See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/259693139

On the Time for the Abandonment of Flooded

Passenger Ships due to Collision Damages

Article in Journal of Marine Science and Technology · September 2014

DOI: 10.1007/s00773-013-0251-0

CITATIONS READS

7 65

2 authors, including:

Apostolos Papanikolaou
Hamburgische Schiffbau-Versuchsanstalt GmbH
369 PUBLICATIONS 1,630 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

NEWDRIFT View project

WAVELODS View project

All content following this page was uploaded by Apostolos Papanikolaou on 12 September 2014.

The user has requested enhancement of the downloaded file.

J Mar Sci Technol (2014) 19:327–337
DOI 10.1007/s00773-013-0251-0
ORIGINAL ARTICLE
On the time for the abandonment of flooded passenger ships due
to collision damages
Dimitris Spanos • Apostolos Papanikolaou
Received: 27 June 2012 / Accepted: 28 December 2013 / Published online: 21 January 2014
Ó JASNAOE 2014
Abstract The time dependency of the survivability of
passenger ships in flooding accidents leading to ship’s
loss is shown to be confined within short times after the
flooding initiation. RoRo ferry and cruise type ships
demonstrate similar behaviour, though vessel types differ
with respect to the subdivision and the flooding process.
The presented research is based on numerical simulations
of the flooding of two sample passenger ships in collision
damages and the estimation of the probability to capsize.
The systematic fast character of capsize events is shown
to be primarily a consequence of the extent of hull
breaches. The IMO regulatory concept for orderly
abandonment of damaged passenger ships, in addition to
the safe return to port regulatory provisions, are discussed
in view of the present results. The timely onboard damage
identification by ship’s master and the enhancement of the
SOLAS survivability requirements for passenger ships
appear to be the most effective operational and design
measures leading to improved survivability of the ship
and to enhanced safety of people onboard.
Keywords Survivability  Flooding  Evacuability 
Damage stability  Time to capsize  Passenger ships
Dimitris Spanos formerly worked with Ship Design Laboratory,
NTUA.
D. Spanos
Athens, Greece
A. Papanikolaou (&)
Ship Design Laboratory of the National Technical University of
Athens (NTUA-SDL), Heroon Polytechniou 9, 15773
Athens, Greece
e-mail: papa@deslab.ntua.gr
1 Introduction
The time duration of probable loss of passenger ships
after the initiation of flooding when losing their water-
tight integrity due to a collision or other hazards has
gained significant research interest in the past decade
in parallel with related international regulatory
developments.
A series of research teams worldwide dealt with the
development of numerical simulation methods and
computer codes determining the behaviour of damaged
ships in waves and estimating the probability of ship’s
capsize and loss in case of (mainly) deterministic and
(rarely) probabilistic damage scenarios and irregular sea
excitation. Some important studies focusing on the time
to ship’s capsize/sinking after flooding, are given next,
though the list cannot be exhaustive: Santos et al. [1, 2],
Jasionowski et al. [3], Van’t Veer et al. [4] and Spanos
et al. [5, 6]. Characteristically, all above studies, except
for Van’t Veer et al. [4], refer to applications to the
capsize of RoRo passenger ships in relation to water on
deck (flooding) problems, though conceptually related
numerical methods and computer codes may be adapted
to the flooding of any type of ship, noting that the com-
plexity of flooding of multiple spaces as found on cruise
ships needs special care. It should be noted that compared
to the estimation of the probability of capsize or sinking
of a ship in case of a collision flooding (assuming a
deterministic damage scenario), the estimation of the
probability density function of the time to capsize (TTC)
of statistically likely collision damage scenarios over
ship’s life cycle (probabilistic damage scenarios) adds
complexity to problem and requests reliable and fast
numerical simulation methods. A review of methods and
software tools related to the simulation of ship’s damage
123
328
stability in waves may be found in Chapter 5 of Papa-
nikolaou (ed) [7].
This article evaluates the probability of slow (and fast)
sinking/capsize of passenger ships (of RoRo ferry and
cruise type), aiming to clarify the feasibility of safe
operations/evacuation in flooding incidents. The present
investigation concerns side collision damages in view of
SOLAS damage stability regulations. First, a brief review
of the basic developments at IMO is given to stress the
leading concept of orderly abandonment as well as the
required time for safe evacuation. Next the ensuing
probabilistic method for the assessment of the probability
of time to capsize is presented together with all the
probabilistic assumptions for the collision damage sce-
narios. Following the application of the set-up probabi-
listic model to two sample ships (one RoPax and one
cruise ship), the interpretation of the obtained results,
their impact on current regulatory provisions and the
future perspective of the relevant research is discussed in
view of the present findings.
2 Review of regulatory developments regarding
the available time for the abandonment of passenger
ships
The IMO regulatory concept of safe return to port
became mandatory for passenger ships (of length 120 m
and over) with the resolution MSC.216 [8], through the
SOLAS Regulation II-1/8-1: ‘‘A passenger ship shall be
designed so that specified systems remain operational
when the ship is subject to flooding of any single
watertight compartment’’.
This regulation implements the first part of the guiding
approach of IMO to passenger ship’s survivability after
flooding, according to which ships should be designed for
improved survivability based on the concept that ‘‘a ship is
her own best lifeboat’’. Depending on the casualty extent,
passengers and crew should be able to either stay safely on
board while the ship would always be able to proceed back
to port at a minimum safe speed or, to orderly evacuate and
abandon the ship, if this is deemed necessary. Thus, two
options are offered to ship’s master for decision in case of
emergency following flooding:
a. Safe return to port
b. Orderly evacuation and abandonment
The need to enhance the safety of passenger ships was
adopted by IMO in response to justified concerns (of late
1990s) regarding the timely evacuation of the large
passenger ships in emergency situations and in view of
their growing size and the huge number of people on
board, particularly for post PANAMAX cruise ships,
123
J Mar Sci Technol (2014) 19:327–337
which have reached capacities of more than 8,500
people.1
Crucial technical parameters in these regulatory devel-
opments are the assumed casualty threshold, namely the
assumed damage extent, and the time frame, namely the
time the ship remains viable and operative after damage.
As long as the damage extent does not exceed the specified
threshold (of one watertight compartment according to the
safe return to port provisions) then passenger ships should
survive infinitely and be able to proceed to a port, whereas
when this threshold is exceeded (i.e. breach of two or more
compartments) then sufficient time should be available to
allow a safely and orderly evacuation and abandonment of
the damaged ship.
So far the 3 h survivability has been assumed indirectly
as a minimum time criterion for the orderly abandonment,
which starts with the initiating flooding event and lasts
until all persons have abandoned the ship, IMO, MSC
78/26 [8]. This time duration is also assumed for the safe
return to port provisions, IMO, MSC 82-24-Add.1 [9].
Meanwhile, the general time requirement for the muster-
ing, evacuation and abandonment is set to 60 min for
passenger ships, or 80 min for non-RoRo ships with more
than three main vertical (fire) zones, IMO, MSC.1/
Circ.1238 and 1369 [10, 11].
Thus, the time (either as time to sink or to flood or to
capsize TTC, or survival time and regardless of the partic-
ular differences) has become a crucial parameter for the
safety of people on damaged passenger ships and forms a
major ship design objective, especially for larger passenger
ships; though TTC is difficult to determine with certainty
(see relevant discussion in SLF 48/6/2 [12]), the compliance
of the ship design with the orderly evacuation requirement
by ensuring sufficiently slow sinking conditions needs to be
demonstrated to the satisfaction of flag authorities.
While the safe return to port provisions are currently in
the phase of development of guidelines for implementation
(SLF 54/17 [13]), the orderly abandonment is still an open
challenge for designers, researchers and regulators; it is
viewed as a long term R&D objective, requiring further
progress in the knowledge about the time-dependent sur-
vivability and particularly the relationship of the time with
the design characteristics of ships, including the life saving
appliances (see, project FLOODSTAND [14] and Jasio-
nowski [15]).
1
It should be noted that since the 70s, the size of cruise ships
(measured in GT) has doubled in size (and correspondingly the
number of people onboard), whereas the requirements of relevant
damage stability regulations did not exceed the deterministic SOLAS
90 two compartment standard (with max. damage length of merely
11 m), until the introduction of the probabilistic SOLAS 2009 on
January 1, 2009 for new buildings, which should be on average
equivalent to SOLAS 90.
J Mar Sci Technol (2014) 19:327–337
Systematic studies by the authors on the time to capsize/
sink of RoRo passenger ships (Spanos et al. [6, 16]) indi-
cated that in cases of ship’s loss, the orderly abandonment
of this type of ships is practically not feasible. We extend
this research herein to cruise ships to demonstrate that they
exhibit similar characteristics for the time-dependent sur-
vivability in flooding.
3 Probabilistic analysis of time to capsize
3.1 Probability simulations
Past ship accidents demonstrate that the time a passenger
ship takes to sink may be notably scattered, e.g. approxi-
mately 20–30 min for a RoRo ferry, e.g. MV Estonia
(1994), MV Express Samina (2000), to 15 h for cruise ships,
e.g. MS Sea Diamond (2007). Though the time scatter is not
yet well attributed, this variation mostly depends on the
particulars of the accident, the extent of hull breach through
which flooding occurs and the status of ship’s watertight
doors (closed according to SOLAS or left open).
For a specific accident the expected time variation is
rather limited as then most of the determining parameters are
also specific (quasi-deterministic problem, except for the
uncertainties of some parameters). But when considering the
survivability of a ship throughout her life cycle, at the stage
of ship design, then a generic random hazard and design
environment that accounts for the wide range of probable
hull damages should be assumed. This results to a much
wider variance for the capsize time (compared to a specific
accident). This wider variance is a characteristic of each
damaged ship and is evaluated in this research for two
sample vessels, one RoRo ferry and one cruise ship.
The time to capsize for a given ship is assumed as a
random variable depending on the random environmental
conditions during the flooding incident, the random shape
and location of the hull breach, the variation and uncer-
tainties of ship’s loading (e.g. KG actual) and local details
of the flooded spaces. Its statistical probability distribution
(when capsize is a possible event) can be directly approa-
ched with a basic Monte Carlo MC simulation, where the
time to capsize is sampled from a deterministic time
domain simulation for the ship flooding and for a suffi-
ciently large number of damage cases to meet statistical
convergence of the results. The damage cases are deter-
mined by the damage openings on the hull, the ship loading
condition (D, GM) and the significant wave height of the
sea waves. The dimensions of the opening and the ship
weight (displacement and centre of mass) and character-
istics of the waves follow probability distributions
according to the assumed design scenarios. Thus, each
sample damage case is determined with a random vector
329
variable (xd, Ld, Bd, Hd, D, KG, Hs). As soon as a sufficient
number of cases are sampled, then the statistical proba-
bility of the time to capsize can be calculated through the
encountered frequencies as below.
P
½cases with ðTTC\tÞ
F ðt Þ ¼ ð1Þ
ðtotal sampled casesÞ
For the present investigation samples of 5000 damage
cases were considered to achieve a satisfactory conver-
gence at the level of probability distributions.
With this approach the domain of possible damage cases
is mapped to the sample space of the random variable of
the time to capsize. Thus, the probability distribution can
be statistically estimated over the sample space. The two-
step function f for the probability simulation is
q p
f : D ! X ! <
where D domain of damage cases
X sample space of time to capsize
q mapping of domain
p probability measure
3.2 Deterministic flooding simulation
For specific damage cases, the time to capsize may be
deterministically estimated with a broadly established
numerical simulation method which is based on first prin-
ciples for the ship’s flooding in the time domain. The ship
flooding problem might be formulated at different levels of
detail, including even higher analysis with CFD techniques
for the local flow through openings. However, such an
approach is computational heavy and not practical for the
probabilistic framework herein addressed. Alternatively, an
efficient formulation for the flooding of single or multiple
compartments can be based on simplified hydraulic models
and the modified Bernoulli equation, an approach which is
adopted by several researchers and the authors of this paper
(see, Papanikolaou et al. [7] for overview). Nevertheless,
this simulation approach is still computationally demand-
ing and results to limited applicability to large-scale
probabilistic analyses or complex internal arrangements;
however, for the present research purposes the computa-
tional efficiency is of low significance.
In the deterministic simulation, the governing process for
the ship flooding and any possible progressive flooding of the
interconnected spaces is the water flow through the assumed
damage openings. The complex arrangement of ship’s
watertight compartments is modelled together with the inter-
connecting openings/watertight doors etc. Openings on the
hull’s outer shell are assumed to be the result of side collision
damage, through which ship’s initial flooding occurs.
The floodwater mass inside each flooded compartment is
time variable and results from the sum of water flux
123
330 J Mar Sci Technol (2014) 19:327–337

through all the interconnected openings. The water flux
across each opening is assumed to be described by a
modified Bernoulli equation, considering some linear dis-
charge coefficient cd, which accounts for all deviations
from the ideal water flow. Then the elementary flux dq is
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
dqðtÞ ¼ cd signðDHÞ 2gjDH jdA ð2Þ
where DH is the difference of water heads in both sides of
the assumed opening and dA the elementary opening area.
Here, plane quadrilateral openings were considered and
constant (=0.65) discharge coefficients were assumed.
For the modelling of the flooding of cabins and
accommodation areas in general, as found on passenger
ships, with multiple interconnecting openings, it is neces-
sary to group a number of cabins into a larger block and to
create an equivalent opening reflecting the total area of all
interconnections (see, Van’t Veer et al. [4]).
The ship will change her position due to the weight
distribution of the accumulated floodwater inside the
compartments. The nonlinear motion of the ship under the
action of the floodwater, and the flooding of the compart-
ments were calculated in coupled way in the time domain,
using NTUA-SDL’s code CAPSIM (Spanos et al. [5]). For
each examined damage scenario, a capsize event was
assumed to occur, when the heel angle exceeded 45°.
Table 1 Main dimensions of the study case passenger ships
Dimension RoRo ferry Cruise ship
Lpp (m) 137.4 216.8
B (m) 24.2 32.2
T (m) 5.4 7.2
DDECK (m) 9.1 9.8
Displ. (tn) 12,300 35,000
4 Study cases of two passenger ships
Two sample passenger ships were considered in the present
investigation, one RoRo ferry (which corresponds to sun-
ken MV Estonia) and a PANAMAX cruise ship (a design
concept introduced in the FLOODSTAND 2009–12, EU
research project [14]). The RoRo complies with past (at
that time in force) stability requirements (SOLAS’74) and
is characterised by a single, undivided car deck of maxi-
mum exploitable area. The sample cruise ship concept was
developed by the project’s yard partner Meyer Werft,
Papenburg in compliance with latest damage stability
regulations (SOLAS’09). The main dimensions of the two
studied ships are given in Table 1.
As shown in Figs. 1 and 2 both vessels have longitudinal
subdivision arrangement comprising of fifteen watertight
compartments abaft the collision bulkhead. The bulkhead
deck is that of the vehicle deck for the RoRo ferry and deck
No. 3 (blue line) for the cruise ship.
For the RoRo ferry, the modelling of superstructures is
extended up to one deck above the car deck. The car deck
space is assumed watertight and the lower compartments
are connected to the upper spaces through the central
casings located on the car deck. For the modelling of the
cruise ship all decks up to No. 7 (three decks above the
bulkhead deck) were taken into account. Thereby, with the
present modelling (of decks) each watertight compartment
(as defined with the transverse bulkheads) is further sub-
divided, mainly vertically, into rooms, which are inter-
connected with openings that correspond to doors,
stairways and lift wells, whereas all watertight doors on the
transverse watertight bulkheads were assumed closed as
required by SOLAS.
The modelled rooms were treated as unrestricted spaces
with homogeneous permeability. Inner boundaries (like

Fig. 1 Subdivision layout of

sample RoRo ferry

123
J Mar Sci Technol (2014) 19:327–337
Fig. 2 Subdivision of the
sample cruise vessel
A-class) were not modelled, though such boundaries might
slow down the flooding in function of the strength (in
deformation and collapse) of the non-watertight doors and
openings present on these boundaries, with evident effect
in lower amounts of floodwater. Since the collision dam-
ages are assumed to extend along the full draught, the
particular impact of inner boundaries on the time to flood
should be limited to the later stages of flooding.
Some design modifications of the original subdivision
layouts were also considered (as possible Risk Control
Options), which were applied above the bulkhead deck and
which may affect the later stages of ship’s flooding too. In
this way, four watertight side casings were introduced on
the car deck of the RoRo ferry (dashed lines in Fig. 1). For
the cruise ship, three transverse bulkheads on deck No. 5
could extend upwards the main vertical zones by two decks
above the original bulkhead deck (double line on deck No.
4 of Fig. 2).
5 Domain of collision damage scenarios
Damage cases that result to ship’s flooding are determined
by the assumed damage openings on the hull shell, the
ship’s loading condition and the sea waves prevailing
during ship’s flooding incidents. These parameters were
assumed as random variables. The assigned probability
models for the dimensions of hull opening and the signif-
icant wave height correspond to collision damages, as they
were defined by the HARDER (2000–2003) [17] research
project and which form the background of the current
SOLAS’09 damage stability regulations.
Regarding the impact of seaway, the assumed probabi-
listic model for the significant wave height in collision
damage scenarios was also defined in the HARDER pro-
ject; it was actually deduced from collision statistics for all
ship types and may be conservative for the operational
profile of passenger ships; in the assumed modelling the
significant wave height is distributed up to 4.0 m, while a
331
1.0
0.8
0.6
CDF
0.4
SIGNIFICANT WAVE HEIGHT
0.2 in collision damages, HARDER 2003
0.0
0 1 2 3 4 5
Hs (m)
Fig. 3 Probability of significant wave height in collision accidents
notable percentage of more than 30 % of collision acci-
dents regards collisions in calm water (Fig. 3). If we add
the low wave height conditions (e.g. significant wave height
\1.5 m)2 to calm water, then the corresponding percentage
of collisions remarkably rises to 80 %, hence calm water
condition damage scenarios are the major contribution to
ship’s probabilistic damage stability.
The assumed location and dimensions of the collision
damage openings are shown in Fig. 4.
The damage location is assumed evenly distributed
along the ship length. The damage length linearly
decreases up to 15 % of ship’s length and exhibits
lower probability for damage lengths up to 30 % of the
ship’s length. Since this distribution is almost of tri-
angular shape, the average damage length is obviously
close to 5 % of ship’s length. A difference of this dis-
tribution with respect to the one adopted by SOLAS
2
It should be noted that ‘low wave conditions’ in terms of absolute
significant wave heights (registered in collision accidents) and their
impact on ship motions (as well as the likely flooding of high decks)
practically cover a much wider range of wave conditions, namely
more than 90 % of the collision cases, when dealing with very large
passenger vessels, e.g. large RoPax and cruise ships in general.
123
332 J Mar Sci Technol (2014) 19:327–337

Fig. 4 Location and 2.0 12

DAMAGE LOCATION from AP
dimensions of collision damage DAMAGE LENGTH
For Side Collisions, HARDER (2003) 10 For Side Collisions, HARDER (2003)
openings according to 1.5
HARDER 8

PDF

PDF
1.0 6

4
0.5
2

0.0 0
-0.20 0.00 0.20 0.40 0.60 0.80 1.00 1.20 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
(x/Ls) Ld / L

5.0 0.20
4.5 DAMAGE PENETRATION DAMAGE HEIGHT above the draught
For Side Collisions, HARDER (2003) For Side Collisions, HARDER (2003)
4.0 0.16
3.5
0.00, 3.20
3.0 0.12
PDF

PDF
7.80, 0.103
2.5
2.0 0.08
1.5
12.50, 0.043
1.0 0.50, 0.80
0.04
0.5
0.0 0.00
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0 5 10 15
Penetration / B (Hd - T) (m)

2009 is that the maximum damage lengths are in
SOLAS 2009 limited to 60 m, which generally affects
ship larger than 200 m. The present cruise ship size
(216 m) is close to this boundary therefore the differ-
ences are limited.
The damage penetration is also linearly decreasing up to
half breadth of the ship (up to ship’s centreplane), while
vertically the damage is extending from the baseline
upwards up to above ship’s waterline according to the
shown double-step function and is limited to below
12.50 m from the waterline.
These distributions were here assumed to have been
generated by independent random variables; however,
combinations of very deep penetrations associated with
very small damage area on the shell were discarded as non-
realistic events.
The probability of ship’s displacement (loading condi-
tions) was distributed between three discrete conditions
and in line with current SOLAS 2009 regulation, namely
40 % for each of the full and the partial draughts and 20 %
for the light draught condition.
The distribution of ship’s initial stability (intact meta-
centric height) was empirically assumed (Fig. 5). The
indicated range of GMs should represent some wider
operational conditions for ship’s life cycle. Both theoretical
distributions are single peak models with the peaks close to
the nominal (design) values of the two studied ships; for
the RoPax ship a smaller GM was herein assumed
reflecting ship’s SOLAS’74 damage stability standard in
compliance, compared to that of the cruise ship, which is
designed with SOLAS’09 compliance; in the distribution,
some variation was included which aims to cover the
operational range of loading conditions and extreme situ-
ations with lower probability. However, because of the
nature of the investigated problem, the increased variance
in the GM did not have a notable impact on the findings.
6 Probability of time to capsize
The probability distribution of the time to capsize for the
two passenger ships, considering the subdivision modifi-
cations (applied Risk Control Options), was estimated with
the outlined probabilistic simulation approach and the
overall findings are summarised in Figs. 6 and 7.
These distributions correspond to ship capsize condi-
tions, whereas for the survival conditions the time is infi-
nite. The probability to capsize within 3 h from the
collision incident converges to 4.9 ± 0.5 and 2.4 ± 0.3 %
for RoPax and cruise vessel, respectively, whereas a 10 %
uncertainty is considered associated with the MC simula-
tion results.3 This is a remarkable outcome, which may be
interpreted as rationally evaluated attained subdivision
indices of A = 0.951 and A = 0.976 for the studied RoPax
and cruise ship, respectively, though their corresponding
attained indices calculated according to SOLAS’09 are
much lower. This indicates that the actual safety level of
passenger ships may be much better than that assumed by
present rules.
3
The uncertainty refers to the confidence limit regarding the
accuracy of the obtained results.

123
J Mar Sci Technol (2014) 19:327–337 333

1.6 12 1.0
RoRo ferry
10 PDF (mod)
0.8
1.2 PDF (org)
CDF(mod)
8
CDF(org) 0.6

CDF
PDF
PDF

0.8 6

Cruise 0.4
4
0.4 RoPax
0.2
2

0 0.0
0.0
0.00 0.05 0.10 0.15 0.20 0.25 0.30
0.0 1.0 2.0 3.0 4.0
intact GM (m) Ld / L

Fig. 5 The assumed probabilities for GM of the intact ships Fig. 8 Estimated probability of the damage length for RoRo ferry,
when it capsizes in calm water conditions—original and modified
design scenarios
0.30 1.0
RoRo ferry
notably shorter than the regulatory provisions for the orderly
0.25 PDF (mod)
PDF (org)
0.8 abandonment of the passenger ships in case of flooding.
0.20
CDF(mod) The effect of the assumed subdivision modifications
CDF(org) 0.6 (applied RCOs) on the time to capsize appears insignificant
CDF
PDF

0.15 (red dashed lines of Figs. 6 and 7). The particular design
0.4 modifications are mainly effective in later stages of ship
0.10
flooding, and since here the capsize events are occurring in
0.05
0.2 short times this effect is eventually limited; nevertheless, it
is still an evidence of the sensitivity of the resulting TTC
0.00 0.0 on the examined design parameters.
0 5 10 15 20 25 30
TTC (min)
Capsize events for the RoRo ferry occur strictly only
for the cases when the car deck gets damaged. When the
Fig. 6 Probability of the Time to Capsize for the damaged RoRo hull breach is limited to below this deck, then this large
ferry, when it capsizes in calm water conditions—original and intact car deck volume forms a large reserve bouyancy
modified design scenarios
and contributes significantly to ship’s floatability and
stability; thus the ship capsize is not feasible any more.
0.30 1.0 Note that up-flooding might happen here only through
CRUISE
the central casing opening (Fig. 1).
0.25 PDF (mod)
PDF (org)
0.8 The time to capsize is distributed within ranges of
0.20
CDF(mod) comparable width for both ships, though the cruise ship is
CDF(org) 0.6 three times larger (in displacement) than the RoRo. This is
CDF
PDF

0.15 related to the size of relevant damage openings, which are

0.4 proportional to ship’s length and then to the corresponding
0.10
flooding rates, because of the dimensionless distribution of
0.05
0.2 Fig. 4. For the RoRo ferry, the actual (absolute) damage
sizes are 36 % smaller than the damages assumed for the
0.00 0.0 cruise ship.
0 5 10 15 20 25 30
The corresponding distribution of the damage length
TTC (min)
for the capsize events is plotted in Figs. 8 and 9, which
Fig. 7 Probability of Time to Capsize for the damaged cruise ship, may well explain the short times that were recorded for
when she capsizes in calm water conditions—original and modified both ships. The capsize cases are obviously related to
design scenarios large damage openings only. The average damage length
for the capsize cases converges to 0.19 L and 0.18 L for
Focusing on the detailed simulation results for the time to the two vessels, respectively, compared to the assumed
capsize (TTC), the capsize events for both ships appear overall average damage length of about 0.066 L, while
limited to within 15 min. Such time is rather short and is 8 % of the recorded cases exceed 0.15 L as may be

123
334 J Mar Sci Technol (2014) 19:327–337

12 1.0
CRUISE
10 PDF (mod)
0.8
PDF (org)
CDF(mod)
8
CDF(org) 0.6

CDF
PDF

6
0.4
4

0.2
2

0 0.0
0.00 0.05 0.10 0.15 0.20 0.25 0.30
Ld / L
Fig. 9 Estimated probability of the damage length for cruise ship,
when she capsizes in calm water conditions—original and modified
design scenarios
inferred from Fig. 4. Such large damage lengths gener-
ally correspond to multiple damaged watertight com-
partments, namely three or more. Thereby, ships capsize
for the large openings only, and then flooding is trivially
fast (due to large damage areas) and the recorded times
are short accordingly.
In earlier investigations, e.g. (Spanos et al. [5, 16]), the
fast capsize was shown to be a characteristic behaviour of
damaged RoRo ferries. These ships are vulnerable to the
flooding of their car deck, which can take place only in
high damages and in the presence of waves exceeding
some critical wave height. As soon as this critical wave
height is exceeded, then the time to capsize rapidly short-
ens. Thus, the time-dependent survivability in waves is
limited to short times similarly to the present results.
Cruise ships, differently to RoRo ferries, usually do not
have large undivided spaces near the sea surface, like the
car deck of RoRo ships, and the significance of waves for
the distribution of the time to capsize is expected to be
limited. The flooding process for well subdivided ships
follows a sequential flooding of rooms in a cascade mode.
The sea waves may not alter the qualitative characteristics
of the distribution of Fig. 7. This argument is further
supported by the low probability of collision in high sea-
ways, as discussed in the previous section for the passenger
ships and this argument applies particularly to cruise ships,
because of their unique operational profile.
Capsize in calm water is trivially correlated with zero or
very low residual stability, as detailed with Figs. 10 and 11.
Approximately 80 % of the damage cases were fully
unstable (with practically zero stability, GZmax less than
about 0.010 m) while for the rest cases GZmax was also low
and below 0.10 m. This limit is directly comparable to the
current limit of 0.050 m used in SOLAS 2009 (Ch. II-1,
Part A, Reg. 7.2) to define the full survival capabilities for
intermediate stages of flooding. The present results indicate
Fig. 10 Probability of GZmax for RoRo ferry, when it capsizes in
calm water conditions—original and modified design scenarios
that current regulation may miss only 10 % of the likely
intermediate stages of flooding capsizes (i.e. the proba-
bility of exceedance of GZmax of 0.050 m).
7 The time to capsize jump phenomenon
In the previous sections, it was shown that short times to
capsize are connected to multiple damaged compartments.
This section elaborates further on this point by focusing on
the characteristic behaviour of the time to capsize at the
range of damage lengths that marginally cause three
compartments damage.4
A characteristic single compartment damage case for the
studied cruise vessel in full load condition is detailed,
which regards the collision incidents associated to damage
at mean location ?13 % from midship (position A in
Fig. 2). This location is defined in the middle of the
respective watertight compartment and as long as the
damage length is less than the compartment’s length, then
one-compartment damage occurs and the flooded ship
survives infinite time. For damage lengths greater than the
compartment’s length then the ship capsizes in fast mode
with three adjacent compartments flooded.
The time to capsize changes from infinite to finite short
time within a very narrow range of damages that corre-
sponds to the marginal situation, when the damage length
just exceeds the compartment length. The joint probability
of the time to capsize and damage length is depicted in the
next Fig. 12 (x-axis: time to capsize in logarithmic scale, y-
axis dimensionless damage length) for this scenario. The
estimated probability TTC is distributed over the threshold
length of one-compartment damage (0.065 L) and for
5 min after the start of flooding. The discontinuity of the
4
This refers to the deterministic compartment notion and not to the
zonal modeling used in the probabilistic damage stability assessment.

123
J Mar Sci Technol (2014) 19:327–337
Fig. 11 Probability of GZmax for Cruise ship, when she capsizes in
calm water conditions—original and modified design scenarios
Fig. 12 Joint probability of time to capsize TTC and damage length
Ld/L in the region of exceedance of one compartment damage
probability distribution defines the lower boundary of the
distribution. Longer times approaching 100 min may be
still observed, but only within the very narrow region of
exceedance of the threshold damage.
Figure 12 demonstrates the characteristic jump of time
at the region of the threshold damage, where the infinite
time below the threshold sharply changes to finite short
times above the threshold.
The interesting point in this behaviour is that the jump
occurs for a very narrow range between survival and non-
survival conditions; this herein theoretically corresponds to
an increase of the damage length by merely 5 cm, indicated
with the arrow in Fig. 12, what practically corresponds to
the damage of the two boundary watertight bulkheads. The
existence of this threshold damage could be well expected
prior to the analysis, but the width of jump has been not
addressed, yet, by others.
The jump of time is the characteristic behaviour for the
transition between survival and non-survival conditions for
the other damage cases too of the cruise vessel. Taking this
into account, the probability of longer times to capsize is
highly limited, whereas slow capsizes may only occur
335
rarely or if ship’s watertight compartmentation is com-
promised by improper operation, e.g. by partly open left
watertight doors and/or of other openings.
8 Research perspectives
The obtained TTC results (Figs. 6, 7) suggest that when a
capsize event proves likely for a passenger ship, then the
ship should be urgently abandoned, as fast as possible.
Capsize occurs in remarkably shorter times than the
required time for an orderly abandonment according to
current regulations. The 3 h objective for orderly aban-
donment or even the 1 h requirement for evacuation that is
closer to the actual evacuation times, are both much longer
times. Thus unconditionally, the orderly abandonment
appears infeasible for flooding accidents leading to capsize,
while optimistically, it might be only partially accom-
plished (namely only for a percentage of people on board).
Currently the objective for orderly abandonment regards
damages which exceed the threshold of one compartment
for the safe return to port. The present research points out
that loss of passenger ships is a fast event and is connected
to three or more compartment damages, whereas for two
compartment damages ships of SOLAS 90 standard do not
capsize and then the abandonment in view of flooding is
unnecessary in principle, unless open left watertight doors
may compromise ship’s survivability.
Therefore, the timely onboard identification of the actual
damage case, namely whether it involves two or three
compartments, proves to be crucial for the onboard safety
assessment. The timely reliable identification of survival
conditions by an onboard decision support system would
allow to keep people onboard as much and as long it is
assumed safe, instead of ordering abandonment and send-
ing people unnecessarily to the sea, where they might be
exposed to new risks due to the hostile sea environment of
high waves and/or low/freezing temperatures.
However, when considering actual damage incidents,
then a reliable onboard assessment is today still a technical
challenge, as the onboard and timely identification of the
actual damage extent and of the related survivability suf-
fers by remarkable uncertainty. To the extent this uncer-
tainty can be minimised, then the estimated ship’s survival
time converges to either too long (survival conditions) or
short time (capsize conditions). Independently of the
identification and assessment capabilities the situation
appears practically binary in terms of time.
The difficulties for the identification are even more
stressed if we take into account that the survivability of the
damaged ships is considered within a probabilistic frame-
work, like the current SOLAS regulations, where deter-
ministic casualty thresholds are not applied (namely full
123
336
survivability for two or three adjacent watertight com-
partments are not required). If such deterministic thresh-
olds would be applied, then the probabilistic principles and
all related advantages related to the flexibility of ship
design would be practically abandoned.
Designing ships with time-dependent survivability and
providing capsize times longer than the present time
requirements appear practically an infeasible objective.
Instead, the improvement of ship’s survivability by shifting
present safe boundaries to higher levels, namely by
enhancing the typical survivability of the damaged ship
through increased survivability requirements appears a
feasible objective. Then the probability for encountering
critical damage scenarios requesting ship’s abandonment
would be greatly reduced and the survival scenarios further
increase.
It should be noted that the presented research regards
side collision damages, and similar analysis should be
extended to other type of hazards causing flooding, like
groundings, for completeness and further strengthening of
the findings. Nonetheless because of the fact that ships’
capsize occurs trivially for large damages, which are con-
nected with large flooding rates, it could be early inferred
that the capsize due to large grounding damages, pene-
trating ship’s double bottom and allowing the flooding of
multiple compartments, will be accordingly of short time.
9 Conclusions
The presented investigation explored the general proba-
bility distribution of the time to capsize of two passenger
ships of different types in collision damages and was
concerned with the detection of possible conditions related
to the rapid character of ship capsize. Despite the fact that
the two investigated ships were of particular design and
stability standard, the obtained results are generic in view
of their fundamental character.
The results demonstrated that sinking and capsize of
passenger ships occur systematically in short times after the
damage incident, as a consequence of the fact that ship’s
loss is generally related to excessive size of damages. The
estimated times seem to be by far less than those assumed in
current regulatory provisions for the orderly evacuation and
abandonment of passenger ships, at least for the subdivision
arrangements tested, which are according to the regulations
in force at the time of ships’ building.
The characteristic jump of time in the transition region
between survival and non-survival conditions clearly sug-
gests that the probability of slow capsize is highly limited,
unless other factors (e.g. partly open left watertight doors
or other openings) introduce more complexity into the
physics of ship’s flooding.
123
J Mar Sci Technol (2014) 19:327–337
The survivability of damaged passenger ships seems
practically to be time-independent and an improvement of
survivability seems not feasible through the control of the
capsize time, but rather ensuring by design measures and
ship’s subdivision that the ship can sustain certain extent of
possible damages. The timely and reliable onboard iden-
tification of the actual extent of damage proves to be the
most important parameter for decision making regarding
ship’s evacuation and towards an improved safety of peo-
ple on passenger ships.
Taking into account current technical difficulties for
the reliable onboard damage identification, the only
effective alternative seems to aim for higher survivability
requirements (subdivision required index) as an effective
counter-measure to the risk of fast capsize (see recom-
mendations of project GOALDS to IMO-SLF55, Febru-
ary 2013 [18]).
Closing this article it is recalled that the present analysis
concerns the collision hazard as strictly defined through the
damage statistics assumed from the current damage sta-
bility regulations for passenger ships (SOLAS’09). Other
hazards leading to ship’s flooding, like groundings, should
be also investigated to complete the research for the time
characteristics of the sinking process of passenger ships
due to flooding.
Acknowledgments The financial grant of this research from the
European Commission research projects FLOODSTAND (Integrated
Flooding Control and Standard for Stability and Crises Management),
FP7-RTD-218532 and GOALDS (Goal-Based Damage Stability),
FP7-RTD-233876, is acknowledged. The European Commission and
the authors shall not in any way be liable or responsible for the use of
any knowledge, information or data presented, or of the consequences
thereof.
References
1. Santos TA, Winkle IE, Guedes Soares C (2002) Time domain
modeling of the transient asymmetric flooding of Ro–Ro ships.
Ocean Eng 29(6):667–688
2. Santos TA, Guedes Soares C (2009) Numerical assessment of
factors affecting the survivability of damaged Ro–Ro ships in
waves. Ocean Eng 36(11):797–809
3. Jasionowski A, Vassalos D, Guarin L (2004), Theoretical
developments on survival time post-damage. In: Proceedings of
the 7th International Ship Stability Workshop, Shanghai,
November
4. Van’t Veer R, Peters W, Rimpela A-L, de Kat J (2004) Exploring
the influence of different arrangements of semi-watertight spaces
on survivability of a damaged large passenger ship. In: Pro-
ceedings of 7th Ship Stability Workshop, Shanghai, November
5. Spanos D, Papanikolaou A, Zaraphonitis G (1997) On a 6 DOF
mathematical model for the simulation of ship capsize in waves.
In: Proceedings of the 8th International Congress on Marine
Technology (IMAM), Istanbul
6. Spanos D, Papanikolaou A (2007) On the time to capsize of a
damaged RoRo/passenger ship in waves. In: Proceedings of the
J Mar Sci Technol (2014) 19:327–337
9th International Ship Stability Workshop, Hamburg, Germany,
August 30–31
7. Papanikolaou A (ed) 2009 Risk-based ship design, methods, tools
and applications. Springer, New York. ISBN 978-3-540-89041-6
8. IMO, MSC 78/26 (2004) Report of the MSC on its 78th session,
by Secretariat-General, May 28
9. IMO, MSC 82-24-Add.1 (2007) Resolution MSC.216 (82),
Adoption of amendments to the international convention for the
safety of life at sea, 1974, as amended, Annex 2, by the Secre-
tariat, January 12
10. IMO, MSC.1/Circ.1238 (2007) Guidelines for evacuation ana-
lysis for new and existing passenger ships, October 30
11. IMO, MSC.1/Circ.1369 (2010) Interim explanatory notes for the
assessment of passenger ship systems’ capabilities after a fire or
flooding casualty, June 22
12. IMO, SLF 48/6/2 (2005) Passenger ship safety, Report of the
SDS Correspondence Group, submitted by Sweden and the
United States, June 24
View publication stats
337
13. IMO, SLF 54/17 (2012) Report to the Maritime Safety Com-
mittee, submitted by the Secretariat, February 3
14. FLOODSTAND (2009–2012), Integrated flooding control and
standard for stability and crises management, EU research pro-
ject, FP7-RTD-218532, 2009–2012
15. Jasionowski A (2011) Decision support for ship flooding crisis
management. Ocean Eng 38(14–15):1568–1581
16. Spanos D, Papanikolaou A (2012) On the time dependent sur-
vivability of ROPAX ships. J Marine Sci Technol 17(1):40–46.
doi:10.1007/s00773-011-0143-0
17. HARDER (2000–2003) Harmonization of rules and design
rational, EU funded Research Project, FP5, DG XII-BRITE
18. GOALDS (2009–2012) Goal-based damage stability, EU funded
research project, FP7-RTD-233876, 2009–2012, http://www.
goads.org
123