You are on page 1of 17

ARTICLE IN PRESS

Reliability Engineering and System Safety 93 (2008) 13281344


www.elsevier.com/locate/ress

Analysing the risk of LNG carrier operations


Erik Vanema,, Pedro Antaob, Ivan stvikc,1, Francisco Del Castillo de Comasd
a

DNV Research, Det Norske Veritas, 1322 Hvik, Norway


Unit of Marine Technology and Engineering, Technical University of Lisbon, Instituto Superior Tecnico, Lisboa, Portugal
c
LMG Marin, Bergen, Norway
d
Navantia, Madrid, Spain

Received in revised form 15 January 2007; accepted 30 July 2007


Available online 17 August 2007

Abstract
This paper presents a generic, high-level risk assessment of the global operation of ocean-going liqueed natural gas (LNG) carriers.
The analysis collects and combines information from several sources such as an initial hazid, a thorough review of historic LNG
accidents, review of previous studies, published damage statistics and expert judgement, and develops modular risk models for critical
accident scenarios. In accordance with these risk models, available information from different sources has been structured in the form of
event trees for different generic accident categories. In this way, high-risk areas pertaining to LNG shipping operations have been
identied. The major contributions to the risk associated with LNG shipping are found to stem from ve generic accident categories, i.e.
collision, grounding, contact, re and explosion, and events occurring while loading or unloading LNG at the terminal. Of these,
collision risk was found to be the highest. According to the risk analysis presented in this paper, both the individual and the societal risk
level associated with LNG carrier operations lie within the As Low As Reasonable Practicable (ALARP) area, meaning that further risk
reduction should be required only if available cost-effective risk control options could be identied. This paper also includes a critical
review of the various components of the risk models and hence identies areas of improvements and suggests topics for further research.
r 2007 Elsevier Ltd. All rights reserved.
Keywords: LNG carriers; Marine transportation; Risk analysis; Maritime safety; Gas tankers; Formal safety assessment; Liqueed natural gas

1. Introduction and background


1.1. The SAFEDOR project
The study presented in this paper was carried out within
SAFEDOR [1]. SAFEDOR is a research project conanced by the European Commission as an integrated
project (IP) in their 6th Framework Programme. One of
the aims of this project is to encourage innovative ship
design for cleaner and safer maritime transport. In order to
facilitate this, concepts for a risk-based regulatory framework will be developed that will open for acceptance of
novel design concepts based on risk analyses. These novel
design solutions might have equal or better safety than

Corresponding author. Tel.: +47 67 57 70 09; fax: +47 67 57 75 20.


1

E-mail address: Erik.Vanem@dnv.com (E. Vanem).


Grieg Logistics, Bergen, Norway since March 2006.

0951-8320/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ress.2007.07.007

conventional designs even though they may violate current


prescriptive requirements.
As a part of the SAFEDOR project, generic Formal
Safety Assessment (FSA) studies on various ship types
have been initiated. The risk assessment of LNG carriers
presented herein represents one partstep 2of one of
these generic FSAs.
1.2. Liquefied natural gas and LNG shipping
Liqueed natural gas (LNG) is composed of mostly
methane and is a cryogenic liquid at approximately
162 1C. When vaporized, its ammability range is
between approximately 5% and 15% by volume, i.e. a
mixture with air within this range of concentrations is
ammable. Thus, in addition to possible damages due to its
cryogenic temperatures, LNG spills are associated with
hazards such as pool res and ignition of drifting vapour
clouds. In its liquid state, LNG is not explosive, and LNG

ARTICLE IN PRESS
E. Vanem et al. / Reliability Engineering and System Safety 93 (2008) 13281344

1329

Fleet development - LNG tankers


400
350
300
# ships

250
200
150
100
50
0
1965

1975

1985

1995

2005

Year

Fig. 1. LNG eet developments 19652005 and forecast until 2010.

vapour will explode only if ignited in a mixture with air


within the ammability range and within an enclosed or
semi-enclosed space. Natural gas may also present an
asphyxiation hazard. LNG is not toxic and will not be
persistent if spilled in a marine environment. LNG weighs
less than water, thus LNG spilled on water will oat.
In liqueed form, the volume of LNG is 600 times less
than the same amount of natural gas at room temperatures. LNG shipping is therefore an economic way of
transporting large quantities of natural gas over long
distances. LNG is transported and stored at normal
atmospheric pressure, and LNG carriers are purpose-built
tank vessels for transporting LNG at sea.
The current world eet of LNG carriers is relatively
small; as of August 2005 it contained 183 ships,2 but it has
been increasing steadily in recent years. A further increase
is expected in the coming years. Fig. 1 illustrates the LNG
eet development, and includes a forecast until 2010.
During the more than 40 years of LNG shipping, the total
number of LNG carrier shipyears is 2838 (including 2005).
Of these, 1857 are accumulated since 1990.
In addition to an increase in numbers, the size of LNG
carriers is also increasing. The average size of the current
eet is almost 120,000 m3, whereas the average size of
vessels currently in the order book is 156,000 m3.2 LNG
super-tankers with capacities of 200,000250,000 m3 are
foreseen in the near future [2].
All types of LNG carriers are double-hull vessels, but
there exist different cargo containment systems of independent or integrated cargo tanks. The current LNG eet
is dominated by two main types of vessel designs, i.e. the
membrane tank designs and the spherical tank designs. In
membrane tank designs, the cargo containment system
consists of a very thin invar or stainless steel double-walled,
insulated cargo envelope that is structurally supported by
the ships hull. The spherical tank carriers, also referred to
as Moss tankers, have spherical aluminium tanks or
prismatic-shaped stainless steel tanks that are self-support2

Information obtained at http://www.lngoneworld.com.

ing within the ships hull. These tanks are insulated


externally. Both tanker alternatives are designed, constructed and equipped with sophisticated systems for
carrying LNG over long distances, stored at temperatures
around 162 1C. Each of the main types of LNG vessel
designs constitute about half of the eet (the actual
distribution is 50% membrane ships, 45% spherical
tankers and 5% of other types of LNG tankers), but
membrane tankers are dominant among LNG new
buildings. LNG vessels are generally well designed, well
maintained and operating with well-trained crew. Thus,
LNG shipping so far has a good safety record. The two
main types of LNG carriers are illustrated in Fig. 2.
1.3. The FSA methodology
FSA is a standard risk assessment, with the aim of
developing maritime safety regulations in a structured and
systematic way. It should not be mistaken for a risk
assessment for a specic ship or a ships safety case. FSA
may better be described as a safety case for the rules and
regulations. FSA can be used in the evaluation of new
regulations and for comparison between existing and
possibly improved regulations, and it aims at balancing
safety and environmental protection levels with costs. Both
technical and operational issues, including the inuence of
the human element on shipping accidents, may be
incorporated in an FSA. Various aspects and applications
of the FSA methodology have been extensively discussed in
different academic journals [39].
IMO has developed guidelines for FSA studies [10]. By
now, a number of FSA studies have been performed and
reported to IMO according to these guidelines, and
decisions have been made based on such submissions
[11,12]. The methodology is described as a 5-step process:
0.
1.
2.
3.

Preparatory steps
Identication of hazards
Risk analysis
Identifying risk control options

ARTICLE IN PRESS
1330

E. Vanem et al. / Reliability Engineering and System Safety 93 (2008) 13281344

Fig. 2. Main types of LNG carriers: moss spherical tankers (top) and membrane tankers (bottom).

4. Costbenet assessment
5. Recommendations for decision-making.
The study presented in this paper corresponds to the
second steprisk analysisof an ongoing FSA on LNG
tankers.
2. Dening the scope of study
This paper describes a high-level, generic risk analysis of
LNG carriers, and the scope of the study is the whole eet
of ocean-going LNG carriers. A generic LNG carrier is
assumed representative for the various types of LNG
vessels. Historic accident experience indicates that the risk
level of the main types of LNG carriers is comparable, so
this assumption seems valid up to a certain level of detail. It
is noted that membrane tanks have experienced cargo tank
leakage through its primary barrier, and are also believed
to be more vulnerable to effects of dynamic loads and
sloshing of LNG cargo, but risk contributions from such
scenarios are assumed small in comparison with the main
accident scenarios investigated in this study. Therefore,
detailed studies of specic ships as well as specic trades
and port environments are dened out of scope.
However, a particular reference vessel will be used where
needed. For the purpose of this generic analysis, a
138,000 m3 membrane LNG vessel currently under construction at Navantia was chosen. The cargo capacity of
this reference vessel is within the main range of LNG
carriers and its speed is typical for this type of vessels. It
has a reversible geared, cross-compound steam-driven
28,000 kW  83 RPM main turbine and a 5-blade xed
pitch-type propeller. The main characteristics of the
reference vessel are given in Table 1.
Only the shipping phase of the LNG chain is considered,
covering loading at the export terminal, the actual voyage
and unloading at the receiving terminal. Risks pertaining
to exploration, production, liquefaction, storage and
regasication of natural gas are considered out of scope.
Furthermore, only the operational phase of an LNG vessel
is considered, and risks associated with construction,
repairs in dock and scrapping of LNG vessels are out of
scope.

Table 1
Main characteristics of reference vessel
Length overall
Length between perpendiculars
Breadth (moulded)
Depth to main deck (moulded)
Depth to trunk deck (moulded)
Draft design (moulded) under keel
Draft scantling (moulded)
Double-bottom depth
Double-side width
Total displacement
Total cargo tank capacity
Total insulation thickness
Service speed at design moulded draft
Accommodation capacity
a

284.40 m
271.00 m
42.50 m
25.40 m
32.20 m
11.40 m
12.30 m
2.9 m
2.32 m
98,500 t
138,000 m3
0.530 m
19.50 knotsa
40 personsb

At NCR (90% of MCR) of main propulsion machinery.


Including 4 Suez canal workers.

Security issues are regarded as out of scope. The


environmental risks associated with LNG shipping are
assumed small since LNG is non-toxic and non-persistent,
and is therefore not relevant to this study. Thus, the focus
of this study has been on safety and risk to human lives.
Third-party risks to people on shore are also left out of the
study. It is argued that such risks should rather be
investigated in specic risk analyses pertaining to specic
LNG terminals or LNG trades, and these issues are more
related to the siting of LNG terminals than to IMO
regulations for maritime safety. Thus, this study only
considers the risks to LNG crew as well as risks to crew and
passengers onboard other vessels. Finally, the focus has
been on accidents of a certain scale, and high-frequency,
small-scale occupational accidents have not been emphasized.
3. Risk acceptance criteria
In order to evaluate the risk as estimated by a risk
analysis, appropriate risk acceptance criteria should be
established prior to and independent of the actual risk
analysis. Among different principles of safety [13], the As
Low As Reasonably Practicable (ALARP) principle will be
employed, i.e. risk acceptance criteria should divide

ARTICLE IN PRESS
E. Vanem et al. / Reliability Engineering and System Safety 93 (2008) 13281344

between three levels of risk: intolerable, ALARP and


negligible. For risks within the ALARP area, costeffectiveness considerations apply to the amount of
resources that should be spent on risk mitigation. The
IMO FSA guidelines [10] recognized this as the current best
practice, although criticism of the ALARP principle has
occurred [14]. For this risk analysis on LNG carriers, risk
acceptance criteria for individual and societal risks were
derived for crew as well as for passengers onboard other
vessels that might be affected by a possible LNG accident.
3.1. Individual risk acceptance criteria
A thorough review of alternative risk acceptance criteria
was presented in Skjong et al. [15]. Based on this review,
the acceptance criteria for individual fatality risk to crew
presented in Table 2 were adopted for the current study.
This corresponds to the risk level experienced by an
exposed crewmember. The individual risk to third parties
(including passengers on other vessels) is intuitively
assumed to be negligible, so only criteria for the LNG
crew are deemed necessary.
3.2. Societal risk acceptance criteria
Societal risk acceptance criteria for crew were established
based on the approach described in Norway [16]. According to this method, acceptance criteria are associated with
the economic importance of LNG shipping, calibrated
against the average fatality rate per unit of economic
production. Based on reasonable estimates of daily rates,
operational costs and capital costs due to initial investTable 2
Individual risk acceptance criteria for LNG crew
4103
106103
o106

Intolerable risk per year


ALARP area per year
Negligible risk per year

1331

ments, the economic value of LNG shipping was estimated


to be USD 1.6 million per shipyear. The risk acceptance
criteria that were derived based on these estimates are
illustrated in Fig. 3. It should be noted that these criteria
are somewhat stricter than the criteria proposed for tankers
in general by Norway [16].
These societal risk acceptance criteria can be assumed
appropriate also for crew onboard other vessels, but for
passengers onboard other vessels more strict criteria might
be appropriate. Thus, for the purpose of this study, societal
risk acceptance criteria for passengers with anchor points
that are one order of magnitude lower than for the criteria
for crew will be used.
3.3. Cost-effectiveness criteria
The results from the risk analysis presented in this paper
may be carried forward to support cost-effectiveness
assessments of new as well as existing safety measures. If
the risk level is found to lie in the ALARP area, all
available cost-effective risk control options should be
implemented according to the FSA philosophy. There exist
various measures of cost effectiveness, and two that are
often used in conjunction with FSA studies are the Gross
Cost of Averting a Fatality (GCAF) and the Net Cost of
Averting a Fatality (NCAF), dened by Eqs. (1) and (2).
DC is the cost, DR is the risk reduction and DB is the
economic benet resulting from implementing the measure,
respectively:
GCAF

DC
,
DR

(1)

NCAF

DC  DB
.
DR

(2)

The risk analysis denes the baseline risk level and


thereby also the potential for risk reduction. For a
prospective measure, i.e. a regulatory option, the risk
models developed in this paper can be used to investigate

Risk acceptance criteria for LNG tankers


F - frequency of N or more fatalities
(per shipyear)

1.E-01

1.E-02

Intolerable
1.E-03

ALARP

1.E-04

1.E-05

Negligible
1.E-06
1

10
N - number of fatalities

Fig. 3. Societal risk acceptance criteria for crew.

100

ARTICLE IN PRESS
E. Vanem et al. / Reliability Engineering and System Safety 93 (2008) 13281344

1332

Historic LNG incidents


Other

Unknown

11 %

1%
Spherical
51 %

Membrane
37 %

LNG incidents, 1985 and later


Other
4%

Unknown
2%
Spherical
33 %

Membrane
61 %

Fig. 4. Breakdown of LNG accidents based on ship type: since the start of LNG shipping (top) and for 1985 and later (bottom).

the expected risk-reducing effect. Further analyses would


also be required in order to estimate the expected cost as
well as any economic benets that would result from
implementing the measure. In accordance with current
practice within IMO and the proposals presented in MSC
72/16 [10], a risk control measure will generally be
recommended for implementation if GCAFpUSD 3
million or NCAFpUSD 3 million (note that by denition,
NCAFpGCAF, so if GCAFpUSD 3 million, NCAF will
always be pUSD 3 million). For risk control options
where the estimated cost effectiveness is close to USD 3
million, further scrutiny might be required.

ing the severity of the accidents or the relative population


of membrane and spherical carriers.
The available material indicates that accidents have
happened more often on spherical-type LNG ships during
the history of LNG shipping, but that accidents have been
more frequent on membrane ships during the last 20 years.
Available statistics are too sparse to draw any denite
conclusions, and conclusions cannot be drawn without
considering the population of each LNG carrier type, but it
is noted that accidents have occurred for all types of LNG
carriers. For the purpose of this high-level study, the
accident frequency is henceforth assumed independent of
LNG carrier type.

4. Accident scenarios
4.2. Generic accident categories
4.1. Operational experience with LNG tankers
Marine transportation of LNG has gradually increased
since the rst LNG cargo was transported by an LNG
tanker in 1959 and the rst purpose-built LNG tanker was
engaged in commercial trade in 1964. Thus, more than 40
years of operational experience with LNG carriers has
accumulated over the years.
A literature survey on the history of LNG shipping
reveals information of 182 events, with or without LNG
spillage, involving LNG carriers larger than 6000 GRT
[17]. Of these, 24 occurred to ships out of regular service
(e.g. in yard during construction or repair or while in tow
or during sea trials) and are considered out of scope. Thus,
previous experience amounts to 158 known relevant
accidents involving LNG carriers. The breakdown of
LNG events on ship types according to these accidents is
illustrated in Fig. 4. It is emphasized that this only
illustrates the total number of accidents without consider-

The 158 known relevant LNG carrier accidents can be


grouped into a few generic accident types. The breakdown
of accidents into accident categories is presented in Table 3,
and Table 4 presents the breakdown into accident
categories and different periods of time.
This categorization of accidents is in general agreement
with the scenarios identied during the hazard identication [18]. Some of the accident categories are not believed
to be associated with severe consequences in terms of
fatalities. For example, the accident category equipment
and machinery failure is associated with incidents where the
failure did not lead to any subsequent events such as
collision, grounding or re. For incidents where subsequent
events occurred, e.g. for rudder failures leading to collision
or grounding, it has been sorted under the accident
category corresponding to this subsequent event. Hence,
even though equipment and machinery failure is associated
with the highest incident frequency, this accident category

ARTICLE IN PRESS
E. Vanem et al. / Reliability Engineering and System Safety 93 (2008) 13281344
Table 3
Distribution of historic LNG accidents on categories
Accident category

Collision risk

Accidents
(#)

Collision
Grounding
Contact
Fire and explosion
Equipment and machinery failure
Heavy weather
Events while loading/unloading cargo
Failure of cargo containment system
Total

Frequency
(per shipyear)

19
8
8
10
55
9
22
27

6.7  103
2.8  103
2.8  103
3.5  103
1.9  102
3.2  103
7.8  103
9.5  103

158

5.6  102

Table 4
Breakdown of historic accident data based on accident categories and
periods of time

Exposure (accumulated
shipyears)
Collision
Grounding
Contact
Fire and explosion
Equipment and
machinery failure
Heavy weather
Events while loading/
unloading cargo
Failure of cargo
containment system
Total

6475
116

7685
585

8695
770

9605
1367

6405
2838

1
1

10
6
4
5
39

4
1
4
3
9

19
8
8
10
55

6
13

3
3

9
22

15

27

15

98

22

23

158

is not believed to be very critical to crew safety. Also,


incidents due to bad weather and failure of cargo containment system that are not leading to subsequent accidents
such as collision, grounding and re are considered to
constitute a small risk compared with the remaining
accident scenarios. Thus, the following ve generic accident
scenarios were selected for further study in the risk
analysis:







1333

Collision
Grounding
Contact3
Fire or explosion
Incidents while loading/unloading of cargo

These are regarded as the main risk contributors, and


risks from other scenarios are assumed to be negligible in
comparison. It is noted that the accident categories listed
above are general maritime accident scenarios that can
occur for all types of ships, but the potential consequences
3
Striking or being struck by any xed or oating objects other than
another ship or the sea bottom.

Grounding risk

Total risk

Contact risk

Fire/explosion risk

Loading/Unloading risk

Fig. 5. Overall risk model for LNG carriers.

and the possible further escalation of the scenarios are


specic to LNG carriers due to the characteristics of its
cargo.
5. Risk models
Selection of accident scenarios for further analysis
corresponds to the overall risk model for LNG carriers
as illustrated in Fig. 5. In the following, overall risk models
for each accident category will be developed and the
frequencies and consequences associated with each of the
sub-models will be further investigated.
5.1. Frequency assessment
The frequency of an initiating event for each of the risk
sub-models is based on the historic frequencies in Table 3.
However, some adjustments to the estimates for the re
and explosion scenario and the loading/unloading scenario
are made.
In all, 50% of the experienced re and explosion
accidents were vent riser res and these do not constitute
any signicant hazard to the crew or the ship. Hence, these
incidents were not regarded and the initiating frequency of
re and explosion is reduced accordingly, i.e. to 1.8  103
per shipyear. A total of 22 loading/unloading incidents
have been reported. However, only 9 of these reported any
leakage of LNG, and only these events are assumed critical
to safety. This corresponds to a frequency of 3.2  103 for
loading/unloading accidents, resulting in leakage of
LNG, and this frequency will be used in the risk
assessment. The initial frequencies used for each accident
scenario are presented in Table 5. Comparing these
frequencies with statistics for other generic vessel types,
such as oil tankers, chemical tankers, LPG tankers and
bulk carriers, they are found to agree reasonably well. In
general, the accident frequencies are found to be somewhat
lower for LNG carriers than for these other types of
vessels, but this was expected considering the high focus
on safety on these ships and the generally high competence
of LNG crew.

ARTICLE IN PRESS
1334

E. Vanem et al. / Reliability Engineering and System Safety 93 (2008) 13281344

5.2. Consequence assessment


The expected consequences for each of the selected
scenarios will be assessed by utilizing event tree techniques.
First, a set of event trees will be constructed, one for each
accident scenario, based on conceptual risk models. Then,
these event trees will be quantied using a variety of
different techniques for different branches. The estimates in
Table 5 will be used as frequencies for the initiating events.
5.2.1. Constructing event trees
As the event trees tend to grow complex, it is not feasible
to describe them in detail within the format of this paper.
However, the conceptual risk model for each accident
scenario will be presented in the following.
A typical collision scenario with an LNG carrier might
develop in the following way. First, a collision occurs. The
LNG vessel might be the struck or the striking ship. If the
LNG vessel is the striking ship, the likelihood of further
escalation of the accident is regarded as small as it will
receive the collision impact in the bow in front of the
Table 5
Frequency estimates for selected accident scenarios
Accident category

Initial frequency
(per shipyear)

Collision
Grounding
Contact
Fire and explosion
Leakage of LNG while loading/unloading

6.7  103
2.8  103
2.8  103
1.8  103
3.2  103

Collision
frequency

collision bulkhead. Furthermore, the collision might occur


when the LNG carrier is in ballast or when it is fully
loaded. When the LNG carrier is the struck ship, the
collision might cause a damage that penetrates the outer
and inner hulls. If penetrating the inner hull, it might cause
leakage of cargo. This might result in materialization of an
LNG hazard such as pool re. Cryogenic temperatures of
LNG or heat generated from a pool re might deteriorate
the strength of the ship and may eventually lead to sinking.
If LNG hazards materialize or the ship sinks, failure to
evacuate in time may lead to a number of fatalities among
the crew. Finally, for some of the possible LNG hazards,
fatalities may occur among crew or passengers onboard the
other vessel. Hence, the risk model illustrated in Fig. 6
describes a typical collision accident, and an event tree is
constructed based on this.
The grounding and contact scenarios will resemble the
collision scenario in many ways. First, a grounding/contact
event might occur in either loaded or ballast condition. The
grounding or contact will result in a certain extent of ship
damage, and this damage might cause leakage of cargo
and/or loss of stability. If LNG is released, one or more
LNG hazards might materialize. Again, the LNG carrier
might sink due to the damage or due to deteriorating
strength caused by LNG leakage. Finally, if the crew is not
able to evacuate in time, there might be fatalities due to
LNG hazards or the sinking of the ship. The grounding
and contact risk models that form the basis for event trees
are illustrated in Fig. 7.
The re or explosion scenario describes an accident
where a re or an explosion is the initiating event. It is
distinguished between three types of re scenarios, namely,

Collision frequency

Loading
condition model

Probability of being loaded/in ballast

Damage extent
model

Probability distribution of damage

Cargo leakage
frequency

Probability of cargo release

LNG hazard
model

Probability distribution of
LNG hazards materializing

Survivability
model
Third parties
model

Probability of
sinking

Evacuation
model
Number of
fatalities, LNG

Number of
fatalities, other
vessel
Consequence

Fig. 6. Risk model for collision of LNG carriers.

ARTICLE IN PRESS
E. Vanem et al. / Reliability Engineering and System Safety 93 (2008) 13281344

Grounding /contact
frequency model

1335

Grounding or contact frequency

Loading
condition model

Probability of being loaded /in ballast

Damage extent
model

Probability distribution of damage

Cargo leakage
frequency

Probability of cargo release

LNG hazard
model

Probability distribution of
LNG hazards materializing

Survivability
model

Probability of
sinking

Evacuation
model
Number of
fatalities, LNG

Consequence

Fig. 7. Risk model for grounding or contact of LNG carriers.

res that start in the machinery spaces, in accommodation


areas or day rooms, and in the cargo area. For res starting
in machinery spaces or in accommodation areas, it is
deemed highly unlikely that it will spread to the cargo area
and no LNG-specic hazards are assumed. Thus, these
scenarios will resemble similar re accidents on other cargo
ships such as oil tankers. The compressor room is the most
likely place in the cargo area for a re to break out, and
such res will be specic to LNG carriers. For a
compressor room re, the following scenario is assumed:
the re protection systems might fail in preventing or
extinguishing the re or explosion, which might lead to a
breach in the cargo containment system and subsequent
leakage of LNG. If there is leakage of LNG, an LNG
hazard might materialize, and there will be a possibility
that the ship will not survive. Finally, in the event of an
escalating re, the crew needs to evacuate and failure to do
so in time might result in a number of fatalities. The re
and explosion risk model adopted in this study is illustrated
in Fig. 8.
Spillage events during loading or unloading of cargo
while in port are generally assumed to be of small scale,
where only a limited number of the crew are exposed to
risks of injuries or death. Fatal accidents are only deemed
likely for crew members directly exposed to the cryogenic
LNG. The resulting risk model is illustrated in Fig. 9,
which was used to construct event trees for this scenario.
5.2.2. Quantifying the event trees
In order to assign probabilities for the various escalating
events, and thereby quantify the probabilities and con-

sequences associated with each scenario, a set of different


approaches and techniques have been used. For each submodel and for each branch of the event three, the method
that was found to be most practical was utilized and the
information source that was assumed to be most relevant
was exploited. Detailed descriptions of all these are not
possible within the scope of this paper, but an outline
describing the main approaches will be provided in the
following.
The event tree for the collision scenario was quantied
based on various techniques. First, some obvious general
assumptions were made, e.g. the probability of being the
striking or struck ship was assumed to be 0.5, and the
probability of being loaded and in ballast was assumed to
be 0.5 (since LNG shipping is principally unilateral in
nature). The damage extent model contains several parts,
describing the probability distribution of damage location,
depth, length and height that again determines the
probabilities of water ingress, receiving the damage in the
cargo area, etc. For these probabilities, results from
previous studies have been utilized, e.g. Olufsen et al. [19]
and Skjong and Vanem [20], and in particular the damage
database developed by the HARDER project have been
exploited [21,22]. The extent of collision damage will
determine the probability of cargo leakage. The remaining
parts of the risk model, related to materialization of LNG
hazards, survivability and evacuation, are associated with
signicant uncertainties and no amount of relevant data
exist. Therefore, a workshop was arranged where expert
opinion was elicited and incorporated in the event tree.
A separate Delphi session was arranged in order to arrive

ARTICLE IN PRESS
1336

E. Vanem et al. / Reliability Engineering and System Safety 93 (2008) 13281344

Fire/explosion frequency model


Engine room fire

Fire/explosion frequency

Accommodation area fire Compressor room fire

Probability of being loaded/in


ballast, at sea/port

Loading
condition model

Fire protection
model

Probability of fire protection


system failing

Cargo leakage
model

Probability of cargo
release

LNG hazard
model

Probability
distribution of LNG
hazards materializing
Probability of
sinking

Survivability
model

Evacuation
model
Number of
fatalities
Consequence

Fig. 8. Risk model for re and explosion on LNG carriers.

Loading/unloading
spill frequency model

Probability of loading/unloading incident with spillage

Spillage extent
model

Probability distribution of spillage extent

LNG hazard
model

Probability distribution of
LNG hazards materializing

Accident model

Number of fatalities, LNG


crew, terminal workers

Consequence

Fig. 9. Risk model for spill events during loading or unloading of LNG carriers.

at consequence estimates related to number of fatalities on


the LNG vessel as well as on the other vessel involved in
the collision. All the estimates were inserted into the event
tree in order to estimate the overall risk related to the
collision scenario. The event tree for collision scenarios is
illustrated in Fig. 10.
For the grounding and contact scenarios, similar
approaches for quantication of the event trees were used
as in the collision scenario, and the same sources of

information were exploited. The event trees were quantied accordingly, as illustrated in Fig. 11 (grounding) and
Fig. 12 (contact).
Due to its relatively small eet, sufcient data on LNG
carrier res are not readily available. In the accident
statistics for LNG carriers only ve res were reported
(disregarding vent riser res), and, of these, 3 started in the
machinery spaces. This would correspond to 60% of the
res originating in the engine rooms, but it is realised that

0.0067

Collision

Yes

fatalities per shipyear

0.00159

3rd party

0.144

0.05

0.95

0.856

fatalities per shipyear

0.65

0.35

0.086

0.05

0.95

0.1

0.9

0.11

0.89

No pool
fire

LNG Hazaard model

No drifting
No cryogenic
vapour
No leakage
damage to
cloud
of LNG
hull
ignition

Cargo
leakage
model

0.914

For Crew 0.004415

0.5

0.65

0.35

0.5

0.5

0.5

critical
damage

Damage extent model

Damage
In ballast outside
(no LNG) cargo area
only

TOTAL RISK

No

0.5

0.5

Striking
ship

Loading
condition
model

A crew of 30 is assumed

Collision frequency
model

0.989

0.989

0.978

0
0.978

0
0.978

0
0.978

16

16

12.9

12.9

12.9

12.9

Probability of
fatalities
# fatalities
among crew

Evacuation model

0.98

0.98

0.98

Probability
of fatalities
on other
vessel

3.11
62.7
627

0.97
0.03

627

0.03

0.13

62.7

0.87

3.11
0.97

627

0.03

0.87

62.7

0.13

3.11
0.97

# fatalities
on other
vessel

0.13

20.000 GRT)

Not large
passenger
ship (

0.87

Other
vessel not
passenger
vessel

Third party model

Fig. 10. Event tree: collision of LNG carriers.

Surviving

Survivability
model

Frequency

Risk

Sum Risk

1.535E-05 0.00024561

LNG Crew 0.00399896

16 1.5506E-05 0.00024809

16

12.9 0.00012282 0.00158435

12.9 2.8668E-05 0.00036981

12.9 9.1573E-05 0.00118129

627

62.7

3.11

627

62.7

3.11

627

62.7

Risk

fatalities per shipyear


fatalities per shipyear

0.00159049

3rd party

Other Crew 0.00041572

5.992E-08 3.7571E-05

1.937E-06 0.00012148

1.337E-05 4.1572E-05

5.932E-08 3.7195E-05

1.918E-06 0.00012026

1.323E-05 4.1156E-05

4.8E-07 0.00030094

1.552E-05 0.00097304

0.0001071 0.00033299

Frequency

Third parties

3.11

Consequence

Risk contribution

12.9 2.8668E-05 0.00036981

Consequence

LNG crew

ARTICLE IN PRESS

E. Vanem et al. / Reliability Engineering and System Safety 93 (2008) 13281344


1337

1338

A crew of 30 is assumed
Grounding
frequency
model

Grounding

In ballast
(no LNG)

Cargo
leakage
model

Damage extent model

Damage
outside
cargo area
only

0.5

0.3

0.7

0.3

0.7

No cryogenic No drifting
No leakage
No pool
damage to
vapour cloud
of LNG
fire
hull
ignition

Probabilities of
fatalities among
crew

Surviving

0.924

0.076

# fatalities

Consequence

Frequency

Risk
contribution

0.978

12.9

12.9 3.1218E-05 0.000402709

12.9

12.9 7.2841E-05 0.000939655

12.9

12.9 3.1218E-05 0.000402709

0.989

16

16 5.9002E-05 0.000944036

0.989

16

16 7.2924E-06 0.000116679

0.989

16

16 7.3661E-06 0.000117857

0.924

0.076

0
1

0.978
0

0.924

0.076

0
1

0.978

0.924

0.076

0.9

0.89

1
Yes

Evacuation m odel

0
1

No

1
0.11

0
1

0.1

0
1

0
1

Sum Risk

Fig. 11. Event tree: grounding of LNG carriers.

0.002923644 Fatalities per shipyear

ARTICLE IN PRESS

0.5

critical damage

Survivability
model

LNG Hazard model

E. Vanem et al. / Reliability Engineering and System Safety 93 (2008) 13281344

0.0028

Loading
condition
model

A crew of 30 is assumed
Grounding
frequency
model

Contact

In ballast
(no LNG)

Cargo
leakage
model

Damage extent model

Damage
outside
cargo area
only

0.5

0.3

0.7

0.3

0.7

No
No leakage cryogenic
of LNG
damage to
hull

No drifting
vapour
No pool
cloud
fire
ignition

Evacuation model

Probabilities
of fatalities # fatalities
among crew

Surviving

0.962

0.038

0
1

0.978

12.9

12.9 1.5609E-05 0.00020135

12.9

12.9 3.6421E-05 0.00046983

12.9

12.9 1.5609E-05 0.00020135

0.989

16

16 2.9501E-05 0.00047202

0.989

16

16 3.6462E-06 5.8339E-05

0.989

16

16

0.038

0
1

0.978
0

0.962

0.038

0.038

0
1

Risk
contribution

Frequency

0.962

0.962

Consequence

0.978
0

0.9

0.89

0.11

0
1

1
Yes
No

1
0.1

0
1

0
1

Sum Risk

3.683E-06 5.8929E-05

0.00146182 Fatalities per shipyear

Fig. 12. Event tree: contacts of LNG carriers.

ARTICLE IN PRESS

0.5

critical damage

Survivability
model

LNG Hazard model

E. Vanem et al. / Reliability Engineering and System Safety 93 (2008) 13281344

0.0028

Loading
condition
model

1339

1340

A crew of 30 is assumed
Loading condition
model

Fire / explosion Model

Fire and explosion


distribution

Compressor room
0.03

Probability of
fatalities
# fatalities
among crew

Surviving

Consequence

Frequency

Risk
contribution

0.1

0.85
0.15

0.37

1.4985E-07

1.4985E-07

0.9

0.85

0.37

7.6424E-06

7.64235E-06

0.15

0.37

1.3487E-06

1.34865E-06

0.85

0.37

8.4915E-07

8.4915E-07

0.9

0.37

1.4985E-07

1.4985E-07

0.1

0.989

4.0055E-08

3.20436E-07

0.1

0.15

Evacuation model

0.9

0.37

8.4915E-07

Yes

8.4915E-07
No

0.85

0.37

7.6424E-06

7.64235E-06

0.15

0.9

0.37

1.3487E-06

1.34865E-06

0.1

0.989

16

16

3.6049E-07

5.76785E-06

0.0018

0
1

Machinery space
0.81

0.37

1 0.00053946

0.00053946

0.37

1 0.00010656

0.00010656

Accommodation area
0.16

Sum Risk

Fig. 13. Event tree: re and explosion of LNG carriers.

0.000672088

fatalities per ship year

ARTICLE IN PRESS

0.5

Survivability
model

Fire Fighting
No leakage no pool
systems
of LNG
fire
successful

In ballast
at port
(no LNG)

0.5

LNG
Hazard
model

Cargo
leakage
model

E. Vanem et al. / Reliability Engineering and System Safety 93 (2008) 13281344

Fire /
Explosion

Fire
protection
model

ARTICLE IN PRESS
E. Vanem et al. / Reliability Engineering and System Safety 93 (2008) 13281344

the statistical base for this estimate is very weak. It was


therefore assumed that the relative distribution of res
starting in machinery spaces, accommodation areas and
cargo areas is similar to that of oil tankers, and recent
results from a study on aframax tankers were utilized [23].
It may be argued that both the cargo area and the
machinery spaces are very different for LNG tankers and
oil tankers, and this suggests that probabilities arrived at
from such comparison are not very accurate. Notwithstanding, re statistics for LNG carriers are too scarce to
be useful, and statistics from oil tankers are believed to be
the best available source of information. The estimates
used for the relative distribution of res should be regarded with a high degree of uncertainty, but it is noted
that the overall results are not overly sensitive to these
uncertainties.
Furthermore, for res in machinery spaces and accommodation areas, statistics from oil tanker accidents were
assumed appropriate for LNG tankers and no further risk
modelling was deemed necessary with regard to how these
res would potentially escalate. For quantication of the
risk model for compressor room res, different sources of
information were used. For example, the assumed failure
rate of the re protection system was based on results from
previous studies on other ship types [24,25]. Estimates on
probabilities related to LNG leakage, LNG hazards
survivability and evacuation performance were again based
on expert judgement. The event tree for re and explosion
is illustrated in Fig. 13.
The event tree for spillage events during loading and
unloading of cargo was quantied based on assumptions
related to typical LNG trading patterns and normal cargo
transfer operations. Furthermore, it was assumed that only
relatively small-scale spillages would occur during loading
and unloading, with the potential to harm only those
members of the crew directly exposed to the spilled LNG.
The event tree for loading and unloading events is
illustrated in Fig. 14.

6. Risk summation
The risk modelling outlined in the previous sections
results in the contributions from the various scenarios to
the total potential loss of lives per shipyear from LNG
shipping operations. This is presented in Table 6.
6.1. Individual risk
Intuitively, the individual risk for passengers onboard
other vessels is not an issue in this context, and only the
individual risk to LNG crew will be considered. Assuming
a crew of 30 for a typical LNG carrier and a 5050 rotation
scheme (meaning that two complete crews are needed for
continuous operation of the vessel), the individual risk to
crew is assessed to be 1.6  104 per person year.
Compared with the individual risk acceptance criteria
established in this study, it is seen that the individual risk
falls within the ALARP area.
It should be noted, however, that this is the individual
risk from ship accidents, and contributions from occupational accidents are not considered. Occupational accidents
were not considered in the present study, but previous
estimates of occupational fatality risks onboard gas tankers
are presented by Hansen et al. [26] as 2.7  102 fatalities
per 10,000 days onboard. Assuming a 5050 rotation

Table 6
Potential loss of lives from LNG carrier operations per shipyear
Accident category

PLL (crew)

PLL (passengers)

Collision
Grounding
Contact
Fire and explosion
Loading/unloading events

4.42  103
2.93  103
1.46  103
6.72  104
2.64  104

1.59  103
0
0
0
0

Total

9.74  103

1.59  103

Yes

A crew of 30 is assumed
No

L/U incident
frequency
model

Freq of L/U
incident

0.00317

Spillage
LNG Hazard
extent model
model

Accident
model

No cryogenic
No spillage
damage to
near to crew
# fatalities
crew
operation
members

Consequence

0.9167

0.0833

1341

Frequency

0.000264167

Sum Risk

Risk
contribution

0.000264167

0.000264167

Fatalities per shipyear

Fig. 14. Event tree: spillage during loading or unloading operations of LNG carriers.

ARTICLE IN PRESS
E. Vanem et al. / Reliability Engineering and System Safety 93 (2008) 13281344

1342

scheme, a crewmember will be on board for approximately


182 days per year, and 10,000 days onboard corresponds to
55 person years for a typical crewmember. Hence,
according to the estimates presented by Hansen et al.
[26], the occupational fatality rate onboard gas tankers is
around 4.9  104 per person year. Assuming this estimate
and adding this to the individual risk from ship accidents,
the total individual risk for crewmembers is estimated to be
approximately 6.5  104 per person year. This estimate is
still within the ALARP area according to the risk
acceptance criteria established in this study. On a nal
note, it is observed that the individual risk to crewmembers
onboard LNG carriers is dominated by occupational
accidents, with a ratio of 3 occupational fatalities to every
fatality due to ship accidents according to the estimates
above.

passenger (Figs. 15 and 16). Compared with the established


risk acceptance criteria, it is found that also the societal
risks lie within the ALARP region. An FN curve that
shows the contribution from each of the main accident
scenarios may also be produced (Fig. 17), and it is readily
seen that the overall risk level is dominated by the collision,
grounding and contact scenarios. However, re and
explosion are dominating the low-consequence risk contributions in the order of one fatality. The frequencies in all
3 gures are in terms of per shipyear.
It is observed that the FN curves resulting from this
study lie slightly above the FN curves for gas tankers
presented in MSC 72/16 [16], but in general the FN curves
are found to be in reasonable agreement.

6.2. Societal risk expressed by FN curves

When the various assumptions and other sources of


uncertainties are reviewed, it becomes evident that some
are conservative while others might be optimistic. Thus, the
overall results might in principle be skewed in either way.

Detailed results from the risk analysis can be used to


produce FN curves for the overall risk to crew and

7. Uncertainties and suggestions for further research

Risk level - LNG carriers

F - frequency of N or more fatalities

0.01

0.001

0.0001

0.00001
1

10

100

N - number of fatalities

Fig. 15. FN curve for crew.


Risk to passengers from LNG carrier operations

F - frequency of N or more fatalities

1.E-04

1.E-05

1.E-06

1.E-07

1.E-08
1

10

100
N - Number of fatalities

Fig. 16. FN curve for passengers.

1000

10000

ARTICLE IN PRESS
E. Vanem et al. / Reliability Engineering and System Safety 93 (2008) 13281344

1343

Risk level - LNG carriers


Broken down on accident categories

F - frequency of N or more fatalities

1.E-02
1.E-03
1.E-04
1.E-05
1.E-06
1.E-07
1

10

100

N - number of fatalities
Risk level - LNG

Collision

Grounding

Contact

Fire/Explosion

Loading/Unloading

Fig. 17. FN curve for crew, broken down into accident types.
Table 7
Conservative and optimistic assumptions made in the presented study
Assumption

Effect

Relevant scenario

HARDER data have been used for damage distributions


Consequential cryogenic damages due to LNG released assumed to sink the ship
Probability of crack in tanks in case of compressor room re 0.1
Fire frequency compared with oil tankers
Fire-ghting systems assumed similar as for HSC and passenger ships
No leakage of LNG in ballast condition
Critical damage penetration in grounding scenarios is 3.4 m
No cryogenic damages to crew
High probability (0.8) of no pool re in the event of LNG leakage

Conservative
Conservative
Conservative
Conservative
Conservative
Optimistic
Optimistic
Optimistic
Optimistic

Collision, grounding and contact


Collision, grounding, contact and re and explosion
Fire and explosion
Fire and explosion
Fire and explosion
Collision, grounding, contact and re and explosion
Grounding
Collision, grounding and contact
Collision, grounding and contact

However, upon further evaluation of the various assumptions, it is believed that the net effect tends to be
conservative, and it is believed that the overall results are
more likely to be conservative than optimistic. Hence, the
risk analysis presented in this paper might be regarded as
somewhat conservative. Table 7 lists some of the assumptions made in this study that might be conservative or
optimistic.
All in all, there are a number of uncertainties associated
with this study, and improvements in any areas would
always be favourable. It is also emphasized that a generic
FSA such as this should be considered as an on-going
process where the results are continuously updated
according to new knowledge, developments in technology,
environment and trading patterns, renements of underlying assumptions, etc. Nevertheless, in spite of the
subjectivity and inherit uncertainties, on a high level, the
results from the current study are believed to be meaningful
and robust for the world eet of LNG carriers.
Even though the risk assessment presented in this paper
is believed to be based on the best available information,
approaches and estimates, parts of the study should
undoubtedly be regarded as somewhat subjective. In some

areas of the analysis there have been sufcient statistical


data available to draw meaningful conclusions, whereas in
other areas no sources of information have been available.
In the latter areas, quantitative estimates have been based
on qualitative considerations and expert judgement,
although it is acknowledged that this is generally associated
with a certain degree of subjectivity [27]. However, this risk
analysis is modular by design and, if new knowledge
becomes available in any area, it should be rather straightforward to replace or update relevant modules accordingly.
It is therefore recommended that such amendments to the
risk models should be made when appropriate.
Uncertainties have been particularly salient in some
areas of this analysis, and it is suggested that further
studies should be carried out in order to bridge the gaps in
fundamental knowledge and available statistics pertaining
to hazards and risks related to LNG shipping. These are in
most cases represented towards the right-hand side of the
event trees, describing events in the later stages of a
scenario. In particular, it is suggested that further studies
related to the following parts of the risk model are
initiated: LNG hazard model, damage extent models,
survivability model, evacuation model and third-party

ARTICLE IN PRESS
1344

E. Vanem et al. / Reliability Engineering and System Safety 93 (2008) 13281344

model. These studies could be in the form of experiments,


calculations or simulations.
8. Conclusions and recommendations
The overall risk associated with LNG carriers was found
to be in the ALARP area. Thus, all risks should be made As
Low As Reasonable Practicable by implementing any costeffective risk reduction measures that may be identied.
Further work on identifying, selecting and assessing prospective risk control options in terms of cost effectiveness is
therefore a logical continuation of this risk analysis. These
issues will be addressed in subsequent SAFEDOR tasks.
Three generic accident scenarios are together responsible
for about 90% of the total risk related to LNG carriers, i.e.
collision, grounding and contact. These scenarios are
related in that they all describe a situation where an
LNG vessel is damaged because of an external impact
from, e.g., another vessel, oating objects, the sea oor,
submerged objects, the quay or other structures. Upon
closer investigation of the risk models associated with these
scenarios, four sub-models in particular stand out where
further risk reduction could be effective. These are the
accident frequency model, the cargo leakage frequency
model, the survivability model and the evacuation model.
Therefore, it is recommended that further efforts in the
next steps of this on-going FSA on LNG carriers focus on
measures related to:








Navigational safety
Manoeuvrability
Collision avoidance
Cargo protection
Damage stability
Evacuation arrangements.

The above list should by no means be considered as


exclusive, and possible cost-effective risk control options
related to, e.g., cargo handling and re protection should
also be investigated. However, it is believed that the most
cost-effective risk control options will be those addressing
the high-risk areas identied in this risk analysis.
Acknowledgements
The work reported in this paper has been carried out
under the SAFEDOR project, IP-516278, with partial
funding from the European Commission. The opinions
expressed are those of the authors and should not be
construed to represent the views of the SAFEDOR
partnership or that of DNV, IST, LMG Marin or Navantia.
The authors wish to acknowledge valuable contributions
to this study in the form of discussions, expert opinion,
management or general support from the following
individuals: R. Skjong, S. Valsgard and S.E. Jacobsen
(DNV) and C. Suedes Soares (IST).

References
[1] Christensen H, Hensel W, Perez de Lucas A, Sames PC, Skjong R,
Strang T, et al. SAFEDORrisk-based ship design, operation and
regulation. In: Proceedings of IMAM 05, 2005.
[2] Scheibach K, Noble P, Broman C. The next generation of large LNG
carriers. In: Proceedings of the 9th International Marine Design
Conference, 2006.
[3] Fang Q, Yang Z, Hu S, Wang J. Formal safety assessment and
application of the navigation simulators for preventing human error
in ship operations. J Mar Sci Appl 2005;4(3):512.
[4] Hu S, Fang Q, Xia H, Xi Y. Formal safety assessment based on
relative risks model in ship navigation. Reliab Eng Syst Saf 2007;
92(3):36977.
[5] Lee J-O, Yeo I-C, Yang Y-S. A trial application of FSA methodology
to the hatchway watertight integrity of bulk carriers. Mar Struct
2001;14(6):65167.
[6] Rosqvist T, Tuominen R. Qualication of formal safety assessment:
an exploratory study. Saf Sci 2004;42(2):99120.
[7] Soares CG, Teixeira AP. Risk assessment in maritime transportation.
Reliab Eng Syst Saf 2001;74(3):299309.
[8] Wang J. The current status and future aspects in formal ship safety
assessment. Saf Sci 2001;38(1):1930.
[9] Wang J, Foinikis P. Formal safety assessment of containerships. Mar
Policy 2001;25:14357.
[10] IMO. Guidelines for Formal Safety Assessment (FSA) for use in the
IMO rule-making process. MSC/Circ.1023-MEPC/Circ.392, IMO,
2002.
[11] IACS. Experience with Formal Safety Assessment at IMO. MSC 78/
19/1, IMO, 2004.
[12] Skjong R. Experience with the use of risk assessment in IMO. In:
Proceedings of the ESREL 2003, 2003.
[13] Le Trung B. The game, mem and alarp principles of safety. Rech
Transports Secur 2000;2000(68):4865.
[14] Melchers RE. On the ALARP approach to risk management. Reliab
Eng Syst Saf 2001;71(2):2018.
[15] Skjong R, Vanem E, Endresen . Risk evaluation criteria. SAFEDOR report D.4.5.2, 2005.
[16] Norway. Decision parameters including risk acceptance criteria.
MSC 72/16, IMO, 2000.
[17] Vanem E, stvik I, Antao P. Risk analysis of LNG tankers.
SAFEDOR report D.4.3.2, 2006.
[18] stvik I, Vanem E, Castello F. HAZID for LNG tankers. SAFEDOR report D.4.3.1, 2005.
[19] Olufsen O, Spouge J, Hovem L. The formal safety assessment methodology applied to the survival capability of passenger
ships. In: Proceedings of the RINA Passenger Ship Conference 2003,
2003.
[20] Skjong R, Vanem E. Damage stability evaluation in collision of bulk
carriers. In: Proceedings of the ICCGS 2004, 2004.
[21] Denmark, Germany, Norway, United Kingdom. Updated statistics
for extent of damagereport from the research project HARDER.
SLF 44/INF.11, IMO, 2001.
[22] Laubenstein L, Mains C, Jost A, Tagg R, Bjrneboe NK. In:
Proceedings of the ICCGS 2001, 2001.
[23] Apostolos P, Eleftheria E, Aimilia A, Seref A, Cantekin T, Severion
D, et al. Critical review of Aframax tankers incidents. In: Proceedings
of the ENSUS 2005, 2005.
[24] MCA. MCA Research ProjectFSA of shippingPhase 2Trial
application to HSC. Deliverable 9Summary report. UK: Maritime
and Coastguard Agency; 1998.
[25] Vanem E, Skjong R. Fire and evacuation risk assessment for
passenger ships. In: Proceedings of the Interam 2004, 2004.
[26] Hansen HL, Nielsen D, Frydenberg M. Occupational accidents
aboard merchant ships. Occup Environ Med 2002;59:8591.
[27] Skjong R, Wentworth BH. Expert judgement and risk perception. In:
Proceedings of the ISOPE-2001, 2001.