Ocean Engineering 129 (2017) 382–388

Contents lists available at ScienceDirect

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

Risk assessment of floating offshore wind turbine based on correlation- MARK

FMEA
⁎
Jichuan Kangb, Liping Sunb, Hai Suna,c, , Chunlin Wub
a
College of Aerospace and Civil Engineering, Harbin Engineering University, China
b
College of Shipbuilding Engineering, Harbin Engineering University, China
c
Marine Renewable Energy Laboratory, Dept. of Naval Architecture & Marine Engineering, University of Michigan, United States

A R T I C L E I N F O A BS T RAC T

Keywords: Floating offshore wind turbine (FOWT) is complexly structured by interdependent subsystems and experiences
Floating offshore wind turbine negative impacts in harsh operating conditions. During the risk and reliability analysis, two issues have to be
Risk assessment addressed: system failure mode complexity and mutual correlation. We conducted risk assessment through a
FMEA modified Failure Modes and Effects Analysis (FMEA) method, named correlation-FMEA,to study the
Correlation
connection between failure modes and its effect on the failure probability of the entire system. A series of
failure modes with high priority were determined by conventional FMEA, and the corresponding connections
were analyzed to obtain the correlation coefficients using the reliability index vector method. The data used in
our research comes from field operation in China. Probability Network Evaluation Technique (PNET) was used
to get the weakest failure modes set of the system based on those coefficients. With the results, suggestions for
floating wind turbine design were provided regarding aspects of safety and reliability.

1. Introduction
In recent decades, FOWT has rapidly developed due to the surge of
renewable energy demand. FOWT, however, is costly and vulnerable.
Its frequent demand for maintenance and the difficulty of such
procedure call for vast expenditure. Due to the location and high
repair time, all capital and operating costs are assumed to escalate. One
way to provide effective maintenance is through risk assessment, which
are predictions of weak points in the system around the design stage.
For a long time this prediction relied on comparison data of similar
systems and statistical study. Operation data from worldwide wind
farms have been collected to acquire key statistical features, such as
failure rate and downtime, revealing wind turbines reliability under
various weathers, locations and configurations (Braam and
Rademakers, 2004; Ribrant, 2006; Polinder et al., 2007; Spinato
et al., 2009; Bussel and Zaaijer, 2011; Zhang et al., 2016).
Failure Modes and Effects Analysis (FMEA) method is widely
applied on power-generating system and proves a sound approach.
Arabian-Hoseynabadi et al. (2010) introduced FMEA method into
wind turbine risk assessment. In order to analyze a failure mode in the
FMEA procedure, three character metrics (severity, detection and
occurrence) are employed. These metrics are all scaled into several
levels and denoted specific values. By multiplying severity, occurrence
and detection, Risk Priority Number (RPN) is obtained to measure the
risk of failure mode and determine the most hazardous subassemblies.
The direct application of FMEA to FOWT, however, has potential
issues. First, FOWT structure and operation condition are more
complex compared with inland wind turbine, reflecting the limitation
of FMEA on complex systems due to overwhelming workload. Second,
similar or equal RPNs appear when the amount of failure modes
increases, increases the difficulties or in some cases, impossible to
identify the risk sequence. Third, the original assumption considers
failure modes as isolated items without correlation, which indeed exists
among failure modes (Sun 2015). The assumption yields over-assess-
ment, leading to higher safety redundancy in the design stage.
In this study, we describe a method for FMEA application on FOWT
by reducing failure modes and concerning correlation between them
after the routine procedure of the algorithm. The FOWT system is
divided into subassemblies and components, which defines failure
causes and failure modes through various studies (Tavner et al., 2007;
Anthony et al., 2015; Madjid and Michailides, 2015), and RPNs are
calculated accordingly. In order to reduce the excessive failure modes,
failure modes with higher RPNs are selected, named the weak modes.
The correlation of two failure modes is defined when both of the modes
simultaneously occur on one component or several, and the magnitude
of correlation depends on the number of such components.

⁎
Corresponding author at: College of Aerospace and Civil Engineering, Harbin Engineering University, No.154 Nantong Ave, Harbin, Heilongjiang, 150001, China.
E-mail address: sunhai2009@gmail.com (H. Sun).

http://dx.doi.org/10.1016/j.oceaneng.2016.11.048
Received 9 January 2015; Received in revised form 11 June 2016; Accepted 27 November 2016
Available online 02 December 2016
0029-8018/ © 2016 Elsevier Ltd. All rights reserved.
J. Kang et al. Ocean Engineering 129 (2017) 382–388

Correlations between the weak modes are then calculated. PNET Table 2
method is used for elimination of weak modes according to the Occurrence rating scale for FOWT FMEA.
correlation values to yield a devastated failure modes group which
Scale Description Criteria
could be relied on for optimization in the design stage.
This paper is organized as follows: The correlation-FMEA algorithm 1–2 Extremely unlikely Probability of occurrence is less than 0.0001
is described in Section 2. System grading of FOWT is proposed in 3–5 Remote Probability of occurrence is more than 0.0001 but
less than 0.001
Section 3. Results and discussion are in Section 4. Conclusions are
6–8 Occasional Probability of occurrence is more than 0.001 but
drawn at the end. less than 0.01
9–10 Frequent Probability of occurrence is more than 0.01
2. Correlation-FMEA

Table 3
FMEA could identify, analyze and estimate possible faults in system Detection rating scale for FOWT FMEA.
and the manifestations (Ahire and Relkar, 2012). After analyzing the
impact and consequences of each failure mode, vulnerabilities of the Scale Description Criteria
system can be confirmed according to the severity and probability of
1–2 Almost certain Current monitoring methods almost always detect
occurrence and detection. Upon this basis, recommendations for the failure
maintenance and improvement of relevant components are given, 3–5 High Good likelihood current monitoring methods will
and therefore system reliability can be ameliorated. The FMEA detect the failure
procedure assigns a numerical value to each risk associated with 6–8 Low Low likelihood current monitoring methods will
detect the failure
causing a failure, using severity, occurrence and detection as metrics.
9–10 Almost impossible No known monitoring methods available to detect
By multiplying severity by occurrence by detection of the risk, the RPN the failure
can be obtained, which reflects criticality rank. Severity refers to the
magnitude of the end effect of a system failure. The more severe the
consequence, the higher the value of severity will be assigned to the then
effect. Occurrence refers to the frequency that a failure cause is likely to n
β1⋅β2 ∑ ai bi σx2i
occur, described in a qualitative way. Detection refers to the likelihood ρ12 = cos θ G1G 2 = cos(β1, β2 ) = = i =1
of detecting cause before a failure occurs (Arabian-Hoseynabadi et al., β1 ⋅ β2 σG1 σG 2 (7)
2010). By targeting high value RPNs, the most risky failure modes can Here we then consider this relationship without losing general-
be addressed. The modified severity, occurrence, detection scale and ization through the entire system, for the ith and jth failure modes,
criteria are tabulated in Tables 1–3. their correlation coefficient is
The present work considers correlation between failure modes on
the foundation of the FMEA method. Let G1, G2 be two arbitrary ρij = cos(βi , βj ) (i , j = 1, 2, ⋯n ) (8)
failure modes. According to Hou and Ou (2002) and Sun et al. (2016), A main problem in the reliability analysis process for complex
the limiting conditions are given in polynomial form system is the excessive failure modes. Finding all failure modes and
G1 = a 0 + a1 x1 + a2 x2 +⋯+an xn (1) inserting them into the model is theoretically possible and provides a
full assessment but would necessitate excessive calculation work.
G1 = b0 + b1 x1 + b 2 x2 +⋯+bn xn (2) Analysis shows that only a small proportion of main modes are
where ai and bi, i=0,…,n are coefficients, and xi are a group of essential and others could be neglected (Tavner, 2012; Pérez et al.,
randomly independent variables. Further, we take µ(xi) the mean value 2013). Generally, for further reduction of failure modes amount the
and σ(xi) the mean square deviation of xi, thus the correlation design of simple system is optimized by controlling a weakest failure
coefficient is mode derived from the main modes based upon RPN.
For complex systems, since one weakest mode is inadequate to
n n n
∑i =1 ∑ j =1 ai bj Cov (xi , xj ) ∑i =1 ai bj σx2i represent the whole, a weakest failure modes group is employed
ρ12 = ρ (G1, G 2 ) = =
σG1 σG 2 σG1 σG 2 (3) containing several weakest modes by using PNET method. The selec-
tion steps are,
The reliable index vector is given by
β1 = (β1 cos θ11, β1 cos θ12, …, β1 cos1n ) (4) (1) Searching main failure modes and the corresponding structure
functions Zi with the FMEA; Applying reliable index vector
β2 = (β2 cos θ21, β2 cos θ22, …, β2 cos2n ) (5) algorithm to calculate the reliable index βi of each failure modes.
(2) Choosing correlation coefficient ρ0 according to practical situation
where θmn is the angle between the m th index vector and the axis of n th to judge the correlation degree between the failure modes. If ρ0 is
variable, hence too small, excessive reliability will be obtained and it is dangerous
ai σxi bi σGxi for the design; it will be conservative if ρ0=1. In general, ρ0 should
cos θ1i = , cos θ2i = be given according to the amount of failure modes and the
σG1 σG 2 (6)
importance of the project. Nugent et al. (1991) suggests the
correlation coefficient should be set between 0.5 and 0.7, Define
Table 1
Severity rating scale for FOWT FMEA.
a standard correlation coefficient threshold ( ρ0 ) as the basis to
identify related degrees between every two failure modes. ρ0 is
Scale Description Criteria usually defined according to practical circumstances (Sun et al.
2016), in our research ρ0 is set to be 0.7
1 Category IV (minor) Electricity can be generated but urgent repair
(3) Listing m weakest failure modes among main failure modes using
is required
2 Category III (marginal) Reduction in ability to generate electricity the PENT method: suppose the failure mode with the smallest
3 Category II (critical) Loss of ability to generate electricity reliable index β1 as 1st failure mode, calculate its correlation
4 Category I (catastrophic) Major damage to the turbine as a capital coefficient ρ1i with other failure modes by reliable index vector
installation
method. If ρ1i > ρ0 the ith failure mode is highly related to the 1st

383
J. Kang et al. Ocean Engineering 129 (2017) 382–388

Table 4 system is a mechanism that turns the blade, or part of the blade, in
Components number. order to adjust the angle of attack of the wind.
Number Component Number Component Number Component
3.2. Generator, electrical and electronic components
1 Slewing 18 Tower 35 Lightning rod
bearing Faults in generators can be caused by electrical or mechanical. The
2 Yaw engine 19 Work bench 36 Ground wire
main electrical faults are due to open-circuits or short-circuit of the
3 Yaw brake 20 Ladder 37 Arrester
4 Counting 21 Lighting 38 Motor winding in the rotor, or overheating can be caused by stator. Many
device device research works demonstrated that, a high failure rate in wind turbines
5 Friction disk 22 Column 39 Oil pump from bearings, rotors and stators (Popa et al., 2003). The main
6 Yaw lubrication 23 Beam 40 Oil tank mechanical defects are the corrosion, dirt and terminal damages. The
7 Yaw sensor 24 Bracing 41 Filter
8 Yaw controller 25 Heave plate 42 Piping
group formed by generator, electrical system and control system, has a
9 Untwist sensor 26 Fairlead 43 Valve relevant rate of failure and downtime in wind turbines (Marquez et al.,
10 Battery holders 27 Anchor 44 Water pump 2016).
11 Pitch controller 28 Hydraulic 45 heat transfer
device tube
3.3. Transmission system
12 Pitch engine 29 Linear winch 46 Water tank
13 Pitch bearing 30 Bending 47 Water pipe
anchor The transmission system are composed by a main bearing, gearbox
14 Control system 31 Chain 48 Radiator and coupling. The main bearing has a very high synthetic mechanical
15 Position sensor 32 Wire 49 Check valve
property, where, the bearing sustains the load/torque during the start
16 regulating 33 Weights
mechanism
and shutdown of the wind turbine.
17 Nacelle 34 Buoy The gearbox functions to transform high-torque to low-torque,
which also transforms low-speed of the main bearing to high-speed of
the generator. The gearbox is prone to be damaged due to the shock
failure mode, then the ith failure mode can be replaced by the 1st load caused by the instability of the wind speed, and the erosion caused
failure mode. If ρ1i < ρ0 the ith failure mode is slightly related to by the salty air. In addition, it is difficult to fix the gearbox because of
the 1st failure mode, and the ith failure mode cannot be replaced. its size and high installation position. Therefore, a long maintenance
Repeat the above steps until all the weakest failure modes are time is needed to fix the gearbox in case of shutdown. The direct-drive
determined. generator will be more widely used in the offshore wind turbine to
reduce the Life Cycle Cost (LCC).
3. FOWT system grading
3.4. Support system
The offshore floating wind turbine is a multi-functional system.
When using FMEA to analyze offshore wind turbine, categorizing The support system (Fig. 2) of FOWT includes nacelle, tower,
failure modes according to different system functions is more accurate floating foundation and mooring system. The main failure causes of the
for assessing risk and reliability. In this section, the subsystems are support structure are fatigue, corrosion, welding cracking and hull
defined according to the main functions of a direct-driven offshore collision. At extreme sea conditions, the floating foundation, the
wind turbine. Partial components of FOWT are listed in Table 4. mooring lines and the tower vibrates intensely to make the FOWT
unstable. And even severe accidents may occur, for instance, the tower
3.1. Blade system fracture, mooring system failure and blades damage.

The blades are the components of the wind turbine with the highest
percentage in term of failures and downtimes. The blades faults are
predominantly related to structural failures, for example, strength and
fatigue of the fibrous composite materials. Other faults, namely, cracks,
erosion and delamination, could appear in the leading and trailing
edges of the blades. A common fault of the blades is associated with the
failure of the yaw and pitch control system (Fig. 1). The yaw and pitch
3.5. Auxiliary system
The auxiliary system (Fig. 3) is comprised of lighting protection
system, hydraulic system and cooling system. Due to the height, FOWT
is prone to lighting strike. Lightning induction and impulse can burn
the electrical devices, control systems and communications. Therefore,
the lightening protection should be wisely designed to deal with a

Fig. 1. Yaw and pitch system.

384
J. Kang et al. Ocean Engineering 129 (2017) 382–388

Fig. 2. Support system.

variety of conditions. The hydraulic system serves as a power source for Table 5
mechanism for brake, pitch system, yaw system etc. The main failure FOWT failure modes.
modes of hydraulic system are hydraulic pressure error, temperature
Structural Electrical
error, timeout of hydraulic pump and motor failure. The cooling system
becomes more significant with the increase of the installed capacity of Fracture Electrical insulation
wind turbine and the heat generation of the generator and the Blockage Electrical failure
converter. So far, the air-blast cooling is the first choice for the Detachment Output inaccuracy
Looseness Software fault
generator unit and the converter of wind turbines. However, as the Deformation Intermittent output
offshore wind turbines are becoming larger, the water-cooling system is Overheat
also becoming more widely used.

severity, which results in related failure RPN. Summing these failure

4. Correlation-FMEA procedure for FOWT
RPNs gives the selected failure mode RPN. Then aggregating the part
RPNs builds the subassembly RPN.
Based on the subdivision of the FOWT, the main failure modes are
The complexity of FOWT system leads to numerous failure modes,
generated. These failure modes were considered for all 91 components
therefore, calculating relevance between every two failure modes will
and many were found to be common between various parts. The FOWT
cause a great amount of effort, besides, the significance of further
system is composed of mechanical structures and electrical compo-
analysis for the failure modes with low risks is also limited. Therefore,
nents. Tables 5 and 6 show the common failure modes and basic causes
in the proposed method, only the failure modes with high RPNs are
identified in these two related columns.
considered in the process of relativity calculation. The components
For each failure mode with several causes, the occurrence and
FMEA and RPN sort of failure modes are shown in Table 7 and Fig. 4.
detection values are multiplied for each cause by the failure mode

Fig. 3. Auxiliary system.

385
J. Kang et al. Ocean Engineering 129 (2017) 382–388

Table 6
Failure causes.

Structural Electrical

Corrosion Electrical overload

Fatigue Electrical short
Mechanical overload Insulation failure
Installation defect Calibration error
Maintenance fault Connection failure
Manufacturing defect Power input failure
Vibration Software design fault
Overheating

The failure modes of convert overheat (G4) and transform winding
failure (G5) are chosen as samples. According to the function for
floating wind turbine, G4 involves rectifier module, power module, and
the other twelve modules, while G5 involves transformer core, trans-
former winding, and another eleven components. Both of these two
failure modes involve all components of coolant system, vacuum fuse,
and four other parts, totaling seven parts. The components have a
series relationship, and the weight of every component are identified in
Fig. 4. RPNs of failure modes.
formula (1) and (2) to be 1. According to He and Wang (1993), assume
that the failure probability of the components subjects to normal
distribution and the variation coefficient is 0.2, the mean values are
divided into three classes, 0–0.01, 0.01-0.001, 0.001-0.0001, then put
them into formula (7) to calculate the correlation.

Table 7
Components FMEA.

Assemblies Failure modes End effects Causes of failure S O D RPN

Generator 1. Bearing deformation Equipment damage 1. Improper grease 4 7 5 676

2. Over-tighten/loosen bearing-shaft matching 6 6
3. Over-tighten/loosen bearing-shaft cap matching 6 5
4. Electric corrosion of rollaway nest 5 5
5. Shaft wear deformation 7 4
6. Cooling system failure 5 3
2. Overheat FOWT shutdown 1. Shaft failure 3 4 3 396
2. Turbine overload 5 5
3. Air cooling system failure 6 3
4. Partial short circuit on stator winding 7 5
5. Excessive system oscillation 7 6
3. Winding failure FOWT shutdown 1. Cable insulation failure 4 8 5 912
2. Connecting plug fall off 7 4
3. Interturn short circuit 6 4
4. Winding corrosion 8 7
5. Long-term overload 7 6
6. Electric sequence reverse 5 4
7. Thunder stroke fault 6 3
Electrical controls 4. Convert failure Disconnect to grid 1. Load mutation 3 6 6 630
2. Invert power input fault 7 8
3. Low voltage on power grid 7 5
4. Overload 8 6
5. Cooling system fault 7 5
Electrical controls 5. Transform winding failure Disconnect to grid 1. Insulation failure 3 7 6 618
2. Constant overload in transformer 6 4
3. Short circuit shock 5 4
4. Iron core corrosion 8 8
5. Overcurrent 7 4
6. Overvoltage 7 4
6. Output voltage inaccuracy Disconnect to grid 1. Rotor-stator friction 3 7 8 411
2. Computer timing failure 4 4
3. Rational speed sensor failure 5 7
4. Wind wheel shaft damage 6 5
7. Yaw positioning inaccuracy Inefficiency 1. Wind direction sensor inaccuracy 3 5 6 333
2. Yaw damper inaccuracy 7 6
3. Brake torque deficiency 6 4
4. Excessive yam gear distance 5 3
Support structure 8. Mooring line fracture FOWT shutdown 1. Extreme marine environment 4 4 3 340
2. Fatigue damage 6 8
3. Improper anchor design 3 3
4. Improper anchor installation 4 2
5. Collision 4 2
Auxiliary system 9. Cooling system failure FOWT shutdown 1. Pipeline or joint leakage 4 6 8 556
2. Coolant liquid pump damage 5 6
3. Coolant liquid deficiency 7 3
4. Fouling clog on radiator or cylinder 8 5

Data used in FMEA process were provided by the Goldwind Sci and Tech Co., Ltd.

386
J. Kang et al. Ocean Engineering 129 (2017) 382–388

Table 8 5. Conclusions
Correlation coefficients.
A correlation FMEA method of FOWT risk assessment is presented
G1 G2 G3 G4 G5 G6 G7 G8 G9
and evaluated. The reliable index vector method is employed to
G1 1 0.600 0 0.414 0.430 0 0 0 1 calculate correlation of failure modes, and to obtain the weakest failure
G2 1 0.323 0.414 0.430 0.258 0 0 1 basis of the FOWT system. The failure modes with the highest RPN
G3 1 0.267 0.277 0.289 0 0 0
values, namely, the most influential failure modes, which are calculated
G4 1 0.741 0.154 0 0 1
G5 1 0.240 0 0 1 by the PNET method and FMEA. Results from quantifying the failure
G6 1 0 0 0 rate from FMEA analysis have shown that.
G7 1 0 0
G8 1 0 1) Failure occurs at lower probability for structural components than
G9 1
for electronic components.
2) Material corrosion is the key factor of failure. These results conform
7 to the empirical data from industry. The study has identified that
β4⋅β5 ∑ ai bi σx2i
ρ45 = cos θ G4 G5 = cos(β4 , β5) = = i =1 = 0.741 reinforcing the preservative treatment of the equipment is effective
β4 ⋅ β5 σG4 σG5
means to enhance the reliability of the system.
3) The analysis shows that direct-driven generator group has the
All the correlation coefficients results are shown in Table 8.
highest RPN value among all sub-systems. The comprehensive RPN
According to Table 8, correlations between different failure modes
value of the floating foundation and mooring system is also
cannot be ignored in risk assessment. For instance, when malfunction
relatively high.
of the convertor happens due to overheat, then failure modes No.4 and
4) The quality of the floating foundation and the mooring system are
No.5 may cause shutdown of transformer. If the calculation is
demanding and even trifling failures on them can cause consider-
conducted merely according to series connection without taking the
able loss. Thus, in the entire system, the generator group, the
correlations of different failure modes into consideration, the failure
floating foundation, and the mooring system are the most signifi-
probability will be overestimated.
cant components that need to be considered discreetly during the
After the correlation analysis, PNET method is applied to the nine
designing process.
major failure modes. The relevant threshold is set to 0.7 (Nugent
1991), the weakest failure modes basis of the floating wind turbine
This correlation FMEA method provides prospective that it can be
system as modes 1, 2, 3, 4, 6, 7, and 8.
adapted to practical applications in order to instruct or optimize the
As mentioned, there are 11 common failure modes and 15 common
designing process of FOWT system and reduce repair costs. The
failure causes. These limited numbers of failure modes and causes were
method shows its potential not only in offshore engineering but other
repeated in 91 parts of the FOWT system. In Fig. 4, among the failure
types of engineering applications.
modes with high RPNs, the most significant one is electrical failure and
the most frequent cause is corrosion. FMEA method could find the
Acknowledgements
failure modes and causes with the highest occurrence probability, and
also identify high-risk components of the system. For the FOWT in this
The authors acknowledge the support of the Development of
paper, control systems and electrics feature most in consideration of
Floating Offshore Wind Turbine Risk Assessment Software project in
failure rates; generators are most frequently cited in connection with
this work, funded by the International S & T Cooperation Program of
failure effects. These results could be useful to FOWT design improve-
China, Contract Number 2013DFE73060. Partial data were provided
ment and maintenance optimization.
by the Goldwind Sci. and Tech. Co., Ltd. Prepared under National
Being a major failure mode of the system, mooring line fracture
Nature Science Foundation of China (No.51609053).
needs particular attention. FOWT mooring lines have a high probability
of failure due to fatigue, corrosion, impact, extreme sea conditions and
References
other risks. Large-scale wind turbine can be built to over 100 m in
height; movement of the floating foundation may cause strong oscilla-
Ahire, C.P., Relkar, A.S. Correlating Failure Mode Effect Analysis (FMEA) & Over all
tion on the upper structure, which brings impact on blades, transmis- Equipment Effectiveness (OEE).
sion and control system. The strength and stiffness of mooring lines Anthony, M.V., Andrew, J.G., Habib, J.D., 2015. Model test of a 1:8-scale floating wind
face challenges and even a minor failure could lead to serious turbine offshore in the Gulf of Maine. J. Offshore Mech. Arct. Eng.-Trans. ASME, 137.
Arabian-Hoseynabadi, H., Oraee, H., Tavner, P.J., 2010. Failure modes and effects
consequences. It is extreme important to not only choose the material analysis (FMEA) for wind turbines. Electr. Power Energy Syst. 32, 817–824.
and layout scheme of the mooring lines, but also optimize floating Braam, H., Rademakers, L.W.M.M., 2004. Models to analyse operation and maintenance
foundation design in order to minimize the impacts to the marine aspects of offshore wind farms. ECN Rep..
Bussel, G.J.W., Zaaijer, M.B., 2011. Reliability, availability and maintenance aspects of
environment. large-scale offshore wind farms. In: Proceedings of the 2001 European wind energy
In the calculation of correlation between main failure modes of conference, Copenhagen, Denmark, pp. 557–560.
FOWT, the theoretical model is appropriately simplified without losing He, S.Q., Wang, S., 1993. Structural Reliability Analysis and Design. National Defence
Industry Press, Beijing, China.
observation of the correlation, particularly due to the high correlation Hou, G.L., Ou, J.P., 2002. Computation and applications of structural reliability index
coefficients between the electronic components. In addition, failure vectors. Chin. J. Comput. Mech. 19, 144–147.
modes of the electronic components account for a large proportion in Madjid, K., Michailides, C., 2015. V-shaped semisubmersible offshore wind turbine: an
alternative concept for offshore wind technology. Renew. Energy 83, 126–143.
the weakest failure modes basis, thus this type of failure modes call for Marquez, F.P.G., Perez, J.M.P., Marugan, A.P., Papaelias, M., 2016. Identification of
development of more than structure failure modes. The correlation critical components of wind turbines using FTA over the time. Renew. Energy 87,
between failure modes demonstrates that we cannot simply add the 869–883.
Nugent, R.P., Krohn, M.A., Hillier, S.L., 1991. Reliability of diagnosing bacterial
probability of the failure modes to obtain an overall result in the
vaginosis is improved by a standardized method of gram stain interpretation. J. Clin.
reliability calculation with series relationship between the failure Microbiol. 29 (2), 297–301.
modes. If an accurate risk assessment of the entire system is required, Pérez, J.M.P., Márquez, F.P.G., Tobias, A., Papaelias, M., 2013. Wind turbine reliability
analysis. Renew. Sustain. Energy Rev. 23, 463–472.
the correlation between each failure mode is imperative. This method
Polinder, H., Bang, D.J., Li, H., Chen, Z., 2007. Concept report on generator topologies,
suits not only the FOWT system, but also other engineering applica- mechanical and electromagnetic optimization. Tech. Rep. Project UpWind,
tions. Mekelweg, the Netherlands and Aalborg East, Denmark.

387
J. Kang et al.
Popa, L.M., Jensen, B.B., Ritchie, E., Boldea, I., 2003. Condition monitoring of wind
generators. In: Proceedings of the 38th IAS Annual Meeting on Industry
Applications Conference, pp. 1839–1846.
Ribrant, J., 2006. Reliability Performance and Maintenance—a Survey of Failures in
Wind Power Systems (Master's thesis). KTH School of Electrical Engineering,
Stockholm.
Spinato, F., Tavner, P.J., Bussel, G.J.W., Koutoulakos, E., 2009. Reliability of wind
turbine subassemblies. Renew. Power Gener. IET 3 (4), 387–401.
388
Ocean Engineering 129 (2017) 382–388
Tavner, P.J., 2012. Offshore wind turbines: reliability, availability and maintenance.
Renew. Energy Ser., 13.
Tavner, P.J., Xiang, J.P., Spinato, F., 2007. Reliability analysis for wind turbines. Wind
Energy 10, 1–18.
Zhang, X., Sun, L., Ma, C., Fassezke, E., Sun, H., 2016. A reliability evaluation method
based on the weakest failure modes for side-by-side offloading mooring system of
FPSO. Journal of Loss Prev. Process Ind. 41, 129–143.