DESIGN FOR SAFETY

CONFERENCE 2016
Hamburg, 28-30 November

6th Conference on Design for Safety

2830 November 2016

Sponsored by:
MONDAY, 28 NOV 2016
08:30 Registration desk opens
09:00 Welcome
Dr Bjrn-Olaf BORTH, Country Chair and Member of Executive
Board, (DNV GL)
Introduction
Dr Pierre C SAMES, Group Technology and Research Director
(DNV GL)
Keynote
Prof Dracos VASSALOS, Chairman of the DfS International
Standing Committee
Session 1: Damage Stability Chair: Prof Apostolos Papanikolaou
09:35 Representing Military Behavior in Naval Ship Evacuation
Simulations including Flooding Damage Scenarios
David L.L. SICURO (Braz. Navy Research Inst.), Jos M.
VASCONCELLOS (UFRJ), Dracos VASSALOS (Univ Strathclyde)
10:00 Design for safety in a competitiveness perspective
Jose Jorge GARCIA AGIS, Ulrikke Brask BRANDT,
Per Olaf BRETT (Ulstein)
10:25 Designing for Damage Stability beyond Design Level
Prof Dracos Vassalos, Evangelos BOULOUGOURIS,
Donald PATTERSON (Univ Strathclyde), Markku KANERVA
10:50 On the Survivability of ROPAX and Cruise Ships:
A New Approach to Differences in Design
George DAFERMOS, Apostolos PAPANIKOLAOU (NTUA)
11:15 Coffee break
Session 2: Intact Stability 1 Chair: Prof Naoya Umeda
11:45 Performance based stability
Govinder Singh CHOPRA (SeaTech Solutions)
12:10 Study on the Correction of Wave Surge Forces to Improve
Surf-riding/Broaching Vulnerability Criteria Check Accuracy
Pei-Yuan FENG, She-Ming FAN (Maric), Liwei YU,
Ning MA (Shanghai JT Univ)
12:35 The Weather Criterion: Experimental Wind Tunnel Results
Arman ARIFFIN, Shuhaimi MANSOR,
Jean-Marc LAURENS (ENSTA Bretagne)
13:00 Lunch
Session 3: Intact Stability 2 Chair: Prof Naoya Umeda
14:00 Safety Level Required by the IMO Second Generation Intact
Stability Criteria for Ships under Dead Ship Conditions and
Parametric Roll Resonance Naho YAMASHITA, Naoya UMEDA,
Masahiro SAKAI (Osaka University)
14:25 Statistical Analysis on Parametric Roll Groups detected by
IR-HHT Method in Irregular Head Seas
Liwei YU, Ning MA (Shanghai Jiao Tong Univ),
Yoshiaki HIRAKAWA (YNU)
14:50 A Meta-Model For Risk Assessment of RoPax Capsizing as
an Alternative Way of Ship Safety Evaluation
Tomasz HINZ (Deltamarin), Przemysaw KRATA,
Jakub MONTEWKA (Gdynia Maritime University)
15:15 Coffee break
Session 4: Design for Safety 1 Chair: Prof Dracos Vassalos
15:45 Design for Safety in a Changing Climate
Elzbieta BITNER-GREGERSEN, Erik VANEM, Odin GRAMSTAD
(DNV GL)
16:10 Naval Auxiliaries, Safe in Design - Safe through Life
Loren ROBERTS, Andrew SMALLER (BMT Cadence)
16:35 Equivalent Level of Safety Approach Fire Safety
Daniel POVEL, Andres OTT (DNV GL)
17:00 Data Analytics for Marine Operations*
Hao WANG, Girts STRAZDINS,Guoyuan LI (NTNU Aalesund)
TUESDAY, 29 NOV 2016
09:30 eSAFE discussion
Session 5: Safe Operations Chair: Prof Yoshiho Ikeda
10:00 Applied Bayesian Network Model for Risk Assessment for
Ship Collision
Takeshi SHINODA (Kyushu Univ)
10:25 Shopera Manoeuvrability in Seaway
Vladimir SHIGUNOV (DNV GL)
10:50 Ship Optimization for Efficiency and Maneuverability in Adverse
Sea Conditions
George ZARAPHONITIS, Aphrodite KANELLOPOULOU,
Apostolos PAPANIKOLAOU (NTUA), Vladimir SHIGUNOV (DNV GL)
11:15 Coffee break
Session 6: Safe Operations II Chair: Takeshi Shinoda
11:45 Probabilistic Aspect on Minimum Propulsion Power Requirement
Issue under Adverse Weather Conditions
Shusuke OHIWA, Naoya UMEDA (Osaka University)
12:10 CFD Based Simulation of LNG Release During Bunkering
and Cargo Loading/Unloading Simultaneous Operations
of a Containership
Hongjun FAN, Kang CHENG, Shunping WU
(China Classification Society)
12:35 Subdivision Optimization of LNG Fueled RoPax Ship
Teemu MANDERBACKA, Pekka RUPONEN, Daniel LINDROTH,
Markus TOMPURI (NAPA)
13:00 Group picture in DNV GL Atrium or outside
Lunch
Session 7: Design for Safety 2 Chair: Prof Dracos Vassalos
14:00 Safety Culture - Beyond the Horizon but Still in View
Martin TOLAND (QinetiQ)
14:25 Challenges for Application of Risk Based Design Approaches
for Arctic and Antarctic Operations
Pentti KUJALA, Jorma KMRINEN, Mikko SUOMINEN (Aalto University)
14:50 Coffee break
Session 8: Man Machine Interfaces Chair: Prof Ning Ma
15:20 Accident Scenario-based Rapid and Interactive Damage Control
System Using Coded Shortcut Keys
Hee Jin KANG, Dongkon LEE, Jin CHOI (KRISO)
15:45 Enhancing SAR Communication and Decision Making
using Vessel TRIAGE: Concept and Developments
Floris GOERLANDT, Jori NORDSTRM,
Pekka RUPONEN (Aalto University)
16:10 SEAHORSE Procedure Improvement System
Rafet E. KURT, Volkan ARSLAN, Emma COMRIE,
Hassan KHALID, Osman TURAN (University of Strathclyde)
WEDNESDAY, 30 NOV 2016
Session 9: Autonomous Systems Chair: Prof Volker Bertram
10:00 Autonomous Shipping A Concept Design for an
Autonomous Firefighting and Rescue Vessel
Simon PULLIN, Samantha HILL (QinetiQ)
10:25 A New Concept for Avoiding Collision of Automatically
Operating Ships and its Evaluation
Yoshiho IKEDA, Kana YOSHIDA, Nobuyuki SHIMIZU, Keiichi
HIRAYAMA, Masakazu ARIMA (Osaka Pref Univ)
10:50 Safety Aspects of Autonomous Ships
Sauli ELORANTA, Andy WHITEHEAD (Rolls-Royce Marine)
11:15 Coffee break
Session 10: Data Analytics Chair: Dr Pierre Sames
11:45 A Method for Describing Ocean Environments for Ship
Assessment
Eirik EISINGER, Jens Bloch HELMERS,
Gaute STORHAUG (DNV GL)
12:10 Hull Monitoring Closing the Gap between Design and
Operation
Gaute STORHAUG, Adrian KAHL (DNV GL)
13:00 Lunch
Workshop: Digital Safety Chair: Dr Pierre C Sames
14:00 Setting the scene
Dr Pierre C SAMES (DNV GL)
14:15 PSC and survey data to understand ship condition
Insight from combining data sources
Andreas MELCHNER (DNV GL)
14:30 Mining casualty data to identify root causes
Insight from recent FSA on container vessels
Rainer HAMANN (DNV GL)
14:45 Fleet operation center how it works to improve safety
Keith DOWDS (Carnival)
15:15 Dynamic Barrier Management What we can learn
from offshore industry
Frank B PEDERSEN (DNV GL)
15:30 Panel discussion
All speakers
16:00 Workshop Conclusions
Dr Pierre C SAMES (DNV GL)
Conference Dinner
18:30
* has been moved from session 10
 Index of Papers

Hongjun Fan, Kang Cheng, Shunping Wu 5

CFD Based Simulation of LNG Release During Bunkering and Cargo Loading/Unloading Simultaneous
Operations of a Containership

Floris Goerlandt, Jori Nordstrm, Pekka Ruponen 11

Enhancing SAR Communication and Decision Making using Vessel TRIAGE: Concept and Developments

Pei-yuan Feng, She-ming Fan, Li-wei Yu, Ning Ma 17

Study on the Correction of Wave Surge Forces to Improve Surf-riding/Broaching Vulnerability Criteria
Check Accuracy

George Dafermos, Apostolos Papanikolaou 22

On the Survivability of ROPAX and Cruise Ships - A New Approach to Differences in Design

Eirik Eisinger, Jens Bloch Helmers, Gaute Storhaug 34

A Method for Describing Ocean Environments for Ship Assessment

Elena-Gratiela Robe-Voinea, Alexandru Pintilie, Raluca Vernic 40

A Short Ship Design Risk Analysis using the Monte Carlo Method

Arman Ariffin, Shuhaimi Mansor, Jean-Marc Laurens 46

The Weather Criterion: Experimental Wind Tunnel Results

Gaute Storhaug, Adrian Kahl 53

Hull Monitoring Closing the Gap between the Design and Operation

Hee Jin Kang, Dongkon Lee, Jin Choi 61

Accident Scenario-based Rapid and Interactive Damage Control System using Coded Shortcut Keys

Govinder Singh Chopra 68

Performance Based Stability

George Zaraphonitis, Aphrodite Kanellopoulou, Apostolos Papanikolaou,Vladimir Shigunov 75

Ship Optimization for Efficiency and Maneuverability in Adverse Sea Conditions

Liwei Yu, Ning Ma, Yoshiaki Hirakawa 85

Statistical Analysis on Parametric Roll Groups Detected by IR-HHT Method in Irregular Head Seas

Teemu Manderbacka, Pekka Ruponen, Daniel Lindroth, Markus Tompuri 90

Subdivision Optimization of LNG Fueled RoPax Ship

David L.L. Sicuro, Jos Marcio Vasconcellos, Dracos Vassalos 98

Representing Military Behavior in Naval Ship Evacuation Simulations including Flooding Damage
Scenarios

Loren Roberts, Andy Smaller 111

Naval Auxiliaries, Safe In Design - Safe Through Life

Shusuke Ohiwa, Naoya Umeda 127

Probabilistic Aspect on Minimum Propulsion Power Requirement Issue under Adverse Weather
Conditions

2
Tomasz Hinz, Przemysaw Krata, Jakub Montewka 132
A Meta-model for Risk Assessment of RoPax Capsizing as an Alternative Way of Ship Safety Evaluation

Yisi Liu, Xiyuan Hou, Olga Sourina, Dimitrios Konovessis, Nieves Endrina Snchez, Gopala Krishnan 141
Human Factors Evaluation in Maritime Simulator-based Assessment: A Case Study

Jose Jorge Garcia Agis, Ulrikke Brask Brandt, Per Olaf Brett 146
Design for Safety in a Competitiveness Perspective

Naho Yamashita, Naoya Umeda, Masahiro Sakai 153

Safety Level Required by the IMO Second Generation Intact Stability Criteria for Ships under Dead Ship
Conditions and Parametric Roll Resonance

Vladimir Shigunov 158

SHOPERA Manoeuvrability in Seaway

Sauli Eloranta, Andy Whitehead 168

Safety Aspects of Autonomous Ships

Daniel Povel, Andreas Ott 176

Equivalent Level of Safety Approach Fire Safety

Elzbieta Bitner-Gregersen, Erik Vanem, Odin Gramstad 183

Design for Safety in a Changing Climate

Pentti Kujala, Joma Kmrinen, Mikko Suominen 196

Challenges for Application of Risk Based Design Approaches for Arctic and Antarctic Operations

Rafet Emek Kurt, Volkan Arslan, Emma Comrie, Hassan Khalid, Osman Turan 204
SEAHORSE Procedure Improvement System

Martin Toland 214

Safety Culture - Beyond the Horizon but Still in View

Simon Pullin, Samantha Hill 219

Autonomous Shipping A Concept Design for an Autonomous Firefighting and Rescue Vessel

Dracos Vassalos, Evangelos Boulougouris, Donald Paterson, Markku Kanerva 225

Designing for Damage Stability beyond Design Level

Hao Wang, Girts Strazdins, Guoyuan Li 232

Data Analytics for Marine Operations

Yoshiho Ikeda, Kana Yoshida, Nobuyuki Shimizu, Keiichi Hirayama, Masakazu Arima 237
A New Concept for Avoiding Collision of Automatically Operating Ships and its Evaluation

Takeshi Shinoda, Ryoma, Nasu, Koji Uru 243

Applied Bayesian Network Model for Risk Assessment for Ship Collision

3
4
 CFD Based Simulation of LNG Release During Bunkering
and Cargo Loading/Unloading Simultaneous Operations of
a Containership

FAN Hongjun, China Classification Society, hjfan@ccs.org.cn

CHENG Kang, China Classification Society, kcheng@ccs.org.cn
WU Shunping, China Classification Society, sp_wu@ccs.org.cn

ABSTRACT stations are under developing, and will be issued in 2016,

In order to reduce waiting time in port for large LNG
fueled ships, it is suggested that LNG ship-to-ship (STS)
bunkering and cargo loading/unloading should be car-
ried out simultaneously. This study investigated the
safety zone of a LNG bunkering vessel with 10,000 cubic
meters capacity transferring LNG fuel to a LNG fueled
18000TEU containership. Four LNG leakage scenarios
were identified based on failure frequencies analysis of
piping systems and severity of consequence. The
three-dimensional CFD software FLACS was adopted
to calculate flammable cloud dispersion after LNG
leakage. As a result, we obtained a rectangular danger
zone; the domain outside of this zone can be considered
as safe. The danger zone of LNG STS bunkering and
cargo loading/unloading simultaneous operations (SI-
MOPS) varies with different results for different designs
and operation locations. Due to high frequencies and
severe consequences involved, two typical scenarios
cannot be ignored in similar studies: the leakage of an
LNG hose and natural gas release from a bunkering
tanks safety relief valve during bunkering.
INTRODUCTION
LNG as marine fuel is a leading alternative for meeting
current and future more stringent air emission requirements.
To date, there are more than 100 seagoing LNG fueled ves-
sels operating mainly in the North Sea and Baltic Sea. In
China, there are more than 600 inland LNG fueled vessels
are under building or on order (Xu, 2015).
The use of LNG requires the development of bunkering
infrastructures and a complete regulatory framework. The
feasible bunkering modes are as follows (China MSA,
2013): 1) Truck to ship; 2) Onshore station to ship; 3) Pon-
toon to ship; 4) Potable fuel tanks; 5) Ship to ship; 6) Off-
shore unit to ship. In terms of the regulatory framework, in
China, the Chinese government issued a regulation for LNG
bunkering pontoons in 2014, the regulation for LNG bun-
kering vessels and Design code for onshore LNG bunkering
the complete regulatory systems will promote the formation
of waterborne LNG supply chains.
However, due to the complexity of LNG bunkering, in
some cases there is a need to provide personalized solutions,
such as relative motion limit between two ships, fender
arrangement, allowed weather conditions, safety operations
zone, etc.
For large ships with large amount LNG fuel, simulta-
neous operations (SIMOPS) of LNG ship to ship (STS)
bunkering and cargo loading/unloading could reduce time
in port. If loading/unloading operations were in the flam-
mable cloud zone, there would be a risk of fire. In this study,
the SIMOPS safety zone of a LNG bunkering vessel with
10,000 cubic meters capacity transferring LNG fuel to a
LNG fueled 18000 TEU containership was addressed.
THE ACCIDENTS OF LNG BUNKERING IN THE
HISTORY AND ENLIGHTENMENT
The accidents in the history are warning significances in
LNG bunkering practice. There is a short history of only 15
years after LNG bunkering appeared in marine sector, only
two accidents in Norway have been reported (IMO, 2015).
In Risavika Harbour, during bunkering of MS Ber-
gensfjord on 9 May 2014, there was a leak in the Quick
Release Coupling located in the bunkering station on board
the ship. The investigation concluded that approximately
130 kg LNG were released. The incident did not result in
personal injuries or damage to property.
The other accident occurred in Moskenes. During bun-
kering of MF Landegode on 13 June 2014, the ships stern
moved away from the quay. The hose was stretched, and a
davit arm was damaged. The ferry managed to return to the
quay, and there was no leak. The bunkering operation con-
tinued. The davit arm broke later that day. The incident did
not result in any other damage to property, and no person
was injured.
The above two accidents shed some light on the safety
of LNG bunkering: a safety zone during bunkering should
be established; LNG drip trays should be installed; hoses
should not be restricted; bunkering station should be con-

5
tinuously monitored; ESD should be tested before every In the field of LNG risk assessment, defining LNG
bunkering operation; the breakaway couplings should be leakage positions and hole sizes based on failure frequen-
installed; mooring procedure should be focus on; Enhance cies analysis is becoming an international consensus. As an
personal training, etc. These safety measures have been example, the U.S. Federal Energy Regulatory Commission
considered in the project in this study. (FERC) issued guidance for the selection of leak sizes
based upon failure frequencies for piping systems; in par-
THE LNG FUELED SHIP AND THE BUNKERING ticular, scenarios with failure frequencies greater than
VESSEL 3E-05 per year must be considered (McInerney et al., 2013).
This criterion is used to define hazard scenarios in this
The LNG fueled ship is an 18000 TEU containership study.
(Fig. 1 and Table 1) under conceptual design according to
CCS (China Classification Society) Rules for natural gas Table 3: Nominal FERC failure frequencies per unit length
fueled ships. for piping
A 10,000 cubic meters LNG bunkering vessel (Table 2) Failure frequency (/yr/m)
with a new type bunkering arm (combination of davit, pipes Pipe Diame-
and hoses) was assumed to provide bunkering service for Catastrophic
ter Dhole=1/3dpipe Dhole=25mm
the containership in this study. Rupture
d pipe50mm 10E-07 - 50E-07
50mmd pipe
5E-07 - 20E-07
149mm
150mmd
2E-07 4E-07 7E-07
pipe299mm

Fig. 1: Basic arrangement of the LNG fueled containership Table 4: Failure frequencies and operation times of LNG
hose
Table 1: Basic parameters of the LNG fueled containership Failure frequen- 4.0E-7 hole diameter:10%D50mm ,
Length, over- ~ 403.40 m Depth, 30.20 m cy (/hour) D: hose diameter
all moulded (Refer to Wolting et al., 2013: Modified
Length be- ~ 384.00 m Design draft, 14.00 m failure frequency according to TNO
tween perpen- moulded purple book)
diculars Operation time 16 h (estimated)
Breadth, 58.50 m LNG storage 10,000 m3 every bunkering
moulded tank Bunkering times 20 times
Service speed 20 kn Main engine 56800 kW every year
power Total operation 320 h
time every year
Table 2: Basic parameters of the LNG bunkering vessel
Length, overall ~113.50 m Depth, 5.80 m Table 5: OGP manual valve failure frequencies (per valve
moulded year)
Length between ~106.00 m Designed 12.00 m DN50 DN150 DN300
perpendiculars draft, 13 2.0E-05 3.1E-05 4.3E-05
moulded 310 7.7E-06 1.2E-05 1.7E-05
Breadth, 20.30 m LNG storage 10,000 m3 Hole diameter 1050 4.9E-06 4.7E-06 6.5E-06
moulded capacity (mm) 50 -- 2.4E-06 1.2E-06
Max. bunkering 600 m3/h Type of Membrane 150
rate LNG tank
150 -- -- 1.7E-06
LNG LEAKAGE SCENARIOS DEFINITION BASED ON
Table 6: OGP actuated valve failure frequencies (per valve
FAILURE FREQUENCIES ANALYSIS
year)
In terms of LNG leakage during STS operations, the DN50 DN150 DN300
main focus is LNG bunkering system related receiving ship 13 2.4E-04 2.2E-04 2.1E-04
and bunkering vessel, besides the bunkering system itself, 310 7.3E-05 6.6E-05 6.3E-05
the valves for isolating other systems are involved as well. Hole diameter 1050 3.0E-05 1.9E-05 1.8E-05
As the safety zone assessment is the goal in this study, only (mm) 50 -- 8.6E-06 2.4E-06
LNG release sources located on open areas were considered. 150
Based on historical data, the flammable gas dispersion after
150 -- -- 6.0E-06
gaseous phase leakage will not influence the safety zone
significantly. Therefore, the liquid phase leakage is the only
situation considered.

6
 Currently, there are some databases with failure fre- Table 8. Hazard scenarios of LNG leakage
quencies for piping systems, such as FERC (McInerney et Hazard Description Cumulative Cumulative
al., 2013), OGP (OGP, 2010), TNO (VROM, 2005), HSE scenarios time of LNG LNG
(HSE, 2012). Failure frequencies data suitable for this pro- leakage leakage
ject are shown in Tables 3, 5 and 6. Table 4 shows the mod- Scenario Hose in bunker- 90 s 1082.7 kg
ified failure frequency and operation times of LNG hose. 1 ing arm (liquid),
Table 7 shows the failure frequencies calculation of hole diameter is
bunkering piping systems which are located on open area. 50 mm
Due to lack of failure date of DN200 valve, the data of Scenario Actuated valve 90 s 91.3 kg
DN150 valve was used conservatively. The failure fre- 2 (liquid) in bun-
quency of the LNG hose is the highest, Table 7. kering station of
receiving ship,
Table 7: Failure frequencies of bunkering piping sys- hole diameter is
tems on open area 10 mm
N Type Length or Specification Failure frequencies Scenario Actuated valve 90 s 8.2 kg
o. amount 3 (liquid) or man-
1 Pipe 25m DN 200 Catastrophic Rupture ual valve (liquid)
(liquid) 5.0E-06 in bunkering
Hole diameter 67 mm: station of re-
1.0E-05 ceiving ship,
Hole diameter 25mm: hole diameter is
1.75E-05 3 mm
2 Pipe in 14m DN 200 Catastrophic Rup- Scenario LNG release 2.5 s (Acc. 15.0 kg
bunker- ture:2.8E-06 4 from safety relief to reseating
ing arm Hole diameter 67mm valve of bunker- pressure of
(liquid) 5.6E-06 ing tank safety valve,
Hole diameter 25mm: see 4.2.3)
9.8E-06
3 Hose in 20m DN 200 Hole diameter CFD ANALYSIS OF LNG FLAMMABLE CLOUD
bunker- 50mm:1.28E-04 DISPERSION
ing arm
(liquid) The 3D CFD software FLACS was used to simulate
4 Actuated 8 sets DN 200 Hole diameter flammable cloud dispersion after LNG leakage. The soft-
valve 1~3mm2.2E-04 ware is an authoritative gas dispersion tool worldwide. It
(liquid) Hole diameter has received approval from the US Pipeline and Hazardous
Materials Safety Administration (PHMSA) for LNG vapor
3~10mm6.6E-05
dispersion modeling scenarios according to federal regula-
Hole diameter
tions 49 CFR 193.2059(a). The environment conditions of
10~50mm1.9E-05 STS operations site are shown in Table 9, and these data are
Hole diameter the input of the CFD analysis.
50~150mm8.6E-06
5 Manual 1 set DN 200 Hole diameter Table 9: Environment conditions of STS operations site
valve 1~3mm3.1E-05 No. Type Parameter
(liquid) Hole diameter 1 Annual mean wind speed 3.15m/s
3~10mm1.2E-05 2 Annual mean temperature 16
Hole diameter 3 Atmospheric pressure 101325Pa
10~50mm4.7E-06 4 Relative humidity 75%
Hole diameter 5 Solar radiation intensity 583w/m2
50~150mm2.4E-06 6 Atmospheric stability D

Hazard scenario 1 ~ scenario 3 identified are listed in Establishment of 3D CFD calculating model
Table 8 according to the criteria (failure frequencies greater
than 3E-05 per year). In addition, two ships (receiving and Fig. 2 shows the calculating model. As the pipe outlet
supplying ship) are connected during LNG bunkering. Nat- type of the vent mast influence gas release and dispersion
ural gas release from the vent mast of the supplying ship significantly (Wu, 2016), the real pipe outlet was estab-
may disperse flammable gas to the receiving ship, bringing lished accurately in the 3D model, Fig. 3.
additional risk. Therefore, scenario 4 is identified, Table 8.
According to experience, ESD total time of 90 s is assumed
to consider LNG leakage continuously (60 s for detection
and initiation, 30 s for isolation).

7
 unloading are carried out simultaneously, LNG
leakage occurs at the beginning of containers un-
loading. Fig. 5 shows gas dispersion after the LNG
was contained by the drip trays.

a) Top view

a) North wind

b) Perspective view
Fig. 2: 3D CFD calculating model

b) East wind

Fig. 3: Pipe outlet model of bunkering vessels vent mast

Calculations and analysis

Half of the lower flammability limit (LFL) (volume c) West wind

concentration value 2.5 percent) of natural gas was accept- Fig. 4: Flammable gas dispersion range for LNG leakage in
ed as the outermost boundary of the flammable zone. bunkering station (volume concentration: 2.5%~15%)

LNG leakage from hose in bunkering arm

LNG leakage from hose in bunkering arm corresponds
to Scenario 1 in Table 8. For this scenario, flammable gas
would disperse a large range because of a relative large
leakage volume, therefore, the dispersion would be sensi-
tive to obstacles surrounding. Accordingly, the two situa-
tions (unloaded and loaded) are analyzed to compare how
the containers (obstacles) influence the gas dispersion. Fig. 5: Flammable gas dispersion range when LNG leakage
When performing calculations, three wind directions, east, occurs in bunkering station (north wind, volume concentra-
west and north, are considered. (South wind is ignored be- tion: 2.5%~15%)
cause of the LNG fueled ship would shelter the gas from
wind, the wind directions are shown in Fig. 2.)

(1) Unloaded situation

In this situation, LNG bunkering and containers
loading are carried out simultaneously. LNG leak-
age occurs at the beginning of containers loading.
Fig.4 shows gas dispersion after the LNG was
contained by drip trays (1.2m1.2m0.2m). The a) East wind
results calculated show that in north wind, flam-
mable gas flowed over the gunwale and spread
27.5 m along the ship transverse direction, Fig.4;
in east and west winds, flammable gas was always
below the gunwale, therefore, there is no influence
on the safety of loading. When LNG leak into the
water rather than in the bunkering station, there is
no harm to the loading operations for flammable
b) West wind
gas was always below the gunwale.
Fig. 6: Flammable gas dispersion scope when LNG leak
(2) Loaded situation
into the water (volume concentration: 2.5%~15%)
In this situation, LNG bunkering and containers

8
 The results calculated show that (Figs.5 and 6), in the
north wind situation, flammable gas flowed over the gun-
wale and spread 14.0 m along the ship transverse direction;
in the east and west wind situation, flammable gas was al-
ways below the gunwale, therefore, there is no influence on
the safety of loading.
LNG leakage from valves
Scenario 2 and scenario 3 are all regarding LNG leak-
age from valves in the bunkering station. LNG would be
contained by the drip trays.
Calculations are carried out only in north wind situation
(In east and west wind situations, the hazards are signifi-
cantly less by trial calculation). The results (Fig.7) demon-
strate that the range of gas dispersion after valve leakage is
very narrow, which does not affect the loading and unload-
ing operations.
a) Actuated valve leak
b) Manual valve leak
Fig. 7: Flammable gas dispersion range after LNG leak
from valves (north wind, volume concentration: 2.5%
~15%)
LNG release from safety relief valve of bunkering tank
LNG bunkering tanks safety valve set and reseat pres-
sures are 0.25 bar and 0.23 bar respectively, and its effec-
tive flow area is 31400 mm2 in this study, the gas release
rate from safety valve was compared by two ways. The
result is 6.76 kg/s by gas flow equation (VROM, 2015), and
the result obtained from FLACS leak wizard tool is shown
in Table 10, the total release time is 2.5 s, the total release
mass is 15 kg, therefore, it is reasonable to obtain natural
gas release parameters based on FLACS software.
The distance from vent outlet to the farthest boundary of
flammable cloud is 10.1 m (Fig.8), however, the distance
from the bunkering vessels vent outlet to the container-
ships deck edge is 12.7 m during bunkering operations,
therefore, ignition hazard due to natural gas release from
safety relief valve can be ignored.
Table 10: Natural gas release rates from safety valve
Time(s) Release rates(kg/s)
0 6.7621
0.5 6.5937
1 6.4242
1.5 6.3364
2 6.0762
Fig. 8: Flammable gas dispersion range after safety valve
takeoff (time: 2s, 3s, and 4.56s)
Safety zone
Based on the above calculations, initial rectangle dan-
gerous zone (27.5 m84.0 m) is obtained by dangerous dis-
tances envelopment, conservatively, the final dangerous
zone (Fig.9) is a new rectangle (41.3 m126.0 m) which is
obtained from every initial rectangle length is multiplied by
1.5. Outside the range of the final dangerous zone can be
defined as safety zone of simultaneous operations.
Fig. 9: Dangerous zone
CONCLUSIONS
The main conclusions obtained in this study are summa-
rized below:
1. Due to LNG leakage and gas dispersion are influenced
by ship design and environment conditions signifi-
cantly, the safety zone of LNG ship to ship (STS)
bunkering and cargo loading/unloading simultaneous
operations (SIMOPS) cannot keep the same, there are
different results for different designs and operation
sites;
2. Because of the failure frequency of LNG hose is high
and flammable gas dispersion range is large after nat-
ural gas released from safety relief valve, the scenarios
of LNG hose rupture and natural gas released from
bunkering tanks safety relief valve cannot be ignored
in similar study.
REFERENCES
China MSA Jiangsu Branch, China Classification Society.
(2013), Research report of LNG bunkering modes and
bunkering stations. (In Chinese)
HSE (2012), Failure Rate and Event Data for use within
Risk Assessments.
9
IMO MSC 95/INF.17 (2015), Information on incidents
during bunkering of LNG.

McInerney, E.H., Hart, R.J. and Kytmaa, H.K. (2013),

New Quantitative Risk Criteria for U.S. LNG Facilities,
9th Global Congress on Process Safety.

OGP (2010), Risk Assessment Data Directory Report No.

4341 Process release frequencies.

VROM (2005), Publication Series on Dangerous Sub-

stances (PGS 3) Guidelines for quantitative risk assess-
ment (Purple book).

VROM (2015), Publicatiereeks Gevaarlijke Stoffen 2

Methods for the Calculation of Physical Effects, Due to
releases of hazardous materials (liquids and gases).

Wolting, A.G. and Vijgen, L.J. (2013), Development of a

QRA method to calculate the risks generated by Lique-
fied Natural Gas (LNG) filling stations for road trucks,
9th Global Congress on Process Safety.

Wu, S.P., Fan, H.J., Luo, X.F. et al. (2016), Evaluation

Study for Vent Mast Height of Inland LNG Fuelled
Ships, Ship Engineering, 2016(01), pp 59-63. (In Chi-
nese)

Xu, J.Y., Fan, H.J., Wu, S.P. et al. (2015), Research on

LNG Ship to Ship (STS) Bunkering Operations, Ship
Engineering, 2015(01), pp 7-10+82. (In Chinese)

10
 Enhancing SAR Communication and Decision Making
using Vessel TRIAGE: Concept and Developments
Floris Goerlandt
Jori Nordstrm
Pekka Ruponen
Aalto University, Marine Technology, floris.goerlandt@aalto.fi
The Finnish Lifeboat Institution, jori.nordstrom@meripelastus.fi
NAPA Ltd, pekka.ruponen@napa.fi

ABSTRACT studies on RoPax vessels, there are significant limits re-

Efficient response to maritime incidents and accidents re-
quires good communication processes and situation awareness
by all involved parties, in particular between the Search and
Rescue (SAR) response operators and the crew of the dis-
tressed vessel. This paper discusses the Vessel TRIAGE
method, a newly developed tool to assess the safety status of a
vessel involved in a maritime accident or incident. The focus of
this the assessment of the vessels condition in terms of its ca-
pability to provide a safe environment for the people onboard,
as well as for response crews potentially assigned to the scene.
The method uses a relatively simple categorization system
based on a number of threat factors, and results in a categori-
zation in one of four safety levels. The levels can be easily
communicated by all involved parties to achieve a shared situ-
ational awareness, enabling enhanced decision-making. This
paper furthermore gives a brief overview of industrial devel-
opments related to Vessel TRIAGE, showing the methods
potential for innovative technical solutions supporting SAR
communication. Finally, further testing needs of the Vessel
TRIAGE method and stakeholder concerns about its imple-
mentation are outlined, based on discussions at IMOs
Sub-Committee on Navigation, Communication and Search
and Rescue. This provides a path for further envisaged work
for implementing the method in practice.
INTRODUCTION
Maritime accidents, particularly those involving cruise and
RoPax vessels and tankers, can have serious consequences
in terms of human casualties, economic losses, and envi-
ronmental damage. Whereas the overall safety performance
of the global maritime fleet has improved significantly over
the last decades (Eliopoulou et al., 2016), its performance is
still significantly lower than other transportation modes.
While measures to prevent accidents are in principle most
desirable, for low probability/high consequence events in
complex socio-technical systems, it is essential to also fo-
cus on strategies to reduce the disaster potential. Technical
design solutions providing inherently safe designs are one
feasible approach to achieve this, but may also lead to
higher economic costs (Klanac, 2011). Based on extensive
garding the extent to which the survivability of current ship
design concepts can be improved in relation to realistic
impact conditions (Hogstrm, 2012).
An important strategy to reduce the disaster potential of
maritime accidents is to enhance post-accident situational
awareness and related decision making. Several accident
investigations (MAIB, 2009; MIT, 2013) point to the
crews uncertainty about the damage extent, which hampers
effective communication both onboard and towards Search
and Rescue (SAR) response operators. This easily leads to
loss of time, during which the event can further escalate,
both in terms of accident progression itself (e.g. more ex-
tensive flooding, larger fires) but also in terms of increased
human casualties. The importance of shared situational
awareness and the quality of information provided to the
involved parties has been identified as a critical feature of
successful emergency response Seppnen and Virrantaus
(2015).
Improving current tools and practices for post-accident
vessel safety assessment and related SAR decision-making
is an important aspect of improving maritime safety in
practice. The emergency phase classification, based on the
SAR Convention, describes the urgency of response and the
reliability of the information by dividing emergencies in
uncertainty, alert and distress phases. However, this method
does not assess the safety level of the ship itself, which
nonetheless is crucial information for onboard contingency
measures and for effective SAR response. In practice, this
means for instance that a man-over-board situation and an
out-of-control fire onboard a vessel would both be classi-
fied as a distress phase, while it is clear that different re-
sponse actions would be needed and the situations differ
substantially in terms of the safety of the persons onboard
the vessel.
Recently, a new method has been developed to improve the
current situation regarding the assessment of the vessels
safety status in maritime accidents and incidents, to im-
prove communication between the involved stakeholders
and to facilitate related SAR response decision making: the
vessel TRIAGE method. This paper first provides an over-
view of this method and then outlines recent technical de-

11
velopments to support the method. Finally, further testing
needs of the Vessel TRIAGE method and stakeholder con-
cerns about its implementation are outlined, partly based on
discussions at IMOs Sub-Committee on Navigation,
Communication and Search and Rescue.
VESSEL TRIAGE METHOD
The Vessel TRIAGE method is a new tool for assessing and
communicating the safety status of vessels in maritime ac-
cidents and incidents, intended to be used both by the
onboard crew and by emergency responders. The focus is
on assessing whether, and to what extent, the subject vessel
can provide a safe environment for the people onboard,
which is linked to different operational focuses for SAR
response actions. Taking a broad view on SAR activities,
the decision to leave persons onboard the vessel is also a
viable option practical SAR operations, making the Vessel
TRIAGE method not only a matter of post-accident risk
assessment but one of SAR planning and operations.
The method is inspired by similar categorization systems
applied in medical emergencies, see e.g. (Azeredo et al.,
2015). It aims to provide a fast but well-justified operation-
al safety status of the vessel, which is then used to com-
municate to the various stakeholders: the vessels onboard
crew, SAR responders, the vessels shipping company
shore personnel, crew of other vessels and salvage compa-
nies, and others. As such, the method can be used to proac-
tively assess the situation and facilitate ensuing decision
making, limiting the further escalation potential of the
event.
The Vessel TRIAGE method consists of three main parts:
the definitions of the vessel safety categories, the corre-
sponding threat factor matrix and a practical form to assess
Fig. 1: Vessel TRIAGE categories, abridged from RAJA et al. (2015)
The Vessel TRIAGE threat factor matrix
The basis for the assessment of the vessel TRIAGE catego-
ry is a threat factor matrix. This matrix lists the possible
threats to the vessels safety, including flooding, list-
ing/decrease of stability, decreased maneuverability,
blackout, fire / explosion and danger posed by hazardous
substances. The matrix includes a general description for
each threat factor, which indicates to the assessor to which
severity level the ongoing situation onboard the vessel cor-
responds, see Figure 2 where a part of the matrix is shown
for the flooding, listing/loss of stability and black-out aspects of the situation, aiming to make the assessment as
threat factors.
Depending on the threat factor, there are either three or four
degrees of severity, linked to the general description of the
vessels safety status of Fig. 1. For instance, blackout can-
12
the vessels status in practice. These are briefly introduced
below; for a more elaborate description of the method, its
development process and preliminary evaluation, the reader
is referred to RAJA et al. (2015) and Nordstrm et al.
(2016). A categorization of operational focuses for SAR
response organizations in line with the vessel safety catego-
ries has also been developed, but is outside the scope of the
current paper.
The Vessel TRIAGE categories
The vessel TRIAGE method applies four categories for the
safety status of a vessel, each of which is given a color code.
Black represents the most unsafe conditions, where the
vessel no longer provides any safety for the people onboard.
Green represents situations in which the vessel safety is
least compromised (acknowledging that, in case no accident
has occurred to the vessel, none of the categories should be
applied). The yellow and red categories represent the most
significant differences regarding the vessels safety status.
In the yellow category, there is a risk that the situation will
get worse, implying that it is still safe for the people to
remain onboard, but it considered possible that the situation
may change. In the red category, the level of safety has
significantly worsened or will worsen, which implies that
the safety of the persons onboard is severely threatened,
either immediately or in the near future. Figure 1 shows the
color codes and their basic definitions. A general descrip-
tion of the conditions onboard has also been made for better
delineating the types of situations covered by these catego-
ries; see RAJA et al. (2015) and Nordstrm et al. (2016).
The methods qualitative nature is intentional, as real-life
decision making requires tools, which are intuitively clear
to the users (Fuchs et al., 2015).
not by itself lead to the worst ship safety level (black cate-
gory). In contrast, stability can be decreased to such an ex-
tent that evacuation operations are no longer possible or
that the ship has capsized or sunk.
In the Vessel TRIAGE method, the categorization of the
severity level for each threat factor is a judgment based on
the best available information at the time of the assessment.
Information can be gathered from diverse systems and
sources, and a list of questions has been proposed alongside
the general description of the severity level of the threat
factors, Nordstrm et al. (2016). The focus is on observable
objective as possible.
 Fig. 2: Vessel TRIAGE threat factor matrix, abridged from RAJA et al. (2015)
Practical application: the Vessel TRIAGE assessment form
The practical application of the vessel TRIAGE method is
supported by a specific, easy-to-use paper form, which
could be implemented e.g. in the onboard copy of the
IAMSAR manual and/or in onboard and shore-side safety
management systems. An example of such a form can be
found in RAJA et al. (2015). The method for assessing the
overall Vessel TRIAGE vessel safety category takes fol-
lowing steps.
First, a number of basic information fields are filled out,
containing basic facts such as the date and time of the acci-
dent occurrence, the vessel type, number of crewmembers
and passengers. Second, it is assessed for each threat factor
whether or not has materialized. Apart from the options
yes or no, the option not known can be used in cases of information, it is beneficial to make use of technological
where it is too time consuming to get a status report on the
specific threat factor. Third, the severity of the materialized
threat factors is assessed, based on the information about
the situation onboard, using the questions appropriate to the
factors. The information is compared with the general de-
scription of the severity level of the threat factors, such as
in Fig. 2. A judgment is made about the severity level, tak-
ing the stance that in case of doubt, the more severe option
is selected. Fourth, an assessment of the crew capabilities
and the weather conditions is made, which could e.g. hinder
the performance of measures which would, under more
favorable conditions, stabilize the vessels safety status.
These are treated separately from the treat factors as such,
because they affect the entire vessels safety status. Fifth
and finally, the overall Vessel TRIAGE category is deter-
mined. Here, the basic rule is applied that the overall vessel
safety status should be at least as high as the severity level
of the highest realized threat factor. It may be judged higher,
when taking the crew capabilities and weather conditions
into account.
In typical maritime accident scenarios, the vessels safety
status typically changes over time, depending for instance
on mitigating actions taken by the crewmembers. Hence, an
integral part of the application of the Vessel TRIAGE
method is to periodically repeat the assessment, to ensure
that appropriate SAR actions are taken.
RELATED TECHNICAL DEVELOPMENTS
While the Vessel TRIAGE method has only been recently
developed (RAJA et al., 2015), the method has already
been taken as a basis for representing information from
onboard decision support systems, in particular for passen-
ger vessels and related to the threat factors flooding and
listing/decrease of stability. Such technological develop-
ments are considered a valuable contribution to the imple-
mentation and utility of the Vessel TRIAGE method in
practical settings. While the overall method is qualitative
and is intended to be flexible to account for various sources
support for classifying the vessels safety status. Arguably,
advanced progressive flooding software can provide better
information about the predicted survivability than crew
members can obtain from checklists of static stability soft-
ware, which is the current industry standard (MAIB, 2015;
MIT, 2013).
Several decision support systems for flooding crisis man-
agement have been proposed, e.g. Jasionowski (2011),
Ruponen et al. (2012) and Varela et al. (2014). An example
of output of a system which assesses the survivability of a
flooded vessel in real-time is shown in Fig. 3, based on
work by Ruponen et al. (2015). This software uses data
received from flood level sensors to detect breaches in the
hull. The real-time information about the water levels is
used in a time-domain progressive flooding simulation
(Ruponen et al., 2007). Thus, the stability and survivability
of the vessel are predicted for a given period into the future,
continuously updating the prediction based on the latest
available data.
Ruponen et al. (2015) presented the vessel survivability as a
value between 0 and 1. However, such values are not very
intuitive for ship crew, who need readily understandable
information in a simple form on which they can act in an
accidental situation. The Vessel TRIAGE categories and
corresponding color codes present an easily interpretable
13
platform for communication. However, linking the calcu-
lated survivability with these categories requires criteria for
delineating the categories. Pennanen et al. (2016) propose
such criteria for passenger vessels. The flooding criteria
relate are based on the number of flooded compartments
and a flooding extent coefficient, similarly as in the Safe
Fig. 3: Survivability assessment for a flooded passenger ship:
Vessel TRIAGE category, time-domain heel angle prediction and flooding extent, (Pennanen et al., 2015)
DISCUSSION: FURTHER RESEARCH AND TESTING
While a preliminary evaluation of the theoretical construct
of the Vessel TRIAGE system was overall very positive,
see Nordstrm et al. (2016), further testing has been rec-
ommended both by the developers of the method (IMO,
2015) and by IMOs Sub-Committee on Navigation, Com-
munication and Search and Rescue (IMO, 2016). Rigorous
testing has also been applied to triage systems for medical
emergencies, see e.g. Parenti et al. (2014).
The method has a definite intuitive appeal, and offers a
simple way forward to improve communication and deci-
sion making in post-accidental situations. However, im-
plementation of the method in international SAR regula-
tions requires taking into account the concerns of various
stakeholders, and further justification that the method does
indeed serve its intended purpose to improve communica-
tion and be useful for practical decision making, without
putting unnecessary burdens on ship crew. It also requires
that the method is aligned with current practices in SAR
communication, e.g. related to the already existing emer-
gency phase classification included in the SAR Convention.
Nordstrm et al. (2016) stress the need for further testing,
but do not specify how this could be done in practice, and
provide no further discussion on the practical concerns
about its implementation.
Some concerns regarding the above issues briefly outlined
in Table 1, partly based also on the discussions of the
working group in the IMOs working group of the
Sub-Committee on Navigation, Communication and Search
and Rescue (IMO, 2016). A number of possible practical
actions are listed as well, providing a way forward to ad-
dress these concerns.
14
Return to Port regulation, see also Vassalos (2007). The
listing/decrease of stability criteria rely on the SOLAS reg-
ulations, in particular the heel angle and the s-factor in
SOLAS II-1 Part II-1 Reg. 7. For details, see Pennanen et al.
(2016), where a case study for a damaged passenger ship is
shown as well.
CONCLUSION
This paper has presented an overview of the Vessel TRI-
AGE method, a recently developed tool for assessing and
communicating the safety status of a vessel in maritime
accident situations. Technical developments related to the
flooding and listing/decrease of stability threat factor
were outlined, showing that Vessel TRIAGE is a simple yet
powerful method to act as a framework for integrating in-
formation from technical decision support systems into an
overall tool for emergency communication and decision
making.
The Vessel TRIAGE method has undergone preliminary
positive evaluation, showing good potential for enhancing
maritime safety in maritime distress situations. This paper
has outlined several possible further tests to show the effi-
cacy of the method to improve SAR decision making and
maritime safety, and further developments to integrate the
method in SAR procedures. It is hoped that this can spur
additional research and testing activity, ultimately leading
to broad acceptance among the different stakeholders and
on the appropriate regulatory levels of decision making.
 Table 1: Vessel TRIAGE: concerns and actions to address these
ID Concerns
1 Vessel TRIAGE should be valid for all acci-
dental situations affecting the vessels status
as a means to provide safety to people
onboard (crew, passengers, responders)
2 Vessel TRIAGE should not lead to additional
workload to the ship crew, which could hinder post-accidental situations, ensuring the Vessel TRIAGE clas-
successful shipboard response efforts
3 Vessel TRIAGE should give a discernible
benefit to emergency communication and de-
cision making
4 Vessel TRIAGE should not impose decisions
on the SAR response resources on the ship
master
5 Vessel TRIAGE should not conflict with ex-
isting onboard and shore-based emergency
assessment and communication methods
6 Vessel TRIAGE should be consistently ap-
plied by different individuals in different situ- narios, to evaluate the inter-rater reliability of the tool.
ations
7 SAR communication procedures should not
conflict when a vessel crosses multiple SAR
regions be harmonized across SAR regions
8 Regulatory implications of implementing
Vessel TRIAGE should be accounted for.
REFERENCES
Azeredo, T.R.M., Guedes, H.M., Rebelo de Almeida, R.A.,
Chianca, T.C.M., Martins, J.C.A. (2015). Efficacy of
the Manchester Triage System: a systematic review. Int.
Emerg. Nurs. 23, pp.4752.
Eliopoulou, E., Papanikolaou, A., Voulgarellis, M. (2016)
Statistical analysis of ship accidents and review of
safety level. Saf. Sci. 85, pp.282292.
Fuchs, H.M., Steigenberger, N., Lbcke, T., (2015) Intui-
tion of deliberation - How do professionals make deci-
sions in action?, in: Proc. ISCRAM, May 24-27. Kris-
tiansand, Norway.
Hogstrm, P., (2012). RoPax ship collision - a methodol-
ogy for survivability analysis, PhD Dissertation.
Chalmers University of Technology, Gothenburg, Swe-
den.
IMO. (2016) Guidelines on harmonized aeronautical and
maritime Search and Rescue procedures, including SAR
training matters. Amendments to the IAMSAR manual.
Revised guidelines for preparing plans for cooperation
between Search and Rescue services and passenger
ships MSC/CIRC.1079 - Report of the working group.
Action(s) to address the concern
Additional theoretical assessments of the construct validity of
the method, through expert judgment and/or through assessing
the threat factors in light of accident reports.
Elaborate on crew communication procedures in
sification contributes to the response effort rather than dis-
tracting resources.
Test the Vessel TRIAGE method in simulated accident sce-
narios, ensuring the crew communication procedures ensure a
successful outcome.
Test and provide guidance on how often the reassessment of
the Vessel TRIAGE category should be performed.
Accident reports can be studied to evaluate how much time
would be gained had Vessel TRIAGE be used.
Test the Vessel TRIAGE method in simulated accident sce-
narios, to evaluate the benefits it brings to practice.
Elaborate on communication and decision roles between ves-
sel crew, on-scene commanders and shore-based SAR re-
sponders.
Other international and national emergency assessment and
communication methods and procedures should be identified,
and integrated into one unified procedure.
Test the Vessel TRIAGE method in simulated accident sce-
SAR emergency communication procedures should be har-
monized across SAR regions.
If and when it is decided to implement Vessel TRIAGE in
international regulations, the effects on relevant IMO and
ICAO documents should be assessed and accounted for.
IMO (2015) Guidelines on harmonized aeronautical and
maritime Search and Rescue procedures, including SAR
training matters. Vessel TRIAGE - a method for as-
sessing and communicating the safety status of vessels in
maritime accidents and incidents, NCSR 3/21/2
Jasionowski, A. (2011) Decision support for ship flooding
crisis management. Ocean Eng. 38, pp.15681581.
Klanac, A. (2011) Design method for safe ship structures
PhD Thesis. Aalto University, Espoo, Finland.
MAIB. (2015) Report on the investigation of the ground-
ing and flooding of the ro-ro ferry Commodore Clipper
in the approaches to St Peter Port, Guernsey on 14 July
2014, No. 18/2015 Marine Accident Investigation
Branch, Southampton, UK.
MAIB. (2009). Report on the investigation into the flood-
ing and foundering of the grab hopper dredger Abigail H
in the Port of Heysham 2 November 2008, No. 15/2009.
Marine Accident Investigation Branch, Southampton,
UK.
MIT. (2013) Cruise Ship Costa Concordia Marine casualty
on January 13, 2012 - Report on the safety technical in-
vestigation, Ministtry of infrastructures and transports.
Nordstrm, J., Goerlandt, F., Sarsama, J., Leppnen, P.,
Nissil, M., Ruponen, P., Lbcke, T., Sonninen, S.,
(2016). Vessel TRIAGE: a method for assessing and
15
 communicating the safety status of vessels in maritime
distress situations, Saf. Sci. 85, pp.117129.
Parenti, N., Reggiani, M.L.B., Iannone, P., Percudani, D.,
Dowding, D., (2014). A systematic review on the valid-
ity and reliability of an emergency department triage
scale, the Manchester Triage System, Int. J. Nurs. Stud.
51, pp.10621069.
Pennanen, P., Ruponen, P., Nordstrm, J., Goerlandt, F.,
(2016) Application of Vessel TRIAGE for a damaged
passenger ship in: Proc. 15th International Ship Stabil-
ity Workshop, ISSW, Stockholm, Sweden, pp.141146.
Pennanen, P., Ruponen, P., Ramm-Schmidt, H. (2015).
Integrated decision support system for increased pas-
senger ship safety. in: Damaged Ship III. Presented at
the Royal Institution of Naval Architects, London, UK.
RAJA, TraFi, LiVi, VTT. (2015). Vessel TRIAGE: A
method for assessing and communicating the safety sta-
tus of vessels in maritime accidents and incidents - user
manual.
Ruponen, P., Larmela, M., Pennanen, P. (2012). Flooding
prediction onboard a damaged ship in: Proc. 11th Int.
Conf. Stability of Ships and Ocean Vehicles (STAB2012),
Athens, Greece, pp. 391400.
Ruponen, P., Lindroth, D., Pennanen, P. (2015). Prediction
of survivability for decision support in ship flooding
emergency. Proc. 12th Int. Conf. Stability of Ships and
Ocean Vehicles, Glasgow, UK.
Ruponen, P., Sundell, T., Larmela, M. (2007). Validation
of a simulation method for progressive flooding. Int.
Shipb. Prog. 54, pp.305321.
Seppnen, H., Virrantaus, K. (2015). Shared situational
awareness and information quality in disaster manage-
ment. Saf. Sci. 77, pp.112122.
Varela, J.M., Rodrigues, J.M., Guedes Soares, C. (2014).
On-board decision support system for ship flooding
emergency response. Procedia Comput. Eng. 29,
pp.16881700.
Vassalos, D. (2007). Safe Return to Port - A framework
for passenger ship safety. in: Proc. 10th Int. Symp.
Practical Design of Ships and Other Floating Structures.
Houston, Texas, USA.

16
 Study on the Correction of Wave Surge Forces to Improve
Surf-riding/Broaching Vulnerability Criteria Check
Accuracy
Pei-yuan FENG1, She-ming FAN1, Li-wei YU2, Ning MA2
1. Science and Technology Laboratory of Water-jet Propulsion,
Marine Design & Research Institute of China, pyfeng23@163.com
2. State Key Laboratory of Ocean Engineering, Shanghai Jiao Tong University, China

ABSTRACT tion for the development of broaching stability assessment

This study focuses on the influence of wave surge
force prediction accuracy on the surf-riding/broaching
vulnerability criteria check. It is demonstrated through
captive model test that the current wave surge force
prediction model adopted in the draft criteria is not ac-
curate enough. To improve the vulnerability criteria
check accuracy, the Green function based 3D panel
method and the empirical correction formula proposed
by Japan are adopted to calculate the wave surge forces
of five sample ships, respectively. It is shown that the
empirical correction formula can help improve the wave
surge force prediction result, despite some discrepancies.
Finally, the surf-riding/broaching vulnerability criteria
check is performed to the sample ships and it is demon-
strated that neglecting diffraction effect in the predic-
tion of wave surge forces may lead to wrong conclu-
sions.
INTRODUCTION
The International Maritime Organization (IMO) is currently
working on the Second Generation Intact Stability Criteria
to ensure ship safety in waves more effectively. Five new
stability failure modes are to be included in the new genera-
tion stability criteria: pure loss of stability, parametric roll,
dead ship condition, surf-riding/broaching and excessive
acceleration (Umeda 2013). Among them, broaching is
considered to be the most complicated one due to its highly
nonlinear nature (Umeda 1999, Hashimoto et al 2004a).
Broaching occurs when a ship cannot keep a constant
course despite the maximum steering effort, typically in
following and quartering waves (Umeda and Renilson 1992,
Spyrou 1995). Surf-riding is usually regarded as the pre-
requisite of broaching when a ship is captured by the wave
approaching from the stern that accelerates the ship to the
wave celerity (Motora et al 1982, Umeda and Vassalos
1996). Small-size high-speed ships are most vulnerable to
this stability failure mode.
Significant theoretical and experimental efforts have been
made by researchers in the recent decades (Hashimoto and
Stern 2007, Maki et al 2010), which form a good founda-
criteria. The latest proposal from U.S. and Japan (SDC
2/INF.10 2014) follows a three-tiered framework: Level 1
evaluation only needs the ship length and speed infor-
mation; Level 2 evaluation is based on a simplified
surf-riding model, the probability of surf-riding occurrence
in irregular seaway is chosen as the criteria for assessment;
Level 3 direct stability assessment procedures are still un-
der discussion, the draft guidelines can be found in
SDC1/INF.8 (2013).
This study focuses on the influence of wave surge force on
the Level 2 surf-riding/broaching check result. Experi-
mental data based on captive model tests have shown that
the correctness of wave force estimation plays an important
role in the prediction model (Hashimoto et al 2004b, Horel
et al 2014). The current prediction model adopted in the
surf-riding/broaching vulnerability check procedure only
takes into account the Froude-Krylov component of the
wave surge force. As pointed out in Annex 35 of the draft
criteria (SDC 2/INF.10, 2014): for calculating the ampli-
tude of the wave surge force, the Froude-Krylov component
on its own could often over-estimate it, in which case, em-
pirical diffraction effect correction should be considered.
Therefore, it is meaningful to investigate more accurate
wave surge force prediction methods and find out a proper
correction model.
In this study, it is demonstrated through captive model test
that the current wave surge force prediction model adopted
in the draft criteria is not accurate enough. To improve the
vulnerability criteria check accuracy, the Green function
based 3D panel method and the empirical correction for-
mula proposed by Japan (SDC 2/INF.10, 2014) are adopted
to calculate the wave surge forces of five sample ships,
respectively. It is shown that the empirical correction for-
mula can help improve the wave surge force prediction
result, despite some discrepancies. Finally, the surf-riding/
broaching vulnerability criteria check is performed to the
sample ships and it is demonstrated that neglecting diffrac-
tion effect in the prediction of wave surge forces may lead
to wrong conclusions.

17
SURF-RIDING/BROACHING CRITERIA N
K T ( J ) i J i = 0 + 1 J + 2 J 2 + L [7]
For the purpose of this study, Level 2 surf-riding/broaching i =0
vulnerability check procedures of Annex 32 and Annex 35 where, R(u) is the calm water resistance of the ship; Te(u; n)
in SDC 2/INF.10 (2014) are relevant. is propeller thrust in calm water; tP is thrust deduction fac-
tor; wP is wake fraction; DP is propeller diameter; J is ad-
For a ship to pass Level 2 vulnerability criteria assessment, vance coefficient; is water density.
it is required that:
C < RSR [1] ncr is the critical propeller revolution solved by Melnikovs
method (Maki et al 2010):
Te ( cw ; ncr ) R ( cw )
where, RSR is the standard value, which has recently been N i
2 = Cij ( 2 ) I j [8]
changed from 0.005 to 0.05. C represents the probability of j

surf-riding occurrence, which is estimated by: f i =1 j =1

N N a

( fk )
j 2
Wij C 2ij c i!
Cij = i j ci j [9]
C = W 2 ( H S , TZ ) N Na j !( i j ) ! ( m + mx )
i =1 j =1 j 2 w
[2] fk
H S TZ
Wij
(1 t P )(1 wP ) i
i
i =1 j =1
ci = +r [10]
ni 2 DPi 4
i
where, W2(HS,TZ) is the weighting factor of each sea state
according to long-term wave statistics; HS is the significant
j +1 j+2
wave height; TZ is the zero-crossing wave period; Wij is the I j = 2 [11]
statistical weight of a wave with steepness (H/) varying 2 2
from 0.03 to 0.15 and wave length to ship length ratio (/L) where, cw is wave celerity; k is wave number; m is ship
varying from 1 to 3. Details concerning the calculation of mass; mx is added mass in surge direction; () is the
these factors are specified in SDC 2/INF.10 (2014). Figure gamma function.
1 shows the occurrence percentage of the long-term wave
statistics. As can be seen, the sea state with the highest oc- f is the amplitude of wave surge force, which according to
currence frequency corresponds to HS=2m and TZ=8s. the draft criteria, is calculated by:
H
f = gk FC2 + FS2 [12]
2
NS
FC = S ( xi ) exp {0.5kd ( xi )} sin kxi xi [13]
i =1
NS
FS = S ( xi ) exp {0.5kd ( xi )} cos kxi xi [14]
i =1
where, Ns is the number of sections; d(xi) and S(xi) are the
draft and submerged area of section i in calm water.

This wave force prediction model is based on linear theory

and the effect of ship forward speed is neglected. Moreover,
this force model only accounts for the Froude-Krylov (FK)
component of the wave surge force, while the diffraction
Fig.1: Long-term wave statistics (occurrence percentage) effect is totally neglected. Japan proposed an empirical
correction formula to correct the diffraction effect in Annex
C2ij indicates whether surf-riding occurs for the designated 35 of SDC 2/INF.10, as shown in equation [16]. The block
wave case, which is defined as: coefficient CB and the mid-ship section coefficient CM are
1 if Fn > Fncri , j used as the featuring parameters for the correction. Detailed
C 2ij = [3] explanation of the reasoning behind this empirical formula
0 if Fn Fncri , j for diffraction effect correction can be found in (Ito et al
where, Fncr is the critical Froude number corresponding to 2014).
the threshold of surf-riding (surf-riding occurs under any H
initial condition); ucr is the critical speed determined by: f = x gk FC2 + FS2 [15]
Te ( ucr ; ncr ) R ( ucr ) = 0
2
[4]
1.46CB 0.05 CM < 0.86
N

R ( u ) ru
i = r1u + r2u + L
i 2
[5] x = ( 5.76 5CM ) CB 0.05 0.86 < CM < 0.94 [16]
i =1 1.06CB 0.05 CM 0.94

Te ( u; n ) = (1 t P ) n 2 DP4 KT ( J ) [6]

18
CAPTIVE MODEL TEST

To validate the wave surge force prediction model adopted

in the latest draft criteria, captive model test in regular fol-
lowing waves is performed to measure the wave surge force
for a purse seiner. The main particulars of the ship model
(scale ratio 17) are listed in Table 1. The experiment is
conducted in the towing tank of MARIC (Marine Design &
Research Institute of China). The basin is 280m long, 10m
wide and 5m deep. The ship model is towed at the service
speed. Several wave frequencies are tested with a constant
wave steepness of 0.02.

Table 1. Main particulars of the purse seiner

Item Symbol Full Scale Model
Length Perpendicular LPP (m) 42.5 2.5000 Fig.3: Comparison of wave surge force results

Breadth B (m) 7.80 0.4588 SAMPLE SHIP CALCULATIONS

Depth D (m) 4.05 0.2382
The 3D panel method and the empirical correction formula
Mean Draft d (m) 3.20 0.1882
proposed by Japan are adopted to calculate the wave surge
Block Coefficient CB 0.6721 0.6721 forces of five sample ships, respectively. Main dimensions
of the sample ships are listed in Table 2, which are mainly
Mid-section Coefficient CM 0.8808 0.8808
fishing vessels with length around 40m and Fn around 0.3.

The experimental model of the purse seiner is shown in Table 2. Main dimensions of sample ships
Fig.2. During the experiment, two poles are used to confine
NO. Ship Type LPP (m) Fn CB CM
the sway and yaw motions of the model. Two force trans-
ducers are installed correspondingly to measure the forces 1 purse seiner 42.5 0.320 0.6721 0.8808
in the surge direction. The summation of the forces from 2 purse seiner 43.0 0.340 0.8011 0.9309
the two sensors results in the measured wave surge force.
3 purse seiner 54.0 0.330 0.7396 0.9269
Besides, numerical result from a frequency domain Green 4 fishery vessel 29.5 0.364 0.4796 0.7416
Function based 3D panel method is also compared. The 3D 5 net fishing boat 41.0 0.318 0.5800 0.8730
panel method can calculate the diffraction component of the
wave surge force, and can reflect the influence of ship
speed. Detailed theoretical explanations and applications of According to the wave surge force prediction results shown
this method can be found in Chen (2004). in Figs.4-8, it can be seen that for No.1, 2 and 3 ships, the
correction formula over-estimates the wave surge force. For
The obtained wave surge force response amplitude opera- No.4 ship, the wave surge force is under-estimated. For
tors (RAOs) are illustrated in Fig.3, where the horizontal No.5 ship, the correction formula can give quite accurate
axis denotes the wave length to ship length ratio and the estimation. Generally speaking, the empirical correction
vertical axis denotes the non-dimensional wave surge force. formula can help improve the wave surge force prediction
result, despite some discrepancies.
From the results, it is clearly demonstrated that the ampli-
tudes of the wave surge force without diffraction correction Finally, both Level 1 and 2 surf-riding/broaching criterion
are over-estimated. Therefore, it is necessary to include the check is performed to the sample ships, with three strate-
diffraction effect in the wave surge force prediction model. gies of wave surge force prediction: (a) without diffraction
correction; (b) using empirical diffraction correction for-
The 3D panel method result agrees quite well with the ex- mula; (c) using 3D panel method.
perimental result, while the empirical formula can help im-
prove the prediction accuracy, but not as accurate. The results are shown in Table 3. As can be seen, although
the criterion check conclusion for all the sample ships are
consistent, the resulting C values are drastically different
whether the diffraction effect is considered or not, thus may
lead to wrong conclusions.

Fig.2: Experimental model of the purse seiner
On the other hand, the check result using the empirical dif-
fraction correction formula is close to that using the 3D
panel method. Therefore, the empirical formula can im-
prove the check accuracy and the diffraction effect correc-
tion is recommended in the surf-riding/broaching vulnera-
bility assessment procedure.

19
 Table 3. Surf-riding/broaching criterion check result
Level 2 Check
No. (a) (b) (c)
C Result C Result C Result
1 8.06E-03 Pass 1.74E-03 Pass 0.00E+00 Pass
2 1.12E-02 Pass 1.09E-03 Pass 0.00E+00 Pass
3 8.75E-03 Pass 0.00E+00 Pass 0.00E+00 Pass
4 3.01E-01 Pass 1.76E-01 Pass 2.10E-01 Pass
5 1.61E-02 Pass 1.51E-03 Pass 1.51E-03 Pass

Fig.7: Wave surge force prediction of No.4 ship

Fig.4: Wave surge force prediction of No.1 ship

Fig.8: Wave surge force prediction of No.5 ship

CONCLUSIONS

In this study, the influence of wave surge force prediction

on surf-riding/broaching vulnerability criteria check is in-
vestigated. The following conclusions can be drawn:

1. It is demonstrated that the wave surge force prediction

method adopted in the current vulnerability criteria
check procedure is not accurate and may lead to
wrong conclusions.
2. Comparison between experimental data from captive
model test and numerical results based on 3D panel
Fig.5: Wave surge force prediction of No.2 ship method shows that 3D panel method can give accurate
prediction of the wave surge forces of the ship in fol-
lowing waves under service speed.
3. The empirical correction formula proposed by Japan
can help improve surf-riding/broaching criterion check
accuracy. However, sample ship calculations show
that it may over-estimate or under-estimate the wave
surge forces.

ACKNOWLEDGEMENTS

The present study is supported by the Ministry of Industry

and Information Technology of China and the National
Natural Science Foundation of China (Grant No.
51579144).
Fig.6: Wave surge force prediction of No.3 ship

20
REFERENCES
Chen, X. B. (2004), Hydrodynamics in offshore and naval
applications, Proc. 6th Int. Conf. on Hydrodynamics
Hashimoto, M., Umeda, N. and Matsuda, A. (2004a), Im-
portance of Several Nonlinear Factors on Broaching Pre-
diction, J. Marine Science and Technology 9, pp.80-93
lowing Sea, Manoeuvring and Control of Marine Craft.
Computational Mechanics
Ito, Y., Umeda, N. and Kubo, H. (2014), Hydrodynamic
Aspects on Vulnerability Criteria for Surf-Riding of
Ships, Jurnal Teknologi, Vol. 66, No. 2, pp 127-132

Hashimoto, M., Umeda, N. and Matsuda, A. (2004b),

Model experiment on heel-induced hydrodynamic forc-
es in waves for broaching prediction, Proc. 7th Int. Ship
Stability Workshop

Hashimoto, M. and Stern, F. (2007), An Application of

CFD for Advanced Broaching Prediction, J. Japan So-
ciety of Naval Architects and Ocean Engineers 5E,
pp.51-62

Horel, B., Guillerm, P. E., Rousset, J. M. and Alessandrini,

B. (2014), Experimental database for surf-riding and
broaching-to quantification based on captive model tests
in waves, Proc. 14th Int. Ship Stability Workshop

Maki, A., Umeda, N., Renilson, M. and Ueta, T. (2010),

Analytical Formulae for Predicting the Surf-Riding
Threshold for a Ship in Following Seas, Jour. of Marine
Science and Technology, No.15, pp 218-229

SDC 1/INF.8, (2013), Draft Guidelines of Direct Stability

Assessment Procedures as a Part of the Second Genera-
tion Intact Stability Criteria, International Maritime
Organization

SDC 2/INF.10, (2014), Proposed Amendments to Part B

of the 2008 IS Code to Assess the Vulnerability of Ships
to the Broaching Stability Failure Mode, International
Maritime Organization

Motora, S. (1982), On the Mechanism of Broaching-to

Phenomena, Proc. 2nd Int. Conf. Stability of Ships and
Ocean Vehicles

Spyrou, K. (1995), Surf-riding, Yaw Instability and Large

Heeling of Ships in Following/Quartering Waves, Ship
Technology Research 42, pp.103-112

Umeda, N. (1999), Nonlinear Dynamics of Ship Capsizing

Due to Broaching in Following and Quartering Seas, J.
Marine Science and Technology 4, pp.16-26

Umeda, N. (2013), Current Status of Second Generation

Intact Stability Criteria Development and Some Recent
Efforts, Proc. 13th International Ship Stability Work-
shop

Umeda, N. and Vassalos, D. (1996), Non-linear Periodic

Motions of A Ship Running in Following and Quartering
Seas, J. Society of Naval Architects of Japan 179,
pp.89-101

Umeda, N. and Renilson, M. (1992), BroachingA Dy-

namic Analysis of Yaw Behaviour of a Vessel in a Fol-

21
 On the Survivability of ROPAX and Cruise Ships
A New Approach to Differences in Design
George Dafermos
Ship Design Laboratory, NTUA, Greece, giorgosdafermos@gmail.com
Apostolos Papanikolaou
Ship Design Laboratory, NTUA, Greece, papa@deslab.ntua.gr

In the regulatory framework of SOLAS 2009, the s factor

ABSTRACT
The present paper summarizes the main results of a
comparative study on the survivability after damage of
a ROPAX and a Pure Passenger ship (or cruise ship), of
the same size. Systematic flooding simulations were
carried out in order to investigate the damage stability
and identify notable differences in the survivability
behavior of the two vessels. In this respect, the present
formulation of attained subdivision index A of SOLAS
2009 is being examined in order to determine whether it
can sufficiently account for the design differences of the
study vessels. On this background, we proceed with the
introduction of a new Survivability Performance Index
(SPI), which better represents ship's behavior during
flooding and properly accounts for notable differences
in the assessment of the damage stability of ROPAX and
cruise type ships of the same size and capacity.
INTRODUCTION
Flooding simulations of ships after the loss of their
watertight integrity are generally used for more refined
investigations of ship's damage stability, as in cases where
the physics of flooding and its effects on the time to sink
(TTS) need to be clarified in the frame of forensic studies.
Previous studies of the authors laboratory (see, e.g., [1],
[11], [14]) have indicated the occurrence of some typical
consequences of the flooding process, such as the transient
heeling developed during the initial phases of flooding
and/or cross flooding. Such phenomena, which are related
primarily to free surface effects, were thoroughly discussed
during the adoption of the new probabilistic damage
stability regulations of SOLAS 2009. This is also reflected
in the damage stability assessment of passenger ships by the
way of calculation of the survival factor; namely, the s
factor is calculated based on two different formulations, one
for the intermediate and one for the final equilibrium stage.
expresses the probability of survival of the ship and
accounts for any stage of flooding. The purpose of the
intermediate s factor implementation is to identify whether
the ship faces a higher risk of capsize in the intermediate
stages, under the effect of the above mentioned phenomena,
than in the final stage of flooding.
Since the inception of the probabilistic approach to ship's
damage stability there were justified concerns regarding the
capability of the present s- factor formulation to properly
account for the actual survivability of the ship under
consideration; this was particularly disputable for passenger
ships of different designs and sizes and is still an
unresolved issue in scientific and practical terms. There
were several attempts to reformulate the s factor for
passenger ships, e.g. by project GOALDS [10], and to
better account for the actual survivability of passenger ships,
with the SLF55 formulation finally prevailing in the latest
amendment of SOLAS [5]. All above methods follow a
quasi-static approach to ship's survivability, where
intermediate stages of flooding are taken into consideration,
but not the time dependent progressive flooding process
itself.
In recent years, the development and coding of
mathematical models for the dynamic flooding and ship
motion simulation enabled the improved modeling and
recording of the behavior of the vessel throughout the
evolution of the flooding process. When a time domain
simulation of the flooding process is performed, which we
like herein to interpret in the sense of the probabilistic
SOLAS s- factor, we could use the same stability
assessment criteria like in SOLAS, but accounting for the
time evolution of the flooding process; this includes the
assessment of ship's GZ curve momentarily, at every time
instance, while determining the probability of survival as a
certain event of survival or sink/capsize, depending on the
simulated TTS predicted by the flooding simulation. This is
an important extension of present thinking regarding ship's
damage stability, namely by introducing a more refined

22
assessment of the individual flooding scenarios by a
modified quasi dynamic s- factor, which also accounts for
simulated TTS. The ship motions following a breach of
ship's outer shell are being in parallel determined by
dynamic simulation methods.
There are a variety of approaches and dynamic simulation
models and computer solvers, in which the flooding process
and ship's motions are addressed by simplified and/or
advanced modeling and numerical techniques ([3], [7]). In a
probabilistic approach to ship's damage stability, in which a
large number of damage scenarios needs to be assessed, it is
essential to dispose methods and simulation tools, which
can be readily used in practice, without compromising on
the accuracy of the obtained results.
In this study, the ship is considered as a rigid body (at zero
forward speed) and her floating position is fully determined
by three degrees of freedom, namely her heel, trim and
mean draft1. We employ the well-established flooding and
motion simulation tool of NAPA and consider the time
dependent GZ-curve for the assessment of ship's stability;
through it, all main stability parameters, such as the
maximum GZ value, the time varying metacentric height
and the s factor can be readily extracted and used for the
assessment of ship's survivability. The employed method
determines at each time step a new floating position, while
the floodwater ingress and spreading is computed by a
hydraulic model for the water fluxes; this is leading
eventually to a new floodwater distribution and a new
equilibrium position, based on the balance of the forces and
moments caused by the floodwater [2].
THE SOLAS 2009 ASSESSMENT
The main scope of this assessment study is to determine the
survivability performance of the two passenger ships of
different design (a ROPAX and a cruise ship), but of the
same size and capacity (in terms of People On Board), with
respect to their behavior in case of flooding; this is
achieved by numerical flooding simulations, while
examining whether the current SOLAS regulation, which
specifies the same Required Subdivision Index for the two
study vessels, can effectively take into account the
difference of the inner subdivision of the vessels.
Initially, the two designs are being evaluated against the
requirements of the SOLAS 2009 criteria on damage
stability. According to this framework, three loading
conditions corresponding to the deepest subdivision draft
(ds), to the partial subdivision draft (dP) and the light service
condition draft (dl) need to be defined. Furthermore, the
subdivision of the vessel needs to be considered in order to
properly apply the zonal approach prescribed in the SOLAS
regulations. For each zone determined via the vessels
subdivision one or more damage cases are generated.
Multi-zone damages are also generated as all possible
1 the estimated time to sink/capsize (TTS),
Herein, it is tacitly assumed that even though the ship
may be operating in a seaway, it will keep her mean
position while she is not drifting away and her wave
induced motions are small so that they can be practically
neglected.
combinations of adjacent one-zone damages. For all
damage cases the corresponding probabilities of damage pi
and vi are assigned. Also, the box-shaped breach SOLAS
concept is adopted and the dimensions of each breach are
determined based on the subdivision and the penetration
limits. For each damage case, the GZ curve is obtained for a
number of intermediate phases and the final one; the s
corresponding factor is calculated based entirely on the
properties of the GZ curve. The s factor represents the
probability of survival of corresponding stages of the
flooding process. The cumulative probability of survival
from each damage scenario is the product of the p, v and s
=
factors.
For each draft specified in the regulation the summation of
the above conditional probabilities, corresponding to each
damage scenario, results to the attained subdivision factor
at the considered draft. The attained subdivision indices are
weighted by the w factors that represent the percentage of
ships lifecycle associated to each condition. The final
attained index is calculated on the basis of the following
= + +
formula:
= 0.2 + 0.4 + 0.4
For each damage case, the initial condition is known as well
as the wi factor. As a result, the A index can be calculated
via the equivalent formulation:
=
, ,
THE ASSESSMENT BASED ON NUMERICAL
SIMULATIONS
Flooding simulations may be used for the assessment of
certain damage scenarios in a deterministic approach, e.g.
when doubts arise on whether the cross flooding completes
within the specified time limit or when a specific damage
scenario is considered of particular importance (e.g.
forensic studies), but rarely for the probabilistic assessment
of ship's damage stability (e.g. [14], [15]). The assessment
is generally both qualitative and quantitative, elucidating
the physics and consequences of the flooding process on the
damaged vessel. However, complicated mathematical
models need to be developed to closely simulate the physics
of the problem and even more in order to enable the
possibility for comparison and optimization amongst
several design options.
In this study a new formulation for the survival s- factor of
SOLAS is introduced. We replace the traditional s-factor by
the new Survivability Performance Index (SPI), which
takes better account of the actual survivability of the vessel
and is determined by the most important parameters of the
flooding process, namely
the time varying s factor at any moment of the
simulation,
the s factor at the 30th minute of the simulation
and
23
 the maximum and/or final heeling angle.
The time to sink TTS is probably the most important
parameter that is herein used. Even if the current SOLAS
assessment may lead to a survival s- factor of zero in case
of sinking/ capsize, the time until the final loss of the vessel
is crucial and needs to be assessed; namely, a design that
delays the time to sink/capsize is safer (and should score
better in the assessment of its survivability), even though
eventually the ship may be lost, but the potential loss of life
is being mitigated.
The specified s- factors used for the SPI formulation are the
s factor at the 30th minute of the simulation and at the final
moment, which is considered as either the moment when a
steady state (equilibrium) is achieved, or when the
maximum simulation time (60 min) is reached. Both these
factors may be calculated on the basis of the time varying
GZ characteristics by the employed simulation method. The
s factor at the 30th minute of the simulation is a
representative parameter indicating that the ship retains her
stability and an initiated evacuation of people onboard can
be completed, as specified in the regulations. The final
equilibrium stability is also considered and expressed by
the final s- factor, so that a design associated to a higher s-
factor should inherently have a higher survivability index,
since it is more stable.
Furthermore, the floating position of the vessel needs also
to be taken into consideration. Heeling angles are
considered as adequately representative and the most
significant parameters characterizing ship's damage stability
and eventually floatability. The heeling angles used herein
are the maximum heeling angle developed during the
flooding process and the finally attained heeling angle. The
calculation of the index, based on the maximum heeling
angle takes into account possible transient phenomena that
emerge during the initial ingress of floodwater or during
cross or down flooding periods. Ship's final floating
position is expressed through the final heeling angle.
The Survivability Performance Index (SPI) is calculated by
use of the following formulations:
= 11
! '()*+
/0
"#$% /0,-*.
'()*+,-*.
= 11
! '()*+ /0
< =
"#$% '()*+,-*. /0,-*.
Where the TTS, sfinal, s30, heelfinal and heelmax were
discussed previously. The TRULE is the total simulation time
and it is herein set equal to 60 minutes (default value). The
sfinal-max and s30-max are the maximum values of the s factor at
the final stage and at the 30th minute, respectively, which
are unity. The capsize angle (heelcapsize) is herein set equal to
45 (default value) and it corresponds to the angle that the
ship is considered as lost due to capsize. The above default
values may be adjusted on the basis of systematic studies.
As observed, all parameters taken into account are
normalized by the maximum value they could attain. The
reason for this is to obtain an index which is comparable to
24
the s factor used in SOLAS and as such its maximum value
must be unity. The SPI represents the survivability
performance of a vessel as percentage of the best possible
performance that is herein characterized by SPI equal to
unity.
Both SPI formulations are herein used in parallel for the
assessment of the vessels survivability performance, in
order to examine which one is more appropriate for use. In
the SPIfinal formulation, the finally attained heeling angle is
used as criterion and the final equilibrium (if any) is the
performance indicator, noting that a smaller final heeling
angle represents a more stable condition after flooding.
When the maximum attained heeling angle is used in
SPImax, the transient flooding effects are more dominant and
the smoothness of the flooding process is favored.
When the SPI index is calculated, a new attained index can
be evaluated likewise to the A index obtained from the
SOLAS assessment. In this case the Asim index is calculated
according to the following formulation:
=
!>?
, ,
It is understood that the survivability performance of the
vessel is weighted based on the probabilities of the damage
breach position and dimensions, as well as the initial
loading conditions, similarly to the SOLAS s- factor
approach.
THE PROPOSED NEW ASSESSMENT PROCEDURE
The present study deals with the comparative assessment of
the survivability of the two selected passenger ship types of
equal size and capacity according to both the current
SOLAS regulatory framework and by use of the herein
proposed new SPI method on the basis of flooding
simulations. In this section the procedure followed for the
evaluation is outlined.
The external hull surface and internal compartmentation of
the study vessels were developed by the Ship Design
Laboratory of NTUA. The difference in the internal
subdivision of the two vessels is that a new deck was
455 '()*+
;
introduced in the RORO space of the ROPAX ship for the
455 6*78(9:
pure passenger ship (cruise ship), with cabins for passenger
accommodation. The external openings, the under-deck
;
455 -*.
subdivision as well as the loading conditions were kept
455 6*78(9:
unchanged. The reason for not altering the loading sets was
to decouple the effect of the denser compartmentation
above the main deck from possible effects of differences in
the hydrostatic characteristics and the initial stability
parameters of the two vessel types. Furthermore, it was
necessary to define a compartment connection table relating
to the flooding of the cruise ship spaces above the main
deck, as in the case of the unobstructed garage space of the
ROPAX ship.
At first, the evaluation of both ship type according to
SOLAS 2009 is carried out. Calculations are made for
damage cases involving alternatively up to five-zone
damages and up to three-zone damages. The former
assessment is made in order to fully explore the survival
capabilities of the vessel, attaining the highest possible A
index and examining practically all possible damage cases.
The latter assessment, which involves fewer damage
scenarios, is performed in order to generate a feasible set of
damage scenarios that can all be examined in parallel by
flooding simulations and the newly proposed method within
reasonable computational times.
For each SOLAS damage case and initial condition, a
damage scenario is defined and a flooding simulation is
performed by NAPA. After the completion of each flooding
simulation, the parameters needed to calculate the SPI
index are obtained. Both formulations of the SPI index are
calculated simultaneously.
Finally, after the s factors and the survivability performance
indices (SPI) are obtained, the attained indices A and Asim
Fig. 1: Flowchart of the new damage stability assessment procedure
DESCRIPTION OF STUDY SHIP MODELS
General
The initial ship study model is a ROPAX vessel of about
117.00m in length and 19.50m in breadth. The depth to the
car deck is 6.70m from the baseline and the top of the
RORO space is at 12.25m from the baseline. The deepest
subdivision draft is at 4.90m and the corresponding initial
freeboard is 1.80m. The number of persons on board is
1560 persons, including the crew members. The ship is
subdivided by 13 transverse bulkheads extending to the
main car deck and there is no lower hold. The employed
can be calculated, for each damage scenario. The
summation of the attained indices from all damage
scenarios yields the final cumulative attained index, as
given in the equations. It needs to be pointed out that the
attained subdivision index A according to SOLAS is the
probability that the ship survives a collision damage, while
the ASIM index is a survivability performance index
expressing more precisely the stability and floatability
behavior of the vessel after flooding within a set time
frame. The comparison of both above attained indices, for
each studied vessel type (ROPAX and pure passenger ship),
should enable to detect the general trend of the actual
survivability performance of each ship against the estimated
performance given through the current SOLAS framework.
hull geometry model extends till the upper limit of the
RORO space, disregarding the superstructures above.
Starting from the above ROPAX parent model, a Pure
Passenger vessel/ cruise type vessel has been generated,
while keeping the external hull and ship's capacity the same.
A new deck was introduced at a height in between the
original main deck (freeboard deck) and the original 2nd
deck, which forms the uppermost limit of the modeled hull.
The main deck (former car deck) and the new deck are
disposed for the accommodation of passengers and crew.
Two structural bulkheads are extended above the main deck,
forming the boundaries of the fire zones needed. A total of
25
38 large passenger cabins and 7 crew cabins are fitted. Also,
some additional public spaces are also provided in the same
region. The number of persons on board, the zone
definitions (subdivision) and the loading conditions are kept
exactly the same with those determined for the ROPAX;
this enables the identification of differences in the
calculated attained index purely due to differences in the
flooding of the above main deck spaces.

The main particulars of the vessel are given in the next

table:
SHIP PARTICULARS
Length Between
LBP [m] 117.18
Perpendiculars
Length Overall LOA [m] 124.20
Breadth B [m] 19.50
Design Draft T [m] 4.90
Car Deck Height D [m] 6.70
Displacement [t] 6400
Table 1: Ship principal particulars Plan 2: Pure Passenger vessel general arrangement plan

General Arrangement Openings

The ship is divided by thirteen transverse bulkheads below
the car deck. Also a ducktail is fitted on the aft end of the
vessel. The general arrangement plans of the two vessel
types are given next.
Regarding the definition of unprotected openings in NAPA,
they were herein modeled as geometric objects. This
enables the software to automatically retrieve the necessary
information concerning the lowest point of the opening and
the cross flow area, which is needed for the flooding
simulations. Fitted ventilators in the RORO space are
generally considered as unprotected external openings and
have an effect on the GZ curve determination. Internal
openings on the car deck were also taken into account and
this helped to define a compartment connection table that
allows the flooding of a dry compartment when the opening
is immersed. Even though these openings are not
necessarily watertight, they were herein considered as such,
so as to prohibit the change on the properties of the GZ
curve.

The same external openings were also used for the

assessment of the Pure Passenger vessel. The internal
openings were herein defined as open watertight doors and
stairs, connecting the adjacent compartments of the new
passenger decks.

The SOLAS assessment requires only the position of the

Plan 1: ROPAX general arrangement plan
lowermost point of each opening and the opening type, in
order to determine the characteristics of the GZ curve, at
every flooding stage and calculate the survivability factor,
accordingly. For the flooding simulations, though, more
information is needed. Firstly, the area of the opening needs
to be defined, as well as the pressure loss on the opening.
The area of the opening is automatically calculated based
on its geometry. The pressure loss is expressed as a
discharge coefficient which reduces the water flow through
the opening. The discharge coefficients are set to 0.5 for
external openings, 0.75 for internal openings and 0.65 for
breach openings. The pressure of the water at the lowermost
point of the opening, the area and the discharge coefficients
allow the calculation of the water flow through the opening,
using a modified Bernoulli equation solution. The opening
type is indifferent for the flooding simulation, since the

26
progressive flooding is inherent to the flooding simulation floodwater volume is determined by the internal free water
concept. surface height and the compartment's geometry. In the final
stage, the internal and the external waterlines are in
Concerning a damage cases generation in the flooding equilibrium and the flooding process is assumed completed.
simulations, some remarks are due at this point. In the Contrary, in the time dependent flooding simulation
NAPA software environment the damage definition is not approach, the time dependent flow through the openings is
topological/geometrical, in the sense that there is not a determined. At each time step, a modified Bernoulli
geometric object representing the breach (damage). The equation is used to estimate the flow through every opening
definition can be made in NAPA by either automatically or and the corresponding water transfer is calculated. The
manually selecting the compartments that are affected in distribution of floodwater leads to a new floating position
each damage scenario or geometrically aided, using a for the vessel and to a new pressure head at the lower point
breach type or defining the extents of the damage breach of the opening and this iterative process continues until the
and allowing the software to decide on which external and internal pressure heads are balanced.
compartments are flooded, while the theoretical geometry
used is not defined as a real geometric object, in the NAPA Initial Loading Conditions
software. In both cases the damage definition results in a
group of compartments that are floodable, with reference to The same initial loading conditions were used for the
the level of the external waterline. This is the concept evaluation of the two vessel types, in order to make the
followed in the SOLAS assessment. However, when alteration in the survivability performance due to the
flooding simulations are performed, regardless of the change of part of compartmentation (above the main deck)
method employed, the definition of the initial condition of visible. The characteristics of the initial loading conditions
the vessel is crucial for the solution of the problem. The are given in the following table:
software allows the flooding to begin either from a NAPA Initial Condition DL DP DS
damage or an opening. In the first case, a new floating
T0 [M] 4.060 4.564 4.900
position after the flooding of the rooms defined in the
damage description is calculated, by the equalization TR0 [M] -0.798 0.000 0.000
between the internal and external water heights and the HEEL0 [DEG] 0.0 0.0 0.0
progressive flooding after the equalization is studied via the DISP0 [T] 5007.3 5790.9 6404.8
flooding simulation. When the flooding starts from an LCB [M] 51.913 53.199 52.827
immersed opening, the flooding simulation begins from the TCB [M] 0.000 0.000 0.000
initial floating position of the vessel. Therefore the breach VCB [M] 2.376 2.637 2.838
must be defined as a geometric object in order to allow the GM0 [M] 2.841 2.281 2.530
software to calculate the water flow through it. GM [M] 2.841 2.281 2.530
KG [M] 8.80 8.86 8.57
The previous discussion makes evident that a passage KMT [M] 11.641 11.141 11.100
from the traditional SOLAS damage definition to a Table 2: Initial conditions description
geometric breach definition is necessary to carry out the
required flooding simulations in NAPA. This is herein RESULTS
achieved by firstly registering the damaged compartments,
from each SOLAS damage definition. Secondly, the
SOLAS Assessment
registered compartments are separated into two categories,
namely the external compartments directly accessible by
In this section the SOLAS assessment results for the two
outside damages and which may be directly hit by the
vessel types are presented.
breach and those which may be within the damage extent,
but they are internal in the sense that they are protected by
Subdivision Length: 123.134 m
the external compartments. The compartments outside the
Breadth at the load line: 19.500 m
reach of damage extents, are interconnected as described in
Breadth at the bulkhead deck: 19.500 m
the compartment connection table, which remains active
Number of persons N1: 0
and their flooding will be taken into account during the
Number of persons N2: 1560
progressive flooding phase.
Required Subdivision Index: R=0.78400
The definition of the lowermost point and the area of the
Attained Subdivision Index- ROPAX: A=0.81850
damage opening can be automatically established, since
Attained Subdivision Index- Pure Passenger: A=0.81932
each opening refers a fully defined geometric NAPA object.
In that way, the theoretical damages prescribed in the
SOLAS evaluation are interpreted as geometric objects, SOLAS ASSESSMENT
representing the breach opening. W*P*V*S
Damages Difference
ROPAX Pure Passenger
One of the main differences between the SOLAS approach 1-Zone Damages 0.38808 0.38808 0.000%
and the herein advanced flooding simulations is the way of 2-Zone Damages 0.36477 0.36549 0.197%
determining the floodwater accumulation. The SOLAS 3-Zone Damages 0.05591 0.05599 0.143%
approach suggests multiple phases of floodwater 4-Zone Damages 0.00854 0.00856 0.234%
accumulation (intermediate stages), where at each phase the 5-Zone Damages 0.00120 0.00120 0.000%

27
 A-Index Total 0.81850 0.81932 0.100%
Table 3: SOLAS assessment for both ship types
Pure Passenger Ship

The attained indices of both designs are by a good margin Similarly, the results obtained for the Pure Passenger vessel
higher than the required one, thus both vessel types comply are summarized in Table 5 and Fig. 3.
with regulations 6 and 7 of SOLAS 2009. The results also
show that the difference between the attained subdivision PURE PASSENGER SHIP RESULTS
index for the ROPAX and the Pure Passenger vessel is
SOLAS 0.80955 0.000%
practically zero or merely 0.1%, despite the differences in
ATTAINED INDEX SPIFINAL 0.87394 7.953%
their design above the main deck. Thus, this difference
should account for the existence of internal openings in the SPIMAX 0.86768 7.180%
above main deck spaces that can mitigate the spreading of Table 5: Pure Passenger ship SOLAS and simulations results
water, as well as the free surface effects on stability. It is
evident that this is not captured by the traditional SOLAS
approach and the two vessels demonstrate about the same
safety level against capsize, what is not reflecting reality2.
Flooding Simulation Assessment

The procedure followed for the assessment by use of the

flooding simulations was earlier described. For each
damage case and breach opening a flooding simulation is
performed by NAPA and the SPI index is calculated
according to both formulations. The new attained
simulation index is calculated as the product of w, p and v
factors and the SPI. The same methodology is repeated for
the Pure Passenger vessel.

For the flooding simulations, we assumed a realistic seaway

environment. For the latter, a regular wave pattern is
assumed. The incident, transversely propagating waves, is
assumed having a wave height of 2.00 m and a wave period
of 5.5 sec. By considering this seaway environment
scenario, we may expect a more pronounced flooding and
free surface effect on ship's main deck (Water On Deck
problem). The obtained results are presented below,
separately for the ROPAX and the Pure Passenger vessel
and then comparatively.

ROPAX

The results for the ROPAX vessel are given in the next
table and the following figures. The percentage difference
given in the fourth column is the difference between the
attained index via simulations and SOLAS A, normalized
by the SOLAS attained index.
ROPAX RESULTS
SOLAS 0.80876 0.000%
ATTAINED INDEX SPIFINAL 0.82541 2.059%
SPIMAX 0.81462 0.725%
Table 4: ROPAX SOLAS and simulations results

The attained index ASIM using the SPI index, following the
formulation with the final heeling angle is denoted as
SPIFINAL, while the attained index related to the SPI index
with the maximum heeling formulation is denoted as
SPIMAX. The correlation between the SPIMAX and SPIFINAL
against the SOLAS s factor are given in Fig. 2, for the case
of the ROPAX ship.

2
A difference could have been detected if the Stockholm
Agreement Water On Deck procedure would have been
implemented in the probabilistic SOLAS approach

28
 ROPAX: SPI- SOLAS s factor
1

0,9

0,8

0,7
R = 0,8852
0,6 R = 0,9026
SPI

0,5

0,4

0,3

0,2

0,1

0
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1
SOLAS s

SPIfinal-s SPImax-s Linear (SPIfinal-s) Linear (SPImax-s)

Fig. 2: SPI indices against SOLAS s factor for various damage cases of the ROPAX ship

Pure Passenger Vessel: SPI- SOLAS s factor

1

0,9

0,8

0,7 R = 0,6759
R = 0,6811
0,6
SPI

0,5

0,4

0,3

0,2

0,1

0
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1
SOLAS s

SPIfinal-s SPImax-s Linear (SPIfinal-s) Linear (SPImax-s)

Fig. 3: SPI indices against SOLAS s factor for various damage cases of the Pure Passenger vessel

29
Comparative Results angle against the SOLAS s factor are given in Fig. 4 and Fig.
5, for both ship types.
In this section the results obtained for the ROPAX and the
Pure Passenger ship are compared. Initially, the SPI index
based on the final heeling angle and the maximum heeling

SPI final - s SOLAS

1

0,9

0,8
R = 0,6759
0,7

0,6
SPI final

0,5 R = 0,8852

0,4

0,3

0,2

0,1

0
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1
SOLAS s

ROPAX PPF Linear (ROPAX) Linear (PPF)

Fig. 4: SPIfinal formulation against SOLAS s factor for ROPAX and Pure Passenger Ferry (PPF)

SPI max - s SOLAS

1

0,9

0,8

0,7

0,6 R = 0,6811
SPI max

0,5

0,4
R = 0,9026
0,3

0,2

0,1

0
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1
SOLAS s

ROPAX PPF Linear (ROPAX) Linear (PPF)

Fig. 5: SPImax formulation against SOLAS s factor for ROPAX and Pure Passenger Ferry (PPF)

30
The next table summarizes the results obtained from this
comprehensive study, along with the differences calculated
for each method and design alternative.

ATTAINED INDICES COMPARISON BY METHOD AND DESIGN ALTERNATIVE

METHOD/ DESIGN ROPAX PURE PASSENGER DIFFERENCE (ROPAX BASE)
SOLAS 0.80876 0.80955 0.099%
SPI (FINAL HEEL) 0.82541 0.87394 5.880%
DIFFERENCE (SOLAS BASE) 2.059% 7.953% -
SPI (MAX HEEL) 0.81462 0.86768 6.513%
DIFFERENCE (SOLAS BASE) 0.725% 7.180% -
Table 6: Attained indices comparison by method and design alternative

surface effects are herein slightly mitigated and thus

DISCUSSION
the transient heeling angles are limited.
4. In Fig. 3 the SPI indices, based on both formulations,
The most important results of this study are summarized in
are plotted against the SOLAS survivability index s.
Table 6. Following the traditional SOLAS approach a
The R2 values are less than 0.7, therefore the
negligibly higher attained subdivision index for the pure
correlation between the survivability performance
passenger vessel is recorded, when compared to the
indices and the survivability factor is herein very weak.
ROPAX of the same size and capacity, whereas the herein
The SPI formulations lead to higher attained indices
proposed new survivability performance index (SPI)
and consequently they account for a better survivability
approach yields a significantly higher index indicating an
of this vessel type.
enhanced survivability behavior for the pure passenger
5. In Fig. 4 and Fig. 5 the SPI based on the final heeling
vessel.
angle and the SPI based on the maximum heeling angle
are depicted against the SOLAS s factor. Some
An elaboration of the obtained results is given in the
comments need to be made, so as to explain the
following.
difference between the two simulation indices and the s
factor. As easily observed, in the Pure Passenger vessel
1. The survivability of the ROPAX design is judged based
case, the SPI results are more scattered for zero s factor
on the results presented in Table 6 and Fig. 2. The
and s factor unity, comparing to the ROPAX case.
attained index based on the SPI, taking into account the
Starting from the s factor unity region, where in both
final heeling angle criterion, is the highest of the
cases the SPI indices give results different from unity.
attained indices for the ROPAX design. Comparing the
The reason for this is that in the SPI formulation the
two SPI indices, which are based on the simulations, it
heeling angle is directly implemented, while in the s
is obvious that the attained index based on the
factor it is indirectly accounted for, through the first
maximum heeling angle yields always a lower index.
stable equilibrium angle predicted in the GZ curve.
This is explained by the fact that the heeling angles for
Consequently, the development of any heeling angle
the intermediate stages of flooding can be quite
has a direct effect on the value of the SPI and leads to
excessive. These heeling angles are closely related to
values lower than unity. When the final heeling angle is
the free surface effects, which are more intensive
taken into account for the SPI calculation, the
during the initial and generally transient phases of the
scattering is less excessive both for the ROPAX and the
flooding process. The percentage difference between
Pure Passenger ferry. That is an indication of the
the two formulations of the SPI, based on the SOLAS
existence of excessive heeling angle during the
results, is about 1.3% which may be considered
intermediate flooding stages, with a consequent more
significant.
severe impact on the SPI. As for the region of zero s
2. Fig. 2 shows a good agreement between the SPI indices
factors, exaggerated differences are observed for the
and the SOLAS s factor, for the ROPAX case study.
case of the Pure Passenger vessel. This is actually what
The R2 values are about 0.9, with the highest value
is identified as the enhanced survivability of this type
corresponding to the SPI that takes into account the
of vessel, which retains its stability even though the
maximum heeling angle value. This suggests that the
properties of the GZ curve yield zero s factor. For the
SPImax approach differs from the SOLAS s-factor
ROPAX case, in most cases the SPI and the s factor are
approach only slightly, at least for the ROPAX case
nullified simultaneously. That is the explanation why
study.
the pure passenger vessel displays significant rise in
3. Considering the Pure Passenger vessel alternative, the
the attained index, compared to the SOLAS assessment
difference between the attained indices from the
results.
simulations and the SOLAS assessment is more
6. In Table 6 the differences between the two design
pronounced. This is an indication that while an
alternatives are recorded, by the evaluation method
enhanced survivability performance of this vessel type
used. As seen in the table, the difference between the
is present, the traditional SOLAS s factor approach
attained subdivision indices for ROPAX and pure
fails to account for it. Also, considering that the
passenger ferry, based on the SOLAS approach is about
difference between the attained indices from the
0.1%. When it comes to the attained indices from the
simulations is 0.8% leads to the conclusion that the free
flooding simulations, the results give a difference of

31
 5.88% for the final heeling angle formulation of SPI
and 6.51% for the maximum heeling angle
formulation. This suggests that the traditional SOLAS
assessment method predicts comparable survivability
capability for both ship types, while the SPI deriving
from flooding simulations of the real physical
phenomenon predicts an enhanced survivability for the
Pure Passenger vessel. Furthermore, the difference of
the attained indices from the simulations, for the two
ship types, is higher when the maximum heeling angle
is used for the SPI calculation. This is an indication of
the importance of the transient heeling angles, which
are significant especially for the large unobstructed
RORO space.
CONCLUSIONS
In this study a new index for the assessment of the
survivability performance of a vessel, via flooding
simulations, is introduced. The SPI index approach takes
better into account the main ship design attributes that
characterize the behavior of a vessel throughout the
flooding process and assesses ships survivability within a
set time frame, which may be defined by SOLAS. The
conducted comprehensive study on a ROPAX and a
modified Pure Passenger vessel of the same size and
capacity suggests the following conclusions:
1. While the present SOLAS probabilistic approach to the
attained subdivision index is practically insensitive
with respect to the compartmentation of spaces above
the main deck, when using the new SPI approach for
the assessment of the survivability through flooding
simulations, a significant difference is observed, when
assessing a ROPAX and an equivalent Pure Passenger
ship design.
2. The obtained attained indices via the SPI, flooding
simulations, are generally higher than the attained
subdivision indices obtained by the SOLAS approach,
which is more conservative.
3. The percentage difference between the SOLAS
assessment and the new SPI approach is for the
ROPAX much lower than the corresponding difference
for the Pure Passenger ship. This suggests that for the
ROPAX ship both formulations deliver similar results,
while for the Pure Passenger the SOLAS approach is
more conservative.
4. When the new SPI formulation is based on the
maximum heeling angle, the obtained attained index is
slightly lower than the index obtained when the final
heeling angle is used. That signifies the importance of
the transient heeling angles developed, which are
related to free surface effects during flooding.
5. Comparing the differences between the ROPAX and
Pure Passenger ship for the two formulations of SPI, it
seems that the higher percentage difference occurs for
the maximum heeling angle formulation. This is
understandable, considering that the denser
compartmentation leads to the free surface effect
limitation and slower compartments flooding.
Consequently, the transient heeling angles are reduced
and that is expressed as the higher percentage
difference between the two arrangements.
32
6. The new Survivability Performance Index approach
better accounts for differences in the design
arrangements, based on calculated parameters from
numerical simulations conducted by a standard naval
architectural software package like NAPA; it
introduces aspects of the Time to Sink in the
assessment process, which are presently not considered
in the s factor formulation and the SOLAS approach.
7. Present Required Subdivision Indices of SOLAS for
passenger ships do not consider basic differences in
ship design, i.e. between ROPAX and Pure Passenger
ships of equal size and capacity, which is reasonable
from the point of view of setting a safety level for the
risk of People On Board. The present study suggests
that there are significant differences in the actual
survivability performance of these two basic ship types
and this can be readily demonstrated by the new SPI
approach elaborated in the present paper
REFERENCES
[1] Dafermos, G., "Investigation on Damage Stability
of ROPAX and Pure Passenger Ferries and
Flooding Simulations," Diploma Thesis, School of
Naval Architecture and Marine Engineering,
National Technical University of Athens, NTUA,
Athens, 2015.
[2] Dankowski, H. and Krger, S., "Dynamic
Extension of a Numerical Flooding Simulation in
Time-Domain," in 12th International Conference
on the Stability of Ship and Ocean Vehicles,
Glasgow, UK, 2015.
[3] Fonfach, J. M., Manderbacka, T., and Neves, M.
A. S., "Numerical sloshing simulations:
Comparison between lagrangian and lumped mass
models applied to two compartments with mass
transfer," Ocean Engineering, vol. 114, pp.
168-184, 3/1/ 2016.
[4] GOALDS. (2009-2012). Goal Based Damage
Stability-EU FP7 RTD project.
[5] IMO-Amendments To SOLAS Regulations II-1/6
And II-1/8-1,Computerized Stability Support For
The Master In Case Of Flooding For Existing
Passenger Ships, Guidelines On Safe Return To
Port For Passenger Ships, Amendments To Solas
Chapter II-1 And Associated Guidelines On
Damage Control Drills For Passenger Ships,
IMO-SDC3/WP.4, 2016.
[6] Kanellopoulou, A., "Study of damage stability
regulations and their impact on the design and
safety of ROPAX ships with the use of NAPA
program," Diploma Thesis, School of Naval
Architecture and Marine Engineering, National
Technical University of Athens, NTUA, Athens,
2013.
[7] Manderbacka, T. and Ruponen, P., "The impact of
the inflow momentum on the transient roll
response of a damaged ship," Ocean Engineering,
vol. 120, pp. 346-352, 7/1/ 2016.
[8] NAPA. (2014). NAPA for Design Manuals for
Release 2014.3.
[9] Papanikolaou, A., Boulougouris, E., and
Sklavenitis, A., "Investigation into the Sinking of
 the RO-RO Passenger Ferry S.S HERAKLION,"
in 11th International Conference on the Stability of
Ships and Ocean Vehicles, Athens, Greece, 2012.
[10] Papanikolaou, A., Hamann, R., Suk Lee, B.,
Mains, C., Olufsen, O., Tvedt, E., Vassalos, D.,
and Zaraphonitis, G., "GOALDS-Goal Based
Damage Stability of Passenger Ships," in SNAME
2013 Annual Meeting, Seattle, 2013.
[11] Papanikolaou, A., Spanos, D., Boulougouris, E.,
Eliopoulou, E., and Alissafaki, A., "Investigation
Into The Sinking Of The Ro-Ro Passenger Ferry
Express Samina," International Shipbuilding
Progress, vol. 51, 2004.
[12] Ruponen, P., "Progressive Flooding Of A Damaged
Passenger Ship," Doctor of Science in Technology
Doctoral Dissertation, Department of Mechanical
Engineering- Ship Laboratory, Helsinki University
of Technology, 2007.
[13] Spanos, D., "Simulation of flooded ship motion in
waves and assessment of Ro-Ro passenger ship
behavior," PhD Doctoral Dissertation, School of
Naval Architecture and Marine Engineering,
National Technical University of Athens, Athens,
2002.
[14] Spanos, D. and Papanikolaou, A., "On the time
dependence of survivability of ROPAX ships,"
Journal of Marine Science and Technology, vol.
17, 2011.
[15] Spanos, D. and Papanikolaou, A., "On the time for
the abandonment of flooded passenger ships due to
collision damages," Journal of Marine Science and
Technology, vol. 19, 2014.
[16] Zaraphonitis, G., Papanikolaou, A., Roussou, C.,
and Kanellopoulou, A., "Comparative study of
damage stability regulations and their impact on
the design and safety of modern ROPAX ships," in
13th International Ship Stability Workshop, Brest,
2013.

33
 A Method for Describing
Ocean Environments for Ship Assessment
Eirik Eisinger, Jens Bloch Helmers and Gaute Storhaug
DNV GL
ABSTRACT
Ship traffic data (AIS) and weather reanalysis data are
emerging technologies for the shipping industry. This paper
examines a method for describing ocean environments for ship
assessment by combing AIS and reanalysis data. The quality
of AIS and reanalysis data is studied, and possible applications
for the proposed method are analyzed and explained.
The AIS data are found to contain some noise in terms of both
wrong and lacking data entries and propose filters to deal with
this lack of quality. These filters seem to filter out enough of
these bad data points for the results to be credible. Reanalysis
data show a trend much similar to buoy measurements, but
seems to underestimate the occurrence of extreme values.
The applications studied using this method are: making scat-
ter diagrams based on trades and vessel type, weather routing
effects, and correlations between significant wave height, ves-
sel speed and heading. It is argued that these applications are
applicable in moderate ocean environments and trend studies.
The ERA-Interim model uses a sparse grid, if the proposed
method would be used with a reanalysis model of finer resolu-
tion there is reason to believe that the method may also work
better for extreme values. Much longer observations are
needed for proper extreme value statistics.
INTRODUCTION
Todays industry standard for environmental data for ship
applications are taken from the BMT Atlas (BMT Atlas,
2011). The BMT atlas provides wave data based on visual
observations done from aboard vessels in the period
1949-1986. The data is statistically handled for improved
quality and is stored for different nautical zones (DNV GL,
2007, appendix B)
Automatic Identification System (AIS) has been mandatory
for most commercial vessels since 2005. The system pro-
vides vessels that are within a range of 20-40km of each
other with data. The most essential AIS data is found in
Table 1. Satellite-based AIS receivers have been used to
store data for all AIS equipped vessels, and provide AIS
data for more than 120 000 vessels. Environmental data is
taken from ERA-Interim, a reanalysis model provided by
the European Center for Midrange Weather Forecast
(ECMWF).
34
Table 1: Essential AIS fields
Field Description
MMSI Unique Identification number
IMO Number International maritime organization
number
Position Longitudinal and latitudinal position
Speed Speed over ground in knots
True heading Relative to true north in degrees
Course over ground Relative to true north in degrees
Studies have been performed on both ERA-interim data and
AIS data. The contribution to this field of study is the com-
bination of the two datasets. The combination of these data
sources could give a good estimate of the environmental
conditions experienced by operating vessels in their exact
position, and gives a range of new applications. This paper
is based on results from a thesis called a method for de-
scribing ocean environment for ship design (Eisinger, 2016).
Before taking a closer look at applications, the method will
be discussed and validated.
METHOD
ERA-Interim is a global reanalysis model calculated on a
reduced Gaussian grid that has spacing of approximately
79km, and with calculation performed for every six hours.
To reduce the complexity of the problem and to help with
calculation speed, an interpolated version of the model has
been used. The interpolated version has a uniform spacing
of one degree for both longitude and latitude axis. The
model provides environmental data like: significant wave
height Hs, average zero crossing period Tz, wave direction ,
wind speed and temperature from 1979 up to today.
ERA-Interim uses 30 wave frequencies and 24 wave direc-
tions, giving the model a total of 720 degrees of freedom.
(ECMWF, 2014)
The AIS data have used is provided by vesseltracker. They
provide data every 10 minutes whenever data is available.
The coordinate system used for position is a geographic
coordinate system, and knots (arc minutes per hour) is used
for speed.

AIS data is used as a basis to combine these sources of

information. Interpolation is performed on the ERA-Interim
data to estimate the environment data to the time and posi-
tion of each vessel. Fig.1 illustrates this process where a
fleet and a time interval is chosen to get a position interval
for each vessel in the fleet. The reanalysis library is then
called to provide the environmental data for these positions
and times.

AIS data is used to find time and position for all vessels of
interest. These times and positions are then used to estimate Fig.2: Time interval between AIS loggings for four AIS
environment data. vessels.

Fig.2 shows the time interval between AIS loggings for four
vessels. The figure shows a time interval of up to 70 hours.
Within this interval there is no AIS data available. The figure
shows that interpolation on AIS data must be done with
caution since the uncertainty grows with the time interval.
Our greatest challenge related to AIS data is that the data
only goes back to 2012. At the time of writing this gives
roughly three years of data. In order to analyze unlikely
events a much larger time span is needed. Preferably 10
years or more.

Fig.1: Process of combining AIS data with environmental ERA-Interim Data

data. In order to validate the ERA-Interim data, moored buoys
have been logged since 3 march 2015. The buoy data is
taken from the national data buoy center (NOAA, 2016).

DATA VALIDATION

In this section the data quality of AIS and ERA-Interim data

is studied. It is also explained how to solve some of these
challenges.

AIS Data
The quality of AIS data can vary. The time interval between
neighboring entries can vary depending on if there is a
receiver in the area. Sometimes vessels report their position
with the wrong IMO number. Two equal data entries are
sometimes stored in the database, positions can be stored
with NaN, and sometimes important fields like speed are
missing.

These are all challenges that need to be considered and

addressed, and filters have been implemented which
cleanses the data for most of the addressed issues, but with
120 000 vessels in the database there is bad quality data
present that is not detected and filtered out. Because of the
vast amount of data used, this may not affect the result
significantly.
Fig.3: Comparison between buoy data and ERA-Interim
data for a buoy in the Bering Sea.
In this section a buoy in the Bering Sea is used as an exam-
ple. The buoy is hereafter referred to by its ID, which is
46073. This is the buoy where the most severe wave heights
were detected among all the logged buoys. Fig.3 shows a
direct comparison between ERA-Interim data and buoy data

35
in the period 3 march 2015 to 26 may 2015. The figure
shows a similar trend between the methods for both Hs and
Tz calculations. The absolute difference is included in light
grey, with the corresponding mean value and standard devi-
ation in the upper right corner of each plot. Notice that the
absolute difference of the wave direction has a mean value
of 23o and a standard deviation of 25o which gives this
value much uncertainty. This high uncertainty can partly be
explained by the interpolation performed for ERA-Interim
data. Buoys are logged every hour and hence ERA-Interim
data must be interpolated to fit all buoy time steps. Notice
also how buoys measure more extreme values than the
ERA-Interim model (see for example Hs for 9 April.
This can be seen more clearly in Fig.4, where all Hs values
are stored in a histogram for buoy 46073. The tail of the
distribution is zoomed into in order to show extreme values.
One can clearly see how ERA-Interim data measures lower
values than buoy measurements. This could be because
ERA-Interim uses a sparse grid and shows only the average
value within the grid. A finer grid can detect local varia-
tions which again could result in more accurate extreme
values (Bitner-Gregersen et al., 2016).
Fig.4: Hs histograms for buoy data and ERA-Interim data.
One can see how buoys measure higher waves than the
ERA-Interim model in extreme sea states from Fig.4. One
can also see that the number of occurrences for the lowest
values is less for ERA-Interim than for Buoy data. The
reason for this could be that the grid of the ERA-Interim
model is too sparse.
APPLICATION
A study of the challenges and limitations of the proposed
method have been shown in previous sections. The follow-
ing section will focus on possible applications of the meth-
od. All results presented in this section is using AIS data
that are 150 nautical miles from shore and each vessel has a
time step of three hours.
36
Scatter Diagrams
Fig.5: Scatter plot for Tz and Hs for all container vessels in
the AIS database sailing in the North Atlantic.
Scatter diagrams are widely used in ship design, as a meas-
ure of the weather conditions a vessel is likely to experi-
ence. The proposed method can potentially provide scatter
diagrams which are based on the weather actually experi-
enced by operating vessels. These scatter diagrams can be
made with arbitrary filters like: vessel type, vessel length,
age, etc. Container vessels and oil tankers in the North
Atlantic Ocean are used as examples. Fig.5 shows a scatter
plot for Tz and Hs for all container vessels in the North
Atlantic. The histograms show the distribution for Hs and
Tz. A scatter diagram could be extracted from this by
counting occurrences in every cell (cells are marked with
dotted lines).
The histograms included in Fig.5 can be fitted with three
parameter Weibull plots. The complementary cumulative
distribution function then gives the probability for exceed-
ing a given value.
The scatter diagram used as industry standard today is
called Rec.34 and uses data from four zones in the BMT
Atlas. Fig.6 compares this data with our method. The figure
also includes data from all the ERA-Interim values in a
nautical sector going back to 1979. In Fig.6 this is shown as
reanalysis (RA) data. The black line at approximately 10-5
shows the 25 year value, meaning a vessel is likely to reach
this value during its life time. Note that our method shows a
25 year value up to 4 m lower than the industry standard.
Even the RA data gives lower Hs values than the Rec.34
data. This is another indication that the ERA-Interim model
underestimates values. Notice also that oil tankers have a
25 year value which is one meter higher than for container
vessels. This indicates that container vessels are routing
more than oil tankers. Rec.34 data is based on visual wave
observations from 1949-1986, which would not include as
much routing as is common today, but even when Rec.34
data are a bit conservative our data is too low, as can be
seen in the RA data which does not take routing into ac-
count.
Oil Tankers Container Vessels
Fig.6: Probability of exceeding for four zones for which
Rec.34 is defined. Rec.34 data is compared with our AIS-
RA method and all Reanalysis data found in each zone.
Routing
Routing is the effect that a vessel does a maneuver to avoid
bad weather. This can be done with speed reduction, speed
increase, moving into sheltered areas, change of course, etc.
In order to measure to what degree vessels are routing on a
big scale, an obstacle free area is chosen for the analysis.
This is done in order to know that when a vessel is moving
away from the shortest distance path it is most likely because
of weather conditions. Fig.7 is showing the boundaries for
such an area. The dashed line is showing the route taken by
an AIS vessel in this area, while the dotted line is showing
the shortest distance path between the end points. The figure
also shows the actual heading and the end point heading of
the vessel in four points. The difference between the lengths
of these routes gives an indication of how much the vessel
was routing while crossing this area. The difference between
headings also gives such an indication, but does not take
drifting into account. This method is therefore not consid-
ered as robust as the distance comparison.
Fig.7: Area with no obstacles. The shortest distance between
the end points can be compared with the route taken by the
vessel.
Fig.8 is showing the routing rate (routing distance dr divid-
ed by shortest distance ds) for all crossings done by con-
tainer vessels. This figure varies with season, as winter
storms in the North Atlantic are most severe and gives good
reason for routing.
Fig.8: Number of crossings versus routing rate (routing
distance divided by shortest distance), for container vessels
on open sea.
Two scatter diagrams can be extracted based on the routing
crossing Sr and the shortest distance crossing Ss. The rela-
tive difference between both scatters can be calculated
using :
The result is shown in Fig.9. The result illustrates at what
conditions captains start routing. The transition where the
relative difference goes from negative to positive gives an
idea of when captains typically start routing. The relation
can be written mathematically as:
. The line produced by the linear equation can
be seen in Fig.9
Fig.9: Relative difference scatter between a routing scatter Sr
and the shortest distance scatter Ss.
37
Hs Speed Relation
The proposed method makes many new relations available,
like relative wave direction vs Hs, Tz vs ship length or Hs vs
speed. Filters can be applied in order to look at these rela-
tions in specific areas, for specific ship types or size.
The Hs speed relation will now be studied as an example,
but other relations can also be extracted.
Fig.10: Hs speed relation for following sea, beam sea and
head sea.
One can see from Fig.10 that speed is reduced as Hs is in-
creased. The effect is most significant for head sea. This may
differ from other sources like Barhoumi and Storhaug (2013)
suggesting that the speed may increase in stern seas at least
for moderate wind speeds.
CONCLUSIONS
The study was set out to examine a method for describing
ocean environments for ship design, by combining ship
traffic data (AIS data) with weather reanalysis data. As the
maritime industry is making a move towards data smart
solutions, AIS databases grow every day and reanalysis
models are becoming better, this is believed to be an im-
portant subject of study. The following questions have been
sought to be answered:
What is the data quality of AIS data, and how can
the data be filtered to improve quality?
What is the data quality of reanalysis data, and
what limitations does this bring?
What are possible applications for such a method?
It has been found that AIS data contain some noise in terms
of both wrong and lacking data entries, and it has been
found challenging to do comparison between different AIS
vessels as their data points are not given for the same time.
Reanalysis data have proven to have a trend much similar
to buoy measurements, but seem to underestimate extreme
values. This proves to be critical when comparing 25 year
values for reanalysis data, buoy data and data taken from
the industry standard Rec.34, as our method gets values
much lower than the methods used to compare.
This paper has looked at three main applications: scatter
diagrams, weather routing and correlations between signif-
38
icant wave height, vessel speed and heading. All applica-
tions have given informative and reasonable results, some
of which are confirmed by captains and vessel owners.
This leads to the conclusion that although AIS data contain
errors, enough of these can be filtered out that in spite of
the poor data quality in some of the cases, the vast amount
of data makes the poor quality data statistically insignifi-
cant as there are many other data points of better quality.
The reanalysis model failing to reproduce extreme values
makes it difficult to use this method for extreme value sta-
tistics without handling extrapolation. Since this is an im-
portant application there is a need to improve the reanalysis
model. The grid was simplified, going from a reduced
Gaussian grid for which the reanalysis was originally done,
to a geographic coordinate system. This was to make the
interpolation scheme faster and less complex. It is recom-
mended to use data in the original grid and to do compari-
sons. This would be likely to improve the method, but the
result would still be based on the coarse grid used by
ERA-Interim. Comparing results with a fine resolution
model like NORA10 would therefore also be necessary in
order to know what difference the grid resolution makes.
Another limitation to the method is the short time range for
which AIS data has been stored. The database will be ex-
tended with time, but it might also be possible to extend the
data backwards in time.
Only vessel position data and time is needed in order to
estimate the weather environment, and values such as
heading and speed can be found based on the change in
position. GPS systems get hold of vessel positions and
these systems became available to the public in the 1980s.
The AIS database could therefore be extended for vessels
where GPS data is made available.
The proposed method can provide an important contribu-
tion to the study of ocean environments for ship assessment,
but the method must still be improved to make it useful for
the whole range of applications. Currently, the method must
be limited to fatigue or trend studies, but the scientific
community is likely to develop better models and with
more computer power reanalysis models will likely im-
prove. With better reanalysis models and more AIS data it
is believed that these principles can solve a range of ques-
tions that are difficult to solve with the current industry
standards.
The data can also be used to study other assumptions which
are important in the design of ships. For fatigue, for in-
stance the time in port or sheltered water can be a factor to
limit the fatigue damage, requiring the scatter diagram to be
based on the time at sea.
ACKNOWLEDGEMENTS
The authors would like to thank DNVGL for support in this
study. Eisinger would also thank the supervisor at UiO and
at DNVGL, in particular Helmers for contributions towards
the master thesis.
REFERENCES

Barhoumi, M. and Storhaug, G. (2013), PRADS

Bitner-Gregersen, E.M. et al. (2016) Sea state conditions

for marine structures analysis and model tests,
http://www.sciencedirect.com/science/article/pii/S0029801
816300105

BMT Atlas (2011),

http://www.globalwavestatisticsonline.com/

DNV GL, Y. (2007), RP-C205, Environmental Conditions

and Environmental Loads ,
https://rules.dnvgl.com/docs/pdf/DNV/codes/docs/2007-10/
RP-C205.pdf

ECMWF IFS Documentation - Wave Model, Y. (2013),

http://www.ecmwf.int/sites/default/files/IFS_CY40R1_Part
7.pdf

Eisinger, E. (2016), A method for describing ocean envi-

ronment for ship design.

IMO (2016),
http://www.imo.org/en/OurWork/Safety/Navigation/Pages/
AIS.aspx

NOAA (2016),
http://www.ndbc.noaa.gov/

Vesseltracker (2016),
https://www.vesseltracker.com/en/static/Company.html

Wikipedia-AIS (2016),
https://en.wikipedia.org/wiki/Automatic_Identification_Sys
tem

39
Authors will not be present
at the conference

A Short Ship Design Risk Analysis using the Monte Carlo Method
Elena-Gratiela Robe-Voinea*, Alexandru Pintilie**, Raluca Vernic***
*PhD Student at Mathematical PhD School, Faculty of Mathematics and Computer Science, University of Bucharest and Project Planner at
SNC Ship Design-Constanta Shipyard, Email: elena_robe@yahoo.com

** PhD Assist. Prof. at Faculty of Maritime, Industrial and Mechanical Engineering,

Ovidius University of Constanta, Email: apintilie@yahoo.com

*** PhD Assoc. Prof. at Faculty of Mathematics and Computer Science, Ovidius University of Constanta and Institute for Mathematical
Statistics and Applied Mathematics, Bucharest, Email: rvernic@univ-ovidius.ro

ABSTRACT
The primary focus during the development of a ship
basic/detailed design is to pay attention to an important
aspect called risk. Taking into consideration that build-
ing a ship implies a lot of risks, an important duty is to
prevent them by maximizing the probability and con-
sequences of positive events and, in the same time, by
minimizing the probability and consequences of adverse
events related to the projects objectives. A well-known
method such as Monte Carlo simulation is quite often
used to analyze the risks in a project development. This
paper aims to present the method using an ongoing ship
design project for a petroleum chemical tanker.
INTRODUCTION
According to studies regarding the evolution of naval
constructions, as evidenced in Market Research Reports,
the industry of naval constructions has registered a series of
major fluctuations lately. Ship orders have constantly risen
during a four year period, from 2003 to 2007 and plum-
meted in 2008-2009, as a consequence of the global finan-
cial crisis. After a slight comeback in 2010, the naval in-
dustry starts experiencing problems again in 2011.
The slow evolution of ship building, worldwide speak-
ing, reflects the weak economical foundations, the debt
crisis within the euro zone and the lack of loans in maritime
transport. The ship building industry is faced with a dark
perspective due to the constant growth of naval shipyards
efficiency, modest investments in the field and also the
market saturation.
Building new ships means substantial costs, a great
consumption of raw matter and materials, significant hu-
man resources as well as launching new dedicated tech-
niques and technologies. Consequently, to start a new pro-
ject, a risk analysis is required to reflect as accurate as pos-
sible the implications and consequences of all the factors
which are about to participate in the task achievement.
RISK ANALYSIS BASIS
Designing a new ship involves the identification and
assuming multiple risks, dependent on the environment, the
mobility of the production process, long execution terms,
the large number of coordinated activities and the complex-
ity of the entire process as well as contractual relations with
third parties.
The risk is the possibility of an event occurring, usually
an unfavorable one, bearing consequences on the cost per-
formances (it usually leads to further expense), on the qual-
ity (not rising up to the necessary quality specifications)
and on time (not meeting deadlines). In other words, risk
can be defined as uncertainty associated with a desired re-
sult.
Risk can appear in three situations:
the event occurs, but the result is uncertain;
the effect of an event is known, but the result is un-
certain;
the event is uncertain as well as the result.
Risk analysis uses the theory of probabilities and it con-
sists of studying all variables of an objective which can
contribute to the obtaining of acceptable performances,
considering all the factors already discussed. This analysis
can be made quantitatively, through stochastic methods by
quantifying the possible value dispersion as previsions in
terms of cost and/or time, and from a qualitative point of
view, through a formal investigation, based on judgments,
experience and diagnostic methods using checklists.
The risk element in a project is represented by any ele-
ment which possesses a measurable probability to deviate
from a well-defined plan. Any element of such a structure
can be used as a potential risk element. The more repre-
sentative a structure is, the more significant the risk ele-
ments will be taken into consideration.
Thus, to achieve a goal, a series of activities must be run.
An activity, (a), can be considered a risk element if two

40
Authors will not be present
at the conference
simultaneous conditions are met: RISK ANALYSIS METHODS

Qualitative methods
(1)
Here, P(a) represents the probability that the event (a) The qualitative method of risk analysis implies using
will occur; E(a) represents the effect of event (a) on the qualitative criteria, using various categories for separating
project; L(a) represents the financial evaluation of E(a). parameters, with definitions establishing the range for every
Generally, evaluating risks means covering the follow- category. Qualitative analysis is based on expert opinions, it
ing stages (see Roman and Andreica 2012 ): doesnt require specialized instruments and it is easily ap-
sensitivity analysis; plied. Interpreting results is clear to most people, not re-
critical variables probability distribution; quiring special knowledge, and the conclusions are suffi-
risk assessment; cient. Qualitative risk analysis elements consist of risk
probability, risk impact and risk assessment matrix (see
evaluating acceptable levels of risk;
Ionescu 2010).
risk prevention. The risk probability is defined as the probability of that
risk occurring. According to the probabilities theory, the
RISK MANAGEMENT ASPECTS probability of the event (a) is defined as the ratio of the
number of favorable results (m) of the event (a) and the
Risk management is a cyclic process which unfolds total number of the event results (n), considered equally
throughout the entire project and it consists of covering at possible, that is
least three distinct stages, as it can be observed in Table 1.
(2)
Table 1 Risk management stages
Nr. Risk management
Activities Risk impact points to the effect that risk has upon the
Crt. stages
RISK - identify the risk project objectives, if it occurs.
1.
IDENTIFICATION - identify significant risks The risk probability and impact, I(a), are seen as very
- eliminate incongruous risks high, high, medium, low or very low.
- in-depth analysis of significant Risk score, S(a) is determined using the formula
2. RISK ANALYSIS risks
- result assessment (3)
-probability assessment
-expected value determination
To determine the actual scores of every risk, one must
REACTION TO - risk reduction
3. draw up the risk score matrix, as can be seen in table 2.
RISK - risk elimination
- risk distribution
Table 2 Risk score matrix
Risk impact
In the risk identifying stage one must evaluate potentials Probability
perils, as well as effects and apparition probabilities in or- 0,05 0,10 0,20 0,40 0,80
der to decide which specific risks need prevention. Actually, 0,90 0,05 0,09 0,18 0,36 0,72
at this stage, one must identify all the elements which meet
the criteria (1). At the same time, one must eliminate in- 0,70 0,04 0,07 0,14 0,28 0,56
congruous risks that are those elements with low probabil- 0,50 0,03 0,05 0,10 0,20 0,40
ity of occurrence or with insignificant consequences. This
means that we shall not take into consideration those ele- 0,30 0,02 0,03 0,06 0,12 0,24
ments whose P(a) or L(a) value tend to zero. 0,10 0,01 0,01 0,02 0,04 0,08
In the risk analysis stage one must take into considera-
tion the risks primarily identified and another in-depth
analysis must be drawn up. To perform a risk analysis, one The significance of the data in Table 2 is as follows:
must use various mathematical instruments, ranging from S(a) < 0,05 = low impact (green);
probability analysis to Monte Carlo analysis. 0,05 S(a) < 0,15 = medium impact (yellow);
In the reaction to risk stage one takes action to eliminate, S(a) > 0,15 = high impact (red).
to reduce or to distribute risks. These actions are carried out The qualitative approach is subjective but it allows a
according to a risk management plan which includes the higher generalization degree, being less restrictive.
procedures to be used in risk management, designating in-
dividuals responsible for certain actions in risk domains, Quantitative methods
resources assigned to this purpose, as well as assessing re-
sults obtained in risk management. The quantitative risk analysis implies the evaluation of
the risk value through numerical methods. Through quanti-
tative risk analysis one takes into consideration the numer-
ical evaluation of probability and impact of each risk upon
the entire project. To this end, several quantitative methods
are employed, such as decision trees analysis or the Monte
Carlo simulation.

41
Authors will not be present
at the conference
The quantitative analysis cannot be made without the
qualitative analysis to determine the initiating events. The
method requires a high volume of data, very well prepared
professionals and expensive software.
Decision trees method represents decisions and random
events, in the way they are viewed by the decision makers.
As represented in Fig. 1 (Ionescu 2010), each probable fu-
ture event (represented by a circle) is matched with an ac-
tion (a square) which can be taken by the decision maker,
resulting in a tree-like graph. The actual occurrence of a
future event may mean that one of more actions needs to be
taken, and the sum of all probabilities equal to unity.
Fig. 1 Decision tree
The significance of notations in Fig. 1 is as follows: D
decision point; E probable event; U profit according to
various uncertainty degrees.
The Monte Carlo analysis, elaborated in 1940, uses re-
peated random sampling to generate simulated data to use
with a mathematical model (see Roman and Andreica 2012,
Kroese et al. 2013). In this context, the simulation is repre-
sented by the approximation process of a result through the
random and repetitive use of the algorithm.
Monte Carlo simulation allows project managers to in-
corporate uncertainty and risk in project planning unlike
other risk analysis methods which cannot quantify these
aspects equally well.
The Monte Carlo simulation is frequently applied, due
to its acknowledged advantages, both by the practitioners
(Hulett and Avalon 2015) and by the academic community
(Vose 2008).
The quantitative approach is much more objective and
accurate, but one must mention the fact that results can be
affected by the precision and the validity of input parame-
ters. Due to this fact, the quantitative results mustnt be
considered exact numbers, but estimative ones, with a var-
iable scale depending on the quality of the data used in the
analysis.
42
MONTE CARLO SIMULATION FOR A 41000 TDW
SHIP DESIGN CHEMICAL TANKER PROJECT
In this study case we consider a short part of a ship de-
sign project which was planned to start on January 4, 2016
and scheduled to be completed on July 18, 2017 at a cost of
$120.000.
Preliminary
The planning of the project was developed in the pro-
gram software Primavera P6, having initially a level 3 work
breakdown structure (WBS) as in figure 2.
Fig. 2 Basic design project schedule in Primavera P6
After the planning was made and the Baseline was as-
signed, the next step was to export our project into a soft-
ware able to help us to make a risk analysis. In consequence,
we imported it in Primavera Risk Analysis (Fig. 3).
Fig. 3 Basic Design project schedule imported to Primavera
Risk Analysis
In order to start an analysis we must assure that aspects
such as: finish date, duration and cost are well completed
and imported from P6.
First of all we will we evaluate the risk only under the
effect of duration uncertainty using two distribution types:
triangle and uniform.
The triangular distributions is commonly used to model
the duration of a task. Its simple set of parameters make it
easy to relate to real life. Requires minimum, maximum and
most likely duration, having the following density
(4)
This distribution is often skewed to the left. This is be-
cause a lot of tasks cannot physically be completed in less
than a certain duration, but all tasks can generally be delayed
Authors will not be present
at the conference
for any number of reasons. This leads to the minimum du-
ration being closer to the most likely than the maximum
duration. A long thin long tail on the triangular distribution
models a range of things that could go wrong but are un-
likely.

The Uniform distribution is used when only the extremes

of uncertainty can be specified and where intermediate val-
ues have equal probability of occurrence (or where no in-
ference can be made as to the likely shape of the distribu-
tion). Its density is given by

(5)
The software allows us to simulate all mentioned above
using also others distribution types such as: Normal, Beta
Pert, Lognormal, Discrete, Trigged, cumulative, etc.
In the second part of the simulation the project will be
constrained in addition and by a risk register. All these is-
sues will be detailed forwards.
Fig. 5 Uniform distribution finish date
From Fig. 5, we can observe that if we chose a uniform
distribution the probability to finish in time is more opti-
mistic, 98%. This is due to the shape of the Uniform distri-
bution, but we would expect a Triangular distribution to be
more realistic for such a study.

Simulation 1

Table 3 Set uncertainties to activities

Distribution
WBS Name Min. Likely Max.
type
Structural design doc-
Triangle 25% 75% 100%
uments
Ships equipment, out-
Triangle 70% 105% 120%
fitting
Accommodation
Triangle 85% 110% 130%

The percentage assigned to the minimum, likely and Fig. 6 Triangle distribution cost
maximum are applied to the total duration/ activity.

Fig. 7 Uniform distribution cost

The costs for Simulation 2 are almost similarly in both

Fig. 4 Triangle distribution finish date
After running the Monte Carlo simulation, we can ob-
serve (Fig. 4) that the possibility of finishing on time is
90%. Also, we can observe that the P-80 date is now July
04, 2017 for a Triangle distribution, compared with May 30,
2017 for the Uniform distribution.
cases: even though we use a Uniform or a Triangle distribu-
tion (Fig.6 and Fig.7), there are 94% chances for the Uni-
form distribution and 96% chances in case we chose a Tri-
angle distribution to maintain the initial budget.
Simulation 2
In this part, as we mentioned before, we use a Risk

43
Authors will not be present
at the conference
Register (Fig. 9). Every risk is defined by four attributes
such as: probability, schedule, cost, performance and score.
Every attributes has assigned a factor scale (low, medium,
high) in accordance with its project impact factor.

Fig. 9 Risk Register

The Matrix Risk Register (Fig. 10) shows the before

Pre-Mitigation period

We assume risks such as:

Changing the input data Fig. 12 Cost - with the default Risk Register
Leaving employees
Class certification Risk Register - Cost and Duration sensitivity
Logistic database
We can also make a classification regarding the Cost
(Fig.13) and Duration (Fig.14) sensitivity.

Fig. 10 Matrix Risk Register

Fig. 13 Cost

Fig. 11 Risk result for Finish date Monte Carlo simulation

We can easily observe that in this case, the P-80 date is

now August 18, 2017 or almost more than one month from
the scheduled date of July 18, 2017, and that the possibility
of finishing on time is in this scenario only 39% (Fig. 11).
Regarding the costs, there are only 19% chances to fin-
ish the project with the initial budget ($120.000), and a
P-80 posibility to have a cost of $282.448, which means a
$ 162.488 amount difference (Fig. 12).
Fig. 14 Duration

It results that the Leaving employees risk is the most

44
Authors will not be present
at the conference
sensitive with a 85% possibility to happen if we take into Schedule Risk Analysis, Long International Inc.
consideration the cost, while Class certification is the
most dangerous risk if we look at the duration issue. Kroese, D.P., Taimre, T. and Botev, Z.I. (2013), Handbook
of Monte Carlo methods (Vol. 706). John Wiley & Sons
CONCLUSIONS
Vose, D. (2008), Risk Analysis: A Quantitative Guide, 2nd
By adding default risk to the task activities, the follow- Edition, John Wiley & Sons
ing aspects changed (see Table 4):
Ionescu (Mndru), L. (2010), Theoretical research and
Table 4 Result of analysis case studies on risk assessment in an integrated manage-
Item Initial Pre-mitigation ment system for quality-risk industrial companies, PhD
Duration 562 d 586 d Thesis Summary, Transilvania University of Brasov
Cost $120.000 $383.283
Start Date 04/01/2016 04/01/2016 Dorp, J.R.v. and Duffey, M.R. (1997), Statistical depend-
Finish Date 18/07/2017 11/08/2017 ence in risk analysis for project networks using Monte
Carlo methods, Elsevier
After a Pre-Mitigation process, it should follow a
Post-Mitigated process. Roman, M. and Andreica, M. (2012), Paper no. 10- De-
Furthermore, in order to improve the possibility to have veloping risk analysis into the cost-benefit analysis of pro-
a project success, it is indicated - and the experience has jects financed by the ERDF and CF, Bucharest, University
shown this- that it must exist a risk analysis periodically of Economics Studies
throughout all stages of a project, namely: project plan-
ning, engineering, procurement, execution and commis-
sioning.
Finally, in order to improve the perspective of a success,
Hulett and Avalon (2015) recommended a plan which im-
plies the following stages:
Create a summary schedule at Level 2 or Level 3
that is integrated, logically-linked, and complies
with scheduling best practices. The budget or cost
estimate is incorporated into the schedule with
summary time-dependent and time-independent
resources assigned to activities.
Gather data about uncertainty and risk. Experience
has shown that individual confidential interviews
of subject matter experts in the project teams and
management, contractors, and other knowledgea-
ble personnel in the organization, can uncover
risks that are either new or unpopular to discuss
risk workshops.
Use a modern Monte Carlo simulation software
package to model both uncertainties to activity
duration and to time-dependent resources burn
rates or time-dependent resources total material
costs.
Prioritize the Risk Drivers to schedule and to cost
(only schedule was illustrated above)
Prepare a pre-mitigated risk result presentation
Conduct a risk mitigation workshop using the priori-
tized risk contained in the pre-mitigation presenta-
tion as a guide to the most important risks
Gain organizational commitment to those mitigation
actions to be adopted, scheduled, budgeted, staffed,
and monitored.
Prepare a post-mitigation model and simulate it.
Prepare a post-mitigation report.

REFERENCES

Hulett, D.T. and Avalon, A. (2015), Integrated Cost and

45
 The Weather Criterion: Experimental Wind Tunnel Results
Arman Ariffin1, Shuhaimi Mansor2, Jean-Marc Laurens1
1
ENSTA Bretagne, IRDL, Brest, France
2
Faculty of Mechanical Engineering, Universiti Teknologi Malaysia, Malaysia
ABSTRACT
The intact stability code, 2008 IS Code, includes a
weather criterion. The vessel is subject to a lateral
strong wind and a roll motion. The wind force on the
sail, the roll amplitude and the bilge keel effect are
computed according to a set of empirical formulae given
in the 2008 IS Code. The maximum resulting list angle is
then computed using the righting arm curve only,
neglecting any damping. To verify how conservative, the
regulation is, experimental trials have been conducted
in the Low Speed Wind Tunnel of the Universiti
Teknologi Malaysia for two models: a simple academic
shape and the DTMB 5415. The experimental setup is
described in the paper. Results have been obtained for
several wind velocities, initial heel angles, yaw angles
and with and without the bilge keel. The maximum
angles of list obtained are compared with results given
by the code as implemented within the stability code
GHS. As expected, the code is sometimes barely
conservative and sometimes, very conservative.
INTRODUCTION
The International Code of Intact Stability 2008 (2008 IS
Code) is based on the best state-of-the-art concepts
available at the time the Code was developed. Since the
design technology for modern ships is rapidly evolving,
2008 IS Code should be revise continuously and
reevaluated as necessary. Generally, 2008 IS Code consists
of two main criteria. First is the criteria regarding righting
lever curve properties as stated in (IMO, 2009) Ch 2
Paragraph 2.2 and second is the severe wind and rolling
criterion (also known as weather criterion) as stated in
(IMO, 2009) Ch 2 Paragraph 2.3.
In 1939, Rohala (Rohala, 1939) wrote a doctoral thesis
which evoked widespread interest throughout the world at
that time because it was the first comprehensive study and
proposed method to evaluate the intact stability which did
not require complex calculation (Ariffin, Mansor, &
Laurens, 2015). The weather criterion adopted as
Resolution A.562 by IMOs Assembly in 1985 was a leap
beyond the statistical approach of the earlier Rohala-type
general intact ship stability criteria.
46
In the MSC 81st, the guidelines for alternative
assessment of the weather criterion were approved by IMO.
This guideline are aiming to provide the industry with
alternative means (in particular, model experiments) for the
assessment of severe wind and rolling criterion (weather
criterion) (IMO MSC.1/Circ.1200, 2006). It consists of the
guidelines for experimental determination of the wind
heeling lever and angle of roll to windward due to wave
action. Wind test and drift test were explained in these
guidelines.
In the MSC 82nd session, IMO approved the explanatory
notes to the interim guidelines for an alternative assessment
of the weather criterion (IMO MSC.1/Cir.1227, 2007). This
explanatory notes provide an example of the alternative
assessment of severe wind and rolling criterion (weather
criterion) based on a series of model tests following the
Interim Guidelines for the alternative assessment of the
weather criterion for better understanding of the alternative
procedures.
When evaluating the stability of a ship in beam seas and
winds, consideration must be given to the modelling of
wind and wave forces. While much research has been
devoted to the hydrodynamics of ship rolling motions,
relatively little work has been devoted to wind heeling
loads on ships. This situation is particularly surprising when
considering that current intact stability criteria are based
largely upon the heeling of ships under wind forces
(McTaggart, 1992).
Naval Engineering Division conducted a wind tunnel
test for naval ship 378-WHEC US Navy ship. The test
was measuring 4 heeling position (upright) 25, 50 and 75
and 4 wind velocities (6, 12, 18 and 24 m/s). This
experiment concluded that wind heeling moment is about
16% less than calculated value (Paul, 1994).
An experimental wind tunnel to obtain the experiment
results of wind forces and moment acting on ship models at
various angle, with beam wind was conducted in 2003
(Bertaglia, Serra, Francescutto, & Bulian, 2003). The
models used were scale down to 1/125th and wedges have
been applied to the model to simulate the correct heeled
position. The measurement has been performed with an
undistributed wind velocity of 13m/s with Reynolds
number of 2x106 (length reference) and 2.3x105 (breadth
reference). This paper concluded that the real centre of
underwater forces was assumed at half draught; and seems
to be a safe assumption, but further test could be made in
order to verify this assumption.
The draft amendment of the IS Code regarding
vulnerability criteria and the standards (levels 1 and 2)
related to dead ship condition and excessive acceleration
are contained in SDC 3/INF.10 Annex 1 and 2. The level
1 check for dead ship condition is basically the same
method used for current IS Code 2.3 which is weather
criteria. If it failed, the design should process to level 2
check and the direct assessment. Direct assessment
procedures for stability failure are intended to employ the
most advanced state-of-the art technology available either
by numerical analysis or experimental work for quantitative
validation as stated in SDC 1/INF.8 Annex 27 (IMO, 2013).
THE WEATHER CRITERION
The IS Code 2008 Part A 2.3 contains the weather
criterion. The ship must be able to withstand the combined
effects of beam wind and rolling. The conditions are:
a. the ship is subjected to a steady wind pressure
acting perpendicular to the ship's centre line
which results in a steady wind heeling lever (lw1).
b. from the resultant angle of equilibrium (0), the
ship is assumed to present an angle of roll (1) to
windward due to wave action. The angle of heel
under action of steady wind (0) should not exceed
16or 80% of the angle of deck edge immersion,
whichever is less.
c. the ship is then subjected to a gust wind pressure
which results in a gust wind heeling lever (lw2);
and under these circumstances, area b shall be
equal to or greater than area a, as indicated in Fig.
1.
Fig. 1 Severe wind and rolling
The heeling lever shall be calculated using formula:
lw1 =(P*A*Z)/(1000*g*) (1)
lw2 = 1.5 lw1 (2)
where lw1 = steady wind heeling angle, lw2 = gust wind
heeling lever, P = wind pressure of 504 Pa, A = projected
lateral area (m2), Z = vertical distance from the centre of A
to the centre of the underwater lateral area or approximately
to a point at one half of the mean draught (m),
=displacement (t) and g = gravitational acceleration).
Direct Assessment (DA) can be used to verify the weather
criterion for unconventional ships. The DA can be
experimental. The present study shows how such an
experimental DA can be conducted for two models, a
civilian ship and a military ship.
In the weather criterion, two main rules are commonly
used. For commercial ship, it uses the IMO weather
criterion and for naval ship, it uses the Naval Rules. The
IMO Weather criterion is shown in Fig. 1 and the weather
criterion for naval ship is shown in Fig. 2. The significant
different between IMO an Naval Rules are presented in the
Table 1.
Fig. 2 Weather criterion for naval ship
Table 1: Comparison IMO and naval rules for weather
criterion
Criterion IMO Naval Rules
Wind velocity 26 m/s 100 knots
Roll back angle various* 25
WHA constant cos2
Ratio A2/A1 1 1.4
Gust Yes No
* roll back angle (phi1) calculated based on IS Code 2008
# WHA wind heeling arm, A2 - restoring energy, A1 capsizing energy
GHS CODE
In this paper, a software code, General Hydro Static
(GHS) is used. This code is for designing and to
evaluate all types of ships and floating structures. It
addresses flotation, trim, stability and strength by
calculating the force involved using
mathematical/geometrical model of the vessels. In GHS,
the weather criterion can be evaluated with the presetting of
the wind condition. There are two methods to set up the
wind in GHS. The first one specifies the wind speed and
the second is using the wind pressure. The main difference
between both function is that using the wind pressure
function, the wind speed is constant at all height. It means
that there is no boundary layer for this setting. For wind
speed (knots) function, the specified wind speed in knots is
given at 10 meters above water plane, assuming a standard
boundary layer above the sea level. This is a standard
coming from maritime weather forecast (B. Chelton &
H.Freilich, 2006).
Since this paper presents a comparison result with
experimental work, the function of wind pressure was used.
It is because of the boundary layer in the wind tunnel used
in this experiment is about 30mm. This boundary layer
can be neglected because the value is relatively small
compare to the model size in wind tunnel test section.
47
SHIP MODEL Both models were constructed at ENSTA Bretagne,
France using the Computer Numerical Control (CNC)
Two models were used for the experimental work. The machine. The material used was polystyrene. Both models
first model is an academic container ship geometry referred were designed in 3D drawing and imported to CNC machine
as ASL shape in the rest of the paper. The second program for fabrication process. The hulls were divided into
model is a research ship model, the well know DTMB 5415 six parts for the cutting process. Then, all parts were glued
(Molgaard, 2000). The 5415 DTMB model is widely used and laminated with a fiberglass. The superstructure used the
for the research study in seakeeping (Begovic, Day, & synthetic glass. The completed ship models are shown in
Incecik, 2011; Jones & Clarke, 2010; Yoon et al., 2015). Fig. 5.
MODEL VERIFICATION
The main particulars of ASL shape are given in Table 2
and for the 5415 DTMB shape in Table 3. The body plan and To determine the correct centre of gravity, inclining
perspective view for ASL shape is shown in Fig. 3. The tests were performed. The inclining test is a procedure
body plan and perspective view for 5415 shape is shown in which involves moving a series of known weights,
Fig. 4.Fig. 4 Body plan (left) and perspective view (right) of normally in transverse direction (Fig. 6), and measuring the
the 5415 shape resulting change in the equilibrium heel angle of the ship.
By using this information and applying basic naval
architecture principles, the ships vertical centre of gravity
Table 2 Main particulars of ASL shape is determined from the GM. We also verified that the
natural roll period is as expected. Two devices were used
Ship model Ship Model for the data recording, first is the Ardu Flyer device and
LOA, (m) 140 1.400 smartphone (Djebli, Hamoudi, Imine, & Adjlout, 2016).
BOA, (m) 20 0.200
Draft, (m) 12 0.120
Displacement, (tonnes) 26,994 0.027
VCG, (m) 10 0.10
LCG, (m) 70.037 0.70
KM, (m) 10.206 0.10
GM, (m) 0.206 0.002

Table 3 Main particulars of 5415 shape

Ship model Ship Model
LOA, (m) 153.3 1.533 (a)
BOA, (m) 20.54 0.205
Draft, (m) 6.15 0.061
Displacement, (tonnes) 8,635 0.0086
VCG, (m) 7.555 0.076
LCG, (m) 70.137 0.70
KM, (m) 9.493 0.09
(b)
GM, (m) 1.938 0.014
Fig. 5 Complete build ship models (a) ASL shape (b) 5415
DTMB shape

Fig. 3 Body plan (left) and perspective view (right) of the

ASL shape

Fig. 4 Body plan (left) and perspective view (right) of the Fig. 6 Moving weight and transverse distance for inclining
5415 shape test

48

EXPERIMENTAL SETUP
A wind tunnel test was conducted at the low speed wind
tunnel facility at Univerisiti Teknologi Malaysia. This
wind tunnel has a test section of 2m (width) x 1.5m (height)
x 5.8m (length). The maximum test velocity is 80m/s
(160 knots). The wind tunnel has a flow uniformity of
less than 0.15%, a temperature uniformity of less than
0.2C, a flow angularity uniformity of less than 0.15 and a b.
turbulence level of less than 0.06% (Ariffin et al., 2015).
The experiment started with the model placed in the
water tank with the correct draft (Fig. 9). A laser light is
used to ensure the vessel is upright. The test started with
measurement of the stable heel. The wind tunnel velocity
was increased slowly while the heel angle was recorded
using the Ardu Flyer device. The Ardu Flyer is a
complete open source autopilot system designed for 3D
robotics. This experiment involved three models
configuration as stated below:
a. ASL shape.
5415 shape.
c. ASL with bilge keel shape.

Wind tunnel setup A roll back angle (2*) measure was performed for all the
models. The definitions of (1) and (2*) are shown in Fig.
The models were allowed to heave and roll freely. It 11. The test steps are as follow:
was not allowed to yaw because the model must be hold at a. Model placed in water tank.
the longitudinal axis to avoid the model bump to water tank b. Wind applied and the wind velocity and heel angle
side. The models were fixed with a rod both at bow and recorded.
stern (Fig. 7). It is passing through the point of c. Roll back angle (1) applied at the model. (using rod
longitudinal centre of buoyancy. Both rods at bow and in Fig. 10)
stern were aligned using laser light to confirm the shafts d. Then model is suddenly released.
positioned at same axis. The arrangement of rod used in e. The maximum counter roll back angle (2*)
this experiment is frictionless therefore, minimum recorded. The is (2*) is measure from the angle of
interaction between the rod and rod stand can be obtained. stable heel. With referring to the value used in IS
Code 2008, the (2*) can be calculate as (2) - (0).
To allow the model to float in the wind tunnel, a water
tank fabricated with glass of 8mm thickness was installed.
The water tank size is 1600mm x 400mm x 240mm (length
x width x depth). Since the wind tunnel is not water tight, to
avoid any leak of water during the experiment, a dummy
pool was placed underneath the platform. The dummy
pool is capable to cope the total volume of water if the glass
water tank gets damaged. The arrangement in the test
section is shown in Fig. 8.

(a)

Fig. 7 Rods fixed at ship models

(b)
Fig. 9 Ship models ready to be tested in wind tunnel test
section (a) ASL shape (b) 5415 shape

Fig. 8 Arrangement in the test section

49

Fig. 10 Rod used to force the model to heel at roll back angle
(1)
Fig. 11 Definition used in this experiment
Scaling criteria
The models used in the experiment were scaled down to
1:100. It is the same scale used by (Begovic et al., 2011)
for the ship motion experiment using DTMB 5415 model.
For the GZ curve, the model and full scale ship has a same
curve shape but values for the model are divided by 102.
For weight calculation, values used for the model are
divided by 106. For the wind velocity, the value used for
the model is divided by 10.
Boundary layer
When the air flow over the ocean surface from any
direction, a natural boundary layer is formed. This means
that the wind velocity at the surface is zero and increase
with higher altitude. The boundary layer thickness in the
test section for this experiment is about 35mm and the
velocity profile is shown in Fig. 12. The comparison of
boundary layer thickness and the ship models is shown in
Fig. 13. The two lines in this figure shows the water line
and boundary layer thickness.
50
Fig. 12 The velocity profile curve
Fig. 13 Comparison of velocity profile and ship models
RESULT AND DISCUSSION
Angle of stable heel (0) vs wind velocity
Fig. 14 shows the graph for angle of stable heel, 0
versus wind velocity for the two models and two methods;
IMO and experimental. The 5415 curves are following a
parabolic shape since as we can see in Fig. 15, the GZ
curve of 5415 shape follows a linear curve up to 30 degrees.
Furthermore, the experimental curve is below the IMO
curve which indicates that the drag coefficient CD, of the
ship silhouette is smaller than 1, the value assumed in the
IMO formula (Fig. 14). The ASL curves present different
shapes and behaviour. At first, they do not present the
parabolic shape because as we can see in Fig. 15, the GZ
curve is only linear up to 5. Furthermore, the experimental
curve for this case is above the IMO curve (Fig. 14). That
is explained by the fact that the drag coefficient CD, for the
box shape of the ASL is bigger than 1. This can be
confirmed by the many references that exist giving the drag
coefficients of basic shapes, see for example (Sadraey,
2009).
Fig. 14 Graph of wind velocity and angle of stable heel for
ASL shape and 5415 shape on the experimental results and
 GHS calculation
Fig. 15 The GZ curves for ASL shape and 5415 shape
Roll back angle (2*) versus roll to windward (1)
Fig. 16 shows the roll back angle (2*) versus roll to
windward (1) for ASL shape for wind velocity range of 2
m/s to 4 m/s. Fig. 17 shows the roll back angle (2*)
versus roll to windward (1) for 5415 shape. In the absence
of damping the results should be like a swing where 2*
follows 1. The results suggest a far more complex
behaviour where the hydrostatic force shape is playing an
important role.
Fig. 16 Roll back angle (2*) vs roll to windward (1)
for ASL shape
Fig. 17: Roll back angle (2*) vs roll to windward (1) for
5415 shape
Ratio 2* and 1 with bilge keel
Fig. 18 shows the ratio (2*/1) for the ASL shape and
the ASL with a bilge keel. Both models were tested at
wind velocity 2m/s. For the bare ASL, the average ratio is
0.55 and for the ASL with bilge keel, the average ratio is
0.43. As expected, the configuration with bilge keel
contributes to more roll damping than configuration without
bilge keel.
Fig. 18 Roll back angle (2*) vs roll to windward (1) for
ASL shape, 5415 shape and ASL with bilge keel
configuration
Yaw angle effect on stable heel
Fig. 19 shows the angle of stable heel for the ASL and
the 5415 both with the wind direction from star board 75
and port 105. For the ASL, the values of 0 are smaller for
the beam wind than those obtained with the yaw angles. In
other words the assumption of the beam wind in the IMO
code is not necessarily conservative. This phenomenon
also appears for the 5415.
Fig. 19 Angle of stable heel for wind from starboard 75 and
port 105
Effect of roll to windward (1) and roll back angle (2*)
with yaw angle
Fig. 20 shows the result for 1 and 2* for the ASL and the
5415 with beam wind and wind from starboard 75. For
the ASL, the beam wind has higher 2* than wind from
starboard 75 and for the 5415, the beam wind has smaller
2* than wind from starboard 75. The two models have a
different response to the yaw angle. The behaviour is a
combination of the superstructure geometry, the GZ curve
and the damping.
51

Fig. 20 Roll back angle (2*) vs roll to windward (1) for
5415 shape with wind from port 105
Comparison IMO GHS and experimental result
Fig. 21 shows the comparison results between IMO and
experimental results. For the ASL, the counter roll back
angle (2*) obtained from experimental results is 24.07,
lower than IMO which is 29.638. Therefore, IMO result
is more conservative. For the 5415, the counter roll back
angle (2*) obtains from experimental results is 16.31,
lower than Naval Rules which is 33.82 for ratio capsizing
and restoring energy 1.0 and 39.45 for ratio capsizing and
restoring energy 1.4. Therefore, the IMO and Naval rules
are always more conservative.
Fig. 21 Comparison result for IMO rules and Naval Rules
CONCLUSIONS
In this paper the authors presented an experimental
Direct Assessments (DA) of the weather criterion for two
different models; a civilian ship with a simple geometry and
a military ship, the well-known DTMB 5415. To conduct
the experiments, the low speed wind tunnel of UTM was
used. Both models were placed in a water tank in the
wind tunnel. Both models were free to roll so the heel
angle could be measured and compared with the IMO and
Navy Rules.
Although the assumptions taken by the rules are not always
conservative, the final results always show that the
experimental values are lower than the values given by the
rules.
ACKNOWLEDGEMENTS
The authors would like to acknowledge the support of
the Government of Malaysia, the Government of the French
Republic and the Direction des Constructions Navales
(DCNS).
52
REFERENCES
Ariffin, A., Mansor, S., & Laurens, J.-M. (2015). A
Numerical Study for Level 1 Second Generation
Intact Stability Criteria. In International Conference
on Stability of Ships and Ocean Vehicles (pp. 183
193). Glasgow, UK.
B. Chelton, D., & H.Freilich, M. (2006). Comments on
Scatterometer-Based Assessment of 10-m Wind
Analyses from the Operational ECMWF and NCEP
Numerical Weather Prediction Models. Monthly
Weather Review, 134(2), 737742.
Begovic, E., Day, a. H., & Incecik, A. (2011).
Experimental Ship Motion and Load Measurements
in Head and Beam Seas. In Proceeding of The 9th
Symposium on High Speed Marine Vehicles (pp. 18).
Naples, Italy.
Bertaglia, G., Serra, A., Francescutto, A., & Bulian, G.
(2003). Experimental Evaluation of the Parameters
for the Weather Criterion. In International
Conference on Stability of Ships and Ocean Vehicles
(pp. 253264). Madrid, Spain.
Djebli, M. A., Hamoudi, B., Imine, O., & Adjlout, L. (2016).
The application of smartphone in ship stability
experiment The Application of Smartphone in Ship
Stability Experiment. Journal of Marine Science and
Application.
IMO. (2009). International Code of Intact Stability, 2008.
London.
IMO. (2013). SDC 1/INF.8 - Development of Second
Generation Intact Stability Criteria.
IMO MSC.1/Cir.1227. (2007). Explanatory Notes to the
Interim Guidelines for Alternative Assessment of the
Weather Criterion.
IMO MSC.1/Circ.1200. (2006). Interim Guidelines for
Alternative Assessment of the Weather Criterion.
Jones, D. a., & Clarke, D. B. (2010). Fluent Code
Simulation of Flow around a Naval Hull: the DTMB
5415. Victoria, Australia.
McTaggart, K. a. (1992). Wind effects on intact ship
stability in beam seas. Journal of Wind Engineering
and Industrial Aerodynamics, 44(1-3), 24872498.
Molgaard, A. (2000). PMM-test with a model of a frigate
class DDG-51. Lyngby, Denmark.
Paul, H. (1994). 378-WHEC Wind Tunnel Tests by Paul
Hirsimaki Naval Engineering Division (G-ENE-5B).
CG Engineers Digest, 2737.
Rohala, J. (1939). The Judging of the Stability of Ships and
the Determination of the Minimum Amount of
Stability. Doctoral Thesis, Technical University of
Finland.
Sadraey, M. (2009). Aircraft Performance Analysis. VDM
Verlag Dr Muller.
Yoon, H., Simonsen, C. D., Benedetti, L., Longo, J., Toda,
Y., & Stern, F. (2015). Benchmark CFD validation
data for surface combatant 5415 in PMM maneuvers
Part I: Force/moment/motion measurements. Ocean
Engineering, 109, 705734.
 Hull Monitoring Closing the Gap between the Design and Operation
Gaute Storhaug & Adrian Kahl
DNVGL, Maritime advisory
ABSTRACT
Sensor technology used for measuring hull girder
loading is becoming more popular. The leading class
societies have already hull monitoring rules, but the
requirements differ. Many hull monitoring suppliers
exist, but not all systems are approved before installed.
While the main objective of hull monitoring is to ensure
safety and reduce maintenance costs, the specific system
with its software configuration, functionality and
installation may not be optimal for the above objective.
It can therefore be difficult for the owner and yard to
choose the right supplier and right rule sets. This paper
will highlight some of the important aspects and
opportunities with examples. It will also outline a few
development needs.
INTRODUCTION
Sensor technology is widely used, but the potential may
be far from fully utilized. Today, only a small fraction of
ships and offshore structures are equipped with hull
monitoring systems. The number of installed systems is,
however, increasing fast as owners/managers realize the
benefits.
Many class societies have already hull monitoring rules.
These give requirements to hull monitoring systems, which
typically need to be approved before installed on board.
The installation of a hull monitoring system is usually
associated with a class notation. DNV GLs class notation
is called HMON (DNVGL, 2016), which is applicable for
both ships and offshore structures. It is more general than
other similar class notations, and HMON can in principle
include any sensor for any purpose. The main objective is,
however, to include strain sensors to measure hull girder
loading for the purpose of giving onboard decision support
related to safety (extreme loading) and maintenance
(fatigue loading), including the effect of wave induced
vibrations. Another important application can be the
monitoring of accelerations with regard to cargo securing,
e.g. on container and multi-purpose ships. An overview of
rule requirements is given with examples of recent and
expected development.
Regarding safety, the system reveals the status of the
response versus design criteria (including warnings), but
also the change of the response due to human interaction,
i.e. seamanship. This can be crucial as the captains gut
feeling and experience may not be reliable, e.g. for new
ship designs, new ship trades, new loading condition or rare
extreme conditions. Key elements are well calibrated
sensors and intuitive display for decision support on board.
This is illustrated.
Related to maintenance, the display on board will show
fatigue rates. Generally, fatigue may not be well understood
by the crew, so the fatigue rate is given a clear definition in
order to build awareness also to the effect of wave induced
vibrations.
Data may be transferred to shore for further big data
analysis. Ship owners may share data with class in relation
to class related services, e.g. risk based inspection, life time
extension, hull integrity management and condition
assessment program. The objective is to ensure safety and
reduce and optimize the inspection efforts and plan
maintenance. Data can also be used to improve assumptions
in ship design rules and class guidelines. Further, the
owner/manager may use the data to improve their company
practice and gain full control, and even as part of
documentation in sale.
Hull monitoring is becoming a standard element in
bridge/control systems on ships and offshore structures and
part of the daily operation for the crew on board and ship
manager onshore. It can provide input to class to ensure
safety and optimize inspection and maintenance costs.
RULE REQUIREMENTS
The hull monitoring rule requirements are found in
DNVGL (2016), Pt.6, Ch.9. Sec.3. These hull monitoring
rules are close to identical with the previous hull
monitoring rules from DNV and GL Legacy before the
merger of the two companies. The first hull monitoring
rules were developed early 1990ies, and in 2005
requirements to identification of the importance of wave
induced vibrations for fatigue and extreme loading were
included. This is one item which separates this rule set from
many others.
The rules text is divided into the following main elements:
53
1. General: This includes scope, application and
definitions. More importantly it includes qualifiers that
needs to be specified by ship operator (owner or
management company) and documentation
requirements to the yard, or supplier, whoever submits
this to class for approval
2. Component requirements: This is referring to general
requirements for any electrical and connected system,
but also to sensors with amplitude ranges, accuracy,
filter frequencies and sampling frequencies
3. System design: This refers to system requirements, e.g.
that un-interruptible power supply should be included in
case of power shut down, but also approximately where
sensors should be located and how data should be
processed with different types of statistics, how the
display should work and how data should be stored. A
Fig. 1 Illustrations of main sensors on an oil tanker
guide to the extent of monitoring depending on ship
type is also provided, but this should also be discussed
For a container ship, there may be additional concerns
with class for the specific ship type.
related to hatch corner fatigue and lashing loads for
4. Installation and testing: Fabrication acceptance testing
container stacks. This can then be covered by additional
should be carried out, and installation and test reports
local strain sensors, L and a motion reference unit, M.
should be delivered. Proper calibration of strain sensors
Engine related sensors, E may also be added, e.g. for
is also an important aspect, which should be witnessed
monitoring fuel consumption. Some of these sensors may
by a surveyor from the Class Society.
already be onboard. The notation may then look like
HMON(A1, C1, E3, G4, L2, M1, O1, W1). There could be
Each qualifier is represented by a letter which refers to a
additional sensors for other purposes related to e.g. engines
specific type or group of sensors. E.g. G refers to global
(systems), propeller-shaft, temperature (fire fighting
strain sensors for hull girder loading. The available number
vessels), and wave conditions, with reference to DNVGL
of qualifies are the largest number of qualifiers given by
(2016).
any rule set. It is thereby considered to be the most flexible
rule set, in principle covering any need.
RECENT AND EXPECTED DEVELOPMENT
The associated class notation is referred to as HMON()
Four elements will be discussed in this section:
where refers to the list of qualifiers and each qualifier
has a number, referring to the number of that type of Monitoring display
sensors. E.g., G4 refers to four global strain sensors. Calibration
Fatigue rate
There are many different combinations of hull monitoring Damping of hull girder vibrations
systems, but a typical minimum sensor set up can be
characterized by HMON(A1, C1, G4, O1, W1) which Recent ship accidents such as of MSC Napoli and MOL
means: Comfort have shaken the container ships industry, and the
A1: One vertical accelerometer close to bow focus on certain load components like whipping has
C1: Connection to loading computer for vertical increased. IACS (2015) have responded by introducing a
bending moment at measurement positions, GM and new longitudinal strength standard UR S11A where
draught whipping is required to be addressed for Post Panamax
G4: Four global deck sensors; two at midship on port container ships according to the individual class society
and starboard side, one at aft quarter length on starboard requirements. The approach is not specified by IACS, but
side and one at forward quarter length on starboard side. already covered by DNVGL rules (DNVGL, 2016b). The
accident report after MSC Napoli also recommended
O1: One navigation sensor such as GPS for position,
research into hull monitoring to handle whipping (MAIB,
speed over ground and course over ground
2008). However, the DNVGLs HMON notation already
W1: One wind sensor for wind speed and heading.
covers whipping, and have done so for more than a decade.
Whipping is basically handled as part of vertical wave
This can be quite useful for an oil tanker, and an illustration
bending. It is included when warning is given at 80% and
of the locations of some of the main sensors are given in
100% of the total moment the vessel is designed against, i.e.
Fig. 1.
the sum of permissible still water and rule wave bending
moments. If whipping is contributing, it eats of the
allowable wave bending moment. There will, however, also
be additional safety margin in the design, so 100% is not
referring to the collapse strength, but to the total design
moment.

54
There are additional aspects that have been recently looked
into. One of them considers the human factor and is related
to the monitoring display on the bridge. It should be
intuitive and easy to understand, focusing first of all on the
main safety aspects. The technology may be new for many
captains, so when coming onboard a new ship the captain or
officer on watch should immediately be able to understand
what the system is doing by looking at the display. An
example of a new display developed in cooperation with
DNVGL is illustrated in Fig. 2. This display has received
much positive feedback from captains. It should be
expected that requirements to the content of the main view
of hull monitoring displays are given in the next rule
revision based on the principles indicated above. For
example, even how the arrow in the display in Fig. 2 moves,
will be an important element, since it cannot follow the
instantaneous loading which varies too fast; whipping
vibrations have a period in the order of 2 seconds. In Fig. 2
the display is relevant for a container ship or a high speed
vessel because of the range from 0 to 100%. For a container
ship hogging is the main concern, while for the high speed
vessel sagging may be the main concern. For an oil tanker
the speedometer will go from -100% to 100%, because
both hogging and sagging are important.
Fig. 2 Mainview display showing hull girder loading
Another aspect is still water bending moment. This was a
topic for both MSC Napoli and MOL Comfort. The
reporting from MOL Comfort was not very clear on this
point, but it was indicated that the ships permissible still
water bending moment was exceeded, but also that the
uncertainty could be 10% (NK, 2014), i.e. possibly even
more overloaded. The uncertainty is an important aspect,
which has been addressed recently by DNVGL in a couple
of projects. One of the elements is related to the loading
computer (qualifier C1), which is often providing the hull
monitoring system with the still water bending moment as
basis for the warning in combination with wave bending
moment. However, in practice on certain ship types it has
been observed that the loading computer can be in
working mode frequently. This may also be related to a
specific type of loading computer. In the worst case on a
container vessel the loading computer did not provide the
hull monitoring system with still water bending moments
for up to 50% of the time. This is not acceptable, and in
such cases the hull monitoring must rely solely on the strain
sensors. A quick solution was to use the latest known
loading condition, but that is not reliable. Thus, proper
calibration of the strain sensors is needed. However, the
loading computer neglects physical effects, and calibration
against the loading computer then needs to be handled with
care. Other physical effects need to be accounted for during
calibration such as:
1. Temperature effects (sun expanding steel)
2. Forward speed (steady wave in bow and stern)
3. Uncertainties in cargo and ballast
4. Sea pressure on the ships end
5. Cargo fluid pressure on bulkheads
6. Hydroelastic effects
7. Heel (distance from neutral axis to sensor)
Most of these effects can be handled relatively easily so
that it can be neglected. For instance the forward speed
setting up a sagging moment can be avoided by performing
the calibration at zero speed. However, the calibration
procedure needs to account for sea pressure at the ships
ends and static hydroelasticity, both which may be
significant according to Storhaug et al. (2016). An
illustration of the static hydroelastic effect, reducing the
still water hogging moment of container ships, is given in
Fig. 3. The static hydroelastic effect comes from changed
buoyancy distribution due to the deflecting hull. It increases
with ship length, and it is in the order of 5%, e.g. for large
container ships.
Fig. 3 Reduction of hogging moment, dM, in percentage of
the hogging moment, M, as a function of ship length
Based on proper calibrated strain sensors, there should be a
deviation between the bending moment from the loading
computer and the measured bending moment. For a
container ship, it is also important that the measured
bending moment is represented by the average of sensor
readings from both port and starboard side, to eliminate
contribution from axial warping stress from the static
torsion moment distribution. If the deviation between the
loading computer and the hull monitoring is large this
should be provided as input information to the captain. That
is done by the vertical columns on the left hand side in Fig.
2, showing both the situation in case the vessel is in harbour
or at sea. In harbour a higher still water bending moment is
allowed and the utilization becomes lower. An explanation
for high deviation could be that the loading computer does
not have automatic reading of ballast tanks and that wrong
filling of ballast tanks have been given as input. For
container ships, the other possibility is that the uncertainties
in the declared container weights affect the actual still water
55
bending moment. It should be expected that the next
revision of the DNVGL rules includes stricter requirements
to calibration of the sensors and how they should be
combined in order to make the comparison with the loading
computer more consistent.

Fatigue damage is a nonlinear process and the crew and

captain onboard may have had little training in
understanding of fatigue loading. They may however have
experience with cracks of the hull structure and the
associated inconvenient repairs. Fatigue damage is caused
by cyclic loading, e.g. induced by waves. By doubling the
loading, fatigue damage becomes at least eight times as
large (power of three). During the last decade also the wave Fig. 4 Five-minute fatigue rates with and without wave
induced vibrations, i.e. whipping and springing, have been induced vibrations in deck on a Post Panamax container
recognized to contribute to fatigue damage, and the amount vessel with a target life of 40
of contribution has been quantified by measurements for
several ship types and sizes and different operational Due to the merger of DNV and GL, and since the topic of
conditions. This effect can more than double the fatigue wave induced vibrations have been high on the agenda
damage in deck of ships, e.g. see Storhaug and Kahl (2015). lately, the two class guidelines dealing with how to account
The main contribution comes from vibrations increasing the for whipping and springing in design of container ships
peaks of the wave induced load cycles (Storhaug et al., have also been merged into a new document (DNVGL,
2007). For the purpose of building awareness for the crew 2015b). In this document also numerical calculations of
onboard a parameter called fatigue rate is required to be wave induced vibration and model tests are allowed for. An
calculated by the hull monitoring system. This is defined as important input parameter, affecting both the contribution
the fatigue damage for a specific time period, like 5 minutes of whipping for extreme bending moment and whipping
or 30 minutes, divided by the average fatigue damage and springing for fatigue damage, is damping. The damping
during this period which gives a damage ratio (Miner sum) affects the resonance level in springing, which may occur
of 1.0 after the target design life, normally being 25 years. more or less all the time, but also the decay of the whipping
A Miner sum of 1.0 corresponds to a through-thickness cycles occurring now and then. An illustration of a
crack after the target life (associated with 2-3% probability). whipping event is given in Fig. 5.
The fatigue rate is then typically below 1.0, but in storms it
can reach 50 to 100. This is displayed on board and the
understanding is that staying at 100 for one day 100 days of
fatigue budget is spent. One should then not encounter
more than three days with such storms during a year. This
will soon build awareness for the officers on watch. An
example of the fatigue rates (each dot) from a container
vessel is illustrated in Fig. 4. In this case fatigue rates of
more than 360 have been reached within a 5 minute interval
based on a target life of 40 years. With a target life of 20
years it would be 180, but is still very large. With duration
of 30 minutes, the fatigue rate will be less affected by
Fig. 5 Illustration of a whipping event
extreme whipping events, but still more than 100 in this
case, which has about 80% contribution from whipping and
It is important to have a damping value as a target value to
springing. One ship operator was interested in keeping the
the numerical analysis or model tests, and this value should
value below 50 based on half hours and 20 years target life.
be based on real life experience. Damping values for
This could be a guide for a captain, because the fatigue rate
container ships or any other ship type are not well covered
may be reduced significantly by avoiding storms or reduce
in the literature, and this represents an uncertainty. A recent
the speed in a storm. The definition of fatigue rate has been
study carried out in cooperation with DNVGL assessing
implemented in the hull monitoring rules (DNVGL, 2016).
more than ten container ship designs, but also other ship
The calculation of fatigue damage is covered by a class
types, confirm that the damping is higher on container ships
guideline DNVGL CG 0129 (DNVGL, 2015).
than other ship types (Edin and Laanemets, 2016). Further,
the sampling rate (min. 20 Hz) and length of the time series
(30 minutes) required by DNVGL (2016) are suitable for
reliable estimates of damping. However, the available
methods for the evaluation of damping are not considered
equally good. The half-band width method should be
abandoned, but spectral method where the spectrum has
been smoothened is regarded reliable but tends to
overestimate the damping slightly. Other more complicated
methods may be more accurate, but difficult to implement,

56
e.g. stochastic subspace identification. This is not very systems
helpful for operation today, but in order to collect damping SST; Naviscan 99 systems; one
http://www.sst21c.com/Eng_Goods_MainFr system approved by
data more quickly and efficiently, the suppliers of hull ame.htm DNVGL
monitoring systems may be required to include methods to BMT Scientific Marine Services; Integrated <10
estimate damping of main hull girder vibrations, including Marine Monitoring System
torsional vibrations for container ships. Two convenient http://www.scimar.com/
Weir-Jones Group; HMONTM <10
methods are then the spectral method or the random http://www.weir-jones.com/products.php?pi
decrement technique. An example of the spectral method d=125&cat=Marine
applied to torsional vibration on a container ship is given in Rouvari Oy; Hullmos <10
Fig. 6. The torsional vibration occurs at 0.28Hz (1.76 rad/s) http://www.rouvari.fi/
in this case and with a damping ratio which from the Totemplus; Hull stress monitoring <10
http://www.totemplus.com/hsm.html
spectrum fit in Fig. 6 was estimated to 5.3%. 2-node HMC; Hull monitoring system 4 systems + several
vertical vibration at about 2.5 rad/s is however more http://www.hmc.nl/mba/hullmosys/ motion systems
important. CPE Systems; Hull stress monitoring system <10
http://www.cpesys.co.nz/
MCA Consultants; Hull monitoring system 8 in service + some
http://www.mcaco.com/products/index.html gone out of service
Miros <10 research systems;
http://www.miros.no/ first of all wave radar
supplier
Marin http://www.marin.nl/web/show <10 research systems
Micron Optics, Hull stress monitoring <10 research systems
system http://www.micronoptics.ru/
Cetena; Sh.A.M.An <10 research systems
http://www.cetena.it/
Automasjon og Data a.s.; HUMS <10 systems
http://automasjon.no/
Fig. 6 Illustration of spectral method by a fit (solid) to the EnduranceConsulting; HSMS <10 systems; navy
measured stress spectrum (dots); fit made to the torsional http://www.endcon.com.au/
vibration mode for a ULCS DNV <10 research systems
GL; GL SeaScout <10 systems

OVERVIEW OF SUPPLIERS AND CLASS NOTATIONS Table 2 Class societies and associated class notations
Class society Class notation
There are many suppliers of hull monitoring, e.g. those ABS HM2
listed in Table 1. The suppliers are given an approximate BV MON-HULL
number of systems delivered. For the suppliers with a low DNVGL HMON()
number of delivered systems < 10 is provided. Additional KR HMS
comments are also included occasionally. As seen from LR SEA(HSS-n)
Table 1, there are few leading suppliers, and the rest of the NK HMS
suppliers have not delivered many standard systems. Even RINA MON-HULL
the class societies like previous DNV and GL have
developed their own systems, which have been used in EXAMPLE RESULTS
research but also partly commercially, e.g. some GL
SeaScout systems are still in operation. Some other class On a 4600TEU Panamax container vessel a GL SeaScout
societies may have done the same. Not all suppliers have system together with a measurement system for research
delivered systems, approved by a class society, despite that was installed and Fig. 7 shows the load history (stress range
many class societies have hull monitoring rules with an spectrum) with and without wave induced vibrations. The
associated class notation. Table 2 lists class notations from stress is taken from a sensor located slightly aft of midship
various class societies. The hull monitoring rules vary in upper deck.
significantly which is a challenge for the suppliers but also
ship operators and yards, and a secondary problem is that
the name of the class notations varies. However, a future
harmonized class notation could earliest be expected in
2019, since there is no ongoing activity in this respect yet.

Table 1 List of suppliers

Supplier; Name of system Number of
systems delivered
Light Structures AS; Sensefib 172; approved systems
http://www.lightstructures.no/
Strainstall UK; StressAlert 300+; approved
http://www.strainstall.com/systems/vessel-m systems
onitoring/stressalert/
BMT Seatech; SmartStress 100+; approved
systems, but stopped
delivering new

57
 Fig. 7 Cumulative spectrum of nominal stress ranges in
upper deck amidships with and without wave induced
vibration

From Fig. 7 it is observed that the stresses are amplified

relatively more due to wave induced vibrations at high
stress levels. Combined with a S-N curve, the damage from
the different cumulative frequency levels can be estimated
as shown in Fig. 8. It is seen that the fatigue damage is
related to lower stress levels occuring frequently. The wave
induced vibrations do not affect significantly the probability
level dominating the fatigue damage. More details are given
in Kahl et al. (2016). The measured effect of wave induced
vibrations on fatigue damage can be used to correct
numerical calculations for all relevant structural details,
such as longitudinal end connections at transverse Fig. 9 Fatigue damage from wave frequencies and vibration
bulkheads as well as deck outfitting details. frequencies for a gas carrier as a function of significant
wave height for stress taken from a sensor amidships just
below upper deck

In Fig. 10 the fatigue damage is displayed versus heading

for these deck details and for only cargo condition. It is
clear that most of the fatigue damage comes from head seas
(sector 1), while not so much from stern (sector 6) and
beam seas (sectors 3 and 4). It is also observed that all
headings have contribution from wave induced vibration,
but that the contribution is most significant from head and
bow quartering seas. More details are included in Storhaug
et al. (2013).

Fig. 8 Importance of fatigue damage at different probability

levels with and without wave induced vibrations

Desipite that Fig. 8 shows that the governing fatigue

damage comes from loads associated with high cumulative
frequencies, it is a common misunderstanding that this
always refers to daily or hourly levels of fatigue loading (it
would also imply that the peaks in Fig. 4 would not matter
which is certainly not the case). In a statistical smeared out
situation in may appear so, but if the damage is shown
differently, e.g. as in Figure 4 or versus the significant wave
height, the understanding is increased. An example taken
Fig. 10 Fatigue damage for the gas carrier in cargo
from a gas carrier is shown in Fig. 9, for which it is
condition as a function of heading from head to stern sea
observed that 5 meter significant wave height is most
important. The wave induced vibrations are contributing at
Previous examples included fatigue loading, but also
all wave heights. The results are not differing between
extreme loading is of importance. An example of the
loading conditions or headings.
extreme dynamic loading measured amidships and at aft
and forward quarter length from a large container ship with
extreme bow flare angle is shown in Fig. 11. The
measurements are normalized with respect to the
longitudinal strength standard IACS UR S11. It should be
noted that this has just been replaced by UR S11A (IACS,
2015) which is slightly stricter and requires whipping to be
addressed for such vessels. It is clear that especially in the
aft part this vessel has encountered high wave loading with
strong contribution from whipping, while the forward part
has good margins. In all cases the wave frequency loading
has been below the design loading. This is because the ship
has not really encountered real severe storms, but it has
been operating at high speed in moderate to small storms.

58
Whipping is quite sensitive to the speed. More details are
given in Barhoumi and Storhaug (2013).
Fig. 11 Deck stress for a large container ship at aft, midship
and foreward quarter length with and without whipping and
normalised with stress based on loads from IACS UR S11
(replaced by UR S11A)
The main point of Fig. 11 is that the design wave load may
be exceeded, i.e. it is important for the officer on watch to
have a functional warning system to stay below the design
loads, although, there are margins against ultimate hull
girder collapse. Further, if the wave loading with whipping
exceeds the design wave load, the total loading may still not
exceed the total design load, when at the same time the still
water bending moment is below the permissible still water
load. It is therefore also important to know the actual still
water bending moment of the ship versus the design
limitation. An illustration is shown in Fig. 12 taken from a
measurement on a large container ships. It demonstrates
that the ship stays below the maximum permissible bending
moment at sea all the time, but that frequently it may be
rather close to it. The time it is above the limitation (100%)
refers to a harbour condition and is not relevant here. The
still water bending moment distribution taken from the
stress rather than the loading computer will also display a
slightly different and more smooth distribution.
Fig. 12 Still water bending moment distribution of a large
container ship versus maximum permissible still water
bending moment midships; taken from the loading
computer
CONCLUSIONS
The number of hull stress monitoring system increases
fast. There are however few standard suppliers with mature
systems approved to the latest hull monitoring rules from
class societies. As the associated class notations also vary,
it is not only important to choose the right supplier, but also
to choose the right class notation. Recent developments
have also been made with focus on good intuitive displays
on the bridge and proper calibration of the strain sensors in
deck including hydroelastic effect and end pressure effects.
Other physical effects must also be handled during
calibration. The hull monitoring systems is also
recommended to also include the determination of fatigue
rates.
A typical minimum installation is illustrated with examples
from such systems. The governing fatigue damage is caused
by loads associated with high cumulative frequencies, but
this does not necessarily imply low associated sea states.
Head and bow quartering seas are most important, but wave
induced vibration contributes to fatigue damage at all sea
states and for all headings. Here the fatigue rates may be
helpful in operation as the crew immediately can relate the
wave loading including wave induced vibrations to budget
fatigue damage.
The extreme loading has relatively higher contribution from
whipping, and occasionally it can exceed design limitations
even though the wave frequency loading stays well below.
It may be difficult to understand when this can happen, and
the seamanship experience may be unreliable in this respect.
A hull stress monitoring system can then be useful to close
the gap between design and operation, both, with respect to
the wave loading including whipping and to verify the
loading computer. The loading computer may indicate
frequent still water loading close to the maximum still
water bending moment without accounting for uncertainties
in the cargo or false input to the loading computer. Hull
monitoring should therefore be used to provide the crew
with decision support on the deviation from the loading
computer.
It should be expected that the hull monitoring rules in IACS
are harmonized within a decade, but with the recent
development, the DNV GL HMON() class notation is
already ensuring safety and building awareness in order to
reduce repair and maintenance costs.
ACKNOWLEDGEMENTS
The authors would like to thank DNV GL for the support
of this paper and previous full scale measurements and hull
monitoring projects.
REFERENCES
Barhoumi, M. and Storhaug, G., Assessment of whipping
and springing on a large container vessel, Proceedings
of the PRADS2013, 20-25 October 2013, CECO,
Chanwon City, Korea
DNVGL (2015), Fatigue assessment of ship structures,
DNVGL-CG-0129, October 2015,
59
 http://rules.dnvgl.com/ServiceDocuments/dnvgl/#!/indus Storhaug. G., Oma, N., Blomberg. B, and Hirota, K., (2013),
try/1/Maritime/1/DNV%20GL%20rules%20for%20class The effect of measured whipping and springing on LNG
ification:%20Ships%20(RU-SHIP) vessels, Proceedings of the ASME 2013 32nd
DNVGL, (2015b), Fatigue and ultimate strength international conference on ocean, offshore and arctic
assessment of container ships including whipping and engineering, OMAE2013-10775, June 9-14, Nantes,
springing, DNVGL-CG-0153, October 2015, France
http://rules.dnvgl.com/ServiceDocuments/dnvgl/#!/indus
try/1/Maritime/1/DNV%20GL%20rules%20for%20class
ification:%20Ships%20(RU-SHIP)
DNVGL, (2016), DNV GL rules for classification: Ships
(RU-SHIP), Part 6 Additional class notations, Ch.9
Survey arrangements, Sec. 9 Hull monitoring systems
HMON, January 2016,
http://rules.dnvgl.com/ServiceDocuments/dnvgl/#!/indus
try/1/Maritime/1/DNV%20GL%20rules%20for%20class
ification:%20Ships%20(RU-SHIP)
DNVGL, (2016b), DNV GL rules for classification: Ships
(RU-SHIP), Part 5, Ch.2, June 2016,
http://rules.dnvgl.com/ServiceDocuments/dnvgl/#!/indus
try/1/Maritime/1/DNV%20GL%20rules%20for%20class
ification:%20Ships%20(RU-SHIP)
Edin, I. and Laanemets, K., (2016), Damping estimation of
wave induced vibrations, evaluation process of damping
estimation methods, Master thesis, Department of
shipping and marine technology, division of marine
technology, Chalmers University of Technology,
2016:X-16/343.
IACS, (2015), S11A, Longitudinal strength standard for
container ships, June 2015, International association of
classification societies, Unified Requirements (UR),
Requirements concerning Strength of ships (S),
http://www.iacs.org.uk/publications/default.aspx
Kahl, A., von Selle, H. and Storhaug, G., (2016), Full
scale measurements and hull monitoring on ships,
Proceedings International institute of welding (IIW),
V1675-15 (XV-1494-15)
MAIB, (2008), Report on the investigation of the
structural failure of MSC Napoli, English Channel on 18
January 2007, Marine Accident Investigation Branch,
Report No. 9, April 2008.
https://www.gov.uk/maib-reports
NK, (2015), Investigation Report on Structural Safety of
Large Container Ships, The investigative panel on large
container ship safety, ClassNK, September 2014,
http://www.classnk.or.jp/hp/pdf/news/Investigation_Rep
ort_on_Structural_Safety_of_Large_Container_Ships_E
N_ClassNK.pdf
Storhaug, G., Aagaard, O. and Fredriksen, ., (2016),
Calibration of hull monitoring strain sensors in deck
including the effect of hydroelasticity, ISOPE2016
TPC-0160, The twenty-sixth International Ocean and
Polar Engineering Conference, Rhodes, Greece, June
26-July 2, 2016
Storhaug, G. and Kahl, A., (2015), Full scale
measurements of torsional vibrations on Post-Panamax
container ships, 7th International conference on
Hydroelasticity in marine tehnology, pp. 293-310, Split,
Croatia September 16-19 2015.
Storhaug, G., Moe, E. and Holtsmark, G., (2007),
Measurements of wave induced hull girder vibrations of
an ore carrier in different trades, Journal of offshore
mechanics and arctic engineering, Volume 129, issue 4,
pp. 279-289, doi:10.1115/1.2746398

60
 Accident Scenario-based Rapid and Interactive Damage
Control System using Coded Shortcut Keys
Hee Jin Kang, Dongkon Lee, Jin Choi
Korea Research Institute of Ships and Ocean engineering (KRISO), Republic of Korea

ABSTRACT the perspective of operators. The table 1 shows a plan for

using a DCS according to five Ws and one H.
When an accident happens on a ship, effective dam-
age control activities are required. Many large ships, Table 1 Required action items for damage control system
especially naval vessels, are equipped with damage con-
trol systems to mitigate damage to life, environment, Necessary measure
and property. In spite of the damage control systems, Immediately upon Check, spread damage condition
When
most onboard recovery activities are still carried out by accident Composition of damage control crew
the crews. In any case, the final decision-maker is Where Accident scene
Input damage control crew
someone on the ship. Therefore, the key to damage con- Conduct damage control
trol rests with well-trained crews on the ship during the Who Crew Damage control
early phase of an accident. For effective use of damage Composition of damage control man-
control systems, each system has its own symbols and ual
static and dynamic kill cards, including many that pre- Establishment of ground cooperation
What Damage control
system for rescue
dict calculation functions for ship performance. How- Record, analyze, and improve
ever, too much information on a graphic user interface, onboard training results
and too many complex tasks that must be done in a very Severe life, envi-
short time for damage control, lower the usability of a ronment, and asset Predict accident influence according
damage control system. For this reason, this paper uses Why damage occurs to time flow
when initial re- Decision-making for damage control
a systems engineering approach to guide the develop- sponse fails
ment of a design alternative for existing damage control Create damage control system based
systems. To guarantee the success of damage control on operator requirements
activities in the early phase of an accident, coded Through system Gain certification to operate ship load
shortcut keys are considered for effective communica- that reflects vari- Establish organic cooperation system
How ous accident sce- of command crew, scene response
tion between members of the damage control team. A narios and onboard team, and ground team
suggested damage control system based on coded training results Establish database for preliminary
shortcut keys is developed using the results of a re- analysis and decision-making with
quirement analysis. The proposed design concept makes regard to flooding and fire accidents
effective damage control possible with rapid and inter-
active kill cards. Therefore, current DCSs were improved based on survey
results of stakeholders related to ship sailing and by ana-
INTRODUCTION lyzing cases of existing accidents. This direction was im-
plemented within regulations and laws related to ship dam-
Damage control systems (DCSs) that minimize life, en- age control.
vironment, and asset damages in naval vessels or large pas-
senger ships are in operation. However, cases that prove the ANALYSIS OF DAMAGE CONTROL SYSTEM BEING
utility of DCS have not yet been reported for severe acci- OPERATED
dent cases such as the rubber boat bomb on the destroyer
USS Cole or the grounding accident of the Italian Costa Most DCSs on large passenger ships and naval vessels
Concordia. Quick external support is difficult in marine are being operated as subsystems to engineering control
accidents, and onboard accident situations must be immedi- systems (ECS). DCSs are linked with onboard fire and
ately controlled. Life, environment, and asset damage can flooding sensors, doors, hatch openings, closing infor-
spread quickly and severely if accident information is not mation, remotely controlled valves, and kill cards that pre-
quickly grasped and responded to. Therefore, it is necessary sent sequential damage control procedures to respond to
to analyze the functions of currently operated systems and various accident situations. A DCS linked with ECS checks
determine how to improve damage control systems from mission operability restrictions according to the damage,

61
and includes functions for responding to attack situations
on vessels and evaluating basic operations on ships such as
destination sailing during accident situations.
Interviews and a survey were conducted for 20 items
dealing with each function of a damage control system.
These included onboard utility, correspondence for safety
management duty procedures, regulations related to func-
tional improvements, possibilities for reflective systems,
and factors that should be considered during ship design by
assuming fire and flooding accidents on a ship with 310
sailors (crewmen), safety management, a regulation system,
and design field experts. Based on the results, requirements
for improvement, and design plans for the effective use of a
DCS, were presented.
ANALYSIS OF REQUIREMENTS FOR FIRE-FIGHTING
Regulations and guidelines for damage control are pre-
pared and used for each shipping company, ship classifica-
tion (IMO 1997; KRS 2004; IMO 2007; Bgh et al. 2009).
The Fig.1 shows an example of the procedures for fire-
fighting and abandon-ship drills, and basic factors to con-
sider in configuring a DCS.
Check ventilation closing of main compartments, and
extinguish areas when fire spreads
Stage 3: Mustering and preparation
Check duty contents of scene damage control crew
according to procedure and remote support
Simplification or automation of personnel check and
report procedure
Stage 4: Strengthen entry into the fire zone, and fire-
fighting
Accurate and quick recording and coordination of ac-
cident situation and response situation
Inspect main extinguishers, emergency extinguishing
pumps, and fuel oil and electrical blocking with re-
mote support when needed
Provide decision-making support information for set-
ting and resetting fire location and extinguishing area,
extinguishing team entrance and exit routes, timing
Stage 5: Rescue operation
Support confirmation of casualty location, suffer in-
formation, and injury information
Support quick transportation of the wounded, and
share information with nearby ships and rescue ships
for treatment

Stage 6: Use of a fixed fire-extinguishing system

Use fixed fire-extinguishing system, check remaining
personnel, and support evacuation when required
Check for blocking of ventilation in areas and remote
support

ANALYSIS OF REQUIREMENTS FOR FLOODING

DAMAGE CONTROL
Fig.1 Firefighting and abandon-ship drill (KRS 2009)
Regarding damage control for flooding, situations where
The requirements and required information for damage DCS can be used to prevent the spread of accident damage
control, by each stage related to fire damage control, are were analyzed by investigating the results for the Italian
summarized from IMO (1997) and KRS (2004) as below: Cost Concordia. The requirements for an effective damage
control are listed in Table 2.
Stage 1: Reporting bridge of initial fire
Register fire accident information, and call and check Table 2 Required actions for flooding damage of a ground-
nearby crew ing damage case (IMIT 2012)
Announce fire alarm considering scene-reported con-
tents, fire class, and size according to location; and Fields that can be improved when
Time Event Heeling
confirmation of scene through other identification applying DCS
tools such as CCTV and reports by crewmen
Monitor site situation and check smoke and heat Check previously defined
accident scenario simulation
spreading route by considering how the fire spreads 09:45
Collision
database
Monitor site situation and check kill card for fire- with Scole
p.m.
reefs Respond according to Kill
spreading situation Cards(damage control guide-
lines)

Stage 2: Initial actions to be taken by an initial caller DCS is used in damage control
Respond according to kill card (decision making of 09:50 Blackout regardless of blackouts (owing
p.m. onboard to separate emergency power
response team in onboard accident scenes, record of
and independent operability)
damage control procedures prioritizing fire- extin-
guishing activity, guidance for decision-maker)

62
 Check ship behavior and risk
of capsize by flooding steps
and considering progress of
flooding
Set waterproof area to prevent
risk of capsize
Recommend evacuation of
passengers to shelter deck
when risk of ship capsize is
judged to be high, and call
crewmen in each area to sup-
10:36 Ship drift-
05 port evacuation
p.m. ing
Perform rescue signaling ac-
cording to defined damage
control procedures
Order mustering on shelter
deck, and transfer information
on accident situation to nearby
ships and ground units (marine
police, VTS automatically)
Call crewmen in each area,
and check remaining personnel
and the wounded
functions were selected by considering the results of a pri-
ority analysis of requirements, technical implementation
difficulties, and the effects of damage control. The func-
tions present a direction for improving existing DCSs.
Table 3. Selected TPMs for effective use of DCS
TPMs
Support shortcut key input to activate main
TPM 1
functions
User inter- Provide wireless communication system
TPM 2
face (WICS)-linked usage onboard
Provide information to predict accident influence,
TPM 3
such as fire and flooding, based on database
Smoke, heat spreading route, and tendencies
TPM 4
during fire can be predicted
Provide operation guidelines for HVAC in main
TPM 5
compartments and extinguishing areas
Provide number of casualties and location
TPM 6
information for passengers and crewmen

Record damage control infor- Check damage control procedures in real time,
mation (including information Strengthen TPM 7 and provide dynamic response guidelines that can
Ship touch- decision be managed
10:44 on casualties that occurred
es the sea 12 making
p.m. during accident and abandon-
bottom support Provide response guidelines that can be updated at
ing procedures), and share with TPM 8
any time
rescue support team

10:48 General Provide static and dynamic restoration predicting

TPM 9
p.m. emergency functions during flooding

First life- Provide ship behavior properties and

10:55 boat flooding-progress predicting information
TPM 10
p.m. launched at according to time flow, assuming flooding
sea situation by accident scenarios

10:58 Ship
p.m. grounding
15 It was found from the results of TPM selection that acci-
dent information for responses must be provided to roughly
440 persons predict the influence of an initial accident on user interfaces.
11:37
still to 20 It was found that a shortcut-key-based operational proce-
p.m.
evacuate
dure (that can omit complicated and diversified procedures
Ship master for menus such as difficult calculations or data input by
00:34 leaves the Analyze accident and damage
7075 keyboard and trackball) was required to secure an addition-
a.m. ship by control results. Improve
lifeboat onboard DCS and training
al communication network based on wireless communica-
process tion. This allows damage control teams or decision-makers
Helicopter who are distant from each other on the ship to effectively
ITCG in-
00:41
tervention 80
share situational recognition and effectively control acci-
a.m. dent situations that spread quickly (Calabrese et al. 2012).
to recover
50 persons To strengthen decision- making support, it was found that
the function of presenting measurements for each step
50 persons
03:44
still to (from the start of the accident to the endpoint of damage
a.m. control) must be strengthened by considering the develop-
evacuate
ment of the damage and aspects of damage control. These
30 persons
04:22
still to
issues include changes in ship behavior considering the
a.m. state of the sea based on the progress of flooding, sugges-
evacuate
tions for operability evaluations and plans for main equip-
06:14 Evacuation ment, area lockout considering smoke and the spread of
a.m. completed
heat onboard, and configurations for effective kill cards to
use in a fixed firefighting system.
Table 3 shows the results of an analysis on the require-
ments for effective use of a DCS. Damage control systems OPERATION CONCEPT OF IMPROVED DCS
in operation must adhere to laws, regulations, and guide-
lines. To select technical performance measures (TPMs) to The concept of using an improved DCS, which assumes
evaluate the utility of a DCS, 10 items in two fields were that TPMs in existing DCSs are applied, is shown in Fig. 2.
selected for improvement. Table 3 also shows functions that
are required for the efficient use of a DCS on a ship. These

63
 Fig. 2 Example of operational concept of DCS including improvements
(Some images of the figure are derived from google.com and edited for noncommercial use)
TPM 1, TPM 7, and TPM 8 are related to the use of kill
cards to respond to accident alarms. These alarms are trig-
gered by flooding and fire information gathered from sen-
sors, damage control personnel on the ship, location infor-
mation of weak traffic, and coordination of various accident
situations in the DCS operations console.
TPM 2 and TPM 6 provide locations of onboard person-
nel and open-and-shut information for watertight doors,
airtight doors, and hatches. These data are linked through a
signal interface that can be checked in the DCS.
Information on the operability or disabling information of
equipment that is essential for ship operation of generators
and distributing boards according to flooding and fire
spreading during accidents is used by linking it with ECS
information. The level of linkage may vary by the type of
ship within the range of laws, regulations, and guidelines
related to the operators requirements.
There is a difference in the number of compartments in
ships or the size of ships for TPM 3, 4, 5, 9, and 10. This
relates to the kill card configuration used to predict damage
development, predict the damage scale, or respond to each
accident situation according to the time flow in accidents.
This uses the simulation results from various damage sce-
narios (Kang et al. 2013). However, flooding analysis is
configured by simulating more than 5,0008,000 cases of
damage scenarios that consider the sea state and the inci-
dent angles of waves. For fire accidents, simulation results
from a smoke and heat-spreading analysis of main areas
and passages with risks of A-, B-, C-, and D-class fires are
used.
64
The DCS works with information such as the initial re-
sponse plan for initial crewmen who find the fire, the tim-
ing of the fire-extinguishing team entering the fire scene,
the temperature of the exiting point, smoke height standards,
area lockout timing, and points where a fixed fire extin-
guishing system is used.
According to preliminary training results, the initial step
to minimize accident damage during fire or flooding acci-
dents is taken by the initial accident finder or crewmen near
the accident scene (Peters et al. 2004). Reports are given
after measurements in minor accidents, but the DCS auto-
matically intervenes when decision making by the captain
is required because the size of the accident spreads over a
certain area.
The accident situation is determined through sensors in-
stalled onboard. The DCS operator reports to the com-
mander and refers to accident scene reports and kill cards
from the DCS. During emergencies, or when the ship is not
large, one person must perform the role of the DCS opera-
tor and decision-maker. When reporting site information
during emergency damage control, it is difficult to effec-
tively record data using diversified menus with existing
keyboards or trackballs. Therefore, it is necessary to con-
sider operating the DCS by using shortcut keys, along with
voice recording and recording events on the DCS operation
screen. Fig. 3 shows a shortcut key configuration using
NSTM 079 and DCAMS 5.02 casualty icon.
Table 4 Example of Use of Fire Damage Control with Coded Shortcut Keys

65

Fig. 3 Concept of NSTM 079 and DCAMS 5.02 casualty
icon-based shortcut key configuration (US navy 2001)
The shortcut keys were configured based on primary in-
formation that must be recorded by the DCS operator. The
shortcut keys are configured based on situations that require
decision making according to existing regulations and
guidelines for damage control procedures, and for events
that require confirmation. They are used to analyze damage
control results or evaluate onboard accident response train-
ing results. They are also used to improve accident response
scenarios during the ships life cycle. The table 4 shows the
concept of operating the coded shortcut keys.
The user interface of a DCS that uses coded shortcut
keys can be roughly configured as shown in Fig. 4. This
configuration of the user interface can minimize the devel-
opment costs of completely new DCSs, and can easily dis-
play the management mode and required database in the
existing DCS for linkage.
Fig. 4 Example of modified GUI for DCS
FUTHER CONSIDERATIONS
The management of a DCS using shortcut keys is a con-
cept used with existing system management methods, and
can be selectively used by considering the skill of the sys-
tem operator.
Accident-scenario-based simulation databases are limited
because they cannot completely reproduce actual accident
situations. Considering the difficulty of calculating damage
restoration or securing reliable calculation results during
flooding accidents on deck, the database can be used to
roughly check changes in ship behavior or safety during
restoration over time after flooding accidents. For fires, it is
impossible to make an accurate quantitative prediction of
the amount of smoke or spreading time of a fire (Henley et
al. 2008). The database can be used for damage control
based on kill cards by checking the entrances for
66
fire-extinguishing crews, exit routes, and the tendency of a
fire to spread.
CONCLUSIONS
In spite of the DCS, most onboard recovery activities are
still carried out by the crews. And DCS is not preferable
system for crews since its difficult and not familiar features.
For this reason, in this proceeding, specifications of DCSs
on large passenger ships and naval vessels are studied. And
requirements of stakeholders are investigated. From this
following conclusions are obtained.
1. Existing DCSs can be improved with coded shortcut
keys and well predefined kill cards.
2. The shortcut keys are configured based on situations
that require decision making according to existing
regulations and guidelines for damage control proce-
dures, and for events that require confirmation.
3. The shortcut keys can be used to analyze damage con-
trol results or evaluate onboard accident response
training results. They are also used to improve Kill
Cards during a ships life cycle.
4. Further study required to prove effectiveness of sug-
gested concept of coded shortcut keys in near future.
ACKNOWLEDGEMENTS
This research was supported by the research projects of
KRISO, PNS2500 and PES2350.
REFERENCES
Bgh, S.A., and Severinsen, T. (2009), Damage manage-
ment control system on the Danish navy ships, In Proc.
14th Int. Ship Control Systems Symp. (SCSS). Ottawa,
Canada.
Calabrese, F., Corallo, A., Margherita, A., and Zizzari, A. A.
(2012), A knowledge-based decision support system for
shipboard damage control, Expert systems with Applica-
tions, Vol. 39, No.9, pp.8204-8211.
Henley, G., Haupt, T., Parihar, B., and Williams, F. W.
(2008), Fire GUI: A Software Tool for Predicting Fire
and Smoke Spread on Ships, In ASNE Symp.: Ship-
building in Support of the Global War on Terrorism
IMIT (2012), Marine Accident Investigation C/S Costa
Concordia, Italian Maritime Investigative Body on Ma-
rine Accidents
IMO, MSC/Circ.919 (1999), Guidelines for Damage Con-
trol Plans
IMO, MSC.1-Circ.1245 (2007), Guidelines for Damage
Control Plans and Information to The Master (Secretari-
at)
Kang, H. J., Yang, Y. S., Choi, J., Lee, J. K., Lee, D. (2013),
Time basis ship safety assessment model for a novel
ship design, Ocean Engineering, Vol.59, No.C,
 pp.179-89

Korea Register of Shipping (2009), "Fire Fighting/Abandon

Ship Drill Checklist

Korea Register of Shipping (2004), "Technical Information

- Onboard Fixed Fire Extinguishing System

Peters, S., Bratt, E.O., Clark, B., Pon-Barry, H., and Schultz,
K. (2004), Intelligent systems for training damage con-
trol assistants, Proc. of ITSE.

US Navy (2001), Damage Controlman-NAVEDTRA

14057, Naval Education and Training Professional De-
velopment and Technology Center

67
 Raising the Bar for Tug Stability
Govinder Singh Chopra
Director, SeaTech Solutions International (S) Pte Ltd, Singapore

ABSTRACT another. This brings in the human factor to the safety equa-
tion. Safety can be considered a function of these 3 inter-
One of the crucial components of safety in the ship de- dependent elements as shown in Fig. 1, and within the lim-
sign spiral is ship stability. The ship must remain stable its specified by the Regulations.
when loaded under different conditions under various
operations performed by its operator in the various sea
conditions. Traditionally stability under different load-
ing conditions has been evaluated and checked for com-
pliance with the IMO stability criteria. However, espe-
cially for workboats which have completely different
operating profiles as compared to commercial ships,
such stability criteria alone may not be adequate. For
the ship to remain stable, in addition to the loading con-
dition, the other variables such as vessel operations, op-
erators competency, and sea condition must also be
considered. In this paper, we identify the limitations of
the present IMO stability criteria, both the first and
second generation, and with some case studies, suggest
more meaningful performance based criteria to be in-
cluded in the design for safety.

Fig. 1. Safety as a function of Human Factor, Operations

INTRODUCTION
Oceans can support the ship, and they can also cap-
size it This is a well-known ancient Chinese saying that
presents the stark reality that every vessel is vulnerable to
capsize. Under extreme operations and or environmental
conditions that exceed acceptable limits, the vessel will turn
over, heel and capsize.
For ship designers, one of the crucial components of
safety has been the vessels stability. Standards for safety
are set by Regulations, and for stability the standard to be
complied is the Intact Stability Code 2008, which applies to
all vessels greater than 24m length.
Unfortunately, the drivers for better regulations is ac-
cidents, and after some serious accidents involving headsea
parametric roll resonance the IMO decided in 2012 to de-
velop the second generation stability criteria which is pres-
ently under discussions.
Guaranteeing a sufficient level of safety to comply
with the Regulations has typically been considered to be a
matter of design. However, it is impossible to ensure safety
only by design as it also depends on how the vessel is oper-
ated and under what operating conditions.
Further, safety itself is a personal perception and
non-absolute. What may be safe to one could be unsafe to
and Design within the limits of Regulations
The vessel must be designed for the intended opera-
tions, operated as intended and have onboard a safety man-
agement system for the human operators taking into con-
sideration not only their human skills and experience but
also their attitudes and behaviors for a more proactive re-
sponse to safety.
The present stability criteria do not adequately con-
sider the intended operations of the vessel. This deficiency
becomes more exposed in the case of workboats which are
in fact intended to operate in diverse roles and in harsh
weather conditions.
PRESENT IMO STABILITY CRITERIA
The conventional fundamental principles for stability eval-
uation remain valid but we need to adapt and adjust these
criteria to the actual operations. The goal is to ensure ade-
quate reserve buoyancy, positive GM, and ensure there is
always sufficient righting energy along with watertight in-
tegrity, in all conditions of operations.

68
The four (4) pillars of stability against capsize are:
1. GM Statical / Initial stability regulations
2. Righting Energy Dynamic Stability Regulations
3. Freeboard Load line Regulations
4. Watertight Integrity Load Line Regulations
The vessel geometry and hull form decides the vessels
stability characteristics, such as KM, KN etc. Each hull
form being unique, the stability form characteristics may be
different, however for a given set of vessel dimensions,
there is little room for designers to drastically change these
stability form characteristics. The vessel arrangement,
depth and weight distribution fixes the vertical centre of
gravity. Together, this decides the vessels stability at the
design.
GM Statical / Initial Stability Regulations
The origin of the first-generation intact stability regu-
lations can be traced to the pioneering works of Rahola in
1939. It is based on purely empirical and statistical ap-
proach. Stability of the ships that capsized was compared
with those with a long accident-free history. The stability
parameters were calculated in calm and still waters as it
was not possible to compute and evaluate ship motions in a
seaway.
GM is the most basic and elementary stability criteri-
on to quantify the ships stability. Specification of a mini-
mum value of initial GM and freeboard sufficed as the sole
numerical stability criterion in use until 1940s. The value of
GM is easy to calculate, its meaning is relatively simple to
understand, and the effect of its magnitude on the heel re-
sponse is readily predictable. However, GM is only a valid
measure within small angles of heel, less than 5 degrees.
For workboats, the required GM in the IMO criteria is only
0.15m. This is extremely inadequate for powerful tugs of
the present and the required value must be increased sub-
stantially and also related to the bollard pull to be realistic.
IMO GENERAL CRITERIA
GM = KM - KG 0.15 m.
Fig. 2. Statical/Initial Regulations under the
IMO General Criteria
In the absence of a valid IMO criterion, designers
have been guided by the Argyriadis formula:
GM = (ft.)
SHP : Shaft Horsepower per shaft
: displacement, long tones
f : minimum freeboard along the length
B : molded beam
h : vertical distance from propeller shaft centreline
to towing bitt
Murphy further modified the basic formula as follows:
GM = (ft.)
N : number of screws
D : propeller diameter
S : fraction of propeller circle cylinder intercepted
by rudder turned to 45
The US Coast Guard used the Murphy Criterion until
1971, when the factor of 76 in the denominator was
changed to 38, resulting in a twofold increase in the re-
quired GM. Another criterion is the standard proposed for
Norway which was experimentally derived by towing a
model with a block coefficient of 0.5 sideways at a speed of
4 knots. The standard requires:
GM = (m)
h1 : height of towing bitt above waterline
d : draught
f : minimum freeboard along length
The major limitation of the GM criteria is that it is
valid only at angles of heel less than about 5 degrees. These
expressions are unable to predict the response of the work-
boat at heel angle near capsize. A more sound approach for
large angles of heel is based on GZ and balancing the heel-
ing or overturning moment induced with the available
righting moment below a critical angle, such as either the
angle of downflooding or the threshold of capsize.
Righting Energy Dynamic Stability Regulation
The concept of energy balance is that the righting en-
ergy must be more than the capsizing or overturning energy.
However, the present IMO criteria do not cater for work-
boat operations and operating conditions, which experience
a different set of forces and overturning moments.
In the absence of suitable criteria from IMO, Classifi-
cation Societies developed and laid down criteria for
workboats energy balance, considering certain operations
such as towing, firefighting and crane operations. Finally in
May 2016, at the 96th session of The Maritime Safety
Committee, IMO has approved the amendments in the IS
code with respect to Stability information for vessels en-
gaged in towing. It includes design criteria and operational
criteria for harbor towing, coastal or ocean towing and Es-
69
cort operations. This is a good starting point, however, 1) 0 < de
these criteria need to be further developed to properly rep-
Fig. 4. Classification society criteria/IMO weather criteria
resent both the operation mode and the weather conditions
under which these operations are carried out. IMO CRITERIA (TOWING)
DYNAMIC

1) G > H
2) 0 < 3
Fig. 5. IMO Towing Criteria
IMO CRITERIA (ESCORT)
IMO - GENERAL CRITERIA
1) 'A' - Area under GZ curve up to 30 0.055 m.-rad
2) 'B' - Area under GZ curve up to 2 0.09 m.-rad
3) 'C' - Area under GZ curve bet. 30 and 2 0.03 m.-rad
4) 'E' - Max. GZ to occur at angle 25
5) 'F' - Max. GZ 0.2 m
Fig. 3. IMO General Criteria for Dynamic Stability

HEEL & RIGHTING

1) Area I 1.25 * Area J

2) Area K 1.40 * Area L; and
3) 0 15 degrees.
Fig. 6. IMO Escort Criteria
The IMO criteria were developed using empirical values
derived from statistics of commercial vessels. Workboats
IMO - WEATHER CRITERIA have different proportions and hull forms and perform dif-
ferent operations. This renders the criteria not suitable and
1) 0 0.8 x de or 16 whichever is less in some cases inadequate. There is an urgent need to in-
2) S2 S1 clude all other operations that the workboat may be re-
quired to perform, and make the criteria more performance
ABS TOWING CRITERION based.
1) S2 > 0.09 m.-rad
ABS FIRE FIGHTING CRITERA Freeboard Load Line Regulation

1) S2 > 0.09 m.-rad The loadline concept was developed for commercial
2) f > dy shipping to prevent overloading vessels and jeopardizing
their safety. Samuel Plimsoll was instrumental in drafting
ABS ESCORT CRITERIA the Merchant Shipping Act and convincing the British Par-
liament to finally enact this in 1890.

70
 The loadline convention followed in 1930:
Freeboard and the minimum bow height has an effect on
various vessel characteristic
1. Shipping green water
2. Deck immersion
3. Reserve buoyancy
4. Righting energy
5. KG Values
The concept of freeboard is still valid. However, the
value in the present regulations is obtained arbitrarily and
empirically without scientific basis. For some workboat
operations, a lower freeboard at the stern may be necessary.
This conflicts with the stipulated freeboard and may result
in a design which is not suitable or unsafe for certain opera-
tions.
On the other hand, workboats with the minimum statu-
tory freeboard may not prove to be very suitable for opera-
tions in harsh sea conditions.
Watertight Integrity Load Line Regulation
One of the crucial means of ensuring adequate stability
involves providing external watertight integrity so that the
hull boundary remains effective in providing buoyant force
and righting energy.
Downflooding point is the point at which water could
enter the hull boundary which provides buoyancy when the
vessel heels due to an external overturning moment.
Fig. 7. Downflooding Points
The requirements for watertight integrity are contained
in the load line regulations, and this is directly related to the
downflooding point which is a very critical factor in the
IMO stability criteria.
There are basically two types of downflooding points
assumed in the calculation of stability:
1. Weathertight openings
2. Unprotected openings.
The most common unprotected opening is the engine
room ventilator, since provision of air to combustion ma-
chinery is necessary for operations. The possibility exists
that in certain conditions, however, some of the unprotected
openings may also need to be closed such as during the
preparation for severe storm or for the duration of the tow
and when the hull is unmanned and not in an operational
condition.
Weathertight openings are those openings which are
provided with weathertight closing appliances that can be
closed in bad weather. Providing weathertight closures on
openings into the buoyant envelope removes them from
consideration as downflooding points because they are as-
sumed to be effective in preventing the ingress of water
during intermittent immersion. Such openings need a min-
imum height (sill height) above deck so that when open,
water ingress is minimized. However, for small workboats
which have also a lower freeboard, this sill height may not
be adequate. There is a need for adopting the loadline regu-
lations to workboats and basing the requirements with
greater emphases on physics and operational requirements.
SECOND GENERATION STABILITY CRITERIA
The development of the second-generation intact stabil-
ity criteria has focused on five dynamical stability capsize
modes during operations:
- Excessive acceleration
- Deadship
- Parametric roll
- Broaching
- Pure loss of stability
Fig. 8. Structure of Second Generation Intact Stability Cri-
teria IMO (2008)
The second generation criteria are ship specific and
consider the stability during certain operations. Further it
provides guidance to the operator on limitations for opera-
tions. Vulnerability has now been defined it is the possi-
bility that the ship may capsize.
The aim of level 1 vulnerability criteria is to check the
susceptibility of a ship to a specific failure mode, whereas
the aim of level 2 vulnerability criteria is to indicate the
degree of ship susceptibility to a particular failure mode.
These criteria again are developed for commercial ship-
ping and do not consider the operations of workboats.
Limitations of present IMO Stability Criteria
From the previous section, it is clear that the present
IMO stability criteria is totally inadequate as they do not
consider the different operations of workboats which may
take place with simultaneous wind, waves and currents to
be superimposed. The reaction or forces from the thrusters
71
to counter the environmental forces/moments resulting in raised and additional decision assist guidance must be pro-
additional heeling moments need to be added in the vided to the operators to ensure safer tug operations.
weather criteria, along with crane, towing, or maneuver-
ing operations, which are undertaken by the workboat.

Fig. 11. High Speed Operations - Escort

OPERABILITY ENVELOPE

For safe operations, a limiting or operability envelope

could be provided for operators guidance for each mode of
Fig. 9. External Forces & Heeling Moment during tug op-
operation. Such an envelope will provide the following
erations
limits which the operators may monitor closely and have
built-in alarms to forewarn the operators as well as provide
In addition to the operations which are usually at low
enough guidance to the operator for decision making.
speeds, on some occasions, the workboat may also perform
operations at higher speeds.
Limiting KG
Report from a recent tragedy concluded that the major
cause of the accident was apparently due to suddenly turn-
The limiting KG is the maximum KG complying with
ing both the thrusters to maximum hard over position
prescribed and applicable set of criteria at a given draft and
while the tug was moving at its maximum full speed ahead
given sea conditions.
during sea trials for a steering test.
Another high speed operation is escorting in both direct
Limiting Heel
and indirect modes. There is adequate awareness of this
operation within the industry and the Classification Socie-
ties have laid good and adequate stability guidelines and This is another useful guidance for operators. For tow-
criteria for escort tugs. This criterion has now been in the ing operation, the heel must be less than the angle of which
IMO stability. water may flood the vessel through openings left without
weathertight closures. For turning operation, the heel needs
to be less than Deck Immersion or say 10 degree, whichev-
er is less. The 10 degree heel has been suggested based on
feedback from operators on a reasonable value.

Limiting Speed

For higher speed operations, it is essential to advise the

operator on the limit speed for safe operations.

Limiting Sea Conditions during Different Modes of

Fig. 10. High Speed Operations Turning Operations

Perhaps, this is the most critical guidance for the opera-

Further, during operations, the watertight integrity may tor - limiting sea conditions i.e. the wind, wave, and current
differ from the as-designed or as-regulated conditions. limitations where the different operations may be carried
For actual realistic operating conditions, a more worst out.
downflooding point for those weathertight openings which
remain open, may need to be considered, and higher sill
heights made mandatory in the regulations. The bar must be

72
CASE STUDIES
Limiting KG curves were then plotted (Figs. 12 and 14)
Investigations were carried out on existing designs of for the following cases:
tugs, in order to develop an operability envelope taking 1. Without wind
account of the actual operating conditions. The parameters 2. With steady wind (35 knots)
of the 3 tugs are shown in Table 1. 3. With steady wind (50 knots)
4. With wind and worst downflooding (DF) point
Table 1. Principal Dimensions of Tug Boats (See Figs. 14 & 15)
5. With wind, worst downflooding (DF) point and
Beam Depth Design Bollard aft trim 1% of LBP (See Figs. 14 & 15)
LOA LBP Limiting heeling angle curves were plotted as follow.
Tug (mld) (mld) Draft Pull
1. For Turning Operation, the heel angle limit was
(m) (m) (m) (m) (m) MT defined as Deck Immersion Angle or Absolute
1 25.80 21.910 9.50 5.00 4.00 40 10 deg whichever is less. (See Fig. 13)
2. For Towing Operation, the Equilibrium Heel An-
2 32.50 24.830 10.50 4.90 3.90 45 gle Limit for each case was plotted. (See Fig. 15)
3 33.00 27.640 11.60 5.60 4.40 65

Dominant Criteria

Limiting KG values were calculated under 4 different

draft conditions for all the criteria as defined in Figs 3-4.
1. General stability criteria
2. Weather criteria
3. Towing & Fire fighting operations
4. Towing Operations with steady wind of 35 knots
and 50 knots Fig. 12. Limiting KG - Dominant Criteria Turning
Criteria for turning operations as follows: Equilibrium
heel on turn <10 or deck immersion.
Investigations revealed a certain pattern in the criteria
which was most dominating at different draft loading con-
ditions (see Table 2).

Table 2. Dominant criteria of tugs under four different

loading conditions

Fig. 13. Heeling Angle Limit for Turning

Limiting KG Curves

As seen in Table 2, the following two criteria are domi-

nant.
1. Towing + Steady Wind
2. Turning + Steady Wind

For Turning Criteria, Tug 1 was selected as it has this as

the dominant criterion for all the 4 draft conditions.
For Towing Criteria, Tug 2 was selected as it has this as
the dominant criterion for all the 4 drafts conditions. Fig. 14. Limiting KG - Dominant Criteria Towing

73
 - Wind, wave and current forces superimposed on the
existing criteria for towing, firefighting operations etc.
- More fail safe means to ensure external watertight in-
tegrity
- Definition of downflooding points to be changed to
include access openings on main deck
- The effect of aft trim to be considered
- Effect of thruster forces leading to additional heeling
moments during turning with wind effect to be includ-
ed.
- Training procedures and standards need to be made
uniform for the industry to ensure the operator is com-
petent and his knowledge is constantly updated.
For operators guidance in decision making, the next
step is to develop easy to use stability advisory tools (soft-
ware) with built-in limits from the limiting envelope. The
operator will be provided with clear instructions on limiting
operating parameters, by the on-board stability alarm and
monitoring system.
Fig. 15. Heeling Angle Limit for Towing Further detailed research would be required and collab-
oration with other stakeholders to analyse a larger sample
Table 3: Limiting KG Values of existing designs with inputs from operators on their op-
erational requirements and finally provide a basis to devel-
op a modified stability criterion relevant for workboats.
Limit-
Reduction Finally, it would be more effective if ship designers
Description ing KG
(%) were to have more direct access to the relevant IMO Com-
(m)
mittees along with other stakeholders to provide the de-
signers perspectives and ensure that the safety and stability
Turning Criteria requirements regulated are realistic and suitable for intend-
Without Wind 4.123 - ed operations. We must not wait for another tragedy to raise
With Wind Speed (35 knots) 4.030 2.3% the safety standards!
With Wind Speed (50 knots) 3.925 4.8%
Towing Criteria
Without Wind (ABS Criteria) 4.612 -
2.9% REFERENCES
With Wind Speed (35 knots) 4.476
With Wind Speed (50 knots) 4.389 4.8%
IMO, 2009, Report of the Maritime Safety Committee on
With Wind Speed (50 knots) its eighty-sixth session.
4.147 10.1%
+ (Worst Downflooding)
Rohr, J., 2003, Stability Management for DP Platform
With Wind Speed (50 knots) Supply Vessels, Dynamic Positioning Committee, Dy-
+ (Worst Downflooding) 4.091 11.3% namic Positioning Conf., Houston, Texas.
+ Trim by aft (1% of LBP)
Marine Technology, April 1980, Vol. 17, No. 2, pp.
During turning, if the wind speed is superimposed, the cor- 163-173
responding impact on the Limiting KG is shown in table 3.
Turning with wind speed (50 knots) reduces the values of William A. Henrickson (n.d.), Assessing Intact Stability,
Limiting KG by 4.8%. University of Michigan. Available online:
http://www.sname.org/HigherLogic/System/DownloadDocume
ntFile.ashx?DocumentFileKey=5980d5a0-e50a-4448-8ca8-2bd
During towing, if the worst downflooding point and 1%
39d7aa1bc (Accessed on: 5 July 2016)
aft trim is superimposed, it has a significant impact on the
reduction in the Limiting KG of up to 11.3%.
Martin, L.L. (1980), Ship Manoeuvring and Control in
Wind. SNAME Trans. 88. [Available online:
As expected, the worst downflooding point at the deep-
http://www.sname.org/HigherLogic/System/DownloadDocume
est drafts and with 1% aft trim has the most significant im- ntFile.ashx?DocumentFileKey=c0c558bf-fff8-4890-ac5c-9242
pact on the reduction in the Limiting KG. 8c56f3c4] (Accessed on: 5 July 2016)

CONCLUSIONS

There appears a strong case for modifying the existing

IMO criteria to include the following:

74
 Ship Optimization for Efficiency and Maneuverability in
Adverse Sea Conditions
George Zaraphonitis1), Aphrodite Kanellopoulou1),
Apostolos Papanikolaou1),Vladimir Shigunov2)
1)
National Technical University of Athens, 2)DNV GL Maritime, Hamburg
ABSTRACT
Aiming to improve energy efficiency and to reduce
greenhouse gas emissions from shipping, IMO
introduced the Energy Efficiency Design Index (EEDI),
MEPC.212(63). In order to meet the gradually
increasing requirements of the new regulation, coming
into force in successive five-year phases, new ships
should be designed with reduced propulsion power. As a
result, questions have been raised regarding the ability
of these designs to operate safely in adverse weather
conditions. Within the EU-funded research project
SHOPERA, alternative methods of variable complexity
and accuracy have been developed for the assessment of
the manoeuvrability of ships under the action of heavy
wind and waves. The present paper focuses on the
development of an optimization procedure for the
design of RoPax ships, which applies the
above-mentioned assessment procedures to identify
designs with adequate powering to ensure safe operation
in adverse weather conditions while keeping the right
balance between economy, efficiency and safety of the
ship and the environment.
INTRODUCTION
Shipping is by far the most efficient and environmentally
friendly means of transportation of significant quantities of
cargo around the world. As a result, even nowadays, despite
all the technological developments of land and air
transportation, more than 80% of the worlds merchandise
trade is transported by sea. According to the United Nations
Review of Maritime Transport (UNCTAD 2015) the
international seaborne trade (millions of tons loaded all
cargoes) increased from 5,984 in year 2000 to 9,842 in year
2014, with an average annual growth rate of 3.62%. During
the same period, the ton-miles of transported cargo
increased from 30,823 (billions of ton-miles all cargoes)
to 52,572 with an average annual growth rate of 3.89%.
Despite the proven efficiency of maritime transportations,
due to the huge volume of transported cargo, the impact of
greenhouse emissions from shipping is inevitably quite
significant. Maritime transport emits around 1000 million
tonnes of CO2 annually and is responsible for about 2.5% of
global greenhouse gas emissions (3rd IMO GHG study,
2014). The increase of emissions from shipping up to year
2050 is estimated from 50% to 250%, depending on
worldwide economic development. Acknowledging the
paramount importance of averting the catastrophic
consequences of climate changes due to global warming,
the control of greenhouse emissions from shipping was
brought before the Marine Environment Protection
Committee and after lengthy debates IMO adopted in 2011
the
Energy Efficiency Design Index (EEDI) and
Ship Energy Efficiency Management Plan (SEEMP).
Based on statistics from the existing fleet, relevant formulae
for the calculation of the Energy Efficiency Design Index
have been developed for each ship type, depending on the
transport capacity, service speed and installed power, and
trend lines were drawn depicting the current status of the
world fleet. Then, requirements for new trend lines in
five-year intervals (phases) have been adopted to ensure the
substantial reduction of greenhouse gas emissions from
ships up to year 2025. Roughly speaking, a ships EEDI is
proportional to its fuel consumption (therefore to the
installed power) divided by its transport capacity and speed.
For the same transport capacity and speed, a reduction of
EEDI might be achieved by either reducing specific oil
consumption or the required installed power. Although
some improvements of specific oil consumption of modern
diesel engines might be achieved in the near future, it is
reasonable to expect that these would be rather small, since
it seems that the design of these engines is nowadays
reaching its optimum. Hopefully, technological
developments will enable the design of more efficient ships
in terms of energy consumption, either by developing more
efficient hullforms and propulsion systems, streamlined and
optimized appendages, installation of energy saving devices
and waste heat recovery systems, improvements in hull
coatings, or by the combination of the above solutions1. If
however the benefits from technological advances are not
enough to enable compliance with the increasingly
demanding EEDI requirements coming into force in
1 Improvements in operational procedures (e.g. trim optimization
or improved weather routing) can also help in the reduction of
ship emissions, but these are not accounted for in the calculation
of EEDI.
75
January 1 2020 (Phase 2) and 2025 (Phase 3), then an
obvious design solution would be to reduce the propulsion
power by an adequate reduction of the ships design speed.
In either case, it is expected that future ships will be
designed with significantly reduced propulsion power in
comparison with those before the introduction of the
Energy Efficiency Design Index. There are, however,
serious concerns regarding the safety of under-powered
ships, related to the sufficiency of the installed propulsion
power and steering devices to maintain their
manoeuvrability in adverse weather conditions. This gave
reason for additional considerations and studies at IMO
(MEPC 64/4/13 and MEPC 64/INF7). As a result,
guidelines for the evaluation of minimum propulsion power
were proposed (IMO MEPC 62/5/19 and MEPC 62/INF.21),
which later resulted in the 2012 Interim Guidelines (IMO
MEPC 64/4/13 and MEPC 64/INF.7), updated in 2013 in
Res. MEPC.232 (65). These guidelines for the minimum
powering evaluation should be considered as a provisional
remedy rather than a permanent solution, since it was
acknowledged that at the time of their adoption, there was a
significant lack of understanding of the impact of the
various underlying parameters affecting the dynamic
behaviour and hence the manoeuvrability of ships in
adverse weather conditions. To address the above
challenges by in-depth research, the EU funded project
SHOPERA (Energy Efficient Safe SHip OPERAtion)
(2013-2016) was launched in October 2013. The ultimate
goal of the SHOPERA project is to develop criteria and
corresponding environmental conditions, along with the
required assessment methods of varying complexity and
accuracy, which can be applied in order to assess
sufficiency of propulsion and steering systems of ships for
manoeuvrability in adverse conditions, including open sea,
coastal waters and restricted areas. The impact of the
proposed new guidelines for the minimum propulsion and
steering efficiency on the design and operational
characteristics of various ship types and sizes is
investigated, among other measures, by
a. Conducting a large number of case studies by design
teams comprising designers, shipyards, shipowners,
classification societies, research institutes and
universities.
b. Developing and applying optimization procedures,
targeting sufficient propulsion and steering
requirements for safe ship operation in adverse weather
conditions, while keeping the right balance between
ship economy, efficiency and safety of ship and the
environment.
The development of such an optimization procedure for the
design of RoPax ships is presented in the following. The
adopted procedures for the assessment of a ships
propulsion and manoeuvrability in adverse weather
conditions are outlined, and the obtained results are
presented and discussed.
MANOEUVRABILITY IN ADVERSE CONDITIONS
Manoeuvrability of ships is presently addressed by IMO
Standards for Ship Manoeuvrability, IMO (2002), which
norm turning, initial turning, yaw-checking, course-keeping
and emergence stopping abilities of ships. These Standards
76
have been often criticized for not addressing ship
manoeuvring characteristics at limited speed, in restricted
areas and in adverse weather conditions. The latter has been
addressed in the project SHOPERA, Papanikolaou et al.
(2015).
Based on the analysis of accident statistics, accident reports,
interviews of ship masters and existing proposals for
manoeuvrability criteria in adverse conditions, Shigunov
and Papanikolaou (2014) proposed to address three
scenarios of adverse conditions and defined the
corresponding functional requirements:
(1) In the open sea, it is sufficient for the ship to be able to
weather-vane, i.e. change the heading to a favourable
with respect to seaway and keep this heading; even
uncontrolled drifting with seaway can be acceptable for
a limited time.
(2) In coastal areas, the requirements to manoeuvrability are
more stringent: first, the ship should be able to perform,
in principle, any manoeuvre, in seaway from any
direction and, second, the ship should be able to leave
the dangerous area before the storm escalates.
(3) Manoeuvring at limited speed in restricted areas is
relevant in situations where forward speed is limited
because of navigational restrictions; the ship should be
able to manoeuvre at reduced speed in the presence of
strong wind and obstacles.
Based on these functional requirements, Shigunov and
Papanikolaou (2014) proposed equivalent, but simple to
evaluate in design and approval practical criteria:
In the open sea, Weather-Vaning criterion: the ship
should be able to keep heading in head to
bow-quartering seaway up to 60 off-bow.
In coastal areas: Steering Ability criterion: the ship
should be able to keep any prescribed course in waves
and wind from any direction, and Propulsion Ability
criterion: the ship should be able to keep a prescribed
advance speed in waves and wind from any direction.
In restricted areas at limited speed, course-keeping at a
specified low speed in strong wind was proposed in
shallow water, shallow water near a bank and shallow
water during overtaking by a quicker ship.
Whereas IMO Manoeuvrability Standards IMO (2002) are
evaluated in full-scale trials, this is impossible for the
proposed criteria in adverse weather conditions. An
alternative, model tests with self-propelled ship models in
simulated irregular waves and wind, for all required
combinations of wave direction and wave period, is
impracticable at the present state of technology, whereas
direct numerical simulations of transient manoeuvres in
irregular waves are not mature enough yet for routine
design and approval. The project SHOPERA proposed an
alternative procedure, referred to as Comprehensive
Assessment, which is based on free choice of separate
simple model tests, numerical simulations or empirical
formulae to define wave, wind and rudder forces and
manoeuvrability coefficients, which are combined in a
simple numerical model to model ship motions, Shigunov
and Papanikolaou (2014). This procedure is not expensive
if empirical formulae are used for all force contributions;
however, it still requires the solution of a nonlinear system
of 3 motion equations for many combinations of forward
speeds and seaway headings. Therefore, an even simpler
alternative assessment, the Simplified Assessment
procedure, was proposed by Shigunov et al. (2016), which
has the complexity of a spreadsheet calculation. This
procedure is especially suitable for multi-objective design
optimization and was used here. For convenience, it is
briefly described below.
The Simplified Assessment procedure is based on a reduced
number of assessment cases and reduced complexity of the
motion equations, but still takes into account all relevant
physics of propulsion and steering, and addresses the same
manoeuvrability criteria as the Comprehensive Assessment.
In this paper, two criteria were considered: Propulsion
Ability and Steering Ability.
The Simplified Propulsion Ability assessment is based on
the observation that bow seaways are most critical for
required power at a given speed (Fig. 1, middle plot),
therefore it is enough to consider only seaway directions
from 0 to about 60 off-bow in the assessment, and, further,
that the influence of drift on the required thrust and
required power can be neglected, thus the system of motion
equations reduces to one surge equation (1).
Fig. 1. Examples of Comprehensive Assessment results in
polar coordinates ship speed (radial coordinate) seaway
direction (circumferential coordinate, head waves and wind
come from the top): lines required power equal to
available power (A), advance speed 4.0 knots (B) and
rudder angle 25 (C)
X s + X w + X d + X R + T (1 tH ) = 0
(1)
where indices s, w, d and R denote calm-water resistance,
wind resistance added wave resistance and rudder
resistance respectively, T is the propeller thrust and tH
is the thrust deduction. It is important to keep in mind that
the added resistance X d in eq. (1) is taken as maximum
over wave directions between 0 and 60 off-bow. Whereas
forces X s , X w , X d , X R and thrust T in eq. (1) can
be found using any method (empirical, numerical or
experimental), it seems logical to allow using even simpler
approximations for these terms in the Simplified
Assessment. In particular (see Shigunov et al. (2016) for
details and validation of the assumptions),
Rudder resistance X R , which may be substantial in
bow-quartering waves, is calculated as a fraction of
propeller thrust X R = tR T , where tR is an
empirical constant.
As the required advance speed, 6.0 knots was used; at
such a low speed, the propeller characteristics may be
evaluated using the bollard pull assumption instead of
full open-water propeller characteristics.
Calm-water resistance X s at 6.0 knots speed can be
defined using the ITTC regression line.
Wind resistance X w = 0.5 X w a (vs + vw ) AF is
' 2
defined by the air density a , wind speed vw , frontal
windage area AF and head wind resistance coefficient
X w' , which can be assumed conservatively as 0.9.
For the maximum added resistance in short-crested
irregular waves X d over the wave directions 0 to 60
off bow, an empirical formula is proposed and validated
(only for container and general cargo ships, bulkers and
tankers) in Shigunov et al. (2016):
( )
X d = 83 Lpp C B1.5 1 + Fr hs2 , where Lpp is the
length between perpendiculars, CB is the block
1/ 2
coefficient, Fr = vs ( gLpp ) is the Froude number
and hs is the significant wave height.
The resulting assessment procedure is implemented as an
MS Excel worksheet and covers diesel and diesel-electric
engines working on fixed- and controlled-pitch propellers.
The evaluation of the Steering Ability criterion is more
complex, because for the steering ability, both the steering
system and propulsion (which influences steering ability)
should be taken into account in the assessment: e.g. ships
with powerful propulsion may have a smaller rudder,
whereas ships with weaker propulsion may compensate this
with larger or more effective steering devices. The
Simplified Steering Ability assessment is based on the
observation that the steering ability is challenged to the
largest degree in seaway directions close to beam (Fig. 1,
right), i.e. the point where both the ratio of the required to
available delivered power and the rudder angle are
simultaneously maximized (further referred to as critical
conditions for steering for brevity) is close to beam
seaway. Therefore, evaluation of the time-average wave
and wind forces can be reduced to beam seaways. Further,
analysis of the terms of the motion equations at the critical
conditions for steering shows that for conventional vessels,
certain terms can be omitted, and the resulting Simplified
Steering Ability assessment reduces to one equation:
X s + X w90 + X d90 + X R + T (1 tH ) = 0 (2)
and one check:
YR b (Yw90 + Yd90 ) (3)
90
In eq. (2), X s denotes calm-water resistance, X w and
90
X added resistance due to beam wind and beam waves,
d
respectively, X R the rudder resistance, T the propeller
thrust and tH the thrust deduction on the hull. The
solution of eq. (2) defines the maximum attainable speed,
the corresponding propulsion point and the propeller thrust,
and thus defines the maximum available lateral steering
77
force on the rudder YR ; this force should be not less than
the lateral steering force required by eq. (3),
b (Yw90 + Yd90 ) , where Yw90 and Yd90 are lateral forces
due to beam wind and waves, respectively, and
b = N s /( N s + 0.5Ys Lpp ) is a coefficient depending on the
ratio between calm-water yaw moment Ns and
calm-water lateral force Ys at the critical conditions for
steering. To provide a conservative recommendation for the
value of b , it was evaluated in critical conditions for
steering using the Comprehensive Assessment for 11 ships
(bulk carriers, container ships and tankers), leading to
b = 0.5 as a default conservative value. For the other terms
in equations (2) and (3), in addition to any methods from
the Comprehensive Assessment (empirical, numerical and
experimental), simplified approximations are proposed by
Shigunov et al. (2016), consistent with the complexity of
the Simplified Assessment:
The increase in rudder resistance X R , significant in
critical conditions for steering, is taken into account
using a simple assumption X R = tR T , where
tR = 0.24 is an empirical constant.
To calculate the available lateral force on rudder YR ,
rudder model by Sding from Brix (1993) was used
with a specified maximum lift coefficient equal to 1.2.
The lateral force due to beam wind is calculated as
Yw90 = 0.5Yw'90 a AL vw2 ; where Yw'90 = 0.9 is used as a
conservative assumption for the lateral wind force
coefficient.
Approximation of the calm-water resistance in eq. (2) is
more difficult than in eq. (1): the ITTC regression line
cannot be used, because it would under-estimate
resistance at the (rather high) forward speeds relevant in
critical conditions for steering. If the resistance curve
is available e.g. from model tests, it can be directly
used; alternatively, resistance curve is calibrated to
match exactly the propulsion point, ship speed and
propeller rotation speed at full MCR.
For the added resistance in irregular short-crested beam
waves X d90 , a simple empirical formula
X d90 = 380 Lpp C B1.5 ( 0.1 + Fr ) hs2 is proposed
Shigunov et al. (2016).
Similarly, a simple empirical formula for the
time-average lateral wave force Yd90 in irregular
short-crested beam seaway is proposed (maximum over
relevant wave periods):
600 Lpp hs2
Yd90 =
1 + 45000 ( CB L0.5
pp )
4.6
This procedure was implemented in a MS Excel
practical use and extended to cover diesel
diesel-electric engines working on fixed-
controlled-pitch propellers, as well as propellers in
drives.
78
DEVELOPMENT OF PARAMETRIC MODEL
During the preliminary ship design stage, the designer, in
search of the optimal vessel for a specified operation
scenario, is faced with a series of important decisions, all of
them having crucial impact on the vessels performance. It
is the nature of the designers work that requires many of
these decisions to be made during the early stages, usually
based on limited, vague or in some extent unreliable
information. In such cases, the designer needs to rely on his
experience, intuition and engineering judgment,
occasionally supported by the exploitation of relevant data
available from past designs. Advanced design tools, making
use of modern CASD technology to facilitate the
elaboration of a vessels preliminary design in limited time
and with limited human effort, yet at the same time with
reasonable detail and accuracy, would be undoubtedly of
great assistance.
A systemic approach to ship design needs to consider the
ship as a complex system, integrating a variety of
subsystems and their components, all serving well-defined
ship functions (Papanikolaou, 2010). Employing the
parametric design, or parametric modelling procedure, the
design of a certain object, component or system may be
automatically elaborated, using specifically developed
software tools for each particular set of values of the design
variables defined by the designer. A parametric design
procedure may be relatively easily implemented in the case
of simple objects or components. In the case of integrated
systems, however, such as an industrial plant or a large
commercial ship, the implementation of a parametric model
is no more a simple task, if possible at all, as the level of
complexity increases exponentially. In such cases, the
design of the parametric model requires particular attention,
in order to ensure its integrity, accuracy, robustness and
functionality. The parametric model should be flexible and
generic, so that it can be applicable to as many design
alternatives as possible, detailed enough to depict all the
essential characteristics of the design, and at the same time
as simple as possible, to avoid any unnecessary
complexities and implications during the development of
the corresponding software tools. Such tools, if available,
would enable the application, testing and verification of
crucial assumptions and decisions regarding the ships main
by technical characteristics on a large number of design
alternatives, in order to identify the most suitable design
according to a set of selected criteria, to serve as the basis
for the subsequent detailed design stages. Going one step
further, once developed, these tools could be used as the
core of a formal optimization procedure, facilitating the
rational exploration of the design space and the
identification of a series of optimal, or near-optimal
design solutions, according to a predetermined set of design
criteria (objective functions or merit functions), while at the
(4)
same time fulfilling a set of design constraints. Inherent to
ship design optimization are the conflicting requirements,
resulting from the design constraints and optimization
for
criteria, reflecting the interests of the various stakeholders:
and
ship owners and operators, ship builders, classification
and
societies, administrations, regulators, insurers, cargo
pod
owners/forwarders, port operators etc. A ship needs to be
optimized for cost effectiveness, operational efficiency,
improved safety and comfort of passengers and crew, and,
last but not least, for minimum environmental impact
(minimization of risk of accidental oil outflow, engine
emissions etc.). Many of these requirements are clearly
conflicting and a decision regarding the optimal ship design
needs to be rationally made. In the course of the SHOPERA
project, a series of parametric models have been developed
for various types and sizes of ships, including RoPax and
cruise ships, tankers, bulk carriers, containerships and
general cargo carriers. Some of them were developed using
the NAPA software by NAPA Oy, others using the
CAESES software by Friendship Framework. In the present
paper, the parametric model for the global optimization of
RoPax ships will be outlined, and the results obtained from
its application to the design and optimization of a small
RoPax will be presented.
The parametric model for the RoPax ships was developed
in NAPA. The main advantage of this software tool, apart
from the availability of a large suite of modules and tasks,
suitable for the detailed 3d modelling and analysis of ship
designs, is the availability of NAPA Basic, an embedded
programming language that facilitates the development of a
parametric model with the level of detail needed for the
work at hand. Using the functionality provided by the
NAPA Basic programming language, it is possible to fully
automate the development and evaluation of a 3d model,
without the need for any user intervention.
The development of each design variant is based on a large
set of design parameters, the most important of which are
listed in the following:
Main particulars (Length BP, Beam, Draught, Depth)
Length of entrance, mid-body and run
Variables controlling local hullform details (e.g. shape
and size of the bulbous bow, the transom and duck tail)
Length and position of engine room(s) (optional)
Number of bulkheads aft and forward of the engine
room(s)
Number of vehicle decks and deck heights below and
above the bulkhead deck
Number of passengers (berthed and total)
Service speed, range
The parametric model is quite generic and can be used for
the design of RoPax ships of small, medium or large size.
Once NAPA is called by the optimization software, the
parametric design methodology is automatically executed
and the following basic tasks are elaborated:
1. Hullform development
2. Resistance and propulsion estimations
3. Development of internal layout
4. Weights estimation - Definition of Loading Conditions
5. Evaluation of transport capacity (lanes length, number
of cars/trucks, payload)
6. Evaluation of Stability Criteria and other Regulatory
Requirements
7. Assessment of Building and Operational Cost, Annual
Income and Selected Economic Indices
8. Evaluation of Energy Efficiency Design Index (EEDI)
9. Evaluation of hydrodynamic/manoeuvring performance
in adverse seaway/weather conditions
The first step of the parametric design procedure is the
development of the hullform. A set of NAPA macros have
been developed, facilitating the fully automatic hullform
definition. Considerable attention has been given to the
development of the corresponding NAPA macros, to ensure
that the resulting hullforms are of adequately high quality,
to serve as the basis of the subsequent tasks, particularly the
hydrodynamic and the intact and damage stability
calculations. The hull is divided in three parts: entrance, run
and parallel mid-body, with the length of the mid-body
being very small, as usual for fast displacement-type RoPax
vessels. The development of each part of the hull is based
on the corresponding set of design parameters. Apart from
the main particulars, specific parameters are introduced to
control local hullform details, such as the size and shape of
the bulbous bow (either conventional or goose-neck type),
the shape of the flat of side and of the flat of bottom, the
immersion of the transom, or the existence of a propeller
tunnel, a duck tail or a stern wedge. Based on these
parameters, three grids of definition curves are created
defining the vessels entrance, mid-body and run. The
resulting hullforms are typical of modern twin-screw RoPax
vessels with fine fore-bodies and buttock-flow sterns. The
definition grid of representative hullform of a small RoPax
ship, developed by the parametric model, is shown in Fig.
2.
Fig.2: Typical hullform (fore and aft parts), developed by
the parametric model
The development of the vessels internal layout starts with
the definition of the watertight subdivision below the main
car deck. A horizontal bulkhead deck and a piecewise
horizontal double bottom deck are created according to
user-defined parameters. The vessel is subsequently divided
in zones, aft and forward of the main engine room. The size
and position of the main engine room are determined first.
The length of the engine room is derived from the size of
the main engines, based on empirical formulae derived
from the statistical analysis of data from existing vessels. In
order to minimize the shaft length, the aft bulkhead of the
engine room is positioned as far aft as possible, based on
the shape of the hullform. Alternatively, the size and
position of the engine room may be defined by the user.
The main transverse bulkheads aft and forward of the
engine room are subsequently positioned, and the
corresponding watertight compartments are created. The
number of car decks and the type of vehicles carried on
each of them (mix of private cars and trucks) are specified
by a series of user-defined parameters, in accordance with
the size of the vessel. Smaller vessels have usually only one
79
deck for the carriage of trucks and private cars. For the
larger vessels, an additional upper deck for trucks and/or
private cars and/or one or two lower holds for private cars,
located forward of the engine room, may be created,
according to the users specifications. Alternative layouts
with central or side casings may be modelled, as specified
by the user. A number of upper decks are then generated,
providing the necessary area for the accommodation of
passengers and crew, according to the passengers transport
capacity specified by the user and the required crew number.
The model of the internal arrangement of accommodation
spaces is rather coarse; however, it is ensured that ample
room for public spaces and cabins for passengers and crew
and for the required service spaces is provided. The upper
decks layout considers also the accommodation of the
vessels lifesaving equipment required for the particular
service (i.e. life boats and/or life rafts). The internal layout
of a small RoPax is presented in Fig. 3.
The prediction of the vessels resistance and calm water
propulsion power are performed by applying the Holtrop
1984 procedure. These predictions may be corrected using
appropriate calibration factors provided by the user. Then,
suitable diesel engine models are selected from a data base.
Fig.3: Internal arrangement of a small RoPax, developed by
the parametric model
The ships light weight is decomposed in the following
main weight categories: structural, propulsion, auxiliary,
deck machinery and outfitting, electrical, piping, heating
80
and air conditioning, accommodation and miscellaneous.
The steel weight is estimated based on enclosed volumes of
the main hull and superstructure using appropriate weight
coefficients. For the main engines, the actual weight
provided by the engine manufacturer is used. The
calculation of the remaining weight groups is based on
empirical formulae and appropriate weight coefficients. The
vessels maximum payload is determined by subtracting the
light ship weight and the various DWT items (consumables,
provisions, stores etc.) from the total displacement. A series
of typical loading conditions are then created, for which the
intact stability assessment is performed based on the
requirements of IMO Resolution A.749. Suitable NAPA
macros for the assessment of damage stability are also
implemented, but they were not applied during the
optimization studies presented below, in order to keep the
calculation time in reasonable limits. The assessment of
manoeuvring performance of the ship in adverse weather
conditions is performed by series of NAPA macros,
applying the assessment procedure presented in the
previous section, developed by the SHOPERA project.
The building cost is divided in labour cost and cost of
material. The latter is decomposed in the cost of steel,
outfitting, main and auxiliary engines, technical and other.
The cost of each of these items is calculated using
appropriate unit cost coefficients. For the calculation of
labour cost, an estimation of the required man-hours is
needed. The latter may be calculated by a similar
decomposition, assuming appropriate man-hour rates per
ton of steel, area of accommodation spaces, kW of installed
power etc. An alternative approach for the estimation of
man-hours is based on the compensated gross tons (cgt)
concept (OECD 2007), jointly developed by the
Community of European Shipyards Associations (CESA),
the Shipbuilders Association of Japan (SAJ) and the
Korean Shipbuilders Association (KSA). The compensated
gross tons is a function of the actual gross tons and may be
calculated by the following formula:
cgt = A gt B (5)
where A represents mainly the influence of the ship type, B
is the influence of the ship size and gt is the gross tonnage
of the vessel. Appropriate values for the above coefficients
are available for various ship types. For Ferries, A is set
equal to 20 and B equal to 0.71. Using an average of 30
effective hours per cgt, an estimation of the required
man-hours for the construction of a RoPax ship may be
readily obtained. Both methods may be used for the
estimation of a ships building cost, according to the users
selection. It is also possible to use the reference (known)
cost of an existing ship (used as the starting point of the
optimization) and to calculate the cost difference of each
design alternative applying either one of the above
procedures.
The financial evaluation of the shipbuilding investment is
based on the calculation of the Net Present Value. To this
end, suitable macros have been developed, based on two
scenarios defined by the user: the investment scenario and
the operating scenario. The investment scenario is defined
by the building instalments, the years to delivery of the ship,
the loan factor (loan as percent of ship price), the loan years,
the interest rate and the discount rate. The operation
scenario is used for the calculation of the annual income
and operating cost and then the assessment of cash flow, tax
allowances, taxable profits and taxes in order to calculate
the annual discounted cash flow throughout the lifetime of
the ship and the selling price after the end of service. The
operating scenario is defined on an annual basis and is
assumed to remain unchanged throughout the economic life
cycle of the ship. In the course of one year, three periods of
economic exploitation (i.e. low, intermediate and high
season) are taken into account. For each of these periods,
the duration in months, the details of the route, including
distance between the ports and operating speed, passenger
and vehicle occupancy rates and the corresponding fares
and port charges are defined by the user. The annual income
is calculated consisting of earnings from passenger and
vehicle fares and from services to passengers on-board the
ship, and any state subsidies per trip, if applicable. The
operating costs consist of the annual port, fuel, crew,
upkeeping (maintenance) and other costs. An inflation rate
and a fuel price escalation pattern, both defined by the user
are taken into consideration.
OPTIMIZATION STUDIES
As already mentioned, the developed parametric model is
quite generic, and has been applied for the design and
optimization of RoPax ships of various sizes. In the
following, results obtained from the optimization of a
relatively small vessel will be presented. The ship is
designed for operation between Piraeus and the island of
Crete, with a roundtrip length of 320 sm, a transport
capacity of 1200 passengers, 200 private cars and 21 trucks
and a service speed of 23.8 kn. The ship will be operated
year-round, considering a high season during summer of
1.5 months with 7 roundtrips per week, a medium season of
5.5 months with 5 roundtrips per week, and a low season of
5 months with 3 roundtrips per week. The corresponding
occupancy rates for the passengers, cars and trucks that
have been assumed for the calculation of annual revenues
are 70%, 65% and 52%, respectively, for the high season,
50%, 30% and 75% for the medium season and 30%, 20%
and 75% for the low season. The ships earnings per trip are
calculated assuming an income of 39 per passenger (ticket
price plus earnings from services to passengers on-board),
50 per car and 557 per truck. The assumed oil prices are
220 for FO, 420 for DO and 950 for LO. A lifetime of
25 years has been used. Over this lifetime, an average
inflation rate of 2% has been assumed, while for the oil
prices in particular, an annual escalation rate of 4% has
been applied. For the above operational scenario, several
studies have been performed with the optimization tool
available in NAPA (i.e. the Optimization Manager), using
Genetic Algorithms. The results presented in the following
were obtained by varying the main dimensions as follows:
Min. value Max. value
Length BP, m 105.0 115.0
Beam, m 18.0 20.0
Design Draught, m 5.0 5.4
The objective of the study was to maximise the Net Present
Value of the owners investment, while ensuring
compliance with safety regulations, EEDI Phase II
requirements and the SHOPERA criteria for
manoeuvrability in adverse weather conditions presented
above, as well as various operational constraints such as
draught and trim constraints for the various loading
conditions, upper limit on building cost, lower limits on
DWT and lanes length requirements, constraints on the
average truck weight etc. A scatter diagram of the Net
Present Value as a function of Length BP is presented in
Fig. 4 (only feasible designs are included). The starting
point of this optimization (the original design) in this and
the following diagrams is marked by a cross. As shown in
Fig. 4, the maximum NPV is obtained with the length at the
upper limit of the design space. The impact of the other
main dimensions on NPV was relatively less significant.
This fact is mainly attributed to the impact of the length on
the required propulsion power (see Fig. 5). Scatter diagrams
of NPV vs. the building cost, DWT, Propulsion Power and
Lanes Length are shown in Figures 6, 7, 8 and 9,
respectively. A negative relationship between NPV and
DWT is shown in Fig. 7. This is because the freight rate of
trucks was assumed to be constant, irrespectively of their
weight. This way, ships capable of carrying the same
number of trucks, but of increased mean weight, are
penalised because of their increased propulsion power
requirements, whereas at the same time they are not
receiving any credit for their increased DWT capacity. A
scatter diagram of building cost vs. DWT is presented in
Fig. 10.
The obtained results indicate that compliance with the
EEDI Phase II requirement was quite demanding for this
study, as most of the unfeasible designs failed to satisfy this
criterion. At the same time, only two designs were
identified, marginally fulfilling the EEDI Phase III
requirement. Scatter diagrams of the calculated margins
with respect to the EEDI Phase II and Phase III are plotted
in Fig. 11. The EEDI margin is herein defined as the
required EEDI, according to the regulations, minus the
value attained. Therefore, a positive margin indicates a
design complying with the corresponding phase, while
designs with negative margin are failing to comply with the
requirement. A scatter diagram of the calculated EEDI
margin with respect to the Phase III requirement versus
NPV is plotted in Fig.12. The results shown in this plot
indicate that increased NPV is associated with improved
performance with respect to EEDI. The Net Present Value
of the two designs capable of complying with Phase III is
equal to m14.35 and m14.37. It is anticipated that
systematic hullform and propeller optimizations,
installation of energy saving devices and waste heat
recovery systems, or other technological advances might
help the designers to improve the performance of future
ships, without the need of significant) compromises with
respect to service speed. However, with the current
parametric model, it was found that, in order to obtain a
significant number of designs in compliance with EEDI
Phase III requirement, the design speed should be reduced
down to 21.6 kn, a reduction that would have a significant
impact on the specified operational scenario. For the design
speed of 23.8 kn four Wrtsil 9V32 engines were selected
for all ships presented herein. These engines were replaced
by four Wrtsil 9L26 engines for the reduced speed of
21.6 kn. As a result, compliance with EEDI Phase III
requirement was achieved (see Fig. 13).
81
 14.6 14.60

14.4 14.40

14.2 14.20
NPV [mEuro]

NPV [mEuro]
14.0 14.00

13.8 13.80

13.6 13.60

13.4 13.40

13.2 13.20
111.5 112 112.5 113 113.5 114 114.5 115 15500 15600 15700 15800 15900 16000 16100 16200 16300
LBP [m] Propulsion Power [kW]

Fig.4: NPV vs. Length BP feasible designs Fig.8: NPV vs. Propulsion Power

16,300 14.6

16,200 14.4

16,100
14.2
Prop. Power [kW]

16,000

NPV [mEuro]
14.0
15,900
13.8
15,800
13.6
15,700

15,600 13.4

15,500 13.2
111.5 112 112.5 113 113.5 114 114.5 115 590 595 600 605 610 615 620 625 630
LBP [m] Lanes Length [m]

Fig.5: Propulsion Power vs. Length BP feasible designs Fig.9: NPV vs. Lanes Length

14.6 65.75

14.4

65.50
14.2
Building Cost [mEuro]
NPV [mEuro]

14.0
65.25
13.8

13.6
65.00

13.4

13.2 64.75
64.9 65 65.1 65.2 65.3 65.4 65.5 65.6 65.7 1800 1900 2000 2100 2200 2300
Building Cost [mEuro] DWT [t]

Fig.6: NPV vs. Building cost Fig.10: Building cost vs. DWT

14.6 5

14.4 4

3
14.2

2
NPV [mEuro]

EEDI Phase 2 Margin

EEDI Margin

14.0
EEDI Phase 3 Margin
1
13.8 Initial Design - Phase II
0 Initial Design - Phase III

13.6
-1

13.4
-2

13.2 -3
1800 1900 2000 2100 2200 2300 1800 1900 2000 2100 2200 2300
DWT [t] DWT [t]

Fig.7: NPV vs. DWT Fig.11: EEDI Phase II and III Margin vs. DWT

82
 0.2 0.97
0.0
0.96
-0.2
-0.4
0.95

Propulsion Criterion
-0.6
EEDI Ph.3 Margin

-0.8 0.94
-1.0
-1.2 0.93
-1.4
-1.6 0.92

-1.8
0.91
-2.0
-2.2
0.90
13.2 13.4 13.6 13.8 14 14.2 14.4 14.6
1800 1900 2000 2100 2200 2300
NPV [mEuros]
DWT [t]
Fig.12: EEDI Phase III Margin vs. NPV
Fig.14: Propulsion Criterion vs. DWT
7
EEDI Phase 2 Margin 6.9
EEDI Phase 3 Margin
6

6.8
5

Hs_steer_max [m]
EEDI Margin

4 6.7

3
6.6
2

1 6.5

0
1800 1900 2000 2100 2200 2300 6.4
DWT [t] 1800 1900 2000 2100 2200 2300
DWT [t]
Fig.13: EEDI Phase II and III Margin vs. DWT for vessels
Fig.15: Steering Criterion: max wave height vs. DWT
with smaller engine, selected for a service speed of 21.6kn

On the other hand, and in contrast to what was observed for CONCLUSIONS
other ship types, compliance with the SHOPERA criteria
for manoeuvrability in adverse weather conditions was A parametric model for the design of RoPax ships has been
easily achieved. The same conclusion was derived by all developed, to be used for the investigation of the impact of
optimization studies for RoPax ships that were carried out, EEDI and of the ability of ships designed to meet Phase II
regardless of their size. This of course was not an and III requirements to operate safely in adverse conditions.
unexpected result, since RoPax ships are highly powered in This parametric model has been used to perform a series of
comparison with other types of ships of equal displacement. formal optimization studies of RoPax ships of various sizes.
Even the designs with reduced service speed of 21.6 kn, Results from the optimization study of a small RoPax ship
complying with EEDI Phase III, had enough power to indicated that compliance with EEDI Phase II requirement
ensure manoeuvrability in adverse weather conditions. was quite demanding, as most of the unfeasible designs
Calculations performed with only one propeller in operation, failed to satisfy this criterion. In order to comply with EEDI
in order to verify redundancy of the propulsion system, and Phase III requirement, the design speed should be reduced
assuming a 30% reduction of the propeller thrust to account by more than 2 kn, a reduction that would have a significant
for ship motions, unsteady conditions and propeller racing, impact on the selected operational scenario, unless
indicated that all feasible designs were able to achieve 12 systematic hullform and propeller optimization, energy
kn at bow waves with a significant wave height of 4.0 m saving devices and waste heat recovery systems, alternative
(see Fig. 14). The propulsion criterion plotted in this figure fuels or other technological advances might help to achieve
corresponds to the ratio of the required to the available Phase III requirements without the need of significant
propulsion power; therefore, values less than 1.0 indicate compromises with respect to service speed.
compliance with the criterion. With both propellers in On the other hand, and in contrast to what was observed for
operation, even with a 50% reduction of thrust, all feasible other ship types (especially bulkers and tankers), systematic
designs could still achieve 12 kn at a significant wave calculations for RoPax ships regardless of size indicated
height of 5.5 m. The maximum wave height for which that compliance with the SHOPERA criteria for
compliance with the steering criterion was obtained is manoeuvrability in adverse weather conditions was easily
shown in Fig. 15. Once again, the presented results are achieved. This was not an unexpected result, since RoPax
obtained with one propeller in operation, and assuming a ships are highly powered in comparison with other types of
30% reduction of the propeller thrust. ships of equal displacement.

83
ACKNOWLEDGEMENTS UNCTAD (2015) Review of Maritime Transport 2015,
United Nations Conference on Trade and Development,
The authors acknowledge the contributions of all http://unctad.org/en/PublicationsLibrary/rmt2015_en.pdf
SHOPERA partners and of numerous researchers who
contributed to the work presented herein. The work NAPA Software, NAPA Oy, https://www.napa.fi/
presented in this paper is partly supported by the
Collaborative Project SHOPERA (Energy Efficient Safe CAESES Software, FRIENDSHIP SYSTEMS,
SHip OPERAtion), Grant Agreement number 605221, https://www.caeses.com/
co-funded by the Research DG of the European
Commission within the RTD activities of the FP7 Thematic
Priority Transport, FP7-SST-2013-RTD-1, Activity 7.2.4
Improving Safety and Security, SST.2013.4-1: Ships in
Operation. The European Community and the authors
shall not in any way be liable or responsible for the use of
any knowledge, information or data of the present paper, or
of the consequences thereof. The views expressed in this
paper are those of the authors and do not necessary reflect
the views and policies of the European Community.

REFERENCES

Brix, J. (1993), Manoeuvring Technical Manual,

Seehafen Verlag

Holtrop, J. (1984), A statistical re-analysis of resistance

and propulsion data, Int. Shipbuilding Progress, vol. 31,
No 363, p. 272-276

Papanikolaou, A. (2010), Holistic ship design

optimization, Journal of Computer-Aided Design, vol.
42 pp 1028-1044

Papanikolaou, A., Zaraphonitis, G., Bitner-Gregersen, E.,

Shigunov, V., El Moctar, O., Guedes Soares, C.,
Devalapalli, R., Sprenger, F. (2015), Energy Efficient
Safe Ship Operation (SHOPERA), Proc. 4th World
Maritime Technology Conference

Shigunov, V. and Papanikolaou, A. (2014), Criteria for

minimum powering and maneuverability in adverse
weather conditions, 14th Int. Ship Stability Workshop
(ISSW 2014), Kuala Lumpur, Malaysia

Shigunov, V., Papanikolaou, A. and Chroni, D. (2016),

Maneuverability in Adverse Conditions: Assessment
Framework and Examples, Proc. 15th Int. Ship Stability
Workshop, Stockholm, Sweden

IMO (2002) Standards for ship manoeuvrability, Res.

MSC.137(76)

OECD Directorate for Science, Technology and Industry

(STI), Council Working Party on Shipbuilding,
Compensated Gross ton (CGT) System,
http://www.oecd.org/officialdocuments/publicdisplaydoc
umentpdf/?cote=C/WP6(2006)7&docLanguage=En

Third IMO GHG Study (2014),

http://www.imo.org/en/OurWork/Environment/Pollution
Prevention/AirPollution/Pages/Greenhouse-Gas-Studies-
2014.aspx

84
 Statistical Analysis on Parametric Roll Groups Detected by
IR-HHT Method in Irregular Head Seas
Liwei Yu1,2, Ning Ma 1,2,3, Yoshiaki Hirakawa3,4
1. State Key Laboratory of Ocean Engineering, Shanghai Jiao Tong University, Shanghai, China
2. Collaborative Innovation Center for Advanced Ship and Deep-Sea Exploration (CISSE),
Shanghai Jiao Tong University, Shanghai, China
3 Institute of Advanced Sciences, Yokohama National University, Yokohama, Japan
4 Systems Design for Ocean-Space, Yokohama National University, Yokohama, Japan

ABSTRACT practical non-ergodicity of parametric roll is one of the

Sufficient test time or enough repeated tests are
needed for statistical post-processing of parametric roll
which has practical non-ergodicity. However, long test
time or many repeated tests are time-consuming espe-
cially for model experiments. A new technique is de-
veloped to reveal more efficiently the statistical
properties of parametric roll in irregular waves.
The time frequency analysis method, Hilbert
Huang Transform (HHT), is applied on the simu-
lated parametric roll records to obtain the
time-dependent instantaneous frequency. Based on
the time-dependent instantaneous frequency, the
roll motion groups where parametric roll occurs in
the records are picked up. Then the statistical
analysis is conducted on these groups which con-
tain most of the roll energy. Results show that the
new technique can obtain a faster convergence on the
running standard deviation of different roll realizations
comparing to the original one.
INTRODUCTION
Parametric roll is a self-resonant roll motion phenome-
non with strong nonlinearity. The large roll angle in para-
metric roll can cause huge cargo damages and losses. Its
mechanism in regular waves has been well addressed by
both numerical simulations (Umeda et al. 2004; Bulian
2005; Neves et al. 1999; Spanos & Papanikolaou 2007;
Hashimoto & Umeda 2010) and model experiments
(Spanos & Papanikolaou 2009; Hashimoto et al. 2006;
Hashimoto et al. 2007).
However, parametric roll in the random seas is a phe-
nomenon with unique statistical properties. It has been con-
firmed by researchers through numerical simulations and
model experiments that parametric rolling in irregular
waves has practical non-ergodicity, due to the nonlinearity
on roll motion, the multi-direction and groupness property
of irregular waves. Belenky et al (2003) pointed out that
governing factors in statistical post-processing of paramet-
ric roll based the numerical simulation results. It was also
suggested that 1220 repeated tests are needed for detailed
analysis of parametric roll in random seas. Bulian et al.
(2006) also confirmed by a comprehensive theoretical, nu-
merical and experimental study that the parametric roll
process is practically non-ergodic and long test time is
needed to achieve the convergence on temporal average
which is probably impracticable. Furthermore Bulian et
al.(2008) investigated the influence of the initial conditions,
the problem of the length of the initial stochastic transient,
the dispersion in the estimated statistical characteristic of
measured processes based on the experimental results.
Large uncertainty was observed from time histories for roll,
especially when the environmental conditions are close to
the parametric roll stability boundaries. Umeda et al. (2012)
conducted model experiments of a car carrier with many
realizations for longtime duration to establish guideline
specifications for model experiment and numerical simula-
tion of parametric rolling in irregular waves. It was sug-
gested that around 12-realizations and more than 360 en-
counter wave cycles should be used in model experiments.
For cases close to the threshold for parametric rolling, ac-
curate estimation is difficult. Kim et al. (2011) conducted a
systematic study on the sensitivity of simulation results to
some parameters of parametric roll in irregular waves. It
was found that the statistical properties of parametric roll in
irregular waves are highly sensitive on simulation time
window, the discretization of wave spectrum in irregular
wave and motion simulation.
Past researches conclude that large number of repeated
tests are needed for the statistical analysis of parametric roll
in random seas. Parametric roll in irregular waves comes
and goes and has a so-called grouping phenomenon
(Belenky et al. 2003; Bulian et al. 2006). In irregular waves,
parametric roll will happen under a certain wave group and
then disappear. It is anticipated that the statistical properties
are easier to converge for the roll motion groups where
parametric roll occurs and most of the roll energy is con-
ceived. Thus by analyzing solely on the roll motion groups

85
where parametric roll occurs, the number of the repeated In the statistical analysis, the running standard devia-
tests may be reduced and the statistical properties of para- tion(STD) of the ith record is calculated as:

( X (t ) X (t ) )
metric roll can be obtained more efficiently. n 2

( tn ) j =1
In this paper, the parametric roll records in irregular i j i n
2
= (1)
heading waves are obtained by numerical simulations on a n 1
3100TEU containership using a 6-DOF weakly nonlinear where n is the number of record points up to the time, tn.
unified model accounting for non-linear restoring and X i (tn ) is the mean of the ith record up to tn. The ensemble
Froude-Krylov forces, maneuvering motion, rudder and
averages of the running standard deviations of all the 20
propeller hydrodynamics (Yu et al. 2012). Based on the
records are calculated to check for the convergence.
parametric roll detection approach proposed by Yu et al.
(2016), the groups where parametric roll occurs in the rec-
ord are picked out. Then the statistical analysis is conducted
on these groups which contain most of the roll energy.
Through this new technique, the statistical properties of
parametric roll are expected to be revealed more efficiently
comparing to the conventional one.

STATISTICAL ANALYSIS BY CONVENTIONAL

TECHNIQUE

The statistical analysis of parametric roll simulation records

by the conventional technique is carried out firstly. The
simulation cases and statistical analysis methodology are
described.

Simulation Cases
Fig.1 Samples of simulation records with a duration of
The 6-DOF unified model coupling both sea-keeping and 1800s
maneuvering motion proposed by Yu et al. (2012) are used
to obtain the parametric roll simulation records of a
3100TEU containership in irregular heading waves. The
model was validated by model experiments to simulate
parametric roll in head sea. Main particulars of the contain-
ership are shown in Table 1.

Table 1. Main particulars of 3100TEU containership

3100 TEU Containership
Length between per-pendiculars, Lpp(m) 214.2
Breadth, B(m) 32.2
Depth, D(m) 18.8
Fore draft, df(m) 11.49
Aft draft, da(m) 12.31
Displacement(t) 55633
Metacentric height, GM(m) 0.85
Block coefficient, CB 0.668
Radius of gyration, roll, kxx(m) 7.09
Radius of gyration, pitch(yaw), kyy(kzz)(m) 53.6 Fig.2 Running STD (colored thin solid line) of the roll (up)
Roll natural period, Troll (s) 18.85 and pitch (down) motion of the 20 records
Service Speed, V (kn) 22.5
The irregular heading waves are obtained using the
JONSWAP spectrum with significant wave height Hs=5m
and Tp=15.3s. Totally 20 simulation records each with a
duration of 1800s are obtained from numerical simulations
under the same condition with different random seeds. The
time duration of 1800s i.e. 0.5 hour is normally enough for
the conventional seakeeping statistical analysis and longer
than the time duration that model experiments can achieve
without wave reflection.

Statistical Analysis Methodology

86
Fig.3 Probability plot for the normal distribution (solid line)
of the roll (left) and pitch (right) motion of the 20 records
The running STDs and their ensemble average of roll and
pitch motions for all the 20 records are plotted in Fig.2 with
colored thin solid line. As shown in the figure, the running
STDs of pitch motion converge to the ensemble average.
Thus the pitch motion in irregular wave is ergodic process
with practical accuracy. However, the running STDs of roll
motion do not converge to the ensemble average with sig-
nificant dispersion. Moreover, the probability plots for
normal distribution of roll motion in Fig.3 also show large
dispersion among different records. These results indicate
that the temporal average obtained from a limited time rec- Fig.6 Running STD of merged records with eight records of
ord cannot converge to the ensemble average. Thus it is the roll motions
concluded (as expected) that parametric roll in irregular
waves is a practically non-ergodic process. STATISTICAL ANALYSIS BY THE NEW TECHNIQUE
In order to find out the numbers of repeated records
needed for the convergence of roll motion STD, the running Given that large number of repeated tests are still diffi-
STD of the merged records are calculated and presented in cult sometimes especially for model experiments, the new
Fig.4,5 and 6. The merged record is defined as the combi- technique is developed to reduce the number of tests re-
nation of records randomly selected from the total 20 rec- quired. In this new technique, The parametric roll detection
ords. From the figures, it is found that the convergence of approach based on the incremental real-time Hilbert-Huang
running STD is improved as the number of merged records transform (IR-HHT) method proposed by Yu et al. (2016)
increasing. When 8 records or more are merged, the disper- are used to pick out the groups where parametric roll occurs.
sion becomes steady and can still be as large as 20% as The IR-HHT method can obtain the time-dependent instan-
shown in Fig.6. taneous frequency. In the instantaneous frequency as shown
in Fig.7, there is a frequency shift before and after the onset
of parametric roll. Based on this frequency shift, the
groups where parametric roll occurs in the record can be
detected. The detection results of record #1 and #6 are
demonstrated in Fig.7 and 8.

Fig.4 Running STD of merged records with four records of

the roll motions

Fig.7 Groups where parametric roll occurs for simulation

record #1 (up: roll motion, down: instantaneous frequency)

Fig.5 Running STD of merged records with six records of

the roll motions

Fig.8 Groups where parametric roll occurs for simulation

record #6 (up: roll motion, down: instantaneous frequency)

87
 During the statistical analysis, the groups where para-
metric roll occurs within one record are combined together
to the new record. Thus 20 new records are formed. The
time durations of the 20 new records are not the same. The
running STDs of roll motion for all the 20 new records are
shown in Fig.9. In records #3, #5 and #7 shown as thick
solid black line in Fig.9, because parametric roll happens
for very short time, the dispersion of these three records is
larger than that of the other records. Except for records #3,
#5 and #7 as shown in Fig.9, the dispersion on the running
STDs of roll motion of the new records is much smaller
than that of the original records in Fig.2. This is further
confirmed by the probability plot presented in Fig.10. Other
than records #3, #5 and #7 which are of very short time
duration, the probability plots of all the other records con-
verge very well with the normal distribution.
Moreover, the ensemble average of the original records
in Fig.2 is plotted in Fig.9. It shows that the running STDs
of the new records is larger than that of the original records.
Because in the new records, the parts where roll angle is
small are eliminated.
Furthermore, the running STD of the merged records are
calculated and presented in Fig.11, 12 and 13. From the Fig.10 Probability plot (solid line) for the normal distribu-
figures, it is found that the convergence of running STD is tion of the new records
improved as the number of merged records increasing.
However, the dispersion can be about 12% when 8 records
are combined as shown in Fig.13. Comparing to the results
in Fig.4,5 and 6 of the original records, the records obtained
based on the new approach converge faster as the number
of merged records increasing, and the dispersion on roll
running STD for the new records are smaller.
Therefore, it is concluded that the new statistical analysis
approach which deals only with the groups where paramet-
ric roll occurs can obtain a faster convergence and smaller
dispersion among the roll running STDs of all the records.
Hence, the number of repeated tests needed for the statisti-
cal analysis of parametric roll can be reduced.
Fig.11 Running STD (colored thin solid line) of merged
records with four records of the roll motions

Fig.9 Running STD (colored thin solid line) of the roll mo-
tion of the 20 records

Fig.12 Running STD (colored thin solid line) of merged

records with six records of the roll motions

88

Fig.13 Running STD (colored thin solid line) of merged
records with eight records of the roll motions
CONCLUSIONS
In order to reveal the statistical properties of parametric
roll more efficiently, a new approach for the statistical
analysis of parametric roll in irregular waves is proposed.
The new approach deals only with the groups where para-
metric roll occurs. These groups are picked out using the
parametric roll detection scheme based on IR-HHT.
Through the comparison on the roll running STDs and
probability plots between the conventional technique and
the newly proposed one, it is concluded that the new tech-
nique can obtain a faster convergence and smaller disper-
sion among the roll running STDs of all the records, despite
some records with small duration of time for parametric roll.
The number of repeated tests needed for the statistical
analysis of parametric roll can be reduced by this new tech-
nique.
However, the new technique shows its advantage in this
specific numerical simulation condition of the 3100TEU
containership. Its effectiveness on other numerical simula-
tion conditions as well as model experiment results of par-
ametric roll is still needed to be confirmed in the future
work.
ACKNOWLEDGEMENTS
The present study is supported by the National Natural
Science Foundation of China (NSFC) research project: No.
51579144 and the Institute of Advanced Sciences in Yoko-
hama National University
REFERENCES
Belenky, V., Weems, K.M. and Pauling, J.R., 2003.
"Probabilistic analysis of roll parametric resonance in
head seas". In 8th Int. Conf. on the Stability of Ships
and Ocean Vehicles. pp. 325349.
Bulian, G., 2005. "Nonlinear parametric rolling in regular
wavesa general procedure for the analytical
approximation of the GZ curve and its use in time
domain simulations". Ocean Engineering, 32(34),
pp.309330.
Bulian, G., Francescutto, A. and Lugni, C., 2006.
"Theoretical, numerical and experimental study on the
problem of ergodicity and 'practical ergodicity' with an
application to parametric roll in longitudinal long
crested irregular sea". Ocean Engineering, 33(8),
pp.10071043.
Bulian, G. et al., 2008. "Qualitative and quantitative
characteristics of parametric ship rolling in random
waves in the light of physical model experiments".
Ocean Engineering, 35(17), pp.16611675.
Hashimoto, H. et al., 2006. "Experimental and numerical
studies on parametric roll of a post-panamax container
ship in irregular waves". In 9th Int. Conf. on the
Stability of Ships and Ocean Vehicles, Rio de Janeiro,
Brazil. pp. 181190.
Hashimoto, H., Umeda, N. and Sakamoto, G., 2007.
"Head-sea parametric rolling of a car carrier". In 9th
Int. Ship Stability Workshop, Hamburg, Germany. pp.
3031.
Hashimoto, H. and Umeda, N., 2010. "A study on
quantitative prediction of parametric roll in regular
waves". In 11th Int. Ship Stability Workshop.
Wageningen, Nether- lands, pp. 295301.
Kim, Y. et al., 2011. "On Uncertainty in numerical analysis
of parametric roll" . In 12th Int. Ship Stability
Workshop. Washington DC, USA.
Neves, M.A.S., Prez, N.A. and Valerio, L., 1999. "Stability
of small fishing vessels in longitudinal waves". Ocean
Engineering, 26(12), pp.13891419.
Spanos, D. and Papanikolaou, A., 2007. "Numerical
simulation of parametric roll in head seas". Int.
Shipbuilding Progress, 54(4), pp.249267.
Spanos, D. and Papanikolaou, A., 2009. "SAFEDOR
international benchmark study on numerical
simulation methods for the prediction of parametric
rolling of ships in wave". NTUA-SDL Report, Rev, 4.
Umeda, N. et al., 2004. "Nonlinear dynamics on parametric
roll resonance with realistic numerical modeling". Int.
Shipbuilding Progress, 51(2), pp.205220.
Umeda, N. et al., 2012. "Estimation of parametric roll in
random seaways". In Parametric Resonance in
Dynamical Systems. pp. 4559.
Yu, L., Ma, N. and Gu, X., 2012. "Study on parametric roll
and its rudder stabilization based on unified
seakeeping and maneuvering model". In 11th Int. Conf.
on the Stability of Ships and Ocean Vehicles. Greece.
Yu, L., Ma, N. and Gu, X., 2016. "Early detection of
parametric roll by application of the incremental
real-time Hilbert Huang Transform". Ocean
Engineering, pp.224236.
89
 Subdivision Optimization of LNG Fueled RoPax Ship
Teemu Manderbacka, Pekka Ruponen, Daniel Lindroth, Markus Tompuri
NAPA, Finland, teemu.manderbacka@napa.fi, pekka.ruponen@napa.fi, daniel.lindroth@napa.fi, markus.tompuri@napa.fi
ABSTRACT
Stricter requirements imposed on ships through
emission control areas (ECA) can be fulfilled by using
LNG as fuel. Low flashpoint fluids onboard have raised
some concerns with respect to the safety level, especially
on passenger vessels. In addition, recently revised SO-
LAS regulations pose more stringent requirements to
the survivability of ships in case of damages involving
flooding of a ro-ro space. The number of LNG fueled
ships is increasing and new projects are initiated. Due to
fairly new development towards LNG fueled ships, de-
signers are still lacking extensive experience and
knowledge on the restrictions and possibilities for LNG
fueled ship. The designer is also faced with an additional
challenge of safe positioning of the LNG tanks. An op-
timal subdivision of an LNG fueled ship is studied by
applying a multi-objective genetic algorithm. The study
concentrates on feasible design alternatives of an LNG
RoPax in order to obtain a design with a good safety
level in case of damage. Calculations are performed ac-
cording to the SOLAS2009 and accounting for the Wa-
ter-on-Deck effects (Stockholm Agreement).
INTRODUCTION
Ship designers are facing stricter requirements with re-
spect to the environmental impact and safety. Emissions of
exhaust gases are limited in the densely operated special
Emission Control Areas (ECA). Moreover, the require-
ments for Energy Efficiency Design Index (EEDI) must be
taken into account. In addition, recently revised SOLAS
regulations pose more stringent requirements to the surviv-
ability of ships in case of damages involving flooding of a
ro-ro space.
Significant reductions in the emissions of NOx, SOx
and CO2 can be reached by using LNG (Burel et al. 2013).
Requirements imposed on ships in the emission control
areas can be fulfilled by using LNG as fuel. With respect to
the energy efficiency, an LNG fueled RoPax can achieve a
significant advantage of ~ 26 % better EEDI than the one
powered by fuel oil (GasPax 2013). However, low flash-
point fluids onboard have raised some concerns with re-
spect to the safety level, especially on passenger vessels.
The shipping companies still perceive the possible impacts
on safety as one of the important barriers impeding to take
90
LNG into use (Acciaro et al. 2013). Regulation and cost are
also perceived as barriers.
The number of LNG fueled RoPax vessels is still rela-
tively low, only a couple are currently in operation. How-
ever, several RoPax newbuilding projects are LNG fueled
or LNG-ready vessels. Altogether, the number of LNG
fueled vessels worldwide is rising, there are 162 confirmed
LNG ship fuel projects worldwide (DNVGL 2016). Special
attention in the design of an LNG fueled ship must be paid.
The new mandatory code for ships fueled by gases or other
low-flashpoint fuels was adopted by IMO in June 2015
(IMO, 2015). The International Code of Safety for Ships
using Gases or other Low-flashpoint Fuels (IGF Code) will
enter into force on 1 January 2017. The IGF Code contains
mandatory provisions for the arrangement, installation,
control and monitoring of machinery, equipment and sys-
tems using low-flashpoint fuels. LNG bunker tanks are
spherical and insulated, and they occupy roughly twice the
space occupied by fuel oil, when the regulatory constraints
are taken into account (Eide DNVreport 2010).
Large void spaces would be required between the LNG
tanks and the ship side shell. Asymmetric flooding in case of
side breaches can be avoided by equalization of floodwater be-
tween void spaces through a cross-flooding duct. Several new
cruise ships have been built with void spaces or double skin
around the engine room compartments, after the Costa
Concordia accident. Because of this recent trend, it was
decided to study also if a similar design would improve the
safety level of the RoPax ship. Besides the void spaces
around the LNG tanks and the engine rooms, damage sta-
bility is also dependent on the subdivision i.e. the locations
of the longitudinal or transversal bulkheads, and the vertical
position of the decks.
Optimization studies can be performed to increase the
general safety level without compromising the operational
requirements of the ship design. Lyu et al. (2015) studied
subdivision of a damaged military ship by applying genetic
algorithm to optimize transversal bulkheads positions. Ge-
netic algorithms have been applied in several studies aim-
ing to optimize ship design (Turkmen and Turan 2007, Pa-
panikolaou et al. 2010, Boulougouris et al. 2011, Kout-
roukis et al. 2014). In the present study the starting point is
an initial design with a suitable hull form and cargo capaci-
ty. The target is to optimize the watertight subdivision in
order to maximize the attained subdivision index of SOLAS.
Parametric ship model created with NAPA software has
been applied in previous studies combined together with an
external optimization tools (Boulougouris et al. 2004, Pa-
panikolaou et al. 2010, Puisa et al. 2010, Zaraphonitis et al.
2012, 2013). In this study optimization is performed apply-
ing genetic algorithm combined with a parametric model of
the ship, both, the genetic algorithm, and the parametric
model are implemented and run in the NAPA software.
PARAMETRIC MODEL OF A ROPAX SHIP
The case study ship is a large RoPax ship for short in-
ternational routes, e.g. in the Baltic Sea, Fig. 1. The ship
has two vehicle decks for trucks and trailers. The main di-
mensions are listed in Table 1. The ship has dual fuel en-
gines, and in addition to LNG the ship is equipped with
MDO (Marine Diesel Oil) tanks.
The following geometrical constraints are applied:
hull form and displacement are kept constant
The total volume of LNG tanks must be at least
600 m3
The total volume of MDO tanks must be at least
600 m3
There must be two engine room compartments
with a minimum length of 13.9 m, in order to ac-
commodate the main engines
If these criteria are not met, the design is considered to
be unfeasible.

Fig.1. LNG fueled RoPax

Table 1. Main dimensions

Gross tonnage (approx.) 45 000
Length over all 215 m
Breadth 30.0 m
Design draft 6.5 m
Service speed 22 kn
Passengers 2000
Crew 250
Fig. 2. Original subdivision and the design variables
Lifeboat capacity (N1) 1000
No. of persons in addition (N2) 1250 IGF CODE
Lane meters (approx.) 2 300
The allowed location of the LNG tanks is limited by the
A parametric 3D model of the ship was created with the IGF Code (IMO, 2015). This restricts both the double bot-
NAPA software. The following design variables were used: tom height and the width of the U-voids around the tanks.
transverse bulkheads (13 bulkheads) The tanks should be located at a minimum distance of B/5
longitudinal bulkheads around the LNG tanks or 11.5 m, whichever is less inboard from the ship side,
double bottom height where B is breadth. The distance of the fuel tanks from the
main deck height (freeboard) bottom should be at least B/15 or 2.0 m. The size of the
double skin around the engine room compartments tanks and the required protective distances from the side
(yes/no) and bottom bring additional requirements for the designer
to take into account with respect to the subdivision and
Altogether the design variables result in 18 parameters. damage stability of the ship.
Table 2 gives initial values, variation range, and variation
steps of the parameters. The parameters are illustrated in DAMAGE STABILITY CALCULATIONS
Fig. 2 along with the subdivision of the initial design.
The damage stability is calculated according to SOLAS
Table 2. Design variables II-1 Part II-1 Reg. 7. There is lifeboat capacity to all per-
variable initial variation variation sons onboard, and thus the required subdivision index is R
value range (m) step (m) = 0.793302.
(m) The damages up to 6 zones are calculated since longer
transverse bulkheads -2.4 2.4 0.8 damages would likely not contribute to the attained subdi-
LNG U-void width 6.6 6.6 8.0 0.2 vision index.
double bottom height 1.6 0.0 0.3 0.1 The proposed changes regarding flooding of ro-ro spac-
(positive upwards) es are applied, so that the s-factor, representing the surviva-
main deck height i.e. 9.7 -0.2 0.4 0.1 bility of a damage case, is calculated as:
freeboard (positive 1

upwards) GZ max range 4

s final = K (1)
ER double skin 2.4 1.0 4.5 0.7 TGZ max Trange
width

91
 where GZmax is limited to TGZmax and range is limited
to Trange. If the damage case involves a ro-ro space
TGZmax = 0.20 m and Trange = 20, otherwise TGZmax =
0.12 m and Trange = 16.
The effect of the heel angle is accounted with the co-
efficient:
15
K= (2)
15 7
when the heeling angle is between 7 and 15. With
smaller heel angles K = 1 and if the heeling exceeds 15
then K = 0.
The attained subdivision index for the initial design is
0.830993, which is larger than R-index, thus making the
original design feasible.
There are also special requirements concerning passen-
ger ship stability, namely Reg. 8.1 for damage in the for-
ward part and Reg. 8.2 for minor side damages. In addition,
there are requirements for bottom damages, Reg. 9. For the
studied ship design this is fulfilled by ensuring a minimum
double bottom height of 1.5 m throughout the length of the
ship.
The studied ship is intended to operate in Europe, and
thus the so-called Stockholm Agreement, EC (2003), for
accumulation of water on the main vehicle deck needs to be
fulfilled. This study concentrates on the subdivision opti-
mization during the initial design, and therefore the three
initial conditions of SOLAS are used instead of the real
loading conditions. Unrestricted operational area is as-
sumed and the applied significant wave height is 4 m.
CROSS-FLOODING
The LNG tanks need to be protected by large voids, and
the IGF Code (IMO, 2015) requires a distance of at least
B/5 to hull surface. The side effect of large voids is the risk
of asymmetric flooding and large transient heeling that
must be equalized rapidly with the use of effective
cross-flooding devices. This needs to be included in the
optimization process since the designs with slow
cross-flooding are not feasible. For the cross-flooding cal-
culations the U-voids are divided into two parts at center-
line, the cross-flooding duct is modelled as shown in Fig. 3.
Fig. 3. Modelling of cross-flooding ducts in the U-void
around the LNG tanks
92
Effect of air compression in a large void space can be
significant, Ruponen et al. (2013). In this study the cross-
flooding calculations are done according to IMO Resolution
MSC.362(92), IMO (2013), specifying that air compression
effects must be accounted if the air pipe area is less than
10% of the cross-flooding device area. However, excessive
air pipes are not feasible since they would take too much
space from the vehicle decks. The pressure losses in the
cross-flooding duct are accounted for with the so-called
k-sum parameter in the simplified method in MSC.362(92).
The paragraph 3.2 of the Resolution provides a very simple
method for estimation of the restrictive effect of air pres-
sure on the cross-flooding. This approach is fast to evaluate
and thus suitable for an optimization study. In practice the
effective total k-value for the cross-flooding device is:
2
S
k e = k w + k a a w (3)
w Sa
where kw is the k-sum for the cross-flooding device
(water) and ka is the effective k-sum for the air pipe. The
applied water density w = 1025 kg/m3 and air density a =
1.225 kg/m3. Sw and Sa are the effective area of the
cross-flooding device and air pipe.
Time-domain simulation would provide more realistic
results, Ruponen et al. (2012), but the simplified method in
MSC.362(92) is considered to be more suitable for optimi-
zation calculations.
The effective total area of a cross-flooding device is ap-
proximately:
S w = 0.35L zone H db (4)
where Lzone is the length of the zone and Hdb is the
height of double bottom from the baseline.
The k-sum for water flow in the cross-flooding devices
is estimated based on the length of the duct:
k w = N 1.778 (5)
where N is the number of longitudinal girders in the
duct:
L
N = int duct + 1 (6)
3Llfr
where Llrf = 0.7 m is the longitudinal frame spacing.
For the effect of air pipes a constant value ka = 3.0 is as-
sumed as an initial assumption, based on the CFD analyses
for a typical air pipe configuration with free discharge,
Ruponen et al. (2012).
The maximum feasible total air pipe size for each side
of one void is considered to be Sa = 0.4 m2, which corre-
sponds to 2 pieces of circular pipes with inner diameter of
about 500 mm.
A-CLASS BULKHEADS
The A-class fireproof bulkheads can seriously restrict
the flooding, and thus they must be accounted for in the
damage stability calculations. Alternative approaches for
this have been studied and compared by Ruponen and
Lindroth (2016).
In this study the A-class structures are accounted for by
calculating all possible combinations of intermediate
flooding stages, by considering the A-class bulkheads either
to collapse immediately or to remain practically watertight.
With this approach the number of alternative intermediate
stages of flooding increases enormously as the function of
the number of modelled A-class bulkheads. Therefore, in
the subdivision optimization it is reasonable to keep the
model as simplified as possible, Fig. 4. The final compli-
ance of the regulations can be done with the selected opti-
mal subdivision, using time-domain simulation in order to
ensure realistic treatment of various non-watertight struc-
tures inside the watertight compartments.
Fig. 4. Modelled A-class bulkheads (marked with red)
WEIGHT AND CENTER OF GRAVITY
The above mentioned damage stability calculations re-
quire the definition of initial conditions for the deepest sub-
division draught DS, lightest service draft DL, and a partial
subdivision draught DP. Since the hull form, lightship and
the total deadweight are kept constant, the draughts of the
initial conditions are also constant. However, the location
of the bulkheads, the tank top and the bulkhead deck are
varied in the optimization process. Naturally this affects
also the center of gravity of the ship, which must be ac-
counted for in the damage stability calculations.
A simplified approach is adopted, and it is assumed that
the changes in the subdivision affect only the vertical center
of gravity. When the height of the bulkhead deck (main
vehicle deck) is changed, also the upper vehicle deck needs
to be correspondingly changed, and this must be included in
the estimation of the changed vertical center of gravity. For
the other decks in the superstructure it is assumed that the
changes are minimal and can be arranged by changing the
free height of the passenger areas.
For the initial conditions DS and DP also the change in
the vertical center of gravity for the deadweight must be
accounted when the main deck location is changed.
OPTIMIZATION PROCESS
The number of design variables is large, and there is no
direct information about the mathematical behavior of the
objective space of the watertight subdivision of a ship re-
garding the attained subdivision index. There are conflict-
ing objectives and constraints, and therefore, adoption of
Multi-Objective Genetic Algorithm (MOGA) appears to be
the only solution to the set optimization problem, Sen and
Yang (1998).
The optimization objectives are to maximize attained
subdivision index A and to minimize the required GMmin
according to the Stockholm Agreement. The order of com-
pliance checks is based on the required computational time.
The fastest checks to calculate are performed first, so that
non-feasible designs are ruled out quickly, in this way ex-
tensive calculations are not performed for non-feasible de-
signs. For example, the Reg. 8.1 requirement for bow dam-
age is much faster to calculate than the Reg. 8.2 minor
damages, and therefore it is checked first. The whole opti-
mization procedure is illustrated in Fig. 5.
Design variation
Check geometry failed
constraints
ok
Multi Obejctive Genetic Algorithm Optimization
failed
Check IGF code
ok
Update CoG and
initial conditions
failed
Check Reg. 8.1
ok
failed
Check Reg. 8.2
ok
Check Stockholm failed
Agreement
ok
SOLAS 2009
index calculation
Fig. 5. Optimization process
RESULTS
Altogether 2673 number of designs were generated by
the optimization algorithm of which 226 were considered
feasible. 115 without double skin and 111 with double skin.
Majority of non-feasible designs were ruled out by the ge-
ometry or IGF-Code checks. The total computation time for
the optimization process was about two days on a single
computer.
There are two groups of designs, namely the ones with-
out side voids and the ones with side voids around the en-
gine rooms. Higher A-indices are obtained for designs
without the side voids than for the designs with side voids.
Moreover there is less scatter in A-index for the designs
without the side void. Some of the generated designs with
side void reach nearly as high A-indices as the ones without
the voids, however in general the designs with the side void
have poorer A-indices. Fig. 6.
Increasing the height of the bulkhead deck i.e. the
freeboard height improves A-index. On the other hand,
increasing the tanktop height has the opposite effect on the
A-index. Increasing the vertical positions of the decks in-
creases KG, this seems to have negative effect with respect
to the tanktop position, however for the bulkhead deck the
positive effect of increased freeboard height results in total
for improved A-index. Looking at the combined effect
namely the distance between the bulkhead deck and the
tanktop improves A-index, see Fig. 6.
Location of the longitudinal bulkheads around LNG
tanks was varied, its effect does not seem to play an im-
portant role for the values of A-index, see Fig. 7.
93
 8.2 are mainly correlated by the height of the main deck,
no double skin
this is more pronounced for the designs without the side
double skin voids.
optimum design without double skin Two different optimal designs were selected, one with
optimum design with double skin double skin in the engine room compartments and one
initial design without. The selection was based on A-index, but in addi-
0.87
tion it was required that the minimum GM due to Stock-
holm Agreement and Reg. 8.2 had to be less than 2.0 m.
0.86
A-index for the optimal design without double skin is
0.864946 and with double skin 0.866617, i.e. only a mar-
ginal difference. There is a notable improvement in the
0.85
A-index of the optimal designs when compared to the initial
design.
0.84
A-index

0.83 0.87

0.82 0.86

0.81 0.85

0.8
0.84

A-index
7.8 8 8.2 8.4 8.6
distance between maindeck and tanktop (m)
0.83

Fig. 6. A-index versus change of the distance between the

main deck and the tanktop. 0.82
no double skin
double skin
0.81 optimal design without double skin
no double skin
optimal design with double skin
double skin intial design
0.8
optimum design without double skin
1.6 1.8 2 2.2
optimum design with double skin
WOD min GM(m)
initial design
0.87 Fig. 8. Minimum required GM for Water on Deck (Stock-
holm agreement) versus obtained A-index for designs with
0.86 and without side void.

0.85 no double skin

double skin
optimal design without double skin
0.84
A-index

optimal design with double skin

initial design
0.83 0.87

0.82 0.86

0.81 0.85

0.8 0.84
A-index

6.5 7 7.5 8
LNG U-void width (m) 0.83

Fig. 7. A-index vs. U-void width around the LNG tanks. 0.82

With respect to minimum required GM in case of Water 0.81

on Deck (Stockholm agreement) the presence of the side
skin gives slightly smaller limitations than the design with- 0.8
out the double skin, Fig. 8. Designs with double skin are 1.6 1.8 2 2.2
less limited by the deterministic regulation 8.2., for which Reg. 8.2. min GM(m)
smaller minimum GM is allowed than for the designs
without the side skin, see Fig. 9. The requirement of the Fig. 9. Minimum required GM for Reg 8.2 versus obtained
minimum GM set by the Stockholm agreement and the Reg A-index for designs with and without side void.

94
 Both optimal designs are compared to the initial design
in Figs. 10 and 11. Interestingly, the length and location of
the main engine rooms is not changed. In the case with the
double skin, the width of the void space around the engine
room was very small, only 1.0 m.
The A-index calculation is limited to collision damages,
and thus the double bottom height has no significant effect.
Including a probabilistic model for grounding damages
might have led to different results in this respect. In both
optimal designs the main deck height has been increased by
0.4 m, i.e. the maximum allowed distance in the optimiza-
tion process. This is an expected result since the freeboard
height has a significant effect on damage stability of a
RoPax ship.
In both optimal designs the LNG tank compartment was
shifted one frame (0.8 m) forward and the width of the
voids around the tanks was increased by 1.0 m. interesting-
ly, the transverse division of the U-void was in both cases
shifted from the center of the void to forward.
In addition, in both aft and fore ship several bulkheads
were shifted, up to 3 frames (2.4 m). Especially in the for-
ward part of the ship the optimal subdivision is notably
changed from the initial design.

Fig. 10. Comparison of the optimum design without double skin and the initial subdivisions.

Fig. 11. Comparison of the optimum design with double skin and the initial subdivisions.

95
CONCLUSIONS
LNG as fuel is a viable option for the fuel of RoPax
ships, especially when operating in the emission control
areas. However, the LNG tanks need to protected against
both collision and grounding damages. This makes the wa-
tertight subdivision more complex than with the tradition-
ally fueled ships. In this study the subdivision of an LNG
fueled RoPax ship was optimized in order to improve the
safety of the ship. The A-index of SOLAS2009 was used
as a measure of the safety level against collision damages.
In addition, other relevant stability regulations need to be
used as constraints. Based on the results, two alternative
optimal subdivision arrangements were selected; one with
double skin around the engine room compartments and one
without.
The strict deterministic requirements of the minimum
distance to the hull surface in the IGF Code leave a little
room for improved safety level within the probabilistic
framework of SOLAS2009. The optimization resulted in
significantly larger A-index, but most of this improvement
was obtained by increasing the freeboard height and by
finding more optimal locations for the transverse bulkheads.
The size of the U-void around the LNG tanks had only a
small effect of the A-index. However, in both of the select-
ed optimal designs the width of the void was increased.
Although large voids may cause problems in ensuring fast
enough cross-flooding, it seems that in the presented case
asymmetric transient flooding was not a major factor, af-
fecting the optimal design of the subdivision. However, this
should not be generalized and further research on the topic
is still needed.
The double skin around the engine room compartments
did not in general improve the A-index since it increased
the asymmetry of flooding. However, for large raking side
damages this kind of arrangement can improve the surviva-
bility. However, it is considered to be extremely important
to correctly account for cross-flooding devices, including
air pipes, in damage stability calculations already in the
initial design phase.
The presented approach for optimizing the subdivision
of a RoPax ship is a fast and efficient way to improve the
safety of the ship in the event of a collision damage. Use of
LNG as a fuel introduces new constraints to the optimiza-
tion process, and especially the large voids around the LNG
tanks need to be properly accounted for the damage stabil-
ity calculations.
REFERENCES
Acciario, M., Hoffmann, P.N., Eide M. S. (2013), The
Energy Efficiency Gap in Maritime Transport, J. Ship-
ping and Ocean Engineering Vol 3, pp 1-10.
Boulougouris, E., Papanikolaou, A., Pavlou, A. (2011),
Energy efficiency parametric design tool in the frame-
work of holistic ship design optimization, J. Engineer-
ing for the Maritime Environment, 225 (3), pp.242-260.
Boulougouris, E.K., Papanikolaou, A.D., Zaraphonitis, G.
(2004), Optimization of Arrangements of Ro-Ro Pas-
senger Ships with Genetic Algorithms, Ship Technology
Research Vol. 51, pp.99-105.
96
Burel, F., Taccani, R., Zuliani, N. (2013), Improving sus-
tainability of maritime transport through utilization of
Liquefied Natural Gas (LNG) for propulsion, Energy,
Vol. 57, pp.412-420.
DNVGL (2016) Full list of LNG vessels in operation and
on order (as of March 2016)
https://www.dnvgl.com/maritime/lngi-business-intelligen
ce-portal-for-lng-industry.html?utm_source=dnvgl.com
&utm_medium=maritime-hero-area&utm_campaign=lng
i [Accessed 31 May 2016]
EC (2003) Directive 2003/25/EC of the European Parlia-
ment and of the Council of 14 April 2003 on specific
stability requirements for ro-ro passenger ships
Eide, M.S. (2010), Assessment of measures to reduce fu-
ture CO2 emissions from shipping, DNV Report,
www.dnv.com
GasPax (2013) note LNG-fuelled ships - Research find-
ings: GasPax - LNG for passenger ships, Naval Archi-
tect, (SEP), pp.S24-S25
IMO (2013), Resolution MSC.362(92): Revised Recom-
mendation on a Standard Method for Evaluating
Cross-Flooding Arrangements. Int. Maritime Org., Lon-
don
IMO (2015) Resolution MSC.391(95): International Code
of Safety for Ships Using Gases or Other
Low-Flashpoint Fuels (IGF Code), Int. Maritime Org.,
London.
Lyu, Z., Ma, K., Liu, F. (2015), Military ships subdivision
optimization for reinforcement of anti-wind capacity af-
ter damage, J. Marine Science and Technology, Vol.
20(3), pp.579-589
Koutroukis, G., Papanikolaou, A., Nikolopoulos, L., Sames,
P., Kpke, M. (2014), Multi-objective optimization of
container ship design Proc. IMAM Conf., 1, pp.477-489.
Papanikolaou, A. (2010), Holistic Ship Design Optimiza-
tion, CAD Computer Aided Design, Vol. 42,
pp.1028-1044. doi:10.1016/j.cad.2009.07.002
Papanikolaou, A., Zaraphonitis, G., Skoupas, S., Boulou-
gouris, E. (2010), An Integrated Methodology for the
Design of Ro-Ro Passenger Ships, Ship Technology
Research Vol. 57, pp.24-37.
Puisa, R., Tsakalakis, N., Vassalos, D. (2010), Reducing
uncertainty in subdivision optimization, Proc. 4th De-
sign for Safety Conference, Trieste, Italy
Ruponen, P., Queutey, P., Kraskowski, M., Jalonen, R.,
Guilmineau, E. (2012), On the Calculation of
Cross-Flooding Time, Ocean Engineering Vol. 40,
pp.27-39. doi:10.1016/j.oceaneng.2011.12.008
Ruponen, P., Kurvinen, P., Saisto, I., Harras, J. (2013), Air
Compression in a Flooded Tank of a Damaged Ship,
Ocean Eng. Vol. 57, pp.64-71.
doi:10.1016/j.oceaneng.2012.09.014

Ruponen, P., Lindroth, D. (2016), Time-Domain Simula-

tion for Regulatory Flooding Analysis, PRADS2016,
Copenhagen, Denmark.

Sen, P., Yang, J.B. (1998), Multiple Criteria decision

Support in Engineering Design, Springer-Verlag, Lon-
don

Turkmen, B.S., Turan, O. (2007), A new integrated mul-

ti-objective optimisation algorithm and its application to
ship design, Ships and Offshore Structures Vol. 2/11,
pp.21-37

Zaraphonitis, G., Skoupas, S., Papanikolaou, A., Cardinale,

M. (2012), Multi-objective Optimization of ROPAX
ships considering the SOLAS 2009 and GOALDS dam-
age stability formulations, Proc. STAB2012, Athens,
Greece.

Zaraphonitis, G., Boulougouris, E., Papanikolaou, A.

(2013), Multi-objective Optimization of Cruise ships
considering the SOLAS 2009 and GOALDS damage
stability formulations, Proc. IDFS 2013, Shanghai,
China, pp.89-97.

97
 Representing Military Behavior in Naval Ship Evacuation
Simulations including Flooding Damage Scenarios

David L. L. Sicuro1
Jos Marcio Vasconcellos2
Dracos Vassalos3
1
Brazilian Navy Research Institute (IPqM), Rio de Janeiro, Brazil david@ipqm.mar.mil.br
2
Rio de Janeiro Federal University (UFRJ), Rio de Janeiro, Brazil jmarcio@ufrj.br
3
University of Strathclyde, Glasgow, UK d.vassalos@strath.ac.uk

ABSTRACT objective is to achieve consistent results in personnel evac-

The application of simulators, able to represent dif-
ferent scenarios, environments and individual charac-
teristics of each occupant enable consideration even of
extreme situations involving breach of the vessel exter-
nal boundary, as a result of an accident or indeed war-
fare. The simulation process is an important decision
support tool to improve the existing evacuation proce-
dures in a number of ways. Re-configuration of com-
partments and gateways are examples of design
measures that could be taken and implemented by using
simulation. Rules and safety guidelines can similarly be
developed by a better understanding of the evacuation
process.
This paper addresses the adequacy of a passenger ship
simulation system for evacuation of personnel in war-
ships considering all the specifics of this type of vessel as
well as the individual characteristics of naval crew. The
evacuation simulation system that is the object of this
study is Evi, developed at the Department of Naval Ar-
chitecture and Ocean and Marine Engineering, Univer-
sity of Strathclyde. This examination of suitability in-
volves modeling adjustments to fit the behavior of a
military crew, changes in the way of modeling the phys-
ical arrangements and inclusion of new functionalities.
The results of this process are presented in this paper,
considering a case study including flooding damage
scenarios.
uation drills in military vessels through an evacuation sim-
ulator typically built for passenger ships adapted for this
different application. This option originated in the absence
of an evacuation simulator developed specifically for mili-
tary ships and crew. The steps to achieve this goal are de-
scribed below and detailed in the body of this paper.
First, the focus is on the understanding of the problem. This
entails characterizing a warship by describing its peculiari-
ties and differences in relation to passenger ships in the
physical, operational and behavioral aspects of her crew.
Next, we will briefly outline the modeling techniques cur-
rently employed for simulators in order to investigate what
happens during the evacuation simulation. Subsequently,
we will explain the reasons for the choice of evacuation
simulator to be adapted to contemplate military ships and
how it was made. Finally, a case study will be presented
which will simulate the evacuation of the crew of a Brazil-
ian Navy Frigate, followed by discussion of the results.
UNDERSTANDING MILITARY SHIPS
A warship has several characteristics that differentiate it
from a passenger ship. Below the most important specifics
considered for the simulation system adaptation process
will be described.
Physical Arrangement

A warship has a very characteristic physical arrangement.

INTRODUCTION
This paper is related to decision support systems using sim-
ulations. This area of research, within the marine engineer-
ing, aims to provide reliable information to assist decision
on the physical configuration of a naval plan and proce-
dures to be adopted by the crew in certain situations. The
There is a huge concern for the water tightness of the com-
partments due to the high probability of damage and the
need for keeping survival and functionality intact. The
compartments are usually small whilst the corridors are
narrow and have well-defined directions of traffic. The type
of the passages between compartments is different from a
passenger ship due to concern about the water tightness and

98
the deck change is usually via stairs with high inclination,
including passage through watertight hatches.
The physical plant also assumes different configurations
according to the mode in which the ship is operating. That
is, because of the mandatory closing of doors and hatches in
each mode of operation, ranging from a more relaxed clos-
ing condition "X", inside a harbor in peace time, through
the closing conditions "Y", "Z" to a very restrictive condi-
tion known as "Z-NBC" in which the ship is in a scenario of
nuclear, biological or chemical warfare.
Operational Modes
There are numerous modes of operation in a warship. Each
specific situation represents a different mode of operation.
Operations in the port, charge transfer between ships, refu-
eling maneuvers, a helicopter landing on a flight deck, etc.,
will determine a different operating mode. All of this, still
under the main condition that defines whether the ship is in
a war or peace scenario.
Crew Features
The crew of a warship presents a demographic homogeneity.
In the Brazilian Navy it is always composed of men only,
all of them well trained and knowledgeable of their tasks in
each mode in which the ship is operating. The reaction time
is very short and age differences usually do not affect the
moving speed, which differs drastically from the occupants
of a passenger ship. In a passenger ship we have from chil-
dren to elderly, men and women with different abilities, low
ship plant knowledge, and the number of people is general-
ly much larger than a warship.
This demographic homogeneity, however, presents other
forms of variation according to the operating mode in
which the ship is placed. The positioning on board, the ac-
tivity performed and the level of stress experienced vary
greatly depending on the current operating mode. For ex-
ample, the same man who works relaxed in an office com-
partment in peace time, in battle stations, will be in charge
of a 40mm cannon, positioned in the compartment ammu-
nition under very high stress.
Another peculiar feature of a warship is the existence of a
group of highly trained military experts with a great diver-
sity of skills, known as Group of Salvation and Destruction
of the ship (GSD). This group represents about 1/3 of the
ship crew (captain included). It is responsible for trying to
save the ship in extreme situations of damage, when the rest
of the crew already headed for their abandonment stations,
or even destroy the ship to make it unusable to the enemy,
in case of capture, if there is no possibility of salvation.
Evacuation Process
The literature related to people evacuation from different
types of plants, buildings or ships, uses the word evacuation
as a general term that includes the proceedings of escape
and abandonment from such facilities. A document that
distinguishes well these two steps is the ISO 13702 (ISO,
1999), which discusses about the control and extinguish-
ment of fires and explosions on offshore oil production fa-
cilities. According to this standard, the escape stage is asso-
ciated with the act of the people turn away from an immi-
nent danger area (fire, smoke, explosion, etc.), heading to a
place where the effects of these hazards are reduced or
eliminated, but still within the accident scene. The aban-
donment stage suggests a change of scenery: to a secure
situation or even to a secondary risk scenario, for example,
survival at sea, since this second scenario is outside of the
influence of the original accident scenario that caused the
evacuation.
Within a naval context, the situations that lead to the evacu-
ation of the crew of a ship usually result from damage im-
posed by the enemy or major incidents such as a fire out of
control, a collision or raking/grounding. Searching the ex-
isting Brazilian Navy's standards and procedures related to
evacuation, we try to separate this information into the two
phases, escape and abandonment, as the current approach.
The procedures specifically related to the escape phase are
based on a document called Master Table (CAAML, 2005).
The Master Table states that each crew member has its spe-
cific muster station, that is, the nearest possible survival raft
from the battle station of that person. The Brazilian Navy
has a policy coming from the US Navy for training its mili-
tary regarding this escape phase. "Exercise Abandonment
Stations" were adapted from US Navy Fleet Exercise Pub-
lication 4 (Rev. A) in MOG-S-7-SF section - Preparation
for Abandon Ship (US Navy, 1993), with the aim of as-
sessing the training of the military at this stage of the vessel
evacuation.
Another unclassified document from Brazilian Navy, "Sur-
vival Guide at Sea" (CAAML, 2007), describes procedures
of the escape and abandonment phases related to three situ-
ations directly related to the ship buoyancy's reserve and the
extent of damage: possible, probable and imminent capsize
or sinking.
In the possible capsize/sinking situation, there is no imme-
diate danger of sinking or capsize. In this case, where there
is the possibility of keeping the ship afloat or save it from
the enemy without great risk of loss of life, the Commander
will give the order "Prepare to Abandon ship", "Go to Mus-
ter Stations" and to the GSD, "Take Ship's Control", which
will be disseminated by the media available. The following
should be noted: 1) the crew, except the GSD, meets at the
muster stations. The crew with the highest rank in each
group checks for presence and sends the absence list to the
Chief Officer who transfers the list to the Commander. If
possible, the GSD will attempt the rescue of the absent
ones; 2) the crew members who stand in the stations to be
relieved by GSD members should only leave their stations
after the relief. If not relieved, they must remain there and
incorporate the GSD team; 3) the highest ranked member of
each group chooses 2 men as raft launchers. The specific
procedures for each type of raft are then started; 4) the
Chief Officer makes sure that radio equipment and GPS are
placed in each raft; 5) the Navigation Officer spreads by
any means available, again and again, the magnetic mark-
ing, the distance from the nearest land, whether friend or
not, the direction and the wind speed, the direction and the
99
speed of the current, the seawater temperature and the esti-
mated survival time in the water, the local depth and the
destruction procedure to be adopted (shallow or deep wa-
ters). The charts and navigation instruments that may be
useful must be prepared and shipped to the GSD raft ; 6)
The GSD continues their effort in an attempt to rescue the
ship, constantly informing the Commander of the situation ;
7) if the situation is critical, the Commander will give the
order to "Abandon Ship" and to the GSD, "Destroy the
Ship". Only after the order of "Abandon Ship" is given, that
survival rafts are launched overboard and personnel embark
them . Rafts intended for the GSD remain tied to the ship
and the others clear away, remaining as close as possible to
each other; 8) the GSD executes the planned actions to de-
stroy the ship, and each of the members reports "done" to
the Commander, then go to the muster stations. After the
destruction of ship by the GSD, the Commander will give
the order "GSD - Abandon Ship"; and 9) if the GSD can
save the ship after the abandonment of the crew, the captain
will give the order "Re-embark the Crew." The action is
coordinated quickly by the staff of the GSD, in order to
restore the priority of the command as soon as possible.
Survival rafts should remain attached to the ship to allow
their subsequent gathering.
In the probable capsize/sinking situation, there is no imme-
diate danger of sinking or capsize, but there is a high prob-
ability of the ship sinking or being captured by the enemy,
leaving only the alternative of destroying it. In this case, the
Commander will give the order "Prepare to Abandon Ship",
"Go to Muster Stations", and "GSD - Destroy the ship",
which will be disseminated by the media available. Proce-
dures 1 to 5 are the same as described in the situation of
possible sinking. Then, 6 upon receipt of the ready of man-
ning the GSD in the muster stations, the Commander will
give the order to "Abandon Ship"; and 7) only after the or-
der "Abandon Ship" survival rafts are launched overboard,
and personnel embark them. Rafts intended for the GSD
remain tied to the ship and the others clear away, remaining
as close as possible to each other. The action ends with the
same procedure (number 8) of the possible sinking case.
Finally, the imminent sinking situation that is, when there is
immediate danger of sinking or capsize, the Commander
will give the order "Abandon Ship" by the media available
and all pass the order on. The crew evacuates the ship, and
if possible, attends to the procedures used in case of proba-
ble sinking.
As can be seen from the above description, the procedures
of escape and abandonment in a military ship has specific
characteristics. The main differences are the action of the
GSD, the relief of crew members by GSD during the escape
phase and the organized way in which the movement of the
military crew happens.
MODELING TECHNIQUES
Several modeling techniques have been studied in order to
represent the pedestrian evacuation of a certain physical
space. Depending on the purpose of the study a different
technique is chosen. Below, we briefly describe the main
characteristics of the most used modeling techniques
100
(Xiaoping et al., 2009).
For example, if the goal is to simulate the physical interac-
tion between individuals and the physical structure of the
environment (for example, a crowd in a stadium forcing a
fence) the most appropriate technique would be the Social
Forces Models. In this technique, at each time t, the move-
ment of each pedestrian is determined by a sum of forces of
attraction and repulsion to other pedestrians and environ-
mental elements. This type of modeling is translated by a
continuous dynamic system for homogeneous groups on a
microscopic scale (Helbing et al. 1995).
Cellular Automata Models are discrete dynamical systems
in time and space where the three-dimensional environment
is represented by a regular grid of cells that can be occupied
by a single individual. Cells may be free (capable of being
occupied by someone) or fixedly engaged representing ob-
stacles, walls, etc. The individual walks (or not) from one
grid cell to another, every predetermined time interval, in
accordance with the values of variables that are assigned to
each cell and its neighbors, fulfilling predetermined move-
ment rules. The modeling using this technique works on a
microscopic scale of interaction. It can consider both ho-
mogeneous groups of individuals and heterogeneous groups
and are used to simulate evacuation in normal situations
and in emergencies (Yang et al., 2005).
Techniques based on fluid dynamics (Fluid-dynamics Mod-
els) describe how the density and velocity of a fluid vary in
time by partial differential equations. They are continuous
dynamic systems for homogeneous groups in macroscopic
scale applied to moving crowds. This type of modeling has
the following assumptions: the movement of pedestrians in
very dense concentrations has characteristics similar to flu-
ids in motion and can be described by nonlinear partial dif-
ferential equations; and, panicked crowds have similar be-
havior to the turbulence of a fluid on an obstacle or a nar-
row passage (Henderson, 1971).
There are also techniques such as Lattice Gas Modeling
where each pedestrian is considered an active gas particle
and the crowd behaves like a gas network in a regular grid
(Fredkin et al., 1982); the Game Theoretical Modeling pro-
vides that each member of the game evaluates all available
options to achieve a goal and rationally chooses to maxim-
ize success in reaching it (Lo et al., 2006); modeling based
on Behaviors of Animals (like ants), etc. Therefore, each
application will determine which modeling technique
should be used.
The technique that has proven most suitable for evacuation
drills on board passenger ships is the Agent-Based Models
(ABM) (Bonabeau, 2002). They are computer models able
to build complex social structures from the multiplicity of
relatively simple interactions between individuals repre-
sented by virtual agents. In other words, ABM is a compu-
tational model class for the simulation of actions and inter-
actions of autonomous agents in order to evaluate their ef-
fects on the system as a whole, i.e., ABM is a "bottom-up"
process that goes from the simple to the complex. Further,
to approximate the simulation of a real case, the Mon-
te-Carlo method is often used to introduce randomness to
the process. From this point of view somewhat chaotic, this
type of modeling shows to be extremely suitable for simu-
lations of confusion and panic situations precisely because
of the possibility of the lack of coordination of actions of
agents. ABM is the dynamic interaction of agents with "de-
cision-making power". It is like having a system based on
rules and facts, that is, an expert system "running" in each
agent in which we can configure features that can make this
agent more or less able to deal with a given situation. This
interaction can create a complexity that makes the simula-
tion system much like reality. Agents, acting as expert sys-
tems, gain intelligence and purpose in their actions. The
location of each agent, its responsive behavior and purposes
are encoded in the form of non-procedural algorithms in
computer programs. Interestingly, the algorithms that reside
on each agent can be different from each other (not only the
parameter settings), turning the agents in real individuals:
some smarter than others, some more responsive than oth-
ers, and so on.
CHOOSING THE SIMULATION SYSTEM
The first attempts to develop evacuation systems applied to
marine facilities began with a European community project
called Mustering and Evacuation of Passengers: Scientific
Basis for Design (MEPdesign) between 1997 and 1999
(Lee et al., 2003). The EVAC simulation program was de-
veloped from the knowledge generated in MEPdesign. It
had microscopic interaction approach and the initial distri-
bution of the passengers was defined by the user. Using
Cellular Automata Models, the movement and the interac-
tion between the passengers were simulated without, how-
ever, considering the movement of the ship.
At the same time, the Gerhard-Mercator-University in
Germany in 1999 began the BYPASS project with the ob-
jective of evaluating the evacuation time on passenger ships
using a simple Cellular Automata Model. The BYPASS was
the basis for the company TraffGo HT to develop the pe-
destrian evacuation simulation system PedGo, which has a
version that incorporates maritime features named AENE-
AS (Klupfel, 2000) using Agent-based Models (ABM). The
AENEAS remains in the market today with several inter-
esting features, but the interaction with hazard models like
flooding models were not implemented.
Already in 2000, the Fire Safety Engineering Group
(FSEG) at the University of Greenwich, England, devel-
oped an evacuation system called EXODUS. In 2003, in
collaboration with the Canadian company BMT FLEET
TECHNOLOGY, FSEG created a version of the evacuation
system for ships called maritimeEXODUS also based on
the ABM with interesting studies about the displacement of
agents under the effects of ship movement (Galea,
2001).This system continues to evolve today.
From 2001, also based on the knowledge generated by
MEPdesign and EVAC, The Ship Stability Research Center
at the University of Strathclyde and Safety at Sea Ltd.,
Glasgow, Scotland, in collaboration with Deltamarin, de-
veloped a so-called evacuation simulation model Evi
(Evacuability Index). Based on ABM, they introduced the
mesoscopic approach, emphasizing that both approaches,
microscopic and macroscopic, are required for a more real-
istic representation of the movement of agents in an evacu-
ation simulation (Vassalos et al., 2001). An interesting ap-
proach was used by modelling the time discretely and the
space continuously for the movement of agents, avoiding
the usual representation of the environment through a grid.
This approach allows more flexibility in the physical repre-
sentation of the environment and is particularly suitable for
environments with small and irregular compartments. An-
other great feature is the capability of integration of damage
scenarios as fire, smoke and flooding.
Also in 2001, the Korea Research Institute of Ships and
Ocean Engineering (KRISO) and Seoul National University
(SNU) developed an evacuation model called MonteDEM
using the Monte-Carlo method in a probabilistic approach.
MonteDEM uses the Fire Safety Evaluation Module
(FSEM) assessing the safety of fire on ships, calculating the
probability of having casualties, which was used as a safety
index. Based on Cellular Automata Models, this mesoscop-
ic model considered the effect of vessel motion forces in six
degrees of freedom acting on the movement of passengers
(Lee et al., 2003). In 2004 the same institutions developed a
model called Intelligent Simulation Model for Extrication
(IMEX) for large ships based on ABM, considering the
attributes of individuals and evaluating the evacuation time
and the procedures being executed. There were no more
recent references related to this work.
Focusing on three main evacuation systems, which remain
in development until the present day, AENEAS, Evi and
maritimeEXODUS, we can point the main technological
trends of this research area: use of agent-based modeling
(ABM) to consider the heterogeneity of individuals; devel-
opment of systems that consider the inherent uncertainty of
human behavior with statistical analysis; use of systems
that enable the integration of damage models such as fire,
smoke, flooding and the effects on individuals; and the de-
velopment of systems that include psychological factors in
the characterization of simulated individuals.
Among the main systems studied, none was entirely appro-
priate to the needs of a warship under damage. Undoubtedly,
the greatest demand is for passenger ships. As previously
mentioned, the main points that distinguish a warship from
other ship types are: the various operating modes and ship
plant configurations during a mission; the high level of risk
exposure leading to high probability of damage, especially
in combat situations; the procedures of the crew during the
evacuation, especially GSD; and the behavioral profile of
its military crew.
The main requirements of an evacuation simulation system
for a warship would be: great flexibility in the configuration
of environments; ability to simulate the movement of the
ship and its effects on crew moving; capacity to support the
integration of models of the main types of damage, such as
fire and smoke spreading and flooding, and its effects on
the crew; ability to simulate the GSD's members behavior
during evacuation; and ability to represent the profile of the
naval military in each participating agent.
Potentially all three major evacuation simulation systems
101
studied were able to incorporate adaptations for use in mil-
itary plants. However, Evi was chosen because it was the
simulation system that met two features considered very
important to achieve the objective of this and future works
in naval area:
Unlike other simulators that model the physical space in
a discrete way through a grid, Evi considers it in a
semi-continuous way. This was considered an important
factor in the modeling of a plant with small, irregular
compartments and narrow passages, characteristic of a
warship, where it would be harder to fit a grid exactly.
The combination of Social Forces (SF) e Cellular Au-
tomata (CA) techniques determined an unique approach
to the agents dynamics. VASSALOS et al., (2001) in-
troduced the mesoscopic modeling, merging the macro-
scopic and microscopic approaches, as explained by
Guarin et al. (2014):
The macroscopic behavior defines the way an agent
will travel from one location to another on board the
ship layout. Building on the graph structure defined
within the model, the process of identifying the
shortest route to a destination is achieved using
Dijkstra's classic shortest path algorithm (Dijkstra,
1959) with the weighting taken as the distance be-
tween doors. Once route information has been gen-
erated for each node, the process of travelling from
one point in the environment to another is just a case
of following the sequence of information laid down
by the search; this is referred to as the path plan.
The microscopic model covers the behavior of
movement of agents within spaces. It dictates the
way agents avoid boundaries of spaces and how it
should avoid other agents. Given these constraints,
the objective is to steer the agent towards a local
destination (waypoint) in an optimal manner without
being un-cooperative towards the other agents in the
space. In order to simplify calculation, a range of
discrete decisions are established around the agent
with the objective of identifying the one which will
allow the agent to travel the greatest distance toward
the local target. In addition, a continuous local (so-
cial/personal) space is established around each agent,
which other agents will aim to avoid. This space is
used to prevent a deadlock situation when the num-
ber of agents in an area becomes high. The agent
makes a decision of the best use of its personal space
to resolve any conflicts that may arise. As a result,
this approach allows the evacuation process to be
modeled in sufficient detail and still run in real time
or faster.
Ease of integration with the models for simulation of
fire, smoke and flooding was considered an essential
feature, aiming future works in naval evacuation analy-
sis under damage. Evi is designed to read and interpret
the data generated by the renowned software Fire Dy-
namics Simulator (FDS) for fire and smoke spreading,
produced by the National Institute of Standards and
Technology (NIST) of the US Department of Commerce,
and the VTT Technical Research Centre of Finland
102
(McGrattan et al., 2013). The results of flooding caused
by hull openings obtained by Proteus 3.1 (Jasionowski
et al., 2001) system produced by the Ship Stability Re-
search Center (SSRC) of the University of Strathclyde
and Safety at Sea Ltd, Glasgow, UK, are quite unique.
The movements of the vessel during the evacuation in-
fluence the speed of the crew, making the simulation
more realistic (Tsychkova, 2000). Of all the systems
considered only Evi is capable of dealing with flooding
scenarios.
The originality of this paper resides in applying the concept
of evacuability of Evi for the first time in naval vessels that
it is unique in this respect.
SUITABILITY ASSESSMENT PROCESS
The work of adequacy of Evi to meet simulations in mili-
tary vessels was performed in a joint effort directly with the
simulator development team at the University of Strath-
clyde and Safety at Sea Ltd in Glasgow, UK. After a de-
tailed study of the characteristics and features of Evi, the
main points identified for this adaptation were: 1) the
change in the form of demographics defining the ship's
crew; 2) the inclusion of a function that could implement an
objective exclusively found in a military ship; 3) a change
in the way of crossing a passage (door or hatch) according
to the ship's closing condition at a given operating mode;
and 4) change of the form of movement of the crew in order
to better represent the behavior of a group of trained mili-
tary.
Setting the Crew
The simulation of an evacuation in Evi uses a script written
in a programming language based on TCL (Tool Command
Language). This script defines what will happen during the
simulation: the crew is characterized and positioned, goals
are set for each member and events can be triggered at spe-
cific points in time.
Regarding the demographic definition of the crew, this task
is performed in quite different ways on a passenger ship and
a warship. Usually in a passenger ship, passengers and crew
are treated in an impersonal way, being able to make use of
a random distribution of simulation agents in the ship envi-
ronment. In a warship, each military has a specific posi-
tioning for each vessel operation mode according to the
current condition being war or peace. Certain military per-
sonnel, especially members of the GSD, have specific tasks
during the escape phase leading to the need of a personal
identification of each crew member. This need was solved
by the creation of new commands for the definition of sim-
ulation agents where we could assign a unique and individ-
ual name to the agent, in addition to the usual demographic
definitions that characterize and place the agents in a pas-
senger ship. Thus, at any time of the simulation, we know
exactly where a particular agent is and if he is managing to
meet a set goal at that moment.
Taking advantage of an Evi feature to differentiate func-
tionally and visually crew members and passengers, we use
this to differentiate members of the GSD from other crew
members in order to facilitate visually the identification of
these agents during the simulation. Therefore, the new
commands created, AddNameCrew and AddNamePax,
identify and place in the ship's plant the GSD's members
and the normal crew, respectively. New commands in TCL
language were also to simplify the arduous task of defining
one by one each member of the crew (name, personal char-
acteristics, initial positioning, and basic objective) reducing
the size of the script and making it more readable, facilitat-
ing the development and maintenance. Below is an example
of a procedure called DefAgEx created to define a GSD
agent in an exercise situation of probable sinking: the agent
is the captain of the ship, male M, log-normal distribution
function for awareness time with offset AT1m, initially
placed at the d-1-CrewService12 compartment with the
basic objective of reaching the muster station known as
MS3_Deck01_Starboard.
DefAgEx Crew M GSD:CAPTAIN d-1-CrewService12
$AT1m MS3_Deck01_Starboard
The reaction time and speed are defined automatically in-
side the procedure by probabilistic log-normal and normal
distributions respectively, thus ensuring the differentiation
of agents within a simulation run and the difference be-
tween each simulation iteration when running in batch, for
statistical analysis.
Creating a new objective in simulation
As described above in the explanation of the procedures in
cases of possible and probable sinking, item (2), there is a
situation where a military should be relieved by a member
of the GSD before heading to muster station. If this member
of the GSD, for any reason (death, for example), did not
come to that place, the crew member who would be re-
lieved assumes the function and will integrate the GSD.
This is really a situation that does not exist in a passenger
ship. So it was necessary to create a new objective called
relieve to implement this. The syntax of this command is
shown below:
Objectives GSD:1103 {{relieve d-1-CrewService58 {evac
MS1_Deck02_Starboard} 1301} {wait 60} {...}}
In this example, the member of GSD identified as
GSD:1103 must relieve the crew member 1301 which is
waiting in the compartment d-1-CrewService58, which then,
must proceed to MS1_Deck02_Starboard abandonment
area. The member of the GSD should wait 60s in this place
before fulfilling his next goals.
Changing the way of crossing a passage
The version of Evi used for this work would assume a pas-
sage as open by default,. If the passage was initially defined
as blocked, it would not be part of a route until its condition
would change to open. In a passenger ship there is no need
to close a door after the passage of people during an evacu-
ation of the ship. In contrast for a warship, in a possible or
probable sinking situation, where there is still the possibil-
ity of the ship to be saved by the staff of GSD, there is the
need to preserve the water tightness of the compartments
and survival sectors for a possible recovery and
re-occupation of the vessel.
As explained above, depending on the ship's closing condi-
tion (X, Y, Z, ...), as a function of the mode in which the
ship is operating, several passages must remain closed
while others may remain open. Therefore, a procedure was
developed in Evi which adds a time to open a passage
which is defined as closed. At the present stage of devel-
opment of this feature, it is assumed that the door closes
automatically after the passage of the agent. In case of a
group of agents wanting to go through this passage, the
system has enough intelligence to allow the passage of the
entire group before closing the passage again. As can be
seen, the agent does not remain next to the passage in order
to close it, what usually needs to occur in most doors and
hatches of a conventional naval ship. Despite this, the new
implementation does quite accurately represent what occurs
in the vessels of the Brazilian Navy, naturally causing a
delay in the movement of agents, turning the simulation
very close to the operation of a military ship.
This version of Evi does not include hatches modeling. This
restriction was solved as follows: where there was a small
isolation lobby to access the hatch, it was simply consid-
ered the duplication of the times for opening and closing of
the lobby door, considering the door + hatch system as a
single entity, taking advantage of the physical proximity of
them; and, in cases where there were no isolation lobby, a
virtual one was introduced in the model. Its virtual door,
then, assumes the properties of real hatch. The simulator
development team already includes hatch modeling for fu-
ture versions.
Representing the movement of a group of well trained
military
The biggest challenge in the process of adaptation of an
evacuation simulator for passenger ships to military ships is
the representation of the crew behavior during the evacua-
tion. The simulators generally develop their behavioral
models in order to reproduce the passengers in panic situa-
tions, which renders the movement disorderly. In these cas-
es, it is common "to fight" for a better position during the
evacuation. The lack of knowledge about the ship is also
explored leading to the simulation of the behavior known as
"herding" where there is a tendency to go where everyone
else is going, avoiding to try an alternative path, among
other aspects.
However, in a warship the picture is totally the opposite.
All crew members have similar skills and level of fitness,
all are trained and prepared to face extreme situations of
danger and all know deeply the ship where they serve.
These reasons would be enough for the development of a
new behavioral model quite different from that used in
passenger ships. However, it was out of the scope of this
work to undertake such big changes in the behavioral mod-
el's paradigms of the current simulator. Therefore, some
attempts have been made in order to achieve a more orga-
nized behavior. Three lines of action were assumed as fol-
lows:
103
 Regarding the heterogeneity of the group: set the be-
havioral parameters of the agents in order to make the
group most homogeneous.
For disorganized behavior: change the physical model-
ing of the plant to create pseudo-corridors in more spa-
cious areas of the ship to induce agents to adopt a be-
havior as if they were walking in a "queue".
define the best route option in case of blockage. In the sec-
ond case, without totally blocking the usual route, conges-
tion not explicitly brings a new attitude to staff members
who are approaching the congested region to seek alterna-
tives. To bring about this attitude it has created a model that
measures the density of agents on that critical region, and
so a new route choice is activated or not. We called this
technique as "ship knowledge".

Concerning the knowledge of the ship in order to To evaluate the effectiveness of the proposed strategy to
choose other paths: use of special features of the simu- represent the organized behavior of a military crew, a small
lator to "induce" agents to choose another path in order plant tests will be used. For ease of presentation, the test
to avoid congestion on the usual route, despite the dis- plant is contained in only one deck. This test plant will be
tance being longer in this alternative route. modeled using the conventional technique used in passen-
ger ships and also with the changes proposed to represent
To make the group the most homogeneous, the age range the military's behavior. The objective is to evaluate the dif-
was considered narrow and the distribution of reaction time ferences in the behavior of agents and changes in time of
had its standard deviation reduced. The configuration of completion of the programmed movement.
speeds was made based on specific test described by Glen
et al. (2003). In this research, velocity measurements in the The plant of test, shown in Fig. 1, includes: the STARTING
plane and on stairs with 10 degrees of inclination, used in ROOM, where the agents are placed at the beginning of the
accessing hatches in naval vessels were made, having as simulations; the ACCESS CORRIDOR, which leads the
target population men and women with and without expe- agents to the KEY ROOM, where the proposed adaptations
rience with this type of equipment. Former military men will take place; and two ways to reach the DESTINATION
and women belonging to the Royal Navy and the Coast ROOM: the MAIN CORRIDOR (shortest route) and the
Guard have volunteered and joined the group with experi- ALTERNATIVE CORRIDOR (longer route).
ence. Among all results, the chosen were related to the male
group with experience, shown in Table 1:
Table 1 - Speed definition
Horizontal speed (m/s) Vertical speed (m/s)
Min Mean Max Min Mean Max
0.90 2.00 2.10 0.40 0.50 0.60
Fig. 1 - Test Plant
Speeds were assessed according to a uniform probability
distribution with the minimum, average and maximum
Three test situations were simulated: 1) conventional mod-
listed above. The small variation characterizes the group as
eling; 2) adapted modeling with "forced queue" implemen-
homogeneous as expected for a military operative crew.
tation; and 3) adapted modeling with "forced queue" and
"ship's knowledge" implementations. All simulations have
The reduction of disorganized behavior of the agents was
the same number of agents, with the same demographic
implemented, when possible, by reconfiguring the internal
characteristics and the same goals. The dynamic to be ob-
representation of the compartment. A compartment is nor-
served is the movement of 80 agents that are within the
mally represented by geometric elements (rectangles and/or
STARTING ROOM and start moving simultaneously to-
polygons) which, when grouped, make up the area of the
wards the DESTINATION ROOM. The congestion gener-
entire compartment. A geometric element is connected to its
ated at the output of the STARTING ROOM to access the
neighbor through a pseudo-passage known as union, occu-
ACCESS CORRIDOR is not taken into account in the
pying the entire length of contact between the two elements.
analysis. The focus of the analysis is what happens in the
The choice of geometric forms of the elements and the
KEY ROOM.
quantity of them is not restricted, allowing for a simplified
representation. Given this flexibility, it was possible, in
Test 1:
some cases, to create sets of elements favoring the for-
In this simulation, the plant is modeled in the conventional
mation of a row of agents making the evacuation more or-
way. The KEY ROOM is represented by only a Basic Ge-
ganized. We called this technique as "forced queue".
ometric Element (BGE), in this case, a rectangle. The Pri-
mary Route is defined including ACCESS CORRIDOR, the
Regarding the ability to knowing the layout of the ship to
KEY ROOM and the MAIN CORRIDOR. Fig. 2 shows the
choose other paths, there are two distinct situations: the
test plant with conventional modeling and the indication of
usual route is blocked; and the usual route is just congested.
the Primary Route.
In the first case, the strategy was to combine a feature of the
simulator for setting primary routes (Primary Route) along
with another feature that allows one to generate alternative
output signals (Primary Exit) in a particular passage during
the modeling phase of plant. This feature was designed to

104
 Fig. 2 - Conventional Modeling (Primary Route)
In this type of modeling, Evi simulator tries to represent the
situation observed in a passenger ship. The dimension of
the door that separates the KEY ROOM to the MAIN
CORRIDOR is far from ideal for the expected flow, caus-
ing a major congestion. The simulation agents representing
passengers, unaware of the existence of an alternative route
and, therefore, they clog, seeking access to the only route
that is being used by all, in a typical situation of herding.
This causes a significant delay in arrival at the destination
and definitely is not a trained military behavior. Fig. 3
shows congestion obtained in the test with conventional
modeling.
Fig. 3 - Congestion in the Conventional Modeling
Test 2:
In this simulation, the plant is modeled in the adapted way.
The KEY ROOM is represented by several Basic Geomet-
ric Elements, with the clear intention of inducing the for-
mation of a queue. The Primary Route is defined including
ACCESS CORRIDOR, some Basic Geometric Elements of
the KEY ROOM and the MAIN CORRIDOR. The alterna-
tive output is also defined, in the event of blockage of the
Primary Route. Fig. 4 shows the new subdivision made in
the KEY ROOM and Fig. 5 shows the test plant with the
adapted modeling, the definition of the Primary Route (now
including the Basic Geometric Elements BGE1, BGE4 and
BGE6) and the alternative output signage.
Fig. 4 - Adapted Modeling of the Key Room
Fig. 5 - Adapted Modeling (Primaries Route and Exit)
The result of this simulation shows a much more organized
behavior of agents. The dimension of the door that sepa-
rates the KEY ROOM to the MAIN CORRIDOR remains
less than the ideal for the expected flow, however, the for-
mation of the queue favors the flow of the agents. As long
as there is not a blockage in the main route, the alternate
route is not used. This model does not include yet the "ship
knowledge" modeled in the script. Agents remain in the
"forced queue" until the end of the flow to the DESTINA-
TION ROOM. A large reduction was observed in the total
moving time, just because now the movement is more or-
ganized. Fig. 6 shows the queuing of agents obtained in the
test with adapted modeling.
Fig. 6 - Queuing in the Adapted Modeling
Test 3:
In this simulation, the plant is modeled as in Test 2. The
KEY ROOM is represented in the same way and the alter-
native exit as well. The Primary Route is also defined in the
same manner as in the previous test. Fig. 7 shows the test
plant with the adapted modeling, the definition of the Pri-
mary Route, the alternative output (Primary Exit) and the
Basic Geometric Element chosen as "conditional BGE".
In this modeling is included a decision-making mechanism
that evaluates the agents density in the "conditional BGE"
(BGE4), to represent the "ship knowledge". This modeling
is performed in the simulation script using the commands:
AtSpaceOccupancyCount BGE4 N {Se-
lectAgentsInSpaces {PreviousRoute}}
AtSpaceOccupancyCount BGE4 N {Objective select
{goto NewRoute}}
The command AtSpaceOccupancyCount assesses the num-
ber of agents in the "conditional BGE" BGE4. At the mo-
ment that the number of agents N is reached, the agents
which are moving in the PreviousRoute are selected and a
new objective is given to them. In this situation, they must
pass thru the NewRoute location. In this example, the Pre-
viousRoute could include the BGE6 and the ACCESS
CORRIDOR, and the NewRoute location could be the AL-
105
TERNATIVE CORRIDOR. The choice of these parameters
will depend on the analysis of each region in which one
wishes to model this behavior. Visually, the simulation rep-
resents better the situation of people arriving where a row is
becoming dense and then, they choose another way know-
ing that it will lead to the same destination, keeping the
organization of the moving. Decreasing the density in
BGE4, the normal path becomes the one chosen again. If
congestion evolve into a blockage, all agents, both those
already in line at KEY ROOM as those approaching it, they
would automatically be redirected to the passage defined as
Primary Exit.
The result of this simulation demonstrates a more organized
behavior by the agents, adding the aspects of knowledge of
an alternative path. The dimension of the door that sepa-
rates the KEY ROOM to the MAIN CORRIDOR remains
less than ideal for the expected flow and the formation of
the queue continues to favor the development of the move-
ment. Although there was not a blockage in the main route,
this time the alternative route is used. However, in a differ-
ent way. As the main route is not blocked, there is not a
total redirection of agents for the alternative route. Some
agents remain in the queue on the main route and some
agents "evaluate" that the queue is large and "choose" the
alternative route. As a consequence, a reduction in total
drive time is observed, featuring aspects of an organized
behavior, showing the knowledge about the paths in the
ship on which they serve. The time reduction was small
because the alternative pathway used by some agents is
longer than the main path, demonstrating that the results
obtained from this modeling are consistent. Fig. 8 shows
agents in a queue in the main route and agents which are
choosing another route.
Fig. 8 - Queuing and "Ship Knowledge" represented
The average times of 50 simulations for each of the three
cases tested are shown in Table 2. A progressive time re-
duction is observed from the case of conventional modeling
(Test 1) to the case with full adapted modeling (Test 3).
The simulations carried out have shown that it is possible to
satisfactorily represent the behavior of a military crew by a
specifically adapted modeling, using a simulator developed
for passenger ships. In the next section, a case study will be
presented in which all the adaptations suggested in this
work will be applied.
Table 2 - Simulation times
Times to accomplish the moving (sec)
Test
Mean Standard Deviation
Test 1 421.6 49.8
Test 2 219.0 26.1
Test 3 201.5 17.2
106
CASE STUDY
The warship chosen for this case study was a "Niteri"
Class Frigate, English design, built in Brazil between 1965
and 1970, with an average crew of 209 men. Ships at that
time were not as focused on the safety of the crew as today.
The focus was concentrated on the functional operative
efficiency, leaving security in the background. Actually, the
main reason probably was because it was very difficult to
predict certain problems associated with a possible evacua-
tion of the ship without the availability of a simulator.
Nowadays, especially in Navies with more resources and
ability to design their own ships, the design for safety of the
crew, of course, is a reality. This is a trend that will include
not only large passenger ships, but the merchant ships and
oil platforms, because of the ease of use of the simulators.
Talking about design for safety in navies with few resources
could be a theme somewhat inapplicable because not much
is done in this area since most of their ships with large ca-
pacity are old and purchased from other navies. However,
the update needs of these ships are frequent. Many sectors
of the ship could pass through a redesign phase to receive a
new weapon or a new sensor. At this point, we can use the
simulation capabilities to assess the impact that a change in
the ship's plant could have on the safety of the crew in a
possible evacuation situation.
The simulations were executed considering two scenarios,
both in a probable sinking situation. The first one repro-
duced the current doctrine used on board in evacuation ex-
ercises (drill case). This case was used to validate the simu-
lation model (adapted modeling) by comparing the real
time spent by all normal crew members (non GSD) to reach
their muster stations to the correspondent simulated time.
This time (Escape Time) was measured from the moment of
the Commander's orders (to the normal crew: "Prepare to
Abandon ship", "Go to Muster Stations" and to the GSD
members, "Take Ship's Control") to the moment when the
last member of the normal crew reached his muster station.
The time spent by the GSD members complete all their
duty and go to their muster stations was not used for this
validation purpose.
Some situations to use the command "relieve" have been
created following suggestions of the operating sector of the
ship. These situations are internally defined for each ship of
the class, with minor variations in procedures from one ship
to another. The procedures related to the application of the
command "relieve" were simulated correctly, following the
sequence of pre-determined actions. There were no prob-
lems for the agents of the GSD when they were going to-
wards the compartments where they were expected by
members of the normal crew. There were no critical situa-
tions of movement in counter-flow.
By military secrecy issues, the results presented are related
to reference values, privileging a qualitative analysis of the
adapted simulator under evaluation. The difference between
the real mean time obtained in the drill case, used as refer-
ence, and the mean simulated time after 50 runs was less
than 5%. This result validated the adapted model.

The Fig. 9 is related to the first scenario (drill case - current
doctrine), showing the congested regions. It represents a
snapshot of the moment of maximum congestion shown in
the graph of Fig. 10.
Fig. 9 - Congested Regions in the Scenario 1
According to the graph of Fig. 10, the maximum value of
congested agents in this case is 60 and the average was
calculated at 32.8 agents throughout the moving time T1.
The vertical axis (No. of Agents) is the number of congested
agents at each simulation time instant.
Fig. 10 - Congested Agents Time History in the Current
Doctrine (Scenario 1)
Once the adapted model has been validated, the simulation
results could be used to improve the evacuation procedures.
The second scenario includes some changes in the evacua-
tion procedures and in the physical plant of the ship in order
to improve the simulation results. The following three sug-
gestions aim to reduce the number of congested regions and
the Escape Time of the previous scenario:
1. Changing of routes for the personnel assigned to the
main deck (Deck 1) and the lower decks (Decks 2, 3 and
4). In this case all these people would have to go to the
stern of the ship using internal routes on decks 1 and 2.
Then they would access the external area of the aft main
deck, going up to the Flight Deck by an external stair-
case. The goal is to reduce the flow of people in the ex-
isting bottlenecks in regions 1 and 2;
2. Changing the ship's physical plant with enlargement or
duplication of external staircase used to access the
Flight Deck (Muster Station), in order to avoid conges-
tion in the stern region;
3. Traffic direction reversal of the outer starboard corridor
of the deck 01 only during the evacuation procedures.
The goal is to avoid going up to deck 02 and then going
down again to the deck 01 to use the port corridor.
The Fig. 11 shows just one congested region in the second
scenario. It represents a snapshot of the moment of maxi-
mum congestion shown in the graph of Fig. 12. In relation
to Scenario 1, one can see in Fig. 11 the elimination of
congestion in the regions 2 and 3 and a significant im-
provement in the region 1.
Fig. 11 - Congested Regions in the Scenario 2
According to the graph of Fig. 12, the maximum value of
congested agents in this case is 30 and the average was
calculated at 16.3 agents throughout the moving time T2.
The vertical axis (No. of Agents) is the number of congested
agents at each simulation time instant.
Fig. 12 - Congested Agents Time History in the new pro-
posal (Scenario 2)
The escape time in scenario 2 (T2) fell by 55% compared to
the reference time (T1) of scenario 1. However, this solu-
tion is only indicated if a paradigm change in the movement
of the crew is accepted, allowing violation of the traffic
rules.
FLOODING SCENARIOS
Evis integration with hydrodinamic simulations
107
The hydrodynamic simulation to be integrated with the
evacuation drill is achieved through the system Proteus 3.1
(Jasionowski et al., 2001) developed at Ship Stability Re-
search Centre (SSRC), University of Strathclyde, Glasgow,
UK.

The Evi simulator system uses the MEPdesign project re-

sults (Tsychkova, 2000) to reduce the speed of the agents
according to the ship's movements

Building the hydrodinamic model

Fig. 13 MONOLAX image for flooding damage
The Protheus System 3.1 requires input files that define the
hydrodynamic model of the ship to be studied: Damage simulations

The file extension ".g" ou ".geo", which contains the
geometrical description of the hull for simulations of the
intact case; and
The file extensions ".sus" e ".dam" which contain the
geometrical description of the hull, the internal com-
partments that could be object of damage study and the
damage characteristics itself, to the damage case.
These files are usually generated by specific computer pro-
grams for marine engineering like NAPA and PolyCAD, for
example.
The System Proteus 3.1 provides the results of hydrody-
namic simulations through the output files with extension
".mot" to the intact case and the files with the extension
".mot" and ".wat1 the .watn" to the damage case. In the file
".mot" we have, in a timescale, the ship's position relative
to the its center of gravity according to the number of de-
grees of freedom (DOF) defined for the analysis. In files
with the extension ".wat1" to ".watn", also in the same time
scale, the data are consistent with the sea water accumula-
tion in n internal compartments defined in geometric mod-
eling of the vessel under study. All these files are in text
format and will be used by Evi simulator for interaction
between simulations.
Three flooding damage cases will be presented: bow,
amidships and aft of the ship. The same initial conditions of
the crew with positioning in "Battle Stations" used in Sce-
nario 1 will also be used in these simulations.
The simulations integrated into the simulator Evi have a
visual representation. In the case of simulations integrated
with hydrodynamic simulations of Proteus 3.1, Evi features
a blue plane representing the sea. The 3D model of the ship
moves according to the movements of the sea and in the
event of a flooding damage, of course, also represents blue
in the flooded compartments. Also on the visual aspect, the
agents who die due to the action of flooding have its color
changed to red and leave the simulation. The Fig. 14 shows,
as an example, the visual representation of an integrated
simulation with flooding damage amidships.
Fig. 14 Evi in a flooding simulation

The SSRC also developed an extremely useful tool for the
evaluation of the results generated by the simulator Proteus
3.1, called MONOLAX (MONOLAX, 2013). Through this
application we can see all the movement of the ship in the
virtual ocean set for simulation within the timescale estab-
lished for that experiment. It is also possible to visualize
and quantify, in real time, the flow direction and quantity of
sea water that enters and leaves the vessel during a hull
damage simulation, and the free surface effects within the
flooded compartments. The visualization of the simulation
is quite interesting and it allows feeling the continuity of
the movement of the ship, discarding the presence of coarse
numerical errors generated in the simulation.
Two types of damage were simulated according to their
extension:
Damages of small proportions whose time for a
commitment of compartments involved were great.
Thus, it was defined as small proportion's damage, sizes
less than 1m, where the compartment's drain capacity is
not sufficient to solve the problem; and
Damages of major proportions whose time for react is
very short. Thus, it was defined as major proportion's
damage, sizes up to 15m, in the form of tear, reaching
more than one compartment in longitudinal direction.

Fig. 13 illustrates MONOLAX image generated by the ap-
plication to the simulated event of damage aft.
The design of military ships give too much emphasis to the
issue of sealing, especially in decks below the waterline. In
the case of vessels in this class, the simulations allowed to
note that this project priority was very well implemented,
basically because of the following reasons:

108
 Most compartments below the waterline have a small
longitudinal dimension; and
In decks below the waterline does not exist longitudinal
communication between compartments.
These two factors make this vessel fairly safe regarding to
the spread of a flooding. There is no possibility of forget-
ting an open door cause the spread of a flooding as there are
no doors in the bulkheads of the ship below the waterline.
The entire moving is in the vertical direction, being neces-
sary moving up to achieve higher decks and then moving
down, to progress in the longitudinal direction.
The reference time for this analysis will be the total time to
reach the abandonment stations in Scenario 1 (TSC1). The
simulations were carried out in the sea state 5 (coarse sea).
Batch of 50 rounds was run with all damages localized be-
low the waterline. The results obtained are shown in Table
3.
Table 3 - Simulation times
Times to reach the
Damage Situation
Casualties
abandonment stations
Rooms
Location Extention Mean
affected
short 1 0 TSC1+17.9%
Bow
long 3 4 TSC1+17.6%
short 1 0 TSC1+18.1%
Midship
long 2 4 TSC1+18.5%
short 1 0 TSC1+19.3%
Stern
long 2 2 TSC1+18.2%
It was observed that, regardless of the scale and location of
the damage (bow, midship or stern), there was no occur-
rence of blocking of escape routes, even considering that
nothing was done to mitigate the problem. The difference
observed according to the extent of damage was the number
of casualties among agents. In damages of small propor-
tions, there was enough time for agents to get access to the
higher decks, containing the flooding only in the affected
compartments. In damages of larger proportions, there was
no time for agents to evade, causing the death of these
agents.
Thus, the escape times showed no significant difference
compared to those obtained in the case of reference, but due
to the movements of the sea, which naturally causes a re-
duction in travel speed. The damage simulations were gen-
erated considering a drifting ship in Sea State 5, in order to
cause a worsening of the vessel's stability. As the ship was
in the closing condition "Z", in which the hatches to lower
decks must remain closed, the vessel remained stable even
in damages of greater extent. In consequence, it was not
verified any ship capsizing situation in 30 minutes of simu-
lation time, in any of the 50 simulation rounds.
CONCLUSIONS
In this article, we analyze the possibility of adapting an
originally designed simulator for passenger ships to simula-
tions in warships. The adaptation of Evi simulator allows
different possibilities to better assess the movement of per-
sonnel in a military ship. Following will be presented the
evaluation of each part of the adaptation process and some
considerations about the flooding cases:
1. The adjustments made to nominally set the simulation
agents allowed the identification of them in possible
functional analyzes in specific shipboard operations;
2. The creation of auxiliary procedures in TCL language
targeting a code compaction contributed to the clarity
and maintenance of simulation scripts;
3. The adaptations for the representation of hatches com-
pensated the absence of this feature in the original sim-
ulator;
4. The adjustments made to pass through a passage (open-
ing and closing doors) did not reproduce exactly the
procedures required for manually operated doors used in
this ship. However, the objective of causing a delay in
the agent's passage was achieved, offsetting this defi-
ciency;
5. The new command to relief a member of the regular
SD crew by a member of the GSD has been met partially.
The situation in which the member of the GSD does not
4.1% appear to assume the station was not implemented and
4.7%
is also expected for the next versions of the simulator;
5.2%
5.0%
4.4%
6. The Primary Exit feature was a good modeling tool to
represent the "ship knowledge" during the evacuation of
4.6%
the ship;
7. The attempts to change the disorganized behavior of the
passengers of a cruise ship into the behavior of trained
military achieve its goal. The test cases using a small
hypothetical plant demonstrate that the techniques of
"forced queue" and "ship knowledge" were able to
model the military organization, even without any
change in the core of the behavioral modeling program
of the simulator Evi. The definition of the demographic
parameters contributed to tune some aspects of the mil-
itary behavior of the agents.
8. In the simulations with flooding damage scenarios was
found that Evi simulation system uses a powerful hy-
drodynamic simulator with great possibilities for inte-
gration between programs. This ease of integration be-
tween the two simulators, by sure, lies in the fact that
they were developed in the same department at the
University of Strathclyde in Glasgow, UK. At the cur-
rent stage of development of this integration, the affect-
ed compartments are blocked and can no longer be part
of possible escape routes. Agents are affected gradually
in accordance with the progression of the flood level.
An agent is considered dead and excluded from the
simulation if the flooding level exceeds its height. This
condition does not accurately reflect the real case for an
analysis of casualties during a damage of flooding.
Resuming, the purpose of obtaining results for decision
109
support and changes through computer simulations was
reached. The results obtained in hypothetical proposals
suggested in the case study proved to be sufficiently con-
sistent and reliable for use in decision support in the areas
of planning and training involving the movement of the
crew on board with or without damage situations.
Some results obtained within the scope of the simulations
were not presented in this paper for confidentiality reasons.
ACKNOWLEDGEMENTS
To all team of the Safety at Sea Ltd., Glasgow, UK, for
their support in the simulator adaptation phase, especially to
my instructor Yasmine Hifi for the attention and dedication
in teaching on the use of simulation software.
REFERENCES
Bonabeau, E., "Agent-based Modeling: Methods and Tech-
niques for Simulating Human Systems", Proceedings of
the National Academy of Sciences of the United States
of America (PNAS), v. 99, n.3, pp. 7280-7287, 2002.
CAAML, 1201 - Damage Control Organization, 1 ed. Rio
de Janeiro, Brasil, Centro de Adestramento Almirante
Marques de Leo, Marinha do Brasil, 2005.
CAAML, 1212 - Manual of Survival in the Sea, 1 ed. Rio
de Janeiro, Brasil, Centro de Adestramento Almirante
Marques de Leo, Marinha do Brasil, 2007.
Dijkstra, E. W., "A Note on Two Problems in Connection
with Graphs", In: Numerische Mathematik, v. 1, pp.
269271, 1959.
Fredkin, E., and Toffoli, T., "Conservative logic", In: Inter-
national Journal of Theoretical Physics, v. 21, pp.
219253, 1982.
Galea, E. R., "Simulating Evacuation and Circulation in
Planes, Trains, Buildings and Ships using the EXODUS
Software". In: Proceedings of Conference on Pedestrian
and Evacuation Dynamics, pp. 203226, Duisburg,
Germany, June 2001
Guarin, L., Hifi, Y., and Vassalos, D., "Passenger Ship
Evacuation Design and Verification". HCI Interna-
tional 2014, Creta, Greece, 22-27th June 2014.
Helbing, D., and Molnar, P., "Social Force Model for Pe-
destrian Dynamics", Physical Review, v. 51, n. 5,
42824286, May 1995.
Henderson, L., "The Statistics of Crowd Fluids", Nature, v.
229, pp. 381-383, Fev. 1971.
ISO 13702, Petroleum and Natural Gas Industries - Con-
trol and Mitigation of Fires and Explosions on Offshore
Production Installations - Requirements and guidelines,
Geneva, Switzerland, 1999.
110
Jasionowski, A., and Dodworth, K., PROTEUS 3.1 User
Manual, 1 ed. Glasgow, UK, The Ship Stability Re-
search Centre (SSRC), 2001.
Klupfel, H., Meyer-Knig, M., Wahle, J., Schreckenberg,
M., "Microscopic Simulation of Evacuation Processes
on Passenger Ships", In: Proceedings of the Fourth In-
ternational Conference on Cellular Automata for Re-
search and Industry: Theoretical and Practical Issues
on Cellular Automata, pp. 63-71, Oct. 2000.
Lee, D., Kim, H., Park, J.H., and Park B.J., "The Current
Status and Future Issues in Human Evacuation from
Ships", Safety Science, v. 41, n. 10, pp. 861-876, Dec.
2003.
Lo, S.M., Huang, H.C., Wang, P., and Yuen, K.K., "A Game
Theory Based Exit Selection Model for Evacuation",
Fire Safety Journal, v. 41, n. 5, pp. 364-373, Jul. 2006.
McGrattan, K., McDermott, R., and Weinschenk, C. et al.,
Fire Dynamics Simulator User's Guide, Special Publi-
cation 1019, 6 ed. United States, National Institute of
Standards and Technology (NIST), 2013.
MONOLAX, MONOLAX Simulation Software User
Manual, 1 ed. Glasgow, UK, University of Strathclyde
and Safety at Sea Ltd, 2013.
Tsychkova, E., Influence of Waves and Ship Motions on
Safe Evacuation of Passenger Ships, Licentiate Thesis,
Kungliga Tekniska Hgskolan (KTH), Stockholm,
Sweden, 2000.
US Navy, FXP 4, Mobility (MOB), Logistics (LOG), Fleet
Support Operations(FSO), Noncombat Operations
(NCO), and Explosive Ordnance Disposal (EOD) Exer-
cises, Rev.A, Norfolk, United States, Naval Warfare
Publication, 1993.
Vassalos, D., Kim, H., Christiansen, G., and Majumder, J.,
"A Mesoscopic Model for Passenger Evacuation in a
Virtual Ship-Sea Environment and Performance- Based
Evaluation". Conference on Pedestrian and Evacuation
Dynamics, Duisburg, Germany, June 2001.
Xiaoping, Z., Tingkuan, Z., and Mengting, L., "Modeling
Crowd Evacuation of a Building Based on Seven
Methodological Approaches", Building and Environ-
ment, v. 44, n. 3, pp. 437-445, Mar. 2009.
Yang, L.Z., Zhao, D.L., Li, J., and Fang T.Y., "Simulation
of the Kin Behavior in Building Occupant Evacuation
Based on Cellular Automaton", Building and Environ-
ment, v. 40, n. 3, pp. 411-415, Mar. 2005.
 Naval Auxiliaries, Safe In Design - Safe Through Life
Loren Roberts, Andy Smaller
BMT Cadence Ltd
ABSTRACT
The increase in the number of Naval Auxiliaries
currently being purchased is requiring a change in
approach to safety in design from the commercial
shipyards being contracted to build these ships.
Modern Naval practice requires a more detailed
approach to proving the safe design and subsequent
operation of the ship. The concept of a Wholeship
Safety Case is being more widely utilised throughout
modern navies and where a prescriptive approach does
not exist there is a requirement to conduct a risk-based
approach and to meet Naval Regulations.
This paper aims to introduce the concept of Safe in
Design through the approach taken in the design of
Naval Auxiliaries and how this fulfils the start of the
Safe Through Life concept. The paper will discuss how
a risk based approach can be utilised where, due to the
unique nature of the ship, current Classification Society
Rules do not apply for all areas of the design. The
paper will cover all of the significant hazardous areas:
Stability, Structures, Escape, Evacuation and Rescue,
Propulsion and Manoeuvring Systems, Fire, Aviation
and Explosives, it will also consider how risks to the
platform and environment can be taken into account,
and the benefits of understanding and managing these
areas.
The other key area to ensure a ship is Safe in Design and
Safe Through Life is to consider the interfaces within
the design of the ship. The interfaces between systems
or capabilities of the ship are often not covered by the
prescriptive approach defined by Classification Societies
to the level of detail required by Naval Regulators. A
discussion will be presented as to how interfaces can be
identified and how a risk-based approach can be utilised
to ensure that the interfaces are considered throughout
the design of the ship.
The paper will conclude with a discussion on how the
safe design of the ship can be used to continually
demonstrate that the ship remains Safe Through Life
utilising ongoing management and review.
Acknowledging that this ongoing management is an
overhead on the through life costs of running the ship is
important, given the existing statutory regulation, IMO
and SOLAS, however an argument will be made for the
benefits of adopting this type of approach to the
management of ships in the commercial world.
INTRODUCTION
Naval Auxiliaries vary in type and size but the main
purpose of each Naval Auxiliary is to provide support to
naval shipping, within their fleet. They may also be
required to support any North Atlantic Treaty Organisation
(NATO) ship during joint operations.
A Naval Auxiliary may have to meet a low capability
requirement (e.g. solely to transport fuel), or alternatively
could be a complex ship with:
liquid and solid cargo capacity;
a hospital;
ability to deploy an organic helicopter;
ability to deploy as a Mothership;
ability to carry deck cargo whether this is in
containers or in the form of vehicles or boats.
Whatever the capability or cargo, one key role of any Naval
Auxiliary is to be able to transfer its cargo to another ship.
Although this transfer can take place alongside or whilst
rafted together at anchor or alongside, it is normal practice
to undertake the transfers whilst underway at sea, in an
activity known as Replenishment At Sea (RAS).
Therefore even a relatively simple tanker has many
complex interfaces.
A naval ship, be it a warship or an auxiliary, is required to
follow command decisions to meet a military operational
imperative, even though this may involve deliberately
putting the ship and her crew at increased risk. This
military operational imperative does not apply to
commercial shipping, and this difference can influence the
decision as to what is considered acceptably safe. This
paper will provide a summary of how the As Low As
Reasonably Practicable (ALARP) principle is used in the
context of military operations.
This paper uses examples of complex interfaces to
demonstrate why a Wholeship Safety Case is essential to
ensuring the ship remains safe throughout its lifecycle. A
Naval Auxiliary that cannot undertake a safe RAS whilst
maintaining a steady course and speed has failed in its key
design requirement. Even a Naval Auxiliary whose sole
111
role is a hospital ship will be required to transfer casualties.
Although this may not normally be through a ship-to-ship
transfer, this method of casualty receipt often needs to be a
capability.
This paper will demonstrate how, through the combination
of prescriptive and risk-based assessment, the development
of a Wholeship Safety Case can provide assurance to the
operator that the ship is safe for all its intended purposes.
Where a risk cannot be reduced through design, mitigations
can be implemented through life to manage the risk to a
level that is tolerable or ALARP. BMT Cadence has
demonstrated how this enhances the safety performance and
provides protection to both the ship owner and operator
based on their experience of developing safety cases,
primarily for the United Kingdom (UK) Ministry of
Defence (MoD). The experience gained in the
development of safety cases for the UK Tide Class Tankers
and the Norwegian Logistics Support Vessel HNoMS Maud
will be used throughout this paper. Both of these vessels are
based on the BMT Aegir Design produced by BMT
Defence Services Ltd and constructed by Daewoo
Shipbuilding and Marine Engineering (DSME).
Although not a new concept, the safety case is not a widely
used construct in the marine industry, whether at Wholeship
or Equipment level. Considering the operating
requirements and the level of input into both the design and
through-life management of commercial shipping from
Classification Societies, International and National
legislative bodies, e.g. the International Maritime
Organisation (IMO), the UK Maritime and Coastguard
Agency (MCA) or the United States Coast Guard (USCG),
this lack of a safety case for the majority of commercial
shipping is understandable. A Wholeship Safety Case
however, becomes essential when a ship becomes more
complex through a military capability and requires a
risk-based approach.
THE CONCEPT OF THE WHOLESHIP SAFETY CASE
The History of Safety Cases in the United Kingdom
The concept of a safety case was introduced in the nuclear
power industry as far back as 1965 [1], but it wasnt until
the introduction of the Control of Industrial Major
Accidents Hazards (CIMAH) Regulations [22] (the
implementation of the 1982 EC Directive 82/501/EEC [8])
that the concept of hazard identification and management
began to be used more widely [5]. CIMAH, later
superseded by the Control Of Major Accident Hazards
(COMAH) Regulations 1999 [23], was developed in
response to the Flixborough disaster of 1974 [11].
The CIMAH regulations implemented a safety assessment
regime mainly within the chemical industry to ensure the
safety of their on-shore activities. Following the Piper
Alpha disaster in 1988, in which 167 workers lost their
lives, a public inquiry led by Lord Cullen was undertaken.
Lord Cullens report [5] concluded that the formal safety
assessments conducted for on-shore facilities should be
developed to cover off-shore installations. He also noted
that these safety assessments were often recorded within a
112
Safety Case and that there was a need to maintain this
Safety Case through life.
Safety Cases were introduced in a naval context for the first
time through a Defence Council Instruction in 1993 [6] and
implemented via regulations detailed in Joint Services
Publication (JSP) 430 Issue 1 [18] in 1996.
In 2006 the UK lost a Nimrod MR2 Aircraft XV230 in
Afghanistan. The loss of this aircraft resulted in an
inquiry, which was conducted by Charles Haddon-Cave QC.
The Nimrod Review [7] highlighted that, although a Safety
Case existed for the aircraft, it was not sufficiently robust
and had failed to identify significant design flaws that, if
rectified, could have potentially avoided the accident. The
recommendations from this review identify the need for a
Safety Case that is succinct, driven by the owner, accessible,
proportionate and easy to understand. The Nimrod
Review has had a significant influence on the development
of both safety regulation and through-life Safety Cases
throughout the UK MoD for all operational platforms.
The safety case for the Nimrod MR2 aircraft was not
developed during the design phase, due to the age of the
platform. This meant that during the development of the
Safety Case, opportunities were not available to influence
the design, and the design information available was legacy
information.
What Is A Safety Case?
Issue 5 of JSP 430 [19] defines a safety case as:
a set of structured arguments, supported by a body of
evidence, that provides a compelling, comprehensible and
valid case that a system is safe to operate and is operated
safely and environmentally soundly for a given application
in a given operating environment.
The body of evidence supporting the Safety Case compiles
all of the evidence that the ship has been designed to
appropriate standards, constructed in accordance with the
design and is maintained and managed safely through life.
Examples of this evidence would be Classification Society
approval for the design, supported by surveys undertaken
during construction and prior to the ship being handed over,
and continuous survey through life with a clearly defined
and implemented maintenance regime.
In this context the concept of safe is subjective and
means freedom from unacceptable or intolerable levels of
harm [19]. It is clear that absolute safety is impossible to
attain and as such UK Case Law employs the ALARP
principle when assessing risk reduction measures
undertaken by Duty Holders or employers. Risks are said
to be ALARP if it can be shown that the cost of reducing
the risk further is grossly disproportionate to the benefit
gained [4]. It should be noted that UK Health and Safety
at Work Act 1974 (HSWA) [2] imposes a duty on
employers to ensure, so far as is reasonably practicable
(SFAIRP), the health, safety and welfare at work of all their
employees. These two concepts are considered to be
equivalent.

As part of a Wholeship Safety Case it is also necessary to
consider equipment and system level safety. From the
design and system architecture there is a logical hierarchy
to the safety documentation as shown in Fig 1. An
Equipment Safety Case provides the body of evidence for
the equipment, this then forms part of the body of evidence
for the System Safety Case, which in turn forms part of the
body of evidence for the Wholeship Safety Case. It is
accepted that not all equipment requires a safety case: for
example, some simple Commercial Off the Shelf
Equipment (COTS) or a simple equipment such as a small
valve. However, safety information would be expected in
a form such as a Material Safety Data Sheet (MSDS), CE
marking or other equivalent marking.
Wholeship
Safety Case
System 1 System 2 System X
Safety Case Safety Case Safety Case
Equipment Equipment Equipment Equipment Equipment
1 Safety 2 Safety 1 Safety 2 Safety X Safety
Case Case case Case Case
Equipment
3 Safety
Information
Fig. 1: Hierarchy of Safety Cases
At various points of the ships lifecycle, a Safety Case
Report is produced which provides a summary of the Safety
Case. These reports usually coincide with milestones in
the ships life, for example at the end of detailed design to
provide assurance that the design is safe; at the start of the
ships in-service phase of the life-cycle to document that
the ship is safe to go to sea; or after a major refit period to
reflect the changes in the ship.
Developing the Body of Evidence
There are several types of evidence that are used to provide
the majority of the body of evidence: Classification Society
Rules, Statutory Regulations such as Safety Of Life At Sea
(SOLAS) and Maritime Pollution (MARPOL), Naval
Authority Certification and, evidence provided by
risk-based hazard identification and risk assessment.
Lloyds Register has defined a specific set of Naval
Rules [13] against which an increasing number of Naval
Auxiliaries and Ships are being designed, constructed and
maintained. These rules provide standards for the design
of naval systems that may not exist on commercial ships,
for example RAS, Nuclear, Biological and Chemical
Defence (NBCD), or that may be used in a different way,
for example the propulsion system. However these rules
only consider the design of the equipment. Within the
Wholeship Safety Case the design of the system would fall
within the certification body of evidence whilst its
operation would remain a gap.
This gap can be filled by using a risk-based approach to the
assessment. A risk-based assessment can take many forms
from a simple desktop hazard identification to a more
complex Hazard and Operability Study (HAZOP). The
approach taken will depend on the complexity of the area
being considered and some examples of different methods
are shown in this paper.
Identifying Hazards and Assessing Risks
There are a range of regulations that are used to ensure the
management of safety within the maritime industry, for
example compliance with Flag State Rules, The ISM
Code [15] and Standards of Training, Certification and
Watchkeeping (STCW) for Seafarers [14]. These policies
cover international navies (depending on manning
arrangements), the offshore oil sector and commercial ship
operation. The areas covered by these regulations provide
methods for the assessment of safety for example:
goal-based regulation, formal safety assessment, functional
safety, health and safety management systems and ship
certification. The aim of these documents is to ensure
hazards are identified and risks are managed for the ship in
all of its operating contexts.
These regulations, as a general rule, consider those areas
that could be classed as the significant hazards or those
hazards that apply to all floating vessels, for example
stability or escape and evacuation. In the maritime
industry, safety is inherently considered during the design
of the ship, for example, by designing a hull that is stable
and will withstand the environmental operating envelope
the personnel working onboard are naturally protected.
These significant hazards, in the main, fall into the category
of basic design requirements and tend to be driven by
prescriptive goal-based requirements. Significant hazards
and the approach to them are covered in more detail later in
this paper.
Hazardous areas or activities that are not covered by these
policies will require a different approach to the
identification of hazards. The risk-based approach
considers how all aspects of a design or operation of a ship
could be hazardous and often interfaces with a significant
hazard. For example, when considering movement around
the ship a fall from height may be identified, the
consequence of which is determined by the surface on
which the falling person lands e.g. a deck, equipment or
overboard into the sea. This has the potential to have a
direct influence on the design in that area. Mitigations
then need to be identified that will help reduce either the
likelihood of occurrence or the consequence of the accident.
To continue with the fall from height example, if this
activity is a frequent event then the designer may consider
the addition of guardrails or safety nets, or it may be more
appropriate to provide fall arrest equipment and connection
points in order to limit the consequence.
When hazards are identified, the risk to personnel needs to
be considered, what level of injury they are likely to suffer
and the frequency of occurrence of this injury. These
assessments can be undertaken using either quantitative or
qualitative methods.
113

Reporting The Findings
In order to provide a body of evidence the design,
construction and through-life management activities
undertaken must be documented. For the design, this
evidence is in the form of approved drawings and
supporting documentation, approval and certification.
Any hazards identified as part of the design also need to be
documented. These are recorded in a hazard log. This
records a description of the hazard, the risk assessment, the
mitigations identified to reduce the hazard to a level that is
ALARP, and the person(s) responsible for managing the
hazard and implementing the mitigations. A Safety Case
Report is used to provide a summary of the body of
evidence.
SIGNIFICANT HAZARDOUS AREAS
What Might Be Considered A Significant Hazardous Area?
Significant hazard areas are those areas of ship design or
operation that are assessed as presenting the greatest risk to
operators, maintainers or others.
The UK MoD Naval Authority Regulations (JSP 430 Part
3) [19] define a significant failure as a defect of a system
that can cause, or has the potential to cause, an adverse
effect on a particular defined capability of a ship. Hazards
that could lead to a significant failure must be examined by
the Naval Authority, through a submission prepared by the
MoD Platform Team. Following their audit, the Naval
Authority will issue a Certificate of Safety.
JSP 430 Part 3 defines the following hazardous areas, each
covered by their own Certificate of Safety:
Structure, Buoyancy and Stability;
Escape, Evacuation and Rescue;
Propulsion, Manoeuvring & Navigation Systems;
Fire Detection & Fighting;
Aviation
Explosives
These areas are also broadly reflected in the structure of
ANEP-77, the NATO Naval Ship Code [21].
How Are Significant Hazards Designed Out?
A Prescriptive Approach
Some significant hazardous areas are addressed through
prescriptive means for both commercial and naval shipping.
Structural strength, buoyancy and stability requirements are
well established rules with which the ship designer must
comply. Similarly for escape and evacuation, SOLAS
provides prescriptive rules about the quantity and nature of
escape routes and the provision of life saving equipment,
based on the role and crew size of the ship.
Where a naval ship is designed, built and maintained
through life to applicable Class, Statutory and Flag State
rules, the Naval Authority seeks evidence that these and
114
any additional Naval Authority requirements have been
complied with. This task is often delegated to a
Recognised Organisation such as a Classification Society.
A Risk-based Approach
Other Naval Authority areas also demand compliance with
Classification, Statutory and Flag State rules. However,
these rules are generally based around a commercial ship
such as a tanker or ferry, and are not necessarily sufficient
for a military application.
For example, in the case of bad weather a cargo ship might
navigate around the storm, or seek shelter in port, whereas a
military ship is required to continue to carry out its duty.
Furthermore, a military ship is expected to be able to
sustain a certain amount of damage without significant
degradation of its key capabilities - for example, by
providing higher levels of redundancy.
As such, the Naval Authority applies a risk-based approach
to meet a wider range of capabilities than is expected for
commercially-operated ships. In order to take account of
the military operational imperative, it expects to see a
Safety Case arguing that the risks have been managed to a
level that is ALARP.
Some Classification Societies have begun to develop
specific rules for military applications, for example Lloyds
Register Naval Ship Rules [13]. These rules are
harmonised across a number of Classification Societies
through the Naval Ship Code [21].
HOW CAN INTERFACES BE MANAGED SAFELY
Methods for Identifying Interfaces
Any Naval Auxiliary design begins with a set of User
Requirements. This is essentially a list of ideal
capabilities the ship would have. These User
Requirements are usually produced once a capability gap
has been identified. Simplistically the User Requirement
for a Naval Auxiliary might be to support other naval
vessels, over time, in both national and international waters.
This User Requirement would then be developed into a set
of System Requirements that would be refined to ensure
that they can be achieved within the budget and would form
the basis of a contract specification. An article written for
the Canadian American Strategic Review [3] describes how
the Norwegian Forsvaret Logistikkorganisasjon, Norways
Defence Logistics Organisation, requested tenders for a
support vessel for the Sjforsvaret, the Norwegian Navy.
The original System Requirement included a
Roll-On/Roll-Off capacity, at sea replenishment, cargo
handling and modular accommodation, but no design could
achieve all of these capabilities for an affordable budget.
After reviewing these requirements, the Defence Logistics
Organisation were able to identify a set of System
Requirements that were developed into Design
Requirements and have resulted in a contract being placed
between Forsvaret, the Ministry of National Defence in the
Government of Norway, and DSME for the design and
production of the first purpose built Logistics Support
Vessel (LSV) for the Norwegian Navy.
Once a set of System Requirements have been identified,
these are developed into Design Requirements. Design
Requirements are then used to formulate a design solution.
Once the Design Requirements have been identified it
becomes possible to begin to identify interfaces that require
safe management. It we once again take the example of
RAS we can begin to identify where interfaces may occur.
For example, to conduct a RAS safely, a steady speed is
required. The production of a steady speed is through the
Propulsion System resulting in a direct interface between
these two systems. Examination of the Propulsion System
reveals that it is supported by a number of auxiliary systems,
resulting in another interface.
Therefore, if we are to consider the movement of liquid
cargo from one ship to another whilst underway there are a
number of interfaces with other systems in the ship. Fig 2
shows an example of the various interfaces within the ship.
Platform
Cargo Pump
Management
System
Electrical System
Other Machinery
Systems
RAS RIG
PROPULSION RAS
SYSTEM OPERATION
NAVIGATION
SYSTEM
Electrical System COMMUNICATION
SYSTEM
Fig. 2: Example of Interfaces That May Be Associated
With RAS Operations
As is demonstrated by developing the interfaces for RAS
operations, it can quickly become a complex network of
interfaces. There are also opportunities for some systems
to interface with more than one system. For example, the
electrical system is required in order to operate the RAS rig,
the propulsion system, the navigation systems and
communications. As Fig 2 illustrates there are also
interfaces external to the ship and its systems, in this
example the Receiving Ship. These interfaces also need to
be considered to establish if they fall within the remit of the
design or are excluded. For example in a RAS operation
the delivering ship is responsible for the RAS rig and hoses,
but the responsibility for connecting this to the receiving
ship falls within the Safety Case for the receiving ship.
Ensuring the ships maintain an appropriate steady course
speed, and separation is part of the Navy-wide operating
procedures.
If this identification of interfaces is undertaken for each
design requirement, a complex network of interfaces is
quickly developed representing the Design Requirements.
Although the outcome of the review of the Design
Requirements appears complicated, the identification
process is a relatively simple exercise. However, this
approach is not without risk. A simple review of the
Design Requirements will not identify the interfaces within
the design solution, but only those between the
requirements. A full review of the design should be
undertaken to identify the interfaces, but this would require
a certain level of maturity of the design. Beginning the
safety review when the design is at this maturity level
would make it difficult to influence the design from a safety
perspective so it is essential to identify the interfaces as
early as possible in the design phase. This can be
achieved by developing the interfaces within the design
requirements and then conducting a validation exercise.
System
This validation exercise should involve the design team as
they will be able to provide appropriate evidence for the use
of interfaces and identify the detailed and specific
interfaces associated with the design solution. This will
also help to identify any areas where design requirements
Other Mechanical
Systems
have been changed or are being met through alternative
design options. The output of this validation process
should not only confirm the interfaces to be considered but
any items excluded from the design, for example the
Receiving Ships RAS equipment.
Hydraulic System
Once the interfaces have been identified, these can then be
used as part of the through-life safety management. If a
change is made to a system during the life of the ship, for
RECEIVING example a new propulsion system, the interfaces can be
SHIP
reviewed to ensure that all aspects of the ship are
considered.
Documenting the Interfaces
Once the interfaces have been identified these should be
documented. It is essential to have this documentation to
ensure that the interfaces are not forgotten or lost during the
design phase. It is also necessary to provide an auditable
document trail for the decision process used as these
interfaces will form the basis of the safety assessment
process. This document trail will also form part of the
body of evidence for the safety case.
The documentation should identify the interfaces, the
boundary of the safety case activities and any assumptions
that have been made. The boundaries of the safety case
will identify any areas excluded from the safety assessment,
such as the Receiving Ship RAS equipment in our example.
Any assumptions that have been identified must also be
recorded for the body of evidence of the safety case. These
assumptions may impose requirements on obtaining further
information, for example that all equipment will be
115
delivered with a safety case, or assumptions about the
operation of the ship, e.g. that RAS will occur at set speeds.

There are many ways in which the interfaces can be

recorded and the chosen method may be strongly
influenced by the complexity of the design and the number
of interfaces. BMT Cadence has developed an approach
to the recording of interfaces and boundaries. An example
of this is shown diagrammatically in Appendix 1.

The boundaries and assumptions diagram was initially

developed for the safety assessment developed as part of
the basic design of the LSV for the Norwegian Navy. The
diagram was used to define the LSV, its capability and
interfaces, as well as exclusions not covered by the hazard
identification up to the preliminary design review at the end
of basic design. The boundaries and assumptions
associated with the safety (including fire safety) and
environmental aspects of the LSV design were defined
during a meeting held with the BMT Defence Services
Design Area Leaders (DAL).

The elements captured define the attributes of the basic

design that were covered under the safety and
environmental management remits. The definitions of the
components in the diagram are as follows:

Platform - the requirements and specifications for

the LSV to fulfil the basic capability of a ship i.e. Fig 3a: Example of Interconnecting Nodes
to accommodate the crew, float and move;
Inter-operability - the requirements and Fixed Line
specifications of the LSV to perform the tasks that
are of value to the Norwegian Defence Logistics
Organisation (the reasons that the ship is being Ship to Air
built); Military
Interfaces - the parts of the design that have
important interactions, in terms of safety and
environmental risks, that need to be considered; (a) COMMS
Exclusions - or boundaries are areas not covered
by the safety and environmental management remit
as they are either:
Civil Ship to Ship
outside of the designer's control at basic
design;
not relevant at this stage of the design
process; or Electrical
are operational aspects of the LSV. (b)

The boundaries diagram visually presents the links and
interfaces between systems. However, for a whole ship
this is a complex diagram with many links between systems.
It can be seen that a number of nodes require linking, but
the complexity of the diagram produced cross-connections
that would cut through other nodes, making the diagram
hard to interpret on paper. To avoid this confusion, links
have been made using nodules labelled with the system
name and a unique identifier as shown in Fig 3a and 3b.
These diagrams are a subset of the Wholeship nodes
presented in Appendix 1.
Fig 3b: Example of Interconnected Node
Appendix 1 shows examples of those areas of the design
that are covered by classification or statutory certification.
This colour-coding demonstrates how much of the ship
design falls within the requirements of certification and
clearly demonstrates that the majority of risk-based
activities are required for the operability of the ship.
Each node on the diagram was identified as a principal
hazard area. These principal hazards formed the basis for
the hazard identification process and were key to
identifying where an interface crossed a system design
boundary. When hazards or accidents were identified that
had a direct influence on another principal hazard area,

116
these were reviewed in tandem. By doing this, we could
ensure that the mitigating actions could be implemented
effectively and proportionally to reduce the risks associated
with both principal hazard areas.
DEVELOPING HAZARDS
Methods for Identifying Hazards
A hazard is defined in JSP 430 [19] as potential to cause
harm e.g. a physical situation or state of a system, often
following from some initiating event, that may lead to an
accident.
In contrast, an accident is defined as an event, or sequence
of events, that causes unintended harm.
There are many techniques to identify hazards, of varying
degrees of rigour. A formal HAZOP is an example of a
rigorous technique. Developed for the chemical process
industry, it follows a methodical sequence to examine every
node of an operation in turn, and consider all the ways in
which the process could operate abnormally at that node.
It uses guide words such as More of, Less of, None
of, Reverse and others to comprehensively analyse how
a system could fail and what the consequences would be.
Ships do not lend themselves to such a rigorous technique.
The number of interacting systems to take into account,
along with external factors (e.g. weather, other ships and
human operators) would be unmanageable. An alternative
is to use Structured What-If Technique (SWIFT) - a
systematic team-oriented technique for hazard identification.
It considers deviations from normal operations identified by
brainstorming, with questions beginning What if? or
How could?. The brainstorming may be supported by
checklists to help avoid overlooking hazards. SWIFT is
dependent upon the team of Suitably Qualified and
Experienced Personnel (SQEP) to bring to the table their
experience of the system under examination, or similar
systems. The meeting facilitator is responsible for
ensuring that everyone has an opportunity to contribute.
Conducting Risk Assessments
Having identified hazards it is then necessary to assess the
associated risk. This provides a means to rank and
prioritise the hazards. This is done by assessing against
two criteria.
The first is severity: supposing that an accident occurs as
a result of the hazard, what is the impact? As well as the
obvious human impact (injury or death), this may be an
impact on the environment (e.g. leakage) or damage to the
ship or platform itself. By considering the platform and
environment at this point in the assessment, changes can be
made to the design and operation of the platform that have
the benefit of reducing these risks thereby improving the
service life and the environmental impact of the ship.
The second is likelihood: What are the chances, given that
the hazard exists, that it will lead to the identified accident
occurring? This may be measured in terms of likelihood per
year, or per ship lifetime, or across the fleet. Likelihood
may be quantified using a Fault Tree or Event Tree or
similar technique, or may be estimated by review of failure
rates or historical evidence, or based on the experience of
those in the room.
Director Ships within MoD Defence Equipment and
Support has established a common Risk Classification
Matrix [17] against which all MoD shipping risks are
assessed. This allows comparisons and aggregations of
risk to be measured across the fleet and this matrix was
used on the UK MoD Tide Class design.
Following risk assessment, the matrix is used to combine
the likelihood and impact to produce a Risk Classification
of A, B, C or D.
The four Risk Classifications are defined as shown in Table
1 below. The military are prepared to accept risks at a
higher level than would be deemed appropriate for a
commercial ship due to the military operational imperative,
but it is still necessary to demonstrate that all reasonably
practicable measures have been taken.
For example, on the Tide Class tankers, the risk of collision
during an abeam RAS is a Class B risk which has been
endorsed as ALARP. This reflects the inherent residual
danger in such an operation, despite the fact that all
reasonable safety measures have been identified and put in
place. The military imperative means that the ship must
deliberately put itself in harms way to meet its mission.
Class Description
A Unacceptable. Risk cannot be justified
except in exceptional circumstances.
B Tolerable only if further risk reduction is
impracticable and if the penalties are grossly
disproportionate to the improvement gained
(ALARP).
C Tolerable if the penalties are grossly
disproportionate to the improvements gained
(ALARP)
D Broadly Acceptable. Must still be ALARP,
but no need for detailed assessment.
Necessary to ensure the risk remains at this
level.
Table 1: Risk Classifications
Developing Mitigations
Of course, the purpose of identifying hazards is not simply
to produce a list of them. The aim is to manage the risk
posed by the hazard to acceptable levels.
In general, it is not possible to eliminate all risk associated
with a particular equipment, system or operation. All
activities inherently carry risk. However, UK Law states
that we have a duty to ensure that the level of risk posed
ALARP. Risk can be minimised by:
Reduce the Likelihood: Even though the
consequences may remain severe, you reduce the
chance of the accident occurring.
117
 Reduce the Impact: This focusses on the outcome
and tries to make it less severe.
Mitigations can be grouped into various types, as shown in
Table 2 below. Those near the top of the list are harder,
and so typically more costly, to achieve, but are more
effective at reducing the risk.
Mitigation Description
Elimination Redesign the equipment/system or
change the task such that the
hazardous circumstance does not arise.
Engineering Provide additional physical measures
Control to reduce the risk - e.g. guard rails, fall
arrestors, ventilation
Procedural Provide instructions/procedures to
Control carry out the task safely. E.g. Job
Instruction Cards (JICs), Standard
Operating Procedures (SOPs),
operator/maintainer manuals, safety
signage
Personal If the risk cannot be reduced by any
Protective other method, ensure the person at risk
Equipment (PPE) is suitably protected - e.g. hard hats,
goggles, boots, etc.
Table 2: Hierarchy of Mitigation
For real-world examples of these mitigations, consider the
hazard of working at height on a ship - perhaps on a mast or
funnel. Clearly, the associated accident would be a fall
from height onto the deck below. Examples of the four
types of mitigations are as shown in Table 3 below:
Mitigation Description
Elimination Design the ship such that there is no
reason to go aloft in the first place -
e.g. no equipment requiring
maintenance in that location.
Engineering Provide a guard rail and working
Control platform where the work is to be
undertaken, provide clip point for
harness; design the equipment such
that minimal or infrequent
maintenance is required.
Procedural Procedure for the task setting out
Control method of work and stressing the risk;
Working at Height to be risk-assessed
and a Permit-To-Work granted by a
responsible person, taking into account
sea conditions and other such factors.
Personal Harness, Hard Hat
Protective
Equipment (PPE)
Table 3: Example of hierarchy of mitigation
It is clear from this that the first two are relatively easy and
cost effective to achieve during the design phase, though
they might be difficult or prohibitively expensive to
retro-fit at a later date. Note that these mitigations are
often interlinked - for instance there is no benefit in
providing a harness with nowhere to clip it onto.
118
Reducing Risk
Whether action is taken to reduce the likelihood or the
impact, this will reduce the overall level of risk. The
reduction is generally a qualitative assessment of the
change to likelihood or impact, with the new risk
classification derived from the matrix.
In reality though, the level of risk is a continuum, not
simply one of the four discrete levels, as shown in Fig 3
below. Action that reduces the risk from a High C to a
Low C may still be desirable - even if it does not appear
to affect the Risk Class, the risk has still been reduced.
Intolerable
Region
Class A
Tolerable and
ALARP
Region
Class B or C
Broadly
Acceptable
Class D
Fig 4: The HSE Framework for the Tolerability of Risk
Reasonably Practicable is generally a subjective
judgement. JSP 430 [19] defines ALARP as being the
state where the cost of any further risk reduction (in terms
of money, time or trouble including the loss of defence
capability), is grossly disproportionate to the benefit
obtained from that risk reduction which involves weighing
a risk against the trouble, time and money needed to control
it.
In order to determine more objectively what constitutes
grossly disproportionate, it can be useful to assign a
monetary value to achieving a reduction in the risk of death.
The Health & Safety Executive [9] provides guidance to
industry sectors (including the MoD) in setting a
benchmark Value to Prevent a Fatality (VPF). This is of
particular use in calculating the safety benefit gained from
implementing a mitigating action if a Cost Benefit Analysis
is the method employed to justify that risk is ALARP. The
cost of implementing engineering solutions can be justified
by calculating the incremental reduction in risk achieved
per fatality avoided per single installation where the system
is installed and used over many years. This can help in
making and justifying a decision as to gross disproportion.
Note that VPF is not the value that society, or the courts,
might put on the life of a real person or the compensation
appropriate to its loss.
As it is subjective, an ALARP justification is required that
sets out why those responsible for the hazard consider its
risks to be ALARP, and that justification is to be endorsed
by a properly-constituted Safety Committee.
Reviewing Risks
There is a requirement, periodically through the life of the
ship, to review the hazards and risks identified. The
review should consider whether the hazard or risk has
changed. A change may occur for many reasons, for
example: a change in the lifecycle phase, a change to the
operating envelope, or a change of role.
The hazards should be reviewed to ensure that no changes
have occurred to any of the hazard information, including
the risk assessment, and that the risk remains ALARP.
DEVELOPING A SAFETY CASE
The Purpose Of The Safety Case
As stated previously the purpose of a safety case is to
provide:
a structured argument, supported by a body of evidence,
that provides a compelling, comprehensible and valid case
that a system is safe for a given application in a given
environment.
By fulfilling this purpose a Wholeship Safety Case provides
assurance to stakeholders that the ship is safe for all its
intended operations. The stakeholders for a ship would
include Ships Staff, Regulators whether in the form of
Classification Societies or other certification bodies, and
owners. For the UK MoD, the ownership is managed by
two separate entities or duty holders.
The Platform Duty Holder (PDH) is responsible for the
design and upkeep of the ship whilst the Operating Duty
Holder (ODH) is responsible for the safe operation of the
ship in accordance with its defined operating envelope. In
the commercial industry this ownership regime might be
seen as the equivalent of a ship leasing arrangement
whereby the ship lease company is responsible for a ship
that is of a safe design and construction (the PDH role)
whilst the lessee is responsible for the safe operation of the
ship (the ODH role).
However, the purpose of a safety case is not only an
assurance role. A comprehensive safety case can also
provide a legal defence that best efforts have been made to
reduce the risk to personnel associated with the operation of
the equipment or ship. It can be seen from the Nimrod
Review [7] that a safety case that is not comprehensive is
not capable of supporting a legal defence and can leave
governments, companies and individuals open to blame.
The Body of Evidence
The body of evidence that forms the basis of any safety
case should encompass all relevant evidence that supports
the argument that the ship is safe in design and safe to
operate. This section aims to describe some of the
evidence that is expected, but this list is not exhaustive and
the evidence should be identified for each individual safety
case. The body of evidence should be maintained through
the life of the ship as information changes.
Design Information
When considering whether a design is safe, the design
information is an essential piece of evidence. This
information would include all drawings, calculations and
reports that provide the details of the design. It should
also include all design approval whether this is internal to
the design, for example checking of calculations, or
external from a Classification Society or other approval
body. The design of naval ships to Classification Society
rules provides a cost-effective solution to meet capability
requirements.
The intent of providing this evidence is twofold: to
demonstrate the underlying integrity of the design; and, by
proving the design meets the design criteria and design
requirements, demonstrating that the ship has been safely
designed.
Certification - Classification and Statutory
Classification and Statutory certification provides a key
aspect of assuring safety both in design and through life.
If we take the new UK Tide Class Tankers currently under
construction for the UK MoD as an example they have
three different types of certification. The ships are being
built to Lloyds Rules as defined in the Technical
Specification [20] and will be delivered with Lloyds
Certification covering areas of Stability, Structures, and
propulsion systems. This certification will be maintained
through life. The ships are also required to meet Statutory
Certification including SOLAS and MARPOL. The ships
also comply with aspects of the Naval Ship Code [21], as
tailored by the Technical Specification [20]. Finally the UK
MoD have their own certification body called the Naval
Authority Group (NAG). This body is independent of
both the Duty Holder and the operator and provides
certification that the ship meets the requirements laid down
in JSP 430 [19]. This MoD certification body was
developed to provide assurance that MoD shipping is
designed and maintained safely and is therefore safe to
operate.
Certification may be granted to a ship from any of these
certification bodies with caveats or limitations. These,
along with the proposed solutions and timeframes should be
recorded as part of the body of evidence. Such evidence
should be referenced in the Safety Case Report.
All of this certification is required to be maintained through
life. Classification Society and Statutory certification
have clearly defined survey and certification regimes and
119
the MoD Naval Authority certification has similar
stipulations. For example, if a ship is docked, the Naval
Authority certification is suspended and evidence has to be
re-submitted to provide assurance that either the changes
made to the ship have no impact on the scope of
certification, or that changes meet the appropriate
requirements prior to certification being re-instated.
Classification Societies maintain certification through a
regime of surveys, the Lloyds Register surveys for Naval
ships are summarised in their document Naval Ship Safety
Assurance Guidance for Navies and Shipbuilders [10].
The intent of certification as part of the body of evidence is
to demonstrate that the design and through-life management
of the ship is independently reviewed and certified.
Technical Studies
Technical studies can provide essential evidence to the
safety case. These studies may concentrate on a specific
design solution, such as a study detailing the benefits of
new or novel equipment and the reasons for selecting it as
part of the design. Alternatively, they might contain
details of how and why technical details have been tailored,.
For instance, a naval auxiliary may tailor the SOLAS Safe
Return To Port requirements to ensure the ship could
undertake a safe return to port, whilst remaining within the
physical and budgetary restrictions placed on the design.
A technical study may also be used to provide evidence of a
mitigation against a hazard, whether this is a design
justification or the results of a Cost Benefit Analysis.
This type of document is essential to a safety case body of
evidence as it often provides information on the design or
decision making process that would otherwise be a
potential gap in evidence.
Hazard Log
The hazard log is a vital part of the body of evidence. The
hazard log records the hazards that have been identified and
their associated risk assessment. It also contains details of
the mitigations that have been identified which, during the
design process, should signpost design changes.
As well as the hazard log recording the hazard and its
associated risk assessment it also records the person
responsible for managing the hazard. Once a hazard has
been detailed it is then possible to start identifying actions
to mitigate the hazard. These actions should mitigate or
reduce one or both of the likelihood or consequence of the
risk and will also need to be allocated to the individual who
has the budget to complete the action. As discussed
previously it is possible at the design stage for a mitigating
action to design out (eliminate) the hazard.
The hazard log provides the currently recorded status of
each hazard, the history of the hazard and any future actions
that are required to be completed at a later date, for
example, outfitting the ship under construction with
lifesaving appliances. The hazard log is also the format
for recording the ALARP statement for each hazard.
120
It should be noted that a comprehensive hazard log will
record closed or non-credible hazards along with
justification for the decision that these hazards are not
pertinent to the design.
The hazard log is a live document that is used, maintained
and reviewed throughout the life of the ship.
Training Records
The competence and currency of competence of individuals
is a significant mitigation to the majority of hazards that
cannot be designed out. For example, if a fire occurs
onboard a naval ship, the competence of the compartment
re-entry team is as important as the firefighting equipment.
If a re-entry team is not appropriately trained then
additional lives may be put at risk whilst attempting to fight
the fire in a compartment.
Therefore, a significant part of the body of evidence is
demonstrated through the provision of relevant training and
the training records of the ships crew. Competence of the
crew in terms of their RFA Seamanship training,
STCW [14] and specific role/equipment training is critical,
both in the safe operation and maintenance of the ship and
its systems, and in an effective response to emergency
situations. These training records will demonstrate the
qualification, experience and therefore competence of the
crew and should be maintained throughout the life of the
ship.
The approval of a design by a Classification Society
provides evidence that the design team is SQEP to produce
the design. The competence of the ship builder is assessed
through the Classification Societys survey and approval of
the build process and as-built ship. UK MoD also
undertake shipyard audits and inspections to provide
assurance that the yard is capable of delivering the ship
safely.
The competence of the designers and builders is also a key
part of the argument. The Safety Case must demonstrate
that those involved in the design and construction of the
ship are competent to do so.
Maintenance Procedures
A significant aspect of maintaining the safety of the ship
and its equipment is to ensure that maintenance is
conducted in accordance with the maintenance regime
required by the equipment manufacturer, Classification
Society and/or ship owner. It is therefore essential that the
required maintenance is documented through appropriate
procedures and that a schedule has been developed for the
maintenance. The body of evidence for the safety case
should consequently include details of the maintenance
procedures
Maintenance support may be required from Original
Equipment Manufacturers (OEMs) or other shore-based
facilities. Although these maintenance items will be
identified as part of the ships maintenance schedule, the
responsibility for the activities being conducted safely falls
outside the boundary of the ships Safety Case, and rests
with the organisation in question. Functional
Claim Area Claim
Operating Procedures

The defined operation of the ship is fundamental to the

safety case as this operational envelope defines the basis of
the boundaries of the safety assessment. To ensure that
the ship is operated within this envelope, operating Arguments are
procedures must be defined. Argument made to justify the
claim
The operating procedures form a key part of the body of
evidence as they detail how the ship should be operated to
remain within the operational envelope. Evidence is provided
to support claims
For a new-build ship, the development of operating Evidence and arguments
procedures significantly influences the competency of the
crew, and may identify specific training requirements. For
example, the Bosch Rexroth abeam RAS stations installed
on the Tide Class Tankers have not been used within the
Fig 4: An Example CAE Diagram.
UK MoD before. This has resulted in the development of
new operating procedures and training courses to support
Identifying Forward Actions
the safe operation of the equipment.
A key role of the SCR is to provide details of the forward
The Safety Case Report
actions that are outstanding. These forward actions may
be mitigating actions that have not been completed or other
The Safety Case Report (SCR) provides a snapshot in time
outstanding activities that may be required depending on
of the status of the safety case and its body of evidence.
where the ship is in its lifecycle.
The SCR records the status of the safety case at a particular
time within the ships lifecycle and is only valid and current
Forward actions should identify what ongoing management
until changes occur to the body of evidence.
is required to complete the action, the owner or person
responsible for completing the action and the due date by
The Safety Argument
which the action is required to be competed.
The body of evidence that supports the safety case is
Independent Review
reported within the SCR as the safety argument. The
safety argument aims to provide a clear and structured
There are two types of independent review - those
argument that the ship is safe for its intended operation.
conducted by Classification Societies, and for the UK MoD
those conducted by an Independent Safety Auditor (ISA).
There are several ways of structuring the safety argument
The role of the Classification Societies in providing
and the structure is often defined by the complexity of the
independent technical assurance is well understood, so this
system. The safety argument for a simple system may be
section considers the benefit of an ISA review of the
addressed entirely through compliance with standards,
risk-based aspects of the Safety Case.
regulations and rules, whereas a complex system such as a
whole ship may require a more structured approach for
The production of a SCR is an opportunity for an
example through the use of Goal Structuring Notation
independent audit of the whole safety assessment process.
(GSN) or Claims Argument Evidence (CAE).
The UK MoD requires all SCRs to go through an
independent review to ensure that the process as described
BMT Cadence has used a CAE approach for the
in JSP 430 [19] has been appropriately followed.
newly-designed Naval Auxiliaries with which it has been
involved. This approach uses a set of claims and
The Nimrod Review [7] details the independent review
sub-claims against which arguments are developed that the
process that was followed and the inadequacies of the
available evidence supports the claim. Fig 4 provides an
process used. This report highlighted that part of the
example of how a CAE diagram is constructed. During
problem was a lack of distinction between the roles of
the design phase a traffic light system was used to
technical assurance and independent audit. It is crucial to
colour-code the CAE diagrams. This traffic light system
understand these differences when contracting for either
is a simple method of identifying where arguments are not
role as part of a ship build project. The Nimrod Review
fully defined or evidence has not been completed.
identified that there is a need for independent review to
Appendix 2 shows how the CAE diagram can be developed
consider both the process and the process output for legacy
and the traffic light system utilised.
equipment.

121
To maintain appropriate independence, an ISA should not
have detailed experience of the ship or system under review,
and therefore audit of the process output will include
validation that evidence exists to support the completed
mitigations identified for each action.
Independent Technical Assurance as undertaken by a
Classification Society, however, will require experience of
similar ship or system design.
References to the independent audit are a significant part of
the body of evidence, since they provide assurance that,
through following the published process, all aspects of ship
safety have been appropriately considered.
Managing the Safety Case With A Safety Management
System
The MOD uses a six-phase acquisition cycle knows as
CADMID to manage the acquisition of defence
equipment [16]. The six stages are:
Concept
Assessment
Demonstration
Manufacture
In-Service
Disposal
The focus of the safety management activity is different at
each phase of the CADMID cycle. During the Concept
and Assessment phase, the focus is on using high-level
hazard identification to derive safety requirements that will
result in a tolerably safe design from the outset. This is a
critical part of design for safety - a safety programme
initiated later in the cycle will have less opportunity to
influence the design.
In the Demonstration phase (for a ship, essentially the Basic
and Detailed Design stages), the hazards may be refined, as
the design emerges further hazards will arise, and again
there is an opportunity to design out (eliminate) hazards
whilst the ship exists only on paper.
For example, during the design of the Tide Class tankers, a
hazard was identified related to the Flight Deck Officers
visibility of the flight deck. At the design stage, it was
reasonably straightforward to include a build-out in the
FDOs compartment to enable 180 visibility of the flight
deck. As a later modification, this would have been
expensive and time-consuming to implement.
A key focus of the Tide Class Safety Case has been to
ensure that such opportunities are identified, recorded and
implemented.
During Manufacture, the focus is on ensuring that the
safety-related design features identified have been reflected
into the as-built product.
Once the ship is accepted into service, responsibility for
safety transfers from the ship designer and builder and
acquisition Project Team (Platform Duty Holder) to the
122
operator (Operating Duty Holder). The safety case should
be maintained and updated to take into account any changes
to the ships operational profile or material state over its
service life.
At Disposal, the safety case serves to provide assurance of
safety, and awareness of the locations of any hazardous
materials, to a new purchaser, dismantler or recycler.
For the Tide Class Tanker, the first ship has been accepted
off contract, and is undergoing customization for the full
military capability prior to entry into service. As such,
responsibility for the safety case has transferred from the
shipbuilder to the customisation contractor, who will update
the hazard log and safety case to reflect the changed
material state of the ship.
In order to ensure this process is followed, and that there
are no gaps, a Safety Management System (SMS) is
required. This is usually documented within a Safety and
Environmental Management Plan (SEMP), which
developed in the early stages of the project. It sets out:
An outline of the equipment/system/ship;
Agreed safety requirements, targets;
Organisation & responsibilities - MOD is a
complex chain of authorities, with different
customers and users depending upon the phase
in the CADMID cycle;
Structure of the Safety Case;
Timescales for production of documentation;
Approval arrangements;
Forward actions
The Safety Case Report Timeline
The SCR is only valid at the time of production and
therefore, over the life of the ship, an SCR should be
produced at key milestones in the ships life, for example:
At the end of the design phase to demonstrate that the
design is safe;
At the end of construction to demonstrate that the ship
has been constructed in accordance with the design and
to ensure it is safe to operate;
At appropriate times through the in-service phase of
the lifecycle, for example after a change in role or
major refit;
At the disposal point of the ship to provide assurance
that the ship is safe to dispose of whether through sale
or dismantling.
MANAGING SAFETY THROUGH LIFE
The safety case is developed during the design phase of the
ships lifecycle and is developed through the construction
phase providing assurance that the design and construction
of the ship is safe. The point at which the ship enters
service is a significant milestone as this is when the
assumptions and hazards are tested, and is the point at
which operational hazards may manifest themselves.
Ongoing management of the safety case becomes an
important aspect of the through-life management of the ship.
The SMS will identify the through-life milestones at which
the safety case requires updating to ensure the SCR remains
live and relevant to the ship.
When the ship enters service the focus of the safety case
changes from the design and construction aspects to the
operational aspects of the ship. Hazards will still be
identified either through the ongoing hazard identification
process or due to an incident, accident or near miss
occurring. These details will need to be recorded and
reported in the SCR at the next update of the document.
During the design and construction phase of the ship,
consideration should be given to whether there is any future
legislation that may impact the ship. For example, the
Marine Pollution (MARPOL) Regulations [10] requiring all
tankers came into force via a phase-in schedule whilst the
regulation was defined in 1993. Therefore any tanker
designed after this regulation came into force should have
considered the impact of meeting this regulation and
equally the impact of not meeting this regulation.
Therefore, when the safety case is reviewed through life,
consideration should be given to the impact of current and
future legislation whether this is national or international.
For example, the Tide Class tankers have been designed
such that they are compliant with MARPOL Tier 2 Nitrous
Oxide (NOx) emissions upon entry into service, but also
such that they may be modified to comply with Tier 3 in the
future with minimal additional work.
When the UK MoD first implemented safety cases, these
were produced for legacy ships and equipment. This
required significant effort to be made to identify the
baseline status for documenting in the SCR. However,
more recently designed and built ships have had safety
cases produced at the design stage resulting in significant
through-life cost savings for the management and update of
this information.
It is acknowledged that during normal operation of a ship,
whether it be military or commercial, there is no direct
financial benefit to producing the body of evidence required
to support a Safety Case. However, in the event of an
incident or accident, the burden of proof that all hazards
were considered remains with the owner of the ship. In
this instance, the benefit of this already-prepared body of
evidence becomes apparent. It can also be used to analyse
the impact of any change, and to make sure that safety
margins are not eroded.
CONCLUSIONS
This paper has focused on the development of a safety case
in line with current best practice for Naval Auxiliaries, and
has, where possible, drawn comparisons with the
commercial shipping industry. It is accepted by the
authors that the majority of commercial shipping is not
complex, unique or novel in its design and so meeting
statutory regulations is sufficient to provide assurance to
the ship owner or operator that the ship is tolerably safe.
As detailed in the Naval Ship Safety Assurance
Guidance [12] the use of Classification Society Rules has
been widespread within the commercial industry for over a
century and building naval ships to these commercial
standards is becoming more common. The development
of Classification Society Rules for naval ships, for example
Lloyds Register Rules and Regulations for the
Classification of Naval Ships [13] has eased the transition
from regulations developed specifically by navies to
Classification Society Regulation.
The use of a commercial design for naval auxiliaries, such
as BMTs Aegir, allows governments to purchase their
ships through commercial shipyards which are experienced
in delivering such ships quickly and efficiently.
There are differences between Naval Auxiliaries and
commercial shipping that can be easily defined, for
example, both the Tide Class and the Norwegian LSV have
a number of roles that would not be combined in the
commercial world. Where the LSV is one ship that
provides a replenishment capability, a container carrying
role, a solid cargo role, an aviation capability, a hospital
and mothership role, these would likely all be separate
ships in the commercial industry and some of these roles
may not even be required in industry. These combinations
of roles are one of the key reasons that naval auxiliaries
require well defined and maintained safety cases to ensure
that areas and interfaces not covered by existing
commercial standards are fully addressed.
For specific significant hazardous areas, the Safety Case is
examined by the relevant Naval Authority to support the
issue of a Certificate of Safety for the ship. This ensures
that the most significant hazards are subject to rigorous
review.
In order to ensure that all areas of the ship design and
operation are assessed for potential hazards it is essential to
identify the boundaries of the ship operation and the
interfaces between the systems onboard. This activity
allows those areas of the design that are not covered by
prescriptive regulation to be identified as principal hazard
areas and managed through a risk-based process.
This paper has detailed the significant level of additional
information that is required to produce and manage a safety
case through life, and the cost benefit may only be seen in
the event of an accident. However, with the growth of a
litigious society the need for a cost effective and
demonstrable baseline providing evidence that all potential
hazards and risks have been considered by a ship owner or
operator may increase and the production of a safety case,
summarized in an SCR, would help provide this evidence.
The operating ethos is different for naval auxiliaries than
commercial shipping, because of the military operational
imperative which means that naval auxiliaries must often
put themselves in harms way. For example, a
commercial ship will plot a course around a severe storm,
or seek shelter in a port, whereas the naval auxiliary may be
123
commanded to a certain location regardless of the weather
conditions.
The military operational imperative also affects the
response to emergency situations. It is important for naval
shipping to recover from an incident and continue their
mission, part of the Float, Fight, Move philosophy.
However, despite all of these differences there are
justifications for the development of a safety case in
commercial shipping. There are vessels or platforms in
the commercial industry that are complex in design and
operation as they have specific roles or alternatively they
have a significant role with passengers.
Some areas of the commercial world such as cruise ships or
LNG carriers are already utilizing a risk-based approach to
safety management. In this case the body of evidence for
the safety case already exists and it would only be a small
step to document this process.
ACKNOWLEDGEMENTS
amendments);
[16] MoD Defence Equipment & Support - Acquisition System
Guidance (ASG), June 2016;
[17] MoD Implementation of Director Ships Common Risk
Classification Matrix, Version 1, June 2009;
[18] MoD Joint Services Publication (JSP) 430, Ship Safety
Management System Handbook, Volume 1, Issue 1,
January 1996 (not publicly available);
[19] MoD Joint Services Publication (JSP) 430, Management of
Ship Safety and Environmental Protection;
[20] MoD MARS Tanker Schedule 3 Technical Specification
(redacted version from
https://data.gov.uk/data/contracts-finder-archive/contract/4
65670);
[21] NATO Naval Ship Code ANEP-77, Edition F Version 1,
August 2014;
[22] SI1984:1902, Control of Industrial Major Accidents
Hazards (CIMAH) Regulations, 1984;
[23] SI1999:743, Control of Major Accidents Hazards
(COMAH) Regulations, 1999;
[24] SI2013:1471, Reporting of Injuries, Diseases and
Dangerous Occurrences Regulations 2013.

The authors are grateful to UK MoD Commercially

Supported Shipping for their permission to use the Tide
Class Tankers as an illustration in this paper.

REFERENCES

[1] Act of Parliament: 1965 C. 57, Nuclear Installations Act

1965;
[2] Act of Parliament: 1974 C. 37, Health and Safety at Work
etc. Act, 1974;
[3] CASR, A Joint Support Ship Comparison - An AOR
Norwegian Style, Joint Support Ship (JSS) Project News
Release July 2013,
http://www.casr.ca/doc-news-aor-norwegian-style.htm;
[4] Court of Appeal Judgment in Edwards v. National Coal
Board, [1949] 1 All ER 743;
[5] Cullen, The Honorable Lord, The Public Enquiry into the
Piper Alpha Disaster, HM Stationery Office, 1990;
[6] DCI GEN 270/93, Arrangements for the Management of
Safety for MoD Ships (not publicly available);
[7] Haddon-Cave, Charles QC, The Nimrod Review, An
Independent Review Into The Broader Issues Surrounding
he Loss of the RAF Nimrod MR2 Aircraft XV230 in
Afghanistan in 2006, October 2009;
[8] European Commission Directive 82/501/EEC issued in
1982 (known as the Seveso Directive);
[9] Health & Safety Executive: Reducing Risks, Protecting
People: HSEs decision making process, 2001, ISBN 0
7176 2151 0;
[10] IMO International Convention for the Prevention of
Pollution from Ships (MARPOL) 1973, as amended;
[11] Kelly, Timothy, Arguing Safety - A systematic Approach
To Managing Safety Cases, 1998;
[12] Lloyds Register Marine, Naval Ship Safety Assurance
Guidance for Navies and Shipbuilders, 2014;
[13] Lloyds Register Rules and Regulations for the
Classification of Naval Ships, January 2015;
[14] MSC.1/Circ.1164/Rev.15, International Convention on
Standards of Training, Certification and Watchkeeping for
Seafarers, 1978, As Amended, 17 June 2015 (and all
previous amendments);
[15] MSC.353(92), Amendments to the International
Management Code for the Safe Operation of Ships and
Pollution Prevention (International Safety Management
(ISM) Code), adopted 21/06/2013 (and all previous

124
Appendix 1 - Interface and Boundary Diagram

125
Appendix 2 - Claims Arguments Evidence diagram

03 - The ship can

conduct RAS, Lift-On,
Lift-Off, and other
defined stores evolutions
safely

02 - The risk of 03 - Stores can be

01 - RAS can be
conducted safely.
05 - VERTREP
causing environmental moved and secured
operations can be
pollution during RAS appropriately on the
conducted safely.
is reduced to ALARP. ship.

The RAS equipment

and RAS design has The ship has been
Stores can be moved Stores can be safely Control of the ship
been designed to designed to allow
safely around the secured on the ship can be maintained
reduce the risk of VERTREP operations
ship whilst at sea. whilst at sea. during RAS.
environmental to be undertaken.
pollution.

Appropriate wheeled The flight deck has Interactions between

Appropriate wheeled Seakeeping and mechanical handling Seakeeping and been arranged to the ship and
The RAS equipment The safe use of the
mechanical handling manoeuvring trials equipment will be manoeuvring trials allow for VERTREP helicopter have been
to be used has been RAS equipment is
equipment will be are being conducted provided with are being conducted operations. considered.
clearly stated. defined.
provided. for the ship. defined storage for the ship.
positions.

The basic design of The basic design of The Power and

The RAS deck has Appropriate deck Visibility of VERTREP Internal and external
the ship has the ship has Propulsion operating
been designed to coatings will be used The ship has defined operations can be communications are
identified stowage identified stowage scenarios have been
contain spillages in way of the stores stores routes. maintained during all defined for RAS
areas for all defined areas for all defined defined for RAS
from equipment. routes. operating conditions. operations.
cargo requirements. cargo requirements. scenarios.

The RAS deck has

The operating been designed for
RAS equipment has Control of the ship 04 - Lifting operations
conditions suitable RAS operations in all
been designed to can be maintained can be conducted
for RAS are clearly defined
operate safely. during RAS. safely on the ship.
defined. environmental
operating conditions.

Cranes suitable for

The Power and the operating The risk of human
Seakeeping and Appropriate weather Visibility of the RAS Internal and external
Propulsion operating environment and error associated with
manoeuvring trials deck coatings will be deck can be communications are
scenarios have been with appropriate safe crane operations has
are being conducted used on the RAS maintained in all defined for RAS
defined for RAS working loads have been considered
for the ship. deck. operating conditions. operations.
scenarios. been selected for the during basic design.
ship.

RAS deck equipment

The RAS deck officer The communications The RAS and
has been designed to The selected crane
and RAS deck crew Crane Operators will The selected crane required for container deck have
The RAS machinery RAS equipment is operate in the will have the safety
The ship has defined have safe havens to be trained in will meet Class operation of the been designed to
space is unmanned guarded as defined features defined in
stores routes. reduce the number accordance with Society crane have been ensure safe crane
during operation. appropriate. environmental Class Society
of personnel on MOD procedures. requirements. identified during operations can be
operating requirements.
deck. basic design. conducted
temperatures.

126
 Probabilistic Aspect on Minimum Propulsion Power
Requirement Issue under Adverse Weather Conditions
Shusuke Ohiwa, Osaka University, Shusuke_ooiwa@naoe.eng.osaka-u.ac.jp
Naoya Umeda, Osaka University, umeda@naoe.eng.osaka-u.ac.jp
ABSTRACT
At the IMO (International Maritime Organization),
minimum propulsion power guidelines are now under
discussion. The provisional guidelines require the min-
imum propulsion power for the specified Beaufort
number. The authors numerically estimated the mini-
mum power for guaranteeing safety against grounding
accidents under sea states specified by the significant
wave height, peak wave period and mean wind velocity
by using a surge-sway-yaw-roll model and its local sta-
bility analysis. The used method was compared with
model experiments using a handy-max bulk carrier and
and the agreement between the two is satisfactory at
least for head wind and waves. As a next step, the au-
thors attempt to establish the relationship between the
estimated results for several different states and the
mariners experience based on the Beaufort scale by
calculating danger probability under the different
Beaufort scales. The used ship is the handy-max bulk
carrier having MCR of 7930 KW. The added resistance
due to waves and the wind forces were predicted by a
potential flow theory and an empirical formula, respec-
tively. All the propulsive and manoeuvring forces are
estimated with the conventional model experiments for
the ship. In conclusion, it was conformed that the tested
ship in head wind and waves can avoid grounding up to
the Beaufort scale of 8.
INTRODUCTION
For reducing greenhouse gas from ships, Annex IV of
the MARPOL Convention (International Convention for the
Prevention of Pollution from Ships) requires us to reduce
the energy efficiency index (EEDI) by 30 % for a ship after
2015. This is because an energy efficient ship shall emit
smaller amount of greenhouse gas such as CO2. The EEDI
defined here is proportional to a ratio of the engine power
and the ship speed. Since the engine power is roughly pro-
portional to the cube of the ship speed, reducing the ship
speed by only 20 % results in reduction of engine power by
50 % so that the EEDI decreases more than 30 % (Takasaki,
2014). Thus simple reduction of the design ship speed and
engine power is easiest response to the EEDI requirement.
However, simple decrease of engine power could result
in operational danger under adverse weather. Typical ex-
amples, include grounding accidents of a Panamax bulk
carrier named Pasha Bulker and a Capesize bulk carrier
named Ocean Victory, although they were built before the
EEDI requirement, off New Castle port in Australia and
Kashima port in Japan, respectively. In both cases, the ships
failed to go ahead because severe head wind and waves,
and then lost their manoeuvrability and finally grounded.
Thus, the Annex IV also requires that installed propul-
sion power shall not be less than the power needed to
maintain the manoeuvrability of the ship under adverse
conditions as defined in the guideline to be developed by
the IMO. Therefore, developing the guideline is urgent
issue. While the EEDI regulation requires the maximum
allowable propulsion power, this guideline requires the
minimum one. As a result it is essential for the minimum
propulsion power guideline should have reasonably small
safety margin for realizing possible design range in propul-
sion power.
For this purpose, the authors attempted to examine op-
erability of a handymax bulk carrier under severe wind and
waves by identifying equilibrium points of a
surge-sway-yaw-roll manoeuvring model with a PD auto-
pilot and examining their local stability. This procedure was
well established in nonlinear dynamics and was already
applied to ship manoeuvrability in wind and waves by Asai
(1980), Kadomatsu (1990) and others. In this work, for
guaranteeing accurate prediction, more accurate modelling
of hydrodynamic elements and experimental validation in
wind and waves were executed. Furthermore, final judge-
ment of the installed propulsion power was made using
probability for allowing actual environmental data of rele-
vant water areas. For the same purpose, Shigunov et al.
(2016) applied a similar procedure but with neither experi-
mental validation in wind and waves nor probabilistic
assessment. Thus this work could supplement their work
with these elements.
EXPERIMENTAL METHOD
The model experiments were conducted using a
handy-max bulk carrier, which has superstructures, at a
seakeeping and manoeuvring basin of the National Re-
search Institute of Fisheries Engineering of Japan. Table 1
127
shows principal particulars of the bulk carrier. Irregular
wind and long crested irregular waves were generated at the X
Wind
same time by three sets of blowers and 80 segmented wave U

maker. The propeller rotation speeds at which the forward x

speed of the model is almost zero, due to winds and
waves, were determined as a function of the significant O Wave
wave heights. The spectrum of irregular waves is JON-
y
SWAP spectrum. Mean wave period is the period at which
the added resistance becomes maximum value. The rela- Y
O
tionship between the mean wind speed and the significant
wave height is based on Table 2. The details of this exper- Fig.1 Coordinates system
iment will be published later.

Table 1 Principal particulars of the bulk carrier

(m + mx )u& (m + my )vr = X H + X P + X R + X A + XW (1)
(m + my )v& + (m + mx )ur myl y = YH + YR + YA + YW (2)
&&
Ship Model
( I z + J z )&& = N H + N R + N A + NW (3)
Lpp(m) 178.0 2.9091
( I x + J x )&& m y l y v& mx l x ur + 2& + & & + mgGM =
B (m) 32.26 0.5273 (4)
K H + K P + K A + KW
D (m) 21.14 0.3455
TE& + = K p ( c ) K pTD& (5)
d (m) 11.57 0.1891
DWT (ton) 47500 0.207 where and are the ship speeds in the and direc-
tions. is the metacentric height. and are the
MCR (kW) 7930
added mass in the and directions. and are the
Dp (m) 5.51 0.09 moment of inertia around the and axes. and
are added moment of inertia around the and axes.
and are linear and quadratic roll damping coefficients.
Table 2 relationship between the mean wind speed and the and are hydrodynamic forces in the and direc-
significant wave height tions. and are hydrodynamic moments around the
Beaufort NO. 6 7 8 9 10 and axes. The subscripts, , , , and , mean
hull, propeller, rudder, wind and wave, respectively. is
H1/3 (m) 3 4 5.5 7 9
the rudder angle. ! is the time constant of the steering
Wind speed (m/s) 12.35 15.55 19 22.65 26.5 gear. " is the specified ship heading angle for the auto
pilot. # and $ are the proportional gain and the time
constant for differential control.
NUMERICAL METHOD The added resistance due to waves, Xw, are estimated by
Maruos theory (Maruo, 1963) with the Kochin function
For predicting required propulsion power under wind and calculated by Ogilvie-Tucks slender body theory. The
waves, a surge-sway-yaw-roll coupled simulation model is wave-induced steady sway force, Yw, and yaw moment, NW,
used with a PD autopilot. This models low-frequency mo- are calculated by a linear 3D panel method using the Green
tion only and ignores high-frequency motions such as function for zero forward speed. These wave-induced
harmonic motions due to incident waves or fluctuating steady forces and moment in irregular waves are calculated
wind. The coordinate systems describing the model are with the wave spectrum by assuming they are proportional
defined as shown in Fig.1. The coordinate system O-XYZ to the square of wave height. The wind forces and moments
represents a space-fixed, while the coordinate system O-xyz are estimated with Fujiwaras empirical formula (Fujiwara
does a body-fixed. Wind is assumed to blow in the negative et al., 1998). All the propulsive and manoeuvring forces,
direction of -axis. Here the ship heading measured except for the added masses and moments of inertia, are
from the -axis is indicated by and the angle be- obtained from the conventional model experiments for the
tween wind and wave directions is represented by . The ship such as circular motion tests using a X-Y towing car-
ship is assumed to run with its forward speed of with the riage. The added masses and moments of inertia are
drift angle of . The heel angle is against -axis. estimated by Motoras empirical charts.
Based on the above simulation model with the system pa-
Then equations of surge-sway-yaw-roll motion and a re- rameters, equilibrium points are determined by the Newton
sponse equation of the steering gear are represented as method. Then we locally linearized the mathematical model
follows. in the vicinity of the equilibrium and calculated the eigen-
values of the Jacobi matrix. If all the real part of the
eigenvalues is negative, the equilibrium is stable so that the
equilibrium point indicates a possible steady running condi-
tion.

128
NUMERICAL PREDICTION OF THE MINIMUM 12000
POWER IN HEAD WINDS AND WAVES Experiment

10000 Experiment(Steady wind)

For validating the above calculation procedure, model ex- Estimation(U=0kt)
periments using wind and waves are required. However, the 8000 Estimation(U=4kt)
scale effects of ship frictional resistance and propeller wake,

BHP (kW)
Plimit(U=4kt)
which depends on the Reynolds number, exist and accurate 6000
measurements of means of time-varying elements requires
longer measurements regardless shorter model basin. To 4000
resolve these difficulties, the authors executed model ex-
periments for a ship losing her forward speed due to severe 2000
head or bow-quartering wind and waves. As a result, the
scale effect due to hull boundary layer and statistical inac- 0
curacy due to the limited size of model basin can be 2 4 6 8 10
avoided. Significant wave height (m)
Fig.2 shows the comparison in the propeller revolution for Fig.3 Comparison of calculated value and required value
zero forward speed between the experimental results and
calculations. Good agreement between the experiment and NUMERICAL PREDICTION OF THE MINIMUM
the calculation for the Beaufort scales 6-10 are found. Fig.3 POWER IN OBLIQUE WINDS AND WAVES
shows the comparison in the brake horsepower. The calcu-
lated results well agree with the experimental ones. These Operability in head wind and waves itself cannot guaran-
good agreements demonstrate that the used calculation tee safety in oblique wind and waves. Thus, we extended
procedure is capable for explaining the required propulsion our calculation to the 4kts runs in bow quartering seas un-
power within practical accuracy. Similar comparison in der the Beaufort scale No. 8. As shown in Fig. 4, the
bow quartering waves using a different model can be found required engine power further increases from that in head
in Umeda et al. (2016). seas and has a peak at the specified heading of 60 degrees
from the wind direction. Even so, the required engine pow-
er is smaller than the engine limit. Fig. 5 shows the required
120
rudder angle for keeping the ship course. The required rud-
Rotational speed of the propeller (rpm)

Resting
der angle is smaller than its limit, i.e. 35 degrees, and the
100 Resting(Steady wind) real parts of eigenvalues for these ranges are negative. Thus,
Estimation the present calculation suggests that the ship can run with 4
80 kts even in bow quartering waves under the Beaufort scale
No. 8.
60
6000
40
5000

20
BHP (kW)

4000
2 4 6 8 10
Significant wave height (m)
3000

Fig.2 Measured propeller revolution at which ship rests

2000 Plimit
in head winds and waves and its estimation
1000 Estimation

As a next step, we apply the same calculation procedure to 0

the case that the ship runs with 4kts in the opposite direc- 0 10 20 30 40 50 60 70 80 90
tion of wind and waves. If a ship can run with 4kts in Course specified for auto pilot (degrees)
severe head wind and waves, she is able to leave lee shore Fig.4 Required engine power and its limit in bow quartering
even slowly for avoiding grounding. Fig.3 includes the waves under the Beaufort scale 8.
predicted engine power with the forward speed of 4 kts in
head wind and waves. The maximum allowable engine In stern quartering waves, the situation to be investigated
power is also shown based on the engine dynamics model- could be different. Since the added resistance due to waves
ling. An intersection between the two can be found at the could be negative in stern quartering waves, there is no
significant wave height of 7.5 metres, in other words, the difficulty to increase the propeller revolution. On the other
Beaufort scale No. 9 or over. Thus, it can be concluded that hand, for keeping the forward speed of 4 kts could require
the ship cannot run with 4kts towards wind and waves un- even negative propeller revolution so that the hydrodynam-
der the Beaufort scale No.10 because the required engine ic situation becomes unnecessarily complicated.
torque exceeds its limit.

129

Fig.5 Required rudder angle and its limit in bow quarter-
ing waves under the Beaufort scale 8.
Thus, we examine operability of the ship in stern quarter-
ing seas with the maximum allowable propeller revolution
number which can be obtained from the engine dynamics.
The results are shown in Figs. 6-7. In case of stern quarter-
ing seas ranging from 90 to 180 degrees from the wave
direction, both the engine power and rudder angle are be-
low their limits. Thus, the ship is able to select any course
in stern quartering waves under the Beaufort scale 8.
9000
PROBABILITY OF FAILURE DUE TO THE LACK OF
PROPULSION POWER
The environmental conditions discussed so far are defined
by the Beaufort scale. However, the Beaufort scale is not
suitable for probabilistic safety assessment because sea
state data for certain water area are usually available as
wave scattering diagram and rigorous relationship between
such sea state data and the Beaufort scale cannot be found.
Thus, we attempt to calculate the probability that a ship
meet dangerous sea states for failure due to the lack of
propulsion power. This probability could be more informa-
tive index than the binomial judgement using the Beaufort
scale.
Firstly, we utilise the wave scattering table for the North
Atlantic as shown in Table 3 as one of the worst situation.
The relationship between the mean wind velocity and the
significant wave height is assumed to be given by the fol-
lowing formula.
++.4
)* +, 0.067173
%&'(
(6)
This could represent the situation that a ship is departing
from a port facing the North Atlantic without any obstacles.
Table 3 Wave scattering diagram of the North Atlantic
H/Tz 1.5 2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 13.5 14.5 15.5 16.5 17.5 18.5
0.5 0 0 1.3 134 866 1186 634 186 36.9 5.6 0.7 0.1 0 0 0 0 0 0

8000 1.5 0 0 0 29.3 986 4976 7738 5570 2376 704 161 30.5 5.1 0.8 0.1 0 0 0
2.5 0 0 0 2.2 198 2159 6230 7450 4860 2066 645 160 33.7 6.3 1.1 0.2 0 0
7000 3.5 0 0 0 0.2 34.9 696 3227 5675 5099 2838 1114 338 84.3 18.2 3.5 0.6 0.1 0
6000 4.5 0 0 0 0 6 196 1354 3289 3858 2686 1275 455 131 31.9 6.9 1.3 0.2 0
BHP (kW)

5.5 0 0 0 0 1 51 498 1603 2373 2008 1126 464 151 41 9.7 2.1 0.4 0.1
5000 6.5 0 0 0 0 0.2 12.6 167 690 1258 1269 826 387 141 42.2 10.9 2.5 0.5 0.1
4000 7.5 0 0 0 0 0 3 52.1 270 594 703 525 277 112 36.7 10.2 2.5 0.6 0.1
Plimit 8.5 0 0 0 0 0 0.7 15.4 97.9 256 351 297 175 77.6 27.7 8.4 2.2 0.5 0.1
3000 9.5 0 0 0 0 0 0.2 4.3 33.2 102 160 152 99.2 48.3 18.7 6.1 1.7 0.4 0.1
2000 Estimation 10.5 0 0 0 0 0 0 1.2 10.7 37.9 67.5 71.7 51.5 27.3 11.4 4 1.2 0.3 0.1
11.5 0 0 0 0 0 0 0.3 3.3 13.3 26.6 31.4 24.7 14.2 6.4 2.4 0.7 0.2 0.1
1000 12.5 0 0 0 0 0 0 0.1 1 4.4 9.9 12.8 11 6.8 3.3 1.3 0.4 0.1 0
0 13.5 0 0 0 0 0 0 0 0.3 1.4 3.5 5 4.6 3.1 1.6 0.7 0.2 0.1 0
0 30 60 90 120 150 180 14.5 0 0 0 0 0 0 0 0.1 0.4 1.2 1.8 1.8 1.3 0.7 0.3 0.1 0 0
Course specified for auto pilot (degrees) 15.5 0 0 0 0 0 0 0 0 0.1 0.4 0.6 0.7 0.5 0.3 0.1 0.1 0 0
16.5 0 0 0 0 0 0 0 0 0 0.1 0.2 0.2 0.2 0.1 0.1 0 0 0
Fig.6 Required engine power and its limit in stern quarter-
ing waves under the Beaufort scale 8 with the maximum
propeller revolution.

Fig.7 Required rudder angle and its limit in stern quartering

waves under the Beaufort scale 8 with the maximum pro- Fig.8 Ratio of required engine power to its limit for the
peller revolution. ship with 4kts in head seas under various sea states.

130
 Secondly, for each sea state defined by the significant sented. It suggests that the bulk carrier should leave a port
wave height, the mean zerocrossing wave period and the sufficiently before the significant wave height exceeds
JONSWAP spectrum, the required engine power in head 7.5m. Further probabilistic analysis using the estimation of
seas is calculated using the present calculation procedure. consequence and uncertainty of loading could be future
The ratio of required engine power to its limit is shown in tasks.
Fig. 8. It is noteworthy here that the required engine power
drastically depends on the mean wave period. Thus, the ACKNOWLEDGEMENTS
assessment using the worst mean wave period could be very This research was supported by ClassNK and JSPS KA-
conservative. If this ratio for a certain sea state is larger KENHI Grant Number (15H02327). It was executed under
than 1, the sea state can be judged as dangerous for opera- the umbrella of the strategic research committee of the Ja-
tion of the ship. Table 4 indicates results of such judgement. pan Society of Naval Architects and Ocean Engineers
The danger due to the lack of propulsion power for the (JASNAOE) chaired by Prof. Yasukawa. For the calcula-
given ship appears when the significant wave height is tion of wave-induced steady forces and manoeuvring forces,
larger than 7.5 m. Profs. Yasukawa and Yoshimura kindly provided their data.
Thirdly, we calculated the failure probability by integrat- For the experiment, Mr. Matsuda and Dr. Terada from Na-
ing the probability density of sea states. The obtained tional Research Institute of Fisheries Engineering provided
results for the North Atlantic is 2.56 % because occur- effective advices and assistance. We thank Prof. Papaniko-
rence probability of the significant wave height larger than laou and Dr. Shigunov from EU Project (SHOPERA) for
7.5m is not so small in the North Atlantic. Of course, if the their valuable discussion as well as Dr. Miyake from
ship leaves the port sufficiently before the significant wave ClassNK and Dr. Tsujimoto from NMRI.
height exceeds 7.5m, the failure probability could be almost
zero.
REFERENCES
Table 4 Binomial judgements of operability for the ship
with 4kts in head seas under various sea states. Asai, S.: A Study on Check Helms for Course Keeping of a
Ship under Steady External Forces, Journal of Society of
H/Tz 1.5 2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 13.5 14.5 15.5 16.5 17.5 18.5 Naval Architects of Japan, Vol. 150, 1981, pp. 245-253.
0.5 P P P P P P P P P P P P P P P P P P
1.5 P P P P P P P P P P P P P P P P P P
Fujiwara. T. et al.: Estimation of Wind Forces and Mo-
2.5 P P P P P P P P P P P P P P P P P P ments acting on Ships, J. Society of Naval Architects of
3.5 P P P P P P P P P P P P P P P P P P Japan, Vol. 183, 1998, pp. 77-90 (in Japanese).
4.5 P P P P P P P P P P P P P P P P P P Kadomatsu, K. et al.: On the Required Minimum Output of
5.5 P P P P P P P P P P P P P P P P P P
6.5 P P P P P P P P P P P P P P P P P P
Main Propulsion Engine for Large Fat Ship with Consid-
7.5 P P P P P P F F F P P P P P P P P P ering Manoeuverability in Rough Seas, J. Society of
8.5 F F F F F F F F F F P P P P P P P P Naval Architects of Japan, Vol. 168, 1990, pp. 171-182
9.5 F F F F F F F F F F F P P P P P P P (in Japanese).
10.5
11.5
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
P
F
P
P
P
P
P
P
P
P
Maruo, H.: Resistance in Waves, 60th anniversary Series,
12.5 F F F F F F F F F F F F F F F P P P vol. 8, The Society of Naval Architects of Japan, 1963,
13.5 F F F F F F F F F F F F F F F F F P pp.67-102.
14.5 F F F F F F F F F F F F F F F F F P Nagai, T.: Long Term Statistics Report on Nationwide
15.5 F F F F F F F F F F F F F F F F F F
16.5 F F F F F F F F F F F F F F F F F F
Ocean Wave information network for Ports and Harbours
(NOWPHAS 1970-1999), Technical Note of the Port and
Furthermore, we use a wave scattering diagram off Airport Research Institute, no. 1035, 2002, pp.276-280,
(in Japanese).
Kashima port, which faces the North Pacific and in which a
Shigunov, V., Papanikolaou, A. and Chroni, D.: Manoeu-
grounding accident was reported, in place of that for the
vrability in adverse conditions: assessment framework
North Atlantic. This data is based on measurements in the
and examples, Proc. 15th Int. Ship Stability Workshop,
period between 1972 and 1999 with ultrasonic wave probes
(Nagai, 2002). The obtained probability is 0 % because the Stockholm, 2016.
maximum significant wave height in Kashima port was 7.5 Takasaki, K.: Reduction of CO2 Emission from Main En-
gine, Bulletin of the Japan Society of Naval Architects
m but the mean wave period is outside the dangerous sea
and Ocean Engineers, Vol. 56, 2014, pp. 23-26 (in Japa-
states for the ship. This means that the ship can leave
nese).
Kashima port anytime, if the present calculation and as-
Umeda, N., S. Koga, et al.: Methodology for Calculating
sumed loading and environmental conditions are correct.
Capsizing Probability for a Ship under Dead Ship Condi-
tion, Proc. 9th Int. Ship Stability Workshop, Hamburg,
CONCULUSIONS
2007, pp. 1.2.1-1.2.19.
A procedure to estimate required propulsion power and
Umeda, N. et al.: Tank Test Results and Simulation under
manoeuvrability under severe wind and waves is presented
Adverse Weather, Proc. Minimum Ship Propulsion Power
and is partly validated with model experiments in wind and
Symposium, Tokyo, 2016 (in Japanese).
waves. The procedure indicates that a handy-max bulk car-
rier can run with 4 kts or over in any direction under the
Beaufort No. 8.
As a more rigorous approach, a calculation procedure of
probability that a ship meets dangerous sea states is pre-

131
 A Meta-model for Risk Assessment of RoPax Capsizing as
an Alternative Way of Ship Safety Evaluation
Tomasz Hinz1,2, Przemysaw Krata2,3, Jakub Montewka3,4,5
1) Deltamarin sp.z o.o, Gdansk, Poland, tomasz.hinz@deltamarin.com
2) Waterborne Transport Innovation Foundation, apino, Poland
3) Gdynia Maritime University, Faculty of Navigation, Poland, p.krata@wn.am.gdynia.pl
4) Aalto University, Research Group on Maritime Risk and Safety, Espoo, Finland,
jakub.montewka@aalto.fi
5) Finnish Geospatial Research Institute, Masala, Finland

ABSTRACT recommendations, (IMO 1999). However, two main disad-

This article presents the application of risk assess-
ment of an intact ship stability failure, called ship cap-
sizing in dead ship conditions. The analysis is based on
the results of simulations of ship motions in regular
waves performed with the use of six degree of freedom
code called LaiDyn. Subsequently the results are orga-
nized into a probabilistic meta-model with the use of
Bayesian learning algorithms. Obtained results in a
form of F-N curve are plotted on top of IMO risk ac-
ceptance criteria. Finally they are compared with the
actual stability criteria. Good agreement is found for
those cases, where a ship fulfills the IMO stability crite-
ria by large margin. However, for the cases where the
prescriptive IMO criteria are barely met, the results of
risk assessment clearly indicate that the risk level is not
acceptable for the analyzed ship.
The approach could be found suitable for safety evalua-
tion of a ship during design stage and operation.
INTRODUCTION
One of the safety features of seagoing ships is ships
stability, that influences ships overall sea-keeping perfor-
mance. The stability related issues are covered by the prop-
er design and the appropriate operation. The loading condi-
tion of a ship causing the insufficient stability may lead her
to capsizing or generally the stability accident. The latter is
defined as exceeding the amplitude of rolling or an angle of
heel at which operating or handling of a ship is risky or
impossible. Generally, such an approach focusing on the
excessive heel avoidance is valid both in a port during car-
go operations and in rough sea when underway, (Krata and
Szapczyska 2012; Krata, Szpytko, and Weintrit 2013).
The design aspect is at least same vital in terms of
providing ships relevant stability as the operational one.
The contemporary approach towards ship stability assess-
ment is rooted in prescriptive criteria based on the IMO
vantages of this approach can be noticed. One objection is a
very simplified nature of these criteria, which are of statis-
tical nature. Only the weather criterion is based on the
mathematical model of ship capsizing even if this model is
only coarse approximation of the capsizing phenomenon.
Second reproach is a binary outcome of the stability as-
sessment procedure, where a ship either complies with the
requirement or not and there is no information provided
about the level of risk related to the considered loading
condition.
Under the pressure of wide criticism of the contempo-
rary stability standards the attempts were started in the mid
of 90ties to develop more rational stability criteria cover-
ing some typical dynamical phenomena inflicting hazards
to ship sailing in rough sea, (Francescutto 2004). As a re-
sponse to such demand the second-generation ship stability
criteria are being developed. An expected implementation
of the second-generation stability criteria shall notably
improve the known shortages of the present standards in
terms of relevant modeling of the dynamical phenomena.
Moreover, numerous postulates were submitted in the
marine society forum to make another step forward towards
the development of s called risk-based stability criteria,
(Kobylinski 2007). To address the risk estimation problem
the second-generation criteria and any other methods for
ship stability performance evaluation can be employed.
Within this approach, focusing on the evaluation of the risk
level, the modeling of ship stability characteristics is con-
sidered as means to an end, not an end itself.
Therefore, this paper presents the results of attempt
taken to develop a risk assessment of a particular stability
failure, i.e. the dead ship conditions (DSC) for a given
Ro-Ro/passenger ship (RoPax). The outcome of this as-
sessment shall provide information about the level of risk
related to the operation, which is appreciable in the ship
design process. The obtained risk can be compared to the
commonly recognized risk acceptance margins, however it
can be utilized in much wider way. The outcome variable

132
enables not only a statement whether the ship complies with
any criteria but rather it provides an extended information
how far from the accepted safety margins the obtained risk
level remains. To achieve this a probabilistic meta-model is
developed and tested. The idea of probabilistic meta-model
has been successfully implemented earlier for the assess-
ment of damage ship safety in waves, see for example
(Papanikolaou et al. 2010; Schreuder et al. 2011).
The process adopted here for the development of me-
ta-model is two-stage, (Montewka, Goerlandt, and Zheng
2015). First, numerous simulations are run for a chosen
RoPax for a number of anticipated operational conditions.
For that purpose we use a state-of-the-art numerical model
of ship dynamics called LaiDyn. This allows overcoming of
the limitations of commonly used static methods. Second,
the obtained results are organized in a probabilistic me-
ta-model with the use of Bayesian Belief Network (BBN).
Subsequently the meta-model is used as a tool for opera-
tional risk assessment of a ship with respect to her capsiz-
ing in DSC, where a ship is exposed to adverse action of
wave and wind. To measure the risk we adopt the probabil-
ity of a given number of fatalities resulting from the
above-mentioned accident. The results are presented in a
form of F-N curve plotted on top of the risk acceptance
criteria. To inform the end-users about the quality of the
meta-model and its validity, two analyses are carried out.
These are cross-validation and sensitivity-uncertainty-bias
assessment. The first determines the accuracy of the me-
ta-model compared to the results of numerical simulation,
the second helps to determine the most relevant variables
for the model in order to point to the direction of the future
research. Such approach may help to develop trust in the
meta-model among the end-users.
Finally the results of the risk assessment for a set of
case studies are compared with the existing prescriptive
stability criteria. Rather good agreement is found for those
cases, where ship fulfills the IMO stability criteria by large
margin. However, for the cases where the prescriptive IMO
criteria are barely met, the results of risk assessment clearly
indicate that the risk level is not acceptable for a ship.
The remainder of the paper is organized as follows: sec-
tion II elaborates on the stability requirements. Section III
presents methods and materials used to develop the me-
ta-model assessing the risk of capsizing in DSC. The ob-
tained results are shown in Section IV and discussed in
Section V. The concluding remarks are provided in Section
VI.
AN OVERVIEW OF SHIP STABILITY REQUIRE-
MENTS
The history of shipping reaching at least a couple of
thousands years dispensed with a deeper understanding and
any standardizing of ship stability for majority of time.
However, in recent decades the subjective value of human
life steady increases resulting in endeavor to improve the
safety of navigation. Historically, the first approach towards
ship stability evaluation was based on an experiment only.
Till the time of boat launching one could not be sure the
design correctness so the design process had to be based on
architects and seamens experience and on earlier projects
similitude.
The works of Bouguer from 1746, Euler from 1749,
Bernoulli from 1757 and Attwood from 1796 are recog-
nized as the foundations of a scientific approach towards
ship stability. The first introduced the metacentric height
notion remaining the only measure utilized to assess ship
stability till the end of the XIX century, (Kobylinski and
Kastner 2003). In the mid-nineteenth century the righting
arm curve was already know and thanks to Moseleys pub-
lication from 1850 also the notion of dynamical stability
was introduced. Although, those theoretical achievements
were not fully applied in practice and none stability stand-
ards were elaborated, (Kobylinski and Kastner 2003).
A significant breakthrough in the field of ship stability
standardization was Raholas doctoral dissertation from
1939, (Rahola 1939). He statistically obtained the margins
between sufficient and insufficient stability characteristics
analyzing the initial metacentric height and the shape of the
righting arm curve including some areas under this curve,
(Womack 2003). In the 1960s under the auspices of the
IMO the contemporary ship stability standards were devel-
oped, mainly covering Raholas results. In the 80s they
were supplemented by the weather criterion inspired by the
Japanese proposal and researches.
However, during some previous IMO - SLF sessions
(from 47th onwards) discussions have taken place concern-
ing the new generation of regulations, so called sec-
ond-generation stability criteria, (Umeda 2013). Four defi-
nitions of criteria are given, which deal with the assessment
of the intact stability failure in different ways (Szozda
2014):
A probabilistic performance-based criterion is a crite-
rion based on a physical model of a stability failure
considering this phenomenon as a random event, see
for example, (IMO 2006a, 2006c)
A deterministic performance-based criterion is a crite-
rion based on a physical model of a stability failure
considering this phenomenon in a deterministic manner,
see for example, (IMO 2006b).
A probabilistic parametric criterion is a criterion based
on a measure of a quantity related to a phenomenon,
but does not contain a physical model of the phenom-
enon, and includes one or more stochastic values for
this criterion (for example, the new probabilistic dam-
age stability criterion contained in resolution (IMO
2005).
A deterministic parametric criterion is a criterion based
on a measure of a quantity related to a phenomenon,
but does not contain a physical model of the phenom-
enon, while all the input values are deterministic (for
example, the GZ curve criterion contained in section
2.2 of the draft revised Intact Stability Code).
Intact stability failure is defined as a state of inability of
a ship to remain within design limits of roll angle and com-
bination of rigid body accelerations, (IMO 2008). The sec-
ond-generation regulation are developed based on the fol-
lowing four scenarios of intact stability failure, however
some authors enumerate more, (Kobylinski 2003, 2008): 1)
Dead Ship Condition, i.e. ship without forward speed, ex-
posed to action of waves and wind; 2) Pure-loss of stability;
3) Parametric roll; 4) Surf-riding and broaching; 5) Exces-
sive accelerations.
The results provided by the second-generation ship sta-
bility criteria are of binary nature, likewise the
133
first-generation criteria. The ship either passes the criteria
or not, without specifying the safety margin for various
loading conditions. The stability criteria are still based on
the comparison of a specific computation result and a lim-
iting value established on the basis of statistics. The limit-
ing values are worked out for contemporary typical conven-
tional ships, which make the approach, suspect to be less
effective in case of non-conventional hulls. Moreover, the
discussed criteria can be used only at the design stage of a
ship and they are not relevant to ship operation, (Szozda
2014). Despite being a significant step forward in stability
assessment development, the second-generation stability
criteria however say nothing about the consequences of a
stability failure. The latter is a substantial element of risk
assessment, which can be seen as a holistic approach to
safety evaluation of a ship.
To address the above shortcomings, while utilizing to a
large extend the achievements of the second-generation
criteria, we propose a meta-model that quantifies the risk of
ship capsizing in DSC allowing for displaying the associ-
ated risk level on top of risk criteria thus showing the dis-
tance of the risk level from the delimiting criteria lines. The
meta-model encompasses the probability of an event along
with its adverse consequences.
METHODS AND MATERIALS
Ship motion simulations
The crucial element of any risk models related to ship
stability is prediction of ship motion when in the rough sea.
Numerous variables characterizing intensity of rolling,
pitching and other motions are introduced into such models
likewise in the case described here.
In recent years majority of significant research centers
makes many attempts to develop accurate and effective
mathematical models of ship motions to solve the complex
problem of ship motions prediction (Hinz 2007; ITTC
2014; Spanos and Papanikolaou 2009). The meta-model
proposed here for risk of ship capsizing in DSC utilizes
LaiDyn code, (Matusiak 2002), to evaluate motions of a
given RoPax when adrift in DSC.
Since the linear theory of ship motion is well acknowl-
edged and the linear computation of ship response are
commonly known, the LaiDyn code is based on the as-
sumption that the complete response of a ship to external
excitation equals the sum of linear and non-linear parts
(Matusiak 2007). The radiation and diffraction forces are
presented by linear equations. The main part of the first
order load is calculated with the linear approximation,
based on the current heading and location in relation to a
wave. Defining the non-linear part, such elements as
non-linearity due to ship shape, hydrostatics and wave force
were taken into consideration (Matusiak 2011). The results
given by the code has been validated with the use of model
scale tests, see (Acanfora and Matusiak 2014; Matusiak and
Stigler 2012). Moreover it took part in several benchmarks,
see for example (Potthoff and Papanikolaou 2016; Spanos
and Papanikolaou 2009).
Meta-model development
The meta-model for risk assessment of a ship in DSC
presented here combines deterministic and probabilistic
134
approach. Relatively high number or simulations that are
run for a number of anticipated operational conditions with
the use of the LaiDyn code overcomes the shortcomings of
the existing static methods that are used for safety assess-
ment of a ship. The simulations method allow the analysis
of numerous scenarios, where a ship is exposed to varying
environmental loads, delivering a wide set of ship responses
in terms of roll angle and time to reach the maximum roll
angle, given hydro-meteorological conditions. The proba-
bilistic approach taken to develop the meta-model recog-
nizes the inherent uncertainties in the explanatory variables.
Thus it overcomes some of the issues of static, determinis-
tic methods adopted nowadays for safety evaluation of a
ship in operation. A concept of the meta-model is depicted
in Fig. 1.
Moreover, the application of the Bayesian network, al-
lows the end-user not only to predict the level of risk given
set of inputs (predictive reasoning) but also the backward
reasoning can be made, from the effect to the causes. The
latter helps to determine the required value for input pa-
rameters in order to obtain the desired value of the output.
Finally, the BNs allow for quick assessment of model sen-
sitivity and uncertainty, determining the paths of uncertain-
ty propagation. This is very useful in risk analysis, where
uncertainty is an inherent, though often neglected, feature
of risk, see for example (Aven 2008; Goerlandt, Montewka,
and Kujala 2014; Goerlandt and Reniers 2016).
Weather Repair 5me
Response of Time to
Evacua5on 5me
ship in DSC capsize
Ship stability Fatality rate
Capsizing
Probability of Number of
F-N curve
DSC people on board
Fig. 1. A concept of meta-models estimating risk of ship
capsizing as a result of DSC, (Hinz, Montewka, and Krata
2016)
Risk model parameters
There is a wide number of interpretation and measures
for risk, see for example (Aven 2010). However the me-
ta-model presented here expresses risk in its most frequently
met manner, namely the probability of the consequences.
The latter is expressed as an expected number of fatalities,
resulting from the stability accident. Mathematically, the
adopted notion of risk can be formulated as follows, (Hinz
2015):
~
= | _ | ,
(1)
where P(DSC) is the probability of a ship being in DSC
when at sea, P(caps|DSC) is the probability for a ship to
capsize when in DSC, P(N_fat|caps,DSC) is the probability
of certain number of fatalities when capsized in DSC.
The probability for a RoPax being in DSC while in op-
erations at sea is taken as 0.14 per year per ship. This number
is derived from the experts judgment on the reliability of
machinery and electrical systems of a ship in operation,
(Brandowski 2009). However this number is most likely
debatable, since it depends on numerous factors, like the
engine type and configuration, mode of ship operation,
moreover the number is subject to a spread, as reported in the
available body of literature, (Antao and Guedes Soares
2006).
Another major parameter of the risk model is
P(caps|DSC), which is determined by ship behavior in
waves. This is governed by inter alia ships loading
condition, which defines stability condition of a ship. For the
purpose of this paper we anticipate four different loading
conditions for the same ship draft, as presented in Table 1.
For each loading conditions a set of simulations with the use
of LaiDyn is conducted. From the time series of a simulation
the relevant information about the maximum roll angle and
the time of reaching it are obtained.
The ship is considered capsizing if the time of reaching
the critical angle is shorter than the anticipated time of repair
time of recovering from the DSC. Otherwise, the ship is
considered safe. Time to repair is modeled with the use of
Weibull distribution with alpha=0.5 and beta=23, (Leva et
al. 2006). In a similar manner the life loss is determined, if
the time to capsize is shorter than the evacuation time, we
assume that the life is perished. Then, the number of fatali-
ties (N) is estimated based on a concept of fatality rate -
(Jasionowski, Vassalos, and Scott 2007) - calculated as
follows, (Montewka et al. 2014):
= (2)
where haz is the exposure time to a hazard (time for the
ship to capsize), resp is the response time, in this case the
evacuation time, Npax is the number of people on board the
ship. The latter is obtained from the available data published
by ship operators in the Baltic Sea regarding the total
monthly volume of passengers transported with the use of
similar RoPax ships as the one considered here. Assuming
an ordered evacuation of a ship in danger, the time to evac-
uate the ship is modelled with the use of triangular distribu-
tions following the IMO recommendations, and a distinction
is made between day and night, for more details see (IMO
2007; Montewka et al. 2014).
Wave and wind parameters are taken from the available
statistics as presented in IMO documents - (IMO 2013a) -
and combined into a three-state variable called weather, as
presented in Table 2. The adopted distribution of this varia-
ble reflects the prevailing wave conditions for the Northern
Atlantic Ocean, however it can be easily updated to reflect
any sea area. The direction of the encountering waves with
respect to a ship heading is modeled with the uniform dis-
tribution, where the variable is discretized in four states
(180-270, 270-360, 360-90, 90-180), covering all pos-
sible directions of wave.
Table 1 Loading condition
Loading condition Draft [m] GM [m]
LC01 6.10 0.30
LC02 6.10 0.98
LC03 6.10 1.11
LC04 6.10 2.29
Table 2 Weather condition
Weather Significant wave height [m]
Good Below 3.33
Moderate Between 3.33 and 7.34
Bad Above 7.34
Evaluation of meta-model suitability for a purpose or risk
assessment
In the classical view on quantitative risk assessment
(QRA) in many technical disciplines, the aim of a QRA is
considered to be an accurate estimate of an underlying true
probability. This true probability is unknown, and uncer-
tainty is expressed about the accuracy of this supposedly
true probability. When probabilities are derived through
engineering models, e.g. through the exceedance of a limit
state in terms of ship capsizing, these engineering models
typically involve a number of simplifications. These can
lead to evidential uncertainty, but the type of simplification
may also involve value-laden choices of a modeler
(Diekmann and Peterson 2013). In particular, certain choic-
es may lead to more conservative or optimistic model re-
sults compared to how the modeled system works in reality.
This can be considered evidential bias and can also be qual-
itatively assessed, as originally proposed in (Rosqvist and
Tuominen 2004).
The importance of evidential uncertainty and bias de-
pends on the overall importance of the BN node to which
the evidential uncertainty and/or bias applies. In this respect,
the sensitivity analysis of the BN provides useful insights:
evidential uncertainty and/or bias of highly sensitive nodes
is more important than for nodes which imply low overall
sensitivity, for more in-depth elaboration of that issue the
reader is referred to our earlier study presented in
(Goerlandt et al. 2014).
Cross validation of the meta-model
K-fold Cross-Validation (CV) is performed to check the
predictive power of the probabilistic meta-model by com-
paring the response of the meta-model with the data rec-
orded in the course of simulation (training dataset).
Within the K-fold CV part of the data is used to develop
the model, and the other part is used to check the predictive
power of the model. The K-fold algorithm works as fol-
lows:
1. randomly divide the training dataset into K sub-
sets;
2. for each subset S:
train on the data but not on the subset S;
test model on the subset S;
3. return the average error over the K subsets.
The results of this analysis are presented in Table 3 with
respect to a variable called Number of fatalities (N), which
is discretized into five states. Very good prediction power
for the meta-model is noted for all the states of this variable
but one. Prediction power is understood as a level of
agreement between the results obtained from the me-
ta-model and the results from the simulation. The lowest
prediction power holds for the state of N covering the high-
est number of fatalities (2500-3000).
Among all conditions leading to that specific number of
fatalities (2500-3000) that are recorded in training dataset,
the obtained meta-model associates properly only 20% of
them. In the remaining 80% of the cases that ended up with
N=(2500-3000) the meta-model delivers the following
results: N=(1500-2500) in 72% - and N=0 in 8%. For
135
 other states of variable N the probability of delivering the
correct answer by the meta-model falls between 0.82 and
0.99.
Sensitivity, uncertainty, bias assessment
Both the uncertainty (U) and the bias (B) of the model
elements can be systematically assessed and put in relation
to the sensitivity of the model outcome to changes in the
individual nodes of the BN (S). Such an assessment is qual-
itative, but it provides an overall picture of how uncertain
the resulting probability assignment in terms of the F-N
curve is expected to be, see (Goerlandt et al. 2014). More-
over, it provides some insight as to whether the F-N curve
is expected to be optimistic or conservative, namely the
model underestimate or overestimate the risk level. Such an
integration of sensitivity, uncertainty and bias in one
framework can be beneficial to support decision-making.
The classification for the level of U, B and S shown in
Table 4, is based on (Flage and Aven 2009).
The value-of-information analysis
This analysis identifies the most informative variables,
with respect to the output variable. It determines the varia-
bles among which the probability mass of the output is
scattered. This analysis can be seen as a tool for analyzing
the potential usefulness of additional information, before
the information source is consulted, i.e., what the effect of
observing one node would have on the other. To visualize
the potential influence that two directly connected nodes
have on each other the thickness of the arcs connecting
them is varied, (Koiter 2006). The thicker arc the stronger
influence, the results of this analysis are shown in Figure 2.
Table 3 The results of CV of the meta-model
Results obtained from the probabilistic meta-model
Number of 0 0- 750- 1500- 2500-
fatalities 750 1500 2500 3000
Training dataset
N
0 0.9999 0.0000 0.0001 0.0001 0.0000
0-750 0.0510 0.8163 0.1122 0.0204 0.0000
750-1500 0.0000 0.0000 0.9611 0.0389 0.0000
1500-2500 0.0101 0.0050 0.0000 0.9246 0.0603
>2500 0.0800 0.0000 0.0000 0.7200 0.2000
RESULTS
To measure the risk associated with the stability acci-
dent of DSC of a ship in operation we adopt a cumulative
probability of fatalities that results from the
above-mentioned accident. The results are presented in a
form of F-N curve plotted on top of the risk acceptance
criteria. However the meta-model presented in Figure 2
determines the probability of a given interval of fatalities (f),
once the accident happens. To get the cumulative frequency
per year (F), the results needs to be multiplied with the
probability of DSC and the frequency of an interval (f)
needs to be transformed into the cumulative frequency (F).
Social risk acceptance criteria are adopted here, as proposed
by IMO for RoPax and passengers ships, (IMO 2013b).
To inform the end-users about the quality of the me-
ta-model and its validity, a set of analyses is performed.
Their results indicate that the structure and the parameters
136
of the model correspond with the data-set obtained from
over 17300 simulations of a ship in DSC (C-V analysis).
This means that the translation of data obtained from nu-
merical simulation is sound.
Subsequently qualitative assessment of the meta-model
regarding its evidential uncertainty, bias and sensitivity is
carried out, and the abridged results are presented in Figure
3. The results of it show that the meta-model is moderately
sensitive to majority of variables, and highly sensitive to
two variables (Evacuation time and Probability of DSC).
However these variables are also characterised with low to
medium uncertainty, with a tendency to overestimate the
modelled parameters. Overestimation of the parameters
means here the overestimation of the probability of fatali-
ties, thus risk. The majority of variables stay within low to
medium level of both sensitivity and uncertainty moreover
they tend to take the middle ground in the bias assessment.
All that mean that the results obtained form the model are
expected to slightly overestimate the risk of ship capsizing
in DSC.
Finally, when we look at the results of the value of in-
formation analysis, it is evident that there is significant
influence of Time of day on the Time for evacuation, which
is a sensitive node. Obviously there is also a significant
influence of Capsizing on Time to capsize and Probability
of fatalities. This means that the most of the uncertainty
associated with the outcome node is passed along those
paths, which are marked in bold arrows in Figure 2.
9 ELEMENTS OF META-MODEL
H A - Wave direction
7 B - Weather
B I
C - Loading conditions
BIAS
5 A,C G D - Roll angle
D
E - Time to max roll angle
E F
3 F - Repair time
G - Evacuation time
1 H - Probability of DSC
L M H I - Number of people on board
UNCERTAINTY
H G,H H G H
SENSITIVITY
SENSITIVITY
F
M C B,E D,F I M E C,D B, I
L
L A A
L M H 1 3 5 7 9
UNCERTAINTY BIAS
Parameter
underestimated overestimated
Fig. 3. Qualitative SUB-assessment of the meta-model of
the risk of RoPax capsizing in DSC.
Subsequently the safety of a ship is evaluated, by plot-
ting the results of risk analysis on top of the risk accepted
criteria given by the IMO. Finally the results are compared
with the standard stability criteria, as currently in force, for
four loading conditions as anticipated by the risk model, as
depicted in Figure 4. The ship is considered safe, when the
risk falls within the acceptable region. On the lower part of
Figure 4, the actual stability parameters in each considered
loading condition of the ship are plotted in color dots and
limiting curves reflect individual stability criteria. If a dot is
above the corresponding line the ship is considered safe,
otherwise she is not.
Comparing the risk picture with the classical stability
criteria picture, it is evident, that good agreement is found
for those cases, where ship fulfils the IMO stability criteria
by large amount (LC4). This is the case for LC4 (however
the curve goes beyond the limiting lines for some numbers
of fatalities). However, for the cases where the prescriptive
IMO criteria are barely met (LC1-3), the risk level is not
acceptable for that ship, since the risk measure falls beyond
the limiting lines of ALARP.
Fig. 4. The results of risk assessment for ship capsizing in
DSC plotted on top of risk acceptance criteria for ships
carrying passenger above; graphical representation of four
loading conditions plotted on top of present stability criteria
- below.
DISCUSSION
The meta-model presented here can be utilized for in-
novative ships and their unusual loading conditions. The
range of effects taken into account and the relation between
them are exceptional. Not only is the dynamical behavior of
the ship considered but also other time-depended phenom-
ena producing the whole scenario of the studied incident.
Despite the unquestionable advantages the presented
approach face some difficulties as well. First of all the BNs
require the quantitative adjustment in terms of probability
distributions of the applied variables which may be prob-
lematic since there is a shortage of well-established reliable
models in the literature. The considered evacuation time is
an important factor influencing the meta-model outcome
while this time may be predicted only roughly. There is a
lack of evacuation models predicting the time required to
abandon a ship with regard to actual motion amplitudes that
seems to be important. On the other hand, the ship in DSC
capsizes rather rapidly, which does not leave much time for
an ordered evacuation. Moreover, the prevailing weather
conditions at the time of capsizing may be very challenging
for successful evacuation. For that reason, a straightforward
assumption of all hands lost in case of DSC accident could
be made instead. However, we decided to follow the same
logic as in case of damage ship capsizing, where we quan-
tify the share of passengers that are likely to perish, once
the time to capsize is shorter that evacuation time see Eq.
2.
Another challenge influencing the final outcome of the
meta-model is ships motion modeling. LaiDyn software
was utilized as well-benchmarked simulation tool, (Spanos
and Papanikolaou 2009;Potthoff and Papanikolaou 2016),
although it contains numerous simplifications. Due to the
complexity of ship 6DOF motion it is difficult to assess to
what degree the probability of breaching the predefined
marginal angle of heel corresponds with the actual ship
behavior in rough sea. The time spent to reach this critical
angle of heel is one of the key variables in the model (Roll
angle, Time to capsize). Reducing uncertainty about varia-
ble Weather, Number of people on board and Loading con-
ditions will also reduce the overall uncertainty of the me-
ta-model. However, the uncertainty associated with the first
and second variables is of aleatory nature, thus can be
deemed unresolvable, (Der Kiureghian and Ditlevsen 2009).
The knowledge about Loading conditions is not available to
us at the moment, but in principle could be gained. The
uncertainty of Repair time depend on design and operation-
al factors, which are not considered here, however the
model is not very sensitive to that node.
The influence of the uncertainty associated with the es-
timates of the probability of a ship being in DSC while in
operation seems to be rather significant. This parameter is
obtained based on the experts judgment, which accounts
for various aspects of ship operational conditions. On one
hand such assessment can be considered less uncertain than
if based on reliability data given by the manufacturer. On
the other the value for that parameter holds for medium size
container ship with one engine, thus for a double-engine
RoPax this value can be lower, pulling down the risk values.
Definitely this parameter needs more detailed investigation.
It is also important how the nodes are discretized, how
many states do they have, and what are the bases for se-
lecting those. This affects the results of cross-validation,
and the overall performance of the meta-model, (Montewka
et al. 2015; Pitchforth and Mengersen 2013).
Last but not least, the time required for the meta-model
preparation can be seemed as a drawback compared to
classical stability criteria. The process of meta-model de-
velopment is at the moment time-consuming since there is a
need for performing a huge number of numerical simula-
tions to obtain the probability distributions for the essential
variables. However, the preparation of the data-set is an
exercise that needs to be performed only once in ships life,
for a wide range of operational conditions.
Notwithstanding the mentioned difficulties the perfor-
mance of the meta-model is decent and the application of
137
the model to evaluation of safety of ship in operation goes
beyond the contemporary ship stability assessment methods.
The holistic approach taking into account the consecutive
events creating scenarios bring the model closer to real life
situation than any other.
CONCLUSIONS
The risk assessment of a specific accident type for a
RoPax ship is presented here, which can be seen as a novel
approach to evaluate the safety of a ship at sea. The proba-
bilistic meta-model developed with the use of the
state-of-the-art ship dynamics model and organized with the
use of Bayesian learning algorithm indicates the level of
risk associated with a given operational scenarios, account-
ing for the uncertainties, providing the end-user with con-
vincing message rather than just stating that the ship fulfils
the criteria or not, as it is presented in the existing methods.
The latter may give a false impression of a ship being safe
in conditions where the criteria are met, but the risk is un-
acceptable. Therefore, the presented model shows its supe-
riority over the existing approach in assessing the safety
level of a ship.
Moreover, the meta-model can be further extended to
comprise other scenarios of stability incidents and
non-stability related failures. The model can be used at the
stage of ship design as well as ship operation. The sensitiv-
ity, uncertainty, bias assessment as well as the value of
information assessment allows for a semi-qualitative evalu-
ation of the model quality. This in turn allows reflecting
upon the probabilities that the meta-model is developed
upon and detecting the main path of uncertainty propaga-
tion.
Further studies could focus on further refinement of
sensitive and uncertain variables as well as quantification
and visualization of their uncertainty on the risk level de-
picted in a form of the F-N curve.
ACKNOWLEDGEMENTS
Professor Jerzy Matusiak from Aalto University is ap-
preciated for making his model LaiDyn available for the
purpose of this research. The probabilistic models intro-
duced in this paper were created using the GeNie modeling
environment developed at the Decision Systems Laboratory,
University of Pittsburgh available from
http://www.bayesfusion.com/.
The Merenkulun sti the Maritime Foundation - IMO. 2006c. SLF 49/INF.7 REVISION OF THE INTACT
from Helsinki is appreciated for the travel grant provided.
REFERENCES
Acanfora, Maria, and Jerzy Matusiak. 2014. Quantitative
Assessment of Ship Behaviour in Critical Stern Quartering
Seas. 14th International Ship Stability Workshop
(October):1927.
Antao, Pedro, and Carlos Guedes Soares. 2006. Fault-Tree
Models of Accident Scenarios of RoPax Vessels.
International Journal of Automation and Computing
3(2):10716.
Aven, Terje. 2008. Risk Analysis: Assessing Uncertainties beyond
Expected Values and Probabilities. Chichester, UK: Wiley.
Aven, Terje. 2010. On How to Define, Understand and Describe
Risk. Reliability Engineering & System Safety 95(6):623
138
31. Retrieved September 17, 2012
(http://www.sciencedirect.com/science/article/pii/S0951832
01000027X).
Brandowski, Alfred. 2009. Estimation of the Probability of
Propulsion Loss by a Seagoing Ship Based on Expert
Opinions. Polish Maritime Research 1(59):7377.
Diekmann, Sven, and Martin Peterson. 2013. The Role of
Non-Epistemic Values in Engineering Models. Science
and Engineering Ethics 19(1):20718.
Flage, Roger, and Terje Aven. 2009. Expressing and
Communicating Uncertainty in Relation to Quantitative
Risk Analysis. Reliability & Risk Analysis: Theory &
Application 2(13):918.
Francescutto, Alberto. 2004. Intact Ship Stability: The Way
Ahead. Marine Technology 41(1):3137.
Goerlandt, F., J. Montewka, and P. Kujala. 2014. Tools for an
Extended Risk Assessment for Ropax Ship-Ship Collision.
Pp. 22922302 in Vulnerability, Uncertainty, and Risk:
Quantification, Mitigation, and Management - Proceedings
of the 2nd International Conference on Vulnerability and
Risk Analysis and Management, ICVRAM 2014 and the 6th
International Symposium on Uncertainty Modeling a.
Goerlandt, Floris, and Genserik Reniers. 2016. On the
Assessment of Uncertainty in Risk Diagrams. Safety
Science 84:6777. Retrieved December 26, 2015
(http://www.sciencedirect.com/science/article/pii/S0925753
515003215).
Hinz, Tomasz. 2007. Mathematical Models in Description of
Capsizing Scenarios. Archives of Civil and Mechanical
Engineering 7(3):12534.
Hinz, Tomasz. 2015. Analiza Ryzyka Jako Alternatywna Metoda
Oceny Bezpieczenstwa Statecznosciowego Statku W Stanie
Nieuszkodzonym. Gdansk University of Technology.
Hinz, Tomasz, Jakub Montewka, and Przemysaw Krata. 2016.
The Use of Bayesian Networks for Risk Assessment of
Ship Capsizing as an Alternative Way of Evaluating the
Operational Safety of a Ship at Sea. in 13th International
Conference on Probabilistic Safety Assessment and
Management. Seoul, Korea.
IMO. 1999. Code on Intact Stability for All Types of Ships
Covered by IMO Instruments, IMO Res. A.749(18),as
Amended by MSC.75(69). London.
IMO. 2005. RESOLUTION MSC.194(80) ADOPTION OF
AMENDMENTS TO THE INTERNATIONAL
CONVENTION FOR THE SAFETY OF LIFE AT SEA,
1974.
IMO. 2006a. SLF 49/5/5 REVISION OF THE INTACT STABILITY
CODE Proposal of Methodology of Direct Assessment for
Stability under Dead Ship Condition Submitted by Japan.
IMO. 2006b. SLF 49/5/6 REVISION OF THE INTACT STABILITY
CODE A Methodology of Direct Assessment for Capsizing
due to Broaching Submitted by Japan.
STABILITY CODE Containership Safety Submitted by the
Netherlands.
IMO. 2007. Guidelines for Evacuation Analysis for New and
Existing Passenger Ships.
IMO. 2008. SLF 51/WP.2 REVISION OF THE INTACT
STABILITY CODE Report of the Working Group (Part 1)
GENERAL. in 51st session of IMO Sub-Committee on
Stability and Load Lines and on Fishing Vessels Safety.
IMO. 2013a. Development of Second Generation Intact Stability
Criteria. Vulnerability Assessment for Dead-Ship Stability
Failure Mode Submitted by Italy and Japan. SDC 1/INF.
London.
IMO. 2013b. Revised Guidelines for Formal Safety Assessment
(FSA) for Use in the IMO Rule-Making Process.
MSC-MEPC.2/Circ.12. LONDON, UK.
ITTC. 2014. Seakeeping Committee Final Report and
Recommendations to the 27th ITTC. Pp. 195262 in 27th
 International Towing Tank Conference. Copenhagen, Environment Protection, edited by Adam Weintrit and
Denmark. Tomasz Neumann. CRC Press, Taylor&Francis Group,.
Jasionowski, Andrzej, Dracos Vassalos, and Andrew Scott. 2007. Papanikolaou, Apostolos et al. 2010. GOALDS - Goal Based
Ship Vulnerability to Flooding. in 3rd international Damage Stability. Pp. 4657 in Proceedings of the 11th
maritime conference on design for safety. Berkeley, International Ship Stability Workshop, 2010, edited by
California. Apostolos Papanikolaou. Wageningen: Springer Berlin
Der Kiureghian, Armen, and Ove Ditlevsen. 2009. Aleatory or Heidelberg.
Epistemic? Does It Matter? Safety Science 31:10512. Pitchforth, Jegar, and Kerrie Mengersen. 2013. A Proposed
Kobylinski, Lech. 2003. Capsizing Scenarios and Hazard Validation Framework for Expert Elicited Bayesian
Identification. 8th International Conference on the Networks. Expert Systems with Applications 40(1):16267.
Stability of Ships and Ocean Vehicles 77786. Retrieved January 21, 2013
Kobylinski, Lech. 2007. System and Risk Approach to Ship (http://www.sciencedirect.com/science/article/pii/S0957417
Safety, with Special Emphasis of Stability. Archives of 412008810).
Civil and Mechanical Engineering 7:97106. Potthoff, Robert, and Apostolos Papanikolaou. 2016. Presentation
Kobylinski, Lech. 2008. Stability and Safety of Ships: Holistic of Benchmark Results. in SHOPERA Benchmark
and Risk Approach. Reliability & Risk Analysis: Theory & Workshop. LONDON, UK.
Applications 1(1):95105. Rahola, Jaakko. 1939. The Judging of the Stability of Ships and
Kobylinski, Lech K., and Sigismund. Kastner. 2003. Stability and the Determination of the Minimum Amount of Stability -
Safety of Ships. Elsevier. Especially Considering the Vessels Navigating Finnish
Koiter, J. R. 2006. Visualizing Inference in Bayesian Networks. Waters. Technical University of Finland.
Delft University of Technology. Retrieved Rosqvist, Tony, and Risto Tuominen. 2004. Qualification of
(http://www.kbs.twi.tudelft.nl/Publications/MSc/2006-JRK Formal Safety Assessment: An Exploratory Study. Safety
oiter-Msc.html). Science 42(2):99120.
Krata, Przemysaw, and Joanna Szapczyska. 2012. Weather Schreuder, Martin, Per Hogstrm, J. W. Ringsberg, Erland
Hazard Avoidance in Modeling Safety of Motor-Driven Johnson, and Janson C-E. 2011. A Method for Assessment
Ship for Multicriteria Weather Routing. TransNav, of the Survival Time of a Ship Damaged by Collision.
International Journal on Marine Navigation and Safety od Journal of ship research 55(2):8699.
Sea Transportation 6(1):7178. Retrieved June 8, 2016 Spanos, Dimitris, and Apostolos Papanikolaou. 2009. Benchmark
(http://www.transnav.eu/Article_Weather_Hazard_Avoidan Study on Numerical Simulation Methods for the Prediction
ce_Krata,21,336.html). of Parametric Roll of Ships in Waves. Pp. 19 in 10th
Krata, Przemysaw, Janusz Szpytko, and Adam Weintrit. 2013. International Conference on Stability of Ships and Ocean
Modelling of Ships Heeling and Rolling for the Purpose Vehicles. St. Petersburg, Russia.
of Gantry Control Improvement in the Course of Cargo Szozda, Z. 2014. Examples of Weaknesses of the 2nd Generation
Handling Operations in Sea Ports. Solid State Phenomena Intact Stability Criteria. Zeszyty Naukowe / Akademia
198:53946. Retrieved June 8, 2016 Morska w Szczecinie (nr 40 (112)):8087.
(http://www.scientific.net/SSP.198.539). Umeda, Naoya. 2013. Current Status of Second Generation Intact
Leva, M. C., P. Friis-Hansen, E. S. Ravn, and A. Lepsoe. 2006. Stability Criteria Development and Some Recent Efforts.
SAFEDOR: A Practical Approach to Model the Action of in 13th International Ship Stability Workshop.
an Officer of the Watch in Collision Scenarios. Pp. 2795 Womack, John. 2003. Small Commercial Fishing Vessel Stability
2804 in European safety and reliability conference; Safety Analysis, Where Are We Now? Where Are We Going?
and reliability for managing risk; ESREL 2006. Francis, Marine Technology 40(4):296302.
London.
Matusiak, Jerzy. 2002. Towards an Unified Theoretical Model of
Ship Dynamics. in The Maritime Research Seminar.
Sjokulla.
Matusiak, Jerzy. 2007. On Certain Types of Ship Responses
Disclosed by the Two-Stage Approach to Ship Dynamics.
Archives of Civil and Mechanical Engineering 7(4):15166.
Matusiak, Jerzy. 2011. On the Non-Linearities of Ships Restoring
and the Froude-Krylov Wave Load Part. International
Journal of Naval Architecture and Ocean Engineering
3(1):11115.
Matusiak, Jerzy, and Claus Stigler. 2012. Ship Roll Motion in
Irregular Waves During a Turning Circle Maneuver. Pp.
29198 in 11th International Conference on Stability of
Ships and Ocean Vehicles.
Montewka, Jakub et al. 2014. A Framework for Risk Assessment
for Maritime Transportation systemsA Case Study for
Open Sea Collisions Involving RoPax Vessels. Reliability
Engineering & System Safety 124:14257. Retrieved
January 9, 2014
(http://www.sciencedirect.com/science/article/pii/S0951832
013003116).
Montewka, Jakub, Floris Goerlandt, Pentti Kujala, and Mikko
Lensu. 2015. Towards Probabilistic Models for the
Prediction of a Ship Performance in Dynamic Ice. Cold
Regions Science and Technology.
Montewka, Jakub, Floris Goerlandt, and X. Zheng. 2015.
Probabilistic Meta-Models Evaluating Accidental Oil Spill
Size from Tankers. Pp. 23141 in Maritime Pollution and

139
 Table 4. Definition of scores in the SUB-matrix framework, (Goerlandt et al. 2014)
Aspect Score Interpretation
U L The assumptions made are seen as very reasonable
Much reliable data are available
There is broad agreement/consensus among experts
The phenomena involved are well understood; models used are known to
give predictions with the required accuracy
M Conditions between those characterizing low and high uncertainty
H Conditions opposite to low uncertainty
B 1 Adopted values lead to a characterization of consequences severity / risk which is
believed to strongly underestimate the true consequence severity / risk
5 Adopted values lead to a characterization of consequences severity / risk which is
believed to accurately reflecting the true value
9 Adopted values lead to a characterization of consequences severity / risk which is
believed to strongly overestimate the true consequence severity / risk
S L Large changes in base values needed to bring about altered conclusions
M Relatively large changes in base values needed to bring about altered conclusions
H Relatively small changes in base values needed to bring about altered conclusions

Fig. 2. Results of sensitivity and value of information analyses

140
Authors will not be present
at the conference
Human Factors Evaluation in Maritime Simulator-based
Assessment: A Case Study
Yisi Liu1, Xiyuan Hou1, Olga Sourina1, Dimitrios Konovessis2, Nieves Endrina
Snchez2, Gopala Krishnan3,
1
Fraunhofer IDM@NTU, Nanyang Technological University, Singapore
2
School of Mechanical and Aerospace Engineering, Nanyang Technological University, Singapore
3
Maritime Institute@Singapore Polytechnic
ABSTRACT
Human factors can be studied in maritime virtual
simulators to analyze the causes of human errors and
failures. The advantage of using maritime simulators is
that the environment is fully controllable, and different
scenarios can be virtually induced. In our work, the
virtual simulator is used to study human factors in mar-
itime. We propose tools to recognize emotion, mental
workload and stress levels from Electroencephalogram
(EEG) to monitor cadets during the simulator-based
assessment. We design a simulator-based experiment to
record EEG signals of cadets, from which we recognize
the changes of their stress levels during the task per-
formance. Here, we present one case study involving the
changes of the stress levels of two crew members when
they encounter demanding events during the tasks per-
formance. Emotion, workload and stress levels are cal-
culated from EEG recording with the time resolution 1
sec. From the data analyses results it can be seen that
the stress levels recognized from EEG correlates accu-
rately with the time of the demanding events. The mag-
nitude of the recognized stress correctly reflects the se-
verity of the situation. Different from the traditional
human factors study like use of surveys or interviews,
the proposed EEG-based method allow researchers
capture the detailed changes of stress and mental work-
load levels during maritime simulator-based assess-
ments.
INTRODUCTION
The shipping industry has been constantly improving the
design of ships and its systems, seeing large advances in
reliability, hull design, safety, comfort, propulsion and
navigational systems. Ships today consist of a fusion of
highly advanced technologies that aim to improve efficien-
cy, reliability and safety. Despite these advances, the mari-
time casualty rate still remains high. It is often quoted that
around 80% of marine casualties are caused by some form
of human error (Rothblum 2000).
While there has been plenty of research on human factors
in the aviation industry, the available literature on human
factors in the maritime industry is neither comprehensive
nor extensively studied. Only in recent years an increased
interest in this field has there been appeared, as researchers
begin to acknowledge the need for further studies in the
maritime domain (Hetherington, Flin et al. 2006).
Existing research into human factors on maritime opera-
tions usually involve statistical analysis of accident data by
categorisation of the causal factors, in-depth analysis of
specific accidents, analysis of work and safety cultures and
focus group studies (Koester 2001). Meanwhile, outside of
the maritime sector, researchers have also conducted ex-
periments in simulated environments to further analyse the
causes of human error as the simulator is fully controllable,
and different scenarios can be virtually induced.
Methods involving the measurement of bio-signals have
also been trialed in various human performance research. In
particular, tests involving the use of Electroencephalogram
(EEG) have shown high accuracies in monitoring various
human effects such as stress levels, workload and emotions.
Furthermore, the use of an EEG device allows researchers
to analyze data on a continuous basis, providing them with
much better time resolution when compared to traditional
methods of administering pre-, mid- and post-test surveys
(Berka, Levendowski et al. 2007, Lei and Roetting 2011,
Borghini, Astolfi et al. 2014).
In this paper, we use the virtual simulator to study the
human factors in maritime. We propose tools to recognize
mental workload and stress levels from Electroencephalo-
gram (EEG) to monitor cadets during the simulator-based
assessment. We design a simulator-based experiment to
record EEG signals of cadets, from which we recognize the
changes of their mental workload and stress levels during
the task performance. We present one case study of two
crew members. Stress and workload levels are calculated
from EEG recording with time resolution 1 sec. From the
data analyses results it can be seen that there is a correlation
between the EEG-based stress recognition results (in terms
of timing and magnitude) and the demanding events that are
happening in the simulator.
141
Authors will not be present
at the conference
METHODS
Device
The EEG device used for all experiments is the Emotiv
EPOC Headset (Emotiv). It is chosen due to its affordabil-
ity, convenience, comfort for the user, and hardware speci-
fications. 14 channels are available from Emotiv, namely
AF3, F7, F3, FC5, T7, P7, O1, O2, P8, T8, FC6, F4, F8,
and AF4.
Emotion Recognition
In this paper, we use the subject-dependent emotion
recognition algorithm proposed in our previous work (Liu
and Sourina 2014, Lan, Sourina et al. 2015). Higuchi fractal
dimension features (Higuchi 1988), statistical features (Pi-
card, Vyzas et al. 2001), and higher order crossings (Pet-
rantonakis and Hadjileontiadis 2010) are extracted from
five channels namely FC5, F4, F7,AF3, T7 and combined
together to form the feature space. A sliding window of 4
seconds with 75% overlapping is used in the feature extrac-
tion step. In the classification step, the Support Vector Ma-
chine with polynomial kernel is used. The value of gamma
of polynomial kernel was set to 1, coef was set to 1 and
order d was set to 5.
The validation of the proposed algorithm shows that pos-
itive and negative emotions can be recognized with a
recognition accuracy of 92.03% (Lan, Sourina et al. 2015).
Workload Recognition
The subject-dependent workload recognition algorithm
proposed in our previous work (W. L. Lim, O. Sourina et al.
2015) is used in this paper. Higuchi fractal dimension fea-
tures and statistical features are retrieved from the EEG
data. Then, these two types of features are combined and
fed into an SVM classifier with polynomial kernel. The
setting of polynomial kernel is the same with the emotion
recognition algorithm. It is shown in (W. L. Lim, O.
Sourina et al. 2015) that this combination of the features
extracted and the classifier used yields the best accuracy
among other combinations, with the reported average accu-
racy of 80.09% for 4 levels of mental workload recognition.
Stress Recognition
Stress has been shown to have significant correlation
with emotional state and workload levels (Hou, Liu et al.).
The stress level can be identified by the combination of
workload and emotion recognition results. For example,
when the subject is having positive emotion and low work-
load, the corresponding stress level is low. By expanding
the method of classifying stress used in research (Hou, Liu
et al.), 8 levels of stress are obtained using results of emo-
tions and mental workload recognition as shown in Table 1.
Table 1. Revised Mapping between Combination of Emo-
tion and Workload and Stress Levels
Emotion Workload Level Stress
Positive 0 1
Negative 0 2
Positive 1 3
142
Negative 1 4
Positive 2 5
Negative 2 6
Positive 3 7
Negative 3 8
EXPERIMENT DESIGN
We carried out a simulator-based experiment to obtain
the EEG signals of cadets. The experiment consists of 4
exercises with different difficulty levels and the detailed
procedure is introduced in the following section. The ex-
periments were conducted in the full mission ships bridge
simulators (Fig. 1). The bridge contains equipment such as
True Motion radar, Automatic Radar Plotting Aid (ARPA),
ship handling controls and electronic navigational aids dis-
play. A 180-degree field of view is provided by large mon-
itors, simulating a highly realistic environment.
Fig.1 Singapore Maritime Academy's bridge simulator.
Calibration
As the proposed recognition algorithms are sub-
ject-dependent, calibration is needed to train the classifier
before the assessment, and the trained model is used to
identify the brain states of the subjects when they perform
the exercise.
Emotion Calibration
Sound clips sourced from IADS were edited together to
form a one-minute segment, including a 15 second silence
at the beginning of the clip with the intent of allowing the
subject to relax. Two recordings were used in the calibra-
tion: first one targeted to elicit positive valence and the
other one targeted evoking negative valence. EEG data was
recorded, and a questionnaire requiring the subject to rate
his valence level is administered after the end of each re-
cording.
Workload Calibration
Stroop colour word test is administered during the EEG
calibration session, in order to reliably induce workload in
the subject. Four levels of workload are determined by four
different test sections. The first section requires the subject
to do nothing, which is considered as no workload. The
second section needs the subject to press the corresponding
key when he sees the word with the congruent color shown
on the screen. For example, press the key that represents for
Authors will not be present
at the conference
green when he saw the word green on the screen with increase to moderate levels, with the 1-minute average
green color. The third section requires the subject react to showing an increase in Fig. 2.
the color regardless of the meaning of the word on the
screen. For example, the subject should press the key that
represents green color when he saw the word red with
green color. In the last section, the subject needs to react to
incongruent case within a limited time which may stress the
subject more. After each section of the test, a short ques-
tionnaire is given in which the subject has to rate mental
workload on a scale of 1 to 9. The subject is also asked to
describe his feeling during the exercise using a word.

Experiment Procedure
Fig.2. Stress status when the subject was overtaking another
The experiment was held over two consecutive days, bulk carrier.
involving 7 subjects. Every subject needed to complete 4
exercises. The exercises were designed to be of increasing At 0:14/ 840, the subjects ship approached crossing a tug
difficulty with the first exercise being the easiest or the boat. In response, subject began to carry out aggressive
least demanding, and the last exercise being the hardest, or course corrections. A spike of stress is observed at this
the most demanding. The difficulty was influenced by the moment as shown in Fig. 3.
inclusion of timed occurrences of mechanical emergencies
or alarms, traffic conditions, and weather conditions.
Before each exercise, the briefing was held by the in-
structors about the simulation exercise. Information about
the type of ship, starting location and destination was given.
The subjects were expected to navigate their ships as per
normal with regards to specific rules and regulations per-
taining to the operational area. No other information about
the exercise was given. After the instruction the subjects
proceeded to their assigned simulator bridges, where cali-
bration of emotion and workload recognition were done.
Then, the subjects were allowed some time to chart their
course before the simulation exercise commenced. Fig. 3. Stress status when the subject began to correct
During the exercise, EEG data and video footage of each course.
subject in the bridge were recorded. Efforts to synchronize
the EEG data recording and video footage were made for At 0:15/ 900, the Subjects ship collided with a fishing
the following case study. vessel. At the same time, the steersman enters the bridge
(he was previously away for a while) and made a joke about
CASE STUDY the situation, eliciting a laugh from the subject. Instantane-
ous workload and stress levels are seen to spike before col-
Since we synchronized the EEG-based stress recognition lision happened then fluctuate between low and high as
results with the video which was taken during the exercises, shown in Fig. 4, which is possibly due to the defusing of
we are able to do a thorough study on the brain states tension by the steersman.
changes during the exercises. Two instances of collision in
the simulator showed the most visible increases in work-
load and stress levels, and the full analysis of these data is
presented in this section as a case study.
Fig. 2-4 shows the sec-by-sec analysis of EEG data rec-
orded during exercise 2, from the subject 6. In the figures,
the x axis is the sample point number and y axis is recog-
nized stress levels (1-8). Since EEG data is assessed using a
4 second moving window with a 3 second overlap, sample
point number n refers to a period of exercise from n sec-
onds to n+3 seconds. The instantaneous stress level recog-
nized from EEG is drawn as a solid red line, and also the Fig. 4. Stress status when the subject collided with a fishing
1-minute average EEG stress levels is plotted using an or- vessel.
ange dotted line in the figures.
At 0:04/ 240 (Hour: Minutes/sample point in the figure), Fig. 5-11 show the sec-by-sec analysis of EEG data rec-
which is marked on the graph via a blue line in Fig. 2, the orded during exercise 4, from subject 9.
subject was attempting to overtake another bulk carrier go-
ing in the same direction, as such, he was very attentive to At 0:07/ 420, the visibility of the simulated environment
the situation. The instantaneous stress level can be seen to was reduced and fog was at 85%. The subject ordered ap-

143
Authors will not be present
at the conference
propriate watch to be set and stress is seen to increase a
little as shown in Fig. 5.
Fig. 5. Stress status when the visibility was reduced.
At 0:09/540, the visibility was reduced further and fog
was at 95%. From Fig. 6 it can be seen that intermittent
spikes of high to moderate stress are detected. The 1-minute
stress average value is seen to begin a steady rise.
Fig. 6. Stress status when the visibility was further reduced.
At 0:10/ 600, the subject was overtaking a small vessel
and was seen to be monitoring the situation carefully. The
recognized stress levels shown in Fig. 7 are seen to spike,
and then increase to moderately high levels as subject ap-
proached the small vessel. The 1 minute average is showing
a consistent rise of stress.
Fig. 7. Stress status when the subject was overtaking a
small vessel.
At 0:17/1020, another small vessel crossed very close to
the bow of subjects ship, and the subject had ordered
maximum steering. From Fig. 8, we can see that the 1 mi-
nute average stress level increases gradually, and then,
generally maintains its levels even after the vessel crossed.
There is a very high instantaneous stress spike when in-
structor informed subject of the near miss.
Fig. 8. Stress status when there was a small vessel crossed
closely to the subjects ship.
At 0:18:30/1109, the subjects ship was on the course for
a collision with a bulk carrier and the subject was unable to
do anything as the ship was already hard to starboard. In-
144
stantaneous stress levels are consistently at the very high
levels as shown in Fig 9 and the 1 minute average stress
levels is seen to increase greatly.
Fig. 9. Stress status when the subjects ship may collide
with a bulk carrier.
At 0:20/1197, the subjects ship managed to avoid the
bulk carrier but another small vessel appeared out of the fog
and was about to have a collision. From the video recorded,
the subject can be heard commenting angrily on the situa-
tion as he tried to decide on the correct course of actions. It
can be seen from Fig. 10, after the initial avoiding the bulk
carrier, the average stress value is increasing again as the
subject identified the next threat. A long period of high
instantaneous stress levels is seen when the small vessels
appeared from the fog.
Fig. 10. Stress status when the subjects ship avoided the
bulk carrier but a new threat appeared.
At 0:20:30/1230, the subjects ship was about to collide
with small vessel and it can be heard from the video that the
subject was confirming that the ship collides. Stress is seen
to be high throughout as shown in Fig. 11, peaking at data
point no. 1268 when the collision occurs. After the collision
till the end of the exercise, the ship was damaged and tilting
to one side, the subject ran through standard emergency
collision procedures and his stress levels remains high for
the rest of the exercise as shown in Fig. 11.
Fig. 11. Stress status when the collision with a small vessel
happened.
CONCLUSION
In this paper, we present a case study done in virtual
simulator based maritime experiment. The EEG signals and
video of the exercise of the cadet in the simulator was rec-
orded. Stress levels of the cadet during the exercise were
recognized. The results show that the stress level recog-
nized from EEG correlates accurately with the timing of
demanding events happened in the experiment. Additional-
ly, the magnitude of the recognized stress correctly reflects
Authors will not be present
at the conference
the severity of the situation. For example, very high levels
of sustained stress are assessed when a collision occurs.
Similarly, a moderate level of stress is assessed when mild-
ly demanding situations such as changing weather condi-
tions occurs.
The preliminary results confirmed that the proposed
EEG-based algorithms can be used in virtual maritime sim-
ulators for human factor study in maritime research. The
correlation between brain states (emotion, workload, stress)
and the performance of cadets can be studied to reveal the
reason of human errors in maritime. It can also be used as
an additional assessment of the cadets besides the perfor-
mance in simulator to decide whether the cadets are
well-trained for the sea service. In the future, the pattern of
optimal stress and mental workload levels during different
task performing should be studied with more subjects.
ACKNOWLEDGEMENTS
This research is supported by Singapore Maritime Insti-
tute, and by the National Research Foundation, Prime Min-
isters Office, Singapore under its international Research
Centres in Singapore Funding Initiative.
We would like to acknowledge the final year project
students of School of MAE of Nanyang Technological
University and personally Gnin Yeow Dong for their assis-
tance in this work.
Workplace. Human Factors and Ergonomics Society
Europe Chapter Annual Meeting.
Lan, Z., Sourina, O., Wang, L. and Liu, Y. (2015).
"Real-time EEG-based emotion monitoring using stable
features." The Visual Computer, pp.1-12.
Lei, S. and Roetting, M. (2011). "Influence of task
combination on EEG spectrum modulation for driver
workload estimation." Human Factors: The Journal of
the Human Factors and Ergonomics Society No.53(2),
pp.168-179.
Lim, W.L., Sourina, O., Wang, L. and Liu, Y. (2015).
EEG-based Mental Workload Recognition Related to
Multitasking. Proc. Int. Conf. on Information,
Communications and Signal Processing (ICICS).
Liu, Y. and Sourina, O. (2014). "Real-time subject-
dependent EEG-based emotion recognition Algorithm."
Trans. Computational Science XII, Lecture Notes in
Computer Science XII, in print.
Petrantonakis, P.C. and Hadjileontiadis, L.J. (2010).
"Emotion recognition from EEG using higher order
crossings." IEEE Trans. on Information Technology in
Biomedicine No.14(2), pp.186-197.

REFERENCES Picard, R.W., Vyzas, E. and Healey, J. (2001). "Toward

Berka, C., Levendowski, D.J., Lumicao, M.N., Yau, A.,
Davis, G., Zivkovic, V.T., Olmstead, R.E., Tremoulet,
P.D., and Craven, P.L. (2007). "EEG correlates of task
engagement and mental workload in vigilance, learning,
and memory tasks." Aviation, Space, and Environmental
Medicine No.78(Supplement 1), pp. B231-B244.
machine emotional intelligence: Analysis of affective
physiological state." IEEE Trans. on Pattern Analysis
and Machine Intelligence No.23(10), pp.1175-1191.
Rothblum, A.M. (2000). Human error and marine safety.
National Safety Council Congress and Expo, Orlando,
Florida.

Borghini, G., Astolfi, L., Vecchiato, G., Mattia, D. and

Babiloni, F. (2014). "Measuring neurophysiological
signals in aircraft pilots and car drivers for the
assessment of mental workload, fatigue and drowsiness."
Neuroscience & Biobehavioral Reviews No.44, pp.58-75.

Emotiv. "http://www.emotiv.com." from

http://www.emotiv.com.

Hetherington, C., Flin, R. and Mearns, K. (2006). "Safety in

shipping: The human element." J. Safety Research
No.37(4), pp. 401-411.

Higuchi, T. (1988). "Approach to an irregular time series on

the basis of the fractal theory." Physica D: Nonlinear
Phenomena No.31(2), pp.277-283.

Hou, X., Liu, Y., Sourina, O. and Mueller-Wittig, W.

"CogniMeter: EEG-based Emotion, Mental Workload
and Stress Visual Monitoring." Trans. on Computational
Science, Lecture Notes in Computer Science
No.accepted.

Koester, T. (2001). Human factors and everyday routine in

the maritime work domain. Human Factors in
Transportation, Communication, Health, and the

145
 Design for Safety in a Competitiveness Perspective

Jose Jorge Garcia Agis1,2, Ulrikke Brask Brandt1, Per Olaf Brett1,2
1, Ulstein International AS, 6065 Ulsteinvik, Norway
2. Norwegian University of Science and Technology, 7491 Trondheim, Norway

ABSTRACT way or in the right place, reflecting recent years

This paper discusses the current practice of handling
uncertainties and risks by the use of extra-margins in
offshore service vessel design and construction and its
consequences for design complexity and cost - what is
the right balance between an adequate safety level and
cost-effectiveness from a competitiveness standpoint?
The current market situation characterized by a lower
oil price, has strongly urged owners and designers of
offshore vessels to contribute to the improvement of the
overall competitiveness level of the offshore oil & gas
(O&G) industry. A reasonable solution-proposition is
therefore, to critically start questioning the setting of
and evaluating the consequences and implications of
such safety factors and uncertainty margins within the
ship design and shipbuilding processes. It is argued in
this paper that as a consequence of the golden era the
trend in design and construction strategies has become:
More is better, where margins have been added to
prepare for almost any uncertainties and eventualities.
Safety and uncertainty margin setting has been one of
the most traditional and common ways of catering for
uncertainties in the design and construction of ships.
Such safety and uncertainty margins are set to handle
both the absolute level of, for example, safety as well as
compensating for uncertainties in the analysis of and
assumptions of boundary conditions given for such
complex design and construction tasks.
The safety requirements from classification societies
and IMO have in most cases and over the years
increased in order to improve the absolute safety level
onboard vessels. In some cases it has also been
introduced to cater for uncertainties in the
pre-conditions for analyses and work practices. But
what is really a good enough safety level? How much
uncertainty and eventualities should be catered for? Is
the industry willing to pay for reaching a zero
accidents level of safety? Higher accuracy of analyses
and higher fidelity of boundary conditions should
perhaps lead to a reduction in such uncertainty margin
practice. Are the safety margins applied in the correct
accident statistics? Introducing a new competitive
design strategy for offshore service vessels where Good
is good enough is argued in this paper. This paper
readdresses the topic and critically scrutinize current
industry practices and suggests new avenues for more
effective vessel design solutions.
INTRODUCTION
Since the significant Brent oil price drop through the
second half of 2014, the headlines in the news have been
dominated by vessel sales and lay ups, workforce reduction,
company merges and bankruptcies, among others. These
effects, initiated by cost cutting and investment stop of oil
companies have trickled down in the value chain.
Therefore, the offshore oil and gas (O&G) industry is now
looking to achieve a higher competitiveness level by
developing more effective solutions in a difficult market
situation where all the stakeholders should contribute.
The oil price and consequently lower Exploration &
Production spending directly and indirectly affects the
design strategy and methodology for offshore vessel
(Ulstein and Brett 2015). After the millennium, offshore
service vessels (OSV) have become more complex and with
higher design requirements both from the designer himself,
the shipowner, the classification society, etc. Vessels and
platforms have been designed with a more is better
strategy, and to a large extent, neglecting commercial and
operational consequences (Garcia, Brandt, and Brett 2016).
But, is a larger and more advanced vessel, with more class
notations, more complex topside equipment, a better
vessel? This view is discussed by Ulstein and Brett (2015),
where it is argued that the definition of a better vessel
should be based on a holistic view of the vessel design
solution and not on a single factor bases or a Design for
X approach. The more is better strategy, together with
stronger technical, operational and documentation
requirements, have affected the cost level of the vessels and,
thereby, typically lowered vessel designs and industry
competitiveness, as demonstrated by Garcia, Brandt and
Brett (Garcia, Brandt, et al. 2016).

146
The offshore O&G industry has experienced a significant
development of novel, innovative and advanced vessel
design solutions over the last 10-15 years. Although
primarily supported by a strong oil price, some other factors
also have influenced this design trend. The need for new
vessels to cover more demanding operations (deeper
waters, artic operations, etc.), supported by growing day
rates and the need for market differentiation, as well as the
introduction of new Rules and Regulations have continually
contributed to the establishment of higher cost vessel
design solution practice.
How can, therefore, the industry develops more effective
vessel design solutions without compromising smartness,
greenness and safety?
Three major costs drive the daily operational cost of a ship
(known as break even rate): manning, fuel and capital
costs (Stopford 2009). In order to make profit in an OSV
market with lower day rates, a significant break even rate
reduction is needed. Hence, making changes on one or
more of these three factors are needed.
One tendency has been to reduce manning cost by means of
contracting less qualified personnel while investing in
technologically advanced vessels (Mallam, Lundh, and
Mackinnon 2015). As a result, we have seen vessels being
detained by port authorities finding that the crew was
unable to properly operate some of the most complex
apparatus and automation and control systems onboard
the vessel (GCaptain 2016).
On the other hand, some ship owners have chosen the use
liquefied natural gas (LNG) as fuel, in order to be prepared
for future emission limitations and as a cost advantage to
marine diesel (MDO). Unfortunately, the recent significant
drop in the oil price has left them with higher fixed capital
costs (CAPEX) and a less attractive advantage regarding
cost differentiation between LNG and MDO.
These examples demonstrate the need for taking a holistic
view and a proper compromise among all the influencing
factors as well as a proper analysis of uncertainty factors
and their influences. This paper focuses principally on
aspects related directly with capital costs (CAPEX), but
also influencing indirectly operational (OPEX) and voyage
(VOYEX) costs.
Some reactions to these developments have already been
presented and discussed within the maritime industry. For
example, the European Boat Design Manifesto Group
(McCartan et al. 2014) argues the need of considering
human performance as a design parameter, in a merchant
fleet that has become more complex and automated, with
reduced manning and higher performance and safety
requirements.
SHIPPING, A RISKY INDUSTRY
Shipping industry, especially the offshore vessel industry, is
characterised by the involvement of multi-million (USD)
assets, the vessels. In addition, ships are designed to
transport cargoes or perform special operations, being those
in some cases even more valuable than the particular vessel
or being potentially polluting the environment. Finally, but
not less important, ships are operated by people and for
people, which makes safety a basic requirement for
practicing best practise shipping operations.
At the same time, ship design is considered a complex
engineering activity (van Bruinessen, Smulders, and
Hopman 2013; Gaspar, Erikstad, and Ross 2012; Jain,
Pruyn, and Hopman 2015). It requires a multidisciplinary
consideration in arriving at relevant design objectives and
constraints (Deb et al. 2015), imposed by the stakeholders
involved, the market needs and conditions prevailing at any
one point in time and over the ship lifetime (Gaspar et al.
2015; Ulstein and Brett 2012). This design complexity
environment and boundary condition involves a high degree
of uncertainty, both technical, commercial and operational
(Garcia, Pettersen, et al. 2016); that induce extra risk, if not
properly handled.
Three major groups of risks are identified affecting the
shipping industry:
- Technical: Regarding the quality and validity of
the technical solution.
- Operational: Regarding the safety of crew and
personnel and pollution of the sea.
- Commercial: Regarding the achievement of
stakeholders expectations and objectives.
Looking at this, one could say that the shipping industry,
and more in general ocean engineering systems, is a risky
industry with a considerable degree of uncertainty
related to key aspects of the systems future operating
context (Erikstad and Rehn 2015).
Historically, the responsibility of holding shipping risks
under proper control has been distributed to different
entities. Classification societies covering technical aspects,
by certifying that the ship has been designed and built
according to their standards. Shipowners hold the
commercial and operational risk, having to pay back
interests and loan to banks or private bonds. Ship designer,
shipyard and suppliers hold the responsibility of ensuring
that their products perform as initially agreed upon and
contracted.
According to the International Association of Classification
Societies (IACS), in addition to the classification societies,
other entities with a responsibility for or interest in
promoting maritime safety include shipowners, shipbuilders,
flag State administrations, port State control authorities,
underwriters, shipping financiers, charterers, and, of course,
seafarers (IACS 2011).
Stakeholders, in order to reduce and compensate for the
associated risks that they have to assume, implement
margins in the final vessel design solution. A description of
those margins and the practicing of them is discussed in the
next section.
147
DESIGN MARGINS
Ship design development is a complex decision-making
process where the designer has to make compromises
among the design variables to come up with a fully
operational system over its lifetime (Ulstein and Brett
2015). There always exists significant uncertainty in the
ship design process and in the future lifecycle of that
particular vessel. Designers and owners should keep in
mind that the vessel has to retain acceptable levels of
performance over its entire life, as needs and expectations
change, and the technical capabilities deteriorate (Garcia,
Pettersen, et al. 2016).
In order to ensure at least a minimum level of system
performance though the operational life-cycle, the shipping
industry has historically appealed to the use of margins
(Meyer 2002) (also denominated safety factors, design
margins or safety margins). Uncontrolled use of margins,
in order to ensure the validity of the results, could lead to
noncompetitive vessel designs. For example, excessive hull
strength will increase the weight of the vessel, hence
reducing the performance and increasing the costs (Garcia,
Pettersen, et al. 2016). Vrijdag, de Jong and van Nuland
(Vrijdag, de Jong, and van Nuland 2013) describe the use
of margins in the calculation of the bollard pull in tugs. The
uncertainty related with the final performance and the
penalties related to such a parameter, requires the use of
margins, in order to ensure the contractual bollard pull.
However, how much uncertainty margins are good enough?
Margins are presently in designers every day job. Either
explicitly in their calculations and decisions or implicitly
when applying rules and regulations or considering the
information provided by the supplier. Every stakeholder in
the value chain introduces their own margins with different
purposes (Meyer 2002):
- Account for uncertainties in the design methods
(principally in the earlier stages of the design);
- Assure safe operation even after degradation of the
structure or individual components during the
ships lifetime;
- Assure performance even if operating conditions
are slightly different from those which the ship
was designed for;
- Allow for future changes;
The importance of properly applied and assessed margins in
ship design was early stated by the US Navy. In the mid 70,
the studies of Hockberger and Gale (among others)
addressed the use of margins and their consequences (Gale
1976; William A Hockberger 1976). Hockberger classifies
ship design margins in three categories: Design and
Construction Margins, Assurance Margins and Future
Growth Margins. Since this categorization was developed
for Naval Ships, the authors of this paper, propose a new
categorization of margins in commercial merchant and
offshore ship design:
Design and Construction Margins: Margins introduced in
order to treat the uncertainty present in the early stages of
the design process. A typical example could be the steel
weight or the speed performance of the vessel.
148
Life-cycle Margins: Margins introduced in order to
account for future events and degradation of the systems. A
typical example are the sea margins.
Market Margins: Margins introduced for commercial
purposes. A common practice in order to increase the
second hand value or as differentiating factor over the
vessels life time. The implementation of such margins
should pursue a commercial purpose and utility value. A
typical example could be an increased DP (dynamic
positioning) capability or an excess use of steel in an
offshore vessel to extend its lifetime and overall robustness
in use.
The lower oil price and the need for developing more
effective offshore vessel design solutions has brought to the
table the need of properly assessment of those design
margins.
Fig. 1 Stakeholders adding margins in the vessel design (The
graph does not represent how much, just who is influencing)
More or less, all stakeholders include margins to protect
and secure their objectives and utilities of the vessel, see
Figure 1. The designer will include margins in order to be
sure the vessel will fulfil the statements given in the
contract by good margin such that legal obligations are
properly met. The suppliers have also introduced margins in
order to ensure their product performance, e.g. steel plates
thicknesses or propeller effects to fulfil their promised
capacities and qualities. Maybe, the specified steel plate is
out of stock and the yard will therefore use a 0,5mm thicker
plate. These examples, belong to the Design and
Construction Margin category. As examples of Life-cycle
margins, could be the shipowner requirements. The
shipowner does not know the operations the vessel will
perform through its whole lifecycle and therefore, adds
additional functional requirements, in order make the vessel
more flexible and be better prepared for the future needs
and eventualities. The implementation of additional class
notations is also a good example of such an approach and
uncertainty margin practice. The consequence is that the
final product is delivered with margins on margins and
thereby the cost level, light ship weight, etc are increased
unnecessarily.
MARGIN SETTING PRACTICES enclosed and isolate fire, among others. Due to the
importance of the hull structure and its integrity, and its
The designers of OSVs like most other segments of the relevance to the cost and operation of the vessel, a
maritime industry are confined by rules and regulations description of how margins are built up is elaborated upon.
defined by among other sources, the International Maritime
Organization (IMO), flag states and classification societies. Class rules define the minimum requirements for the
In addition, oil companies, charterers, operators and owners structural strength of the hull. Rules and regulations are
will include respectively their own constrains and developed based on empirical observation, service
requirements; based on their experience (or novelty) in the experience and research and they already include set
sector, market tradicion or speculatively, among others. margins. The philosophy of the class rules can be described
by the sketch in Figure 4. A new delivered vessel will have
an initial level of margins built into the vessel to make the
distance to the ultimate limit strength large enough. In the
first years in operation, little trouble and surprises are
expected w.r.t. hull structure failures or structural capacity
deterioration. Typically, by the third or fourth periodical
survey (15 and 20 years) the vessels structural capacity will
be partly upgraded but never to its original level. This
Fig. 2 Schematic representation of margin setting practices cycle continues through the design life-time of the vessel.
At the end of the design life time it can be discussed if the
The philosophy and dynamics of the market (product of remaining structural capacity of the vessel still has a too big
stakeholders behaviour) has been the primary cause of of a margin to the ultimate strength of the material/structure.
margin additionality. The more is better strategy has Independent of the margins size at the end of the vessel life
motivated shipowners and designers to invest in time, this philosophy illustrates that the class rules include
state-of-the-art vessels, whereby having more, even if not margins over the vessel life time.
needed, was generously compensated for in the time charter
rate. As example, the large difference between dayrates for Meanwhile the ship owner will include margins in the form
medium and large size AHTS (Fig. 3), which originally of additional class notation, extra capacities and capabilities
led to a rise of the contracting activity for +18k BHP and/or increased minimum plate thickness. If the they are
vessels. Consequently, the market has seen vessels built added because they are only nice to have and it is
with Ice class to have more steel and vessels with the unknown if the vessel over its life time will operated such
highest level of dynamic positioning capability because that the extra capabilities are relevant, they will represent a
the other vessels have it. Todays market does not accept loss of competitiveness compared to a lower specified
such a luxary and is definitely not properly compensated for vessel.
(Dahle and Kvalsvik 2016) and a new era of medium When the designer has received the shipowners requirement,
specified have led to more cost-effective solutions at they like to design the best solution fulfilling these
present. requirements. In the OSV segment no standard loading
condition requirements are set or specified Hence, no or
28K 220

200
only maximum (full flexibility/utility of the vessel) limits
Dayrate difference between +18k BHP and 13-18k BHP (GBP)

26K

24K
for loadings are defined. This is very different compared to
180
22K
for example, the loading condition practice of bulk carriers.
160
Therefore, OSVs typical are designed with higher hogging
Adjusted Brent Oil price (USD)

20K

18K 140 and sagging moments than defined by the Hulls Structural
16K 120 Design rules. Consequently, the steel weight is increased
14K
100
and strictly speaking higher than they have to be if more
12K
80
specific loading conditions were specified in advance of
10K
vessel design process.
8K 60

6K
40
4K
The steel manufactures have also included margins in order
2K
20 to secure the level of ultimate strength and the thickness of
0K
1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014
0
their rolled plates. The intention with this example is to
Year
illustrate how margins are adding on top of each other in
Fig. 3 Comparison of Brent oil price and dayrate difference the design and building phase of a vessel. The hull is
between medium and large size AHTS approximately 25% of the vessel cost (typically for OSVs)
and by reducing these margins the final cost of the vessel
will be reduced. However, a balance between margins,
Example of Structural Margins safety, maintenance, vessel capability and cost level must
be carefully balanced w.r.t. competitiveness level.
The hull is the most fundamental part of the vessel. Without
the hull the vessel would not float and therefore could not
perform the operation that it is intended for. In addition, the
hull structure is responsible for keeping cargo, the offshore
operation intact and people safe, polluting substances

149

Fig. 4 Sketch of the philosophy behind class rules
DESIGN FOR SAFETY
IMO, classification societies and flag states establish safety
and pollution prevention rules and regulations for ships.
IMO is mainly concerned with the environment and the
safety of the main ship functions and safety of equipment,
whilst the classification societies address the major
systems and equipment safety level, optional sub systems
and components, and defines detailed hull rule
requirements (Han 2004). The class societies can be defined
as the technical advisors to the maritime regulations
authorities (Stopford, 2009). The primarily purpose of IMO
is to improve the safety in maritime operations at
international level, as is expressed in their slogan: Safe,
secure and efficient shipping on clean oceans (IMO 2013).
The role of the classification societies is to approve the
design solution of the vessels, ensure that they are built as
designed and approved and that their upkeep is retrained
according to agreement over the vessel life time. Thus, the
vessel can be insured and a construction certificate to be
issued, which entail that the requirements are met in order
to reduce risk of failure and consequences if still failures
occur. Technical standards are defined by each
classification society in their own right and through IACS
collaboration. The societies are paid for their service, but
have no legal powers of enforcement beyond withdrawing
their service.
It is important to mention that rules and regulations are not
only defined for the design and construction phase, but also
subjects as maintenance standards and surveys, crewing
costs, employment conditions, taxation etc. are linked to
regulations.
The reason to introduce new rules or modify the existing
ones varies. The rule demand may be a political decision or
often an expectation from industry actors and by the society
at large when a serious accident has occurred. The incidents
with Titanic, Torrey Canyon, Herald of Free Enterprise,
Exxon Valdez are all examples on incidents which resulted
in new regulations. For instance, the first version of SOLAS
Convention (The international Convention for Safety of
Life at Sea) was adopted in 1914 in response to the Titanic
disaster.
150
Vassalos and Papanikolaou describe how traditional
prescriptive rule-based approaches do not handle properly
safety when developing innovate solutions, due to their
experience-based characterization (Vassalos and
Papanikolaou 2015). For that reason, in the early 2000s,
the need for changing the way safety was being dealt with
in shipping, proliferated the consideration of the marine
industry as a risk industry. This categorization led to the
need of adopting risk-based approaches. The growing
Design for Safety philosophy, led to the development of a
formal state-of-the-art design methodology, Risk-Based
Design (RBD) (Papanikolaou et al. 2009).
Fig. 5 Risk-Based Design (Papanikolaou et al. 2009)
The RBD methodology changed the way safety was being
treated, considering it as a design objective rather than a
constrain (Papanikolaou et al. 2009), see Figure 5. RBD as
goal-based design approach (Andrews and Erikstad 2015),
consists in quantify the life-cycle risk of a vessel during the
concept design development by considering passive
(design) and active (operational) safety measures
(Papanikolaou et al. 2009). In its implementation has been
pioneer the International Maritime Organization,
introducing Goal-Based Standards and probabilistic
assessment methods (Andrews and Erikstad 2015;
Papanikolaou et al. 2009; Vassalos and Papanikolaou
2015).
This methodology opens the possibility of considering
safety in a holistic view. A proper compromise between
safety and effectiveness has to be done, in order to achieve
more competitive vessel design solutions with a good
enough safety level.
ACCIDENT DATA
The principal purpose of rules and regulations is to improve
and promote maritime safety operation, including both of
human life and environment protection. In this section we
review the historical evolution of accident data with focus
on OSVs, as presented in Figure 6. In the last decade the
accident level has been around 0,5-1%. A clear positive
accident reduction is predicted from Figure 6, where in
2014 the industry has achieved minimum historic values,
reducing to 1/3 the statistics from late 70s and early 80s.
 Table 1 Casualties in the period of 1978-2015 per vessel rules and regulations have been continuallt introduced,
type (IHS Fairplay 2016) which at the same time have constrained the
Frequency
competitiveness retaintion of offshore vessel designs.
Vessel type Dominant casualty (%)
(Casualties*/vessel years**)
Accommodation Ship 2,57 % Hull/Mchy. Damage (45%) New calculation tools, improved building processes and
Anchor Handling Tug Ship 0,73 % Fire/Explosion (22%)
Cable Layer 2,07 % Fire/Explosion (30%) rules and regulations have not led to a real reduction of
Diving Support Vessel 1,06 % Fire/Explosion (25%) safety and uncertainty margins in offshore support/service
Offshore ships

Drilling Ship 0,65 % Fire/Explosion (28%)

Offshore Support Vessel
Offshore Tug/Supply Ship
Pipe Carrier
Pipe Layer
Platform Supply Vessel
Seismic Survey Vessel
Standby Safety Vessel
Well Stimulation Vessel
Bulk Carrier
Chemical/Products Tanker
Merchant ships
0,97 % Hull/Mchy. Damage (37%)
vessel design. So, what is the acceptable lowered level of
0,66 % Collision (19%) safety and for how long is the industry willing to pay for a
1,32 % Collision (27%)
0,89 % Fire/Explosion (38%)
continued high margin?
0,45 % Collision (20%)
1,07 % Fire/Explosion (36%) A goal-based approach moving the regulatory framework
1,43 % Hull/Mchy. Damage (48%)
1,37 % Fire/Explosion (40%) from a culture of compliance, governed by Rules, to a
3,67 % Hull/Mchy. Damage (31%) culture of benchmarking, backed by functional first
5,61 % Hull/Mchy. Damage (38%)
principle risk-based approach is considered the way

Cellular Container Ship 2,75 % Collision (34%)

Crude Oil Tanker 8,85 % Hull/Mchy. Damage (47%) forward. By establishing clear, demonstrable and verifiable
General Cargo Ship 3,50 % Hull/Mchy. Damage (30%)
LPG Tanker 4,63 % Hull/Mchy. Damage (43%)
goals based on what the vessel MUST have and those that
Passenger/Cruise 4,19 % Hull/Mchy. Damage (36%) are NICE to have, the ship designer could achieve better
Passenger/Ro-Ro Ship 3,08 % Hull/Mchy. Damage (36%) alternative designs to todays solutions with equivalent or
Products Tanker 3,86 % Hull/Mchy. Damage (34%)
Tug 0,38 % Collision (28%) higher levels of safety but also higher levels of
competitiveness. In addition, it would promote the
* All casualties, including fatalities /total loss implementation of new technology and greater innovation
** Considering 351 days per year and the utilization factor of the fleet
within the shipping/offshore industry.
The number of accidents for different types of OSVs is in
CONCLUSIONS
the range of 0,5% to 2,6% being the dominating types
Fire/Explosion, Collision and Hull/machinery damage,
respectively, see Table 1. Comparing with the merchant The design process of a new vessel design involves
fleet where the frequency of casualties is typically around uncertainty. Nobody, can easily predict or foresee all the
3,0-4,0%, the frequency of casualty for OSV is barely half. operations or situations the vessel might experience over its
Tugs have the lowest frequency (0,4%) while crude oil life time. Today, the vessel design and construction
tankers have the highest frequency (8,9%), see Table 1. For industry practices margins to cater for such uncertainties.
merchant vessels, the dominating types of casualties are However, this often result in an expensive complex and
Hull/machinery damage and collision, respectively. advanced multi-functional vessel, which easily could tip
over to become a multi-useless vessel; due to its higher
costs and thereby reduced competitiveness. From a safety
perspective, margins are added in order to reduce the risk of
undesirable incidents and accidents. These margins are
principally set by IMO and classifications societies through
rules and regulations, although other stakeholders (designer,
supplier, yard, port authority etc.) also contribute. Today,
the frequency of accidents for OSVs is below 0,5% so the
question is: Is the industry willing to continue paying for a
such uncertainty margins and at the same time be penalized
on their vessels competitiveness.

REFERENCES

Andrews, D. and Erikstad, S.O. 2015. State of the

Fig. 6 Registered accidents for OSVs in the period 1978 Art Report on Design Methodology. Pp.
2015 compared to the global operating OSV fleet, (IHS 89105 in 12th Int. Marine Design Conf.
Fairplay 2016) (IMDC), vol. 1. Tokyo, Japan.
Bruinessen, T.v, Smulders, F., and Hopman, H. 2013.
Towards a Different View on Ship Design - The
DISCUSSION
Development of Ships Observed Though a
Social-Technological Perspective. in 20th Int.
Looking at the historical evolution of accidents in the OSV
fleet (Fig. 6), one can not find a strong reason to not Product Management Conf. Paris, France.
scruitinise todays safety margin practices and critically Dahle, A. and Kvalsvik, D.J. 2016. Freight Rate
define a new set of standars and margin practice. Despite Determinants in the Offshore Market: Does
the fact that accidents statistics for OSVs havereduced to a Energy Efficiency Pay? Norwegian School of
1/3 of the level of accidents of previous years it is Economics.
somewhat surprising to notice that more and more strict Deb, K., Lu, Z.C., McKesson, C.B., Trumbach, C.C.,

151
 and Decan, L. 2015. Optimal Ship Design and (http://dx.doi.org/10.1016/j.ergon.2015.09.007).
Valuable Knowledge Discovery Under McCartan, S. et al. 2014. European Boat Design
Uncertain Conditions. Innovation Group: The Marine Design
Erikstad, S.O., and Rehn, C.F. 2015. Handling Manifesto. in Marine Design. Coventry, UK:
Uncertainty in Marine Systems Design - RINA.
State-of-the-Art and Need for Research. Pp. Meyer, J. 2002. A Risk-Based Approach to Optimal
32442 in 12th Int. Marine Design Conf. Margins in Ship Design. Massachusetts
(IMDC), vol. 2. Institute of Technology.
Gale, P.A. 1976. Margins in Naval Surface Ship Papanikolaou, A. et al. 2009. State of the Art Report
Design. Naval Engineers J. 87(2), pp.17488. on Design for X. in 10th Int. Marine Design
Garcia, J.J., Brandt, U.B., and Brett, P.O. 2016. Conf. (IMDC). Trondheim, Norway.
Unintentional Consequences of the Golden Stopford, M. 2009. Maritime Economics. Abingdon,
Era of the Offshore Oil & Gas Industry. in Int. UK: Taylor & Francis.
Conf. on Ships and Offshore Structures. Ulstein, T. and Brett, P.O. 2012. Critical Systems
Hamburg, Germany. Thinking in Ship Design Approaches. in Int.
Garcia, J.J., Pettersen, S.S., Rehn, C.F., and eMarine Design Conf. (IMDC). Glasgow- UK.
Ebrahimi, A. 2016. Handling Commercial, Ulstein, T. and Brett, P.O. 2015. What Is a Better
Operational and Technical Uncertainty in Ship? It All Depends . in Int. Marine
Early Stage Offshore Ship Design. in Conf. Design Conf. (IMDC).
System of Systems Engineering. Kongsberg, Vassalos, D. and Papanikolaou, A. 2015. State of
Norway. the Art Report on Design for Safety,
Gaspar, H.M., Brett, P.O., Erikstad, S.O., and Ross, Risk-Based Design, Life-Cycle Risk
A.M. 2015. Quantifying Value Robustness of Management. in Int. Marine Design Conf.
OSV Designs Taking into Consideration (IMDC). Tokyo, Japan.
Medium to Long Term Stakeholders Arthur, V., Jong, J.de and Nuland, H.v. 2013.
Expectations. in 12th Int. Marine Design Conf. Uncertainty in Bollard Pull Predictions. pp.
(IMDC). 44753 in 3rd Int. Symp. Marine Propulsors.
Gaspar, H.M., Erikstad, S.O.,and Ross, A.M. 2012. Tasmania, Australia.
Handling Temporal Complexity in the Design Hockberger, W.A. 1976. Ship Design Margins -
of Non-Transport Ships Using Epoch-Era Issues and Impacts. Naval Engineers Journal
Analysis. Trans. RINA Part A: Int. J. Maritime 88(2):15770.
Engineering 154(3), pp.10920.
GCaptain. 2016. gCaptain Newsletter. Australia
Finds Detained Bulk Carriers Crew Was
Unable to Operate the Ships ECDIS.
Retrieved 27.5.2016
(http://gcaptain.com/australia-finds-detained-b
ulk-carriers-crew-was-unable-to-operate-the-s
hips-ecdis/).
Han, G.Y.. 2004. Ship Design Rules and
Regulations. in Rogue Waves Workshop. Brest,
France.
IACS. 2011. CLASSIFICATION SOCIETIES -
WHAT, WHY and HOW? London, United
Kingdom.
IHS Fairplay. 2016. IHS Maritime World Register
of Ships.
IMO. 2013. IMO What It Is. London, UK. Retrieved
(http://www.imo.org/en/About/Documents/Wha
t it is Oct 2013_Web.pdf).
Jain, K.P., Pruyn, J.F.J. and Hopman, J.J. 2015.
Influence of Ship Design on Ship Recycling.
pp. 26976 in Maritime Technology and
Engineering. London, UK.
Mallam, S.C., Monica Lundh, and Scott N.
Mackinnon. 2015. Integrating Human Factors
& Ergonomics in Large-Scale Engineering
Projects: Investigating a Practical Approach
for Ship Design. Int. J. Industrial Ergonomics
50, pp.6272. Retrieved

152
 Safety Level Required by the IMO Second Generation
Intact Stability Criteria for Ships under Dead Ship
Conditions and Parametric Roll Resonance
Naho Yamashita, Naoya Umeda, Masahiro Sakai
Department of Naval Architecture and Ocean Engineering, Osaka University,
umeda@naoe.eng.osaka-u.ac.jp

ABSTRACT beyond the limitations of the formulae of the weather crite-

At IMO (International Maritime Organization), the
second generation intact stability criteria are now un-
der development. They consist of three-levelled criteria
for major failure modes including synchronous reso-
nance under dead ship condition and parametric rolling.
For stability under the dead ship condition, the current
IMO Intact Stability Code already includes the weather
criterion, which is based on a simplified roll model in
irregular beam wind and waves. The level 2 criterion of
the second generation intact stability criteria for the
dead ship conditions also deals with capsizing in irregu-
lar beam wind and waves, but in a more sophisticated
manner. The authors compared in systematic calcula-
tions for four ships the weather criterion and the new
criterion. The results of the probabilistic safety level in
the North Atlantic indicate that the stability index used
in the weather criterion is more reasonable than the
metacentric height. This is because the weather criteri-
on includes the dynamics, but hydrostatic restoring
curves do not. Effects of GM and draught on the safety
level of stability under the dead ship condition and
parametric roll resonance are further discussed through
numerical calculations.
INTRODUCTION
IMO (International Maritime Organization) adopted the
weather criterion as Res A. 562 in 1985 for guaranteeing
intact stability for ships in severe wind and waves without
forward velocity. This semi-empirical criterion is based on
a domestic stability regulation established in 1958 in Japan
for passenger ships. Passenger ship designs have drastically
changed since the great success of the M.S. Sovereign of
the Seas in 1988, which was larger than the R.M.S. Queen
Elizabeth 2 known as one of the last North Atlantic pas-
senger liners. Then the size of cruise ships increases year by
year. Because of their large superstructure accommodating
many passengers, the breadth to draught ratio and
above-water height of ship gravitational centre increase
rion. As a result, cruise ship designers claimed that large
cruise ships designed with the weather criterion could be
excessively safe.
In response, IMO started to revise the weather criterion
in 2002. After long discussions and many sample calcula-
tions, IMO concluded to keep the weather criterion as it
was, but also to allow the use of model tests for estimating
aero- and hydrodynamic parameters as alternatives to em-
pirical estimation methods with a slightly extended wave
steepness table. The weather criterion was tuned with casu-
alty statistics for finalising the mean wind velocity so that
even partial revision could require recollection of casualty
statistics. The weather criterion based on a beam-sea roll
model was tuned with capsizing accidents due to not only
beam seas but also following seas. However, because of the
stringent weather criterion, there has been no major intact
stability accident for large cruise ships. Thus we cannot use
the same semi-empirical approach at the development of
the current weather criterion.
The answer from the IMO is to develop physics-based
intact stability criteria. In other words, we will tune intact
stability criteria without accident data but with scientific
evidence. These are the second generation intact stability
criteria. They deal with not only the dead ship situation (i.e.
synchronous rolling in beam wind and waves without for-
ward velocity), which the weather criterion assumed for its
modelling, but also parametric roll, pure loss of stability,
broaching and excessive acceleration. In order to avoid
frequent use of advanced numerical simulations for failure
probability in irregular wind and waves, simplified but still
physics-based criteria, which are named as vulnerability
criteria, for each mode were drafted by January 2016 and
will be finalised soon (Umeda and Francescutto, 2016). A
ship complying with the simplified criteria for a certain
mode is judged as safe for that mode.
The second generation intact stability criteria will be
substituted into the current weather criterion in the long run.
Thus it is important to evaluate the safety level obtained by
the new criteria, as the new criteria could relax the re-
quirements of the current weather criterion. With quantita-
tive data, we can make final approval of new intact stability

153
criteria with confidence. For this purpose, we have per-
formed systematic sample calculations of the current
weather criterion, the dead ship level 2 vulnerability crite-
rion and the second check of parametric roll level 2 criteri-
on for four ships and many loading conditions. Since the
results for the dead ship level 2 vulnerability criterion and
the second check of parametric roll level 2 criterion yield
the probabilistic index, the safety level can be evaluated.
STABILITY ASSESSMENT METHODS
Weather Criterion
The technical background of the IMO weather criterion
is described in the explanatory notes of the 2008 Intact Sta-
bility Code (IMO, 2009). In this criterion, an uncoupled roll
equation is applied to a ship in beam wind and waves. If the
external energy exceeds the restoring energy due to the
calm-water nonlinear restoring moment, i.e. b/a, a ship is
judged as unsafe. The wave exciting moment and wind in-
duced moment are empirically estimated based on the
Froude-Krylov assumption and wind tunnel experiments,
respectively. The roll amplitude due to waves is calculated
with quadratic roll damping and a linear restoring moment.
The estimation method of roll damping including effect of
bilge keels is based on Russian model experiments. The
wave steepness is determined with Sverdrup and Munks
observation as the maximum amplitude out of 20 to 50 roll
wave height and the mean wave period are determined with
the North Atlantic wave scattering diagram (IACS, 2001)
and the relation between the mean wind velocity and the
significant wave height is obtained from the WMO (World
Meteorological Organization) reference wave heights
(WMO Code 1100).
In summary, the dead ship level 2 criterion uses the
same framework of ship dynamics but the failure probabil-
ity is determined without intermediate probability assump-
tions for wind and waves. Estimation methods of hydrody-
namic parameters are updated, but those for aerodynamic
parameters.
Parametric Roll Level 2 Criterion
Since parametric roll is a resonant phenomenon, roll
damping is crucial. Thus, it is appropriate to deal with
parametric roll and dead ship stability together.
The second check of parametric roll level 2 criterion
(IMO, 2015) uses an uncoupled model with restoring varia-
tion as follows.
( e
)
&& + 2& + &3 + 2 + 2 F + M cos t 1 / 2 3
(1)
+ l3 3 + l5 5 = 0
2 2
is the roll angle. and are the roll damping coefficients.
is the natural roll circular frequency. The term
2 + 2 l 3 3 + 2 l 5 5

cycles. The fluctuating wind is represented by a gustiness describes the restoring moment in
factor based on Watanabes observation. Using the model ( e t ) 1 / 2 3
2 F + M cos
calm water, the term the
mentioned above and casualty statistics of ships before the
restoring variation due to waves. The roll damping is esti-
1950s, the mean wind velocity of 26 m/s was determined
mated with Ikedas simplified method including bilge keel
to distinguish safe and unsafe ships.
effects or its equivalent and the restoring variation is calcu-
lated under the Froude-Krylov assumption.
Deadship level 2 criterion
If a solution of Eq. (1) is assumed to have following
form:
The dead ship level 2 criterion (IMO, 2016) is in princi-
ple an updated version of the weather criterion. It also uti- = A cos ( t ), A 0 (2),
lises an uncoupled roll model in fluctuating beam wind and
irregular beam waves. However, the final judgement is the amplitude A can be determined by solving the following
made with the calculated probability of stability failure for algebraic equation (Maki et al., 2011) :
one hour in the North Atlantic (i.e. C value). If the proba-
2
( ) 8 2 M
2
bility exceeds the acceptable values, a ship is regarded as 3 A 2 2 + 8 6 A 2 2 2

+
vulnerable to dead ship stability failure. The roll restoring ( 2 A )
2 2 2

2
(
4 A
2 2
)
moment is modelled as linear. For the critical roll angle, the
5 2 A4 l5 6 2 A2 l3 + 8 2 2 8 2
2 2 2 2
area of approximated GZ is equal to that of original GZ up
+ =F
2
to the angle of vanishing stability. If the roll angle exceeds (
4 2 A2
2
)
the critical roll angle, stability failure occurs (Bulian, 2004).
This means that the approximated potential energy is kept (3).
as the same as the original one. Thus the energy balance
concept still survives in the new criterion. This is known as an averaging method (Sato, 1977). If
The roll damping moment is estimated using Ikedas multiple solutions exist, local stability analyses of the solu-
simplified formula, which takes bilge keel effects into ac- tions are executed to select the most feasible amplitude. If
count (Kawahara et al., 2009), or an equivalent approach. the amplitude exceeds 25, the ship is judged as unsafe for
The wave exciting moment is calculated under the Froude- the relevant regular waves.
Krylov assumption and assuming rectangular hull sections The relation between regular waves and irregular waves
of same breadths and section area as the original sections is based on Grims effective wave concept (Grim, 1961)
(Umeda and Tsukamoto, 2008). The wind induced moment and a Pierson-Moskowitz type spectrum. The joint proba-
is estimated using the empirical method in the weather cri- bility density of the significant wave height and the mean
terion. The fluctuating wind is modelled with the Davenport wave period are determined with the North Atlantic wave
spectrum and the irregular waves with a Pierson-Moskowitz scattering diagram. Integrating the joint probability density
spectrum. The joint probability density of the significant within the dangerous sea states, the probability that a ship

154
meets a dangerous sea state can be determined. Assuming
the probability density of the wave heading to be uniform,
this probability is averaged from head seas to following
waves. Since the ship speed at sea is normally equal to the
service speed, the effect of wave heading is approximately
equivalent to that of ship speed in longitudinal waves. If
this finally obtained probability, i.e. C2 value, exceeds 0.06,
the ship is judged as vulnerable to parametric roll.
RESULTS AND DISCUSSION
We apply the above three methods to some sample
ships: a cruise ship, two RoPax ships and an LNG carrier,
all built in Japanese shipyards.
As a first step, the draughts are fixed as their fully
loaded conditions and light-ship conditions and GMs are
widely changed for covering even the cases b/a<1, i.e. cases
where ships fail with the curent weather criterion. The
relation between GM and b/a is shown in Fig.1. No clear
correlation can be found. While b/a <1 happens only for
smaller GM, small GM can give also large b/a.
Fig. 1. Relationship between weather criterion and GM for four
sample ships under fully loaded and light ship conditions.
Fig. 2. Relationship between dead ship level 2 criterion and GM
for four sample ships under fully loaded and light ship conditions.
The relationship between GM and the C value is shown
in Fig. 2. Again, there is no clear correlation. This suggests
that GM is not a good measure for dead ship stability. This
is partly because GM is an index for initial stability and not
for stability at large heel angle and partly because
synchronous resonance is related to natural roll period as
well.
The relation between b/a and C is plotted in Fig. 3.
Roughly speaking, there is negative correlation here. If b/a
increases, C decreases. This is because both the weather
criterion and the dead ship level 2 criterion are based on the
same physical modelling. The scatter in the intermediate b/a
range is probably due to differences in hydrodynamic or
environment modelling between the two methods.
Fig. 3. Relationship between dead ship level 2 criterion and
weather criterion for four sample ships under fully loaded and
light ship conditions.
Fig. 4. Relationship between dead ship level 2 criterion and
weather criterion for the cruise ship for all operational conditions.
As a next step, both draught and GM are systematically
changed within operational ranges for the cruise ship and
the RoPax ship 1, for which the weather criterion could be
dominant. The relation between b/a and C values for the
cruise ship is shown in Fig. 4. If we apply the dead ship
155
level 2 criterion to this ship, some loading condition under
which the ship fails to comply with the weather criterion
can be acceptable for operation. Note that such extended
operational zone has sufficiently small failure probability in
the North Atlantic. Thus we can say that the relaxation of
the weather criterion due to the dead ship level 2 criterion
does not change the safety level for dead ship stability. This
supports the discussion for the weather criterion at the
IMOs earlier days. Fig. 5 shows the case of RoPax ship 1.
It also provides similar results.
Fig. 5. Relationship between dead ship level 2 criterion and
weather criterion for RoPax ship 1 for all operational conditions.
It is also important to evaluate the effect of the relaxa-
tion of the weather criterion on parametric roll. The rela-
tionship between b/a and C2 values for the cruise ship and
the RoPax ship 1 are shown Figs. 6 and 7, respectively. The
RoPax ship 1 is critical to parametric roll danger under
certain loading conditions. However, these conditions
sufficiently comply with the weather criterion. In loading
conditions allowed by the dead ship level 2 criterion, the C2
values are sufficiently small, i.e. the ship is safe in terms of
parametric roll.
Fig. 6. Relationship between parametric roll level 2 criterion and
weather criterion for the cruise ship for all operational conditions.
156
Fig. 7. Relationship between parametric roll level 2 criterion and
weather criterion for RoPax ship 1 for all operational conditions.
CONCLUSIONS
If we apply the dead ship level 2 criterion instead of the
weather criterion, some additional loading conditions for a
cruise ship can be allowed for operation. The probabilistic
assessment using level 2 vulnerability criteria suggests that
these loading conditions do not change the safety level for
dead ship stability or parametric roll.
ACKNOWLEDGEMENTS
This work was supported by a Grant-in Aid for Scien-
tific Research from the Japan Society for Promotion of
Science (JSPS KAKENHI Grant Number 15H02327). It
was partly carried out as a research activity of Goal-based
Stability Criteria Project of Japan Ship Technology Re-
search Association in the fiscal years of 2015 funded by the
Nippon Foundation.
REFERENCES
Bulian, G. and Francescutto, A. (2004) A simplified mod-
ular approach for the prediction of the roll motion due to
the combined action of wind and waves, J. Engineer-
ing for the Maritime Environment 218, pp. 189-212.
Grim, O. (1961) Beitrag zu dem Problem der Sicherheit
des Schiffes in Seegang, Schiff und Hafen 6, pp.490-
497.
IACS (2001) "Standard Wave Data", Rec. No. 34, Rev.1.
IMO (2009) International Code on Intact Stability, 2008.
IMO (2015) Report of the Working Group (Part 1), SDC
2/WP.4.
IMO (2016) Report of the Working Group (Part 1), SDC
3/WP.5.
Kawahara, Y., Maekawa, K. and Ikeda, Y. (2009), A sim-
 ple prediction formula of a roll damping of conventional
cargo ships on the basis of Ikedas method and its limi-
tation, Proc. 10th Int. Conf. on Stability of Ships and
Ocean Vehicles, pp.387-398.

Maki, A., Umeda, N., Shiotani, S. and Kobayashi, E. (2011),

Parametric rolling prediction in irregular seas using
combination of deterministic ship dynamics and proba-
bilistic wave theory, J. Marine Science and Technology,
16/3, pp.294-310.

Sato, C. (1977) Nonlinear Vibration, Asakurashoten (To-

kyo).

Umeda, N. and Francescutto, A. (2016), Current state of

the second generation intact stability criteria Achieve-
ments and remaining issues, Proc. 15th Int. Ship Stabil-
ity Workshop, Stockholm, pp.3-9.

Umeda, N. and Tsukamoto, I. (2008), Simplified formula

for calculating effective wave coefficient and its impact
on ship stability assessment, Proc. 6th Osaka Collo-
quium on Seakeeping and Stability of Ships, Osaka, pp.
329-333.

157
 SHOPERA Manoeuvrability in Seaway
Vladimir Shigunov
DNVGL SE, Hamburg
ABSTRACT
Maneuverability of ships is regulated by the IMO
Standards for Ship Maneuverability, not addressing
ships maneuverability in adverse conditions. The
importance of this increased after the introduction of
EEDI, which led to concerns that fulfilling EEDI by
simply reducing ships installed power may lead to
insufficient maneuverability in adverse conditions.
Responding to the need for norming the
maneuverability in adverse conditions within EEDI
regulations, EU-funded project SHOPERA developed
rational criteria and practical assessment procedures,
which are outlined in this paper together with some
application examples.
INTRODUCTION
Introduction of the Energy Efficiency Design Index (EEDI)
to improve energy efficiency and reduce Greenhouse Gas
(GHG) emissions of shipping, MEPC.203(62), invoked
concerns regarding sufficiency of propulsion and steering
abilities of ships to maintain manoeuvrability in adverse
weather conditions if the EEDI requirements are achieved
by simple reduction of the installed engine power.
Following a proposal from the International Association of
Classification Societies (IACS), Reg.21, Ch.4 of MARPOL
Annex VI was augmented with the requirement For each
ship to which this regulation applies, the installed
propulsion power shall not be less than the propulsion
power needed to maintain the manoeuvrability of the ship
under adverse conditions as defined in the guidelines to be
developed by the Organization. Studies by IACS led to
development of first such guidelines in 2011, MEPC
62/5/19 and 62/INF.21, and approved by IMO 2012,
MSC-MEPC.2/Circ.1 (2012). These guidelines were based
on a three-level assessment: Comprehensive Assessment
(Level 3), Simplified Assessment (Level 2) and Minimum
Power Lines (Level 1). They were revised in 2013,
MEPC.232(65) Level 3 was removed; in Level 2, the
option to use numerical methods to define added resistance
was removed and only model tests were left; in Level 1, a
formulation was accepted that did not relate to propulsion
or steering characteristics of ships; besides, standard wave
height was re-calibrated for small vessels and approved
for Phase 0 of EEDI implementation (until 2014-12-31).
158
Later, the 2013 Interim Guidelines were updated, MEPC.
262(68), and extended into EEDI implementation Phase 1
(until 2019-12-31). Presently, these Guidelines are the only
provision preventing new built ships from under-powering;
still, some elements can be improved, e.g. the necessity to
do model tests in Level 2 assessment is too impractical;
Level 1 assessment was not rationally derived; absence of
Level 3 assessment makes the Guidelines unsuitable for
innovative propulsion and steering concepts, i.e. contradicts
the idea of EEDI introduction.
This led to several research initiatives in EU (project
SHOPERA, Energy Efficient Safe Ship Operation),
Germany (PerSee Performance of Ships in Seaway), The
Netherlands (Short-sea Shipping Requirements: EEDI &
Minimum Power Requirements), Japan (Strategy Research
Committee on Minimum Propulsion Power) and other
countries, aiming at improving these guidelines. The 3-year
research project SHOPERA, launched in October 2013 and
funded by the European Commission, aimed at submitting
its results for consideration to MEPC 71 (spring 2017),
whereas some results are to be submitted to MEPC 70
(October 2016). The project involves a strong European
RTD consortium 1 including classification societies,
universities, research organisations, model basins, ship
designers, shipyards and ship operators.
MANOEUVRABILITY CRITERIA
Manoeuvrability of ships is presently normed by IMO
Standards for Ship Manoeuvrability, IMO (2002), which
address turning, initial turning, yaw checking, course
keeping and emergence stopping abilities, evaluated in
simple standard manoeuvres in calm water. These
Standards have been often criticized for not addressing ship
manoeuvring characteristics at low speed, in restricted areas
and in adverse weather conditions; the importance of the
latter increased after the introduction of EEDI.
In EE-WG 1/4 (2010), IACS put together the requirements
of classification societies related to the redundancy or
1
http://www.shopera.org, National Technical University of Athens,
DNVGL, Lloyds Register, MARINTEK, Instituto Superior Tecnico,
University Duisburg-Essen, Registro Italiano, Uljanik Shipyard, Flensburg
Schiffbautechnische Gesellschaft, VTT, Flanders Hydraulics Research,
CEHIPAR, Strathclyde University, Denmark Technical University,
Technical University Berlin, Delft University of Technology, Naval
Architecture Progress, Danaos Shipping Company, FOINIKAS Shipping,
CALMAC Ferries
duplication of the propulsion system to indicate relevant
criteria and environmental conditions for steering and
propulsion in adverse weather conditions; a summary of
these requirements in Table 1 (vd means design speed, vw
wind speed, hs significant wave height) indicates, basically,
two requirements: to change (or keep) heading and
maintain some minimum advance speed (or position, i.e.
zero speed).
Table 1. Criteria and weather conditions for redundancy and
duplication of propulsion system according to requirements of
classification societies, EE-WG 1/4 (2010)
Class Criteria vw hs
GL Change and keep heading 21 m/s 5.4 m
(weather-vaning)
GL Advance speed 11 m/s 2.8 m
min(7 knots, 0.5vd )
LR Steering ability, advance - -
speed 7 knots
BV Advance speed Bft 5 (8.0 to corresp. to
7.0 knots 10.7 m/s) vw
ABS Weather-vaning without 33 knots 4.5 m
drifting (17.0 m/s)
DNV Weather-vaning at Bft 8 (17.2 to corresp. to
advance speed 6 knots 20.7 m/s) vw
Studies by IACS on minimum power requirements for
manoeuvrability in adverse weather conditions started with
analysis of functional requirements to manoeuvrability in
the open sea and coastal areas, MEPC 62/5/19 (2011) and
62/INF.21 (2011), concluding that manoeuvring in coastal
waters is more challenging than in the open sea; the
resulting criteria for ship propulsion and steering abilities
were formulated in MEPC 64/4/13 (2012) and 64/INF.7
(2012): the ship should be able, in seaway from any
direction, to (1) keep course and (2) keep advance speed of
at least 4.0 knots. The corresponding weather conditions are
not too severe, because ship masters do not stay near the
coast in an increasing storm and either search for a shelter
or leave to the open sea and take a position with enough
room for drifting. The environmental conditions selected
(wind speed 15.7 m/s at significant wave height 4.0 m for
ships with Lpp=200 m to 19.0 m/s and 5.5 m, respectively,
for Lpp=250 m and greater) were derived by benchmarking
of tankers, bulk carriers and container ships in the EEDI
database against these two criteria.
In SHOPERA, available detailed investigations of accidents
related to insufficient manoeuvrability in adverse weather
conditions were studied. Summarising, as the most frequent
cause of heavy weather-related grounding accidents is
waiting at anchor in heavy weather and too late starting of
the engine. In several accidents, MAIB (1996,2009,2012),
ATSB (2008), vessels were not able to move away from the
coast or turn into seaway despite full engine power applied.
Table 2 summarises relevant criteria as related to
insufficient propulsion or course-changing abilities and
weather conditions.
Interviews of masters of about 50 container ships, bulk
carriers and tankers conducted in the projects PerSee and
SHOPERA, indicate that in the open sea, the captain
usually can decide what severity of weather conditions is
acceptable, depending on the freeboard, cargo, stability and
propulsion and steering characteristics of the vessel. On the
Table 2. Summary of accident reports
Ref. Relevant criteria Environmental Conditions
MAIB (2009) Course Bft 9-10 (20.8 to 28.4 m/s),
changing hs > 10 m
MAIB (1996) Course gale force wind (17.3 to
changing 20.8 m/s), low waves
MAIB (2012) Propulsion vw = 40 knots (20.6 m/s),
hmax = 4 m
ATSB (2008) Propulsion vw = 38-46 knots (19.5 to
23.7 m/s), hs = 6.0-6.6 m
MAIB (2002) Propulsion Bft 5-7 (8.0 to 17.1 m/s),
low waves
MAIB (2009a)0 Propulsion Bft 10 (24.5 to 28.4 m/s)
MAIB (2012a) Course vw = 38-45 knots (19.5 to
changing 23.2 m/s)
other hand, when caught in most violent storms, steering
against seaway may be impossible for any vessel, but
drifting with seaway can be an option for a limited time.
However, power matters to escape the storm and bring the
ship into safe weather conditions. Manoeuvring in coastal
areas was reported as more challenging than in the open sea,
requiring, in principle, any manoeuvre, sometimes at
unfavourable seaway direction with respect to the ship.
Environmental conditions are less severe than in the open
sea, because ship masters do not remain near the coast in a
growing storm, but either search for shelter or leave to the
open sea. As relevant manoeuvring problems, steering
problems were mentioned insignificantly more often than
propulsion problems (83% vs. 60% of cases, respectively);
insufficient engine power was mentioned more frequently
for bulk carriers and tankers, whereas insufficient rudder
capability more frequently for container vessels. As a very
specific manoeuvring problem in restricted waters,
manoeuvrability at limited speed (due to navigational
restrictions, e.g. during approaching ports) was mentioned,
in strong wind and, sometimes, strong current, but usually
without large waves in protected areas.
Based on the above results, Shigunov and Papanikolaou
(2014) proposed to differentiate three scenarios, in which
steering and propulsion abilities of ships are challenged in a
different way and which, therefore, require specific criteria:
open sea, coastal areas and restricted areas at limited speed.
For the open sea, the ability of ship to weather-vane, i.e.
keep heading in head to bow-quartering seaway, was
proposed as a criterion. Regarding corresponding
environmental conditions, it was noted that none of existing
ships can steer against waves and wind in most severe
possible storms, therefore, benchmarking of the existing
world fleet with respect to the weather-vaning criterion was
proposed to define the standard wave height. For the coastal
areas, two criteria were proposed: the ability of the ship to
perform any manoeuvre and the ability to maintain some
speed over ground to enable leaving the coastal area before
the storm escalates; due to navigational restrictions, both
should be possible in seaway from any direction. The
corresponding environmental conditions are less severe
than in the open sea and should also be defined by
benchmarking of existing ships against these criteria.
Manoeuvrability at limited speed in restricted areas refers
to situations where the ship master has to reduce the applied
engine power (and thus forward speed) significantly below
available power because of navigational restrictions, e.g.
during approaching to or entering ports, navigation in
159
channels and rivers etc. Because full available power
cannot be applied in this scenario, the corresponding
manoeuvrability criteria will not impose any restrictions on
minimum propulsion power, thus this scenario is not
considered here.
SHOPERA proposed the following three criteria for
manoeuvrability in adverse weather conditions: weather
vaning ability in heavy weather in the open sea and steering
and propulsion abilities in increasing storm in coastal areas.
In the practical assessment, weather-vaning ability was
treated in a very simplified way, as the ability of the ship to
keep position in bow to bow-quartering seaway; this
simplification follows from the observation that the ship
(with the steering devices only at the stern) is not able to
keep heading at some small forward speed, because of the
reduced forces on hull and rudder.
The steering ability in increasing storm in coastal areas is
understood as the ability of the ship to perform any
manoeuvre in seaway from any direction. An equivalent,
but easier to verify in practice criterion was used, that the
ship should be able to start or continue course change in
seaway from any direction. This formulation should be
distinguished from the traditional course-keeping problem:
here, the ability of the steering system to overcome
environmental forces and start (or continue) course change
during an arbitrary manoeuvre is normed (i.e. capability of
the steering system); i.e. it does not matter whether each
intermediate state during manoeuvre is stable or not, thus
the stability of ship on a course is not addressed, whereas
the traditional definition of course-keeping addresses
stability of straightforward motion. Note that the proposed
criterion does not exclude the ships ability to perform also
straightforward motion (which can be considered as one of
required manoeuvres): if a ship is directionally unstable on
some course, it is still able to follow this course using
rudder for course corrections.
The propulsion ability in increasing storm in coastal areas
ensures that the ship is able to leave coastal area in a
sufficient time before the storm escalates. As the minimum
required advance speed, 6.0 knots was chosen by
SHOPERA instead of 4.0 knots used in 2013 Interim
Guidelines to take into account possibly strong currents in
coastal areas.
The corresponding environmental conditions for these three
criteria are proposed to be defined by benchmarking of
existing vessels against these criteria.
ASSESSMENT PROCEDURES
Introduction
The idea proposed by SHOPERA is to allow free choice of
assessment procedures of different complexity (similarly to
2012 Interim Guidelines), ranging from simple empirical
formulae at Level 1 to advanced assessment procedures at
Level 3, so that the designer can select the most suitable
procedure depending on the particular design needs. Simple
assessment procedures are sufficient for the majority of
conventional vessels, whereas more accurate assessment
procedures and evaluation methods are required for cases
with large uncertainties, such as innovative propulsion and
steering solutions.
160
SHOPERA proposes three assessment procedures:
- Comprehensive Assessment allows the best
accuracy, solving coupled nonlinear motion
equations. Still, the designer does not have to use
expensive evaluation methods for different forces,
but can choose between numerical, experimental or
empirical methods for different force components.
This level is necessary for ships with innovative
propulsion and steering arrangements.
- Simplified Assessment, a first-principle assessment
with reduced number of considered situations and
reduced complexity of motion equations, also
allowing choosing between experimental, numerical
or empirical methods to evaluate force components,
and having complexity of a spreadsheet calculation.
- Sufficient Propulsion and Steering Ability Check is
based on pure empirical formulae to define the
required installed power as a function of main ship
parameters (deadweight, block coefficient, windage
area, rudder area, engine and propulsion type), of a
complexity of a pocket calculator.
Comprehensive Assessment
Whereas the compliance with the IMO Manoeuvrability
Standards, IMO (2002), is demonstrated in full-scale trials,
this is impossible for manoeuvrability in adverse weather
conditions. Alternatively, the proposed criteria (weather
vaning, steering and propulsion abilities) can be evaluated,
in principle, directly in transient model experiments with
self-propelled ship models in simulated irregular waves and
wind, for all required combinations of wave directions and
periods. This is, however, presently unfeasible for several
reasons for practical purposes: First, providing reliable
statistical predictions in irregular seaway requires repeating
tests in hundreds of long realisations of each seaway, which
is too expensive. Besides, few facilities exist worldwide
able to carry out such tests, which makes them impractical
for routine design and approval. Finally, results of such
tests very much depend on the time history of steering,
which causes too large variability of the results, which thus
cannot be reliably verified especially in marginal cases
(which are the cases of interest in approval). Alternative to
such model tests, direct numerical simulations of transient
manoeuvres in irregular seaway, are not mature enough for
routine design and approval yet, ITTC Manoeuvring
Committee (2008).
The Comprehensive Assessment procedure proposed by
SHOPERA is based on neglecting oscillatory forces and
moments due to waves and thus considering only time
average forces and moments, assuming that the time scale
of such oscillations is shorter than the time scale of
manoeuvring motions. This effectively reduces the
evaluation of manoeuvrability criteria to a solution of
coupled motion equations in the horizontal plane under the
action of time-average wave-induced forces, as well as
wind forces, calm-water forces, rudder forces and propeller
thrust. Projecting forces on the x - and y -axes and
moments on the z -axis of the ship-fixed coordinate system,
Fig. 1, leads to a system of equations, converging to a
steady state described by the following system (note that
achieving converged solution can be realised in different
ways, including time-domain simulation):
Xs + Xw + Xd + XR + T (1 t ) = 0 (1)
Ys + Yw + Yd + YR = 0 (2)
Ns + Nw + Nd YR lR = 0 (3)
Fig. 1. Coordinate system and
definitions
Indices d, w, s, and R denote, respectively, wave, wind,
calm-water and rudder-induced forces and moments; T is
propeller thrust. The coordinate system has an origin O in
the main section at the water plane; x-, y- and z-axes point
towards bow, starboard and downward, respectively
(positive rotations and moments with respect to z-axis are
clockwise when seen from above). The ship sails with the
speed vs; its heading deviates from the course by the drift
angle . The mean wave and wind directions are specified
by angles e and w, respectively (0, 90 and 180 for waves
and wind from the north, east and south, respectively);
rudder angle is positive to port; lR is the lever of the yaw
moment due to rudder. A converged solution, described by
system (1)-(3), provides the required propeller thrust (from
which, advance ratio J, rotation speed n of the propeller and
required PD and available PDav delivered power are found),
drift angle and rudder angle .
Figure 2 shows examples of converged solutions and
application of the steering and propulsion criteria in polar
coordinates ship speed (radial) seaway direction
(circumferential, head waves and wind come from the top):
along line A, the required delivered power is equal to the
available delivered power, line B corresponds to the
required advance speed (here 4.0 knots), and line C limits
the highlighted area, in which the required steering effort
exceeds the available one (here, rudder angle exceeds 25o).
The left plot illustrates a seaway in which the vessel fulfils
both criteria (line A does not cross lines B and C); in the
middle plot, the installed power is marginally sufficient to
provide advance speed of 4.0 knots in head seaway, where
line A crosses line B; in the right plot, the installed power is
marginally sufficient for course-keeping in nearly beam
seaway, where line A crosses line C.
Fig. 2. Result examples of Comprehensive Assessment
The advantage of the proposed Comprehensive Assessment,
and the main difference from the Comprehensive
Assessment in the 2012 Interim Guidelines is that the
different effects (wind, waves, calm water, rudder, propeller
and engine) can be measured or computed separately, if
necessary with different methods (experimental, numerical
or empirical). Note that even if model tests or complex
numerical computations are used for some of contributions,
they are done in stationary setups under well-controlled
conditions and combined in a simple mathematical model.
Because the proposed assessment procedure allows
definition of different forces separately, using different
(experimental, numerical or empirical) methods, an
important question is how much freedom we have in the
definition of each force component. Table 3 shows the
percentage of the change of the required installed power to
fulfil propulsion and steering ability requirements due to
changes of each force or moment component in turn by
10% at significant wave height of 5.5 m (for example,
change of the x-calm water force by 10% changes the
required installed power by 3.0%).
Table 3. Percentage change of required installed power due to
change of components of forces and moments
Contributions x-force y-force z-moment
Calm-water 3.0 3.4 3.5
Wind 2.5 1.6 0.6
Wave 3.8 3.0 0.3
Rudder 1.5 3.4 -
The table shows that the most important contributors (bold)
are the time-average x- (added resistance) and y-forces,
calm-water x- and y- forces and z-moment and lateral
rudder force. For these forces, the error of about 15% leads
to the error in the definition of the required power of 5%.
Opposite to the transient manoeuvring tests in simulated
wind and irregular waves, the proposed procedure relies on
measurements of separate forces in well-controlled steady
conditions. For calm-water, wind, rudder and propeller
forces, such experiments are well established and can be
done in many facilities world-wide (calm-water resistance
and propulsion characteristics have to be measured for
EEDI verification anyway). However, measurement of the
time-average wave forces requires advanced measurements
in a seakeeping basin, therefore, cannot be used routinely.
Numerical methods are presently available, in principle, for
the same forces (calm-water, wind, rudder and propeller) as
well; however, their use in regulatory assessment needs a
significant effort of administrations and recognised
organisations. The availability of numerical methods for
time-average wave forces is one of the most critical issues:
the absence of suitable numerical methods for added
resistance led the removal of numerical methods from the
2013 Interim Guidelines. Development and validation of
numerical methods for time-average wave forces and
moments was one of the major tasks of SHOPERA. Model
test measurements of such forces were done for three ship
models (a 14000 TEU container vessel, a VLCC tanker and
a RoRo) for various ship speeds and wave directions and
periods (with particular emphasis on short waves); results
of these measurements were used for an international
benchmarking of the available numerical methods for
time-average wave forces. The results of this benchmarking
161
(will be published elsewhere) show a significant progress of
numerical methods in the last years and indicate the
principal availability of numerical methods for regulatory
purposes, if applied correctly.
Especially important for the practical implementation of the
proposed procedure is the availability of empirical methods
for different force components. In addition to the well-
established empirical methods for wind forces,
Blendermann (1993), Fujiwara et al. (2006), extensive
validation studies were carried out together with the project
from Japan for calm-water reactions and rudder forces in
propeller race, which indicate availability of such empirical
methods for the practical use. Again, empirical methods for
the time-average wave forces and moments have required
particular attention and the significant progress achieved in
the project will be reported elsewhere.
An important aspect of propulsion and steering in adverse
conditions is a correct description of the engine under high
load. Note that frequently used wrong assumptions (e.g.
constant torque or constant rotation rate of the engine) lead
to strongly misleading results. In SHOPERA, air/surge
limits of diesel engines (two- and four-stroke) were verified
and recommendations were provided for practical
assessment procedures.
Simplified Assessment
Principles
The aim of the development of the Simplified Assessment
was to have a simple enough procedure for routine use by
Administrations by reducing the number of calculations
(solution cases) as well as the number of terms in motion
equations (1)-(3), while keeping all relevant physics. In
particular, the Simplified Assessment addresses the same
criteria as those enforced in the Comprehensive Assessment
(weather-vaning, steering and propulsion). In this paper,
Simplified Assessment procedures concerning steering
ability (ability to keep or change course in all seaway
directions) and propulsion ability (ability to keep a
minimum advance speed in all seaway directions) are
summarised; details can be found in Shigunov et al. (2014).
Propulsion Ability
The starting point is system (1)-(3), solved for all possible
seaway directions to check whether the ship is able to keep
forward speed of at least 6.0 knots in seaway from any
direction. Noting, first, that bow seaways are most critical
for required power (Fig. 2, middle plot), it will be enough
to consider only seaways from 0 to about 60 off-bow.
Second, neglecting the influence of drift on the required
thrust and power allows dropping equations (2) and (3). As
a result, only eq. (1) needs to be considered,
X s + X w + X d + X R + T (1 tH ) = 0 , (4)
in which, however, the time-average surge force due to
waves Xd means maximum force over all mean wave
directions between 0 and 60 off-bow.
Whereas the contributions Xs, Xw, Xd and XR can be found
using any method (empirical, numerical or experimental)
from the Comprehensive Assessment, it seems logical to
allow using even simpler empirical approximations in eq.
(4), consistent with the complexity of the Simplified
Assessment. For example, assuming rudder resistance as
162
XR=tRT, where tR is an empirical constant, results in a
simple non-iterative equation for the required thrust T:
X + Xw + Xd
T = s (5)
1 tH tR
Again, the most challenging term is the time-average
longitudinal force in short-crested irregular waves (added
resistance) Xd. According to the 2013 Interim Guidelines,
it can be defined only using model tests. According to the
proposal by SHOPERA, it can be defined using any method
from the Comprehensive Assessment (empirical, numerical
or experimental) to define quadratic transfer functions of
added resistance in regular waves, combined with a spectral
integration. For the use in the Simplified Assessment, a
simple empirical expression was proposed in Shigunov et al.
(2014) directly for the maximum surge force over mean
wave directions 0 to 60 off-bow in irregular short-crested
waves, based on computations for many ships with the
software GL Rankine, Sding and Shigunov (2015):
(
X d = 83 Lpp CB1.5 1 + Fr hs2 ,) (6)
1/ 2
where Fr = vs ( gLpp ) is the Froude number.
Figure 3 compares the ratio of the required to available
delivered power according to the Simplified Propulsion
Ability assessment (y axis) with the same ratio from the
Comprehensive Assessment (x axis) for 4 bulk carriers
( , , , ), 3 tankers ( , , ) and 4 container ships
( , , ,) at significant wave heights from 0.0 to 9.5 m.
The Simplified Assessment procedure is sufficiently
accurate to slightly conservative (dashed line means exact
agreement); the conservativeness increases for the ratio of
the required to available power greater than one, which is
not relevant anyway. This procedure was implemented for
practical use in MS Excel, including vessels with
conventional propulsion (diesel with fixed pitch propeller),
as well as vessels with diesel-electric propulsion and
controlled-pitch propeller.
Fig. 3: Ratio of required to available delivered power
according to Simplified (y-axis) vs. Comprehensive
(x-axis) Propulsion Ability Assessment
Steering Ability
The starting point is system (1)-(3), solved for all relevant
forward speeds and all seaway directions to check that the
ship is able to keep course in seaway from any direction.
Results of Comprehensive Assessment for many ships show
that the dimensioning condition for the installed power, at
which the ratio of the required to available delivered power
is maximised along the line of maximum rudder angle
(further referred to for brevity as critical condition for
steering) is close to beam seaway, Fig. 2, right. (Note that
from experience, as well as from the results of
Comprehensive Assessment for many ships, the critical
conditions for steering occur most frequently in stern
quartering waves, like in Fig. 2, middle; however, in such
situations the required power is defined by the Propulsion
Ability, i.e. by the crossing point of lines A and B; when
Steering Ability is dominating for the definition of the
installed power, i.e. when line A crosses line C, the critical
conditions are always close to beam seaway situations.)
This allows reducing the evaluation of the time-average
wave and wind forces to beam seaways. The second Fig. 4: Ratio of required to available delivered power
simplification stems from the observation that in critical according to Simplified (y-axis) vs. Comprehensive
conditions for steering, the levers of time-average wave and (x-axis) Steering Ability Assessment
wind yaw moment are much smaller than the lever of the
calm-water yaw moment; this allows replacing equations drift angle in critical conditions for steering ability. A
(2) and (3) with a single equation conservative recommendation, based on Comprehensive
(
YR = b Yw90 + Yd90 , ) (7) Assessment for 15 vessels, is b = 0.5 ; an empirical formula
for b as a function of main ship particulars needs to be
where b ls ( ls + lR ) N s (N s + 0.5Ys Lpp ) . Thus, system developed.
(1)-(3) reduces to one equation The procedure was implemented in MS Excel for practical
use for vessels with diesel and diesel-electric propulsion,
X s + X w90 + X d90 + X R + T (1 tH ) = 0 (8) fixed- and controllable-pitch propellers, conventional
and one check (7). The solution of eq. (8), i.e. the rudders and pods.
maximum attainable ship speed and corresponding
propeller rotation speed and propeller thrust, defines in turn
the maximum available lateral steering force on the rudder PROPULSION AND STEERING ABILITY CHECK
YRav , which, according to the proposed Simplified Steering
Ability assessment, should be not less than the required The simplest assessment procedure, Sufficient Propulsion
lateral steering force defined by eq. (7). and Steering Ability Check (Level 1), is based on pure
To define terms in equations (7)-(8), in addition to methods empirical formulae to define the required installed power as
from the Comprehensive Assessment (empirical, numerical function of main ship parameters (deadweight, block
and experimental), simplified empirical formulae were coefficient, windage area, rudder area, engine and
developed, consistent with the complexity of the Simplified propulsion type etc.). Its development was still ongoing at
Assessment. The increase in rudder resistance XR is the time of writing this paper; Fig. 5 shows preliminary
calculated as XR=tRT, where tR is an empirical constant. A results: logarithm of the required installed MCR according
simple empirical formula is proposed for the added to the Sufficient Propulsion and Steering Ability Check,
denoted as Level 1, vs. logarithm of the required installed
resistance in irregular beam waves X d90 , obtained from
MCR according to the Simplified Assessment, denoted as
numerical computations with GL Rankine and spectral Level 2, for the propulsion (top) and steering (bottom)
integration with JONSWAP spectrum (=3.3) and cos2 abilities.
spreading as a maximum over peak wave periods from 7.0
to 15.0 s:
B ( 0.1 + Fr ) hs
X d90 = 380 Lpp C 1.5 2
(9) COMPARISON OF CRITERIA
Similarly, a simple empirical formula is proposed for the
time-average lateral wave force in irregular short-crested Obviously, the number of any new criteria should be
beam seaway, minimised: criteria can be dropped if they are redundant,

{ }
5 1
e.g. dominated by other criteria or well correlated with
Yd90 = 540 Lpp hs2 1 + Tp5 ( CB L0.5
pp ) (10) other criteria. Figure 6 compares marginal wave heights
according to weather-vaning and propulsion requirements.
Figure 4 compares the ratio of the required to available
The used weather-vaning criterion (position keeping in bow
delivered power according to approximation (7)-(8), y-axis,
seaway) obviously perfectly correlates with propulsion
with the same ratio from the Comprehensive Assessment,
(6.0 knots advance speed in bow seaway) and is, therefore,
x-axis, for the same ships and wave heights as in Fig. 3.
not required. It is, however, important to compare
The value of b is taken here from the Comprehensive
SHOPERA weather-vaning assessment results with the
Assessment in critical conditions for steering ability. Note
results of a more sophisticated criterion proposed by the
that the approximation (7)-(8) provides accurate to slightly
project from The Netherlands (heading recovery ability
conservative results.
of the ship to change heading from beam into head seaway).
Factor b takes into account calm-water reactions at the
Figure 7 compares marginal wave heights according to the
critical conditions for steering; obviously, it depends on
steering and propulsion criteria. For bulk carriers (BC) and

163

Fig. 5: Logarithm of required MCR according to Sufficient
Propulsion and Steering Ability Check (y-axis) for the
same wave height as provided by Simplified Assessment
at MCR shown along the x-axis, for propulsion (top) and
steering (bottom) requirements
Fig. 6: Maximum attainable significant wave height
according to weather-vaning criterion (y-axis) vs.
propulsion criterion (x-axis)
Fig. 7: Maximum attainable significant wave height
according to steering ability criterion (y-axis) vs.
propulsion criterion (x-axis)
164
tankers (TA), propulsion requirement dominates for smaller
vessels, whereas steering requirement dominates for large
ones. For the general cargo vessels (GC), steering and
propulsion requirements are well correlated. For container
vessels, propulsion requirement dominates for smaller
vessels (feeders), whereas steering requirement strongly
dominates for large vessels, due to large lateral windage
area. Other vessel types were in assessment during the
writing of this paper and results will be presented later.
As the way ahead to finalise selection of the criteria
(weather-vaning, steering and propulsion), sample
calculations by other stakeholders and comparison of
results can be proposed.
STANDARD WAVE HEIGHTS
Whereas criteria identify relevant ships characteristics
(ability to turn, ability to keep course etc.) and
corresponding measures (turning diameter, maximum wave
height for course-changing etc.) quantify performance of
ships with respect to these criteria, standards (norms) set
the acceptance limits on these measures for the ship to fulfil
the defined requirements. With respect to the weather
vaning, steering and propulsion criteria defined above,
obvious standard is the maximum wave height (criterion
dependent) at which the ship is able to fulfil the criteria.
Several sources have been used in SHOPERA to estimate
the standard wave heights. Statistics of weather conditions
during manoeuvring-related accidents in adverse weather
conditions, Ventikos et al. (2014), shows, first, that the
accident rates concerning grounding, collision and contact
in heavy weather are of the order of magnitude in the range
of 10-5 to 10-3 per ship per year, i.e. much less frequent than
other types of accidents. Second, environmental conditions
during accidents are remarkably mild: mean wind speed
about 10 m/s and mean significant wave height about 1.5 m.
This agrees with the well-known statistics from the
HARDER project, according to which more than 80% of
collisions happen at significant wave height below 2.0 m,
and significant wave height greater than 4.0 m is practically
not encountered during collisions.
On the other hand, maximum significant wave heights and
wind speeds during manoeuvrability-related accidents
achieve, according to SHOPERA statistics, 7.0 m and
20 m/s, respectively. Detailed accident reports, Table 2,
indicate wind force up to Bft 10 (in excess of 23 m/s) and
significant wave heights above 6.0 m during few accidents
in coastal areas. However, the well-documented case ATSB
(2008), where the significant wave height exceeded 6.0 m,
clearly indicates unacceptably long waiting of the vessel at
anchor in an increasing storm as the reason of the accident.
Figure 8 shows the number of ships at anchor as percentage
of the initial number of vessels at anchor depending on the
significant wave height during this storm. Whereas the two
vessels, which experienced the accident and a near-accident,
have been waiting at anchor until the significant wave
height exceeded 6 m, 50% of the vessels left anchorage
before significant wave height achieved about 5.5 m.
Similar conclusions follow from the interviews of ship
masters, Fig. 9: 50% of the ship masters prefer to leave
coastal areas before wind increases to Bft 8 and significant
wave height achieves 5.0 m.

Fig. 8. Number of vessels at anchor as percentage of the initial
number of anchored vessels vs. significant wave height during an
increasing storm according to data in ATSB (2008)
Fig. 9. Wind force (top) and significant wave height (bottom)
during leaving coastal areas ( ) and during encountering steering
and propulsion problems ( ) from interviews of ship masters
Of course, it is impossible to design ships for the worst
possible storms that can be encountered even in coastal
areas. The standard wave heights should be defined in such
a way that the majority of the existing vessels fulfil the
requirements, because, first, present design and operational
practices cannot be changed abruptly, and, second, the
present safety level with respect to manoeuvrability-related
accidents in heavy weather is satisfactory, Ventikos et al.
(2014). Notably, the study by IACS, EE-WG 1/4 (2010),
Table 1, identifies much milder standard wave heights for
weather-vaning (wind speed up to 21 m/s and significant
wave height up to 5.4 m) and advance speed (up to Bft 8 at
6.0 knots advance speed) requirements.
Therefore, benchmarking of the existing fleet with respect
to the new criteria appears as the most rational way to
define standard wave height. Such an approach was used in
the IACS studies on minimum power requirements and led
to the environmental conditions in the 2013 Interim
Guidelines: wind speed 15.7 m/s at significant wave height
4.0 m for ships with Lpp=200 m to 19.0 m/s and 5.5 m,
respectively, for Lpp=250 m and greater.
Therefore, case studies were undertaken in SHOPERA
concerning all ship types considered in the EEDI
regulations. At the time of writing, some preliminary results
were available for bulk carriers, tankers, general cargo
vessels and container carriers, which are discussed here;
further results will be presented elsewhere.
Because it is impossible to apply the new criteria to all
existing vessels, an important question is how to select the
representative vessels for the definition of the standard
wave height. An approach demonstrated here is to use the
marginally low powered vessels from the FairPlay
database: for each ship type and size, a vessel at the low
limit of the installed power is selected from all existing
vessels. Note that this means, effectively, selecting
slow-steaming vessels; other possibilities of selection will
be used and reported for comparison elsewhere.
The selected representative vessels are shown in Fig. 10 in
axes Lpp (x axis) installed MCR (y axis), separately for
bulk carriers, tankers, general cargo vessels and container
carriers. Grey dots indicate vessels from the FairPlay
database, black squares vessels used in the case studies and
blue circles the selected representative vessels.
Figure 11 shows the marginal significant wave heights
according to propulsion (top) and steering (bottom) ability
requirements for the selected representative bulk carriers,
tankers, general cargo vessels and container carriers.
One observation from these results is that the marginal
wave heights are ship size-dependent: larger vessels are
able to fulfil both propulsion end steering requirements at
greater significant wave heights than smaller vessels. This
is understandable physically; in principle, ship size
dependent standard wave heights may be acceptable from
the pragmatic point of view: because consequences of
accidents are greater for larger vessels, acceptable
probability of accidents should be lower for larger vessels.
Besides, ship size-dependent standard wave height would
reflect existing design and operational practices: smaller
vessels, obviously, do not operate in storms of the same
severity as larger vessels. Note that standard wave heights
in the 2013 Interim Guidelines are also ship-size dependent,
however, this is a subject of ongoing discussion.
Another observation from Fig. 11 is the dependency of the
marginal significant wave height on the ship type. In
particular, the selected representative bulk carriers and
tankers show remarkably similar marginal significant wave
heights, which are lowest over all ship types. General cargo
vessels and, especially, container carriers achieve
significantly greater marginal significant wave heights than
bulk carriers and tankers.
From the ongoing case studies for other ship types (the
results will be shown elsewhere) it is also obvious that
RoRo Cargo vessels also demonstrate greater marginal
wave heights than bulk carriers and tankers, whereas
passenger vessels (RoPax and cruise vessels) show
significantly higher marginal wave heights than other vessel
types, due to advanced propulsion and steering systems
(twin screw, diesel-electric main propulsion, controlled
pitch propeller, pods).
165

Fig. 10. Installed MCR vs. Lpp for (top to bottom) bulk carriers,
tankers, general cargo vessels and container carriers carriers from
FairPlay database (),used in SHOPERA case studies ( ) and
representative vessels used to define standard wave height ()
At the time of writing, the discussion was still ongoing
whether the standard wave heights should depend on ship
type. On the one hand consequences of accidents, as well as
operational practices, differ between different ship types,
166
especially between passenger and cargo vessels. Besides,
the revealed differences in the marginal wave heights
between different ship types reflect established design and
operational practices, which should not be drastically
influenced by new regulations. Both these arguments are in
favour of correspondingly different standard wave heights.
On the other hand, the revealed difference in the
manoeuvring characteristics in heavy weather reflects,
obviously, also different requirements to the operational
performance of the propulsion and steering systems of ships
of different types, which should not influence the definition
of the minimum safety standards. Obviously, reaching
conclusion regarding ship type-dependency of standard
wave heights requires a prolonged discussion with all
interested stakeholders.
Fig. 11. Marginal significant wave heights according to
Propulsion (top) and Steering (bottom) Ability requirements for
representative bulk carriers (BC), tankers (TA), general cargo
vessels (GC) and container vessels (CV)
Besides the standard significant wave height, also the
relation between wind speed and significant wave height,
seaway spectrum and directional spreading function need to
be defined. The relation between the wind speed and
significant wave height in coastal areas, given in tabular
form in the 2013 Interim Guidelines (resulting from the
IACS studies on minimum power), was formulated by
Shigunov (2015) as an equation hs = 0.316vw , where hs
is significant wave height in m and vw is wind speed in
m/s and confirmed in SHOPERA by the available statistics
of wave height and wind speed in coastal areas. JONSWAP
spectrum with the peak parameter 3.3 and cos2-directional
spreading of wave energy are assumed, as in 2013 Interim
Guidelines.
A final note concerns possible contradiction between
fulfilling the proposed requirements to manoeuvrability in
adverse weather conditions and the possibility to fulfil
EEDI requirements, which are progressively strengthening
from Phase 0 to Phase III. It is interesting to note that
whereas the selected representative general cargo vessels
and container carriers easily fulfil the requirements to Phase
III of EEDI implementation, as presently formulated, the
selected representative bulk carriers and tankers are able to
fulfil requirements of Phase II, while only marginally, if at
all, fulfil Phase III requirements. Note, however, that the
method used here for the selection of representative vessels
picks out, effectively, slow-steaming vessels; whether such
vessels can be considered as representative vessels of fleet
in service was a subject of ongoing discussion at the time of
writing.
REFERENCES
ATSB (2008) Independent investigation into the grounding
of the Panamian registered bulk carrier Pasha Bulker on
Nobbys Beach, Newcastle, New South Wales, 8 June
2007, Australian Transport Safety Bureau (ATSB) Rep.
Marine Occurrence Investigation No. 243
Blendermann, W. (1993) Schiffsform und Windlast-
Korrelations- und Regressionsanalyse von Windkanal-
messungen am Modell, Rep. 533, Institut fr Schiffbau,
Harburg
Brix, J. E. (1993) Manoeuvring Technical Manual.
Seehafen Verlag, Hamburg
EE-WG 1/4 (2010) Minimum required speed to ensure safe
navigation in adverse conditions, submitted by IACS
Fujiwara, T., Ueno, M. and Ikeda, Y. (2006) Cruising
performance of a large passenger ship in heavy sea, Proc.
16-th Int. Offshore and Polar Engineering Conf., Vol. III,
pp. 304-311
IMO (2002) Standards for ship manoeuvrability, Res.
MSC.137(76)
IMO (2012) Interim Guidelines for Determining Minimum
Propulsion Power to Maintain the Manoeuvrability of
Ships in Adverse Conditions, MSC-MEPC.2/Circ.11
IMO (2013) Interim guidelines for determining minimum
propulsion power to maintain the Manoeuvrability in
adverse conditions, IMO Res. MEPC.232(65)
ITTC Manoeuvring Committee (2008) Final report and
recommendations to 25th ITTC, Proc. 25th ITTC, Vol. I
MAIB (1996) Report on the investigation into the
grounding of the passenger roro ferry Stena Challenger
on 19 September 1995, Blriot-Plage, Calais, Marine
Accidents Investigation Branch
MAIB (2002) Report on the investigation of the grounding
of mv Willy, Marine Accidents Investigation Branch
MAIB (2009) Report on the investigation into the
grounding, and subsequent loss, of the roro cargo vessel
Reverdance, Shell Flats Cleveleys Beach, Lancashire,
31. January 2008, Marine Accidents Investigation
Branch
MAIB (2009a) Report on the investigation of the grounding
of Astral on Princessa Shoal, East Isle of Wight, 10
March 2008, Marine Accidents Investigation Branch
MAIB (2012) Report on the investigation of the grounding
of the cargo ship Carrier at Raynes Jetty in Llanddulas,
North Wales on 3 April 2012, Marine Accidents
Investigation Branch
MAIB (2012a) Report on the investigation of windlass
damage, grounding and accident to person on the ro-ro
ferry Norcape, Firth of Clyde and Troon, Scotland, on
26-27 November 2011, Marine Accidents Investigation
Branch
MEPC 62/5/19 (2011) Reduction of GHG emissions from
ships - Consideration of the Energy Efficiency Design
Index for New Ships. Minimum propulsion power to
ensure safe manoeuvring in adverse conditions,
Submitted by IACS, BIMCO, CESA, INTERCARGO,
INTERTANKO, WSC
MEPC 62/INF.21 (2011) Reduction of GHG emissions
from ships - Consideration of the Energy Efficiency
Design Index for New Ships. Minimum propulsion
power to ensure safe manoeuvring in adverse conditions,
Submitted by IACS, BIMCO, CESA, INTERCARGO,
INTERTANKO, WSC
MEPC 64/4/13 (2012) Consideration of the Energy
Efficiency Design Index for new ships Minimum
propulsion power to maintain the manoeuvrability in
adverse conditions, Submitted by IACS, BIMCO,
INTERCARGO, INTERTANKO and OCIMF
MEPC 64/INF.7 (2012) Background information to
document MEPC 64/4/13, Submitted by IACS
MEPC (2014) EU Project Energy Efficient Safe SHip
OPERAtion (SHOPERA), Paper MEPC 67/INF.14
submitted by Germany, Norway and United Kingdom
Sding, H. and Shigunov, V. (2015) Added resistance of
ships in waves, Ship Technology Research -
Schiffstechnik 62(1) 2-13
Shigunov, V. and Papanikolaou, A. (2014) Criteria for
minimum powering and maneuverability in adverse
weather conditions, 14th Int. Ship Stability Workshop
(ISSW 2014), Kuala Lumpur, Malaysia
Shigunov, V. (2015) Manoeuvrability in adverse conditions,
Proc. 34th Int. Conf. on Ocean, Offshore and Arctic
Engineering OMAE2015, St. John's, Newfoundland,
Canada; Paper Nr. OMAE2015-41628
Shigunov, V., Papanikolaou, A. and Chroni, D. (2016)
Maneuverability in adverse conditions: Assessment
framework and examples, Proc. 15th Int. Ship Stability
Workshop, 13-15 June, Stockholm, Sweden
Ventikos, N., Koimtzoglou, A., Louzis, K. and Eliopoulou,
E. (2014) Database of ships and accidents and Risk
analysis of accidents related to manoeuvring in adverse
weather conditions, SHOPERA D1.3, D.1.4, NTUA
167
 Safety Aspects of Autonomous Ships
Sauli Eloranta, Andy Whitehead
Rolls-Royce Marine, Turku, Finland, Sauli.eloranta@rolls-royce.com

ABSTRACT AEROSPACE SAFETY APPROACH

Autonomous shipping represents major challenges to Safety approach and statistics

a wide array of current regulatory framework. Auton-
omy means that the ship senses internal and external The aerospace industry safety approach is based on the
situations and reacts to them by giving commands to probability of risk and the potential severity of the risk:
the Automation & Control system without local or
remote human control. Autonomy can be a way to sim-
plify unmanned ship operations in order to minimize
the need for remote control.
Related technology evolves fast due to the rapid in-
troduction of unmanned vehicles in other areas, such as
aerial drones and unmanned cars. The degree of au-
tonomy is increasing from enhanced and adaptive con-
trols to full situational awareness and machine-based
decision-making. Regulatory changes typically lag be-
hind technical readiness and they are created and ad-
justed based on full-scale piloting and accumulated
safety performance.
Autonomous vessels reduce or completely remove
the risk for onboard crew, having an impact on most
safety rules for ship design and operation. This has a
major impact on risk-based design criteria when there
are no personal safety consequences onboard. Fig.1. Aerospace risk/consequence categorization, FAA
The coexistence of autonomous and manned ships (2008).
needs to be assessed as this phase will take several dec-
ades with the exception of special operating zones for The aviation industrys safety culture differs from ma-
unmanned ships. Autonomy means that there is a more rine industry and highlights the following issues:
pronounced major risk area of Connectivity and cyber - the safety philosophy is holistic and based on
security on top of Powering, Propelling, Steering and safety management systems
Controlling a ship. Also the importance of software de- - the overall aviation safety level is approx. 100
velopment validation and system reliability and func- times better compared to shipping based on the
tionality is increased. The legal liability and responsibil- annual number of total losses / number of fleet,
ity of system providers will increase with the introduc- Sudarshan (2011)
tion of more autonomous applications. - civil aviation incidents are exposed to international
The paper introduces key aspects of aerospace safety investigations
approach and recommends best practices to supplement - aircraft equipment is designed to operate without
current, mainly deterministic safety rules and manual intervention during flight
regulations in Shipping. Then conflicts of various - aircraft maintenance is mandated and based on
shipping related safety rules to autonomous ships are preventative measures
discussed. The objective is to give practical guidance to - aircraft component suppliers are certified and crit-
all involved parties to have a sound approach to safety ical components have special requirements (parts
in autonomous shipping classification, traceability, etc.)
- critical aircraft equipment is exposed to stringent
airworthiness certification approvals and expen-
sive testing (e.g. engine reliability) therefore,
there is a limited number of prototype designs in
the market

168
 Aviation industry has gone through three major phases
in safety culture, Sudarshan (2011):
1. 1920s to 1970s
a. mainly reactive
b. focuses on accidents
c. focus on technology
2. 1970s to 1990s
a. mainly proactive
b. focuses on incidents
c. focus on human factors and regulations
d. regulatory compliance highlighted
3. 1990s onwards
a. mainly predictive
b. focus on organization (leadership, cul-
ture)
c. holistic safety management system
d. data-based identification and prioritiza-
tion of safety risks
e. direct accountability for safety on the part
of senior management
The civil aviation safety records only suffered 2 total
losses in 2014 out of 24,000 aircraft, both due to loss of
control, Boeing (2014). By comparison, there were 85 total
losses of ships in 2015, out of a fleet of approx. 100,000
vessels, Allianz (2016).
International shipping is on a similar safety culture
journey with aviation but with a delay. Fast changes in
technology require risk/goal based approach and predic-
tive measures instead of deterministic legislation.
Only few Marine companies use the aerospace safety
approach in their operations, Rolls-Royce being one of
them.
MARINE SAFETY APPROACH
The Marine regulatory framework consists mainly of
deterministic international and national legislation, tech-
nical rules related to vessel classification and various other
rules and standards. Recently, the trend has been towards
probabilistic and goal-based rules that can be applied to
novel applications.
The introduction of novel technologies highlights con-
tractual obligations and liabilities when the regulatory
framework is not fully in place.
Fig.2. Marine regulatory framework, source: Western Eco-
nomic Diversification Canada.
Vessel autonomy highlights the liability of technology
providers. Digital solutions are typically safety critical and
may represent a single cause of failure (e.g. loss of control
or communication), highlighting the safety impact of all
autonomy-related solutions.
The most important rules affecting national legislation
are provided by IMO. Recently, some nations have given
input to IMO 2018-2023 strategy to include aspects of un-
manned shipping, Berner (2016).
Vessels operating within a single country can be exempt
from international rules by the approval of a national au-
thority. International rules (IMO) evolve slowly due to the
complex approval process by member states and the re-
quired minimum country adoption rate before new rules or
amendments take place (e.g. Ballast Water treatment regu-
lations).
UNMANNED VEHICLES
The introduction of unmanned vehicles is increasing
rapidly in all areas. Only recently these vehicles have start-
ed to blend in with manned vehicles instead of using them
in isolation in subsea activities, mining industry, nuclear
power stations, military applications, etc. The level of au-
tonomy is quite low with all applications and most activities
are remotely human controlled. However, there are en-
hanced controls with automation, such as computer sup-
ported in-flight stability.
The operational safety level needs to be proven to au-
thorities based on long-term full-scale realistic testing. With
small aerial drones, many countries allow them to be oper-
ated within sight, at a maximum altitude and away from
city areas, airports, etc. This is in order to separate un-
manned and manned vehicles. Driverless cars are more
challenging as it is not practical to create a separate traffic
infrastructure for them. Therefore, full-scale tests need to
prove the administration that an acceptable level of opera-
tional safety can be achieved in a hybrid traffic with co-
existing manned and unmanned vehicles. This validation
approach will be needed in all traffic modes. The first fatal
accident with a Tesla car happened in 2016 after 130 mil-
lion miles of autonomous operation this is statistically just
1.5 times safer than with manned cars in the USA, Musk
(2016).
typical
potential # of speed reaction
VEHICLE vehicles km/h time s note
Unmanned car 1200000000 100 0.1s full-scale demonstration relatively difficult
Aerial drone (big) 25000 800 1s small drones allowed and abundant
Unmanned ship 100000 40 10s full-scale demonstration relatively easy
Fig.3. Unmanned vehicle characteristics.
169
LEVELS OF AUTONOMY

Level Description
The computer does everything autonomously,
10 ignores human
The computer informs human only if it (the
9 computer) decides so
8 The computer informs human only if asked
The computer executes automatically, when
7 neccssary informing human
The computer allows human a restricted time to
6 veto before automatic execution Fig.6. Bowtie approach,
https://www.caa.co.uk/Safety-Initiatives-and-Resources/Wo
The computer executes the suggested action if
rking-with-industry/Bowtie/.
5 human approves
4 Computer suggests single alternative The Bowtie approach describes threats as events that
3 Computer narrows alternatives down to a few may cause an unsafe state if not managed with preventative
The computer offers a complete set of decision controls, Fig.6. The top event describes the point where we
2 alternatives no longer have adequate control over the hazard. A top
The computer offers no assistance, human in event can be an unsafe condition but have the potential to
become disaster if nothing is done to control it. Conse-
1 charge of all decisions and actions
quences describe the undesirable events that may potential-
Fig.4. Levels of autonomy, Sheridan, NN (2016). ly result from the top event if the event is not managed with
recovery controls.
UNMANNED SHIPS In unmanned shipping, this approach highlights the im-
portance of threat identification and the creation of preven-
SAFETY REQUIREMENTS tative controls to manage the level of risk (top event taking
place) and recovery controls to reduce the consequences of
Unmanned shipping represents a quantum step in ship- risk.
ping as we know it. Therefore, a mere regulatory compli- The maximum probability of risk leading to any per-
ance is not a sufficient approach. It is recommended that the manent human injuries or other drastic impact needs to be
marine industry applies best practices from the aviation connected to the severity of consequences.
industry by implementing holistic safety management. The
extensive Haddon-Cave (2009) report highlights the fol- Personal Ship Environment
lowing main areas for managing safety (airworthiness): 1 Minor Minor injury Local equipment Local spill
- Leadership Non-severe ship Significant
- Independence 2 Significant Several injuries damage local spill
- People (not just Process and Paper). Severe local
3 Severe One fatality Severe damage spill
- Simplicity
Severe large
4 Very severe 10 fatalities Total loss spill
100 fatalities or Loss of several
5 Catastrophic more ships Very large spill
Fig.7. Severity index, Kretschmann, L. et al. (2016).

Fig.8. Proposed acceptable level of risk for unmanned op-

erations, Kretschmann, L. et al. (2016).

The interpretation of the risk matrix for unmanned op-

Fig.5. Heinrichs Triangle, Haddon-Cave (2009).
erations follows the aerospace model but sets target levels
that are appropriate for unmanned operations, Fig.8. The
broadly acceptable probability of a very severe conse-
quence (total loss) is at 10-3 (100 total losses / year in world
fleet). The total loss of an unmanned ship would cause
damage to financial assets (ship) and the environment but
not on the ships personnel avoiding personal risk would

170
be a new concept in categorizing the maximum probability
of risks.
The probabilistic safety approach requires comprehen-
sive testing and validation to demonstrate the anticipated
safety level of unmanned ships.
DNVGL estimates that 80% of ship accidents are
caused by human error and autonomy may be key in re-
ducing accidents, Tvete and Engelhardtsen (2014). MU-
NIN reports discuss the impact of unmanned operations in
risk probabilities in more detail, MUNIN (2016).
The incident categories collision and foundering are re- Fig.9. Unmanned ship degree of autonomy, MUNIN
sponsible for almost 50% of all total losses in the 2005 to (2016).
2014 period. Thus it clearly represents the category with the
highest incident probability. Furthermore, human errors are
The AAWA project, NN (2016), has stated that un-
a crucial part of the root cause of most maritime accidents.
manned ships must be capable of:
Based on an analysis of collision and foundering scenarios
- generating, or at least using, a valid voyage plan for
for a MUNIN concept vessel and given a proper operational a foreseen sea voyage and assuring the ships read-
and robustness testing, a decrease of collision and founder- iness for the voyage before departure;
ing risk of around ten times compared to manned shipping
- navigating accurately according to the predefined
was found to be possible, mainly due to the elimination of
voyage plan, and avoiding collisions with other
fatigue issues, MUNIN (2016).
traffic and obstacles both fixed and floating - en-
The level of autonomy within the world shipping fleet countered during voyage;
has major impact on risk scenarios: - maintaining its sea worthiness and operability over
1. Manned ships only the voyage as carried out in varying sea states;
a. current basis for practices, culture, regula-
- responding safely to critical events and adjusting its
tions, technology, etc.
operation to potentially dangerous changes in the
b. humans are an integral part of all safety
operating environment and ship conditions;
systems onboard - facilitating emergency interventions for recovery
c. automation exists within human control and rescue at sea; and
(e.g. DP) - resisting unauthorized intrusions into ship systems,
d. human error covers for 80% of accidents
either physical or virtual, with the aim of malicious
e. prevailing scenario for most operating ar-
acts or illegal exploitation.
eas for the next 10 years
2. Hybrid Manned and unmanned ships The EU project MUNIN has identified the following safety
a. challenging co-existence of machine and related scenarios for unmanned shipping, Rdseth and
human controlled ships (mixture of algo-
Burmeister (2015):
rithm and human based decision making)
b. remote operations needed for problem
Normal operational scenarios
solving
1. Open sea mode without malfunctions
c. unmanned ships causing risks for manned 2. Small object detection
vessels (but manned vessels can react as 3. Weather routing
appropriate)
4. Collision detection and deviation
d. scenario for some operating areas in 5-10
5. Periodic status updates to shore control
years time
6. Periodic updates of navigational data
3. Unmanned ships only 7. Release vessel from/to autonomous operation
a. human onboard element can be excluded 8. Manoeuvring mode normal
(remote operation maintains human con-
trol)
Possible problematic scenarios identified
b. algorithm-based decision making in case
9. Flooding detected
of autonomy
10. GNSS (GPS/GLONASS) malfunction
c. increased level of autonomy 11. Manoeuvring mode with malfunctions
d. loss of life related to collisions with 12. Communication failure
manned vessels only
13. On-board system failure and resolution
e. some operating areas in >5 years of time
14. Pilot unavailable: Remote control to safety
15. Piracy, boarding and ship retrieval
Unmanned vessels will still need human input from land
16. Rope in propeller
making connectivity between the ship and the shore crucial. 17. Open sea mode with malfunction
Such communication will need to be bidirectional, accurate, 18. Unmanned ship in search and rescue (SAR)
scalable and supported by multiple systems creating redun-
dancy and minimising risk. Sufficient communication link
The MUNIN project has categorized the following situ-
capacity for ship sensor monitoring and remote control,
ations as safety critical, Kretschmann, L. et al. (2016):
when necessary, has to be guaranteed, NN (2016).

171
 1. Interaction with other ships, whether they follow
COLREGS or not, is a critical issue. Navigation
and anti-collision software must be thoroughly
tested.
2. Errors in detection and classification of small to
medium size objects is critical as it may be
wreckage, persons, life boats or other objects that
need to be reported to authorities. This function
must be carefully tested.
3. Failure in object detection, particularly in low vis-
ibility, can cause powered collisions. The ad-
vanced sensor module must be verified to be able
to do all relevant types of object detection, also in
adverse weather.
4. Propulsion system breakdown will render the ship
unable to move. It is necessary to have a very good
condition monitoring and forecasting system to
reduce such incidents to an acceptable minimum.
5. Very heavy weather may make it difficult to ma-
noeuvre the ship safely. It is necessary to avoid
excessive weather and it is also required to inves-
tigate improved methods for remote control if such
conditions should be encountered.
RELIABILITY
Stochastic reliability
The maximum vessel level failure rate is determined by
the risk / consequence table, Fig.1. As the failure rates are
typically once in 1-100 years, it is not feasible to only per-
form full-scale tests in real operations. Advanced analysis
and simulations are needed to expose autonomous systems
to an accelerated rate of possible scenarios. Full-scale tests
are needed to validate simulations and all anomalies need
to be carefully logged and fed back to the system.
Most reliability issues onboard manned ships are not
safety critical as rules and regulations state requirements for
redundancy of critical equipment (emergency electrical
power, emergency steering, redundant systems, etc.).
Manned ships are routinely maintained during sea voyage,
especially for smaller maintenance issues and it is unclear
what the actual overall reliability level is.
Critical functions
The critical functions of a ship can be categorized as:
1. Communication (remote operations)
2. Control (onboard)
3. Steering
4. Powering
5. Propulsion
6. Navigation
On top of this, several passive functions safeguard the
seaworthiness of a ship, such as stability, watertight or fire
integrity, etc.
Each critical function consists of systems, sub-systems
and components. The overall reliability is the product of
each function, system, sub-system and component.
172
If the intended overall ship reliability level is at 0.99, it
requires (based on each system to comprise of 10
sub-systems):
- each critical function to be roughly at 0.999
- each system under critical function to be 0.9999
- each sub-system to be 0.99999
- each component to be 0.999999
HUMAN IN THE LOOP IN VESSELS SAFETY SYS-
TEMS
Current rules and regulations include human in the loop
for safety:
- humans in control of non-trivial and trivial actions
onboard (activities that require deep domain under-
standing or no domain understanding at all and
anything in between)
- humans in charge of situational awareness (SA), the
interpretation of related data and decisions on ac-
tions (internal and external SA)
- human taking control in case of malfunction as part
of active safety control (e.g. fire hoses and crew
drills to supplement automatic sprinkler system)
- humans doing problem solving (multiple criteria
issues although subjectively)
- humans doing proactive or preventative actions (e.g.
changing operational decisions based on multiple
criteria)
Conventional ships appear to rely strongly on the crew
on-board as an in situ resource for timely failure recovery at
sea and execution of preventive maintenance programs
online during the sea voyage. This allows using less costly
machinery configurations that require frequent preventive
maintenance actions and have lower reliability with respect
to failures repairable at sea.
Lack of permanent crew on-board would essentially di-
minish the capability to perform preventive and corrective
manual maintenance tasks on ship equipment during sea
voyages. This implies that systems essential for operation
need to be designed to be resilient to failure and extended
maintenance intervals. Lack of permanent on-board crew
also creates higher demands for scheduling of maintenance
actions on harbour stays. This calls for the introduction of
efficient diagnostics and new predictive prognostic algo-
rithms to help assessing and controlling the risk of failures
and prescheduling of required maintenance actions as part
of overall ship operation planning. Designing easily main-
tainable systems would help to minimise the time and re-
sources required and to assure that the actions are correctly
performed.
Future vessels will still need human input from land
making connectivity between the ship and the shore crucial.
Such communication will need to be bidirectional, accurate,
scalable and supported by multiple systems creating redun-
dancy and minimising risk. Sufficient communication link
capacity for ship sensor monitoring and remote control,
when necessary, has to be guaranteed, NN (2016).
 Known to us Unknown to us

Known to Known Unknown

others knowns knowns
Unknown Known Unknown
to others unknowns unknowns
Fig.11. A Johari Window, Luft and Ingham (1955).

It is worth noting that risks can be:

- technical
- economical
- organizational
Fig.10. Human or Machine intelligence as part of closed
- behavioural / cognitive
loop control.
- cultural
- legislative
However, humans pose a significant risk to operations
- latent / accumulating (creeping)
due to non-rational behavior:
- various level of probability
- varying degree of knowledge, competence and ex-
perience in world fleet
CYBERSECURITY
- different perception to safety and risk
- limitations in human data processing capacity
Cybersecurity may represent a major risk as a single
- skill degradation (e.g. automation & controls skills
source of failure in unmanned shipping. It can cause several
are more important in the future than mechanics)
critical functions to fail in operating as intended. The foun-
- misinterpretations of data
dation for a sound cybersecurity is created early in the sys-
- unpredictability
tem design process when creating control system architec-
- vessel behavior based on captains personal charac-
ture.
teristics
The level of autonomy impacts the type of cybersecurity
- limited communication and logging of observations
arrangements. Remote operation from a shore control cen-
& decisions
tre typically requires the ship and the centre to be connected
- fatigue, boredom
real-time whereas autonomy lets the vessel operate without
active controlling from shore.
RISKS
SYSTEM DEVELOPMENT PROCESS
Potential risks
An unmanned ship is a complex system, comprising of
Potential risks related to unmanned ships are not all
several physical and virtual critical functions, systems,
known:
sub-systems and components. Safety related requirements
1. Known knowns: Generally well identified risk ar-
are defined at concept level and flowed all the way down to
eas across the industry, such as automation of
component level. Testing, verification, validation and
COLREGS maneuvers, data communication over
demonstration takes place again from component level all
the satellite including cybersecurity, etc. High lev-
the way up to operational concept level.
el of reliability of Formal Safety Assessment
For example, the requirements for Situational Aware-
2. Known unknowns: Limited knowledge but identi-
ness system are based on current human-based and ma-
fied areas of risk: interaction between manned and
chine-enhanced observation capability as a minimum. The
unmanned ships, operations in harsh weather con-
requirements for individual sensors ensure that the fused
ditions, complex operating situations.
sensor information can produce as reliable and as accurate
3. Unknown knowns: Areas that other unmanned ve-
awareness as with humans.
hicles have identified and possibly solved but we
As digital technologies evolve fast, the use of a tradi-
are unaware of: autonomy algorithms of other un-
tional gated engineering process needs to be considered.
manned vehicles, etc.
The concept level solution needs to support frequent up-
4. Unknown unknowns: Areas of unmanned opera-
dates of sub-systems and components together with data
tions with yet unidentified risk areas, e.g. special
analytics improvements without having to redesign concept
cases of cybersecurity.
level architecture.

173

Fig.12. System development V-model, NN (2005)
Fig.13. Example of a system description (controls princi-
ples, NN (2016)
SUMMARY AND CONCLUSIONS
Unmanned shipping represents a quantum step in ship-
ping, affecting all areas of the maritime business ecosystem.
It is imperative that the industry takes a responsible role in
implementing novel technology. Regulatory rules devel-
opment takes place with delay, highlighting the importance
of the industrys self-regulation (industry standard). It is
recommended that all maritime regulatory bodies start to
de-human shipping related safety rules and to remove
explicit rules to demand human presence onboard. The goal
shall be to gain appropriate level of safety for unmanned
operations as there is no personal risk onboard the un-
manned ship.
The automation level onboard ships is steadily increas-
ing as it is not possible for humans to manually control the
functionality of all onboard equipment. Most onboard
equipment is already remotely controlled from control
rooms / navigation bridge. Most controls have embedded
automation for managing multiple control criteria and for
making the control more user-friendly. Some operations are
already autonomous in practice, such as Dynamic Position-
ing where the human takes over only as a back-up. It is
suggested that this human intervention capability is main-
tained long into the future with autonomous vessels through
the use of remote control if needed.
It is important to see autonomous ships as a long-term
vision that accelerates the introduction of several safe-
ty-enhancing solutions. For example, Situational Awareness
is an integral part of an autonomous vessels navigation
system but it can also enhance the awareness of the crew in
a manned ship in limited visibility. These solutions can be
rapidly implemented onboard existing vessels. Situational
awareness requires machine learning and eventually artifi-
cial intelligence in interpreting the data from multiple sen-
sors. Installations that gather as much data from the ships
environment as possible would help the systems learn
174
especially when the systems are used to support an onboard
crew.
Autonomous ships require a step change in ship system
reliability as there is no possibility to manually repair
equipment underway. Reliability issues that are typically
not safety critical onboard manned vessels may be safety
critical in unmanned operations if they lead to loss of one or
more critical functions. The development of more reliable
onboard equipment and systems benefits the ship owner
even in manned operations, resulting in less down time and
reduced maintenance work.
RECOMMENDATIONS
1. Ensure that unmanned ships have equal or better
likelihood of overall safety risks compared to
manned ships.
2. Develop technology that can transfer ships crew ac-
tivities onshore.
3. Progress towards full autonomy in small steps, mak-
ing sure related technology is proven over a suffi-
ciently long time period in real operating conditions
maintain possibility for remote operation.
4. Analyze, simulate, verify and demonstrate the relia-
bility of unmanned ships using the system develop-
ment V-model.
5. Support International Maritime Organization IMO
and Classification Societies to rapidly develop re-
lated rules and regulations to set clear goal-based
safety criteria for unmanned operations.
6. Develop an industry standard to complement inter-
national rules and regulations, in order to self-govern
sustainable and safe solutions.
7. Accelerate technology development of unmanned
ship systems to enhance safety and reliability of
manned ships.
8. Engage specialists from the ICT industry to develop
cybersecurity, communications and software issues.
REFERENCES
Allianz (2016), Safety and Shipping Review, Allianz
Global Corporate & Specialty
Berner, A. (2016), Finlands input on IMO strategy, Inter-
national Maritime Organization
Boeing (2014), Statistical Summary of Commercial Jet
Airplane Accidents Worldwide Operations 1959
2014, Boeing Commercial Airplanes
FAA (2008), Develop Preliminary Vulnerability and Risk
Assessment, Federal Aviation Administration
Haddon-Cave, C. (2009), The Nimrod Review, UK Gov-
ernment
Kretschmann, L. et al. (2016), MUNIN D9.2 Qualitative
Assessment, FP7 GA-No 314286
Luft, J. and Ingham, H. (1955). "The Johari window, a
graphic model of interpersonal awareness". Proc.
Western Training Laboratory in Group Development ,
UCLA, Los Angeles
Musk, E. (2016), Tesla blog, Tesla Motors UK
MUNIN (2016), Research in maritime autonomous sys-
tems - project Results and technology potentials
NN (2005), Clarus Concept of Operations, Publ. FHWA-
JPO-05-072, Federal Highway Administration
(FHWA)
NN (2016), Remote and Autonomous Ships The Next
Steps, AAWA White Paper, Rolls-Royce
Rdseth, .J. and Burmeister, H.C. (2015), Risk Assess-
ment for an Unmanned Merchant Ship, Int. J. Marine
Navigation and Safety of Sea Transportation 9/3, pp.
357-364
Sudarshan, H.V (2011), ICAO Safety Management princi-
ples, NPF/SIP/2011-WP/19
Tvete, H.A. and Engelhardtsen, . (2014), DNV GLs
research within Autonomous Systems, Nor-Shipping
workshop, Professional Insight on Unmanned Ships

175
 Equivalent Level of Safety Approach Fire Safety
Daniel Povel and Andreas Ott
DNV GL Maritime Advisory, Hamburg
daniel.povel@dnvgl.com, andreas.ott@dnvgl.com
ABSTRACT
SOLAS convention allows the assessment and use of
alternative design and arrangements for fire safety by
regulation II-2/17 since 1st July 2002. Various designs
and arrangements have been assessed and approved by
flag states on the basis of the SOLAS regulation II-2/17
and the corresponding guidelines MSC/Circ.1002 so far.
Since 2013 in addition MSC.1/Circ. 1455 serves with
guidelines for the approval of alternatives and equiva-
lents.
The alternative design and arrangements address
alternatives on passive and active fire protection
measures, means of escape and organizational fire safe-
ty measures but also the use of innovative materials or
arrangements.
Until today we have seen for alternative design and
arrangement assessments the application of different
assessment methods. The comparative risk assessment
method uses a comparison of the risks between the al-
ternative design and a prescriptive reference design.
The performance based risk assessment method uses
pre-stated risk performance criteria and the conse-
quence assessment method pre-stated consequence per-
formance criteria. All three methods have their benefits
while the performance based assessment method has to
be applied carefully to fulfil the safety level required by
SOLAS and to validate the assumptions.
This paper highlights the methods available and ex-
perience gained by the performed alternative design
assessment. Further on, the paper will summarize life
safety performance levels calculated for alternative and
reference designs and arrangements.
INTRODUCTION
The equivalent level of safety approach is well known in
various industries. The approach gets relevant if a design or
arrangement does not comply fully with the given prescrip-
tive regulations. The reason for such rule deviation can be
various and is not limited to new developments exceeding
the basis of a regulation or the use of innovative materials
and designs.
176
Within the maritime regulations the equivalent level of
safety approach is applicable in various locations of the
regulatory regime:
1. SOLAS I/5 addressing equivalents in general
2. SOLAS II-1/55 addressing alternative design and
arrangements for machinery and electrical installa-
tions
3. SOLAS II-2/17 addressing alternative design and
arrangements for fire safety
4. SOLAS III/38 addressing alternative design and
arrangements for life safety appliances and ar-
rangements
5. HSC-Code 2000 is based on the principle to serve
in general an equivalent level of safety for high
speed craft than SOLAS and Load Line Conven-
tion is stating for conventional vessels. Further on
the HSC-Code regulation 1.11 allows equivalents
to its rules in general.
6. Diverse rules of classification societies allow
equivalents as well.
The fire safety regulations for conventional vessels are
internationally regulated in SOLAS chapter II-2, FSS-Code
and rules of diverse classification societies. The fire safety
on conventional vessels is based on passive and active fire
protection measures, operational requirements and on
measures addressing adequate escape routes. The active and
passive fire protection measures include preventive and
mitigating measures. SOLAS chapter II-2 provides funda-
mental fire safety objectives and functional requirements
which serve as basis for any prescriptive regulation of that
chapter.
Fire safety objectives of SOLAS II-2:
.1 prevent the occurrence of fire and explosion;
.2 reduce the risk to life caused by fire;
.3 reduce the risk of damage caused by fire to the ship,
its cargo and the environment;
.4 contain, control and suppress fire and explosion in
the compartment of origin; and
.5 provide adequate and readily accessible means of
escape for passengers and crew.
On the basis of the fire safety objectives the following
functional requirements are named:
 .1 division of the ship into main vertical and horizontal
zones by thermal and structural boundaries;
.2 separation of accommodation spaces from the re-
mainder of the ship by thermal and structural
boundaries;
.3 restricted use of combustible materials;
.4 detection of any fire in the zone of origin;
.5 containment and extinction of any fire in the space
of origin;
.6 protection of means of escape and access for
fire-fighting;
.7 ready availability of fire-extinguishing appliances;
and
.8 minimization of possibility of ignition of flammable
cargo vapour

As mentioned before SOLAS regulation II-2/17 allows to

deviate from prescriptive regulations of the corresponding
SOLAS chapter on condition that the given fire safety ob-
jectives and functional requirements are fulfilled. Following
chapter gives a picture of the regulatory background.

REGULATORY BACKGROUND

SOLAS II-2/17
Regulation 17 has been introduced to SOLAS chapter
II-2 back in 2002 with the intent to open the prescriptive
SOLAS rule set for alternative designs and arrangements.
Alternative design and arrangements are fire safety
measures which deviate from the prescriptive requirements
of SOLAS chapter II-2, but are suitable to satisfy the fire
safety objectives and the functional requirements of that
chapter. Proof of this can either be done by a comparative
or performance based approach.
Reasons to implement an alternative design may be e.g.
cost effectiveness, design issues or the introduction of a
novel technology/arrangement that has not been considered
by the prescriptive rule set yet.
Fig.1 Alternative design and arrangements process
flowchart (MSC/Circ.1002)1
MSC.1/Circ.1455
The IMO Guideline for the approval of alternative and
equivalents as provided for in various IMO instruments is
intended for use of both administrations and submitters
when dealing with an approval request for an alternative
and/or equivalent design.
The guidelines serve to outline the methodology for the
analysis and approval process of an alternative and/or
equivalent design and are intended for application when
approving alternative and/or equivalency designs.

MSC/Circ.1002 Amendments to MSC/Circ.1002

The IMO Guidelines on alternative design and ar-
rangements for fire safety are intended to outline the
methodology for the engineering analysis required by SO-
LAS regulation II-2/17. The guideline is meant to give a
clear definition of the applicable process, methods, in-
volved stakeholders and documentation for approval by the
administration. Figure 1 gives the flowchart of the process
which is to be followed.
The scope of any alternative design assessment according
to MSC/Circ.1002 is to show that the alternative design and
arrangements provide the equivalent level of safety to the
prescriptive requirements of SOLAS chapter II-2.
The IMO Maritime Safety Sub-Committee on Ship Sys-
tems and Equipment (SSE) proposed in their 3rd session in
March 2016 amendments to the Guidelines on alternative
design and arrangements for fire safety (SSE 3/WP.4 An-
nex 1). Those amendments are subject for approval at the
ninety-seventh session of the Martime Safety Committee in
November 2016.
The amendment gives a guideline for the selection of life
safety performance criteria. The use of performance criteria
follows the approach not to benchmark the level of safety
of the alternative design against a rule compliant prescrip-
tive design, but against specific criteria in quantitative
terms. Impact of radiant heat exposure, air temperature,
carbon monoxide concentration and reduced visibility on
the available safe egress time (ASET) or alternatively the
fractional effect dose model (FED) should be considered.

1
Formatting error in circular, 5th box says Evaluate performance
of prescriptive vs. proposed system

177

ASSESSMENT METHODS
Comparative risk assessment
The alternative design and arrangements assessment ac-
cording to SOLAS regulation II-2/17 and current version of
MSC/Circ.1002 follows the approach to show that the al-
ternative design and arrangements provide an equivalent
level of safety to the fire safety objectives and the function-
al requirements. In general parlance the scope of an alterna-
tive design assessment is to show, that the risk of the alter-
native design does not exceed the accepted risk level of a
comparable SOLAS compliant reference design. To quan-
titatively compare alternative and reference design, the ef-
fective level of safety of each design has to be determined.
If performance criteria for the assessment of the alternative
design cannot be determined directly from the prescriptive
regulations they may be developed from a prescriptive de-
sign. The performance of prescriptive and alternative de-
sign may then be compared against each other.
In general SOLAS does not quantify the level of safety
or better accepted risk of a prescriptive design. Rather the
in SOLAS regulation II-2/2 given safety objectives and
functional requirements have to be achieved in order to
comply with the by SOLAS accepted risk level. To archive
this risk level the prescriptive requirements of SOLAS II-2
have to be followed. As for different space categories indi-
vidual prescriptive requirements apply, the corresponding
safety level of a given SOLAS rule compliant space cannot
be the exact same as for any other. This results in the need
to individually quantify what the accepted risk or level of
safety for the considered space in a fully rule compliant
design actually would be. To do this, the effects of agreed
design fire scenarios should be assessed under considera-
tion of the applicable frequency of the scenario.
The consequence of a design fire scenario may be deter-
mined with the aid of computer-based fire and evacuation
simulations. Under consideration of several scenarios and
their corresponding probability a value for the risk or level
of safety can be calculated. If the individually calculated
risk of the alternative design is lower or equal than the in-
dividually calculated risk of the reference design, the alter-
native design can be considered to suitable satisfy the fire
safety objectives and the functional requirements.
Summary: The comparative risk assessment requires an
individual assessment of alternative and reference design to
identify and compare the actual level of safety of each de-
sign. Benefit of this approach is that almost any rule devia-
tion may be considered. On the other hand the quantifica-
tion of the rule compliant design requires additional effort
in the assessment process.
Performance based risk assessment
The performance based risk assessment follows in con-
trast to the comparative risk assessment not the approach to
compare an alternative design with prescriptive require-
ments but with the compliance to given risk performance
criteria. Performance criteria serve as benchmarks and can
be used to estimate at what point the acceptable level of
safety is exceeded.
178
To estimate the actual level of safety, equaling the pre-
vailing risk, severity has to be put in context with the fore-
seen frequency of (a) scenario(s). The consideration of sev-
eral scenarios of individual extent and frequency leads to a
quantified value for the overall risk of a design. A prior to
the assessment agreed risk acceptance serves as benchmark
for this approach.
Summary: The performance based risk assessment faces
the challenge to choose the appropriate risk performance
criteria. Advantageous of this approach is that due to the
consideration of risk several design fire scenarios with dif-
ferent severities and frequencies are included. Another
benefit is that the alternative design may be evaluated
against fixed performance criteria which results in the fact
that the assessment of a comparable rule compliant refer-
ence design can be neglected.
Consequence assessment
Major difference of the consequence assessment to the
comparative and performance based risk assessment is that
the latter do consider risk as product of severity and fre-
quency. Or the other way around, a consequence assess-
ment does not consider the frequency of an event.
One way to carry out a consequence assessment is to
assess exclusively one defined scenario. This means that
this approach does not necessarily include a variety of fail-
ure/design fire scenarios.
Aim of the assessment is to identify the severity of the
failure scenario, evaluate if the resulting severity is in line
with the general acceptance and if not to introduce risk
control measures. Disadvantageous about this is that the
worst case, worst credible or whatever chosen scenario does
not necessarily give the full picture of all incident varieties.
Due to neglecting of the frequency of the event it also does
not give the necessary weighting. A scenario with cata-
strophic outcome may have compared to other scenarios a
low risk as its frequency may be almost negligible. An as-
sessment with exclusive focus on consequence ignores this
fact completely.
Summary: The consequence assessment does not consid-
er the frequency of a given scenario. Other than in a risk
assessment, scenarios are not weighted by their probability.
This leads to potential misinterpretation of the results as the
attention of the assessment is not necessarily given to the
most significant incident scenario, design characteristic or
risk control measure.
LIMITATIONS OF FIRE RISK ASSESSMENTS
As any method, model or assumption fire risk assess-
ments are subject to limitations in somehow extent. Limita-
tions may originate from uncertainties of input data or by
limits and boundaries of the applied models. In the follow-
ing a short overview of the most important limitations is
given.
Data available
Risk is defined as the product of frequency times the
probable magnitude of a future loss. The magnitude of a
loss, or severity, can be quantitatively estimated by use of
fire safety engineering methods. The frequency of an inci-
dent has to be derived either from statistical data or esti-
mated by experts opinion. In general statistical data serves
as more reliable source and should be favored.
Fortunately severe fire incident on board of ships are the
exception - unfortunately this results in a small database
with incidents to learn from. Fire incident databases will
never capture the whole spread of fire incidents as most
near miss incidents will most likely never be documented
publicly. The estimated number of unreported cases has to
be derived by exerts opinion. Due to the small number of
reported cases statistical uncertainties for the resulting fre-
quencies are noteworthy.
Fire and Smoke Simulation
Computational fluid dynamics (CFD) is widely used as
an analysis tool in fire safety engineering as it possesses the
ability to handle complex geometries and characteristics of
fire and movement of smoke. Those models solve numeri-
cally thermally-driven flow which gives the ability to fore-
cast the movement of smoke, heat and toxic gases. From
this knowledge the conditions inside the compartment can
be predicted and the available safe egress time estimated.
CFD calculations require high computational effort. It is
not uncommon that the fire and smoke simulation of a large
public space takes several days, even if the simulation is
done on a high performance computing cluster using sever-
al hundred CPUs. This leads to the fact that the number and
variety of predefined scenarios that may be simulated is
restricted.
The experts carrying out the assessment have to ensure
that the most significant design fire scenario has been cho-
sen. On the other hand sensitivity studies are inherent part
of any simulation as models may contain uncertainties.
Further any model has boundaries and restrictions which
need to be known and obeyed.
Evacuation Simulation
Evacuation simulations are used to estimate the required
safe egress time (RSET) to a not by the scenario affected
safe area. MSC.1/Circ.1533 Revised guideline for evacua-
tion analysis for new and existing passenger ships offers
the possibility of using either a simplified or advanced
evacuation method.
The simplified method is a hand calculation which uses
hydraulic equations to predict passenger flows and the re-
quired time for escape. Biggest uncertainty with this meth-
od is that temporal effects cannot be taken into account.
Benefit is that it is fast and easy to carry out, but applied to
an entire ship the calculation may become easily unclear
and defective.
The advanced method requires the use of simulation
software. Each person is presented in the model individual-
ly with a unique set of parameters varying among the indi-
viduals of the population. Those parameters are e.g. re-
sponse time or walking speed and are randomly assigned to
the individuals based on a pre-given statistical distribution.
Any unexpected characteristics like irrational behavior,
panic or deviations of operational procedures can only be
considered up to a very limited extent. The scenarios de-
fined in the guideline are meant to serve as benchmark to
evaluate the overall evacuation performance of a vessel. If
it is the intention to use evacuation simulations in the con-
text of fire risk assessments, assumptions/adoptions of indi-
vidual parameters may have to be taken considering the
casualty scenario which is subject of the assessment.
Risk Calculations
To perform risk calculations an individual risk model
suitable for the application has to be established. Such a
model may be based on e.g. Monte Carlo or risk contribu-
tion trees. Inputs to these models are quantified values for
frequency and severity. The reasons and effects of uncer-
tainties in these factors have been discussed before. On top
additional uncertainties may be introduced due to improper
selection or application of the risk model. Models are in any
case a simplification of reality which results in a system
given limitation of accuracy.
EXAMPLES AND PERFORMANCE LEVEL
Cabin Deck Area
A standard cabin deck area on a cruise ship is normally
designed with one or two longitudinal corridors and may
have in addition transversal corridors. The cabin deck area
is subdivided by various C and B class boundaries. Ac-
commodation spaces on passenger ships like cabin decks
have to be equipped with a dedicated fixed fire & smoke
detection system and with a fixed fire extinguishing sys-
tems.
Fig.2 Exemplary layout of a cabin deck area
within one MVZ
The fire risk for a cabin deck area is represented by dif-
ferent fire scenarios affecting single cabins, other single
spaces or multiple spaces. Highest consequences may be
expected for fire scenarios which affect the corridor and
block all means of escape due to fast smoke propagation.
This may happen in case of a fully developed fire within
one cabin and a blocked open cabin door or in case of a fire
179
within the corridor area itself with a local failure of one
sprinkler nozzle, e.g. by a fire of a housekeeping trolley.
Fig.3 Smoke propagation (A-A) t = 120 s
Fig.4 Smoke propagation (A-A) t = 180 s
In cases of severe fires with smoke propagation within a
cabin corridor during night time smoke is likely to propa-
gate to the whole corridor area before any (sleeping) person
is able to react. Such scenarios can occur even under con-
sideration of the high SOLAS requirements on fire safety
systems like detection and suppression systems.
Fig.5 Smoke propagation cabin deck t = 60 s
Fig.6 Smoke propagation cabin deck t = 120 s
180
Fig.7 Smoke propagation cabin deck t = 180 s
A comparative risk assessment for the cabin deck area
evaluates the fire risks to persons in comparison for alterna-
tive and reference design. The calculated fire risk has to
include different possible fire scenarios as well as their
probabilities. The design fire scenarios have to consider the
failure of fire safety systems. The calculated fire risks for
the cabin area will be based on more frequent fire scenarios
e.g. in a single cabin affecting a limited number of persons
and on more seldom scenarios affecting a higher number of
persons by e.g. smoke propagation into the corridor. The
comparative risk assessment has to compare the calculated
risks for both designs, alternative and reference design,
under consideration.
The performance based risk assessment evaluates the
alternative cabin deck design or layout on the basis of a
chosen acceptance criteria. The fire risk in a cabin deck
varies a lot due to e.g. size, layout, number of accommo-
dated persons, fire load and ignition sources. The prescrip-
tive regulations for cabin areas serve with basic require-
ments for the above mentioned items but do not state a
generally acceptable fire risk for accommodation or cabin
deck areas. Thus the chosen acceptance criterion has to be
well reasonable. Further on the performance based risk as-
sessment is more sensitive on assumptions and input pa-
rameter than a comparative risk assessment.
The alternative design assessment of a cabin deck area by
use of a consequence assessment is much more sensitive
than the comparative or performance based risk assessment.
The result is highly dependent on the chosen design fire
scenarios and the acceptable consequences. As stated be-
fore a cabin deck area may be affected by reasonable fre-
quent fire scenarios affecting a limited number of persons
or by very infrequent (but still possible) scenarios with a
higher number of affected persons. For the second case the
proposed consequence performance criteria as given in the
latest draft amendments of Circ. 1002 are for a SOLAS rule
compliant prescriptive design not fulfilled at all. Thus the
application of the consequence assessment for an alterna-
tive design of a cabin deck area is questionable.
For the alternative design analysis of a cabin deck area it
can be concluded that a risk assessment is required to assess
the acceptability and to state SOLAS equivalent level of
safety. A reduction of the assessment to a consequence as-
sessment of defined design fire scenarios may not be suffi-
cient to state acceptable or equivalent safety.
Large Public Spaces
Cruise ships have a variety of public spaces on board.
Especially large cruise ships have the need on large public
spaces to accommodate a large number of passengers for
dining or entertainment. Exemplary spaces are one or two
deck dining rooms, large theatres and atria.
In public spaces the furniture and decoration material
normally represent a relevant volume of combustibles.
Compared to the cabin areas and according to the nature of
the space we dont have any subdividing boundaries within
the space itself. In case of a fire event the smoke can prop-
agate nearly unhindered throughout the whole space.
Fig.8 Exemplary layout of a public space
Similar to the cabin deck area we may have incidents
with only small consequences due to a prompt reaction of
nearby crew or passengers suppressing the incident fire or a
prompt activation of the fixed fire extinguishing system. In
case of technical or operational failures the fire may devel-
op to a size and a smoke production rate which may result
in higher consequences. Heat, smoke and toxic gases will
propagate below the ceiling throughout the space and po-
tentially block escape routes and affect escaping persons.
Fig.9 Smoke propagation within a public space
A comparative risk assessment is able to consider in the
prescriptive and alternative design the specifics of the as-
sessed public space and of the rule deviation, e.g.: A large
public space within a large main vertical zone (MVZ) will
be assessed by comparing the alternative design with a ref-
erence design exactly with the same use and interior but
scaled down to an acceptable reference size. The risk as-
sessment has to consider the probabilities of the different
scenarios during the quantitative analysis. This is essential
as for large public spaces there will be possible but infre-
quent scenarios with relevant consequences to be consid-
ered.
The performance based risk assessment considers also
the frequency and consequences of different scenarios. As
stated before large public spaces can be different in the use,
size, layout and occupation. When applying risk acceptance
criteria the variation of fire risk within different public
spaces on board the same ship has to be considered.
The consequence assessment will serve with results for
individual design fire scenarios. For large public spaces the
results and conclusions are highly dependent on the chosen
design fire scenarios. Slightly different scenario develop-
ments may result in totally different conclusions when as-
sessing the individual results with pre-given consequence
performance criteria. Due to this the application of the FED
criteria will be fine for some scenarios but may be not ac-
ceptable for some other scenarios in which the smoke starts
propagating and blocking escape routes. Thus the FED cri-
terium has to be connected to clear defined scenario defini-
tions otherwise this seems not to be an objective assess-
ment.
Large public spaces on board of passenger ships should
be as safe as possible as they can accommodate a high
number of persons. The nature of certain risks is that we
cannot reduce them to zero. Any scenario may be worse if
additional failures of safety systems or human errors are
considered. Thus it is seen necessary to do a complete risk
assessment including the assessment of frequencies of in-
cident scenarios or the authorities have to clearly define
pre-given design scenarios.
Life Safety Performance Level
The life safety performance level is not given explicitly
in the SOLAS regulations. The fire safety assessments done
so far for alternative design and arrangements on passenger
ships have resulted in different personal risk figures. The
calculated risk figures for different type of spaces are listed
in Table 1. The given figures only represent spaces assessed
so far by alternative design and arrangement assessments.
They are not validated for all possible kind of layouts and
sizes acceptable by prescriptive SOLAS regulations.
Table 1 Personal Risk Figures on Passenger Ships
Type of Space Personal Risk Figures [year-1]
Cabin deck area 0.4*10-4 to 1.5*10-4
Large public spaces (single deck) 0.3*10-4 to 0.7*10-4
Large multi-deck public spaces
- with smoke extraction system 0.04*10-4 to 0.20*10-4
- without smoke extraction system 0.12*10-4 to 0.25*10-4
The figures shows a range as the personal risk for per-
sons within the different spaces deviates due to the size,
occupation and safety systems implemented in the areas.
The calculated personal risk for cabin areas is calculated
with the highest value. This is mainly driven by fire cases
during the night and a slow response time of sleeping per-
sons. Large public spaces can be distinguished by the type
of spaces. Large single deck spaces have a higher personal
risk as smoke from a fire incident may affect all escape
routes. For multi-deck spaces, especially atriums with a
181
smoke extraction system, the personal risk for passenger MANCE CRITERIA, draft amendments to guidelines
and crew to be affected by smoke is significant lower. on alternative design and arrangements for fire safety
(MSC/Circ.1002), SSE subcommittee, London, 2016.

CONCLUSIONS International Maritime Organisation, MSC.1/Circ.1533

REVISED GUIDELINES ON EVACUATION ANAL-
This paper summarizes the gained experience with alter- YSIS FOR NEW AND EXISTING PASSENGER
native design and arrangement assessment and the corre- SHIPS, London, 2016.
sponding IMO regulations and guidelines.
The three presented methods to assess alternative design Ott, A., Brandtechnologische Sensitivittsanalyse mit nu-
and arrangements have each their benefits and disad- merischer Simulation im Passagierschiffbau, master
vantages. The comparative risk assessment is seen to be the thesis, Hamburg, 2013.
classic assessment method as it compares the performance
of prescriptive design vs. proposed alternative design. Fur- Povel, D. and Bulk, J., Fire and Evacuation Simulations in
ther on the scope of the assessment can be individually ad- Alternative Ship Design, VIB-Tagung, Hamburg, 2011.
dress the proposed rule deviation of the design and thus the
method is most flexible. Povel, D., Ullrich, A. and Ott, A. Risk Assessment for
The performance based risk assessment assesses the al- Passenger Ships, Proc. Design and Operation of Pas-
ternative design on basis of risk performance criteria. These senger Ships, 2013, London, UK, 2013.
may be general or more individual depending on the scope
of the proposed alternative design. The benefits of a per- Povel, D., Evacuation Preparedness of Modern Cruise
formance based risk assessment are to address the risk in a Ship Designs, Cruise Ship Safety Forum, NTSB - Na-
more general manner and not having the need of the com- tional Transportation Safety Board, Washington, 2014.
parison with a prescriptive design. This may help for alter-
native design or arrangements when more than one rule Ullrich. A. and Povel, D. Safety behind the Scenes, GL
deviation has to be assessed or for innovations without any Nonstop 01/2011, Germanischer Lloyd, Hamburg, 2011.
availability of a comparable prescriptive design or ar-
rangement.
The consequence assessment limits the alternative design
and arrangement assessment on the assessment of conse-
quences arising during one or more design fire scenarios.
As the results of the consequence assessment are highly
dependent on the chosen design fire scenarios the scenarios
and the corresponding consequence performance criteria
have to clearly pre-defined. Still there is the risk that a pure
consequence assessment does not consider frequencies of
scenarios and may focus on not relevant scenarios.
For all assessment method it is essential that the experts
who carry out the analysis ensure to choose the applicable
assessment method, quantification model, suitable assump-
tions and the most significant design fire scenarios. Alter-
native design assessments require high effort in the design
and approval process but they give a suitable way to enable
innovation and to strengthen intended designs.

REFERENCES

International Maritime Organisation, SOLAS Safety of

Life at Sea, as amended, London, 2016.

International Maritime Organisation, MSC/Circ. 1002

GUIDELINES ON ALTERNATIVE DESIGN AND
ARRANGEMENTS FOR FIRE SAFETY, London,
2001.

International Maritime Organisation, MSC.1/Circ.1455

GUIDELINES FOR THE APPROVAL OF ALTERNA-
TIVES AND EQUIVALENTS AS PROVIDED FOR IN
VARIOUS IMO INSTRUMENTS, London, 2013.

International Maritime Organisation, GUIDELINES FOR

THE SELECTION OF LIFE SAFETY PERFOR-

182
 Design for Safety in a Changing Climate
Elzbieta Bitner-Gregersen, Erik Vanem and Odin Gramstad
DNV GL Strategic Research and Innovation, 1322-NO Hvik, Norway
Elzbieta.Bitner-Gregersen@dnvgl.com, Erik.Vanem@dnvgl.com, Odin.Gramstad@dnvgl.com
ABSTRACT
The paper addresses the impact of climate change and
associated uncertainties on the safe design of ships. The
Fifth Assessment Report of the Intergovernmental Pan-
el on Climate Change (IPCC) uses four scenarios for
future greenhouse gas concentrations in the atmosphere
called Representative Concentration Pathways (RCP).
Two of these scenarios are applied to investigate how
sensitive the future North Atlantic wave climate is to the
emissions they represent. Winds obtained from six glob-
al climate models have been used to simulate waves for
a historical period at the end of last century and project
waves for a future period towards the end of this centu-
ry for these two scenarios. Based on these projections,
possible changes in extreme wind and waves are inves-
tigated and the associated uncertainties are discussed.
The occurrence of rogue-prone sea states which may
trigger generation of rogue waves in the past and future
climate is also studied. It is shown how the scientific
findings on uncertainties related to climate change pro-
jections and rogue waves can be incorporated in the
risk-based approach used in current design practice of
tankers, and ship structures in general. The potential
effect of climate change on the safety level of current
design practice for tankers is demonstrated. Finally, the
paper discusses how structural design of ships can be
upgraded to account for climate change and rogue
waves without necessarily leading to significant eco-
nomic consequences.
INTRODUCTION
Marine safety is one of the main concerns of the shipping
and offshore industry in general and Classification Societies
in particular. The importance of including the
state-of-the-art knowledge about meteorological
(temperature, pressure, wind) and oceanographic (waves,
current) conditions in ship and offshore standards has been
discussed increasingly by industry and academia in the last
decades in several international forums.
Global warming and extreme weather events reported in
the last years have attracted a lot of attention not only in
academia and media but also in the shipping and offshore
industry. A central question for the maritime industries is:
to what degree will climate change affect future ship traffic
and design of ship structures?
The observed climate changes include natural variability
of climate and anthropogenic climate change. Natural
climate variability is due to the Earths system dynamics,
short term externally forced climate changes (volcanic
activity, short term changes in solar radiation) and long
term external forcing such as tectonic movement, solar
radiation, changes in the Earths orbit and asteroid
bombardment. It has always been present. Antropogenic
climate change is due to human activities and is mostly
associated with emissions of greenhouse gases to the
atmosphere from burning of fossile fuels, but other factors
such as land usage changes and deforestation also play a
role. These two types of climate variations, natural and
anthropogenic, interact with each other. By now, a
consensus has been reached within the scientific
community that human activities contribute significantly to
the observed climate changes (IPCC, 2007, 2013). They
lead to changes in metocean conditions. Notice that natural
variability of climate has been taken into account when
relevant return values are calculated for use in design by
considering sufficiently long meteorological and
oceanographic (metocean) data records.
The Fifth Assessment Report (AR5), which was issued by
the Intergovernmental Panel on Climate Change (IPCC,
2013), is summarising the latest scientific findings regard-
ing climate change. A significant development has taken
place since the issue of the Fourth Assessment Report
(AR4) (IPCC, 2007), particularly in the increased use of
quantitative statistical measures simplifying synthesis and
visualization of climate model performance (see e.g. Sa-
hany et al., 2012). AR5 uses four scenarios for future
greenhouse gas concentrations in the atmosphere called
Representative Concentration Pathways (RCP) with radia-
tive forcing of 2.6, 4.5, 6.0 and 8.5 W/m2 by the end of the
21st century referred to as RCP 2.6, RCP 4.5, RCP 6.0 and
RCP 8.5, respectively. AR5 confirms the conclusion of AR4,
that there are large regional variations in observed and
projected climate driven changes in metocean conditions.
However, some variables such as temperature, sea water
183
level, precipitation, and ice extent received more attention
compared to wind and waves.
Changes in wave and wind climate are expected to have
the largest impact on marine structure design in comparison
with other environmental phenomena. Changes in sea level
combined with storm surge have little potential to affect
ship design directly but may impact offshore and coastal
installations. Further, due to changes in sea level, changes
in harbour and offloading water depths may need to be
taken into account in the future climate. The predicted
increase in fouling from growth of ocean organisms may
increase loads on marine structures in some ocean regions
resulting in higher environmental loads and increase of fuel
consumption. On the opportunity side, retreating arctic sea
ice may open new areas for commercial ship traffic, but this
is associated with additional risks related to operations in
remote and harsh environments.
Climate change in terms of increased storm activity
(intensity, duration and fetch) in some ocean areas, and
changes of storm tracks, may lead to secondary effects such
as increased frequency of occurrence of extreme wave
events (abnormal waves also called rogue or freak waves),
see e.g. Toffoli et al. (2011); Bitner-Gregersen and Toffoli
(2014, 2015).
At present climate change and rogue waves are not
explicitly included in Classification Societies rules and
Offshore Standards due to lack of sufficient knowledge
about uncertainties associated with climate change
projections and no full consensus about the probability of
occurrence of rogue waves.
A decision about possible updates of Classification
Societies rules and standards for ships should be based on
the state-of-art knowledge about climate change projections.
The maritime industry needs to know what changes in
metocean conditions can be expected due to climate change
in different parts of the ocean, what potential consequences
they may have on ship design and the safety of marine
operations, and what methods can be used to account for
these changes. Further, to be able to provide design and
operational criteria accounting for climate change, relevant
uncertainties associated with climate change projections
need to be identified and quantified. This is of particular
importance as uncertainties associated with climate change
projections are currently large.
It may be argued that the potential effects of climate
change can partly be compensated for by adjusting ship
routes based on improved weather forecasts and with the
support of advanced routing systems. However, weather
forecast is affected by uncertainties and may not ensure
avoidance of every storm, and not all types of ships are
routing to the same degree. Further, the technology for
deriving wave heights from marine radars has not yet been
demonstrated in a satisfactory approach; this is under
investigations by, e.g. the JCOMM Wave measurement and
Evaluation (WET) project (www.jcomm.info). Finally,
some ocean areas where an incerase of significant wave
height can take place may be difficult to avoid, e.g. for
ferries, supply ships and FPSOs. Notice that a discussion of
184
whether ship routing can be accounted for in current design
practice is outside the scope of the present study.
Although large uncertainties are associated with climate
change projections, adaptation processes to climate change
has already started in the marine industry. In the Norwegian
Standard NORSOK (2007), being under revision, it is
suggested to increase extreme significant wave height and
wind speed by 4% on q-probability values due to climate
change. This number may need to be updated in the future
when new findings are becoming available.
Further, it is also worth to mention that the oil company
STATOIL has already introduced an internal requirement
accounting in a simplified way for rogue waves when
designing the height of a platform deck (see e.g.
Bitner-Gregersen et al., 2013a). This requirement is now
suggested for implementation in the revised version of
NORSOK (2007) being under hearing. Furthermore, some
revisions of the DNV GL rules for design of superstructures
of passenger ships that account for rogue waves have also
taken place as a result of the EC EXTREME SEAS (2013)
project coordinated by legacy DNV.
The need to address climate changes in guiding
documents became increasingly evident for legacy DNV
and was a motivation for proposing the ExWaCli (Extreme
waves and climate change: Accounting for uncertainties in
design of marine structures) project in 2013
(Bitner-Gregersen et al., 2013b). One of the main objectives
of this project has been to get better insight into climate
changes of wave conditions in the North Atlantic, as well as
the associated uncertainties. Note that the North Atlantic
wave climate is used today as a basis for ship design. Some
results from this project are presented herein with particular
focus on uncertainties associated with climate change
projections.
The paper is organised as follows. Section 2 is dedicated
to definitions of uncertainties; Section 3 shows projected
changes of wind and wave climate in the North Atlantic and
discusses associated uncertainties of these projections.
Occurrence of rogue waves in the future climate is
addressed in Section 4, while impact of climate change on
design is discussed in Section 5. The paper closes with
recommendations for future research needs and
conclusions.
DEFINITION OF UNCERTAINTIES
Different scientific and engineering groups use different
definitions of uncertainties. Herein we adopt the definitions
used by the structural reliability community which have
been developed having Structural Reliability Analysis
(SRA) in mind.
Classification of uncertainties for metocean description
has been suggested by Bitner-Gregersen and Hagen (1990).
The proposed definitions were later generalised and in 1992
included in DNV Rules (DNV, 1992).
Generally, uncertainty related to an environmental de-
scription may be divided into two groups: aleatory (natural
variability) uncertainty and epistemic (knowledge) uncer-
tainty. Aleatory uncertainty (see e.g. DNV, 1992) represents
a natural randomness of a quantity, also known as intrinsic
or inherent uncertainty, e.g. the variability in wave height
over time. Aleatory uncertainty cannot be reduced or elim-
inated.
Epistemic (knowledge) uncertainty represents errors
which can be reduced by collecting more information about
a considered quantity or by improving the measurement
methods. In accordance with Bitner-Gregersen and Hagen
(1990), this uncertainty may be classified into: data uncer-
tainty, statistical uncertainty, model uncertainty and climatic
uncertainty.
- Data uncertainty is due to imperfection of an instrument
used to measure a quantity, and/or a model used for
generating data. If a quantity is not obtained directly
from the measurements but via some estimation process,
e.g. significant wave height, then the measurement
uncertainty must be combined with the estimation or
model uncertainty by appropriate means.
- Statistical uncertainty, often referred to as estimation
uncertainty is due to limited information such as a
limited number of observations of a quantity (sampling
variability) and is also due to the estimation technique
applied for evaluation of the distribution parameters.
The latter can be regarded as the model uncertainty.
- Model uncertainty is due to imperfections and
idealisations made in physical process formulations as
well as in choices of probability distribution types for
representation of uncertainties.
- Climatic uncertainty addresses the representativeness of
measured or simulated wave history for the (future) time
period and area for which design conditions need to be
Fig.1 Flowchart illustrating uncertainties associated with different steps of projections of wind and wave conditions.
PROJECTED CHANGES OF WIND AND WAVE IN THE
NORTH ATLNTIC
Several investigations carried out in the last decades have
provided. A data set has to be sufficiently long to
eliminate climatic uncertainty, e.g. to avoid biasing
towards years characterized by severe winds or by calm
weather only.
To characterise the accuracy of a quantity, e.g. significant
wave height, Hs, it is necessary to distinguish between sys-
tematic error (bias) and precision (random error) with ref-
erence to the true value , which usually is unknown (see
e.g. Bitner-Gregersen and Hagen, 1990).
In the present study we will discuss uncertainties related
to climate change projections of wind and wave character-
istics in the perspectives of the above definitions.
Projections of future wind and wave climate include sev-
eral steps shown schematically in Fig. 1. Different uncer-
tainties are associated with these steps as listed in the figure.
They will all impact the generated wind and wave data, but
in a various degree, and will consequently affect design and
operations of ship structures.
It is noted in particular that uncertainties are especially
large for extremes due to limited number of extreme data
points. Different methods for extreme value analysis were
compared in Vanem et al. (2015) and Vanem (2015), high-
lighting this large uncertainty. One consequence of such
large uncertainties for high quantiles is that it will be very
difficult to find statistically significant changes in extremes.
This is indeed confirmed by the analyses presented in this
paper.
shown that there has been a statistically significant increase
in the mean and extremes of significant wave heights in
several ocean regions over the last half of the 20th century,
e.g. Neu (1984), Bouws et al. (1996), Gulev and Hasse
185
(1998), Gnther et al. (1998), Sterl et al. (1998), WASA
Group (1998), Gulev and Grigorienka (2004), Caires and
Swail (2004). Wang and Swail (2006a, 2006b), Young et al.
(2011), Dobrynin et al. (2012), IPCC (2007, 2013). For a
review of the existing findings see also IPCC (2013), Bit-
ner-Gregersen et al. (2013c). The observed increase of Hs is
shown to be very regional dependent. An overall negative
trend has also been observed in some ocean areas; see e.g.
Dobrynin et al. (2012). Much less attention has been given
to wave period but this is also an important parameter for
ship design.
Two IPPC emission scenarios RCP 4.5 and RCP 8.5 are
analyzed herein and the sensitivity of the future North At-
lantic wind and wave climate to these scenarios is investi-
gated.
The third generation wave model WAM (Reistad et al.,
2011) forced by winds and with ice coverage obtained from
six different global climate models (GFDL-CM3, EC-Earth,
HADGEM2, IPS-CM5A-MR, MRI-GCGCM3 and MI-
ROC5), has been used to project waves for a historical
period and a future period. The grid resolution of the wave
model is 50 km x 50 km and the data are sampled every 3rd
hour. The WAM50 data are obtained by interpolation of the
wind field from the climate models to the 50 km grid. The
spatial resolution of the wind output varies between climate
models and some models have multiple runs (ensemble
members). An overview of the various climate model out-
puts that has been used in this study is presented in Table 1.
The resulting wave data used in the analysis have been
generated by the MET Norway (Aarnes and Reistad, 2015).
Table 1. The CMIP5 global climate models used to force the
wave simulations.
Historical RCP 4.5 /
8.5
Model Res Period # Period # The plots presented in this paper are mostly included for
EC-EARTH 1.125x1.125 1971-2000 3 2071-2100 3 illustrative purposes, and it is re-emphasized that the vari-
HADGEM2-ES 1.875x1.25 1970-1999 1 2081-2099 1 ous models, different ensemble members and different
IPSL-CM5A-MIR 2.5x2.5 1971-1999 3 2071-2099 1 emission scenarios yield variable results. For example the
MRI-CGCM3 1.125x1.125 1971-2000 2 2071-2100 1 spatial maximum mean significant wave height for the
GFDL-CM3 2.5x2.0 1970-1999 1 2071-2100 1 historical period ranges from 2.8 to 4.3 m across the various
MIROC5 1.4x1.4 1971-1999 5 2070-2099 3
The study has shown that the applied emission scenarios,
the selected climate models and their ensemble members
give rather different results, and the variations in the pro-
jections are significant. This makes it difficult to draw firm
conclusions regarding the future wind and wave climate in
the North Atlantic. However, some general observations can
be drawn. Fig. 2 shows the multi-model-multi-run average
of mean significant wave height for the historical period as
well as the projected changes assuming RCP 4.5 and RCP
8.5, respectively. These figures indicate that there might be
a decrease of the mean Hs in large parts of the North Atlan-
tic Ocean, with perhaps a slight increase in Arctic regions.
The range of changes for the multi-model average mean Hs
over the region is from a decrease of 22 cm to an increase
of 61 cm (RCP 4.5) and a decrease of 33 cm to an increase
of 75 cm (RCP 8.5).
186
However, the figures show the average over all runs, and
results vary considerably across model runs. For example,
Fig. 3 shows the projected changes in mean Hs from three
different ensemble members from the same climate model,
the EC-EARTH, for scenario RCP 4.5. It is observed that
there are notable differences, for example along the north-
ern Norwegian coast; one of the runs suggests a slight in-
crease in this area, one indicates a slight decrease and one
indicates no significant differences. Notwithstanding these
differences, most models suggest a general tendency to-
wards lower mean Hs in a future climate over most of the
North Atlantic Ocean. These changes in mean Hs are highly
correlated with projected changes in mean wind speed, as
shown in Fig. 4 (for the same model runs as the plots for
mean Hs). Typically, the correlation between wind speed
and significant wave height is in the order of 0.7-0.8 with
higher correlations in restricted waters.
However, for design and operational purposes changes in
long-term joint distributions of metocean parameters and
extreme values are of most importance. Return periods of
interest may vary from 1-year to 100-years. Obviously,
estimates of return values corresponding to very large re-
turn periods are more uncertain since they depend on ex-
trapolation beyond the support of the data, and there would
be larger variability in higher quantiles of the distributions.
In Fig. 5, the spatial pattern of annual maximum Hs is
shown together with projected changes for the RCP 4.5 and
RCP 8.5 future scenarios, respectively (from one of the
EC-EARTH runs). Now, a much more variable tendency is
seen, with some ocean regions experiencing an expected
increase in extreme Hs. However, it should be noted that in
most locations, the expected increases or decreases in the
extremes are not statistically significant at any reasonable
significance levels, due to the high uncertainties.
model runs. With regards to the estimated differences in the
mean values between the historical and future runs, most
models predict a predominantly decreasing trend of less
than one meter. For the higher quantiles, the differences
between model runs are larger and there are much more
variable differences in the extremes between historical and
future runs. However, due to the large uncertainties these
differences are mostly not statistically significant, although
some exceptions are found with significantly decreasing or
increasing trends at different locations for certain wave
model runs. For an overview of all the individual simula-
tions developed by MET Norway, reference is made to
Vanem (2016a).
As an example, an increase of the 20-year extreme Hs
from 0.2 m up to ca. 2.0 m can be expected in the future
period compared to the historical period when the wave
model is forced by the GFDL-CM3 or EC-EARTH winds.
This increase is slightly higher when the 100-year extreme
Hs is considered, (Bitner-Gregersen, 2016). Further, differ-
ences between Hs extremes calculated for the three
EC-EARTH ensemble members are larger than differences
between projections given by the GFDL-CM3 and EC-
EARTH models. This indicates that natural variability,
represented by ensemble members, may dominate climate
change projections making evaluation of the effect of an-
thropogenic climate changes even more difficult. Again,
these changes are highly regional- and model-dependent.

Fig. 2 Average mean Hs over all models and all runs for the historic period (left) and average projected differences in
mean Hs for RCP 4.5 (center) and RCP 8.5 (right) scenarios, respectively.

Fig. 3 Projected changes in mean Hs for three different model runs from the EC-EARTH model.

Fig.4 Projected changes in mean wind speed for three different model runs from the EC-EARTH model.

187
Fig.5 Annual max Hs for historical period (left) and projected changes in RCP 4.5 (center) and RCP 8.5 (right) from an
EC-EARTH model run.

For marine design applications, the joint distribution of

significant wave height and spectral/zero-crossing wave
period is of importance. Notice that long-term metocean
description will not be affected significantly by sampling
variability (statistical uncertainty due to limited number of
data) but a choice of distributions used to fit the wave data
and a fitting technique applied will affect the results.

Different joint distributions were applied to the ExWaCli

data for a particular location (Vanem, 2016b; Bit-
ner-Gregersen 2016), and the effect of climate change on
the joint description was investigated. Results from this one
location and model runs have indicated that the correlation
between significant wave height and wave period might
increase in the future scenarios, most notably in the ex-
tremes. Fig. 6 shows the fitted joint distribution of Hs and
Tp for a historic and a future run of the EC-EARTH model Fig. 7 Changes in 25-year environmental contours in a
using the RCP 4.5 scenario for location 6 in Fig. 9. This changing climate, GFDL-3CM model, location 6.
figure suggests that the joint distribution may become nar-
rower in a future climate and in some ocean regions the
Such comparison is shown in Fig. 7, where wave data for
conditional standard deviation of wave period given Hs
location 6 from the GFDL-CM3 model have been used to
maybe reduced by up to 70% compared to the historical
construct environmental contours for the 25-year extreme
data.
sea states for the historical period as well as the RCP 4.5
and RCP 8.5 scenarios, respectively. These environmental
contours are notably different for the future scenarios, indi-
cating that ship design might be affected by climatic
changes. Still, these results need to be confirmed by other
studies before firm conclusions can be made.

Intuitively, one would expect that the emission scenario

Fig. 6 Joint (Hs, Tp) distribution fitted to the WAM50 data
from the period (1971-2000) (left) and (2071-2100) (right),
the North Atlantic location 6 (see Fig. 9), total sea, the
EC-EARTH model, ensemble member 12.
Often, environmental contours are used for marine design
to characterize extreme wave conditions. Hence, it could be
of interest to compare estimated contours based on histori-
cal wave data and projected future wave data.
RCP 8.5 should affect the extreme wind and wave values
more than the scenario RCP 4.5, however, the investiga-
tions carried out in ExWaCli have shown that this may not
always be the case. This will depend on wind and wave
characteristics analyzed, and return periods and locations
considered. Further, deviation between predictions of wind
and wave extremes for the RCP 4.5 and RCP 8.5 scenario
will be a function of a return period, as illustrated in Fig. 8
for the GFDL-CM3 model run for the location 6.

188
 Extreme Hs as a function of the return period
GFDL-CM3 model, location 6
22
20
Hs (m)
18
16
RCP 4.5
RCP 8.5
14
0 20 40 60 80 100
Return period (years)
Fig. 8 Extreme Hs as a function of the return period, the
GFDL-CM3 model, location 6.
It is interesting to note also that in some ocean regions the
number of severe winds is increased in the future period.
Further, there are some severe storms in the future period
with wind speeds considerably higher than the largest wind
speed reported in the historical period. This may partly
explain an increase of extreme significant wave heights in
the future climate in some areas of the North Atlantic.
OCCURRENCE OF ROGUE WAVES IN THE FUTURE
CLIMATE
The occurrence of rogue waves, their mechanisms and
detailed dynamic properties are through recent years re-
search efforts now becoming clearer, and consistency be-
tween numerical models and experimental data has been
documented by several studies. A number of different
mechanisms may be responsible for generating these waves
such as linear focusing (frequency or angular), wave
current interactions, crossing seas, quasi-resonant nonlinear
interactions (modulational instability), shallow water effects
and wind (Onorato et al., 2013). Although climate change
and rogue waves represent two different phenomena, the
ongoing work on rogue waves poses an important question
for the shipping industries: Will we see more of these
waves in a changing climate?
Climate change in terms of increase in storm activity
(intensity, duration and fetch) in some regions, and changes
of storm tracks, may lead to secondary effects such as in-
creased frequency of occurrence of extreme wave events
such as rogue waves. The topic has been addressed by Bit-
ner-Gregersen and Toffoli (2015) giving particular focus to
modulational instability.
It has been shown that the occurrence of modulational in-
stability in deep water is characterized by high wave steep-
ness and a narrow wave spectrum, both in frequency and
direction, and can be identified by the Benjamin-Feir Index
(BFI) (see e.g. Onorato et al. (2006, 2013). The BFI is a
measure of the relative importance of nonlinearity and
dispersion. It can be defined as BFI=(kpHs/2)/(/p),
where /p is the frequency spectral bandwidth ( is
often measured as the half-width at half-maximum of the
spectrum and p is the spectral peak frequency). Provided
the wave field is sufficiently steep and narrow banded in
frequency and direction, rogue waves are expected to be
generated when BFI=O(1). Thus reported increase of steep
sea states in the future climate by Bitner-Gregersen and
Toffoli (2015) may increase occurrence of rogue waves.
However, as mentioned the occurrence of a steep sea state
is not alone sufficient to trigger modulational instability. In
addition, the wave spectrum needs to be narrow-banded in
frequency and direction when one wave system is present.
Interestingly enough, however, Onorato et al. (2006,
2010) have shown using Nonlinear Schrdringer (NLS)
equations that the modulational instability and rogue waves
120 can be triggered by a peculiar form of directional sea state,
where two identical, crossing, narrow-banded random wave
systems interact with each other. Such results have been
supported by recent numerical simulations of the Euler
equations by HOSM (Higher Order Spectral Method) and
experimental work carried out at the MARINTEK Labora-
tories (Toffoli et al., 2011). These results also indicated a
dependence of the angle between the mean directions of
propagation of the two crossing wave systems, with a
maximum rogue wave probability for angles of approxi-
mately 40 degrees. Such an unusual sea state of two almost
identical wave systems (approximately the same significant
wave height and mean frequency) with high steepness and
different directions was observed during the accident to the
cruise ship Louis Majesty (Cavaleri et al. 2012).
Occurrence of wind sea and swell having almost the same
spectral period and significant wave height and crossing at
the angle 40o < < 60o has been investigated later by
Bitner-Gregersen and Toffoli (2014). The study has shown
that occurrence of such sea states is location specific, de-
pending strongly on local features of wave climate. These
seas have been observed in the North Atlantic as well as in
the North and Norwegian Seas but only in low and inter-
mediate sea states. Further, the numerical simulations car-
ried out by HOSM have shown that although directionality
has an effect on the occurrence of extreme waves in cross-
ing seas, rogue waves can occur not only for nar-
row-banded wave directional spreading but also when it is
broader. It seems that the most critical condition for occur-
rence of rogue waves in crossing seas is associated with
energy and frequency of two wave systems while the angle
between the wave systems and directional spreading will
decide how large extreme waves will grow. The 40 degree
angle and narrow-banded directional spreading seem to be
generating the largest waves.
In addition to providing various integrated wave parame-
ters such as significant wave height Hs and wave periods Tp
and Tz, the WAM50 runs also provided the full directional
wave spectrum at selected locations (indicated with green
crosses in Fig. 9). From analysis of the full wave spectra we
have been able to investigate the occurrence of sea states
that are known to be related to higher probability of ex-
treme and rogue waves, in the past and future wave climate.
For this particular study 12 locations in the North-Atlantic
were selected, as shown by red squares in Fig. 10.
Based on the current knowledge on rogue waves summa-
rized above, we have considered two types of rogue-prone
sea states. Firstly, sea states that have large BFI (BFI > 1)
and narrow directional spreading < 30. The second type
of sea state we have considered is crossing seas where the
two wave systems propagate with a crossing angle of
189
40-60 and where the two systems have similar values for
Hs and Tp.
Fig. 9 Map showing position for which wave spectra are
available (green crosses). The numbered locations indicated
by red squares show the 12 points selected for further anal-
ysis.
It should first be noticed that such sea states are relatively
rare in the WAM50 data, occurring in about 0.1-1% of all
sea states, depending on location. It is also interesting to
notice that while the sea states with large BFI and narrow
directional spreading mainly occur in high sea sates and are
highly overrepresented among sea states with Hs > 10 m,
the rogue-prone crossing sea states tend to occur in lower
sea states with Hs < 6-8 m, confirming the earlier findings
of Bitner-Gregersen and Toffoli (2014).
When looking at the occurrence of the rogue-prone sea
states in the future climate compared to the historical period
1970-2000, our results show that there are very large varia-
tions, both with respect to different locations and with re-
spect to the different climate models. Fig. 10 shows the
observed change in occurrence of rogue-prone crossing sea
states (in percent) between historical period and future
climate using the RCP 8.5 emission scenario for the twelve
locations given in Fig. 9 and using the six different climate
models. Empty cells indicate cases where no significant
change is detected according to the chi-squared test for
statistical difference between historical and future periods.
As indicated by Fig. 10 there seems to be large uncertain-
ties related to the projections of these sea state properties.
Nevertheless, in certain locations we do see some more
clear results. For example in the northernmost location that
we considered, outside the Norwegian coast of Finnmark
(location 12 in Fig. 9), a significant increase in the range
110%-360% in the number of rogue-prone crossing seas is
observed for all six climate models. This shows that in
some ocean regions one may see a significant change in the
number of sea states with higher probability of rogue waves,
but that this is strongly location dependent, and that in gen-
eral the uncertainties related to the climate projections for
190
such parameters are still very large at present.
Fig. 10 The observed change in occurrence of rogue-prone
crossing sea states (in percent) between historical period and
future climate using the RCP8.5 scenario, in 12 different
locations in the North-Atlantic for the six different climate
models. Empty cells indicate cases where no significant
change is detected.
IMPACT OF CLIMATE CHANGE ON DESIGN
The traditional format of Classification Societies Rules is
mainly prescriptive, without a transparent link to an overall
safety objective. The International Maritime Organization
(IMO, 1997, 2001) has developed guidelines for use of the
Formal Safety Assessment (FSA) methodology in rule
development, which will provide risk based goal-oriented
regulations that are well balanced with respect to acceptable
risk levels and economic considerations. Formal Safety
Assessment consists of five inter-linked steps; (1) a hazard
identification, (2) a risk analysis (3) identification of risk
control options, (4) a cost benefit assessment of the risk
control options, and (5) a recommendation to decision
makers. Risk is defined in FSA as a product of probability
of occurrence of a hazard multiplied by its consequence
estimated usually as a monetary value (see e.g. Skjong and
Bitner-Gregersen, 2002). FSA will be used when consider-
ing implementation of climatic changes and rogue waves in
design practice. In this case an increase of Hs or occurrence
of a rogue wave will represent a hazard.
When performing FSA for ship structures it is beneficial
to apply Structural Reliability Analysis (SRA) in the risk
assessment (step 2) and the cost-benefit assessment (step 4).
The reliability methods (see Madsen et al., 1986) allow
quantifying in a probabilistic way the uncertainties in the
different parameters that govern the structural integrity.
This allows reliability assessment of structural components
or a structure. Further, reliability-based design of a struc-
tural component (or a structure) provides a means to satisfy
target reliability with respect to specific modes of failure.
The probabilistic approach can be used for calibration of
partial safety factors in the development of LRFD (Load
and Resistance Factor Design) codes, and for development
of acceptance criteria for structural designs, confer DNV
CN 30.6 (1992), ISO 2394 (1998), Bitner-Gregersen et al.
(2002), Skjong and Bitner-Gregersen (2002) and Hrte et al.
(2007a, 2007b). Standard software allowing carrying out
structural reliability calculations is available within industry.
Also, complicated nonlinear effects can be included by
embedding a time domain simulation code in a reliability
code, such as the PROBabilistic ANalysis code PROBAN
(DNV, 2002).
Fig. 11 shows schematically how climate change trends
and rogue waves can be accounted for in the risk approach
based on the FSA methodology. Rogue waves can be in-
cluded directly in the metocean description (Alternative 1)
or as a correction in a load distribution (Alternative 2).
When considering accounting for climate change and
rogue waves on marine structure design a distinction will
need to be made between existing structures and new ones.
SRA is recommended to be used for checking whether the
existing structures will maintain the same safety level as
Fig. 11 Risk based approach, overview of interfaces where climate change and rogue waves are integrated.
The North Atlantic joint (Hs, Tz) model adopted by CSR
for Tankers (IACS, 2010; IACS, 2000), with a 3-parameter
Weibull distribution for Hs and a conditional lognormal
distribution for the zero-crossing wave period Tz given Hs,
has been modified to reflect the climate change. The Hs
distribution has been shifted by a constant value corre-
sponding to the specified increase of Hs. The formulation of
the conditional distribution of Tz has been kept unchanged
(Bitner-Gregersen et al., 2011, 2013c, 2015). This simplifi-
cation is considered to be acceptable for extremes but not
for fatigue calculations. Three cases have been considered:
Case a) increasing extreme Hs by a constant value, Case b)
modifying the joint distribution using closed form expres-
sions for the Weibull scale and location parameter (the
shape parameter is kept constant) proposed by Vanem and
Bitner-Gregersen (2012), Case c) assuming that the in-
creased extreme significant wave height instead of being a
fixed value (as in Case a) is normally distributed with
COV=0.25.
current design.
In light of the findings regarding potential changes of sig-
nificant wave height summed up in the previous section and
the results of the EC EXTREME SEAS project, projected
impacts that climate change and rogue waves may have on
the design of tankers is illustrated below as an example.
The IACS Common Structural Rules for Tankers, IACS
(2010), has been used to demonstrate these effects on hull
girder collapse of tankers. Five oil tankers, ranging from
Product Tanker to VLCC (Very Large Crude Carrier) have
been considered. Notice that the ship length of the tankers
is ranging from 174.5 m (Product Tanker) to 320 m
(VLCC). An increase of the extreme significant wave
height by 0.5 m, 1.0 m, 1.5 m and 2.0 m has been investi-
gated, see Bitner-Gregersen et al. (2015) for details.
The above described approaches can be used when only
historical wave data are available. Alternatively, if climate
data exist, the joint (Hs, Tz) distribution could directly be
fitted to climate data and applied in the reliability calcula-
tions; this is not done herein due to large uncertainties asso-
ciated with climate change projections.
The three approaches for including effects of climate
changes in the joint (Hs, Tz) distribution give approximately
the same results. Case b shows a slightly higher probability
of failure for the largest increase of Hs compared to Cases a
and c. The difference between Case a and Case c is negligi-
ble indicating that the increase of the most probable ex-
treme Hs, reflecting systematic error, has much larger effect
on current design practice than the random error associated
with the extreme Hs prediction.
As an example the annual probability of failure for the
considered Aframax (length 263 m) as a function of the
deck area (total cross sectional area of deck plate and stiff-
191
eners) when modification of (Hs, Tz) distribution according
to Case c is applied is shown in Fig. 12. The deck area
equal 1 refers to the initial input design of the ship (Base
Case) as used by CSR for Tankers (IACS, 2010).
Fig. 12 Annual probability of failure, Aframax, the modi-
fied (Hs, Tz), Case c.
The investigations have shown that an increase of the sig-
nificant wave height beyond 0.5 m in the North Atlantic
will have a large impact on the current design practice for
tankers. The steel weight of the deck in the midship region
may need to be increased to maintain the current safety
level. Longer ships would require a larger increase in the
ship deck area. If the significant wave height increases by
1.0 m, the deck area will need to be increased by 9% for a
VLCC and by 5% for a Product Tanker. Increasing the hull
girder strength in a more optimal way without increasing
the steel weight, not addressed herein, could also be con-
sidered.
In this context it should be noticed that the results pre-
sented here are for "net scantlings", which is part of the
explanation of, what some would say, relatively high failure
probabilities. Gross scantling, which is the net scantling
plus the corrosion addition, would reduce the failure proba-
bilities by approximately an order of magnitude.
Similar trends as for Hs increase due to climate change
have been found when the bending moment due to presence
of rogue waves is increased (Bitner-Gregersen et al., 2015).
A 10% increase of the bending moment would increase the
cost of ship deck by approximately 10% in order to main-
tain the safety level. Rogue waves leading to increase of
wave bending moment below 5% will not have significant
impact on reliability level of current design practice of
tankers.
DISCUSSION AND FUTURE RESEARCH NEEDS
In this paper, climatic changes in wind and waves condi-
tions that are important for the safe design of ship and other
marine structures have been addressed. These are mostly
related to potential changes in various wave- and wind
parameters. In order to assess the expected changes in these
parameters, one needs to rely on climate simulations and
projections from global climate model as well as numerical
wave simulations from physical wave models. The global
192
climate models provide historical and future wind fields
that can be used to force the numerical wave models to-
gether with simulated ice coverage from the climate mod-
els.
The results indicate that the range of possible changes in
the operational environment of ships is sufficiently large to
have an effect on the safety of marine operations and ship
design. However, the study also reveals that there are large
uncertainties associated with the expected changes in rele-
vant wave and wind parameters.
There is large variability across different climate models
and across different runs from the same model. This is due
to the fact that current state-of-the-art climate models per-
form poorly when it comes to predicting surface winds.
Compared to other climate variables such as temperature,
precipitation and sea water level rise, wind fields are much
harder to describe (see de Winter et al., 2013; Zappa et al.,
2013). Also, the extent of sea ice coverage is highly uncer-
tain, and for some of the climate models explored in this
study, the ice coverage seems to be clearly inaccurate. The
large variability in wind output, as well as the ice coverage,
translates directly to the high uncertainty of simulated wave
fields. It should be noticed that physical wave models are
affected also by various uncertainties associated with mod-
els assumptions and models resolution that will impact
final results. The latter may give biased extremes (Bit-
ner-Gregersen and Toffoli, 215).
Previously, the climate community has put more emphasis
on temperatures, sea level rise and precipitation, and less on
the prediction of surface winds. In order to obtain improved
wave projections it is important to develop climate models
with improved capabilities of predicting winds. It is there-
fore necessary for the wave modelling community as well
as the various stakeholders within the maritime industries to
communicate this clearly to the climate scientists so that
next generation climate models improve on this. There are
currently initiatives to develop fully coupled climate and
wave models and combined with enhanced spatial resolu-
tion so there are reasons to believe that the uncertainties
associated with wave climate projections may be signifi-
cantly reduced in the future. The maritime industries need
to pay close attention to such developments in order push
state-of-the-art knowledge about climate changes relevant
for the safety of marine operations and for ship design.
Secondary effects of climate change such as increased
frequency of occurrence of rogue waves needs also to get
attention. This will require further insight into potential
changes of wave spectra. Recently met-offices have started
to archive wave spectra from hindcast runs providing new
opportunity for such studies. The marine industry would
much welcome access to these data.
In the study presented herein, the ship deck area of tank-
ers is increased following a relatively simple recipe where
both the stiffeners and plate thicknesses are modified, but
not the frame spacing or number of stiffeners. It might be
possible to increase the hull girder strength in a more opti-
mum way using less steel; this should get attention in the
coming years.

Further, the Portuguese ship yard ENVC has demonstrat-
ed in the EXTREME SEAS (2013) project innovative de-
sign of three vessels built recently by ENVC (Ro-Pax Ferry,
Heavy-Lift Container and Asphalt Tank Carrier) and has
shown that extra building costs related to accounting for
rogue waves could be marginal or none, depending on how
initial design and High Strength steel quality is used. Simi-
larly, the German ship yard Meyer Werft (MW) in collabo-
ration with legacy GL Hamburg and the University of
Duisburg -Essen has shown in EXTREME SEAS that rede-
sign of a cruise ship superstructure to stand against rogue
waves can be done when the total price for a cruise liner
will increase slightly but the operational costs will remain
unchanged.
CONCLUSIONS
The study has demonstrated that there are large uncertain-
ties associated with climate change projections of the North
Atlantic wind and wave conditions at present. The various
models, different ensemble members and different emission
scenarios yield variable results which are much ocean re-
gion and location dependent. These uncertainties will also
affect predictions of occurrence of rogue waves. Therefore
firm conclusions regarding climate changes are difficult to
reach today.
Although considerable uncertainty remains in prediction
of the future wind and wave climate and there may be sig-
nificant regional differences, current best estimates of ob-
served and projected climate change indicate that in the
coming decades it is likely that marine structures may ex-
perience higher environmental loads in some regions.
The maritime industry has methods and various tools,
although some may need to be enhanced, to deal with cli-
mate changes but systematic evaluation of their impact on
loads and responses of different ship types is still lacking.
Such studies are needed to quantify potential implications
of climate change on design and marine operations as well
as on related economic consequences. Further, they are
necessary for specification of the existing margins in cur-
rent design practice. Retaining the current safety level in
rules and standards during this process is crucial. In addi-
tion, costs associated with possible accounting for climate
change should be kept low through the introduction of in-
novative designs.
A decision regarding possible updates of Classification
Societies rules and standards for ships, and marine struc-
tures in general, should be based on the state-of-art
knowledge about climate change projections. Therefore the
marine industry needs to closely follow developments on
climate change and collaborate with the climate and wave
community. The latter is essential in order to push
state-of-the-art knowledge about climate changes relevant
for design and the safety of marine operations. Establish-
ment of expert panel(s), including external researchers and
users, to discuss results and exchange information about
climate changes is highly recommended. It should be men-
tioned that the shipping community is interested not only in
climate change projections but also in uncertainties associ-
ated with them.
ACKNOWLEDGEMENTS
This work has been supported by the Norwegian Re-
search Council project ExWaCli (the Project No. 226239).
The authors are indebted to the Norwegian Meteorological
Institute for providing the wind and wave climate data.
REFERENCES
Aarnes, O.J., Reistad, M. (2015), Projection of Wind and
Wave Climate in the North Atlantic. ExWaCli Memo.
Bitner-Gregersen, E.M. and Hagen, . (1990), Uncertain-
ties in Data for the Offshore Environment, Struct. Safety,
Vol. 7, pp 1134, doi:10.1016/0167-4730 (90)90010-M.
Bitner-Gregersen, E.M., Hovem, L. and Skjong, R. (2002),
Implicit Reliability of Ship Structures. Proc. OMAE
2002 Conf., Oslo.
Bitner-Gregersen, E.M., Hrte, T.F. and Skjong, R. (2011),
Potential Impact on Climate Change on Tanker De-
sign. Proc. OMAE 2011 Conf., Rotterdam.
Bitner-Gregersen, E.M. (chairman), Bhattacharya S. K.,
Chatjigeorgiouc, I.K, Eames, I., Ellermann, K. Ewans, K.
Hermanski, G., Johnson, M.C., Ma, N., Maisondieu, Ch,
Nilva, A., Rychlik, I., Waseda, T. (2013a), Reply to the
Official Discusser, ISSC 2012 Committee I.1 Environment.
Proc. ISSC 2012, Vol.3, Rostock, Germany.
Bitner-Gregersen, E.M., Eide, L.I., Reistad, M., Huseby
Bang, A., (2013b), ExWaCli : Extreme waves and cli-
mate change: Accounting for uncertainties in design of
marine structures. Proc.13th Int. Workshop on Wave
Hindcasting and Forecasting, Banff, Canada).
Bitner-Gregersen, E.M., Eide, L. I., Hrte, T., Skjong, R.
(2013c), Ship and Offshore Structure Design in Climate
Change Perspective. SpringerBrief in Climate Studies,
Monograph, Springer, ISBN 978-3-642-34137-3.
Bitner-Gregersen, E.M. and Toffoli, A. (2014), Probability
of Occurrence of Rogue Sea States and Consequences for
Design of Marine Structures. Ocean Dynamics, Vol. 64
pp 1457-1468.
Bitner-Gregersen, E. M., Eide, L. I., Hrte, T. and Vanem, E.
(2015). Impact of Climate Change and Extreme Waves
on Tanker Design, SNAME Transactions 2014, USA.
Bitner-Gregersen, E.M. and Toffoli, A. (2015). Wave
Steepness and Rogue Waves in the Changing Climate in
the North Atlantic, Proc. OMAE 2015, St. John's, Can-
ada.
Bitner-Gregersen, E.M., (2016). Potential Changes of the
North Atlantic Wind and Wave Climate and Occurrence
of Rogue Waves, Proc. OMAE 2016 Conf., Busan,
South Korea.
Bouws, E., Jannink, D. and Komen, G. J. (1996), The
Increasing Wave Height in the North Atlantic Ocean,
Bull. Amer. Meteor. Soc., Vol. 77, pp 2275-2277.
Caires, S. and Swail, V. R. (2004), Global Wave Climate
Trend and variability Analysis. Proc. 8th Int. Workshop
on wave Hindcasting and Forecasting. Oahu, USA.
Cavaleri, L., Bertotti, L., Torrisi, L. Bitner-Gregersen, E.,
Serio, M. and Onorato, M. (2012), Rogue Waves in
Crossing Seas: The Louis Majesty accident, J. Geo-
193
 physical Research, Vol. 117, CXXXXX,
doi:10.1029/2012JC007923
DNV (1992), Structural Reliability Analysis of Marine
Structures. Classification Note 30.6. July 1992.
DNV (2012), PROBAN Theory, General Purpose Proba-
bilistic Analysis Program, L. Tvedt, DNV Report, Ver-
sion 4.4, Hvik, Norway.
Dobrynin, M., Murawsky, J. and Yang, S. (2012),
Evolution of the Global Wind Wave Climate in CMIP5
Experiments. Geophysical Research Letters, Vol. 39:
L18606.
EXTREME SEAS (2013), Design for Ship Safety in Ex-
treme Seas (EXTREME SEAS) [Online]. Contact per-
son: Bitner-Gregersen, E. M. Available:
http://cordis.europa.eu/result/rcn/55382_ en.html.
Gulev, S.K. and Grigorieva, V. (2004), Last Century
Changes in Ocean Wind Wave Height from Global Vis-
ual Wave Data, Geophysical Research Letters, Vol. 31,
L24302, doi:10.1029/2004GL021040.
Gulev, S.K. and Hasse, L. (1998), North Atlantic Wind
Waves and Wind Stress fromVoluntary Observations, J.
Phys. Oceanogr., 28, pp 1107-1130.
Gnther, H., Rosenthal, W., Stawarz, M., Carretero, J. C.,
Gomez, M., Lozano, I., Serano, O. and Reistad, M.,
(1998), The wave climate of the Northeast Atlantic over
the Period 1955-94: the WASA wave hindcast, Global
Atmos. Ocean System, Vol. 6, pp 121-163.
Hrte, T., Skjong, R., Friis-Hansen, P., Teixeira, A.P. and
Viejo de Francisco, F. (2007), Probabilistic Methods
Applied to Structural Design and Rule Development.
Proc. Conf. Development of Classification & Interna-
tional Regulations, London, UK.
Hrte, T., Wang, G. and White N. (2007b),Calibration of
the Hull Girder Ultimate Capacity Criterion for Double
Hull Tankers, Proc. 10th Int. Symp. Practical Design of
Ships and Other Floating Structures, Houston, USA.
IACS, (2000), Standard Wave Data for Direct Wave Load
Analysis. IACS Recommendation No.34, (the 2000 ver-
sion updated in 2001).
IACS (2010), Common Structural Rules for Double Hull
Oil Tankers with Length 150 metres and above. Rules
for Classification of Ships., Vol. 8, 1, July 2010.
IMO, (1997), "Interim Guidelines for the Application of
Formal Safety Assessment (FSA) to the IMO Rule
Making Process", Maritime Safety Committee, 68th ses-
sion, June 1997; and Marine Environment Protection
Committee, 40th session, September 1997.
IMO, (2001), "Guidelines for Formal Safety Assessment for
the IMO Rule Making Process", IMO/Marine Safety
Committee 74/WP.19, 2001.
IPCC, (2007), The Fourth Assessment Report: Climate
Change (AR4): The AR4 Synthesis Report, the Working
Group I Report: The Physical Science Basis (ISBN 978
0521 88009-1 Hardback; 978 0521 70596-7 Paperback),
the Working Group II Report Impacts: Adaptation and
Vulnerability, the Working Group III Report: Mitigation
of Climate Change, Cambridge University Press.
IPCC, (2013),Climate Change 2013: The Physical Science.
Contribution of Working group l to the Fifth Assessment
Report of the Intergovernmental Panel on Climate
Change, Cambridge University Press, Cambridge, UK
and New York, USA.
194
ISO 2394, (1998), General Principles on Reliability for
Structures.
Neu, H.J.A. (1984), Interannual Variations and Longer-
term Changes in the Sea State of the North Atlantic from
1970 to 1982, J. Geophys. Res., Vol.89(4), pp.6397
6402.
NORSOK, (2007), Standard N-003: Action and action
effects, Rev. 2, under revision.
http://www.standard.no/pronorm-3/data/f/0/03/78/7_107
04_0/N-003d2r2.pdf.
Madsen, H.O., Krenk, S. and Lind, N.C. (1986), Methods
of Structural Safety. Prentice-Hall, Enlewood Cliffs,
USA.
Onorato, M., Osborne, A., Serio, M., Cavaleri, L., Brandini,
C.,and Stansberg, C. Extreme waves, Modulational In-
stability and Second Order Theory: Wave Flume Exper-
iments on Irregular waves, Europ. J. Mech. B/Fluids,
Vol. 25 (2006), pp 586601.
Onorato, M., Proment, D. and Toffoli, A. (2010), Freak
Waves in Crossing Seas, European Physical Journal,
Vol. 185, pp 45-55.
Onorato, M., Residori, S., Bortolozzo, U., Montina, A. and
Arecchi, F. T. (2013), Rogue Waves and their Generat-
ing Mechanisms in Different Physical Contexts. Physics
Reports, Vol. 528, pp 4789.
Reistad, M., . Breivik, H. Haakenstad, O. J. Aarnes,
Furevik, B.R., and Bidlot, J.-R. (2011), A
high-resolution hindcast of wind and waves for the North
Sea, the Norwegian Sea and the Barents Sea, J. Geo-
physical Research, Vol. 116: C05019.
Sahany, S., Neelin, J.D., Hales, K. and Neale, R.B., (2012),
Temperaturemoisture Dependence of the Deep
Convective Transition as a Constraint on Entrainment in
Climate Models, J. Atmos. Sci., Vol. 69, pp 13401358.
Skjong, R. and Bitner-Gregersen, E.M. (2002) Cost
Effectiveness of Hull Girder Safety, Proc. OMAE 2002
Conf., Oslo, Norway.
Sterl, A., Komen, G. J. and Cotton, P. D. (1998), Fifteen
Years of Global Wave Hindcasts using Winds from the
European Centre for Medium-Range Weather Forecast
Reanalysis: Validating the Reanalyzed Winds and As-
sessing the Wave Climate, J. Geophys. Res., Vol. 103,
pp 5477-5494.
Toffoli A., Bitner-Gregersen, E.M., Osborne, A. Serio, M.,
Monbaliu, J. , Onorato, M. (2011), Extreme waves in
random crossing seas: Laboratory experiments and nu-
merical simulations. Geophys. Res. Lett., Vol. 38(2011):
L06605, doi: 10.1029/201.
Vanem, E., Bitner-Gregersen, E.M. (2012), Stochastic
Modelling of Long-term Trends in the Wave Climate and
its Potential Impact on Ship Structural Loads. Applied
Ocean Research, Vol. 37(2012), pp 235 248.
Vanem, E., Bitner-Gregersen, E. M., Eide, L. I., Garr, L.
and Friis Hansen, P. (2015), Uncertainties of Climate
Modeling and Effects on Wave Induced Bending Mo-
ment, SNAME Transactions 2014, USA.
Vanem, E. (2015). Uncertainties in Extreme Value Mod-
elling of Wave Data in a Climate Change Perspective, J.
Ocean Engineering and Marine Energy, Vol. 1(4), p.
339-359.
Vanem, E. (2016a). Initial Analysis of ExWaCli data.
DNV-GL Report, No. 2016-0423.
Vanem, E. (2016b), Joint Statistical Models for Significant
Wave Height and Wave Period in a Changing Climate,
Marine Structures, Vol. 49, pp 180-205.
Wasa Group, (1998), Changing Waves and Storms in the
Northeast Atlantic?, Bulletin of the American Metero-
logical Society, Vol. 79(5).
Wang, X.L. and Swail, V. R. (2006a), Historical and Pos-
sible Future Changes of Wave Heights in Northern
Hemisphere Ocean. Atmosphere-Ocean Interactions
[Perrie, W. (ed.)]. 2, Wessex Institute of Technology
Press, Southampton, pp 240.
Wang, X.L. and Swail, V.R. (2006b), Climate Change
Signal and Uncertainty in Projections of Ocean Wave
Heights, Climate Dynamics, Vol. 26, 109126 DOI
10.1007/s00382-005-0080-x.
de Winter, R.C., Sterl, A. and B.G. Ruessink. "Wind
extremes in the North Sea Basin under Climate Change:
An Ensemble Study of 12 CMIP5 GCMs." J. Geophysical
Research: Atmospheres, Vol. 118(2013), pp 1601-1612.
Young, R.I, Zieger, S. and Babanin, A.V., 2011. Global
trends in wind speed and wave height. Science, 332, 22
April 2011, pp 451455.
Zappa, G., Shaffrey, L.C., Hodges, K.I., Sansom, P.G. and
Stephenson, D.B., 2013. A Multimodel Assessment of
Future Projections of North Atlantic and European Ex-
tratropical Cyclones in the CMIP5 Climate Models, J.
Climate, Vol. 26, pp 58465862. doi: http://dx.doi.org/
10.1175/ JCLI-D-12-00573.1.

195
 Challenges for Application of Risk Based Design
Approaches for Arctic and Antarctic Operations

Kujala Pentti1, Kmrinen Jorma2, Suominen Mikko1

1. Aalto University, School of Engineering, D. of Mech. Eng., Finland, pentti.kujala@aalto.fi
2. Finnish Transport Safety Agency, TRAFI, Finland

ABSTRACT
Selection of suitable ice class for ships operation is an
important but not simple task. The process of selecting
an appropriate ice class is of high importance both from
a safety as well as an economical perspective, but the
selection process is still based on accumulated experi-
ence and traditions within the areas of existing opera-
tions. The increased exploitation of the Polar waters,
both seasonal periods and geographical areas, as well as
the introduction of new international design standards
such as Polar Code, reduces the relevancy of using ex-
isting experience as basis for the selection, and new
methods and knowledge have to be developed. This
paper will critically review the current knowledge of
applying risk based design methodologies for ice cov-
ered waters. To illustrate the current best practice, the
new polar code and especially the new Polar Operation-
al Limit Assessment Risk Indexing System (POLARIS)
is applied for two vessel: one navigating independently
in the Antarctica (SA Agulhas II) and one navigating
with icebreaker assistance on the Kara Sea on the Rus-
sian Arctic (MT Uikku) to evaluate the suitable ice class
for these waters. It is found the PC 3 is the most suitable
ice class for ships navigating in harsh Antarctic ice con-
ditions and PC 6 is enough behind an icebreaker on the
Kara Sea during hard winter conditions.
for charting are relatively scarce. These Arctic hazards are
compounded by the fact that the rate of recovery of the
Arctic nature is slow, meaning that environmental hazards
are made more serious.
The scope of this paper is first to shortly review the
present knowledge for risk based ship design (RBSD) for
ice conditions as RBSD is the current trend also in the other
maritime activities. In addition, the current best practice is
described by analyzing the functionality of the proposed
POLARIS through comparison of the risk index output with
some full scale observations onboard SA Agulhas II in
Antarctica and onboard MT Uikku on Russian Arctic. The
International Maritime Organization (IMO) has been de-
veloping a mandatory International Code of Safety for
Ships Operating in Polar Waters (Polar Code), which will
come into force on 1st January 2017. The guidance is
planned to include an example of an acceptable methodol-
ogy for assessing operational capabilities and limitations for
ships operating in ice, the so called Polar Operational Limit
Assessment Risk Indexing System (POLARIS). POLARIS
provides a standard approach for the evaluation of risks to
the ship and the ice conditions encountered/expected (ice
regime).
RISK BASED DESIGN FOR ICE CONDITIONS
Arctic shipping and accident statistics

There is a long history of Arctic marine transport and

INTRODUCTION
The increased activity in the Arctic involves hazards
such as harsh environment, especially the ice cover and
cold temperature, remoteness and lack of infrastructure,
lack of information about bathymetry, among others. Ice
cover is also highly variable and dynamic with increasing
variation in future as due to the changing effects of the
world climate, also ice conditions on all ice-covered areas
are under dynamic change. This effect on Arctic operations
is a complicated task to solve. The remoteness of the Arctic
areas means that in case of an accident, the search and res-
cue (SAR) capability is low. Also the fairways are not
marked very extensively and especially the soundings taken
year-round navigation has been maintained since 1978-79
in the ice-covered western regions of the Northern Sea
Route (between the port of Dudinka on the Yenisei River
and Murmansk), see e.g. AMSA (2009). Arctic transport
today is destinational, conducted for community re-supply,
marine tourism and moving natural resources out of the
Arctic. Still, due to the limited amount of traffic, the acci-
dent statistics form the Arctic activities are rare
Kubat and Timco (2003) compiled and analyzed a da-
tabase of ship damage in the Canadian Arctic for the period
between 1978 and 2002. Marchenkos (2012) book on the
Russian Arctic waters goes into some detail regarding ac-
cidents in the Arctic seas bordering Russia. Her coverage is
fairly comprehensive as she provides details on accidents

196
since 1900 up until the early 1990s, when the volume of
Arctic traffic in Russian waters dropped off dramatically
due to the collapse of the Soviet Union.
As part of the Arctic Councils Arctic Marine Shipping
Assessment (AMSA) 2009 Report, a database of a sum-
mary of the incidents and accidents occurring in the Arctic
region between 1995-2004 was developed from several
other existing data sets. While this data set is one of the
more comprehensive available, it does not include any in-
cidents that occurred in Russian waters, which is unfortu-
nate.
The challenge with all this type of damage databases is
that they are not detail enough to evaluate the sequence of
events causing the accidents and therefore they cannot be
easily applied in the development of proper risk based de-
sign methodologies. In the development of the Finn-
ish-Swedish ice class for the Baltic Sea, damage statistics
have played an important role and there exist extensive data
bases starting from the early work of Johansson (1967) and
continuing by Kujala (1991) and Hnninen (2004). There is
a lot of useful information about possible damages for ships
in ice.
Present ice class approaches
Ice class rules are based on the long term experience of
navigation in ice in various sea areas and typically the used
design scenarios are not clearly specified. The most obvious
definition of the design scenario is given for the IACS Uni-
fied Requirements for Polar Class Ships (IACS UR I.1-3),
in which a glancing impact between ship and ice is used.
This is based on the original theory developed by Popov et
al. (1967) and developed further by Daley (1999) assuming
that the ship kinetic energy is spent for ice crushing when a
ship hits ice edge on a glancing impact. The scantling for-
mulae are set with the use of plastic limit state equations for
plating and frames (Daley et al. 2001, Kendrik 2015).
For other relevant Arctic ice class rules such as Cana-
dian ASPPR (Carter et al., 1992) and Russian rules (RMRS,
2010), the design scenarios are not so clearly illustrated.
Both are using so called ice numerals to relate the required
ice class with the prevailing ice conditions, which are based
mainly on the long term experience of navigation in ice in
these waters. For Finnish-Swedish ice class the design sce-
nario can be described as hitting the ice edge with level ice
thickness of about 1.0 m for ice class IA Super and this is
decreasing down to 0.4 m for ice lass IC even though there
is still a remarkable uncertainty which ship-ice interaction
scenario causes the highest loads, (Riska et al., 2012).
Challenges to apply risk based ship design for ships in ice
The risk based design for ice-going ships have not been
extensively studied. The main challenges are related to the
definition of the ice environment and the ship-ice interac-
tion on this varying environment. Unfortunately, the ice
environment is a complicated obstacle to define properly.
When comparing ice environment to open water, in which
typically only two parameters are needed i.e. the wave
height and period, it is far more complex as ice can have
various forms and typically at least the following parame-
ters are needed to describe it: level ice thickness, floe size,
ice concentration, amount of rafted or ridge ice, frequency
and height of ridges, amount of compression in the ice field.
In addition, ice field are very dynamic and changes on the
ice cover characteristics can happen rapidly e.g. due to the
wind and currents. In addition, ice can be first, second or
multiyear ice, which will have large impact on the strength
properties of ice as well as on the possible thickness of ice.
Ship-ice interaction has been under extensive research
at least during the last 50 years. Russian scientists made the
pioneering work as was summarised by Popov et al. (1967).
A number of fundamental theoretical approaches were de-
veloped both to analyse the ship-ice contact and the possi-
ble highest loads occurring using the energy approach i.e.
assuming that the ship kinetic energy is spent in ice crush-
ing and changes in the kinetic and potential energy of the
vessel hitting the ice edge. After that there have been a lot
research trying to observe and model the ship-ice contact as
summarised e.g. by (Enkvist et al., 1979, Daley et al., 1990,
Jordaan et al. 1993, Kendrick et al., 2011). Typically this
research can consists of laboratory measurements of ice
failure, full scale field tests of ice failure or full scale meas-
urements of ice induced pressure contact or local loads on
board various ships. Unfortunately still the knowledge of
the variation of the contact on the ship hull and this relation
to the ice induced loads in various ice conditions and opera-
tion situations is limited to form a good basis for the risk
based design. Due to this complexity, no sound and reliable
approach to evaluate the risks exists and the few existing
approaches to evaluate risks are summarized next.
Daley and Ferregut (1989) presented a model of struc-
tural risk for ice going ships, called ASPEN (Arctic Ship-
ping Probability Evaluation Network). The ASPEN model
used a cell grid map of the arctic, with ice statistics in each
cell for each month. A user would specify a route in terms
of cells (and month). The model calculated the encoun-
ter-detection-avoidance-impact damage probabilities using
a set of probability algorithms. The program could evaluate
the sensitivities of aspects such as route selection, detection
strategies, and structural capacity. Buzuev and Fedyakov
(1997) examined the reliability and risk of shipping in ice
along the northern sea route (NSR) in Russia. The focus
was more about transportation reliability than structural
risks, though both rely on similar models of ice conditions.
Loughnane et.al (1995) examined the risks for an arctic oil
tanker with a focus on oil spill risks and mitigation costs
and strategies.
More recently there have been studies related to the
probabilistic analysis of the ship ramming through an Arc-
tic ice field (Ralph and Jordaan, 2013), analysis of the
probability of a ship to get stuck in ice (Montewka et al.,
2015) and risk analysis of the winter navigation system of
Finland (Valdez et al., 2015, Goerlandt el., 2016).
The best present practice for the risk based design is the
new POLARIS system as part of the new Polar Code and
therefore this is presented next with two practical examples
how to apply it.
DESCRIPTION OF THE POLARIS APPROACH
POLARIS uses a Risk Index Outcome (RIO) value to
assess limitations for operation in ice (MSC. 1/Circ 1519,
2016). Risk Index Values (RIVs) are assigned to the ship
based on ice class. For each ice regime encountered the
Risk Index Values are used to determine a RIO that forms
197
the basis of the decision to operate or limitation for opera-
tion. The RIO is determined by a summation of the RIVs
for each ice type present in the ice regime multiplied by its
concentration (expressed in tenths):
RIO=(C1xRIV1)+(C2xRIV2)+(CnxRIVn) (1)
where C1Cn are the concentrations (in tenths) of ice types
within the ice regime and RIV1RIVn are the Risk Index
values corresponding to the ships ice class, Table 1.
Table 1. Risk index values (RIV) for various ice classes,
winter conditions
ious ice classes of PC1 to PC7 as shown in Table 3 and for
other ice classes operations in ice should be avoided for
voyage planning purposes when RIO is negative. For ships
under escort by an icebreaker where the track created by the
icebreaker(s) is wider than the beam of the ship under es-
cort operational limitations are basically based on the ice
conditions on the channel behind an icebreaker. For voyage
planning purposes when icebreaker is intended to be used,
the RIO derived from non-escorted historical ice data may
be assumed to be modified by adding 10 to its calculated
value, given that due caution of the Mariner will be exer-
cised, taking into account such factors as changes in weath-
er and visibility.

SHORT DESCRIPTION OF THE FULL SCALE MEAS-

UREMENTS

Ship particulars

The POLARSIS approach is validated next by compar-

ing this approach with the full scale observations onboard
two vessels: Research vessel SA Agulhas II in Antarctica
and tanker Uikku in the Russian Arctic. Table 4 summaris-
es the main characteristics of the vessels and Fig.1 illus-
trates the ships.

Table 4. Main particulars of SA Agulhas II and MT Uikku

SA Agulhas II MT Uikku
Length, Lpp (m) 121,8 150
Breadth (m) 21,7 22,2
Draught (m) 7,65 9,5
Deadweight (t) 5000 15748
Displacement (t) 13632 22654
Speed (kn) 14 17
Propulsion power 9 11,4
(MW)

Table 2. Risk index outcome (RIO) evaluation criteria for

independent navigation

Fig.1a. Profile of SA Agulhas II showing also the instru-

mented areas

Table 3. Recommended speed limits for elevated risk oper-

ations

Fig.1b. Profile of MT Uikku.

The operational limitations for ships operating inde-
pendently are based on the evaluation criteria shown in
Table 2. The RIO should be positive to indicate normal
operations and when the value of RIO is below 0 but higher
than -10, limited speed requirements is established for var-
Instrumentation
PSRV S.A. Agulhas II was built in STX Finland in
Rauma into the Polar ice class PC5 and the hull was con-
structed in accordance with DNV ICE-10. Three areas of
the starboard side of the hull were instrumented with strain

198
gauges when she was under construction in 2011/2012. The
upper and lower parts of the frame were instrumented with
V-shaped strain gauges, which measured the shear strains
occurring in the frame. The instrumentation consists of two,
three and four adjacent frames at the bow, bow shoulder
and stern shoulder respectively. In addition, the hull plating
was instrumented with strain gauges in these areas. See
Suominen et al. (2013) for more detailed description of the
instrumentation.
MT Uikku is classified by DNV for class +1 A 1 Tanker
for Oil, and by Finnish Board of Navigation to ice class 1 A
Super. Ship was built 1976 in Werft Nobiskrug Gmbh, and
Helsinki New Shipyard rebuilt Azipod conversion 1993.
Ship has a diesel electric propulsion system with four diesel
generators. The ship hull and propulsion system was in-
strumented on 1997 for the EU funded ARCDEV project
and the instrumentation was extensive including: load on
the shell transverse frame at bow area, at bow shoulder area,
at midship area and at aftship area (measured by shear
strain gauges), load on the shell longitudinal frames at
midship area (measured by shear strain gauges), stresses on
the shell plating and frames at waterline at bow area, at bow
shoulder area, at midship area and at aftship area, the lon-
gitudinal bending stresses on deck and vertical accelera-
tions at the bow and stern of the ship and longitudinal ac-
celeration at bow. More detail description of the instru-
mentation can be found in Kotisalo and Kujala (1999).
Description of the measuring voyage
SA Agulhas II
Full scale measurement data on ice conditions in Ant-
arctic waters was collected in the Antarctic waters onboard
S.A. Agulhas II between Dec 6, 2013 and Feb 2, 2014. The
ice conditions were observed visually and the machinery
control and navigational data were recorded continuously.
The iceinduced loads were also measured at the bow,
bow shoulder and stern shoulder of the ship hull.
Fig.2a. Sea ice extent on Antarctica on 7th December 2013.
The ship departed Cape Town on November 28th, 2013
(Suominen et al., 2015a, 2015b). From Cape Town, the ship
headed to the zero Meridian, which she followed to Antarc-
tica, see Fig.2. The first time the ship encountered ice was
on December 7th. On December 23rd, the ship arrived the
Akta Bukta close to the Neumayer III (the German Antarc-
tic research station, see Fig.2a).
Fig.2b. Route of the convoy during ARCDEV voyage
Between December 24th and 30th, the ship operated
between the Akta Bukta and the Penguin Bukta (the loca-
tion at the ice shelf with the shortest distance to the South
African Antarctic research station, SANAE IV, see Fig.2a)
close the ice shelf. On December 30th, the ship headed to-
wards South Georgia and the South Sandwich Islands. The
ship arrived at the South Sandwich Islands January 4th
when also the extent of sea ice ended.
The ship returned in iceinfested waters January 23rd
to get through the pack ice to Penguin Bukta for the pas-
sengers and cargo unloading, where she arrived January
25th. The ship operated between the Penguin Bukta and
Akta Bukta between January 26th and 31st. The ship head-
ed back to Cape Town on January, 31st and the last ice was
observed 1st February.
MT Uikku
The voyage started from the port of Murmansk on 26th
of April 1998, so this presents hard winter time ice condi-
tions for the studied area. The route of the convoy is pre-
sented in Fig.2b. Due the heavy ice conditions and the east
wind the passage through the Kara Gate was blocked and
the convoy MT Uikku and IB Kapitan Dranitsyn- entered
the Kara Sea using the northern route. North edge of the
Novaya Zemlja was passed on 29th of April and the nuclear
icebreaker Rossiya joined to the convoy.
While the convoy was proceeding through the Kara Sea
the IBN Vaygach broke a channel through the fast ice of the
bay of Ob to the town of Sabeta. The convoy reached the
entrance of this channel on 3rd of May, when the IBN Ros-
siya left the convoy. MT Uikku and IB Kapitan Dranitsyn
proceeded on their own to Sabeta, and the convoy reached
subice loading terminal on 4th of May.
Loading was completed on 8th of May and the convoy
proceeded back to the Kara Sea. When the convoy reached
to the Kara Sea IBN Rossiya joined to the convoy and as-
sisted the convoy through the Kara Sea and through the
Kara Gate to the Barents Sea, where the convoy arrived on
12th of May. After light ice conditions and open water were
reached MT Uikku proceeded independently to Murmansk,
where she arrived on 13th of May.
199
Observed ice conditions
SA Agulhas II
Ice conditions were observed continuously by visual
observations on the ship and reported in 15 minutes inter-
vals during the first day and in 10 minutes intervals there-
after. The ice concentration was observed in tenths during
each observation period. The mean ice concentration during
each observation period was calculated as a weighted aver-
age and the results are presented in Fig.3. The sea ice con-
centration was high, almost 10, most of the time.
Fig.3. Mean ice concentration in each observation period
onboard SA Agulhas II
Ice thickness was estimated by observing the thickness
of the turning ice floes at the bow of the ship both visually
and by using stereo-camera photos. Observations of ice
thickness have been reported at a 20 cm ice thickness in-
terval from 0 to 2 meters and at 50 cm intervals thereafter
for each observation period. This was possible when the ice
thickness was up to about 2 m as when the ice thickness
exceeded 2 m, the ice pieces did not turn anymore and the
thickness estimation was more complicated. Thick snow
cover made also the thickness observations difficult. Fig.4
shows the observed mean ice thickness for each period.
Fig.4. Mean ice thickness in each observation period
onboard SA Agulhas II
Most of the time, the level ice thickness encountered at
a certain 10 minutes observation period varied quite much.
Typically 4 to 6 different ice thicknesses were observed
during each 10 minutes periods. Obviously, the ice fields
are often moving out to the open sea from Antarctica due to
the wind, waves and currents and therefore the level ice
field is broken into ice floes. New ice is then formed be-
tween the broken ice floes, which can explain the existence
of different thicknesses of level ice at open sea.
Fig.5. The speed of SA Agulhas II during each 10 minutes
period.
Fig.5 illustrates the ships speed during the voyage. As
can be seen, the ship had large difficulties to get through
the thick ice regime when approaching Akta Bukta before
Christmas 2013. The ship was stuck a number of times, but
200
got however slowly forward independently by ramming
through the thick ice.
MT Uikku
Ice conditions were observed continuously by visual
observations on the ship and reported in 20 minutes inter-
vals. The ice concentration was observed in tenths during
each observation period. The mean ice concentration during
each observation period was calculated as a weighted aver-
age and the results are presented in Fig.6.
Ice thickness was estimated by observing the thickness
of the turning ice floes at the bow of the ship visually. Ob-
servations of ice thickness have been reported on five clas-
ses: below 10 cm, 10- 30 cm, 30-70 cm, 70-120 cm and
above 120 cm. Fig.7 shows the observed mean ice thickness
for each period. Fig.8 illustrates the ships speed during the
voyage. As can be seen, the ship could keep fairly high
speed of 10-15 kn most of the time as there was always
icebreakers to assist the ship.
Fig.6. Mean ice concentration in each observation period
onboard MT Uikku
Fig.7. Mean ice thickness in each observation period
onboard MT Uikku
Fig.8. The speed of MT Uikku during each 20 minutes pe-
riod.
POLARIS CALCULATIONS
Risk index values for SA Agulhas II
The Risk Index Values (RIV) for the various ice thick-
ness values were determined in accordance with Table 1
(MSC. 1/Circ 1519, 2016) and Table 5 below shows the
results. Table 5 gives the Risk Index Values for decayed ice
conditions. It was decided to use the RIV values for de-
cayed ice conditions, because the measured sea ice strength
was rather low, the temperature measurements made during
the voyage showed positive or close to zero air tempera-
tures and the sea ice extent around Antarctica was dimin-
ishing during the voyage. It should be noted that the range
of the reported ice thicknesses do not always coincide with
the range of ice thicknesses in Table 1, and therefore the
stages of ice development defined by WMO do not fully independent Antarctic operation, ice class PC 3 is obviously
comply with the range of ice thicknesses given in Table 5. most suitable to be in the safe side, for more detail discus-
As it is mentioned above, it was not quite clear, if ice sion of this topic, Kujala et al. (2015).
thicker than 2 m was second year ice or multiyear ice.
However, it was decided to consider ice thicker than 2 m
but less than 2.5 m as second year ice, ice at least 2.5 m
thick but less than 3 m thick as light multiyear ice, and
ice 3 m thick or thicker as heavy multiyear ice. The
POLARIS calculations were done for IACS PC classes PC3,
PC5 and for the FinnishSwedish ice class IA Super.

Table 5. Risk Index Values (RIV) for ice classes PC3, PC5
and IAS for decayed ice conditions during SA Agulhas II
voyage.

Fig.9. The results of POLARIS calculations for PC3, PC 5

and IAS using SA Agulhas II data.

Results of POLARIS calculations for MT Uikku

Risk index values for MT Uikku
The POLARIS calculations for ice class PC3, PC 5, PC
The Risk Index Values (RIV) for the various ice thick-
6, IAS and IA were done in accordance with the formula
ness values were determined in accordance with Table 1
(1) for assessing operational capabilities and limitations in
MSC. 1/Circ 1519, 2016) and Table 6 below shows the
ice: Polar Operational Limit Assessment Risk Indexing
results. In this case the winter ice conditions are used as
System (POLARIS) using the Risk Index Values (RIV)
during this voyage it was still hard winter in the Russian
given in Table 6 above and the ice concentration data. The
Arctic.
results of the calculations are given in Fig.10.
Table 6. Risk Index Values (RIV) for ice classes PC3, PC5
PC6, IAS and IAS during MT Uikku voyage.

It should be noted that the range of the reported ice

thicknesses do not always coincide with the range of ice
thicknesses in Table 1, as e.g. for the observation range of
30-70 cm and 70-120 cm, the lower value is used for the
whole range.

Results of POLARIS calculations for SA Agulhas II

The POLARIS calculations for ice class PC3, PC 5 and

IAS were done in accordance with the formula (1) for as-
sessing operational capabilities and limitations in ice: Polar
Operational Limit Assessment Risk Indexing System
(POLARIS) using the Risk Index Values (RIV) given in
Table 5 above and the ice concentration data. The results of
the calculations are given in Fig.9 below.
The results of Fig.9 indicate that for most of the time the
Fig.10. The results of POLARIS calculations for PC3, PC 5,
RIO values are positive for PC3, but occasionally negative
PC6, IAS and IA using MT Uikku data.
up to -20 for PC 5 and mostly negative to IAS ice class
getting values even close to 40. This indicates that for

201
 Comparing Fig.10 with the allowable values given in
Table 2 and adding the value 10 due to the icebreaker as-
sistance, it seems that ice classes PC 6 is the most suitable
for navigating in the Kara Sea behind an icebreaker during
hard winter time ice conditions.
COMPARISON OF POLARIS CALCULATIONS WITH
MEASURED ICE LOAD DATA
The ice load measuring system onboard both SA Agul-
has II and MT Uikku enable also the evaluation of the
proper ice class by calculating the required hull strength so
that the damage probability will be on a right level. This
has been done on a recent master thesis by Kurmiste (2016).
Using the long term measurements onboard SA Agulhas
during 2013-2015 and the three week data on MT Uikku, he
fitted Gumbel 1 extreme value distribution on these data
and forecasted the life time extreme loads for these ships.
Then by comparing these load values with the ultimate
strength of the frames using various ice classes to determine
the scantlings, he could evaluate the probability of reaching
the planned limit state. Similar analysis has been conducted
before e.g. by Kujala (1991) and Kaldasaun and Kujala
(2011). As a result, a reliability index is obtained by de-
fining a probability of failure for the frames. For
ice-strengthened ships, typically it is accepted that the
values should be in the range 1-2, i.e. approximately 1 in 10
to 100 ships will experience some local damages on the
shell structures during ship life time. Fig.11 shows the ob-
tained safety index values using SA Agulhas II data and
Fig.12 using the MT Uikku data.
Fig.11. Safety indices of the bow frames of SA Agulhas II
(Kurmiste, 2016)
Fig.12. Safety indices of the bow frames of MT Uikku
(Kurmiste, 2016)
202
It is really interesting that also based on this analysis PC
3 seems to be the proper ice class for independent naviga-
tion on the Antarctic ice conditions and similarly PC 6 is
the proper ice class for ships following an icebreaker on the
Kara Sea during hard winter time ice conditions. This is
strong evidence that the developed POLARIS gives reliable
estimates for the safety of ships in various ice conditions.
CONCLUSIONS
The POLARIS calculations for ice classes PC3, PC5,
PC6 IA Super and IA were done in accordance with the
formula given in the Polar Operational Limit Assessment
Risk Indexing System (POLARIS) using the Risk Index
Values (RIV) and the ice concentration data observed
onboard SA Agulhas II and MT Uikku. The results indicate
that PC 3 is the most suitable ice class for Antarctic opera-
tions and PC6 for a ship navigating behind an icebreaker on
the Kara Sea during hard winter time ice conditions. Simi-
lar results for also obtained by calculating the probability of
hull damage based on the measured full scale ice induced
load data onboard SA Agulhas II and MT Uikku. The PO-
LARIS approach is fairly easy to apply and it seems to give
a reasonable estimate for the risk index to indicate whether
navigation on the planned ice conditions is safe with the
chosen ice class.
ACKNOWLEDGEMENTS
The work was funded by the Finnish Funding Agency
for Technology and Innovation (Tekes), Academy of Fin-
land, and the Lloyds Register Foundation. The Lloyd's
Register Foundation supports the advancement of engi-
neering-related education, and funds research and devel-
opment that enhances the safety of life at sea, on land, and
in the air. Their financial support is greatly acknowledged.
In addition, all the partners of the Tekes project NB1369
PSRV full scale ice trial, namely the University of Oulu,
University of Stellenbosch, Aker Arctic, Rolls-Royce, STX
Finland, Wrtsil, and the Department of Environmental
Affairs (South Africa), and Finnish Meteorological Institute,
as a partner of the Academy of Finland project called
ANTLOAD, are gratefully acknowledged. Also the finan-
cial support of TRAFI is acknowledged with pleasure.
REFERENCES
Arctic Council (2009). Arctic Marine Shipping Assessment
(AMSA) 2009 Report.
Buetzow, M.R., Athanasiadis, S., Ximenes, M.C., Cinelli,
T.G. and Lofaro, R.Jr. (1997), Structural Design and Ice
Strengthening of the Hibernia Shuttle Tanker M.T.
KOMETIK. Proc. 1997 SNAME Annual Meeting,
p20.1-20.21.
Carter, J.E., Frederking, R., Jordaan, I.J., Milne, W.J.,
Nessim, M.A., and Brown, P.W. (1992). Review and
Verification of Proposals for the Revision of the Arctic
Shipping Pollution Prevention Regulations. Report for
Canadian Coast Guard Northern, by Memorial Univer-
sity of Newfoundland Ocean Engineering Research Cen-
ter. Report TP11366E.
Daley, C.G. and Ferregut, C. (1989), Structural Risk Mod-
el of Arctic Shipping, IUTAM/IAHR Symp.
Ice-Structure Interaction, St. Johns, Canada.
Daley, C.G. (1999). Energy Based Ice Collision Forces,
POAC 1999, Helsinki, Finland.
Daley, C. and Riska, K. (1990). Review of ship-ice inter-
action mechanics. TKK Naval Architecture report
M-102.
Daley, C.G., Kendrick, A. and Appolonov, E. (2001).
Plating and framing design in the unified requirements
for Polar class ships. POAC 2001, Ottawa, Canada.
Enkvist, E., Varsta, P. and Riska, K. (1979). The ship-ice
interaction. POAC 1979, Trondheim, Norway
Goerlandt, F., Montewka, J., Zhang, W. and Kujala, P.
(2016), An analysis of ship escort and convoy opera-
tions in ice conditions. Safety Science, Special Issue
'Safety and Integrity in Harsh Environments
Jordaan, I., Maes, M., Brown, P., and Hermans, I., (1993).
Probabilistic Analysis of Local Ice Pressures. ASME J.
of Offshore Mechanics and Arctic Engineering. Vol. 115,
pp. 83-89.
Hnninen, S. (2004). Incidents and Accidents in Winter
Navigation in the Baltic Sea, Winter 2002-2003 (un-
published). Winter Navigation Research Board, Re-
search Report No 54.
Johansson, B. (1967). On the ice-strengthening of ship
hulls. Int. Shipbuilding Progress, Vol. 14. pp. 231-245.
Kaldasaun, J., and Kujala, P. (2011). Risk-Based Ap-
proach for Structural Design of Ice-Strengthened Vessels
Navigating in the Baltic Sea. Proc. 21st Int. Conf. on
Port and Ocean Engineering under Arctic Conditions.
Montral, Canada
Kendrick, A.M., and Daley, C.G. (2011). Structural chal-
lenges faced by Arctic ships. SSC-461. Ship Structure
Committee. US Coast Guard. Washington. USA.
Kendrick, A.M. (2015). Polar Ship Design Standards
State of the Art, and Way Forward. ICETECH 2014.
Banff, Canada.
Kubat, I. and Timco, G. (2003). Vessel Damage in the
Canadian Arctic. Proc. 17th Int. Conf. Port and Ocean
Engineering under Arctic Conditions, POAC'03, 1, pp.
203-212
Kujala, P. (1991). Damage statistics of ice-strengthened
ships in the Baltic Sea 1984-1987. Winter Navigation
Research Board. Report. No. 50. 61 p. + app. 5 p.
Kujala, P. (1991). Safety of ice-strengthened ships in the
Baltic Sea. London. RINA Trans.
Kujala, P., Kmrinen, J., and Suominen, M.
(2015). Analysis of a suitable ice class of ship hull for
Antarctic operations. 5th World Maritime Technology
Conf., Rhode Island, USA.
Kotisalo, K. and Kujala, P. (1999). Ice load measurements
onboard MT Uikku during the Arcdev voyage. Proc.
POAC99 Vol 3, Espoo, pp. 974-987.
Kurmiste, A. (2016). Analysis of structural safety of
ice-going vessels in the Arctic and Antarctic. Master
thesis. Aalto University, School of Engineering.
Loughnane, D., Judson, B. and Reid, J. (1995), Arctic
Tanker Risk Analysis Project, Maritime Policy
&Management, Volume 22, Issue 1, pp. 3-12
Marchenko, N. (2012). Russian Arctic Seas. Springer-
Verlag Berlin Heidelberg. DOI
10.1007/978-3-642-22125-5.
Montewka J., Goerlandt F., Kujala P. and Lensu, M. (2015),
Towards probabilistic models for the prediction of a
ship performance in dynamic ice. Cold Regions Science
and Technology 112, pp.14-28.
MSC. 1/Circ 1519, 6 June 2016. Guidance on methdodolo-
gies for assessing operational capabilities and limitations
in ice. IMO-
Popov, Yu., Faddeyev, O., Kheisin, D., and Yalovlev, A.
(1967). Strength of Ships Sailing in Ice. Sudostroyeniye
Publishing House, Leningrad, Russia. 223 p. (English
Translation by U.S. Army Foreign Science and Technol-
ogy Centre).
Ralph, F., and Jordaan, I., 2013. Probabilistic Methodoly
for Design of Arctic Ships. Proc. of the 32nd Int. Conf. on
Ocean, Offshore and Arctic Engineering, June 9-14,
2013. Nantes, France.
Riska, K., and Kmrinen, J. (2012). A Review of Ice
Loading and the Evolution of the Finnish-Swedish Ice
Class Rules, SNAME Annual meeting 2012 and SNAME
Trans.
RMRS. (2010). Rules for the Classification and Construc-
tion of Sea-Going Ships. St. Petersburg.
Suominen, M., Karhunen, J., Bekker, A., Kujala, P., Elo,
M., von Boch und Polach, R., Enlund, H., and Saarinen,
S. (2013). Full-Scale Measurements on Board PSRV
S.A. Agulhas II in the Baltic Sea. Proc. 22nd Int. Conf.
Port and Ocean Engineering under Arctic Conditions,
Espoo, Finland.
Suominen, M., Kujala, P. and Kotilainen, M. (2015a). The
encountered extreme events and predicted maximum ice
induced loads on the ship hull in the Southern Ocean.
Proc. ASME 2015 34th Int. Conf. Ocean, Offshore and
Arctic Engineering, OMAE2015, St Johns, Canada.
Suominen, M. and Kujala, P., 2015b. The measured line
load as a function of the load length in the Antarctic wa-
ters. Proc. 23rd Int. Conf. Port and Ocean Engineering
under Arctic Conditions, Trondheim, Norway
Valdez, B., Osiris, A., Goerlandt, F., Montewka, J., and
Kujala, P. (2015), A risk analysis of winter navigation
in Finnish sea areas. Accident analysis and preven-
tion,Vol. 79, pp.100-116.
203
 SEAHORSE Procedure Improvement System
Rafet Emek Kurt, Volkan Arslan, Emma Comrie, Hassan Khalid, Osman Turan
University of Strathclyde, Department of Naval Architecture Marine and Ocean Engineering, 100
Montrose Street G4 0LZ Glasgow, United Kingdom, rafet.kurt@strath.ac.uk
ABSTRACT
Standardisation of operating procedures has been
used in variety of different sectors with the aim of
achieving more reliable operations hence operational
safety. Likewise, Standard Operating Procedures
(SOPs) have been adopted by the maritime sector and
enforced through the regulatory framework in an at-
tempt to achieve safer shipping operations. Despite the
fact that regulations enforce the SOPs, it is often ob-
served that during shipping operations these procedures
are not followed due to various reasons. One of the most
common reasons for not following the SOPs is because
shipping companies suffer from poorly designed proce-
dures, which are impractical, unclear or sometimes
factually wrong. These poorly designed procedures are
disregarded by crewmembers, which prevent the prac-
tical implementation of these SOPs during shipping
operations. Therefore, it is of key importance that a
systematic approach is needed to identify and improve
the current SOPs as well as preventing potentially
harmful workarounds.
The EU FP7 SEAHORSE project is developing a
Procedure Improvement System which will be active-
ly used by the crewmembers any time anonymously. In
order to achieve this, over 400 questionnaires have been
collected from seafarers across Europe where they were
asked to report on impractical SOPs and common
workarounds conducted on board ships. These ques-
tionnaires were organised and a risk benefit analysis
was conducted. This paper will report the initial results,
as well as demonstrating the overall methodology for
the SEAHORSE Projects procedure improvement sys-
tem.
INTRODUCTION
Maritime accident investigations indicate that human
factors are attributed to most of the accidents. So far, the
dominant approach taken to overcome and manage this root
cause of accidents is the development of a set of Standard
Operating Procedures (SOPs). These SOPs are intended to
provide a standardised means of working within a given
organisation and is an attempt to make the system less de-
pendent on human operators. However, to date, the envis-
204
aged impact of SOPs upon safety/accidents has not been
achieved in the maritime industry. This may be attributable
to the lack of standardisation between vessels, operations,
environmental conditions, crew numbers and so on. Due to
the aforementioned factors, together with poorly designed
SOPs, it is known that within the shipping industry the
SOPs do not always match with operational realities and as
such, seafarers, in some cases, deviate from the SOPs to
complete their duties. Deviations conducted by crew mem-
bers to overcome a problem or limitation presented by the
SOPs, are hereby defined as workarounds. In some cases
workarounds can present smarter means of carrying out
duties; however, they may also result in higher risks. Thus,
a methodology is required to collect workaround data and
to inform decision-making about whether to make amend-
ments to the SOP.
This paper presents a methodology to perform a
risk-benefit analysis between a SOP and its workaround to
support decision-making. The methodology can be summa-
rised as follows. Seafarers anonymously submit worka-
rounds in the form of a questionnaire. Seafarers identify
and provide details of the workaround being practiced
on-board vessels. To assess a SOP and a workaround, a
group of experts are asked to rate the workaround based on
a number of pre-defined attributes. The uncertainty and
vagueness of the selection between alternatives (SOP or
workaround) problem leads to the use of the Fuzzy Multiple
Attribute Group Decision Making (FMAGDM) method to
provide an assessment of the two alternatives (SOP and
workaround). This method is employed to compare current
SOPs and maritime workarounds and provide a value of
how much better one alternative is over the other. The
Technique for Order of Preference by Similarity to Ideal
Solution (TOPSIS) Method is utilised to find the best alter-
native between SOP and workaround. Results from the
group of experts are provided to the administrator/decision
maker. The administrator/decision maker considers the
outcome of the comparison and a decision is made regard-
ing the workaround. Feedback of the review is provided to
every member of the organisation, thus closing the loop.
The methodology was tested on 107 workaround cases
in a workshop. To date, the methodology has received posi-
tive feedback from experts in the maritime industry and the
need for a methodology which collects, assesses and sup-
ports decision making in the organisation is clearly required.
Furthermore, the methodology can also be easily adapted
and applied in other areas such as the aviation industry.
MARITIME STANDARD OPERATING PROCEDURES
Shipping is commonly regarded as one of the most
dangerous industries in the world and according to the IMO,
there is the necessity of a set of international regulations to
be followed by shipping organisations, globally, to improve
safety. Importantly, the IMO has taken steps to prevent ship
operators cutting expenditure at the cost of shipping safety
(IMO 2012). A series of incidents representing significant
potential impact of shipping accidents is given in
Hetherington et al. (2006). Most importantly they conclude
that increased work load and deviations from standards may
lead to huge repercussions.
In order to comply with International Safety Manage-
ment (ISM), each shipping company has its own safety
management system (SMS). SMS is widely used across
industries to help direct companies to take a systematic
approach to managing safety within their organisation.
However, Mearns et al. (2013) recognise that a SMS with-
out a suitable level of organisational culture does not fulfil
the objective. Bhattacharya (2012) observes that there is a
significant gap between the mangers and seafarers per-
ceived meaning of the ISM code thus leading to a gap be-
tween the intended purpose of the code and the way in
which it is operationalised in daily seafaring.
Mearns et al. (2013) note: safety culture and safety
management go hand in hand to achieve safe practices in
an organisation. One is less tangible than the other is, but
both are required. If there is only an SMS but no real com-
mitment to safety, then the SMS will not be effective, as
decisions will not really prioritise safety, and the SMS will
be merely a paper exercise. Similarly, if there is a good
safety culture but no SMS, then in a complex organisation
the way safety is applied runs the risk of being inconsistent,
under or mis-resourced, and not seen as business driven
(because it will not be part of the business plan). The
aforesaid highlights that although processes can be formal-
ised within a company, should the company not implement
these processes properly, they are a waste of time and re-
sources.
The above provides a basis for the necessity of a con-
tinual review of SOPs within shipping organisations. There
is a fundamental need that this continual review is partnered
with a good communication flow. Clearly, there is a need of
communication between workers and management to dis-
cuss safety aspects and issues within the SOPs. Indeed, the
involvement of seafarers in the review and development of
SOPs will increase their respect and adherence to the SOPs.
However, any rejection of proposed workarounds has to be
adequately communicated to seafarers in order to avoid loss
of interest and confidence in the system.
SEAHORSE PROCEDURE IMPROVEMENT SYSTEM
(SPIS)
SEAHORSE procedure improvement system (SPIS)
aims to develop a comprehensive methodology to capture
workarounds performed by seafarers within a company,
assess them and compare them to SOPs in order to find the
most effective and safe way of working. A small group of
expert reviewers is assigned by the company to assess the
workaround and SOP. All assessments are aggregated into a
result that captures how much better (or worse) a specific
workaround is than the SOP. These results are then distrib-
uted within the company. SPIS has been embedded in a
software-based platform to ease the work of the managers
and improve SOPs in a structured way.
A general overview of SPIS is shown in Fig. 1. It con-
sists of three main stages: 1) gathering of workaround data
and development of attributes, 2) ranking and selection of
alternatives using FMAGDM and TOPSIS and 3) final
decision-making by administrator and feedback provided to
seafarer and reviewers.
Fig. 1 Overview of SEAHORSE Procedure Improvement
System (SPIS)
SPIS is a pioneering system that could also be applied to
other kinds of problems already existing in the maritime
industry. Additionally, it could also be utilised within other
domains such as the aviation industry.
Stage 1: Identification and Review of Workaround
The blame culture is still a predominant factor in the
maritime industry and seafarers are reluctant to share in-
formation about workarounds because they fear of negative
repercussions (Ek* & Akselsson 2005). A questionnaire
was developed to anonymously collect appropriate data
related to the workaround and SOP from the seafarer. SPIS
questionnaire composed of two distinctive parts: 1) de-
mographics and attitude section and 2) workaround section.
The structure of workaround section within SPIS question-
naire is shown in Fig. 2.
To begin with, seafarers are expected to give a brief de-
scription of the SOP from which they deviate, also the par-
ticipant is required to provide a description of the worka-
round being practiced. This is done to provide a written
account of the alternative means of carrying out the task to
facilitate reviewers in comprehending and assessing the
workaround.
The workaround frequency question aims to identify
how often a workaround happens amongst seafarers. The
identification of common deviations is a clear warning that
a SOP is not being followed, suggesting that there are po-
tential issues with the SOP. At the same time, this question
also provides insight into whether there may be challenges
if the workaround needs to be prohibited.
205
 important attributes for SOP & workaround benchmarking
process. In that, five attributes were defined as benefits,
which try to identify advantages of performing a worka-
round. The other four attributes were defined as risks which
may arise with the implementation of a workaround. The
list of selected attributes is shown in Fig. 3.

Fig. 2 Structure of workaround question within SPIS ques-

tionnaire
Fig. 3 The list of attributes
It was expected that certain type of operations may in-
Safety is paramount in the maritime domain. In order to
cite workarounds, thereby seafarers were expected to iden-
maintain safety of shipping, many regulations came into
tify it along with the location onboard where such operation
force including: International Convention for the Safety of
was being undertaken. There are always several reasons
why SOPs are not working effectively in the maritime in- Life (SOLAS), International Convention for the Prevention
of Pollution from Ships (MARPOL), Convention on the
dustry. Therefore, it is vital to capture the reasons behind
International Regulations for Preventing Collisions at Sea
the deviations from SOPs. This allows insight into the sea-
(COLREG), International Convention on Loadlines
farer and potential issues with the existing SOP.
(LOADLINE), The International Ship and Port Facility
Presumably, SOPs are designed to describe the best way
of working, however, sometimes seafarers may identify Security Code (ISPS), The International Safety Manage-
smarter, more efficient ways of performing the same task ment Code (ISM) International Convention on Standards of
Training, Certification and Watchkeeping for Seafarers,
without compromising safety. The identification of the
(STCW) and lastly, The ILO Merchant Shipping (Minimum
benefits allows the seafarers to express why the worka-
Standards) Convention (ILO) (ICS 2015).
round is being deviated from, showing the positive reasons.
Practicality attribute aims to elicit the experts opinion
Naturally, there are possible risks associated with per-
forming the task in the alternative way. In order to gain whether the proposed workaround may provide more bene-
insight into the risks, the seafarers are asked to give their fits to the practicality of an operation compared to the ex-
isting SOP. Here, practically is the quality or state of being
own account of the possible risks associated with the work-
practical. A SOP can be safe, cost effective etc.; however,
arounds. Lastly, any additional information worth consid-
it may not be practical and may induce operational difficul-
ering for the review of SOP is enquired.
ties. Seafaring is known as one of the hardest occupations
To summarise, the above questions allow seafarers to
provide a description of both workaround and SOP along with an excessive workload. Therefore, SOPs and work
with associated information including their perceived risks instructions should be prepared in a practical and user-
friendly way to decrease the workload.
and benefits of performing the workaround. Open-ended
Time efficiency attribute is meant to establish the ex-
questions were determined as most suitable for this ques-
perts opinion whether proposed workaround may make a
tionnaire as it was identified that seafarers would have to
shipping operation more time efficient compared to the
detail many aspects to allow assessment in the second stage
of the methodology by the experts. The questionnaire pur- existing SOP. Shipping companies operate at a very tight
schedule and need to deliver their cargo in a limited time.
posely has a limited number of questions to encourage the
Due to the long and exhaustive work hours, seafarers some-
reporting of the workaround and reduce the time required to
times come up with time efficient solutions to ensure they
complete the questionnaire. It however does elicit sufficient
have the specified rest hours.
information needed for the subsequent assessment.
The second aspect of this stage is the development of a Cost is a crucial parameter not only in shipping but also
set of attributes to measure the workaround. Measuring the in other industries. Companies always try to decrease their
investment or maintenance costs by implementing smarter
subjects like health, safety and environment is not easy and
technologies or conducting shipping operations effectively.
requires some measures to gain insight into company sys-
Cost effectiveness can be achieved in a company by utilis-
tems (Groeneweg et al. 2013). The performance of shipping
ing fewer resources such as labour, time or equipment.
companies is typically evaluated by key performance indi-
cators (KPIs). Decision makers need several attributes to Therefore, companies promote seafarers cost effective
assess problems and come up with a result. In SPIS, nine solutions on daily basis without putting safety at stake.
Regulatory compliance is essential to maintain safety
attributes were selected to evaluate the performance of
within the maritime industry. Shipping companies put
workaround according to safety critical performance drivers.
enormous effort to comply with international regulations.
An expert workshop was arranged to identify the most
Complying with the rules not only increases the safety and

206
reputation of a firm but also protects from costly fines.
Companies, who are successful at regulatory compliance,
can avoid catastrophic accidents undoubtedly. The Interna-
tional Maritime Organization (IMO) is known as the big-
gest regulatory body in the maritime industry. The IMO
made the International Safety Management Code (ISM)
mandatory after the accident of Herald of Free Enterprise to
avoid reoccurrence of these accidents (Goulielmos &
Goulielmos 2005).
Risk to person attribute aims to extract the experts
opinion whether proposed workaround may create more
risks to persons health and safety compared to the existing
SOP. Numerous casualties and injuries occur because sea-
farers do not strictly follow the SOPs and are performing
risky workarounds. All risks associated to human should be
eliminated or reduced to a level that is as low as practicable
on board ships to avoid such circumstances.
Around 90% of the goods transportation is seaborne and
shipping is very complex and high risk environment. The
shipping industry had catastrophic accidents throughout the
years such as Herald of Free Enterprise, Costa Concordia
and Deep Water Horizon. Risks to ships should be elimi- Fig. 4 Overview of FMAGDM methodology
nated or reduced appropriately to avoid occurrences of such
accidents. A total ship loss with numerous casualties is the This method leverages reviews by experts, which are
worst scenario a company can have, therefore, there are elicited through an established workaround assessment
various detailed SOPs to maintain safety on critical opera- form to provide an assessment of the workaround based on
tions such as fire protection, manoeuvring, etc. A tragic a number of attributes. Reviewers/experts are defined as
accident can also affect companys image negatively and individuals with substantive knowledge of a given area. In
may even lead to bankruptcy. Considering the aforemen- order to identify appropriate experts in the organisation the
tioned factors above, risk to ship is a very important factor questionnaire completed by the seafarer categorises the
that affects SOP development. workaround into one of four operations: deck operations,
Risk to environment attribute aims to elicit the experts cargo operations, engine room operations and bridge opera-
opinion whether the proposed workaround may create more tions. Experts are categorised as having expertise in one of
risks to environment (such as oil spills, gas/chemical waste, the four categories and experts are internal within the or-
garbage, etc.) compared to the existing SOP. Once there is ganisation. Experts are given workarounds to assess based
an oil spill at sea, even though the prompt actions take on their expertise decided by the administrator.
place after these events, the residuals keep contaminating
the marine environment. Scientists found subsurface oil Table 1 Experts rank the attributes to determine the
residues twenty years after the Exxon Valdez accident weighting of the attributes
(Boehm et al. 2011). Besides the oil spills, emissions of
CO2 emissions and other wastes also constitutes a great
4. Important
1. Very Un-

5. Very Im-
3. Neither

threat to the environment.

important
2. Unim-
portant

portant
Shipping is known as a very complex occupation and in
order to ease the complexity, the workload is broken down
into different shipping operations such as manoeuvring,
bunkering, cargo loading and mooring, etc. Each shipping Practicality
operation require different types of expertise and high de-
gree of situational awareness. It is crucial to consider all Time Efficiency
hazards a shipping operation might have. Risk assessment
Cost Efficiency
should be carefully conducted in order to address all types
of risks of a specific operation. Regulatory Compliance
Safety
Stage 2: FMAGDM Method for Ranking and Selection
of Alternatives Risk to Person
Risk to Ship
In order to compare the workaround and the SOP, SPIS
has adapted the Fuzzy Multiple Attributive Group Decision Risk to Environment
Making (FMAGDM) method proposed and outlined by Risk to Operation:
ler and Odabai (2005). This method consists of three
distinct states, namely (1) the rating state, (2) the attribute Other:
based aggregation state and (3) the selection state. A visual
representation of the FMAGDM methodology is shown in Experts are firstly asked to rate the workaround with
Fig. 4. respect to each attribute using a Likert-type scale (very

207
unimportant, unimportant, neither important nor unim- ment matrix (AM) is calculated. AM displays the degree of
portant, important, and very important), thus giving an agreement between every pair of experts. The diagonal is
assessment in linguistic terms as shown in Table 1. the degree of agreement of an expert with themselves
Naturally, each expert has different levels of expertise therefore values on the diagonal are always 1.
across the attributes (for instance one expert can have more Average degree of agreement (AA) is then calculated by
knowledge about safety but has less knowledge about com- using AM. The average degree of agreement of expert u,
pliance). Therefore, it is important to have different denoted by is calculated as
knowledge levels on each attribute and to utilize heteroge- 1
neous group of experts in the aggregation. In the aggrega- ( )= ( , )
tion process, the person calculated as having the most ex- 1
pertise was weighted as 1 and others were compared and (2)
weighted relatively with this person. The linguistic terms
were converted to standardised trapezoidal fuzzy numbers is the number of experts and corresponds to ex-
pert , 1 and 1 .
because linguistic terms are not mathematically operable.
Established conversion scales exist in the conversion of The relative degree of agreement (RA) is next calculat-
fuzzy data to fuzzy numbers (Chen et al. 1992). The con- ed as:
version scale used in SPIS is shown in Table 2.
( )
Table 2 Selected Scale for Aggregation ( ) =
( )
Likert Scale Statements Scale 3 (3)
1. Very Poor (0, 0, 0.1, 0.2)
Consensus degree coefficient denoted by CC( ) of
2. Poor (0.1, 0.25, 0.4)
expert is calculated by
3. Fair (0.3, 0.5, 0.7)
4. Good (0.6, 0.75, 0.9) CC( ) = " $% + (1 ") ( ),
(4)
5. Very Good (0.8, 0.9, 1, 1)
"(0 " 1) represents the relaxation factor. Natu-
The selected experts are asked to rate the workaround in rally, when " = 0 all experts are considered equally and
a questionnaire format. Experts are required to assess the this will occur in a homogeneous group of experts. It is
workaround based on a number of predefined subjective evident that " acts as a weighting of $% , which denotes
attributes, these attributes (Table 1) are generic and used for the importance of the expert and ( ) which is the rela-
all workarounds regardless of their operation categorisation. tive degree of agreement of the expert. ler and Odabai
The second state is the attribute based aggregation state. (2005) suggest that one way to assign weightings to experts
Its purpose is to provide an aggregated result for the work- is to use a moderator who assigns weights to each expert.
around. In this state, a score is calculated and assigned to Lastly, the aggregation results, '( , of fuzzy opinions
each expert for each attribute capturing the expertise of are calculated as
each expert that is a potential means of weighting of each
expert within the analysis. This calculation is performed to
'( = ))( ) ))( ) ))( )
allow appropriate weighting of the expert opinion and then (5)
provide robust results and a higher degree of confidence in
the calculations. The aggregation process of SPIS follows denotes the fuzzy addition operator and denotes
the sequence of ler and Odabai (2005) and is para- the fuzzy multiplication operator.
phrased from their paper: To summarise, MAFGDM method aggregates the rat-
Firstly, the degree of agreement (degree of similarity) is ings provided by the group of experts for each alternative
calculated, this is denoted by S(R1, R2). In this stage, the according to subjective attributes. All of experts ratings for
method developed by Chen (1995) is utilised to calculate each alternative are aggregated according to subjective
degree of similarity between all possible sets of experts. attributes and both attributes and experts are weighted ac-
The degree of agreement is calculated as follows: given we cording to their importance in the decision-making context
have the opinion of Expert A who gives in trapezoidal expertise.
number, say = ( , , , ) and then Expert B who The third state is the selection state which aims to pro-
gives = ( , , , ), S(A, B) is calculated as: vide a ranking of the alternatives. After State 2, all aggre-
gated trapezoidal fuzzy numbers are defuzzied to rank the
( , ) best alternative. Fuzzy numbers are transformed into crisp
| |+| |+| |+| |
= 1 numbers for evaluation by implementing fuzzy scoring
4 approach (Chen et al. 1992). Weighting of the attributes is
(1) considered in this state.
Finally, TOPSIS is utilized as a MADM method in the
An increase in S(A,B) corresponds to a higher degree of ranking stage to rank the order of the alternatives. TOPSIS,
agreement between the experts with a maximum possible developed by Hwang and Yoon (1981) is well-known with
value of 1. After the calculation of degree of similarity is its broad acceptability in many problematic areas and effec-
performed between all possible pairs of experts the agree- tive for determining best alternatives quickly. The working

208
algorithm of TOPSIS is given below (ler & Odabai
2005):
Attribute dimensions are converted into
non-dimensional attributes in order to benchmark
the attributes and obtain normalised weightings;
The normalised decision matrix multiplied with its
associated attribute weight is done in order to
calculate weighted normalised ratings. There are
several methods to calculate the weightings of the
attributes such as Weighted Evaluation Technique
(WET), eigenvector method and entropy method
(ler & Odabai 2005). Here, the WET has been
adopted;
Positive ideal and negative ideal solutions are
calculated;
Separation measures are calculated by Euclidean
distance;
Similarities to positive ideal solution are
calculated;
Preference order is ranked amongst alternatives.
Stage 3: Finalised Results and Feedback of Workaround
Evaluation
It is recognised that decision making predominantly in-
volves the consideration of more than one criterion or at-
tribute (Pidd 2009). An interesting aspect of research inves-
tigates the way in which people make decisions, and the
work of Kahneman and Tversky (1982) highlights that
individuals do not make decisions in a systematic way and
there are inherent biases in the decision making processes
of individuals. In the decision making process the decision
maker has to identify and consider the relevant stakeholders.
The finalised decision ends up to be a compromise between
needs and expectations of the different stakeholders identi-
fied and, sometimes, it needs to prioritise the wishes of the
stakeholders with the most power (Pidd 2009). Indeed,
decision-making is a challenging task, especially when
there are varying criteria with which to measure the differ-
ent alternatives. In order to ensure consistency across ex-
perts and facilitate a structured and repeated process of
decision-making, the experts judgments are elicited with a
formalised means based on a questionnaire. This helps
experts think through the problem based on agreed
pre-defined attributes.
In order to allow the decision maker to make a judgment
on the workaround a summary is provided to the decision
maker to assist them. The decision maker is provided: a
written summary of the workaround; a summary of the SOP
which is being deviated from; a written summary of the
risks; a written summary of the benefits; the aggregated
results from the heterogeneous group of experts; a break-
down of the attribute based values from the experts as-
sessments; a breakdown of the experts assessments. The
decision maker is also given associated information about
the calculation of the score.
Table 3 Guidance Scale (where x denotes the value pro-
duced by the FMAGDM method)
Value of Guidance
Workaround
SOP should be strictly followed. Value
of the workaround highlights severe
0 - < 0.4 lack of adherence to key attributes.
These types of workaround should be
prohibited.
SOP should be kept as a template, but
the information in the workaround can
0.4 - < 0.5
be considered and there is potentially a
need for improvement in the SOP.
Workaround is better but requires dis-
0.5 - < 0.6
cussion of amendments to SOP.
The workaround should be converted
into SOP to enhance operational safety
0.6 - < 1 and meets both the operational realities
on-board vessels as well as meeting
desired safety levels.
The communication of uncertain information is a chal-
lenging task and decision-making is also somewhat difficult.
In order to communicate clearly the results, the information
is provided in a visual format to enable decision makers to
better understand the variation between experts as well as
variation between attribute scores. A generic scale has also
been developed to assist the decision maker; the scale is
used to determine which category the result of the experts
assessments belongs and provide guidance about the ap-
propriate action for the decision maker to choose, this is
shown in Table 3 above. The decision maker will formulate
a decision based on the information.
TEST OF SPIS & INITIAL RESULTS
SPIS was tested in several stages and this section pro-
vides an overview of the various tests performed.
Collection of Workaround Data through Online Survey
The SEAHORSE Smart Procedures questionnaire was
made available online via (http://seahorseproject.eu/) and
distributed through partners of the SEAHORSE project.
This was done so to gain a rich database of workarounds. In
total, 451 responses were received. The survey is made up
of two sections, the first section of the survey focussed on
the collection of attitude data. The second section, pertain-
ing to workarounds, was completed by 295 participants.
After a careful review/ cleansing of the workaround data,
107 distinct workarounds were identified. These 107 work-
arounds were identified as one of four operations: deck
operations, cargo operations, bridge operations and engine
room operations. The categorisation of the workaround
assisted when assigning an expert to review the workaround
in the second stage of the methodology.
It was noted that, despite asking seafarers to fill out both
parts of the survey most of them missed workarounds part.
The reasons for only one part of the survey being complet-
ed is unknown, however, the researchers identified that
providing a large survey may lead to questions not being
completed. Therefore, the decision was made that the sur-
209
vey to be administered to seafarers would only capture data
about the workaround and some essential demographic and
attitude data.
Initial Results of Expert Workshop Assessing Worka-
rounds
Even though SOPs are designed to reflect the best way
of working, our results showed that in many cases the per-
formed workaround is better than actual SOP (Fig. 7). In
total 23 out of 107 (22%) workarounds were better than the
actual SOP according to FMAGDM method.

A workshop was held in Glasgow on March 3rd, 2015.

This reflects the testing of second stage of the methodology
where experts are asked to assess the workaround based on
a predefined set of attributes and an appropriate associated
Likert-scale. At the workshop 40 experts from both air and
maritime industry were present, these experts were assigned
into one of the four groups (deck operations, cargo opera-
tions, bridge operations and engine room operations) based
on their expertise.
Due to large in numbers, deck operation workarounds
were divided into two different booklets while the cargo
and bridge operations were consolidated as shown in Fig. 5. Fig. 7 Distribution of better procedure

In the following sample SOPs and associated worka-

rounds, pertaining to each type of operations are presented:

Cargo & Bridge Operations

A sample SOP & Workaround data of cargo and bridge

operations is outlined in Table 4. Here, as shown in Fig. 8,
SOP is demonstrating as a better option for the first case
with the ratio of 0.55 and workaround is showing a better
option for the second case with the ratio of 0.76.

Table 4 Sample Cargo & Bridge Operations SOPs and

Fig. 5 The categorization of the booklets
Workarounds Data
In all of the working groups, different experts from dif- a. Define the standard b. Define the alternative
ferent backgrounds were distributed equally and each procedure (If you find it practical way adopted by
No
working group consists of the given experts below in Fig. 6. easier copy and paste the crew to perform same
standard procedure) operation
The alternative it is filled
in with tick boxes and
Simple example - there
date only for recording /
are so many checklist for
audit purpose. But not
1 each navigational activity.
used as the objective in
This is to be done prior
ensuring the actual
commencing the activity.
checks have been com-
pleted.
Consolidated checklists
Completing separate
should be established for
checklists for daily
daily weekly and
2 weekly and monthly tests
monthly tests for naviga-
for navigation and com-
tion and communication
munication equipment
equipment
Fig. 6 Expertise of each working group
In the first case, crew members complain about filling
Experts were asked to self-assess their expertise in the an excessive amount of navigational checklists while they
workbook, thus differing a little from the proposed meth- have other commitments. The workaround aims to prepare
odology. Nevertheless, this highlights how experts can be a checklist of a navigational duty before performing the
weighted in an alternative means should a moderator not be actual duty just for the audit purposes. Checklists are very
available. crucial for the operational safety and they have potential to
The FMAGDM and TOPSIS method were implemented prevent failures that can turn into catastrophic accidents.
in Matlab to allow the assessment of the expert data and the Even though workaround seems a wise idea, not following
outcome of the FMAGDM and TOPSIS were provided. checklists can lead to the skipping of a crucial step and can
create a dangerous situation in shipping operations.

210

Fig. 8 Comparisons between sample Cargo & Bridge Oper-
ations SOPs and Workarounds
High workload is a well-known fact in the maritime in-
dustry. In addition to seafarers heavy duties, they are also
expected to fill in numerous papers for regulatory compli-
ance purposes. Companies can work on this issue in order
to decrease amount of paperwork, for instance by automat-
ing and collecting the same information just for once.
Risk assessment is a very crucial task to determine haz-
ardous operations and foresee possible dangers related to
them. Companies should provide sufficient training ex-
plaining the importance of conducting risk assessments and
their impact on avoiding accidents and incidents. Copying
and pasting from previous risk assessment data sheet is too
dangerous for a shipping company.

Second case is approved as an acceptable workaround Engine Room Operation

by the majority of the participants in the workshop. A sea-
farer suggested a consolidated version of navigational Table 6 gives a comparison between the sample SOP &
checklist which can decrease the workload without putting Workaround data pertaining to engine room operation. As
safety at stake. Instead of performing a repetitive work, a could be seen from Fig. 10, workaround is a better option
seafarer found a better solution without causing any dan- for the case 93 with the ratio of 0.52 and SOP is a better
gerous situation. option for the case 94 with the ratio of 0.72.
It is evident from case 93 that some of standard operat-
Deck Operation ing procedures do not comprise the details of a required
operation. Therefore, crew members ask help from other
Table 5, depicts a comparison between the SOP & seafarers who have higher ranks. Companies always review
Workaround data. As shown in Fig. 9, workaround is a their operating procedures and update the one which are not
better option for the case 32 with the ratio of 0.66 and SOP operable by considering crew members feedbacks.
is a better option for the case 33 with the ratio of 1.00.
Table 6: Engine Room Operation Workaround Data for
Table 5 Sample Deck Operations SOPs and Workarounds Case 93 and 94.
Data a. Define the standard b. Define the alternative
a. Define the standard b. Define the alternative procedure (If you find it practical way adopted by
No
procedure (If you find it practical way adopted by easier copy and paste the crew to perform same
No standard procedure) operation
easier copy and paste the crew to perform same
standard procedure) operation When carrying out the
Final reports are prepared task of checking flexible
At the end of each task. couplings between a
Reports will be prepared pump and motor, the
All reports are brought standard operating pro-
32 during the business
into a single report cedures differ in many Common sense, obtain
work-related tasks. These
reports are requested 93 ways to the equipment information from some-
between 2 -24 hours and isolating methods we one in a higher rank.
Crew often copy and have on-board. They are
paste risk assessments for written as a general
Conducting risk assess- guidelines but do not
33 other tasks - enter only
ment for a specific task apply specific procedures
'standard' risks and con-
trol measures for that unit.
During bunker operations
on large sea-going vessels
the vessel crew need to
visit the bunker delivery
barge (which is at a lower On some ships the junior
height) and check their Engine crew members
meters or do some paper (particularity the rather
work. After so many fatal shy ones) are encouraged
94
accidents around the to skip the gangway and
world in using mon- accept climbing down a
key/embarkation ladders monkey/embarkation
for this purpose, many ladder!
Fig. 9 Comparisons between sample Deck Operations SOPs shipping companies have
and Workarounds adopted the procedure of
using the ship's gangway
for such operations.

211

Fig. 10 Comparisons between sample Engine Room Opera-
tions SOPs and Workarounds
Case 94 demonstrates an unacceptable workaround
which can create risks for performing the duty. Even
though the companies advanced procedures and technolo-
gies, utilizing old and risky methods for fun should be
eliminated and prohibited. The company procedure should
be followed and the officers encouraging these kinds of
behaviours should be stopped by the company.
DISCUSSIONS
The role of SOPs within the maritime industry are rec-
ognised as paramount in ensuring safety of shipping. Within
the aviation industry SOPs have also been adopted and to
date have had much success. The introduction of SOPs has
not provided the same level of success as the aviation in-
dustry. Therefore, there is work to be carried out in im-
proving SOPs and it is proposed here that the development
of SOPs is to be an iterative process. SOPs are to be devel-
oped, then used in daily operations and based on the use
seafarers provide feedback on SOPs which are being devi-
ated from. From this, it highlights which SOPs may need to
be reviewed, this process is shown in Fig. 11.
Fig. 11: Summary of iterative process of design, imple-
mentation and refinement of SOPs.
A key issue related to SOPs and shipping safety relates
to the fact that seafarers may choose an alternative means
of performing a task or operation, thus deviating from the
SOP. It is assumed that managers in the organisation know
a proportion of the workarounds being practiced, whilst
others are unknown. To date, there is no methodology
which captures workaround data from seafarers, yet the
benefits and necessity of this are abundantly clear.
On-going engagement with several shipping organisa-
tions allows the demonstration of the proposed methodolo-
gy, these industry partners explicitly expressed there is a
212
significant need for a methodology such as SPIS. Without a
formalised procedures to capture and assess workarounds,
risky practices may be carried out in the daily operation of
ships putting seafarers in even greater danger in their daily
working life.
The unique contribution of SPIS is the development of a
means to collect workaround data from seafarers, the iden-
tification of the key attributes by which to assess worka-
rounds and the assistance provided for the decision making
process and subsequent communication. It is a well-known
fact that SOPs have crucial impact on operational safety
and by utilising the SPIS presented here, companies are
provided with a structured means for the elicitation of
workaround data and the assessment of workarounds by
experts in the organisation. In addition to the seafarer col-
lection of data, the methodology provides a means to bring
together the assessments of multiple experts to provide a
value of the extent in which a SOP is better than the work-
around.
In any case, the use of a workaround indicates an un-
derlying issue in the organisation and even if the SOP is
judged as the best alternative, decision makers need to
review the underlying reasons as to why the workaround is
being practiced. Also the decision making process is sup-
ported by the methodology. The recommendation may be to
adopt the workaround as it is safer and more effective than
the SOP. It will also identify risky workarounds which
should be prohibited for operational safety. Usually these
highlight an SOP which does not meet the operational reali-
ties on-board vessels, hence a review of the existing SOP or
the design of a new one might be necessary.
The methodology was tested during the Glasgow
SEAHORSE workshop by inviting 40 experts from the
maritime and aviation industries to attend. The methodolo-
gy received positive feedback on the real need of the pro-
posed approach in the maritime industry as well as on the
format of evaluation and the outcomes provided. The
workshop enabled further refinement of the methodology as
experts identified aspects causing issues in the assessment
of workarounds.
The test of the methodology revealed that there are
some workarounds which are conducted due to limitations
in the design of the SOP, however the chosen workaround is
not the best way to address the problem. Even though SOPs
are designed with the intention of providing the best way of
working, the results have shown that in many cases, per-
formed workaround is better than actual SOP. In total, 23
out of 107 (22%) workarounds were identified as better
than the SOP according to FMAGDM method.
CONCLUSION
Prevalence of workarounds shows that the maritime in-
dustry still has issues with regards to safety. Good practices
are required to be turned into formalised procedures while
bad practices should be eliminated. Identification of work-
arounds may also identify underlying reasons for the work-
around being practiced and this may provide valuable in-
sight into the limitations in the design of the SOPs.
To conclude, this paper provides a methodology to ad-
dress the existence of workarounds in the maritime industry.
It provides a structured way to support the identification,
review and decision making related to workarounds and it
can be easily transferred to other industries utilising SOPs Attributive Group Decision Making Methodology
such as the air industry where this gap in knowledge also and Its Application to Propulsion/Manoeuvring
exist. System Selection Problem. European Journal of
Operational Research 166(1): 93-114.
ACKNOWLEDGEMENTS Pidd, M. (2009). Tools for Thinking; Modelling in
Management Science, John Wiley and Sons Ltd.
The research presented in this paper, was carried out as part
of FP 7 Research project SEAHORSE; Safety Enhance-
ment in Transport by Achieving Human Orientated
Resilient Shipping Environment under the call
FP7-SST-2013-RTD-1, Technology Transfer in the area of
transport; Grant Agreement Number: 605639

REFERENCES

Bhattacharya, S. (2012). The Effectiveness of the Ism

Code: A Qualitative Enquiry. Marine Policy 36(2):
528-535.
Boehm, P., Page, D., Neff, J. and Brown, J. (2011). Are Sea
Otters Being Exposed to Subsurface Intertidal Oil
Residues from the Exxon Valdez Oil Spill? Marine
pollution bulletin 62(3): 581-589.
Chen, S., Hwang, C. and Hwang, F. P. (1992). Fuzzy
Multiple Attribute Decision Making(Methods and
Applications). Lecture Notes in Economics and
Mathematical Systems.
Chen, S. M., Lin, S.Y. (1995). A New Method for Fuzzy
Risk Analysis. Artificial Intelligence Workshop,
Taipei, Taiwan, Republic of China.
Ek*, . and Akselsson, R. (2005). Safety Culture on Board
Six Swedish Passenger Ships. Maritime Policy &
Management 32(2): 159-176.
Goulielmos, A. M. and Goulielmos, M. A. (2005). The
Accident of M/V Herald of Free Enterprise: A
Failure of the Ship or of the Management?
Disaster Prevention and Management: An
International Journal 14(4): 479-492.
Groeneweg, J., Van Kampen, J. and van der Beek, D.
(2013). The Value of Safety Indicators. European
HSE Conference and Exhibition, Society of
Petroleum Engineers.
Hetherington, C., Flin, R. and Mearns, K. (2006). Safety in
Shipping: The Human Element. Journal of safety
research 37(4): 401-411.
Hwang, C. L. and Yoon, K. P. (1981). Multiple Attribute
Decision-Making: Methods and Applications.
Springer-Verlag.
ICS. (2015). The Principal Regulations Governing
Maritime Safety. from
http://www.ics-shipping.org/shipping-facts/safety-
and-regulation/the-principal-regulations-governing
-maritime-safety.
IMO. (2012). Maritime Safety. from
http://www.imo.org/OurWork/Safety/Pages/Defaul
t.aspx.
Kahneman, D. and Tversky, A. (1982). On the Study of
Statistical Intuitions. Cognition 11(2): 123-141.
Mearns, K., Kirwan, B., Reader, T. W., Jackson, J.,
Kennedy, R. and Gordon, R. (2013). Development
of a Methodology for Understanding and
Enhancing Safety Culture in Air Traffic
Management. Safety science 53: 123-133.
ler and Odabai (2005). A New Fuzzy Multiple

213
 Safety Culture - Beyond the Horizon but Still in View
Martin Toland
QinetiQ, International Maritime Consultancy and Software, mjtoland@qinetiq.com
ABSTRACT
Safety Culture is frequently described as what you
do when nobody is looking, however in a world of ever
increasing regulation and supervision the occasions
when nobody is looking are increasingly rare. Tradi-
tional approaches to safety have been to empower those
on board to execute an agreed plan or set of standards
while allowing flexibility in the application to achieve
the requirements of a vessel or purpose, but this is fre-
quently no longer the case. Even a ship out at sea,
seemingly beyond the eyes of management and regula-
tor, is subject to numerous company safety policies as
well as national and international legislation and stand-
ards. Any failure to adhere or comply will be identified
through an audit or inspection programme and report-
ed for remedial action. The focus on compliance can be
a means of delivering metrics to demonstrate safety
culture to the satisfaction of shareholders, but poten-
tially impacts to the detriment of a descriptive approach
and development of a comprehensive safety case.
Learning from experience is essential to prevent re-
peated mistakes and enables sharing of best practice
both within a domain and across multiple industries;
but it is important to make sure the right lessons are
identified. Incorrect responses or increasing the extent
of compliance that is required increases the risk of
overly onerous measures being bypassed reducing the
level of protection provided instead of enhancing it. This
paper will discuss the challenges that this prescriptive
approach to safety can cause and examine the effect it
can have on the safety culture when compared to the
model of the DuPont Bradley Curve. The paper will go
on to discuss the relationship between the different ways
in which lessons are identified including accident re-
porting.
INTRODUCTION
The safety culture of an organisation is notorious diffi-
cult to define or gauge, but is perhaps most easily described
as what you do when no one is looking. The ability of
trained seafarers to follow the processes and protocols laid
down when supervised is not in doubt, but when the oppor-
tunity is there to make life simpler, albeit not necessarily
safer; corners are often cut to achieve a job in a quicker or
214
easier way. Historically the ability to depart from procedure
has been plentiful as the seafarer has been able to operate
away from the prying eyes of regulators or company offi-
cials, but the greater connectivity desired by crews has led
to a new challenge for them. A company can now get real
time feedback on activities while underway and identify
where there is a departure from approved practice. This
increase in accountability has a definite upside in that it
leads to greater levels of compliance with the rules as
transgressions will be identified and remedial action taken.
However the down side is also present in that it means ac-
tion related to safety is always taken to appease a supervi-
sor and stay out of trouble instead of progressing the safety
culture and encourage a positive attitude to safety leading to
continuous improvement.
DUPONT BRADLEY CURVE
The DuPont Bradley Curve (and variations thereof) has
been used as a way of demonstrating the culture of an or-
ganisation since it was devised (DuPont Sustainable Solu-
tions, 2012). Development of the idea means it can be ap-
plied to any organisational culture, in this case its attitude to
safety. Figure one shows the typical phases of culture de-
velopment on the curve. It shows that as improvements in
safety occur the motivation of an individual changes from
one of self to that of team.
Accident rate
Fig.1: Example of DuPont Bradley Curve
The first section on the curve covers the natural instincts
of a person to stay safe using their own knowledge and ex-
perience it is labelled as reactive as it people reactions to
the situations they face and have faced previously that is
keeping them safe. The overarching company goal is com-
pliance and is driven by the safety manager meeting stand-
ards defined by external organisations e.g. ISO or industry
bodies.
As the management commitment improves the company
goal progresses to the delivery of metrics or key perfor-
mance indicators (KPIs) the safety manager is joined by a
team of supervisors to impart the company safety messages.
Safety is included in inductions, routines, training and fur-
ther to this it will be enforced, potentially with the threat of
discipline for transgressions. Compliance by the individual
is therefore through the desire to not be in trouble, hence
the rules and procedures are followed. This is a stage which
many companies find themselves in; delivering safety
through rules and process, demonstrated by metrics report-
ed to stakeholders. Safety remains dependent on the com-
pany structures continuing to enforce the safety manage-
ment systems. The challenge then comes in moving to the
next steps.
In order to develop personnel independence for safety
an individual needs to receive the training and support to
understand the risks and standards to be achieved, but also
the empowerment to act accordingly. The ability to recog-
nize a hazardous situation developing and address it before
an incident or accident occurs requires commitment and the
safe habits of an organisation to improve. In this case the
management commitment is provide the work environment
where people have autonomy to act in the right manner, not
the setting where compliance is demanded without consid-
eration or thought. At this stage the individual is still think-
ing self and how to keep himself safe or out of trouble.
The final stage on the DuPont Bradley Curve is the abil-
ity of a group of individuals to act as a team to keep every-
one safe, not through compliance with a code, but through
understanding your own responsibility and how to keep
each other safe. The development of the interdependence is
not a short process; development of a safety culture from
the beginning can be a decade long process to fully imple-
ment a safety culture with genuine interdependence and
team members acting for the good of all for reasons other
than compliance and because the management says so. The
development of this culture is made more challenging by
the turnover of crews, time away from the ship and the in-
tegration of different nationalities and creeds. Throughout
the curve the improvement of safety culture sees the acci-
dent rates fall with an aspiration of zero incidents and it is
only when genuine interdependence is achieved that zero
incidents is a realistic goal.
The aspiration of an organisation should always be to
improve its safety culture to a point where it is operating in
the interdependence section of the curve. However, there
may be obstacles to overcome. The connectivity referred to
above means that delivery of KPIs can be near constant and
the level of supervision such that the ability to move into a
self and then team motivated phase is limited. Compliance
remains the target as this is what is being measured and
reviewed in audits and a standard that is safe enough be-
comes the acceptable level.
The safety culture is not just limited to the professional
seafarers; people are going to sea in ever greater numbers
on cruise ships. Lessons of the past have led to the devel-
opment of Safe Return to Port requirements and the princi-
ple that the ship is the best lifeboat, while undeniably true
measures must still be put in place to allow to the evacua-
tion of several thousand people in a short period of time in
potentially hazardous settings. Notwithstanding the inves-
tigation into the cause of the emergency and how to prevent
that we can study the behaviour of people during evacua-
tions, to ensure the Life Saving Appliances (LSA) and
briefings given to those who are not seafarers can be im-
proved to increase the survivability in accidents.
COMPLIANCE
The achieving of safety criteria can generally be con-
sidered to be achieved in two ways, through compliance
with standards or through risk management. The ability to
conform to a standard is, relatively, straightforward to do as
standards offer clarity on both the level to be achieved and
the manner in which compliance can be achieved. As such
compliance can be demonstrated to the satisfaction of audi-
tors, stakeholders and a certification issued accordingly.
Standards are best demonstrated when applied to
equipment, for example a life raft will be designed and built
to achieve performance parameters, a defined testing re-
gime will have been applied in build and specified equip-
ment is contained inside meeting a requirement dictated by
the environment and application. Although the standards
with the associated documentation is often seen as red tape
the ability to show how a standard is reached is simple and
any equivalence between multiple standards universally
agreed if operating across multiple jurisdictions.
Within a company ongoing compliance can be moni-
tored with the use of check off lists or internal audits and
metrics easily produced. The use of ISO, SOLAS and MED
standards give confidence to an organisation that when they
procure equipment it will be fit for purpose and can be used
in the environment and manner described within the stand-
ard. The greater challenge is present when equipment built
to a set standard is intended to be used in a manner or envi-
ronment outside the defined scope of supply. In this in-
stance a risk based approach is necessary to identify the gap
between the standard and the intended use. Resources are
available to enable commensurate regulations to be identi-
fied and analysed including the extract from the MED web-
site (MarED, 2016) in Figure 2.
Fig.2: MED standards for liferafts A.1/1.12
RISK MANAGEMENT
Risk based safety techniques are used when a standard
or established best practice is not available to set a bench-
mark against which equipment or processes can be assessed.
It can often be the case that it is harder to demonstrate a
safety objective is achieved in a risk based approach as
215
there is always a further analysis to be done or an extra
mitigation to be applied albeit with diminishing returns on
each extra safeguard. Even when following a prescribed
hierarchy of risk control the ability to demonstrate that
enough has been done could be a challenge. This can lead
to the requirement to perform a cost benefit analysis before
a risk can be shown to be as low as reasonable practicable
(ALARP).
It is also potentially complicated to show a requirement
is met or how a risk based assessment such as Failure Mode
and Effects Analysis (FMEA) or Failure Modes, Effects and
Criticality Analysis (FMECA) has delivered results com-
mensurate with a specification or standard. For this reason
some organisations do not like risk based safety because of
the challenge of showing that safe is safe enough, but when
fully understood a risk based approach gives many benefits.
Adherence to an over specified standard can be an onerous
and time consuming issue, but if the risk is assessed a lower
specification can be opted for and still deliver a level of
safety commensurate with the requirement. An example of
this is the EU funded Safepec project which is looking at
risk based approaches to maintenance at the moment with
an aim enabling ship owners and operators to target re-
source where it is most needed. Risk based approaches need
a safety culture where the flexibility is provided to find the
most applicable set of safety mitigations to achieve a re-
quirement. This needs a developed safety culture where
engagement occurs at all levels in an organisation.
The ideal balance between standards and risk based ap-
proaches is a hybrid of both schools of thought. A standard
e.g. ISM code (IMO, 2015) will specify a requirement, but
not how it is to be achieved or demonstrated. This enables
a company to have adherence to a standard, to the satisfac-
tion of regulators, stakeholders and interested parties, while
tailoring the safety management solution to suit the applica-
tion. For example ISM Code, Part A, 7 Shipboard Opera-
tions states: The Company should establish procedures,
plans and instructions, including checklist as appropriate,
for key shipboard operations concerning the safety of the
personnel, ship and protection of the environment. The
various tasks should be defined and assigned to qualified
personnel.
The presence of the procedures, plans instructions etc.
will enable compliance to be demonstrated, but the extent
of the plans, level of detail, which operations are identified
as key, the qualifications required etc. can be established
through a risk based approach commensurate with the en-
vironment and operations involved. This level of flexibility
is essential especially when working with a standard such
as ISM code which can be applied to many different ship
types ranging from a small workboat through to a panamax
container ship.
As previously identified above reliance on compliance
can limit the ability of a safety culture to grow as no op-
portunity is given to challenge the conventions and laid
down practices to make a safer or more efficient set of sys-
tems. The use of frameworks and codes such as ISM can
deliver an optimum balance of compliance and risk based
approaches and enables the risk based methodologies to
supplement the prescribed requirements. This is not to sug-
gest that this solution will by default identify and manage
all risks to a zero accident level. Any system is only as good
as the implementation and effective management required
216
to deliver the stated objectives.
NEAR MISSES AND INCIDENT REPORTING
When people are encouraged to apply a proportionate
and risk based approach to safety they are also encouraged
to objectively look at their surroundings. This will result in
seafarers identifying scenarios that are unsafe or require
resolution; these could be accidents or incidents that have
occurred or hazardous situations which may still be devel-
oping. In a strong safety culture these will be reported for
rectification but to allow this culture to grow it is important
that certain rules are observed. Reporting must be part of a
no blame culture; notwithstanding negligence or serious
shortfalls in the required standards reporting should not
lead to disciplinary action. Incidents should be viewed as a
learning opportunity and as such the more reports received
the more opportunities to learn. To that end using a down-
ward trend in accidents/incidents as a KPI can be detri-
mental as incidents will go unreported and lessons not iden-
tified and subsequently an accident occurs; a failing many
companies do. The management of information and what it
means is important too. A recent increase in incident re-
porting in the Royal Navy nuclear submarine fleet was
greeted with alarm in the media, but was actually the prod-
uct of a concerted effort to increase the reporting of unsafe
incidents to prevent accidents from occurring in the future.
Where lessons are identified it is important to share com-
mon faults and best practice, especially where a fleet of
ships operate with common procedures or crew rotations
means not everyone will be present for training or work-
shops post event. A strong safety culture will also allow for
external sharing of information. Programmes such as the
Nautical Institutes Mariners Alerting and Reporting
Scheme (MARS) (Nautical Institute, 2016) enable other
owners and operators to learn from others outside of their
organisation.
The culture also facilitates an environment where the
reporting of problems is not seen as whistle blowing or
criticism of a given course of action. The lookout on the
bridge is not picking fault with the Officer of the Watch
when he points out a previously unobserved ship; he is
performing his role in the team to allow for the safe opera-
tion of a ship. Likewise the guidance to use Personal Pro-
tective Equipment (PPE) or follow a procedure is not fo-
cusing on shortcoming it is preventing an unsafe or haz-
ardous situation form occurring. The ability to challenge
behaviours is essential in a good safety culture; it is through
such challenges and the receptive nature of the response
that continuous improvement can be achieved.
ACCIDENT INVESTIGATION
Traditionally accident investigations have found it dif-
ficult to gather the information that is needed. The fear dis-
cussed above that a scapegoat is being sought or that re-
criminations or disciplinary action will follow can lead to
the facts being obscured. The increase in digital technology
while alleviating one aspect of this problem, enabling the
investigator to deal in facts rather than incorrectly recalled
memories or distortions of events, can lead to other chal-
lenges. The volume of CCTV imagery, black box data and
voyage reconstruction information can provide information
overload making it harder to identify the root cause cor-
rectly. In the era of big data this is a problem experienced
in many aspects of the marine industry, not just accident
investigation; how much information do you need to make
an assessment, and just because you can collect data will it
be useful once you have it? (Swanson, 2013).
One option for investigations is to perform a virtual
re-enactment of the event and understand the promulgation
of the emergency. The same software which is used to per-
form a pre-emptive assessment during the design stage can
be used to model the actual damage seen and understand its
cause. This information can then be fed back into subse-
quent designs for new build vessels or give detail for modi-
fications and retrofitting of solutions for an operational fleet.
In adherence with established hierarchies of risk control the
focus of any finding should be on prevention through engi-
neered solution, not the addition of check off list require-
ments or retrospective application of PPE.
The ability of an investigator, be that internal or external,
to interpret the facts and produce a series of observations or
recommendations to prevent future occurrence may be lim-
ited by another factor of safety culture; the attitude of the
seafarer themselves. There are numerous examples of Per-
sonal Flotation Devices (PFD) saving peoples lives, and
conversely examples of the failure to wear an appropriate
life jacket or PFD being cited as a factor in fatality follow-
ing a man overboard incident (MAIB, 2014a). In the com-
mercial fishing fleets there have been safety initiatives
(RNLI, 2013) to provide life jackets at PFD to people free
of charge, supported by education campaigns, to encourage
their use. Despite this many people are still using the pres-
ence of the equipment to appease an auditor or surveyor
instead of actually wearing it for protection in the event of
an accident. The attitude of it will never happen to me or
the complacency brought about by experience can be a bar-
rier to the progression along the safety culture curve. Also
damaging is the effect this behaviour can have on others. If
an experienced crew member witnesses a veteran seafarer
foregoing safety equipment, deeming it unnecessary that
attitude will be inherited and taken forward.
When an accident occurs it is important to capture all
the lessons that are available. To achieve this external help
can be very valuable. This external help may come from
another department or division of a company or a third par-
ty for example an external consultancy. In many regards a
truly independent third party has advantages as it will better
facilitate an objective review free from internal politics and
the personalities involved. A poorly performed investigation
in worse than none at all and becomes a box ticking exer-
cise explaining away issues and non-conformities instead of
highlighting them to be addressed. External investigations
may also have access to test and evaluation facilities not
available to an internal investigation. For example QinetiQ
conducted a report for the MAIB into the effectiveness of
radar reflectors fitted to yachts after an accident involving
the yacht Ouzo and the passenger ro-ro ferry Pride of Bil-
bao the report showed how different geometries and sizes
of reflector perform and the variation in Radar Cross Sec-
tion that was achieved (QinetiQ, 2007). QinetiQ were able
to use test equipment otherwise employed for military test-
ing to deliver a comprehensive set of data and give recom-
mendations regarding the options available in the market to
provide radar reflection and prevent future accident scenar-
ios.
A secondary benefit of using an outside agency to con-
duct an audit or investigation is the ability to view a prob-
lem and the possible solutions in a new way. An internal
investigation will have the inherent bias brought about by
the prior knowledge of company operations. For example
an accident on board a fishing vessel investigated by an
owner or another skipper will be viewed with a fishing
vessel centric opinion. If the investigation is conducted by
someone with a workboat or rescue boat background they
might reach different findings, potentially identifying a root
cause or solution which has been used successfully in a
different sector or whole industry. The ability to identify
solutions from other specialisations prevents companies
operating in isolation from new initiatives. For example the
use of simulators in aviation training has been common
place for a long time, but it is being adopted into new do-
mains such as medical to improve diagnosis or conduct a
virtual operation prior to a patient being operated on for
real.
ROOT CAUSE
The purpose of any investigation is to identify the root
cause in order to prevent reoccurrence, but must not fall
into the trap of preventing the last accident; it is the next
accident which must be addressed. If an investigation does
not find the root cause you prevent the same accident oc-
curring, but do not stop commensurate accidents with the
same hazard. For example a current topic of concern in the
marine industry is the number of deaths associated with
confined spaces. The accidents have varied; some with fail-
ures in checking atmosphere is safe to breathe, some failing
to provide trained personnel and some within sufficient
rescue equipment, but each has the same end state. Each
accident has been accompanied by investigations by the
authorities, but lessons have not been effectively captured
in order to update processes, ensure adequate training and
provision of rescue equipment hence accident still occur.
If the root cause is not identified the wrong corrective
action may be applied; a man overboard scenario where the
efforts to get the casualty back on-board are protracted and
ineffective may lead to a conclusion that the man overboard
equipment is insufficient or the training in its use ineffec-
tive. Subsequently there may be investment in rescue boats,
recovery equipment and remedial training. The root cause
however may be the insufficient safety rails which failed to
prevent the fall with the contributing factors of man over-
board equipment suitability or familiarity. By responding to
the wrong indicators or looking to fix a problem in the
quickest way an organisation may find itself preventing the
last accident instead of preventing the next. This can lead to
adding new or superfluous barriers to an accident scenario
instead of addressing flaws in the existing system.
A common response to an accident is to add an addi-
tional item to a check off list or operating procedure. As
stated above this may only tackle one step of the accident
sequence and not have the desired effect. Worse still the
addition of a check off list item may be given to someone
who is already overloaded or preoccupied with other re-
sponsibilities; in this eventuality the check off list may not
be completed owing to other pressures as in the case of the
Finnarrow where stabilizers were not correctly retracted in
217
preparation for entering harbour (MAIB, 2013) or its com-
pletion becomes a nominal act with no added value to the
safety of the ship or personnel.
Fig.3. Finnarrow (MAIB, 2013)
Either way the desired effect of providing an additional
barrier to prevent the accident will not be achieved. More
preferable to check off lists which may be ignored or incor-
rectly administered is to produce an engineering solution to
prevent the accident scenario. This could be in the form of a
physical guard or barrier around machinery or, with the
increasing level of technology involved, a safe mode or
software inhibition alerting the user to an unsafe scenario or
initiating a preprogramed response.
Failing to address root cause may also occur through
failing to take into account a scenario changing from the
accepted normal. The marine environment is evolving and
facing new challenges; ships are larger, piracy threat, in-
crease in migrant traffic consequently the established miti-
gations may not remain as effective and must be evaluated
in any investigation to ensure a proper root cause is under-
stood and the relevant mitigation applied and if necessary
systems updated, not merely added to an existing system.
A further complication can be seen when considering
remedial training if there is deemed to be a shortfall in
competency as part of the root cause. Drills and emergency
scenario training often only deal with initial response and
not follow up actions. While it is acknowledged this is be-
cause of time constraints and lean manning reducing the
ability to run realistic scenarios it is still a consideration
which reduces the training benefit. For example a drill to
respond to a galley fire will rarely extend beyond the fire
being extinguished and perhaps smoke clearance, the issue
of feeding a hungry crew and potentially passengers with
no galley is not often discussed.
By taking the opportunity to learn the right lessons from
accidents we get the opportunity to challenge the paradigm
and consider disruptive technologies which can change the
levels of risk experienced. New concept designs are emerg-
ing on a regular basis as naval architects consider new ways
of tackling existing problems or new operational methods.
A thorough accident investigation may be able to highlight
the greatest hazard and provide the areas most applicable
for autonomy to remove personnel form the area of risk.
The frequency of incidents involving fishermen caught in
winches is increasing this is especially a concern as more
fishing vessels are responding to the financial pressures of
the industry by engaging in solo manned fishing (MAIB
2014b). Consequently when accidents occur there is no one
on hand to assist and prevent entanglement. If automated
218
solutions are able to reduce the exposure to the entangle-
ment hazard and reduce the accident frequency as in in ac-
cordance with a hierarchy of control where the removal and
isolation of the hazard is more important than guarding or
providing PPE (HSE, 2002).
CONCLUSIONS
It is clear that as organisations seek to grow they need to
invest in their people and ensure not only their physical
safety, but also the environment in which a positive safety
culture is encouraged. A safety culture cannot be bought in
as a simple commodity; it takes investment in training, fa-
cilities, but above all time. A strong safety culture can take
up to a decade to be fully implemented and self-sustaining;
it can however be lost in a far shorter period of time. Per-
sonnel changes, lack of management commitment or pres-
sures being applied to achieve targets without a focus on
safety and compliance can all lead to a rapid decline in
safety culture and accidents. One product of a good safety
culture is the confidence to report near misses in a no blame
culture and the aspiration to make a working environment
safer for all. Despite all this accidents will still occur, albeit
infrequently. The thorough investigation of accidents is
crucial to identify the right lessons and deal with the root
cause of an event to prevent future accidents instead of
forever being caught in the trap of preventing the last acci-
dent not the next one.
REFERENCES
DuPont Sustainable Solutions (2012), The DuPont inte-
grated approach (DNA) for safety: A catalyst to acceler-
ate performance
Health and Safety Executive (2002), Marine Risk Assess-
ment, Offshore Technology report 2001/063
International Maritime Organisation (2015), ISM Code
International Management Code Resolution A.741(18)
Marine Accident Investigation Branch (2013), MV Fin-
narrow Contact with berth and subsequent flooding Re-
port no 24/2013
Marine Accident Investigation Branch (2014a), Case 20
Kill cords and Lifejackets Your Tools for Survival
Safety Digest Lessons from marine Accident Reports
1/2014
Marine Accident Investigation Branch (2014b), Case 17
Drum Roll But No Fanfare Safety Digest Fishing
2014
MarED (2016), The group of Notified Bodies for the im-
plementation of the Marine Equipment Directive EU
Council Directive 96/98/EC as amended,www.mared.org
Nautical Institute (2016), 201638 Ships passing in the night
- just
QinetiQ (2007), QINETIQ/D&TS/SEA/CR0704527/2.0
Performance investigation of marine radar reflectors on
the market
Royal National Lifeboat Institution (2013), Press release:
RNLI supports free lifejackets for Scottish fishermen ini-
tiative
Swanson, P (2013), AIS Information overload; Too Much
of a Good Thing Power and Motoryachting Magazine,
March
 Autonomous Shipping A Concept Design for an
Autonomous Firefighting and Rescue Vessel
Simon Pullin, Samantha Hill
QinetiQ International Maritime Consultancy and Software, sihill@qinetiq.com
ABSTRACT
QinetiQ have developed a concept design for an au-
tonomous firefighting and rescue vessel to illustrate the
possibilities for autonomous shipping to undertake dull,
dirty and dangerous roles, and to enable discussion of
the considerations needed for the design of autonomous
vessels. The role of firefighting and rescue was identified
as one which is required on an irregular basis, yet is
required for the protection of significant assets in ports
or offshore. The attractiveness of an autonomous vessel
for a role which involves considerable periods of loiter-
ing is clear, but also has the potential to remove people
from the potentially dangerous firefighting activity. In
considering the firefighting role, it is easy to see a situa-
tion where the firefighting vessel would be faced with
people in the water or in liferafts/lifeboats, where a
manned vessel would provide a rescue facility and
transit to shore. The challenge was to incorporate this
kind of function into an autonomous vessel and to iden-
tify any potential limitations.
More generally, the design of an autonomous vessel
poses questions in terms of the standards and regula-
tions in place designed to provide safety at sea. A case
can be made that the vessel design need not incorporate
many of the features which provide a safe workplace for
those on board, but there will need to be consideration
of staff coming on board to conduct servicing and
maintenance, and also consideration of any interaction
with manned vessels.
INTRODUCTION
Unmanned machines and vehicles have entered the pub-
lic consciousness in recent years. Developments in sensor
and control technologies mean that autonomous applica-
tions are expanding rapidly. The aerospace industry has
been developing and advancing unmanned products for
some time and unmanned aerial vehicles are available to
buy as off the shelf products from the small drones avail-
able to the general public to the payload carrying unmanned
systems in widespread military use.
Development in the aerospace industry reflects the gen-
eral attitude to unmanned systems - currently even in the
military sector although the technology would enable in-
creased levels of true autonomy with the vehicle taking
decisions based on situational awareness, there is always a
human in the loop. The first steps into true autonomy are
currently being taken by the automotive industry, trialling
driverless vehicles on public roads. To date, development
shows that there is still work to do mostly due to the un-
predictability of human operators of nearby vehicles, and to
convince the public that driverless vehicles are desirable.
The marine sector is some way behind other sectors in
developing and implementing unmanned technology. To
date unmanned technology has only been utilised for very
simple pre-set operations within the marine market such as
topographic surveys, met-ocean data gathering and military
target practice for example. The technology and systems
developed by these other industries are clearly transferable
to the marine industry together with the lessons learnt on
regulation, safety and public acceptance.
QinetiQ IMCS has been established to focus on work
within the commercial marine market and has the enviable
position of having the backing of a significant amount of
knowledge and experience from military applications in
Land, Marine and Air sectors. The naval architecture con-
sultancy division of IMCS has developed a concept design
for an autonomous offshore standby firefighting vessel.
QinetiQ established the requirements for the vessel after
considering what applications would most benefit from
utilising an autonomous ship. The concept design has been
produced as a practical platform to consider the challenges
and benefits of the application of autonomy to a particular
role.
THE ROLE OF AUTONOMOUS VESSELS
When considering suitable applications for an autono-
mous ship there are a number of considerations that make
an application particularly suitable for early adoption.
These are commonly tasks which are dull, dirty or danger-
ous. It clearly makes sense to remove humans from dan-
gerous operations, and in developed nations particularly, it
is difficult to recruit and retain personnel to undertake these
tasks. Because of this, these roles are likely to be the first to
gain an acceptance of the utility of autonomy by the con-
servative marine market.
Firefighting and rescue vessel fits into this category. For
the majority of the time vessels allocated for this role are on
standby, but when called to action, the role is highly dan-
gerous for the crew. In addition, there are other factors
which make autonomy an attractive proposition:
Clearly defined tasks Autonomous applications
are far easier to establish if there is a well-defined
task with known constraints and boundaries. The
computer systems can be much more easily coded
to deal with all eventualities if the task is well
bounded. The firefighting ship concept provides a
very well-defined task of being on standby within
a given area and when alerted sprinting to a de-
fined location and beginning firefighting duties.
Whilst the retrieval of survivors is not quite so
concisely defined it still remains a task which is
potentially achievable by autonomous systems;
Improved capability There are applications
which simply cannot be carried out by crewed
ships or which are too hazardous to justify putting
219
 a crew into the situation. If there is a demand for
such an application the adoption of autonomous
technology for a ship could be more easily sup-
ported. Additionally there are many applications
where the human element limits the capability of
the asset. For example, having to return to port on
a regular basis for food supplies and for a change
of crew limits the time a ship can be carrying out
its tasks. The firefighting concept could benefit
from longer endurance on standby with no need to
change crew or replenish food stores. The ship
could also arguably remain on standby in worse
weather conditions than a crewed ship thereby
providing a greater capability. When fighting a fire
the ship could be put in more hazardous positions
in order to better fight the fire than a crewed ship
could be, simply because the concern for the lives
of the crew has been removed and the vessel could
even be considered a disposable asset;
Removing human error Many accidents and in-
cidents are caused by human error. Whilst human
input has certain benefits over computer controls,
in many applications computer control offers sig-
nificant benefits whether it is more accurate or re-
liable or whether the task is completed more
quickly. Applying computer control to crewing
functions could be seen merely a progression of
the adoption of technology. Removal of humans
from the control of the ship could potentially pre-
vent accidents, especially those caused by fatigue
which can occur especially in stormy conditions
when the crew need to remain in an alert state to
respond to an incident;
Cost saving crew costs are significant and re-
moving this cost could realise sizeable cost sav-
ings for an application. Where there is a demon-
strable and significant cost benefit by removing
the crew from a ship there will be an incentive to
achieve the cost savings by moving towards un-
manned / autonomous ships. On the contrary,
where the crew cost is insignificant to the overall
cost of an operation there is no need to risk the
adoption of autonomous technology which could
result in complexities and capital cost increases. In
addition to the direct cost of employing a crew
there are also associated cost savings such as re-
maining at sea for longer periods reducing time
transiting to port etc. Any operational cost savings
need to be assessed against any increase in the
capital or support costs of providing an autono-
mous ship. In the early adoption time period the
capital costs of designing, building and classing an
autonomous ship will potentially be significantly
higher than equivalent traditional ships.
Dedicated firefighting vessels are currently confined to
port and inshore operations. The autonomous firefighting
and rescue vessel concept proposes a larger vessel aimed at
offshore operations, where this role is currently undertaken
as just one of those an offshore or platform support vessel is
able to undertake. Offshore oil and gas installations are
particularly challenging in firefighting, due to the intense
heat of hydrocarbon fires and the potential for harsh open
water environmental conditions.
220
For offshore installations, a standby safety role is re-
quired by UK legislation when necessary to ensure com-
pliance with the regulations for effective personnel recov-
ery and rescue. The Prevention of Fire, Explosion and
Emergency Response on Offshore Installations (PFEER)
UK HSE regulations 1995 require plans to be in place to
ensure safe evacuation to a place of safety or to a place
where personnel can be recovered to a place of safety. Cur-
rently standby vessels are used as mitigation in an offshore
safety case when operations of higher risk are conducted.
In the US, responsibility for firefighting lies with the
owner and operator of offshore installations. US law re-
quires through Code of Federal Regulations 33 that installa-
tions within 12nm of land have plans in place to ensure a 12
person fire team will be in attendance within 8 hours and
for installations between 12 and 50nm of land the same
team must be in attendance within 12 hours.
The scenario for the firefighting application requires the
ship to be on standby, in all weather conditions, and ready
to respond to a fire on any of a number of platforms in a
given area. The ship then needs to get to the scene as
quickly as possible and commence firefighting duties, re-
quiring fire monitors and dynamic positioning capability.
There is also a high likelihood that in the event of a
large fire, survivors would have evacuated the platform and
be in the water, in liferafts and in lifeboats. SOLAS
amendment in 2014 to regulation III/17-1 requires all ships,
(including those to which SOLAS chapter III does not ap-
ply) to plan to be able to recover persons from the water
and from survival craft. It was decided therefore to consider
how the ship could have the facility to recover and provide
refuge for survivors in addition to the primary firefighting
capabilities.
The following basic requirements were established for
the design:
Designed to be usually unmanned and autonomous
in operation
Capability to transit 150Nm @ 12 knots to and
from operational area
Maximum speed of 20knots
Loiter on standby in field for minimum of 90 days
Minimise fuel consumption whilst loitering
Dynamic Positioning capability in accordance with
DNV DP II requirements
Firefighting capabilities in accordance with DNV
FiFi III (I+)
Provide means of recovering survivors and suita-
ble temporary accommodation
Provide a bridge for manned navigation for limited
time periods, e.g. pilotage
Suitable for operation in the North Sea
OUR DESIGN
The chosen hull-form is a standard monohull design that
can cope well with heavy seas, achieve the maximum speed
with a reasonable installed power and offers a simple and
well understood construction. A large roll period when
compared to a multihull design is also favourable for main-
taining accurate targeting with the fire monitors. It was also
chosen to look very similar to most other offshore support
vessels so that attention is not focussed on a novel design
without really considering the main benefit of the design,
that it is autonomous in operation.
The design has the following principal dimensions:
Length overall 80.2 m
Beam 12.0 m
Depth 7.0 m
Draught 4.6 m
Where possible, well established and relatively simple
technologies have been used. Reliability of equipment
aboard an autonomous vessel is vital as there are no means
of fixing simple failures whilst at sea and even overcoming
false alarms can be significantly more challenging than on a
manned vessel. The prime movers are all medium speed
diesel engines; the main propulsion is through direct drive
electrical motors driving fixed pitch propellers with con-
ventional rudders behind. All the thrusters are tunnel
thrusters rather than drop down azimuthing thrusters to
keep the systems as simple as possible.
The size of the vessel has been kept to a minimum to
fulfil the requirements, the size of the prime movers, fire-
fighting equipment, main propulsion and thrusters effec-
tively determined the minimum dimensions of the design
however additional length was added to reduce the power
requirement to achieve the 20 kn maximum speed. Whilst
the length is comparable to small traditional offshore sup-
port vessels the beam is somewhat less as there is no need
for any significant deck payload capacity or specific stabil-
ity design drivers. The size of the design allows it to carry
the required fuel for the task and comfortably passes all the
standard IMO and Class stability requirements for a ship of
its type.
The machinery has been chosen to allow energy effi-
cient operation, the role profile of the vessel demands that
in the loiter mode very little power is needed but in both the
firefighting and sprint modes there is a high power demand
from different consumers in each case. Therefore a full
diesel electric arrangement has been chosen with 6 prime
movers of 2 different sizes, 4 x 2.7MW and 2 x 0.6MW
generator sets. In addition the top of the superstructure is
covered with a large array of photovoltaic panels and a
sizeable battery bank is fitted to minimise the need to run
the diesel generators whilst loitering. Whilst sprinting the
vast majority of the power is required by the twin 5MW
main shaft electric motors to achieve a 20knots top speed.
Once on station and fighting a fire in dynamic positioning
mode the power is required by the electrically driven
thrusters and fire pumps.
Redundancy of systems is important on-board an au-
tonomous vessel as there is no means to repair or physically
interact with equipment that has developed a fault. For
many items of mechanical equipment the final reversionary
mode is manual override, which is clearly not an option on
an autonomous ship. Whilst individual component design
can be improved to increase reliability the simplest way to
improve reliability of the ship is to provide redundancy in
the design. Therefore where possible a number of smaller
units have been fitted, for example twin shafts and rudders,
three bow and two stern thrusters, four fire pumps, four
water monitors, two foam monitors and six prime movers.
Whilst the details of supporting systems have not been es-
tablished for this concept the same philosophy would be
used throughout the system designs. For example multiple
fuel filters, cooling systems, AC units etc. typically with
bypass lines or switchable units. Whilst this is not signifi-
cantly different to the design of a manned ship the level of
redundancy for an unmanned ship will invariably need to be
greater especially whilst the technologies are still develop-
ing to facilitate the progression of unmanned ships.
The firefighting equipment has been selected to achieve
compliance with FiFi III (I+) requirements, which includes,
4 electrically driven water pumps, 4 water monitors, 2 foam
monitors and 1 whole boat external water mist system.
Tank capacity is provided for the storage of up to 100
tonnes of foam concentrate, providing at least 1 hour of
maximum foam generation. To satisfy the minimum re-
quirements of FiFi III regulations the ship needs approxi-
mately 170te of fuel storage reserved for remaining on sta-
tion and fighting a fire for the minimum 96 hours. The total
fuel storage capacity of the ship is 500te therefore the ulti-
mate duration for which firefighting activity can be sus-
tained is dependent upon how long the ship has been on
standby and the distance from port.
To enable the design to be used in North Sea oil fields
the fuel tank arrangements are compliant with CLEAN-
DESIGN regulations.
A small bridge is included which will allow temporary
navigation from on board. No provision is made for crew
accommodation with the intended use of the bridge being
only for navigation when entering docks, highly congested
ports or ports that will not accept entry of an unmanned
vessel.
Recovery of survivors is carried out via the rescue zones
positioned port and starboard on the main deck. These res-
cue zones are 10 m long and comply with the DNV rescue
zone requirements for offshore standby vessels. They are
provided with drop down scramble nets to enable survivors
in the water or in liferafts to climb out and embark onto the
ship. Lifeboats are recovered with survivors onboard via the
A-frame over the transom. The lifeboat is held on deck
whilst survivors disembark and is then returned to the water
to allow another lifeboat to be recovered. This facility has
been provided as the majority of lifeboats used on offshore
platforms are freefall lifeboats which would be difficult to
disembark from and climb onto the ship via the rescue
zones.
Twelve 50 man liferafts are positioned on the aft deck
some of which would be released for survivors in the water
to get aboard. The remaining liferafts are retained to ensure
that the vessel has liferaft capacity for the survivors
on-board in the event of an emergency on board meaning
that the autonomous ship itself has to be abandoned.
Accommodation is provided within the superstructure at
main deck level for up to 150 survivors in accordance with
Class regulations, including treatment and recovery suites.
A helicopter winching area is provided atop the main su-
perstructure, compliant where possible with the UK CAA
CAP437 standards. The winching point allows for shore
based staff and medical personnel can be quickly brought
out to the vessel to provide treatment for survivors as well
as allowing urgent transfer of medical emergencies to hos-
pital ashore.
221

Fig.1 Renders of concept firefighting and rescue vessel
DESIGNING FOR SHIPS WITH NO CREW ON-BOARD
Designing a ship in the 21st century is as much about
meeting regulation, and providing for the interaction which
people will have with the vessel as it is about the naval ar-
222
chitecture and marine engineering. Designing an autono-
mous ship removes much of the human element and the
aspects of regulation which are all about human safety and
comfort on board.
This freedom has allowed us to consider each aspect of
the structure and outfit of the vessel and its operation. We
recognise that there will need to be regular access for
planned and unplanned maintenance, so the requirement for
access to machinery and electrical equipment remains the
same as for a manned vessel. There is a limited bridge fa-
cility included in the design to allow for a scenario where
local manned control of the vessel is necessary. We can see
a similar scenario to UAV control, where the vessel is con-
trolled in this situation by a remote operator, and indeed,
envisage semi automation and remote control as the logical
stepping stone to full autonomy.
By having no accommodation not only does it free up
space but the concerns over noise and vibration or ship
motions affecting the occupants are removed which allows
machinery arrangements to be optimised for other purposes.
Ultimately, depending upon the role of the ship, au-
tonomous ships have the potential to be somewhat smaller
than conventional manned counterparts. Whilst the ship
needs to maintain its integrity and functionality and for
some tasks provide a stable platform, ultimately the ship
can experience much greater motions without there being
the concern for a crew only for the systems on board. This
potentially allows the ship size to be reduced as a common
approach for a ship to provide a habitable platform in more
severe sea conditions is to make it bigger to limit the mo-
tions.
CHALLENGES TO BE OVERCOME
The single largest challenge to overcome for autono-
mous ships to become reality is in the regulatory framework.
As identified, regulations which ensure safety and comfort
on board may be able to be dispensed with, but like the
current scenario with driverless cars, the challenge is the
interaction between unmanned and manned vessels.
The most obvious and possibly the most challenging set
of regulations being COLREGS, which is quite under-
standably worded assuming that tasks are being carried out
by humans on-board all ships. Arguably SOLAS regula-
tions need not apply to autonomous ships but again there
would need to be new standards for some of the areas cov-
ered by SOLAS to prevent, if nothing else, the oceans being
littered with stricken ships. The firefighting concept has a
particular complication in that whilst being an autonomous
ship accommodation is also provided for survivors. How-
ever it is not possible to fully comply with SOLAS regula-
tions as for example there is no crew to man a rescue craft
or operate lifeboats. Therefore only liferafts have been fit-
ted. It could be reasonably argued that providing a refuge
for survivors that does not fully meet SOLAS regulations is
better than providing no refuge at all just because the ship
cannot fully comply with SOLAS. Ultimately the only time
anyone would be on-board would be in a serious emergency,
except a small crew that could potentially operate the vessel
from the bridge, but as this is only intended for use within
the confines of a port there is no need to comply with full
SOLAS regulations for the ship type. Either obtaining
agreed exceptions to these regulations or the production and
acceptance of an alternative set of regulations released for
unmanned ships will be a major undertaking and is likely to
need technical demonstration and operationally derived
evidence to enable it to be achieved.
To provide this meaningful evidence, experience of op-
erating autonomous/unmanned demonstrator vessels will be
needed before formal changes are agreed to regulations.
This poses its own short term challenges as to how a ship
will be allowed to go to sea whilst not complying with all
relevant national and international regulations. Classifica-
tion societies and national administrations will need to
support demonstration ships in the short term to enable
projects to be successfully realised.
Demonstration ships are likely to be either full scale
projects with a very simple role / limited capability or
moderate scale ships to keep the costs to a minimum whilst
providing a suitable test facility and an ability to demon-
strate the capability of autonomous ships.
Whilst there are already existing technologies that can
be employed to facilitate an autonomous ship there remains
a need to further develop these as well as new technologies
to allow the creation and operation of unmanned and au-
tonomous ships.
Possibly the most important technology challenge to be
solved is that of ensuring that the ship has the required situ-
ational awareness to be able to ensure correct decision
making in the wide range of weather and sea conditions
which will be experienced, day and night. The current gen-
eration of sensors and sensor fusion is sufficiently mature to
allow good decision making in good weather day and night,
however the limitations associated with poor weather and
sea conditions mean that development is required to enable
even the remote control level of autonomy. It is consid-
ered that sensor suites will need to be developed specifical-
ly for the needs of autonomous shipping. Until the situa-
tional awareness of an autonomous ship can be proven to be
equal to or better than a manned crew in all weathers, the
level of autonomy and likely the roles and areas of opera-
tion of ships with no onboard crew will have to be limited.
In addition, the reliance of such shipping on software to
achieve safe operation means that almost all the software on
board and in any control centre will need to be produced to
the highest level of software integrity, requiring extensive
testing and certification. More attention will also need to be
paid to Cyber-security, to ensure that the ship and any con-
trol centre will not be vulnerable to cyber interference,
whether malicious or accidental.
Machinery reliability and maintenance requirements
will also need to be reviewed. Current prime movers typi-
cally require significant human interaction on a frequent
basis, from changing filters, checking false alarms, tighten-
ing fastenings to general condition monitoring. In order for
autonomous ships to become truly useful, ships systems are
going to be needed that require infrequent human interac-
tion whilst remaining very reliable. Other technologies that
exist in other industries will need to be made more robust to
cope with the harsh marine environment.
Maintenance regimes of many types of equipment will
need to be changed as weekly or even monthly maintenance
by mechanics / technicians is not possible aboard autono-
mous ships. Ideally all maintenance will be carried out in a
small number of maintenance periods through the year but
more likely whenever a vessel visits a port mechanics will
board the ship to either carry out maintenance in situ or just
replace components so that those removed can be serviced
in a workshop offline thereby reducing the downtime of the
ship. For some ships and systems this will lead to a drive on
design and arrangements with a different focus to tradition-
al ships. Components that are traditionally serviced and
repaired in situ can be located in inaccessible parts of the
ship but on a ship where those components are better main-
tained ashore they need to be mounted to allow them to be
decoupled and replaced quickly.
The security of an unmanned asset is a particular chal-
lenge; both physical security and cyber security. Without a
crew the ship could be a greater target for pirates believing
it to be an easier vessel to capture than a crewed vessel.
Whilst there is no concern about crew safety the loss of the
ship and any cargo would be a significant insurance cost if
it became a common problem. Cyber security is perhaps a
more significant challenge, with no means to revert to a
manual mode if the electronic systems are hijacked or mali-
ciously damaged. Whilst cyber security is already a focus
of concern in the shipping industry the scale of the conse-
quence for unmanned ships will demand ever more robust
systems to allow autonomous shipping to come to fruition
on a significant scale.
Apart from complying with prescriptive regulations and
ensuring security there is a wider assurance aspect that is
likely to be needed to convince the typically conservative
marine community that autonomous ships are a viable, cost
effective, sustainable and safe alternative to manned ships.
In this example having an autonomous ship operating in oil
fields and in close proximity to high value and potentially
hazardous assets would require significant justification that
it was going to remain under control in all conditions.
Equally the role of the vessel is critical and assurance will
be needed that when it is needed to tackle a fire it will reli-
ably perform the task to at least the same standard as an
existing crewed ship. For autonomous shipping to become
accepted it will also be necessary to ensure that autonomous
ships are not going to pose additional operational hazards to
manned ships either through collisions or interfering with
navigation in congested waters.
In order to establish autonomous ships in the market
there will undoubtedly have to be a progressive approach to
removing the human from the control system. In a similar
way to how driverless cars have developed, rather than go-
ing from a crewed ship to an unmanned autonomous ship it
is likely there will be a progressive path with steps that
could include:
I. on-board oversight by a crew so that in the event
that something doesnt function as it should the
crew can take over;
II. remotely controlled; in that there would be a full
time watch but based onshore controlling the
ship with the use of the on-board sensors to pro-
vide the required information to make decisions
and send instructions to the ship;
III. remote monitoring where a small number of
people have oversight of a number of vessels
keeping a periodic check on each vessel to en-
sure that it is functioning as it should be.
Ultimately autonomous ships should not need real-time
monitoring but will self-diagnose problems and send an
alert to a shore based facility to make a decision on what
223
action to take, whether that is to change destination, termi-
nate the current task, return for maintenance period at an
earlier time, continue as planned etc.
QinetiQ agree with the many commentators who con-
sider that autonomous shipping will become a reality, and
that significant collaboration and support is needed from
the entire maritime community to ensure a practical and
safe realisation.

224
 Designing for Damage Stability beyond Design Level
Dracos Vassalos1, Evangelos Boulougouris2, Donald Paterson3and Markku Kanerva4
1
University of Strathclyde, d.vassalos@strath.ac.uk
2
University of Strathclyde, evangelos.boulougouris@strath.ac.uk
3
University of Strathclyde, donald.paterson@strath.ac.uk
4
Consultant, markkukkanerva@gmail.com
ABSTRACT
This paper describes the background and provides the
rationale and the framework to embrace all feasible
measures (passive/design and active/operational nor-
mal and emergencies) for improving the damage sur-
vivability of RoRo Passenger ships. The ideas elaborat-
ed in the paper is an attempt to elucidate and assess the
impact on options for new and existing ships of increas-
ing the required subdivision index R, the former in re-
sponse to the higher damage stability standards recom-
mended following the conclusion of the EMSA III pro-
ject and the latter in case IMO decided to apply higher
damage stability requirements retrospectively, particu-
larly in the aftermath of an accident. Such a framework
would provide the motivation for instigating and estab-
lishing novel damage stability enhancing paradigms in
line with IMO Circular 1455 on equivalents, for alter-
native compliance. This, in turn, would enable the in-
dustry to focus on all credible measures for damage
stability enhancement in case of a flooding accident.
This represents a step change both in the mind-set of
naval architects and in safety legislation but the impact
will be immense and mostly positive. This paper paves
the way in this direction by providing the background
and rationale for such a framework and by introducing
an alternative system for damage stability enhancement
that involves injecting highly expandable foam in the
compartment(s) undergoing flooding during the intimal
post-accident flooding phase thus enhancing damage
stability and survivability of RoPax vessels well beyond
the design levels in the most cost-effective way currently
available. This is a mind-set changing innovation that is
likely to revolutionise design and operation of most ship
types and RoPax, in particular. A number of applica-
tions are considered in the paper for a range of ship
sizes with impressive results that will challenge the cur-
rent established practice.
INTRODUCTION
Back in 1912 when RMS Titanic was sailing from South-
ampton, UK to New York, USA, the airplane had just been
invented 9 years ago by the Wright brothers and ships were
still using coal. The tragic loss of 1,513 people onboard a
state-of-the-art for its time ship was a shock for the society
and the maritime industry. This led to the adoption of the
first International Convention for the Safety of Life at Sea
(SOLAS) of 1914, which over the last century is constantly
improved and enhanced in its latest 1974 form. As a result,
the shipping losses have been reduced from 1 ship per 100
per year back in 1912 to about 1 ship per 672 per year in
2014 (Allianz, 2015). Still, the societal outcry that follows
every accident, especially in the case of passenger ships is
tremendous with more recent examples the loss of Costa
Concordia and Sewol. Tragedies such as these remind to
both industry and academia that we have to do more in
order to reduce the vulnerability of our ships in case of
flooding. Unfortunately, our arsenal is still an enhanced
version of the one that the naval architects had back in the
dawn of the 20th century. Furthermore, due to the grandfa-
ther clause, major changes to SOLAS are applied mainly
to newbuildings, which represent obviously a small fraction
of the existing fleet. Thus, the state-of-the-art knowledge on
damage stability inoculates very slowly the fleet at risk,
leaving most of the ships with severe vulnerabilities and
their passengers and crew unnecessarily to higher risk. This
becomes even more woeful considering the continuously
accelerating pace of todays scientific and technological
developments. As a result, our regulatory framework is
becoming progressively less relevant and unable to keep up
with this pace of development.
However, the introduction of the probabilistic damage sta-
bility concept in its latest SOLAS 2009 (IMO
MSC.Res.216(82), 2006) form provides an objective, au-
ditable way to measure the current risk and the improve-
ments made by any proposed risk reduction measures. Fur-
thermore, the realisation in IMO that the prescriptive regu-
lations prohibit the introduction of innovations in the design
and the adoption of the Guidelines for the approval of al-
ternatives and equivalents (IMO MSC.1/Circ. 1455) pro-
vide the tools for a step improvement in the way ships are
225
designed with regards to their damage stability characteris- rence of each damage case, the total risk of losing the ves-
tics. sel in case of damage can be calculated:

This paper presents an alternative system for damage stabil- Total Risk =( ) (2)
ity enhancement that involves injecting highly expandable
foam in the compartment(s) undergoing flooding during the Similarly, the local risk associated with a particular damage
initial post-accident flooding phase thus enhancing damage case i can be calculated as:
stability and survivability of ships and especially RoPax
vessels well beyond the current design levels in the most Local Risk = = (1 ) (3)
cost-effective way possible.
Vassalos (2012) underlines that the ship, as a system with
BACKGROUND multiple operational modes and conditions, has some de-
sign (nominal) characteristics and a number of operational
Since the introduction of the probabilistic model for the ones which may modify its survivability at any given time
assessment of the survivability of ships and especially dur- (e.g. loading condition, open watertight doors, addition of
ing the last 15 years there have been multiple attempts from cross flooding valves etc.). In that respect we should dis-
the industry and the academia to optimise the design of tinguish the design vulnerability of the ship to the opera-
RoPax ships with multi-criteria design optimisation, using tional one. An excellent example is the vulnerability distri-
survivability after damage as an objective (Boulougouris, bution of MV Estonia shown at her design condition (Fig.
2004). In R&D projects such as ROROPROB (2000-2003), 1) and at the time of her accident (Fig. 2).
GOALDS (2009-2012) and more recently in the EMSA 3
study (EMSA, 2015), the designers attempted to maximise
the attained subdivision index for RoPax vessels of vari-
ous sizes. In most of these cases, the designers had to in-
crease the breadth or the freeboard of the vessel by few
centimetres, add bulkheads under the main deck, subdivide
the car deck or add cross flooding devices in order to
achieve tangible improvements in the attained subdivision
index (EMSA, 2015). The problem is that many of these
design changes proved to be not cost effective risk control
options (RCOs) due to their high cost or their low risk re-
duction, surpassing the Net Cost of Averting a Fatality
(NCAF) of 8mil USD (5.9mil) which is widely accepted
nowadays in the maritime industry (EMSA, 2015). These
solutions have a significant impact on the CAPEX, OPEX,
Fig. 1 MV Estonia - Design Vulnerability distribution
FUELEX, loss of revenues, but they may also have impact
(Vassalos, 2012)
to the costs related to air emissions, upstream processes,
climate change, harbour fees, salvage and loss of cargo. It is
therefore obvious that the designers need new tools in
order to optimise their designs in a cost-effective way.
Damage Stability Recovery System (DSRS) is such a solu-
tion.

VULNERABILITY

"Vulnerability" as concept is used extensively in the sur-

vivability assessment of naval ships (Boulougouris et.al.,
2016) but in merchant shipping is used as a term in the 2nd
Generation of Intact Stability Criteria without been directly
associated to probability. Hence, a definition here is in
order. The way this term has been used for merchant ships
at Strathclyde relates to "the probability that a ship may Fig. 2 MV Estonia - As Operated at the Time of her Loss
capsize or sink within a certain time when subjected to a (Vassalos, 2012)
feasible extreme flooding case" (Vassalos, 2014). In the
probabilistic framework, it could be defined as: Reducing the vulnerability beyond the levels achievable by
adding another bulkhead, closing the watertight doors or
vi = 1 si (1) using cross flooding, requires a system that can provide
some of lost buoyancy. DSRS is that system.
where si is the probability of survival of the vessel in a
particular damage case i. Therefore, vulnerability contains DAMAGE STABILITY RECOVERY SYSTEM (DSRS)
(and provides) information on every parameter that affects
damage ship survivability. As a consequence, vulnerability DSRS is a bespoke system releasing highly expandable
is directly linked to risk. Using the probabilities of occur- foam in the flooding compartment(s) during the initial

226
post-accident flooding phase. It has been developed (patent
pending) by the University of Strathclyde with the support
from Scottish Enterprise. It can be fitted to new or retrofit-
ted to existing ships in order to reduce the likelihood of
capsize/sinking and progressive flooding following a major
flooding accident.
The working principle of the proposed system is simple:
when a vessel is subjected to a critical damage, stability is
recovered through the reduction of floodable volume within
the vessels high risk compartment(s). This is achieved by
rapidly distributing fast setting, high expansion foam to the
protected compartment(s), regaining lost buoyancy whilst
also eliminating free surface effects and forming a near
watertight seal over unprotected openings. Moreover, with
water being constrained low in the ship, it actually increas-
es damage stability (higher GM).
The system itself consists of a fixed supply of both foam
resin and hardener agents; each stored within an individual
tank and connected to a piping network for distribution.
Different foam types have been considered and can be used
depending on the specification of the system and the re-
quirements of the owner. For generic version of the system
(see Fig. 3 and 4) the operation starts when two distribution
pumps supply a flow of filtered sea water into individual
resin and hardener lines. Both streams are then dosed with
concentrated resin and hardener agents, before they each
pass through a static mixer in order to produce a homoge-
neous solution of each component.
Fig. 3 System Representation
The two lines are then fed to the protected compartment
where they meet and enter a foam generator. Here both
streams mix and compressed air is introduced into the sys-
tem for the in situ production of foam. The foam is then
passed in to a branched piping network within the vulnera-
ble compartment where both port and starboard side
branches allow the foam distribution to the flooded room.
The whole process is monitored and controlled by a central
system linked to vital components and sensors. The system
can cover one or multiple compartments and may have
sufficient capacity for one or more compartments. The use
of the system is under the full control of the crew, with a
decision support system (DSS) available to help the ships
master decide where and when the system will act as well
as inform all concerned of the ensuing actions.
The foam compound, the resin and the hardener meet all the
environmental and health criteria, they is not harmful to
humans and the foams release does not pose any danger to
the people onboard or the environment. Furthermore the
foam is non-flammable and in this respect could reduce risk
by other event sequences such as a fire ignited in collision.
The residual clean-up post system discharge is also aided
by a foam dissolving agent ensuring minimal business in-
terruption.
METHODOLOGY
For the purposes of this study two ROPAX vessels, cur-
rently operating in European waters, have been investigated
with a view to assess the effectiveness of the proposed
Damaged Stability Recovery System (DSRS) as a risk re-
duction technology. The probabilistic approach to damage
stability (SOLAS 2009) has been used as a means of estab-
lishing the initial level of risk associated with the designs.
The effects of the DSRS have then been modelled and the
vessels were re-examined in order to assess the risk reduc-
tion achieved by the system.
DSRS IMPLEMENTATION & MODELLING
In order to ascertain the impact of the proposed system on
vessel safety, the designer has to identify the overall risk
level associated with the vessel. From Eq. 2 we can easily
associate the attained index A with the overall risk as:
= 1 (4)
This provides a benchmark to gauge any improvements on
the vessels safety achieved by the use of the DSRS.
In order to ensure the system is designed in the most effi-
cient manner, the system must target those compartments
which represent the Achilles heel for the vessel i.e. con-
stitute the greatest risk. As such, a risk profile of the vessel
is created in order to aid in the identification of design vul-
nerabilities.
Fig 4 Example Risk Profile
227
The results from the probabilistic damage stability assess- Following the analysis of a significant number of existing
ment provide a straightforward way of determining the designs it has been proven that in the large majority, using
vessels risk profile by firstly considering the local risk the foam in one and two compartment would be sufficient
associate with each damage scenario, as calculated by Eq. 3. for reducing substantially the risk of capsizing or founder-
These local risk values are then mapped across the vessel ing in case of damage. Two of these studies will be pre-
according to damage centre, in order to form the risk profile sented here below.
as shown in Fig. 4.
CASE STUDY: LARGE ROPAX
In the above risk profile, risk is plotted on the vertical axis
and the damage position along the horizontal. Differing Overview
lengths of damage, as measured by multiples of adjacent
zones, are distinguished by marker type and colour. This Starting from the most demanding in terms of foam volume
enables the identification of both safety critical design spots requirements case, the DSRS team studied a large ROPAX
and opportunities where safety could be improved most with a central cased ro-ro deck suitable for drive-through
significantly and efficiently. Two cases in particular, circled operations, with a large lower hold spanning eight com-
in Fig. 4, are identified as large risk contributors. As such, it partments under the main deck (see Fig. 6). The vessel is
can be reasoned that the DSRS would be best applied in the equipped with a hoistable car deck suitable for additional
protection of one if not both of the compartments which car storage.
give rise to this risk. Following this methodology for the
sample vessel, the system could be applied in the most The vessel was built in 1998 to a two-compartment subdi-
efficient and effective manner. vision standard according to SOLAS 90 along with Stock-
holm agreement compliance with a significant wave height
For the analysis, the vessel should initially be modelled of 2.9m. Below the bulkhead deck the vessel is divided into
accurately according to its nominal (design) conditions. a total of twenty water tight compartments and has pro-
Data such as lines and general arrangement plans are re- nounced B/5 subdivision spanning almost the entire length
quired. The relevant stability documentation is used in of the vessel and cross flooding ducts fitted to enable sym-
order to ensure that all unprotected and weather tight open- metrical flooding. The vessels principal particulars and
ings are taken into account. Loading condition information general arrangement are provided in Table 1 and Fig. 6.
within the vessels stability booklet is used in conjunction
with the damage stability metacentric height (GM) limiting Table 1. Principal Particulars
curves in order to select the SOLAS 2009 initial loading
conditions. Length o.a (m) 200.65
Length b.p (m) 185.4
The required volume for the DSRS system is calculated
through trade-off analysis of protected compartments per- Breadth (m) 25.8
meability. The required volume of foam is then estimated Draught MLD. (m) 6.8
based on the minimum volume required to save the ship in
Displacement (t) 19468
the most demanding high risk damage scenario. The opti-
mum volume can be estimated based on an Cost-Benefit Deadweight (t) 5830
Analysis (CBA) taking into account the cost of the system, Crew Number 200 persons
the additional weight, the loss of carrying capacity (if any),
Passenger Number 1500 persons
and the achieved reduction of the risk as a function of foam
volume, up to the threshold of the NCAF. Benefits from e.g.
changing the payload distribution (e.g. more passengers) or
increase in the earning potential of the ship due to im-
provements to the hotel arrangements should also be in-
cluded. An example of the trade-off between foam volume
and risk is shown in Fig. 5.

Fig. 6 General Arrangement

Fig. 5. Trade-off analysis between foam volume and Risk

228
DSRS impact assessment was then re-assessed following a permeability change to the
lower hold to account for the effects of the foam.
In order to assess the damage stability performance of the
vessel a total of 942 damage cases have been analysed The new attained index values calculated in this case can be
under three loading conditions as outlined in Table 2. found in Table 4 along with the updated risk profile of the
vessel highlighted in Fig. 8.
Table 2. Loading Conditions
Displacement (t) Draft (m) GM (m) Table 4. Attained Index after DSRS usage in 1-comp
Al 0.96
LC1 (dl) 19468 6.8 2.226
Ap 0.85
LC2 (dp) 17412 6.4 2.003
As 0.84
LC3 (ds) 15087 5.733 3.191
New Attained Index A 0.87
The results of the SOLAS 2009 damage stability assess-
ment along with the required index value calculated for this
vessel can be found in Table 3 below. The risk profile de-
rived for the vessel is also provided in Fig. 7. It is obvious
that the SOLAS 90 + WOD requirements provide the ves-
sels with sufficient survivability in order to fulfil easily the
SOLAS 2009 requirements.

Table 3. SOLAS 2009 Results

As 0.79
Ap 0.80
Al 0.96
Attained index A 0.83
Required index R 0.795
Fig. 8. Updated Risk Profile

It is clear from the newly calculated results that the effects

Fig. 7. Risk Profile
Still, a closer inspection of the vessels risk profile reveals
several vulnerabilities existing within the vessels design.
This risk is founded primarily by damages that penetrate
beyond the B/5 longitudinal bulkhead of the lower hold.
Damages involving this space were not covered by the
regulations in place at the time although they do however
present a significant threat to the vessels safety.
Damage to the lower hold gives rise to large scale flooding
leading to a significant reduction in the vessels residual
stability. Having been identified as the largest risk contrib-
utor this space was selected for the application of the sys-
tem.
The volume of foam required in this case was defined as
that required to mitigate the risk stemming from two com-
partment damages involving the lower hold, equating 2000
m3 expanded volume. The damage stability performance
of the system have resulted in a substantial reduction of risk.
This is evident in the eradication of the risk contribution
made by one and two compartment damages involving the
vessels lower hold. The risk stemming from three com-
partment damages to this space has also been mitigated,
particularly in those damages located closer to amidships.
However, there still exists a series of high risk three com-
partment damages towards the fore of the lower hold and
mitigation of these risks would call for a larger volume of
foam to be utilised. In total the system has resulted in a
24% risk reduction from 0.17 to 0.13 for a 1-compartment
DSRS application.
Selection of the second compartment for system protection
involved re-evaluation of the vessels risk profile. Through
doing so, the vessels main engine room was identified as
the largest of the remaining risk contributors. This particu-
lar space has a large volume coupled with a high permea-
bility value leading to large scale flooding when damaged
and serious diminishment of the vessels residual stability.
As the one compartment system application required an
already large volume of foam the decision was made to use
a constant volume of available foam in the investigation of
two compartment protection. As such, the volume of foam
was shared between the two protected compartments in
such cases that they were simultaneously damaged. When
either of the protected compartments was damaged inde-
pendently the entire volume of foam was assumed to be
used for the damaged compartment in question.

229
The damage stability results following this process are Table 11. Small ROPAX Principal Particulars
provided in Table 5 and the vessels updated risk profile is Length Overall 89.48 m
provided in Fig. 9. Length Btwn Perpendiculars 81.80 m
Table 5. Attained Index after DSRS usage in 2-comp Breadth 16.40 m
Freeboard Draught 3.40 m
Al 0.97
Displacement 3434.6 t
Ap 0.86
Deadweight 749.6 t
As 0.85 Crew Number 34 persons
New Attained Index A 0.88 Passenger Number 550 persons

Fig. 9. Updated Risk Profile

The results in this case show that the protection of two

compartments has worked to mitigate the risk stemming
from damages to the main engine room but failed to eradi-
cate these risks. In total, there has been a relative 5% addi-
tional risk reduction afforded by this further protection. In
order to generate a more meaningful reduction in risk, ei-
ther a larger volume of foam would be required or the range
of compartments served by the system would have to be
increased. The system was however able to produce an
overall risk reduction of almost 30% from 0.17 to 0.12.
Fig. 10. Small ROPAX GA
DSRS impact assessment
For the assessment of the vessels damage stability perfor-
mance a total 533 damage cases were considered at three
loading conditions. The results of the risk profile generated
for the vessel in Fig. 11.

CASE STUDY: SMALL ROPAX The results from the damage stability assessment show a
large disparity between the attained index value calculated
Overview and the required index value for the vessel (see Table 12).
In this case the vessels is GM is limited dominantly across
Investigating the effectiveness of the DSRS on the other the three loading conditions by the requirements of the
end of the scale, a small ROPAX vessel was studied. It has 2008 IS code and as such the attained index value is much
side casings that run its length and the aft portion is open to higher than that required by SOLAS 2009 regulation 7.
allow the transportation of hazardous cargo. The vessel can
accommodate a maximum of 550 passengers and is operat-
ed by a total of 30 crew members. The vessel was launched
in 2010 fulfilling the probabilistic SOLAS 2009 standard
along with the water on deck deterministic requirements
mandated by the EU passenger ship directive 2003/25/EC
(EC, 2003). The vessel is divided into three main vertical
fire zones and subdivided into 12 watertight compartments
below the bulkhead deck. Lifesaving appliances are pro-
vided for all 584 persons on board for domestic voyage, as
a Class B vessel according the EU passenger ship directive
2009/45/EC. The vessel is not equipped with life boats. The
principal particulars of the vessel are provided in Table 11
below along with the GA in Fig. 10. Fig. 11. Small ROPAX original risk profile

Still, vulnerabilities are identifiable within the vessels

design. Observation of the vessels risk profile reveals that
the stabiliser compartment and damages involving this

230
space give rise to the large risk contributions. This particu-
lar space has a large volume coupled with a high permea-
bility value leading to large scale flooding when damaged
and a serious diminishment of the vessels residual stability.
As such, it was decided that this space should be investi-
gated for the 1-compartment application of the system.
With a goal of eradicating the risk associated with 2 com-
partment damages to this space, the volume of foam af-
forded to the system was set at 365 m3. The damage stabil-
ity assessment was then re-conducted producing the results
given in Table 12 and the updated risk profile in Fig. 12.
Table 12. Attained Index before and after DSRS
Original DSRS 1-comp
Al 0.97 0.98
Ap 0.9 0.926
As 0.9 0.925
New Attained Index A 0.91 0.94
Required Index R 0.73
Fig. 12. Small ROPAX new risk profile with DSRS 1-comp
The system has again proven here to be greatly effective in
increasing the vessels safety level. Observation of the
vessels updated risk profile shows that the risk stemming
from two compartment damages to the stabiliser compart-
ment has been eradicated. Furthermore, the risk associated
with three compartment damages involving this space has
been considerably reduced. In total the DSRS has achieved
a risk reduction from 0.09 to 0.06 or 33%.
In case a second compartment is protected by the system
with additional foam, the updated risk profile suggests that
this should be the shaft alternator room and the fin stabiliser
compartment. Assuming that the foam could be delivered to
either of the given compartments if damaged, a new 2
DSRS compartment protection risk profile has been pre-
pared and the new total risk has been estimated to 0.047 or
48% reduction from its original value.
CONCLUSIONS
The challenge faced by the maritime industry in the 21st
century is to reduce drastically the loss of life in maritime
transportation. The approach used in the 20th century has
reached its plateau and a step change is required. The au-
thors argue that DSRS is the solution to this problem. This
has been proven by the results presented herein. By com-
bining expertise in ship damage stability and specialist
knowledge in expanding foams, a non-intrusive
cost-effective solution to the damage stability problem of
ROPAX vessels has been identified that does not interfere
with the existing characteristics of the vessel, its functional-
ity or business model, enabling the vessel to remain com-
petitive while being above all safer. The system can be
easily installed in new and existing ships and its technology
is proven and reliable.
REFERENCES
Allianz Global Corporate & Specialty SE (2015), Safety
and Shipping Review 2015
https://www.allianz.com/v_1427190309000/media/press/do
cument/other/Shipping-Review-2015.pdf (accessed 7 July
2016).
Boulougouris, E., Winnie, S. and Papanikolaou, A. (2016),
Assessment of survivability of surface combatants after
damage in the sea environment, J. Ship Production and
Design 32/3, pp. 110.
Boulougouris, E.K., Papanikolaou, A. and Zaraphonitis G.,
(2004), Optimisation of arrangements of Ro-Ro passenger
ships with genetic algorithms, Ship Technology Research
51/3, pp.99-105.
EC (2003), DIRECTIVE 2003/25/EC of the European
Parliament and of the Council on 14 April 2003 on specific
stability requirements for ro-ro passenger ships, Official J.
of the European Union
EMSA (2015), EMSA/OP/10/2013 (EMSA 3), Impact
assessment compilation part 1; Impact assessment in ac-
cordance with the EC IA, Report No. 2015-1194, Rev. 3.
GOALDS (2009-2012). Goal-based Damage Stability,
Project funded by the European Commission, FP7-DG
Research, 233876.
IMO MSC.1/Circ. 1455 (2013), Guidelines for the ap-
proval of alternatives and equivalents as provided for in
various IMO instruments, 24 June 2013.
IMO MSC.Res.216(82) (2006), Adoption of Amendments
to the International Convention for the Safety Of Life At
Sea, 1974, As Amended, IMO, London
ROROPROB (2000-2003), Probabilistic Rules-Based
Optimal Design of Ro-Ro Passenger Ships, FP5-DG Re-
search, Contract Number G3RD-CT-2000-00030.
Vassalos, D. (2012), Damage stability of passenger ships
Notions and truths, STAB 2012, Athens pp.775-789.
Vassalos, D., Boulougouris, E., Guarin, L., Jasionowski, A.
and Garner, J., (2014), Regulatory, design, operational and
emergency response measures for improving the damage
survivability of existing passenger ships, 14th Int. Ship
Stability Workshop, Kuala Lumpur, pp.292-300
231
 Data Analytics for Marine Operations
Hao Wang, Girts Strazdins
Big Data Lab, Faculty of Engineering and Natural Sciences,
Norwegian University of Science and Technology, Aalesund, Norway
Guoyuan Li
Mechatronics Lab, Faculty of Maritime Technology and Operations,
Norwegian University of Science and Technology, Aalesund, Norway

ABSTRACT pervasive computing, wired/wireless sensors, networks, and

Much more data are being collected for diagnostic
and monitoring purposes with sensors on board a vessel.
This incurs the information overload problem. We are
building a new Visual Analytics framework to integrate
and visualize the monitoring data, embedded with
advanced intelligent algorithms for effective detection
and prediction of ship behaviors. This paper reports our
recent works in collaboration with the industry.
1. INTRODUCTION
Marine operations face the major challenge of increased
complexity. Much more data are being collected for
diagnostic and monitoring purposes with sensors on board a
vessel. Due to the lack of easy data presentation and
interaction, the crewmembers are overwhelmed by the large
volume of information and alert messages, causing serious
information overload problems.
Norway is ranked as the worlds second largest nation in
marine operations. In particular, the northwestern Mre
region of Norway is one of the leading clusters in the global
Offshore Support Vessel (OSV) shipbuilding and shipping
(ship service) market, consisting of almost 200 companies
with a total turnover in excess of NOK 50 billion. As
marine operations face the major challenges of increased
complexity in more demanding waters, technologies are
being adopted for acquiring monitoring data about how the
vehicle and different components are behaving. Recently,
with the intention of remote ship monitoring for better
services for shipping customers, vessel builders started to
adopt new sensor technology by installing different sensors
for different components on board a vehicle and transmit
data using satellite communications to land-based service
centres, e.g., the HEalth MOnitoring System (HEMOS) by
Rolls-Royce Marine AS. This development has been
strengthened by the quickly increasing awareness of the
new Internet of Things (IoT) paradigm (Borgia, 2014). IoT
covers technologies including ubiquitous computing,
embedded systems, forming a communicating-actuating
network of a large amount of things. In this way, physical
environment and resources could have presence in the
digital world.
These systems provide more accurate and timely
operational data, but they also introduce new danger to the
operations: information overload problem (IOP) (Keim et
al., 2008, Wang, Zhuge, et al., 2016) -- the crew members
receive a large volume of monitoring information and alert
messages that s/he can easily overlook important/vital ones.
Therefore, it is urgently needed to develop and implement a
new framework to integrate and visualize the monitoring
data in an informative way. In this way, the crewmembers
can examine the massive, multi-dimensional, multi-source,
time-varying information streams to make effective
decisions in time-critical situations.
The aviation industry is more mature in monitoring systems
on board an aircraft. Flight Data Monitoring (FDM) is a
systematic method of accessing, analyzing and acting upon
information obtained from digital flight data records of
routine flight operations to improve safety (Larder, 2002).
It is a well-established practice in many fixed wing airlines
for airline safety management processes. Because of the
similarities between marine vehicles and aircrafts, a lot of
knowledge, know-how, and best practice can be transferred
into data integration and visualization for marine operations.
However, existing techniques must be adapted and new
methods must be developed for the specific requirements
for the setting of marine operations. Marine operation
experts play a critical role in the knowledge transfer and
adaptation.
The new paradigm Big Data Analytics (BDA) is quickly
rising. It is highlighted in a statement from the United
Nations that the world is experiencing a data revolution
(UN Global Pulse, 2012). As a key step in BDA, data
integration combines data from various sources and
provides a unified view of these data, including data fusion,
data cleansing/purification, and data validation.

232
The monitoring data on a vessel comes from many sources,
in different formats and frequencies. i.e., the data are
multifaceted (Kehrer & Hauser, 2013). For example, low
frequency data include heading, speed, and GPS; and high
frequency data include vibration and torque. In addition, a
major challenge for vessel data integration is the poor
quality of the raw sensor data. Data cleaning aims to
improve data quality using for example statistics, integrity
constraints. However, it is generally very difficult to
guarantee the accuracy of the data cleaning process without
verifying it via experts or external sources (Chu et al.,
2015). Therefore, close collaboration with domain experts
is essential for cleansing the monitoring data from vessels.

Another major challenge is the presentation and user

interaction of the monitoring data. Marine operations are
complicated and the monitoring data can be used in many
different, even unforeseeable, ways. The user-interaction
functionality in existing data management and visualizing
tools are either very limited or involved with complex
query commands, which is not applicable for marine
operations. Therefore, easy and intuitive data visualization
and interaction for both 1) on-board captains and crew
members and 2) onshore online support teams and offline
analysts are essential.
In this paper, we present an overview of the Visual
Analytics (VA) framework (Wang, Fossen, et al., 2016,
Wang, Zhuge, et al., 2016, Li, Kawan, et al., 2016) that the
Big Data Lab (BDL) in NTNU lesund is building for
marine operations.
2. VA FOR MARINE OPERATIONS
2.1 Data Analytics
The city lesund is located at the heart of the
northwestern Mre cluster of maritime industry. The SFI1
Centre for marine operations has been set up in Norwegian
University of Science and Technology in lesund in 2015
to support innovation for demanding marine operations. In
the framework of our Innovation Norway project GCE
Blue Maritime Big Data (Wang, Osen, et al., 2015), we
obtained the monitoring data from HEMOS by Rolls-Royce
Marine AS covering three years operation of one vessel,
including high frequency machinery and low frequency
vessel behavior monitoring data.
Fig. 1 Data acquisition, integration, and visualization for
marine operations
2.2 VA Framework
Keim et al. (2008) in their seminal paper, explored the
definition, process, and challenges of visual analytics. They
stated that the goal of VA is the creation of tools and
techniques to enable people to:
synthesize information and derive insight from
data;
detect the expected and discover the unexpected;
provide timely and understandable assessments;
communicate assessment effectively for action.
Shneiderman (1996) proposed a well-known information
seeking mantra: Overview first, zoom/filter, details on
demand. However, when the data are too large or
complicated, more iterations with the human analyst will be
necessary. Therefore, Keim et al. (2008) extended the
mantra to be: Analyze first, show the important,
zoom/filter, analyze further, details on demands and
introduced the seminal VA framework, depicted in Fig. 2.

The overview of data acquisition, integration and

visualization for marine operations is depicted in Fig. 1.
Monitoring data is collected from various sensors on board
a vehicle, and the data will go through the integration layer
for data fusion, cleansing/purification, and validation. The
output from the integration layer is visualized with intuitive
user interaction, including data correlations exploration, Fig. 2 Visual analytics framework [6, Fig.1]
spatial and temporal zooming, and etc.
The use of visualization can steer the analytical process
(Kerren & Schreiber, 2012): the human analyst can
interactively choose parts of data or change parameters of
automatic analysis methods, the results of which can be
further visualized and analyzed with other methods.
1 SFI is a prestigious research program in Norway to build up a Centre
for Research-based Innovation for a certain application area

233
One key element of VA, compared to the currently
dominating approach of automatic (algorithmic) analysis, is
the recognition of the importance of information
visualization in the human understanding and analyzing
process. Fekete et al. (2008) has rigorously explored the
value and benefits of information visualization. Even as the
fully-automated algorithmic methods can quickly identify
useful information and provide more accurate prediction,
they lack the ability to interact with human and deliver
effectively the knowledge. The combination of
visualization and automated algorithmic methods enables a
virtuous cycle of user interaction, parameter refinement for
algorithmic models so as to achieve rapid correction and
improvement of humans knowledge and decisions.
Visual analytics is still relatively new for the maritime
community. Riveiros Ph.D. thesis (2011) on the detection
of anomalous vessel behavior in traffic and some
following-up research on maritime traffic are the closest
work we can find. The subject area, requirements, and
sources of data of their work are different from our project
and visual analytics framework, but we did get plenty of
information and inspirations from them.
2.3 An Interactive Visualization Prototype
The architecture and the screenshot of our interactive
visualization prototype (Wang, Zhuge, et al., 2016) are
depicted in Fig. 3 and Fig. 4 respectively. D3.js is an
open-sourced JavaScript library. The library provides a
large collection of data manipulation and visualization
components, allowing developers to bind arbitrary data to a
Document Object Model (DOM), without being tied to a
proprietary framework.
Fig. 3 Architecture of the interactive visualization prototype
Fig. 4 Screenshot of the interactive visualization prototype
For the multivariate facet of the data, we have used the
force-directed graph and the chord diagram to visualize the
correlations between data variables. In the force-directed
234
graph, a physics engine is embedded to enable dynamic
interaction. In the graph, 1) each node represents a different
variable; 2) the thickness of the link between two nodes
represents the correlation coefficient between the two
variables; 3) the user can select and drag any node, if a
node is more correlated with more nodes, then this variable
has a stronger effect or force on all other variables; 4)
when one node is selected, the links to all variables that are
correlated to it are highlighted. The chord diagram is
another graph visualizing the quantified correlations. When
the user selects one variable, all variables that are correlated
to it are highlighted.
We implemented an interactive line chart. With WebGL
and Three.js, we implemented the 3D vessel motion
animation, in which we embedded an interactive 2D
trajectory visualization using the Google Maps JavaScript
API. It is important to note that these visualization
components are coordinated multiple views of the same
data source. The user can use the force-directed graph or
the chord diagram to explore the variables and their
correlations; the user can select one or several variables to
be shown in the line chart. Once the user selects a time
period in the line chart and starts the animation, the changes
of variables and the motion of the vessel will be animated
synchronously.
The prototype allows easy data exploration w.r.t.
spatiotemporal features, data correlations, and etc. We have
conducted both qualitative and quantitative evaluation in a
small scale on the prototype (Zhuge, 2016). The evaluation
showed that the prototype enables users to gain insights
about the behavior and status of the vessel. The prototype
has received positive feedback from our industrial partners
and is undergoing further improvements.
2.4 Adversary Event Detection
The detection of adversary events is of paramount
importance during marine operations, and effective use of
the monitoring data can greatly relieve the burden on the
crewmembers in this respect.
Propeller ventilation (Smogeli, 2006) occurs when air from
the free surface or exhaust gas being drawn into the rotating
propeller blades, which will incur large loss of propeller
thrust and possibly dynamic loads, noise and vibration,
causing significant tear and wear of the propulsion unit. It is
well accepted that propeller ventilation is one of the main
reasons for the widespread mechanical failures of thrusters,
with costly repairs and increased vessel down-time. Most of
existing ventilation detection methods have been designed
based on the dynamics/control modeling of water and the
propeller, and they generally require information or
accurate estimates of vessel specific parameters.
The main challenges for such methods are 1) the dynamics
models have become increasing complicated, especially in
more demanding waters; 2) the vessel specific parameters
are often not available in practice.
We aim to address these challenges using our VA
framework, combining automated analysis algorithms with
interactive data visualization, aided by the knowledge of
dynamics/control models. As an initial result, we have
developed detections methods (Wang, Fossen, et al., 2016)
that require no a priori information about the subject vessel,
so we can efficiently monitor a larger number of vessels.
2.5 Ship Motion Prediction
For ship maneuvering, health, safety, environment,
security and cost will be given high priority during
maritime operations. Ship data analysis and modeling are
therefore essential for ship maneuvering, especially for the
emergence of new demands in offshore operations.
With our VA framework, we aim to combine data mining
with modeling methods to design and implement ship
motion prediction for different complicated maneuvering
cases. We have implemented a flexible neural network
(NN) based modeling mechanism (Li, Kawan, et al., 2016),
in which the user can design the model based on the
correlation analysis. By training and testing the neural
network, we obtained the ship motion predictive model.
Through a case study of fine maneuvering, the scheme is
verified for analyzing and modeling sensor data for ship
motion prediction.
In order to generate predictive models, we utilize a NN with
three layers: the input layer, the hidden layer and the output
layer. The concept used for modeling is primarily focused
on giving maximum flexibility to users. To this end, the
three layers are designed to be dynamic and changeable.
The output parameters are selected according to the specific
application. In addition, users can decide the number of
hidden nodes, as well as determining the input parameters
with the help of correlation analysis. The learning rate and
the type of activation function are also optional for
modeling. As a result of off-line training, the predictive
model is established. To verify the model, users can further
design a testing set by applying the same query to another
time series and test the constructed NN. Fig. 5 illustrates
the modeling procedure through user interaction.
Fig. 5 Neural network modeling with user interaction
3. CONCLUSIONS
In this paper, we report the recent progress in our VA
framework for marine operations. Visual analytics
combines visualization and automated algorithmic methods
to achieve rapid correction and improvement of humans
knowledge and decisions. Based on the real-world needs
and input from the industry, we have studied and
implemented prototypes for data visualization, event
detection, and ship behavior prediction. These works have
received positive feedback from our industrial partners.
Further improvement and integration have been planned
and under development.
ACKNOWLEDGEMENTS
A part of the present study is supported by the project
GCE BLUE Maritime Big Data funded by Innovation
Norway and the project AIS and Maritime Transport (No.
4506002) funded by MarKom 2020. The authors would like
to thank Bikram Kawan, Xu Zhuge, Sindre Fossen, Rune
Valle and other members from BDL; Houxiang Zhang of
Mechatronics Lab; Rune Garen, Leif R. Sols, Thomas
Oksavik, Krzysztof Swider, and Are Folkestad from Rolls-
Royce Marine AS; Per Erik Dalen from lesund
Knowledge Park AS and GCE BLUE Maritime.
REFERENCES
Borgia, E. (2014) The Internet of Things vision : Key
features , applications and open issues, Computer
Communications, vol. 54, pp. 131.
Chu, X., Morcos, J., Ilyas, I. F., Ouzzani, M., Papotti, P.,
Tang, N., and Ye, Y. (2015) KATARA: Reliable data
cleaning with knowledge bases and crowdsourcing, Proc.
VLDB Endow., vol. 8, no. 12, pp. 19521955.
Fekete, J.-D., Van Wijk, J. J., Stasko, J. T., and North, C.
(2008) The value of information visualization, in
Information Visualization, LNCS, vol. 4950. Springer, pp.
154175.
Keim, D. A., Andrienko, G., Fekete, J.-D., Grg, C.,
Kohlhammer, J., and Melanon, G. (2008), Visual
analytics: Definition, Process, and Challenges, in
Information Visualization, LNCS, vol. 4950. Springer,
pp.154175.
Kehrer, J. and Hauser, H. (2013) Visualization and visual
analysis of multifaceted scientific data: A survey,
Visualization and Computer Graphics, IEEE Transactions
on, vol. 19, no. 3, pp. 495513.
Kerren, A. and Schreiber, F. (2012) Toward the role of
interaction in visual analytics, in Proc. Winter
Simulation Conf, p. 420.
Larder, B.D. (2002) Final report on the helicopter
operations monitoring programme (HOMP) trial, Civil
Aviation Authority paper 2002/02, UK.
Li, G., Kawan, B., Wang, H., Styve, A., Osen, O. L., and
Zhang, H. (2016) Analysis and Modelling of Sensor
Data for Ship Motion Prediction, in MTS/IEEE Oceans
Conference.
235
Riveiro, M. J. (2011) Visual analytics for maritime
anomaly detection, Ph.D. dissertation, rebro Univ.

Shneiderman, B. (1996) The eyes have it: A task by data

type taxonomy for information visualizations, in Visual
Languages, Proc. IEEE Symp., pp. 336343.

Smogeli, . N. (2006) Control of marine propellers: from

normal to extreme conditions, Ph.D. dissertation,
Norwegian University of Science and Technology.

Sun, G.-D., Wu, Y.-C., Liang, R.-H., and Liu, S.-X. (2013)
A survey of visual analytics techniques and applications:
State-of-the-art research and future challenges, J.
Computer Science and Technology 28/5, pp. 852867.

UN Global Pulse (2012), Big Data for Development:

Challenges & Opportunities, United Nations.

Wang, H., Fossen, S., Han, F., and Hameed, I.A. (2016)
Data-driven Identification and Analysis of Propeller
Ventilation, in MTS/IEEE Oceans Conf.

Wang, H., Osen, O., Li, G., Li, W., Dai, H.-N., and Zeng,
W. (2015) Big data and industrial internet of things for
the maritime industry in northwestern Norway, in
TENCON 2015: IEEE Region 10 Conf.

Wang, H., Zhuge, X., Strazdins, G., Wei, Z., Li, G., and
Zhang, H. (2016) Data Integration and Visualisation for
Demanding Marine Operations, in MTS/IEEE Oceans
Conf.

Zhuge, X. (2016) A visual analytics framework of marine

operations, Master thesis, Norwegian University of
Science & Technology.

236
 A New Concept for Avoiding Collision of Automatically
Operating Ships and its Evaluation
Yoshiho Ikeda1), Kana Yoshida2), Nobuyuki Shimizu3), Keiichi Hirayama4) and
Masakazu Arima1)
1) Osaka Prefecture University, Yoshiho Ikeda (tyi27568@Osakafu-u.ac.jp)
2) Kawasaki Heavy Industry, 3) Imabari Shipbuilding Co., Ltd., 4) Japan Radio Co. Ltd.
ABSTRACT
For realization of an automatically operating ship,
many barriers of, not only technical ones but also legal,
customary and social ones, should be broken. To re-
move the worry of navigators against an unmanned ship,
the ship should safely sail giving no fear to the naviga-
tors of other ships around the ship. A simpler collision
avoidance navigation method than sophisticated one
commonly used by sailors may preferable for the auto-
matically operating ships. The present paper proposes a
simple concept to avoid collision. The ship basically
keeps course, and avoid all crossing ships by changing
speed. If the avoidance of collision is difficult in 10%
change of speed during six nautical miles, the ship
changes course by the offset method proposed in the
paper. Using the proposed collision avoidance method,
the automatically operating ship can sail without giving
any fear to other ships.
INTRODUCTION
In Japanese domestic shipping, decrease of high crew cost and
safety of operation have been in the most important issues. About
30 years ago, a research and development project of autonomous
ships was carried out in Japan and many fruitful results were
obtained (Oshima 1989). However, undeveloped sensor technolo-
gy and IT technology at that time could not realize a practical
autonomous ship.
Recently, developments of navigation apparatuses like GPS,
AIS, ECDIS etc., and of high-speed information tools between a
ship and an office in land (Shoji, R. 2013) may able to realize an
automatically operating ship.
In such situations, some concept designs of autonomous ships
were developed by DNVGL (Tvete, H., 2015) and Rolls-Royce
(Ackerman 2014).
The Institute for Developing Advance Ships, Osaka Prefecture
University, started a joint project for developing an automatically
operating ship with JRC and Imabari shipbuilding. The target ship
is not an autonomous ship which can independently sail, but a
semi-automatic sailing ship under control and support of the land
operation center. The ship has ability to safely sail without colli-
sions, grounding or capsizing due to rough weathers. As the first
report of the research project, a concept for avoiding collisions
with other ships is proposed and evaluated.
BASIC CONCEPT FOR COLLISION AVOIDANCE
PROPOSAL OF COLLISION AVOIDANCE METHOD
A basic concept of the collision avoidance method is set as No
Fear for Navigators of Other Ships. This means that the developed
ship automatically and gently escapes from collisions with other
ships. The authors consider only such unmanned ships can be
accepted in the present real shipping world.
Collisions of ships often occur due to human errors under brave
mind of the navigators. The navigators of the ship which has a
right to sail in her course may be in strong stress to understand the
way of thinking and decisions of the navigator of the oncoming
ship and forced to make the final decision in order to avoid the
collisions even if his own-ship has the right to sail straight under
the international law. These circumstance may occur because the
collision avoiding navigation commonly used are complex be-
cause it is composed by changing the course and changing the
speed of the non-right ship.
Therefore, in the present study, a simpler way is adopted as
follows. The developed ship lets an oncoming ship, which can be
collided in some minutes, go in her original course by changing
speed of the own ship. This means that both ships keep their
courses, and the developed ship automatically changes her speed
to avoid the collision. In Fig. 1, a conventional way of collision
avoidance by using veering and the proposed way by keeping their
courses and adjusting the own speed for collision avoidance navi-
gation are shown in the left and right figures, respectively.
veering adjusting speed
Fig. 1 Methods for avoiding collision.
237
SIMULATION OF COLLISION AVOIDANCE
CONDITIONS OF SIMULATION
Definitions of variables used in the simulation are shown in
Fig.2. The developed ship, or own ship will meet the oncoming
ship at the predicted collision point, which is located at L miles
ahead of the own ship. The length and speed of the own ship are
defined as L1 and V1, and those of the oncoming ship are L2 and V2,
respectively. Encounter angle of the two ships is defined as .
Collision of the two ships is identified by contact of the bumpers
of them.
The bumper of each ship is defined as shown in Fig. 3. A
bumper model is used for evaluating safe distance between ships,
and the bumper is usually defined as 6.4L in front of a ship, and
1.6L in both sides and aft of the ship. To identify the collision of
two ships, the bumper is assumed to be halved as shown in Fig. 3
in the present simulation.
Note that the needed reduction of the speed is larger for a slower
oncoming ship of 10 knots. This is because a slower ship takes
more time to cross the belt of the sailing course of the own ship.
It is also possible for the ship by increase of the sailing speed to
avoid the collision with the oncoming ship. In Fig. 5 the obtained
speeds to require to avoid the collision by increasing the own
speed are shown. Similarly, large change of speed is needed when
the sailing distance in higher speed is shorter than five nautical
miles. Note that the dependency of sailing speed of an oncoming
ship on the required change of speed is more significant than the
case of deceleration shown in Fig.4.
These results show us how much deceleration or acceleration
are needed to avoid collisions.
15

encounter angle 1590 deg. 10

predicted
collision point
oncoming vessel
L2120 m V [kt] V2: speed of oncoming ships
V2 V21040 kt 5
10 kt 20 kt
Acceleration or deceler- V
ation distanceL own vessel 30 kt 40 kt
L1142 m
0
V115kt
0 5 10 15 20
deceleration distance L [mile]
Fig. 2 Definitions of parameters in simulation of collision
Avoidance. Fig. 4 Required speed drop from navigating speed (15kt) for
avoiding collision with ships running in 10-40kt.

0.8L1 3.2L1 30

0.8L1 V2: speed of oncoming ships

10 kt
25
20 kt
V1 [kt]

Fig. 3 Bumper model for identifying collision of two ships. 30 kt

40 kt
20
REQUIRED CHANGE OF SPEED
In the first simulation, necessary amount of change of speed to
15
avoid the collision with an oncoming ship is calculated. In the 0 5 10 15 20
simulation, the own-ship is assumed to be a RORO ship with deceleration distance L [mile]
142m of length, and is sailing in 15 knots, and the oncoming ship
is 120m length and her sailing speed is systematically changed Fig. 5 Required speed up from navigating speed (15kt) for avoiding
from 10knots to 40knots. The simulation is carried out for 15 collision with ships running in 10-40kt.
90 of encounter angles of the two ships.
EVALUATION OF PROPOSED COLLISION AVOIDACE
Examples of the simulated results are shown in Figs.4 and 5.
METHOD
The results are those at the condition of 30 degrees of encounter
angle between the two ships. Required speed V of the automati- SPEED CHANE WITHIN 10%
cally sailing ship to avoid collision with the oncoming ship is In collision avoidance sailing, it may be better for the developed
shown for various distances during changing speed in nautical automatically sailing ship that the change of speed is shorter and
miles. Fig. 4 shows the results when the ship sails to avoid colli- smaller. Therefore, the limit of the changing speed should be
sion by sailing just behind the oncoming ship without any contact determined. In the present study, 10% from the sailing speed is
of the both bumpers of them by reducing the speed of the assumed as a criterion. Required distances for changing speed are
own-ship. We can see that large reduction of speed is needed if the calculated by the simulation. As examples of the simulated results,
sailing distance in slower speed is shorter than five nautical miles. the calculated required distances for change of speed at an en-

238
counter angle of 30 are shown in Fig.6. Size and speed of the L2: Length of oncoming ships
oncoming ship are systematically changed from 60 m to 240 m, acceleration deceleration
and 10 to 40 knots, respectively. The results for collision avoid-
ance by acceleration are shown in solid lines, and those by decel-
eration are shown in dotted lines in the figure.
22
L2: Length of oncoming ships 15
20

speed of own vessel V1[kt]

acceleration deceleration
18

16

15
30 14

12

10 10
0 10 20 30 40
L [mile]

22
30
5 20

speed of own vessel V1[kt]

18

0 16
0 10 20 30 40 50
speed of oncoming vessel [kt] 14

Fig. 6 Required distances of 10% accelerated or 10% decelerated 12

navigations to avoid collision with oncoming ships with var-
ious length (60-240m) and at various speed (10-40kt). 10
0 10 20 30 40
22
As the oncoming ship is larger and slower, longer change of 60
20
speed of own vessel V1[kt]

speed is needed to avoid the collisions. These are because it takes

longer time for such ships to cross the belt of the sailing course of 18
the own-ship. For smaller oncoming ships, the collision avoidance
16
method by acceleration needs shorter distance of speed change.
For medium and larger oncoming ships, the acceleration method 14
needs shorter distance of change speed at higher speed, but the 12
deceleration method needs shorter distance at lower speed. This
results may suggest that an appropriate method to avoid collision 10
0 10 20 30 40
depends on the size and speed of the oncoming ship which can be
22
collided in short minutes. 90
20
speed of own vessel V1[kt]

DISTANCE OF CHANGING SPEED 18

In the master thesis of one of the authors (Yoshida 2016), speed
change during more than 20 nautical miles is sometimes needed 16

by collision avoidance using 10% speed change from the sailing 14

speed. However, shorter distance of changing speed may be better,
12
but too short distance may give the navigator of the oncoming ship
some doubt and fear. Therefore, for the developed automatically 10
operating ship should change her speed in shorter but in enough 0 10 20 30 40 50
speed of oncoming vessel V2 [kt]
distance which the ship does not give the doubt and fear. Fuchi
(2014) reported that most of navigators start to consider collision Fig. 7 Required speed change during 6 nautical miles from 15kt
navigating speed to avoid collision with oncoming ships
avoidance in mind when the distance of the two ships is near about navigating from various directions (15-90).
four nautical miles. On the basis of the report, it can be said that
six nautical miles may be an enough distance not to give fear the
navigator of other ships to start the collision avoidance action. REQUIRED SPEED DURING SIX NAUTICAL MILES TO
AVOID COLLISION

The simulated results of required speed to avoid collision by

changing speed during six nautical miles are shown in Fig.7. In
the figure, the sailing speed of 15 knots is shown read solid lines,

239
and the ship speeds which are changed by 10% are green broken
lines and orange broken lines, respectively. The results demon- L2/L1 0.25 0.5 1 2 3
strate that in most cases at the encounter angle larger than 30 100
degrees, the ship can avoid collisions with other ships by changing 80
deceleration
speed within 10% during six nautical miles keeping her straight 60

[deg.]
original course. 40
But for slower oncoming ships, larger speed change more than
20
10% or longer speed change than six miles are needed. Course
0
change, or veering, can be also selected in such cases. 0 0,5 1 1,5 2 2,5
V2/V1
100
LIMITATION OF PROPOSED METHOD acceleration
80
As mentioned before, the proposed method for collision avoid- 60
ance has a limitation in practical usage, like too long or too large

[deg.]
40
change of speed. One of the authors (Yoshida K., 2016) found that
20
ratios of lengths and speeds of the two ships are key factors to
0
determine the required change of speed for avoiding collision. An 0 0,5 1 V2/V1 1,5 2 2,5

example of the simulated results is shown in Fig. 8. The figure

demonstrates that even if the speeds of the own ship are different,
like 10 and 20 knots in the figure, the ratios of the required speed
to the original sailing speed, V/V1, are almost same.
Using the ratios of speed and size of the two ships, figures to
show criteria of the proposed method can be made as shown in Fig.
9. Lines to connect the same marks show the criteria of the meth-
od, and in regions above the lines the ship cannot avoid collision
by 10% change of speed during six miles. We can see that as the
oncoming ship is larger and slower than the own ship the regions
where the proposed collision avoidance method can work become
smaller.
L2/L1
V1=10 kt
V1=20 kt
100
V2/V1 = 1
V/V1 100 [%]
Fig. 9 Critical encounter angle between two ships to avoid colli-
sion by 10% change of speed during six nautical miles.
COURSE CHANGE TO COLLISION AVOIDANCE
The simulated results clearly demonstrate that in cases of small
encounter angle, for slower ships and larger ships, the proposed
collision avoidance method by changing speed in keeping the
straight course does not work well. Then, course change is needed.
When there are ships at anchor or at rest, working ships like a
dredge boat and fishing vessels operating in the course, the ship
must change her course. Usually a ship avoids an oncoming ship
with the veering method shown in the left figure of Fig. 10, but the
method sometimes does not tell the intention of the collision
avoidance action of the ship to the navigators of an oncoming ship.
To tell the intention clearly to the navigator of the oncoming ship,
the developed ship in the present study changes her course as
follows. At first, the ship changes her course to be parallel to the
course of the oncoming ship. At the point where the offset from
the original course of the own ship is enough to be able to pass
behind the oncoming ship, the own-ship changes her course in the

90
parallel course to the original one. Then the ship changes the
course again to return on the original course of her. The method is
80
called the Offset Method in this paper. The minimum offset can be
calculated in advance. This method slightly increases the running
70 distance, but the intention of collision avoidance action of the ship
0 20 40 60 80 100
[deg.] can be easily understood by the navigators of an oncoming ship.

150
V2/V1 = 1
140
V/V1 100 [%]

predicted
130 oncoming vessel
collision point
120

110
6mile V1 V
1

100
0 20 40 60 80 100 own vessel
[deg.]

veering method offset method

Fig. 8 An example of simulated results of required change of
speed during 6 nautical miles to avoid collision of ships
Fig. 10 Collision Avoidance navigations; conventional veering
of different sizes (L2/L1=0.25-3). method (left) and proposed offset method.

240
 Examples of the calculated offsets from the original course of
7000
the ship are shown in Fig. 11. In the calculations, the two ships
have the same length of 120m, and the own-ship sails in 15 knots.

EHP
Encounter angle of the two ships and the sailing speed of the
oncoming ship are systematically changed as shown in the figure.
The results show that as encounter angle becomes larger, the
0
required offset increases. The calculated results also show that the 0 5 10 15 20
Us (kt)
required offset distances are within 0.4-0.8 nautical miles.
Fig. 12 EHP curve used in prediction of increase energy
consumption due to proposed collision avoidance methods.
15 30 60
1
L2120 m
0,8 Keeping course method Changing course method
Loff [mile]

0,6

0,4

0,2
120

consumption(%)
0 15
0 10 20 30 40 50
110

energy
speed of oncoming vessel V2 [kt]
100

Fig.11 Simulated offset distance from the original sailing route

90
for avoiding collision by changing course. 0 10 20 30 40 50

EVALUATION OF PROPOSED METHOD IN TERM OF EN- consumption(%) 120

30
ERGY CONSUMPTION
110
energy

Energy consumption is a key factor for shipping economy and

protections of the global environment. Therefore, the proposed 100

collision avoidance method is evaluated in terms of energy con-

90
sumption in this section. 0 10 20 30 40 50

In the evaluation, a RORO cargo ship with 140m of length is

selected. The collision avoidance action starts at the point where is 120
consumption(%)

located at six nautical miles from the predicted meeting point of 60

110
energy

the two ships, and the own-ship returns on the original course on
the point where is far twelve nautical miles from the starting point 100

of the collision avoidance action to keep the original schedule. 90

0 10 20 30 40 50
This means the ship keeps her original scheduled at the point after speed of oncoming vessel V2 [kt]
collision avoidance navigation during twelve nautical miles.
Fig. 13 Increase of energy consumption due to collision
avoidance navigation in percentage (L2:120 m).
The energy consumption during the collision avoidance naviga-
tion is calculated by using the EHP curve shown in Fig. 12. Some
120
examples of the calculated results are shown in Figs. 13 and 14.
L260m
consumption(%)

Fig. 13 shows the results of increase of energy consumption due to 110

two methods of collision avoidance navigations, a keeping-course
energy

methods (deceleration or acceleration), and the change course 100

method (offset method). Length of the oncoming ship is assumed

90
to be 120 m. The results show that at small encounter angle the 0 10 20 30 40 50

energy consumption by the change of course method is lower than 120

L2240m
those by the keeping-course method. At larger encounter angle,
consumption(%)

110
however, the energy consumption by the keeping-course method
energy

is lower than those by the changing-course method. Increase of the 100

energy consumption by the proposed method is only several per- 90

0 10 20 30 40 50
cent.
speed of oncoming vessel V2 [kt]

In Fig.14, effects of size of the oncoming ships on the calculat-
ed energy consumptions are shown. The encounter angle of the
two ships are 60. The energy consumptions by the changing-
course method are larger than those by the keeping-course meth-
ods.
Fig. 14 Effect of oncoming ship length on energy consumption due
to collision avoidance navigation ( : 60).
CONCLUSIONS
An automatically navigating ship under a concept of gentle
navigation which means the ship gives no fear to the navigators of

241
other ships is proposed and the collision avoidance method with a Ships and Its Evaluation, Master thesis of Osaka Prefecture
simple navigating way is evaluated. Following conclusions have University, (in Japanese)
been obtained.
1) When the ship avoids a collision with other ship by reducing
or increasing speed during six nautical miles with keeping the
original straight course, the ship can avoid the collision by
changing her speed by 10% from the original speed if en-
counter angle is over 30degrees.
2) Required amounts of the speed change depend on the ratios
of speed, size and encounter angle of two ships.
3) When the two ships will meet with encounter angle below 15
degrees and the coming ship is larger or lower than the own
ship, the collision cannot be avoided by 10 change of speed
during six nautical miles. In this case, larger or longer speed
change, or change of course are needed. The required chang-
es can be obtained by a simple simulation.
4) Evaluation in term of energy consumption show that the
present proposal of collision avoidance method can work
well in smaller increase of energy consumption.
5) To avoid a coming ship in small angle, change of the course
of the own ship is better from the point of view of energy
saving.

Ways for avoiding collision with ships staying for working in

the planned course of the automatically navigating ship will be
proposed in next paper.

ACKNOWLEGMENTS

The present study is carried out in the joint project on

development of an automatically operating ship. The authors
appreciate Imabari Shipbuilding Co., Ltd and Japan Radio Co.,
Ltd. for their sponsorship for the project.

REFERENCES

Ackerman, E. (2014), Unmanned Ships Face Industry Resistance

Are a Good Idea AnywayHomepage of Rolls-Royce

Fuchi, M., (2014), Characteristics and problems of collision

avoidance by radar at Sea, Ergonomics, Vol.50pp54-55 (in
Japanese)

Fujii, Y., Makishima T., Hara, K., (1981), Ocean Transport

Engineering, published by Kaibundo Shoten (in Japanese)

Hasegawa, K. and Kouzuki, A. (1987), Automatic Collision

Avoidance System for Ships Using Fuzzy Control Proc. of 8th
Ship Control Systems SymposiumVol.2pp.34-58

Oshima, H., (1989) High-reliability intelligent ship (advanced

automatic operation system), Bulletin of SNAJ, No.721 (in Jap-
anese)

Shoji, R. (2013), Navigation Support and Advanced Navigation

System, KANRIN47, (in Japanese)

Tvete, H. (2015), The ReVolt - A new inspirational ship con-

cept,
https://www.dnvgl.com/technology-innovation/revolt/index.html

Yoshida, K. (2016), Collision Avoidance Methods for Automatic

242
 Applied Bayesian Network Model for Risk Assessment for
Ship Collision
Takeshi SHINODA*, Ryoma, NASU***, Koji URU**
* Department of Marine Engineering Systems, Kyushu University, Japan, shinoda@nams.kyushu-u.ac.jp
** Development Operation Headquarter, Class NK, Japan, r-nasu@classnk.or.jp
*** Human Resource Group, NYK Line, Japan, koji_uryu@jp.nykline.com

ABSTRACT Accidents Inquiry Agency Japan made public to extract

In this study, a risk assessment model for ship colli-
sions is proposed according to the guidelines for Formal
Safety Assessment (FSA) approved by IMO in 2002. The
analysis is applied to ship collisions between cargo ves-
sels owing to their high frequency and enormous dam-
age. Bayesian network theory for risk analysis has been
applied to reveal a causal relationship on human factors.
A trial evaluation of Risk Control Options (RCOs) for
collisions is attempted through the calculation of the
dominance index. Finally, a trial cost benefit analysis
for RCOs is considered through Gross Cost of Averting
Fatality (GCAF) in FSA.
INTRODUCTION
Most vessel-related marine accidents result from human
factors such as misdetection, misjudgment, and faulty oper-
ation in using functional system that enable safe navigation.
Recently, collision accidents by cargo vessels have fre-
quently occurred in seas close to Japan and have been
widely reported in the media. This type of accident is still
frequent and has led to huge damages including risk of loss
of life and environmental disruption.
The annual report released by the Japan Marine Acci-
dent Tribunal in 2015 shows that collision accidents occur
most frequently (22%) among all marine casualties (JMAT
2015). According to our survey for the ship traffic survey
and constructed accident database, occurrence probability
in Seto Inland Sea in Japan were 4.55x10-5 (cases/ship
year )
Owing to high ship traffic amount and high accident
occurrence frequency in Seto Inland Sea, collision acci-
dents between cargo vessels are investigated in this study
and a risk assessment is proposed in accordance with For-
mal Safety Assessment (FSA) procedure approved by IMO
in 2002 (IMO 2002).
In our previous research, first, we mainly investigated
collisions between fishing and cargo vessels. we used the
records of judgment of the marine accidents that the Marine
data for human-related factors and categorize the data ac-
cording to facts or the human factors involved in the colli-
sion. Second, we created a database of collisions between
fishing and cargo vessels. Third, we developed a risk anal-
ysis methodology on the basis of the probability of colli-
sions, which are taken the ship traffic data and estimated by
combining the analyses of marine casualty data and the
navigation track data obtained by examining the ship traffic
survey in Kanmon Channel in Japan. Event tree analysis
was applied to evaluate the collision risk. Finally, we ex-
amined the data to evaluate the risk control options (RCOs)
using the developed methodology (T. Shinoda and Y.
Tamura 2011).
In this study, by extending our previous research, we
examine the application of Bayesian network methodology
for evaluation of the effectiveness of the RCOs as safety
measures to reduce the collision risks.
Here, cargo vessels are defined as merchant vessels that
are generally of a large size.
FLOW OF RISK ASSESSMENT
The proposed risk assessment is in accordance with the
FSA procedure approved by IMO in 2002. In this research,
human factors are considered in connection with the FSA.
The proposed risk assessment consists of five steps: identi-
fication of hazards, risk analysis, generation of RCOs, as-
sessment of cost, and benefit and recommendation of the
decision maker. The flow of the proposed risk assessment is
shown in Fig. 1. The details of each step are described as
follows.
(Step 1) Identification of hazards:
To identify the hazards associated with human factors in
the ships maneuvering, a collision database is constructed.
Moreover, to analyze the hazard associated with traffic
volume, a ship traffic survey was conducted and data de-
scribing harmful situations were extracted.

243
 Step 1: Process of identification of h azards

Analyzing records of adjudication Application of variation tree

of marine accident inquiry in Japan analysis (VTA) Observation / AIS data
analysis
Extracting human factor through VTA

Configuration of i tems of database Extraction and estimation

Cognition and decision-making factors of traffic volume
Construction of database for collision
Input and consolidate casualty data Estimated occurrence
probability for col lision

Step 2: Process of risk analysis Step 3: Process of generation of risk control options

Consideration of relation of cause and e ffect of accident

Causal relationship of human Conditional Probability
factors by Bayesian network Table (CPT) Cost increase C Risk reduction R

Step 5: Process of recomme ndation to decision maker Step 4: Process of assessment of cost benefit analysis

Fig. 1 Flow of proposal method of safety assessment for collision accidents

Auditory Visual
perception perception Ordinary operation

Perception Automatically updating Correspondence

Ordinary Estimation Decision of
of the situation the action Response
Environment
Attention Direct perception of peril Redefinition Correspondence
of the situation Decision of
Uncommon Activation of script about peril (Expectation the action
Avoidance schema from peril of the situation ) Response

Detection Watch Judgment Avoidance Collision

keeping of peril of Operation
collision

Fig. 2 Model of human information processing on maneuvering

To configure the collision database, the hazards at-
tributable to human factors relating to collisions are identi-
fied using the records of the adjudication of marine accident
inquiries, application of the variation tree analysis (VTA)
method, and definition of maneuvering processes based on
human cognitive information processing in an emergency
situation.
On the other hand, for the ship traffic survey, we used
the software (Ship finder) obtained for cargo vessels by
marine radars, automatic identification system (AIS), and
visual observation in the Seto Inland Sea in Japan in 2014
to 2015.
The trial calculations of occurrence probabilities of col-
lision accidents were performed using data obtained from
the constructed database and the ship traffic survey. The
hazard of collision is estimated by the use of occurrence
probabilities of collision.
(Step 2) Risk analysis
Risk analysis model applied by a Bayesian network is
created for collision accidents to elucidate a causal rela-
tionship with human factors in maneuvering and analyze
the ship collision risk using a conditional probability table
that is calculated from data in the constructed collision
database.
(Step 3) Generation of risk control options
RCOs to prevent collision accidents are generated from
the results of the database analysis. Safety measures such as
improvement of the radar reflector, shapes with high level
visibility at night time, collision warning equipment, and
installation of AIS are considered practical RCO examples
in this study, and the effectiveness of each RCO is analyzed
by the proposed risk analysis model for collision accidents.
(Step 4) Assessment of cost benefit
The cost benefit assessment is an important step in order
to decide to implement the RCOs in ship operation. Trial
cost benefit analysis for RCOs is considered through gross
cost of averting fatality (GCAF) in the FSA process. The
risk reduction effect by risk analysis model and cost evalu-
ation through step 3 are used in the calculation of GCAF.

244
 Table 1 Code of cognition and decision-making factors on collision accident database
AP1 Success of detection
Category A AP2 Success of detection by radar
Failure of detection
Detection AP3
of opponent AP31 Distraction by other vessels
vessel AP32 Distraction by other works
AP33 Improper watch
AP34 Sleeping
AP35 No watch at anchoring
AP36 Watch fore side only
AJ1 Shipping route
AJ2 Judge as safe situation
Judge from experiences
AJ21
AJ22 Detect no vessels by radar
AJ23 Recognize no vessels
AJ24 Expect other vessels action
AJ3 Non compliance of regulations
AJ31 Improper watch assignment
AJ32 No report on other vessels
AJ33 Improper watch
AJ34 No attention to blind zone
AJ4 Fatigue
AJ41 Feel sleepiness
AJ42 Fatigue
AJ43 Posture during watch (sitting...)
AJ44 Under the influence of alcohol
Watchkeeping
Category B BP1
BP2 Watchkeeping by radar detection
Watch Failure of detection
BP3
keeping BP4 Insufficient watchkeeping
BP411 Distraction by other vessels
BP412 Mind out by change of course
BP42 Distraction by other works
BP43 Sleeping
BP43 Postponent of decision
BJ1 Shipping route
BJ5 Misjudge on passing innocuusly
BJ51 Misjudgment on situation
BJ511 Judge from cursory inspection
BJ512 Improper watch assignment
BJ513 Correspond to MARTIS
BJ52 Misjudgment on crossing situation
BJ53 Take the action to avoid collision
BJ54 Change of course
BJ55 Non compliance of regulations
BJ6 Expect other vessel avoidance action
BJ61 Role by regulations
BJ62 Take action to encourage other
vessels avoidance action
(Step 5) Recommendation of decision maker
Recommendations are discussed and four steps are pre-
sented to the relevant decision makers in the relevant sec-
tion of the IMO.
This study examined the feasibility of steps 1-4 while
providing a trial risk assessment for shipship collisions
between cargo vessels that are related to typical human
factors.
CONSTRUCTION OF MARINE CASUALTY DATA-
BASE
VTA is a method to establish safety measures for acci-
dents and extract variation factors from the sequence of
operational procedures used in the accident. The variation
factors depart from the usual procedures and involve hazard
factors that should be excluded for safe operation. The
CP2 Watchkeeping
Category C CP3 Failure of detection
CP4 Insufficient watchkeeping
Judgment CP5 Lose sight of other vessels
of peril of CM1 No action to avoid collision
collision CJ5 Misjudge on passing innocuusly
CJ52 Misjudgment on crossing situation
CJ54 Change of course
CJ6 Expect other vessel avoidance action
CJ611 Stand-on vessel (crossing vessel )
CJ612 Stand-on vessel (overtaking vessel )
CJ62 Take action to encourage
other vessels avoidance action
CJ67 Failure of judgement
on other vessel condition
CJ7 Failure of judgement on situation
CJ71 Failure of judgement (late action)
CJ8 Improper watch assignment
CJ81 Expect co-watchers judge
CM2 Failure of avoidance action
CJ51 Failure of judgment
DP3 Failure of detection
Category D DP4 Insufficient watchkeeping
Avoidance DP5 Lose sight of other vessels
operation DM1 No avoidance action
DJ5 Misjudge on passing innocuusly
DJ513 Correspond to MARTIS
DJ52 Misjudgment on crossing situation
DJ54 Change of course
DJ55 Non compliance of regulations
DJ6 Expect other vessel avoidance action
DJ7 Failure of judgement on situation
DJ8 Improper watch assignment
DM2 Improper/Insufficient avoidance action
No avoidance action
EM1
Category E
EM12 Notice at close range,
Avoidance but unable to take action
operation EM13 Expect other vessel action
at close EM14 Tie down
range EM15 Can not avoiding
EM2 Improper/Insufficient avoidance action
EM3 Action at close range
EM31 Engine operation
EM32 Steering operation with signal
EM33 Steering operation without signal
EM34 Reminder signal (whistle/light)
EM35 Warning signal
extracted variation factors are considered as causes of the
accident and are sequentially arranged to show their flow;
the safety measures for accident prevention are investigated
by cutting the flow of the variation factors. To extract the
variation of the human factors in the collisions between
cargo vessels, the VTA is applied to the records of 100
cases judged at the Marine Accident Inquiry from 2002 to
2007. (Y. Tamura, T. Shinoda 2009)
The marine casualty database was constructed with the
extracted variation human factors obtained from the adju-
dication records of marine accident inquiry, such as acci-
dent characteristics (place, time, season, weather, sea con-
dition, visibility), vessel characteristics (type, length, gross
tonnage, speed, course, bridge resources, roles at crossing
situation such as give-way or stand-on), and accident cate-
gories (head-on situation, crossing situation and overtaking
situation).
245
 80 (B)Watch keeping (C) Judgment of peril of collision
(%)
70 (A)Detection (D,E) Avoidance operation
60
50 Remarks

40
30
20
10
0
tion esse
l)
esse
l)
esse
l) ssel
)
Oth
ers rage
n situa - on v ay v k en v i ng ve Ave
d - o n d e - w r ta rta k
Hea sta (giv (ove (ove
on ( tion tion
s i tuati it ua i tu a u a t ion
g gs s s it
ssin ssin rtak
in g ing
Cr o Cr o Ove v e rtak
O Type of accident

Fig. 3 Frequency of cognition and decision-making factors in collision by database analysis

Table 2 Definition of random valuable based on cognition level and prescriptive maneuvering

Random Distance division between Ordinary level of cognition and judgment

variable two vessels in the prescriptive maneuvering
x1 Equal and more than 6 miles Detection of other vessel at collision risk

x2 Watch keeping of other vessel

Less than 6 miles
at collision risk
x3 Less than 4 mile Judgment of peril of collision
x4 Less than 2 mile Taking avoiding action
x5 1 mile or less and close range -
x6 Failure -

The total database consists of 4,867 cases that were
judged at Marine Accident Inquiry from 1988 to 2014. The
items in the database are considered for the ships maneu-
vering process based on human cognitive information pro-
cessing in emergency situations that is defined as percep-
tion of an unusual situation, redefinition of the situation,
decision of the action, and correspondence to the environ-
ment, as shown in Fig. 2. Following such information pro-
cessing, the flow of ships maneuvering process is defined
using five major categories: detection, watch keeping,
judgment of peril of collision, avoidance action and avoid-
ance action at close range.
Finally, the influences of human factors on the collision
are reconstructed for the four major categories, and the
details of the categories are summarized as cognition and
decision-making factors on collision accident in Table 1.
Fig. 3 shows an example of the analysis of the con-
structed collision database using 360 collision cases. Fre-
quency of cognition and decision-making factors in colli-
sion that is categorized by accident type: head-on situation,
crossing situation (stand-on and give way) and overtaking
situation (overtaking and overtaken). The cognition and
decision-making factors concerning detection is major part
in every accident type, and it is as much as 40 % on average.
It is important to reduce the detection factors.
CONSTRUCTION OF EVALUATION OF EFFECTIVE-
NESS OF RCOs BY APPLIED BAYESIAN NETWORK
Accident Occurrence Model by Applied Bayesian Network
The Bayesian network model is a convenient method-
ology for revealing a causal relationship and calculating the
posterior probability by conditional probability. In this
study, risk analysis models using Bayesian network are
created for the collision accident to reveal a causal rela-
tionship with human factors in maneuvering and analyze
the risk of collision using the conditional probability table
calculated using the constructed collision database. The
application of Bayesian network to the evaluation of RCO
effectiveness is particularly effective.

246
 Fig. 4 Bayesian network model for risk assessment for collision

The Bayesian network model for risk assessment for
collision consists of three major models such as the model
of ships maneuvering, the model of collision factors, and
the model of effective factors for RCOs, as shown in Fig. 4.
The details of the models are as follows (Shinoda et al.,
2012).
In the model of the ships maneuvering process, four
major random variables are used. These are Situation A:
sensibility, Situation B: watch keeping, Situation C: judg-
ment of peril of collision, and Situation D: avoidance ma-
neuvering. While in principle, many characteristics of the
ships state such as the distance, angle, vessel velocity, and

247
ships operation can be used to set the random variables, in
this study we use the distance between the two ships in-
volved in possible collision to set the random variables set
because inquiry records include detailed descriptions of the
distance between the two ships until collision.
Table 2 shows the categories of vesselvessel distances
such as more than 6 miles, 4 to 6 miles, 2 to 4 mile, 1 to 2
mile, and 1 mile or less and close range. Each distance
category is related to the ordinary level of cognition and
judgment in the prescribed maneuvering. This model con-
siders the distance category as the status of random varia-
bles ranging from x1 to x6, as shown in Table 2. All random
variables and their states are listed in Fig. 4.
In the model of cognition and decision-making collision
factors, although all factors shown in Table 1 should be
considered as random variables, due to the limitation of
Bayesian network software, in this study only the repre-
sentative factors are selected and their existence probability
is calculated.
In the model of effective factors for RCO, collision fac-
tors associated with the effectiveness of RCO are reduced
when the RCO functions effectively, and a new model of
collision factors is then reconstructed. Calculation with the
new model results in a change of existence probability; this
change represents the effectiveness of the RCO. The details
of the method for evaluation of the RCO effectiveness are
described in the following section.
Method of Evaluation of Effectiveness of RCOs
Setting of CPT and impact of RCO
In the beginning of calculation of Bayesian network, the
conditional probability tables (CPT) for the model of ships
maneuvering process are created using the constructed
collision database. These are given as follows:
(1)
The tables relate the variables APi, AJi, BPi, BJi, CPi,
CJi, CMi, DPi, DJi, DMi, EPi, EJi, and EMi that are de-
fined in the list of random variables in Fig. 4 and described
in Table 1.
For the evaluation of the effect of safety improvement
of RCO we define a set of safety impact factors on RCO as
SIRCOi. The elements of SIRCOi consist of effective fac-
tors for all collision factors. For example, when the RCO
influences the collision factor on APi, the CPT of APi takes
the value 1. Otherwise, it takes the value 0. The formula for
the CPT is expressed as follows:
(2)
248
Estimation of State of Maneuvering
Let PRCOi,jk denote the estimated existence probability
after taking RCO No.i based on each situation Xj on ships
maneuvering and each distance between two vessels xk.
The formulae for PRCOi,jk are given as follows:
(3)
The values of existence probability of each situation on
ships maneuvering are calculated using these conditional
probabilities and the safety state of maneuvering can then
be obtained. In this study, the Byonet software is used for
Bayesian network calculation (Y. Motomura, 2006).
RCO Effectiveness Index
The effectiveness of an RCO can be obtained by calcu-
lating the difference in existence probability before and
after RCO measures.
Let P0,jk denote the existence probability before taking
the RCO; the matrix for evaluation on dominance D of
RCOi,jk that shows the existence probability differences is
defined as follows:
, (4)
where sgn is the signum function, with values taking
1 or 0 or 1 for positive and 0 and negative values in the
brackets of sgn, respectively. Let MD,jk denote the matrix
defining the dominance of PRCOi,jk and given by
, (5)
where the line of matrix represents each situation Xj of
the ships maneuvering process and the column represents
each category of the distance between two vessels xk. Each
element of MD,jk is assigned either 1 or 1 to represent the
effectiveness of the ships maneuvering. Here, the elements
of MD,jk are set through an interview survey of the captains
of the cargo ships.
For example, the first line of matrix MD,jk treats the
situation of X1 showing detection, and detection on ship s
maneuvering generally should be finished earlier. The ele-
ments of MD,jk expression take the positive value 1 until x2
to evaluate the effectiveness of detection by the safety im-
provement on RCO.
Furthermore, other elements after x3 take the negative
value 1 because the difference of PRCOi,jk and P0,jk is
negative, changing the negative sign to evaluate DRCOi,jk.
Let DRCOi,j denote the dominance index for the situa-
tion that expresses the effectiveness of RCO in each ma-
neuvering process. DRCOi,j can then be calculated using
 0.7
(a) Before taking RCO (b) After taking RCO #1
0.6 State of random variable

Existence probability
x1 x2 x3 x4 x5 x6=Failure
0.5
0.4 Remarks
0.3
0.2
0.1
0
on ing ion n on pin
g on
isi : erati
on
te cti e ep ol lis : ratio t ecti e e ll
e o n D op
De hk fc n D op :D
e k fc
A: atc : eril o atio nce A a tch il o uatioance
n W u a n : r t
tio : C fp Sit void tio : W C pe Si oid
itua io n B ation nt o A itua o n B ation nt of Av
S at tu e S at i tu e
u i m u Si dgm
Sit S dg
Ju Sit Ju
Maneuvering situation

Fig. 5 Calculation results of posterior probability of each maneuvering situation in collision

Eq. (4) as the sum of DRCOi,jk over xk. It is given as
follows:
(6)
Let DRCOi denote the RCO dominance index that ex-
presses the total effectiveness of RCO. Using Eq. (6),
DRCOi is then calculated as the sum of DRCOi,j over j. It is
given as follows:
(7)
The dominance index must be able to evaluate the total
effectiveness of RCO. Let Dall denote total effectiveness
when the RCO has effectiveness on all factors. The value of
RCO effectiveness VROCi can then be defined as follows:
(8)
The existence probability values on detection suggest
that the appropriate on-board detection was not operating
when accidents occurred.
On the other hand, P0,15 is 0.06 when the detection is
abnormally carried out in the state of close range (X1 = x5)
and P0,1f is almost 0.13 when detection failed.
The existence probability values in each situation of
watch keeping, judgment of peril of collision, and avoid-
ance operation are similar to the value of the situation of
detection, indicating that the operational delays appear in
each situation.
Finally, the sum of the probability values of P0,45, and
P0,4f increases to 0.82, suggesting that delay or failure of
avoidance operation appear for occurrence of accidents.
This trend suggests that earlier detection on cargo ves-
sels is not effective for watch keeping and judgment of peril
of collision.
RCO EFFECTIVE EVALUATION
RCOs

According to the trend exhibited by the probability val-

Risk analysis of collision accidents
ues as described in the previous section, RCOs for support-
ing detection are effective for collision prevention. The
In this section, we present practical examples of the
effectiveness of RCO is evaluated by the proposed method
calculation method described in the previous section.
using the RCO effectiveness index.
Fig. 5(a) shows the existence probability values P0,ik at
random value xk in each state Xi before taking RCOs. The
RCO #1) Installation of class-A AIS:
P0,11 that represents the existence probability value at a
According to regulation 19 of IMO SOLAS Chapter V,
random value x1 in state X1 is 0.27. While the value of P0,11
the regulation requires AIS (Automatic Identification Sys-
is high when detection (X1) is normally performed in the
tem) to be installed aboard all international ships of 300
state of the shipship distance of more than 6 mile (X1 = x1),
gross tonnage and upwards engaged on international voy-
the probability value on the detection is low for the occur-
ages, all cargo ships of 500 gross tonnage and upwards not
rence of accidents. Even when the detection is conducted at
engaged on international voyages, and all passenger ships
less than 6 mile (X1 = x2) that is the range of 4 mile to 6
irrespective of size (IMO 2015). The installation of the
mile, the existence probability value is 0.23; however, this
class-A AIS on all cargo vessels is considered as RCO. The
is not still high level.
system that provides the current information such as ship
name, speed, and course to monitor other vessels are the
minimal requirements for this system.

249
 Table 3 Estimated cost and impacts to cognition and
decision-making factors on each RCOs
RCO Expected impact to cognition Expected cost
Measures
No. and decision-making factors (USD)
AP31, AP33, AP35, AP36
RCO #1 Installation of AJ21, AJ22, AJ23,AJ24, 25,000
AIS
AJ31,AJ32,AJ33,AJ34
Installation of AP34, AJ31, AJ41, AJ42
RCO #2 doze prevention 30,000
device AJ43, AJ44

AP31, AP33, AP36, AJ1,

RCO #3 Installation of AJ22, AJ23, AJ24, AJ32, 50,000
the second radar
AJ33, AJ34

Installation of AP31, AP33, AP36,

RCO #4 GPS plotter 150,000
AJ21, AJ22, AJ23, AJ34

of peril of collision increase in the early phase where both

RCO #2) Installation of doze preventing device:
The collision frequently occurs by doze at the wheel.
This result suggests that installation of doze preventing
device is one of improvements to reduce collisions.
RCO #3) Installation of the second radar:
According to regulation 19 of IMO SOLAS Chapter V,
the regulation requires, X-Band radar is required on any
vessel of 300 gross tonnage and over. S-Band radar is also
required on ships of 3,000 gross tonnage and over, and it is
fitted as the second radar. It is expected that the second
radar built on any vessel has a redundant functional system.
RCO #4) Installation of GPS plotter:
GPS plotter is a device used in marine navigation that
integrates GPS data with an electronic navigational chart
(ENC). The GPS plotter displays the position, heading and
speed of the ship, and can display additional information
from radar, AIS or other sensors. It is helpful for ship oper-
ator to evaluate a situation of the ship.
vessels are sufficiently far from each other, showing that
RCO #1 is effective. In addition, this result shows that the
impact of the RCO spreads to the other situations and re-
duces failures in judgment of peril of collision and avoid-
ance operation. Since the shape of existence probabilities
after introducing RCO illustrated in Fig. 5(b) is different
from that obtained before introducing the RCO shown in
Fig. 5(a), the RCO contribution can be evaluated by calcu-
lating the RCO effectiveness index.
Fig. 6 shows the calculation result for each effectiveness
index defined by Eqs. (6), (7), and (8). RCOs #1, #3 and #4
are effective in detection of other vessels. When these ef-
fects are combined, RCO #1 is evaluated to be the most
effective safety improvement.
Evaluation of Cost Benefit of RCOs
GCAF, an index of cost effectiveness in FSA, is used
for evaluation of cost benefit. The formula of GCAF is
defined as follows:

Table 3 shows the list of estimated costs and impacts on (10)

the cognition and decision-making factors on RCOs. In the
RCO effectiveness model, these impact items are consid-
ered as the elements of SIRCOi described in the Method of
Evaluation of Effectiveness of RCOs section.
For example, for RCO #1, AP31, AP33, AP35, AP36,
etc. that are described in Table 3 are cognition and deci-
sion-making factors defined in Table 1, and it is assumed
that these failure factors are reduced when AIS is intro-
duced in the vessels. In calculation of efficiency of RCO #4,
these factors are eliminated from the original Bayesian
network model for risk assessment. The new network is
reconstructed and existence probabilities are calculated.
Fig. 5(b) shows the calculation result of estimated ex-
istence probability values on each state after introducing the
safety measures of RCO #1.
According to this figure, the existence probability val-
ues on the state of detection, watch keeping and judgment
Where C is the cost increase factor and R represents
the risk reduction by introducing RCOs. In this study, C is
set to the price of RCO including installation fee and R is
set to the total RCO effectiveness index. The cost effec-
tiveness is shown by the ratio of both factors, and RCO is
more cost effective when for smaller GCAF values. Ac-
cording to the definition of risk reduction by a RCO is es-
timated as the reduced number of loss of life in the process
of evaluation by GCAF. However, the collision accidents
between cargo vessels are rarely linked to loss of life for
crews. In this research, all accidents including property
damage accident, fatal and injury accident should be ana-
lyzed.

250
 0.8 1.0
RCO #1 #2 #3 #4
0.7
0.8

Effectiveness index
0.6

Dominance indices
Remarks
0.5 0.6
0.4
0.3 0.4

0.2
0.2
0.1
0 0

n g ex ex
te ctio eepin ision tion ind s ind
De hk l a e s
col D: er nc ne
A: Watc il of tion ce op i na tive
r a
C: f pe Situ oida n m ec
ion do eff
ituat on B: uationent o Av
CO O
S i Situdgm
uat R RC
Sit J

DRCOi,jk DRCOi VRCOi

Fig. 6 Comparison of RCO dominance indices at each maneuvering situation and RCO effectiveness indices

Table 4 Calculation results of GCAF on each RCOs

RCO No. Expected cost, Total RCO GCAF

C effectiveness index, R (total cost basis)
Unit (USD) (USD)
RCO #1 2,500,000 0.816 3,064,719
RCO #2 300,000 0.007 41,644,118
RCO #3 5,000,000 0.482 10,372,894
RCO #4 1,500,000 0.600 2,501,885

Table 4 shows the calculated results of GCAF on each
RCO. In this table, according to the cost comparison data,
since cost on RCO #2 is less than cost on the other RCOs,
shipping companies can accept the introduction of these as
safety measures. However, the total RCO effectiveness
index on RCO #1 is the highest effectiveness index in
RCOs, and RCO #2 is the second one, and RCO #3 is ex-
tremely low. When taken these together, GCAF on RCO #2
and #3 are much bigger than the index on RCO #1 and #4,
shipping companies hesitate to accept that the introduction
of these RCOs, particularly #2. It is necessary to improve
safety performance.
From the view of GCAF on RCO, shipping companies
can accept the introduction of RCO #2 and #1 as safety
measures.
CONCLUSIONS
A risk assessment is proposed in accordance with the
FSA procedure and is applied to the problem of collisions
between cargo vessels. Through the application of the mod-
el to practical vessel operations, a risk analysis model for
ship collisions using applied Bayesian network and a
method of evaluation of effectiveness of RCO are proposed.
These analytical methods are then applied to the prob-
lem of collisions between cargo vessels. Cost benefit as-
sessment based on GCAF in FSA is calculated and the
possibility of acceptance of RCO by shipping companies
can be evaluated from comparison of GCAF.
The results obtained indicate that the proposed method-
ologies are efficient for providing risk assessment for the
problem of collisions.
REFERENCES
IMO (2013), IMO MSC-MEPC. 2/ Circ. 12, International
Maritime Organization
IMO (2016), International Maritime Organization
http://www.imo.org/OurWork/Safety/Navigation/Pages/
AIS.aspx (accessed on 17th May 2016).
Japan Marine Accident Tribunal. The report of Marine
Accident Inquiry (2016)
http://www.mlit.go.jp/jmat/saiketsu/saiketsu.htm
(accessed on 10th June 2016).

251
Shinoda, T. and Tamura, Y. (2011), Risk Assessment of
Collision Accidents at Sea in Consideration of Human
Factor -Application of Collision between Fishing Ves-
sels and Cargo Vessels, J. Institute Navigation, vol.
124, pp.11-19
Shinoda, T., Shimogawa, K., and Tamura, Y. (2012), Ap-
plied Bayesian Network Risk Assessment for Collision
Accidents between Fishing Vessels and Cargo Vessels,
J. Institute Navigation, vol. 127, pp.165-174
Motomura, Y. (2006), Technology of Bayesian network,
Tokyo Denki University Press
Tamura, Y. and Shinoda, T. (2009), Safety Assessment
with Marine Casualty Data on Collision Accidents be-
tween Fishing Vessels and Cargo Vessels, Proc.
ISOPE, pp.641-648
Tamura, Y. and Shinoda, T. (2013), Hearing Investigation
to Fishermen on Collisions with Cargo Vessels, J. In-
stitute Navigation, vol. 128, pp.81-88

252