FPSO
FPSO
FPSO
F P S O - TA N K E R C O L L I S I O N I N
TA N D E M O F F L O A D I N G O P E R AT I O N
By
Haibo Chen
January 2003
Trondheim Norway
URN:NBN:no-3369
URN:NBN:no-3369
A B S T R A C T
Collisions between FPSO and shuttle tanker in tandem offloading operation have caused
a growing concern in the North Sea. Several recent contact incidents between
FPSO/FSU and shuttle tanker have clearly demonstrated a high likelihood of contact
between vessels in tandem offloading. The large masses involved, i.e. the high potential
impact energy, make the collision risk large. Traditional ship/platform collision
frequency modeling may not be applicable in the tandem offloading context. Moreover,
offshore quantitative risk analyses generally focus more on technical aspects, little on
human and organizational aspects. This leads to a hardware-dominated risk reduction
approach, and it has been proved not to be effective to mitigate risks involved in
complex marine operations in general.
Frequency modeling of collision between FPSO and shuttle tanker in offloading
operation is carried out in this study. The collision frequency model is structured in two
stages, i.e. the initiating stage and the recovery stage, where the former involves an
uncontrolled forward movement of tanker, and the latter involves the recovery actions
initiated from tanker and FPSO to avoid the collision.
In the initiating stage, this study focuses on tanker drive-off forward scenarios.
Macroscopically, the frequency of tanker drive-off ahead during offshore loading and
specifically during tandem offloading is portrayed by statistical data from an earlier
study, recent SYNERGI incident data, and expert judgments made by tanker DP
operators. Relatively high frequency values of tanker drive-off in tandem offloading are
found. Microscopically, the tanker drive-off ahead scenario is investigated by
examining 9 such events in tandem offloading based on investigation reports, interviews
and discussions with individuals who directly or indirectly were involved. Findings
reveal that in order to effectively reduce tanker drive-off in tandem offloading, efforts
should be targeted on minimizing those failure prone situations, i.e. the excessive
relative motions (termed as surging and yawing) between FPSO and tanker. A
simulation-based study is carried out to quantitatively assess and effectively minimize
the occurrence of excessive surging and yawing events. Horizontal motions of FPSO
and tanker in tandem configuration are simulated via a state-of-the-art time-domain
simulation code SIMO. Findings demonstrate that excessive surging and yawing events
can be effectively minimized via measures such as minimizing FPSO surge and yaw
motions in offloading, coordinating mean heading between FPSO and tanker, and using
the dedicated DP software with the tandem loading function on tanker. Ultimately, these
measures may provide a sound operational environment where the possibility of tanker
drive-off can be minimized.
In the recovery stage, this study is focused on the recovery action initiated by the tanker
DP operator. Possible recovery actions are identified and evaluated. Based on calibrated
tanker motion simulations, the allowable time for DP operator to initiate recovery
URN:NBN:no-3369
action, so that tanker can be stopped within a separation distance, e.g. 80 m to FPSO, is
found to be critically short. A 3-stage information-decision-execution model is
generalized to model the DP operators information processing stages regarding action
initiation when in a drive-off scenario. Based on this human information-processing
model, expert judgment by simulator trainer and questionnaire survey among shuttle
tanker captains and DP officers are conducted, reasonable estimates of the time needed
for action initiation are obtained. The estimates are found to be convergent to the facts
in the incidents. Findings suggest that tanker DP operators in general need more time to
initiate recovery action than the allowable time window, i.e. recovery failure is likely
due to lack of reaction time. Two principal recommendations are proposed to reduce the
recovery failure probability, i.e. to provide a longer time window for the operator to
initiate recovery action, and/or to provide various kinds of assistance to the operator to
reduce the recovery action initiation time.
To increase the time window, a promising measure is to substantially increase the
separation distance between FPSO and tanker, e.g. from 80 m to 150 m. The feasibility
of this measure is discussed from a number of perspectives. Recovery improvement
gains are assessed. The key question concerning implementation is to know how much
separation distance should be configured in the operation. This has to be based on
considerations of both human operators need for reaction time, and tanker drive-off
behavior. Parametric tanker drive-off motion simulations are carried out in which
human action at various times are imposed. The necessary distance values to stop the
tanker are then obtained, and ideally these should correspond to the separation distance
values between FPSO and tanker in tandem offloading. These findings provide
decision-making support to select an optimum field configuration for FPSO-tanker
tandem offloading, which may inherently minimize the collision risk.
Effective reduction of reaction time can be achieved by early detection and/or quick
decision-making. This is based on the operator information-processing model
generalized earlier in this study. Measures to improve early detection are identified.
Discussions are guided by the human signal detection theory, and supported by the
operational facts of alarm and non-alarm signals in the operation. Measures to
effectively reduce the operators time involved in diagnosis and situation awareness are
also identified. They are theoretically built on the generic human decision-making
theory, and specifically designed for drive-off intervention based on the facts collected
via a questionnaire survey among shuttle tanker captains and DP officers. These
findings illuminate a broad area in the human factor perspective, i.e. training, procedure,
crew resource management, human-machine interface, and automation support, where
measures to reduce operator reaction time should be targeted. These measures may
directly reduce the FPSO-tanker collision risk in tandem offloading.
ii
URN:NBN:no-3369
A C K N O W L E D G E M E N T S
I first want to thank my supervisor, Professor Torgeir Moan. In the past four years
Torgeir has been a supervisor, a mentor, and gradually also a close friend to me. We
have our traditional Mondays Dr.Ing Talk which was started in the early days of my
Dr.Ing journey when everything, e.g. risk analysis, human and technical factors, and
so on, were flying around in the air, and has lasted up to the present where fact-based
models and analyses are rooted down in the soil. Together we had those inspired
moments when we had a break-through in the research, and low periods when progress
was stagnant. Torgeir has always been the first (and in many cases also the last) helping
hand for me, supporting initiatives and of course helped me back on track in my
digressions.
Risk modeling in this study has to be built on a thorough understanding of the practical
operation. An important yet particularly challenging step for me is to setup contacts
with the industry. I am very grateful to have met Professor Jan Erik Vinnem who
introduced me to the FPSO Operational Safety JIP. This is a precious link between the
university and the industry, which was initiated when Jan Erik invited me to give a brief
talk at the JIP steering committee meeting on 8 December 1999. After this meeting, I
joined the group as an observer, and several months later I became a member of the JIP
research team. The activities in this JIP have benefited my Dr.Ing study significantly,
and I am deeply indebted to Jan Erik for his continuous help, support, and
encouragement throughout these years.
Needless to say, FPSO and shuttle tanker operators are of vital importance to my Dr.Ing
study. Statoil and Navion are two organizations to acknowledge as a whole; but I would
also like to thank the following people in particular.
Dr.Ing Sverre Haver at Statoil. I am very grateful and very lucky to have met Sverre.
We formally started our contact on my first visit to the Statoil Head Office in Stavanger
on 30 June 2000. The hydrodynamics involved in tandem offloading is a crucial risk
contributor. Time-domain simulation of FPSO-tanker motion was the approach Sverre
and I identified in our meeting on that day. Since then, I have always looked forward to
every status meeting that we have had, which proceeded passing each milestone, i.e.
computer program familiarization, vessel model and environment input setup, model
calibration based on full-scale motion measurement, statistical modeling, and later
tanker drive-off simulation, and so on. Every single step involved would have been a
daunting task if without generous and continuous help from Sverre, and later also his
colleagues, Mr. Kjell Larsen and Mr. Harald Kleppest, and also Dr.Ing Trond S.
Meling at the early stage. These kind people are gratefully acknowledged here.
Mr. Leif Ivar Tnnessen in Navion is one that I am very much indebted to. We first met
on the steering committee meeting of FPSO Operational Safety JIP on 13 November
iii
URN:NBN:no-3369
2001. Leif Ivar provided me with a great deal of tandem offloading operational data
regarding, for example, shuttle tanker crash-stop characteristics, tanker propulsion
response data, recovery alternatives from DP operator in tanker drive-off. He also
introduced me to his colleague, Mr. Kjell Helgy, a resourceful Navion specialist on DP
systems and incident investigation, whom I would also like to thank. And my intended
questionnaire survey with shuttle tanker captains and DP officers would never have
been possible if without Leif Ivars warm-hearted support and help. Those valuable
operational data offered me a concrete basis for an important part of my Dr.Ing study
the human initiated recovery in tanker drive-off.
In human factor and human reliability studies, nothing is probably more frequently
mentioned than training. The training institute Ship Manoeuvring Simulator Centre
(SMS) in Trondheim is sincerely acknowledged. Specifically, I want to give a big
thank-you to Instructor/Captain Helge Samuelsen in the SMS for supporting me a lot of
valuable tandem offloading operational data. Helge has years of extensive experience of
tanker offshore loading simulator training, and he has been very supportive since day
one, 21 September 2000, when I visited him. I still remember vividly the 5-day
simulator training course Offshore Loading Phase 2 given by Helge, which I followed
as an observer in the SMS during 11-15 December 2000. This fresh experience bridged
my mental gap between written procedures and actual operation. I also directly
observed the human responses in those simulated tanker drive-off scenarios. Afterwards
I realized that the situation often could be far more complicated than what had tidily
been stated in the final investigation report.
The observation of a real tandem offloading operation, discussions of hands-on
experience and problematic areas in the operation came from the 5-day visit onboard
M/T Navion Oceania during 28 June 02 July 2001. First I want to thank Captain Bjrn
Kre Hammersvik and 1st Officer Petter Johan Ellingsen for their patient explanation of
operational details, detailed walk through and talk through practices, and insightful
information about incidents and near misses. I also would like to thank the Navion
people and Professor Jan Erik Vinnem (again) for arranging this valuable and
memorable trip, and other FPSO Operational Safety JIP team members Senior
Research Scientist Per Hokstad and Ms. Hilde K. Sle in SINTEF, for their help in
many ways during this trip.
Research cant be done in a vacuum. The connections with other risk analysts have
proved valuable. I am particularly grateful that I met the former managing director of
Safetec (now within CorrOcean), Dr.Ing Stein Haugen, on 5 March 1999. That was only
one month after I had started this Dr.Ing study. He helped me with an arrangement of
literature reading regarding various FPSO risk projects and safety cases documentation
in the Safetec Trondheim office in July 1999. This early exposure to the offshore risk
analysis projects was very valuable to me, and it set the tone of my study. I also talked
with Haugens colleagues, Mr. Frank Vollen and Mr. Jon Daniel Nesje, on various
topics. I appreciated all the help, insights and suggestions from the people at Safetec at
that stage of my study.
With the involvement in FPSO Operational Safety JIP, I had contact with several people
in SINTEF Industrial Management, Safety and Reliability Section. I would like to thank
iv
URN:NBN:no-3369
Mr. Stein Hauge and Dr.Ing Ragnar Rosness for their suggestions in relation to my
study until Autumn 2000. Afterwards I am grateful to have had many discussions with
Senior Research Scientist Per Hokstad, Ms. Hilde K. Sle, and later Mr. Terje Dammen
regarding the FPSO Operational Safety JIP work. They inspired many ideas that have
been implemented in this Dr.ing study. I also appreciate to have contact with Mr. Oliver
Kieran in Offshore Safety Division in Health & Safety Executive (HSE) in the UK.
Many thanks for his warm encouragements since the day we met in December 1999,
and his kind help when I was in the HSE London Office for incident data collection.
Olivers colleague, Mr. Andrew D. Moyse in HSE, is also acknowledged for helping me
on several occasions on various UK FPSO data since February 2001.
Some people may just have a very brief appearance in this Dr.Ing journey, but their
contributions have made my work very different. Professor Asgeir Srensen reviewed
the dissertation draft and made many valuable technical (and linguistic) comments. Mr.
Svein-Arne Reinholdtsen at Marintek/NTNU helped me with the SIMO program, and
Marintek is acknowledged for granting the use of this computer program in my study.
Mr. Torbjrn Hals and Mr. Kjetil Gudmestad in Kongsberg Simrad clarified DP
software details and provided tanker propulsion information to me. Mr. Tony Read in
IMCA (International Marine Contractors Association) offered me a free copy of that
valuable report: Quantified frequency of shuttle tanker collision during offtake
operations. Finally, Ms. Anja Angelsen proof-read the dissertation draft thoroughly and
provided many corrections. I want to thank all these people for their kind help.
Last but not least, my Dr.Ing study was financially supported partly by the fellowship
from the Research Council of Norway and partly by the stipend from Statoil. These
supports are very appreciated.
Haibo Chen
19 January 2003
URN:NBN:no-3369
C O N T E N T S
1.
INTRODUCTION ............................................................................................................................1
1.1
1.2
1.3
1.4
2.
MODELING OF COLLISION........................................................................................................8
2.1
2.2
3.
BACKGROUND ............................................................................................................................1
MOTIVATION ..............................................................................................................................4
OBJECTIVES ...............................................................................................................................6
LIMITATIONS ..............................................................................................................................6
MODEL CONSTRUCTION .............................................................................................................8
MODEL APPLICATIONS ................................................................................................................9
4.
5.
6.
.............................................................................................................................55
Separation distance extension ............................................................................................55
FEASIBILITY
vi
URN:NBN:no-3369
6.1.2
6.2
6.3
7.
8.
..........................................................................................................................74
Modeling of FPSO-Tanker collision...................................................................................74
Probability of tanker drive-off ............................................................................................75
Probability of recovery failure ...........................................................................................75
SUGGESTIONS FOR FUTURE WORK ............................................................................................76
Level I .................................................................................................................................76
Level II................................................................................................................................77
Level III ..............................................................................................................................78
CONCLUSIONS
REFERENCES .........................................................................................................................................80
A.
THEORETICAL BACKGROUND...............................................................................................87
A.1
RATIONALE BEHIND THE COLLISION FREQUENCY MODEL .........................................................87
A.1.1 Human supervisory control ................................................................................................87
A.1.2 Human-Machine system dynamics......................................................................................89
A.1.3 Operational safety modeling...............................................................................................90
A.2
HUMAN ACTION, ERROR, AND RELIABILITY ..............................................................................91
A.2.1 Human action .....................................................................................................................91
A.2.2 Human error.......................................................................................................................92
A.2.3 Human reliability................................................................................................................93
A.3
HUMAN INFORMATION PROCESSING MODELS ............................................................................95
A.3.1 S-O-R models......................................................................................................................95
A.3.2 Step-ladder model...............................................................................................................96
A.3.3 Wickens model...................................................................................................................98
A.4
REFERENCES .............................................................................................................................99
B.
C.
vii
URN:NBN:no-3369
E.
viii
URN:NBN:no-3369
A B B R E V I A T I O N S
BLS
CPP
DARPS
DFM
DP
DWT
EOP
EPS
ESD
FPSO
FSU
HCR
HMI
HRA
HSE
IMCA
NPP
QRA
OSM
PFM
PRS
SIMO
SMS
SPM
ST
TRC
UFM
UKOOA
ix
URN:NBN:no-3369
URN:NBN:no-3369
C H A P T E R
1.
INTRODUCTION
Deal with the hard while it is still easy.1
L A O T Z U (571 B. C.)
This chapter outlines the background, motivation, objectives, scope and limitations of
this Dr.Ing study. It starts with a brief introduction to the floating production, storage
and offloading (FPSO) concept and an outline of tandem offloading operations between
FPSO and shuttle tanker. There is a practical need to reduce the collision risk between
FPSO and shuttle tanker in tandem offloading. Moreover, ship-platform collision risk
models from previous offshore quantitative risk assessment (QRA) studies may not be
applicable to the tandem offloading context, and there is little consideration of human
and organizational contributions in those QRA models. Further development of
collision risk modeling and, more importantly, identification of effective measures for
reducing the occurrence of collision in tandem offloading are therefore two major
objectives of this study. The study scope and limitations are also discussed.
1.1 BACKGROUND
The FPSO concept is based on a combination of traditional ship building technology
and platform design. The following definition with respect to FPSO are found in the
NORSOK Standard (NTS, 1998):
FPSO - Ship Shaped Floating Production, Storage and Offloading Unit
A floating unit can be relocated, but is generally located on the same location for
a prolonged period of time. Inspections and maintenance are carried out on
location. The Floating Production, Storage and Offloading unit normally
consists of a ship shaped hull, with an internal or external turret, and production
equipment on the deck. The unit is also equipped for crude oil storage. The
crude may be transported to shore by shuttle tankers via an offloading
arrangement.
Lao Tzu (571 BC). Tao Te Ching. Translated by Arthur Waley, Wordsworth Editions Ltd., 1997
URN:NBN:no-3369
Chapter 1
The overall arrangement of a North Sea FPSO is shown in Figure 1-1. The living
quarter and control room are located in the bow, upwind of any hydrocarbon fire. The
turret is installed forward of mid-ship. The process area is aft of the turret, elevated
from the main deck with natural ventilation. The oil storage is provided by storage
tanks, mainly located aft the turret. The offloading system is installed at the stern.
Figure 1-1
FPSOs have been used in the Far East, Africa, and South-America for some decades
since the mid-1970s, but these areas are in general benign waters. The wide use of
FPSOs in the North Sea, West of Shetland, as well as in other hostile environments
actually started in the1990s, despite that the first North Sea FPSO, Petrojarl I, had been
used since 1986. FPSOs are probably by far the most popular floating production
system in offshore oil and gas fields worldwide. In the near future, almost 60% of the
floating production systems now on order have ship-shape hulls (McCaul, 2001).
With an increasing number of FPSOs in use, the number of shuttle tankers performing
crude oil offloading from these FPSOs is growing too. Though some FPSOs may
offload oil to shuttle tankers indirectly via remote loading buoy connected to the FPSO
by a pipeline, the majority of FPSOs currently do rely on direct offloading to shuttle
tanker to transfer oil to the shore. This direct offloading operation is carried out
generally via a tandem configuration as shown in Figure 1-2. Alongside offloading is
another possibility, but a less-adopted configuration in harsh environments. The tandem
offloading is dominant in the North Sea, and is discussed in this study.
URN:NBN:no-3369
Picture adapted from the Journal of Offshore Technology, pp.18, Vol.3, No.2, May 1995.
Introduction
The tandem offloading means that the shuttle tanker is positioned at some distance, e.g.
80 m, behind the FPSO. The two vessels are physically connected by a mooring hawser
and a loading hose through which cargo is offloaded. The tanker may position itself by
its own dynamic positioning system so that the hawser is not tensioned (DP mode), or
by applying certain astern thrust and maintain a small tension on hawser (Taut hawser
mode). Tug or standby vessel assistance may be required for taut hawser mode. The DP
tankers have greater uptime in harsh environments and therefore are widely applied in
the North Sea. This is the case considered in this study.
Tanker (DP)
Hose
50-90 m
Figure 1-2
URN:NBN:no-3369
Chapter 1
to the weather) around its turret located either internally or externally, and it may also
have significant low frequency motions in the horizontal plane (surge, sway and yaw)
due to waves and wind if in harsh environments. In order to stay connected for loading
and at the same time maintain a separation distance, e.g. 50-90 m behind FPSO stern,
the DP shuttle tanker has to position itself according to the FPSO position.
Offshore loading by shuttle tankers has been carried out in the North Sea for more than
two decades (HSE, 1997). Traditionally, this involves shuttle tanker with an articulated
loading platform or a spread-moored loading buoy. The situation is dramatically
changed in the tandem offloading operation in terms of positioning complexities and
damage potential, i.e. the significant amount of mass involved (a 150,000 dwt shuttle
tanker, for example) in close distance to an installation (FPSO) for a long duration.
However, offloading hardware, software, operational procedures, and so on, which
evolve from experience and lessons learned before, largely remain the same in this new
context. Shuttle tanker loss of position in powered condition and subsequently collided
with FPSO/FSU had been reported a few times. As commented in a recent study by
Global Maritime to IMCA (IMCA, 1999), the most significant risk, in terms of tanker
offtake, is associated with tandem loading operations.
1.2 MOTIVATION
There are likely five collision incidents between FPSO/FSU and DP shuttle tanker
occurred in the North Sea in recent years, based on reference information from Vinnem
(1999) and Leonhardsen et al. (2001).
-
The collision frequency is relatively large based on the above incident record. The
estimated total number of tandem offloading operations by DP shuttle tanker in the
North Sea is around two thousand between the years 1996 to 2000 (Helgy, 2002). This
indicates one collision every four hundred offloading operations. For a DP shuttle
tanker undertaking fifty tandem loading operations per year, this equals to one collision
in the order of every ten years. However, a reasonable interpretation of these statistical
results should also include the following fact: The tandem offloading operation between
FPSO/FSU and DP shuttle tanker has been in continuous evolution during recent years.
The high frequency averaged over these years cannot reflect the significant amount of
improvements and efforts made by shuttle tanker and FPSO operators in the mean time.
URN:NBN:no-3369
Introduction
The collision damage potential is large, due to the large masses, and consequently the
large impact energy involved in possible tanker-FPSO collisions. The impact energy in
one of the collisions had reached 31 MJ. This is estimated for a 154,000 dwt shuttle
tanker at a 0.6 m/s impact velocity from the information in the investigation report
(Statoil, 2000). Stern damage on the FPSO may cause penetration and flooding in the
machine room. Moreover, with the widely adopted FPSO design, e.g. Gryphon,
Captain, Norne, sgard, etc. (Addy et al., 1995; Odland, 1995), the living quarters are
located in the bow area, thus the flare towers, which have to be located in the stern area,
are vulnerable to tanker impact. A worst-case scenario could therefore be a major tanker
collision that topples down the flare tower of the FPSO. This can initiate a chain of
events with severe fire and explosion on both vessels. From one of the occurred
collision incidents, damage of members and bracings of the flare structure did happen
(Leonhardsen et al., 2001).
A wide range of parties in the offshore industry, i.e. regulators, technical system
designers, shuttle tanker and FPSO operators, training institutions, and risk analysts, are
involved in combating the collision problem in tandem offloading. Continuous efforts
have been made in recent years to reduce collision frequency and/or consequence, see
publications from HSE (1999), the FPSO Operational Safety JIP (Vinnem, 2000), and
tandem offloading guidelines by UKOOA (2002). Yet, to control the collision risk in
tandem offloading, and particularly to reduce the frequency of collision, involve two
basic difficulties arising from offshore QRA methodology. Note that the collision
consequence modeling is generally based on energy method and non-linear structural
mechanics. Details on this subject can be found in Skallerud and Amdahl (2002).
First, the traditional ship-platform collision frequency modeling may not be suitable for
the tandem offloading context which involves FPSO and DP shuttle tanker. Quantitative
frequency modeling of ship-platform collision is not a new issue. Furnes and Amdahl
(1980) developed their quantitative collision frequency model based on the geometric
consideration in the early 1980s. Haugen and Moan (1992) presented in the early 1990s
a collision frequency model based on considerations of traffic number, navigation
course, and recovery actions from ship and platform. However, these models were
primarily developed for a fixed platform and passing vessels. There are some offshore
risk studies of FPSO and tanker collision, e.g. by MacDonald et al. (1999) in the recent
JIP on Risk and Reliability of a FPSO in Deepwater Gulf of Mexico. However, in
general those studies are few in number, and their applicability to the North Sea and the
level of detail of collision frequency modeling are questioned.
Second, offshore quantitative risk studies have traditionally focused more (if not solely)
on technical failure events, little on human (and organizational) failure events.
However, there is a growing recognition that humans (and organization) do play a role,
in connection to structural failures as documented by Bea (1997) and Kvitrud et al.
(2001). In tandem offloading operation, the HSE UK also identified that the majority of
contact incidents between shuttle tankers and installations during 1997 and 1998
involved DP problems and human factors (HSE, 1999). This calls for an integration
of the modeling of human (and organizational) contributions into the collision
URN:NBN:no-3369
Chapter 1
frequency model which is valid for the offloading context between FPSO and DP
shuttle tanker.
In summary, there is a practical need to carry out risk analyses in order to reduce the
collision frequency in tandem offloading operation, given the relatively large contact
frequency and large damage potential at present. To fulfill this need, it is necessary to
further develop a quantitative collision frequency model, in which the uniqueness of the
tandem offloading context, as well as the human and organizational contributions to the
collision frequency can be taken into account.
1.3 OBJECTIVES
The overall objectives of this Dr.ing study are to:
1. Develop a quantitative frequency model to analyze the collision risk between FPSO
and shuttle tanker in the tandem offloading operation, taking both technical aspects
and operational aspects into account.
2. Exemplify the above modeling approach by case studies based on the collected
operational data from the North Sea practices; identify measures to reduce the
collision occurrence in tandem offloading.
1.4 LIMITATIONS
Different technical systems, procedures and environments for the tandem offloading
operation may imply different collision risk pictures. In this study the technical systems
considered are a purpose-built FPSO and a DP shuttle tanker, both for North Sea
operations. The operational procedures are in general applicable to the tandem
offloadings performed in the North Sea. However, note that the details of the procedures
may vary from field to field. The environmental condition applies to the Haltenbanken
area in the Norwegian Sea.
Given the fact that pure technical failure events have been exhaustively modeled and
analyzed in traditional offshore QRA models, this study emphasizes more on modeling
the operational failure events and the interaction between technical and operational
events.
There are a few modeling approaches apparently close to the above purpose, e.g.
CRIOP (Ingstad and Bodsberg, 1990), Influence Diagram Approach (Embrey, 1992),
Risk Influence Analysis (Rosness, 1998), Bayesian Probabilistic Networks (Hansen and
Pedersen, 1998; Faber et al., 2001). However, note that each modeling approach is
developed from its own original context. To collect facts in tandem offloading, which in
itself is unique, and then try to fit the facts into a modeling approach that appears
suitable, may not guarantee an effective way of preventing collision. Therefore, none of
the above modeling approaches is taken for granted in the study. Instead, a fact-based
URN:NBN:no-3369
Introduction
URN:NBN:no-3369
C H A P T E R
2.
MODELING OF COLLISION
Use models by all means if you find them useful but do not
become a slave to them.1
TREVOR KLETZ
This chapter presents the overall frequency model of FPSO and shuttle tanker collision
in tandem offloading. The rationale behind constructing this model is briefly outlined.
The in-depth theoretical background behind the modeling is presented in Appendix A.
By a coarse evaluation of possible scenarios, a practical top-level collision frequency
model is formulated. Implications of this model are discussed. Though short in size, this
chapter is the main thread linking all the following chapters in the study.
2.1 MODEL CONSTRUCTION
For a collision between tanker and FPSO in tandem offloading to happen, irrespective
of operational phase, there are two necessary conditions:
-
Recovery actions (initiated from tanker and/or FPSO) fail to avoid the collision
(2-1)
where P(UFM i ) is the probability of tanker uncontrolled forward movement type i; and
P(Failure of Recovery | UFMi ) is the probability of recovery failure initiated from
tanker and FPSO, conditioned on tanker UFM type i.
The principle behind this overall modeling is the operational safety modeling (OSM)
concept. The theoretical background of the OSM concept, as well as theories that guide
1
Kletz T. Learning from Accidents. 3rd Ed, pp.7, Gulf Professional Publishing, 2001a.
URN:NBN:no-3369
Modeling of Collision
the detailed, further modeling in the following chapters, are provided in Appendix A.
Descriptions of theory and methodology are presented in a concise manner which
mainly include the following: human-machine system dynamics (Sheridan, 1992;
Hollnagel, 1998), quantitative risk modeling (Vinnem, 1999), human reliability and
error analysis (Reason, 1990; Kirwan, 1994), and modeling of operator action
(Rasmussen, 1986; Wickens and Hollands, 2000).
The tanker uncontrolled forward movement (UFM) may be initiated in powered
condition, termed as powered forward movement (PFM) scenario which is also called
the drive-off forward scenario in this study. The PFM may be initiated by various
technical system failures, erroneous operational actions, or a combination of both. The
tanker UFM may also be initiated in drift condition when the tanker loses all its power,
and the resultant environmental forces push the tanker towards the FPSO. This is
termed the drift forward movement (DFM) scenario.
The DFM scenario is considered a low probability and low consequence event, and it is
therefore excluded from the further modeling. The reasons are: Firstly, tanker blackout
during offloading is not a frequent event. Secondly, given that event, the resultant
environmental forces, due to weathervane, will typically drift the tanker away from
instead of towards the FPSO. Certainly, there may be cases where the tanker is heavily
loaded, and wind and waves are small, where the tanker is under dominant influence
from current, and current may drift the tanker ahead. However, in such cases, the tanker
typically will not gain much speed within an 80-100 m distance to the FPSO stern.
The recovery actions are mainly initiated from tanker, specifically by the tanker DP
operator, to stop the tanker or steer it away from the FPSO stern. The FPSO crew may
also take action, e.g. using the main screw to create current to blow the tanker away, or,
in principle, change heading. However, in a tanker PFM scenario, the FPSO-created
current cannot effectively blow the tanker away, and time is generally too short for the
FPSO to change heading dramatically. Therefore, these FPSO-initiated actions in
general have limited effect and are excluded from the further modeling.
In summary, the frequency model of collision between FPSO and shuttle tanker in
tandem offloading can be practically formulated as in Eq.2-2. This collision frequency
model is intuitively simple, however, its applications, as discussed in the following
section, form the main contents in this study.
P Collision = P(PFM) u P(Failure of Tanker Initiated Recovery | PFM)
(2-2)
URN:NBN:no-3369
10
Chapter 2
1. The initiating stage Identify and minimize all possible sources and situations that
may cause tanker drive-off forward.
2. The recovery stage Evaluate and improve the recovery actions initiated from
tanker to avoid collision, should drive-off forward happen.
Note that this top-level model should not be interpreted as risk analysis and risk
reduction efforts are solely directed to shuttle tanker. FPSO does play an important role
in the initiating stage, contributing to a tanker drive-off forward scenario. This is
revealed via studying tanker drive-off events, which is presented in Chapter 3. Findings
suggest that in order to effectively reduce the probability of tanker drive-off, excessive
relative horizontal motions (surging and yawing) between FPSO and tanker should be
minimized. Subsequently, a simulation-based approach to quantitatively analyzing the
occurrence of the excessive surging and yawing events is presented in Chapter 4.
Recommendations to the design and operation of FPSO and tanker to minimize these
excessive relative motion events are proposed, which ultimately may provide a sound
operational condition where the initiation of tanker drive-off can be minimized.
The modeling and analyses of the tanker DP operator initiated recovery are elaborated
in Chapter 5. Findings show that given the recovery strategy favored by tanker DP
operators in general, the failure of recovery may be still significant due to lack of
reaction time. Two principal risk reduction recommendations are then identified: one is
to increase the time window (provide enough time) for the tanker DP operator to initiate
recovery action; the other is to effectively reduce the operator reaction time.
The feasibility and implementation of the two risk reduction recommendations are
investigated in Chapter 6 and Chapter 7, respectively. To increase the time window, a
promising measure is to increase the nominal separation distance between FPSO and
tanker in offloading. A brief discussion about another possible measure, i.e. to limit the
forward thrust from main engine/propeller(s) that can potentially be involved in tanker
drive-off, is also included. To reduce reaction time, two themes are focused on, i.e. early
detection, and quick situation awareness. The findings illuminate a broad area in the
human factor perspective, i.e. training, procedure, crew resource management, humanmachine interface, and automation support, where measures to reduce operator reaction
time may be targeted on.
URN:NBN:no-3369
C H A P T E R
3.
D R I V E - O F F I N I T I AT I O N
This chapter deals with the tanker drive-off scenario. Macroscopically, the frequency of
tanker drive-off forward during offshore loading, and specifically during tandem
offloading, is portrayed by statistical data from an earlier study, recent SYNERGI
incident data, and expert judgments made by shuttle tanker captains and DP officers. It
is found that tanker drive-off forward frequency is high in tandem offloading, likely
ranging from 5.4E-03 to 2.0E-02 per loading. Microscopically, the tanker drive-off
forward scenario is investigated by examining 9 such events in tandem offloading based
on investigation reports, interviews and discussions with individuals. Findings show
that the initiation of tanker drive-off involves a complex human-machine interaction,
potentially involving DP hardware and software, position reference systems and vessel
sensors, local thruster control system, and DP operator. The event analyses reveal that
in order to effectively reduce tanker drive-off in tandem offloading, efforts should be
targeted on minimizing those failure prone situations, i.e. excessive relative motions
between FPSO and tanker.
3.1 FREQUENCY OF DRIVE-OFF
Tanker drive-off event is defined in this report as: Tanker is driven away from its
target/wanted position by its own thrusters in offloading operation. This is not a
planned or wanted movement. Note in principle, drive-off can be forward, astern, or
sideway, and it is the drive-off forward that may lead to collision. When drive-off
appears in discussions below, it refers by default to drive-off forward unless other is
stated.
Estimation of the frequency of DP shuttle tanker drive-off during tandem offloading is
not straightforward. First, available and applicable data are scarce. There are few
published statistical data of DP shuttle tanker drive-off frequency in offshore loadings
1
11
URN:NBN:no-3369
12
Chapter 3
in general. Further, tandem offloading between DP shuttle tanker and FPSO/FSU in the
North Sea was not carried out on a large scale until mid-1990s. Statistical data which
either contains little information about tandem offloading, or hardly reflect the recent
status of technical and operational systems are therefore not applicable. Second, underreporting can be a problem. Note that a tanker drive-off event does not imply that there
is a collision or other serious incident; it could turn out to be a collision near-miss under
operator intervention, and subsequently it may not be reported. This was likely to be the
case in the early years of tandem offloading, when the incident reporting from shuttle
tankers was not as strict as it is today.
A reasonable estimation of the frequency of drive-off therefore has to be based not only
on the hard statistical data (given that they do exist), but also potentially the soft data
from experts subjective judgments from their direct/indirect operational experiences. In
the following sections, results derived from an earlier study made by Global Maritime
for IMCA (the International Marine Contractors Association) provide some useful
references to the present study. The frequency of tanker drive-off applicable to the
tandem offloading operations in the North Sea in recent years (1996-2000) is estimated
via two sources, namely the statistical data from the SYNERGI incident database and
the expert judgments from 17 shuttle tanker captains and DP officers.
The results (based on findings in all subsections below) reveal that the tanker drive-off
frequency is high in tandem offloadings, likely ranging from 5.4E-03 to 2.0E-02 per
loading, equivalent to one drive-off in every 50 to 185 loadings. There is also evidence
suggesting that the drive-off frequency in tandem offloading is significantly higher than
the drive-off frequency averaged over all offshore loadings.
It is important to note that the tanker drive-off frequency results derived in this section
do not refer to any specific shuttle tanker operator, nor to tandem offloadings from any
specific FPSO/FSU field. These results are obtained by pooling information from a
number of shuttle tankers performing offloadings from a number of FPSO/FSU fields in
the North Sea, and should be viewed as representative (or sample) values applicable to
the North Sea tandem offloading operations.
3.1.1 Earlier study
Global Maritime made a frequency study of shuttle tanker collision during offloading
for IMCA (IMCA, 1999). This study, probably the most complete and relevant one
published so far, includes all the station keeping data on tanker offshore loading
operations supplied to IMCA, up to August 1998. Among these, there are 134 station
keeping incidents from 9946 offloadings made by DP shuttle tankers from offshore
export facilities, including FPSO/FSU fields. Most of these export facilities are believed
to be located in the North Sea.
Among these 134 incidents, there are 16 forward drive-offs, which ultimately caused 12
collisions with loading point. The resulting frequencies are listed in Table 3-1.
URN:NBN:no-3369
Drive-off Initiation
Number of Offloading
Operations: 9946
Incident Number
Frequency (per loading)
Table 3-1
13
Station Keeping
Incident
Drive-off
Forward
Collision
134
16
12
1.3E-02
1.6E-03
1.2E-03
The above results provide a valuable statistical reference regarding DP shuttle tanker
drive-off in offshore loading operations. However, one must be careful in applying
these results directly to the present study. This is because of the following: First, the
incidents in the IMCA data are collected from various offshore export facilities,
including, but not limited to, FPSO and FSU fields. The uniqueness of the tandem
offloading operation may affect the tanker drive-off frequency, and this cannot be
reflected if average frequency over various offshore loading concepts is applied.
Second, the IMCA data have a long time span dating back to as early as 1979, and this
may not be of relevance to the present study. Again, the technical and operational
systems are different in recent years compared to those of the 1980s. For the present
study, the interested time span starts from 1996 when tandem offloading started to
boom in the North Sea. Third, though all collision incidents are probably included in the
IMCA data, near-miss data, as said in the report, are undoubtedly missing. This implies
a higher actual number of drive-off(s), and subsequently a higher frequency. Practically,
it is difficult to further estimate how much increase is reasonable, given the available
information from that study.
3.1.2 SYNERGI incident data study
DP shuttle tanker tandem offloading incident data from the SYNERGI1 database were
collected and analyzed. The data cover the 5-year period from the beginning of 1996 to
the end of 2000. In total there are 61 tandem offloading incident entries, involving 10
FPSO/FSU fields and 17 DP shuttle tankers in the North Sea, both in the Norwegian
and the UK sector. Note that these incident data do not cover all FPSO/FSU fields, nor
all DP shuttle tankers, in the North Sea. Therefore statistics should not be viewed as a
complete picture, but rather as a reasonable sample, in regard to the tanker drive-off
frequency in tandem offloadings in the North Sea.
Among those 61 incidents, there are 49 station keeping incidents, i.e. the incidents are
related to propulsion (thruster, propeller, engine, generator, pitch-control device), DP
system, position reference sensors, mooring system, and operation or maintenance of
these systems. The remaining 12 are solely related to the loading system (green line,
manifold, telemetry, etc.). Among those 49 station-keeping incidents, 7 are identified as
drive-off forward (2 are drive-off astern, in addition). These 7 drive-offs ultimately
caused 4 collisions with FPSO/FSU, and the remaining 3 are near misses.
URN:NBN:no-3369
SYNERGI database is operated by Pride ASA. For more information, see http://www.pride.no
14
Chapter 3
There does not exist a readily available number of how many tandem offloadings that
were conducted corresponding to these incidents. Combining records and estimations
made by Tveit (1998) and Tnnessen (2002), a conservative estimation is reached at
1300 tandem offloadings during this 5-year-period from 1996 to 2000. The resulting
frequencies of station keeping incident, drive-off, and collision in the recent 5-yearperiod are listed in Table 3-2.
Number of Tandem
Offloadings: 1300
Incident Number
Frequency (per loading)
Table 3-2
Station Keeping
Incident
Drive-off
Forward
Collision
49
3.8E-02
5.4E-03
3.1E-03
High frequencies of drive-off and collision per loading are observed from the above
results. There is approximately one tanker drive-off in every 185 tandem offloadings,
and consequently one collision in every 325 tandem offloadings. To put numbers in an
annual perspective, an FPSO which annually has 40-50 offloadings to DP shuttle
tankers may have one tanker drive-off during offloading in the order of every 5 years
and one collision in the order of every 10 years.
Again, under-reporting did possibly exist, and taking this factor into account may
further increase the frequency of tanker drive-off. There is only one incident entry (a
collision) for the whole year of 1996, while there are 26 incident entries (including
collision, near misses, safety problems) for the year 2000. The comparison of the
number and also the contents of the reported incidents between the two years may
reflect a possible under-reporting in the earlier years, though the number of tandem
offloadings in 2000 significantly surpasses that of 1996.
By comparing to the results in Table 3-1, we may observe that the tanker drive-off
frequency here is 3.3 times higher. One might conclude that tanker drive-off has
occurred 3 times more frequently in tandem offloading than in all types of offshore
loadings as a whole. However, we must notice the potential under-reporting in both
statistical sources, which may have biased drive-off frequency numbers to a varying
degree. We may further observe that for DP shuttle tankers, collision frequency in
tandem offloading is 2.6 times higher than the averaged value for all offshore loadings.
Unreported collision events are less likely in both statistical sources. The comparisons
may consistently indicate an alarming message, i.e. that the frequencies of tanker driveoff and subsequent collision are higher (likely 3 times higher) in tandem offloading,
than the averaged values of all types of offshore loadings.
It is important to notice a hidden limitation when interpreting the above frequency
results. Basically, the tandem offloading operation between FPSO/FSU and DP shuttle
tanker had been in continuous evolution during those 5 years. New hardware and
software were designed and put into use. Operational procedures were optimized, e.g.,
the DP watch pattern, and more operator trainings were carried out. Moreover, the
URN:NBN:no-3369
Drive-off Initiation
15
safety awareness has been improved, as reflected by the increasing number of near
misses and safety problems reported. However, the frequency numbers averaged over
these 5 years (which are high in magnitude) cannot reflect the significant amount of
improvements (and efforts, too) made by shuttle tanker and FPSO operators in the mean
time.
Last but not least, it is worth to note the limitations in the analyses. First, the analyzed
SYNERGI data only span over 5 years, and the data applicable to the study are
considered no more than 6 years up to now since 1996. All in all, the applicable data are
scarce. This brings statistical uncertainty to the resulting frequencies. Second, there are
uncertainties connected to the estimation of tandem offloading numbers corresponding
to the recorded incidents. Other independent sources that can verify the estimated
loading numbers have not been found in this study.
3.1.3 Expert judgments from tanker operators
The frequency of tanker drive-off in tandem offloading is estimated via findings from a
questionnaire survey concerning the tandem offloading safety. This survey, as a part of
this Dr.Ing study, was conducted in the spring of 2002. A total of 17 shuttle tanker DP
operators (captains and DP officers) participated in the survey. The questionnaire is
attached in Appendix E.
The number of tandem offloadings and drive-offs that each operator had experienced
were obtained. The drive-offs considered here include both forward and astern ones. 17
drive-off frequencies in tandem offloadings are derived. These estimations are
considered independent since they are based on the operators individual operational
experience. Overall, the operational experiences from participants sum up to 1293
tandem offloadings.
The 17 estimations of drive-off frequency in tandem offloading are plotted in Figure
3-1. The averaged drive-off frequency is 8.2E-02 per loading, and the maximum value
is 3.3E-01 per loading. These values should be handled cautiously since operators with
limited operational experience may provide some non-representative values of drive-off
frequency, and subsequently the average value of all estimations can be biased.
URN:NBN:no-3369
16
Chapter 3
0,35
0,30
0,25
0,20
0,15
0,10
0,05
0,00
1
10 11
12 13
14 15
16 17
Feedback ID.
Figure 3-1
It is expected that the more loading operations are involved, the more representative is
the derived drive-off frequency, since statistical uncertainty due to a limited number of
operations can then be minimized. Accordingly, three groups of operators are defined
with reference to the number of loadings they have performed. The drive-off
frequencies are averaged for each operator group, as shown in Table 3-3.
Individual Operator
Experience
Sum of
Operator
Sum of
Operation
Averaged Drive-off
Frequency per Loading
0 < Loading d 50
10
430
1.2E-01
263
3.2E-02
150 d Loading
600
2.0E-02
Table 3-3
The averaged tanker drive-off frequencies in tandem offloading range from 2.0E-02 to
1.2E-01 per loading. The lower limit is the average from estimates by the three most
experienced operators with total experiences of 600 tandem offloadings, and it implies
one drive-off every 50 tandem loading. The upper limit is the average from estimates
by 10 operators with total experiences of 430 tandem offloadings, and it implies one
drive-off every 8 tandem loading. These frequency values are both high. The lower
limit value, which is believed to have the best credibility, is still close to four times
higher than the results in SYNERGI incident data study. However, we have to note that
both forward and astern drive-offs are included here.
Reasonable interpretations of the results are important. The above drive-off frequency
results are more in the expert judgment domain than in the historical data domain.
The reasons for this are: First, one drive-off may be experienced both by the captain
URN:NBN:no-3369
Drive-off Initiation
17
and the officer, and subsequently would appear twice in the survey. The number of
drive-offs will therefore be higher than what really happened. The same goes for the
offloading operation number. Second, some operators stressed that the number of driveoffs is an approximation, not a precise record. Subjective elements are clearly involved.
This may contribute to the difference between the SYNERGI incident study results and
the operator estimates. We also have to take into consideration the historical element
that is involved. That is, though the survey was conducted recently, the drive-off
frequency results are derived based on operators past experiences which cannot be
viewed as a direct reflection of the present technical system and operational
configuration, nor as a fully representative future prediction.
3.2 ANALYSIS OF INCIDENTS
Facts and findings in Section 3.1 may have documented clearly that tanker drive-off
during tandem offloading was frequent in the past. Facing the future, with more tandem
offloadings to come, the important questions, at least in this study, are not to debate
whether tandem offloadings by DP shuttle tankers should be banned or not, but to
clarify what may go wrong that can cause the tanker drive-off, and afterwards to
identify how to reduce the occurrence of tanker drive-off effectively. These are the
objectives in analyzing those occurred incidents and near misses.
A study of 9 previous DP tanker drive-off events in tandem offloading is carried out,
hopefully to achieve the above objectives. It is true that studying the past may
illuminate only the hazards one has passed through, rather than those that lie ahead. It is
however better to see the hazards afterwards than not seeing them at all, as one may
pass the same way again (Kletz, 2001). Among these 9 tanker drive-offs, five resulted
in collisions with FPSO, the remaining four were near-misses, and all happened in the
North Sea between 1996 and 2001. The incident data are mainly from the investigation
reports made by field operators and/or regulators. Note that an incident investigation
report may be inaccurate or incomplete, even when prepared by experienced
investigators. There are two difficulties for incident analysis based on written reports, as
commented by Reason (1990): First, an accident report will always contain less
information than was potentially available. Second, a written account has the effect of
digitizing what in originally was a complex and continuous set of analogue events.
For these reasons, interviews and discussions with the individuals that have direct or
indirect information were also conducted during the study. Data are pooled together and
analyzed anonymously to preserve confidentiality.
The data from each incident and near miss are coded in a tabular format which is
structured according to the event development sequence. It is from these facts of what
had happened that the principles of the collision frequency model presented in Chapter
2 are drawn. Each analyzed event forms a table, and 9 resulting tables are attached in
Appendix B. A brief summary is presented in Table 3-4. The meaning of the terms used
in the summary table is clarified below.
URN:NBN:no-3369
18
Chapter 3
The initiation of tanker drive-off is structured as Initiating Process and Context at the
top level. The Initiating Process is structured into Link I, Link II, and Link III based on
the event development. The Link I refers to the traceable origin of the event chain which
finally leads to the drive-off, based on the available information. The Link II and III
refer to the sequential contributing events in the event chain. The events cover various
technical failures and operator actions. Note that operator actions contributing to the
initiation of drive-off are not necessarily the operator errors, as discussed in Section
3.3.
The Context addresses the circumstances during which the technical failure and/or
operator action were initiated. It consists of Weather and Relative Motion. The Weather
refers to the environmental conditions, and only incidents during which the field
operational weather criteria were exceeded are counted. The Relative Motion refers to
the relative horizontal motions between FPSO and tanker, i.e. surging, fishtailing and
heading deviation, as discussed in detail in Section 3.4.
Context
Initiating Process
Collision
Incident
Weather
Surging, fishtailing
Hawser sensor DP
Heading deviation
Operator
Action
DP
Thruster
capacity
PRS
Operator
Action
DP
Surging, fishtailing
Heading deviation
Operator
Action
DP
Thruster
capacity
Heading deviation
Operator
Action
DP
Above
criteria
Relative Motion
Link I
Link II
Link III
Near Misses
Weather
Relative Motion
DP
CPP
Oper. Action
DP
DP
PRS
Operator
Action
DP
Table 3-4
Link I
Link II
Link III
The identified technical failures are listed below. This information may pinpoint those
vulnerable technical areas in tandem offloading, which can also be found in more detail
in other risk studies (HSE, 1997; IMCA, 1999).
URN:NBN:no-3369
Drive-off Initiation
19
Failure of local thruster control system, e.g. pitch control failure of main
controllable pitch propeller (CPP). Note that this failure mode mainly affects DP1
vessels. For DP2 vessels, two CPPs are unlikely to fail at the same time. If one CPP
fails, another one will generate astern thrust to balance the forward thrust from the
failed one. This is an advantage of using DP2 tanker.
Failure of vessel sensors. Wind sensor, hawser tension sensor, vessel draught sensor,
and gyros failures may initiate a tanker drive-off. A recent tanker drive-off event
happened because an erroneous wind speed generated by a faulty wind sensor was
given to the Wind Feed Forward module in DP, and subsequently the DP initiated
drive-off, though this was not in tandem offloading (Helgy, 2002). Hawser tension
sensor may also feed DP an abnormal high tension which can (and did) cause the
DP to drive tanker forward.
The identified operator erroneous actions may be roughly grouped into the following
three types: a) Actions due to wrong expectation of technical system function, e.g.
erroneous use of DP manual bias function (however, the DP manual bias is not
applicable to tandem offloading now); b) Actions due to improper use of technical
system, e.g. erroneous calibration of DP mathematical model, erroneous selection of
PRS; and c) Actions due to wrong assessment of internal and external situation, e.g.
weather criteria and vessel positioning capability.
Regarding the context in which drive-off occurred, severe weather conditions only
contributed to one incident. However, the relative motion is observed in four incidents.
Further, these comprise four out of the five collision incidents. This implies a potential
correlation between relative motion and collision.
By knowing technical failure, operator erroneous action, and context separately as
above, may help to clarify what went wrong. However, it is not very helpful in
answering the question related to the future, i.e. How to reduce the occurrence of
As a position reference system, DARPS (Differential Absolute & Relative Positioning System) carries
relative position signals from FPSO to shuttle tanker. DARPS interference is caused by frequency
interference with other DARPS units used in the vicinity, and abnormal relative distance and/or heading
data can be produced. The shuttle tanker DP system can automatically detect and reject the abnormal
position signals from DARPS.
URN:NBN:no-3369
20
Chapter 3
drive-off. This question requires a detailed modeling and analyzing contributions from
technical failures, operator actions and context in a joint, rather than separated, manner.
In this perspective, and based on findings from this section, probabilistic modeling of
tanker PFM scenario is addressed from the point of view of human-machine interaction
in Section 3.3 below.
3.3 PROBABILISTIC MODELING OF DRIVE-OFF
The initiation of tanker drive-off involves a complex human-machine interaction (HMI).
This is the main observation after analyzing those drive-off events. Evidences can be
found from the event links (I, II, and III) in Table 3-4 and in each analyzed drive-off
event in Appendix B. Subsequently, the probabilistic model of P(PFM) should not
address technical events only, as did in many offshore risk studies, but also include the
modeling of human actions and their interaction with technical events. This leads to the
resulting probabilistic model for tanker PFM scenario as presented in Eq.3-1. Note that
the term human actions is used here instead of human errors. This is because of the
following.
Human action may cause or contribute to system failure, for example, inappropriate
action, action taken at the wrong time, necessary action omitted, and so on. However,
not all actions that contribute to system failure can be termed as human errors. As
argued by Macwan and Mosleh (1994), whether an action is termed an error or a nonerror should be defined with respect to some reference point. For example, in a nuclear
power plant, turning off high pressure safety injection is an error for a loss of coolant
event, while the same action is not an error for a steam generator tube rupture event.
Further, in complex systems, there may be gray areas in which the distinction between
error and appropriate action is unclear, for instance, when goals conflict or the operator
lacks information (Murphy & Pate-Cornell, 1996).
In the occurred drive-offs, for example, the DARPS (Differential Absolute & Relative
Positioning System) unit, due to interference, generated an erroneous signal and was deselected automatically by the DP. The operator re-selected the DARPS signal into the
DP as required by procedure when that signal was observed normal. However, during
the re-selection process interference occurred again, and the erroneous signal due to
operators re-selection was then accepted by the DP, and subsequently drive-off was
initiated based on the wrong distance calculated by the DP. It is basically not
appropriate to assert that the operators action was a human error, however, it is fair to
say that the action did contribute to the initiation of drive-off.
The human actions and their interaction with technical failure events can be categorized
into the following three categories in the present study. This is theoretically guided by
the human reliability principles described in Appendix A (Section A.2.3), and
practically based on the observations from the incidents and near misses.
1. Initiating action An action initiates a failure event in the system.
URN:NBN:no-3369
Drive-off Initiation
21
P PFM | AI u P AI
i
(3-1)
P PFM 2 =
P PFM | AR
j
P(TFj )
P(AR k | TFj )
P(PFM | AR k , TFj )
P(AIi )
P(PFM | AIi )
The first type of human action, i.e. the initiating action, can directly initiate a tanker
PFM scenario. It is modeled by P PFM1 in Eq.3-1.
URN:NBN:no-3369
22
Chapter 3
The second type of human action, i.e. the response action, may interact with technical
failures to initiate a tanker PFM scenario. It is modeled by P PFM 2 in Eq.3-1.
In a narrow sense, the third type of human action, i.e. the latent action, may be viewed
as the human action that influences the technical failure probability, e.g. during the
maintenance. This type of human action is not included in the probabilistic model of the
PFM scenario in Eq.3-1. It is more suitable to address this issue in a dedicated
components risk study, e.g. CPP failure study (IMCA, 1995). It is also believed that the
failure rates for components largely have included this type of human action
contribution.
However, in a broader sense, the latent human action has a vast span in terms of time
and contents. It may occur in design, construction, installation, operation and/or
maintenance. It may interact not only with technical failure, but also with the other two
types of human actions. In isolation, it may not be enough to initiate an event, and
subsequently it can lie in the system for a long time before it strikes. Modeling of the
latent action therefore has to be based on an organizational approach, i.e. we have to not
only consider front-line operators, but also include maintenance personnel, management
teams, company safety culture, and so on. This is a research challenge. Whether or not
the probabilistic modeling can capture the subtle interdependency relationships between
various factors at various levels for a dynamic (not static) organization is a problem yet
to be clarified. Therefore, in this study, latent human action as a whole is not included in
the probabilistic model. For further information about latent human action in an
organizational perspective, see Reason (1997).
3.4 QUANTIFICATION FAILURE PRONE SITUATIONS
The probabilistic model (Eq.3-1) for the tanker PFM scenario in the initiating stage is
apparently elegant, e.g. this model incorporates not only the PFM scenarios that are
rooted in human initiating actions, but also those that are rooted in technical failures
which interact with human response actions. However, knowing the above model does
not directly offer much insight into the practical world regarding how to effectively
prevent drive-off initiation in tandem offloading.
A traditional way to proceed is to perform quantifications of the proposed P(PFM). In
principle the work will involve evaluating the following three terms according to the
model: technical failures, human response actions that interact with technical failures,
and human initiating actions. The technical failures identified in this study (Section 3.2)
may have their failure rates derived from various offshore risk studies, e.g. IMCA
(1995) and DPVOA (1994). However, identification and subsequent quantitative
evaluation of human initiating and response actions are not easy to achieve. Qualitative
models, e.g. in operational HAZOP studies, may exist and are effective for the purpose
of identification. For quantification purpose, expert judgment may be the only choice in
practice. Information of expert judgment techniques and their application in offshore
quantitative risk analysis may be found from Hokstad et al. (1998), Skjong and
Wentworth (2001), and Gudmestad (2001). However, comparison and integration of
URN:NBN:no-3369
Drive-off Initiation
23
probabilities generated from different domains, i.e. historical technical data and
subjective human action data, can be problematic. The difference in scale regarding
drive-off frequencies estimated by statistical data and the expert judgments in Section
3.1 can serve as an example. This may lead to practical difficulties to identify where to
target the effective risk reduction efforts, even if quantifications are somehow magically
achieved.
Quantification is difficult, however, it is even more difficult to know what to quantify.
The above outlined quantification efforts are likely futile, while a promising way
forward is identified via examining (and later quantifying) the context in which
technical failures and human actions occur. This is, in spirit, inspired by the Error Prone
Situation (EPS) concept proposed by Fujita (1992). Technical failures are generally
considered random, however, there are exceptions. For example, in high-pressure
weather, DARPS interference often occurred. The high-pressure weather can be viewed
as a failure prone situation for DARPS units (though in this case we cannot do much
about the weather). Human actions are based on situations and how these situations,
including technical failures, are recognized, i.e. they are situation dependent rather than
completely random.
Technical failures and operator actions did not occur in random situations, but mostly in
the situation when relative motions between tanker and FPSO in horizontal plane were
excessive. This is found from the incident analyses, particularly from those occurred
collision incidents. Near misses were generally investigated in a much superficial level
and available information then offers little trace of the context. Four out of five collision
incidents actually were resulted from drive-off(s) which happened when relative
motions were excessive. It is during one or several combined modes of these excessive
relative motions that a human-machine interaction process was initiated, e.g. by a
technical failure event or an operator action (not necessarily an erroneous one), and the
human-machine interaction eventually led to tanker drive-off. The overall event
development is schematically illustrated in Figure 3-2.
These evidences point out the failure prone situation in tandem offloading, i.e. excessive
relative motions in the horizontal plane between FPSO and tanker. A closer examination
of these excessive relative motions reveals that there are two dominant motion modes,
namely the surging, which occurs due to surge motions of the two vessels; and the
yawing, which includes heading deviation and fishtailing motions of the two vessels.
Under excessive surging and/or yawing, a number of failures may happen (or have
happened). For example, the main CPP may fail since there is a frequent pitch shift
from astern to ahead. The tanker may lack enough thruster capacity to maintain a sound
heading, i.e. heading deviation occurs, and operator may take manual control to align
heading. Subsequently, erroneous action may be made, or technical failure may occur,
so that tanker drive-off is initiated. This exemplifies, as pointed out by Reason (1990),
that accidents may begin in a conventional way, but they rarely proceed along
predictable lines. A more systematic and detailed explanation of hazards caused by
surging and yawing can be found in Chapter 4.
URN:NBN:no-3369
24
Chapter 3
FPSO
Excessive
Relative
Motion
HumanMachine
Interaction
TANKER
Drive-off
Tanker
Four out of five collision incidents actually resulted from drive-offs which happened when relative motions
were excessive.
The remaining collision incident resulted from a drive-off that had nothing to do with the excessive relative
motion. The man-machine interaction that led to the drive-off was originated from a technical failure of one
position reference unit on the tanker (ultimately may be viewed from FPSOs failed gyro). This is reflected by
the dashed connection lines between FPSO and Man-Machine Interaction in the figure.
There are also near miss collisions in which the man-machine interaction that led to the tanker drive-off was
initiated by the tankers local technical system, e.g. a failed main controllable pitch propeller. This is reflected
by the connection line between Tanker and Man-Machine Interaction in the figure.
Figure 3-2
To conclude this chapter, risk reduction efforts should urgently be directed to minimize
the occurrence of excessive surging and yawing events. These are failure prone
situations in tandem offloading. Doing so will hit the bottom of tanker drive-off, and it
will hit hard. The surging and yawing are influenced by a number of factors, e.g. the
environmental condition, technical system capacity, configuration, operational
philosophy, and so on. Quantitative studies of these failure prone situations are carried
out in this Dr.Ing study by simulating motions between FPSO and tanker. This is
presented in Chapter 4.
URN:NBN:no-3369
C H A P T E R
4.
S U R G I N G A N D YA W I N G
Excessive relative motions between FPSO and tanker, categorized in surging and
yawing modes, have been identified in Chapter 3 as the failure (drive-off) prone
situation in tandem offloading. This chapter presents a study aimed at quantitatively
assessing and effectively minimizing the occurrence of excessive surging and yawing
events.
The approach is built on a state-of-the-art time-domain simulation code SIMO. The
simulation models are setup and calibrated mainly based on full-scale measurements for
a typical North Sea FPSO and a DP shuttle tanker. The calibration work and the timedomain simulation theory used in SIMO are documented in Appendix C. The simulated
relative distance and relative heading between FPSO and tanker are analyzed by fitting
their extreme values into statistical models which then give out probabilities of
excessive surging and yawing events. Sensitivity studies are performed to pinpoint
contributions from various technical and operational factors. Findings indicate that
excessive surging and yawing events can be effectively minimized via three principal
measures; i.e. minimizing FPSO surge and yaw motions in offloading, coordinating
mean heading between FPSO and tanker, and using the dedicated DP software with the
tandem loading function on tanker. Ultimately, these measures may provide a sound
operational environment where the possibility of tanker drive-off can be minimized.
4.1 INTRODUCTION
Surging refers to the relative surge motion between FPSO and tanker. The surging
becomes a problem when the two vessels oscillate fore and aft in an asynchronous
manner, i.e. the FPSO moves astern at the same time as the tanker moves ahead, as
illustrated in Figure 4-1.
1
Heinrich HW. Industrial Accident Prevention. 4th Ed., pp4, McGraw-Hill Book Company, Inc., 1959.
25
URN:NBN:no-3369
26
Chapter 4
Surging may lead to a rapid change of separation distance between tanker and FPSO. In
order to maintain a wanted separation distance, tanker will try to follow the FPSO
movement. This is generally the case if tandem offloading is performed in DP mode,
and the DP software has no dedicated tandem loading function (See Appendix D for
details of this special DP function.). The situation then requires a relatively rapid change
of tanker propulsion force between ahead and astern. This is a very stressed condition
for the tanker main propeller pitch control system. Failure may occur, e.g. failure of
pitch shift from ahead to astern, and it then leads to tanker drive-off. The situation may
also get worse by response time lag due to the big inertia of a tanker.
One incident may vividly illustrate the danger of surging, which was described by a
shuttle tanker captain during an interview (Chen, 2001) regarding the tandem offloading
safety. This incident happened in a marginal environmental condition (in which tanker
probably should not have been in connection). The FPSO surged astern. This made the
tanker move backwards. While at a time when the FPSO started to surge ahead, the
tanker was still moving backwards. This made the separation distance significantly
longer than the mooring hawser and loading hose can sustain, and both lines were
parted in a very short time.
Normal
Hawser
FPSO
Tanker
Hose
FPSO
Tanker
Surging
60 m
80 m
Figure 4-1
Surging illustration
Yawing refers to the relative yaw motion between FPSO and tanker. Both mean and
instantaneous values of yaw motion are considered here. The former is often called
heading deviation, and the latter fishtailing. Yawing is resulted from the different
weathervane characteristics between FPSO and tanker, and it becomes a problem when
a significant difference of mean headings is developed, and worsened by asynchronous
yaw motion between the two vessels, as shown in Figure 4-2.
Yawing, and specifically the heading deviation, could in principle result in loss of
relative position reference signals between tanker and FPSO. Moreover, tanker DP
officer may have to perform a difficult maneuvering of the tanker in close distance to
the FPSO to correct the heading deviation, if the FPSO does not (or cannot) adjust the
URN:NBN:no-3369
27
heading to fit in with the tanker. Typically, with a limited sideway thruster capacity, the
tanker DP system may initiate forward pitch from main propulsion and use rudder to
provide the necessary turning moment. This can cause tanker drive-off.
The following collision incident, which happened in the North Sea (Statoil, 2000), is
briefly outlined to exemplify the danger of yawing. At the final stage of loading, tanker
had a significant heading difference (about 24) to the FPSO. To align the tanker with
the FPSO, which is a necessary operation to send back the hose, the tanker DP operator
took action to maneuver the vessel. In a combination of technical failure and
inappropriate DP operation, a drive-off was initiated, which ultimately resulted in a
collision.
Normal
Wind
Hawser
FPSO
Tanker
Hose
Wave, Current
Fishtailing
Heading
Deviation
FPSO
Yawing
Tanker
Figure 4-2
Yawing illustration
There are very limited offshore QRA studies that have addressed the excessive surging
and yawing events in tandem offloading operations. For those studies that have included
the risk modeling of these events, expert judgment is a typical approach. The occurrence
probabilities of excessive surging and yawing are estimated by a group of experts, and
event development after these two basic events is modeled. However, this approach may
not offer much information about how to reduce the occurrence of the two basic events.
At best, qualitative measures may be identified based on the experiences of the experts
involved.
Given the status described above, a systematic approach to predict the occurrence of
excessive surging and yawing and consistently minimize their occurrence is clearly
needed. The approach adopted in this Dr.Ing study is based on time-domain motion
simulations of a joint FPSO-tanker system, as described in the following section.
URN:NBN:no-3369
28
Chapter 4
URN:NBN:no-3369
29
measurements (Andersen, 2000; Blom, 2002) before being applied to the study of
surging and yawing.
4.2.2 Procedures
The simulation procedures for analyses of surging and yawing in tandem offloading are
formulated with the following two objectives in mind:
1. To predicate the likelihood of the excessive surging and yawing; and
2. To pinpoint contributions from various technical and operational factors, and
identify measures to reduce the occurrence of such events in the operation.
These two objectives may be different from traditional applications of time-domain
motion simulations, e.g. mooring and riser system analyses and thruster consumption
studies, in the sense that the interested parameters are the relative distance and the
relative heading between the two vessels. Specifically, given the objectives and what the
SIMO code actually can perform, simulations of surging and yawing are structured in
the three steps described below.
First, the offloading operation is simulated for three hours. Note that the whole
operation may take well above twenty hours. The three-hour simulation is considered as
a sample of operation, through which we are able to pinpoint which factors (in
technical and operational categories) have contributed to the occurrence of surging and
yawing.
Second, the three-hour time-domain simulation is performed twenty times with random
seeds for generating time series of wind and wave. The simulated relative distance
(tanker bow to FPSO stern mooring point) and relative heading between FPSO and
tanker are analyzed by fitting their extreme values (from 20 simulations) into the
statistical models, i.e. the first type extreme value distribution. The fitted distributions
are used to assess the occurrence probability of excessive surging and yawing events.
Third, a marginal operational weather condition is selected in simulations. Sensitivity
studies are performed to analyze contributions from various technical configurations
and operational philosophies this weather condition. Various environmental conditions,
given that the weather criteria are satisfied and thrust demands on both vessels are
within their thruster capacity limits, mainly influence thruster power consumption, and
have limited influence on vessel motions. This is because the dynamic positioning
control in the operation keeps vessel motions in a similar order of magnitude under
various weather conditions, as can be observed from the full-scale FPSO and tanker
motion measurements (Andersen, 2000; Blom, 2002). Therefore, the present study does
not include the environmental sensitivity.
URN:NBN:no-3369
30
Chapter 4
4.2.3 Limitations
It is also important to notice the following limitations in the simulation-based approach
at present. Basically, the simulation work is carried by SIMO, and this program itself
has idealizations, for example, in the modeling of the DP control system. However,
given the calibration work performed, those program idealizations do not change the
main conclusions in this study. Regarding the simulation model, further work is needed
to take the following into account.
The thruster power limitations on both the FPSO and the tanker are not considered, i.e.
both vessels have abundant positioning capacities. Subsequently, the environmental
impact on vessel motion behavior is considered small and not studied in the present
sensitivity studies. However, in actual operation there are positioning capacity
limitations, especially on the tanker side. Environmental conditions, such as collinear
vs. non-collinear wind-wave-current, abnormally large current, and etc., will inevitably
influence the surging and yawing events.
Weather is assumed stable (i.e. no weather change) during the three-hour simulation.
Therefore scenarios involving for example a sudden wind change, will not be included.
In such cases, the FPSO may start to change heading, and the tanker has to follow.
During the transition period to the next equilibrium (stable) condition, yawing might be
a problem. Alternatively if the FPSO keeps heading, the tanker may be drifted to a
new heading by the changed environmental forces. In that case, significant yawing
(heading deviation) may be developed in a short time.
The draught changes of FPSO and tanker during the 3-hour simulation (and subsequent
changes of two vessels hydrodynamic coefficients) are not incorporated in the present
simulation model. In principle, this could be done by making a number of simulation
models and simulate motions respectively, e.g. ballast FPSO + fully loaded tanker, and
vice versa. The hydrodynamic interactions between FPSO and tanker, e.g. the shadow
effect (Fucatu et al. 2001) of wind and current from FPSO on tanker, are not included in
the simulation.
4.3 VESSEL MODELS
The two vessel models in simulation of tandem offloading are illustrated in Figure 4-3.
The FPSO is a purpose-built vessel for operation in the North Sea. The FPSO model in
SIMO consists of hull and positioning system. The main particulars of the vessel are
listed in Table 4-1. The hydrodynamic data for the FPSO hull are synthesized by results
from WADAM1 calculations, model tests, and calibration work. Detailed data and their
sources are listed in Table 4-2. Note that these FPSO data are for a medium draught
loading condition.
URN:NBN:no-3369
Figure 4-3
31
The positioning system of the FPSO consists of a internal turret mooring system, three
DP operated thrusters, and a DP control system. There are twelve equally spaced
catenary mooring lines. Each mooring line is a combination of chain and wire rope, and
the breaking strength is 10000 kN. Three azimuth thrusters are modeled as an
approximation to the five real life thrusters, one bow thruster (860 kN) and two stern
thrusters (430 kN and 860 kN, respectively). The FPSO DP system is modeled by a PID
Controller in SIMO, which is based on conventional PID control theory (Reinholdtsen
and Falkenberg, 2001). Geo-stationary reference position and reference heading are
specified for vessel offset and heading control operations.
The tanker is a North Sea DP2 class shuttle tanker. It is positioned 80 m behind the
FPSO in DP mode. Similar to the FPSO model, the tanker model in SIMO consists of
hull and positioning system. The hull data of the tanker are listed in Tables 4-1 and 4-2.
Note that these tanker data are for a deep draught loading condition.
The tanker positioning system in principle consists of bow and stern thrusters, main
propellers and rudders, the DP control system, and the hawser. The hawser is modeled
as a non-load bearing soft spring which has tension of around 20 kN due to self-weight.
The thrusters, main propellers and rudders are modeled in SIMO with an idealization
based on information from the DP system designer (Gudmestad and Aanonsen, 2001).
One bow azimuth thruster (620 kN) and one stern tunnel thruster (190 kN) are modeled.
For simplicity, the possible thrust generated by the combination of two rudders and two
main propellers are modeled by one fictitious azimuth thruster (770 kN) at stern.
Similar to the FPSO model, the tanker DP system is modeled by the PID Controller in
SIMO. It is possible in SIMO to use the moving FPSO stern and heading as the
positioning references for the tanker, in addition to the geo-stationary references. This
makes simulation of different DP software and various DP operational strategies used
on the tanker in tandem offloading possible.
URN:NBN:no-3369
32
Main Particulars
Chapter 4
FPSO
Length (m)
Breadth (m)
Depth (m)
Draught (m)
Max. Draught (m)
Mass (Mg)
Table 4-1
Main particulars
Tanker
260
41
25
15.5
19
119,600
265
42.5
22
14
15
112,000
Sources
Model test
WADAM
Model test, empirical estimation
Calibration
WADAM
WADAM
WADAM
Wind tunnel test
Model test
To calibrate the vessel models so that motions simulated are physically reasonable is an
important task. In this study, full-scale measurements of the horizontal motions of an
FPSO and a DP shuttle tanker during offloading are obtained, and used for the model
calibration work. Given its length and contents and the overall theme in this chapter, the
calibration work is elaborated in detail in Appendix C. The conclusion made after the
calibration is that the present two-vessel model is able to reasonably simulate the
physical horizontal motions between FPSO and tanker in offloading.
4.4 ENVIRONMENTAL CONDITIONS
The environment used in simulation is close to the operational limit. Mild and medium
environmental conditions in tandem offloading are not included at present. The
marginal operational environment is modeled based on information from onboard
measurements during offloading operation (Andersen, 2000) and supplemented with
hindcast data from the Norwegian Meteorological Institute (DNMI, 2000).
The wave is modeled as an irregular short-crested wave by a three-parameter
JONSWAP spectrum with a cos spreading function. The wind is modeled by a one-hour
mean wind speed plus a NPD gust spectrum. Note that neither measurement nor
hindcast includes current data. Therefore the current velocity profile and direction are
based on simple assumptions, i.e. the current is mainly the wind- and wave-generated
current. According to the DNV Class Note 30.5 (1991) the wind- and wave-driven
current may be estimated as 1.5 % of the wind speed. The current direction is ideally
URN:NBN:no-3369
33
assumed to be close to the wind and wave directions, i.e. propagating 180q relative to
the mean FPSO heading. The swell is possible to model based on hindcast information.
However, it is disregarded due to the present SIMO limitation, i.e. the swell affects only
high-frequency (HF) motion in simulation, while the low frequency (LF) motion is the
main interest in this study.
The resulting wind, wave, and current data are summarized in Table 4-3 together with
their sources. The wind, wave and current directions are further illustrated in Figure 4-4.
Environmental Parameters
Significant wave height (m)
5.4
Peak period (sec)
12.6
Spectrum
JONSWAP
Wave direction1 (deg)
171q
Wind speed2 (m/s)
16.3
Wind direction (deg)
-174q
Current velocity (m/s)
0.24
Current direction (deg)
180q
Table 4-3
Offloading environmental condition
Sources
Measurement
Hindcast
Assumption
Hindcast + Assumption
Measurement
Measurement
Assumption
Assumption
Last but not least, as a sample of operation, the above vessel configuration and
environmental condition reflect the following operational picture: The tanker has been
loading from the FPSO to take part of its storage, and the operation is approaching the
end. So the FPSO and tanker are in medium and fully loaded conditions, respectively.
Meanwhile, weather is deteriorating, and close to (but within) the operational limits.
FPSO Heading
Current
180q
Wind
Wave
Figure 4-4
URN:NBN:no-3369
34
Chapter 4
X 76.22
Pmin X 1 exp exp
0.6523
(4-1)
T 5.277
Pmax T exp exp
0.5602
(4-2)
To facilitate the quantitative discussion below, the excessive surging event is defined as
the minimum separation distance between tanker and FPSO smaller than 60 m, i.e. a 20
m reduction of nominal separation distance. Similarly, the excessive yawing event is
defined as the maximum heading difference between the two vessels larger than 20q.
Accordingly, the base case probabilities for excessive surging and yawing events are
presented in Table 4-4.
Excessive Surging in 3-hour simulation
1.59E-11
Probability (per 3-hour)
3.86E-12
The results in Table 4-4 show that in the base case, excessive surging and yawing event
probabilities are negligible. A number of technical and operational factors have
URN:NBN:no-3369
35
contributed to these good phenomena, and their contributions are analyzed in the
following sensitivity studies. Furthermore, we have to make note of the fact that when
an operation is repeated frequently, it is the very low probabilities that are of main
interest in the risk analysis.
4.6 SENSITIVITY STUDIES
The following assumptions are introduced in the sensitivity studies. In the surging event
study, we assume that both vessels use the same heading and yaw motion control as in
the base case. Similarly in the yawing event study, we assume that both vessels use the
same position reference and surge motion control as in the base case. This is because a
floating vessels surge and yaw motions are correlated. With different vessel mean
headings, the environmental forces acting on the vessel are different, and subsequently
different surge mean values and amplitudes will occur. The above assumptions are
therefore to minimize this correlation.
4.6.1 Surging and contact events
The following three factors that influence the surging event are identified.
1. How the tanker positions itself relative to the FPSO, i.e. use a geo-stationary motion
control window around the FPSO stern hawser terminal point, or a moving FPSO
stern point, for horizontal position reference. Note that in both cases the mean
separation distance is maintained around 80 m.
2. How the FPSO controls its surge motion, i.e. whether or not the FPSO uses its DP
thrusters to dampen down the surge amplitude.
3. How the tanker controls its surge motion, i.e. how much surge damping the tanker
thrusters are able to provide to dampen down the surge motion amplitude.
Note that the first and second factors may be influenced by design or by operational
strategy. For example, some FPSOs may only have one or two stern thrusters installed
for heading control. Its surge motion can only be restrained by the turret mooring
system. In other cases, the control room operator on FPSO may not use thrusters to
reduce the surge motion amplitude, even though enough thruster capacity is available.
Similar situations exist regarding the motion window function in tanker DP software,
i.e. no installation, or installed but no utilization. For the third factor, ideally it should
be investigated in the base case and then in each sensitivity case for various tanker surge
damping levels. This will increase the number of simulations significantly. However,
the comparison between the base case and the sensitivity cases should be made with the
same tanker surge damping level. The base case results in Table 4-4 imply that tanker
has effectively dampened down its surge motion. Therefore, for simplicity, the third
factor is assumed constant, i.e. the tanker surge damping provided by thrusters remains
same in all sensitivity cases as in the base case.
URN:NBN:no-3369
36
Chapter 4
According to the first and second surging influencing factors, three surging sensitivity
cases are formulated in Table 4-5.
FPSO surge motion control
Tanker position reference
As in the base case, twenty 3-hour simulations are performed for each sensitivity case,
and the simulated minimum separation distance values are fitted into the first type
extreme value distribution. The probabilities of surging events are then obtained. Values
corresponding to the excessive surging event are presented in Table 4-6.
X P
Pmin X 1 exp exp
P
V
Base Case
76.22
0.6523
1.59E-11
Surging Case 1
75.56
0.7509
1.00E-9
Surging Case 2
74.64
1.084
1.36E-6
Surging Case 3
72.27
1.184
3.16E-5
Table 4-6
Excessive surging event probabilities
Surging
Sensitivity
Probability of excessive
surging in 3-hour simulation
The contribution from the first factor, i.e. how the tanker positions itself relative to the
FPSO, is shown by a comparison between the base case and surging case 1, as well as
between surging case 2 and surging case 3. The probability of surging can roughly be
decreased 10-102 times if the tanker uses a geo-stationary window instead of a moving
FPSO stern point as the position reference. The contribution from the second factor, i.e.
how the FPSO controls its surge motion, is shown by a comparison between the base
case and surging case 2, as well as between surging case 1 and surging case 3. The
probability of surging may increase 104-105 times if the FPSO does not use thrusters (or
does not have enough thruster capacity) to dampen down its surge motion.
By a comparison between the base case and surging case 3, the probability of excessive
surging increases 106 times, and it is not as negligible as in the base case. Practically the
result reflects the difficulties of tandem offloading in a situation where the tanker uses a
moving FPSO stern point as the position reference, and the FPSO has a large surge
motion.
The contact event can be defined as a minimum separation distance between the FPSO
and tanker smaller than zero. The probability of contact, derived from the above
minimum separation distance distributions, is virtually zero for the base case and all
three sensitivity cases. This implies that a collision between FPSO and tanker caused by
a very excessive surging event alone is not possible, i.e. excessive surging does not
directly cause contact between FPSO and tanker.
URN:NBN:no-3369
37
Weathervane heading
Base Case
/
/
Yawing Case 1
Note that when studying the mean heading control in yawing case 1, the same FPSO
yaw motion control is assumed; while similarly the same FPSO weathervane mean
heading is assumed when studying the yaw motion control in yawing case 2. In practice,
however, FPSO mean heading and yaw motion are controlled by a single heading
control function in the DP system. The differentiation made here is an idealization in
order to pinpoint the relative contributions from the first and second influencing factors.
Based on twenty 3-hour simulations, the probabilities of the yawing events and the first
type of extreme value distribution parameters are obtained. Excessive yawing event
probabilities are presented in Table 4-8.
1
2
URN:NBN:no-3369
This reflects one possible FPSO heading operational strategy likely demanded by the production needs.
Tanker DP system determines the optimum weathervane mean heading.
38
Chapter 4
Yawing
Sensitivity
Probability of excessive
yawing in 3-hour simulation
Base Case
3.86E-12
Yawing Case 1
3.04E-07
Yawing Case 2
6.18E-05
Table 4-8
Excessive yawing event probabilities
T P
Pmax T exp exp
P
V
5.277
0.5602
15.05
0.3299
12.80
0.7431
The contribution of the first factor, i.e. how the FPSO and tanker position their mean
heading relative to each other, is shown by a comparison between the base case and
yawing case 1. The probability of an excessive yawing event is significantly increased
(105 times higher than in the base case) to a non-negligible level. These results
illuminate the importance of joint mean heading operation between FPSO and tanker. In
the given environmental condition, the tankers weathervane mean heading is about 6q
different from the FPSOs weathervane mean heading. If the FPSO has to be headed on
the wind, the mean heading difference between the two vessels further increases to 12q.
We may further infer from the simulation results that under certain environmental
condition, the tankers weathervane mean heading is potentially very different (e.g. 20q
or 25q) from the FPSO weathervane mean heading. If there is no mean heading
coordination between the two vessels, excessive yawing events (i.e. 20q heading
difference) may occur several times even in one single offloading operation.
The contribution of the second factor, i.e. how the FPSO controls its yaw motion, is to a
large extent shown by a comparison between the base case and yawing case 2. A very
significant increase of excessive yawing probability (107 times higher than in the base
case) is observed. These results largely demonstrate the importance of FPSO yaw
motion control. The large FPSO yaw motion may impact the probability of an excessive
yawing event, at least as much as (if not more than) the situation when there is no joint
mean heading operation. This is observed by a comparison between yawing case 2 and
yawing case 3.
4.7 FINDINGS AND RECOMMENDATIONS
To conclude the first objective, i.e. the likelihood of excessive surging and yawing,
results show that, given a rather extreme offloading environment, the optimum technical
configuration and operational strategy on both FPSO and tanker may result in negligible
frequencies of excessive surging and yawing events (1.59E-11 and 3.86E-12,
respectively, per 3-hour duration). However, from sensitivity results, the excessive
surging and yawing frequencies can vary up to 3.16E-05 and 6.18E-05, respectively, per
3-hour duration. Assume that each year there are 300 loading hours under weather
condition similar to the one that are chosen, then the frequencies of excessive surging
and yawing can both be in the order of 10-3 per year, high enough to be an important
safety concern. Findings indicate that surging and yawing events should, and can be,
effectively minimized.
URN:NBN:no-3369
39
In addition, simulation results confirm that the excessive surging event, though it may
lead to a frequent reduction of the separation distance, does not directly cause the
contact between FPSO and tanker, given that the mean separation distance is around 50
to 80 meters. However, it may potentially initiate technical failures and/or human errors
as discussed earlier in this chapter and according to facts of incidents shown in
Appendix B.
To conclude the second objective, i.e. how to reduce the occurrence of excessive
surging and yawing events, findings and recommendations are as follows:
1. Significant contributions to excessive surging and yawing come from the FPSOs
surge and yaw motions if these motion amplitudes are not properly dampened down.
Efforts, e.g. in the form of operational guidelines, should be made to reduce the
FPSO surge and yaw motions during the offloading operation, given that the FPSO
has such thruster capacity.
2. The coordination of mean heading control between FPSO and tanker is important to
minimize the probability of excessive yawing. Tanker and FPSO should align with
the same mean heading. This heading can be determined through communication
between tanker DP operator and FPSO control room operator regarding each
vessels positioning preferences and capabilities. In some cases it may be that the
tanker heading is aligned with the optimum FPSO heading, while in other cases it
can be that the FPSO heading is adjusted to align with the optimum tanker heading.
This recommendation is valid as long as the FPSO has thruster capacity for heading
control.
3. Tanker using a geo-stationary motion control window around a mean FPSO stern
point for position reference can reduce the probability of excessive surging,
compared to the case when a moving FPSO stern point is used for positioning. This
is the measure that targets on tanker side solely, especially if the FPSO do not have
thruster capacity to dampen down surge motions.
Last but not least, for tandem offloading operations involving those passively moored
FPSOs which do not have thruster capacity to dampen down the surge motions, nor the
heading control capability, findings show that the excessive surging and yawing events
are likely to happen even if best practices and equipment are adopted on the DP shuttle
tanker side. A contingency operational plan, e.g. special criteria for disconnection,
should be considered in order to handle the excessive surging and yawing these likely
(with such FPSOs) and failure prone situations.
URN:NBN:no-3369
C H A P T E R
5.
FA I L U R E O F R E C O V E RY
accidents occur, not because they (pilots) have been sloppy,
careless, or willfully disobedient, but because we on the ground
have laid booby traps for them, into which they have finally
fallen.1
R. H U R S T
A tanker drive-off forward event does not turn into a collision incident with the FPSO,
if recovery initiated from the tanker is successful (See Eq.2-2 in Chapter 2). In this
chapter, recovery actions initiated by the tanker DP operator in drive-off scenarios are
studied.
Recovery actions are guided by three possible recovery strategies. The one that is
favored by the majority of shuttle tanker DP operators and safety specialists is to stop
tanker, and meanwhile combine the effort to rotate the vessel bow away from the FPSO
stern. Based on calibrated tanker motion simulations, the time available for the DP
operator to initiate recovery action so that tanker can be stopped within a separation
distance to FPSO, e.g. 80 m, is found to be critically short. A 3-stage InformationDecision-Execution model is generalized to model the DP operators informationprocessing stages regarding action initiation when in a drive-off scenario. Based on this
human information-processing model, expert judgment by simulator trainer and a
questionnaire survey with shuttle tanker DP operators are conducted, to obtain a
reasonable estimate of the time needed for action initiation. The estimates are found
convergent to the facts in the incidents. Findings may imply that tanker DP operators in
general need more time to initiate recovery action than the allowable time window. In
other words, recovery failure is likely due to lack of reaction time. Two principal
recommendations are proposed accordingly, i.e. to increase the available time window
and/or to reduce the DP operator reaction time.
Hurst R and L.R. (editors) Pilot Error. 2nd Ed., Aaronson, New York, U.S.A. 1982. The sentences was
quoted in Kletz (2001).
40
URN:NBN:no-3369
Failure of Recovery
41
P ( collision | AR , PFM ) P ( AR
j
| PFM i ) P(PFM i )
(5-1)
Incident and near miss information (Appendix B). Talk through and walk through of
recovery action onboard a shuttle tanker (Chen, 2001).
Interviews with operators and experts, including: shuttle tanker captain and DP
officer (Chen, 2001), DP software designer (Hals, 2001), simulator DP training
instructor at Ship Manoeuvring Simulator Centre in Trondheim (Chen, 2002b), and
safety specialist (Helgy, 2002).
Questionnaire survey with shuttle tanker captains and DP officers (Chen, 2002c &
2002d).
The recovery actions performed by the tanker DP operator can be considered as guided
by the three recovery strategies shown in Figure 5-1. The recovery actions and the event
URN:NBN:no-3369
42
Chapter 5
development accordingly are modeled by combining an event tree model with a time
axis in Figure 5-2. The three recovery strategies are exemplified by routines (1), (2), and
(3) in the event tree model. The event tree provides an overview of how an event may
develop (into collision or near miss) under various recovery actions from the DP
operator. The time axis starts at the initiation of drive-off. Operator action timing is
represented from T1 to T5, respectively.
The No.1 strategy is to maximize the rudder and thruster effect so that maximum
turning moment is generated, and tanker is steered away from FPSO stern. Note that
during this strategy, no efforts are made to stop the tanker. The No.2 strategy is to try to
gain local thruster control and command full astern thrust so that tanker could be
stopped within the separation distance to FPSO stern. The No.3 strategy can be seen as
a combination of the above two, i.e. try to gain full astern thrust and initiate maximum
turning moment from rudder and thrusters.
FPSO Stern
FPSO Stern
FPSO Stern
favored
(1)
Tanker
Bow
Figure 5-1
(2)
Tanker
Bow
(3)
Tanker
Bow
Recovery strategies
There are different views regarding which recovery strategy is the optimum one in
drive-off scenarios. For example, in the previous study by HSE (1997), the No.1
strategy was favored, while the No.2 was claimed to be most likely to fail. However,
in this study, we found that No.3 strategy is favored by the majority of shuttle tanker DP
operators and safety specialists. The main reason for selecting this strategy is that it
appears to be the safest (at least minimizing the impact energy), and it is natural to
perform in a high-stress situation. The questionnaire survey and interview further reveal
that to stop the tanker is the primary objective in the operators mind, and to steer tanker
away from FPSO stern is mainly done to help achieving this goal.
The recovery actions may be carried out by using the DP joystick in Manual DP mode,
or by switching off the DP and maneuvering the tanker via manual steering gear. Both
had happened during actual collision incidents and near misses (Appendix B). Again,
findings in the questionnaire survey show that the majority of operators prefer to use the
DP joystick, or at least try this alternative first unless it is found not to work. The main
reason for using the DP joystick is that it is considered time saving, and the DP console
(and associated position reference system screens) offers a better overview.
URN:NBN:no-3369
Failure of Recovery
Action initiated
?
Manual Takeover
43
Astern Thrust
?
Initiated
Rotation Initiated
?
Thrusters & Rudders
Outcome
Collision
Collision
N
N
(1)
Collision
Near Miss
Collision
Near Miss
Collision
Near Miss
Collision
Near Miss
Collision
Near Miss
Drive-off
Forward
N
N
Y
Y
Y
(2)
N
Y
(3)
T1
Figure 5-2
T2
T3
T4
T5
Time
50
80
120
150
53
72
81
Table 5-1
The estimations are derived from a calibrated motion simulation for a generic North Sea
shuttle tanker. The simulation work is carried out by a time-domain simulation code,
URN:NBN:no-3369
44
Chapter 5
SIMO. The tanker simulation model is adapted as the one used in Chapter 4, and it is
calibrated based on the full-scale measurement of a North Sea shuttle tanker drive-off
behavior recorded by the BLOM PMS system1. The calibration work is documented at
the end of this chapter.
It is clear from Table 5-1 that the time window for successful recovery is critically
short. Note that the whole tandem offloading operation may last over 20 hours.
Meanwhile, a full ahead tanker drive-off initiated by the DP system for example, may
only need 2 minutes to develop into a collision with the FPSO. The tanker DP operator
has to initiate recovery action within the first 53 seconds after drive-off to make a
successful intervention, given an 80 m separation distance. This is a very stressful
situation.
5.4 MODELING OF ACTION INITIATION
A quantitative estimate of how much time a tanker DP operator in general needs to
initiate action is needed, i.e. the T1 corresponding to the Manual Takeover action in
Figure 5-2. A quantitative estimation of T1 has to be based on a sound qualitative
understanding of the information-processing stages that a tanker DP operator undergoes
before he or she2 acts.
There have been many studies of operator action and time in emergency situations in the
past two decades, for example Time Reliability Correlation (TRC) (Hannaman and
Worledge, 1988) and Human Cognitive Reliability (HCR) (Dougherty and Fragola,
1988) in the late 1980s, operator cognitive model and response action analysis under
accident conditions (Parry, 1995; Hollnagel, 1996; Smidts et al., 1997) in the mid1990s, and human reliability of emergency tasks in nuclear power plants (Pyy, 2000;
Jung et al., 2001) in the early 2000s.
However, the context of a shuttle tanker is different from that of a nuclear power plant
(NPP), which many of the above studies are largely rooted. The nature of the task,
human-machine interface, and safety culture, to list a few, are basically different.
Moreover, during an emergency drive-off scenario on tanker in tandem offloading, a
tanker DP operator is not, as the control room operator in NPP is when under
emergency, guided to take corrective actions with various emergency operating
procedures (EOPs). EOPs for DP operator in a tanker drive-off scenario in tandem
offloading are in general scarce.
A simple 3-stage Information-Decision-Execution model is generalized to model the
information-processing stages that a DP operator generally experiences from 0 to T1 in
a tanker drive-off scenario. Note that the objective of this human action model is to
1
BLOM PMS system is a position monitoring system which is installed on shuttle tankers. It has a
position data log, e.g. tanker position and speed.
2
I use he/him/his thereafter when referring to the DP operator, although the person may be either man or
woman.
URN:NBN:no-3369
Failure of Recovery
45
provide the basis for the following quantitative estimation of the response time T1 in
the present scenario. Further work will be needed if this human action model is used for
other purposes, such as human error analysis. A qualitative description of operator
activities prior to recovery action initiation in a tanker drive-off scenario is provided
below. This is based on the collected operational information, and it is the factual
background of the generalized model.
Information During DP watch, the operator may detect abnormal signals (detection).
For example, he may be alerted by a distance alarm. Or he may, when monitoring the
offloading, observe an abnormal thrust output, or he may observe that the vessel starts
to gain forward speed. After the DP operator detects the first abnormal signals, he may
start to actively search for information (observation) to clarify the situation (state
evaluation), i.e., whether there merely is a wrong signal or whether a drive-off actually
takes place. He may perform crosschecks of four information sources, i.e. position,
speed, thrust output, and alarm, to detect the situation. Other sources, e.g. noise from
engine and vibrations, may also be paid attention to.
Decision This stage involves interaction between state evaluation and task
formulation. During state evaluation, the DP operator processes the information
obtained. He may find that it merely is a wrong signal, and then select a minor,
correcting task. Or he may find that this is a drive-off, and he will have to check the
vessel position (distance to FPSO), speed, and the thrust output all over again. The
information helps him to decide on how critical the situation is, how much time window
he has, and this helps him to formulate the appropriate tasks which he believes will
prevent the collision. He will also consider the environmental conditions and vessel
thruster, rudder capacities, and response time, when planning the tasks.
Execution The last stage is the task execution. The formulated tasks are transformed
into sequenced muscle commands, and the DP operator subsequently confirms (by
observation) that the execution is being achieved. Note that this stage may be rather
brief if the command is quickly confirmed as intended. However, in some cases when
there have been some technical failures, a command may result no effect at all, or in a
stressed situation, the command can even be performed on a wrong object. The operator
may try again and wait (search information to confirm that the command is being
achieved) and try until he decides to perform another task or identifies the right object
for command. In those cases, the execution stage may involve a longer time span.
The Information-Decision-Execution model is presented in Figure 5-3 with the time
reference. Note that these three stages do not happen in a purely linear, sequential
manner. The estimation of DP operator action initiation time T1 in a drive-off scenario
is accordingly based on estimations of the following three characteristic time interval
values as shown in Figure 5-3:
URN:NBN:no-3369
46
Chapter 5
Drive-off
Initiation
First Abnormal
Signal
Drive off
Confirmation
Action
Initiation
DECISION
State Evaluation
Task Formulation
Data
Figure 5-3
Ta
INFORMATION
EXECUTION
Observation
Detection
Muscle Command
Td
Action
T1
Time
The theoretical background for constructing this simple 3-stage operator action model is
briefly described here. First, this model is largely adapted from the human informationprocessing model used in a study of pilot action in aviation operations by Wickens and
Flach (1988). More details concerning this human information-processing model can
also be found in Appendix A, or from the recent engineering psychology book by
Wickens and Hollands (2000). The aviation pilots perform operations in a context
which is considered to be of significant similarity to the one shuttle tanker DP operators
face in drive-off scenarios. For example, both cases involve receiving external
information, assessing the situation and performing action under critical time pressure.
Both have a few action alternatives to choose, and actions are all performed in a
confined area (airplane cockpit vs. tanker bridge) with various steering gears. However,
significant simplifications have been made in our model due to its present objective.
Second, the hierarchy and interaction between Information and Decision in our model
are rooted in the Step-Ladder model developed by Rasmussen (1986). The interactions
between various stages in the Step-Ladder model are reflected in our model between
Information and Decision. Note that there is no direct link from Information to
Execution in our model. This is because this type of skill-based behavior is not
considered possible, i.e., the tanker DP operator will not simply disconnect and initiate
full astern maneuvering by reacting automatically to one or several signals.
5.5 TIME NEEDED FOR ACTION INITIATION
On average, experienced tanker DP operators may need 60 to 90 s to initiate recovery
action. This is the conclusion derived from the following three sources: a) information
from incidents and near misses; b) expert judgments by a DP training instructor based
on his extensive experiences with tanker offshore loading training, in particular the
drive-off scenarios training in simulator; and c) a questionnaire survey among North
Sea shuttle tanker captains and DP officers.
URN:NBN:no-3369
Failure of Recovery
47
Collisions
Close to 120
120
91
143
167
Not available
58
125
Near Misses
f
i
Table 5-2
45
Very short
75
Operator recovery action initiation time (incident data)
The time span of the action initiation is observed between 58 to 167 s in collision
incidents, and from very short to 45 s in near misses. Further, incidents b, c, and g
could be considered as collisions due to full ahead tanker drive-offs, based on the time
from drive-off initiation to collision. These three incidents may imply that the time
needed to initiate recovery action lies between 60 to 120 s.
5.5.2 Expert judgment
The operator action initiation time is estimated by a DP training instructor in the Ship
Manoeuvring Simulator Centre (SMS) in Trondheim. This instructor has experience
from hundreds of tandem offloading DP training courses performed in the past a few
years. During the courses, the participants are at a random time exposed to various
failures that may cause (or combine with) tanker drive-off. The instructor observes the
participants responses in the Bridge simulator via a video camera. (There are however
no records of performance during training. This rules out the possibility of a statistical
analysis of response time in training.)
The estimation process is built on the simple operator response action model presented
in Section 5.4. Detailed estimates are presented in Table 5-3. The approach is outlined
below. 100 times of training with experienced DP operators are considered. The
population excludes training of officers who are first-time participants in tanker
offshore loading DP training course, and who seldom have operated the DP onboard.
They spend much more time to initiate action in simulated drive-off scenarios. After
initiation of tanker drive-off, percentage of training is estimated for different time
URN:NBN:no-3369
48
Chapter 5
intervals for the Information, Decision, and Execution stages (0-Ta, Ta-Td, Td-T1,
respectively). The Information and Decision stages are distinguished by observing that
the trainee has detected the first abnormal signal relating to drive-off. The time
intervals were pre-made and were updated by the expert during the estimation process.
In practice, it is difficult to differentiate the Decision and Execution stages from the
observation of trainee performance in the simulator, therefore these two stages are
grouped together.
Information Stage
Time (sec)
0 - 10 10 - 20 20 - 30 30 - 50 0 - 20 20 - 30 30 - 60 60 - 90
10
20
20
50
20
30
50
Probability
0.1
0.2
0.2
0.5
0.0
0.2
0.3
0.5
Table 5-3
Mean values are calculated based on the results in Table 5-3. The mean Information
time is 29 s, and the mean Decision and Execution time is 56 s. The recovery action is
then averagely initiated about 85 s after the initiation of drive-off. These results largely
converge with the incident data.
During estimation, the expert commented that the work attitude heavily influences the
detection of abnormal signals indicating tanker drive-off. The experience gained from
emergency training and knowledge of the system (hardware) are two factors that have
significant impact on the time involved in decision and action execution.
5.5.3 Questionnaire survey
The questionnaire survey conducted in the spring of 2002 with shuttle tanker DP
operators provides quantitative estimates of the time needed for recovery action
initiation directly from the front-line operators. The questionnaire is attached in
Appendix E. A total of 17 shuttle tanker DP operators (captains and DP officers)
participated, and 16 of them provided applicable feedbacks for time estimation. The
operational experience behind these 16 feedbacks involves 1093 tandem offloadings.
The questions are designed according to the 3-stage human action model presented
above. Specifically, the answers to the below questions are essential for an estimation of
a reasonable action initiation time. The term reasonable is important in this context.
The questions are not for performance evaluation, i.e., to find out who is the best. They
are formulated to clarify what the reasonable human capability is.
1. What is the first abnormal signal in the Information stage?
2. What is the decision process, and how much Decision time (Ta-Td) is needed in the
Decision stage?
URN:NBN:no-3369
Failure of Recovery
49
3. What is the preferred recovery strategy, where should the needed actions be
performed, and how much Execution time (Td-T1) is needed in the Execution stage?
The mean time for recovery action initiation is found to be about 60 s after drive-off.
The 16 individual estimates are plotted in Figure 5-4. As shown in the figure, the
reasonable Decision time and Execution time are directly provided by each DP operator,
and an indirect estimate of reasonable Information time is then added uniformly.
Reasonable Time Needed To Initiate Recovery Action
150
Time (sec)
120
Execution Tim e
90
Decision Tim e
Inform ation Tim e
60
30
0
1
10 11 12 13 14 15 16
Feedback ID
Figure 5-4
The reasonable Information time is indirectly found as 30 s after drive-off. This is based
on the following reasoning: The most likely first abnormal signal is identified in the
survey. In non-alarm category, it is the Thruster output on DP console before any
alarm. In alarm category, it is the DP short distance warning.
The detection time of the alarm signal, assuming the DP short distance warning alarm
goes off at 10 m of setpoint distance, is estimated from the calibrated simulation of the
shuttle tanker drive-off behavior which is presented at the end of this chapter. This
alarm goes off at 31 s after drive-off. The 10 m assumption comes from a generic
tandem offloading field configuration as shown in the tandem offloading guideline by
(UKOOA, 2002).
The detection time of the non-alarm signal, i.e., the thruster output on the DP console,
may vary from person to person depending on e.g. job attitude and current attention
level. More discussions on factors that influence the non-alarm signal detection are
provided in Chapter 7. The non-alarm signal may likely be detected earlier than any
warning/alarm, but it may not be as reliable as the warning/alarm to prompt the
operators attention. The estimation made by the simulator instructor shows that 29 s is
a mean time involved. A representative Information time is therefore derived as 30 s
after drive-off.
URN:NBN:no-3369
50
Chapter 5
URN:NBN:no-3369
Failure of Recovery
51
URN:NBN:no-3369
52
Chapter 5
C H A P T E R
T A N K E R
M O D E L
A N N E X
C A L I B R A T I O N
The shuttle tanker model used in the time window simulation here is the same vessel
used in the surging and yawing simulation study. Its main particulars and hydrodynamic
data in SIMO are provided in Chapter 4, Section 4.3.
The calibration work is based on a full-scale measurement of a North Sea shuttle tanker
drive-off event recorded by the onboard BLOM PMS system. The setpoint distance,
bearing, speed, and vessel heading data for a North Sea shuttle tanker in a collision
incident with an FPSO were recorded, and the episode is reconstructed into a video
show (Blom, 2002).
The FPSO heading was 250 clockwise relative to the North, and this heading was kept
more or less constant. The tanker heading was around 228 (with variation from 227 to
228) throughout the drive-off. The minimum distance between tanker and FPSO, as
illustrated in Figure 5-5, can be derived by setpoint distance and bearing information as
in Eq.5-2.
Dmin = Dsetpoint
= Dsetpoint
Minimum Distance
(5-2)
Tanker Bow
FPSO Stern
Set Point Distance
FPSO
Tanker
Bearing
Figure 5-5
The drive-off distance is derived from the calculated minimum distance values. The
starting point is taken from the first Blom data point, and it is 2 s after drive-off
initiation. Note that the time of drive-off initiation is found from investigation
URN:NBN:no-3369
Failure of Recovery
53
information. This position can therefore be viewed as the approximate initial position of
the tanker. The calculated drive-off distance values together with the tanker speed data
are listed in Table 5-4.
Time (s) after
Setpoint
Drive-off
distance (m)
initiation
2
20
36
43
46
51
58
63
68
74
84
94
100
110
115
120
125
Table 5-4
75.2
71.0
64.8
60.8
59.5
55.2
50.4
47.0
45.1
39.6
32.1
27.0
25.1
23.5
25.1
25.2
26.7
Bearing (q)
214
216
211
213
211
212
208
206
203
199
195
186
178
165
155
149
144
0.1
0.3
0.7
0.8
0.9
1.0
1.2
1.3
1.4
1.5
1.5
1.5
1.5
1.4
1.3
1.2
1.2
0.05
0.17
0.36
0.41
0.46
0.51
0.62
0.67
0.72
0.77
0.77
0.77
0.77
0.72
0.67
0.62
0.62
The warning/alarm events can be observed directly from the reconstructed Blom video.
The collision took place at around 125 s after drive-off, since after this time the driveoff distance stopped to increase with the time. The recovery action was initiated at 58 s
after drive-off initiation.
Idealized simulations are made in the still water cases. The same operator action is
imposed at the exact time in simulation as happened in the measurement. The tanker
propeller force is assumed to vary linearly in simulation. The rising and falling time for
ahead/astern propeller forces are both 30 seconds. The steady forward thrust and astern
thrust involved in drive-off are estimated as 1650 kN and 1500 kN, respectively. The
simulated distance-time and speed-time plots are presented in Figure 5-6 and Figure
5-7. Reasonable agreement between the simulation results and the measurement data is
observed in these plots. Therefore, the calibrated tanker model is considered to be able
to simulate a representative tanker drive-off behavior.
By varying one parameter, i.e., the operator action time, various distances needed to
stop the tanker can be found. The times corresponding to a 50 m, 80 m, 120 m, and 150
m stop-distance are in this way estimated, as listed in Table 5-1.
URN:NBN:no-3369
54
Chapter 5
Figure 5-6
Figure 5-7
URN:NBN:no-3369
C H A P T E R
6.
Kletz T. An Engineers View of Human Error. 3rd Ed., Institution of Chemical Engineers, 2001.
55
URN:NBN:no-3369
56
Chapter 6
It is true that with a long separation distance, the tanker will be able to pick up a high
speed in drive-off and subsequently result in high impact energy in collision, given that
URN:NBN:no-3369
57
URN:NBN:no-3369
58
Chapter 6
URN:NBN:no-3369
59
50
80
120
150
53
72
81
1,0
0,8
0,6
0,4
0,2
0,0
0
20
40
60
80
100
120
140
Time (sec)
0.93
0.73
0.34
0.18
60
80
100
120
140
160
Figure 6-1
In Chapter 3 the tanker drive-off frequency is found in the range of 5.4E-03 to 2.0E-02
per tandem loading. Assume that there are 50 tandem offloadings from a FPSO
installation annually, and given the lower limit value, the annual tanker drive-off
frequency in tandem offloading is roughly 0.27. Taking into account the failure
probability of recovery obtained above, the 80 m separation distance may have an
annual collision frequency 1.98E-01, while the 120 m and 150 m distances will have
annual collision frequencies of 9.13E-02 and 4.89E-02, respectively. These values are
still much higher than a classic 1.0E-04 risk acceptance criterion. Partly, we have to
note that the failure probabilities of recovery come from the fitted distribution based on
operators judgment. These values are valid for pointing out the relative differences, but
URN:NBN:no-3369
60
Chapter 6
may not be valid in an absolute sense for the acceptability issue. Partly, the still high
collision frequencies imply that other risk reduction measures, i.e., reducing the driveoff frequency, and/or changing operator reaction time characteristics, are also important
in order to further reduce the collision frequency.
6.3 DESIGN A REASONABLE SEPARATION DISTANCE
The key question concerning implementation of the separation distance extension is to
know how much separation distance should be configured in a tandem offloading
operation. This has to be based on considerations of both the human operators need for
reaction time, and tanker drive-off behavior. These two considerations are integrated in
a parametric tanker drive-off motion simulation in which human action at various times
are imposed. The simulation work is carried out in a time-domain simulation code
SIMO. The calibrated tanker model in Chapter 5 is continually used here. The operator
reaction time is modeled based on the 3-stage human information processing model of
recovery action initiation (see Section 5.4), and findings from operator questionnaire
survey (see Section 5.5). The recovery action introduced in the simulation is to initiate
astern pitch from main propulsion so that the tanker can be stopped in a similar driveoff scenario as happened in one occurred incident.
Parametric cases are formulated in Table 6-2. In a drive-off scenario, the operator will
firstly detect an abnormal signal. This abnormal signal can either be an alarm signal or a
non-alarm signal. Two types of parametric cases are then formulated. One assumes that
the operator manages a detection via the alarm signal DP Short Distance Warning (31
seconds after drive-off). The other assumes that the operator manages an early detection
(16 s after drive-off) which may be based on one or several non-alarm signals. After the
first abnormal signal has been detected, the operator needs a certain time to achieve
situation awareness, formulate, and execute recovery action. Based on findings in the
questionnaire survey, three further parametric cases are formulated by assuming that the
operator needs 20 s, 40 s, or 60 s after detecting the first abnormal signal.
Likely detection - 31 s after drive-off
Case I-1
Case I-2
Case I-3
36
56
76
The simulated tanker drive-off behavior and distance to stop for the six parametric cases
are plotted in Figure 6-2 and Figure 6-3.
URN:NBN:no-3369
URN:NBN:no-3369
Figure 6-2
Figure 6-3
61
62
Chapter 6
URN:NBN:no-3369
C H A P T E R
7.
ANON
E A R LY D E T E C T I O N
To reduce the Information time, as defined in Figure 5-3, detection of abnormal signal
should be made as early as possible after drive-off. The early detection has a special
importance in a time-critical drive-off scenario, since late detection may result in an
63
URN:NBN:no-3369
64
Chapter 7
inevitable collision, even though quick situation awareness and swift task
selection/execution can be made. In the following sections, those possible signals that
indicate something is going wrong are identified. Based on those alarm and non-alarm
signals, improvements on early detection of abnormal signal are identified and
discussed.
7.1 ABNORMAL SIGNALS
The abnormal signal can be an alarm signal, related to vessel speed, position, and
thruster output. For example, in tanker DP Weather Vane mode during loading phase,
there are in total six warning/alarm signals, should tanker drive-off occur (SMS, 2000;
Helgy, 2002).
-
Vessel speed alarms: 1. DP (operator-set, e.g. 0.3 kn); 2. DP (max., e.g. 0.5 kn);
3. BLOM (independent speed alarm)
In addition, the abnormal signal may be a non-alarm signal, e.g. sound from the
engine, vibration on the bridge, abnormal thruster output showing on the DP console,
and so on. A non-alarm signal may typically appear earlier than a warning/alarm, and
subsequently may be detected earlier.
The 17 feedbacks from the operator questionnaire survey regarding what is the first
abnormal signal are presented in Figure 7-1 below. Note that it is not always possible
to clearly pinpoint the first. Therefore most operators indicated a few possible first
signals.
First Abnorm al Signal
Camera
6%
Artemis
6%
DARPS
35 %
35 %
53 %
29 %
12 %
6%
10 %
20 %
30 %
40 %
50 %
60 %
70 %
80 %
90 % 100 %
Percentage of 17 Operators
Figure 7-1
URN:NBN:no-3369
65
The survey results show that Thruster output on DP console before any alarm is
probably the first signal that prompts the operator to notice that something is going
wrong. 9 out of 17 operators indicated this signal as one of the first or the first.
Two other likely first-signal are DARPS screen which shows the speed and the
position before any alarm and Engine sound and vibration. All three signals are nonalarm ones and are supposed to be detected before any warning/alarm.
In the warning/alarm category, DP Short Distance Warning is identified as the most
probable signal to first tell the operator that something is going wrong. 5 out of 17
operators indicated this signal as one of the first or the first.
To detect a non-alarm signal involves a vigilant signal detection process. It requires that
the operator is constantly alert since signals may be unpredictable and infrequent in the
operation. An alarm signal is readily detectable. However, which parameters (e.g. speed
or distance) that are chosen and which limits that are set for alarms may influence the
time of detection. Initiatives regarding how to achieve an earlier detection are therefore
separately addressed for alarm and non-alarm signals in the following sections.
7.2 IMPROVING EARLY DETECTION ALARM SIGNAL
Alarm signals are readily detectable. An alarm sound and a corresponding screen text
which indicates that something is going wrong will most likely prompt the operators
immediate attention, hopefully in early stage of tanker drive-off. Note however that the
so-called alarm inflation, i.e. a significant number of alarms go off in a very short
time period potentially freezing the operators attention, is not applicable to the present
context. When being asked about the present alarm settings, the majority of the
operators (78 %) in the questionnaire survey indicated that the present alarm settings
practically are effective to help an early detection of a drive-off situation. Given this
positive feedback, is there any room for further improvement of alarm design and
setting? The answer is probably yes.
If early detection can be effectively achieved, speed, rather than distance, must be a
parameter to be paid attention to. The speed alarms should also be set so that they will
go off earlier than distance alarms if drive-off happens. This is because of the following
facts. As shown by the incident data in Chapter 6, a DP2 shuttle tanker may pick up
speed quickly within a relatively short distance. In other words, when the distance alarm
(which operators are likely pay attention to) goes off, the tanker may already have
picked up a high speed. Consequently, successful recovery will inevitably be difficult,
since to stop a tanker with a relatively high speed (e.g. 1.0 knots) within a relatively
short distance (e.g. 60 m) is very difficult.
However, from the survey results, the speed alarms in DP (both operator-selected and
system-selected maximum values) and BLOM PMS are not considered by most
operators as the first abnormal signal, although they are indeed effective early-warning
signals. In the incident information in Chapter 6, speed alarm from BLOM PMS
actually went off before the distance warning/alarm. Therefore, it appears strange that
speed alarms are not widely recognized as the first abnormal signal or at least among
URN:NBN:no-3369
66
Chapter 7
those first signs. The optimistic explanation is that the abnormal speed has already
been shown from the DARPS unit before any alarm, and operator has got this signal as
the first abnormal signal (so they did not select speed alarm as the first abnormal
signal). The pessimistic view, however, may imply that most operators do not actively
utilize speed alarms in their efforts to make an early detection.
A consensus among operators regarding the use of speed alarms for early detection of
drive-off has to be setup. This is the first recommendation. The importance of speed
alarms may possibly be reinforced during simulator training and illustrated in an
emergency operation guideline. If there is no such guideline, it is worthwhile to compile
one to handle drive-off in tandem offloading. A number of the recommendations in this
study and good practices from operators could be included and constantly updated in
such a guideline.
Secondly, a calibration of the limits of speed alarms for a specific field operation is
important. Speed alarms should not be set with too high speed limits, otherwise they
may go off after the distance alarm (e.g. DP short distance warning), and no earliness of
detection is achieved. At the same time, speed alarms should not be set with too low
speed limits, otherwise they may go off too frequently due to tanker surge motion in
normal operation. If so, the alarms will be neglected. The implementation of such
calibration work has to be field specific and vessel specific. The present alarm system
also enables the operator to do this calibration (select own abnormal speed limit) based
on field situation, personal experience, and preference.
7.3 IMPROVING EARLY DETECTION NON-ALARM SIGNAL
The real effective early detection has to be based on the early detection of an abnormal
non-alarm signal, rather than an alarm signal. This is because of the nature of the
human-machine system that we are considering now. As argued by Kletz et al. (1995)
for the human-computer system, operators should be looking for changes or trends
at the displays under their control, rather than managing by exception, i.e. waiting until
an alarm shows that something is wrong. We would need much more reliable alarms
than we have if operators always waited for them to operate.
The fact that the majority of operators consider non-alarm signals as the first abnormal
signal is a very positive and promising finding. This reflects that the operational
principle is reasonably maintained. The thruster output on DP console, speed and
position shown in DARPS screen, and engine sound and vibration are the non-alarm
signals that most likely are being detected first. To pay attention to these signals
certainly should be stated in the above proposed emergency operation guideline.
However, to detect non-alarm abnormal signals which are infrequent in a long duration
operation is not as easy as to acknowledge an alarm. Researchers in the human factor
area (signal detection) have concluded that this type of task (vigilance task) imposes a
sustained load on the working memory of the human brain and demands a continuous
supply of processing resources (Deaton and Parasuraman, 1988). This mental demand
may be as fatiguing as the sustained demand to keep ones eyes open and fixated. The
URN:NBN:no-3369
67
fatigue may eventually lead to the loss of vigilance, while the resource-demanding
nature makes this type of task susceptible to interference from other concurrent
activities (Wickens and Hollands, 2000).
Measures to improve the detection of a non-alarm signal are therefore identified as:
[1] to prevent DP operator fatigue.
[2] to prevent interference from other concurrent activities around DP operator.
[3] to provide observation training.
DP operator fatigue probably originated from a non-efficient bridge resource
management. In earlier practices in tandem offloading, it was typically only the captain
that stayed in charge of the DP watch for the whole offloading operation. It is hard to
believe that a captain can sustain a 20+ hours DP watch while keeping the same high
vigilance level. Shuttle tanker operators have realized this problem in recent years and
now better bridge resource management practices have been developed. An ideal
example might be: two DP operators work together for 8 hours maximum, one takes
charge of loading and one performs DP watch, and these two people shift their tasks
once every 2 hours. This example implicitly requires that 2nd officers onboard who
usually take charge of loading should also have DP certificates and be allowed to
perform DP watch.
DP operator fatigue may also be caused by the present human-machine interface design.
As illustrated by the figure to the left in Figure 7-2, the DP operator is sitting in an
awkward position to observe the two above-mentioned non-alarm signals from the
DARPS unit screen and the DP console, respectively. S/He probably has to shift
attention between these two vertically located screens frequently, in order to judge if the
present thruster output looks reasonable to the vessel position (relative to FPSO) and if
vessel speed appears reasonable (close to zero). The BLOM PMS screen is located
further away, and to observe detailed (but valuable) information shown on that screen
requires more effort. A possible way to improve the working environment here is to redesign all PRS screens (reduce size) and fit them as a screen wall on top of the DP
screen, as shown by the gray area in the figure to the right in Figure 7-2. It will be an
integrated DP console and PRS information center where the operator has access to
most of the information he needs without having to make repeated, large physical
movements. The height of these re-located PRS screens should also not interfere the
operators view from the bridge window.
Potential activities on bridge that may interfere DP operators attention during DP
watch should be clarified and avoided by operational procedure. From incidents and
near misses it is also clear that late detection of drive-off happened if DP operators
primary attention was not on the DP watch (interrupted by other activities), or not on
those non-alarm signals that he needs to pay attention to (lack of experience). Special
training for observation of non-alarm signal may be emphasized in simulator training of
drive-off intervention.
URN:NBN:no-3369
68
Chapter 7
BLOM PMS
DARPS
Figure 7-2
Q U I C K S I T U AT I O N AWA R E N E S S
The Decision time, as defined in Figure 5-3 in Chapter 5, refers to the time that the
operator needs to diagnose and understand the situation (situation awareness) after
getting the first abnormal signal. The diagnosis and situation awareness form the
foundation for effective choice or formulation of tasks to be performed. In a timecritical drive-off scenario, the present goal is to identify measures that may effectively
reduce the operators time involved in diagnosis and situation awareness. This goal is
achieved through a process built on both a generic human factor theory and specific
facts of the drive-off scenario in tandem offloading. The theory comes from the human
information-processing model of diagnosis and situation awareness in decision-making,
presented by Wickens and Hollands (2000). The facts are from findings in the operator
questionnaire survey.
Note that it is mainly the time issues that are addressed. No human error issues such as
diagnosis errors, wrong situation awareness, etc., are explicitly included in the
discussion. However, implicitly measures to help diagnosis and subsequently to reduce
time needed to understand the situation may also contribute to human error reduction.
7.4 MODELING OF DIAGNOSIS AND SITUATION AWARENESS
A human information processing model for DP operators diagnosis and status
evaluation process is presented in Figure 7-3. This model is adapted from the human
information processing model for decision making developed by Wickens and Hollands
(2000).
A generic interpretation of DP operators diagnosis and status evaluation process in
drive-off scenario, based on the model in Figure 7-3, is described as follows. After
detection of the first abnormal sign, the DP operator will start to search for more
information (observation). Here selective attention is involved for choosing which
information to observe and which to filter out. Such selection is based on knowledge
URN:NBN:no-3369
69
and experience in long-term memory, and it requires attentional efforts and resources.
The observed information is processed in working memory for updating and revising
beliefs or hypotheses, i.e. diagnosis. These beliefs and hypotheses originate from
knowledge and experience in long-term memory. There is generally an iterative process
involved in diagnosis, because initial hypotheses will trigger the search for further
information to either confirm or refute them. This forms a feedback loop, as reflected by
the information flow labeled confirmation in the figure.
Drive-off
Initiation
First Sign of
Abnormal
Drive off
Confirmation
Action
Initiation
Long-Term
Memory
DECISION
State Evaluation
Task Formulation
Data
INFORMATION
EXECUTION
Observation
Detection
Muscle Command
Ta
Td
Action
T1
Time
Selective
Attention
Confirmation
Working
Memory
Clue Filtering
Observation
Figure 7-3
Theoretically, in order to improve diagnosis and status awareness in terms of less time
and better quality, we may focus on the four main information processing components
in the above model as follows.
-
Facts concerning the above four components for the tanker DP operator to diagnose and
understand the situation after getting the first abnormal signal in drive-off are collected,
and documented in the following section. Based on these facts, potential improvements
are identified and discussed.
URN:NBN:no-3369
70
Chapter 7
6%
Wave information
6%
6%
35 %
71 %
76 %
Tanker speed
82 %
0%
10 %
20 %
30 %
40 %
50 %
60 %
70 %
80 %
90 % 100 %
Percentage of 17 operators
Figure 7-4
The observation process, which is guided by the selective attention, and the information
input to the working memory, are reflected in the following facts. To gather the wanted
data, the operator may check a number of equipment screens, e.g. DP console, BLOM
PMS screen, DARPS screen, wind sensor, and so on (see details in questionnaire in
Appendix D). The number of screen checks each operator performs in the state
evaluation process is presented in Figure 7-5.
There are two types of shuttle tanker main propeller: controllable pitch propeller (CPP) and fixed pitch
propeller (FPP). The majority of vessels performing tandem offloading in the North Sea use the CPP
concept. The facts collected via the questionnaire survey accordingly refer to the CPP concept only.
URN:NBN:no-3369
71
Number of Checks
10
8
7
6
5
4
5
4
5
4
4
3
2
2
0
1
10
11
12
13
14
15
16
17
Feedback ID
Figure 7-5
Results show that most operators perform four data checks, and the four most checked
equipment screens as reflected in Figure 7-6 are:
-
DP console
DARPS screen
Artemis screen
Screens Checked in diagnosis and situation awareness
6%
18 %
Wave Measurement
24 %
Wind Sensor
18 %
6%
94 %
ARTEMIS Screen
82 %
DARPS Screen
88 %
35 %
DP Console
0%
94 %
10 %
20 %
30 %
40 %
50 %
60 %
70 %
80 %
90 % 100 %
Percentage of 17 operators
Figure 7-6
URN:NBN:no-3369
72
Chapter 7
contents may be reflected indirectly in the above findings from the selective attention
and the observation. We have to note that an operator with more knowledge and
experience is generally able to allocate his attention in a better way, and this speeds up
the observation, and thus the necessary information to evaluate the hypotheses is
quickly accessible in the working memory, and ultimately this leads to a quick and
quality diagnosis.
7.6 MEASURES TO REDUCE DECISION TIME
Based on the theoretical model and the operational facts, three types of measures that
may effectively reduce operators time involved in diagnosis and situation awareness
are identified.
[1] Training - targeted to gain and accumulate knowledge and experience in the
long-term memory, as well as to improve the evaluating process in the working
memory.
[2] Proceduralization focused on best practice in the selective attention and
observation process.
[3] Automation designed to keep track of the essential vessel data, aggregate
evidence, and automatically diagnose (and intervene) if it is a drive-off situation.
7.6.1 Training
Extensive experience and practice in general help to achieve a better performance. As a
necessary way to practice and accumulate experience, training has been used
extensively to improve diagnosis and situation awareness. In both simulator training and
onboard DP-play training, by given drive-off scenarios originated from various sources,
a trainee may improve his ability of quick diagnosis and situation awareness
significantly. For further information on training, see SMS (2000) and UKOOA (2002).
7.6.2 Proceduralization
Generally, proceduralization is a technique for outlining prescriptions of techniques that
should be followed to improve the quality of decision-making (Bazerman, 1998).
Specifically, given the time pressure in drive-off scenario in tandem loading, it will be
beneficial if an emergency operation guideline (as proposed in Section 7.2 & 7.3) is
issued to standardize operator attention on those data that are essential to drive-off
diagnosis (Figure 7-4), and minimize the needed screen checks (Figure 7-5 and Figure
7-6). This is because the unfocused attention may result in more data checks that may
contain unnecessary information, and ultimately require longer time in diagnosis.
Note that this emergency operation guideline is not designed to let the DP operator
read and then follow in a time-critical drive-off scenario. Rather, it is to prepare the
operators mind for where to focus and what to check, should a drive-off occur. Further,
the recommended actions should be practiced extensively during training so that they
become a series of natural responses.
URN:NBN:no-3369
73
7.6.3 Automation
Computer automation is potentially a powerful tool to offer operators various types of
support in the diagnosis process. At present, alarms in DP and BLOM PMS systems are
set based on individual parameters, i.e., speed, position, and engine output (see Section
7.1). According to findings in the questionnaire survey, operators have to integrate
tanker speed, position, and propeller pitch information together in order to diagnose the
situation. It is therefore possible, by using the aggregated alarm concept (NPD, 2001),
to design a drive-off detection system to assist the operator in diagnosis and situation
awareness. This is in principle because when several units of information can be
combined into one single meaningful representation, the human brain capacity required
for handling this particular information will be reduced, and the brain will be able to
handle more information effectively (NPD, 2001).
The drive-off detection system should ideally be designed as an aggregated alarm which
takes input from the above three key data, i.e. (tanker) speed, position and heading
relative to FPSO, and main propeller pitch. This aggregated alarm may preferably be
imbedded in the BLOM PMS system, and it could function as a detector for a potential
drive-off, rather than for a single abnormal parameter. Based on its internal algorithm,
the drive-off detector can keep track of the three key data, analyze their trend, and
aggregate evidence if one, two, or three of those data exceed the pre-defined range.
Accordingly, recommendations to the operator may be given via a screen text and/or
voice message, for example:
-
If speed is increasing and over the limit, operator will be reminded to notice the
speed.
If speed exceeds the limit and the main propeller forward pitch is still
increasing, the operator will be asked to notice the early sign of drive-off, close
situation monitoring, please!
If speed exceeds the limit, the main propeller forward pitch is continuously
increasing, and distance to the FPSO is decreasing and finally reaches the shortdistance warning limit, the operator will be informed that there is a likely
drive-off, full astern maneuvering, please!.
After getting the above alarm messages, the operator should accordingly perform the
suggested action within, e.g. 5-10 seconds.
Given this drive-off detector is implemented in the system, it may also be beneficial to
further design a technical system that can automatically full-astern maneuver the tanker
if there is no operator reaction within, e.g. 10 seconds after the drive-off alarm. This
will be a technical barrier, in addition to the current human barrier, to prevent escalation
from tanker drive-off scenarios.
URN:NBN:no-3369
C H A P T E R
8.
This chapter summarizes the conclusions of this Dr.Ing study and proposes directions
for the future research. The conclusion is naturally divided into three areas: 1)
probabilistic modeling of FPSO and tanker collision in tandem offloading, 2)
evaluation/reduction of tanker drive-off probability, and 3) evaluation/reduction of
failure probability of recovery initiated by the tanker DP operator. The future research is
suggested in the following three levels. At level I, it is the supplementary modeling and
analyses in the context of FPSO and tanker collision in tandem offloading. At level II, it
is the application of the rationale behind the probabilistic model of the FPSOtanker
collision (which is the proposed operational safety modeling concept in Appendix A) in
risk analyses of other similar marine operations. At level III, the challenge is to integrate
generally the human element in offshore quantitative risk assessment methodology.
8.1 CONCLUSIONS
8.1.1 Modeling of FPSO-Tanker collision
An applicable probabilistic model of FPSO and tanker collision in the tandem
offloading operation is developed. Theoretically this model is based on the proposed
Operational Safety Modeling concept (see Appendix A), and practically it is built from
the collected operational data. By a coarse evaluation of possible scenarios, a practical
formulation of FPSO and tanker collision frequency is (Eq.2-2 in Chapter 2):
P Collision = P(PFM) u P(Failure of Tanker Initiated Recovery | PFM)
Evaluation, and ultimately reduction of FPSO and tanker collision probabilities in
tandem offloading, as suggested in the above model, focuses on the following two
stages in this study:
74
URN:NBN:no-3369
75
1. The initiating stage Identify and minimize all possible sources and situations
that may cause tanker drive-off forward.
2. The recovery stage Evaluate and improve the recovery actions initiated by the
tanker DP operator to avoid collision, should drive-off forward happen.
8.1.2 Probability of tanker drive-off
Tanker drive-off probability is portrayed by statistical data from an earlier study, recent
SYNERGI incident data, and judgments made by tanker DP operators. A relatively high
frequency of tanker drive-off in tandem offloading is found, likely ranging from 5.4E03 to 2.0E-02 per loading, or 0.25 to 1 time per year if there are 50 tandem offloadings
from FPSO annually.
The tanker drive-off scenario is investigated by examining nine such events in tandem
offloading based on investigation reports, interviews and discussions with individuals
who have direct or indirect information. It becomes clear that the initiation of tanker
drive-off involves a complex human-machine interaction, typically involving DP
hardware/software, position reference systems and vessel sensors, local thruster control
system, and DP operator. A significant finding is the evidence of the failure prone
situations, i.e. excessive relative horizontal motions between FPSO and tanker. It is
during one or several combined modes of these excessive relative motions that a
human-machine interaction process is initiated, e.g. by a technical failure event, or an
operator action (though not necessarily an erroneous one), which eventually causes
tanker drive-off. Among five collision incidents, four tanker drive-offs actually were
initiated in this manner.
The failure prone situations, termed as excessive surging and yawing events in this
study, are quantitatively analyzed based on simulated FPSO and tanker horizontal
motions in a time-domain code SIMO. The simulation models are calibrated mainly
based on the full-scale measurements. Three principle measures are identified to
effectively minimize the excessive surging and yawing events, i.e. minimizing FPSO
surge and yaw motions in offloading, coordinating mean heading between FPSO and
tanker, and using the dedicated DP software with the tandem loading function on tanker.
Ultimately, these measures may provide a sound operational environment where the
probability of tanker drive-off can be minimized.
8.1.3 Probability of recovery failure
Recovery actions initiated by tanker DP operator in the drive-off scenario are studied.
The allowable time for DP operator to initiate recovery action so that tanker can be
stopped in a drive-off scenario within the present separation distance, e.g. 50-90 m to
FPSO, is found to be critically short. Expert judgment by simulator trainer and
questionnaire survey with shuttle tanker DP operators are conducted to estimate how
much time is reasonably needed for tanker DP operator to initiate recovery action in a
drive-off scenario. The estimates are found to converge with the facts in the incidents,
and they imply that tanker DP operators in general need more time to initiate recovery
URN:NBN:no-3369
76
Chapter 8
action than the allowable time window. In other words, recovery failure is likely due to
lack of reaction time. Two principal recommendations are proposed accordingly, i.e. to
increase the available time window and/or to reduce the operator reaction time.
To increase the available time window, a promising measure is to substantially increase
of the separation distance between FPSO and tanker. The feasibility is confirmed from
several perspectives, i.e. limited hardware modification, virtually no additional
operational complexity, and positive opinions from operators. The gain is quantified and
shows a over 50 % reduction of the recovery failure probability if distance is increased
from 80 m to 120 m, and a 75 % reduction if the distance is increased from 80 m to 150
m. In practice, the separation distance between FPSO and tanker should be configured
with dual considerations of tanker drive-off behavior and the time needed by the tanker
DP operator to initiate recovery action, should a tanker drive-off occur. Doing so may
lead to an optimum field configuration for the FPSO-tanker tandem offloading which
inherently minimizes the collision risk.
Effective reduction of operator reaction time can be achieved by early detection and
quick situation awareness. Proposed measures are built on the generic human signal
detection and decision-making theory, and specifically designed for drive-off
intervention based on the operational facts collected via the questionnaire survey. The
use of speed alarms and their calibration for early detection of drive-off are emphasized.
Potential measures to improve early detection of non-alarm signals are identified as: 1)
to prevent DP operator fatigue, 2) to prevent interference from other concurrent
activities around DP operator, and 3) to provide observation training. For the purpose of
quick diagnosis and situation awareness, measures are identified as 1) to provide driveoff intervention training, 2) to issue an emergency operation guideline to standardize
operators attention on those data that are essential to drive-off diagnosis and minimize
the needed screen checks, and 3) to design an automatic drive-off detection and
intervention system to assist operator in detection, diagnosis/situation awareness, and
recovery action execution. These measures may directly reduce the probability of
FPSO-tanker collision in tandem offloading.
8.2 SUGGESTIONS FOR FUTURE WORK
8.2.1 Level I
The possible future research work at level I consists of the following modeling and
analyses supplementary to the present work, in the context of FPSO and tanker collision
in tandem offloading.
The probabilistic modeling of tanker drive-off scenario in Chapter 3 needs further
development. The present model does point out the failure prone situations that need to
be minimized. However, the area between the failure prone situations and the actual
tanker drive-off has not yet been modeled, though we do see a strong link between the
two, based on what had happened in the collision incidents. Risk reduction measures
targeted in this area are lacking at present. However, unwanted failure prone situations
URN:NBN:no-3369
77
may nevertheless happen in the operation, and operators do need contingency plans in
such situations.
Modeling of human-machine interaction involved in tanker drive-off in Chapter 3 also
needs further development. The probabilistic model of tanker drive-off in principle
takes into account three types of human action. However, the latent human action which
may interact with technical failure as well as with the other two types of human action
is not modeled. The latent human action has a vast span in terms of time and contents. It
may occur in design, construction, installation, operation, and maintenance, and it
involves front-line operators, maintenance personnel, and management teams. Modeling
of the latent human action, and the human-machine interaction in general, therefore
have to be based on an organizational point of view. Whether or not the probabilistic
modeling can capture the subtle dependent relationship between various factors at
various levels in a dynamic (not static) organization is a subject to debate. Other
modeling techniques should be tested to see if they are applicable to the tandem
offloading context.
The FPSO-tanker simulation models used in Chapter 4 may require some further work.
Basically the thruster power limitations on both FPSO and tanker are not considered,
i.e., both vessels have abundant positioning capacities. Subsequently, the environmental
impact on vessel motion behavior is considered small and not studied in the sensitivity
studies. However, in actual operation, there are positioning capacity limitations,
especially on the tanker side. Environmental conditions, such as collinear vs. noncollinear wind-wave-current, abnormal current, sudden changes of wind direction, etc.
will inevitable influence the surging and yawing events. The future simulation models
should take these scenarios and the positioning capacity factor into account. It is also
worth to note that the simulation models are calibrated to the North Sea conditions in
terms of technical specifications, operational philosophies, and environmental
conditions. Possible re-calibrations are needed if the models are used in other
geographic locations.
8.2.2 Level II
The possible future research work at level II is to apply the rationale behind the
probabilistic model of FPSO-tanker collision into the risk analyses of other similar
marine operations. It is actually the application of the Operational Safety Modeling
concept which is proposed in Appendix A. Similar marine operations may be
considered as positioning operation of ships and/or floating platforms. A typical
example may be the positioning of the DP drilling rig. In a broader scope, marine
operations with similar dynamics of human supervisory control in the human-machine
system are all applicable. Reference is made to Appendix A for details of the human
supervisory control and human-machine system dynamics that are considered here.
An application example may in principle be outlined as follows. Note that detailed
modeling requires a specific operational context, and the following are only generic
discussions. Given a certain portion of the operational data that have been collected,
scenarios that are breaching the designed performance limit may be identified and
URN:NBN:no-3369
78
Chapter 8
evaluated. The modeling of a critical scenario can similarly be structured into the
initiating stage and the recovery stage. The former models the occurrence of the
scenario and analyzes the occurrence probability, and the latter models the scenario
development and analyzes the systems recovery potential.
-
It may still be possible, though not necessarily, to identify some failure prone
situations where human-machine interaction happens which eventually leads to the
critical scenario. Analyzing and minimizing those failure prone situations in the
operation is where the risk analysis and reduction efforts should be placed first.
Human operators may play a vital role in the recovery stage, though the technical
system may also be involved. Operators positive contributions should be identified
and strengthened, while their negative impact should also be spotted and minimized.
Note that risk analyses typically focus on the negative contributions of human operators,
i.e., the human error. This mind-set should be corrected. Humans are not placed in the
system in order to bring in errors; in most cases they are actually the last safety barriers
to save the situation. Without taking into account (and systematically analyzing) the
positive human contributions, we will hardly know how to strengthen this last-safetybarrier in the operation. Constraints, which prevent operators from being able to do
what they are supposed to do, may escape attention and continue to exist in the system.
The existing time constraint (critically short) on tanker DP operator to intervene driveoff in tandem offloading is a typical example of this.
8.2.3 Level III
The future research work at level III is a vision of systematically integrating the human
element in offshore quantitative risk assessment (QRA). This vision is by all means
vague at present. The following thoughts serve as an early stage sketch.
The human element has been considered in QRA mainly in the consequence modeling,
i.e., fatality (both individual and group) risk. However, there are rooms in other areas
besides the consequence modeling, especially in the frequency modeling (and risk
reduction area accordingly), where human contributions may be identified and taken
into account. It is true that some component failure frequencies may have implicitly
included human contributions occurred in design, operation, and maintenance.
However, as revealed in this study, the frequency of FPSO and tanker collision in
tandem offloading cannot be obtained without considering the potential human recovery
failure, unless we directly assume that human recovery will fail. Then it is the tanker
drive-off frequency that is conservatively taken as the collision frequency. However
risk analysis carried out in this manner will miss a whole area of potential risk reduction
measures due to the incomplete modeling.
It is also true that if the human element is included, quantification becomes a very
difficult task. Even if quantification is achieved, those large human failure probability
numbers derived from subjective judgments or expert guestimations are difficult to
compare with the probability numbers derived from the technical area, not mention to
URN:NBN:no-3369
79
integrate both under some rule-of-thumb 10-4 annual frequency acceptance criterion.
These are challenges. However, quantification is not the ultimate goal in offshore QRA,
the ultimate goal is the risk reduction. Hard-to-quantify does not justify the reason to
exclude human contributions in the modeling. And risk reduction can be made effective
only when both positive and negative human contributions to the system have been
assessed. Lessons learned in other safety-critical domains like the aviation industry will
probably again pave the way for offshore QRA methodology to develop in the face of
these challenges.
URN:NBN:no-3369
R E F E R E N C E S
Addy PD, et al. Gryphon A, the first purpose built permanently moored FPSO in the
North Sea. OTC 7424, Offshore Technology Conference, Houston, 1994.
Andersen OJ. Environmental condition and ship motions during tandem offloading from
FPSO (In Norwegian). Statoil internal report, 2000. [restricted]
Bazerman M. Judgment in managerial decision making. 4th Ed., Wiley New York, 1998.
Bea RG. Human and organization errors in reliability of offshore structures. Journal of
Offshore Mechanics and Arctic Engineering, Vol. 119, February 1997.
Blom. Data Preparation XXXX FPSO & XXXX Tanker. Doc. 0214022-0200075,
Rev.01, 2002a.
Blom. Video re-construction of a collision incident between FPSO and tanker. 2002b.
Chen H. SMS-Offshore Loading Phase 2: Information from Simulator Training
Observation, 11.12.00 15.12.00. Restricted Internal Memo, Dept. of Marine
Structures, NTNU, 2000.
Chen H. FSU-Tanker Offloading Operation: Information from offshore trip. Restricted
internal memo 03-07-01, Revision 24-07-01, Dept. of Marine Structures, NTNU, 2001.
Chen H, Moan T, Haver S, and Larsen K. Predication of relative motions and
probability of contact between FPSO and shuttle tanker in tandem offloading operation.
OMAE2002-28101, Proceedings of the 21st OMAE Conference, Oslo, Norway, 2002a.
[Under review by Journal of Offshore Mechanics and Arctic Engineering]
Chen H and Moan T. Collision risk analysis of FPSO-Tanker offloading operation.
OMAE2002-28103, Proceedings of the 21st OMAE Conference, Oslo, Norway, 2002b.
Chen H and Moan T. Human intervention of tanker drive-off in tandem offloading
operation. The 2nd International Conference on Human Factors in Ship Design and
Operation, RINA HQ, London, UK, October, 2002c.
Chen H. Captain and DP Officers Questionnaire Survey Summary Report 21-05-2002.
Dept. of Marine Structures, NTNU, 2002d. [restricted]
Deaton JE and Parasuraman R. Effects of task demands and age on vigilance and
subjective workload. Proceedings of the 32nd annual meeting of the Human Factor
Society, Human Factor Society, 1988.
80
URN:NBN:no-3369
References
81
DNMI (Tnnessen H). Hindcast data for Haltenbanken area. Norwegian Meteorological
Institute, 1999.
DNV. Environmental conditions and environmental loads. Classification Note 30.5, Det
Norske Veritas, 1991.
Dougherty EM and Fragola JR. Human reliability analysis: a systems engineering
approach with nuclear power plant applications. Wiley, New York, 1988.
DPVOA. Dynamic Positioning Systems Incidents, 1980-1992. Reference listed in the
HSE report OTO 97055, 1994.
Embrey DE. Incorporating management and organizational factors into probabilistic
safety assessment. Reliability Engineering and System Safety, Vol. 38, pp. 199-208,
1992.
Faber MH, Kroon IB, Kragh E, Bayly D, Decosemaeker, P. Risk assessment of
decommissioning options using Bayesian Networks. Proceedings of the 20th OMAE
Conference, Rio de Janeiro, Brazil, 2001.
Fucatu CH, Nishimoto K, Maeda H, and Masetti IQ. The shadow effect on the
dynamics of a shuttle tanker. OMAE2001/OFT-1132, Proceedings of the 20th OMAE
Conference, Rio de Janeiro, Brazil, 2001.
Fujita Y. Human reliability analysis: a human point of view. Reliability Engineering
and System Safety, 38, pp71-79, 1992.
Furnes O and Amdahl J. Computer simulation study of offshore collisions and analysis
of ship-platform impacts. Applied Ocean Research, Vol. 2, No. 3, 1980.
Fylling I, Stansberg CT and Mo K. Extreme motions and anchor line loads in turret
mooring systems. Vol. 1, BOSS92, 1992.
Gudmestad K and Aanonsen G. (Kongsberg Simrad). Private communication regarding
shuttle tanker thruster, main propeller and rudder information. Sept.-Nov., 2001
Gudmestad K. (Kongsberg Simrad). Private communication regarding information of
shuttle tanker propulsion. 2001
Gudmestad OT. Risk assessment tools for use during fabrication of offshore structures
and in marine operations projects. OMAE2001/S&R-2106, Proceedings of the 20th
OMAE Conference, Rio de Janeiro, Brazil, 2001.
Hannaman GW and Worledge DH. Some developments in human reliability analysis
approaches and tools. Reliability Engineering and System safety, 22, 235-256, 1988.
Hals T. Customer acceptance test procedure - Tandem Loading. Kongsberg Simrad,
2000. [restricted]
URN:NBN:no-3369
82
Reference
URN:NBN:no-3369
References
83
URN:NBN:no-3369
84
Reference
URN:NBN:no-3369
References
85
URN:NBN:no-3369
86
Reference
Wickens CD and Hollands JG. Engineering Psychology and Human Performance. 3rd
Ed., ISBN 0-321-04711-7, Prentice-Hall Inc., 2000.
URN:NBN:no-3369
A P P E N D I X
A. THEORETICAL BACKGROUND
This appendix provides the rationale behind the collision frequency model presented in
Chapter 2. The tandem offloading operation is carried out through a joint humanmachine system where human supervisory control is involved. The basics of the human
supervisory control and the dynamics of the human-machine system are discussed. The
operational safety modeling (OSM) concept is then proposed based on the adapted
offshore quantitative risk analysis methodology. The constructed collision frequency
model is an application of this OSM concept. In addition, this appendix also includes
the basics of human actions and human errors, since these are bottom blocks for
evaluating human reliability in a human-machine system. Theoretical evidences and
operational facts make it clear that the modeling in Chapter 3 and Chapter 5 should
focus on human actions instead of various types of human errors. The last part of
this appendix deals with human information processing models. These can be viewed as
support to the proposed model for DP operator action initiation in Chapter 5.
A.1
The FPSO and shuttle tanker tandem offloading operation is carried out through a joint
human-machine system. For those who are new to the tandem offloading operation,
they are suggested to read the Appendix D first. Operational safety analysis of tandem
offloading in general, and modeling of the collision between FPSO and tanker in the
operation in particular, require that the basics of this joint human-machine system are
adequately understood. I therefore start with the human supervisory control concept
which is involved in the human-machine system in the tandem offloading context.
Afterwards the dynamics of human-machine system are discussed, and their
implications, combined with the adapted offshore quantitative risk analysis approach,
provide the basic formulation of the collision frequency model presented in Chapter 2.
A.1.1
87
URN:NBN:no-3369
88
Appendix A
In manual control, operators employ direct sensing and manipulation. Computers may
be an aid in either or both sensing and acting. However, all control decisions depend on
the human operator. In supervisory control, a minor or a major fraction of control is
accomplished by control loops closed directly through the computer and exclusive of
the human. Specifically in tandem offloading operation, the third figure from the left
may be applicable to the Approach, Connection, Disconnection and Departure phases,
while the fourth one is applicable to the Loading phase. In full automatic control, the
human operator can observe but cannot influence the control process.
H
I
S
T
I
S
URN:NBN:no-3369
Theoretical Background
89
knowledge-based reasoning which can be used to cope with system emergencies, and/or
to do tasks which designers do not know how to automate.
There are several difficulties at the heart of human supervisory control. For example,
operators are required to monitor that the system functions properly. It is well known
that even highly motivated operators cannot maintain effective vigilance for anything
more than quite short periods, thus they are actually ill-suited to carry out the task of
monitoring for rare and abnormal events (Reason, 1990). Another example is regarding
one of tasks that justify operators existence in the system, i.e. operator is supposed to
take over manual control when the automatic control system fails. Manual control is a
highly skilled activity, and skills must be practiced continuously to be upheld. However,
an automatic control system that fails only rarely denies operators the opportunity for
practicing these basic control skills. Moreover, when manual takeover is necessary,
something usually has gone wrong. This means that operators need to be more rather
than less skilled in order to cope with these abnormal conditions (Reason, 1990).
A.1.2
Operational safety analyses of tandem offloading require that the dynamics of this joint
human-machine system in the operation be adequately understood. The following
discussions based on Hollnagel (1996) are considered applicable to the present study.
If it can be assumed that the joint system will perform as expected and remain
within the envelope of design performance, then there may be no need to go
beyond a technologically based description. The operators act as regulators
(although as more complex regulators than those which can be designed), but
since their performance remains within the prescribed band, there is no need to
model them differently from how a technical regulator is modeled.
However, it is likely that the system will encounter situations where human
performance begins to play a separate role, then there is a need to understand it
in its own term. This is the case both when human actions are seen as the root
cause(s) or triggering causes of unwanted system events, and when human
actions are seen as the events that save the system.
The above human-machine system dynamics implies that there are two main issues to
be addressed in the operational safety modeling in general.
1. To predict what could possibly go wrong so that the joint human-machine
system will perform outside the design envelope. All types of causes, e.g.,
operation actions, technical system failures, measurement and instrumentation
failures, and so on, have to be included and evaluated.
2. To assess the systems recovery ability when outside the design performance
envelope. The human operator plays a separate role here, and both the technical
system and human operator have to be taken into account when assessing the
systems recovery ability.
URN:NBN:no-3369
90
A.1.3
Appendix A
The operational safety of human-machine systems with human supervisory control may
be modeled based on the adapted offshore quantitative risk analysis approach. The
proposed operational safety modeling (OSM) is formulated in the light of the humanmachine system dynamics described above. The analytical elements in OSM are
schematically shown as a bow-tie-diagram in Figure A - 2. This figure is adapted from
Vinnem (1999).
In the offshore quantitative risk analysis, the starting point is the identification of
initiating events. This involves a broad review of possible hazards and sources of
accidents to ensure that no relevant hazards are overlooked. The cause analysis (left) is
to identify causes that may lead to the initiating events and quantify probabilities of the
initiating events. The fault tree analysis is a typical tool. The consequence analysis
(right) comprises modeling of accident sequences and analyzing the consequences, in
terms of personnel fatality, damage to environment, and assets loss. The even tree
analysis is often used.
Quantitative Risk Analysis
CONSEQUENCE
ANALYSIS
CAUSE ANALYSIS
Initiating Events
SCENARIOS
Designed
Performance Limit
INITIATING STAGE
CAUSE ANALYSIS
RECOVERY STAGE
RECOVERY ANALYSIS
URN:NBN:no-3369
Theoretical Background
91
There are tons of books and papers on human reliability and related areas. We may also
safely say that each book or paper dealing with the human reliability has its original
context, in terms of technical system, operator, organization, etc. It is not my objective
to write a review of human reliability that covers every (or most) major issue under
this heading. Readers are directed to Ch.1 and Ch.4 in Kirwan (1994) and Ch.1 and
Ch.2 in Hollnagel (1998) for a quick grasp of generic theories and principles regarding
human reliability. In this study I have a unique context to consider, i.e. shuttle tanker
tandem offloading from FPSO.
The collision frequency model presented in Chapter 2 has directed our attention to the
scenario of tanker powered forward movement (drive-off). On the left side, human
operators are involved in the initiation of tanker drive-off. On the right side, there is
the recovery ability of the joint human-machine system in which the human operator
plays an indispensable role. The human reliability issues addressed in this section are
closely related to the two areas of interest outlined above. I focus on the basics of
human actions and human errors, since these constitute the fundament for evaluating
human reliability in a human-machine system. Theoretical evidences in this section
guide the operational safety modeling in Chapter 3 and Chapter 5 to three types of
human actions, i.e. initiating action, response action, and latent action, rather than to
various types of human errors.
A.2.1
Human action
URN:NBN:no-3369
92
Appendix A
Rule-Based
Behavior
Skill-Based
Behavior
Identification
(system state)
Decision
Task Planning
Observation
Stored
Procedure
Detection
Execution
Sensory Input
Actions
Figure A - 3 Skill, Rule, and Knowledge based behavior (from Rasmussen et al.,
1981)
A.2.2
Human error
There are many definitions and classifications of human error. Each reflects its own
practical concern and theoretical orientation. The contents of this section are mainly
from Reason (1990) due to its psychological merit and the general practicability.
The term human error is defined by Reason (1990) as follows:
Errors will be taken as a generic term to encompass all those occasions in which
a planned sequence of mental or physical activities fails to achieve its intended
outcome, and when these failures cannot be attributed to the intervention of
some agency. (Page 9).
Reason emphasizes that the notions of intention and error are inseparable. Human action
can be categorized as intentional action and nonintentional action. Reason argues that
human error is only associated with the intentional action, and it has no psychological
meaning in relation to nonintentional behavior. This view is also accepted here;
although, from safety point of view, nonintentional human behavior may contribute to
system failure too.
URN:NBN:no-3369
Theoretical Background
93
Human reliability
URN:NBN:no-3369
94
Appendix A
the reliability of the whole system. These reasons lead to the following scope of human
reliability analyses in this study.
Analyses of human reliability in this study consist of modeling and analyzing human
contributions to the initiation of tanker drive-off and to the recovery operation if tanker
drive-off happens. In the early stages of this study, the modeling attempts were almost
solely focused on human errors and violation. The assumption was that the interested
human contributions would solely be from these two categories, as defined in Section
A.2.2. After all, it seems to be generally accepted that about 80% of offshore accidents
are due to human error, and 80% of those occur during operations. However, there are
different interpretations of what have been included under the label of human error in
that assertion. Human errors, i.e., slips, lapses, and mistakes, and of course violations,
do contribute to the initiation and recovery of tanker drive-off. However, other human
actions may also contribute.
The following is a real example (also briefly described in Chapter 3) showing how a
human action (neither error nor violation type) may contribute to the initiation of tanker
drive-off. During offloading, the DARPS (Differential Absolute & Relative Position
System) unit generated an erroneous signal due to interference. Afterwards the DARPS
signal was de-selected automatically from DP. Based on procedure1, the operator reselected the DARPS signal into DP when the signal was observed normal. This reselection action had been done a few times already since DARPS interference had
occurred a few times. However, during this re-selection process the interference
occurred again, and the erroneous position signal, due to operators re-selection, was
accepted by DP, and subsequently the tanker was driven forward based on the wrong
distance calculated by DP. In this event, the operators action went on as intended, and
the goal, i.e. re-selection of the DARPS signal into DP, was achieved too. There was no
human error involved, and the action was according to the procedure too, i.e., no
violation either. However, we do see that the operators action contributed to the
initiation of tanker drive-off.
The above case exemplifies what Murphy and Pate-Cornell (1996) have commented on;
i.e., that there are gray areas in which the distinction between error and appropriate
action is unclear in a complex system, when for instance goals conflict or the operator
lacks information. We may also note the dynamic nature of such complex systems,
which causes the same action to be at one time appropriate while at other times
initiating a drive-off event.
It is clear that focusing on modeling of human error and violation in this study may
potentially rule out other possible human contributions. This is a modeling challenge.
As an initial step, human actions, regardless of whether they are errors, violations, or
gray actions, are considered in the modeling. Given the nature of the human supervisory
control involved, and the dynamics of the considered human-machine system as
The operational procedure requires that minimum three position reference signals during tandem
offloading.
URN:NBN:no-3369
Theoretical Background
95
described in Section A.1, the human actions are generalized into the following three
types in this study:
1. Initiating action An action initiates a failure event.
2. Response action An action is a response to system demands, typically during
technical failure events or special external situations. It may save or worsen the
situation or cause a transition to another event.
3. Latent action An action influences the status (failure probability) of technical
system components, e.g. maintenance action, and potentially interacts with the
above two types of human actions.
The resulting probabilistic models of tanker drive-off are presented in Chapter 3.
Assessing and reducing human contributions are carried out in the light of the failure
prone situation concept.
A.3
The objective is to reasonably model the reaction from DP operator during the tanker
drive-off scenario. To achieve this objective, lessons learned from operator models
developed in other similar context should be taken into account. This is the motivation
to include this section in the theory appendix.
The most pervasive model of humans in human-machine systems is probably the
Stimulus-Organism-Response paradigm. The human operator is located in the middle,
as Organism, receiving Stimulus, and performing Response. This view is largely
accepted in the context in this study. As Dougherty (1993) stated, the concern is how to
handle the immense richness of the O in S-O-R paradigm that makes human versus
machine performance so interesting. A few operator models based on the S-O-R model
in human reliability studies are briefly outlined.
Taking a closer look at the O in S-O-R, the human can be viewed as an Information
Processing System. According to this view, human cognitive processes could be broken
down to several (not necessarily sequential) procedures and mental states which are
linked by their causal relations with sensory input, muscle behavior, and other mental
states. This view is considered valid in the context in this study. Two typical examples,
i.e. the step-ladder model for process plant operators by Rasmussen (1986) and
Wickens model for aviation pilot by Wickens and Flach (1988), are briefly presented
below. These models form the foundation for the proposed Information-DecisionExecution model for tanker DP operator in drive-off scenario in Chapter 5.
A.3.1
S-O-R models
In early 1980s, the SHARP (Hannamann & Spurgin, 1984) approach suggested a
decomposition of human actions into: observation, diagnosis, and manual actions, after
URN:NBN:no-3369
96
Appendix A
an initiating event. This type of analysis later became common in many human
reliability methods.
Nagel (1988), in his safety study of human error in aviation operations, proposed a
three-stage human action model: information, decision, and action. He argued that most
purposive, skilled behavior in a somewhat constrained environment like the airplane
cockpit can reasonably be described in terms of the simple 3-stage model of behavior as
follows.
-
Decision stage: decisions are made and specific intents or plans to act are
determined.
Smidts et al. (1997) set up a cognitive model (IDA) for the analysis of the nuclear
power plant operator response under accident conditions. The model of the single
operator consists of three major components:
-
Information Module
Action Module
The linear fashion of some of the models above cannot account for the shortcuts that
human decision makers take in a real-life situation. Instead of a straight-line sequence of
stages, Rasmussen proposed a model analogous to a step-ladder. This is discussed in the
following subsection.
A.3.2
Step-ladder model
For control of a physical system such as in industrial process plant, Rasmussen (1986)
proposed a human information processing model which includes the following eight
URN:NBN:no-3369
Theoretical Background
97
URN:NBN:no-3369
98
A.3.3
Appendix A
Wickens model
Wickens and Flach (1988) proposed the following information processing model for
aviation pilots. It consists of the following main processes: sensory store, pattern
recognition, decision and response selection, and response execution, as presented in
Figure A - 5.
Figure A - 5 Wickens human information processing model (Wickens & Flach, 1988)
The first stage in the model is the sensory store. In sensory store, physical energy is
transformed into neural energy. Information is represented in the sensory store in terms
of physical features.
The second stage, pattern recognition, is probably the most important yet least
understood of all the stages. It is at this stage that the physical stimulation in the sensory
stores is integrated into meaningful elements. This pattern recognition process involves
mapping the physical codes of the sensory store into meaningful codes from memory.
This mapping is very complex in that many different physical codes may all map to a
single memory code and a single physical code may map to different memory codes.
Perceptual processes are often limited by the supply of attention resources.
The next stage is the decision and response selection stage. At this stage, a stimulus has
been recognized and a decision must be made as to what to do with it. A number of
options are available at this point. The information can be stored for future use, it can be
integrated with other available information, or it may initiate a response. Each of these
options will generally be associated with potential costs and benefits which must be
considered when choosing among them.
URN:NBN:no-3369
Theoretical Background
99
The last stage, response execution, interprets what may be a generally specified
intention into precisely sequenced muscle commands. The resulting responses, by way
of the feedback loop, then become input to the sensory stores, which can be interpreted
and entered as data relevant to selecting the next response.
In addition to the processing stages, the human action model contains three ways to
store information: sensory store (mentioned earlier), working memory, and long-term
memory. Working memory represents the information currently being used by the
information processor. Long-term memory represents information available to the
information processor, but not currently in use. Long-term memory is the storehouse of
all accumulated knowledge.
A.4
REFERENCES
URN:NBN:no-3369
100
Appendix A
URN:NBN:no-3369
A P P E N D I X
B . I N C I D E N T A N A LY S I S
Nine tanker drive-off events are analyzed and results are provided in this appendix.
Relevant information of this incident study is given in Chapter 3 (Section 3.2) in the
main report.
B.1
COLLISION INCIDENT A
Weather
Relative
Motion
(FPSO & ST)
Thruster
Main
Propeller
(FPSO & ST)
PRS &
Sensors
(FPSO & ST)
Tanker DP
Operator
No action
Action
worsen
situation
Action mitigate
situation
Collision
Too late
Time since
drive-off
In time
Emergency stop of ST succeeded, no contact, but FSU then surged back and
collision happened.
ST operator was not able to decide to
rotate ship to either port or starboard
due to FSU fishtailing.
101
URN:NBN:no-3369
Tanker DP
System
Speed at
contact
Near miss
102
Appendix B
B.2
COLLISION INCIDENT B
Weather
Relative
Motion
(FPSO & ST)
PRS &
Sensors
(FPSO & ST)
Tanker DP
System
Tanker DP
Operator
More than 30q heading difference developed between tanker and FPSO under light
wind and strong spring tides condition.
Operator selected Manual DP with surge and yaw controlled. He used joystick to
steer tanker in sway to align it with FPSO.
DP responded by put ahead pitch to get enough turning moment from rudder
(sideway thruster capacity is not enough), this drove tanker ahead.
Outcome: P (collision | action & PFM)
No action
URN:NBN:no-3369
Thruster
Main
Propeller
(FPSO & ST)
Action
worsen
situation
Action mitigate
situation
Collision
Too late
Time since
drive-off
In time
Speed at
contact
Near miss
Incident Analysis
103
B.3
COLLISION INCIDENT C
Weather
Relative
Motion
(FPSO & ST)
PRS &
Sensors
(FPSO & ST)
Tanker DP
System
Tanker DP
Operator
DP operator re-select DARPS into DP, DP accept wrong position data and
calculated wrong distance.
URN:NBN:no-3369
Thruster
Main
Propeller
(FPSO & ST)
Action
worsen
situation
Action mitigate
situation
Collision
Too late
Time since
drive-off
In time
DP operator selected DP auto heading and used joystick full astern, approx. 1.5 min later after position warning.
Speed at
contact
Near miss
104
Appendix B
B.4
COLLISION INCIDENT D
Weather
Relative
Motion
(FPSO & ST)
PRS &
Sensors
(FPSO & ST)
Tanker DP
System
Tanker DP
Operator
Tanker changed heading to follow FPSO, however due to limited thruster power
and different weathervaning capability, it had difficulties positioning itself relative
to FPSO. Therefore tanker DP operator took manual control by using DP Manual
Bias function to position tanker heading, this action prioritized heading over
distance control.
DP responded by put 40 % forward pitch to get a turning moment from rudder, this
drove tanker ahead.
Outcome: P (collision | action & PFM)
No action
URN:NBN:no-3369
Thruster
Main
Propeller
(FPSO & ST)
Action
worsen
situation
Action mitigate
situation
Collision
Too late
Time since
drive-off
In time
Speed at
contact
Near miss
Incident Analysis
105
B.5
COLLISION INCIDENT G
Weather
Relative
Motion
(FPSO & ST)
PRS &
Sensors
(FPSO & ST)
Tanker DP
System
Tanker DP
Operator
No action
URN:NBN:no-3369
Thruster
Main
Propeller
(FPSO & ST)
Action
worsen
situation
Action mitigate
situation
Collision
Too late
Time since
drive-off
In time
Speed at
contact
Near miss
106
Appendix B
B.6
NEAR MISS E
Weather
Relative
Motion
(FPSO & ST)
PRS &
Sensors
(FPSO & ST)
Tanker DP
System
Tanker DP
Operator
No action
URN:NBN:no-3369
Thruster
Main
Propeller
(FPSO & ST)
Action
worsen
situation
Action mitigate
situation
Collision
Too late
Time since
drive-off
In time
Speed at
contact
Near miss
Incident Analysis
107
B.7
NEAR MISS F
Weather
Relative
Motion
(FPSO & ST)
PRS &
Sensors
(FPSO & ST)
Thruster
Main
Propeller
(FPSO & ST)
Tanker DP
System
Tanker DP
Operator
DP operator re-selected the main propeller (which had been restarted) into DP.
DP then demanded 100 % pitch forward from the main propeller. This drove
tanker ahead.
Outcome: P (collision | action & PFM)
No action
Action
worsen
situation
Action mitigate
situation
Collision
Too late
Time since
drive-off
In time
Speed at
contact
Near miss
1
The estimate comes from DP alarm prints combined with the event description in the investigation
report. The drive-off was initiated after reselection of failed thruster No.4 at around 17:52:40. At
17:53:21, i.e. 41 seconds later, there was an alarm indicating high thruster force. Operator took action
seconds after this alarm, and this gives the 45 s evidence.
2
FPSO control was informed that the tanker had a problem but that it had been solved, at 17:55:00.
Tanker started to pull back at this time. These facts indicate the 140 s stoppage.
URN:NBN:no-3369
108
Appendix B
B.8
NEAR MISS H
Weather
Relative
Motion
(FPSO & ST)
PRS &
Sensors
(FPSO & ST)
No action
URN:NBN:no-3369
Tanker DP
System
Tanker DP
Operator
Thruster
Main
Propeller
(FPSO & ST)
Action
worsen
situation
Action mitigate
situation
Too late
In time
Collision
Time since
drive-off
Speed at
contact
Near miss
Ship stopped at 45 m
Incident Analysis
109
B.9
NEAR MISS I
Weather
Relative
Motion
(FPSO & ST)
PRS &
Sensors
(FPSO & ST)
Thruster
Main
Propeller
(FPSO & ST)
Tanker DP
System
Tanker DP
Operator
FSU heading function was deselected by the DP. DP operator reselected the
heading signal into DP, during which a possible interference occurred again2.
DP accepted erratic FSU heading data, calculated wrong relative position between
FSU and drove tanker ahead.
Outcome: P (collision | action & PFM)
No action
Action
worsen
situation
Action mitigate
situation
Collision
Too late
Time since
drive-off
In time
Speed at
contact
Near miss
Currently there are only 4 frequencies available for use with all 44 DARPS units installed throughout
the North Sea area. With relative short distance between some of the installations, interference between
DARPS units may happen. This interference problem may be solved by designing equipment that is free
of co-channel interference. It is current under development by Kongsberg Seatex.
2
Note that there was a repeated de-selection and re-selection process between DP and DP operator, due to
that DARPS interference continually happened, prior to drive-off.
URN:NBN:no-3369
A P P E N D I X
C . S I M U L AT I O N M O D E L
C A L I B R AT I O N
The objective is to calibrate the simulation models of FPSO and tanker so that the
simulated horizontal motions of these two vessels in SIMO can agree reasonably with
the physical phenomena. The approach starts from validating every group of vessel
input information in SIMO based on model test data, design information, and other
empirical data if relevant. Afterwards, given the similar environmental conditions,
positioning system (DP and mooring) and vessel (linear, quadratic) damping in surge,
sway and yaw are tuned, and motions of the joint FPSO-tanker system are simulated
and compared to the full-scale measurements. The comparison criteria are statistical
values of surge, sway and yaw motions. After tuning the simulation models, reasonable
matches between simulation and measurement should be observed. The model
calibration work is detailed in Section C.1. Background and pre-processing of the fullscale motion measurements of FPSO-tanker during tandem offloading in the North Sea
in winter 2001-2002 are provided in Section C.2. In addition, the basics of time-domain
motion simulation theory used in SIMO are briefly outlined in Section C.3.
C.1
The simulation model calibration work starts from a passive FPSO model as described
in Section C.1.1. Relevant model test results and empirical data are used. Then a DP
FPSO model is calibrated in Section C.1.2 based on full-scale measurements.
Calibrations of the tanker model and the joint FPSO-tanker model take into account of
qualitative operational information and quantitative full-scale measurement. These are
presented in Section C.1.3 and Section C.1.4.
Note that the precise match between simulation and full-scale measurement is close to a
mission-impossible task. We have to be aware of the idealizations contained in the
simulation program itself. In addition, significant amount of information is not available
or not possible to obtain in the full-scale measurement, e.g. vessel loading condition,
current information, onboard usage of thrusters, to list a few. Accordingly, idealizations
and assumptions are introduced into the simulation to account for these unknowns.
However, it is still considered possible to judge the reasonability of the simulation
model via comparing statistical values of motions and behavior of motion trace plots to
the measurements. After all, the simulation model is to be used for the surging and
110
URN:NBN:no-3369
111
The FPSO turret mooring system model in SIMO is setup based on the design
information from the model test report (Marintek, 1994). This gives out a reasonable
natural period in surge compared to the full-scale measurement (Andersen, 2000).
The passive FPSO surge damping is considered mainly from linear damping provided
by turret mooring system. 15 % of critical surge damping is estimated which is roughly
1500 kN/(m/s). The passive FPSO yaw damping is considered mainly from quadratic
damping provided by vessel hull. Based on the quadratic current coefficient, a simple
estimation of quadratic yaw damping is made, and it is calculated as 8.57E+08
kNm/(rad/s)2.
Based on the environmental conditions in Chapter 4 (Section 4.5), the simulated 3-hour
FPSO stern motion trace at hawser connection point is plotted with a 20-minute fullscale measurement in Figure C - 1. Due to normalization according to the mean stern
point in the plot, the mean position is not reflected in the figure. Statistical values of
motion are presented in Table C - 1. Clearly we observe relatively larger surge and yaw
motions. This is because thrusters were used onboard to reduce the motion amplitudes
when measurements were taken, i.e. the 20-minute full-scale measurements were
actually from a DP FPSO.
Figure C - 1
URN:NBN:no-3369
112
Appendix C
From the model test information which extrapolated to the full-scale (Marintek, 1999), a
passive FPSO under a similar offloading environmental (Hs = 4.5 m) condition has a
yaw motion standard deviation of 2.39. In our simulation, the standard deviation of yaw
motion is 2.24. This may roughly confirm that the yaw motion simulated in the passive
FPSO model is reasonable. However, there is no other information available for further
validation of the yaw motion, nor the surge motion.
Mean
Surge (m)
Yaw (q)
Table C - 1
C.1.2
Std.
Max.
Min.
-3.40
1.36
3.68
-5.03
-1.81
2.24
5.34
-6.32
Passive FPSO motion statistics in 3-hour simulation
DP FPSO model calibration
A brief description of full-scale measurement is given here. The 20-minute FPSO surge
and yaw motions are measured. The environmental conditions at the same time are also
measured which have been described in Chapter 4 (Section 4.5). The motion statistical
values and natural periods are presented in Table C - 2.
Mean
Surge (m)
Yaw (q)
Table C - 2
Std.
Max.
Min.
Tn (s)
-3.31
171
0.68
1.77
-2.25
0.03
600
1.45
3.70
-2.72
FPSO motion statistics in 20-minute measurements
The stiffness, damping and integral terms in the FPSO PID controller in SIMO are
tuned, and finally the simulated FPSO surge and yaw motions give out reasonable
statistics that are close to the measurements. The 20-minute is a relatively short duration
for low frequency motions. Longer duration measurements are not available for the used
environmental conditions. To account for statistical variations, 10 simulations with 20minute duration are carried out, and we can observe a convergent match of motion
standard deviation between simulations and measurement as presented in Figure C - 2.
In summary, the FPSO model is tuned and considered reasonable for the study purpose.
The simulated 3-hour FPSO stern motions are shown in Figure C - 3. Statistical values
of motions are presented in Table C - 3.
Mean
Surge (m)
Yaw (q)
Table C - 3
URN:NBN:no-3369
Std.
Max.
Min.
-3.57
0.83
2.32
-4.53
0.09
1.35
3.66
-5.19
DP FPSO motion statistics in 3-hour simulation
URN:NBN:no-3369
Figure C - 2
Figure C - 3
113
114
C.1.3
Appendix C
Mean
Std.
Max.
Min.
FPSO
TANKER
-3.48
-3.45
0.76
1.24
-1.40
-0.83
-6.48
-9.53
Yaw (q)
Mean
Std.
Max.
Min.
FPSO
TANKER
Table C - 4
0.03
1.27
3.42
-3.67
-0.02
1.15
3.40
-3.49
FPSO-Tanker motion statistics in 3-hour simulation
After full-scale FPSO and tanker motion data in tandem offloading and the
corresponding environmental conditions were received in the spring of 2002 (these data
are presented in Section C.2), the joint FPSO-tanker model was further calibrated based
on the quantitative criteria from the measurements. The following is a calibration
example based on the 2-hour motion measurements taken on 30 January 2002, which is
described as Case A in Section C.2 in the following.
The base case FPSO and tanker configurations as described in Chapter 4 are assumed.
Similar environmental conditions are imposed in the simulation as in the 2-hour
measurements. The simulation results and measurements are compared via the statistical
values of surge and yaw motions presented in Table C - 5, as well as the FPSO stern and
tanker bow motion trace plot presented in Figure C - 5. Reasonable matches are
observed between simulation results and measurements. For completeness, simulated
time series of surge and yaw motions for both vessels are presented in Figure C - 6.
URN:NBN:no-3369
115
In summary, the conclusion after the above calibration work is that the present FPSOTanker model is able to reasonably simulate the physical horizontal motions between
FPSO and tanker in tandem offloading.
Figure C - 4
Mean
Std.
Max.
Min.
Measurement
Simulation
NA
-2.13
1.06
1.67
3.68
4.77
-5.72
-5.63
Mean
Std.
Max.
Min.
Measurement
Simulation
0.00
0.07
0.59
0.76
1.90
2.55
-1.50
-2.26
Mean
Std.
Max.
Min.
Measurement
Simulation
NA
-2.51
1.50
1.37
3.85
4.46
-7.91
-5.59
Mean
Std.
Max.
Min.
Measurement
6.00
1.77
11.70
3.20
Simulation
4.70
1.26
8.13
1.65
Table C - 5 FPSO-Tanker 2-hour motion statistics (Simulation vs. Measurement)
URN:NBN:no-3369
116
URN:NBN:no-3369
Appendix C
Figure C - 5
Figure C - 6
C.2
C.2.1
117
FPSO and tanker motion measurement data are partly provided by BLOM A/S (Blom,
2002a). The data are obtained from a North Sea DP2 shuttle tanker during loading from
an FPSO in winter 2001-2002. There are in total 9 episodes as listed in Table C - 6.
Each of them lasts for three hours. The sampling frequency is the best that BLOM PMS
can provide, and it varies roughly between 1.00 0.25 Hz.
Episode
1
2
3
4
5
6
7
8
9
Table C - 6
Date
2001-11-21
2001-11-21
2001-12-26
2001-12-27
2002-01-22
2002-01-24
2002-01-30
2002-02-01
2002-02-09
Time and date for all episodes
Time
15 18
18 21
12 15
00 03
12 15
05 08
00 03
15 18
01 04
The instantaneous Northing and Easting coordinates of FPSO stern reference point
(Artemis station in this case) can be derived based on Distance, Bearing and Tanker
bow position in BLOM data. The Artemis station is located at: longitudinally 160.93 m
from turret center, 0.98 m from middle to the port side.
The FPSO heading values in the BLOM data are however, do not reflect the
instantaneous yaw motion of the FPSO. It looks as if the FPSO has a constant heading
plus many sudden jumps in 3-hour duration. This is not physically true. The
constant/jumping behavior is probably due to the fact that FPSO heading values were
taken from the DARPS unit on tanker, and only some kind of averaged FPSO heading
values were recorded. Therefore FPSO heading together with the heading difference
values in the BLOM data are excluded. The instantaneous FPSO heading data are
obtained from STATOIL (Andersen, 2002).
URN:NBN:no-3369
118
Appendix C
The environmental data in each of the above measurement are provided by STATOIL
(Haver, 2002). The data consist of 20-minute (or in some episodes 40-minute) mean
values of the following items:
-
C.2.2
There are 7 time series that are selected from these 9 episodes as qualified for use in the
calibration of the joint FPSO-tanker model in SIMO. They are listed in Table C - 7.
Note there are basically three types of cases marked with A, B, and C, corresponding to
Hs between 4-5 m, 3-4 m, and 2-3 m, respectively. To further elaborate FPSO and
tanker motion behavior in class A and B, sensitivity cases with shorter duration are
identified.
Case No.
A
A1
A2
B
B1
B2
C
Table C - 7
Duration
Hs (mean)
Uw (mean)
9.8 m/s
9.5 m/s
6.4 m/s
6.6 m/s
7.9 m/s
10 m/s
17 m/s
The criteria for selecting these 7 time series are as follows. First, there should not exist
any apparent measurement errors such as sudden incredible number or miss of data.
(Note that the instantaneous FPSO heading data in case A2 and B2 are partly lacking,
however, these two cases are still considered representative to illustrate tanker motions,
and therefore, they are included.) Second, the FPSO and tanker in simulation are
keeping their mean heading, i.e. no operated heading change in simulation. Therefore,
in measurement, there should not be any dramatic mean heading change on either FPSO
or tanker. Third, only stable weather can be simulated, so in measurement, especially
the significant wave height Hs and wind speed Uw, there should ideally be as less
variation as possible. Fourth, the qualified low frequency motion oscillations should be
kept as many as possible so that statistical uncertainty can be minimized. This means we
have to keep as long as possible the duration for the qualified episode.
C.2.3
URN:NBN:no-3369
119
coordinate is defined based on FPSO mean heading and mean stern point as stated
below. Tanker heading and environmental parameters directions are accordingly
presented clockwise relative to the mean FPSO heading as positive, instead of relative
to the North.
-
Longitudinal (surge) X: pointing along FPSO mean heading from stern to bow.
Since the original measurements are presented in earth coordinate, i.e. Northing, Easting
and heading clockwise relative to the North, measurement data are converted into the
above coordinate system. As an example, the converted measurement data in Case A
are plotted in Figure C - 8 together with the environmental parameters that are used in
the simulation. Wave and wind data are mainly from the measurement, while current
data are based on assumption.
FPSO mean
Heading
F
P
S
O
FPSO mean
Heading
X
Tanker mean
Heading
direction
definition
Wave, Wind or
Current
FPSO mean
Stern Point
S
T
Figure C - 7
URN:NBN:no-3369
120
Appendix C
Figure C - 8
C.3
C.3.1
THEORY
Method overview
The equation of motion for one or several bodies in general may be written as:
x Cx D1 x D2 f x Kx q t , x, x
M
(C-1)
URN:NBN:no-3369
M
xHF C D1 x HF KxHF
121
q (1)
M
xLF D1 xLF D2 f xLF KxLF
q (2)
(C-2)
The mass term contains body mass matrix and added mass (at zero frequency) matrix.
The body mass matrix may be found from design information. The added mass matrix is
calculated by WADAM, and is given as input to SIMO.
The damping term includes linear, quadratic and wave drift damping matrices. Linear
and quadratic damping values are given as input and tuned in the model calibration
process. The wave-drift damping matrix is however not included in the model, since the
environmental condition used in simulation of offloading operation in general has the
significant wave height smaller than 5.5 m. Wave-drift damping contribution is believed
to be small.
The stiffness term includes hydrostatic stiffness, and external stiffness provided by
mooring system and DP system. There is no hydrostatic stiffness contribution to the
horizontal motions. Mooring and DP stiffness contributions are addressed as stiffness
force in the section of mooring and DP force models.
Overall, the excitation force vector may have the following components, see Eq.C-3.
q1wa is the first order wave excitation force. q 2wa is the second order wave excitation
force. Higher order wave excitation, i.e. ringing force is not considered. q wi is the wind
force. q cu is the current force. q ext is the external excitation force, including
URN:NBN:no-3369
122
Appendix C
contributions from mooring system, DP (acting via thruster) and coupling forces
between two vessels.
q q (1) q (2)
q (1)
q1wa
(2)
2
wa
(C-3)
q wi q cu q ext
The HF excitation force, i.e. the first order wave excitation force, q1wa , is obtained in
frequency domain by multiplication of linear transfer function H (1) Z and wave
amplitude ] Z in 6 degrees of freedom. The needed transfer functions are given as
input for a number of frequencies and directions.
Modeling of the remaining LF excitation forces is discussed in the following sections.
C.3.3
URN:NBN:no-3369
Wind force
123
Current force
The current is modeled by a profile with specified direction and speed at different
levels. Linear interpolation is used between the levels.
The current viscous force/moment on hull in surge, sway and yaw are calculated by
multiplication of the current coefficients Ccu D and square of the instantaneous value
of the relative velocity between body and current. Note the body has low frequency
velocity and this LF velocity is included in the model.
The above model does not account for the effect from the yaw-induced cross-flow. This
cross flow is included as a separate quadratic yaw damping estimated from current
coefficient.
C.3.6
Station-keeping forces
The external excitation force q ext includes station-keeping forces and coupling force.
The station-keeping forces are discussed here, which has contributions from mooring
and DP.
The turret mooring system provides FPSO with stiffness in surge, sway and yaw. The
reason that yaw stiffness is provided is based on the fact that in normal operation, FPSO
has its turret locked. The rotation of vessel (though in very small magnitudes) will then
cause rotation of the turret where restoring moment is provided by twisted mooring
lines (and risers). The total stiffness contribution provided by the turret mooring system
comes from the sum from each mooring line. Each mooring line is modeled by a
catenary equation. The procedure for calculating mooring line configuration is based on
a shooting method (Lie, 1990). Mooring line dynamic tension is calculated based on
the model developed by Larsen and Sandvik (1990). Further details are not included
here since in motion simulation the tension of individual mooring line is not of interest.
URN:NBN:no-3369
124
Appendix C
The dynamic positioning system is modeled as a control module with the input from
position measurement and thrust measurement. The control module converts position
and velocity errors into a demand for thrust forces to correct the errors. The output is the
desired resultant forces and moment from thrusters.
The PID (Proportional + Integral + Derivative) controller module in SIMO is applied in
the study. A decoupling approach allows PID control parameters to be specified
separately for surge, sway, and yaw. The control law of the PID controller is presented
in Eq.C-4.
FT 0
K PH t KV H t K I H W dW
FT 0 is
H t
(C-4)
the
wanted
thrust
from
thrusters.
H t is
the
position
error,
as
velocity error, as H t
Coupling force
The coupling element is the hawser. The hawser is modeled as a non-load bearing
spring in the study. It connects the FPSO stern mooring point and shuttle tanker bow.
URN:NBN:no-3369
125
REFERENCES
Andersen OJ. FPSO motions in offloading. Report PTT MA MKS, (in Norwegian),
Statoil, 2000.
Andersen OJ. E-mail attached ASCII files with FPSO heading, surge and sway
measurement data. 22-05-2002.
Blom AS. Data Preparation XXXX FPSO & XXXX Tanker. Doc. 0214022-0200075,
Rev.01, 2002.
Gudmestad K. (Kongsberg Simrad). Private communication regarding shuttle tanker
motion in tandem offloading. 2002.
Haver S. E-mail attached spreadsheet with environmental conditions (wind and wave
data) and ship heading measurements corresponding to BLOM data. 02-04-2002.
Larsen K and Sandvik PC. Efficient methods for the calculation of dynamic mooring
line tension. The 1st European Offshore Mechanical Symposium, Trondheim, Norway,
1990.
Lie H. MIMOSA-2 Users Documentation. MT51 F90-0358, Marintek, 1990.
Marintek. FPSO motion model test results. Report excerpts received from Statoil, 1994
& 1999.
Newman JN. Second-order, slowly-varying forces on vessels in irregular waves.
International Symposium on the Dynamics of Marine Vehicles and Structures in Waves,
University College, London, 1974.
Stansberg CT. On the use, modeling and analysis of irregular water surface waves.
Proceedings of the 8th OMAE Conference, Vol. 2, 1989.
Reinholdtsen SA and Falkenberg E. SIMO Theory / User Manual. MT51 F93-0184,
Marintek, 2000.
URN:NBN:no-3369
A P P E N D I X
Shuttle tanker bridge layout and relevant technical systems involved in tandem
offloading are described. The objective is to introduce readers to the context (i.e.
human-machine system and interface) on shuttle tanker, which have been referred to in
the main text. The tandem offloading operational process is also described based on
onboard observation. This may supplement the briefly mentioned five tandem
offloading operational phases in Chapter 1. In addition, special DP positioning features,
namely the tandem loading function in tanker DP software, are described. This may
clarify the details of how the tanker positions itself relative to the FPSO as discussed in
Chapter 4.
D.1
D.1.1
Bridge layout
The shuttle tanker bridge layout with positions of relevant equipment for tandem
offloading is sketched in Figure D - 1.
Tanker Bow Direction
WINDOW
n
p
q
DP OPERATOR
o
LOADING
OPERATOR
126
URN:NBN:no-3369
b
c
127
DP console
There are two DP computers installed onboard since this is DP2 class shuttle tanker. In
operation, one computer (left one in below figure) is selected as Master and it does the
actual positioning job. The other (right one in below figure) is selected as Slave, and it
works as backup.
Figure D - 2 DP console
D.1.3
The position reference system screens are hung above the DP console. It contains two
DARPS screens, one BLOM PMS screen and one Artemis screen from left to right
(Figure D - 3). The two DARPS(s) and Artemis form the three position reference units
used in the tandem offloading operation.
URN:NBN:no-3369
128
Appendix D
The Bow Loading System (BLS) console is located beside the DP console on the
Bridge. ESD buttons can also be found there (shown in Figure D - 5). The operation of
ESD I II is to press the first and second (from left) buttons, the third button is to start
deluge in bow loading area. The established green line is shown on the screen of bow
loading console, as indicated below. Note that the picture located in upper right was
taken during actually offloading.
URN:NBN:no-3369
129
Three video cameras are installed in the tanker bow area. Visual information is shown
on Bridge regarding: i.) the distance information between FSU and ST, ii.) hawser,
winch information, and iii.) loading hose connection information. Note that the right
pictures were taken during actual offloading.
D.2
A closer look of FPSO and DP shuttle tanker in tandem offloading is provided in Figure
D - 7.
URN:NBN:no-3369
130
Appendix D
DP Shuttle Tanker
FPSO
hawser
hose
Operational activities
Phase
Approach
phase starts
Duration:
1 h 20 min
Picture adapted from Offshore Technology website, Bow Loading System by Hitec Marine.
http://www.offshore-technology.com/contractors/floating_production/hitec_marine/hitec_marine2.html
URN:NBN:no-3369
Time / Distance
Operational activities
131
Phase
Approach
phase
continues
URN:NBN:no-3369
Connection
phase starts
Duration:
1 h 46 min
132
Appendix D
Time / Distance
Operational activities
Phase
8:36 am
ST starts loading
FSU 194q, ST 198q.
Environmental condition:
Hs 1.1 m, Current 2.5 m/s, Wind 9 kn.
Loading
phase starts
9:00 am
9:10 am
9:25 am
12:45 pm
3:30 pm
4:00 pm
6:00 pm
Duration:
11 h 24 min
Loading is stopped
Start to flush hose from FSU.
Finish
flushing hose.
7:50 pm / 75 m
Close coupler valve.
Close crude valve.
Hose is dropped.
8:00 pm
Send back hose messenger line.
Chain stopper is opened.
8:11 pm
Send back hawser, chain and messenger line.
Start DP Approach mode.
8:14 pm
100 m set as set point distance.
7:00 pm
URN:NBN:no-3369
Disconnection
phase starts
Duration:
21 min
Departure
phase starts
Duration:
11 min
D.3
133
SPECIAL DP FEATURES
The tandem loading functions in the tanker DP system have two basic functions
(Hals, 2000):
-
The first one is to make tanker not follow all FSU movements. The second one is to
ratify large heading differences between tanker and FSU.
D.3.1
The FSU Surge/Sway function enables the shuttle tanker to automatically follow a
moving FSU and keeps the shuttle tanker at a constant position relative to the FSU
hawser terminal point. This significantly reduce thruster utilization on the shuttle tanker.
Like the ordinary Weather Vane mode, the heading of the shuttle tanker is always kept
pointing towards the stern of the FSU.
The background of this function is explained as follows. The shuttle tanker has the stern
of the FSU (hawser terminal point) as base point. In ordinary Weather Vane and
Approach modes, the tanker will believe that the stern of the FSU is an earth fixed
point, like in SPM loading. However, the FSU stern will move due to surge and fishtail
movement. This implies that the DP will estimate any FSU movement as movement of
the shuttle tanker. This will lead to wrong current estimates and potentially unstable DP
positioning.
The features of this function are described below.
1. A rectangle window is defined within which the FSU stern position can move
without causing the shuttle tanker to also move. When the FSU moves outside the
border of the rectangle, the rectangle is moved to the actual FSU position and the
shuttle tanker setpoint is updated accordingly. The shuttle tanker then moves to the
new position.
2. The size of rectangle could be defined by the operator. However, the upper limits
are pre-fixed in the DP system. It is important to keep the bow of the shuttle tanker
pointing towards the FSU hawser terminal point if the FSU moves due to
weathervaning. The sway window must then be small, typically 4-8 m. The surge
window is adjusted so that normal surging of FSU does not pass the limit, typically
8-15 m.
3. The shuttle tanker uses earth fixed reference systems (DGPS) for its DP model,
while relative position reference systems (Artemis and DARPS) are used for
monitoring the FSU movement relative to the shuttle tanker. This relative position
information is used to update shuttle tanker to the wanted position. Thus, relative
and absolute systems are used together for overall position keeping.
URN:NBN:no-3369
134
D.3.2
Appendix D
This function is developed in order to ratify large heading differences between tanker
and FSU. It is implemented as a sway position control where the shuttle tanker heading
is kept pointing on the hawser terminal point of the FSU.
The background of this function is explained as follows. The rapid heading change of
the FSU is a problem in tandem loading. This problem is more noticeable if the FSU has
no heading control. Heading changes of 90q in approximately 30 minutes have been
reported during calm weather at the change of tidal stream. Fishtailing movement of 5q10q at a period of 15 minutes is also commonly experienced. Further, when the draught
of the two vessels is very different, i.e. one is full and the other is empty, the optimum
(Weather Vane) headings on FSU and tanker can be significantly different, and this
imposes significant operational difficulties.
The features of this function are described below.
1. The heading of FSU is transferred over the data-link of the DARPS system to the
shuttle tanker. When the operator defined heading difference is exceeded, the tanker
thrusters are activated in order to align itself with the FSU. The resulting force is
mainly in sway direction, but the heading is continuously adjusted to keep the bow
of the tanker pointing towards the FSU stern hawser terminal point.
2. Operator can turn this function ON/OFF, and also define the limits for activating
and stopping the sway/heading control.
D.4
REFERENCES
URN:NBN:no-3369
A P P E N D I X
E. QUESTIONNAIRE SURVEY
or
E-mail: haibo@marin.ntnu.no
135
URN:NBN:no-3369
136
Appendix E
field.
The tanker drive-off is defined in the following context as: Tanker (in DP) moves ahead or
astern from its target/wanted position in tandem loading. This is not a planned or wanted
movement.
Further note: Tanker drive-off does not mean that the collision or other serious incidents in
tandem loading will happen. Successful intervention from Captain/DP officer can save the
situation.
1. Approximately how many times of tandem loading operations (with FPSO/FSU in
the North Sea) have you been involved in?
Answer:
2. Based on your past operational experience, have you ever experienced any tanker
drive-off situation (both ahead and astern) in the tandem loading operation?
A: No
Answer:
3. If your have experienced tanker drive-off situation(s) in tandem loading, what was
your position when the drive-off(s) happened (please fill in the corresponding
number of drive-off(s) with the position you were at that time)?
A: Captain: ____
____
D: 2nd Officer:
A: No
Answer:
URN:NBN:no-3369
Questionnaire Survey
137
During the loading phase in DP Weathervane Mode, due to some unknown failure, your
vessel is starting to have a 40% forward thrust drive-off.
1. What will be the first signal that makes you notice something is going wrong?
URN:NBN:no-3369
138
Appendix E
1. Below you find a list (A-I) of information sources, in order to clarify the situation,
i.e. to find out whether or not this is a drive-off, what is the sequence of your
check? (For example: A-B-C-E means A first, followed by B, and then C, and E is
the last.)
A: DP console
B: BLOM PMS screen
C: DARPS screen
D: Artemis screen
E: Main propeller(s) pitch indicator
F: Main engine(s) RPM indicator
G: Wind sensor
H: Wave measurement
I: Other equipment please specify: _______________
Probable sequence checks:
2. To finish the data check in Question 8 and confirm that this is a real drive-off, how
long time do you think is reasonable, starting from the first signal you get which
indicates something is going wrong?
A: within 20 seconds
B: within 40 seconds
C: within 60 seconds
D: other time please specify: _________________
Answer:
3. Assume that you have identified that you are in the drive-off situation, manual
intervention is supposed to be performed. Which way do you prefer to take action?
And why?
URN:NBN:no-3369
Questionnaire Survey
139
1. If you are working on a DP2 tanker: Which recovery actions do you prefer? and
Why?
A: Try to rotate the vessel by using max. thruster and max. rudder capacities, and steer
the vessel away from FPSO stern, no effort is made to initiate the astern pitch.
B: Try to stop the vessel by initiating astern pitch, no effort is made to initiate vessel
rotation.
C: Try to stop the vessel by initiating astern pitch, combined with the effort to rotate
vessel by using max. thruster and rudder capacities.
Answer:
Reason:
2. If you are working on a DP1 tanker: Which recovery actions do you prefer? and
Why?
A: Try to rotate the vessel by using max. thruster and max. rudder capacities, and steer
the vessel away from FPSO stern, no effort is made to initiate the astern pitch.
B: Try to stop the vessel by initiating astern pitch, no effort is made to initiate vessel
rotation.
C: Try to stop the vessel by initiating astern pitch, combined with the effort to rotate
vessel by using max. thruster and rudder capacities.
Answer:
Reason:
3. How long time do you think is reasonable to decide what to do, in which way, and
then initiate the recovery action, starting from the confirmation of drive-off
situation?
URN:NBN:no-3369
140
Appendix E
Several potential measures to improve tandem loading safety are listed below. Please comment upon
them.
1. To help early detection of drive-off situation, the present alarm settings (including speed
alarms, distance alarm/warning, engine output alarm) in your opinion are:
A: Practically effective
B: Not practically effective
C: Other viewpoints please specify, e.g. what should be done for alarms?
Answer:
Other viewpoints:
2. What is your opinion regarding increasing the separation distance between FPSO and
shuttle tanker in tandem loading, in order to give more time for decision making and
action formulation to avoid collision in tanker drive-off scenario?
3. What else do you think should be done to improve the tandem loading safety?
Comment:
URN:NBN:no-3369
URN:NBN:no-3369
Kavlie, Dag
Hansen, Hans R.
Gisvold, Kaare M.
Lund, Sverre
Vinje, Tor
Lorentz, Jan D.
Carlsen, Carl A.
Larsen, Carl M.
Hatlestad, Brigt
Valsgrd, Sverre
Pettersen, Erik
Nordsve, Nils T.
Fylling, Ivar J.
Haver, Sverre
Odland, Jonas
Engesvik, Knut
Mo, Olav
Amdahl, Jrgen
Czujko, Jerzy
URN:NBN:no-3369
Soares, C. Guedes
Mrch, Morten
Engseth, Alf G.
Baadshaug, Ola
Hessen, Gunnar
Taby, Jon
Wessel, Heinz-J.
Xu Jun
Guoyang Jiao
Olufsen, Arnt
Wu Yu-Lin
Farnes, Knut-Aril
Sotberg, Torbjrn
Hoen, Christopher
Sdahl, Nils
Haugen, Stein
Ormberg, Harald
Marley, Mark J.
Bessason, Bjarni
URN:NBN:no-3369
Svik, Svein
Karunakaran, Daniel
Nonlinear Dynamic Response and Reliability Analysis of Dragdominated Offshore Platforms. 1993.
Passano, Elizabeth
Bech, Sidsel M.
Wang, Xiaozhi
Hellan, yvind
Hermundstad, Ole A.
Vikestad, Kyrre
Azadi, Mohammad R. E. :
Mainon, Philippe
Sagli, Gro
URN:NBN:no-3369
Tronstad, Harald
Wang, Lihua
Heggelund, Svein E.
Babalola, Olusegun T.
kland, Ole
Ge, Chunhua
Byklum, Eirik