Applied Ocean Research 71 (2018) 77–89

Contents lists available at ScienceDirect

Applied Ocean Research

journal homepage: www.elsevier.com/locate/apor

Fatigue reliability analysis of mooring system for fish cage

Hui-Min Hou a , Guo-Hai Dong a , Tiao-Jian Xu a,∗ , Yun-Peng Zhao a , Chun-Wei Bi a ,
Fu-Kun Gui b
a
State Key Laboratory of Coastal and Offshore Engineering, Dalian University of Technology, Dalian 116024, China
b
Marine Science and Technology School, Zhejiang Ocean University, Zhoushan 316000, China

a r t i c l e i n f o a b s t r a c t

Article history: Fish cages in the open sea are exposed to cycle loads due to irregular wave climate during their service
Received 4 June 2017 life, and thus the fatigue reliability assessment of mooring system should be conducted to ensure the safe
Received in revised form operation. The aim of this study is to evaluate the fatigue failure probability of mooring system for fish
19 December 2017
cage. Numerical simulation of net cage in random waves is performed and the time dependent approach
Accepted 20 December 2017
is applied to conduct the fatigue reliability analysis of shackle chains based on S-N curve method. The
sensitivity analysis of fatigue reliability of mooring line to the uncertainty of random variables in the
Keywords:
fatigue limit state is conducted. In addition, the system reliability for mooring system is analyzed and the
Failure probability
Fatigue
effect of the initial pretension and safety factor on system reliability is investigated. The results indicate
Mooring system that a case without the initial pretension on anchor lines is helpful to decrease the failure probability of
Sensitivity analysis mooring system and the safety factor of mooring lines in the current regulation is conservative for the
system reliability against fatigue damage.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction complicated mooring system of fish farms exposed to wave loads

Aquaculture is today the fastest growing sector of the world
food industry, increasing in volume by more than 10% per year,
and currently accounting for more than 35% of all fish consumed,
and now it is becoming the world’s largest protein source. In the
past decades, the aquacultural farms have been located in shel-
tered waters inside the fjords, where the farms are protected from
extreme weather. However, the expansion of near-shore aquacul-
ture is becoming more difficult due to coastal multi-use issues and
environmental impact concerns [1]. Thus, the fish farm is recently
forced to move into the offshore area, which means that the more
exposed area will be utilized and the more severe environmen-
tal load will be applied. Therefore, the future designs of fish farm
must be significantly more robust than the present designs, which
frequently experience the escape of fish. From an engineering per-
spective, the main focus will be to design a system which has an
overall acceptable reliability. Time-dependent reliability analysis
of mooring lines for single-cage and multi-cage system was con-
ducted for the ultimate limit state considering the corrosion effect
and the uncertainties of the significant wave height and period,
and the corrosion depth of chains [2]. Although the extreme envi-
ronmental event may govern the design on some occasions, the
∗ Corresponding author.
is vulnerable to cumulative fatigue damage due to the cyclic nature
of the wave loading, and the fatigue failure may occur earlier than
the emergence of the extreme environmental event. Thus, for both
existing and future fish farms, the integrity of the mooring system
should be investigated in order to withstand the environmental
cyclic loads.
Numerous numerical studies have been conducted on the
hydrodynamic analysis of fish cage structures, including net
panels, floating collars, net cages and mooring systems. Lader and
Fredheim [3] established a numerical model using super element to
investigate the dynamics properties of a flexible net sheet under the
wave loads. Balash et al. [4] analyzed the steady loads on the plane
net through numerical simulation, in which the net is considered as
an inter-connected system of lumped masses and springs. Paturs-
son et al. [5] modeled the net as a sheet of porous media to improve
the computational efficiency and obtained the flow characteristics
through and around net panel. Bouhoubeiny et al. [6] performed
Time-Resolved Particle Image Velocimetry measurements to study
the hydrodynamic flow interaction with fishing net structure
and demonstrated the influence of fluttering net structure. Zhou
et al. [7] investigated the hydrodynamic characteristics of knot-
less nylon netting with the variation of solidity ratio in normal,
parallel and angle of incline to free stream. Kristiansen [8] ana-
lyzed fully nonlinear wave body interaction problems by numerical

https://doi.org/10.1016/j.apor.2017.12.008
0141-1187/© 2017 Elsevier Ltd. All rights reserved.
78 H.-M. Hou et al. / Applied Ocean Research 71 (2018) 77–89

Nomenclature A Standard deviation of scale parameter

B Standard deviation of shape parameter
T Tension force (·) Gamma function
l Elongation of mooring lines X Random variables
l Initial length of mooring lines z(X) Limit state function for variable vector X
S(f) Input wave spectrum P{z(X)≤0} Probability of z(X)≤0
Hs Significant wave height f(X) Joint probability density function of X
Ts Significant wave period pf Failure probability of structures
Tp Spectral peak period R Reliability of structures
f Wave frequency Y Independent variables
fp Spectral wave frequency ˇ Reliability index
 Peak enhancement factor X* =(x1 ,x2 ) Design point in X space
 Peak shape factor Y* =(y1 ,y2 ) Design point in Y space
u(x, z, t)Horizontal velocity of water particles at time t Ф−1 (·) Inverse function of the standard normal probability
w(x, z, t) Vertical velocity of water particles at time t distribution
ai Wave amplitude of ith component wave
Xi Expected value of random variable Xi
h Water depth  Xi Standard deviation of random variable Xi
fi Wave frequency of ith component wave εXi Important factor of random variable Xi
ki Wave number of ith component wave SFi Sensitivity factor for random variable Xi
εi Random phase of ith component wave P(Ci ) Failure probability of the ith component
x Horizontal coordinate of water particles Plower Lower bound for system failure probability
z Vertical coordinate of water particles Pupper Upper bound for system failure probability
p Occurrence probability of each sea state P(Cij ) Joint failure probability of the ith and the jth com-
CDj Drag coefficient in the direction of the j component, ponents
j = ,, ˇi , ˇj Reliability indices for the ith and jth components
Cmj Added mass coefficient in the direction of the j com- Ф(·) Cumulative probability distribution function for 1-D
ponent, j = ,, standard normal distribution
Fj External forces on the net twine for the j component, ij Correlation coefficient between the ith and the jth
j = ,, components
uj Fluid particle velocity vector at the element center F Safety factor
for the j component, j = ,, f Partial safety factor for fatigue load
Aj Projected area for the j component, j = ,, Kc Characteristic value of K
V0 Water displaced volume of an element v0c + Characteristic value of v0 +
Ṙj Central velocity vector of element for the j compo- Bc Characteristic value of B
nent, j = ,, Ac Characteristic value of A
u̇j Fluid particle acceleration vector at the element c Characteristic value of 
center for the j component, j = ,,
R̈j Central acceleration vector of element for the j com-
ponent, j = ,,
Density of water wave tank to calculate wave loads on a floating horizontal

Viscosity of water collar. Fu and Moan [9] predicted the dynamic response of 5 by
C Normal drag coefficient for mesh bar 2 floating collars by the application of an extended 3D hydro-
C Tangential drag coefficient for mesh bar elasticity theory in regular waves. Huang et al. [10] developed a
VR Normal component of the fluid velocity relative to finite element model to investigate the elastic deformations and
the bar mooring line tensions of floating collar in waves. Lee et al. [11]
S Stress range (double amplitude) in MPa developed a mass-spring model to analyze the performance of
N Number of cycles for the stress range S to failure fish cage system in current and waves. Moe et al. [12] performed
K Intercept parameter of S-N curve strength analysis to obtain the loads distribution in the net cage
m Slope parameter of S-N curve due to current, weights and gravity. Xu et al. [13] investigated
D Accumulative fatigue damage the hydrodynamic behavior of multi-cage and mooring system
TL Design lifetime by lumped-mass model under the action of waves combined
N(Sj ) Number of cycles to failure at stress range Sj with current. Li et al. [14] analyzed the nonlinear hydro-elastic
N(TL ) Number of stress cycles in total time TL response by finite element model of a deep-water gravity cage in
A Scale parameter in Weibull distribution irregular waves. Grue [15] predicted the mooring line loads for
B Shape parameter in Weibull distribution two systems of gravity net cages under the action of wave and
v0 + Load cycle per unit time currents through numerical simulation. Kim et al. [16] analyzed
 Allowable fatigue damage the flow field characteristics within the abalone containment
N(e, sd) Normal distribution with the expected value e and structure with computational fluid dynamic software and inves-
standard deviation sd tigated the hydrodynamic response of the moored containment
LN(e, sd) Log-normal distribution with the expected value e structure with a Morison equation type finite element model. Ito
and standard deviation sd et al. [17] investigated the hydrodynamic behaviors of a cubic
H Hessian matrix shaped elastic net structure and estimated the mooring forces and
CA,B Covariance matrix mooring displacements. Yao et al. [18] proposed a hybrid volume
AB Correlation coefficient of scale and shape parame- approach to add the resistance force of the net cage into the flow
ters field for coupling the fluid and net. Winthereig-Rasmussen et al.
 H.-M. Hou et al. / Applied Ocean Research 71 (2018) 77–89 79

[19] performed numerical simulations by solving the Reynolds Table 1

Parameters of simplified fish cage and mooring lines.
average Navier-Stokes equations using finite volume approach
and discussed the effect of cage layouts and net solidities on the Component Parameter Value
flow through and around sea cages. Outer circle General diameter (m) 16.92
Due to cyclic loading in irregular wave climate, fatigue becomes Pipe diameter (m) 0.013
potentially one of the main problems causing reduction in the Unit mass (kg/m) 11.36
integrity of mooring system. The net cage and mooring system Material HDPE
Inner circle General diameter (m) 15.92
will be deployed in the offshore area for a long period, thus the
Pipe diameter (m) 0.013
fatigue damage analysis of the mooring line is a significant part Unit mass (kg/m) 11.36
of the design of mooring system. To our knowledge, the fatigue Material HDPE
characteristic of mooring line was analyzed through experimental Mooring lines Diameter (m) 0.0364
Density (g/cm3 ) 1.14
tests and numerical simulations. Williams et al. [20] investigated
Grid line length (m) 40
experimentally how to account for the damage, fatigue lifetime Anchor line length (m) 58.8
and reliability index of synthetic fiber mooring ropes. Olsen [21] Bridle line length (m) 20.86
estimated the annual failure probability of mooring lines as a func- Buoy line length (m) 4.0
tion of safty factors through physical model tests. Thies et al. [22] Material PP
Bottom ring General diameter (m) 15.92
and [23] analyzed the fatigue load spectrum of mooring system for
Pipe diameter (m) 0.06
the floating marine energy converter, and compared the expected Unit mass (kg/m) 4.05
fatigue damage with the design limits of mooring line. Davies et al. Material HDPE
[24] proposed an empirical model to predict the lifetime of braided
HMPE handling ropes under cyclic bend. Liu et al. [25] conducted
a series of experimental investigation on the dynamic stiffness of cal uncertainty of long term stress range distribution parameters
damaged fiber mooring ropes, and the effect of mean load, strain obtained from our numerical model and to extend the knowledge
amplitude and loading cycles on the dynamic stiffness of dam- of the effect of initial pretension and safety factor on fatigue failure
aged ropes was analyzed. Larsen and Mathisen [26] performed the probability of the whole mooring system for fish farms. And the
reliability-based fatigue analysis of mooring chains and wire rope overview of this study is shown in Fig. 1.
mooring lines, in which the uncertainties of the fatigue loads and The study is organized as follows. Section 2 describes the numer-
fatigue strength of mooring lines were considered. Siddiqui and ical model of net cage and mooring system, and then the fatigue
Ahmad [27] conducted fatigue reliability analysis for tether joints of reliability analysis for the mooring system is presented in Section
TLP tethers based on non-linear dynamic analysis using Palmgren- 3. After that, some results and discussion of the case study are given
Miner’s rule and fracture mechanics method. Gao and Moan [28] in Section 4. Finally, some conclusions are drawn.
proposed an effective frequency-domain method for fatigue analy-
sis of a chain-wire-chain mooring system, which was validated by 2. Numerical model for net cage and mooring system
using the rainflow counting method and the time-domain fatigue
analysis. Han et al. [29] developed a time-domain computational In this study a time-domain numerical simulation of the moor-
method to determine the extreme tension and proposed a spectral ing system for net cage in the open sea was conducted. The water
analysis method for estimating the fatigue life of mooring system. depth for the moored fish cage system is 20 m. Following previ-
Toft and Sørensen [30] demonstrated the reliability-based design ous works [35], this study adopted the self-developed numerical
of the wind turbines in fatigue limit states. Low [31] extended model based on the lumped-mass method and rigid body kinemat-
a time/frequency domain hybrid method for the fatigue analysis ics principle to obtain the tension force on mooring lines for fish
of moorings and risers, and the fatigue damage obtained agreed cage. The numerical model was validated by physical model test
well with the results from the time domain analysis using the in our previous research. A brief description of numerical model is
rainflow counting method. Jameel and Ahmad [32] conducted the given here.
fatigue reliability assessment of mooring lines in random waves
and current considering the damping effect of spar. Dong et al. [33] 2.1. Net cage and mooring system
predicted the contact fatigue damage of the gears for a wind tur-
bine. Li et al. [34] investigated the probabilistic density function The net cage and mooring system is composed of a gravity cage
for the amplitude of tension forces on mooring lines to assess the and a submerged grid mooring, as shown in Fig. 2. The fish cage
fatigue damage with the spectral method. consists of float collar, net pen, bottom ring and sinkers. The cage
The prediction of fatigue reliability of the mooring system for net is made of PE with a mass density of 953 kg/m3 . The detailed
the floating marine structures is becoming an important issue to parameters of the fish cage and mooring system are given in Table 1.
quantify the safety status. The mooring system for fish cage belongs There are 1800 meshes in the circumferential direction and 240
to a multiple combination of series and parallel connection for dif- meshes in the depth direction. The net is knotless, with a mesh size
ferent type of lines, rather than the normal parallel connection for of 46.8 mm and a twine thickness of 1.44 mm. The net mounted as
other floating structures. And this complicate connected way of the diamond meshes forms an open vertical cylinder with a diameter
mooring lines may lead to the load re-distribution with the varia- of 15.92 m and a height of 9 m. There are 10 sinkers, and the mass
tion of the initial pretension. The failure of one mooring line for of each sinker is 14.6 kg. The shape of the sinkers is spherical, with
fish cage would result in the failure of mooring system. And the a diameter of 30 cm.
system reliability aspect of the integrity of mooring system is not The mooring system consists of a submerged (4 m below the
well emphasized in the regulation. The current practice is to adopt surface), pre-tensioned square grid (40 m × 40 m in plane). The
a uniform safety factor for mooring lines, but whether the proba- mooring system includes bridle lines, grid lines, buoy lines and
bility of failure for mooring system against the fatigue damage is anchor lines and the lines are named as shown in Fig. 3. Gravity
acceptable should be check. However, the fatigue reliability analy- cage is located at the middle of each grid. It is connected to the sub-
sis of mooring system for fish cage structures is very limited. Thus merged mooring grid by four bridle lines. The grid is anchored to the
the aim of this study is to investigate the reliability of mooring bottom using eight anchor lines. The motion of the net cage induced
system for fish cage against fatigue damage based on the statisti- by wave loads will cause the deformation of mooring system and
80 H.-M. Hou et al. / Applied Ocean Research 71 (2018) 77–89

Fig. 1. Flow chart of the overview.

Fig. 2. Schematic of: (a) net cage and mooring system; (b) net cage.

the extension/slack of the bridle lines, the grid lines, the buoy lines
and the anchor lines. The deviation of net cage from their balance
position is mainly caused by external loads, and the net cage will
return to their balance position due to the restoring force after the
external loads are removed. When the excitation frequency of net
cage under wave loads is close to the natural frequency of the sys-
tem, the response of net cage and mooring system is extremely
large.
The material of the fiber rope is Polypropylene (PP) and the elas-
ticity of the fiber rope is T = 562(l/l)1.132 , where T is the tension
of fiber rope with the unit of kN; l is the elongation of mooring
lines; l is the initial length of mooring lines (referring to: Xu et al.
[36]). The broken shackle on one of Kikeh’s anchors came as a sur-
prise to the industry in 2007, and the connectors were the second
component in terms of the number of incidents [37]. The PP rope is
assumed to have a better fatigue behavior than the chain or chain
accessories (shackles) [38]. Thus shackle chains are considered as
the fatigue critical components of mooring lines in this paper. The
diameter of shackle chains is selected as 16 mm. For the extreme
wave conditions of significant wave height 6.25 m and period 7.5 s,
the maximum stress of shackle chains is 209 MPa, less than the yield
strength 335 MPa of steel significantly.
2.2. Brief description of wave field
Wave force is the main environmental cycle loading acting on
the fish cage structure in the open sea. Since a small part of floating
fish cage is exposed in the air, the wind load is negligible here. The
modified JONSWAP spectrum is used as the input wave spectrum,
as follows,
  −4   2

exp −(f/fp −1) /(2 2 )
S (f ) = ˇJ Hs2 Tp−4 f −5 exp −1.25 Tp f · (1)
where  is the peak enhancement factor and normally a
value of 3.3 can be used; ˇJ = 0.06238/(0.230 + 0.0336-
0.185(1.9 + )−1 )·(1.094–0.01915ln); Tp is the spectral peak
period, Tp = Ts /(1–0.132( + 0.2)−0.559 ); Hs and Ts are the signif-
icant wave height and wave period, respectively; f is the wave
frequency; fp is the spectral peak frequency;  is the peak shape
factor,  = 0.07 (f ≤ fp ),  = 0.09 (f > fp ).
 H.-M. Hou et al. / Applied Ocean Research 71 (2018) 77–89 81

on the net using the modified Morison equation considering the

relative motion between net element and flow field,
1
F = CD A |u − Ṙ | · (u − Ṙ ) + V0 u̇ + Cm V0 (u̇ − R̈ )
2
1
F = CD A |u − Ṙ | · (u − Ṙ ) + V0 u̇ + Cm V0 (u̇ − R̈ ) (4)
2
1
F = C A |u − Ṙ | · (u − Ṙ ) + V0 u̇ + Cm V0 (u̇ − R̈ )
2 D  
where F and F are the normal components of external forces on
the net twine, and F is the tangential component of external forces
→  
on the net twine; u = u , u , u is the fluid particle velocity vec-
→  
tor at the element center, Ṙ = Ṙ , Ṙ , Ṙ is the central velocity
 
 = u̇ , u̇ , u̇ is the fluid particle acceleration
vector of element, u̇ 
→  
vector at the element center, R̈ = R̈ , R̈ , R̈ is the central accel-
eration vector of element, is the density of water, V0 is the water
displaced volume of an element; A , A and A are the effective
projected areas of an element in the direction of the , ,  com-
ponents, respectively; CD , CD and CD are the drag coefficients in
the direction of the , ,  components, respectively and Cm , Cm
and Cm are the added mass coefficients in the direction of the , ,
Fig. 3. Top view of mooring system.  components, respectively.
For the floating collar, the hydrodynamic coefficients are
obtained based on a series of physical model test. Herein, accord-
ing to our previous research [40], the hydrodynamic coefficients
The horizontal and vertical velocity of water particles for the of float collar are taken as constants: the drag coefficients are
random waves is calculated as the superposition of multiple chosen as CD = 0.4 for tangential component, CD = CD = 0.6 for nor-
monochromatic waves based on the linear Airy wave theory, mal component; and the added mass coefficients are chosen as
Cm = 0.0, Cm = Cm = 0.2. The hydrodynamic coefficients of bottom

n
cosh(ki (z + h))
ring are also taken as constants: the drag coefficients are chosen as
u(x, z, t) = ai · 2fi cos(ki x − 2fi t + εi ) (2) CD = 0.1 for tangential component, CD = CD = 1.2 for normal com-
sinh(ki h) ponent; and the added mass coefficients are chosen as Cm = 0.0,
i=1
Cm = Cm = 1.0 referring to [41]. The hydrodynamic coefficients of
sinkers are: the drag coefficients are 0.5 and the added mass coef-

n
sinh(ki (z + h))
w(x, z, t) = ai · 2fi sin(ki x − 2fi t + εi ) (3) ficients are chosen as 1.0 when modeling the sinker as a sphere.
sinh(ki h) For the net twine and the mooring lines, the hydrodynamic
i=1
coefficients are calculated using a method described by Choo and
Casarella [42].
For the case study in Section 4, the long term wave conditions in
⎧      
one year are represented by 27 short-term sea states in P4 of Bohai ⎪ 8/ Re s · 1 − 0.87s−2 0 < Re ≤ 1
Sea according to Wang et al. [39]. The sea states are assumed to
⎨  
C = 1.45 + 8.55Re−0.90 1 < Re ≤ 30 (5)
propagate from the same direction all the time. The environmental ⎪
⎩  
parameters are presented in Table 2, where Hs is the significant 1.1 + 4Re−0.50
 30 < Re ≤ 10 5
wave height, Ts is the significant wave period and p is the occurrence 
probability of each sea state. C = 
0.55Re
1/2
+ 0.084Re
2/3
(6)

where Re = VR D/

; s = −0.077215665 + ln(8/Re );
is the vis-
2.3. Calculation of environmental loadings on fish cage and cosity of water; C and C are the normal and tangential drag
mooring system coefficients for mesh bar; VR is the normal component of the fluid
velocity relative to the bar, and is the density of water.
To calculate the hydrodynamic forces acting on the fish cage
system, the fish cage system is divided into numerous elements that 3. Fatigue reliability analysis
are subjected to wave loadings, as shown in Figs. 4 and 5. A local
coordinate system (--) is defined to describe the direction of In this study, the fatigue damage and the fatigue limit state func-
the hydrodynamic forces acting on elements. Concerning the local tions of shackle chains are given using the S-N curve approach.
coordinate system --, the -axis lies on the plane that includes In addition, the uncertainties of parameters and the methods of
the -axis and V, and the velocity vector of the water particle at the fatigue reliability estimation for mooring system are given as
the mid-point of the element can be divided into  (tangential) and follows.
 (normal) components. Based on the lumped-mass method, the
external forces on each element are calculated first and then evenly
3.1. Definition of limit state function
distributed to the corresponding nodes. The scattering effect of the
net twine on the flow field can be neglected because the diameter
In general the S-N curve takes the following form:
of net twine is relatively small compared with the characteristic
m
wave length, and thus, it is appropriate to calculate the wave loads NS =K (7)
82 H.-M. Hou et al. / Applied Ocean Research 71 (2018) 77–89

Table 2
A wave scatter diagram for the Bohai Sea representing with 27 short-term sea states.

No. Hs (m) Ts (s) p(%) No. Hs (m) Ts (s) p(%) No. Hs (m) Ts (s) p(%)

1 0.25 1.5 0.35 10 1.75 4.5 10.63 19 4.25 5.5 0.10

2 0.25 2.5 23.88 11 1.75 5.5 0.01 20 4.25 6.5 0.19
3 0.25 3.5 0.43 12 2.25 4.5 6.08 21 4.75 5.5 0.01
4 0.75 2.5 15.20 13 2.25 5.5 0.24 22 4.75 6.5 0.10
5 0.75 3.5 17.26 14 2.75 4.5 0.53 23 5.25 6.5 0.05
6 0.75 4.5 0.09 15 2.75 5.5 2.70 24 5.25 7.5 0.04
7 1.25 3.5 17.44 16 3.25 5.5 1.51 25 5.75 6.5 0.02
8 1.25 4.5 1.46 17 3.75 5.5 0.65 26 5.75 7.5 0.25
9 1.75 3.5 0.70 18 3.75 6.5 0.23 27 6.25 7.5 0.06

Fig. 4. Schematic of mass-spring model: (a) net pen; (b) mooring line.

where N is the number of cycles of stress range S to failure, K is where N(TL ) is the number of stress cycles in total time TL , N(Sj ) is
the intercept parameter of S-N curve, m is the slope of S-N curve. the number of cycles to failure at stress range Sj . By the application
Considering that the fatigue life of shackles is comparable to that of a single slope S-N curve, the damage can be estimated as
of the common links of the same size and grade from the limited ⎡ ⎤
test data [43], the adoption of the design curve for studless chain 
N(TL )
1 1 
N(TL )
1  
is conservative and acceptable, although there are insufficient data D= Sjm · = ·E⎣ Sjm ⎦ ≈ · E [N (TL )] E S m (9)
K K K
for the design curve of shackles. Here, m is 3.0 and K is assumed as j=1 j=1
6.0 × 1010 for shackle chains according to DNV OS-E301 [44] and
Standards Norway [45]. The Weibull distribution ([46,47]) is assumed to represent the
The accumulated fatigue damage D in time TL by adopting the distribution of stress range in long term. The expected value of the
Miner-Palmgren rule is given as mth moment for stress range is calculated as follows,
 
  ∞
m
E Sm = sm · f (s) ds = Am · 1+ (10)
0
B

where A and B are the scale and shape parameters in Weibull dis-
tribution, and (·) is the gamma function. Thus the damage D in a
period of time TL is

N(TL )
1
D=   (8) v+ T  
0 L
j=1
N Sj D= · Am · 1 + m/B (11)
K

Fig. 5. Schematic of coordinate system for floating collar and bottom ring.
 H.-M. Hou et al. / Applied Ocean Research 71 (2018) 77–89 83

Table 3
Basic random variables. N(e, sd) ∼Normal, LN(e, sd) ∼Lognormal (e and sd represent
expected value and standard deviation).

Variables Symbol Distribution

Allowable fatigue damage  LN(1.0,0.03)

Load cycle per unit time v0 + N(0.4197,0.008)
S-N curve parameter ln(K) N(25.51,0.345)

where v0 + is the load cycle per unit time in long term and can be
estimated by the summation of the product of the occurrence fre-
quency and the corresponding load cycle per unit time for each sea
state. Thus the limit state equation using the Miner’s rule in a period
of time TL can be written as follows,
 
v+ T
0 L
 
z =1− · 1 + m/B · Am = 0 (12)
K

Including the uncertainty of the allowable fatigue damage, the

limit state function can be written as,
 1/m
1/m
v+ T
0 L
 
z= − · 1 + m/B ·A (13)
K

where  is the uncertainty for allowable fatigue damage, which

depends on the long-term distribution of stress cycles, S-N curve
and Miner’s rule as well as accessibility for inspection and conse-
quences of fatigue failure.

3.2. Uncertainties of parameters in limit state function

Fig. 6. Flow chart of FORM.

The uncertainties that exist in the fatigue reliability analysis

mainly consist of the scatter of S-N experimental data, environmen- 3.3. Calculation of fatigue reliability
tal loading, and the Miner’s rule. In this context, the uncertainties
of parameters in the stress range distribution, parameter K in the Reliability can be defined as the probability of a performance
S-N curve, load cycle per unit time and allowable fatigue damage function z(X) greater than zero, i.e. P{z(X) > 0}. In other words, reli-
are considered. ability is the probability that the random variables X are in the
In order to obtain the statistical uncertainty of parameters of safe region that is defined by z(X) > 0. The failure probability can be
stress range distribution in long term, the long term stress range defined as the probability P{z(X) ≤ 0}. If the joint probability density
is calculated by the stress range in short term weighted with the function of X is f(X), the failure probability of structures is calculated
occurrence of each sea state referring to [48]. The statistical uncer- as follows,
tainty of the parameters in the distribution of the stress range is 
evaluated in the classic statistics analysis. If the sample size is large  
pf = P z (X) ≤ 0 = f (X) dX (15)
enough, the estimated parameters will follow Normal distribution z(X)≤0
with the expected value obtained from Maximum-Likelihood esti-
mators and the covariance matrix CA,B related to the Hessian matrix The reliability is then calculated as follows,
H which is the second order derivatives of the Log-likelihood func- 
 
tion according to Lindley [49], R = 1 − pf = P z (X) > 0 = f (X) dX (16)
  z(X)>0
A2 AB A B
CA,B = [−H]−1 = (14) The FORM is applied to calculate the reliability index and the
AB A B B2 flow chart of the whole procedure for FORM is shown in Fig. 6. The
reliability analysis includes the definition of the failure function and
where  A and  B are the standard deviations of the scale and shape the random variables, establishing distributions of various random
parameters; AB is the correlation coefficient of the scale and shape variables and transformation of the dependent variables into the
parameters. independent standard normal variables, and calculating the fail-
The uncertainties of allowable fatigue damage, load cycle per ure probability. Since the scale parameter and shape parameter are
unit time v0 + and parameter K are also analyzed and quantified in correlative, Rosenblatt transformation is adopted to transform the
this paper. The allowable fatigue damage is typically modeled as a correlated variables X following the aforementioned joint proba-
log-normal distributed random variable, for instance [50–53]. The bility density distribution into a space of independent variables Y
normal function is assumed to describe the probability distribution which follow the standard normal distribution. In the Y space, the
of v0 + according to [50]. And the parameter K is assumed to be reliability index ˇ can be calculated as
log-normal distributed referring to [27]. Therefore, the involved
uncertainties of the allowable fatigue damage, load cycle per unit zX (X∗ ) − Y∗ T ∇ zY (Y∗ )
ˇ= (17)
time and parameter K in the S-N curve are presented in Table 3.
∇ zY (Y∗ )

84 H.-M. Hou et al. / Applied Ocean Research 71 (2018) 77–89

random variable Xi . Referring to [54], the sensitivity factor SFi can

be calculated as
εXi
SFi =  (20)

 n
 εX 2
i
i=1

3.5. Fatigue failure probability of mooring system

The fatigue reliability of the whole mooring system is evaluated

after obtaining the fatigue reliability index of each mooring line. If
one mooring line is failure, the whole mooring system is considered
as failure. The bounding technique proposed by Ditlevsen [55] is
adopted here to estimate the fatigue reliability of the whole system,
and Garbatov and Guedes Soares [56] suggested that the Ditlevsen
bounds are more practical and useful way than the exact expres-
sions. The lower and upper bounds for the probability of failure are
calculated as follows,
⎧ ⎫

n ⎨ 
i−1
  ⎬
Plower = P (C1 ) + max P (Ci ) − P Cij , 0
⎩ ⎭
i=2 j=1 (21)
 n

n
  
Pupper = P (Ci ) − maxj<i P Cij
Fig. 7. Flow chart of crude Monte-Carlo method. i=1 i=2

where P(Ci ) is the failure probability of the ith component of the

where X∗ = (x1 , x2 ) is the design point in X space and Y∗ = (y1 , y2 ) is mooring system, P(Cij ) is the joint failure probability of the ith and
the corresponding design point in Y space. The reliability index ˇ the jth components of the mooring system. Ditlevsen [55] sug-
can be calculated as follows, gested that the lower and upper bounds of P(Cij ) can be calculated
  as
ˇ = ˚−1 1 − pf (18)    
max Pij , Pji ≤ P Cij ≤ Pij + Pji (22)
where Ф−1 (·) is the inverse function of the standard normal prob- where
ability distribution. ⎛ ⎞
In addition, Monte-Carlo method is applied for calculating the   ˇj − ij ˇi
failure probability in complex limit state function and verifying the Pij = ˚ −ˇi · ˚ ⎝− " ⎠
efficiency of FORM. The integral in Eq. (15) can be calculated directly 1 − ij2
by Monte-Carlo method instead of the analytical or iteration solu-
⎛ ⎞ (23)
  ˇi − ji ˇj
Pji = ˚ −ˇj · ˚ ⎝− " ⎠
tion. The flow chart of the crude Monte-Carlo method is shown
in Fig. 7. The procedures of crude Monte-Carlo method involve
1 − ji2
randomly sampling a plenty of realizations for the uncertain vari-
ables, calculating the sign of limit state function and evaluating where ˇi and ˇj are the reliability indices for the ith and jth com-
the occurrence of positive signs in simulation samples as the fail- ponents for mooring system, Ф(·) is the cumulative probability
ure probability. For rare events, the sample size should be large distribution function for one-dimensional standard normal distri-
enough and even exceeds the capacity of the computer some- bution and ij is the correlation coefficient between the ith and the
times to obtain the accurate results. Thus, the important sampling jth components failure for mooring system.
method is adopted to improve the sampling efficiency. The main
idea of the important sampling method is to sample from a differ-
3.6. Safety factor
ent distribution which has a lower variance than the initial variance.
It is a convenient way to use the design point obtained from FORM
The safety factor is commonly used in conventional engineer-
as the expected value of the new distribution in the important
ing practice to account for the uncertainties. More recently, the
sampling method.
reliability-based design of structures is a more rigorous and reason-
able way for tackling the uncertainties and making the structures to
3.4. Sensitivity analysis of the fatigue reliability be safe in a probabilistic sense. However, the relationship between
the reliability index and the safety factor is still ambiguous for engi-
The sensitivity analysis has great implications for understanding neers. In addition, the impact of the safety factor of mooring lines on
the effect of the random variables’ uncertainties. The important the whole mooring system is currently unknown. In this study, the
factor of the random variables represents the relative importance of safety factor of mooring lines is calibrated with the target reliability
the uncertainty of each random variable and it is defined as follows, index of the whole mooring system. The safety factor is defined as
the ratio of the design fatigue damage to the characteristic fatigue
∂ˇ damage, as follows:
εXi = · Xi (19)
∂
Xi
c
F= v+ T   (24)
where
Xi and  Xi are the expected value and the standard devi- 0c L
· 1 + m/Bc · Ac m
Kc
ation of the random variables, εXi is the important factor of the
 H.-M. Hou et al. / Applied Ocean Research 71 (2018) 77–89 85

where Kc is the characteristic value which the natural logarithm

of is the expected value of ln(K) minus two standard deviations
of ln(K) [51], which is equal to the design value in DNV OS-E301
[44] considering that the characteristic fatigue damage in DNV OS-
E301 [44] is calculated by the design value K; v0c + , Bc , Ac , c are
the characteristic value of v0 + , B, A, , respectively, which are the
expected values of the probability distribution. The design equation
can be given as,

  
v+ T
0c L
  m v+ T
0c L
c − F · 1 + m/Bc · Ac = c − ·
Kc Kc
   m
1 + m/Bc · f Ac ≥0 (25)

Note that the F value is correlated to the partial safety factor

 f for fatigue load by F = ( f )m according to [51] and [30]. More
details about the relationship between the failure probability and
the fatigue safety factor can refer to many references, e.g. [57] and
Fig. 8. Joint probability density distribution for the scale and shape parameters.
[58].

4. Case study

Numerical simulations of the fish cage and mooring system

in random waves were performed and the tension histories of
mooring system were obtained. The time dependent approach [59]
is employed to estimate the fatigue reliability based on the S-N
approach. For simplicity, the reliability index of each type mooring
lines is calculated as the reliability index of the shackle chains of
the corresponding type mooring lines in this study.

4.1. Parameter uncertainty of Weibull distribution

The rainflow counting method is adopted to calculate the stress

ranges of shackle chains based on the time history of tension forces
on mooring lines. The representative wave conditions in long term
list in Table 2 are applied. The long-term stress ranges for the Grid-
1 line in Fig. 3 are calculated by weighting the stress ranges with a
duration of 3 h in each sea state with the corresponding occurrence.
Then long term stress ranges are used to estimate the uncertainty
of the shape and scale parameters of the Weibull distribution. The
total fatigue damage of Grid-1 line for one year is 0.036. The joint
probability distribution of the shape and scale parameters is a two-
dimensional Normal distribution with expected value obtained
from Maximum-Likelihood method. The covariance matrix is cal-
culated through the second order derivatives of the Log-likelihood
function. The joint probability density distribution of the scale and
shape parameters is shown in Fig. 8. The coefficients of variation of
the scale and shape parameters for Grid-1 line are 0.044 and 0.012,
respectively. Tarp-Johansen [60], Nejad et al. [61] and Dong et al.
[50] used an uncertainty with coefficient of variation of about 5% in
environmental condition model, which is close to the magnitude
of the uncertainty obtained in this paper. This indicates that it is
profound to investigate the effect of the statistical uncertainty on
fatigue reliability.
4.2. Fatigue reliability of mooring line with different methods
The crude Monte-Carlo method (CMCM), the important sam-
pling method (ISM) and the first order reliability method (FORM)
are the common methods to calculate the reliability index. The
FORM demands the minimum computational resource; however,
it may be not suitable to estimate the reliability index if there is
a strong nonlinear relationship between the random variables and
the limit state function.
Fig. 9. Time dependent reliability index calculated by FORM and ISM.
For the uncertainty of the scale and shape parameters (in Eq.
(13)) of Weibull distribution for stress range of mooring line, the
fatigue reliability of mooring line for net cage system is analyzed
using the ISM and FORM. The design point obtained from FORM is
selected as the expected value of the new distribution in the ISM. A
total of 1000000 samples are considered here. Fig. 9 shows the time
dependent fatigue reliability calculated using the FORM and the
ISM. It indicates that the difference in the fatigue reliability index
calculated using the FORM and the ISM is less than 0.7%. Thus FORM
is an effective method to evaluate the reliability index although the
limit state function is approximately close to the design points by
a linear function. In other words, a weakly nonlinear characteristic
of fatigue limit state function around the design points is demon-
strated. If we just consider the uncertainty of the scale and shape
parameters of Weibull distribution of stress range of mooring line,
it is appropriate to use the FORM to calculate the fatigue reliability
of mooring line for net cage system.
For the uncertainties of random variables in Table 3 and the
scale and shape parameters, the fatigue reliability of mooring line
is analyzed. The fatigue reliability index of mooring line for 20-
year service time obtained from the CMCM and the FORM is 2.4796
and 2.4800, respectively. The FORM will overestimate the reliability
86 H.-M. Hou et al. / Applied Ocean Research 71 (2018) 77–89
Table 4
Sensitivity factors for random variables.
Variables Symbol Sensitivity factor SFi
Allowable fatigue damage  +0.1312
Load cycle per unit time v0 + −0.0866
S-N curve parameter ln(K) +0.1050
Weibull distribution A −0.5560
B +0.8094
index only about 0.16% due to the weakly nonlinear relationship
between the limit state function and the random variables. Thus
considering the uncertainty of all variables in Table 3 and the stress
range distribution parameters, it is suitable to apply the FORM to
calculate the reliability index.
4.3. Parameter sensitivity of fatigue reliability
The sensitivity of fatigue reliability for the Grid-1 line (in Fig. 3)
during 20 years’ service time is considered, and the first order relia-
bility method (FORM) is adopted to calculate the fatigue reliability
in this section. Table 4 shows the sensitivity factors for different
random variables. The negative and positive signs of sensitivity
factors indicate the positive and negative influence on the fatigue
reliability, respectively. Table 4 indicates that the sensitivity factor
for the shape parameter (B) of Weibull distribution of stress range
is largest, which means that the fatigue reliability is the most nega-
tively sensitive to the variation of shape parameter. In addition, the
scale parameter (A) of Weibull distribution has a significant posi-
tive influence on the fatigue reliability, and the fatigue reliability is
also sensitive to the variation of the parameter K in the S-N curve
and the allowable fatigue damage . The absolute value of the sen-
sitivity factor of the load cycle per unit time v0 + is small, which
means that the fatigue reliability is not sensitive to the variation of
the load cycle per unit time v0 + . Therefore, it is acceptable to take
a constant value of the load cycle per unit time v0 + in the evalua-
tion of the fatigue reliability of mooring line for net cage system.
In general, the fatigue reliability of mooring line for net cage sys-
tem is more sensitive to the parameters (A, B) in the stress range
distribution than the other parameters.
4.4. Fatigue failure probability for mooring system
The fatigue failure probability of grid mooring system for net
cage structure is analyzed through Eq. (21) after obtaining the
fatigue failure probability of each mooring line. The grid mooring
system for net cage structure includes 20 mooring lines, as shown in
Fig. 3. If the failure of one mooring line occurs, the grid mooring sys-
tem is considered to be failure. However, it is difficult to determine
the correlation coefficient of the component failure against fatigue
damage for mooring system precisely due to the fabrication, trans-
portation, the installation and operation. In this context, the effect
of the correlation between the failures of different components is
calculated. For simplify, the correlation between the failure of dif-
ferent components is deemed to be the same even for different line
types.
The fatigue failure probability of each mooring line may affect
the bounds of the failure probability of the whole mooring sys-
tem. Madsen [62] and Melchers [63] suggested that the descending
order of the failure probability of individual components leads to
a narrower bound for the failure probability of the whole system,
which means that a relatively accurate failure probability of system
can be obtained. Thus the descending order of failure probability of
mooring lines is adopted here. Fig. 10 shows the failure probability
of mooring system as a function of correlation coefficient between
component failures for different service time of mooring system
Fig. 10. System failure probability bounds as a function of correlation coefficient.
considering the uncertainties of the parameters in stress range dis-
tribution and the variables in Table 3. The results indicate that the
failure probability of mooring system dramatically increases with
increasing service time. The upper bounds are by and large the
same as the lower bounds for independent component failure (i.e.
the correlation coefficient equals to zero). For 5-year service time,
the system reliability is insensitive to the correlation coefficient
when the correlation coefficient is lower than 0.7. However, the
fatigue failure probability for mooring system decreases rapidly
when the correlation coefficient is greater than 0.8. It indicates
that the full-dependent relationship for failure of components will
give an un-conservative result for the fatigue reliability of mooring
system for net cage structure.
4.5. Sensitivity of fatigue reliability of mooring system to initial
pretension
The initial pretension is defined as the initial load of lines by
shortening the initial length and changing the initial elongation
[64]. The level of initial pretension will affect the tension response
and the corresponding stress range of mooring lines. Hence, it is
necessary to analyze the effect of the initial pretension on the
fatigue reliability of mooring system. Five levels of initial pretension
on anchor lines are considered to carry out the fatigue reliabil-
ity analysis of mooring system, including 0 kN, 7.72 kN, 21.6 kN,
28.2 kN and 40.1 kN. Only the uncertainty of the scale and shape
parameters in stress range distribution is considered in this context.
Fig. 11 provides the fatigue damage in one year for different lines
in the mooring system. Here 11 lines are given since the mooring
system is symmetrical with respect to the wave propagation direc-
tion. It can be referred that the re-distribution of the load occurs in
the mooring system, and the initial pretension affects the tension
response of bridle lines, buoy lines and grid lines significantly. As
shown in Fig. 11, the fatigue damage of Brid-2, Anch-2 and Grid-
2 lines decreases with the increase of the initial pretension. This
indicates that the fatigue load of Brid-2, Anch-2 and Grid-2 lines
(and Brid-3, Anch-7 and Grid-4 lines symmetrically) decreases and
the load transmission develops dramatically with increasing initial
pretension.
Fig. 12 shows the fatigue failure reliability of mooring system for
2-year service time as a function of initial pretension considering
the uncertainty of the distribution parameters of stress ranges. The
failure probability of mooring system is related to the largest fatigue
damage in the mooring system. And failure probability increases
 H.-M. Hou et al. / Applied Ocean Research 71 (2018) 77–89
Fig. 11. Fatigue damage of lines in mooring system considering different initial
pretensions.
Fig. 12. Initial pretension effect on fatigue failure probability and reliability index.
dramatically with the increase of initial pretension when the ini-
tial pretension less than 28.2 kN. The reliability index decreases
75% when the initial pretension varies from 0 kN to 28.2 kN. How-
ever, reliability index increases 38% when the initial pretension
increases from 28.2 kN to 40.1 kN, which may be due to the load
re-distribution of bridle lines in the mooring system from Fig. 11.
Fig. 13. System failure probability and reliability bounds with the variation of safety factors.
87
Table 5
Required safety factors for different target reliability levels.
Target Reliability Level Failure Probability Safety Factor
3.72 0.0001 2.31
3.09 0.001 1.85
2.33 0.01 1.43
When designing the grid mooring system for the net cage structure,
it is beneficial to select a case without the initial pretension (i.e.
the initial pretension corresponding to the peak reliability index in
Fig. 12) for obtaining a larger reliability index of mooring system.
However, zero initial pretension may have some negative effect
on the horizontal motion of net cage in some extent. Thus two
independent limit state functions for the reliability analysis of the
fatigue damage of mooring system and the mean offset of net cage
system will be investigated in the future research to obtain a bal-
ance between the deflection of net cage and the initial pretension
of mooring lines.
4.6. Effect of safety factor on system reliability
The effect of the safety factor of mooring lines on the fatigue reli-
ability of the mooring system is analyzed. Fig. 13 shows the fatigue
reliability of mooring system for 25-year service life as a function
of the safety factor of mooring lines. The results indicate that the
logarithmic value of fatigue failure probability almost decreases lin-
early with increasing safety factor. It is convenient to achieve the
reliability-based design of mooring system for net cage after obtain-
ing the relationship between the safety factor of mooring lines and
the fatigue failure probability of mooring system. The safety factor
of mooring lines can be obtained when a target reliability index
of mooring system is required. Table 5 shows the safety factor of
mooring lines for the required target reliability index (the dotted
line in Fig. 13) of mooring system.
In DNV OS-E301 [44], the fatigue safety factor of mooring lines
which is not regularly inspected ashore is more than 5.0, while
this value is 3.0 for the regularly inspected mooring lines. Based
on our results in Fig. 13, the fatigue failure probability of mooring
system is 5 × 10−6 for safety factor 3.0. This indicates that the failure
probability of mooring system against fatigue damage is less than
10−4 significantly when the fatigue safety factor of mooring lines
follows the DNV OS-E301 [44], which means that the design for
mooring system is conservative.
88 H.-M. Hou et al. / Applied Ocean Research 71 (2018) 77–89
5. Conclusions
The time dependent fatigue reliability of mooring system for fish
cage is analyzed in this study. The sensitivity analysis demonstrates
that the fatigue reliability of mooring line for net cage system is
more sensitive to the uncertainty of the parameters (A, B) in the
stress range distribution than that of other parameters. In addition,
the fatigue reliability bounds of mooring system are conservative
for independent relationship for the failure of components of moor-
ing system. A case without initial pretension on anchor lines is
beneficial for improving the fatigue reliability of mooring system.
The safety factor of mooring lines must be carefully assigned since
the failure probability of mooring system is sensitive to the vari-
ation of the safety factor. Furthermore, the failure probability of
mooring system against the fatigue damage is acceptable for the
safety factor in DNV OS-E301 [44].
Through the case study, a zero initial pretension may be effec-
tive to improve the fatigue reliability of mooring system for similar
fish farms arrangements in similar environmental conditions. In
order to ensure enough redundancy of the system reliability against
fatigue damage, the other practical design considerations also need
to be considered.
Acknowledgements
This work was financially supported by the National Natural
Science Foundation (NSFC) Projects No. 51239002, 51409037,
51579037, 51609035 and 51221961, China Postdoctoral Sci-
ence Foundation (No.2014M560211, No.2015T80254 and
2016M590224), the Fundamental Research Funds for the Central
Universities No. DUT16RC (4)25 and Cultivation plan for young
agriculture science and technology innovation talents of Liaoning
province (No.2014008).
References
[1] D.W. Fredriksson, Open Ocean Fish Cage and Mooring System Dynamics,
University of New Hampshire, 2001 (Doctoral Thesis).
[2] H.M. Hou, T.J. Xu, G.H. Dong, Y.P. Zhao, C.W. Bi, Time-dependent reliability
analysis of mooring lines for fish cage under corrosion effect, Aquacult. Eng.
77 (2017) 42–52.
[3] P.F. Lader, A. Fredheim, Dynamic properties of a flexible net sheet in waves
and current—a numerical approach, Aquacult. Eng. 35 (3) (2006) 228–238.
[4] C. Balash, B. Colbourne, N. Bose, W. Raman-Nair, Aquaculture net drag force
and added mass, Aquacult. Eng. 41 (2009) 14–21.
[5] O. Patursson, M.R. Swift, I. Tsukrov, K. Simonsen, K. Baldwin, D.W.
Fredriksson, et al., Development of a porous media model with application to
flow through and around a net panel, Ocean Eng. 37 (2-3) (2010) 314–324.
[6] E. Bouhoubeiny, P. Druault, G. Germain, Phase-averaged mean properties of
turbulent flow developing around a fluttering sheet of net, Ocean Eng. 82
(2014) 160–168.
[7] C. Zhou, L.X. Xu, F.X. Hu, X.Y. Qu, Hydrodynamic characteristics of knotless
nylon netting normal to free stream and effect of inclination, Ocean Eng. 110
(2015) 89–97.
[8] D. Kristiansen, Wave Induced Effects on Floaters of Aquaculture Plants,
Norwegian University of Science and Technology, 2010 (Doctoral Thesis).
[9] S.X. Fu, T. Moan, Dynamic analyses of floating fish cage collars in waves,
Aquacult. Eng. 47 (2011) 7–15.
[10] X.H. Huang, G.X. Guo, Q.Y. Tao, Y. Hu, H.Y. Liu, S.M. Wang, et al., Numerical
simulation of deformations and forces of a floating fish cage collar in waves,
Aquacult. Eng. 74 (2016) 111–119.
[11] C.W. Lee, Y.B. Kim, G.H. Lee, M.Y. Choe, M.K. Lee, K.Y. Koo, Dynamic simulation
of a fish cage system subjected to currents and waves, Ocean Eng. 35 (14–15)
(2008) 1521–1532.
[12] H. Moe, A. Fredheim, O.S. Hopperstad, Structural analysis of aquaculture net
cages in current, J. Fluids Struct. 26 (3) (2010) 503–516.
[13] T.J. Xu, Y.P. Zhao, G.H. Dong, Y.C. Li, F.K. Gui, Analysis of hydrodynamic
behaviors of multiple net cages in combined wave-current flow, J. Fluids
Struct. 39 (2013) 222–236.
[14] L. Li, S.X. Fu, Y.W. Xu, Nonlinear hydroelastic analysis of an aquaculture fish
cage in irregular waves, Mar. struct. 34 (2013) 56–73.
[15] I.H. Grue, Loads on the Gravity-net-cage from Waves and Currents,
Norwegian University of Science and Technology, 2014 (Master’s Thesis).
[16] T. Kim, J. Lee, D.W. Fredriksson, J. DeCew, A. Drach, K. Moon, Engineering
analysis of a submersible abalone aquaculture cage system for deployment in
exposed marine environments, Aquacult. Eng. 63 (2014) 72–88.
[17] S. Ito, T. Kinoshia, W.G. Bao, Hydrodynamic behaviors of an elastic net
strcture, Ocean Eng. 92 (2014) 188–197.
[18] Y. Yao, Y. Chen, H. Zhou, H. Yang, Numerical modeling of current loads on a
net cage considering fluid-structure interaction, J. Fluids Struct. 62 (2016)
350–366.
[19] H. Winthereig-Rasmussen, K. Simonsen, O. Patursson, Flow through fish
farming sea cages: comparing computational fluid dynamics simulations with
scaled and full-scaled experimental data, Ocean Eng. 124 (2016) 21–31.
[20] J.G. Williams, A. Miyase, D.H. Li, S.S. Wang, Small-scale testing of damaged
synthetic fiber mooring ropes, in: Proceedings of the Offshore Technology
Conference, Houston, Texas, USA, OTC 14308, 2002, pp. 2717–2729.
[21] M.K. Olsen, Estimation of Annual Probability of Mooring Line Failure as a
Function of Safety Factor, Norwegian University of Science and Technology,
2011 (Master’s Thesis).
[22] P. Thies, L. Johanning, G. Smith, Lifecycle fatigue load spectrum estimation for
mooring lines of a floating marine energy converter, Offshore and Artic
Engineering, Rio de Janeiro, Brazil, in: Proceedings of ASME 31 st International
Conference on Ocean, 2, 2012, pp. 667–676.
[23] P. Thies, L. Johanning, V. Harnois, H. Smith, D. Parish, Mooring line fatigue
evaluation for floating marine energy converters: field measurements and
prediction, Renew. Energ. 63 (2014) 133–144.
[24] P. Davies, M. Francois, N. Lacotte, T.D. Vu, D. Durville, An empirical model to
predict the lifetime of braided HMPE handling ropes under cyclic bend over
sheave (CBOS) loading, Ocean Eng. 97 (2015) 74–81.
[25] H. Liu, Y. Lian, L. Li, Y. Zhang, Experimental investigation on dynamic stiffness
of damaged synthetic fiber ropes for deepwater moorings, J. Offshore Mech.
Arct. Eng. 137 (6) (2015) 061401.
[26] K. Larsen, J. Mathisen, Reliability-based fatigue analysis of mooring lines,
Florence, Italy, in: Proceedings of the International Conference on Offshore
Mechanics and Arctic Engineering, 2, 1996, pp. 277–285.
[27] N.A. Siddiqui, S. Ahmad, Fatigue and fracture reliability of TLP tethers under
random loading, Mar. struct. 14 (2001) 331–352.
[28] Z. Gao, T. Moan, Fatigue damage induced by non-Gaussian bimodal wave
loading in mooring lines, Appl. Ocean Res. 29 (2007) 45–54.
[29] J.S. Han, Y.H. Kim, Y.J. Son, H.S. Choi, A comparative study on the fatigue life of
mooring systems with different composition, J. Hydrodyn. 22 (5) (2010)
452–456.
[30] H.S. Toft, J.D. Sørensen, Reliability-based design of wind turbine blades,
Struct. Saf. 33 (2011) 333–342.
[31] Y.M. Low, Extending a time/frequency domain hybrid method for riser fatigue
analysis, Appl. Ocean Res. 33 (2011) 79–87.
[32] M. Jameel, S. Ahmad, Fatigue reliability assessment of coupled spar-mooring
system, Rotterdam, The Netherlands, in: Proceedings of the International
Conference on Offshore Mechanics and Arctic Engineering, 2, 2011, pp.
497–505.
[33] W. Dong, Y. Xing, T. Moan, Z. Gao, Time domain-based gear contact fatigue
analysis of a wind turbine drivetrain under dynamic conditions, Int. J. Fatigue
48 (2013) 133–146.
[34] H. Li, J. Du, S. Wang, M. Sun, A. Chang, Investigation on the probabilistic
distribution of mooring line tension for fatigue damage assessment, Ocean
Eng. 124 (2016) 204–214.
[35] T.J. Xu, G.H. Dong, Y.P. Zhao, Y.C. Li, F.K. Gui, Analysis of hydrodynamic
behaviors of gravity net cage in irregular waves, Ocean Eng. 38 (2011)
1545–1554.
[36] T.J. Xu, Hydrodynamics of Net Cage Group in the Open Sea, Dalian University
of Technology, 2014 (Doctoral Thesis).
[37] K. Ma, A. Duggal, P. Smedley, D. L’Hostis, H. Shu, A historical review on
integrity issues of permanent mooring systems, in: Proceedings of the
Offshore Technology Conference, Houston, USA, OTC 24025, 2013.
[38] S.D. Weller, L. Johanning, P. Davies, S.J. Banfield, Synthetic mooring ropes for
marine renewable energy applications, Renew. Energ. 83 (2015) 1268–1278.
[39] Z. Wang, S. Dong, X. Li, C. Guedes Soares, Assessments of wave energy in bohai
sea, China, Renew. Energ. 90 (2016) 145–156.
[40] Y.C. Li, F.K. Gui, B. Teng, Hydrodynamic behavior of a straight floating pipe
under wave conditions, Ocean Eng. 34 (3–4) (2007) 552–559.
[41] F. Théret, Etude De l’équilibre De Surfaces Reticules Places Dansun Courant
Uniforme (application Aux Chalets), Université de Nantes, Ecole Centrale de
Nantes, Franc, 1993 (PhD Thesis).
[42] Y.I. Choo, M.J. Casarella, Hydrodynamic resistance of towed cables, J. Hydron.
(1971) 126–131.
[43] API RP 2SK, Recommended Practice for Design and Analysis of Station
Keeping Systems for Floating Structures, 2005.
[44] DNV OS-E301, Offshore Standard DNV-OS-E301 Position Mooring, Det norske
Veritas (DNV), Høvik, Norway, 2001.
[45] NS 9415 Marine Fish Farms: Requirements for Site Survey, Risk Analyses,
Design, Dimensioning, Production, Installation and Operation, Standards
Norway, 2009.
[46] A. Almar-Næss, Fatigue Handbook, Tapir Publishers, Trondheim, 1985.
[47] A.J. Kappos, Wave loading, in: T. Moan (Ed.), Dynamic Loading and Design of
Structures, Taylor & Francis, London, 2001, pp. 175–230.
[48] C. Guedes Soares, T. Moan, Model uncertainty in the long term distribution of
wave induced bending moments for fatigue design of ship structures, Mar.
struct. 4 (1991) 295–315.
 H.-M. Hou et al. / Applied Ocean Research 71 (2018) 77–89
[49] D.V. Lindley, Introduction to Probability and Statistics from a Bayesian
Viewpoint, Cambridge University Press, Cambridge, 1976.
[50] W. Dong, T. Moan, Z. Gao, Fatigue reliability analysis of the jacket support
structure for offshore wind turbine considering the effect of corrosion and
inspection, Reliab. Eng. Syst. Saf. 106 (2012) 11–27.
[51] S. Márquez-Domínguez, J.D. Sørensen, Fatigue reliability and calibration of
fatigue design factors for offshore wind turbines, Energies 5 (6) (2012)
1816–1834.
[52] M.E. Mousavi, Z. Reza, S. Upadhye, V. Vijayaraghavan, K. Haverty, A simplified
method for quantitative reliability and integrity analysis of steel catenary
risers, J. Offshore Mech. Eng. 138 (1) (2016) 011601.
[53] C. Guedes Soares, Y. Garbatov, Reliability of maintained ship hulls subjected
to corrosion and fatigue under combined loading, J. Constr. Steel Res. 52 (1)
(1999) 93–115.
[54] W. Huang, Y. Garbatov, C. Guedes Soares, Fatigue reliability of a web frame
subjected to random non-uniform corrosion wastage, Struct. Saf. 48 (2014)
51–62.
[55] O. Ditlevsen, Narrow reliability bounds for structural systems, J. Struct. Mech.
7 (4) (1979) 453–472.
[56] Y. Garbatov, C. Guedes Soares, Fatigue reliability assessment of welded joints
of very fast ferry accounting for vehicle loads, Int. J. Marit. Eng. 153 (A4)
(2011) 231–241.
89
[57] T. Moan, E. Ayala-Uraga, Reliability-based assessment of deteriorating ship
structures operating in multiple sea loading climate, Reliab. Eng. Syst. Saf. 93
(2008) 433–446.
[58] Det Norske Veritas, Fatigue Assessment of Ship Structures—DNV Classification
Notes No 30.7, Det Norske Veritas, Oslo, 2001.
[59] C. Guedes Soares, L.D. Ivanov, Time-dependent reliability of the primary ship
structure, Reliab. Eng. Syst. Saf. 26 (1989) 59–71.
[60] N.J. Tarp-Johansen, Examples of Fatigue Lifetime and Reliability Evaluation of
Larger Wind Turbine Components, Risø National Laboratory, Report No.:
Risø-R-1418, Roskilde, Denmark, 2003.
[61] A.R. Nejad, Z. Gao, T. Moan, On long-term fatigue damage and reliability
analysis of gears under wind loads in offshore wind turbine drivetrains, Int. J.
Fatigue 61 (2014) 116–128.
[62] H.O. Madsen, S. Krenk, N.C. Lind, Methods of structural safety, Prentice Hall,
New Jersey, 1986.
[63] R.E. Melchers, Structural Reliability Analysis and Prediction, 2nd edition, John
Wiley, New York, 1999.
[64] J. Ferreras, E. Peña, A. López, F. López, Structural performance of a floating
breakwater for different mooring line typologies, J. Waterw. Port Coastal
Ocean Eng. 140 (3) (2014) 158–163.