Ocean Engineering 111 (2016) 56–66

Contents lists available at ScienceDirect

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

Stochastic extreme motion analysis of jack-up responses during

wet towing
Won-Hee Kang a, Chunwei Zhang a,b,n, Jian-Xing Yu c
a
Institute for Infrastructure Engineering, Western Sydney University, Penrith, NSW 2751, Australia
b
School of Civil Engineering, Qingdao University of Technology, Qingdao 266033, China
c
State Key Laboratory of Hydraulic Engineering Simulation and Safety, Tianjin University, Tianjin 300072, China

art ic l e i nf o a b s t r a c t

Article history: This paper aims to perform stochastic motion analysis and the system probability estimation of excee-
Received 5 August 2014 dance in motion of a jack-up unit in wet tow mode to assess its safety in terms of its translational and
Received in revised form rotational motions. When a jack-up unit is afloat and towed in seaways, it is important to determine the
28 October 2015
extreme motion response for the stability of the platform and to ensure the safety of equipment on the
Accepted 29 October 2015
deck. This paper discusses the hydrodynamic characteristics and motion behavior of a jack-up unit with
mat at a certain towing speed in various sea states with significant wave height ranging from 2 m to 5 m
Keywords: and wave directions ranging from 0° to 90°. The significant motion responses, such as heave, roll and
Jack-up pitch, are estimated by the frequency domain method. The probability of exceedance of the motion
Wet towing
responses of the jack-up in random seas is estimated by means of system reliability analysis, considering
Extreme motion response
wave overtopping, instability for roll, instability for pitch, and their series system. In order to efficiently
Subset simulation
System reliability determine the limit state exceedance probabilities due to time varying random variables, the Monte Carlo
simulation technique and the subset simulation method have been employed in this study. The results
from deterministic analysis and probabilistic analysis are comparably discussed, and the estimation of
the extreme value statistics of motion responses in various sea states is provided. The approach devel-
oped in this paper can be applied to predict an unsafe condition of the jack-up and other platforms
during wet towing in seaways supporting the officer's decisions on the action taken if excessive
responses are foreseen for the present course and speed.
& 2015 Published by Elsevier Ltd.

1. Introduction
A jack-up unit is an offshore structure composed of a hull, legs,
footings, and self-elevating systems. The unit allows the platform
to be towed to a site, to lower its legs into the seabed, and to lift its
hull to support a stable work deck capable of withstanding the
environmental loads. Generally, a jack-up unit operates in the
three main modes: (i) the transit mode from one location to
another, (ii) the elevation mode with its legs, and (iii) the jacking
up or down mode between (i) and (ii). There are two main footing
types: mat footings and spud cans (Zhao and Sun, 2014). Generally,
mat footings exert a lower bearing pressure on the soil and pro-
vide considerable buoyancy than spud cans in the afloat transit
mode (Vazquez et al., 2005). Based on this provided buoyancy, a
jack-up can be transported from one location to another, taking
n
Corresponding author. Tel.: þ 61 2 4736 0182; fax: þ61 2 4736 0054.
E-mail address: chunwei.zhang@uws.edu.au (C. Zhang).
the wet tow approach, in which the jack-up floats on its own hull
and mat with its legs raised.
Many researchers have studied the dynamic response of a jack-
up with footings on/into the seabed, and its structural risks during
each stage of a jack-up deployment. Wang et al. Wang et al. (2014)
applied sub-structural identification and genetic algorithms for
the structural health monitoring of a jack-up unit in time and
frequency domains. Vlahos et al. (2008) presented experimental
investigation of the system behavior of a three-legged jack-up
model on clay. Bienen and Cassidy (2006) advanced a three-
dimensional fluid–structure–soil interaction analysis of offshore
jack-up structures. Leira and Karunakaran (1995) analytically
estimated fatigue damage and the extreme response exceedance
probability of a jack-up platform with respect to relative operation
time at two sites different in terms of wave environment and
water depth. Jensen and Capul (2006) predicted the extreme
response for jack-up units in the second order stochastic waves
based on the theory of random vibrations and the first order
reliability method (FORM). Jensen (2011) further investigated the

http://dx.doi.org/10.1016/j.oceaneng.2015.10.046
0029-8018/& 2015 Published by Elsevier Ltd.
 W.-H. Kang et al. / Ocean Engineering 111 (2016) 56–66 57

reliability index estimation for overturning of a jack-up rig and a
nonlinear wave induced hogging bending moment in a ship based
on a FORM approach in combination with proper scaling function
obtained from a Monte Carlo simulation analysis.
The management of risk for an in-situ jack-up is important,
such as the drilling risks, the in-transit, jacking, and pre-loading
phases of a jack-up under operation (Hunt and Marsh, 2004). In
particular, a moving jack-up over long distances appears to be
dangerous because it often encounters substantial risks, including
a loss of towline or headway, green water damage, flooding, and
capsize. World Offshore Accident Databank, commonly known as
WOAD (Det Norske Veritas, 2007), even records that many jack-
ups have been lost while afloat. Many jack-ups have sunk while
they were under tow in the North Sea when a single vessel was
used for tow. The well-known examples were Rowan Gorilla I
provided in this paper are useful to ensure the safety margins
included in the design of the application jack-up unit under tow
operation.
2. Theoretical background
A floating body in waves is normally assumed rigid and has six
degrees of freedom motions named surge, sway, heave, roll, pitch,
and yaw. In order to determine the floating body's responses, the
following equations of coupled motions in various degrees of
freedom are required to be solved.
X
6   
M jk þ Ajk x€ k ðtÞ þ Bjk x_ k ðtÞ þ C jk xk ðtÞ ¼ F j e  iωt ðj ¼ 1; ⋯; 6; k ¼ 1; ⋯; 6Þ
k¼1

located off the east coast of Canada when being towed by M/V Smit ð1Þ
London, and West Gamma when being towed by Normand Drott
(Gibson, 2000). As can be learnt from these examples, the roll and where xk is the kth motion of the six-degree of freedom motions;
pitch motions of the platform can lead to various types of failures, M jk are the components of the generalized mass matrix for the
structure; Ajk and Bjk are added mass and damping coefficients; C jk
ranging from seasickness and equipment shift to capsize of the
are the restoring force coefficients; and F j are the complex
platform. Also, green water will occur with a large heave, roll and
amplitudes of the exciting forces and moment-components with
pitch in uncalmed seas, causing severe structural and equipment
the force and moment-components given by the real part of F j
damages. A tow analysis should be conducted to verify if a hull has
e  iωt (i is the complex unit). If the floating body has its forward
sufficient buoyancy and stability to withstand the target motions,
speed, the wave frequency ω in the above equation is replaced by
which can typically be defined as the probability of deck wetness,
ωe , which is the circular frequency of encounter obtained by:
or a maximum pitch/roll angle at a given oscillating period, acting
in conjunction with heave accelerations. As such, many studies ωe ¼ κ ðc  V cos μÞ ð2Þ
regarding the development of stochastic analysis techniques for where κ is the wave number, c is the wave speed of propagation, V
offshore platforms including a jack-up platform under transit is the forward speed of floating body, and μ is the wave direction
condition have already been carried out. However, they have not with an angle relative to the jack-up's speed direction.
been essentially extended to a reliability analysis of a specific jack- These hydrodynamic coefficients can be obtained by solving a
up model during a wet towing condition in practice considering boundary value problem for a velocity potential, which is usually
the uncertainty in the stochastic wave input process and the joint- called the potential flow theory (Newman, 1977). Since the motion
first passage probability of multiple limit states. In addition, few is assumed harmonic, the solution can be performed both in the
studies exist in the literature regarding the dynamic motion
frequency domain and in the time domain. This study mainly
response of a jack-up unit with mat foundation in transit by the
employs the frequency domain analysis for the deterministic
wet tow approach.
response estimations in Section 3, which has proven to be a sui-
This paper aims to present the hydrodynamic performance of a
table approach to provide reasonable predictions for analysis at
jack-up unit in wet tow mode in terms of the events of wave
preliminary stages of ship or floating platform motion control
overtopping and instability for roll and pitch and their system
system design (Perez, 2005).
limit state exceedance condition by estimating the unit's respon-
Short-term stationary irregular sea states can be expressed by a
ses and limit state exceedance probability. The analysis is targeting
wave spectrum Sζ ðωÞ, which is the power spectral density function
at evaluating the joint first passage probability of the jack-up unit's
of the vertical sea surface displacement. Through spectrum ana-
responses with respect to four limit states for these four kinds of
lysis, we can set up the mathematical models of the structural
events against a stochastic wave input process. This study con-
responses in a seaway, normally based on irregular and random
siders extreme-motion related limit states that can be violated
waves, by using a superposition of regular wave components
more likely in a system-level reliability estimation. This study
(Journée and Massie, 2001). A block diagram of this principle is
limits the scope from the consideration of the structural failure,
shown in Fig. 1.
secondary or long term effects such as wind-load induced fatigue
Many investigators have attempted to describe a wave fre-
failure although they are important issues in a jack-up unit design
quency spectrum in terms of measured spectra or parameterized
(Shabakhty, 2011). Also, they should definitely be considered for a
severe sea states mainly for permanent platforms, which will
Input (t ) , waves Output R(t ) , motions
affect the structural integrity. Furthermore, stability issues
regarding multiple compartment flooding is also an important
issue especially for naval ships rather than commercial ships Ships &
Floating structures
against a much higher load from extreme flooding events. They
can be considered as additional limit states within the proposed
t (time) t (time)
framework if a proper analysis algorithm is available, but at the
same time they also need to be analyzed based on the number of Frequency
cycles to failure and have different nature and considered time Wave spectrum response Motion spectrum
range from the extreme response analyses. function
S( ) SR ( )
The proposed analysis framework will be applied to a specific
jack-up unit to be operating in Bohai Bay in China. These analysis
results are denoted as decision support tools for the application
jack-up unit, which are represented by dangerous sea states in
terms of significant wave heights and wave headings. The results Fig. 1. Relation between sea waves and ship motions.
58 W.-H. Kang et al. / Ocean Engineering 111 (2016) 56–66

analytic formulas in a table form (Det Norske Veritas, 2007). The

mathematical formulations of these normalized uni-directional
wave energy spectra are based on the two environmental para-
meters: the significant wave height H s and the average wave
periods T in terms of the mean zero-up-crossing wave period T z or Topside
the spectral peak period T p . The significant wave height H s is
approximately equal to the average height (trough to crest) of the
highest one-third waves in the specified time period. The zero-up-
crossing period T z is the average time interval between two suc- Hull (42m×30m×5m)

cessive up-crossings of the mean sea level. The spectral peak

Leg (65m×Ø2m)
period T p is the wave period determined by the inverse of the
Waterline
frequency at which a wave energy spectrum has its maximum
value. In this paper, the Pierson-Moskowitz (PM) spectrum is
applied for wind seas, which often occur for the most severe sea
states.
  !
5 ω 4
Seabed Mat (42m×38m×3.5m)
5
SPM ðωÞ ¼ UH 2s ω4p U ω  5 exp  ð3Þ
16 4 ωp
Fig. 2. Jack-up in on-bottom operation.

where ωp ¼ 2π =T p is the angular spectral peak frequency.

To discuss the stochastic dynamic transfer function further, the
following information on theoretical basis, equations, and used Total Weight=6750t
software for RAO evaluation is provided. Assuming a steady state,
Leg (65m×Ø2m)
with all transient effects neglected, the motion response of the
floating structure in regular waves is oscillating harmonically with
Crane
the same frequency as the incident waves, or with the frequency of
Helicopter Deck
encounter in the case of a forward speed. Therefore, the response
in irregular sea can be described as:

X
N   Topside
RðtÞ ¼ ζ an Hðωn Þ sin ðωn t þ δðωn Þ þ εn Þ ð4Þ

3.96m
n¼1
  Waterline Hull (42m×30m×5m)
where Hðωn Þ is called the transfer function, also denoted the

5.54m
Response Amplitude Operator (RAO), which is the response Mat (42m×38m×3.5m)
1m

amplitude per unit wave amplitude. δðωn Þ is a phase angle asso-
ciated with the response. These parameters are functions of the
frequency of oscillation ωn . ζ an and εn are the phase and the wave
amplitude of the nth regular wave, respectively.
Here, the key of the motion response is the solution of the
transfer function. The RAO HðωÞ is defined by the following for-
mulation:
F j ð ωÞ
HðωÞ ¼
 ω2 ½M jk þ Ajk ðωÞþ iωBjk ðωÞ þ K jk
In this study, to evaluate RAO, a frequency domain analysis
software package AQWA-FER (Ansys, 2010) is employed. It takes
linearized hydrodynamic fluid wave loading for a floating or fixed
rigid bodies and static equilibrium configuration as part of inputs
to estimate RAO of an offshore platform.
3. Case study
3.1. Application description
The application jack-up platform in this study is designed to be
operating in Bohai Bay in China, and its average water depth is
40 m. The platform will be transported to the offshore oil field by
wet tow with a forward speed of 4 knots. The on-bottom operation
condition and components of the jack-up are illustrated in Fig. 2.
The footing type in this study is the mat, and the towing condition
is illustrated in Fig. 3. The geometry and weight parameters of the
jack-up are also illustrated in Figs. 2 and 3. While towing, the
vertical center of gravity (VCG) is located at 7.24 m height from the
base-line of the mat.
Baseline
Fig. 3. Jack-up during towage.
3.2. Numerical model
The coordinate systems for the motions of the jack-up platform
are shown in Fig. 4, in which OXYZ is a fixed coordinate system
with origin on the still water surface, OXYZ is a (moving) coordi-
ð5Þ nate system with the origin at the center of gravity of the jack-up
(the circular black point in Fig. 4) while the directions for the x, y,
and z axis are the same as those in the OXYZ axis. In this model, a
total of 2192 diffracting elements are meshed in this model.
3.3. Response Amplitude Operators (RAO)
The motion of the jack-up towing in waves is a result of the
wave excitation due to the varying distribution of pressure on the
hull. The wave excitation and the motion response will depend not
only on the characteristics of the waves, including wave amplitude
and wave frequency, but also on the towing conditions, such as
encounter angle and towing speed. Response Amplitude Operator
(RAO) is usually used to represent the transfer functions of vari-
ables, for instance, exciting forces/moments and platform motions
per unit wave amplitude. Fig. 5 shows the frequency character-
istics of the coupled heave and pitch motions of the jack-up in the
three different wave headings: a following wave (μ ¼ 0o ), an
oblique wave (μ ¼ 45o ), and a beam wave (μ ¼ 90o ).
In very long waves, the heave amplitude of the platform
becomes equal to the wave amplitude, and the pitch amplitude
becomes equal to the wave slope amplitude; both dimensionless
transfer functions become to 1. At frequencies higher than the
natural frequencies of heave and pitch, the transfer function values
 W.-H. Kang et al. / Ocean Engineering 111 (2016) 56–66 59

Fig. 4. Definitions of coordinate systems for the motions of the jack-up platform.

2.0 14

12
RAO of heave motion (-)

1.5

RAO of pitch motion (-)

10

8
1.0
6

4
0.5

0.0 0
0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0

Wave frequency (rad/s) Wave frequency (rad/s)

Fig. 5. RAO heave and pitch of the jack-up with the speed of 4 knots.

decrease, and at even higher frequencies with the wave lengths
shorter than the platform length or width, the transfer functions
become zero. Resonance, which is usually marked by a local
maximum in the RAO curves, occurs in the neighborhood of the
natural frequencies, and the values are 0.78 rad/s for heave motion
and 0.55 rad/s for pitch motion, respectively. It is also observed
that the resonance does not always appear at the natural fre-
quency due to the coupling effects, and RAO values are generally
larger than 1.0 at the wave length of approximately the platform
length, when a platform has a forward moving speed.
3.4. Significant motions of the jack-up
In this section, we focus on the statistically significant motions
of the jack-up unit in a natural seaway, which is assumed to be
stationary with known spectrum Sζ ðωÞ as mentioned in Section 2.
The random response spectrum SR ðωÞ can be computed directly
from RAO and the wave spectrum as follows (Chakrabarti, 1987):
 2
SR ðωÞ ¼ Sζ ðωÞRAOðωÞ ð6Þ
The significant motion amplitudes of the jack-up for various sea
states are shown in Fig. 6.
These figures show that all motions, including heave, roll and
pitch, increase with the significant wave height, and the motions
of the jack-up at a certain forward speed are significantly influ-
enced by the wave direction. It is certain that the pitch motion
becomes the largest in the following waves (μ ¼ 0o ) than that in
other sea conditions. The roll motion shows generally greater
value than the pitch motion in an oblique wave, because it has a
greater frequency of encounter than the other motions depending
on the wave frequency, wave direction, and platform forward
speed. In a level 5 sea state, where Hs ¼ 2.5 m and smoother sea
states, the rotational motions in different wave angles are com-
monly negligible, and the vertical motion of the platform is mostly
dependent upon the heave amplitude. In rough seas (Hs 4 2.5 m),
the rotational motions are significantly affected by the wave
headings, and the gaps between the curves are increasing with the
significant wave height. This makes the vertical motion of the
platform more dependent on the pitch and roll amplitudes.

Similar to the wave motions, the variance of the response can

be estimated:
Z 1 4. Probability analysis of exceedance of structural responses
 2
σ 2R ¼ Sζ ðωÞRAOðωÞ dω ð7Þ
0 In the previous section, the resonance frequency that should be
The short-term response statistics can be obtained based on avoided in the actual sea state was estimated from the hydro-
the response spectrum. For example, the significant motion dynamic Response Amplitude Operators and the significant
amplitude is defined as the mean value of the highest one-third of motions of the jack-up unit. Likewise, the trends of heave, roll, and
the amplitudes, given by: pitch motions with the level of the sea state were seen. Yet, a
specific criteria still needs to be established to define the critic-
rs ¼ 2 σ R ð8Þ ality of the jack-up in various sea states considering possible
60 W.-H. Kang et al. / Ocean Engineering 111 (2016) 56–66

Significant values of pitch motion (degree)

50

Significant values of heave motion (m)

2.5
45

40
2.0
35

30
1.5
25

1.0 20

15

0.5 10

5
0.0 0
2.0 2.5 3.0 3.5 4.0 4.5 5.0 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Hs (m) Hs (m)
Significant values of roll motion (degree)

50

40

30

20

10

2.0 2.5 3.0 3.5 4.0 4.5 5.0

Hs (m)

Fig. 6. Short-term significant motion amplitudes of the jack-up in various sea states.

uncertainties in wave motions. This section will implement the interval, S(  )¼wave spectrum, κ i ¼ wave numbers, and
probability analysis of exceedance of motion responses. This study
σ 2i ¼ Sðωi ÞΔωi ð10Þ
limits the scope to the consideration of uncertainties in random
sea waves including the uncertainties both in wave amplitude and In this wave motion model, ui ; ui are independent standard
phase only (Jensen and Capul, 2006; Dostal et al., 2013). For a more normal random variables introducing randomness to the wave
realistic modeling, all the parameter uncertainties and modeling amplitude Ai and the phase εi in Eq. (9) according to the following
uncertainties related to the model needs to be considered usually relation:
considered in the design stage. The additional consideration of qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
those parameters can be easily implemented in the outer loop or Ai ¼ σ i ui 2 þ ui 2
an extra dimension of the Monte Carlo simulation analysis used in
ui
this study, but it can create significant computational demand tan ðεi Þ ¼ ð11Þ
ui
requiring repeated updating of RAO. It can be expected that the
addition of those uncertainties can slightly reduce the overall where Ai and εi follow Rayleigh and uniform distributions,
reliability but their contribution will be less significant than the respectively.
stochastic characteristic of the input wave. In this study, all the other parameter uncertainties and mod-
eling uncertainties related to the jack-up unit model are not con-
sidered for simplicity.
4.1. Stochastic motion models of sea waves and structural responses

In this study, the assumption for relatively moderate sea states 4.2. The definitions of limit states
is made such that the excitation of wave input is a stationary
4.2.1. Limit state formulation
stochastic process and the wave elevation is normally distributed
In the reliability analysis of a time-varying system response, a
(Jensen, 2007). The normally distributed wave elevation can be
limit state function is often defined by the first exceedance of a
expressed as a function of space X and time t as follows:
random response within a specified time window. If a threshold
X
n level for defining an exceedance event is given as b, the corre-
HðX; tÞ ¼ Ai cos ðωi t  κ i X þ εi Þ sponding failure event is defined as follows:
i¼1



    
X
n
F ¼ [ nk ¼ 1 Xðt k Þ 4 b ¼ max Xðt k Þ 4 b ð12Þ
¼ ui σ i cos ðωi t  κ i XÞ ui σ i sin ðωi t  κ i XÞ ð9Þ k ¼ 1;…;n
i¼1
where k¼ 1,…,n denotes n discrete time points within the specified
where ωi ¼ discrete wave frequencies, Δω ¼wave frequency time window, and X(tk) is a time-varying random variable. The
 W.-H. Kang et al. / Ocean Engineering 111 (2016) 56–66
union event represents the exceedance event of the random
variable X(tk) at least once within the specified time window.
4.2.2. Limit state of wave overtopping
A wave overtopping event on the deck of an offshore platform
can cause severe structural and equipment damages due to the
considerable wave force and can increase safety risk (Zhou et al.,
 1X N
1999; Bea et al., 1999). For a more reliable design and prompt
decision support during the operation of an offshore platform, it is
essential to make a stochastic prediction of a wave-overtopping
event. A limit state function for a wave-overtopping event is
defined in terms of an instantaneous air gap, a(x, y, t), which is
defined as follows:
aðx; y; tÞ ¼ a0 þzðx; y; tÞ  ηðx; y; tÞ
61
4.3. Monte Carlo simulation (MCS)
4.3.1. MCS formulation for a failure probability estimation
A failure probability P(F) for a given function g(X) having
negative values for a failure state can simply be estimated using
MCS as follows:
PðFÞ ¼ P gðXÞ o 0  I ðgðxi Þ o 0Þ ð16Þ
Ni¼1
where N ¼the number of samples used in MCS, xi ¼the ith set of
samples of random variables X, and I¼ an indicator function giving
binary value according to the limit state exceedance to a failure
region. For a first passage limit state function with time-varying
ð13Þ random variables X(t), the MCS formulation can be made as

 
where a0 ¼ the still-water air gap, which is the difference in ele- 1X N
PðFÞ ¼ min gðXðtÞÞ o0  I min gðxi ðtÞÞ o 0 ð17Þ
vation between the bottom of the deck or other equivalent parts, t ¼ 0;:::;T N i ¼ 1 t ¼ 0;:::;T
and the mean water level, (x, y)¼a given horizontal location, z(x, y,
t) ¼the vertical displacement of the platform at the given hor- where T ¼the end time point within a discrete time range con-
izontal location, and ηðx; y; tÞ is the instantaneous wave surface sidered. This MCS formulation is highly convenient in terms of
elevation at the same location. The vertical displacement of the implementation as it does not need complex identification of a
platform z(x, y, t) can be calculated based on the assumption that failure event.
the platform make a rigid body motion as follows:
  4.3.2. Subset simulation formulation for a failure probability
 
zðx; y; tÞ ¼ ξh ðtÞ  x sin ξp ðtÞ þ y sin ξr ðtÞ ð14Þ estimation
However, the convergence rate of MCS is well known to criti-
where ξh ðtÞ ¼heave translational motion, ξr ðtÞ ¼roll rotational cally be dependent on the number of samples, and it is difficult for
motion, and ξp ðtÞ ¼pitch rotational motion. The failure event from MCS to achieve a sufficient convergence using a limited number of
this instantaneous air gap is defined as the air gap becoming a samples, especially for very low probability estimation. This study
negative value, i.e., a(x, y, t) o0, meaning that the elevation of a adaptively employs the subset simulation method developed by
wave exceeds that of a deck or an equivalent part, and an over- Au and Beck (2001), which achieves much faster convergence
topping occurs. particularly for low probability estimation. The basic idea behind
the subset simulation is that for given failure events F0⊃F1⊃…⊃Fk
4.2.3. Limit states of stability based on roll and pitch motions …⊃Fm ¼F, the failure probability P(F) can be represented by the
Two more limit states are also defined in this study in terms of product of the conditional probabilities of the intermediated fail-
the instability of the platform according to the roll and pitch ure events as follows:
rotational motions. Since the threshold for this instability of a jack- m1  
up unit is not clearly specified in international design code pro- PðFÞ ¼ PðF 0 Þ U ∏ P F k þ 1 j F k ð18Þ
visions, this study adopts the threshold value 17° equally for roll k¼1

and pitch rotational motions from DNV-OS-C301 (Det Norske where P(F0)¼ the probability of the intermediate failure event F0,
Veritas, 2013) for mobile offshore units and floating offshore and P(Fk þ 1|Fk)¼the conditional probability of the intermediate
installations. Any exceedance of this threshold value either by the failure event Fk þ 1 for the given former intermediate failure event
roll or the pitch rotational motion is assumed to result in the Fk. In this formulation, the probabilities P(F0) and P(Fk þ 1|Fk) are
instable state of the jack-up unit. respectively estimated using MCS and the modified MCMC algo-
rithm (Au and Beck, 2001) for time-varying random variables as
4.2.4. System limit state follows:
Additionally, a system limit state is defined as the series system  
of the abovementioned three limit states for wave overtopping, 1X N
PðFÞ  I min gðxi ðtÞÞ o 0 ð19Þ
instability for roll, and instability for pitch. This is in order to N i ¼ 1 t ¼ 0;:::;T
determine the system failure state of the platform at a given
 
horizontal location. This series system failure is defined by the 1XN
PðF k þ 1 j F k Þ  I min gðxki ðtÞÞ o 0 ð20Þ
following limit state: Ni¼1 t ¼ 0;:::;T
  

 
F sys ¼ [ nk ¼ 1 min aðt k Þ o 0 where xki ðtÞ ¼ conditional samples generated by the modified
k ¼ 1;…;n MCMC algorithm. The algorithm of the subset simulation method
  

  and the implementation of the Eqs. (19) and (20) are summarized
[ [ nk ¼ 1 max ξr ðt k Þ 4 17 3
k ¼ 1;…;n as follows: (1) N samples are simulated using direct MCS to
  
compute P(F0). (2) Define an intermediate failure criterion b1, and
 
[ [ nk ¼ 1 max ξp ðt k Þ 4 17 3 ð15Þ select a subset of the N samples that have N1 samples according to
k ¼ 1;…;n
the criterion, g(X) ob1. (3) Based on these N1 samples, generate
In sum, this study considers a total of four failure events for additional (N-N1) samples using the modified MCMC algorithm,
wave overtopping, instability for roll, instability for pitch, and their and combine the N1 and (N  N1) samples to make a new set of N
series event. Other limit states that are considered to be less likely samples. (4) Define a next intermediate failure criterion b2 and
to be happened compared to the four limit states considered in calculate the conditional probability by counting the N2 samples
this study such as wind-induced fatigue failure are not considered that satisfy g(X) o b2. Then, generate (N  N2) additional samples
in this study. using the modified MCMC algorithm to generate a full set of N
62 W.-H. Kang et al. / Ocean Engineering 111 (2016) 56–66
samples for the ith level. (5) Repeat these processes for further
conditional probability estimations until it reaches the mth level.
Finally, the probability of interest is estimated using Eq. (18).
4.4. Reliability analysis
The subset simulation method is applied to estimate the
probability curves of the application jack-up unit according to the
four limit states for wave overtopping, instability of roll, instability
of pitch, and their series system as defined in Section 4.2. The
analysis has been repeated for varying sea wave conditions
represented by the changing value of Hs. Using the P-M spectrum
(Zhu, 1991), the other wave input parameters Tp and Tz are
obtained from their deterministic relationship with Hs. The dis-
cretised time window of 20 min is considered with the sampling
interval t¼ 0.5 s; thus the number of time instants is 2401. Three
horizontal wave angles, 90°, 45°, and 0°, are considered. For the air
gap estimation, the still-water air gap a0 ¼3.96 m is used to
represent the elevation of the deck of the jack-up unit. The subset
simulation method summarized in Section 4.3 is applied to solve
the first-exceedance failure probabilities. In the subset simulation,
the intermediate thresholds are adaptively selected such that the
intermediate conditional probabilities P(Fk þ 1|Fk) are estimated to
be 0.1. The proposal PDFs in the subset simulation are chosen to be
uniform distribution with the width of 2.
The conditional probability curves of the four limit states for a
given value of Hs are shown in Figs. 7–9 for different wave direc-
tions and different horizontal platform locations (three square
black points on the deck shown in Fig. 4). Each of them shows four
curves for the four limit states discussed in the previous section.
The proposed conditional probability curves can be used to sup-
port the decision on maintaining the platform operation according
to the estimated Hs level. In addition, each plot has a zoomed-in
plot for the range of Hs ¼2–2.6 m, which is intended to clearly
present the small probability estimated using the subset simula-
tion method. Note that the y-axis of this zoomed-in plot has a log-
scale to clearly show the small numbers, which ranges from 10  1
to 10  6. Therefore, any probabilities below 10  6 are not shown in
this zoomed-in plot.
Fig. 7(a) shows the conditional probability for the four limit
states at the horizontal location (0, 15) in beam seas, where the
wave direction is 90° from the moving direction of the jack-up
unit. It is seen from the curves that the curves for wave over-
topping and the series system are very close to each other, while
the two instability curves have small probabilities for the range of
Hs ¼2–5 m. This means that the wave overtopping event is more
likely than instability events, and the series system event is mainly
represented by the wave-overtopping event, which is the most
dominant. Notice also that the wave-overtopping event curve is
sharply increasing within the range of Hs ¼ 3.5–4 m, while that of
the instability curves regarding roll are within the range of Hs ¼4–
4.5 m, which again shows the dominance of the wave-overtopping
event. For Hs 45 m, the probabilities of all curves almost reach
1 except the instability probability of pitch. In this wave direction,
the exceedance event regarding pitch is not significant as the wave
direction and the platform-moving direction are perpendicular to
each other.
Fig. 7(b and c) show the conditional probability for the four
limit states at the horizontal locations (21, 0) and (21, 15) in beam
seas. These plots show a similar trend with Fig. 7(a), except that
the conditional probability of the series system event starts
increasing from a bit earlier than Fig. 7(a) in terms of Hs. This is
because the horizontal locations considered in Fig. 7(b and c) are
further from the center of gravity of the jack-up unit than that in
Fig. 7(a), and consequently, the vertical displacement of the plat-
form becomes greater under the rigid body motion assumption
and the chance of a negative air gap increases. Note that, for
Hs ¼2.5 m, the probability of exceedance event regarding wave
overtopping is less than 5% as suggested in reference (Faltinsen,
1990), which shows the safe operation of the jack-up unit under a
less than-extreme sea condition. However, the wave overtopping
event becomes larger than 5% probability for Hs 42.7 m, which
leads the jack-up unit to the unsafe operation condition according
to the limit state definition considered in this study. The stability
of the jack-up unit is well ensured until Hs ¼ 3.8 m.
Fig. 8 shows the exceedance probability curves for the four
limit states at the three horizontal locations, but the angle
between the wave direction and the moving direction is 45°. In
this case, Fig. 8(b and c) at the horizontal locations (21, 0) and (21,
15) show a similar trend with Fig. 7 such that the wave over-
topping is the most dominant event defining the system failure,
and the series system probability is mostly determined by this
limit state. However, for Fig. 8(a) at the horizontal location (0, 15),
the trend is a bit changing, and the three limit states, i.e., wave
overtopping, and roll and pitch exceedances, contribute together
to the series system probability. This is because the wave-
overtopping event is not as likely as other horizontal points in
this wave direction, and its probability is relatively reduced. Fig. 8
shows that the series system probability starts increasing when
Hs ¼2.2 m at the locations (21, 0) and (21, 15), which means that
the 45° angle between the directions of wave and platform moving
create a higher likelihood of system failure than the 90° angle
cases in Fig. 7.
Fig. 9 again shows the repeated analyses of four curves at the
three horizontal locations, but it considers the following sea
waves, where the wave angle and the moving direction of the jack-
up unit are now in the same direction. The main difference caused
by the change of the wave direction is that the instability event
due to pitch exceedance becomes much more dominant, while the
roll exceedance event becomes almost negligible. In Fig. 9(a), at
the horizontal location (0, 15), the conditional probability for the
series system limit state is mostly affected by the curve for the
pitch instability, and the other events do not contribute sig-
nificantly. Although the wave-overtopping event shows increasing
conditional probability for Hs 44 m, this is mainly caused by the
uncertainties in the wave motion and does little to affect the
system event probability. In Fig. 9(b) and (c), at the horizontal
locations (21, 0) and (21, 15), the probability of the wave over-
topping becomes the most dominant in terms of the contribution
to the system event, as the large pitch of the platform generates
large vertical displacements under the rigid body motion
assumption, and it increases the likelihood of a negative air gap
event. In these curves, the instability event due to roll is again
almost negligible. Fig. 9 shows in general the highest series system
probability compared to Figs. 7 and 8, representing the most risky
case in terms of the limit states considered.
From the discussion of the results shown in Figs. 7–9 above,
several observations can be made. The probability of exceedance of
motion responses is exactly proportional to the significant wave
height with non-linear characteristics. Comparing with the results
shown in Fig. 6, the system reliability index can provide further
information on dangerous conditions. For example, when the jack-
up is towed on the oblique waves with the significant wave height
2.4 m, the motions are usually small. The significant value of heave
is 0.539 m, roll is 1.145°, and pitch is 5.532°, respectively. These
motions are safe for the stability of the jack-up in waves since they
neglect the uncertainties in the wave motion and the corre-
sponding extreme cases. Nevertheless, the failure probability of
the system at the deck location (21, 0) is up to 30% as shown in
Fig. 8(b), which can be considered as a relatively high probability
of exceedance and dangerous for the jack-up unit considered in
this study. It can be noted that the four extreme ship-wave motion
 W.-H. Kang et al. / Ocean Engineering 111 (2016) 56–66 63

Fig. 7. Probability of exceedance of platform responses for four limit states in beam seas.

related limit states considered in this study show the non- In these analyses, the subset simulation method was adap-
operational risk of a jack-up unit for not-too-extreme severe sea tively used because its use is not necessary for relatively large
states with Hs ¼2–5 before considering other types of structural probabilities for which the use of MCS is generally acceptable.
limit states. Thus, in each analysis, MCS has firstly been performed, and the
64 W.-H. Kang et al. / Ocean Engineering 111 (2016) 56–66

Fig. 8. Probability of exceedance of platform responses for four limit states in oblique seas.

results were checked based on their coefficients of variation if equation:

there is a need for the use of the subset simulation. In MCS, it is rffiffiffiffiffiffiffiffiffiffiffi
well known that the coefficient of variation (c.o.v.) of the esti- 1P
δ¼ ð21Þ
mated probability can be obtained based on the following NP
 W.-H. Kang et al. / Ocean Engineering 111 (2016) 56–66 65

Fig. 9. Probability of exceedance of platform responses for four limit states in following seas.

where P ¼ the estimated probability and N ¼the number of sam- of RAM. If the estimated probability is smaller than 2/N, which
ples in MCS. In this paper, we used 500 samples for MCS, and the corresponds to approximately 0.7 of c.o.v. from Eq. (20), addi-
computational cost is approximately 125 seconds with MATLAB tional 500 samples have been generated based on MCMC within
on a computer with an Intel Core i7 CPU 930 (2.80 GHz) and 3 GB the subset sampling framework, which requires 194 s. If the
66 W.-H. Kang et al. / Ocean Engineering 111 (2016) 56–66
estimated conditional probability using the additional 500 MCMC
samples is still too small, additional 500 MCMC samples are
generated again. The maximum number of the addition of new
500 MCMC samples is limited to three times, and if we still get a
probability very close to zero, we assume that the final prob-
ability can be regarded as zero.
When we use the subset simulation, the coefficient of variation
(c.o.v.) of the estimate P(Fk þ 1|Fk) in Eq. (20) is estimated in an
asymptotically unbiased manner as (Au and Beck, 2001):
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 P k  
δk ¼ 1 þ γj ð22Þ
NP k
where
Lc 
X 
jN s
γj ¼ 2 1 ρk ðjÞ ð23Þ
N
j¼1
where Lc ¼the length of each chain, Ns ¼the number of seeds at
subset level k, and ρk(j) is the average j-lag auto-correlation
coefficient of the stationary sequences represented by MCMC
samples (Au and Beck, 2001). By assuming the statistical inde-
pendence of the estimates of the conditional probabilities, c.o.v. of
the probability P(F) in Eq. (18) can be estimated as:
X
m References
δ2PðFÞ  δ2l ð24Þ
l¼1
Using Eq. (24), c.o.v.'s of several small probabilities have been
estimated as follows. For the 90 degree wave angle and the hor-
izontal location (0,15), when Hs ¼ 2.4 m, the c.o.v. of the probability
of system limit state exceedance is estimated as 0.43. In this case,
1500 samples are required to estimate the probability 0.00036. If
we use MCS instead of the subset simulation, we only obtain zero
probability even with the same number of samples. As another
example, for 0 degree wave angle and the horizontal location
(21,15), pitch instability probability is estimated as 0.00014 and the
c.o.v. is estimated as 0.68586, which requires a total of 2,000
samples. This probability cannot be estimated using MCS even
with the same number of samples.
5. Conclusions
A system reliability-based stochastic analysis approach was
proposed, which was built on a deterministic offshore platform
motion estimation framework to estimate the probability of the
extreme motions of a floating platform subjected to stationary
stochastic wave loads. The approach was applied to a jack-up unit
under towing condition depending on the buoyancy provided by
the hull and mat. First, the RAOs were calculated to represent the
frequency characteristics of the coupled motions of the jack-up at
4 knot forward speed in different wave directions, and some
resonance phenomena from the RAO curves were found. The
surge, sway, and yaw motions were ignored because the vertical
motions have the largest influence on the stability of the jack-up
towing in waves. The statistically significant motions of the jack-
up were obtained. To overcome the fact that the significant
motions could not determine the dangerous sea condition speci-
fically, the probability of exceedance of jack-up motion responses
was also evaluated, considering the probability of deck over-
topping, and instability for roll and instability for pitch. These
analyses were conducted based on the dynamic responses and
computed using the Monte Carlo simulation (MCS) technique in
the time domain. The subset simulation method was also
employed adaptively together to achieve a sufficient convergence
with a limited number of samples, especially for low probability
estimation. From this probabilistic analysis, the safety of the jack-
up unit in terms of its motions was evaluated considering possible
uncertainties in wave motions in various sea states, and the con-
ditional probability curves were generated to support the deci-
sions on the action taken if excessive responses are foreseen for
the present course and speed. The proposed framework and pro-
cedures for deterministic and probabilistic analysis of a jack-up
under towing condition can extensively be applied to other
floating platforms or ships in a natural seaway.
Acknowledgments
This research was supported by the National Natural Science
Foundation of China (Grant no. 51350110230), “973” Project (Grant
no. 2014CB046804), Australian Research Council (Project no:
LP140100030), and the Western Sydney University Research
Partnerships Program. The authors would like to acknowledge Dr
Zunfeng Du from Tianjin University for his involvement in the
application.
Zhao, T., Sun, C., 2014. A new spudcan with buoyancy modules for mobile jack-up
rigs. Appl. Ocean Res. 47, 154–161.
Vazquez, J.H., Michel, R.P., Alford, J.H., Quah, C.K., Foo, K.S., 2005. Jack-up Units: A
Technical Primer for the Offshore Industry Professional, 1st ed. Bennett Off-
shore and Offshore Technology, Los Angeles.
Wang, X.M., Koh, C.G., Zhang, J., 2014. Substructural identification of jack-up plat-
form in time and frequency domains. Appl. Ocean Res. 44, 53–62.
Vlahos, G., Cassidy, M.J., Martin, C.M., 2008. Experimental investigation of the
system behaviour of a model three-legged jack-up on clay. Appl. Ocean Res. 30,
323–337.
Bienen, B., Cassidy, M.J., 2006. Advances in the three-dimensional fluid–structure–
soil interaction analysis of offshore jack-up structures. Mar. Struct. 19 (2–3),
110–140.
Leira, B.J., Karunakaran, D., 1995. Site-dependent reliability of a mobile jack-up
platform. Mar. Struct 8, 151–169.
Jensen, J.J., Capul, J., 2006. Extreme response predictions for jack-up units in second
order stochastic waves by FORM. Probab. Eng. Mech. 21 (4), 330–337.
Jensen, J.J., 2011. Extreme value predictions using Monte Carlo simulations with
artificially increased load spectrum. Probab. Eng. Mech. 26, 399–404.
Hunt, R.J., Marsh, P.D., 2004. Opportunities to improve the operational and tech-
nical management of jack-up deployments. Mar. Struct. 17 (3–4), 261–273.
Det Norske Veritas, 2007. WOAD – Worldwide Offshore Accident Databank.
Gibson, Vic, 2000. Jack-up moving – an overview. Offshore Support J (internet), p. 1.
Available from: 〈http://www.shipsandoil.com/features/Featjumv.htm〉.
Newman, J.N., 1977. Marine Hydrodynamics, 1st ed. MIT Press, Cambridge.
Perez, T., 2005. Ship Motion Control: Autopilots With Rudder Roll Stabilisation and
Combined Rudder-fin Stabilisers, 1st ed. Springer, London.
Journée, J.M.J., Massie, W.W., 2001. Introduction in Offshore Hydromechanics, 1st
ed. Delft University of Technology, Delft.
Det Norske Veritas, 2007. DNV-RP-C205: Environmental conditions and Environ-
mental Loads.
Ansys, 2010. AQWA Reference Manual Version 13.0, USA.
Chakrabarti, S.K., 1987. Hydrodynamics of Offshore Structures, 1st ed. Computa-
tional Mechanics Publication, Southampton.
Jensen, J.J., 2007. Efficient estimation of extreme non-linear roll motions using the
first-order reliability method (FORM). J. Mar. Sci. Technol. 12 (4), 191–202.
Zhou, Z.Q., De K.J.O., Buchner B., 1999. A nonlinear 3-D approach to simulate green
water dynamics on deck. In: Proceedings of the 7th International Conference
Numerical Ship Hydrodynamics, Nantes.
Bea, R.G., Xu, T., Stear, J., Ramos, R., 1999. Wave forces on decks of offshore plat-
forms. J. Waterw. Port C 125 (3), 136–144.
Det Norske Veritas, 2013. DNV-OS-C301: Stability and Watertight Integrity; 2013.
Au, S.K., Beck, J.L., 2001. Estimation of small failure probabilities in high dimensions
by subset simulation. Probab. Eng. Mech. 16 (4), 263–277.
Zhu, Y.R., 1991. Ocean Engineering Wave Mechanics, 1st ed. Tianjin University
Press, Tianjin, Chinese.
Faltinsen, O.M., 1990. Sea Loads on Ships and Offshore Structures, 1st ed. Cambridge
University Press, Cambridge.
Dostal, L., Kreuzer, E., Namachchivaya, N.S., 2013. Stochastic averaging of roll-pitch
and roll-heave motion in random seas. Procedia IUTAM 6, 132–140.
Shabakhty, N., 2011. System failure probability of offshore jack-up platforms in the
combination of fatigue and fracture. Eng. Fail. Anal. 18 (1), 223–243.