of the rotor to the base of the tower. If
uncertainties12) in the model and statistics of the Variable Distribution Expected COV
variables are taken into account, Equation (7) type value
becomes: D (m) 3.0 to 4.0
t (mm) 50 and 75
S n = 0.5CT ρv 2 A(1 + 2k p Ic amp X dyn ) 0.5 h × (8) A (m2) 2123
h (m) 60
X aero X exp X st X str kp 3.3
camp 1.35
fy (Mpa) ln (lognormal) 220 0.05
where Xdyn models uncertainties from modeling of the
E (Mpa) ln 2.1×105 0.02
dynamical response characteristics of the turbine,
Xy,ss ln 1 0.05
Xexp models uncertainties associated with the XE,ss ln 1 0.02
exposure coefficient, Xst statistical uncertainty Xcr ln 1 0.10
connected with the assessment of the wind climate, Xdyn ln 1 0.05
Xstr models uncertainty from computation of the Xaero Gumbel 1 0.10
stresses from the loads and Xaero models uncertainties Xexp ln 1 0.20
related to the assessment of the lift and drag Xst ln 1 0.10
coefficients. Xstr ln 1 0.03
From blade element momentum theory the
relationship9),13) between the differential thrust, lift
and drag coefficients is shown below Table 2 Results of Buckling Resistance Simulation

2CT v 2 Fπrδr = W 2 N (C L sin φ + C D cos φ )cδr (9) D t D/t ratio µRn(v) σRn(v) δRn(v)
3.0 50 60 88.12 12.21 0.139
75 40 107.76 14.95 0.139
where F is a correction factor for hub and tip loss, r is
3.5 50 70 102.71 14.18 0.138
the radius of blade element, W is the resultant relative 75 47 125.89 17.46 0.139
velocity at the blade, N is the number of blades, CL is 4.0 50 80 117.16 16.12 0.139
the lift coefficient, φ is the inflow angle, CD is the 75 53 144.01 19.91 0.138

183
 Doboku Gakkai Ronbunshuu F Vol.63 No.2, 181-189, 2007. 5

(3) Extreme wind load based on a proposed γ

distribution of IEC’s Ve50 µ R (v) = µs (17)
φ
IEC stipulates that wind turbines should be able to
withstand extreme wind speeds with a recurrence δ R ( v ) = average(δ R ( v ) ) n
(18)
period of 50 years. Further, it also states that the Ve50
for a class I wind turbine is 70 m/s. Therefore from σ R ( v ) = µ R ( v )δ R ( v ) (19)
these information, we know Ve50 is a deterministic
wind speed with 0.02 probability of being exceeded
Using the results in this section the probability of
each year.
buckling failure PF1 is calculated using Equation
However for a noteworthy buckling failure
(20).
analysis of a wind turbine tower it is essential that the
extreme wind load be described by some probability
distribution fs. This distribution must conform to the PF1 = P[ R(v) ≤ S ] (20)
condition stated above, i.e., the 98% percentile value
is 70 m/s. There are several distributions to choose This is an iterative calculation, and terminates
from, but it is practical to choose a two-parameter when PFT = PF1. The only variable that is changed
distribution so that if one assumes the value of one during the iteration is λs. Once this equality is
parameter the other is easily derived. Although one satisfied the statistics of R(v) and the value of PFT
can choose to use a Gumbel or Weibull distribution, will be used in the subsequent sections. To
the authors chose a lognormal distribution for fs. This summarize the calculation flow in this section a
assumption simplifies the analysis since both flowchart is provided in Fig. 2.
resistance and load are described by a lognormal
distribution. Accordingly the parameters of fs are (5) Numerical simulation
shown in Equation (11). At first the value of φ = 0.77209 is calculated from
Equation (15) where δR(v) ≡ average(δRn(v)) = 0.139.
⎛ ln(70) − λ s ⎞ Next the values of the parameter ζs are calculated by
Φ⎜⎜ ⎟⎟ = 0.98 (11)
assuming a range of values for λs in Equation (11). In
⎝ ζs ⎠ this case, the values ranged from 0< λs ≤ ln(70) with
increment of 0.01 generating a total of 425 values.
The mean µs, standard deviation σs and coefficient
of variation δs of fs are determined from Equations
(12), (13) and (14), respectively.

µ s = exp(λ s + 0.5ζ s2 ) (12)

σ s = exp(ζ s2 − 1) µ s (13)

δ s = σ s / µs (14)

(4) Partial safety factors

Assuming that all the variables defined in the
previous sub-sections are calculated, the partial
safety factors for the load and resistance are
determined as shown in Equations (15) and (16). For
the value of α it is assumed equal to 0.75.

φ = exp(−αβδ R (v ) ) (15)

γ = exp(αβδ s ) (16)

As a result the mean, coefficient of variation and

standard deviation of R(v) are determined using Fig. 2 Flowchart in Obtaining the Design Parameters of a Wind
Equations (17), (18) and (19), respectively. Turbine Tower

184
 Doboku Gakkai Ronbunshuu F Vol.63 No.2, 181-189, 2007. 5

Table 3 Some of the Results of the Wind Turbine Tower Design

Parameters Analysis

λs ζs µs σs γ µR(v) σR(v) PF1

3.95 0.14 52.49 7.67 1.31 89.18 12.36 0.0058
3.96 0.14 52.98 7.48 1.30 89.20 12.35 0.0061
3.97 0.14 53.47 7.28 1.29 89.22 12.36 0.0063
3.98 0.13 53.98 7.09 1.28 89.25 12.36 0.0067
3.99 0.13 54.48 6.89 1.27 89.27 12.36 0.0070

The rest of the variables are easily obtained using

the provided equations together with the flowchart.
Since the results are large only the essential results
are shown in Table 3.
From the analysis, it was observed that λs should Fig. 3 Annual Extreme Wind Speeds at Virac Radar Station
be 3.97 and ζs equal to 0.136 so that PFT ≈ PF1. With
these parameters known the statistics of the proposed
IEC Ve50 distribution can be determined and are as

0.0 0.2 0.4 0.6 0.8

follows: µs = 53.47 and σs = 7.28.
Consequently the statistics of the critical buckling
resistance are also determined and these are µR(v) =
Model
89.22 and σR(v) = 12.36.
Now that the design parameters of a wind turbine
tower have been established we proceed to the next
step. In the succeeding sections, the extreme wind
speeds (typhoons) in the Philippines are modeled
using a GEV distribution. Subsequently the
probability of buckling failure of a wind turbine 0.0 0.2 0.4 0.6 0.8 1.0
tower is calculated using a typhoon distribution as Empirical
the extreme load.
Fig. 4 Probability Plot of Wind Data Versus GEV Model

3. EXTREME WIND DATA IN THE

PHILIPPINES
The wind data which will be used in modeling the
typhoon load are taken from the fifty wind stations of
the Philippine Atmospheric, Geophysical and
Astronomical Services Administration (PAGASA).
The details can be found in Garciano et al15) with
additional recent annual maximum wind speed data
for each station from 2001 to 2005.
4. TYPHOON LOAD ON TOWER
(1) GEV distribution
There are several extreme value models that can be
used to characterize the behaviour of extremes. In
this paper the GEV distribution16) is used to model the
annual maximum wind speeds at specific sites in the
Philippines as shown in Equation (21).
The parameters µ, σ (> 0) and ξ are referred to as
the location, scale and shape parameters,
respectively. The above parameters are estimated
using the maximum likelihood (ML) method.
(2) GEV model of annual maximum wind speeds
To illustrate the modeling, a data set from Virac
Radar Catanduanes Station (Station No. 447) is
analyzed (see Fig. 3). The S-PLUS functions16) were
used to analyse the data set. The ML estimate of the
parameters from this station are µˆ , σˆ , ξˆ = (31.99,
14.63, -0.14). The standard error of the parameters
are 2.82, 2.11 and 0.17 for µ̂ , σˆ and ξˆ respectively.
The probability plot of the data in Fig. 4 shows it is
close to the unit diagonal which lends support to the
GEV as an acceptable model for this data. Equation
(22) is used to obtain estimates of extreme quantiles
of the distribution

G ( x; µ , σ , ξ ) = exp(−[1 + ξ ( x − µ ) / σ ] −1 / ξ ) (21) x p = µ − σ [1 − {ln(1 − p )}−ξ ] / ξ (22)

185
 Doboku Gakkai Ronbunshuu F Vol.63 No.2, 181-189, 2007. 5

where xp is the return level associated with the return (3) Typhoon load distribution
period 1/p. In particular xp is exceeded by the annual The annual wind data were gathered at 10 meters
maximum in any particular year with probability p. above the ground. Therefore to model the typhoon
Therefore using the ML estimate of the parameters load that will be applied at the hub height of a wind
from Station No. 447, the 50-year return wind speed turbine, random samples generated from the GEV
is 76 m/s (see Fig. 5). It is also worthwhile to note density function for each station are extrapolated to
that the return wind speed curve is bounded because the hub height using the logarithmic law shown
the shape parameter is negative. The ML estimate of below
the parameters µˆ , σˆ and ξˆ for all the stations are ln( z / z o )
v' = x r (23)
shown in Fig. 6, Fig. 7 and Fig. 8 respectively. ln( z r / z o )

where v’ is the wind speed at hub height z = 60m, xr

90

is the sample realization at reference height zr = 10 m

Return wind speed in m/s

and zo is the roughness length equal to 0.01. A total of

70

100000 random samples were generated from the

GEV distribution of each station. These samples
50

were then interpolated to hub height and fitted to a

GEV distribution. As an example, the ML estimate of
the parameters for Station No. 447 are µˆ S ( v ') , σˆ S ( v ')
30

95% confidence interval

and ξˆS ( v ') = (40.37, 18.22, -0.13). The ML estimate
10

0.5 1 5 10 50 100
of the parameters for all stations are shown in Fig. 9,
Return Period in Years Fig. 10 and Fig. 11, respectively.
Fig. 5 Return Wind Speed Curve

Fig. 9 Location Parameter of Typhoon Load Model for all

Fig. 6 Location Parameter of GEV Model for all Stations Stations

Fig. 7 Scale Parameter of GEV Model for all Stations Fig. 10 Scale Parameter of Typhoon Load Model for all Stations

Fig. 8 Shape Parameter of GEV Model for all Stations Fig. 11 Shape Parameter of Typhoon Load Model for all Stations

186
 Doboku Gakkai Ronbunshuu F Vol.63 No.2, 181-189, 2007. 5

Fig. 12 Probability Density Function of Annual Maximum Wind

Speeds and Typhoon Load from Virac Radar Station

Fig. 14 Classification of 50-year Extreme Wind Speeds in

the Philippines at 60 m Hub Height

Fig. 13 50-year Return Wind Speeds at 60 m Hub Heights

Based on the above parameters the probability

density function (pdf) of the typhoon load fS(v’) can be
estimated for each station.
In Fig. 12 the result of the pdf estimate of fS(v’)
along with the pdf of the annual wind speed maxima
fG from Station No. 447 are shown. The mean and
standard deviation of each distribution are also
shown in the same figure. Note that fS(v’) is also GEV
distributed. Fig. 15 Probability Density Function of the Load and Resistance
To calculate the 50-year return wind speeds from of a Wind Turbine for Buckling Failure Analysis
each station, the the parameters from each station are
used as input in Equation (22). The results show that
the return wind speeds from sixteen stations proposed class Ia (see Fig. 14).
exceeded Ve50, the dashed line as shown in Fig. 13. This map is useful information for power planners
From the above results, kriging method of ArcGIS who may want to construct commercial size or small
Geostatistical Analyst17) was used to estimate the size wind farm projects in the Philippines.
50-year return wind speeds of the entire country. Let us briefly highlight some of the important
Several runs were done to create a prediction surface results that will be used in the next section. From
which gives the smallest prediction error as possible. section 2 and this section the following results were
The prediction surface is divided into areas according obtained PFT, fR(v), fs and fS(v’). We show these density
to the IEC WTGs classes. Areas with 50-year return functions in Fig. 15. Note that the parameters of fS(v’)
wind speeds exceeding the Ve50 are classified into a in this figure were taken from Station No. 447.

187

5. PROBABILITY OF BUCKLING
FAILURE DUE TO TYPHOON LOADS
To evaluate the probability of buckling failure
PF2 of a wind turbine tower with typhoon as the
extreme load we use Equation (24).
PF2 = P[ R (v) ≤ s (v' )]
Using first order reliability method (FORM) PF2 is
evaluated for each station. The results show that
highest value for PF2 is 0.053 while the smallest
value is 1.45×10-11. The former result is taken from
Station No. 447 while latter is from Station No. 741.
The results also show that at least seventeen stations
have PF2 > PFT. The complete results are shown in
Fig. 16.
In order to maintain the same probability of
buckling failure of a wind turbine designed for
European wind conditions to the Philippine wind
climate a new wind speed resistance f R ( v ) is
new
introduced as shown below.
PFT = P[ R(v) new ≤ s(v' )]
Several distributions for f Rnew
(v)
may exist which
satisfies the above equation.
However to simplify the analysis, it is assumed
that µR(v) is increased by a factor c as shown below.
µ Rnew
(v) = c × µ R (v)
It is also assumed that σ Rnew
(v) = σ R(v)
. buckling failure instead of using accident data from
Fig. 16 Probability of Buckling Failure of a Wind Turbine Due
to Typhoon Loads
Doboku Gakkai Ronbunshuu F Vol.63 No.2, 181-189, 2007. 5
Using FORM, Equation (25) is evaluated with an
initial value of c = 1. This value is adjusted until the
equality is satisfied. The complete results of the
evaluation are shown in Fig. 17.
The results show that an increase in the mean
buckling strength from as high as 1.74 to as low as
1.02 is needed if the target probability of failure for
(24) European wind condition is maintained.
6. CONCLUDING REMARKS
It is a fact that strong typhoons in Asian countries
often occur. As such if wind turbines are constructed
in these areas this extreme wind phenomenon should
be taken into account.
Due to this scenario, the authors determined
design parameters of a wind turbine tower for
buckling analysis. Typhoon loads derived from
extreme wind data in the Philippines were also
established. Based on these results typhoon resistant
design parameters of a wind turbine tower for use in
the Philippines were developed. These design
parameters are compatible with the target probability
(25) of failure of existing wind turbines designed for
European wind conditions.
Another important contribution of this research is
the classification of the wind climate in the
Philippines according to the IEC WTG classes and a
proposed Class Ia.
For future work, other strategies can also be
pursued to manage the risk of typhoons, like risk
transfer to insurance or a combination of both for
(26)
wind power projects in the Philippines. Further work
in finding the most appropriate target probability of
European countries can also be pursued.
Fig. 17 Values of Factor c* for Each Station
188
 Doboku Gakkai Ronbunshuu F Vol.63 No.2, 181-189, 2007. 5

ACKNOWLEDGMENT: The authors wish to 53A, pp. 165 – 171, 2007.

thank the anonymous referees for their valuable 8) Braam, H. and Rademakers, L. W. M. M.: “Guidelines on
suggestions and comments. the Environmental Risk of Wind Turbines in the
Netherlands”. ECN, The Netherlands, 2002.
9) Burton, T., Sharpe, D., Jenkins, N. and Bossanyi, E.: Wind
Energy Handbook, John Wiley & Sons Ltd., England, 2001.
REFERENCES
10) ISO TC 67/SC 7 N 288: Petroleum and Natural Gas
1) Elliott, D., Schwartz, R., George, R., Haymes, S., Heimiller,
Industries – Fixed Steel Offshore Structures, 2001.
G., Scott, G. and McCarthy, E.: “Wind Energy Resource
11) Kato, T., Akiyama, H. and Suzuki, H.: “Plastic Local
Atlas of the Philippines”, US National Renewable
Buckling Capacity of Steel Pipes under Axial
Laboratory, 2001.
Compression”, J. of the Architechtural Institute of Japan,
2) Clausen, N. E., Ott, S., Tarp-Johansen, N. J., Norgard, P.,
No. 204, pp. 9 – 17, 1973. (in Japanese)
Larsen, X. G., Pagalilawan, E. and Hernando, S.: “Design
12) Sorensen, J. D. and Tarp-Johansen, N. J.: “Reliability-based
of Wind Turbines in an Area with Tropical Cyclones”,
Optimization of and Optimal Reliability Level of Offshore
Proc. of European Wind Energy Conference, Greece, 2006.
Wind Turbines”, International Journal of Offshore and
3) Okinawa Electric Power Company Inc.: “Investigative
Polar Engineering, Vol. 15, No. 2, pp. 141 – 145, 2003.
Report on the Collapse of Wind Turbine Generators due to
13) Duquette, M. and Visser, K.: “Numerical Implications of
Typhoon No. 14”, 22 pages, 2004. (in Japanese)
Solidity and Blade Number on Rotor Performance of
4) IEC 61400-1, ed. 2, Wind Turbine Generator Systems - Part
Horizontal-axis Wind Turbines”, ASME Journal of Solar
I; Safety Requirements, 1999(E).
Engineering, Vol. 125, pp. 425 – 432, 2003.
5) Garciano, L. E., Koike, T. and Maruyama, O.:“Performance
14) Garciano, L. E. and Koike, T.: “New Design Wind Speed
-based Design of Wind Turbines for Typhoons”, Proc. of 9th
for a Wind Turbine Generator (WTG) considering Typhoon
Structural Safety and Reliability of Engineering Systems
Loads”, Proc. of European Wind Energy Conference,
and Structures, G. Augusti, G. I. Schueller and M.
Greece, 2006.
Ciampoli, eds., Millpress, Rotterdam, Rome, pp. 1235 –
15) Garciano, L. E., Hoshiya, M. and Maruyama, O.:
1242, 2005.
“Development of a Regional Map of Extreme Wind Speeds
6) Garciano, L. E. and Koike, T.: “Managing the Risks of a
in the Philipines”, Structural Eng. / Earthquake Eng., JSCE,
Wind Farm in Typhoon-prone Areas”, Proc. of 10th East
Vol. 22, No. 1, 15s – 26s, 2005.
Asia-Pacific Conference on Structural Eng. and
16) Coles, S.: An Introduction to Statistical Modeling of
Construction - Wind and Earthquake Engineering, W.
Extreme Values, Springer-Verlag, 2001.
Kanok-Nukulchai, S. Munasinghe and N. Anwar, eds.,
17) Johnston, K., Ver Hoef, J. M., Krivoruchko, K. and Lucas,
Thailand, pp. 219 – 224, 2006.
N.: Using ArcGIS Geostatistical Analyst, ESRI 2001.
7) Garciano, L. E. and Koike, T.: “Stochastic capital budgeting
approach to obtain the target probability of buckling failure
(Received August 18, 2006)
of a wind turbine tower subjected to typhoons in the
Philippines”, Journal of Structural Engineering, JSCE, Vol.

189