Proceedings of the ASME 2014 33rd International Conference onOcean, Offshore and Arctic Engineering

OMAE2014June 8-13, 2014, San Francisco, USA

OMAE2014-23323

New Insights into the Flooding Sequence of the Costa Concordia Accident

Hendrik DankowskiResearch Assistant

Institute of Ship Design and Ship SafetyHamburg University of Technology

Hamburg, GermanyEmail: dankowski@tu-harburg.de

Philipp RussellStudent

Hamburg University of TechnologyHamburg, Germany

Email: russell@tu-harburg.de

Stefan KrügerProfessor

Institute of Ship Design and Ship SafetyHamburg University of Technology

Hamburg, GermanyEmail: krueger@tu-harburg.de

ABSTRACTThe tragic accident of the Costa Concordia in January 2012

was one of the most severe large passenger ship accident in Eu-rope in recent times followed by a tremendous public interest.We present the results of an in-depth technical investigation ofthe flooding sequence which lead to the heeling and groundingof the ship.

A fast and explicit numerical flooding simulation methodhas been developed in the last years to better understand acci-dents like this one caused by complex and large scale floodingevents. The flooding simulation is validated with the help of re-sults from model tests and has been successfully applied to theinvestigation of several other severe ship accidents. It is basedon a quasi-static approach in the time domain which evaluatesthe hydrostatic equilibrium at each time step. The water fluxesthrough the openings are computed by a hydraulic model basedon the Bernoulli equation. Large and partly flooded openings aretaken into account as well as conditional openings like the open-ing, closing and breaking of doors. The fluxes are integrated inthe time domain by a predictor-corrector integration scheme toobtain the water volumes in each compartment involved in theflooding sequence.

Due to the fact that the accident happened in calm waterat moderate wind speeds close to the shore of the island Gigliothis quasi-static numerical flooding simulation can be applied.The results of the technical investigation of the Costa Concor-dia accident obtained with the help of the developed method arepresented. These results match well with the heel and trim mo-

tions observed during the accident and the chain of events whichlead to the final position of the vessel on the rocks in front of theisland Giglio.

The explicit and direct approach of the method leads to a fastcomputational run-time of the numerical method. This allows tostudy several possible accident scenarios within a short period toinvestigate for example the influence of the opening and closingof watertight doors and to identify a most likely flooding scenariowhich lead to this tragic accident.

INTRODUCTIONThe COSTA CONCORDIA sank in the early morning of Jan-

uary 14, 2012 near the island of Giglio in the Mediterranean Sea.A few hours earlier she had hit an underwater rock when per-forming a tight turn at high speed very close to the shore. Asa result of this collision she drifted powerless near to the har-bour of Giglio where she grounded a second time and evacuationprocedures were started. The list gradually increased until shefinally capsized and came to rest on the rocks in shallow waters.Thirty-two of the 4229 persons on board lost their lives duringthe flooding and capsizing of the ship.

In the scope of this investigation our institute strongly co-operates with the German Federal Bureau of Maritime CasualtyInvestigation (BSU). This paper represents a short summary ofthe most interesting results of the elaborate investigation carriedout in [1].

1 Copyright c© 2014 by ASME

FLOODING MODELThe applied numerical flooding simulation determines the

fluxes through the openings by means of a hydraulic model. Thepropagation of the water volumes is computed by a predictor-corrector scheme for the integration of the volume fluxes to eachcompartment in question. Further details about the numericalmethod can be found in [2].

Flux DeterminationThe in- or egress of flood water through the internal and ex-

ternal openings can be idealized by the incompressible, station-ary and viscous- and rotational free Bernoulli equation given inEqn. 1 formulated for a streamline connecting point a and pointb:

dz =pa − pb

ρ g+

u2a −u2

b2g

+ za − zb −ϕab

g. (1)

The dissipative energy term ϕab accounts for pressure lossesthrough the openings and is assumed to be proportional to the ki-netic energy of the flow. This loss is modeled by a semi-empiricaldischarge coefficient Cd . The pressure height difference dz yieldsthe fluid velocity

u =Cd ·√

2g · dz. (2)

For a free outflow, the pressure height difference becomes:

dz =pa − pb

ρ g+

u2a

2g+ za − z0. (3)

The integration of the velocity over the area of the openingassuming a perpendicular flow direction leads to the total flux:

V̇ =∂V∂ t

= Q =∫

Au · dA=

∫Au ·ndA =

∫A

udA. (4)

The solution of this integral becomes more complicated if theopening is large and of arbitrary shape and orientation. There-fore, larger openings are discretized in smaller, elementary partsfor which an analytical solution of the volume flux can be deter-mined as described for example in [2].

Flooding PathsThe possible paths of the floodwater through the inner subdi-

vision of the vessel is described with the help of directed graphs.The openings are the edges of the graph connecting the differ-ent compartments representing the nodes. This gives a required

sign convention for the opening fluxes and allows to obtain themass balance for one compartment by the sum over all edges con-nected to one node. In addition, further algorithms from graphtheory can be applied. An example of such a flooding graph isshown in Fig. 1, which is the simplified flooding graph of the ini-tial damage of the COSTA CONCORDIA. The complete floodinggraph is much more complex.

(1) Outside(2) WB.DB.10C

(3) WB.DB.11C (4) WB.DB.12C (5) VO.DB.6C

(9) Aft D/G PS (11) ElectricMotors

(12) Refrigera-tion Compressors

(8) Fwd D/G Stairs

(6) Fwd D/G PS

(7) Fwd D/G SB

(13) Switchboard PS

(14) Switchboard SB

(10) Aft D/G SB

1

2 3 4

6 7

11

12

8

5

9 10

13

1415

1617

FIGURE 1: Flooding graph of COSTA CONCORDIA’s initialdamage [1]

Integration SchemeThe amount of water dV propagated in one time step to one

compartment from its neighbor(s) is given by the integration ofthe sum of the volume fluxes Qo(t) over all connected openings:

dV =∫ t2

t1Qo(t)dt. (5)

The propagation of water leads to a new distribution of theflood water and new volume fillings in the compartments. Thesevolumes depend nonlinear on the water levels of the compart-ments and has to be determined iteratively. The flux functionQo(t) is also a nonlinear function over time. To account for thisnonlinear characteristic, a predictor-corrector scheme is appliedfor the integration over time, which is sketched as follows:

1. Predict opening fluxes2. Propagate predicted volumes assuming a constant flux

2 Copyright c© 2014 by ASME

3. Calculate new filling levels for the new volumes4. Recompute opening fluxes based on these new fillings5. Estimate mean flux by relaxation6. Reset volumes and propagate again based on the mean flux,

update the filling levels and proceed

This scheme efficiently stabilizes the integration of the non-linear, ordinary differential equation in the time domain.

Conditional OpeningsThe flooding sequence through complex inner subdivisions

of ships is highly influenced by flooding barriers like doors. Es-pecially for the COSTA CONCORDIA accident, the time depen-dent closure and opening of the watertight doors has a high im-pact on the flooding sequence. In addition, windows or doorsonly withstand a certain pressure head before these collapse.This conditional behavior of doors and windows can be modeledby certain opening conditions and events.

The current pressure head acting on one opening is updatedat each time step. If this pressure exceeds a certain threshold, rec-ommended values can be found for example in [3], the openingbreaks and it is assumed to be open for the remaining time.

Especially during accident situations, watertight doors arenot always kept closed. The event of activating these doors areaccounted for in the numerical simulation. These are definedas events, which changes the current status of a certain openingdepending on a given condition like time or pressure head.

An very simple example for three conditions is shown inFig. 2. This shows the change of the discharge coefficient forcertain events. After 25 secs the value is lowered to 20 % in atime interval of 10 secs. The value is increased again up to 60 %at 55 secs. A breaking of the opening is modeled by the lastevent, which depends on the current pressure height.

0

0.2

0.4

0.6

0.8

1

0 20 40 60 80 100

non-

dim

ensi

onal

dis

char

ge c

oeffi

cien

t [-]

condition [s/m]

FIGURE 2: Simple example of Opening Events

This allows also to define a leakage of the opening, if thecondition is chosen appropriately. For example, the dischargecoefficient varies between 0 and 2 m from 0 to 5 % and from2 m to 2 m between 5 and 100 %, were the last event models thecollapsing of the opening.

Simulation OverviewOne time step of the simulation is sketched as follows:

1. Check conditional openings2. Pressure iteration of full compartments3. Fluxes of remaining openings4. Inner iteration for higher-order integration of fluxes5. Propagation of water volumes6. Update of filling levels and determination of full tanks7. Optional air compression8. Iteration of new floating equilibrium9. Check of convergence

This is repeated for each time-step until either the requested sim-ulation time or convergence is reached.

The pressure iteration for completely flooded compartmentsis required, because the water must be still propagated throughthe full compartment. This leads to a coupling of the neighboringcompartments. The remaining free variable here is the internalpressure of the full compartment, which is iterated until the totalflux in one time step becomes zero.

In most of the cases, the influence of air compression is ofminor importance and can be neglected. However, if required, apressure increase caused by the air volume reduction accordingto the Boyle-Marriot law is applied.

More details about additional features like the influence of acontact with the sea bottom can be found for example in [4, 2].

ValidationThe flooding simulation has been successfully validated

with the ITTC benchmark model test for the prediction of time toflood [5] as described in more detail in [6]. The obtained numer-ical results match very well with the measured ones. In addition,the described method has been applied to the reconstruction ofthree real ship accidents. The computed motion of the vesselsafter water ingress and the general time line of events matcheswell with the observations of the witnesses and the other investi-gations carried out for these accidents. Details of the validationpart and the investigations of the accidents can be found in [2].

THE COSTA CONCORDIA ACCIDENTHaving left the port of Civitavecchia near Rome on the

evening of January 13, 2012 the ship was on her way to Savona inNorthern Italy with 3206 passengers and 1023 crew on board. Enroute she changed her planned course and headed for Giglio at a

3 Copyright c© 2014 by ASME

speed of over 15 knots where she was to perform a tight starboardturn near the coast. The final course of the COSTA CONCORDIAis shown as an extract of the AIS track in Fig. 3.

Approaching the shore, she collided with the “Scole Rocks”below the waterline on her port side, which lead to the damage offive watertight compartments. These contained amongst othersthe electric propulsion motors, all diesel generators as well asthe main switchboard. The initial list was to port due to the leakbeing on that side and because of the heeling moment caused bythe rock, which stuck inside the vessel after the impact. After awhile the flooding became almost symmetrical and the ship wentupright again.

Having been damaged in compartments vital for power gen-eration, power distribution and propulsion, she was soon adriftwithout electricity. Even though the emergency diesel genera-tor started up, it did not work reliable enough to provide power.Thus, emergency power was supplied by UPS batteries. How-ever, the steering gear did not function and thrusters needed morethan the emergency power provided. Due to wind and current thevessel was eventually moved north of Giglio harbor before theforces of nature turned her around 180 degrees and pushed her inthe direction of the island until she grounded a second time.

At this time the evacuation procedures were started, whilethe heeling angle to starboard continuously increased. In theearly hours of the next morning, she had finally capsized andsunk onto the seabed approximately 25 meters deep until she wasat last raised on September 16, 2013.

Overview of the Time Line of EventsThe time line of important events are summarized in Tab. 1

to give a brief overview of the accident. Only events related to thetechnical investigation of the flooding sequence are given, fur-ther interesting events concerning passenger safety and the evac-uation procedure are omitted here. An interesting sociologicalstudy of the events can be found in [8] including a reference tothe Voyage Data Recorder (VDR) replay released by the Italianconsumer protection organization CodaCons [9].

Immediately after the accident the ship heels to port due tothe impact by the rocks below the center of gravity and due to thestarboard turn. Shortly thereafter, the blackout happens, indicat-ing the contact of flood water with vital electrical components.In some parts of the Italian investigation there is an earlier timestated for the blackout, namely at 20:45:58 UTC after 51 sec-onds. The final blackout however shows clearly after 108 sec-onds according to the monitoring of the watertight doors on theVDR [10]. Water in the Main Switchboard Room on Deck A isreported seven minutes after the collision, and another eight min-utes later three (5, 6 and 7) of the five damaged compartments areindicated to be flooded. Compartment 4 is also heavily flooded,while Compartment 8 only suffered a minor leak in the doublebottom. During the flooding of these compartments the ship lists

FIGURE 3: AIS track of the COSTA CONCORDIA [7] (giventimes are local time)

about 10 to 15 degrees to port, as can be seen in several pas-sengers’ video recordings. This is supported by the wind, whichblows from north-north-east at approximately 10 knots onto thestarboard side of the vessel.

Some 25 minutes into the accident, the list of the shipchanges from port to starboard due to the flooding becomingmore symmetrical. This coincides with the turning of the vesseland her drifting towards the island, which means that the wind isnow coming from the port side. Approximately 40 minutes pastthe collision water reaches the freeboard deck, Deck 0, while itheels the ship almost 10 degrees to starboard.

The starboard anchor is dropped one hour and a few minutesafter the accident after several failed attempts. According to thephotometric evaluation [11] the ship has a starboard list of about14 to 14.5 degrees at that time. This order of magnitude of theheeling angle is supported by the statement of a bridge officeraround the same time. The trim angle is determined to be 1.2degrees, which results in a trim of 5.19 meters to the aft. There isof course some measuring inaccuracy so that the real angles maydiffer slightly. Nevertheless these values are the most reliableones because they are not based on observations made by the

4 Copyright c© 2014 by ASME

TABLE 1: Important events of the accident [1]

UTC t Heel Event

[hh:mm:ss] [s] [deg] [-]

20:45:07 0 <0 Collision

20:46:55 108 -10 to -15 Blackout occurred

20:52:25 438 <0 Water reported Deck A

22:00:57 950 <0 Compartments 5, 6 and 7confirmed flooded

21:10:00 1493 >0 Ship begins to drift to-wards shore

21:26:30 2483 <10 Water reported Deck 0

21:48:00 3773 14 - 14.5 Starboard anchordropped, -1.2deg trim

21:55:00 4193 15 Second grounding /Lifeboats launched

22:11:26 5179 25 to 30 25 to 30 degrees star-board heel reached

23:34:00 10133 90 Capsize

crew in a stressful situation. Thus, they will serve as a referencepoint.

One hour and ten minutes subsequent to the first groundingthe vessel grounds a second time with the aft starboard side. Atthe same time the first lifeboats are launched, starting on the star-board side to reduce the list. The last heeling angle stated by thecrew on the VDR is 25 to 30 degrees to starboard at 22:11:26UTC. Almost three hours after taking the damage the vessel cap-sizes assuming a heeling angle of around 90 degrees as confirmedby a coastguard helicopter. From here on she sinks completelyand comes to rest on the rocks with a list of approximately 70degrees.

Description of the VesselThe COSTA CONCORDIA was built in 2004 for the Italian

cruise line Costa Crociere by Fincantieri in Sestri Ponente. Hermain particulars are given in Tab. 2.

Computational Data ModelThe hull form and the compartmentation is thoroughly re-

constructed. A complete definition of the hull up to the highestdeck is required for the complex flooding calculations. The hy-

TABLE 2: Main particulars of the COSTA CONCORDIA

Length over all LOA 290.20 m

Length between perpendiculars LPP 247.70 m

Breadth B 35.50 m

Depth D 11.20 m

Summer draught T 8.30 m

Service speed VS 19.60 kn

drostatic model is shown in Fig. 4.

FIGURE 4: Hydrostatic Model

In contrast to classical damage and intact stability calcula-tions, the inner subdivision has to be refined, because additionalflooding barriers like A-class walls or cabin walls have to betaken into account. The permeability is set to SOLAS standard.In addition, opening elements are defined to model doors, win-dows and other objects, which allow a water exchange betweenthe compartments. In total, 642 compartments and 1580 open-ings are included in the data model.

Loading Condition prior to SinkingThe loading condition of the vessel was reconstructed based

on the information given in [7]. All intact and damage criteriaincluding the ones for passenger vessels are fulfilled. The cor-responding righting lever arm curve for the intact condition isshown in Fig. 5. This does not include the influence of the upperdecks.

Damage to the HullThe impact with the rock caused a large and narrow under-

water damage to the hull extending over a length of approxi-mately 45 m from the bottom up to deck B. The assumed extendof damage together with immediately damaged compartments isshown in Fig. 6.

5 Copyright c© 2014 by ASME

xz

Deck 0

Deck A

Deck B

Deck C

WB.DB.10CWB.DB.11CWB.DB.12CVO.DB.6C

Fwd D/G PSAft D/G PSElectric MotorsRefr. Compr.

Swbd. PS

Fwd

D/G

Strs

.

4 5 6 7 8

FIGURE 6: Schematic side view of the real leak (blue surface) [7], the defined openings (red rectangles) and the initially damagedcompartments (blue pattern)

TABLE 3: Loading Condition of the COSTA CONCORDIA beforesinking

Shell Plating Factor 1.002 -

Density of Sea Water 1.025 t/m3

Ships Weight 54998.680 t

Longit. Center of Gravity 118.610 m.b.AP

Transv. Center of Gravity -0.010 m.f.CL

Vertic. Center of Gravity (Solid) 16.850 m.a.BL

Free Surface Correction of V.C.G. 0.344 m

Draft at A.P (moulded) 7.901 m

Trim (pos. fwd) 0.427 m

Heel (pos. stbd) 0.320 Deg.

Longit. Center of Buoyancy 118.632 m.b.AP

Transv. Center of Buoyancy -0.081 m.f.CL

Vertic. Center of Buoyancy 4.388 m.a.BL

Metacentric Height 1.711 m

The damage is modeled by eleven smaller openings, whichallow an immediate water ingress to the sketched compartments.A discharge coefficient of 0.6 is assumed for most of them, onlythe rearmost is lowered to 0.2, because the flow is obstructed bythe rock.

The Watertight DoorsThe COSTA CONCORDIA is equipped with 25 watertight

doors (WTD) to allow the crew to pass bulkheads in the water-tight part below the freeboard deck. The doors are positioned on

-1

-0.5

0

0.5

1

1.5

2

0 10 20 30 40 50 60 70

Rig

htin

g le

ve

r [m

]

Heeling angle [deg]

LC PRIOR DAMAGE starboard side

GZ [m]Requ. or Max. h: 0.971 m

Progfl. or Max.: 56.945 Deg.

GM at Equilib. : 1.711 mArea under GZ [mrad]

FIGURE 5: Intact lever arm curve of the COSTA CONCORDIA

the three decks C, B and A. The specific location of the doorscan be found in [1].

Even though, these doors shall be closed at sea, some ofthe doors were allowed by the Administration to be kept open ifdeemed necessary. These exceptions are the watertight doors 7,8, 12, 13 and 24. Any activation of a watertight door is loggedin the VDR, which data were made available after the accident.This means, that the status of each door during the accident isknown and can be taken into account for the flooding simulation.Details of the watertight door activity can be found in [10, 9].

Heeling MomentsBeside the water ingress, further aspects were identified

which had an influence on the time line of events of the accident.These are summarized as follows:

- Activation of the watertight doors- Heeling moment due to the rock- External moment caused by the wind

6 Copyright c© 2014 by ASME

- Moments exerted by the stabilizer fins- Impact of the grounding event

The activation of the watertight doors is discussed in moredetail in the results section, because it plays a very important roleto explain the observed flooding sequence.

Parts of the underwater rock, which penetrated the hull werestuck inside the steel structure. This caused a permanent heelingof about 650 tm to the port side.

At the time of the accident the wind were measured on amountain near Giglio at a height of 600 meters blowing fromnorth-north-east at 18 knots. A realistic value at sea levelamounts to 10 knots. The caused wind force on the COSTA CON-CORDIA has only little effect on the heel of the vessel, becausemost of the time the wind was blowing almost from the front [1,p. 39 ff]. The wind moment will therefore be excluded from thefurther calculations. The main influence of the wind combinedwith the current is the forced drift back to the shore, which pre-vented the vessel and its passengers from a much more severefate on the open sea.

Another aspect of interest is the influence of the stabilizer,which can be seen in several photos taken after the accident. Itis found that the moment exerted by the port side stabilizer ei-ther cancels with the wind moment or becomes so small withdecreasing speed that it almost vanishes why its influence is alsoexcluded from further calculations.

The impact of the second grounding can only be briefly dis-cussed due to limited available bathymetric information aboutthe real geometry of the seabed below the vessel. With an as-sumed flat seabed here, the later development of the heel anglecannot be explained [1, page 76]. From the salvage it is knownthat two main underwater pinning points exist. The vessel wasprobably pushed by wind and current against these points andthe friction force prevented the hull from sliding further downthe seabed profile. These effects combined lead probably to thefinal position of the COSTA CONCORDIA. Due to the unclearunderwater situation, further calculations regarding the influenceof the second grounding are not performed.

MOST LIKELY SCENARIOThe detailed analysis of the sinking sequence can be found

in the results section of [1]. That part of the report elaboratelydescribes the water progression inside the vessel, the influence ofthe different heeling moments, the impact of the activated water-tight doors and the effect of the second grounding. Furthermore,three hypothetical scenarios are investigated. In the following,only a brief summary of the most important results can be given.

The first scenario presented here takes into account all pos-sible information available thus representing the most likely one.The activity of the watertight doors are taken from the VDR data.The rock moment is applied, the wind and the moment from the

stabilizer fins are neglected as argued in the previous section. Thenumerical simulation ends at the point of the grounding impact.The leakage and collapsing of the openings are modeled accord-ing to results obtained in [3]. The only watertight door which isassumed to leak is No. 24, since it is reported by witnesses thatthe corresponding watertight bulkhead has been deformed duringthe collision.

Development of the floating positionThe first theoretical scenario established serves as a refer-

ence, because neither external influences nor the activity of thewatertight doors are considered. It is only used to demonstratethe influence of the different contributing factors on the floodingsequence.

This reference case is compared with the scenario, which isbelieved to represent the most likely one. The development ofthe floating position for both scenarios is shown in Fig. 7.

-15

-10

-5

0

5

10

15

20:45 20:55 21:05 21:15 21:25 21:35 21:45 21:55

Dra

ught

[m], T

rim

[m

], H

eel [d

eg]

UTC [H:M]

Draught referenceTrim reference

Heeling angle referenceDraught in most likely scenario

Trim in most likely scenarioHeeling angle in most likely scenario

FIGURE 7: Development of the floating position for the mostlikely scenario

First, the impact with the rock immediately followed by thesudden water ingress exerted a heel angle of 12-15 degrees toport with a draught increase by around 2 meters in less thana minute. Due to the almost symmetric layout of the initiallyflooded compartments, the heel angle decreases until it changesto starboard after around 18 minutes past the impact. The rockmoment delays the change to starboard, while the activation ofthe watertight doors accelerates this change later, because a moreunsymmetrical flooding is induced. In addition, the sinkage andthe aft trim is slightly larger in the most likely scenario. Thiseffect is just large enough to allow further progressive floodingabove the freeboard Deck 0. The heel now constantly increasesuntil the second grounding at around 21:50 UTC.

7 Copyright c© 2014 by ASME

In addition, the floating position of the vessel just beforegrounding the second time in front of the Giglio harbor is com-pared in Fig. 8. It shows the position obtained from the numer-ical flooding simulation compared with a photo taken by GiglioNews in that night. Looking for example at the submergence ofthe aft windows as a reference, both positions differ only verylittle. This indicates that the constructed most likely scenariomatches very well with the real accident situation and that theused numerical flooding simulation is a very valuable tool to in-vestigate such accidents, if used appropriately.

FIGURE 8: Comparison of the floating position of the COSTACONCORDIA at 21:48 UTC (Source: Giglio News)

Change of the List to StarboardThe large free surfaces alone developed during the flooding

did not lead to the observed change of the heel to starboard. Thischange is initiated mainly by two effects: First, the Garbage Planton Deck 0 located on the starboard side is connected with theIncinerators Room on Deck C by a vertical trunk. As soon as thewater level reaches Deck 0, the Garbage Plant is flooded throughthis opening, which further increases the heel to starboard due tothe location of the Garbage Plant on this side.

Secondly, the leak of watertight door 24 allows to flood thecrew spaces on Deck A on the starboard side. A possible pathof the water is sketched in Fig. 9 in a blue color and the WTD24 marked in red. The flooding becomes unsymmetrical here,because the reefer doors in the Buffet Preparation on the port sideprevents further water spreading to this side. Instead, the waterfloods the crew spaces and further up to the freeboard deck overthe staircases located here. This area has also been inspected bythe BSU on the sister vessel COSTA PACIFICA [12].

These two effects together with the large free surfaces aresufficient to explain the transition of the heeling angle to star-

Workshops pt

Workshops ct

Aft Staircase

Workshops sb

Buffet Preparation

Deck A Crew 12

Deck A Crew 11

Staircase MFZ 1+2

FIGURE 9: Deck A in Compartment 3 [13]

board by compensating the heeling moment induced by the rock.It should be stressed that no further external moments like windor current are required to explain the observed change of the heel-ing angle to starboard.

Influence of the Watertight DoorsA short summary of the most critical watertight doors ac-

tivities with regard to the flooding sequence will be given in thefollowing. Only the watertight doors located in the flooded partsof the vessel have a reasonable influence here.

The watertight door 6 located on Deck C connects the im-mediately flooded port Forward Engine Room with the SewageRoom. The opening of this door for only 25 seconds allows wa-ter also to enter the Sewage Room. This leads to a quite largeadditional heeling moment of approximately 750 tm, which firstdelays the change of the heel to starboard. However, the stabilityof the vessel is decreased by the additional free surface, whichleads to the very sudden change of the heel angle to starboardobserved at around 21:03 UTC, when the crew spaces on DeckA are flooded.

Another watertight door in the flooded zone, which was acti-vated is No. 9. This one connects the Electric Motors Room andthe Refrigeration Compressors Room. Both of these rooms areimmediately flooded through the leak caused by the rock. Theadditional connection between both established by the openingof WTD 9 merely effects the flooding process. It only bypassesthe water progression without having much further influence.

8 Copyright c© 2014 by ASME

Every watertight compartment requires a separate escapetrunk. However, in the case of this accident one situation re-garding the watertight door 10 is of special interest. Directlynext to the door such an Emergency Escape Exit is located (seeFig. 10). This exit might have been immediately blocked after theincident by the very sudden water ingress and the correspondingwater column which prevented the opening of the door to thisexit. Instead of taking the escape exit, anyone trying to escapefrom here might have tried to find another way out through thewatertight door 10 allowing more water to spread in the aft partof the vessel. In consequence the opening of this door for around45 seconds lead to a further increase of the aft trim and sinkageand to a further reduction of the stability of the vessel.

11 10 9

8

ER Exit

FIGURE 10: Watertight doors on Deck C, aft [13]

As already mentioned, the leakage and opening of the wa-tertight door 24 lead to a further unsymmetrical flooding to star-board. In addition, this small additional water ingress plays avital role for the overall flooding sequence. An additional sce-nario, with the only difference of no leakage of WTD 24, showsthat the heel angle would not have changed to starboard. Instead,the whole flooding sequence takes now about one hour longerand the direction of the final capsize changes to port. This showshow large the impact of a small detail on the overall floodingsequence can be.

Furthermore, if all watertight doors would have been closedand no leakage would have occurred, the time until capsize is fur-ther delayed and computed to be around twice as long as for themost likely scenario. Again, further details on these hypotheticalscenarios can be found in [1].

SINKING AT SEAOne of the most interesting hypothetical scenarios investi-

gated is assuming that the COSTA CONCORDIA did not made itsway back to the shore but drifted further away from it. It shouldbe stressed that only the wind caused the vessel to ground at therocks in front of the shore of Giglio Island.

The development of the floating position for this scenariois compared with the most likely one in Fig. 11. It shows thatwithout the second grounding only about 15-20 minutes later,the vessel would capsize very suddenly in a few minutes. Theknuckles which can be observed in the heel angle developmentare the collapsing of critical large openings on the upper decks 0to 2. These openings lead to the flooding of likewise large freeareas.

-20

0

20

40

60

80

100

20:45 20:55 21:05 21:15 21:25 21:35 21:45 21:55 22:05 22:15

Dra

ught

[m], T

rim

[m

], H

eel [d

eg]

UTC [H:M]

Draught in most likely scenarioTrim in most likely scenarioHeeling angle in most likely scenarioDraught while sinking at seaTrim while sinking at seaHeeling angle while sinking at sea

FIGURE 11: Development of the floating position for the sce-nario of sinking at sea

This is also one possible explanation why many of the vic-tims where found in the aft area of Deck 4. This is the deck withaccess to the lifeboats and it becomes quickly submerged by thecollapsing of the openings and the water traps everyone belowthis area.

This hypothetical scenario would most probably have lead tomany more victims and it was actually pure luck that the currentweather condition pushed the vessel back to the shore.

CONCLUSIONThe results of the performed numerical flooding simulations

match well with the observed behavior of the COSTA CONCOR-DIA during this severe accident. This demonstrates the utility ofthe method for the investigation of such accidents. It has beenshown that the change of the heel angle from initially port side to

9 Copyright c© 2014 by ASME

starboard can reasonably be explained by the complex spreadingof the flood water through the inner subdivision.

Even though the damage occurred to the COSTA CONCOR-DIA was much larger compared to the maximum “allowable”damage according to the applicable damage rules, the vessel stillsurvived for a relatively long time. This survival time could evenbe extended if all watertight doors were actually closed and donot leak in such critical situations. The opening of the doorscould be avoided if the overall subdivision layout of such vesselswould be designed in such a way that the number of watertightdoors are in principle reduced. Another critical aspect are verti-cal escape trunks, which in addition should not be blocked in thecase of entering water.

It is also very interesting to observe how much impact evensmall changes can have on the overall flooding process, if onelooks on the impact of the leakage of the watertight door 24. Fur-thermore, also any unsymmetrical layout of the compartmentsbelow and above the freeboard deck together with any verticalpenetration like the duct between the Incinerators Room on DeckC and the Garbage Plant on Deck 0 play a very important and vi-tal role on the capability to evacuate such large cruise ships in asafe manner. If the decks are not subdivided in a symmetric way,even small deviations from this like the compartment 3 on DeckA can suddenly change the heel angle development especially ifthe stability has already been reduced due to large free surfaces.In addition, the number of openings in the freeboard deck shouldbe kept to a minimum to reduce possible up-flooding ways ofthe water. Otherwise larger heeling angles drastically reduce theavailable time for a save evacuation procedure.

ACKNOWLEDGMENTWe gratefully acknowledge the provided photo material

from Giglio News, which allowed to estimate the floating po-sition of the COSTA CONCORDIA at the final phase of flooding.

We also like to thank the German BSU for the great cooper-ation and the very valuable information made available for us.

Furthermore, the authors would like to thank Anna Culjakfor giving us the advise about the VDR data made available byCodaCons.

In addition, we would like to thank the German Federal Min-istry of Economics and Technology (BMWi) for supporting ourresearch work.

REFERENCES[1] Russell, P., 2013. “The Sinking Sequence of M.V. Costa

Concordia”. Masterarbeit, Institute of Ship Design andShip Safety, Hamburg University of Technology, 11.

[2] Dankowski, H., 2013. “A Fast and Explicit Method for theSimulation of Flooding and Sinkage Scenarios on Ships”.Ph.D. Thesis, Hamburg University of Technology, Institute

of Ship Design and Ship Safety, 8. ISBN 978-3-89220-668-2.

[3] Ruponen, P., and Routi, A.-L., 2011. “Guidelines and cri-teria on leakage occurrence modelling”. In FLOODSTANDProject. Napa Ltd and STX Finland.

[4] Dankowski, H., and Dilger, H., 2013. “Investigation ofthe Mighty Servant 3 Accident by a Progressive FloodingMethod”. In Proceedings of the ASME 2013 32nd Interna-tional Conference on Ocean, Offshore and Arctic Engineer-ing, no. ISBN 978-0-7918-4492-2, ASME, ASME.

[5] van Walree, F., and Papanikolaou, A., 2007. “Benchmarkstudy of numerical codes for the prediction of time to floodof ships: Phase I”. In Proceedings of the 9th Interna-tional Ship Stability Workshop, Maritime Research Insti-tute Netherlands (MARIN), National Technical Universityof Athens (NTUA), pp. 45–52.

[6] Ruponen, P., 2006. Model Tests for the Progressive Flood-ing of a Boxshaped Barge. Tech. rep., Helsinki Universityof Technology.

[7] Marine Casualties Investigative Body, 2013. Report on theSafety Technical Investigation of the Cruise Ship COSTACONCORDIA. Technical report, Ministry of Infrastruc-tures and Transports, Italy.

[8] Culjak, A., 2013. “Organisation und Devianz - Eine em-pirische Fallrekonstruktion der Havarie der Costa Concor-dia”. Diploma Thesis, Bielefeld University.

[9] CodaCons, 2013. Ricostruzione del Naufragio delCosta Concordia. Published on Youtube by Mario Pic-cinelli: http://www.youtube.com/watch?v=csZzD-HfX8E,February.

[10] MCIB, 2013. “Watertight Doors Activity - VDR”. [7].Appendix 6.

[11] Wiggenhagen, M., 2013. “Berechnung des Trimmwinkelsund der Neigung der COSTA CONCORDIA aus Havariefo-tos zum Zeitpunkt des Ankersetzens um 21:48 Uhr”. Pho-togrammetric Expertise, Leibniz University Hannover. pro-vided by the BSU.

[12] BSU, 2013. Besichtigungsprotokoll der COSTA PACI-FICA. Tech. Rep. Az. 330-12, Bundesstelle für Seeunfal-luntersuchung, September. provided by the BSU.

[13] York, A., 2012. COSTA CONCORDIA Flooding Simu-lations. Technical Report S110833.002.2, Safety at Sea,September. Appendix 10, Enclosure No 1a.

10 Copyright c© 2014 by ASME