KOMTECH Technology Review 2016

Technology Review 2016
About KOMtech
Launched in December 2007, Keppel Offshore & Marine Technology Centre (KOMtech) is
an extension and strengthening of Keppel Offshore & Marine’s (Keppel O&M) research and
development initiatives. KOMtech comprises two arms, Shallow Water and Deep Water. The
Shallow Water segment looks into Arctic jackup designs, offshore wind solutions, LNG
applications, environmental and drilling solutions for the oil and gas industry; while the
Deep Water segment looks into ultra-harsh environment semisubmersibles, tension leg
platforms, innovative ship designs, as well as shipyard technology to improve shipyard
processes and productivity.
KOMtech complements and augments the work of the existing three design and engineering
units within Keppel O&M - Offshore Technology Development (OTD), Deepwater
Technology Group (DTG) and Marine Technology Development (MTD). Leveraging
existing and proprietary technologies, and in collaboration with universities, research
institutes and industry partners worldwide, KOMtech continues to develop innovative
solutions that are commercially viable and adaptable to the needs of the industry.
Image on cover: (From the Top) 3D renderings of: Keppel’s 3MMTPA Dolphin Moored FLNG and
FSU Arrangement; Liftboat built to Keppel’s Properietary design ORCA 2500R.
CEO’s Message
In line with Keppel Offshore & Marine's (Keppel O&M) corporate vision of being the provider of choice
and partner for solutions in the global offshore and marine industry, Keppel O&M remains at the
forefront of our industry, working closely with our stakeholders to develop and deliver quality, innovative
and value-adding solutions to our customers.
2015 has been a challenging year. As Keppel O&M braces itself for a tough 2016 in what is turning out
to be a "lower for longer" oil price environment, we recognise the importance of always keeping an eye
on opportunities for longer-term value and efficiency gains, even as we focus on immediate cost
management efforts and near-term value generation.
To ensure that we stay relevant and ahead in the market, Keppel O&M continues to make prudent
investments in research and development (R&D) in order to enhance our core competencies and develop
new solutions. On the back of the current depressed market conditions, we have the versatility to leverage
our deep industrial know-how to create innovative solutions. At the same time, we continue to work
closely with industry partners, clients and academic institutes to produce robust solutions that meet the
needs of the industry.
One foreseeable need will be to improve the industry's process efficiency and productivity in order to
achieve much-needed cost savings. This is one of the research focuses of the Keppel-NUS Corporate
Laboratory, a collaboration between Keppel O&M and the National University of Singapore. In
particular, the research emphasises on productivity enhancements to yard operations to reduce reliance on
manual labour.
Beyond our current core competencies, Keppel O&M has been actively exploring non-drilling businesses
including the Liquefied Natural Gas (LNG) market. Apart from technological solutions, Keppel O&M
is also working with Shell to provide end-to-end LNG bunkering services in the Port of Singapore,
starting in 2017. Our broad range of solutions will stand us in good stead to capture opportunities that
arise from the growth of the LNG market.
Keppel O&M will also continue to diversify by exploring non-oil and gas-related projects whereby we are
able to utilise our existing competencies and experience, such as those pertaining to renewable energy and
offshore desalination.
Moving forward, Keppel O&M Technology Centre (KOMtech) will continue to drive Keppel O&M's
technology and innovation efforts, augmenting the design and engineering activities of the various
business units by providing specialised capabilities and upstream R&D. With a team of about 70
researchers, KOMtech is committed to developing competencies, promoting innovation, stimulating and
carrying out application research, as well as engaging in technology foresight to keep Keppel O&M ahead
of the game.
Technology Review 2016 showcases KOMtech's multi-disciplinary R&D expertise, with a closer look at
trends covering a wide spectrum including offshore, marine and environmental-related developments,
the renewable energy space, as well as the global LNG market. It is a diverse line-up that reflects the
complexity of the offshore and marine industry, and the wide range of competencies that Keppel O&M
is enhancing or developing.
This publication includes valuable inputs from our research partners, clients and other stakeholders who
are also industry front-runners. As such, I hope that you will find this book valuable, and that it will
continue to be a source of inspiration and encouragement for us to deliver quality and innovative
solutions that are commercially viable and relevant to the needs of the global market.
Yours sincerely,
Chow Yew Yuen

CEO
Keppel Offshore & Marine
2 KOMtech Technology Review 2015
Contents
ABOUT KOMtech
1 CHAIRMAN’S MESSAGE
5 EDITORIAL NOTE
FEATURED ARTICLES
1 Jackup Going On Location Analysis

This paper presents a recent development of analysis methods applied to jackup units
going on location.
2 Design Considerations of DOL And VFD Jacking System

The paper discusses the characteristics and peculiarities to be considered for both
DOL and VFD Jacking systems.
3 The Impact of Refrigerant Selection and Precooling Condition

on LNG Production Robustness and Efficiency
This papers describes the selection of precooling refrigerant is a tri-angular exercise
where the physical properties has to fulfill the required precooling process condition.
The energy efficiency, safety and environmental impact should also be taken into
consideration to result in an efficient, robust and safe process, such as Keppel’s PreNEx.
FEATURED HIGHLIGHT
4 Experimental Investigation on Flow Induced Motion of

Tension Leg Platform
The paper discusses and presents a novel arrangement to represent the tendon stiffness
in vertical and horizontal modes in a shallow current channel.
5 Effect of Drillship’s Moonpool Shape on The Resistance

by CFD and Ultra-Reduced Model Experiments
The paper presents the comparison between simulation of the CFD Models and
experimentation results of the effects of a drillship's moonpool shape.

Contents 3
6 Design and Installation of an Active Wave Absorber

in a Wave Channel
The objective of the present work is to design and install a new active wave absorption
system into fKN@LOC, Laboratory of Waves and Currents, at COPPE/UFRJ.
7 Portable Fuel Tank - The Application of a New Method of

Bunkdeing for Small Scale Gas Fuelled Vessesl
This paper describes a new approach of LNG bunkering for gas fuelled tug by replacing the
empty fuel tanks with preloaded fuel tanks which are transported by truck, ship or rail.
8 Innovative Fatigue Prediction Model for Improved Fatigue

Performance of Welded Steel Connections in Offshore
Wind Turbine Jackets
This paper describes the present day fatigue prediction method and outlook of
an innovative model in which the present day method is potentially adapted to be
sensitive to load sequence effects.
9 OWEC Pile Stopper Solutions

This paper explores the solutions in the portfolio of OWEC Tower towards increasing
the operability and reliability of OWT jacket installation while lowering the associated costs.
10 Weight Monitoring System

This paper talks about a real-time system which displays the distribution of
weight applied to legs of a Jack Up rig and helps in making rig’s stability decisions.
11 Numerical Prediction of Wind Loads on Topside of

Semi-submersible Platform
This paper presents CFD (computational fluid dynamics) validation and prediction of
wind loads on complicated topside of a semi-submersible platform.
12 Review of Floating Offshore Structure Concepts for

Exploration in the Arctic Region
This paper discuss the main challenges for the design of Arctic Floaters and the
main principles applied in development of these designs. An overview of available
design solutions is also provided.
13 Drillship DLA/SFA Notation Design Review

DLA/SFA notation is the first vessel to have that notation in keppel design & build history,
this is the global strength and fatigue analysis for ship. This design review is to share the
calculation Competence for ship design areas.
Managing Director’s Message
Since the establishment of Keppel Offshore & Marine Technology Centre

(KOMtech) in 2007, KOMtech has continued to strive for excellence in
developing competencies, promoting innovation, carrying out application
research as well as product and process development to keep Keppel Offshore
& Marine (Keppel O&M) at the forefront of its industry. By anticipating the
industry’s requirements through technology foresight and partnering trendsetting
customers and institutions, KOMtech has to-date delivered a wide range of high
quality and cost-effective solutions that benefit all our stakeholders.
As we move into 2016, which continues to be challenging for the offshore

and marine industry given the low oil prices, managing capital expenditure
and operating costs will be among the top concerns for the industry. However,
rightsizing should not mean just cutting costs – it should also be about value
engineering and investing carefully in new or enhanced solutions so that the
needs of the industry can continue to be met in a robust and cost-effective
manner. Doing so will help ensure companies who continue to invest wisely in
Research and Development (R&D) to emerge stronger from a down cycle.
We are doing precisely that by prudently allocating resources and looking

closely at areas that we can continue to improve in, such as enhancing our core
competencies and building up new strengths to keep ourselves fighting fit.
Commercial viability has always been the key principle of Keppel's technology
innovation efforts, and it will continue to guide the Group in its R&D work.
When pursuing new concepts and solutions, we consider their commercial
viability, our customers’ needs, as well as how they can grow the company’s
knowledge pool and achieve process improvements.
Technology Review 2016 highlights KOMtech’s research in multidisciplinary

areas such as offshore, marine and green/renewable energy, and also its research
in developing new competencies which can better assess and improve operability.
For instance, the focus of jackup technology, which for many years has been on
elevated performance, has been gradually shifting towards enhanced operability.
In this aspect, KOMtech is working with classification societies and Keppel O&M
business units to develop robust methodologies for assessing operability under
transient conditions, which include the use of Computational Fluid Dynamics
(CFD) to model and predict wind loads. Developments in analysis capability to
better predict structural and motion responses will enhance our suite of rig move
tools and support available to customers.

Managing Director’s Message 5
The 2016 edition of our Technology Review also highlights our ongoing research
to better address the current and future needs of the Liquefied Natural Gas (LNG)
industry. Despite a slowdown in global economic growth and weak medium-
term outlook for gas, we believe the LNG market continues to have a bright
future. In fact, the use of LNG as an alternative marine fuel has been increasing
due to emissions reduction requirements set by the International Maritime
Organisation. Among the LNG solutions we have developed our proprietary
liquefaction technology, PreNEx, which when combined with our over 35 years
of experience and expertise in conversion and modularisation, can partner client
to monetise LNG in a fast and cost effective way.
My appreciation goes to the industry partners, partnering universities and

Keppel O&M’s business units who have contributed to the research work and
development of this publication. The researchers and writers of Technology
Review 2016 have put in many hours of hard work to ensure that this publication
is insightful and thought-provoking. Through the Technology Review, we hope
we can stimulate further discussions and innovation amongst industry partners
and customers, and be of value to the technology community in the offshore and
marine industry.
Michael Chia
Managing Director
Keppel O&M Technology Centre
EDITORIAL COMMITTEE
Advisors Mr Michael Chia, Dr Foo Kok Seng, Mr Aziz Merchant,
Mr Charles Sim and Mr Charles Foo
Editorial members Ms Tan Mui Yen, Ms Elaine Toh, Mr Ronald Chua,
Mr Syed Abidi, Mr Wu WenJin, Dr Liu Ming,
Mr Ankit Choudhary and Mr Say Huan Yuan.

Jackup Going On Location Analysis 7
Jackup Going On
Location Analysis
 Paolin TAN, Xiaohong CHEN, Qing YU

ABS
Michael PERRY, Han MU

Keppel Offshore & Marine Technology Centre
Ted CHANG, Marc WONG

Bennett Offshore
Donghui CHEN
Genesis Engineering
INSTEAD OF USING THE TRADITIONAL WAVE HEIGHT LIMIT, OFTEN

REFERRED TO AS THE 5-FT WAVE, it is possible to develop a more rational
definition of operating limits for jackup units going on location through simulations
of jackup leg touchdown on the seabed. This paper presents a recent development
of analysis methods applied to jackup units going on location. In addition to the
detailed analysis method involving the direct time-domain simulation of global
motions and dynamics of the jackup whose legs are approaching and impacting
the seabed, two simplified methods based on different simplification schemes are
also studied. Verification of these simplified methods is carried out using a generic
jackup design and representative soil conditions. Results show that both simplified
methods perform reasonably well in predicting maximum vertical impact forces in
comparison to the more sophisticated method.
KEY WORDS: Jackup Going On Location; Jackup Installation; Design Wave

Method; Coupled Analysis.
ACKNOWLEDGEMENTS We appreciate support from ABS, Bennett Offshore

and Genesis Energy for republishing this work, which has been reproduced with
kind permission of the Organising Committee of the International Conference for
The Jack-Up Platform: Design, Construction and Operation. Passing its 30th year
in 2015, the international conference on the Jack-Up platform is held at City
University London and each successive biennial event has been a great success.
The conferences have proved to be an excellent forum for the exchange of ideas
and the discussion of recent developments in the offshore oil and gas industry
particularly related to Jack-Up technology. They have also contributed to the
development of codes of practice.
No part of the materials published in this journal may be reproduced, stored in a retrieval system or transmitted in any form whatsoever without the prior written permission of KOMtech
8 KOMtech Technology Review 2016 LIQUEFIED NATURAL GAS
INTRODUCTION between the spudcan and the seabed were explicitly

While the methods for assessing jackup units modeled. Rather than simulating the continuous
in elevated mode during operations are well lowering of the legs in each simulation run, the
established, a better understanding is still needed legs were assumed to be locked to the hull and a
for the temporary modes, as it is in these temporary prescribed initial clearance between the spudcan
modes that the majority of productivity and and the seabed representing a single time instant of
property losses take place. The temporary mode the jacking operation was applied in a simulation.
explored in this paper is the so-called ‘jackup going In order to model the entire jacking operation for
on location’, also known as ‘jackup installation’ or a given sea state, multiple initial clearances between
‘jackup emplacement', the operation that involves a the spudcan and the seafloor for the same realization
floating jackup lowering its legs down to the seabed of that sea state were analyzed in separate simulation
till the hull starts being elevated from the water. runs.
Most jackup operating manuals provide only It has been recognized that many factors can affect
minimum guidance for the going on location the loads and responses during jackup going on
operation often in the form of a single limiting location. Four physical behaviors are considered
wave height and allowable maximum pitch and critical in assessing the maximum responses resulting
roll motions. Attempts were made in the past to from jackup leg impact on the seabed. These are
develop analysis methods that can better assist the jackup global motions in random seas, the leg
jackup going on location operation. One of the early lowering process, transient impact of the spudcan
developments, as part of a joint industry project, on the seabed, and jackup structural dynamic
was reported by Lai, et al. (1993) [1], Miller et al. responses. Recent advances in simulation software
(1993) [2], and Smith et al. (1994) [3]. The jackup enable sophisticated simultaneous modeling of
going on location analysis was carried out using a these behaviors in the time domain and, therefore,
2-D finite element model (stick model), while the provide a more detailed simulation of the jackup
afloat jackup global motions were pre-calculated leg touchdown process. While this method may
using a frequency domain hydrodynamic analysis not be practical for operational support or routine
program and then used as boundary conditions for engineering design given the enormous extent of
the finite element analysis. The basic assumption the associated computational efforts, it does provide
is that the spudcan impact on the seabed does not a viable means of deriving verification data that
affect the hydrodynamic responses of the afloat can help in developing more efficient and less
jackup. complex analysis methods that can still yield
comparable results.
Similar methods but with more refined structural
models for the jackup going on location analysis This paper presents a recent development of
were later studied by Matter, et al. (2005) [4], analysis methods applicable to jackup units going
Chakrabarti (2012) [5] and Dong et al. (2013) [6]. A on location. In addition to the detailed analysis
common assumption made in these studies is that, method, which involves the direct simulation of
during the entire impact analysis, the spudcan is the aforementioned four critical behaviors, two
always in contact with the soil springs representing simplified methods based on different simplification
the seabed. This may be reasonable for the case schemes are also studied in the present research.
of very soft soil, but is obviously not capable of The first method follows the so-called design wave
simulating the rebound when the spudcan touches approach where a random sea is represented by a
down on the stiff seabed. number of regular waves (i.e. design waves) derived
to replicate the extremes of selected global motions
Most recently, Daun and Olsson (2014) [7] explored that potentially could govern the impact responses
the possibility of adapting the software originally due to the jackup leg touchdown. A jackup with
developed for floating offshore structure global a continuously lowering leg is simulated to each
performance analysis to simulate the jackup leg design wave in the time domain to calculate the
touchdown on the seabed. Effects of clearance maximum impact responses. The second method

simplifies the leg lowering process to a number of

discrete spudcan positions, each of which remains
unchanged relative to the hull during the time (2)
domain simulation of the jackup in random seas.
Verification of these two simplified methods is
carried out using a generic jackup design and Equation (1) shows that the natural frequency
representative soil conditions. Results show that of the system under consideration depends on
both simplified methods perform reasonably well both the initial gap distance and the motion
in predicting maximum vertical impact forces in amplitude of the mass. When the initial clearance
comparison to the more sophisticated method. vanishes, i.e., ρ=xc/x0=0, the system natural period
simplifies to the well-known bilinear frequency
MODELING CONSIDERATIONS 2ω-ω+/(ω-+ω+). Another interesting observation is
One of the subjects often neglected in previous for the case of α ➛ ∞ where the stiffness of the free
studies on the jackup going on location analysis spring (i.e. soil) is much greater than the stiffness of
is the process during which the jackup spudcan the spring attached to the mass, the limiting natural
approaches the seabed and begins contacting the frequency of the system becomes
seabed intermittently. To start with the discussion
without losing generality, consider a simple 2ω-/(1+2/πsin-1ρ) for any initial clearance satisfying
representation of the jackup going on location, as ρ ∈[-1, 1].
illustrated in FIGURE 1, using a single degree of
freedom (SDOF) mass-clearance-springs system. The implication of Equation (1) is that, for a jackup
with its legs being continuously lowered toward the
x seabed during the installation, the natural period of
xc the jackup-clearance-seabed system after an initial
k kc impact occurs will keep changing along with the
m varying spudcan-seabed clearance relative to the
motions of the jackup (i.e.ρ). Given the importance
of the natural period in determining the responses of
the jackup during installation, Equation (1) suggests
Figure 1. A Bilinear Free Oscillation System with
Clearance (Butcher, 1999 [8]) that the effect of spudcan-seabed clearance and the
change of such clearance should be considered in
the going on location analysis.
Assuming linear spring stiffness and no damping,
the analytical solution of the natural frequency, Ω, In addition to the spudcan-seabed clearance, there
of this bilinear free oscillation system is determined are other aspects that could affect the impact
by Equation (1), as derived by Butcher, 1999 [8]: responses of the jackup during installation. These
may include:
(1) • The jackup is subjected to continuous excitation

forces exerted by random environmental actions
where ρ=xc/x0 is the ratio of the clearance, xc, to the and holding tugs. The intermittent impact of
initial displacement magnitude, x0, and must lie in spudcan on the seabed introduces additional
the interval [-1, 1] in order for contact/separation to impulsive excitation forces that will interact
occur; ω- and ω+, as given in Equation (2), are the with those due to environmental actions and
natural frequencies of the mass-clearance-springs holding tugs through the transient responses of
system before and after contact with the free spring; the jackup. The significance of such a coupling
α is the ratio of ω- to ω+, representing the level of effect varies with the overall damping level and
frequency separation; m is the mass and k is the natural period of the jackup-clearance-seabed
linear stiffness of the spring attached to the mass; kc system as well as with the time interval of two
is the linear stiffness of the free spring. consecutive impacts.
• Structural dynamics of the legs and leg-hull analysis methods. The vertical impact force on
connections could affect the hull responses to the leg-spudcan interface is chosen as a critical
the impact load exerted on the spudcan. In design parameter to present the developed analysis
contrast to the analysis for elevated mode where methods and verification results in this paper.
attention is primarily given to structural
responses, the jackup going on location analysis JACKUP MODEL FOR THE CASE STUDY
also concerns about the effect of impact-induced The jackup selected to demonstrate the direct
dynamics of the legs and leg-hull connections simulation approach and the simplified analysis
on the floating hull motions. methods is adapted from BASS350 design, a
moderate environment jackup comparable in
• Impact responses of the jackup hull may capabilities to the B-Class, Super 116E, and Pacific
introduce additional changes to the water plane 375 jackups. Major dimensions and relative mass
area when the equilibrium position is no longer properties of the jackup are depicted in FIGURE
on even keel and, as a result, the hydrostatic 2. The water depth assumed for the present study is
stiffness. They can also affect the application 61 m (200 ft). FIGURE 3 shows the natural periods
of the frequency-dependent added mass and of the free-floating jackup whose legs are lowered
hydrodynamic damping of the jackup hull. to the depth where the spudcans are 2 m above the
mudline in calm water. The leg lowering speed is set
• Nonlinearity of the spudcan-soil interaction and to be 0.01 m/s (2 ft/min).
potential side sliding of the spudcan on
the seabed may further alter impact responses. Two homogeneous soil conditions are selected in
Assuming all other conditions are the same, the study, one represents very stiff soil with the
stiffer seabed typically leads to higher impact undrained shear strength (Su) of 450 kN/m2 on
loads on the spudcan. the mudline and another is for very soft soil with
Su = 5 kN/m2 on the mudline. For the results
A direct time-domain simulation approach, presented in this paper, linear soil model is applied;
as described below, is developed to explicitly the soil spring stiffness is derived in terms of
model most of these aspects to the extent that undrained shear strength of the soil.
their interactions can be simulated at each time
incremental step. Simplifications do apply, The software used to simulate the jackup global
especially to the spudcan-soil interaction model, motions, structural dynamic responses and impact
due mostly to the software limitations, although the on the seabed is OrcaFlex, a 3D nonlinear time-
developed simulation approach itself is intended domain finite element program used mostly
for general applications. It is also noted that the for global motion analysis for floating offshore
effect of holding tug and current drag are currently structures, moorings and risers. The main jackup
not considered in the simulation. structural components, i.e. the hull, legs (hybrid
models), spudcans, guides and pinions, are
In addition to this numerical intensive direct included in the analysis model, as illustrated
simulation approach, efforts are also made to develop in FIGURE 4. The jackup legs can be moved
simplified methods that can achieve an acceptable relative to the hull such that the spudcan
level of accuracy for much less computational cost. elevation can be adjusted at each time step in the
The results obtained from the direct simulation time-domain simulation of the leg touchdown
approach are used in this study as the reference on the seabed. The spudcan is modeled using
data for verifying the suitability of simplified a group of pipe elements, as shown in
Efforts are also made to develop simplified methods that

can achieve an acceptable level of accuracy for much less
computational cost.

70 m Mass Roll Moment of Inertia - Ixxg
39.3 m (129 ft) Leg

15% Leg
Hull
28% 31%
Spudcan
16%
Hull
69%
Spudcan
43.3 m (142 ft)
41%
64 m
Pitch Moment of Inertia - Iyyg Yaw Moment of Inertia - Izzg

Leg
12%
Hull Leg
31% 30%
Spudcan
13%
Hull
75%
Spudcan
39%
Figure 2. Major Dimensions and Mass Properties of the Case Study Jackup Model.
Natural Period
HEAVE(Z) ROLL(RX) PITCH(RY)
20
18
16
14
Natural Period (s)
12
10
0
0 5 10 15 20
Wave Period (s)
Figure 3. Natural Periods of the Jackup in Figure 4. Jackup and Spudcan Models in OrcaFlex.
Free-Floating Mode.
FIGURE 5. Collectively, the properties of these hydrodynamic responses in random seas, structural
pipe elements are tuned to match the leg dynamics of legs and leg-hull connections, and
penetration curve determined in accordance with transient impact on the seabed in a ‘fully coupled’
the recommended method in ISO 19905-1[9] manner, meaning their interactions are taken into
for a given soil condition. The hydrodynamic account at each incremental simulation time step.
coefficients required as inputs to OrcaFlex The jackup model as shown in FIGURE 4 is used in
are calculated using ANSYS AQWA LINE, the direct time-domain simulation. In addition, the
which is a frequency-domain diffraction-radiation randomness of waves and jackup global responses
hydrodynamic analysis program. as well as the randomness of the moment when the
jackup legs start being lowered down are explicitly
DIRECT TIME-DOMAIN SIMULATION OF modeled. In so doing, the coupling effect of random
JACKUP GOING ON LOCATION wave induced global motions and the randomness
The direct time-domain simulation of jackup going of spudcan-seabed clearance can be modeled in the
on location involves detailed modeling of jackup’s impact analysis.
The direct simulation starts with generating a legs till the legs settle on the seabed and their
3-hour wave elevation time series for a given sea transient movements subside. Since jackup motions
state and a selected random seed. To limit the in free-floating mode are governed by the linear
calculation time, a 1-hour segment of wave wave-frequency responses of the hull, it is not
elevation time series containing the maximum necessary to run the simulation of free floating
wave height is extracted as the input for the time- motions for the entire time duration before a
domain motions analysis in OrcaFlex. The initial selected starting moment of leg lowering. Instead,
elevation of the jackup legs are chosen to allow only a short lead time sufficient to develop the
for a sufficient clearance to the seabed such that, hull motions is needed. This essentially leads
if the legs were locked to the hull, no impact on to applying a sliding time window at 2-second
the seabed would occur in the 1-hour simulation interval to carve out segments of wave elevation
and the jackup would be in free floating mode. time series as the inputs to the OrcaFlex simulations
Since the jackup could start lowering its legs at of leg touchdown.
any moment during that 1-hour time window,
consecutive starting moments at an interval of An example of a 1-hour time series for a random
2 seconds are selected for the initiation of leg sea state (Hs = 2 m, Tp = 10 s, PM spectrum) and
lowering. This results in 1800 simulation cases, a zoom-in plot of wave elevation history showing
each of which is associated with a distinct moment the starting moment of leg lowering at the local
when the jackup switches from the locked-legs, time scale t = 30 are presented in FIGURE 5.
free-floating mode to the process of lowering the The third plot in FIGURE 5 depicts the resultant
total vertical impact force recorded at the connection
between the forward leg and its spudcan in the
case of very stiff soil. The maximum impact force,
as marked in the plot, represents an occurrence
of extreme event. In the present study, there are
1800 records of such maximum impact force for
a 1-hour realization of a random sea state.
FIGURE 6 depicts the normalized (by their mean

value) zero-mean values of the maximum vertical
impact forces for two realizations of the given sea
state and for the seabed with very stiff and very
soft soil conditions. It shows that the variation of
Figure 5. Example Wave Elevation and Vertical Impact Forces during Spudcan Touchdown on the Seabed (The jackup
legs start being lowered down at the local time scale t = 30 at a constant speed of 0.01 m/s. The jackup is
in free floating mode with locked legs prior to t = 30 s. The seabed has very stiff soil.)

normalized vertical impact force is greater in the with various starting moment of leg lowering.
soft soil case, although the associated mean value The extreme value of the impact force associated
of the impact force is much lower in comparison with a specified statistic metric can be predicted
to the stiff soil case. It further shows that the using the records of these extreme events – for
maximum impact forces obtained from the instance, each positive data point of zero-mean
simulations associated with two adjacent starting samples for a given sea state and a soil condition
moment of leg lowering at a 2-second interval in FIGURE 6 may be considered as a maxima
could vary significantly, in particular for the soft and used in determining the most probable
soil case. This difference is not only caused by the maximum extreme value.
leg lowering speed, which is fairly slow (0.01m/s),
but more prominently the 6-DOF motions of the SIMPLIFIED METHODS
jackup hull that could lead to a very different initial The direct time-domain simulation, as described
elevation of the leg when it starts lowering down. above, allows for ‘fully coupled’ modeling of
Note that 2 seconds is 1/5 of peak wave period jackup going on location in a random sea and can
Tp = 10 s. explicitly simulate the interaction of sea state and
leg lowering. The major drawback of this method,
Note further that each data point in FIGURE 6 however, is high computational costs. Assume using
represents an occurrence of extreme event. The a 1-hour segment of wave elevation time series
plots should not be interpreted as time series. screened from a 3-hour time history, the total
Rather, each plot presents a collection of discrete number of time-domain leg touchdown simulations
independent extreme events that could occur for would be 1800 times the number of realizations of
a realization of a given sea state in conjunction each random sea state times the number of sea states
Normalized Zero-Mean FWD Leg Impact Force - Stiff Soil
Seed-1 Seed-2
0.8
Normalized Zero-Mean Impact Force
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
0 200 400 600 800 1000 1200 1400 1600 1800
Load Case No
Normalized Zero-Mean FWD Leg Impact Force - Stiff Soil
Seed-1 Seed-2
2.0
Normalized Zero-Mean Impact Force
1.5
1.0
0.5
0.0
-0.5
-1.0
0 200 400 600 800 1000 1200 1400 1600 1800
Load Case No
Figure 6. Normalized Zero-Mean Maximum Veridical Impact Forces at the Connection of the Forward Leg and the
Spudcan.
under consideration. This leads to a significant floating mode. Since these jackup global motions in
requirement of computing resources as well as efforts free floating mode are governed by wave forces on
for data pre- and post-processing. As such, this the hull, the linear spectral analysis method should
method may not be practical for operational support be adequate for calculating the magnitude of these
or routine engineering design. However, given motions. The design waves can then be determined
the level of details this method can model, it does using the procedure outlined in the ABS MODU
provide a viable means of deriving reference data Rules [10] and as summarized below:
that can help in developing more efficient and
less complex analysis methods that can still yield • Calculate Response Amplitude Operators
comparable results. (RAOs) for each critical response. In the present
study, the critical responses include heave and
Two simplified methods are evaluated in the pitch velocities at the jackup center of gravity
present study. The first one is inspired by the and the velocities at the tip of the spudcan.
‘Design Wave’ approach commonly used in the
design of ships and floating offshore platforms • Select a random sea state and calculate the
[10]
. Another is similar to the approach studied by response spectrum of each critical response
Daun and Olsson (2014) [7], where the legs are
locked to the hull at various prescribed elevations • Calculate the maximum value for each critical
and remain unchanged in each time-domain response using the spectral analysis method. In
simulation. Details of these two simplified methods the present study, the maximum value is chosen
are described in the following two sections. as the most probable extreme value derived from
the response spectrum.
SIMPLIFIED METHOD – DESIGN WAVE
APPROACH • Derive the design wave height curve as a
The basic concept of the design wave approach is function of wave period for each critical response
to identify a number of individual regular waves, by dividing the maximum response by the
i.e. design waves, such that the specified critical RAOs of the critical response
responses of the jackup subjected to these design
waves can match the related extreme responses of • Select design waves in terms of wave height,
the jackup subjected to a given random wave. Once period and direction from the design wave
the design waves are defined, the jackup going on height curves
location analysis can be simplified to a deterministic
regular wave analysis for each design wave. An example of design wave curves is given in
Normally the critical responses are those that are FIGURE 7 for the jackup in a sea state with
directly computable from the global motion analysis Hs = 2 m, Tp = 10s and 180 degree wave heading.
rather than the final target responses for checking The two solid curves are for the horizontal and
against design or assessment criteria. It is essential vertical velocity at the tip of the forward spudcan.
that the selection of the critical responses be based The two dashed curves are for the heave and pitch
on the understanding of correlations between velocity at the jackup center of gravity. The limiting
global responses and a specified target response. wave curve has the wave height Hmax =3.48 m.
It should be noted that, for different target
responses, the correlated global critical responses Also shown in the figure are the ranges of the pitch
could be different as well. and heave natural periods of the jackup in free
floating mode. Applying a regular wave defined by
For the study presented in this paper, the target any point on a design wave curve associated with
response is the vertical impact force at the leg- a specified critical response will generate the same
spudcan interface. The correlated global critical maximum value of that critical response when the
responses are found to be heave and pitch velocities jackup is subjected to the actual random sea state
at the jackup center of gravity as well as the velocities used to derive the design wave curve. Since the
at the tip of the spudcan when the jackup is in free regular waves defined using the points inside the area

bounded by a design wave curve overpredicts the the spudcan-seabed clearance varying along with
maximum value of the critical response, the design the leg movement could introduce additional
waves should be selected on the lowest envelope changes to the natural periods.
of all the design wave curves and the limiting
wave curve. The periods of the selected design After selecting design waves from the design wave
waves should cover the typical wave period range, curves, the jackup going on location analysis is
while more attention needs to be paid to the ranges performed using the same OrcaFlex model as shown
of pitch, heave and roll natural periods. Note that in FIGURE 4. Since the wave conditions used in
the simulation are no longer from the random
waves but the individual deterministic regular
waves defined by the selected design waves, the total
Design Wave Height - 180deg Wave Heading
5.00
number of simulations and simulation time can be
4.50
greatly reduced. Typical vertical impact force time
4.00
histories obtained using the design wave approach
3.50
are shown in FIGURE 8, in which the effect of
using two different phase angles of the design wave
Wave Height (mm)
3.00
2.50
is also illustrated. The maximum impact forces can
2.00
therefore be determined from the time histories
1.50
calculated for all the selected design waves.
1.00
0.50
Heave Pitch SIMPLIFIED METHOD –
Period Period
0.00
DISCRETE LEG ELEVATION APPROACH
5 6 7 8 9 10 11 12 13 14 15
Period (s)
16 17 18 19 20 Given that significant computational efforts
required by the direct time-domain simulation
H(VHFWD) H(VZFWD) H(cg_vz) H(cg_vry) Hmax are mostly inflicted by the need to address the
Figure 7. Design Wave Curves for the Jakcup

(Hs = 2 m, Tp = 10 s, 180 deg Wave Heading).
FWD Leg GZ Force (kN) - Stiff Soil
LC1 - 01 (0 deg) LC1 - 02 (45 deg)

2.0E+04
FWD Leg GZ force (kN)
0.0E+00
-2.0E+04
-4.0E+04
-6.0E+04
-8.0E+04
-1.0E+05
-1.2E+05
-1.4E+05
-50 0 50 100 150 200 250 300 350
t (seconds)
Load Case No
FWD Leg GZ Force (kN) - Soft Soil

LC1 - 01 (0 deg) LC1 - 02 (45 deg)
1.0E+04
5.0E+03
0.0E+00
-5.0E+03
-1.0E+04
-1.50E+04
-2.0E+04
-50 0 50 100 150 200 250 300 350 400 450
t (seconds)
Figure 8. Typical Vertical Impact Forces at the Connection of the Forward Leg and Its Spudcan Calculated Using the
Design Wave Approach (Two phase angles, 0 deg and 45 deg, of the design waves are considered.)
coupling of two sources of randomness (random the mechanism for moving the legs. The analysis
sea and random starting moment of leg lowering), model becomes similar to those used for site-specific
one of the simplification schemes is to approximate assessment for jackups in elevated conditions; the
the continuous leg lowering using multiple discrete main difference is the inclusion of the clearance
leg elevations such that, for each time-domain between the spudcan and the seabed.
simulation, the legs are locked to the hull while
the randomness of sea state is explicitly modeled. FIGURE 9 shows a typical three-hour (with
Multiple simulations with different predefined additional ramp-up time) vertical impact force
discrete leg elevations are required for the same time history for the stiff soil case obtained using
realization of a sea state. It is perceived that, with the discrete leg elevation approach. The extreme
the selection of a sufficient number of discrete value of the impact force associated with a specified
leg elevations and a long enough simulation of statistic metric can be predicted using these time
random wave induced global motions, the effect of histories calculated for each sea state.
changing spudcan-seabed clearance and its interaction
with a random sea state can be suitably modeled. COMPARISONS OF
Recall that the direct time-domain simulation applies CASE STUDY RESULTS
a sliding time window at a two-second interval to The comparison between the direct time-domain
carve out segments of wave elevation time series as simulation approach and the two simplified
the inputs to the simulations of leg touchdown; the methods is carried out using the maximum
discrete leg elevation approach is schematically akin vertical impact force measured at the forward leg-
to applying a sliding observation window at each spudcan interface. The analysis model is shown in
discrete leg elevation while all random sea states run FIGURE 4. One sea state (Hs = 2 m, Tp = 10 s,
through their courses. PM spectrum) at 180 degree wave heading is
applied when using the direct time-domain
Since the continuous leg lowering operation is simulation method and the discrete leg elevation
approximated by multiple discrete leg elevations, approach, while the design waves derived from this
the number of simulations is reduced to the sea state are applied to the simulations using the
number of discrete leg elevations times the number design wave approach. For the direct time-domain
of realizations of each random sea state times the simulation method, 2 random seeds for generating
number of sea states under consideration. In the wave crest elevation time series of the sea state
addition, since the leg lowering does not need to are selected. A combination of 7 random seeds for
be explicitly modeled, the OrcaFlex model in the sea state and 5 discrete leg elevations are applied
FIGURE 4 can be further simplified by removing to the simulations using the discrete leg elevation
FWD Leg GZ Force (kN) - Stiff Soil
2.0E+04
0.0E+00
-2.0E+04
-4.0E+04
-6.0E+04
-8.0E+04
-1.0E+05
-1.2E+05
0 2000 4000 6000 8000 10000
Time (s)
Figure 9. Typical Vertical Impact Forces Calculated Using the Discrete Leg Elevation Approach.

approach. A total of 14 design waves are derived for TABLE 1 shows that, for the very stiff soil case, the
using the design wave approach. maximum values of the vertical impact forces obtained
using two different random seeds have very small
Comparative results for the very stiff and very soft soil variations. The average of maximum values from
cases are listed in TABLE 1 and TABLE 2, respectively, using the discrete leg elevation approach are slightly
in which the maximum values of the vertical impact smaller than those from the direct time-domain
forces are derived using different methods. For each simulation methods, while in general the discrete leg
set of discrete occurrences of maximum vertical elevation approach and the design wave approach give
impact force calculated using the direct time-domain comparable results and show modest conservativeness.
simulation method, the curve fitting for the three-
parameter Weibull distribution is applied to the block For the very soft soil case, as shown in TABLE 2, the
data whose exceedance probability is below 20%. For maximum values of the vertical impact forces appear
the vertical impact force time histories calculated much more sensitive to the change of random seed.
using the discrete leg elevation approach, the Weibull The Weibull curve fitting could sometimes yield a
distribution curve fitting follows the recommended relative low confidence level. See, for example, the
method in ISO 19905-1. case of Seed #1. Similar to the very stiff soil case, the
Table 1. Comparison of Vertical Impact Forces at the Leg-Spudcan Interface – Very Stiff Soil Case
Vertical Impact Force (MN)

Method of Deriving the Direct Time-Domain Discrete Leg
Design Wave
Maximum Value Simulation Method Elevation
Approach
Seed #1 Seed #2 Approach
Maximum Value of All Sampling
106 107 115 119
Point Sets
Average of Maximum Values of
106 103 --
Individual Sampling Point Sets
MPM Values from Weibull
109 109 -- --
Curve Fitting
Average of MPM Values from
109 120 --
Weibull Curve Fitting
Table 2. Comparison Of Vertical Impact Forces At The Leg-Spudcan Interface – Very Soft Soil Case
Vertical Impact Force (MN)

Method of Deriving the Direct Time-Domain Discrete Leg
Design Wave
Maximum Value Simulation Method Elevation
Approach
Seed #1 Seed #2 Approach
Maximum Value of All Sampling
17 15 19 18
Point Sets
Average of Maximum Values of
16 15 --
Individual Sampling Point Sets
MPM Values from Weibull
28* 16 -- --
Curve Fitting
Average of MPM Values from
22* 17 --
Weibull Curve Fitting
Note: Weibull curve fitting has relatively low confidence for the impact forces obtained using random seed #1 in the
direct time-domain simulation
average of maximum values from using the discrete relative to the hull in the time domain simulation of
leg elevation approach are slightly smaller than those the jackup in random seas. The main difference of
from the direct time-domain simulation methods, the three approaches resides in how the randomness
while the discrete leg elevation approach and the of sea state and starting moment of leg lowering is
design wave approach give comparable results and addressed in the analysis.
show modest conservativeness.
Among the three different approaches, the direct
CONCLUDING REMARKS time-domain simulation is considered as a state-
This paper presents a recent development of analysis of-the-art modeling approach for the jackup
methods for jackup units going on location. The going on location; the design wave approach
analysis model as shown in FIGURE 4 is used in is computationally most efficient because all
implementing the direct time-domain simulation the stochastic features are represented through
approach and the two simplified methods. The direct purposefully selected deterministic design waves,
time-domain simulation involves fully-coupled while the discrete leg elevation approach offers a
modeling of jackup’s hydrodynamic responses in straightforward analysis method with a modest
random seas, structural dynamics of legs and leg- requirement of computational effort since only
hull connections, and transient impact of spudcan random waves need to be explicitly modeled in each
on the seabed. The first simplified method follows simulation. Using the maximum vertical impact
the concept of design wave approach where a force at the leg-spudcan interface as an indicator,
random sea is represented by a number of regular the two simplified methods appear to be able to
waves derived to replicate the extremes of selected produce the results that compare reasonably well
global responses that potentially could govern the with those from the direct time-domain simulation.
critical impact responses due to the jackup leg Further verifications using other indicators, such
touchdown. The second simplified method models as the pinion load, and sensitivity analyses with
the leg lowering process using multiple discrete consideration of variations of jackup’s main design
leg elevations, each of which remains unchanged and operational parameters are still required.
AUTHOR’S CONTACT  Michael.Perry@KOMtech.com.sg
REFERENCES
[1] Lai PSK, Lewis TC, Frieze PA, Miller BL, Smith IAA, Limiting Motions for Jackup Moving onto Locations. In: Proceeding, The 4th International Conference on
the Jack-up Platform Design, Construction & Operation, London, UK, 1993.
[2] Miller BL, Frieze PA, Lai PSK, Lewis TC, Smith IAA, Motions and Impact Responses of Jackup Moving onto Location. In: Proceeding of the 25th Offshore
Technology Conference (OTC), Houston, Texas, 1993.
[3] Smith IAA, Frieze PA, Lai PSK, Lewis TC, Miller BL, Evaluation of Leg Damage Risk for Jackups Going on Location. In: Proceeding of the 26th Offshore
Technology Conference (OTC), Houston, Texas, 1994.
[4] Matter G, da Silva R, Tan P, Touchdown Analysis of Jack-up Units for the Definition of the Installation and Retrieval Operational Limits, In: Proceedings of the
24th International Conference on Ocean, Offshore and Arctic Engineering (OMAE 2005), Halkidiki, Greece, 2005.
[5] Chakrabarti P, Going on Location Study for a Jack-up Rig. In: Proceedings of the 31st International Conference on Ocean, Offshore and Arctic Engineering
(OMAE 2012), Rio de Janeiro, Brazil, 2012.
[6] Dong W, Wang J, Song L, Li J, Leg to Seabed Impact Analysis for Jackup during Installation. In: Proceedings of the 23rd International Offshore and Polar
Engineering Conference (ISOPE2013), Anchorage, Alaska, 2013.
[7] Daun V, Olsson F, Impact Loads on a Self-elevating Unit during Jacking Operation. Master’s Thesis, Department of Shipping and Marine Technology,
Chalmers University of Technology, Göteborg, Sweden, 2014.
[8] Butcher EA, Clearance Effects on Bilinear Normal Model Frequencies, Journal of Sound and Vibration. 1999: 224(2):305-328.
[9] ISO 19905-1, Petroleum and Natural Gas Industry – Site-specific Assessment of Mobile Offshore Units – Part 1: Jack-ups, 2012.
[10] ABS, Rules for Building and Classing Mobile Offshore Drilling Units, Part 3 Chapter 2 Appendix 2, 2014.
Design Considerations of DOL
and VFD Jacking System 19
Design Considerations
of DOL and VFD
Jacking System
 Bernard Voon Ee HOW*, PhD, B.Eng

Hai Dong XU*, MSc, B. Eng
Satish MENON*, MSc, B.Eng
Kok Seng FOO*, PhD, B.Eng
* Keppel Offshore & Marine Technology Centre Offshore Technology Development
THE JACKING SYSTEM IS A KEY MISSION CRITICAL AND SAFETY CRITICAL

SYSTEM ON A JACK UP PLATFORM (JUP). Due to its massive hull, a typical
jacking system uses thirty six, fifty four or seventy two motors working in
combination to share the load in elevating or lowering the hull against gravity.
Distinct characteristics for jacking system loads include high inertia, large
variable loads due to different loading conditions of the rig, guide and pinion
friction, cyclic loads due to pinion action, and that the system can experience
instantaneous load change due to the environment.
In this paper, characteristics and peculiarities to be considered for both DOL

and VFD Jacking systems are discussed. Design considerations include start
and stop sequences, interaction of the DOL or VFD drive with the braking system,
load sharing, system reliability, short circuit current ratings and protection, encoder
feedback, shaft voltages, bearing currents, vibrations, cable length and reflected
wave phenomena, thermal insulation and inverter duty motors. Owners and rig
designers should be aware of the tradeoffs and peculiarities when selecting the
type of jacking system to be used in the JUP.
KEY WORDS: Jacking System, Direct-On-Line, Variable Frequency Drives
INTRODUCTION and is not able to hold the gravity load, the rig
A typical JUP has three truss legs with each leg will accelerate and drop downwards into the water
having three mutually parallel chord members. until the gravitational forces are balanced by
Each chord member is provided with a pair of the buoyancy. If the jacking system on one leg
opposing rack members which are engaged via malfunctions, that side of the hull will start to
pinions to four (refer to Fig. 1), six or eight jacking descend causing the rig to tilt and the legs to be
assemblies with gearbox, motor, and brakes. Due bent. This hull drop or tilt could cause economic
to its massive hull, a typical jacking system uses loss or injuries and as a worst case, fatality to
thirty six, fifty four or seventy two motors working onboard personnel.
in combination to share the load in elevating or
lowering the hull against gravity. When designing a jacking system, the characteristic
of the load needs to be considered due to the rig
The jacking system is a key mission and safety structure and the jacking system interaction which
critical system and is responsible for elevating and can contribute to a large variation in loads in each
lowering the hull in a safe and reliable manner. leg, chord and pinion. There can be significant
During jacking, the hull is being displaced difference in loading between pinions at each
vertically. Such loads, also called overhauling instance. A robust jacking system design will need
loads, invite motion of the load due to the effect to consider the effects of these load characteristics
of gravity even at zero speed. In elevated condition, for safe and reliable operation.
if the jacking system malfunctions, loss control
Long-established Direct-on-Line (DOL) is a
Brake
simple and robust method to drive the jacking
Gear box
Motor system. The disadvantage of DOL is the high
inrush current and the mechanical system has
to be designed to withstand the stress and the
torque generated. Variable Frequency Drives
(VFD) jacking systems claim advantages such as
smooth start and stress-reduced jacking due to
the ability to control the current, step-less speed
control and accurate control of the motor torque
and pinion torque. Hence the possibility to optimize
the mechanical system. In fact, if the VFD drive
system could be designed to be robust and reliable,
it could also have the advantage that the hull can
be self levelling and also be able to have better
control of the RPD of the leg [1]. However, in our
investigation, the design and control philosophy is
Pinion
significantly more complex. The system architecture,
type of VFD control, brake control, programming
and setting of the drives have to be considered to
Pinions engage have a safe and reliable system, and does not pose
Chord rack at
Rack
difference phase a challenge in the commissioning and operation of
the system.
In the following sections, the paper will try to

highlight some of the consideration when starting
and stopping the jacking system and the interaction
with the mechanical brakes, load sharing while
Figure 1. (top) Jacking system with four jacking units jacking and considerations related to system
per chord (bottom) Rack and Pinion System. reliability and protection.
START, STOP JACKING OPERATION AND one motor and individual encoder feedback.
MECAHNCIAL BRAKES
DOL JACKING SYSTEM For starting the jacking operation, both V/f and Vector
When starting the jacking operation, pressing the Control are able to offer smooth start by controlling
start jacking button on the jacking console causes the inrush current through adjusting voltage and
the contactors in Fig. 2(a) to be closed. Once the frequency input to the motors. In order for the brake
contactors are closed, full line voltages are applied to open while generating the necessary torque for the
to the motor terminals. To stop jacking, the brakes load to maintain at a stationary position, the start
are sized to stop and hold the hull and the motors sequence has to be parameterized and implemented.
are de-energized by opening the main contactors.
Ref to Fig. 3, using V/f control as an example,
VFD JACKING SYSTEM pressing the start jacking button on the jacking
As there are many types of VFD controls and console will initiate the drive to energize the motor.
configurations available, two sample techniques, As the current is a function of voltage and frequency
with simplified electrical representation, are shown across the motor, a low frequency and voltage can
in Fig. 2(b) VFD: Volts per Hertz (V/f ) Control be set initially to prevent the current inrush. By
with one inverter driving a group of motors and Fig. reducing the current, the torque is reduced. This is
2(c) VFD: Vector Control with one inverter driving also known as smooth start.
Main Switchboard Main Switchboard Main Switchboard
Main Main
Main Circuit Breaker Circuit Breaker
Circuit
Breaker
AC to DC
AC to DC Converter
Converter
Main
Contactor DC bus DC bus
Thermal
Overload
Relay DC to AC DC to AC
Inverter Inverter
M
M M M M
Brake
Encoder
Brake Brake Brake Brake
(a) DOL (b) VFD : V/f contro (c) VFD : Vector Control
Figure 2. Simplified electrical representation of (a) DOL, (b) VFD: V/f Control with one inverter driving a group of
motors and (c) VFD: Vector Control with one inverter driving one motor and individual encoder.
In elevated condition, if the jacking system loss control

and is not able to hold the gravity load, the rig will
accelerate and drop downwards into the water, causing
economic loss or injuries, and as a wost case, fatalities.
When using a VFD to accelerate a load, the ii. If the Brake Release Current is too high, the
amount of torque and therefore current required to drive will work against the brake and the brake
accelerate a load increases as the acceleration time is slips at the start.
reduced. The total amount of necessary accelerating
torque is the sum of the torque required by the load iii. If the V/f setting is too high, resulting in the
plus the torque required to overcome the inertia of torque being too high, the drive will be working
the rotating assembly[4, 5]. During the release of the against the brake and the brake slips at the start.
brakes, if the stator is not sufficiently energized, If the V/f setting is too low, the torque produced
there will not be enough current flowing through in the motor may be insufficient resulting in
the stator. Also, if the current is limited, then there load drop.
is also limited power to overcome the load. In the
case when the hull is elevated, at the instance when Similarly, the stop sequence is dependent on several
the brake is released, and if motors have not built key parameters which need to be tuned:
up sufficient torque, the motor may not be about to
hold the load and the hull may drop. i. If the Slip Prevention Frequency is too high, the
drive works against the brakes and the brake slips
The start sequence is dependent on several key as the drive is stopping. If the Slip Prevention
parameters[2-3] which will need to be tuned, Frequency is too low, the amount of current
specifically: flowing through the motor will be insufficient to
generate enough torque to hold the load.
i. If the Brake Delay frequency or time is set too
low, the brake can be released before the current ii. If the Slip Prevention Time is too long, the drive
flowing through the motor generates enough works against the brakes and the brake slips as
torque to hold the load. On the other hand, the drive is stopping. If the Slip Prevention
if the Brake Delay is set too high, then excessive Time is too short, the drive is ramped down
current at the start may cause a trip. before the brakes are completely shut which
may result in the load dropping.
3P VAC
Drive f
Central ref
V/f Control DC
Control
gate
signals Smooth Start
PWM
Concept
motor overheat sensor

• external
speed / dynamic
torque motor speed feedback overhauling
reference M jacking load
encoder
• Start / Stop
Signals brake signals
brake
Figure 3. Simplified representation of V/f control and considerations.

iii. Similarly, if the V/f setting is too high, resulting of motor parameters is critical as the control method
in the torque being too high, the drive will be uses the drive internal torque reference. Other
working against the brake and the brake slips at parameters for closed-loop vector control include
the start. If the V/f setting is too low, the torque stop timers and torque compensation delay time [2-3].
produced in the motor may be insufficient,
resulting in hull drop. REMARK 2
In UK HSE report [6], two hoisting incidents were
SIMULATION ILLUSTRATION identified attributed to electrical/control system
To illustrate the effects of control parameters on motor issues. In a 2008 Rotterdam incident, the settings of
starting, a squirrel-cage induction motor controlled the sophisticated PLC/inverter motor drive system
by VFD with closed-loop V/f control method is were found to be at fault. An instance of a similar
simulated. The effects of motor acceleration time problem in the U.K. occurred in 2008. In this case
and brake release command trigger time on motor it was the hoist motor drive settings at fault and
starting process are investigated using a simple single the crane could drop a lifted load without
motor single drive system with constant load. From warning. The two incidents are for cranes and
Fig. 4, it is noted the load will drop (negative rotor similar to the jacking system whereby a load is lifted
speed) when the brake release command is triggered. against gravity. It was report that this scenario has
Due to the overhauling type of load the rotor is driven the potential for becoming more prominent as
in the opposite direction and the motor is operating newer systems having ever more sophisticated
in the plugging region. Since the torque is positive control systems come on the market and enter service.
but the speed is negative, the plugging torque appears Much depends on the training/competence of
as a braking torque. This illustration is an example individuals setting up this type of control system and
to show how suitable parameter settings are critical. replacement of spare parts once in service to ensure
If the settings are wrong or can result in the load that the internal settings of all motor drives are
drop for overhauling loads. correct for the application. The settings required
to properly control a motor may well differ from
REMARK 1 the driver manufacturer’s default settings and so
In this section, V/f control is discussed. However, adjustment during initial erection/commissioning
similar consideration applies to vector control and/or spare part replacement exercises may well
schemes. For vector control schemes, accurate tuning be required.
40 160
Motor Electromagnetic Torque (% of Nominal Torque)
140
Rotor Speed (% of Synchronous Speed)
20
120 Brake Release Command Triggered at 0.5s
Brake Release Command Triggered at 1s
Brake Release Command Triggered at 1.5s
0 100
Brake Release Command Triggered at 0.5s
Brake Release Command Triggered at 1s
Brake Release Command Triggered at 1.5s 80
-20
60
-40 40
20
-60
0
-80 -20
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Time (s) Time (s)
Figure 4. (left) Motor Speed and (right) torque with Brake Release Command at different timing.
JACKING AND LOAD SHARING VFD WITH TORQUE FOLLOWER / SPEED

As the motors are mechanically coupled through TRIM FOLLOWER
the racks and pinions, each motor must contribute From Fig. 5(b), the master drive is operated in speed
its proportional share of power to the driven load. control mode. The follower drives are operated
Load sharing using DOL and VFD, i.e. droop in torque control mode. In speed trim follower
control, torque follower, and speed trim [7-9] will be configuration as shown in Fig. 5(c) and 5(d), the
discussed. master drive is operated in speed control. The
follower drives are operated in speed control mode
DOL PASSIVE LOAD SHARING with a speed trim using the torque reference.
For speed regulation and load sharing, the DOL
system is passive as it depends on the natural The master and follower drives will have to be
characteristics of the motor and the designed slip to selected. The consideration is that the torque
regulate the torque and speed reference to the follower drives need to consider the
effects of variation in load between pairs of pinions,
VFD WITH DROOP CONTROL pinions along a chord, and also between legs, the
Droop Control load sharing operates similar to phase difference due to the mechanical offset and
DOL. All drives use a common speed reference as also instantaneous load change. For the follower
shown in Fig. 5(a). A droop can be program into the motors, due to variation in loads between pinions,
drives to allow the higher loaded motor to slip. If if the load decreases, the torque reference can causes
all drives are set to droop, then all drives may share the motor speed to increase rapidly. The design also
load by reducing speed. This could prevent the has to consider the interaction of the load sharing
system from operating with a rigid speed regulator algorithm with the start and stop phases of release
and may result in poor speed regulation. In a master and engagement of the mechanical brake. For
and follower configuration, if the master drive is not example, when starting jacking, at zero speed, the
allowed to droop, all follower drives will be “pulled” master drive torque could be low due to the uneven
by the master [9]. distribution of loads, as such, the torque reference
set point sent to other drives could be insufficient to
hold the load while the brakes are released.
Follower 1 Follower 2 Master Follower 1
Speed Speed Torque Torque

Reference Reference
Reference Reference
(a) (b)
Master Follower 1 Follower 2 Master Follower 1 Follower 2
Torque
Torque Torque Torque Torque
Reference Reference Reference Reference
Speed
Reference Speed
(c) Reference (d)
Figure 5. (a) Speed reference with droop, (b) torque follower, (c) and (d) speed trim follower.
For the speed trim, filtering may be required to at the point where a VFD is connected to electrical
soften the reaction of the trim regulator. However, power can have a significant impact on the safety,
if the filter is not properly tuned, the output signal longevity, and cost of a VFD installation.
will rapidly and continuously move between the
plus and minus limits causing instability [9]. Also important is the impact of aging and power
supply harmonics on the DC bus capacitors. As
SYSTEM RELIABILITY AND PROTECTION evidenced in the catastrophic capacitor failure in
SYSTEM RELIABILITY cruise ship RMS QM2 [12], the long term degradation
Referring to Fig. 2, for the DOL system, the and eventual failure of the VFD may lead to shut
protection is simple with the main circuit breaker, down of surrounding switchboards. Accessibility to
main contactor and the thermal overload relay. With the health status DC link capacitors are important
the addition of more components into the critical mitigation steps in the failure analysis of the VFD.
path for the VFD jacking system, such as the AC
to DC converter, the DC Bus, DC to AC inverter ENCODER FEEDBACK
and the encoder, the system reliability is expected Within a single-encoder system, the possible fault
to decrease. The mean time to failure (MTTF) for include (i) a single electrical fault in the encoder,
each component will have to be considered and the and (ii) a break of the encoder shaft (or loose
reliability and availability of the system has to be encoder shaft coupling), or a loose encoder housing
assessed. which will cause a static state of the encoder signals
(that is, they no longer follow a movement while
SHORT CIRCUIT CURRENT RATINGS AND still returning a correct level), and prevent fault
PROTECTION detection while the drive is in stop state.
For VFD, It is estimated in [10] that about 38% of
the faults in inverter systems are due to failures of As the drive is likely to be operated under closed
power semiconductor devices such as the insulated loop control, the risk analysis of the machine
gate bipolar transistor (IGBT), which is mostly used manufacturer must include all of the faults
in the inverter. Its failures can be broadly categorized described above. Additional safety measures have
as short-circuit and open circuit faults in the power to be taken at suspended/vertical drives or for
devices. Open circuit faults occurs mainly due to handling overhauling loads in order to exclude
the lifting of a bonding wire caused by thermal faults [13]. In a jacking system, which uses thirty six
stress or by an extremely high collector current and and up to seventy two or more motors, inverters and
does not cause serious damages. On the other hand, encoders operating in unison but with each inverter
the short circuit failure in the power devices can be operating on individual encoder speed feedback, an
catastrophic because the failure propagation to the incorrect signal fed back from one encoder to one
complementary power device of same inverter pole drive may result in that drive motor operating in
occurs in less than 10 μs. a “tug of war” with other motors and drives. If a
two-encoder system per motor is implemented, the
Short-circuit current ratings (SCCRs) for VFD is encoders may not be mounted on the same shaft.
a topic that has been discussed often without The algorithm for decision making and response
clarity [11]. Some manufacturers provide SCCRs needs to be considered in the case where the two
based on testing only the output section of VFDs. encoders return different values.
While this method of test may be suitable for
DOL starters used for motor control, the strictest SHAFT VOLTAGES AND BEARING CURRENTS
interpretation of applicable standards conducts In many VFD driven motors, shaft currents have
SCCR tests based on the most likely failure points been found to discharge through the bearings,
in the VFD, which is not the output section. breaking down the bearing grease and causing a
Also, due to the electronic nature of VFDs, their severe wear pattern in the bearing called "fluting."
characteristics may change depending on their If allowed to continue, these shaft currents will lead
electrical power system connection. The level of to high motor vibration levels and eventual bearing
prospective short-circuit current (PSCC) available failure [4, 5].
VIBRATIONS VBUS
Due to the complexity of the mechanical system,

the regulators within the VFD controlling speed OV
over a range, flux or current generated could -VBUS

interact with torsional oscillations and cause
2VBUS
amplification. The motor and mechanical system
may possess several natural frequencies that may
be excited by the control when operating within OV
a particular frequency range [4]. These natural
frequencies may exist in the horizontal, vertical,
axial, or torsional directions and are affected by
factors such as the mass and stiffness of the motor -2VBUS
base, the type of control, the natural frequencies Time

on the mechanical structure, motor structure, the 160
electromagnetic design of the motor and coupling 150
vibration. A case study in [14] described a multi- 140
motor load sharing system which experiencing 130
severe vibrations at certain speeds and loads. It

Temp Rise [ºC]
120
should be noted that, since mechanical vibration 110
and interaction frequencies are different in every 100
90
installation, site-specific adjustments of the control
80
software could be required.
70
60
CABLE LENGTH AND REFLECTED WAVE
50
Location of the MCC where the drives are installed 0 10 20 30 40 50 60 70
and motors needs to be considered in the design. Freq [Hz]
The cable length needs to be considered due to the Figure 6. (a) PWM Voltage at Drive and Motor Terminals
reflected wave phenomena which causes voltage and (b) Motor Temperature Rise on Sine Wave
overshoots at the motor terminals as shown in and Inverter Power.
Fig. 6(a) [15]. Significant damage to the motor
insulation can occur if these overshoots are greater The motor’s insulation system is what makes a
than the maximum rated voltage of the motor [4]. motor capable of operating with an inverter power
supply. For VFD, motor manufacturers have long
THERMAL INSULATION AND INVERTER DUTY recognized that 3-phase ac squirrel-cage induction
MOTORS motors that are fed by PWM inverters experience
It should be recognized that motors generate higher dielectric stresses than do equivalent line-
substantial amounts of heat during operation. fed motors. The fast switching transistors used in
This heat must be dissipated to maintain cool modern PWM inverters have helped to produce
motor operation and motor longevity. Motor more sinusoidal currents as well as more efficient and
operation at reduced speeds results in reduction of compact inverters. The fast transitions (both turn-
fan cooling effectiveness which can cause the motor on and turn-off ) of the transistors helped to enable
to heat excessively and bring about premature these improvements but also create higher stresses in
motor failure. Applications that require extremely the motor insulation. These higher stresses are seen
slow (below 6 Hz) or extremely high speeds may phase to phase, phase to ground, and turn to turn.
require a custom motor design [4]. Harmonics due
to short rise times of the synthesized waveforms Fig. 6(b) shows that a motor, at rated frequency
add to motor heat dissipation requirements. Short and load, under inverter operation may have a
rise times can also cause high voltage spikes that 10°C higher temperature rise than the same motor
can damage the motor insulation. on sinewave power [14, 16]. Constant Torque (CT)
type loads further increase motor temperature rise sharing, system reliability, short circuit current
in Fig. 6(b) as drive output frequency, and thus ratings and protection, encoder feedback, shaft
motor speed, is reduced. This is due to reduced voltages, bearing currents, vibrations, cable length
motor cooling since the internal motor fan speed is and reflected wave phenomena, thermal insulation
also reduced. At some frequency an external fan is and inverter duty motors.
required for CT loads. Variable Torque (VT) rated
loads usually have maximum temperature rise at The DOL jacking system is a simple and robust
rated load and frequency. system. The key design consideration of the DOL
system is on the mechanical system to cater for the
The failure mechanism associated with these higher high current inrush and the starting torque. The DOL
stresses is primarily related to partial discharge (PD) system is mechanically oversized. For VFD jacking
activity, also known as corona inception. This is a system, there can be advantages such as smooth
type of localized emission resulting from transient start, step-less speed control and optimization in the
gaseous ionization in an insulation system when the mechanical system. The tradeoff is in the increase in
voltage stress exceeds a critical value. The damage is complexity, including the system architecture, type of
a cumulative process, much like fatigue failures on a VFD control philosophy, programming and setting
mechanical component. of the drive parameters. Other complexities such as
system reliability due to an increase of components
CONCLUSION in the critical path, vibrations due to operation over
In this paper, characteristics and peculiarities to a speed range, thermal insulation, cable lengths,
be considered for both DOL and VFD jacking commissioning, testing and maintenance have to
systems have been discussed. Design considerations be considered. The design considerations need to
include interaction of the DOL or VFD drive with be addressed to further enhance jacking system
the braking system, start and stop sequences, load robustness and safety.
AUTHOR’S CONTACT  bernard.how@komtech.com.sg
REFERENCES
[1] K. S. Foo, C. B. Tay, “Self-regulating jacking system”, US Patent 7,186,061 B2, filed 15 Sep 2005, issued 6 Mar 2007.
[2] Schneider Electric, Altivar 71 Variable Speed Drive for Asynchronous Motors Programming Manual, Software V1.2.
[3] Yaskawa AC Drive A1000, Crane Software Application Manual, Software No. VSA90507X.
[4] D. M. Bezesky and S. Kreitzer, “NEMA Application Guide for AC adjustable speed drive systems", IEEE Industry Applications Society 48th Annual Petroleum
and Chemical Industry Conf., 2001.
[5] National Electric Manufacturers Association, Application Guide for AC Adjustable Speed Drive Systems, NEMA Standards Publication, 2001.
[6] Health and Safety Executive, “Tower crane incidents worldwide”, RR 820, 2010, UK
[7] N. Mitrovic, et. al., “Electrical Drives for Crane Application”, Mechanical Engineering, InTech, 2012.
[8] Allen-Bradley, Load Sharing for AC Drives, Rockwell Automation.
[9] Allen-Bradley, Load Sharing for the 1336 IMPACT AC Drive, Rockwell Automation.
[10] M. A. Rodríguez, et. al., “Fault Detection for IGBT Using Adaptive Thresholds During the Turn-on Transient, IEEE Transactions on Industrial Electronics, pp.
1975-1983, Vol. 62, No. 3, Mar 2015.
[11] Schneider Electric, Variable Frequency Drives and Short-Circuit Current Ratings, Data Bulletin 8800DB1203, Mar 2013.
[12] Sinamics S120 Safety integrated Function Manual, Apr 2014.
[13] Marine Accident Investigation Branch, Catastrophic failure of capacitor in the aft harmonic filter room on passenger cruise ship RMS Queen Mary 2, Report
28/2011, GOV.UK, Dec 2011.
[14] G. L. Skibinski et. al., “System design of adjustable speed drives, Part 1”, pp. 47-60, IEEE industry Application Magazine, Aug 2012.
[15] G. L. Skibinski, “Installation considerations for IGBT AC Drives”, IEEE Textile, fiber and film conference, Charlotte, NC, May 1997.
[16] J. Andreas, Energy Efficient Electric Motors - Selection and Application, Marcel Dekker.
This paper has been presented at the International Conference: The Jack-Up Platform, September 2015, City University
London, UK.
The Impact of Refrigerant Selection and Precooling
Condition on LNG Production Robustness and Efficiency 29
The Impact of Refrigerant

Selection and Precooling
Condition on LNG Production
Robustness and Efficiency
 Erine TEH Siew Pheng*, Ph.D.

Ravindu ATAPATTU*, B. Eng., M.Sc.
* KOMtech / Keppel Gas Technology Development Pte. Ltd.
KEPPEL’S PreNEx (PRE-COOLED NITROGEN EXPANSION) IS A

PROPRIETARY NATURAL GAS LIQUEFACTION TECHNOLOGY USING
THE WIDELY ACCEPTED NITROGEN EXPANSION SYSTEM. It is being
deployed by Gas Technology Development (GTD) as part of Keppel’s move
in gas strategy. It employs a single Nitrogen Expansion cycle together with
an independent bi-phasic non-flammable refrigeration cycle for pre-cooling
to achieve comparable liquefaction efficiency to that of multiple expander system.
Its power efficiency and better safety level makes it competitive to Mixed
Refrigerant Process (MR) which uses hydrocarbon refrigerants. The nitrogen
refrigeration loop makes use of the standard and proven compressor and
expander; there are no intermediate gas-liquid separators or pumps. Its simplicity
enables conversion of retired LNG carriers with advantages in cost and lead time.
The overall improvement in the efficiency of PreNEx has to be demonstrated and

educated to the market in order to actually benefit the clients. Major factors that
impacts the liquefaction performance have been discussed in previous works
where the feed condition, ambient condition have demonstrated critical impact in
optimizing this simple and robust technology to its highest efficiency. This study
intends to demonstrate how the choice of refrigerant and the precooling condition
is studied and designed by KOMtech. This effort puts PreNEx at the advantaged
position as the technology provider in natural gas liquefaction.
The physical, safety and environmental characteristics of the refrigerants are

researched and summerised in this study. Four commonly used refrigerants are
selected for the subsequent performance study. The simulations are based on
nitrogen compressors powered by GE LM2500 gas turbine, under same feed
and process conditions, with differences in the process configuration.
It compares PreNEx alongside various nitrogen expansion cycle configurations,
focusing on the heat flow distribution, production capacity, thermal efficiency
and feed shrinkage.
INTRODUCTION turbine, which is a highly efficient and compact

Keppel’s Precooled Nitrogen Expansion aero-derivative turbine driver proven in both onshore
Process PreNEx and offshore applications. This configuration avoids
Keppel's leadership in offshore and marine industries complexity and provide for repeatable design.
marks a new milestone with the strategy to provide Figure 1 shows the schematic diagram of PreNEx,
solutions for the global LNG market. The newly setup which utilizes the simple single nitrogen expansion
Gas Technology Development Pte. Ltd. (GTD) in process with an additional refrigeration system as
2015 is geared to provide robust solutions that can a precooling stage. Major factors that impacts the
effectively bring natural gas related products from liquefaction performance have been discussed in
the source to the users. This requires technology previous work [3, 4, 5], this study aims to demonstrate
development and deployment in the areas of gas the choice of HFC refrigerant and its impact on
liquefaction, storage, transportation, gasification, Liquefaction performance.
power generation and distribution.
Refrigeration Systems and Selection
One of the technologies deployed by GTD currently of Refrigerants [6]
is the PreNEx process developed by KOMtech. Refrigeration systems are used in the natural gas
The Precooled Nitrogen Expansion process, liquefaction process to remove heat from the feed
PreNEx [1, 2] is an improved version of the nitrogen stream, with the purpose to condense the feed gas
expansion liquefaction technology. It is ideal for to liquid phase at cryogenic condition of less than
small to mid-sized FLNG solutions and offers a -150 °C and near atmospheric pressure. Compared to
faster and more cost effective liquefaction solution to the highly pressurised CNG (compressed natural gas
Keppel’s clients. While preserving the use of a single at 150 to 200 bar), LNG (liquefied natural gas) has
nitrogen expander, the efficiency is improved via 2.4 times higher energy density that promises
the bi-phasic hydrofluorocarbon (HFC) refrigerant. extended usage and reduced refueling frequency.
HFC refrigeration is commonly used throughout the Besides delivering the natural gas in its physical
offshore and gas processing industries in hydrocarbon condition as per specification, refrigeration is also
dew point control applications. The refrigerant crucial in removing condensate and certain impurities
compressors is driven by a single GE LM2500 gas in order to meet the LNG product specification.
The LNG is required to have a certain composition
range to deliver homogeneous heating value during
downstream regasification. Accumulation of heavy
hydrocarbons and impurities will also cause freezing
and choking along the process and transportation.
The refrigeration systems found in gas processing

plants are mostly based on the Vapour Compression
Cycle and Absorption Cycle. Besides Natural Gas
Liquefaction, processes that require refrigeration
are Natural Gas Liquid extraction (NGL), dew
point control, LPG storage and condensation for
demethanizer and deethanizer. Keppel’s PreNEx
employs basically the vapour compression cycle but
does not rule out the use of absorption cycle which
recycles waste heat from other topside modules.
An example of a single stage vapour compression

type refrigeration system is found in Figure 2. The
Pressure-Enthalpy diagram illustrates its physical
state at different refrigeration point. With careful
Figure 1. PreNEx schematic diagram. [1, 2] heat load and flowrate design, the refrigerant enters
he Impact of Refrigerant
Review
Selection
of Riser
andTechniques
Precooling
Condition on LNG Production Robustness
for DeepwaterandApplication
Efficiency 31
Figure 2. Ideal Vapour Compression Refrigeration Cycle and Pressure-Enthalpy Diagram.
the evaporator as a saturated liquid-vapour mixture

(state 1 of enthalpy diagram). At the evaporator the
refrigerant boils when it absorbs heat from the process
stream and it enters the compressor as saturated
vapour (state 2). The refrigerant vapour is superheated
and pressurized after compression (state 3). It will be
cooled in the condenser to remove the sensible heat
and eventually the latent heat to arrive at liquid state
(state 4). The cooling medium temperature must
be able to cool the pressurized vapour below its Figure 3. Ideal Aborption Refrigeration Cycle.
saturation temperature. The saturated liquid leaves
the condenser and expanded to near atmospheric consideration the necessary temperature approach
pressure via the expansion valve before arriving at the for the heat exchanger/evaporator to be effective. For
evaporator again. This gives a saturated liquid-vapour cryogenic application below -100 °C, few gases can
mixture which is at its boiling point. The refrigerant be utilized such as nitrogen, oxygen, helium, and
is selected based on the boiling point at this state methane.
which normally ranges from 5 to -100 °C, according
to the refrigeration condition required by the process. The refrigerant selection takes into account the
efficiency and meeting process requirement, as well
Absorption Cycle is less common but the principle as the regulations and hazardous characteristic.
is similar to Vapour Compression Cycle where the Process and efficiency considers the coefficient of
compressor is replaced by the absorber-generator in performance (COP), boiling and freezing point,
Figure 3. Heat source is used to generate high pressure evaporator and condenser pressure, and temperature
refrigerant vapour from the absorption liquid. The difference between liquid saturation temperature and
saturated vapour is then cooled by the condenser to the condensing medium temperature. The physical
produce saturated liquid which will be throttled to properties, performance, safety and environmental
near atmospheric pressure and forms low temperature characteristics of selected refrigerants are tabulated
boiling liquid-vapour mixture. When waste heat is for reference of this study, it can be found in
available i.e. from the exhaust of power generating Table 1 and Table 2.
equipment or combustion engine, the less efficient
Absorption Cycle can be justified on a cost basis. The coefficient of performance (COP) of a Vapour
Compression Cycle is the ratio of the heat taken up
LNG liquefaction process separates the refrigeration by the refrigerant from the process stream (Q Evap,
circuits according to the temperature zone. The in) to the compressor work (WC),
temperature zones are generally categorised as h2 – h1
above -50°C, -50°C to -100°C and below -100 °C. COP = QEvap,in / WC =
h3 – h2
The refrigerants selected shall have the matching
temperature zone as boiling points, taking into (Figure 2, P-H diagram)
Table 1. Physical properties, safety and environmental characteristics of selected refrigerants [6]
ASHRAE Name Chemical Molecular Boiling Pt. Freezing ODP Safety

Designation Formula Weight (°F) at Pt. (°F) Group
14.7 psia
R-11 Trichloromonofluoromethane CCl3F 137.37 74.87 –168.0 1 A1
(CFC)
R-12 Dichlorodifluoromethane CCl2F2 120.93 –21.62 –252 1 A1
(CFC)
R-22 Chlorodifluoromethane CHClF2 86.48 –41.36 –256 0.055 A1
(HCFC)
R-123 2,2-dichloro-1,1,1- CHCl2CF3 152.93 82.17 –160.87 0.02 A1
(HCFC) trifluoroethane
R-134a Tetrafluoroethane CH2FCF3 102.3 –15.08 –141.9 0 A1
(HFC)
401A R-22/152a/124 (53/13/34) 94.44 –27.6a — 0.037 A1
(Blend)
404A R-125/143a/134a (44/52/4) 97.6 –51.7a — 0 A1
(Blend)
407C R-32/125/134a (23/25/52) 86.2 –46.5a — 0 A1
(Blend)
410A R-32/125 (50/50) 72.56 –58.9a — 0 A1
(Blend)
R-32 Difluoromethane CH2F2 52.02 –61.1 –213 0 A2
(HFC)
R-124 Chlorotetrafluoroethane CHClFCF3 8.26 –326.47 252.5 0 A1
(HFC)
R-125 Pentafluoroethane CHF2CF3 120.03 –55.43 –153.67 0 —
(HFC)
R-143a 1,1,1-trifluoroethane CF3CH3 84.04 –53.3 — 0 —
(HFC)
R-152a (HFC) Difluoroethane CHF2CH3 66.05 –13.0 –178.6 0 —
R-717 Ammonia NH3 17.03 –28.0 –107.9 0 B2
(Inorganic)
R-744 Carbon Dioxide CO2 44.01 –109.2 –69.9 0 A1
(Inorganic)
R-170 Ethane C2H6 30.07 –127.85 –297 0 A3
(Hydrocarbon)
R-290 Propane C3H8 44.1 –43.73 –305.8 0 A3
(Hydrocarbon)
Compound: chlorofluorocarbons (CFC), hydrogenated halocarbons (HCFC), hydrofluorocarbons (HFC)

Safety: Toxicity - “A” (non toxic) and “B” (toxic). Flammability - “3” (highly flammable), “2” (moderately flammable)
and “1” (non flammable).
ODP: Ozone Depletion Potential (1 is highest, 0 is none)

Table 2. Performance Characteristics of selected refrigerants [6]
ASHRAE Name Chemical Evap. Condens. COP

Designation Formula Pres Pres (psia)
(psia)
R-11 (CFC) Trichloromonofluoromethane CCl3F 2.9 18.3 5.02a
R-12 (CFC) Dichlorodifluoromethane CCl2F2 26.5 108 4.75a
R-22 (HCFC) Chlorodifluoromethane CHClF2 43 172.9 4.67a
3.898b
R-123 (HCFC) 2,2-dichloro-1,1,1- CHCl2CF3 2.3 15.9 4.84a
trifluoroethane
R-134a (HFC) Tetrafluoroethane CH2FCF3 23.8 111.6 4.41a
3.993b
410A (Blend) R-32/125 (50/50) CF3CH3 3.703b
R-32 (HFC) Difluoromethane CH2F2 3.701b
R-125 (HFC) Pentafluoroethane CHF2CF3 58.9 228.1 3.67a
R-717 (Inorganic) Ammonia NH3 34.2 168.8 4.77a
R-744 (Inorganic) Carbon Dioxide CO2 332.4 1045.4 2.81a
R-170 Ethane C2H6 236.4 674.7 2.72a
(Hydrocarbon)
R-290 Propane C3H8 42.4 156.8 4.41a
(Hydrocarbon) 3.929b
a
[6]
b
[7] Table 5 performance with condensing temperature 38 °C.
Figure 4 . Schematic Diagrams of Precooling and Liquefaction Circuits – Precooler in the external and internal of MCHE.
Hazard evaluation considers the toxicity and maintenance of the liquefaction. Temperature run in
flammability of the refrigerant used, and the more the MCHE can potentially freeze the refrigerant and
recent environment impact that is characterized by impurities to be removed by the precooler.
the ozone depletion potential (ODP) and the global
warning potential (GWP). According to ASHRAE COMPARISON STUDY ON
Standard 34-1992, toxicity has two levels, “A” (non REFRIGERATION SYSTEM AND
toxic) and “B” (toxic). The flammability is divided into NITROGEN EXPANSION PROCESSES
three classes: “3” (highly flammable), “2” (moderately Assumptions and Conditions used
flammable) and “1” (non flammable). ODP is the The simulations use treated natural gas with
relative capability of the refrigerant to destroy ozone composition: 96.7% C1, 2.4% C2, 0.4% C3, 0.18%
in the atmosphere of the earth. The ODP for R-11 C4, 0.07% C5+, 0.25% N2 and 30ppm CO2. Aspen
is defined as 1.00. Compounds with significant HYSYS® is used for simulation and optimization
ODP such as the chlorofluorocarbons (CFC) and of the LNG process and the Peng-Robinson fluid
hydrogenated halocarbons (HCFC) have been or package is used as the Equation of State (EOS).
will be phased out due to their role in depleting The base feed gas pressure is set at 65barg and at an
the ozone layer. The survival time of a CFC in the ambient temperature of 15°C has been considered.
atmosphere is very long as a result of its high chemical The process stream temperature after the compressor
stability. This permits the CFC molecules to rise inter-coolers and after-coolers are assumed to be
to the upper atmosphere and move through the ozone 10°C above the cooling water temperatures. The
layer. The ultraviolet radiation from the sun causes minimum approach of the MCHE is kept at 2°C.
the chlorine atom to break free and chemically reacts First part of the study is based on cases having the
and breaks down the ozone layer. The stable chlorine same feed gas flow rate and produces same train
compound continues the damage to subsequent capacity of LNG production at 0.5 MTPA. Second
ozone molecules. The substituting compounds part of the study is based on simulations whereby the
made available are the hydrofluorocarbons (HFC). main refrigerant compressor train is driven by a gas
GWP is the capability of the refrigerant to warm turbine (GE LM2500) with available shaft powers
the atmosphere of the Earth relative to that caused of 28 and 29MW for ambient temperatures 30°C
by carbon dioxide over a period of 100 years. and 15°C respectively. Shaft power of the nitrogen
Chemicals in the atmosphere have a unique average compressors and refrigerant compressors are tabulated
residence time which in turn influences the GWP. and analyzed for the liquefaction efficiency, feed
shrinkage due to liquefaction power requirement.
Precooling and Nitrogen Based
Expansion Technologies Simulated Processes and Configurations
The characteristics of Nitrogen expansion The first part of this study is to compare the impact
technologies and the performance of the various of refrigerant selection and precooling temperature
configurations have been discussed in previous on the liquefaction performance. It is based on the
Technology Review Paper [4, 5]. This paper intend configuration of PreNEx where an external precooler
to focus on the impact of precooling and the to a single nitrogen expansion cycle (Figure 4A). The
configurations of precooler within the Natural Gas process is simulated to produce 0.5MTPA LNG at
Liquefaction process. Figure 4 shows two simplified ambient condition of 15 °C. The refrigerants selected
ways of incorporating precooling step into the for this study are: R134a, R410A, Propane and
Natural Gas Liquefaction process. PreNEx prefers CO2. These refrigerants have zero Ozone Depletion
the external refrigeration system (Figure 4A) for Potential (ODP). The properties of these refrigerants
simplicity in condensate handling and robust process are extracted from the introduction and summarized
control in the Main Cryogenic Heat Exchanger in Table 3.
(MCHE). Certain technology provider incorporate
the precooling circuit into the MCHE to maximize In the second part of the study, four nitrogen
heat exchanger across the entire temperature zone, expansion processes and one single mixed refrigerant
as well as to enhance insulation with centralized process were selected for this comparison study with
cold box. This may complicate the operation and PreNEx process (Figure 4). It aims to evaluate the
Table 3. Physical, performance and safety properties of the refrigerants selected for comparison study
ASHRAE Name MW Boiling Pt. Freezing Safety Evap. Condens. COP

Designation (°F), Pt. (°F) Group Pres Pres
14.7 psia (psia) psia)
R-134a Tetrafluoroethane 102.3 –15.08 –141.9 A1 23.8 111.6 4.41a
(HFC) 3.993b
410A R-32/125 (50/50) 72.56 –58.9a — A1 3.703b
(HFC Blend)
R-32 Difluoromethane 52.02 –61.1 –213 A1 3.701b
(HFC)
R-125 Pentafluoroethane 120.03 –55.43 –153.67 — 58.9 228.1 3.67a
(HFC)
R-744 Carbon Dioxide 44.01 –109.2 –69.9 A1 332.4 1045.4 2.81a
(Inorganic)
R-290 Propane 44.1 –43.73 –305.8 A3 42.4 156.8 4.41a
(Hydrocarbon) 3.929b
a
[6]
b
[7] Table 5 performance with condensing temperature 38 °C.
(5A) Single N2 Expander Process (5B) Dual N2 Expander Process [9]
(5C) Propane Precooled Single N2 Expander Process (5D) Single Mixed Refrigerant Process
Figure 5. Schematic diagram of the selected Liquefaction processes studied.

effectiveness of precooling in improving the thermal hydrocarbon based liquefaction processes. Topside
efficiency of the LNG liquefaction in a single nitrogen with Hydrocarbon based process would require all
expander process. Figure 5 shows the selected process equipment to be classified for hazardous areas.
configurations modeled based on their respective
patents and from Marak’s study [8]. The conventional A similar feed gas condition is applied to all the cases
single nitrogen expansion configuration (Figure 5A) including PreNEx. The results from the simulations
serves as a baseline reference. The dual expansion are compiled and analyzed against published results of
process [9] in Figure 5B has two Nitrogen streams of the processes to validate the model and assumptions.
different temperature at the inlet of the two nitrogen
expanders. This allows two temperature zone in the RESULTS AND DISCUSSIONS
cold composite which closes the temperature gap The performance in terms of energy efficiency of
with the hot composite. The Propane precooled the four precooling refrigerants is summarized in
configuration is also studied and presented in Figure Table 4 where the specific power and COP are
5C, to demonstrate the effectiveness of an external compared. COP is taken as precooler thermal duty
Propane precooler to a single Nitrogen expander over precooler compressor shaft power (Figure 7).
system. It has the same configuration as PreNEx These refrigerants are not able to function across -20
but using Propane as refrigerant, with a precooled °C to -40 °C. For R134a, the saturation properties
temperature of -34 °C. For comparison purpose, only allows it to function not lower than -20 °C.
the PreNEx process is also simulated at -34 °C Propane works till not lower than -30 °C. R410A
(comparison with Propane) and -45 °C (optimized and CO2 could be used to chill the process until
design condition) using R410A as refrigerant. The -45 °C. Hence the comparison can only be done
use of a flammable refrigerant such as propane is at -20 °C and -30 °C, with exception for R134a
disadvantageous compared to the non-flammable where performance data can only be obtained at
R410A. This study is also intended to compare -20 °C. A correlation for Liquefaction specific power
PreNEx process with Single Mixed Refrigerants is observed when the data is analysed in the function
(SMR) process (Figure 5D), which is the most of COP and precool temperature. The correlation
compact and relatively efficient version of all the is presented in Figure 6 and it shows that higher
Table 4. Performance by refrigerant and precool temperature for 0.5MTPA LNG at 15°C ambient
Refrigerant R134a Propane R410A CO2 Precool

Temperature
Specific Power (kW.h/kg) 0.419 0.421 0.422 0.429 -20 °C
COP 2.54 2.16 1.49
Specific Power (kW.h/kg) 0.404 0.406 0.414 -30 °C
COP 2.16 2.03 1.49
Stringent refrigerant selection and process conditions

in precooling stage can improve the overall liquefaction
process efficiency significantly. The selection of precooling
refrigerant is a tri-angular exercise where the physical
properties has to fulfill the required precooling process
condition. The energy efficiency, safety and environmental
impact should also be taken into consideration to
result in an efficient, robust and safe process, such as
Keppel’s PreNEx.
COP gives better efficiency, which improves with Thermal Efficiency, Thermal Duty,
lower precool temperature. COP is also related to Power Consumption and Shrinkage
the molecular weight as shown in the introduction The objective of the second study is to demonstrate
(Table 2). R134a has the highest COP and is the that PreNEx, which uses the single nitrogen expander
most efficient for a -20 °C system. For process and an external precooler, has comparable or even
that requires lower precooling such as -30 °C to higher thermal efficiency to other multi-expander
-40 °C, R410A is preferred over Propane because of systems, without compromising the simplicity and
its non-flammability. CO2 is not a desired refrigerant robustness of single N2 expander process.
due to its low COP hence low efficiency. In terms
of operating condition, Table 3 lists the evaporator The selected processes are first simulated for a
and condenser pressure required for the refrigerant to fix train capacity of 0.5 MTPA. The compressors
function well. The condenser for CO2 system runs at shaft power for precooling refrigerant and main
70 bar pressure, which is five times higher than other refrigerant are compiled and presented in Figure 8
systems and this would require the use of special as the liquefaction power consumption. The main
high pressure barrel type casing for the compressor. refrigerant compressor power requirement reduces
This basic study explains the choice of R134a as the process evolves. Single expander consumes
and R410A of PreNEx liquefaction technology. up to 31 MW and it drops drastically to around
23MW for dual expander and externally precooled
processes i.e. PreNEx. SMR consumes 20MW to
produce 0.5 MTPA of LNG. The external precoolers
are powered by electricity generated onboard,
which requires 2-3MW of power depending on
the thermal duty. Figure 9 shows the specific power of
the liquefaction technologies and its corresponding
shrinkage as liquefaction fuel. Shrinkage is defined
as the fuel gas supply to power the liquefaction
and precooling compressors over the feed gas used
for the LNG production. It has considered ~35%
of fuel to shaft power conversion efficiency. The
single N2 expander has a specific power of 0.5 kW.h/
kg and 9% shrinkage. Dual Expander, Propane
precooled (at -34°C) and PreNEx precooled at
-34°C share very similar efficiency at around
Figure 6. COP and precool temperature for 0.5MTPA 0.41 kW.h/kg and 7.2 to 7.5% in shrinkage.
LNG at 15°C ambient.
PreNEx precooled to -45°C has the best efficiency
and the lowest shrinkage among the N2 expander
processes, at 0.38 kW.h/kg and 6.9% Shrinkage.
The reference SMR has an efficiency of
0.34 kW.h/kg and 6.1% Shrinkage.
Instead of a fix capacity requirement, it is of greater

interest to maximize the production with the
mechanical driver selected. In these simulations
the main refrigerant compressor train is driven by
a gas turbine (GE LM2500) with available shaft
powers of 28 and 29MW for ambient temperatures
30°C and 15°C respectively. The LNG production
Figure 7. Precooler Thermal Duty, Compressor Shaft
is maximized in each case within the required criteria,
Power and COP for refrigerant comparison such as the MCHE temperature approach and
study.
Figure 8. Liquefaction Power Consumption at Figure 9. Specific Power and Shrinkage (liquefaction
Train Capacity of 0.5MTPA. fuel) at Train Capacity of 0.5MTPA.
Figure 10. Liquefaction Power Consumption and Train Figure 11. Liquefaction Power Consumption and Train
Capacity - Available GT Power 29MW, Capacity - Available GT Power 28MW,
ambient 15°C. ambient 30°C.
Figure 12. Specific Power at ambient 15°C and 30°C. Figure 13. Shrinkage as Liquefaction Fuel at ambient
15°C and 30 °C.
compressor coolers design condition. The results in shows comparable efficiency with HFC precooler but
Figure 10 shows the precooler power consumption its flammability makes it less desirable as precooling
and the gas turbine available power to drive the main refrigerant. PreNEx technology has demonstrated
refrigerant compressors at ambient temperature better performance in production capacity and
of 15°C; the line graph shows the maximum train thermal efficiency against the studied processes while
capacity delivered by the selected processes. Similar maintaining simplicity and process flexibility.
for Figure 11 which is the results for ambient
temperature of 30°C. Figure 12 presents the specific Heat Flow Profile Analysis
power and Figure 13 gives the corresponding Figure 14 shows the heat flow of hot and cold
shrinkage as fuel for liquefaction and precooling composites across the MCHE in all the processes
compressors. studied. The wide gap between the composite
curves in Single N2 Expander process Figure 14A
It has clearly shown that the external precooler in indicates undesirable heat exchange profile. While
PreNEx has extended the train capacity of single N2 the temperature of the cold composite increases in
expander from 0.5MTPA to 0.7MTPA, overcoming straight line, the hot composite at temperature zone
the limitation in the Gas Turbine driven refrigerant -80 to -160°C has a plateau caused by slow change
compressors. With an additional 10% or 3MW in temperature as the natural gas condenses. This
electricity consumed in the precooler compressor, occurs when a lot of heat load is being removed from
the production capacity increases by 40%. In the natural gas stream during the liquefaction stage.
addition, the external precooler selected is of Composite curve for temperature zone above -80°C
high energy efficiency hence it helps to improve is mainly contributed by the cooling of Nitrogen
the overall specific power from 0.5 kW.h/kg to refrigerant and gaseous natural gas (in cases where
0.38 kW.h/kg as in the case of PreNEx (ambient there is no external precooler). With additional N2
15°C condition). Its shrinkage has also been greatly expander introduced (Figure 14B) a kink is observed
reduced from 9% to 6.9% (ambient 15°C condition). at the cold composite curves in the region of -100°C
The energy efficiency of the PreNEx process is only and closes the temperature gap between cold and hot
slightly lower than that of SMR, but it has greater streams. This kink is a result of additional nitrogen
process simplicity, control robustness and safety stream from the second expander. It contributes to
features. The SMR requires facility to separate and higher thermal efficiency. A triple expander process
store the mixed hydrocarbon refrigerant which will show two kinks along the cold composite, to
is flammable and demands safety measures to be reduce the gap between the cold and hot composites.
incorporated. The composition of mixed refrigerant
would also require close tuning and control to achieve Figure 14C and Figure 14D are very similar, the
the promised efficiency. Nitrogen expansion process objective is to show that a single N2 expander
uses single component and gas phase refrigerant process precooled externally by a Propane
which is very flexible and straightforward during refrigeration system and a HFC refrigeration
operation. Another advantage of PreNEx lies in system (as in PreNEx) is comparable as long as the
the highly efficient and simple hydrofluorocarbon precooling temperature is the same. The temperature
refrigeration system as a precooling package. The gap is reduced compared to Single N2 Expander
system is independent from the MCHE, whereby without precooler. This is to convince users of the
the precooling of the natural gas stream takes place advantages of an HFC Refrigeration system which
in an evaporator. The heavy condensate separation is non-flammable and efficient. Figure 14E is the
is completed before the vapour stream enters the temperature composite curves for PreNEx at the
MCHE. The MCHE configuration in PreNEx is recommended -40 to -45°C. The gap between the
simpler compared to the other processes: there is only hot and cold composite curves are even closer and the
one treated feed stream, one flash gas recycle stream, total heat load is also reduced. This can be explained
and two nitrogen streams. The liquefaction pressure by the fact that a greater portion of heat load has
and temperature can therefore easily be adjusted and been removed by the precooler which operates
optimized. The comparison with Propane precooler at around 10°C lower than the other precooled
processes mentioned. The heat removal process was could reach the temperature range of Methane
transferred from the less efficient N2 circuit to the liquefaction (-150°C to -160°C depending on the
more efficient dual-phasic HFC circuit. Hence the pressure). Other processes tends to rely on Methane
overall efficiency improvement. PreNEx insists on and/or Nitrogen to reach this temperature zone,
the use of N2 as refrigerant is very important as it turning this process area into a hazardous zone. The
is the only economical and non-hazardous gas that reference process of SMR is simulated (Figure 14F)
(14A) Single Nitrogen Expander (14B) Dual Nitrogen Expander
(14C) PreNEx precooled to -34 degC (14D) Single Nitrogen Expander precooled by
Propane to -34 degC
(14E) PreNEx precooled to -45 degC (14F) Single Mixed Refrigerant
Figure 14. Hot and Cold composite curves to represent Heat Flow in the Main Cryogenic Heat Exchanger of the
processes simulated.
Lorem Dolor
Lorem dolor ipsum 41
Figure 15. MCHE Heat Flow profiles for PreNEx and selected Nitrogen Expander processes (Hot composites only)
for 0.5MTPA.
using 6 components of Methane, Ethane, Ethylene, for the 0.5 MTPA production capacity (see top right
Butane, Propane and Nitrogen in order to achieve a graph), which is contributed by the required cooling
smooth cold composite curve that has rather uniform of the biphasic refrigerant in order to achieve the
temperature approach between the hot and cold expanded refrigerant temperature of -150°C.
composite curves. The bi-phasic hydrocarbon mixed
refrigerant is more energy efficient. In Nitrogen circuit, the compression and intercooling
have removed significant amount of heat load from
One of the challenges in Natural Gas liquefaction the refrigerant, hence less heat removal from the
is in preparing low temperature cooling medium refrigerant in MCHE for the refrigerant to achieve
(refrigerant) to drive the heat removal of the natural low temperature upon expansion. Figure 15 depicts
gas stream. Part of the role of the MCHE is also the total MCHE load for the Nitrogen Expander
to cool down the refrigerant (be it Nitrogen or Processes to be between 17MW to 25MW. By
Mixed Refrigerant) so that it may reach the low having the highest heat removal in precooling,
temperature range (-145°C to -155°C) upon PreNEx would have a smaller MCHE thermal duty
expansion. Figure 15 shows the heat load removal (30% lower) and hence smaller dimension. The dual
rate evolution in the MCHE for the processes nitrogen expansion processes tend to use higher flow
studied, with the full scale graph reduced to the top rates of nitrogen resulting in higher total heat loads
right corner and zoomed scale graph as the main in the hot composite during the cooling of nitrogen
graph. Large heat load of 65MW is observed in SMR in the higher temperature region.
CONCLUSIONS The results analysis shows that the advantage of

The study has demonstrated that careful refrigerant the PreNEx process comes from the shifting of
selection and process conditions in precooling heat load from the main nitrogen cycle to bi-phasic
stage can improve the overall liquefaction process HFC Refrigerant Cycle in the higher temperature
efficiency significantly. The selection of precooling region, and to the Flash Gas/Boil Off stream cycle
refrigerant is a tri-angular exercise where the physical in lower temperature region. This lightens the
properties has to fulfill the required precooling impact of the less efficient region of the nitrogen
process condition. The energy efficiency, safety expansion cycle on the overall efficiency. The
and environmental impact should also be taken shift of heat load to a pre-cooler also frees up gas
into consideration to result in an efficient, robust turbine energy which can be utilized to increase
and safe process, such as Keppel’s PreNEx. This LNG production, hence the higher capacity
would greatly reduce the CAPEX and OPEX. The achievable with a single train of a GE LM2500
incorporation of such precooler gives PreNEx the gas turbine.
comparable performance with existing dual nitrogen
expansion process technologies in the market, and This study concludes that PreNEx is a promising
possibly better. The validity of this study comes from technology differentiator for Keppel’s core LNG
the consistency of the simulated performance of the ship conversion business. Its inherent robustness
dual expander processes with the published data. and simplicity gives Keppel another excellent
The choice of R134a and R410A for the respective opportunity to offer innovative solutions to help
precooling temperature window grants PreNEx the address the growing midstream needs in bringing
ease of operation and its inherent safety contribute small and mid-scale LNG to market to meet
to the simplicity and robustness, with advantages of the rising global demand for energy. This
specific power consumption being 2-3% lower than technology being ideal for small to mid-sized
the dual expander processes studied. From this study, FLNG solutions offers a faster and more cost
PreNEx has demonstrated best performance in terms effective liquefaction solution to Keppel’s clients.
of production capacity and shrinkage as well.
ACKNOWLEDGEMENTS
The authors would like to thank Mr. Charles Sim for technical advice in developing PreNEx to resolve the
technical challenges and to meet market needs.
AUTHOR’S CONTACT  erine.teh@komtech.com.sg
REFERENCES
[[1] M. Barclay, P. Campbell, X. Sheng and W. S. Chong.Patent International application no. PCT/GB2010/050532, 2010.
[2] X. Sheng, W. S. Chong, R. Atapattu and K. S. Foo.Patent International application no. PCT/SG2012/000206, 2012.
[3] X. Sheng and W. S. Chong, "Selection & Development of Offshore LNG Liquefaction Process for Associated Gas," KOMtech Technology Review, pp. 68-75,
2009.
[4] E. S. P. Teh and R. Atapattu, "Effect of Feed Gas Pressure and Ambient Temperature on LNG Production," KOMtech Technology Review, 2014.
[5] E. S. P. Teh and R. Atapattu, "Quest for Simplicity and Efficiency – How Keppel PreNEx Differentiates From Other Nitrogen Expansion Technologies,"
KOMtech Technology Review, 2015.
[6] F. Keith (Editor), The CRC Handbook of Thermal Engineering, CRC Press, 1999.
[7] P. Domanski, D. Yashar and M. Kim, "Performance of HC and HFC refrigerants in a finned-tube evaporator and its effect on system efficiency," in 6th IIR
Gustav Lorentzen Conference on Natural Working Fluids, Glasgow, 2004.
[8] K. A. Marak and B. O. Neeraas, "Comparison of Expander Processes for Natural Gas," in Statoil Research Center Poster PO1-5, Trondheim, Norway.
[9] I. S. Lund Nilsen, "Method and System for Producing LNG". US Patent 2010/0132405, 2010.
Experimental Investigation on
Flow Induced Motion of Tension Leg Platform 43
Experimental Investigation
on Flow Induced Motion
of Tension Leg Platform
 Miguel Angel Moya RAMIREZ , Ms.C. ^
Antonio Carlos FERNANDES*, Ph.D.

^
KOMTECH
* LOC-COPPE/UFRJ
RECENTLY, A GROWING INTEREST HAS BEEN SEEN IN FLOW-INDUCED

MOTIONS OF OFFSHORE UNITS. These induced motions are significantly
relevant for the design of mooring systems and risers of offshore platforms. This
work discusses the dynamic behavior in current of TLP unit (Tension Leg Platform).
The model tests were carried out in the fKN@LOC/COPPE-UFRJ laboratory
facilities. A novel arrangement to represent the tendon stiffness in vertical and
horizontal modes in a shallow current channel is presented and discussed. The
experimental set-up is quite innovative since, to avoid the set down, a tower has
been implemented and with horizontal restoring forces provide by springs.
The ultra-reduced model was built using at scale of 1:200 with four columns and
four pontoons in closed configuration. The range of current velocities was from
0.33 to 2.59 m/s. and five angles of attack were considered: 0, 11.25, 22.50,
33.75, 45 degrees. The reduced velocity reached a very high value of 32 for
45 degrees of heading. Results have shown that the induced motions are
dependent on the current heading and current speed. For zero degrees a typical
bell curve of VIM was observed. On the other hand, for 45 degrees, sway motions
increase steadily with current speed.
INTRODUCTION n Number of TLP tendons

In the recent years the importance of the VIM
phenomena has been recognized as an important Tn Pre-tension of the n-th tendon
factor in the design of floating structures, mooring
Ln Tendon length of the n-th tendon
lines and risers.
KSWAY Sway stiffness
Several works have been developed and can be
founded on the literature, most of them has so far KYAW Yaw stifness
been focused on spars. For instance, [Bai, 2004],
[Constantinides, 2006] [Yung, 2004], [Huang, TSWAY Sway natural period
2003] and [Irani, 2004]. In recent years some
TYAW Yaw natural period
works have presented interesting results about TLP
such as [Leverette, 2003] and [Waals, 2007] and ASWAY Sway added mass in calm water
Semi-submersibles [Goncalves, 2012] and
[Goncalves, 2013]. AYAW Yaw added mass in calm water
General agreement about VIM response is that it Xn Coordinate of the n-th tendon
is very complex topic and, unfortunately, small connection in the direction X
differences in the geometry may vary the results.
Thus, experimental test, have to reproduce faithfully Yn Coordinate of the n-th tendon
the specific floater. connection in the direction Y
VC Current Velocity
However, experimental tests are very expensive which
difficult the optimization of platform geometry to VREDUCED Reduced Velocity
reduce the VIM response.
LPROJECTED Projected length or characteristic
According to this scenario, the scope of this work is length of the body
propose an experimental setup that can be used to
analyze a deep draft TLP in a shallow water tank LC Face dimension of the TLP column
considered ultra-reduced scales. For that, and specific
θ Angle of attack
TLP hull was tested, and the results are presented and
discussed bellow. Y or Y(t) Time domain response of the TLP in
relation to the axis Y. Axes Y is always
This innovative experimental set-up could allow perpendicular to the current velocity.
future works to optimize the geometry floaters and
test devices to reduce the VIM response as well. σ(Y) Standard Deviation of Y(t)
NOMENCLATURE YMAX Maximum value of Y(t)

TLP Tension Leg Platform
YMIN Minimum value of Y(t)
VIM Vortex Induced Motions
A/DNOM Nominal A/D Response
ρ Salt water density
A/DMAX Maximum A/D response
g Gravity acceleration
YAWMAX Maximum value of yaw angle
VTLP TLP submerse volume
YAWMIN Minimum value of yaw angle
MTLP TLP mass
ψMAX Maximum yaw response
IZZ Moment of Inertia about vertical axe
ZZ of TLP
EXPERIMENTAL SETUP
The experimental tests were carried out in the (1)
current channel of the fKN@LOC/COPPE-UFRJ
Laboratory. The current channel has 22 m. length, where:
1.5 m. width and 0.6 m depth.
ρ is the salt water density
The TLP hull was built using 3D printing g is gravity acceleration
techniques at scale of 1:200 and it consists of VTLP is TLP submerse volume
four square columns and four pontoons in closed Tn is pre-tension of the n-th tendon
configuration as is showed in Figure 01. Given the MTLP is the TLP mass
draft of the TLP, the bottom and side clearance
are, approximately, 70 and 110 meters in full scale, Vertical motions of TLP (heave, roll and pitch) have
respectively. Therefore, the possible effects of the very low natural periods, due to the very high axial
channel walls were ignored. Table 01 presents some stiffness of tendons. On the other hand, natural
general properties of the particular case studied on periods of the horizontal motions (surge, sway and
this work. yaw) are, normally, very large.
Table 1. TLP properties
[Abou-Rayan, 2012] have demonstrated that the
Water depth 1200 [m] stiffness in sway and yaw are dependent of tendon
Natural Period in Sway 185.2 [s]
properties, as is showed in Equation 02 and 03.
Natural Period in Yaw 129.9 [s]

(2)
Draft 31 [m]
Tension Leg Platforms are permanently moored

by means of tendons. Tendons are pre-tensioned (3)
to the sea floor due to the excess buoyancy of the
platform [Abou-Rayan, 2012]. Equation 01 presents where:
the relation between the tendon pre-tension, TLP Tn is the pre-tension of each tendon.
buoyancy and TLP mass. Ln is the length of tendon.
Xn is the coordinate of the tendon connection in the
direction X.
Yn is the coordinate of the tendon connection in the
direction Y.
n is the number of tendons.
Consequently, natural periods of sway and yaw can

be estimated by the following expression
(4)
(5)
where:
M is mass of the TLP.
ASWAY is added mass of the TLP in sway.
IZZ is moment of inertia of the TLP in relation to
the vertical axe ZZ.
AYAW is added mass of the TLP in yaw.
Figure 1. Ultra-reduced TLP model.
Additionally, TLP tendons provide a mechanical Taking into accounts this relevant aspects of the
coupling between heave and surge movements dynamic behavior of TLP, two alternatives were
(set-down) [Abou-Rayan, 2012]. It means that proposed to perform the VIM tests.
horizontal offsets can produce a draft variation on
the TLP. Figure 02 presents a comparison of two The first concept uses four vertical springs connected
cases. The first one, a TLP moored on water depth to each corner of TLP hull as showed on Figure 03.
equal to 2H. The second one, the same TLP moored This arrangement has the following advantages and
on water depth equal to H. It is interesting to see, disadvantages:
that an arbitrary horizontal offset D produced on
• Tendon length is considerable minor than the
the first case a draft variation equal to ∆T. However,
required. At least 6 m of spring are required.
the second case the draft variation is approximately
2.65∆T. Hence, draft variation is dependent of the • A consequence of the previous item is the draft
tendon length. variation due to the set-down is significantly
higher. See Figure 02.
• As showed on Equation 04 and 05, sway and yaw
natural periods are only dependent of pre-tension
and length of tendons. Additionally, Equation
01 shows that, at the equilibrium position,
the pre-tension for a specific draft is constant.
consequently, with the maximum available
tendon length on the current channel, the natural
periods will be considerable lower. See Table 02.
• The installation is very simple. Change the
angle of attack is very simple as well.
• Vertical springs can be modeled to represent
the exactly geometric shape of tendons and,
consequently, include a portion of the added
mass and drag forces on the tendons.
Table 2. Sway natural periods of TLP
Parameters Prototype Equivalent

System
Tendon length [m] 1175 95(*)
KSWAY [kN/m] 81.50 1008.5
Figure 2. Set-down comparison for different TSWAY [sec] 185.2 52.66
water depths.
(*) It is the maximum tendon length available on the current channel.
For the second arrangement, the pendulum concept

was used. Thus, a carbon fiber pipe was connected
to one fixed point of metal pipe tower. This tower
with 5 m. height has a joint rod where the carbon
fiber pipe was connected. This connection allows
the surge and say of TLP. Then, the carbon fiber
pipe is connected to the TLP model by a bearing
connection, which allow the rotation in yaw. Figure
04 shows the joint rod and bearing connection used
Figure 3. Equivalent mooring system using vertical
on this arrangement, Figure 05 presents a schematic
springs. representation of this concept. Finally, four horizontal
springs were calibrated to provide the correct stiffness

in surge, sway and yaw. Figure 06 shows the 4
horizontal springs in the general arrangement.
This arrangement has the following advantages and

disadvantages:
• With the pendulum concept the draft variation

of the TLP could have same order of magnitude
that the real case.
• The stiffness can be calibrated by means four Figure 4. Bearing connection (left) and Joint Rod (right)
horizontal springs, Consequently, natural used on the second arrangement.
periods of model and prototype are consistent.

• The installation is quite complicated and must
be re-installed for all the angles of attack.
• There are not contributions of tendons in the
dynamic behavior of TLP. It means that added
mass and drag force on tendons are neglected.
Taking into account the major number of advantages of

the second arrangement, it was chosen to perform the
VIM tests at the current channel. As was mentioned
before, five angles of attack were used in this study.
Figure 07 and 08 present the test setup for the VIM tests.
Tests description
The TLP was tested for five angles of attack and
eleven different current velocities. Additionally,
each velocity was tested three times, to verify the
results. A total of 165 VIM test were performed
on the current channel. The current velocity was
controlled by electronic panel of four pumps.
The stiffness curve for sway and yaw were confirmed

experimentally applying a known force, or moment, and
then measure the linear, or angular, displacement. Figure 5. Equivalent mooring system using pendulum
concept. Join rod allows 2DOF (surge and
sway). Bearing connection allows 1 DOF (Yaw).
Decay tests were performed to measure natural
frequencies of sway and yaw of TLP to compared it with
the target values. It was performed given an initial linear,
or angular, offset to the TLP model and then released
The TLP movements in six degrees of freedom were

measured by Qualisys Track Manager. This is an
optical tracking system that uses infrared light sources.
For the post-processing and analysis of the TLP movements

only the steady part of the measurements were considered, Figure 6. Equivalent mooring system using four
and it is approximately 3 hours in full scale. horizontal springs.
Reduced Velocity
The reduced velocity is defined in Equation 06. (7)
(6)
Where LC is the face dimension of the column and θ
where: is the angle of attack. See Figure 09.
VC is the current velocity
TSWAY is the natural period of TLP in sway. Table 03 presents the projected length to the five
LPROJECTED is the projected length of the TLP. angles of attack. Table 04 presents the values of all
current velocities tested for the five angles of attack
In the case of the TLP with square columns, the and their respective reduced velocity.
value LPROJECTED can be written as a function of the
Table 3. Projected Length for each angle of attack.
angle of attack as is showed in Equation 07.
Angle of attack [deg] Projected Length [m]
0.00 21.21
11.25 20.81
22.50 19.60
33.75 17.74
45.00 15.00
Figure 7. Arrangement of the equivalent mooring Figure 8. Arrangement of the equivalent mooring
system for (a) Zero (b) 11.25 (c) 22.5 degrees. system for (a) 33.75 and (b) 45 degrees.
Table 4. Reduced velocities for each angle of attack. RESULTS

Reduced Velocity Stiffness in Sway and Yaw
Vc Stiffness values for Sway and Yaw are presented,
Angle of Attack
[m/s] respectively, on Figure 10 and 11.
0.0 11.3 22.5 33.8 45.0
0.3 2.8 2.9 3.0 3.4 4.0 The spring constant and the pre-tension values of
0.4 3.6 3.7 3.9 4.3 5.1 the horizontal springs showed in Figure 07 and 08
0.5 4.5 4.6 4.9 5.4 6.4 were adjusted to achieved the target value. Once, a
measured value was reasonable accurate these values
0.7 6.4 6.5 6.9 7.6 9.0
were kept to all the angles of attack.
0.9 7.7 7.9 8.4 9.2 10.9
1.1 9.6 9.8 10.4 11.4 13.5 Sway stiffness agree very well with the target value.
1.4 11.6 11.8 12.6 13.9 16.4 However, yaw stiffness is slightly higher compared
1.6 13.6 13.8 14.7 16.2 19.2 with the target value due to the friction on the
connection which allows the TLP yaw rotation.
1.7 14.9 15.1 16.1 17.8 21.0
2.2 18.5 18.9 20.0 22.1 26.2 Natural Periods
2.6 22.0 22.4 23.8 26.3 31.1 The natural periods of the TLP in calm water are
given in Table 05. Experimental values were slightly
different, due to the friction in the mechanical
joints, with respect to the target values. It is
important to mention that the natural period here
presented was used on the calculations of Ur as is
showed in Equation 06.
Table 5. Natural Periods of TLP.
Natural Periods
Modes
Target Experimental
Sway 185.2 180.4
Figure 9. Projected length. Yaw 129.9 116.0
VIM response
The results from the test are analyzed by taking the
statistics from the time traces as follows.
(8)
(9)
Figure 10. Stiffness Curve for Sway.

(10)
where:
σ(Y) is standard deviation of time response Y(t).
YMAX is the maximum value of Y(t).
YMIN is the minimum value of Y(t).
YAWMIN is the minimum value of yaw motion.
YAWMAX is the maximum value of yaw motion.
LPROJECTED is the projected length, this value is
dependent of the angle of attack and was showed
Figure 11. Stiffness Curve for Yaw. in Table 03.
The comparison of the maximum and nominal Responses for 22.5 and 33.75 degrees have the
response a measure for the regularity of the response same tendency of 45 degrees, however, the order
signal can be obtained. If both are the same the of magnitude were, relatively, lower. Additionally,
behavior would be a sinusoidal motion [Waals, Response for 11.25 degrees presents a typical bell
2007]. VIM curve until reduced velocity equal to 15, then,
for higher reduced velocities the response increase
Figure 12,13 and 14 show the comparison between with the velocity.
the Nominal A/D for the five angles of attack.
Figure 15, 16 and 17 show the comparison between
It is important to mention that the behavior of the Maximum A/D for the five angles of attack.
TLP is complete different for zero in relation to 45
degrees. While in zero degrees a typical bell VIM Maximum A/D response presents the same tendency
curve was observed. For 45 degrees the response of the Nominal A/D response for all angles of
increase steadily with the reduced velocity. attack. Additionally, the order of magnitude is
Figure 12. Nominal A/D Response for zero, 22.5 and Figure 15. Maximum A/D Response for zero, 22.5 and
45 degrees. 45 degrees.
Figure 13. Nominal A/D Response for zero, 11.25 and Figure 16. Maximum A/D Response for zero, 11.25 and
22.5 degrees. 22.5 degrees.
Figure 14. Nominal A/D Response for 22.5, 33.75 and Figure 17. Maximum A/D Response for 22.5, 33.75 and
45 degrees. 45 degrees.
quite the same between maximum and nominal CONCLUSIONS

response to lower reduced velocities. However, to First of all, the results of this work has demonstrated
higher reduced velocities the differences increase. It that the VIM tests of TLP unit are feasible on the
means that the responses follow more accurate the fKN@LOC/COPPE-UFRJ Lab facilities.
sinusoidal shape for lower reduced velocities and,
consequently, present more distortion for higher Moreover, the equivalent mooring system proposed
reduced velocities. here was able to represent the main characteristics
of the TLP tendons on the horizontal modes with
Figure 18, 19 and 20 show the comparison between reasonable accuracy. Experimental values of stiffness
the Maximum Yaw response for the five angles of for sway and yaw present excellent agreement with
attack. the target values. Additionally, the natural periods
of TLP in sway and yaw were very close to the target
The maximum yaw response increase steadily with values.
the current speed for all the angles of attack, and
zero degrees have been showed as the critical angle However, it is important to mention that the
of attack because presents the higher response. friction on the connection that allow the rotation of
the TLP in yaw is very critical. Several mechanical
bearings were tested until find the one with a very
low friction coefficient. However, even with this
bearing the stiffness and, consequently, the natural
period in yaw are slightly different.
For the particular TLP tested, we can conclude:
• The VIM response in sway is dependent on the

current heading. For instance, very different
Figure 18. Maximum Yaw Response for Zero, 22.5, and
45 degrees. behaviors were founded for zero and 45 degrees
of angle of attack.
• The magnitude of the VIM response in yaw is

dependent on the current heading. However,
is very interested to see that the tendency is
the same for all the angles of attack. It is, the
maximum angle increase steadily with the
current speed.
Figure 19. Maximum Yaw Response for Zero, 11.25 and • The TLP response for an angle of attack equal
22.5 degrees. zero degrees presents a typical bell VIM curve.
The highest responses were founded between
reduced velocities equal to 6 and 10.
• Sway response for angles of attack 22.5, 33.75

and 45 degrees increase steadily with the
reduced velocity.
• The lower response on sway was observed to the

angle of attack equal to 22.5 degrees.
Figure 20. Maximum Yaw Response for 22.5, 33.75 and • The critical angle of attack to the yaw response
45 degrees. was zero degrees.
ACKNOWLEDGEMENTS
This work is a joint effort between Wave Current Laboratory LOC - COPPE/UFRJ and Keppel Offshore &
Marine Technology Center - KOMtech
AUTHOR’S CONTACT  miguel.ramirez@kfelsbrasil.com.br
REFERENCES
BAI, Y.; TANG, A.; O’SULLIVAN, E.; UPPA, K. C., RAMAKRISHNAN, S. Steel Catenary Riser Fatigue Due to Vortex Induced Spar Motions, OTC, 2004
CONSTANTINIDES, Y.; OAKLEY, O.H.; HOLMESN, S. Analysis of Turbulent Flows and VIV of Truss Spar Risers. OMAE, 2006
YUNG, W-T.; SANDSTROM, R.E.; SLOCUM, S.T.; DING, Z.J.; LOKKEN, R.T. Advanced of Spar VIV Prediction, OTC, 2004
HUANG, K.; CHEN, X.; KWAN, C-T. The impact of Vortex-Induced Motion on Mooring System Design for Spar-based Installation, OTC, 2003.
IRANI, M.; FINN, L.; Model Testing for Vortex Induced Motions of Spar Platform. OMAE, 2004.
LEVERETTE, S.; RIJKEN, O.; Dooley, W.; THOMPSON, H. Analysis of TLP VIV Responses to Eddy Currents, OTC, 2003.
WAALS, O.; PHADKE, A.; BULTEMA, S. Flow Induced Motion of Multi Columns Floaters, OMAE, 2007.
GONCALVES, R.; ROSETTI, G.; FUJARRA, A.; OLIVEIRA, A. Experimental Study on Vortex-induced Motions of a Semi-submersible platform with four square
columns, Part I: Effects of current incidence angle and hull appendages. Ocean Engineering, 2012.
GONCALVES, R.; ROSETTI, G.; FUJARRA, A.; OLIVEIRA, A. Experimental Study on Vortex-induced Motions of a Semi-submersible platform with four square
columns, Part II: Effects of current incidence angle and hull appendages. Ocean Engineering, 2013.
ABOU-RAYAN, A.M.; SELEEMAH, A.; EL-GAMMAL, A. Response of Square Tension Leg Platforms to Hydrodynamics Forces, Ocean Engineering, 2012.
Effect Of Drillship’s Moonpool Shape on the Resistance
by CFD and Ultra-Reduced Model Experiments 53
Effect of Drillship’s Moonpool

Shape on the Resistance by
CFD and Ultra-Reduced
Model Experiments
 Lucas do Vale MACHADO, M.Sc. (KOMtech)

Antonio Carlos FERNANDES, Ph.D. (LOC – COPPE/UFRJ)
Anis Altaf HUSSAIN, M.Sc. (KOMtech)
DRILLSHIP IS A MARINE VESSEL DESIGNED FOR DRILLING PURPOSES

OF OIL AND GAS WELLS. This kind of vessel has what is called a moonpool that
is an opening on the base of the hull used for drilling operation. Nowadays, in
search of better process efficiency, some of these drillships are being constructed
with dual-derrick and are in need of a larger moonpool, which results in some
effects on the floating structure.
This work investigate shape and size of the moonpool in a drillship dual derrick to
have its influence on ship resistance decreased. It presents the application of a
commercial CFD (Computational Fluid Dynamics) software as a numerical
approach to calculate the flow around a drillship without neglecting free surface
effects. Throughout this work, the made assumptions, applied boundary
conditions and appropriate mesh density studies are discussed. and appropriate
mesh density studies are discussed. Verification assessment is part of the work.
In addition to the numerical study, some experimental tests were done at LOC/
UFRJ (Laboratório de Ondas e Corrente) in order to validate the numerical
approach. The comparison between simulation results and experiments allows
the analysis of the present CFD models benefits and limitations, providing
guidelines for future similar studies. The overall match between laboratory and
virtual tests results supports the expansion of this procedure to other vessels and
offshore floating units.
The results of this work clarify the motion inside the moonpool and its effects.
Furthermore, it gives the results of several different moonpool profiles that were
optimized for this specific hull.
INTRODUCTION design drillships with longer moonpools, increasing

A conventional drillship has a vertical opening the unwanted effects of this. This topic is not well
in the hull from deck to baseline in order to give understood due to enormous complexity involved
passage to the drill pipe. This opening is called in the water motions.
“moonpool” and it has effects on the vessel when
it is sailing or stationary. These effects have to be Goals
taken into account when designing new ships. The main goal of this present work is to investigate
the shape and size of a large moonpool, considering
They can be better understand when we consider dual-derrick drillships, to have its influence on ship
the water motion inside the moonpool under wave- resistance well understood and to provide concrete
induced pressures and ship motion. The water solutions for evaluating the effect of moonpool on
motion happens mainly at natural modes of the vessel speed. With this, it is sought to find an optimized
moonpool as shown in Figure 1: the sloshing mode, moonpool profile at the transit point of view.
back and forth in between the vertical wall and the
piston mode where the water heaves up and down [1]. In order to accomplish this goal, both numerical
works with emphasis in CFD (Computational
Fluid Dynamics) and experimental work in the
available facilities of LOC (Laboratório de Ondas
e Correntes), at the Universidade Federal do Rio
de Janeiro (UFRJ) in Brazil were developed for a
specific design of drillship that hereafter will be
called fKN Ship.
For many years moonpool motions been studied in

an empirical way using model tests, intuition and
Figure 1. Water motion in a) piston mode and b)
sloshing mode (Hammargren et al. [2]). trials. Allying the development of the computers
with improvements algorithms of CFD codes,
These water motions can be excited through such as volume of fluid, became more popular the
different mechanism either in waves or in calm numerical approach to study the moonpool problem.
water with forward speed of the vessel. This dynamic This present work used a commercial CFD code that
magnification can cause unwanted effects like uses finite-volume method to simulate a drillship in
slamming, green water and even increase drastically calm water with forward speed, considering effects
the total resistance (hereafter will be call just as of free surface, resolving the details of boundary
resistance) of the drillship at transit [3]. Drillships layer profile and with the ship motion in two degrees
spend most of its time stationary, in Dynamic of freedom, translation in Z (Heave) and rotational
Positioning (DP) mode, drilling in one location. in Y (Pitch).This commercial software is the STAR-
The other mode is a transit mode, in which it spends CCM+ (CD-adapco).
relatively less time than the stationary but the
performance with moonpool should be a concern There was a need to validate these results discovered
according to the above mentioned. using CFD and this is the reason that experimental
works were done. The ship was constructed in a
Due the importance of the offshore exploration specific scale, calculated for LOC’s water channel,
caused by the need of petroleum, the industry is considering Froude number of the real scale. Tests
doing many improvements in drillship pursuing were done with free ship motion in two degree
drilling deeper water, faster drilling process, and of freedom (Pitch and Heave), measuring the
better subsea well control, among others. One of resistance with a load cell. A camera to capture the
these improvements is the presence of dual-derrick water motion dynamic inside the moonpool was
that allows build-up of the riser in parallel with used too. With the resistance variable measured, it
drilling bringing savings in operation time. With was possible to validate the numerical results and
this dual-derrick setup, it became necessary to improve the understanding of the phenomenon.
Geometry Besides these two ship, another ship was used to

As aforementioned, this work is studying a specific test other moonpool profiles in order to find the
design of drillship, the fKN Ship. As long as the optimized one. This last ship design is called here
moonpool is part of the design of the ship, we as fKN3 Ship, and is followed by a letter starting
gave different names for each ship according to its in “A” and continuing in alphabetical order. These
moonpool profile, normally using the acronym fKN moonpool profiles are shown in Figure 3.
followed by a number and sometimes by a letter too.
The simple design, without any moonpool, is called NUMERICAL APPROACH
fKN1 Ship. The design of the drillship with its With the advent of more powerful computers and
original moonpool, is called fKN2 Ship (Figure 2). improvements in the algorithms for the solution of
engineer problems, more this approach is becoming
cheaper and effective. To solve and analyze problems
that involve fluid flows numerically, there is a branch
of the fluid dynamics called Computational Fluid
Dynamics (CFD).
It is a very important tool in fluid dynamics

Figure 2. Design of the fKN2 Ship. nowadays, used to solve fluid flow problems by the
use of numerical methods and algorithms. The main
idea is to get the conceptual model that represents
the reality and discretize to a mathematical model
that can be solved numerically. In the cases studied
in this work, the governing equations of the flow are
just the continuity equation (conservation of mass)
and the Navier-Stokes equation (conservation of
momentum).
Considering the irrotational flow the equations can

be simplified, which is the case of the potential flow.
Potential solvers are faster and less demanding on
computer resources, but neglects the turbulence
effects.
In the case of moonpool a flow separation occurs

at leading edge and it affects the free surface as
aforementioned, because of it a Reynolds Averaged
Navier-Stokes (RANS) method is more suitable
to take account of these turbulent effects that has
a great impact on the flow behavior. In addition,
Kristiansen et al. [4] assured that the potential theory
highly over-predicts the water motion and did not
recommend for analyzing moonpool.
All the algorithms of numerical simulation have some

similarities. First of all is needed to create a design of
the problem you want to study, then is necessary to
generate a mesh in this domain in order to discretize
the space, as long as the governing differential
Figure 3. Moonpool profiles (Flow comes from right equations are in a algebraic form. The next step is to
to the left).
setup the boundary conditions, initial conditions, Sum of forces Time rate of
and physics of the problem, in other words, all the acting on the change of
=
cells receive an initial value for all the parameters volume control momentum
that are going to be simulated, and some criteria are
set to be obey while the simulation evolves. After it,
the solver process can be started and the software do Then, we have the following mathematical equation,
the math described in the program code. The last
step is to get the values of each parameter in each (2)
cell (if it is a transient simulation, at some time steps
too) and using a specific software analyze the results. Where Tij is the total stress tensor and δij is the
Kronecker’s delta.
Thus, the CFD algorithms normally are divided in
three main steps: (i) preprocessor; (ii) solver; (iii) Again, no other assumption besides that of the fluid
postprocessor. is continuum.
All the simulations of this present work were run in a Turbulence Modelling
very modest computer with Windows 7 Professional When considering turbulence, the algorithm for
(64 bits), processor Intel® Core™ i5-2310 CPU @ the solution needs more equations to represent the
2.90 GHz, 8 Gb RAM and 1 Tb of HD. turbulent properties of the flow.
Governing Equations For this purpose there are some models in the
The Conservation of Mass (Continuity Equation) literature. The most common in the literature are
law states that the mass of an isolated system does the Two Equation Models, based on the Boussinesq
not change as the system evolves, i.e., mass can eddy viscosity assumption. They have become
be neither created nor destroyed. Considering a industry standard and are commonly used for
fixed control volume, the net mass flow through most types of engineering problems. It include two
boundaries (control surfaces) has to be numerically extra transport equation to represent the turbulent
equal to the time rate of decrease of mass inside the properties of the flow without any prior knowledge
same control volume. of the turbulence structure [5].
Net mass flow Time rate of
decrease of Most often one of the transported variables is the
out of control =
mass inside turbulence kinetic energy per unit mass k. The
volume through
control volume V second transported variables varies depending on
surface S
what type of two-equation model it is. Common
choices for this second transported variables are the
Thus, the mathematical equation in the form of
rate of turbulence dissipation or the specific rate
partial differential equation is given by
of dissipation ω.
(1)
Verification and Validation
According to The American Institute of Aeronautics
It is important to notice that the only assumption
and Astronautics [6] the credibility of a solution in
about the nature of fluid is that it is continuum.
CFD is obtained by demonstrating acceptable levels
Therefore, this hold in general for three-dimensional,
of uncertainty and error. The levels of uncertainties
unsteady, viscous and compressible flow.
and errors are defined through verification
assessment and validation assessment.
The Conservation of Momentum (Navier-Stokes
Equation) law states that the sum of forces acting
Verification assessment determines if the
on the volume control is equal to the time rate of
programming and computational implementation
change of momentum. It is based on the Newton’s
of the conceptual model is correct. It examines the
second law applied in a fluid volume control.
mathematics in the models through comparison to
exact analytical results. Verification is the process for • Step 5: Calculate the following error estimates,
assessing simulation numerical uncertainty USN as along with apparent order p:
aforementioned.
• Approximate relative error
Validation assessment determines if the
computational simulation agrees with physical • Extrapolated relative error
reality. It examines the science in the models
through comparison to experimental results. • Step 6: Calculate the fine GCI
The ITTC [7] purpose a guideline for the overall According to the ITTC [10]
definition of variables,
CFD verification and validation procedures that we have
are conveniently grouped in four consecutive steps:
(1) preparation; (2) verification; (3) validation; (4) (3)
documentation.
where UG is the grid uncertainty.
The verification procedure is done by iterative
and input parameters convergence studies that are The numerical simulation uncertainty (USN) is
conducted using multiple solutions with systematic composed of the grid uncertainty (UG), time step
parameters. The validation procedure is done uncertainty (UT), iteration number uncertainty
comparing data and simulation values. (UI) and other possible parameter uncertainty (UP),
which gives the following expression
Verification
The recommended method for discretization error Validation
estimation is the Richardson Extrapolation (RE) Validation is the process for assessing simulation
method. Roache [8] has provided a methodology modelling uncertainty. It is performed by finding the
for the uniform reporting of grid refinement comparison error |E| that is given by the difference
studies that is to find a Grid Convergence Index in the data D and simulation S values
(GCI). This GCI method is based upon a grid
refinement error estimator derived from the theory | E | = | S - D| . (4)
of generalized Richardson Extrapolation. This
method can be summarized in some steps that will To determine if the validation has been achieved,
be described hereinafter. This sum of the method |E| is compared to the validation uncertainty
can be find better explained in the Journal of Fluids UV given by
Engineering Editorial Policy [9]. The ITTC [10]
brings a similar procedure that at the end brought (5)
the same result.
where UD is the experimental uncertainty and USN is
• Step 1: Define a representative mesh size h. the simulation numerical uncertainty.
• Step 2: Select at least three significantly different If |E|<UV, the combination of all the errors in D
set of grids, and run simulations to determine the and S is smaller than UV and validation is achieved
values of key variables. Define the grid refinement at the UV level. If |E|>>UV, the sign and magnitude
factor r=hcoarse⁄hfine >1.3. This value of 1.3 is based on of E can be used to make modelling improvements.
experience and not on formal derivation.
NUMERICAL ANALYSIS
• Step 3: Let h1 < h2 < h3 and r21=h2 ⁄ h1, Domain
r32 = h3 ⁄ h2 and calculate the apparent order p The geometry of the fKN Ship (Figure 2) was
of the method. designed in a CAD software in the model scale in
order to have the results that will be validated with
• Step 4: Calculate the extrapolated values the experimental tests.
For the mesh generation is necessary a domain, i.e., Domain A Domain B Domain C
the region in space where the flow will be simulated.
Fore 1 2 2,5
The idea is to generate a domain that is not so small
that the boundary conditions can change the results Aft 2 3 4
neither so big that increase the computational cost Breadth 0,5 1,5 2,5
in vain. Normally, a size domain study is done. This Depth 0,5 2 3
means that for a specific simulation are constructed Height 1 2,5 4
some geometries with different domains, they are
Figure 5. Dimensions of the three domains generated
simulated with the same physics and after the results as a function of LPP.
are compared. Then it is possible to see when the
results start to be size independent for a specific In Figure 6 we can see the difference in size between
uncertainty. each domain.
Another matter to focus when building the domain

is that the geometry is symmetrical. Although the
flow is not symmetrical because of vortices in the
wake, we can decrease computational cost and
have similar results if we simulated just half ship.
This hypothesis is based on the ITTC Practical
Guidelines for Ship CFD Application [11] and it is
used in a variety of papers ( [12], [13]) in the area to justify
Domain A Domain B
the reduction of computational effort. Another
important parameter to prove that the symmetry
is a reasonable condition, is that the vortices in
the wake may cause only variation in the pressure
resistance, but for this case, friction resistance is
bigger than the pressure resistance (can be seen in
Figure 31 that the friction resistance is almost twice
the pressure resistance). Because of that, it was used
a domain geometry with symmetry condition and Domain C
the main measurements are shown in Figure 4. Figure 6. Three different domains for the size domain
study.
The result parameter that is our most concern is

the resistance of the ship. Calculating the resistance
in each domain (Figure 7), we could notice the
dependency between the results and size. Since
there was a small variation (less than 5%) between
the results with Domain B and Domain C, and this
last took more computational effort, it was decided
to continue the analysis with the Domain B size.
Figure 4. Main particulars of domain.

Variation (ε)
Domain Resistance
(N) Absolute Relative
After some initial tries to get the right setup for the
Domain A 0.463 - -
simulation, it was generated three different domains
to do the size domain study aforementioned. In the Domain B 0.408 -0.056 12.00%
Figure 5 is possible to see the main particulars of Domain C 0.399 -0.009 2.14%
these domains generated as a function of model’s Figure 7. Variation of the resistance result between the
LPP (length between perpendiculars). three domains generated.
Mesh The refined region shown in Figure 9 is to capture

Following the ITTC specifications for the
[11]
the ship wave prediction around the ship.
mesh, in the three dimensional case, hexahedral
elements are the most popular type supported as Some mesh parameters are shown in Figure 10
long as this type have been found to be beneficial
Base Size 0.02354 m
to spatial accuracy of numerical solutions. The
Volumetric Control Parameter Value Unit
main downside of these structural mesh elements is
Custom Size 1.471E-03 m
that it is often very hard to generate high-quality
structured meshes for complex geometry. Bow
Number of Prism Layer 5 -
Prism Layer Thickness 7.365E-04 m
STAR-CCM+ works with a type of mesh that is Anisotropic X Size 1.883E-01 m

called of trimmed mesh. This method generates a FreeSurface
Anisotropic Y Size 1.883E-01 m
high percentage of hexahedral cells and splitting Anisotropic Z Size 2.943E-03 m
them is possible to get quickly a body-fitted mesh.
Anisotropic X Size 4.708E-02 m
A mesh of this type was generated in the mesh Hull
generator algorithm of STAR-CCM+ and is shown
Anisotropic Z Size 2.354E-03 m
in Figure 8.
Custom Size 2.943E-03 m

Hull Surface
Stem Custom Size 1.471E-03 m
Anisotropic X Size 4.708E-02 m

Wake
Anisotropic Z Size 1.177E-02 m
Figure 10: Mesh parameters.

Looking in the Figure 11 is possible to understand
these Volumetric Control shown in Figure 10.
Figure 8. Trimmed mesh generated in STAR-CCM+

mesher.
A good advantage of this type of mesh is the

opportunity to create anisotropic meshes, which
we can set for some elements different sizes in each
direction. The Figure 9 shows a cut plane in where
the free surface is located at the initial condition of
the simulation.
Figure 9. Anisotropic mesh with different dimensions in

XY Plane and same dimension in Z.
Figure 11: Volumetric controls for mesh refinement.
Physical Setup In the Figure 14 are presented the details of each

Besides the mesh generation, this is a very important boundary condition
step for a CFD simulation. It is when the boundary
and initial conditions are set so as the whole physics Velocity Inlet
of the domain. The Figure 12 shows the main U Depend on the case
parameters set. V 0 m/s
W 0 m/s
Density 1,18415
Air Pressure Outlet
Dynamic viscosity 1,86E-05
Fluid Relative pressure Hydrostatic Pressure
Density 997,0476
Water
Dynami viscosity 8,90E-04
Wall
Time model Implicit Unsteady No slip wall

Time step 0.012 s Smooth wall
Temporal discretization 2nd order
Simulation time 90.0 s Figure 14. Boundary conditions setup.
Solver forumulation Segregated In a transient problem, there is a need to set a
Coupling algorithm SIMPLE
condition at the beginning of the time scale. This
Discretization scheme 2nd order
condition is called initial condition. At permanent
Reconstruction gradients Hybrid Gauss-LSQ gradient method
simulation there is a need to set an initial condition
Multiphase model VOF
too, but this just change the convergence speed as
Convention scheme HRIC Convection Discretization Scheme
long as we are not interested on how the problem
Reference pressure 1 atm
evolves in this case.
Reference temperatur 25ºC
Gravity (Z Axis) 9.8 m/s2
Heat transfer Isothermal
For the simulation of this present work the initial
conditions that was set are basically the water level
Turbulence model k-ε
Motion Two degree of freedom (Pitch and Heave)
of the calm water as the necessary for the draft
required and the velocity of the water that we were
Figure 12: Physical modelling general setup. interested to study.
The need of boundary conditions to solve the
governing equations was commented before. When EXPERIMENTAL ANALYSIS
setting up the physics in the software is necessary the Experimental Model
addition of these conditions in each surface of the Following to the ITTC [14] specifications of materials
domain to have a well-posed problem. It consist of used for ship hull models, the surface smoothness
flow inlets and exit boundaries, wall and repeating and manufacturing tolerances for breadth, depth
face boundaries. and openings in the hull (moonpool), the model
was build. To assure the complete similarity between
In the Figure 13 is possible to see the boundary model and prototype, the following criteria are
conditions of the domain. required:
Velocity Inlet
• Geometric Similarity – The model and the
Velocity Inlet
prototype have the same geometry, with all
Pressure Outlet
the parts scaled. According to ITTC [14], the
model should be geometrically similar to the
ship wherever it is in contact with the water.
Velocity Inlet
• Kinematic Similarity – The length and
Wall time scale are similar implying that velocities
at corresponding points are similar too. This
Velocity Inlet
Symmetry means that the streamline patterns for the flow
Figure 13. Boundary conditions in the domain. will be similar.
• Dynamic Similarity – The forces acting on In Figure 17, it is shown the particulars of the model
corresponding fluid particles and boundary fKN2 Ship and in Figure 18 the particulars of the
surfaces are similar, that is, the ratio of all forces real ship.
in the two systems are constant. It is the most
restrictive criteria. Model
Characteristic Symb.
Unit Value
Thus, to achieve dynamic similarity all the Length Overall LOA m 1.476
dimensionless numbers relevant to the flow must
Length between LPP m 1.391
be preserved between the model and the prototype. Perpendiculars
The most important dimensionless numbers for this
Breadth Moulded B m 0.256
study are:
Depth Moulded D m 0.120
Reynolds (6) Draft d m 0.061
Weight displacement kg 17.613

Froude Number (7)
Volume displacement ∇ m 3
0.018
Velocity U m/s 0.45

It is well known that is not possible to have the Density (ITTC) ρ kg/m 3
997.05
same value for prototype and model in both Viscosity (ITTC) μ kg/(s.m) 8.9 x 10-4
dimensionless numbers. In the literature, the most
Average Temperature T °C 25
convenient solution is to scale the model according
to the Froude number and then, using a turbulence Gy-radius around X axis Kxx m 0.083
stimulator, improve the results for the real scale Gy-radius around Y axis Kyy m 0.397
comparison. The scale of model is limited by the Gy-radius around Z axis Kzz m 0.402
experimental facility available.
Figure 17: Main particulars of the model scale of the
fKN2 Ship (original moonpool).
The design of the water channel at LOC in UFRJ,
where the experiments were carried out, are shown Model
in Figure 15. Its mains particulars are shown in the Characteristic Symb.
Unit Value
Figure 16.
Length Overall LOA m 240
Length between LPP m 228

Perpendiculars
Breadth Moulded B m 42
Depth Moulded D m 12
Draft d m 10
Weight displacement mt 79766
Volume displacement ∇ m 3
77834,89
Figure 15. Main particulars of the water channel at LOC
(UFRJ). Velocity U knots 11,2
Density (ITTC) ρ kg/m 3

1024,81
Characteristic Symb. Dim. Value
Length I m 22
Viscosity (ITTC) μ kg/(s.m) 1,077 x 10-3
Width w m 1.4 Average Temperature T °C 20
Draft d m 0.50
Gy-radius around X axis Kxx m 13,53
Maximum velocity Umax m/ s 0.45
Density (ITTC) ρ kg/m3 997.05
Gy-radius around Y axis Kyy m 65,08
Viscosity (ITTC) µ kg/(s.m) 8.9 x 10 -4
Gy-radius around Z axis Kzz m 65,93
Temperature T °C 25
Figure 18: Main particulars of the real scale of the fKN2
Figure 16. Design of the water channel at LOC (UFRJ). Ship (original moonpool).
According to Manoj et al. [15], when the non- Other equipments

dimensionalized parameter that is the ratio of the • Load Cell: This equipment is a transducer that,
width of the channel and the breadth of the model when subjected to a force, create an electrical signal.
ship (w⁄B) is greater than 5.0, for a studied barge, it It was calibrated using known weights to generate
is possible to assume that the residuary resistance is a calibration curve. With this curve and what is
free from tank wall influence. Because of the limited measured during the experiment, is possible to
width of the channel, and to avoid the maximum as calculate the resistance force on the ship.
possible the wall effects, the breadth of the model
ship was the dimension limiting. The dimension • Signal Conditioning: It manipulates the analog
was calculated respecting the paper results for a signal that comes from the load cell in order to
barge with a conservative margin. meet the requirements of the data acquisition
system. The main characteristics of this process
(8) are filter and amplify the signal making it suitable
for processing after conditioning.
Comparing the LPP of the model and the prototype,
the scale ratio is • Data Acquisition System: The output of the signal
conditioning follows to the data acquisition system,
(9) in which the signal is converted into digital numeric
values that can be manipulated by a computer.
That high scale ratio will lead to high uncertainty
and because of that, an uncertainty analysis was • Propeller-type Current Meter: As long as it was
done in this work and will be shown ahead. necessary to measure the speed of the water flow,
it was used a very common mechanical current
Apparatus meter, that is the propeller-type current meter. The
Arm flow passes through the blades of the propeller and
The model ship has the six degrees of freedom like rotate it in a certain angular speed. Using a previous
any other ship. The arm that holds the ship must industrial calibration, a software manipulates its
restrict two of the motions that are sway and yaw. output and prints the current speed.
The transducer equipment (load cell) that will
measure the resistance force in the ship must restrict Procedure
the surge motion. Heave and pitch must be free The basic idea of this experiment is to measure the
because of the conclusion of Veer et al. [16] that, resistance of the ship in transit in certain speeds. Thus,
for a drillship model experiment, restraining these the only quantities to be measured are speed and force.
ship motions increases the water level oscillation
amplitudes and increases the resistance by a After the model and load cell calibration, the model
significant value. While the roll, despite being free, ship is fixed to the channel using the arm, taking care
is very small due to the symmetry of the model. to center it in the center of the water channel beam,
and observing if the trim of the model is correct.
In the ship is fixed one rail and the arm goes inside
of this rail with the bearings region. The top part of All the equipment cited above, except the current
the arm is fixed in the water channel (Figure 19). meter, are assembled together with a computer
and with the ship model in the water channel.
The current meter is fixed in a bump region of the
channel and it is connected to another computer.
The experimental apparatus is shown in Figure 20.

The ship with the arm and load cell are placed in
the upstream and the current meter downstream, in
a far position, so the ship wake does not interfere in
Figure 19. Ship in the channel with the arm. the measurement of flow speed.
When everything is assembled like aforementioned, resistance results and get the variation between each
the pumps that control the flow in the channel are results (Figure 21).
turned on and then, controlling the torque of each
pump, it is possible to control the speed of the flow. Mesh Resistance (N) Variation (ε)
After setting up the pumps for a small speed in the Coarse 0,374 -
channel, it is necessary to wait some time for the flow Medium 0,367 -0,007
regularization, and then, we get about 1 minute of Fine 0,363 -0,004
sample of the speed and the force in the load cell. Figure 21. Total resistance and its variation between the
Setting a new configuration of pumps, the speed is meshes.
increased and then, again, needs to wait a time for
the flow regularization and new measurement. These Following the same procedure explained above, it
steps will be repeated till he maximum configuration was possible to get the same results that are shown
of the pumps, in other words, maximum speed of the in Figure 22.
water channel, that is around 0.45 m⁄s.
Mesh GCI
The current meter is placed in a bump in order to get Coarse -
mote precision of the speed. According continuity Medium 3.562%
equation, that infers the mass conservation, we can Fine 2.145%
calculate the flow speed in the ship region (Vship)
Figure 22. Grid Convergence Index of each mesh.
using the current meter measured speed (Vprop).
As long as the most refined mesh took much longer
time to run than the medium one, it was preferred
(10) to use the medium mesh.
RESULTS AND DISCUSSION This calculated uncertainty is just related to the

Verification refinement of the mesh, so it is what was called the
Using the domain size specified after the size study, grid uncertainty (UG). Time step uncertainty (UT),
it was generated three different meshes for the iteration number uncertainty (UI) and other possible
fKN1 ship simulation. These meshes have different parameter uncertainty (UP), was not quantified
refinement and as the dimensions of the domain are due the lack of machine resources to run countless
the same, they have different number of cells too. simulations in a reasonable time. The most common
analysis of uncertainty is based on grid refinement, so
After generated the first mesh, the coarser, it was it was done. Therefore, we can infer that
possible to calculate its representative mesh size h.
Obeying the grid refinement factor r minimum (11)
value of 1.3, the two other meshes was generated.
The verification procedure of assessing simulation
The setup for all the meshes were the same, numerical uncertainty USN was completed for the
considering the maximum ship speed (0.45 m⁄s). most complex case only and the result will be used
After the solver process, it was possible to take the for the others vessel speed.
Free Surface
Current Meter Load Cell Arm
Flow Direction
Figure 20. Experimental apparatus.
Validation With all these values calculated is possible to

For the validation procedure, it was necessary complete the validation procedure. The Figure 25
to calculate the experimental uncertainty. This summarize these values and confirm that all the
experimental uncertainty is given by the standard results were validated as can be noted that for all the
deviation (σ) of the measured resistance signal cases |E|<UV .
times a coverage factor K, that is chosen based on
Ship Speed E Relative E Uv UD USN
the desired level of confidence. For the level of (m/s)
confidence of approximately 0.95%, we choose
0,15 0,024 28,9% 0,032 0,032 0,002
K=2, i.e., fKN1 Ship
0,30 0,053 23,7% 0,088 0,088 0,006
(12) 0,45 0,030 7,4% 0,084 0,083 0,013
0,15 0,021 46,7% 0,028 0,028 0,002

There are some other uncertainties to take in fKN2 Ship
consideration, related to the other experimental 0,30 0,004 1,8% 0,107 0,106 0,007
measurements, but they were negligible compared 0,45 0,060 12,8% 0,162 0,162 0,015
to the standard deviation in this work.
Figure 25. Validation of total resistance for fKN1 Ship
and fKN2 Ship.
The results with the errors bars will be shown ahead. One of the reasons of the validation of all initial
Now, we are just concerned in the comparison values, without any correction, is due to the
between the numerical simulation and the experimental uncertainty that is relatively high.
experimental data considering the uncertainties Improvements in the water channel of the LOC
of this comparison values. The values that were are being done to minimize these fluctuations.
compared with the numerical results were linear Better comments about this can be found in the
interpolation between the two closest measured conclusions and future proposals.
values. In Figure 23, we can see the numerical
resistance for three speeds of the fKN1 Ship, the Experimental
correspondent experimental resistance interpolated With all the measured values of speed and resistance
and the uncertainties of both, numerical and was possible to do a chart of Resistance vs. Speed for
experimental results. each experimental ship. In Figure 26 it is presented
the chart for the fKN1 Ship, while in the Figure 27
Speed
CFD Experiment* it is the chart for the fKN2 Ship.
Resistance Uncertainty Resistance Uncertainty
(m/s) 0,5
(N) (N) (N) (N)
0,45
0,15 0,059 0,002 0,083 0,032 0,4 Experimental
Resistance Force (N)
CFD
0,35
0,30 0,172 0,006 0,225 0,083 Interpolated
0,3
0,45 0,374 0,013 0,404 0,083 0,25

0,2
* Interpolated Values 0,15
0,1
Figure 23. Numerical and experimental resistance with
0,05
its respective uncertainty for the fKN1 Ship. 0
0 0,05 0,1 0,15 0,2 0,25 0,3 0,35 0,4 0,45 0,5
Speed (m/s)
In Figure 24 we have the same comparison, but for
Figure 26. Resistance vs. Speed of the fKN1 Ship model.
the fKN2 Ship.
0,6
CFD Experiment* 0,5

Speed Experimental
Resistance Force (N)
Resistance Uncertainty Resistance Uncertainty 0,4

CFD
(m/s)
(N) (N) (N) (N) Interpolated
0,3
0,15 0,065 0,002 0,044 0,028
0,2
0,30 0,192 0,007 0,196 0,106
0,1
0,45 0,408 0,015 0,468 0,162
0
* Interpolated Values 0 0,05 0,1 0,15 0,2 0,25 0,3 0,35 0,4 0,45 0,5
Speed (m/s)
Figure 24. Numerical and experimental resistance with

its respective uncertainty for the fKN2 Ship. Figure 27. Resistance vs. Speed of the fKN2 Ship model.
Numerical the implicit scheme for the time-discretization,

Y Plus what brings a theoretical unconditionally stable
The importance of the y+ is related with the capacity solution, the CFL number close to the free surface
to solve the boundary layer in the body. We can take need to be not much higher than an unity to
a look in this value for some simulations. In Figure ensure sharp interface. The Figure 29 shows us the
28 is shown the y+ for each simulation done with CFL number on the free surface of the fKN2
the fKN1 Ship. The same procedure was done for Ship simulation.
the fKN2 Ship.
This result shows that the time step choose for the
Just in some part of the bow, at the maximum speed, transient simulation is reasonable. The others
that the y+ was greater than 1. Anyway, this value simulations did not show much different values of
not exceeded 2 in any simulation done in this work. the one shown above.
It is possible to see that the moonpool does not

cause great variations in the y+ field comparing with
the case without moonpool. The others simulations
did not show much different values of the ones
shown above.
Figure 29. Courant number (CFL) in the free surface of

the fKN2 Ship simulation at cruise speed.
Wave pattern
A qualitative validation procedure consists in
check the wave pattern to assure that simulation is
capturing the surface elevation. In the Figure 30 is
shown the wave pattern of fKN2 Ship.
Figure 28. Y plus in the wet hull for different speeds of

the fKN1 Ship.
Courant-Friedrichs-Lewy (CFL) Condition Figure 30. Wave pattern of fKN2 Ship

This parameter is important when we are
working with a transient simulation, giving
us an idea of the possibility method stability. It is possible to see that the ship wave field was
Despite the simulation setup was considering captured in the numerical analysis.
Resistance Ship Resistance (N) Sm ∆CRI

The result that we were more interested since fKN1 0.374±0.013 - -
the beginning of this present work, was the total
fKN2 0.408±0.015 2.36E-02 1.43E-02
resistance of the ship, with different moonpools, at
the cruise speed. fKN3A 0.439±0.016 2.51E-02 2.59E-02
fKN3B 0.416±0.015 2.36E-02 1.78E-02
Due to the moonpool effect, the drillship resistance fKN3C 0.433±0.015 2.38E-02 2.46E-02
may vary dramatically in time. The resistance fKN3D 0.420±0.015 2.41E-02 1.90E-02
graphics of the simulations, after an amount of time fKN3E 0.416±0.015 2.31E-02 1.83E-02
(about 20.0s), started to show some periodization,
fKN3F 0.436±0.016 2.51E-02 2.46E-02
as we can see in Figure 31. The total duration of the
fKN3G 0.434±0.015 2.51E-02 2.40E-02
simulations was chosen of 90.0s so we could take an
average of the resistance in this periodic range. fKN3H 0.453±0.016 2.36E-02 3.34E-02
fKN3I 0.432±0.015 2.40E-02 2.41E-02
In the Figure 32 it is shown the value of the averaged fKN3J 0.405±0.014 2.36E-02 1.33E-02
resistance of each ship simulated at the cruise speed Figure 32. Numerical results of the resistance and
(Model speed - 0.45m⁄s). In the last column, it comparison with the case without moonpool
was calculated the moonpool resistance increased at cruise speed.
coefficient.
It is possible to see in the Figure 32 that the moonpool
(13) design of the fKN3J Ship had less resistance than
the original one. It is hard to explain why the others
profiles did not reduce the resistance, but we can
infer that these values are about the same because
where Sm = l.b the main cause of the resistance increase still there
Figure 31. Time history of the calm water resistance (fKN3J Ship).
in all these profiles, that is the vortex generated one idea to reduce this high-pressure zone was not
inside the moonpool, creating a stagnation point letting the flow stops. With that idea came up the
in the trailing edge of the moonpool (Figure 33). design of the fKN3J Ship. Besides the decrease of
The only one that could reduce a little bit the the high-pressure zone, the ship has less wall area
average resistance is one profile that try to change to suffer this pressure, so it further helps the final
drastically this stagnation point condition, that is decrease of the total resistance.
the fKN3J profile.
To elucidate what is being told, Figure 34 presents
Vortices inside the Moonpool the pressure resistance for the fKN1 Ship, fKN2
Not just the resistance value is one goal for this Ship and the fKN3J Ship. It is possible to see that
work, but understanding of moonpool’s influence the moonpool causes the increase in the resistance
in the ship. For that, we can see Figure 33 that increasing the pressure factor, not the shear resistance.
shows the flow and the pressure field inside of each It already was an expected completion, but it always
moonpool simulated at cruise speed. interested to prove. Anyway, the focus here is to show
that the decrease in the optimum design is caused by
a decrease in the pressure resistance.
Pressure Shear
Ship Resistance Resistance Resistance
(N) (N) (N)
fKN1 0,085 0,289 0,374

fKN2 0,155 0,253 0,408
fKN3J 0,142 0264 0,405
Figure 34. Numerical results of the resistance

components for the cases without moonpool,
with the original one and the optimum design.
It was proved experimentally that these vortices are

really inside the moonpool. At the free surface, is
possible to see in the simulations results, that the
water flow in the direction of the bow, against the
ship motion. In Figure 35 we can see some snapshots
of an experiment done with the fKN2 Ship and it is
noticeable that the water in the experiment do the
same motion of the simulation.
Figure 33. Pressure field and vortices inside the

CONCLUSIONS
moonpool. The main purpose of this work was to gather all
the available information about moonpool that is
Analyzing the flow inside the moonpool is possible published, try to understand how the physics occurs
to see the generation of the leading edge vortices. inside it and think in a way to optimize the design
These vortices create a recirculation zone inside of the moonpool for a specific drillship in transit at
of the whole moonpool. The average velocity of cruise speed. Besides that all, we had to think in an
the water tends to zero (relative to the ship), but effective way to do this.
a big high-pressure region appears. As previously
mentioned, this high-pressure region caused The cheap way came with the purpose of using
mainly by these vortices hitting the aft wall of a powerful numerical tool that is the CFD,
the moonpool generates forces that are opposed in which we can do many experiments necessary
to the ship’s direction improving the resistance. It for an optimization process without the high costs
was being observed since the first simulations and of an experimental facility and manufacturing
Figure 35: Experiment showing the motion inside the moonpool.
of models. Anyway, to really trust in what we are Now, that we possess the knowledge and the basics
studying in CFD, it is necessary to verify and validate for this kind of simulation, the idea is to continue
the code for the specific purpose of the work. For the trying some new profile, trying to think in drastic
validation, it is necessary to do some experiments to variations that could be applicable and use it to
compare the numerical result with the experimental improve the industry.
one. Even these experiments were the cheapest as
possible because we dedicated the resources for an PERSPECTIVES
ultra-reduced model test at the LOC facilities. This work was concerned with the drillship in
transit and not in operation. It was analyzed all
Returning to the purposes of the work, it started the time just the cases with flat water, just
with a review of the most important works current and no waves. The flow was always in
published in the area. It was possible to see that the bow-stern direction. An enrichment for this
it is not a so early problem in the Oil & Gas work would be to consider the operational case,
Industry and continues to be studied nowadays. when the drillship is in dynamic positioning (DP)
After the verification and validation procedures and waves come in any direction, hit the ship and
we could trust in the numerical results and with cause motions inside the moonpool. Waves can
that confidence, we could analyze the physics cause resonance in at least one of the modes of water
behind this phenomenon. After try some moonpool motion inside the moonpool and it would cause
designs that do not present drastic variations, some stop of working and can even be dangerous
we came with the idea of a profile that would for the workers.
minimize the way that the water would hit the
trailing edge wall of the moonpool, the fKN3J More analyzes need to be done in the optimized
Ship. This profile was the first one that brought a moonpool profile (fKN3J), considering some
result of resistance that is less than the resistance other parameters, like wake factor. Consider other
of the original moonpool. Some experiments need to variations of this specific moonpool design.
be done with one experimental model of this profile.
Another proposal is about the numerical analysis. For the experimental, we had some high standard
As long as we did not have powerful cluster available deviation in the measurements done. These
and it is costly, we could not do all the simulations experimental issues were already identified, are being
we wanted for the verification assessment. Time step studied by other students of the laboratory and soon
uncertainty (UT), iteration number uncertainty (UI) will be improved for better experimental quality.
and other possible parameter uncertainty (UP), was One of the studies started with this present work
not quantified due the lack of machine resources and it is related to the arm that holds the ship in the
and one proposal would be to continue this analysis water channel. Another parallel work is to minimize
as one more enrichment in the numerical point of some fluctuation in the free surface. There is a grid
view. Even in the numerical area, another interest and a honeycomb in the upstream of the channel to
step would be to run the simulations in the other suppress the turbulence intensity and its optimum
commercial CFD software. profile is being studied in CFD.
ACKNOWLEDGEMENTS
This work is a joint effort between KOMtech and Wave Current Laboratory LOC – COPPE/UFRJ.
AUTHOR’S CONTACT 
• Lucas.Machado@kfelsbrasil.com.br • ACFernandes@oceanica.ufrj.br
• Anis.Hussain@keppelfels.com
REFERENCES
[1] B. MOLIN, “On the piston and sloshing modes in moonpools,” Journal of Fluid Mechanics, vol. 430, pp. 27-50, 2001.
[2] E. HAMMARGREN e J. TÖRNBLOM, “Effect of the Moonpool on the Total Resistance of a Drillship,” MSc Thesis Chalmers University of Technology, 2012.
[3] G. GAILLARDE e A. COTTELEER, “Water Motion in Moonpools Empirical and Theoretical Approach,” ATMA, 2004.
[4] T. KRISTIANSEN, T. SAUDER e R. FIROOZKOOHI, “Validation of a Hybrid Code Combining Potential and Viscous Flow with Application to 3D Moonpool,”
OMAE, 9-14 June 2013.
[5] D. C. WILCOX, Turbulence Modeling for CFD, DCW Industries, 1998.
[6] AIAA, Guide for the Verification and Validation of Computational Fluid Dynamics Simulations, AIAA, 1998.
[7] ITTC, “CFD, General Uncertainty Analysis in CFD, Guidelines for RANS Codes,” em International Towing Tank Conference - ITTC, 1999.
[8] P. J. ROACHE, “Verification and Validation in Computational Science and Engineering,” Hermosa publishers, 1998.
[9] B. I. CELIK, Procedure for Estimation and Reporting of Discretization Error in CFD Applications, Morgantown.
[10] ITTC, “Uncertainty Analysis in CFD - Verification and Validation Methodology and Procedures,” em International Towing Tank Conference - ITTC, 2008.
[11] ITTC, “Practical Guidelines for Ship CFD Applications,” em International Towing Tank Conference - ITTC, 2011.
[12] A. T. P. ALHO, “Validation of CFD Predictions of the Hull Resistance and the Wave System of a Catamaran,” OMAE, 1-6 July 2012.
[13] T. TEZDOGAN, Y. K. DEMIREL, P. KELLETT, M. KHORASANCHI, A. INCECIK e O. TURAN, “Full-scale Unsteady RANS CFD Simulations of Ship Behaviour
and Performance in Head Seas due to Slow Steaming,” Ocean Engineering, vol. 97, pp. 186-206, 2015.
[14] ITTC, “Ship Models,” em International Towing Tank Conference - ITTC, 2011.
[15] M. KUMAR e V. A. SUBRAMANIAN, “A Numerical and Experimental Study on Tank Wall Influences in Drag Estimation,” Ocean Engineering, pp. 192-205,
2007.
[16] R. V. VEER e H. J. THOLEN, “Added Resistance of Moonpools in Calm Water,” OMAE, 15-20 June 2008.
Design and Installation of
An Active Wave Absorber in a Wave Channel 71

An Active Wave Absorber
in a Wave Channel
 Rafael de Barros PASSOS, Ms.C. (BrasFELS)

Antonio Carlos FERNANDES, Ph.D. (LOC – COPPE/UFRJ)
AN IMPORTANT STEP IN THE DEVELOPMENT OF NEW OCEAN

STRUCTURES is to conduct tests on reduced scale models in tanks that
reproduce the environmental conditions of their operation. For the wave modeling
in a 2D channel, waves are generated in one of its end by the wave-maker and
they should not be reflected back to the tested vessel. If these reflected waves
were not absorbed, they would interfere into the wave field that is under
investigation. The simplest solution is the use of a passive beach to absorb the
reflected waves, composed by a long ramp at the far end of the channel. An
alternative solution is the use of active wave absorption system. In these systems,
the reflected waves are absorbed by the motion of a second actuator, controlled
by the measured propagating wave at the channel. The objective of the present
work is to design and install a new active wave absorption system into fKN@LOC,
Laboratory of Waves and Currents, at COPPE/UFRJ. First results showed an
absorption level comparable with the passive beach, however a careful tuning
process is necessary to minimize the reflection coefficient.
INTRODUCTION main difference is upon the boundary condition

An important step in the development of new on the wave-maker wall (x=0). In this boundary,
ocean structures is to conduct tests on reduced scale considering small amplitudes for the wave-maker,
models in tanks that reproduce the environmental the displacement profile can be described as a
conditions of their operation. Recent advances function of the depth z:
in offshore industry have led to oil exploration
increasingly in water depth, requiring laboratories (1)
all over the world to find an appropriated way of
representing these new technical challenges with the Considering small amplitude of the oscillatory
precise accuracy. motion at the free surface level (z=0), θ(t) is:
For the wave modeling in a 2D channel, waves are (2)

generated in one of its extremity by the wave-maker
and should be absorbed at the far field by some kind The condition that must be satisfied at boundary
of beach. A tank equipped with a wave-maker is x=0 is that fluid particles have the same normal
called a wave making basin from the perspective of velocity as the paddle. With that, the solution to
ocean engineering. the velocity potential can be written as:
One of the most common conceptions of

attenuating reflected water waves is the use of a
passive beach absorber in the opposite extremity of
the channel. It could be done in a simplified way by
a long ramp in the channel, where the waves finally (3)
break, dissipating their energy. As an alternative to
these conventional beaches, wave-makers equipped
with a designed control system can be used as
active absorbers. These so called active beaches The first part of the potential represents the
have the advantage that they do not extend into the progressive waves, which propagates to infinity. The
experimental domain of the basin and, at least in second part represents a series of stationary waves,
linear theory, low or zero reflection can be achieved. usually called as evanescent modes. The amplitude
of these evanescent waves decay exponentially with
This article focus on the design and installation of the distance from the wave-maker and its effect can
a new active wave absorber system to be tested at be neglect at a distance of two to three times the
fKN@LOC, Laboratory of Waves and Currents, at
COPPE/UFRJ. Moreover, the proposed system is
evaluated in the generation and absorption mode,
compounding the complete system.
REVIEW OF THE HYDRODYNAMIC WAVE-

MAKER PROBLEM
The new wave-maker generation system will be
of a flap type, powered by an electric engine. The
actuation will be done by servo-motor, guided by a
ball screw, connected to a rigid arm. Figure 1 shows
a simplified model of a flap type wave-maker and
the nomenclature used from now on.
The development of this problem follows the

approach of a progressive wave propagating in an
infinity domain (Dean and Dalrymple, 1991). The Figure 1. Wave-maker model and nomenclature.
water depth h. The wave number k0 is related with

the progressive mode as a solution of the dispersion
equation:
(4) (9)
On the other hand, the wave numbers kj, which

are related to the evanescent modes, are found as
solutions of the trigonometric equation: The equation of the rotational moment can be
rewritten as a function of the angular speed, θ(t),
(5) and the angular acceleration θ(t). The proportional
parameters to these coefficients are the so called
After applying the condition at x=0 the orthogonality damping and added inertia of the structure, in the
between the cosine and hyperbolic cosine functions rotational motion.
is applied to isolate the constants of the problem.
The problem is defined obtaining the remaining Now this study can calculate the propagating
coefficients bellow: incident wave height, far from the wave-maker. For
that, only the progressive mode is necessary. Finally,
it is possible to show that the correct transfer
(6) function to the wave-maker under design is the
following (Frigaard and Andersen, 2010):
(10)
(7)
The graph of the proposed transfer function is
obtained when the wave parameters are known. In
Figure 2 the graph of this function is presented with
The integration of the pressure field over the plate the horizontal axis representing the wave frequency,
from the bottom (z=-h) to the free surface (z=0) although for naval architectures, sometimes the wave
gives the horizontal force, Fx: period appears to have a more physical meaning.
SYSTEM CAPACITY
In order to reproduce a wide range of wave periods
at LOC wave channel, the new system will be
analyzed between the following wave period (Tp)
range: 0.5s≤Tp≤2.5s.
Transfer function
(8)
2.0
1.7
1.4
Ratio H/S
1.1
0.8
Considering the rotational motion of a flap type
0.5
wave-maker, the total moment with respect to the
0.2
pivot point is then: 0.5 1.5 2.5 3.5 4.5
Frequency of the progressive wave (Hz)
Figure 2. Theoretical transfer function of the wave

generation, H/S ratio.
MAXIMUM WAVE HEIGHT AND MOMENTS According to Souza and Morishita (2003) the
To estimate the maximum wave height expected to necessary total moment to drive the paddle plate
each wave period one should attend the following is composed by the hydrodynamic reaction, the
requirements: flap inertia and the weight moment of the plate.
Considering now the rotational motion around
• Maximum steepness before wave breaking the axis of the wave-maker, the calculated required
(H/L<1/7); moments are shown in Figure 5.
• Maximum wave height that the physical limits

Maximum stroke at actuator height (hc)
of the water channel can admit (H=0.4 to all wave
0.8
periods, considering a water depth of 0.8m);
0.7
• Maximum stroke of the actuator (Sc=0.5m); 0.6
Actuatror Stroke, Sc (m)

Wave break
0.5
Physical actuator
• Linear Acceleration limit of the actuator 0.4
(9.81m/s^2); 0.3
Screw maximum
aceleration
0.2 Maximum wave
To obtain maximum linear acceleration of the height in the tank
0.1
stroke, considering sinusoidal motions of the cursor,
0.0
which drives the flap, it is derived twice to find: 0.5 1.0 1.5 2.0 2.5
Wave Period (s)
(11) Figure 3. Limitations of the wave-maker stroke.
(12) Maximum Wave Height

0.35
0.30
(13)
Wave height (m)
0.25
0.20
As the maximum admitted acceleration of the ball
screw is established by the engine manufacturer as 0.15
1g (9.81m/s^2), the following condition must be 0.10
respected to any wave period: 0.05

0.5 1.0 1.5 2.0 2.5
Wave Period (s)
(14)
Figure 4. Maximum expected wave height.
Figure 3 presents all the above limitations. Maximum

wave height would be the area below to all the
Moments
curves. The previous calculated transfer function 600
is used to linearly transfer all the conditions to the
limit of the actuator stroke, at the connection point 500
with the paddle plate. 400

Inertia
Moments (n,m)
Weight
Basically the wave breaking limit is the strictest 300
condition until wave period of 1.2 seconds, and after Hydrodynamics

200
that, the limitation is due to the maximum length
Total
of the piston (0.5m). After defining the maximum 100
stroke at each period, the transfer function can

0
be used again, now to find maximum wave height 0.5 1.0 1.5 2.0 2.5
Wave Period (s)
expected to each wave period, as presented in
Figure 4. Figure 5. Calculated total moments of the actuator.
Added Inertia HYDRODYNAMICS COEFFICIENTS OF

15 THE WAVE-MAKER
The methodology to determine the hydrodynamics
coefficients (added mass and damping) for the
Added Inertia (Kg.m^2)
10 wave-maker in an oscillatory motion was previously

explained. Detailed description is performed by
Passos et. al. (2015). Figure 6 and Figure 7 show
5 these two coefficients as a function of the wave
period of the generated wave over the wave channel.
0
0.5 1.0 1.5 2.0 2.5
CONTROL STRATEGY FOR THE
Wave Period (s) WAVE-MAKER
This work used feedfoward and feedback controllers
Figure 6 – Added inertia hydrodynamic parameter. based on the second order dynamic problem of
the rotational motion, since the added mass and
damping coefficients were previously calculated
Damping
200
through linear wave theory. Figure 8 shows a block
diagram of the control system, designed for the new
wave-maker. Figure 9 shows the final arrangement of
150
this new electromechanical system installed at LOC
Damping (Kg.m^2)
wave channel. The computational tools necessary to

100 develop the current work were built by means of the
commercial software LabVIEW (2014).
50
0
0.5 1.0 1.5 2.0 2.5
Wave Period (s)
Figure 7. Damping hydrodynamic parameter.
Figure 9. Final arrangement at LOC wave channel.
Freed Forward
Controller
n(k) Y
u(k)
Servo Driver Actuator
Feed Back
Yd(k) error(k) Controller
Sensor
Figure 8. Block diagram of the control system of the wave-maker.

EXPERIMENTAL ANALYSES wave steepness is the ratio between wave height and
The active absorption differs from the wave wave length and is a non-dimensional parameter.
generation by considering the absorbing wave- A comparison between linear wave theory and the
maker as a transparent boundary between the limited experimental analyses is presented in Figure 12 and
physical facility and the “sea” outside. A transparent the relative errors are presented in Figure 13. It must
boundary means that the velocity profile of the be noted that the theoretical linear prediction are the
paddle is the same as the fluid particles in contact same as presented in Figure 2 but the horizontal axis
with the paddle. Figure 10 shows a considerable is now representing the wave periods, in seconds.
difference between wave generation and wave
absorption. Forcing the paddle to oscillate results LWT AND EXPERIMENTAL RESULTS
H/S (linear) H/S (Experimental)
in the emission of waves on both sides, travelling in 2.0

1.8
opposite directions. The wave amplitude is the same 1.6
1.4
H/S RATIO
on both sides but at the paddle, the wave phases 1.2
1.0
0.8
are opposite. On the other hand, absorption mode 0.6
0.4
gives continuity for the wave profile, with only one 0.2
0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3 2.5
travelling direction. WAVE PERIOD (S)
Figure 12. Comparison between linear wave theory and

The experimental arrangement along fKN@ experimental analysis (steepness equal to 0.02).
LOC wave channel is presented in Figure 11.
Good agreement with the sinusoidal function was
obtained, especially to the small amplitude waves. PERCENTAGE ERROR ANALYSIS
The experiments were conducted using a passive 20%
15%
beach in the opposite end of the channel, to absorb 10%
the waves generated by the new wave-maker, getting 5%
longer duration for the time series under analysis. 0%

0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3 2.5
The wave period and amplitudes were calculated by -5%
-10%
peak statistical analyses in the permanent state of WAVE PERIOD (S)
each time series record.

Figure 13. Percentage error of the previous comparison.
In order to validate the predictions of wave height
by linear wave theory the research generated small The data analysis found that better fit is obtained
amplitude waves with steepness less than 0.02. The to long waves, which here means larger periods.
For the small period waves, since the wave length
is extremely reduced, the amplitude became so
small that the signal-noisy ration became relevant.
In the perspective of the linear wave theory, which
assumes infinity long waves, the results show good
agreement between theory and experiments, with
less than 10% difference to almost all of the tested
waves.
Figure 10. (a) Wave-generation. (b) Wave absorption.
After validating the wave generation mode, the
study performed some tests to investigate the use
New Wave Maker of this wave-maker as an active wave absorber.
(Generation Mode)
LOC laboratory provides an additional wave-maker
system to generate waves, while the proposed new
system was tested in the absorption mode.
Passive Beach
To achieve the objective of active absorption, the

Figure 11. Wave generation along the 2D wave channel. tests made use of wave-probes sensors located far
from the paddle, to measure the wave height and Figure 15 and Figure 16 show a sequence of shots
period. From these parameters, and assuming linear at specific positions of the actuator during the
wave theory, the motion of the actuator is controlled absorption tests. It is clear that the wave profile
by the incident wave to be absorbed. Figure 14 seems to have continuity from both sides of the
shows the laboratory arrangement to perform tests plate. The active wave absorber works indeed as
with the active wave absorber simultaneously with a a transparent wall. In Figure 17 the incident and
wave-generation system. reflected time series are presented together, after an
on-line separation, using the actual time series from
(Generation Mode) Progressive Wave (Absorption Mode)
both wave-probes. This separations was obtained
WP1 WP2
using the SIRW method (Frigaard, 1995).
The calculated value for the reflection coefficient

Figure 14. Active wave absorption along the 2D wave
found for the presented wave (1.0 second wave
channel. period) was Cr=4.91%. This absorption level is
comparable with those calculated with the passive
beach feature (Mello, 2006). Other regular waves
with periods varying from 0.8 to 1.3 seconds were
tested either, with reflection coefficients presenting
results from 10% to 15% in most of the cases of
reflection.
CONCLUSIONS AND FUTURE WORKS

This work presented some steps to design a new
wave-maker system, essential for reduced scale
model tests, mainly applied to the offshore industry.
Figure 15. Active wave absorption along the 2D wave The system was initially tested with small steepness
channel.
waves, generated in the 2D wave channel, at fKN@
LOC laboratory (COPPE/UFRJ).
Results show good agreement between linear

wave theory predictions and experimental
waves generated by the new system. The control
architecture was designed exclusively to the present
work and discrete implemented with the plant by
commercial software LabVIEW.
Figure 16. Actuator near the peak position of a

1.0 second wave period.
Incident and Reflected Time Series

15
10
Surface Elevation (mm)
5
Incident Wave
0
Reflected Wave
-5
-10
-15
Time (s)
Figure 17. Incident and Reflected time series measured at WP1 position.
The results indicate that it is possible to reach reflection. It is clear now that the availability of
reduced values for the reflection coefficient, even active absorption systems helps laboratories around
from the same magnitude of the passive beach, the world to offer good quality works with reduced
although both solutions treats different problems. tank dimensions.
The employment of the passive beach does not
require from the user a previous tuning process. Future works could include the validation of the
same system when the random waves are to be
All the testes performed at the present work required absorbed by the active beach. In addition, other
a reduced length for the wave channel (20m) while kinds of sensors could be used as well as other
the use of the passive beach requires from fKN@ position for them. A wave absorber with a wave
LOC wave channel a total length of 30m, from probe mounted together with the paddle plate is
which 10m is exclusively dedicated to kill the wave one of the possibilities.
ACKNOWLEDGEMENTS
Thanks for all support provided by Keppel Fels during the period of this research. In addition, be grateful
to BrasFels Shipyard for the construction of the paddle plate used as the flap wave-maker.
AUTHOR’S CONTACT  RAFAEL.PASSOS@kfelsbrasil.com.br
REFERENCES
[1] Dean, R. G. and Dalrymple, R. A., 1991. Water wave mechanics for engineers and scientists. Singapore: World Scientific.
[2] Frigaard, P., & Andersen, T. L., 2010. Technical Background Material for the Wave Generation Software. Aalborg: Department of Civil Engineering, Aalborg
University , DCE Technical Reports; No. 64.
[3] Frigaard, P., & Brorsen, M., 1995. A time-domain method for separating incident and reflected irregular waves. Coastal Engineering 24 , 205-2015.
[4] Mello, P. C., 2006. Um aparato para redução de reflexão de ondas em praia parabólica. Master Dissertation-COPPE/PENO
[5] Passos, R. B. et. al., 2015. Design and construction of an active wave absorber in a wave channel. To be published by COPPE/UFRJ" for only COPPE/
UFRJ.
[6] Souza, C. A. and Morishita, H. M., 2003. Evaluation of a flap type wave generator. 17th Int.Cong. of Mechanical Engineering.
Portable Fuel Tank - The Application of a
New Method of Bunkering for Small Scale Gas Fuelled Vessels 79
Portable Fuel Tank - The Application

of a New Method of Bunkering for
Small Scale Gas Fuelled Vessels
 XU Junwei Kelvin, B.Eng

Abul BASHAR, BSc Eng
Sreekala KUMAR, B.Tech
Nigel KOH, BEng, MSc, PhD
THIS PAPER PRESENTS THE APPLICATION OF A NEW METHOD OF

BUNKERING FOR SMALL SCALE GAS FUELLED VESSELS in which a portable
LNG tank is used as the ship’s fuel tanks. The bunkering could be a simple
procedure as replacing the empty tanks with preloaded fuel tanks which are
transported by truck, ship or rail. This can greatly reduce the overall bunkering
duration and ship will be no longer required to travel to remote designed shore
facilities for bunkering. This bunkering method has been adopted in the KOMtech
Ocean-going/Harbour LNG Fuelled (Dual Fuel) Tug design. This is achieved
through a carefully designed and well-utilised deck layout without reducing the
accommodation spaces, maintaining the vessel within the 500GT cap to avoid
more implementation from SOLAS regulation.
INTRODUCTION tank container (IMO - container) is arranged on the

As many LNG terminals over the world are either still aft deck area to avoid obstructing either the towing
under-development or planning for development, operation or the visibility of the master at bridge.
the supply of LNG for use as marine fuel via shore This arrangement of the portable tanks on deck
based terminals can be an issue. Other bunkering does not sacrifice spaces below the main deck which
methods such as ship-to-ship, LNG bunker vessels could otherwise be used as accommodation space or
(carrying dangerous cargo) are restricted to entering engine room and also eliminates the risk of enclosed
non-Petroleum harbour areas lest with authorized. vapour cloud explosions.
Additionally, the industry is hesitant to invest in
such vessels due to its high investment costs and This arrangement is best suited for dual fuel vessels,
limited alternative operations when LNG bunker in which the diesel tanks can be arranged within
demand is limited. the hull and the LNG tanks can be arranged on the
deck. However, this makes the tank vulnerable to
Due to lack of fully developed LNG bunkering high saline seawater which might cause corrosion of
infrastructure, the most preferred method will the exposed metal over a period of time.
be perhaps either the conventional truck-to-ship
bunkering method or the portable fuel tank transfer KOMtech – Ship & Marine Technology have
which is adopted in the KOMtech LNG fuelled developed the dual fuel Tug for worldwide application
tug. Both methods are best suited for smaller-sized as well for harbour use. Both versions are integrated
LNG-fuelled vessels with up to 90m3 tank capacity with the new hull form which is developed by MTD
due to the limited capacity of trucks with 44.5 m3 (Marine & Offshore Technology) for Diesel Escort
and LNG is only to transfer at a low rate. Refer to Tug version in 2015.
Table 1.
The 32m tugboat is designed as dual fuelled vessel
PORTABLE FUEL TANKS – APPLICATION which can run on both marine diesel oil and natural
ON A LNG FUELLED (DF) HARBOUR TUG gas. The hull tanks for the diesel oil storage has
KOMtech has introduced the use of portable tanks as been maintained as with its parent diesel version.
primary fuel tank on a 32m tugboat. The cryogenic The fuel LNG is carried in two portable fuel tanks
arranged on the aft deck of the vessel.
Table 1. Comparison of LNG bunkering method for inland vessel
Description Shore-to-ship Ship-to-ship Truck-to-ship Portable tank

transfer
Operation Cost High Very High Average Average
Maintenance Cost High Very High Average Low
Personnel Training Yes Yes Yes Only Ship Crews
Affect Other Trade Yes Yes Yes Yes
Ship Shore Link Yes Yes Yes No
(with ESD)
Bunkering Time Max. 55 min Max. 45 min Max. 80 min *Max. 30min
(for 40m3 filling at
low rate)
Bunkering Rate 60 ~ 1000m3/hr 80 ~ 500 m3/hr 35 ~ 60m3/hr NA
Capacity Up to 2000m 3
Up to 1000m 3
45m (each)
3
20m ~ 40m3 (each)
3
Suitable Ship Scale 500 GT and above 500 GT and above 250 ≤ 1500 GT 250 ≤ 1500 GT
*Esitmated
Note: Ship to Ship bunkering process time is longer depend on sea state and weather condition. Refer to [1] & [6]
The main areas of interest while having a gas fuelled BUNKERING

vessel are the following: Portable Fuel Tanks - Description
A cryogenic tank (e.g. IMO-container) comprises
1) Bunkering a vacuum insulated container that is designed with
2) Supply of LNG from tank to the prime mover performance, ease of operation and safety in mind.
3) Gas venting arrangement The cryogenic tank can store LNG. The cryogenic
tank is designed in accordance with ISO standards
Vent Mast
and its size and dimensions ensure that the tanks
can be conveniently transported by trailers, ship and
railway. These tanks are to be certified and should
comply with International Maritime Organization
(IMO) regulations, Classification Societies and ISO
1496-3:1995(E) [4].
LNG Tanks
Bunkering Station
Figure 1. 3D View of the 32m LNG Fuelled (DF) Harbour Tugboat.
Table 2. Principal Particulars of LNG Fuelled (DF) Harbour Tug.
Length Overall 32.00 m

Moulded Breadth 12.48 m
Moulded Depth 5.30 m
Design Draft 4.00 m
Extreme Draft (With Skeg) 5.40 m
Service Speed 13 knots
Bollard Pull (Astern) 65 Tons
Main Propulsion Engine 2 x 1920kW @ 800 RPM
Thruster (ASD) 2.7 m diameter
LNG Tank Capacity 2 x 20 m3
LNG System LNG Vaporizer, Buffer tank
Classification ABS +A1, + AMS, E , Towing Vessel, Fire Fighting Vessel Class 1, BP ( ),
GFS (DFD), UWLD.
Figure 2. Profile View of the 32m LNG Fuelled (DF) Harbour Tugboat.
Figure 3. Plan View of the 32m LNG Fuelled (DF) Harbour Tugboat.
The structural frame around the tank helps to The nitrogen on-board is used for the purging of
protect against mechanical damage. The tank is the pipeline to ensure gas-free before connecting/
secured on the deck of the vessel using a twist-lock disconnecting cryogenic hoses/ stainless steel pipe
securing mechanism that is commonly used in connection. The cryogenic tank is fitted with o the
container ships as illustrated in Figure 4. pressure relief valve when the tank pressure increased
above safe level.
The main components of the portable tanks are
mainly LNG transfer line and the pressure relief
valves.
Figure 4. Typical Portable Fuel tank and the twist lock arrangement.
Table 3. Typical Portable Tank Specification
Type Portable, IMO independent type C tank container

Insulation type Vacuum insulation
Size 20’ IGF ISO container
Dimensions 6,058 x 2,438 x 2,591 mm (L xW x H)
Vacuum ≤0.1Pa (Warm)
Net evaporation rate 0.34%/d (LIN) (ambient conditions: 100kPa and 15ºC
Total geometric volume Abt 20m3
MARVS (inner vessel) 1.0 MPa gauge
MAWP (inner vessel) 0.9 MPa gauge
Design pressure (outer vessel) -0.1 MPa gauge
Design temperature -196ºC
Medium density 426~470 kg/m3
Baffles Two transverse baffles
Frame material Stainless steel 304
Main materials Low alloy steel
Tare Weight 8200kg
Load Weight 8460kg
Maximum gross weight 16600kg
Vacuum design life 5 years
Production design life Not less than 20 years
Connections Quick- Coupling cryogenic hoses
Portable Fuel Tank Transfer – SUPPLY OF LNG FROM TANK TO THE

A new bunkering method PRIME MOVER
Portable fuel tank transfer is not considered as a In the regasification unit, liquefied natural gas is
typical bunkering process. The process is treated boiled off then to regasified to form natural gas, in
as loading and unloading of dangerous materials. a circulation of an intermediate fluid (freshwater/
Such operation will have to follow the stowage and glycol mixture) between a vaporizer and ambient
handling requirements for portable tanks containing air heater. The gas from the regasification unit is
hazardous materials. Under CG-OES Policy letter supplied to the engine room through double walled
02-15 [7][8]. LNG delivered in Portable Tanks fall pipes. The annular space in double wall pipe is
under the definition of “Dangerous Cargo” in ventilated by air/insert gas supplied either from
33 CFR Part 126 and must be loaded from a engine room or via pipeline outside the engine
designed Waterfront Facility inspected under room (to keep away from hazardous zone). This is
33 CFR Part 126. to ensure maximum safety in event of gas leakage.
The fuel supply from the vaporizer passes through
The advantage of portable fuel tank transfer a buffer tank for maintaining a stable supply to the
compared to truck-to ship are mainly lesser engine to reduce the risk of knocking phenomenon
maintenance and operation cost (including training in case of not within the range of fuel quality and
of personnel with LNG handling training) and temperature as recommended by engine maker.
lesser bunkering time. Additionally, there is no
requirement for vessel which adopts a Portable Tank GAS VENTING ARRANGEMENT
transfer to have the Ship Shore Link (SSL) installed Excessive Pressure Build Up
on board. However, the shore facility will need a The changes in temperature result in higher vapour
crane facility to transfer these tanks. pressure in the tank especially if the tank is nearly
Figure 5. Truck-to-Ship LNG Bunkering.
Figure 6. Portable Fuel Tank Transfer.

full with reduce vapour space in tank. The cryogenic Vent Mast Arrangement
tank container is fitted with at least two lines exiting Hence, directing the vent mast outlet as high as the
the tank top into the pressure relief valve. In case of foremast allows the vapours to rapidly disperse to
excessive pressure build up, gas passes through the atmosphere vertically, especially in windy condition.
lines into a main vent mast where it is vented to The vent mast outlet is directed to the foremast for
atmosphere. achieving a safe distance radius of 10m away from
the opening to accommodation and others. The safe
Maintaining a minimum safe distance of the venting distance can be reduced to 4.5m radius as long as a
outlet from openings such as accommodation gas dispersion analysis is made to justify the risk and
entrances, engine exhaust outlets and others, is done in case of vapour cloud form beyond the 4.5m, a gas
to avoid the risk of flammable or toxic mixtures detector system is to be provided to acknowledge
accumulating in little air movement condition and operator of the risk for precaution. Refer to Figure
drawn into machinery spaces or accommodation. 7, hazardous zone are define as per description in
These vapours can be heavier than air when in cold IGF-code MSC 95/WP.7 [3] and ABS Guideline [2].
condition and will accumulate in bilges and other
low areas. Protecting the areas affected by Hazard area on
vessel
Commonly observed when flammable or toxic gases Any equipment within the hazard area is to be
can eddy and cause pockets of gas present in most explosive proof certified. For example, the walkie-
unexpected places such as the aft of the superstructure talkie, navigation lights and other lights etc. The
when wind blowing from forward. The operation main challenge in a small vessel like this tugboat is
profile of a tug demands frequent turning and how to restrict the hazard zone.
manoeuvring the vessel during towing, helps to avoid
the accumulation of vapors on those areas.
Vent Mast Exit
Zone 2
3.0m
Engine
exhaust
outlet
4.5m
Entrance/ exit
Zone 1 Engine room
ventilation
inlet/outlet
Tank
connection
pipelines
Figure 7. Illustration of the Hazard Area Plan.

Efficient deck and equipment layout restricting the tanks on deck. Portable fuel tank transfer is best
hazard zone to a small central area on the aft deck suited for small vessels and in countries where facility
is the method adopted in the design of this tugboat. is far from vessel operation region or facilities still
The LNG related equipment, tanks and pipelines under development. The advantage over truck-to-
are carefully arranged to avoid hazard zone area for ship bunkering method is the reduction in bunkering
tanks on deck. This also allows the people to work time with minimum equipment, in a safer manner
at the wharf/berth while vessel is alongside. and also eliminates the ship shore link.
CONCLUSION The design concept have been filed for

The purposes of this paper is to introduce the portable Singapore Provisional Patent (Application no.
fuel tank as an efficient method of fuel storage in 10201508833W) – “AN ARRANGEMNT OF
small scale gas fuelled vessels and highlighting the CYROGENIC CONTAINERS FOR LNG OR
issues related to installation and handling of such CNG ON A DECK OF A TUGBOAT”.
ACKNOWLEDGEMENTS
• ABS Classification Society
• Marine Technology and Development (MTD)
We also acknowledge the information input from one of the LNG system suppliers; SHANGAI HANSAIL
MARINE & OFFSHORE DESIGN CO., LTD
AUTHOR’S CONTACT  Kelvin.xu@komtech.com.sg
REFERENCES
[1] American Bureau of Shipping (ABS), Bunkering of Liquefied Natural Gas-Fuel Marine Vessels in North America 2th Edition, USA, 2015.
[2] American Bureau of Shipping (ABS), Guide for Propulsion and Auxiliary System for Gas Fueled Ships, Houston, USA, May 2015.
[3] International Maritime Organization (IMO), IGF-Code MSC 95/WP.7 Adoption of the International Code of Safety for Ships using Gases or other Low-Flashing
Point (IGF CODE), NORWAY, 10 June 2015.
[4] International Standardization Organization (ISO), ISO 1496-3:1995 (E) Series 1 freight containers –Specification and testing; Part 3: Tank containers for
liquids, gases and pressurized dry bulk, Switzerland, 1995.
[5] John L. Woodard and Rohin M Pitblado, LNG Risk Based Safety – Modelling and Consequence Analysis, Hoboken, New Jersey, John Wiley & Son INC.,
2010.
[6] Swedish Marine Technology Forum (SMTF), LNG Bunkering Ship to Ship Procedure, Uddevalla, 2013.
[7] United States Coast Guard (USCG), Guidance Related to Vessles and Waterfront Facilites Conductiong Liqefied Natural Gas (LNG) Marine Fuel Transfer
(Bunkering) Operation, Washington, 19 Feb 2015
[8] United States Coast Guard (USCG), Guidelines for Liquified Natural Gas Fuel Transfer Operations and Training of Personnel on Vessel Using Natural Gas as
Fuel, Washington, 19 Feb 2015
Innovative Fatigue Prediction Model for Improved Fatigue Performance
of Welded Steel Connections in Offshore Wind Turbine Jackets 87
Innovative Fatigue Prediction Model

for Improved Fatigue Performance
of Welded Steel Connections in
Offshore Wind Turbine Jackets
 Yilmaz SALMAN , MSc 1
Mariano E. OTHEGUY1, PhD, MSc
Prof. Johan MALJAARS2, PhD, MSc
Sander DRAGT2, MSc
Sjoerd van der PUTTEN2, MSc
1
KOMtech Europe
2
Netherlands Organisation for Applied Scientific Research (TNO)
THE SERVICE LIFE OF OFFSHORE WIND TURBINE (OWT) JACKETS IS

SIGNIFICANTLY AFFECTED BY FATIGUE DEGRADATION OF THE WELDED
JOINTS. The challenge in the design of OWT jackets is thus to balance fatigue
performance, structural weight and operating and maintenance (O&M) strategies.
The present day methods for predicting the fatigue performance, according to
offshore wind design standards and guidelines, have certain known conservatism.
Welding technology, materials, material degradation understanding and structural
health monitoring techniques have evolved over the years giving rise to reassess
the potential sources of conservatism in order to reach a more efficient design and
maintenance of OWT jackets.
This paper describes the present day fatigue prediction method and outlook of
an innovative model in which the present day method is potentially adapted to be
sensitive to load sequence effects. The load sequence dependent calculation of
the fatigue life has the potential to give less conservative results than the current
techniques in the context of offshore wind turbine support design life.
INTRODUCTION circumferential butt connections according to the

Offshore Wind power is considered one of the fastest hot spot stress method and fatigue life calculations
growing maritime sectors and the most promising based on SN-curves and the Palmgren-Miner rule of
sources of ‘clean’ energy towards meeting the linear accumulation of fatigue damage.
European targets for 2020 and 2050. The growth of
the sector has also been mirrored in other continents CURRENT FATIGUE PREDICTION METHOD
like China and US. The technology keeps rapidly The detailed fatigue design of an OWT support
growing and the top offshore turbine makers in structure is a process that involves in general the
the industry, Siemens and MHI Vestas, are already wind turbine manufacturer (WTM), the foundation
developing wind turbines with power ratings up to 7.0 designer (FD) and a certification body. The full set of
MW and 8.0 MW, respectively. The combination of load cases, predefined by the industry-wide accepted
increasing wind tower heights, rotor blade diameters IEC-61400 standard and followed by DNV, are
and water depths, needed for the multi-megawatt simulated by the WTM by using the aero-elastic
wind turbines, makes the need for complex jacket model of the wind turbine, in which the model
support structures inevitable. As the design of jacket of the foundation is included. The results of these
support structures is fatigue driven, the challenge is simulations together with the synchronized wave
to balance fatigue performance, structural weight loads are used by the FD to perform simulations
and the operating and maintenance (O&M) on the detailed foundation model. Subsequently a
strategies in order to reduce the Capital Expenditure fatigue assessment is carried out in order to determine
(CAPEX) and the Operating Expenditure (OPEX). whether the foundation design fulfils the design
The current industry-wide accepted methods of criteria for the Fatigue Limit State (FLS). Thereby,
fatigue life prediction, according to offshore wind considering each member of the structure which is
design standards and guidelines, have certain known subjected to fatigue loading, e.g. welded joint and
conservatism. This is due to the uncertainty in attachment or other form of stress concentration.
load and resistance variables, required safety and
assumptions and simplifications in the calculation LOAD CASES
procedures. The number of load cases to be assessed in the
structural analysis of the foundation structure is
This paper presents the current mainstream industry specified in the industry-wide accepted IEC-61400
practice of Fatigue Design of Offshore Wind Turbine standard and followed by the offshore wind-specific
Foundation Structures (OWTFS) and outlines the DNV-OS-J101 guideline [1]. Of the total 34 load
possibility of improving this method, in the context cases defined in the DNV guideline, 7 correspond to
of the on-going Joint Industry Project FeLoSeFi. FLS [1], these being:
FeLoSeFi stands for “Fatigue Life Load Sequence
effects and Failure-probability driven Inspection” and • Power production: load case 1.2.
aims at reducing the CAPEX and OPEX for welded
OWT jacket type support structures by reducing • Power production plus occurrence of fault: load
conservatism of both the fatigue life prediction and case 2.4.
the inspection intervals compared to the present day
methods. • Start up: load case 3.1.
The innovative fatigue prediction model proposed • Normal shut down: load case 4.1.
in this paper considers load sequence dependent
calculation of the fatigue life. This has the potential • Parked (stand still/idling): load case 6.4.
to give less conservative results than the current
techniques in the context of offshore wind turbine • Parked and fault condition: load case 7.2.
support design life. The current mainstream industry
practice described here covers the assessment • Transport, assembly, maintenance and repair:
of Fatigue Limit States at chord-brace and load case 8.3.
Depending on different yaw angles of the wind A constraint mode is the static deformation shape
turbine nacelle, wind speeds and directions, due to a unit displacement applied to one of the
corresponding waves, as it applies misaligned wind boundary DOFs, while the remaining boundary
and waves, and additionally with different wave DOFs are restrained and no forces are applied
probabilistic seeds, the randomness and hence the at the internal DOFs. Hence, the set of constraint
number of simulations increases enormously. The modes contains the reduced foundation structure’s
total number of ten minutes long simulations may static response to applied boundary displacement.
add up to several hundreds, generating amounts of Since only static deformation shapes, i.e. constraint
data in the order of GB’s. modes, are considered in this method, accurate
or exact solutions are only found if this method
ANALYSIS MODEL is applied to static problems. An extension of the
The wind loads exerted on the turbine and the Guyan reduction method is the Craig-Bampton
added effect of wind, waves and current on the reduction method. In addition to the constraint
foundation structure are traditionally computed modes, this method also takes into account the
using different models and software packages. The internal dynamics of the foundation support
main reasons behind this are, firstly, traditionally structure by including tthe vibration modes of
only the turbine manufacturers had the expertise the foundation support structure with fixed
and capability to conduct time-domain aero-elastic interface. Depending on the detail of this
simulations. Secondly, both turbine manufacturers reduction model, i.e. the number of chosen
and foundation designers want to protect their vibration modes, the dynamics of the foundation
respective intellectual property. Finally, there is structure can be captured to a higher extent.
also the inability to easily share the aero-elastic Research suggests that Craig-Bampton super element
and foundation models, as the software packages with only ten vibration modes provides accurate
and modelling approaches used by different parties results for jackets [9].
could be incompatible. In order to overcome these
issues, the different parties prefer to exchange ANALYSIS PERFORMED BY WTM
only interface data and reduced or equivalent The reduced foundation structure together with
models. Thereby, first, the most commonly used the force time histories on the foundation structure
substructuring and reduction techniques found in boundary DOFs are sent by the FD to the WTM.
the class of Component Mode Synthesis (CMS) These force time histories are obtained by applying
methods are used in order to reduce the number the prescribed wave load cases on the detailed
of degrees of freedom (DOFs) of the foundation foundation structure. The reduced foundation
structure FE model, without modifying the mesh. structure and the aero-elastic model are reassembled
The basic idea behind the CMS reduction method by the WTM and the received force time histories
is modal superposition, where the foundation nodal on the interface are applied together with the
displacements are written in terms of modes and synchronized wind loads on the turbine in order
their associated modal amplitudes. One of the oldest to perform integrated simulations for the full set of
component model reduction methods still used predefined load cases, see Figure 1. Subsequently,
nowadays is the Guyan reduction, where the internal the interface responses between the tower and
DOFs of the foundation structure are condensed on reduced foundation structure are retrieved from
the boundary DOFs by using the constraint modes. simulation results and sent to the FD.
Load sequence dependent fatigue life calculation has

the potential to give less conservative results than the
current techniques applied to offshore wind turbine
support structures.
Figure 1. Steps required for obtaining the interface responses on the jacket structure in order to apply
retrieval analysis.
ANALYSIS PERFORMED BY FD linear accumulation of fatigue damage using the

The interface responses, retrieved from integrated hot spot S-N curve corresponding to the structural
simulation results, are together with the synchronized detail. Thereby, the partial fatigue damage caused by
waves used by the FD in order to perform retrieval each hot spot stress histogram is first factored by the
analysis on the detailed foundation model. probabilistic distribution of the corresponding load
Thereby, each simulation typically corresponds to case after which linear accumulation of all partial
a 10-minute long period of the service life of the damages takes place. The obtained total service
structural system. After obtaining the full nodal life fatigue damage, per hot spot per joint, is then
foundation response for each simulation, structural multiplied by the Design Fatigue Factor (DFF)
mechanics is applied in order to obtain the hot spot corresponding to each location on the jacket. The
stresses at hot spots (eight hot spots on chord side DFF amplifies the fatigue damage according to the
and eight on brace side) around the circumference difficulty to inspect and repair eventual initial fatigue/
of every tubular joint or butt weld intersection in coating damages. A given joint in the structure is fit
the structure, see Figure 2. The semi-empirical for purpose if the total service life fatigue damage is
formulas of Efthymiou as indicated in DNV- less or equal to 1.
RP-C203 [2] are used to compute the stress
concentration factors (SCFs), which are multiplied ASSUMPTIONS MADE
by the nominal stresses in in order to obtain the It is known that the current industry-wide accepted
hot spot stress time histories from the stress time method of fatigue life calculation has a number of
histories. Subsequently, Rainflow counting is assumptions which do not reflect the reality. For
applied to the resulting hot spot stress time-histories example, due to uncertainties on residual stresses
in order to transform them into a hot spot stress only the stress ranges are considered in the fatigue
range histogram in which the discrete number calculations. Compressive and tensile cycles are
of hot spot stress ranges is classified into hot spot considered to exert the same effects on welds. Also,
stress range bins. The results are given in terms of it seems that the S-N curves for free corrosion in
FLS compliance, involving the use of hot spot S-N seawater lack experimental evidence and therefore
curves corresponding to each structural detail and possibly affecting the curves for seawater with
joint location. The resulting collection of hot spot cathodic protection. Furthermore, thickness
stress range histograms per nodal hot spot is further correction applied to the S-N curves may not apply
processed according to the Palmgren-Miner rule of to thick sections.
Figure 2. Hot spots (eight on chord side (leg side) and eight on brace side) around the circumference of a particular
joint intersection at which hot spot stress should be evaluated.
A more important assumption made in the present MODEL(S) INCLUDING RETARDATION

day fatigue calculations is the use of Palmgren-Miner EFFECT
rule of linear accumulation of fatigue damage. In In metallic structures, large stress peaks have been
this model the load sequence effects are not taken observed to cause a crack growth reduction, when
into account in the fatigue calculations, i.e. the compared with the same loading without these
resulting hot spot stress range damages are linearly peaks. This is understood and modelled to be due
superposed independently of their actual occurrence to a plasticisation at the fatigue crack tip and along
in time. Nevertheless, the maritime and the oil the crack wake. The enlarged plastic zone resulting
and gas industry, which has decades of experience from a relatively large deformation or load peaks
has proven that the method delivers structures with causes that more energy is required to re-open the
an appropriate fatigue life. However real OWT crack or to advance the crack into the plastic zone
response measurements show clear differences in during subsequent smaller stress cycles, leading
load and response characteristics compared to for to an increase in the crack growth resistance of
example ships and bridges, see Figure 3. the deformed material. This effect is known as
“retardation effect” and is observed in many tests
As the wind force is one of the key drivers for the on various types of material. The opposite effect,
force on the structure, a change in wind direction crack growth acceleration due to relatively large load
will greatly influence the stresses in a given detail. To valleys also occurs but is generally found to be of
give a practical example, a change in wind direction lower significance as retardation.
of 180 degrees can cause a detail, in the bottom of
the tower, to transit from tension into compression Based on these observations, the current fatigue
or vice versa. Smaller changes in wind direction calculation method for OWT jackets may be
can have similar effect, which explains the large regarded conservative. Therefore an alternative
fluctuations in the OWT’s response measurements. fatigue calculation model is proposed within the
FeLoSeFi project in which the currently industry-
wide accepted method for fatigue damage calculation
OWT could potentially be adapted to be sensitive to
load sequence effect. The models including the
retardation effect and one route to produce a feasible
Stress (MPa)
add-on of this effect to the current industry standard

for fatigue life calculation are presented in the
following sections.
01 07 14 21 28
Time (days)
Ship Bridge
Stress (MPa)
Stress (MPa)
01 07 14 21 28 01 07 14 21 28
Time (days) Time (days)
Figure 3. Load response characteristics Offshore Wind Turbine (top) vs Ships (left) and Bridge (right).
Two prerequisites apply for taking crack growth Fracture mechanics models have been put forward
retardation and acceleration into account: in literature in which the effects of retardation are
considered. The models can generally be divided
• The crack growth itself has to be considered. For into three groups:
this purpose, fracture mechanics or finite
element models in which the crack itself is • Models that consider the difficulty of crack
modelled, are suited evaluation methods. advancement in the compressed plastic zone
in front of the crack tip as being responsible for
• The actual load sequences have to be considered. retardation. One of the most applied models
For this purpose, cycle counting procedures is the Generalised Willenborg model [4], which
generally applied in fatigue design, such as the assumes that retardation occurs as long as
Rainflow counting or reservoir procedure are the plastic zone in front of the crack tip is smaller
not suited because load sequences are then lost. than the plastic zone of a larger load cycle that
A cycle-by-cycle procedure is required instead. has occurred in recent past, Figure 4.
overload • Models that consider closure of the crack due

Plastic zone to plasticity in the wake of the crack. A large
Crack tip
due to load cycle generates a large plastic zone in the
overload wake. Subsequent smaller load cycles are less
effective for crack growth because a portion
of the stress cycle is consumed before the crack
is re-opened, Figure 5. One of the most used
overload type of model considering this re-opening are
strip yield models [5].
normal stress
Plastic zone • Models that consider - to a certain extent - both of

extend of the the above mentioned aspects. Notably finite
current cycle element models belong to this group. Essential
in this respect are a sufficiently dense mesh
with elements at the crack tip in the order of
normal stress
a few micrometres and an accurate yield model
accounting for kinematic and isotropic
Figure 4. Schematic representation of the concept of hardening.
the Willenborg model
enlarged plastic zone in the wake

resulting from preceding overload
crack tip
crack wake
Sit 1: no external stress
Plane of symmetry
Sit 2: external stress required

to re-open the crack
Sit 3: external stress exceeds

stress required to re-open
the carck
Figure 5. Schematic representation of the concept of the plastic wake models.

All of these models represent an approximation of consuming. Therefore, at the end of the project, it
reality. The actual retardation and acceleration effects will be investigated whether a simple modification to
are complex and not yet fully understood. Most the S-N curve procedure or Miners sum accounting
models provide reasonable to accurate results for a for load sequence effects is feasible. A flowchart of
single large stress peak - often referred to as overload the procedure is provided in Figure 7.
- in a further constant amplitude loading. The effects
of large stress valleys – underloads – are not predicted
in the Generalised Willenborg model. Strip yield
models are developed for a two-dimensional plain
stress situation and are not applicable in case of the
semi-elliptical surface flaws in thick-walled welded
steel structures that are most often encountered in
practice for offshore wind turbine structures. Finite
element models are probably the most accurate
models but they require a significant computation
time, especially when a 3-dimensional geometry is
considered, and are therefore not suited to be used in
design environments.
Figure 6. Crack growth test on a representative OWT
Within FeLoSeFi, a procedure is being developed jacket welded joint specimen welded by
that does take into account of the effects of Keppel Verolme.
overloads and underloads. The so-called state
space model [6], [7] – a fracture mechanics model Results of a study preceeding this JIP, performed
that originally used the results of simulations with by TNO, show promising results [8]. Some
strip yield models as input – is used as a basis. In important additional findings of the current
FeLoSeFi, contrary to the original model, the model programme – as far as carried out at this moment – are:
parameters are calibrated on the basis of the results
of finite element models. This results in a model that • The retardation of crack growth due to a single
considers load sequence effects on crack growth with overload is much larger than the acceleration
reasonable accuracy and with reasonable effort. due to a single underload.
Complex variable amplitude (VA) load sequences • An underload followed by an overload gives equal
such as those observed in offshore wind are too crack growth retardation as a single overload.
complex for most types of model, including the Hence the effect of the underload is cancelled.
finite element model. To the knowledge of the This is different in the case of an overload
authors, models that are able to predict the effects of followed by an underload. Depending on the
any load cycle are not yet available. For this reason, stress ratio, this either results in retardation or in
a substantial test programme is being carried out acceleration.
at TNO in which the effects of combinations of
over- and underloads and complex variable • High residual tensile stresses are present in the
amplitude (VA) load are determined on simple test vicinity of weld toes. Crack growth retardation is
specimens and on tubular joints representative for much smaller in this area as compared to a stress
Offshore Wind Turbine Jackets (Figure 6). relieved detail due to these residual stresses. For
a somewhat deeper crack, however, residual
These tests are used on the one hand to validate compressive stresses are present and crack
the finite element models and on the other hand – growth retardation is larger in this area.
together with the finite element models – to calibrate
the model parameters of the state space model. OUTLOOK TO IMPLEMENTATION
For design purposes, a fracture mechanics model The present day fatigue prediction method for
such as the state space model may still be too time OWT jackets is based on a full set of predefined load
Figure 7. Procedure followed in FeLoSeFi to model the retardation effect.
cases as input and on the Palmgren-Miner rule for • Broadbandedness of response: short term
the linear cumulative damage, thereby neglecting response to a non-changing environmental
the possibility of positive effects of retardation on condition will be broad banded, e.g. amplitude
the fatigue life. The following subsections describe of every oscillation will be different.
how to implement an innovative fatigue prediction
method, which will take into account the possible • Day-to-day (week to week) variations of
positive effects of retardation to OWT jackets. amplitude of response: the average amplitude for
each operational condition will be different, e.g.
Reduction of load cycles average amplitude will be much higher during
Within the FeLoSeFI project, a novel analytical days with high wind speeds than during days
crack growth model is being developed, which can with mild wind speeds.
take these retardation and acceleration effects into
account. As is the case with the models described • Day-to-day (week to week) variations of mean
earlier, these models work on a cycle-by-cycle bases. stress: difference in mean stress can result from
This puts a large strain on computational resources, changes of environmental conditions. For
especially when very large stress traces are to be example, depending on wind direction the mean
processed. To be able to process a measurement bending moment, as a result of wind loading at
signal, which is highly variable in nature, a the turbine, results in a mean tensile stress or a
representative signal needs to be created. This mean compression stress.
signal has to contain far less individual variable
amplitude cycles while containing the same (history • Exceptional high response (months/years): the
dependent) fatigue damage. stresses during exceptional events, e.g. starting,
stopping, emergency stops and exceptional high
Since existing structures offer the most insight into winds, may be much larger than during normal
the dynamic response of the structure to certain operation of the OWT.
environmental conditions, their measured response
data will be used to develop the reduction algorithm. Depending on the level these phenomena are
Thereby, the actual measurement responses of an contributing to the load sequence effects, they are
OWT are coupled to measured environmental and taken into account when developing the reduction
or operational conditions in order to gain inside into algorithm. This way, the reduced signal will still
the following load sequence phenomena and their contain the information required to estimate the
effect on the OWT’s fatigue life: load sequence effects.
The reduction strategy The developed algorithm, which creates a

The prosed reduction strategy consists of a couple representative load history signal, will be
of steps. First of all, filtering is to be applied to experimentally validated. The load sequence history
the original signal, to remove noise and very small obtained from the reduction algorithm will then
stress cycles. This is expected to already result in a be used as input for the analytical load sequence
large reduction in stress cycles with only a limited effect model.
sustained error. Reasoning behind this can be found
in the SN-curve, where it is seen that small cycles Implementation into the design stage
will only contribute marginally to the total fatigue However, since in the design of a new OWT jacket
damage sustained over a lifetime. support structure measured structural responses are
not available, the stress history should be determined
The second step is to divide the filtered signal into using design load cases and modelled structural
blocks, each block representing an event. These responses. The main challenge thereby is to relate
events are based on the phenomena described above different design load cases to each other, based on
and their influence on the fatigue damage and the insight gained from measurement data of existing
load sequence effects. These events are also related structures. Furthermore, multiple sequences of
to environmental and operational events e.g. storms, events are to be taken into account.
shutdowns and changes of wind direction.
In the current design standards the number of
The next step is to reduce each block, based on the prescribed load cases for fatigue life prediction may
type of event it represents. The idea is to reduce the be incomplete to match the load sequence
parts that will not induce load sequence effects by a phenomena described above. Therefore, for
great deal, while the events that do are to be looked the development of the load history algorithm,
at more precisely. additional load cases could be added in order to
obtain a sufficiently representative load sequence
Then, the signal is reconstructed, based on the history. The calculated change in crack growth
reduced blocks. In the end, the fatigue damage of due to the load sequence effects could then be
the reduced signal is calculated and compared to the translated into, for example, a shifted S-N curve,
damage estimation form the original signal to verify which in turn can be fit in the fatigue assessment
the correctness of the reduction. tool as shown in Figure 8.
Design load cases
➔ Load history algorithm

➔
Wind, waves, turbine settings

➔
Series of stress oscillations

Design response cases
Retardation model
➔
Stress histories
Rainflow count
➔
Stress range Histogram Shifted S-N curve
(shifted) S-N curve

➔
➔
Fatigue damage
Probability condition
➔
Fatigue lifetime prediction
Figure 8. Proposed innovative OWT fatigue assessment approach

CONCLUSIONS in literature in which the effects of retardation are

Although, maritime and the oil and gas industry has considered. However all of these models represent
proven that currently industry-wide accepted hot spot an approximation of reality. The actual retardation
method of fatigue life calculation delivers structures and acceleration effects are complex and not yet fully
with an appropriate fatigue life, this method may understood. Therefore within the Joint Industry
be regarded conservative for OWT jackets. Real Project FeLoSeFi, a procedure is being developed
response measurements of OWT jackets show clear that does take into account these effects. Thereby,
differences in load and response characteristics a model is developed that considers these load
compared to for example ships and bridges. sequence effects on crack growth with reasonable
The occurrence of large stress peaks and troughs accuracy and with reasonable effort.
and the fluctuations in mean stress values in the
OWT jacket response measurements has indicated One route to produce a feasible add-on to the
their increased sensitivity to load sequence effects, current industry standard in order to include
as in metallic structures, large stress peaks have these retardation effect could be to translate the
been observed to cause a crack growth reduction, calculated change in crack growth due to the load
the so called retardation effects, when compared sequence into, for example, a shifted S-N curve,
with the same loading without these peaks. which in turn can be fit in the current fatigue
Fracture mechanics models have been put forward assessment tools.
ACKNOWLEDGEMENTS
This paper is a result of the on-going Joint Industry Project “FeLoSeFi” between Keppel Offshore & Marine
Technology Centre, Netherlands Organisation for Applied Scientific Research (TNO), Arcelor Mittal and
VGB PowerTech. Professor Johan Maljaars and his team are acknowledged for their technical support.
In addition the authors would like to thank OWEC Tower AS for sharing detailed data of their OWEC
Quattropod® substructure.
AUTHOR’S CONTACT  YSalman@Keppelverolme.nl
REFERENCES
[1] Det Norske Veritas AS, “Offshore Standard DNV-OS-J101- Design of Offshore Wind Turbine Structures”, Det Norske Vertitas, Norway, 2014
[2] Det Norske Veritas AS, Germanischer Lloyd SE, “Recommanded Practice DNVGL-RP-0005-Fatiuge Design of Offshore Steel Structures”, Det Norske Veritas
AS, Germanischer Lloyd SE, Norway, 2014
[3] Maljaars J., Pijpers R.J.M., “FaiMoS WP2 – Fatigue of steel structures”, Internal TNO report, 2012.
[4] Willenborg J, Engle RM, Wood HA. A crack growth retardation model using an effective stress concept. Wright-Patterson: Air Force Flight Dynamics
Laboratory; 1971.
[5] Newman JC Jr. Prediction of fatigue crack growth under variable-amplitude and spectrum loading using a closure model, In: Design of fatigue and fracture
resistant structures, ASTM STP 761; 1982, p. 255–77.
[6] Ray A, Patankar R. Fatigue crack growth under variable amplitude loading: Part I – Model formulation in state-space setting, Appl Math Model 2001;25:979–
94.
[7] Patankar R, Ray A. Fatigue crack growth under variable amplitude loading: Part II – Code development and model validation, Appl Math Model
2001;25:995–1015.
[8] Maljaars J. Pijpers R.J.M., Slot H. Load sequence effects in fatigue crack growth of thick-walled welded C–Mn steel members. International Journal of
Fatigue 79 (2015) 10-24.
[9] Lindegaard P, The Relationship between Turbine and Foundation, Enforcing An Integrated Design Approach, Future Offshore Foundation Forum 2015, MHI
Vestas Offshore Wind

OWEC Pile Stopper Solutions 99
OWEC Pile Stopper Solutions
 Mariano E. OTHEGUY , PhD, MSc 1
Yilmaz SALMAN1, MSc
Gunnar FOSS2, MSc
Torbjørn Ruud HAGEN2, MSc
1
KOMtech Europe
2
OWEC Tower AS
THE COST OF FABRICATION AND INSTALLATION OF OFFSHORE WIND

TURBINE JACKETS is affected by a number of factors ranging from choice of
installation partner and the technology they use to motions during grout curing
and new strict rules applying (e.g. DNVGL). Currently common installation
possibilities include e.g. jackup vessels or floating sheerlegs, tolerances
associated to pre-piling that affect as-driven pile top position and measurements
thereof, etc. As a result, jackets are installed with various impact velocities
and as-driven pile top elevation tolerances that affect design details. In turn,
design decisions affect installation costs. This paper explores a range of design
solutions in the portfolio of OWEC Tower and how they can increase the operability
and reliability of OWT jacket installation while lowering the associated costs.
INTRODUCTION OWT jacket market share
The first Offshore Wind Turbine (OWT) jacket was 25%

installed in 2006 off the east coast of Scotland, in the
20%
Beatrice Demonstrator project (see Figure 1). Ever
since, some 175 OWT jackets have been successfully 15%
installed, of which 86 have been designed by
10%
OWEC Tower.
5%
Although still representing a relatively small part of
the OWT foundation market, jackets are reported 0%
Operating Under construction Consent authorised
to be 10-20% more cost-effective than monopiles Figure 2. OWT jacket % over total offshore wind
for larger distances from shore, water depths and foundation installations as per 2015 [1].
turbine capacities [1] [2] (see Figure 3). Indeed a surge
in projects using OWT jackets is taking place, with 60
two times as many projects under construction as 50
already operating, and 3-4 times consented projects
Water depth (m)
40
as per September 2015 [1], as shown in Figure 2.
30
Jacket design is in continuous evolution in search 20
of ever-lower costs of engineering, procurement and 10
construction (EPC). However, the installation (I) 0

3 4 5 6 7 8
costs have a similar order of magnitude, leading to a Turbine rating (MW)
combined EPCI proportion of approximately 20%
of the Total Cost of Ownership of the wind farm [3] Figure 3. Most cost-effective substructures vs
(see Figure 4). Thus, being able to understand and combinations of water depth and turbine
rating (source: BVG Associates [2]).
Project development and

consenting up to FID
Decommissioning
1% Project management and
1%
Operation and planned
contingencies from FID
maintenance
to WCD
5%
3%
Turbine
installation
3%
Array cables
installation
4%
Array cables
2%
Figure 1. OWEC Tower jacket installed at the Beatrice Figure 4. Distribution of Total Cost of Ownership of a
Demonstrator Wind Farm (courtesy of wind farm from design to decommissioning (source:
Scottish Enterprise). BVG Associates [3]).

have a positive impact on the installation costs puts jacket is installed – the most used technique used
OWEC Tower in an advantage position to reduce for the typically one-off oil and gas jackets). Once
the total cost of offshore wind energy. the piles are driven, the jacket is installed on top
of them, by inserting the docking legs (vertical leg
PRE-PILED JACKET SOLUTION tubular extensions) into the piles sticking up from
The installation of OWT jackets, as well as their the seabed (see Figure 5). The jacket is lowered until
design itself, is based on- and evolved from those of the pile stoppers make contact with the pile tops
oil and gas offshore steel jackets, of which thousands (see examples of pile stoppers in Figure 9 and
have been installed worldwide since the 50’s. OWT Figure 10 below). Pre-piling was pioneered by
jackets are installed on the sea bottom by means OWEC Tower in the Alpha Ventus Wind Farm
of a lowering operation from an installation vessel in 2008.
and a subsequent attachment operation, which
fixes the structure to the seabed and enables load Grouting refers to the way the jacket structure is
transfer from turbine and jacket into the seabed. attached to the piles. After sliding the jacket into
There are two main types of installation vessel: the piles, the annular space between the docking legs
floating sheerlegs and jackup vessels. Daily rates and the piles is filled with grout. Grout is a cement-
for sheerlegs are lower than those of jackup vessels, like hardened substance used to ensure transfer of
however sheerlegs are more affected by wave, compressive shear loads from the jacket into the pile.
current and wind loading. The resulting maximum Appropriate shear keys are built on the inner surface
jacket lowering velocities often lead to dedicated of the pile and the outer surface of the docking leg,
impact energy dissipation in order to protect the which are the two side boundaries that enclose the
jacket from impact damage during installation. grout annulus. These keys, usually fabricated as
circumferential weld beads (see shear keys detail on
The most common mechanism used to attach the docking leg in Figure 9), ensure that an appropriate
jacket to the seabed is piling and grouting. Piling mechanical locking is achieved once the grout is
refers to driving large steel tubular pipes (piles) into hardened, transferring the vertical loads from the
the seabed. This is typically done with a hydraulic jacket leg into the grout as shear forces. There is a
hammer, prior to installing the jacket (pre-piling). variety of grout materials in the market with various
An appropriate piling template is required in order advantages and prices. OWEC jackets are designed
to ensure that the piles are driven at the right location so that the cost-effective Portland cement can be
within the required tolerances. For the number of used to deliver the required ultimate and fatigue load
jackets typical of an offshore wind farm, pre-piling bearing capacity, reducing costs at the grout side.
is cost-effective and environmentally friendly in
comparison with post-piling (piles driven once the The pre-piling sequence is shown in Figure 5.
Piling vessel
Hydraulic hammer
Pile
Piling template
Figure 5. Pre-piling sequence, from left to right – 1) Piling through template from floating platform (also possible from
jackup vessel); 2) As-driven pile tops, with applying tolerances: Tpos –horizontal position–, Tt –tilt–,
Tpe –elevation– and Tpem –elevation measurement–; 3) Grouting the annulus between the docking leg and
the as-driven pile –EAC vertical relative movement is limited for the first 24h of grout curing.
Early Age Cycling of grout and DNVGL limits OWEC Tower pile stoppers
Grouted connections in the offshore wind industry The installation of the OWEC jackets is aided by the
have also been evolving along time. After reported OWEC pile stoppers. The stoppers stop the docking
failures of grouted connections in monopiles were leg from further penetration into the pile, keeping
acknowledged and published in 2010 [4], substantial it at the right elevation to ensure jacket verticality.
research work was conducted on grouted connections. They also facilitate contact and load transfer from the
One of the focus points was the so-called Early jackets into the piles, prior to further and definitive
Age Cycling (EAC), i.e. cyclic movement range of jacket-to-pile joining attachment. The pile stoppers
one tubular of the grouted connection relative to have four major functions:
the other during the first 24h of grout curing. As
a result, the DNV-OS-J101 standard was updated • Impact load
in May 2014, requiring that “the possible [axial] Potential and kinetic energy exchange lead to
movements between the inner and the outer tubular certain impact load during jacket landing on
steel members during the 24-hour period after the pile tops. The stoppers protect the jacket
grouting [EAC] shall be […] limited to maximum from structural damage during landing by
1 mm by implementing suitable means” [5]. dissipating the impact energy in the form of
elastic and thermal energy. This helps avoiding
Pile top elevation measurement unnecessary structure overdimensioning and
After piling, the resulting absolute pile top elevation associated costs.
usually varies across the piles driven for each one of
the jackets. Depending on a number of factors where • On-bottom stability
uneven soil resistance plays an important role, pile The stoppers ensure that all legs make
top elevation variation within the piles corresponding full contact with the pile tops at all times,
to a single jacket is typically in the region of tens of avoiding rocking (pivotal movement on the
centimetres. This means that the pile tops may be leg-pile contact points of only two jacket
anywhere within that window. After piling and prior legs). This includes compensation of pile
to installation, the absolute elevation and inclination top elevation differences by tailor-made
of all pile top sections are measured with a precision steel plating (shimming). Between landing
of few millimetres. and grouting the jacket will be subjected
Figure 6. OWEC Quattropod® jacket design examples for 6MW turbine on 40m water depth: with turbine and on
piles (left); showing further and alternative details including J-tubes, cathodic protection, boat landings,
optimised midsection design, pile tops sticking up from the seabed and pile stoppers (right).

to environmental loading. Under these Such a bracket easily delivers the low EAC
circumstances, overturning and the derived required for grout curing. However, landing
loading must be avoided. Also, especially impact on the pile tops requires more attention,
during grouting and grout curing, sliding on because impact velocities typical of installation
the pile tops must also be looked at. All three with floating sheerlegs may lead to local yield
on-bottom stability requirements: full support damage at the stopper. A jackup installation
on legs, overturning and sliding, are related to vessel can typically deliver the needed low
certain sea state limits, thus weather window. impact velocity. However, the higher cost
of this type of installation vessel must be
• EAC vertical relative movement taken into consideration when assessing the
A low EAC vertical relative movement is project-wide cost-effectiveness of this stopper
required to enable appropriate grout curing solution. Also, its inherent high stiffness limits
(see above). This is achieved by the stopper the jacket’s ability to compensate for pile top
by means of providing an appropriate vertical elevation differences and ensure on-bottom
stiffness. stability. As a result, for 4-leg jackets and including
typical structural flexibility, the pile top elevation
• Fatigue load transfer measurement precision would need to tighten
If the stoppers keep in touch with the pile to values ~5-10 times those seen to date.
top after grouting, certain cyclic loads will This, however, does not occur to 3-leg jackets
inevitably travel through them. In order to because 3 legs have always full contact in all
avoid overdimensioning and associated costs, their supports. Additionally, attention must
the stoppers need to be much more flexible be paid to avoid fatigue damage due to the
than the hardened grouted connection so that bracket remaining in stiff contact with the pile
negligible fatigue loading goes through them. top along the service life of the jacket. This
is challenging but may be feasible depending
The OWEC Pile Stopper comes in 3 variants, on the project particulars, eventually requiring
namely: local plate thickening or/and other specific design
1. Bracket Stopper for 3-leg jackets techniques.
2. Elastomeric Stopper for 3- and 4-leg jackets
3. Hydraulic Stopper for 3- and 4-leg jackets
Bracket Stopper
The simplest possible stopper consists of a reinforced
steel bracket attached to the docking leg, which
fulfils its functions by means of metal-metal contact
between jacket and pile (see Figure 7).
Figure 7. Impression of the docking leg with Figure 8. Screen capture of single Bracket Stopper FE
the Bracket Stopper. model, showing the most stressed area in red.
The OWEC Hydraulic Stopper increases the allowable

sea state during grouting and grout curing, leading to
approximately 35% less weather downtime.
Elastomeric Stopper interrelated, in such way that keeping EAC<1mm

The OWEC Elastomeric Stopper has been is technically feasible but often requires improving
successfully installed in 83 OWEC jackets to the pile top elevation measurement precision to
date. This stopper consists of a single- or double- values ~2-3 times tighter than seen in past projects.
ring reinforced steel structure incorporating a This may eventually occur, albeit probably at
number of parallelepipedic elastomeric bearings higher installation costs. After grouting, fatigue
radially arranged under the stopper lower ring load transfer is very low, due to the stopper stiffness
and around the docking leg (see Figure 9). being a fraction of that of the grouted connection.
The bearings are vulcanised to surface-treated steel For example, the stiffness of the stopper shown
plates on both faces to ease further welding to the in Figure 9 is ~70 times smaller than that of the
stopper structure. Once the elevation of the pile corresponding grouted connection. This means that
tops has been measured, steel plate shimming takes only ~1.4% of the fatigue load is going through
place. The final match between stopper and pile the stopper. Thus, achieving satisfactory fatigue
top, however, will still be affected by the tolerance lives of structural details of the stopper without the
of the pile top elevation measurement, which tends need of overdesign.
to be in the region of several millimetres.
Careful design and choice of design constraint
The first requirement of the stopper is to protect cases are key to a successful stopper and jacket
the jacket from landing impact loading during installation. The elastomer chosen, the bearings
installation. The impact energy is first converted number and dimensions, the compression at
into elastic energy in the elastomer bearings and each bearing, and very significantly on the leg-
then dissipated as heat. During the ~0.1 seconds pile eccentricity and pile tilt, it all determines
that the impact typically takes, the load level at all the stopper stiffness, which in turn must satisfy all
structural details must be lower than the structural four functions of the stopper. The design of the
safety limit. After the impact, the jacket must attain stopper is best addressed by computer simulations
and keep on-bottom stability and at the same time using suitable software packages with models
ensure a low EAC. On-bottom stability and EAC are capturing the jacket rotation and own elasticity,
together with hydrodynamic drag and inertia
forces as well as the elastomer non-linearity.
Hydraulic Stopper
In order to keep EAC values below 1mm during
grout curing in a wide range of installation variables,
OWEC Tower in partnership with KOMtech
Europe have come with a robust innovative solution:
the OWEC Hydraulic Stopper.
The OWEC Hydraulic Stopper combines elastomeric

bearings with hydraulic cylinders, making an
EAC<1mm compatible with high lowering
velocities and also with large pile top elevation
differences. A set of elastomeric bearings and
hydraulic cylinders are fixed to a steel structure
similarly as in previous stopper designs (Figure 10).
The cylinders are connected to a hydraulic system
including power and control piping and redundant
Figure 9. Image of an OWEC Elastomeric Stopper valves, Hydraulic Control Units (HCUs) on jacket
featuring 12 bearings – note the grouting
lines, shear keys, leg-pile spacing plates,
deck, electronic spirit level and a containerised
shim plates and elastomeric bearings, local generator set for power supply, plus a
amongst other details. remote monitoring unit to be operated from

the installation vessel. The functioning principle All equipment is relatively small and fits well on
is as follows: board a typical installation vessel or barge. All
1) During jacket lowering and landing, the HCUs and the generator set can be removed from
cylinders are retracted behind the elastomeric the jacket deck and be reused in another jacket.
bearings. This way these can take and
dissipate the landing impact energy. During the impact, the hydraulic cylinders are
retracted and the elastomeric bearings absorb
2) Subsequently, stable weight compression the impact load protecting the jacket structure
is reached at the bearings. Right after, the (see Figure 10). The design and behaviour of
cylinders are stroked down so that the the bearings and load transfer steel structure are
load is removed from the bearings and similar to those of the stopper incorporating
transferred into the cylinders. only the elastomeric bearings. However, in this
case these bearings are used only for the landing.
3) The jacket is then lifted a few millimetres Hence, on-bottom stability relies on the hydraulic
using the hydraulic cylinders, up to jacket jacks and is compatible with the pile top elevation
levelling and approximately uniform measurement tolerances as seen in past projects.
loading on the cylinders. Afterwards, the
cylinders are locked using a valve system at EAC is easily kept <1mm by means of the intrinsic
the jacket deck so that no further hydraulic high stiffness of the locked hydraulic cylinders.
fluid can flow back from them. The fluid Furthermore, limiting EAC during grouting and
locked in the cylinders acts like a very stiff grout curing typically requires a weather window of
linear spring, which keeps the EAC at very 24h with significant wave height (Hs) ≤1.5m. The
low levels, enabling an appropriate grout higher stiffness of the Hydraulic Stopper allows the
curing at the jacket-pile connections. grouting operation to be limited by the grouting
vessel (Hs=2-3m) and the grout curing to occur
4) By the time the jacket rests on the locked with EAC<1mm in sea states in excess of Hs=3.5m.
cylinders, a pressure monitoring device In fact, this limit is often set by jacket rocking and
connected to a transceiver is left on deck or sliding, because these become the design drivers.
in order to remotely keep track of the EAC This extended sea state limit leads to significantly
during the first 24hrs of grouting curing. reduced weather downtime. For example, in a
Figure 10. OWEC Hydraulic Stopper - see elastomeric bearings compressed (middle) and uncompressed with
hydraulic cylinders taking over the load (right).
typical 45-jacket project in the Scottish North Sea, CONCLUDING REMARKS

increasing the allowable sea state from Hs=1.5m The current trend in water depth and turbine
to 2m during grouting and to 3.5m during grout rating are leading to a steady increase in projects
curing leads to approximately 35% less weather demanding OWT jackets, whose supply and
downtime1, meaning some 16% savings in vessel installation cost represents approximately 20%
total hire time and associated costs. of the total lifetime costs of an offshore wind
farm. The current portfolio of OWEC pile stopper
After the grout is cured the cylinders are no longer solutions incorporating bracket, elastomeric and
useful and the hydraulic circuits are unlocked. hydraulic technology, fits all the currently used
This removes any resistance and load bearing at the installation methods and turbine sizes and are
jacks, hence any load transfer through the stopper. designed to be tailored to the project particulars,
leading to cost savings at several stages of the
Any environmental risk of hydraulic fluid leakage combined process of supply and installation.
during the service life of the OWT is mitigated by
using off-the-shelf environmentally-friendly fluids. Our work at OWEC Tower and KOMtech
Europe is on-going to push the technology limits
The Hydraulic pile stopper includes an option and further reduce the cost of offshore
where the hydraulic cylinders are designed to wind energy.
take the impact loading. This option raises the
requirements on the cylinders but reduces steel and
eliminates the need for the elastomeric bearings,
reducing overall costs.
1 Simulations carried out using ABPmer SEASTATES Weather Downtime Express Services [6] for two installation scenarios with weather limit non-exceedance
probability 50%: 1) Installation and grouting from jackup vessel; 2) Grouting from dedicated grouting vessel. Reduction in weather downtime was found to be
similar and approximately 35% in both cases, leading to a reduction in total hire time of 13-19%.
AUTHOR’S CONTACT  mariano.otheguy@komtech.com.sg
REFERENCES
[1] Wind Energy Update, “WEU Offshore Foundations & Supporting Structures Report”, FC Business Intelligence Ltd, 3 September 2015.
[2] BVG Associates, “Offshore wind cost reduction pathways - Technology work stream”, June 2012.
[3] BVG Associates, “Offshore wind: Industry’s journey to £100/MWh”, May 2013.
[4] J. Deign, “Monopile failures put grout in doubt”, Wind Energy Update, 18 March 2011.
[5] DNV GL AS, “Offshore Standard DNV-OS-J101: Design of Offshore Wind Turbine Structures”, p. 166, May 2014.
[6] ABP Marine Environmental Research Ltd., “SEASTATES Weather Downtime Express Service”, 2016. Available: http://www.seastates.net/weather-
downtime/.

Weight Monitoring System 107
Weight Monitoring System
 Syed ABIDI, MSc, B.Eng

Dr. Bernard HOW, PhD, B.Eng
Keppel Offshore & Marine Technology Centre
Satish MENON, MSc, B.Eng*
Umair SULTAN, MSc, B.Eng*
Dr. Matthew QUAH, PhD, CEng, CMarEng, FIMarEST
Dr. FOO Kok Seng, PhD, B.Eng
* Offshore Technology Development Pte Ltd

REAL-TIME DISPLAYS OF THE DISTRIBUTION OF WEIGHT APPLIED TO

LEGS OF A JACK UP RIG HELPS IN MAKING RIG’S STABILITY DECISIONS.
This can also assist in the protection of valuable jacking equipment, which can
be damaged due to overloading if the loads are unevenly distributed, and also
for the protection of the deck structure. If a system has the capability to
generate a comprehensive report, it helps the barge engineer to perform daily
load sheet calculations.
WEIGHT MONITORING SYSTEM KEY FEATURES

The Weight Monitoring System has the capability Real-time Displays of Calculated & Enhanced
to incorporate comprehensive loading information Monitoring System (EMS) Loads
from various sources and calculating the loads exerted The WMS is capable of communicating with on
on individual legs. Operator has the option to operate board control systems to incorporate information
the WMS in any of the two modes, real-time or into the Human Machine Interface (HMI).
overridden mode. In real-time mode, the distributed Capability of the WMS to communicate with on
weight is calculated based on the user inputs for board jacking control system has made it possible to
liquid and solid variables as well as the changes in the monitor the individual leg loads based on the load
tank levels. In the overridden mode, if consolidated transducers readings, installed on each leg. To
liquid and solid variables are known, then they can monitor the leg loads, the WMS accounts for
be inputted directly and the distributed weight will distribution of total weights on the jackup which
then be calculated based on those inputs. Barge includes the weight of the hull, cantilever, drill floor,
engineer and other concerned personal can print out solid variable loads, liquid variable loads, preload
the load configuration for each operation for their ballast, legs and spudcans, together with their respective
record and use. lateral, transverse and vertical centers of gravity.
Real-time displays of the distribution of weight applied to

legs of a Jack Up rig helps in making rig’s stability decisions.
Figure 1. Weight Monitoring System Interface.

Weight Monitoring System 109
Figure 2. Overview of the Tanks On-Board.
Tank Table Database and Configuration Utility Features

The WMS has the option of a customized tank The WMS provides additional facilities for the
table database based on the jackup design. This operator such as data logging, historical trend views
comprehensive database running in the backend etc. Operator can extract the data in csv format
enables user to manually update the tank levels for any past duration for further processing and
to calculate the equivalent weight of the liquid performing meaningful analysis. Operator has the
distributed on each leg. Subsequently, the backend luxury to zoom in to the historic trends of the key
algorithm helps in accurate calculations of the total parameters such as leg loads.
load exerted on each leg. The WMS also provides
the customization option to connect to the on- CONCLUSION
board tank gauging system. If the communication The WMS is a support system which can aid in
protocols are known, a communication link can monitoring the changes in the loading conditions
be setup to fetch the real-time tank level values of the rig. The WMS facilitates easy visualization
from the existing system. This can reduce the task for the operator and allows a quick glance to
of manually changing the values of the levels if determine if the individual leg loads are within
required by making use of the on-board sensors. prescribed limits.
AUTHOR’S CONTACT  syed.abidi@komtech.com.sg
REFERENCES
[1] IADC/SPE Paper 135970, an Integrated System for Improving Geotechnical Performances of Jackup Rig Installation
[2] JACK UP UNITS: A Technical Primer for the Offshore Industry Professional, by Vazquez, J.H., Michel, R.P., Alford, J.H., Quah, C.K. and Foo, K.S. Copyright 2005.
Numerical Prediction of Wind Loads on
Topside of a Semi-submersible Platform 111
Numerical Prediction of
Wind Loads on Topside of
a Semi-submersible Platform
 LIU Jing*, PhD, M.Eng., B.Eng

MA Peifeng*, PhD, M.Eng., B.Eng
XU Haihua*, PhD, M.Eng., B.Eng
Ankit CHOUHARY*, M.Sc., B.Eng
Anis HUSSAIN*, M.Sc., B.Eng
*KOMtech
COMPUTATIONAL FLUID DYNAMICS (CFD) MODELLING HAS BEEN

CONDUCTED TO PREDICT WIND LOADS ON A SEMI-SUBMERSIBLE
PLATFORM. The same scale (1:200) of the platform and the flow chindition as
in the wind tunnel testing have been considered in order to mimic the physical
testing phenomenon. The unsteady Reynolds Averaged Navier-Stokes (RANS)
model is utilized to calculate the transient flow field of the semi platform.
The k-ω sst turbulence model is employed for capturing the turbulence.
The trimmered grid is used in the computational domain, and the effect of the
grid resolution on the accuracy of the solution is investigated.
The obtained numerical results are validated by the wind tunnel test data in the
different wind directions. The force coefficients and overturning moment are
quantitatively compared. Furthermore, the contribution of each components of
semi-submersible platform are discussed for future optimization.
INTRODUCTION not effective and economic compared to numerical

An offshore semi-submersible platform usually modelling method. Alternatively, CFD method
comprises buoyancy pontoons, columns, deck, costs relatively low as mainly requires coputer
helideck, living quarters, gangway and etc. It works resource and manpower. Additionaly, no constraints
in the sea environment and provides relatively stable on the size of the modeling geometries, either model
platform for drilling/production operations. In scale or full scale are acceoptable.
recent years, maintenance requirement to the FPSOs
is increasing because some FPSOs are reaching In recent decade, more and more engineers pay
their end of working design life. Semi would be a attention to utilize a powerful tool CFD method
satisfied solution to implement this task because to predict the wind and current loads on offshore
of its mobility and capability of working in deep vessels as the fast development of the computer
water. However, its station keeping and stability are resources. A lot of complicated physical problems
the most vital issue during operation. Threat mainly can be solved and the results are acceptable in
comes from extermal forces acting on semi platform the industrial application. For example, the
due to random wind, current and wave. wind load on the offshore vessels were calculated
numerically in the reference [3]. It has been pointed
The wind load contributes around 25% of the out that the wind profile has major effect on the
total lateral loads and even higher in the cyclonic numerical results. Both wind load and the wake
condition [1]. It is essentially important to accurately field behind the vessels were discussed and a
predict the wind drag and corresponding overturning good agreement with the wind tunnel testing
moment in terms of semi-submersible station data was achieved. However, they only focused on
keeping and stability. Overestimation may cause steady state flow instead of transient flow. So the
the reduction of the semi loading, sea environment vortex shedding behavior may not be captured. The
working limitation, waste of the material and etc. study of the scale effect between the wind tunnel
The over design of the mooring system and thrusters testing and the actual sea trial were executed by
could be also yielded because the station keeping Christian [10]. The CFD method can also predict
is usually achieved by employing mooring lines the shielding and lift effect of the helideck as
and thrusters to balance the external forces in the discussed in the published literature [11].
sea environment. Therefore, the inefficient design
leads to resource waste, low industry profit and low In this article, the numerical method is used
productivity. In contrast, underestimation is much to calculate wind load on the topside of a semi
dangerous and threatens the life. platform. The wind tunnel test condition and
physical definition on the loads is described in the
In practice, the wind load is usually estimated second section. Followed by that is the numerical
through the rules like ABS (American Bureau of methodology description and the wind profile
Shipping) based on the empirical method in the benchmarking in order to eliminate the effect
offshore floater design. A description on the society of inlet velocity on the results. Then the effect
rules and the wind tunnel testing is addressed in the of the grid resolution on the results is presented
reference [2]. The overall wind loads are calculated in the forth section. The comparison between
by summing up the forces of each component the numerical results and the tunnel test data is
and taking into account of shapes, projected area, performed in terms of different wind directions of
elevation level, and drag coefficient. This method 0°, 45°, and 90°. Contribution of each structure
is efficient but usually lack of accuracy, so it is component on drag and lift are discussed by
really hard to meet the designer’s requirement having a physical insight on the flow behaiours.
and the actual working conditions. Therefore, the Influence of the helideck on the lift force is also
designed semi platform has to be verified again investigated. The obtained results can provide a clue
through wind tunnel testing. A literature review to optimize the semi platform design on mimizing
has been presented by Gomathinayagam et al. on the wind load. The conclustions are given in
wind tunnel testing [1]. However, this procedure is the final section.
WIND TUNNEL TESTING The airflow with approximate atmospheric wind

The wind load test was carried out in BMT Fluid profile was used to mimic the actual semi-submersible
Mechanics Limited (BMT) boundary layer wind working in the sea environment. The density and
tunnel at a scale of 1:200 for the semi platform. The dynamic viscosity of air are ρ=1.2269kg/m3 and
overview of the semi geometry is shown in Figure 1. µ=1.7984 Pas at temperature 15°C, respectively.
The testing semi was at the maximum operational The wind force coefficients CFx CFy and CFz in x-, y-
draft. The waterline was align to the tunnel floor and z- directions acting on the semi are defined as,
surface, namely the topside of the semi is above the
floor and embedded part of column and pontoon (1)
are under the floor. The wind direction is fixed. The
Semi was mounted on a turntable (4.4m in diameter)
in order to vary the heading angle continuously from (2)
0° to 360°. The semi together with the mounted
measurement device can rotate in a counter clock-
wise during experimental testing. The definition of (3)
wind direction, force, and moment axis system are
indicated in Figure 2. The wind load were measured
with a six-component high frequency piezo-electric where Fx, Fy and Fz are forces at three axis directions,
force balance mounted under the model. The origin V is the wind velocity at a reference elevation
of the moment is at the center of the deck. of 10 m (0.05 m for model scale). Ax, Ay and Az
are projected area normal to the flow direction at
different heading angles. The moment coefficients
with respect to the origin of the semi in x-, y- and
z- directions are defined as,
(4)
(5)
(6)
Figure 1. Scale model of a semi-submersible platform

mounted in the wind tunnel.
0º
Fz
Aft
My
Mz
90º Stbd 270º Port
Mz Fy
Mx
Fx My Water Line
Fx
180º Mx
Fwd (a) (b)
Figure 2. Definition of the wind direction and force and moment axis system at the top view (a) and side view (b).
Mx, My and Mz are moment. Mx = Fz y - Fy z, The independence study of influence of Reynolds

My = Fx z - Fz x, and Mz= Fy x -Fx y . Lx, Ly number (Re) on the wind load was carried out
and Lz are reference length. The Reynolds number in the wind tunnel firstly. A number of wind
(Re) is calculated as speeds 5 m/s, 10 m/s, 15 m/s, 20 m/s, 25 m/s,
30 m/s, 32 m/s was used. The corresponding
(7)
Re numbers versus force and moment coefficient
are shown in Figure 3. The trend becomes
Lref is reference length of the semi, and its length
consistent after the wind speed higher than 15 m/s.
is from front of helideck to the gangway. The
Thus, the wind speed 30 m/s was selected to do
overturning moment can be calculated as,
experimental testing for other wind directions.
(8)
Wind Direction 0º Wind Direction 0º

Cx Cy Cz Mx My Mz
Moment Coefficient
Force Coefficient
0.00E+00 5.00E+04 1.00E+05 1.50E+05 2.00E+05 2.50E+05 3.00E+05 3.50E+05 4.00E+05 4.50E+05 5.00E+05 0.0E+00 5.0E+04 1.0E+05 1.5E+05 2.0E+05 2.5E+05 3.0E+05 3.5E+05 4.0E+05 4.5E+05 5.0E+05
Re Re
30 m/s 30 m/s

Cx Cy Cz Mx My Mz
Moment Coefficient
Force Coefficient
0.0E+00 5.0E+04 1.0E+05 1.5E+05 2.0E+05 2.5E+05 3.0E+05 3.5E+05 4.0E+05 4.5E+05 5.0E+05
0.0E+00 5.0E+04 1.0E+05 1.5E+05 2.0E+05 2.5E+05 3.0E+05 3.5E+05 4.0E+05 4.5E+05 5.0E+05
Re
Re
30 m/s
30 m/s
Cx Cy Cz Mx My Mz
Moment Coefficient
Force Coefficient
0.0E+00 5.0E+04 1.0E+05 1.5E+05 2.0E+05 2.5E+05 3.0E+05 3.5E+05 4.0E+05 4.5E+05 5.0E+05 0.0E+00 5.0E+04 1.0E+05 1.5E+05 2.0E+05 2.5E+05 3.0E+05 3.5E+05 4.0E+05 4.5E+05 5.0E+05
Re Re
30 m/s
30 m/s
Figure 3. The effect of Re on the force and moment coefficient.
CFD method is less costly and capable of providing a

physical insight of flow characteristics. The wind load
contribution of each component of the platform can be
readily predicted whereas it is much more challenging
in model testing. This is crucial in optimizing a platform
design to achieve minimal wind load.
NUMERICAL METHODOLOGY Rhino. The small structure pieces like handrails were
Problem Description ignored in the modelling.Their contribution to the
The same dimension of the scale model and Re total wind load is assumed to be negligible. The
(wind speed is 30 m/s) as in wind tunnel were applied cantilevers of two cranes were simplified as a long
in the CFD simulation. The topside of the semi and solid beam with square cross section (Figure 5). The
part of hull were simulated at the maximum drag coefficient of the original cantilever is assumed
operational draft as shown in Figure 4. The CAD to be 0.5 [2]. So the width and height of the cantilevers
model was prepared with the commercial software were reduced to be half as shown in Figure 5.
Figure 4. Geometry of the semi-submersible platform.
(a) (b)
Figure 5 (a) the original cantilever of the crane and (b) the simplified shape of cantilever of cranes.
Computational Domain and Numerical Method This economic method costs less in computational
The computational domain has the same dimension time and computer resources. Second order implicit
as the wind tunnel except the length is to be half discretization scheme was implemented in this
of tunnel length, and namely 4.8 m (W) × 2.4 m transient case calculation. Closure of the k-ω SST
(H) × 7.5 m (L). The blockage ratio is about 3% (shear-stress transport) turbulence model was used
between the semi and the wind tunnel. Blockage to take account for the turbulence. This model
effect is not to be significant. For the computational performs better over k-ε one in handling flow
domain, there is four times semi length spacing in separation over smooth surface. The turbulence
the upstream and ten times in the downstream. intensity is set at 7%, and turbulence length scale is
The applied boundary conditions are indicated in 0.005 m for the scale model. All y+ wall treatment
Figure 6. The tunnel floor is assumed to be a no-slip is used and it is capable to handle both fine mesh
wall. The atmospheric velocity profile discussed in and coarse mesh near the wall. Thus, prism layers
section 3.4 was used at the inlet. The commercial aligned to the walls were created and tunnel floor to
software Star-CCM+ was utilized to create grid and minimize numerical diffusion. The most fine grid
solve the flow field. Trimmed grid was generated fall in the near wall region which mainly consists
for storing the values of the fluid parameters at three layers [4][6]. The viscous sub-layer or laminar
different coordinates. High grid resolution was layer is resolved when y+ ≤ 5 in CFD solver. The
created towards the rigid body in order to capture buffer layer employs wall function when 5 < y+ <
the steep variable gradients precisely and represent 30. The last one is logarithmic layer which applies
the geometry details as well such as gangway logarithmic law y+ ≥ 30. y+ is a dimensionless wall
handrails and trusses, the relatively small pieces. A distance and is approximately equal to ((ρuτ ∆d))⁄μ,
full discussion is provided in section 4.1. here uτ is frictional velocity, and ∆d is first prism
layer thickness. The creating of the prism layers are
The three-dimensional RANS equation models through defining total thickness, stretching ratio,
were utilized to solve the flow field with the Finite and number of layers.
Volume Method (FVM) in a segregated manner.
Atmospheric Wind Profile
There are two ways to define inlet velocity in the
CFD simulation to benchmark the actual wind
profile in the wind tunnel. The first one is to give
a uniform velocity at the inlet boundary, and allow
the flow to develop by itself. However, the length
of the upstream should be long enough to achieve
fully developed situation. The second method is to
use measured wind profile and apply it to the inlet
boundary. The length of the upstream should be
not more than 4 times semi length, otherwise the
velocity will deviate far from the measured profile
Figure 6 . Computational domain and applied boundary
[1]
. The second method was adopted in this article.
condtions.
The trimmed grid was generated in the
computational domain without the semi platform
as shown in Figure 7 and the total cells count is 6
million. The flow direction is parallel to x-axis. The
fine grid was created in the critical region the tunnel
floor and the turntable where the semi was mounted
in order to capture the flow details. The same grid
Figure 7 . The cross-sectional view of the
generation strategy is used as in the reference [4].
computational mesh at y = 0 for wind profile Transient flow condition was simulated in this case.
benchmarking. The time step is 0.01s.
The atmospheric wind profile of ISO is applied and

its velocity U(z) at different sea leave z is defined as:
0.3
Measured Data (9)
ISO Wind Profile where U0 is the reference velocity at a reference
3%
sea elevation z0. C is a U0 dependent coefficient,
0.25
if z0=10 m. The profile from
-3%
Eq. (9) is plotted in Figure 8 (a) together with the
measured data which falls in ±3% band range.
0.2 The wind tunnel height 0.3 m is mainly concerned.
Height Above Sea Lever
It should be noted that the total height of the

scaled semi from floor to the top of antenna is
H = 0.21 m. The region of -0.5 m < x < 0 is the
0.15
semi platform location. The user defined function
of Equation (9) is applied at the inlet of the wind
tunnel at x = -2.5 m. CFD results of wind profile
0.1 along the tunnel height within 0.3 m is also
plotted in Figure 8 (b). The green curves are
± 4% band of the measured data. The profiles at
x = -0.5 m and x = 0 fall in the band except
0.05 Hc < 0.04 m which is within the column height
0.0435 m (model scale). The relative error is
approximately less than 6%. That means that
0 the deck and living quarters of the semi fall in
0 0.5 1 1.5 2 more accurate elevation level from 0.0435 m to
U/U0
0.21 m where resists high percentage of wind
load. Thus, it is reasonable to use the ISO wind
(a)
profile to calculate the wind loads since the
column contributs less over the total resistance.
0.3
Measured Data
X=-2.49
X=-0.5
0.25 X=0
4%
4%
Height Above Sea Lever
0.2
0.15
0.1
0.05
0
0.0 0.5 1.0 1.5 2.0
U/U0
(b)
Figure 8. (a) Comparison between measured data and ISO wind profile; (b) benchmarking numerical results of the
wind turnnel with the measured data.
RESULTS AND DISCUSSION domain grid remains the same as in Figure 9(b).
Grid Independence Study and y+ Discussion The values of aspect ratios and the overall number
Influence of computational grid resolution on of cells for different cases are tabulated in Table
CFD results will be discussed in this part. The 1. The stretching ratio of prism layer is increased
generated grids for different cases are presented in to 1.2 from 1.1 which is sufficient as pointed out
Figure 9. For case 1 and case 2 (Figure 9(a) (b)) the in the reference [7] in capturing the steep gradient
grids were purposely refined in the computational and discontinuities for FVM. Number of prism
domain away from walls in order to study influence layers is 6 determined by applied turbulence
of domain grid on the accuracy of the wind loads. models [6]. The total prism thickness depends on the
Case 3-5 (Figure 9(c)-(e)) are zoomed in grid of Re (Reynolds number). Control of the maximum
the region indicated with a red rectangle of core/prism transient ratio (σ) is applied in case 5
Figure 9(f ) a cross-sectional view at z = 0.06 m and case 6. It guarantees the transient grid from
(z is elevation direction). They are differentiated prism layer to neighbouring domain grid is not
by aspect ratio of the first prism layer while the dramatically stretched.
Table 1. Number of grid cells and the prim layer setting.
Cells Prism Prism 1st Prism Layer Aspect Ratio of

Layers Stretching Thickness (m) 1st Layer
Case 1 9M 2 1.1 4.0x10-4 3.9
Case 2 23M 6 1.1 4.0x10 -4
2.0
Case 3 15M 6 1.2 3.78x10-5 41.3
Case 4 22M 6 1.2 3.78x10-5 20.7
Case 5 40M 6 1.2 3.78x10 -5
10.3
(a) (b)
(c) (d) (e)
(f)
Figure 9. (a) case 1, aspect ratio = 3.9; (b) case 2, aspect ratio = 2.0; (c) case 3, aspect ratio = 41.3; (d) case 4, σ = 6,
and aspect ratio = 20.7; (e) case 5, σ = 5, and aspect ratio = 10.3; (f) cross-sectional view of the
computational grid of z = 0.06 m plane.
For case 1 and 2, prism stretching ratio and the case 2 as shown in Figure 10 (b) comparing with
first prism layer thickness are fixed. Cells number case 1 in Figure 10 (a), for example, the helideck
is increased in case 2 through refining the domain and the main deck surface. In orer to investigate the
grid and increasing prism layers. The surface influence of y+ values on the wind load prediction,
grid is also refined in case 2 due to the reduction different ranges of y+ was considered. Case 1 and
of the aspect ratio which is defined as the length 2 is in the range of y+ = (0, 30) (Figure 10(a) and
over the width of a control volume. More details (b)) while case 3 - 5 is in the range of (0, 5) (Figure
on y+ distribution can be captured with grid in 10(c)-(e)). The reduction of y+ range was achieved
by reducing the 1st prism layer thickness. Hence,

the aspect ratio is higher in case 3 - 5 with much
thiner 1st prism layer thickness (Table 1). It has
to be pointed out that the surface grid is refined
gradually from case 3-5 by control the ratio σ. This
ratio is off in case 3, and 5 and 6 in case 4 and 5,
respectively. In Figure 10 (c) - (e), y+ distribution
shows that the lower the aspect ratio, the more
details calculated as highlighted with the red dashed
circles. It seems that high aspect ratio can smear the
y+ discontinuities.
Figure 10. y+ values of the different cases with respect to the prism layer setting in Table 1: (a) y+ = (0, 30), σ is off; (b)
y+ = (0, 30), σ is off; (c) y+ = (0, 5), σ is off; (d) y+ = (0, 30), σ is 6; (e) y+ = (0, 5), σ is 5.
Convergence of the simulation is determined by significant comparing with the measured data.
monitoring total forces and moments acting on the Thus, it is reasonable to refine the prism layer in
semi-submersible platform. An example of case 4 is case 3-5 to resolve the viscous sub-layer region for
presented in Figure 11. The slight fluctuations are capturing steep flow variable gradients and reduce
observed and this might be attributed to the vortex numerical diffusion. The thickness of 1st prism
shedding behind the object. The computational layer was reduced by 3.78×10-5 m (Table 1). The
time is less than 12 hours with 72 cores on HPC number of grid cells was increased from case 3 to
(High Performance Computers) clusters. The case 5 by reducing the aspect ratio near the wall as
results are fully converged after about the physical shown in Figure 9. Finally, 8% relative difference
time 0.6s. The wind load are calculated by averaging for case 4 was achieved with respect to the tunnel
the values between 1 - 2 s. meaured data. There is only 1% difference between
case 4 and case 5. It could be concluded that the
The values of Fx, Fz, and My, are used for grid influence of both y+ and grid density near the wall
independence study as they are vital important to is significant for wind load prediction. The grid in
evaluate the semi-submersible at 0o wind direction. case 4 are acceptable to do further study at other
Fx is the drag force which affects station keeping. different wind directions.
Fz is the lift and mainly affects semi stabilization
especially in harsh sea environment. My accounts There is 13% difference in drag for case 4 comparing
for the overturning moment. The numerical results with experimental results as shown in Table 2.
are tabulated in Table 2 and relative difference It might be due to simplification to the scale
with respect to the measured data is also given. For model geometry and some small pieces of the
case 1 and case 2, no significant changes for Fx geometry excluded, for example, the protection
and My. This proves that the number of grid handrails. It also might be 6% mismatch of ISO
cells 9 million is fine enough to capture the wind wind profile comparing to the measured data as
load except the lift. Although Fz is improved to pointed in early section in Figure 8. The difference
-32% from -57% for case 2, the difference is still in Fz and My may also due to this mismatch.
(a) (b)
25 5
20 4
15 3
Moment (Nm)
Force (N)
10 2
5 1
0
0
-5
-1
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Physical Time (s) Physical Time (s)
Figure 11. Monitors of (a) force and (b) moment on the semi platform versus physical time for case 4.
Table 2. Comparison between experimental measured data and CFD results on Fx, Fz, and My at 0º wind direction.
Difference between CFD results and experimental results

Wind Loads
Case 1 Case 2 Case 3 Case 4 Case 5
Fx (-13%) (-13%) (-10%) (-13%) (-14%)
Fz (-57%) (-32%) (-32%) (-8%) (-9%)
My (-15.6%) (-14%) (-12%) (-5%) (-6%)

Investigation of Numerical Results at 0° resisted by the semi is computed by summaring

Heading Angle of pressure and friction drag.
Contours of velocity magnitude of case 4 are
presented in Figure 12. The wind direction is 0° and It is well known that magnitude of the wind load
aligns to the x-direction. All plotted cross-sectional mainly depends on the pressure acting on the semi,
views are parallel to the flow direction. The flow the projected area and the elevation level of the
separations are observed at the sharp corners of the structure components. Table 3 tabulates drag
semi-submersible such as front of helideck, living contribution of each component of the semi platform
quarter and bottom of deck which are indicated with at 0° wind direction. Living quarter, equipment
red curves (Figure 12 (a) (b)). The negative pressure rooms and deck resist almost 69% of the total drag
consequently forms at the corresponding area as due to their big projected area. Two cranes contribute
shown in Figure 12 (c) (d), contours of pressure almost 1% higher than four columns beneath the
distribution. The big blue area behind the semi is main deck, althrough columns have big projected
the wake field with full of revers flow. Drag forces area. This might be because of their high position.
Table 3. Contribution of each component of semi platform to the drag at 0º wind direction.
Case 4 Parts Contribution (Fx) Total

1 LQ & Equip. Rooms 47.6%
2 Main Deck 21.2% 87.1%
3 Column 8.8%
4 Cranes 9.6%
5 Left Platform 5.2%
9.4%
6 Helideck 4.2%
Others (Gangway, Right platform,
7 3.4% 3%
Antenna, bollard…)
(a) (b)
(c) (d)
Figure 12. Contours of velocity magnitude (m/s) at different cutting planes paralle to the incoming flow direction:
(a) y = 0 and (b) y = 0.1 m. Pressure distribution (Pa): (c) y = 0 and (d) y = 0.1 m.
Table 4 tabulates lift contribution of each contribute 6% higher lift than that of the helideck.
component at 0° wind direction. 44.5% is On the contrary, the presence of the helideck seems
contributed by helideck as the pressure difference lead to the high pressure on the deck surface as
is big between under and above its surface as shown illustrated in Figure 12 (c), the dashed rectangular
in Figure 12 (c). The yielding upward force would area. For the deck area far from the helideck in
intend to raise the helideck. Hence, it creates high spanwise, the pressure is negative as shown in
overturning moment which threatens the stability Figure 12 (d). If the pressure is high on the deck
of the semi platform. So the lift caused by the surface, the force is downward, otherwise upward
helideck should be minimized through parametric and thereby high lift. Downward force benefits to
study. In the other hand, it is interesting to find that the stability of the platform. Therefore, the presence
the main body including living quarter, equipment of the helideck seems to be helpful to reduce the lift
rooms, deck and beneath connected columns only of the main body.
Table 4. Lift contribution of each component of the semi platform at 0º wind direction.
Case 4 Parts Contribution (Fz)
1 LQ & Equip. Rooms, Deck, and Columns 50.2%
2 Helideck 44.5%
3 Cranes 3%
4 Gangway 1.1%
5 Antenna 1.1%
6 Others (Platforms, Bollard…) 3.2%
(a) (b)
(c) (d)
Figure 13. (a) Countours of velocity magnitude (m/s) of cross-sectional view at z = 0.12334 m (scale model) across
the air gap between helideck and living quarters; (b) Pressure distribution (Pa) of plane z = 0.12334 m;
(c) Countours of velocity magnitude (m/s) of plane z = 0.12757 m just above the helideck; (d) Pressure
distribution (Pa) of plane z = 0.12757 m.
(a) (b)
(c) (d)
Figure 14. Countours of velocity magnitude (m/s) at different elevation levels: (a) z = 0.025 m and (c) z = 0.1 m.
The pressure distribution of (b) z = 0.025 and (d) z = 0.1 m.
Contours of velocity magnitude and pressure high dynamic pressure. Figure 15 illustrates the
distribution of sectional views under the helideck vorticity magnitude of the different cutting planes
planes are presented in Figure 13(a) (b) and above parallel to the flow direction. The high values are
the helideck planes in Figure 13(c) (d). As mentioned mostly caused from the sharp edges of the platform.
early, high pressure is visible in the area behind Addtionally, the vorticity is extreamly stronger at
the helideck as illustrated with a dashed rectangle high elevation level, for example, caused by antenna
(see Figure 13 (b) (d)). Then downward force is and cranes.
created. However, the averaged flow speed is lower
under the helideck than that of above the helideck.
So the high lift is generated by the helideck. In a
conclusion, the helideck has advantage of reducing
lift on the main body and meanwhile generating
high lift on itself.
Cross-sectional views of contours of velocity

magnitude and pressure distribution at different
elevation levels z = 0.025 m and z = 0.1 m are plotted
in Figure 14. Asymmetric wake behind the columns
might be due to vortex shedding (Figure 14 (a)). The
flow slows down towards the columns and speeds
up between columns formed accerated zone in the
flow direction (Figure 14 (b)). Big wake behind LQ
is formed with full of reverse flow (Figure 14 (c)).
The pressure is higher in front of LQ and lower in Figure 15. Vorticity of the difference cross-sectional
the wake region (Figure 14 (d)). So LQ has to resist view parallel to the incoming flow.
Investigation of Numerical Results at Table 5 shows contribution of each component

Different Heading Angles of the semi platform on the resultant horizontal
Comparison between the numerical results and forces at the different heading angles. The percentage
experimental data on wind load coefficients in x-, of the wind loads acting on the living quarter
y- and z- directions are shown in Figure 16. Four and equipment rooms decrease gradually from
difference heading angles of the semi platform 0° to 90° because of reduction of the projected area.
were simulated at 0°, 45°, 60°, and 90°. One In contrast, the loads on the deck, columns and
quarter of (0, 360°) can represent a full circle cranes increase from 0° to 90° due to the
wind loading characteristics. Both trends are in enlarged projected area. Contours of velocity
a good agreement. But force coefficients CFx , CFy magnitude at different heading angles are
and CFz were underestimated if compared with presented in Figure 17. The blue region
the wind tunnel data except CFx and CFz at 90°. always represents the wake. The percentage of
This could be the simplificaton made to the scale the drag acting on the columns is higher at 45°
model geometry. For moment coefficients, it also and 90°. The left platform only contributes
quantitatively shows a close agreement between wind 0.2% drag at 45° and 90° due to the shielding
tunnel data and numerical results except My at 90°. effect from LQ. Table 6 shows the lift contribution.
Table 5. Resultant horizontal force contribution of each part of semi-submersible at different wind directions.
Parts 0º 45º 90º
1 LQ & Equip. Rooms 47.6% 37.8% 24.3%
2 Main Deck 21.2% 21.6% 27.5%
3 Column 8.94% 17.9% 17.4%
4 Cranes 9.6% 10.6% 18.9%
5 Left Platform 5.26% 0.2% 0.2%
6 Helideck 4.2% 8.7% 7.3%
7 Others (Gangway, Right platform, Antenna, bollard…) 3.56% 6.3% 6.1%
Figure 16. Comparison between wind tunnel test results and the CFD numerical results.
In all, about 90% is contributed by the main body platform. For overturning moment, 37% comes
and the helideck. But at 0°, nearly 45% lift caused from helideck as shown in Table 7. The helideck
by the helideck, which must be critical wind does not play significant effect on the lift at other
direction in terms of the stabilization of semi heading angles.
Table 6. Lift contribution of each part of semi-submersible at different wind directions.
Parts 0° 45° 90°
1 LQ & Equip. Rooms, Deck, and Columns 50.2% 76% 79.7%
2 Helideck 44.5% 18.4% 10.4%
3 Cranes 3% 1.8% 2.8%
4 Gangway 1.1% 1.2% 4%
5 Antenna 1.1% <0.2% <0.05%
6 Others (Platforms, Bollard…) 3.2% 2.4% 3%
Table 7. Overturning moment (Nm).
0° 45° 90°
Helideck 1.2 1.01 0.344
Semi-submersible 3.2 3.25 2.47
Percentage 37.1% 31.1% 13.95%
(a) (b)
(c) (d)
(e) (f)
Figure 17. Contours of velocity magniturde of the cutting planes perpendicular to the columns: (a) 0º, z = 0.025;
(b) 0º z = 0.1; (c) 45º, z = 0.025; (d) 45º z = 0.1; (e) 90º, z = 0.025; (f) 90º z = 0.1.
A comparison between experimental measured be due to the simplification to the geometry in

data and the CFD results on the overall horizontal the modelling and mismatch of the wind profiles
force, overturning moment and lift at different below the deck. Figure 19 shows the contours of
heading angles are also shown in a radar chart in velocity magnitude and pressure of cross-sectional
Figure 18. Their trends have a good agreement. views at x-z plane. The pressure is lower at 45° than
The discrepancy is about 20% at 45° for the other heading angles because of the corner of the
horizontal force. As mentioned early, this might platform facing the wind at this heading angle as
Figure 18. Radar chart of the horizontal force and overturning moment.
(a) (b)
(c) (d)
(e) (f)
Figure 19. The velocity magnitude (m/s) and the pressure distribution (Pa) at different heading angles:
(a) 0°, y = 0; (b) 0°, y = 0; (c) 45°, y = 0; (d) 45°, y = 0; (e) 90°, y = 0; (f) 90°, y = 0.
Table 8. Quantitative comparison of forces and moment at different heading angles.
Heading angles (degree) Difference (%)

0º -12.7
Horizontal force per speed2 45º -20.
90º -17.5
0º -8.49
Lift per speed2 45º -7.55
90º 33.21
0º -5.40
Overturning moment per speed2 45º 1.06
90º 13.13
illustrated in Figure 14 (a). However, the project CONCLUSION

area is quite large, and thus the wind load is much CFD method was validated with the wind tunnel
higher according to equation (1). Figure 20 shows measured data in this article. The same scale
the streamline and the pressure distribution over the model 200:1 and Re number were taken as in the
semi platform. The fierce turbulence is observed at experimental test. The wind profile applied at the
the 45° in the wake. High pressure is observed at the inlet in the numerical simuation was benchmarked
surface of three columns and large area of the deck over the measured data. It was proved that the ISO
and LQ as shown in Figure 20 (b). This is critical profile is sufficient to be applied in the semi platform
angle for the platform working in the strong wind simulation. After that grid independence study was
sea environment. carried out, it was found that both domain grid
and y+ had significant effect on the accuracy of the
wind load calculation. The value y+ is decreased by
reducing the aspect ratio and the first layer thickness
which results in the comparable results to the
experimental data, especially the lift values.
The CFD numerical results at the different heading

angles were also discussed. They show the same
trend as in the experimental measured data. Less
20% difference was obtained. The discrepancy is
(a) 0º
(b) 45º (c) 90º
Figure 20. Stream lines and the pressure distribution on the semi surface at heading angle of (a) 0º, (b) 45º, and (c) 90º.
observed over the horizontal drag force which could high overturning moment. The physics behind the
be caused by the simplification made to the semi hydrodynamics was discussed in detail.
geometry, for example, the ignoring of handrail and
modification to the cranes. However, this would not In a word, CFD method plays more and more
be the obstruction on continuous study on the semi important role in the engineer application. It is less
optimization in future as the results are comparable costly and capable of providing a physical insight
to the experimental data on the overturning and of flow characteristics. The wind load contribution
lift magnitude. Moreover, the contribution of each of each component of the platform can be readily
component of the semi on the wind load was also predicted whereas it is much more challenging
calculated and analyzed from the numerical results. in model testing. This is crucial in optimizing a
The horizontal force depends on the projected platform design to achieve minimal wind load.
area and the elevation level which is consistent as Moreover, the numerical results provide a physical
the ABS rules. It is also observed that the helideck insight on the flow behavior which is very helpful to
causes high lift at 0°, and consequently creates explore the aerodynamics behind.
AUTHOR’S CONTACT  Jing.Liu@komtech.com.sg
REFERENCES
[1] S. Gomathinayagam, C. P. Vendhan, J. Shanmugasundaram. Dynamic effects of wind loads on offshore deck structures – a critical evaluation of provisions
and practices. J. Wind Eng. Ind. Aerodyn. 84: 345-367. 2000.
[2] ABS, American Bureau Shipping Rules for building and classing mobile offshore drilling units, 2012.
[3] A. Koop, B. Rossin, and G. Vaz. Predicting Wind Loads on Typical Offshore Vessels Using CFD. In Proceedings of ASME 31th International Conference on
Ocean, Offshore and Arctic Engineering, OMAE2012, July 2-6, 2012, Rio De Janeiro, Brazil, OMAE2012-83449.
[4] B. Blocken, T. Stathopoulos, J. Carmeliet. CFD simulationof the atmospheric boundary layer: wall function problems. Atmospheric Environment 41 (2): 238-
252. 2007.
[5] I. Sahin. A survey on semisubmersible wind loads. Ocean Engng. 12 (3): 253-261. 1985.
[6] P. L. Davis, A. T. Rinehimer, M. Uddin. A comparison of RANS-Based Turbulence Modeling for Flow over a wall-mounted square cylinder. http://www.
cdadapco.com/sites/default/files/technical_document/pdf/PRU_2012.pdf
[7] A. Koop, A. Bereznitski. Model-scale and Full-scale CFD Calculations for Current Loads on SEMI-submersible. Proceedings of the ASME 2011 30th
International Conference on Ocean, Offshore and Arctic Engineering OMAE2011, June 19-24, 2011, Rotterdam, The Netherlands, OMAE2011-49204.
[8] Vaz, G., Waals, O.J., Ottens, H. Fathi, F., Le Souef, T., and Kiu, K., Current Affairs: Model Tests, Semi-Empirical Predictions and CFD Computations for
Current Coefficients of Semi-Submersibles. In Proceedings of the 28the international conference on Ocean, Offshore and Arctic Engineering, OMAE2009-
80216, Hawaii, USA 2009.
[9] Marc Belzile, Jeff Patten, Brian McAuliffe, William Mayda, Bernard Tanguay, Review of Aerodynamic Drag Reduction Devices for Heavy Trucks and Buses.
Technical Report, National Research Council Canada. 2012.
[10] A. Christian, S. L. Hvid, P. H. Hughes, M. Leer-Andersen, Wind Loads on Ships and Offshore Structures Estimated by CFD. Proceedings of The 8th
international conference on the behavior of offshore structures. 1997.
[11] Y. Lin, A. Hu, F. Xiong, W. Jiang, Discussion on Key Points of Wind Lad of Jack-up Unit. China Ocean Eng. 28 (1) pp. 127-138. 2014.
Review of Floating Offshore Structure
Concepts for Exploration in the Arctic Region 129
Review of Floating Offshore

Structure Concepts for
Exploration in the Arctic Region
 CHEN Zhuo, PhD, M.Eng, B.Eng

WANG Wenping, M.Eng, B.Eng
Anis HUSSAIN, M.Sc, B.Eng*
* KOMtech
DRILLING OFFSHORE IN THE ARCTIC POSES MANY CHALLENGES TO

DRILLING CONTRACTORS. More capable solutions than available today
are required to enable commercial feasible exploration in the Arctic. The design
of the drilling rig plays a key role in this and new designs for Arctic floaters will
be required to achieve this goal. The key factor to the drilling part is the total
well cost, while in the Arctic the ability to drill one or more wells in one season
or extending the season is the key factor to lower well cost. Efficient Arctic
floater design, which allows maximum use of available drilling season, can
further enhance economics. This paper discusses the main challenges, such as
harsh environmental conditions, ice loads, deep water challenges and their
influence on the design of Arctic Floaters. It also discusses the main principles
applied in development of these designs, related to subjects such as technology
qualification, safety and environmental concerns. Finally, it provides an overview
and outline of available design solutions
INTRODUCTION The United States, Canada, Russia, Kingdom

The Arctic can be defined as areas north of the of Denmark (Greenland), and Norway possess
Arctic Circle (Figure 1), which consists of the the majority of the resource potential within
Arctic Ocean and parts of Alaska (United States), this region. According to a 2008 Circum-Arctic
Canada, Finland, Greenland (Denmark), Iceland, Resource Assessment by the U.S. Geological Survey
Norway, Russia, and Sweden. The arctic oceans are (USGS) (Bird et al., 2008), the Arctic could hold
highly diverse, from shallow continental shelf with about 25% of the world’s remaining undiscovered
less than 100 meters depth and up to a width of hydrocarbons, which is around 525 Billion barrels
700 kilometers, to deep waters with water depths of oil, or oil equivalent for gas(BBOE).
down to 4000 - 5000 meters. The continental
shelf is made up by Beaufort Sea, Barents Sea, Kara The majority of the Arctic resource potential is
Sea, Laptev Sea, East Siberia Sea and the Chukchi expected to be gas with about 30% estimated to
Sea. They surround the Arctic Ocean, which is be liquids. Russia is estimated to have by far the
divided into two basins by an underwater ridge, the largest Arctic resource potential and will continue
Lomonosov ridge. to be a dominant player in Arctic oil and gas
development. When considering only Arctic oil
potential, however, the United States and Russia are
assessed to have approximately equal portions of the
conventional resource potential with approximately
35 billion barrels of oil each, which represents about
15 years of current U.S. net oil imports (National
Petroleum Council, 2015).
It is estimated that approximately 75% of the

total global Arctic conventional resource potential
is offshore. If assume uniform distribution of oil
within all potential provinces and compute fraction
of area of each province in water depth >100m,
it is estimated that there are over 40 billion barrels
of oil in water depth >100m and 75% of these
distribute in 4 provinces, namely, (1) Beaufort
Sea-Canada Basin; (2) West Greenland – East
Canada Basin; (3) East Greenland and (4)
East Barents Basin (J.M. Hamilton, 2011) as
Figure 1. Arctic Circumpolar Map (National Petroleum

Council, 2015).
The United States, Canada, Russia, Kingdom of

Denmark (Greenland), and Norway possess the majority
of the resource potential within this region. According
to a 2008 Circum-Arctic Resource Assessment by the
U.S. Geological Survey (USGS) (Bird et al., 2008), the
Arctic could hold about 25% of the world’s remaining
undiscovered hydrocarbons, which is around 525 Billion
barrels of oil, or oil equivalent for gas(BBOE).
demonstrated in Figure 2. On the other hand, in the Arctic and the potentially transitory nature
Current industry production experience is limited of the current world oil supply/demand situation
to bottom-founded structures and gravel islands may provide a different prospective.
in depths less than 40m. As there are practical
water depth limits for bottom-founded structures Firstly, the cycle of leasing, exploration, appraisal,
in Arctic conditions, floating or all-subsea development, and production, takes much longer
systems are required for water depth beyond 100m. in the Arctic than in other offshore regions.
The expanding search into the deep-water Arctic For instance, Northstar, the only U.S. offshore
is to some extent underpinned by industry OCS Arctic project, took 22 years from lease
confidence that the necessary enabling technology sale to start of production, while recent Gulf of
will emerge if the prize is large enough. Such Mexico deep water projects such as Mars and
confidence is supported by industry’s long-standing Atlantis only took 11 and 12 years respectively.
record of overcoming major technological hurdles The longer time frame required for the Arctic
to safely and economically produce large projects is the result of remoteness, long supply
hydrocarbon accumulations in harsh environmental chains, short exploration seasons due to ice, r
conditions around the globe. In this paper, the egulatory complexity, and potential for litigation.
main challenges for drilling in Arctic deep water The time frame for developing any significant
are reviewed. The main principles and standards offshore Arctic opportunity would likely be
applied to the development of the Arctic floater between 10 to 30 years or longer.
are discussed. An overview and outline of
available design solutions is summarized in the end. On the other hand, according to a recent U.S.
EIA Reference Case outlook for U.S. crude oil
production, total U.S. crude oil production
increased from 5 million barrels per day in 2008
to 8.5 million barrels per day in 2014, and is
projected to increase to a maximum of 9.6 million
barrels per day in 2019. Crude oil imports are
expected to decline from 9.8 million barrels per
day in 2008 to a minimum of 5.8 million barrels
per day in 2019. But in the Reference Case
after 2019, U.S. crude oil production is expected
to decline to about 7.5 million barrels per day
and imports rise to 7.7 million barrels per day
by 2040. U.S. domestic crude oil production is
57% of domestic demand in 2014, but declines
to 49% in 2040, reversing the improvements in
the economy and energy security from the recent
production increase (National Petroleum Council,
Figure 2. Oil Potential in the Arctic deep water 2015).
(J.M. Hamilton, 2011).
Thus, if development starts now, the long lead
MOTIVATION FOR THE times necessary to bring on new crude oil
ARCTIC EXPLORATION production from the Arctic would coincide with
U.S. and world oil prices have dropped significantly a long-term expected decline of U.S. Lower
during the past one year. In this current context 48 production. Given the resource potential and
of increasing oil supply and declining oil prices, it long timelines required to bring Arctic resources
may looks unrealistic to pursue Arctic exploration to market, Arctic exploration today may provide
and development now. However, the long lead a material impact to U.S, as well as to the rest of
times involved in exploration and development the world.
MAIN CHALLENGE FOR THE DESIGN OF late winter to about 4 to 6 million sq. km in late
THE ARCTIC FLOATER summer. The average sea ice extent for the Arctic is
The physical environment in the Arctic presents shown in figure 3 below, with the potential oil and
unique challenges which increase the complexity gas provinces within the extent of at least the winter
and cost of offshore oil and gas development. These sea ice extent illustrated.
challenges include,
The types of ice that influence the design and
The presence of sea ice operation of offshore facilities in the Arctic include
Potential presence of sea ice and/or icebergs is one first-year ice, multi-year ice, icebergs, and drifting
of the most obvious challenges for Arctic offshore fragments of shelf ice known as ice islands, as
drilling. The ice in the Arctic Ocean is mobile due demonstrated in Figure 4. Depending on local
to prevailing winds and currents, and the area that conditions and overall Arctic Ocean ice transport,
it covers undergoes an annual cycle that varies from these primary ice types may or may not be present
about 15 million square kilometers (sq. km) in in a given Arctic basin.
Figure 3. Sea ice extent in summer and winter conditions (Wassink & List, 2013).
First-year ice Multi-year ice ridge Iceberg Ice Island
Figure 4. Ice Features (NPC, 2015).

First-year ice grows each winter to cover essentially such as Franz Joseph Land, Novaya Zemlya, and
the entire Arctic Ocean surface. Depending on Severnaya Zemlya. Icebergs can range in size from
weather conditions, first-year ice will generally grow a few meters to hundreds of meters with a mass of
to a thickness of about 1.5 to 2 meters (m). Wind tens of millions of tons, and they present impact
and current-induced movement within the Arctic risks for facilities and operations. Under the current
pack compresses some zones of the ice to the point climatology, icebergs are not present in the U.S.
of failure, which forms localized thickened areas Beaufort and Chukchi Seas.
of pressure ridges and/or rubble fields within the
surrounding level ice. The thickened features, once Extreme weather
refrozen, present the primary design and operational Similar with the present of sea ice, extreme weather
challenges for a first-year ice environment. conditions are the also the key challenges for the
development of Arctic floater. Ranging from
Multi-year ice is ice that has survived two or more extreme low temperatures to polar lows (local low
melt seasons. Each year around mid-September, pressure complexes which result in very high winds),
about 5 million sq. km of Arctic pack ice persists 24 hours of darkness, fog etc., these pose significant
after the summer melt season and enters the next challenges to operations in the Arctic. These
freeze-up cycle. Surviving first-year ice becomes challenges require rigs to be specifically adapted for
second-year ice; surviving second-year ice begins use in the Arctic. Whereas sea ice conditions can
its third-year and is then defined as multi-year ice. be avoided to a certain extent, extreme weather
During each freezing cycle, multi-year ice thickens can be encountered in the open water season as
and gains strength, and the loosely consolidated well, specifically at the end of the season when the
blocks that once made up the first-year ridge keels temperature already drops while sea ice is not yet
freeze into a solid mass. This multi-year ice presents formed.
greater challenges to design and operations than first-
year ice due to its greater thickness and strength. For Weather conditions in the Arctic also come with
example, multi-year ice that is more than 4 m thick significant uncertainties. As a remote and sparsely
is very difficult to break with even the largest nuclear inhabited area, there is limited (statistical) weather
icebreakers. The largest and oldest concentrations of data available for the Arctic. Also, effects of
multi-year ice lie northwest of the Canadian High climate challenge appear to already have significant
Arctic islands and north of Greenland. Some of this influence on Arctic weather conditions, which make
ice is transported southwestward and then westward the use of the already limited data for predictions
across the Beaufort and Chukchi Seas by the less accurate.
Beaufort Gyre and southward along the northeast
coast of Greenland by the East Greenland Current. Water Depth
Water depth within the prospective Arctic oil and
Ice islands are rare but extreme ice features that gas basins varies from zero to more than a thousand
occur in the U.S. Beaufort and Chukchi Seas. They meters. Most of the U.S and Russian Arctic
are large pieces of thick, multi-year shelf ice that offshore oil and gas potential lies in water depths
breaks off of areas such as the northwestern coastline of less than 100 meters. Water depths offshore
of Ellesmere Island. Ice islands can be tens of meters Arctic Canada and Greenland, on the other hand,
thick and several kilometers across and may become fall off to more than 100 meters closer to shore.
entrained in the Beaufort Gyre and drift into U.S. Water depth predominantly impacts the type of
Arctic waters. drilling and production platforms that can be used.
Beyond about 100 meters, a technology transition
Icebergs are freshwater ice masses that calve off of from bottom-founded to floating platforms would
glaciers that terminate in the ocean. Icebergs do not be required because the overturning forces of the
occur everywhere in the Arctic; glaciers that produce floating ice become too large for practically sized
large icebergs exist on both coasts of Greenland and bottom-founded structures. Unlike for temperate
on most of the islands across the Russian Arctic waters, where floating drilling facilities are routinely
used in thousands of meters of water, the primary in much higher costs for exploratory drilling.
challenge for floating solutions in Arctic shallow Likewise, development technology requirements
water is related to the allowable maximum offset become more challenging and costs increase with
from the well center due to limitations in allowable decreasing open water season.
angle of the Marine Drilling Riser and drill pipe, as
illustrated in Figure 5 below. In shallow water, the Remoteness
resulting allowable offset becomes very small which, The Arctic is a huge area with little or no existing
coupled to the increased loads in sea ice conditions, infrastructure. The supply routes for all activities
poses significant challenges to the station keeping are very long. When moving further north, satellite
system. will disappear, which will be a crucial challenge for
drilling vessel utilizing dynamical positioning to
Open Water Season stay in place and for communication with the rest
Additionally, the length of the open water season— of the world.
the time without ice coverage—has a significant
impact on the types of technologies that can be used Vulnerable Environment
for exploration and development. The length of the Last but not the least, the Arctic is a pristine and
open water season can vary considerably from year vulnerable environment, and therefore requires an
to year. Over most of the U.S. Chukchi Sea lease even higher focus on environmental protection
area, the average open water season is about 3 to 4 compared with conventional drilling operations.
months long, but has been as short as 1 to 2 months. The unique eco system of the Arctic is more
If the open water season is 3 months or more, it may vulnerable to pollution than most other systems
be possible to complete the drilling of an exploration due to the short and less diverse food chains and
well in a single season using conventional technology the difficulties associated with incident response
that would be used in any open water setting. Shorter in Arctic conditions due to the remoteness and the
open water seasons or deeper reservoirs may require possible presence of sea ice. Two main challenges
multiple seasons to complete a single well, resulting can be identified as below,
Maximum 2º (2.5 / 6%) waterdepth for drilling

Maximum 6º - 10º (8 - 12%) waterdepth for disconnect
Figure 5. Offset limitations for floating rigs related to water depth (Wassink & List, 2013).
• Reduction of the operational environmental The scenario includes initial ship/ice edge contact
footprint, i.e. reduction or prevention of emissions over a small area, with growing contact area until
and discharges during regular operations; the entire structural grillage is loaded. The design
ice load is characterized as an average pressure
• Oil spill and incident response. uniformly distributed over a rectangular patch load
of height and width. Since concentrated pressures
REVIEW OF RULES FOR THE DESIGN OF exist within the load patch, peak pressure factors are
OFFSHORE STRUCTURES IN THE ARCTIC assigned to account for pressure concentration on
There is no ready set of standards for the design localized structural members.
of offshore floating structures operating in the
Arctic or ice infested water from the classification The new ice class by the Russian Maritime Register
society. The only relevant rules is the new ISO of Shipping (RMRS, 2010 are named as ARC4
19906 “Petroleum and natural gas industries–Arctic to ARC 9. ARC 4 and ARC 5 result in similar
offshore structures”, International Organization for scantlings as the Finnish Swedish ice class IA and
Standardization, 2010. , however, this standard is IA Super and ARC 9 is the highest ice class enabling
developed primarily for fixed structures. In brief, navigation in all ice conditions. The new rules
there are three “Exposure Levels” (L1, L2, and L3) define the link between ice class requirements and
that depend on factors such as whether the platform the ice conditions in winter/spring or summer/
is manned, whether it is planned to evacuate, and autumn for the different Russian Arctic seas.
consequence categories that relate to the potential In addition, the type of navigation has also an
risk to life and the environment and to possible effect on the ice lass needed and is thus divided
economic loss. The standard was calibrated to target into the following items: independent navigation,
safety levels, using a special calibration exercise. icebreaker escorted navigation, extreme navigation
(average periodicity once in ten years), hard,
The Finnish-Swedish Ice Class Rules (FSICR) can medium and easy navigation (average periodicity
be considered as the industry standard for ships once in three years). The scantling requirements
operating in first year ice. The FSICR have been are mainly based on the ship’s displacement and
developed as part of the Finnish-Swedish winter shape of the hull.
navigation system, because during an average winter
every Finnish port and every Swedish port north The Lloyd’s Register (LR) rules include the
of Stockholm is ice bound. Although these rules Canadian Arctic Shipping Pollution Prevention
appear to be made for level ice breaking, they in Regulations, Finnish and Swedish Ice Class Rules
fact design the ship for brash ice conditions, e.g. a and the Russian Maritime Register of Shipping. LR’s
broken channel in level ice, with the design point rule philosophy is to align with these regulations
being the impact of the ship with the channel edge. supplemented by LR’s service experiences and
technology development in the form of additional
In August 2006, the International Association of notations and guidance. For ships operate in first
Classification Societies (IACS) released the “Unified year ice condition, Lloyd’s Register rules are aligned
with the FSICR. For ships, which transit waters in
Requirements (UR) for Polar Ships”, which which multi-year ice is present, LR rules incorporated
standardized global ice classification specifications. the IACS UR for Polar ships. A notable feature of
A general description of seven polar classes is given the LR rules is that a wider range of shell, including
therein in terms of nominal ice conditions where the bottom shell, is required to be ice-strengthened.
highest class is approximately equal to the highest In addition, LR released the new procedure, Ship
Russian Register of Shipping ice class and the lowest Right Fatigue Direct Assessment (FDA) ICE, in
to FSICR IA Super. For each polar class, factors are July 2011, which assesses the potential for structural
provided to scale design forces according to the fatigue in vessels designed to trade in ice-covered
expected severity of ice loading conditions. The waters and thereby helps to reduce the risk of
glancing shoulder bow impact scenario has been fatigue damage in the hull structures of their ice-
selected as the basis for the scantling requirements. strengthened vessels.
DNV GL has three categories of their Rules for guidelines for the propeller blade design using direct
Classification of Ships for operations in ice. These calculations and ballast system design are provided.
range from the lowest ice classes intended for very The third category is for IACS Polar Class vessels.
light ice conditions to the highest ice classes covering The requirements are in general equivalent to the
year-round operation in all ice-covered waters. The IACS Unified Requirements for Polar Ships. Ships
basic ice classes (ICE-C and ICE-E) are intended designed for ice breaking may be given the additional
for service in waters with light ice conditions. Icebreaker notation. This notation sets additional
This category typically applies for vessels, which requirements for the propulsion machinery system
are not designed for continuous operation in ice, and bow form. Further, requirements for the steel
but need ice class for commercial or operational are provided and for winterization of ships to ensure
reasons. The structural reinforcements for these compliance to cold climate. Three different levels
classes are generally limited to the bow area. The of winterization are defined in the rules covering
structural arrangements are often similar to open operations in light winter to extreme Arctic
water ships, but plating is thicker and intermediate conditions. Additional notations for redundancy,
ice frames are placed between the main frames in noise and vibration and reduction of the risk to the
the waterline at the bow area. The second category environment are provided.
sets requirements for service in the northern Baltic
Sea in the winter or areas with similar ice conditions REVIEW OF ARCTIC FLOATER CONCEPT
to the FSICR. The majority of commercial ice- There are only a limited number of floating
strengthened vessels fall in this category. Additional exploration or production structures that have been
General characteristics
Type: Drill barge
Tonnage: 27,968 GT / 8,391 NT/ 9,902 DWT
Water depth for Operation: 120m
Drilling Depth: 6100m
Displacement: 17,500 tonnes (lightship) / 28,000 tonnes (full)
Diameter: 81 m (266ft) (main deck)
Draught: 8 m (lightship) / 10 - 12.5 m (operating)
Depth: 18.5 m (61 ft)
Ice class: Arctic Class 4 (PC2)
Installed power: Four diesel engines
Propulsion : None
Crew: 108
Diameter 266ft
Liftboats Bow
Liftboats
Liftboats Liftboats
Port thruster
Stern
Figure 6. Drilling Rig Kulluk (Timco and Johnston, 2002).

used in ice environments. Seasonal exploration can winter season. The conical-shaped hull is nearly
be carried out in the Alaskan OCS using drillships circular in plan view and was designed to deflect
and drilling barges and, in areas without multi- ice downward. It has 24 compartments, forming
year ice, semisubmersibles or a TLP. However, 24 sides, and a double hull and bottom to prevent
for exploration, the only location that a floating hull breaches and pollution. The 12-point mooring
structure might be capable of staying on station system was designed to resist the force of 4ft (1.5m)
year-round might be the Bering Sea under light ice of ice moving in any direction. This structure was
conditions. A Semi-rigid Floater structure might exposed to moving pack ice. Active ice management
work year-round under first-year ice conditions around the Kulluk ensured that the ice conditions
but would need to have the ability to disconnect were not severe. This structure was instrumented to
and leave station in the event of potentially higher measure mooring line forces. Wright (2000, 2001)
loads. Floating production systems for the Beaufort summarized the measured forces on the Kulluk.
Sea, Chukchi Sea and North Bering Sea are not Measured loads were up to 4 MN depending upon
considered to be technically feasible, even with the ice thickness, floe size and ice concentration.
continuous ice management. No floating production
structures could be economically designed to stay Drillship Stena DrillMAX ICE
on station with multi-year ice loads found in the The Stena DrillMax Ice is the most expensive
Beaufort and Chukchi Seas, and possibly northern drillship ever built and the world’s first dynamically
Bering Sea depending on local ice conditions. positioned, dual-mast drillship with ice-class
However, floating systems may have some merit in certification, as shown in Figure 7. The ice-
southern Alaskan OCS areas (IMVPA, 2008). strengthened hull unis has been optimized for
Arctic conditions. Six ice-classed thrusters provided
Some selected existing arctic floater design are maximum maneuverability. Anti-icing equipment
summarized as below protects the unit’s anchors, deck piping, lifeboat
escape exits, scuppers and drains while enhanced
Kulluk de-icing machines keeps decks, gangways, and
In 1983, Gulf Canada Resources Ltd. designed and handrails clear.. The drillship is able to break the
built an inverted-cone shaped floating structure ice, although it is not a typical icebreaker. When
(the “Kulluk”) that allowed drilling later into the operating in the Arctic the drillship will be escorted
Figure 7. Drillship Stena DrillMAX ICE (Karl and Sven, 2015).

by icebreakers and/or icebreaking Offshore Supply the promising areas for future development. The
Vessels (OSV) for ice management purpose. When concepts include drillship for year-round operation,
the drillship is operating in other areas like North Semisubmersible, Spar and other floaters for
Sea or Gulf of Mexico there might be a threat due extended operation.
to frequent storms and heavy seas with large waves,
however the DrillMAX ICE can survive waves up An Arctic floater concept shown in Figure 9
to 30 m height. The drillship’s hull form is based has been recently developed by ConocoPhillips
on Stena’s proven DrillMAX design, some topside (Iyerusalimskiy, 2012). The system is a
modifications were included. Because the drillship disconnectable floating platform concept for deep
is likely to operate in the environmentally sensitive water Arctic regions and is claimed to have the
Arctic region, space was created on deck for an capability to perform exploratory and development
extra six-RAM blowout preventer (BOP), providing drilling and to potentially carry out year round
critical redundancy. The drillship is also equipped production in various deep water Arctic locations.
with DP3 station-keeping and related automation
systems provided by Kongsberg for operating in ice
conditions, Knuckle boom deck cranes are designed
for temperature conditions of -30°C.
The Polyarnaya Zvezda (Polar Star)

semisubmersible
The Polyarnaya Zvezda (Polar Star) SSDR (see
Figure 8) was the first to arrive in the Kirinskoye
field. It was developed based on moss CS50 Arctic.
The units are designed for operation in arctic
conditions in the presence of broken ice thickness
up to 70 cm, designed for the outdoor temperature
to – 30oC and are capable of drilling to a depth of
7500 m at a depth of 500 m. The riser and hose
connections are protect from ice impact. The anchor
chain is protected through ice belt and zone exposed
to freezing sea spray.
Recent Arctic Floater Concepts

Several concepts for Arctic drilling and production
systems are being promoted for application in
Figure 8. The Polyarnaya Zvezda (Polar Star) Figure 9. ConocoPhillips Deepwater Arctic Floater
semisubmersible (Karl and Sven, 2015). Concept (Iyerusalimskiy, 2012).
Figure 10 shows a few spar concepts proposed by the drifting ice with thickness up to approximately
Sablok and Barras (2009) and Murray and Yang 2.0- 3.0m.
(2009). The spars are disconnectable upon encounter
of ice features that exceed their design limitations and Figure 12 shows an Arctic drillship concept by
could be constructed of steel or concrete. Gusto MSC (Alain and Remco, 2013) namely,
NanuQ 5,000 TM. It is claimed to be the most
Figure 11 shows an Arctic Semi concept proposed by capable Arctic drillship, capable of extended season
Huisman (Alexei, 2011). The major feature of the to year round operations in up to 4m of multi-year
unit is the combination of exceptional sea keeping ice. This unit is turret moored, with DP capability
characteristics and excellent ice resistance. This is for station keeping during mooring system hook-
achieved by utilization of a dual draft concept. In up. The turret position is selected to combine
open water the unit has a draft which is typical for good weather- and ice-vaning properties with
column stabilized units. In ice the unit is submerged good motion characteristics at the well center, thus
to a deeper draft where a cone shaped heavily allowing both sea ice and open water operations.
reinforced upper structure receives the ice loads. A Suitable for exploration and development drilling,
20 point mooring system is utilized and the floater it is self-propelled and offers ice class up to PC2,
is claimed to be able to withstand the forces from allowing year round access to all Arctic areas.
(Sablok and Barras, 2009) (Murray and Yang, 2009)

Figure 10. Disconnectable Arctic Spar.
Operating in waves Draft Operating in Ice Draft
Figure 11. Huisman Arctic Semi Concept (Alexei, 2011).

Figure 12. Gusto MSC Arctic Drillship Concept (Alain and Remco, 2013).
SUMMARY The ISO 19906 standard issued in December 2010

An attempt was made in this paper to assess the covers a very wide scope and is invaluable to the designers
present status of Arctic floater design and whether and operators of Arctic structures. The standard may
there is sufficient confidence to support oil and still be lacking detail for ice load calculation for
gas exploration and production development of floating systems but that does not reduce its current
the promising and pristine Arctic provinces. As value and benefit to the industry and all stakeholders.
always this is a balance between risks and benefits It is the best available Arctic standard to-date utilized.
for all stakeholders. Depending on the water
depth and the ice conditions, many Arctic designs The current state of existing Arctic technology indicates
have been proven in the past and the experience that a strong capability exists for year-round drilling in
gained is going to be beneficial in designing all the contemplated Arctic regions. As more experience
new structures for deeper waters and more severe and data become available the action calibration factors
ice conditions. and the ice conditions will need to be updated.
AUTHOR’S CONTACT  Zhuo.chen@komtech.com.sg
REFERENCES
[Alain Wassink, Remco can der List (2013). Development of Solutions for Arctic Offshore drilling. SPE Arctic and Extreme Environments Conference. Moscow,
Russia.
Alexei Bereznitski (2011). A Novel design o mobile offshore drilling unit for Arctic conditions. International Conference on Ocean, Offshore and Arctic Engineering.
Rotterdam, the Netherlands
HIS (2014). http://www.ihs.com/products/oil-gas/epdata/sets/international.aspx.
IMVPA (2008). Arctic Offshore Technology Assessment of Exploration and Production Options for Cold Regions of the US Outer Continental Shelf. Prepared for
United States Minerals Management Service. IMV Projects Atlantic Project No. C-0506-15, Technical Report No. TR-001, January 31st.
National Petroleum Council (NPC)(2015). Arctic Potential – Realizing the promise of U.S. Arctic oil and gas resources.
Jed M. Hamilton (2011). The challenges of Deep-Water Arctic Development. International journal of offshore and polar engineering.
Iyerusalimskiy, A., Gu, G., Zou, J., Chainis, J., Kumar, B., and Sayed, M. (2012). A Conceptual Study of Deepwater Arctic Floater for Year-Round Drilling and
Production, ICETECH 2012, Banff, Canada
J.J., Murray and C.K., Yang (2009). A Comparison of Spar and Single Column Floater in an Arctic Environment, OTC 19797, Houston, USA
Karl Ulrich Evers, Sven Hoog (2015). Report on fixed as well as floating offshore structure concepts. Prepared for Arctic Climate Change, Economy and Society
(ACCESS), Project No. 265863
Sablok, A. and S. Barras (2009). The Internationalization of the Spar Platform, OTC 20234, Houston, USA
Timco, G. W. and Johnston; M. E., (2002) . Caisson structures in the Beaufort Sea 1982-199 : Characteristics, instrumentation and ice loads, Technical Report
CHCTR-003

Drillship DLA/SFA Notation Design Review 143
Drillship DLA/SFA
Notation Design Review
 WANG WenPing, M.Eng, B.Eng

Chen ZHUO, PhD, M.Eng, B.Eng
Anis HUSSAIN, M.Sc, B.Eng
STRUCTURAL DESIGN OF DRILLSHIP IS MORE OPERABILITY AND

CONSTRUCTABILITY DRIVEN NOWADAYS, which makes structural design
requirements more comprehensive. Keppel’s first drillship, CANDO, is one example
to reflect this trend. Apart from traditional midship three tanks hold based analysis,
global strength analysis based on full ship model has also been performed during
the project. The strength capacity check has adopted direct load approach to
obtain wave load to apply to global finite element (FE) model. The fatigue
assessment has taken spectral fatigue method for accumulative damage
calculation under given wave scatter diagram. These works lead to two additional
class notations, Direct Load Approach (DLA) and Spectral Fatigue Approach
(SFA). These two notations give additional drillship structural quality insurance to
owners and operators.
Although full ship model verification is a time consuming process, a few major
advantages can be clearly seen. Firstly, complete full ship strength verification
results are available, including vessel’s forward and aft parts. Secondly, topside
modules and main hull interface studies are beneficial from more accurate vessel
deformation load when vessel is at hogging or sagging. Furthermore, this gives
quite accurate results on dealing with local connection details. Some simple
examples are illustrated to help readers to understand the work flow.
In summary, this paper not only gives general review of our past design work,
but also explain a standard full ship verification procedure we developed through
the project. This direct load approach gives the most accurate results on strength
and fatigue requirement. It is an unified solution to deal with any innovative new
ship design.
INTRODUCTION Table 1. Principal Dimension of drillship

CANDO drillship, Keppel FELS new generation
Length Overall 242.2m
drillship design, focus more on vessel’s operability
function and constructability [1]. The design is a Length Between Perpendiculars 232.6m
reflection of many drillship operators’ views and Breadth 42.0m
requirements. Meanwhile, the design also takes Depth 19.6m
into account of construction philosophy on how Scantling Draft 12.0m
to shorten delivery time to compete with several
Block Coefficient 0.83
major competitors.
Function requirements, like large moonpool
This drillship has been designed in accordance opening, deep riser bay in hull, heavy concentrated
with ABS Guide for Building and Classing hook load and stool concept to support topsides,
Drillships [2]. Apart from traditional class notation make the drillship transverse frames quite different
of XA1,DRILLSHIP, notations of Direct Load in terms of loading and connection details along
Approach (DLA) and Spectral Fatigue Approach the longitudinal direction. This generates a big
(SFA) have also been obtained. These additional workload for structural verification comparing with
two notations require designers to perform full ship traditional ships which have quite similar transverse
verification in terms of strength and fatigue. frames throughout the length. This also makes
designers to decide to perform full ship model
Cando drillship (Figure 1) is designed for global strength and fatigue verification.
unrestricted service, including Offshore Brazil
and Gulf of Mexico. It has capacity of performing Comparing with traditional midship three tank
drilling operations in water depths up to hold analysis method, a few major advantages
12,000 ft. The total drilling depth below the can be clearly seen, especially for vessel’s forward
rotary table is 40,000 ft. It has more than 25 years’ and aft parts. Calculation results can give better
fatigue life against given environmental conditions. result indication than just following class rule
The principal dimensions are listed in table1. recommendations. Furthermore, top side modules
and hull interface studies are another beneficial area
in the full ship verification. Direct load calculation
gives more accurate main hull deformation load to
topside modules due to vessel’s hogging and sagging.
Figure 1. Overview of drillship model.

As ship finite element (FE) models are usually SEAS is for environmental severity check, while ISE
analyzed in the details of stiffener panel size [10], thus, is rule based scantling design check. See reference [2]
a FE model with approximate total 500K elements for detail explanation. Here, TSA is the only FEM
(Figure 10) has been created. To manage the model analysis requirement, which is based on three tank
size when dealing with local connection details, hold model. As drillship function requirements
submodeling skills [9] in finite element method has create quite a lot section capacity change in the
been adopted. This is a time consuming verification longitudinal direction along the ship, three midship
procedure. However, it gives owners and operators models (Figure 2) have been prepared instead of
additional quality insurance in terms of structural one. This includes the verification requirement for
strength and fatigue. topside and hull interface analysis. All yielding,
buckling and fatigue need to be verified. However,
This paper is for general design review of our no FE analysis requirement in reference [2] exists
past project work. For simplicity, only a few key for vessel’s forward and aft parts, apart from some
beneficial areas are highlighted. Some simple work slamming load check based on rule requirements.
flow and examples are illustrated to help readers
to further understand the procedure. Please refer For simplicity, no further illustration on traditional
to documents [7, 8] of complete DLA & SFA design procedure is given in this paper. Readers can
verification reports for detail descriptions. On refer to [2] for design requirement in detail.
the other hand, the standard full ship verification
procedure we developed through the project can
be used for any new innovative ship for strength
and fatigue verification against given environmental
conditions.
TRANDITIONAL SHIP DESIGN

PROCEDURE
According to ABS Guide for building and classing
drillship [2], the following three main works need Figure 2. Midship models required in the traditional
to be carried out first during design stages. design procedure.
– Environmental conditions (SEAS) FULL SHIP VERIFICATION METHOD

Obviously, the traditional analysis on three tank
– Initial Scantling Evaluation (ISE) hold model brings in the assumptions for the
boundary condition. This is usually valid for
– Total Strength Assessment (TSA) typical ship. When drillship is designed at function
driven, large stiffness of topside will weaken the
This guide [2] is a comprehensive guide (published rigid assumptions of transverse bulkhead locations.
on 2012) by combining Mobile Offshore Drilling To overcome this, direct load analysis based on
Units (MODU Rules) [3] and Steel Vessels Rules full vessel models including topside as shown in
[4] targeting mainly for drillship. Cando drillship Figure 10 shall be considered as the most accurate
is one of those first batch drillship which has been approach nowadays. A short description of the
designed, built and classed under this rule. procedure is given as below.
Full ship hydrodynamic and global strength verification

gives a unified solution to all innovative new design.
However, this process needs to be performed with
a good plan between accuracy and efficiency.
To achieve full ship global strength analysis, three FEMAP, and then solved with NASTRAN solver.
separate type of model are needed for these analyses: Figure 4 gives a simple work flow for spectral fatigue
- Hydrodynamic model assessment procedure.
- Mass model
- Structural FE model Direct load approach (DLA) is the strength
verification for assessment on buckling and yielding
They work as different roles in the calculation. capacities. Spectral Fatigue Approach (SFA) is the
Please see Figure 3 for overall illustration of the work fatigue analyses for assessment of all welding and
flow. PRECAL [5] is used for hydrodynamic analysis. connections under given site wave scatter diagram.
Mass model is a reflection of weight distribution Among both DLA and SFA notations, the wave
along the ship. Structural FE model is prepared in loadings has been calculated from direct wave load
analysis, of which the results have been processed in
both short term response and the long term response
A. Determine wave load B. Determine instantaneous analysis to define maximum responses. Then, the
& ship motion in load components
PRECAL design wave selections are needed to cover drillship
critical frames to ensure the extreme loading being
experienced in the calculation. The design loading
conditions, such as hull girder bending and torsion
D. Strength assessments & E. FEMAP
Dynamic Loading model & moment, shear forces and accelerations has been
acceptance criteria
Nastran derived to reflect the design basis requirement.
Stress solver
Deflections Full ship verification gives many beneficial areas

Structure in this project. Here, discussions mainly focus on
Fatigue
C. Determine structural
wave load determination, and topside and main hull
Assessment
responses integration. They are the main focus in this design
F. Spectral fatigue procedure
review.
Figure 3. Work flow for DLA & SFA.
Short term: Given sea state, wave direction and operation condition
Wave Spectrum Stress Transfer Stress Spectrum

Stress PDF
Function
Long term: All sea states, wave directions and operation conditions:
Fatigue Damage
Scatter Diagram
Operation Condition (Time Ratios)
Direction. Prob.
Figure 4. Fatigue damage assessment work flow.

forward pipe rack deck
mud module
living quarters
substructure & two modules nearby
Figure 10. Full ship model and its major topside module local models.
Local Analysis
Critical Areas
Critical areas in
Detailed FE Model transverse web
frame
Critical areas in
horizontal supporting
member on transverse
bulkhead
• Loc 1. Moonpool aft corner at main deck
• Loc 2. Riser bay aft corner at main deck
• Loc 3. Moonpool aft at box bottom
• Loc 4. Moonpool fwd at box bottom Fatigue Analysis

• Loc 5 & 8. Sub aft leg interface with hull
• Loc 6. Fwd deck column interface w/hull
• Loc 7. Long. CL aft bracket at riser bay
• Loc 9. Fr236 trans bracket at riser bay
Figure 11. Sample for local and fatigue assessment detail models.
HYDRO DYNAMIC LOAD DETERMINATION decided by a simple beam theory with global
The linear sea keeping program PRECAL [5] stress concentration factors. Such as, large moon
calculate the transfer function for 6-DOF motions, pool structure, riser bay inside the hull. Local
accelerations and sectional forces. The transfer longitudinal stress can be determined according
functions give the response per unit wave height as to design hull girder bending moment considering
a function of wave period for each wave heading. stress concentration effect. Such as, axial stress
The peak of transfer function shows the wave period and vertical bending stress due to hull girder end
which gives the largest response for the heading pressure and tank boundary conditions.
considered. The transfer functions combine with
a wave spectrum shape, a significant wave height The vessel operation modes, like transit or
and zero crossing periods to calculate the short- operational, also produce different critical dynamic
term statistics. Combining this with the long-term load at different locations. In general, transit
distribution of the different sea state, represented by condition gives more critical load for vessel overall
the wave scatter diagram, the long-term response because the wave loading are much more critical
can be determined. due to unrestricted service requirement. However,
the midship moonpool area is more critical for
To perform structural design, a set of design waves operational condition because concentrated load
have been selected to simulate the critical load which for substructure and moonpool rails generate many
vessel experience in sea. Wave profile stretches as high stress peaks to be paid attention to.
shown in figure 5&6 have been used to make the
loading condition more close to the real scenarios. In ship design procedure, one still water design
This improves the loading determination accuracy envelop was firstly prepared to cover all the
for full ship model. operation condition and transit conditions.
However, full ship model verification requires actual
OVERAL STRUCTURAL LOADCASE loading conditions for the direct load calculation.
CONSIDERATION This requires DLA calculation to be performed in
In general, buckling check or shear capacity check more loading conditions to covers all possibility
are governing in scantling of longitudinal members. of loading conditions of drilling, transit, stand-by
Overall longitudinal stress can be preliminary and survival condition.
To achieve this, a review has been made based on

the vessel stability related loading conditions to
determine the still water bending moment and
shear force envelops which ensure that the maximum
and minimum bending moments and shear
force are covered in the calculation. A sample
indication for bending moment comparison plot is
given in Figure 7.
Figure 5. External hydrostatic pressure mapping for a
load case.
Figure 6. External hydrodynamic pressure mapping for

a load case. Figure 7. Design envelop for still water bending moment

Our present DLA verification takes total five loading DLA gives quite good confirmation for the midship
conditions. Those are, two loading conditions area which has been firstly done by TSA. In any case,
for transit (one for maximum hogging and one the basic notation is XA1,DRILLSHIP, thus, the
maximum sagging), two loading conditions for designed scantling is not less than TSA requirement
operation (one for critical moments and one for of dillship guide.
critical shear), and one more loading conditions
for survival. These make the DLA verification quite The verification results for SFA are screening of
comprehensive. fatigue life as shown in Figure 9b. Consequently,
only critical details need to be further studied as
RESULT PRESENTATION FORMAT local strength and damage assessments.
The verification results for DLA are yielding and
buckling check for all the drillship panels. Figure TOPSIDE DESIGN AND INTERFACE
8 gives a typical full ship deformation shape under WITH HULL
various wave loading conditions. Figure 9a gives an Construction friendly driven design gives quite
example for strength verification in terms of yielding big challenges in the stool designs which support
and buckling unity check for all the panels. all topside modules. However, full ship verification
prove to be the right tool to solve this type of problem.
Our fore and aft strength have been detailed verified This is mainly because direct load calculation can
in DLA calculation stages. The DLA verification gives accuracy vessel hogging and sagging induced
results give less conservative result when the FE deformation load as shown in Figure 8.
model at beyond 0.4L amidships. Furthermore,
Figure 9a. Sample for yielding & buckling unity check.
Figure 9b. Example for fatigue life screening for

Figure 8. Full ship deformation shapes. hull-topside stools.
FE models have been prepared to consider directly Atlantic, Gulf of Mexico, Offshore Brazil, West
topside effects on the hull interface. ABS drillship Africa and UK Sector. Reasonable selection of the
guide [2] gives quite strict requirement for the profiles of loading conditions and wave conditions
interface between hull and topside. To ensure this, are of critical importance in fatigue analysis.
several detail hull-topside interface modeling has The wave scatter diagrams of the main areas
been done and submitted to class for review. Figure have been determined by the typical database
10 gives the main topside FE models has been accumulated from previous design requirements of
studied in DLA & SFA notation. Each model has owners and operators.
approximately 300K ~ 500K element in model size.
To manage the computational time and FE model Generally details calculation procedure for SFA
size, sub-modeling scheme had to be used, where is in accordance with ABS guide [6]. More than
target hull girder loading and acceleration effects sixty local connection details have been refined
were transmitted to sub-models through their with plate thickness size mesh to perform detail
boundary nodes. fatigue calculations. SFA notation is the most time
consuming part for this project.
As for stool fatigue assessment, ABS rule [2] does not
give clear definition for long-term Weibull distribution CONCLUSION
parameter for hull-topside interface location, This paper is for general design review of our
therefore, performing SFA to the topside stools can verification practice for class notations based on
develop this parameter, and give much more accurate full ship model. Full ship verification is a time
results than TSA in fatigue life assessment. consuming process, should only be used for
verification stage. However, it gives better accuracy
LOCAL STRENGTH AND DAMAGE in terms of strength and fatigue capacity. This is
ASSESMENT additional quality insurance in the design.
Several critical locations show peak stresses. Their
fatigue lifes are also required to be studied in details Innovation design and function requirements might
to meet 25 years requirement. Such as: large opening lead to comprehensive structure design challenges.
corner brackets and substructure leg connection to Full ship model based global analysis is a preferred
hull and etc. Figure 11 gives some examples for the solution.
local analysis location and high stress profiles.
With the verification procedure used for DLA and
Typically, critical strength locations are related to SFA notation, our design team has established a
critical fatigue locations. To ensure fatigue life for standard practice can be used for any new innovative
these local details, SFA is required. The calculation ship with direct load calculations approach in terms
covers various wave conditions, such as, North of given environmental conditions.

ACKNOWLEDGEMENTS
Authors would like to thank all engineers’ effort in working as a team for the drillship project delivered.
Meanwhile, thanks for ABS class review team for valuable design comments and software support.
AUTHOR’S CONTACT  Wenping.wang@keppelfels.com
REFERENCES
[1] Designing Next Drillship, How Keppel Sees Exploration, Development and Completions in Ultra Deep Water, April 2004, SNAME, Marine Technology.
[2] ABS Guide for Building and Classing Drillship, August 2011.
[3] ABS Rules for Building and Classing Mobile Offshore Drilling Units (MODU Rules), 2012
[4] ABS Rules for Building and Classing Steel Vessels, 2012
[5] PRECAL V4.0 User’s Manual, Dec2000, Marin
[6] ABS Guide for Fatigue Assessment Of Offshore Structures. April 2003.
[7] Global Hull Verification with DLA Method (Vol 1&2), Jan 2014
[8] Fatigue Analysis with Spectral Fatigue Approach (Vol 1&2), Sept 2014
[9] FEMAP user manual, V11.0,
[10] Guidance Notes on Safehull Finite Element Analysis of Hull Structures, Dec 2004
©2016 Keppel Offshore & Marine Technology Centre Pte Ltd (“KOMtech”)
31 Shipyard Road
Singapore 628130
Tel: (65) 6591 5450
Fax: (65) 6265 9513
GCC04.2016
Email: KOMtech@keppelom.com
Co Reg No: 200615559N

KOMTECH Technology Review 2016

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

KOMTECH Technology Review 2016

Uploaded by

Copyright:

Available Formats

Technology Review 2016

Chow Yew Yuen

1 Jackup Going On Location Analysis

2 Design Considerations of DOL And VFD Jacking System

3 The Impact of Refrigerant Selection and Precooling Condition

4 Experimental Investigation on Flow Induced Motion of

5 Effect of Drillship’s Moonpool Shape on The Resistance

6 Design and Installation of an Active Wave Absorber

7 Portable Fuel Tank - The Application of a New Method of

8 Innovative Fatigue Prediction Model for Improved Fatigue

9 OWEC Pile Stopper Solutions

10 Weight Monitoring System

11 Numerical Prediction of Wind Loads on Topside of

12 Review of Floating Offshore Structure Concepts for

13 Drillship DLA/SFA Notation Design Review

Managing Director’s Message

Since the establishment of Keppel Offshore & Marine Technology Centre

As we move into 2016, which continues to be challenging for the offshore

We are doing precisely that by prudently allocating resources and looking

Technology Review 2016 highlights KOMtech’s research in multidisciplinary

My appreciation goes to the industry partners, partnering universities and

 Paolin TAN, Xiaohong CHEN, Qing YU

Michael PERRY, Han MU

Ted CHANG, Marc WONG

INSTEAD OF USING THE TRADITIONAL WAVE HEIGHT LIMIT, OFTEN

KEY WORDS: Jackup Going On Location; Jackup Installation; Design Wave

ACKNOWLEDGEMENTS We appreciate support from ABS, Bennett Offshore

INTRODUCTION between the spudcan and the seabed were explicitly

simplifies the leg lowering process to a number of

(1) • The jackup is subjected to continuous excitation

Efforts are also made to develop simplified methods that

70 m Mass Roll Moment of Inertia - Ixxg

39.3 m (129 ft) Leg

Pitch Moment of Inertia - Iyyg Yaw Moment of Inertia - Izzg

FIGURE 6 depicts the normalized (by their mean

Normalized Zero-Mean FWD Leg Impact Force - Stiff Soil

Normalized Zero-Mean FWD Leg Impact Force - Stiff Soil

Figure 7. Design Wave Curves for the Jakcup

FWD Leg GZ Force (kN) - Stiff Soil

LC1 - 01 (0 deg) LC1 - 02 (45 deg)

FWD Leg GZ Force (kN) - Soft Soil

FWD Leg GZ Force (kN) - Stiff Soil

Vertical Impact Force (MN)

Vertical Impact Force (MN)

AUTHOR’S CONTACT  Michael.Perry@KOMtech.com.sg

 Bernard Voon Ee HOW*, PhD, B.Eng

Satish MENON*, MSc, B.Eng

Kok Seng FOO*, PhD, B.Eng

* Keppel Offshore & Marine Technology Centre Offshore Technology Development

THE JACKING SYSTEM IS A KEY MISSION CRITICAL AND SAFETY CRITICAL

In this paper, characteristics and peculiarities to be considered for both DOL

KEY WORDS: Jacking System, Direct-On-Line, Variable Frequency Drives

In the following sections, the paper will try to

Main Switchboard Main Switchboard Main Switchboard

In elevated condition, if the jacking system loss control

motor overheat sensor

Figure 3. Simplified representation of V/f control and considerations.

JACKING AND LOAD SHARING VFD WITH TORQUE FOLLOWER / SPEED

Follower 1 Follower 2 Master Follower 1

Speed Speed Torque Torque

Master Follower 1 Follower 2 Master Follower 1 Follower 2

Due to the complexity of the mechanical system,

over a range, flux or current generated could -VBUS

base, the type of control, the natural frequencies Time