Proceedings of the ASME 213 32nd International Conference on Ocean, Offshore and Arctic Engineering OMAE213 June 9-14, 213, Nantes, France OMAE213-1342 INVESTIGATION OF THE MIGHTY SERVANT 3 ACCIDENT BY A PROGRESSIVE FLOODING METHOD Hendrik Dankowski Research Assistant Institute of Ship Design and Ship Safety Hamburg University of Technology Hamburg, Germany Email: dankowski@tu-harburg.de Hendrik Dilger Student Hamburg University of Technology Hamburg, Germany Email: hendrik.dilger@tu-harburg.de ABSTRACT The semi-submersible heavy-lift vessel MIGHTY SERVANT 3 sank off the port of Luanda, Angola in the morning of December 6th, 26 during a ballast operation to offload the drilling platform Aleutian Key. The official investigations carried out after the accident identified an error in the control of the submerging ballast operations as the direct cause of the sinking. However, the detailed phenomenons and reasons for the sudden excessive trim development has not been investigated further. This paper intends to identify the most likely sceneario which lead to the hydrostatic stability failure during the discharge operation by computing the flooding process during the ballast operation in the time domain. A numerical progressive flooding simulation method is presented for applications like accident investigations or damage stability assessments. This method is modified to fit the special requirements of simulating the operational procedures of semi-submersible vessels in the time domain. Extensions like the inclusion of pump elements but also the multi-body interaction of the cargo and the vessel with regard to the hydrostatics is presented. The direct flooding simulation computes the flux between the compartments based on the Bernoulli equation and the current pressure heads at each intermediate step. Large and partly flooded holes are taken into account as well as optional air compression and flooding through completely filled rooms. Pressure losses due to viscous effects are taken into account by applying semi-empirical discharge coefficients to each opening. The flooding paths are modeled by directed graphs. A detailed investigation of the MIGHTY SERVANT 3 accident and an identification of the possible failure modes leading to the sinking of the vessel is presented. This will help to better understand the phenomenons leading to critical situations during the submerging procedure of semi-submersible heavy-lift vessels and to avoid such accidents in the future. Applying time domain flooding simulations allows to predict the ship behavior during ballast operations to identify critical situations and to better schedule the different steps of such an operation in advance. FLOODING MODEL The developed numerical flooding simulation determines the fluxes through the openings by means of a hydraulic model. The propagation of the water volumes is computed by a predictorcorrector scheme for the integration of the volume fluxes to each compartment in question. Flux Determination The in- or egress of flood water through the internal and external openings can be idealized by the incompressible, stationary and viscous- and rotational free Bernoulli equation given in Eqn. 1 formulated for a streamline connecting point a and point b: dz = p a p b ρ g + u2 a u 2 b 2g + z a z b ϕ ab g. (1) 1 Copyright c 213 by ASME
The dissipative energy term ϕ ab accounts for pressure losses through the openings and is assumed to be proportional to the kinetic energy of the flow. This loss is modeled by a semi-empirical discharge coefficient C d reducing the flux velocity. The pressure height difference dz yields the fluid velocity u = C d 2g dz. (2) For a free outflow (left side of Fig. 1), the pressure height difference becomes: Flooding Paths Directed graphs are used to describe the flooding paths. The openings are the edges of the graph connecting the different compartments representing the nodes. The direction of the edges defines the sign convention for the opening fluxes. The mass balance for one compartment is given by the sum over all edges connected to one node. An example of such a directed graph is given in Fig. 2. This graph describes the flooding paths from the validation test case B, which is described in more detail in the validation section of this paper. The geometric setup associated to this graph is shown in Fig. 3. dz = p a p b ρ g + u2 a 2g + z a z. (3) (1) Outside 1 (3) DB2 (4) R11 The flow velocity through a deeply submerged opening (right side of Fig. 1) is independent of the location z of the opening driven by the following pressure height difference: 2 3 6 8 (2) DB1 (5) R21 (8) R12 dz = p a p b ρ g + u2 a u 2 b 2g + z a z b. (4) 4 5 9 7 (6) R21P (7) R21S (9) R22 z a p a a p b z a p a a p b FIGURE 2: Directed graph for test case B z z b z p b FIGURE 1: SMALL OPENINGS z b b Propagation Step The amount of water propagated in one time step to one compartment from its neighbor(s) is given by the integration of the sum of the volume fluxes Q o (t) obtained from all connected openings: The integration of the velocity over the area of the opening assuming a perpendicular flow direction leads to the total flux: V = V t = Q = u da = u nda = uda. (5) A A A The solution of this integral becomes more complicated if the opening is large and of arbitrary shape and orientation. Therefore, larger openings are discretized in smaller, elementary parts for which an analytical solution of the volume flux can be determined as described for example in [1]. t2 dv = Q o (t)dt. (6) t 1 The propagation of water leads to a new distribution of the flood water and new volume fillings in the compartments. The water levels of the compartments depend nonlinear on the volume fillings and has to be determined iteratively. The flux function Q o (t) is a also nonlinear function over time, since the opening fluxes depend mainly on the water levels. To account for this nonlinear characteristic at the time integration, a predictorcorrector scheme is applied, which is sketched as follows: 1. Predict opening fluxes 2. Propagate predicted volumes assuming a constant flux 2 Copyright c 213 by ASME
3. Calculate new filling levels 4. Recompute opening fluxes based on these new fillings 5. Estimate mean flux by relaxation 6. Reset volumes and propagate again, recompute filling levels based on the mean flux and proceed This scheme avoids efficiently the flux direction change during one time-step caused by the explicit characteristic of the method and stabilizes the whole numerical simulation. Propagation of Fully Flooded Compartments A completly flooded compartment introduces a coupling with the neighboring compartments, because the flux determination and propagation step can no longer be handled independently. The sum of all water fluxes to this compartment must be zero, which also means that the water has to be propagated through this compartment from the neighboring, partly filled compartments. The only free variable remaining is the pressure of the full compartment. The resulting nonlinear equation has to be solved iteratively to find a pressure value, which leads to a zero overall water flux through all connected openings. Conditional Openings The computation of complex flooding scenario requires the possbility to apply certain conditions to the openings involved in the flooding sequence. These conditions are for example: - Pressure head conditions for breaking windows - Time dependent discharge coefficients - Pump elements The pump elements are especially useful as control elements to simulate ballast sequences. The volume flux through these openings is kept constant during a given time interval, which represents the influence of a ballast water pump. Gravity flooding of tanks can be controled by closing or opening the valves to these tanks at a certain time. This is modelled by a time dependent discharge coefficient of the corresponding opening. Air Compression The compression of entrapped air is modeled by Boyle s law stated as follows: p V = p 1 V 1. (7) Simulation Overview One time step of the simulation consists of the following parts: 1. Check conditional openings 2. Pressure iteration of full compartments 3. Fluxes of remaining openings 4. Inner iteration for higher-order integration of fluxes 5. Propagation of water volumes 6. Update of filling levels and determination of full tanks 7. Optional air compression 8. Iteration of new floating equilibrium 9. Check of convergence This is repeated for each time step until either the requested simulation time or convergence is reached. Validation The flooding simulation is validated with the ITTC benchmark model test for the prediction of time to flood [2]. These model tests of a box-shaped barge has been carried out by Pekka Ruponen at the Ship Laboratory of the Helsinki University of Technology [3]. The results obtained by the developed flooding simulation are compared to the experimental and the numerical results obtained by Ruponen [4]. Parts of the results for validation case B are presented in the following. The geometric setup of this test case is shown in Fig. 3 and the derived flooding graph in Fig. 2. R12 R11 Side View R22 R21 z DB1 2 DB2 x 9 7 6 A 1 A 8 3 ps R22 Front Section View A-A R21P R21 R21S 4 5 DB2 FIGURE 3: Test case B - Side and section view of openings and compartments 8 y z CL 3 sb The decrease in the air volume leads to an increase in air pressure. The air flow itself is not taken into account, but the automatic determination of entrapped air pockets is included. The influence is discussed in more detail in the validation section of this paper. This test case represents a complex upflooding sequence involving air compression in the double bottom compartment DB1. For the numerical simulation a time-step of.1 seconds is chosen. The simulated results from the developed method are marked with the prefix calc. D., the measured values with 3 Copyright c 213 by ASME
meas. R. and the computed values by Ruponen with calc R.. The comparison of the simulated trim and heave motion is shown in Fig. 4. Trim angle (deg) 2. 1.5 1..5. 1 2 3 4 5 6 7 8 9 6 12 18 24 3 36 42 48 Time (s) meas. trim R. calc. trim R. calc. trim D. meas. heave R. calc. heave R. calc. heave D. FIGURE 4: TEST CASE B - TRIM AND HEAVE MOTION The obtained results match very well with the measured values. At the beginning of the flooding sequence the motion values are slightly overpredicted but in the later phase these differences vanish. These variations can be explained by the treatment of the entrapped air in the double bottom compartment DB1. The development of the water levels in the double bottom compartments are shown in Fig. 5. Water height (mm) 16 14 12 1 8 6 meas. DB1 R. 4 calc. DB1 R. (1) calc. DB1 D. (2) calc. DB1 D. 2 meas. DB2 R. calc. DB2 R. calc. DB2 D. 6 12 18 24 3 36 42 48 Time (s) FIGURE 5: TEST CASE B - WATER LEVELS OF THE DOU- BLE BOTTOM COMPARTMENTS Heave (mm) Two different approaches to determine entrapped air pockets are compared. In the first case, marked with (1) in Fig. 5, it is assumed that the air pockets above the water level in a compartment compresses, if all openings leading to this compartment are submerged from both sides. The obtained results for the water levels match very well at the end of the simulation, even though larger differences are present at the beginning. The differences are explained by the fact that no air flow is taken into account, which delays the air compression, because the amount of air mass to be compressed decreases. The second approach marked with (2) in Fig. 5 is less conservative and assumes that the compression starts as soon as the openings are submerged from one side only. In the experiment, an air bubble flow has been observed raising from DB2, which means that air still escapes even though the opening is submerged from the side of the DB1 compartment. This leads to an earlier start of the air compression and allows less water to enter the model. To conclude, the first approach leads to appropriate results without the necessarity to include air flow in the model and is used in the presented method. The numerical method has further been validated by the reinvestigation of three real ship accidents from the past: The accident of the EUROPEAN GATEWAY [5] in the English Channel, the loss of the greek passenger ferry HERAKLION [6] and the tragic accident of the ESTONIA [not yet published] ferry in the baltic sea. For all three investigations, reasonable results for the chain of events leading to the loss of these vessels has been obtained by applying the developed numerical flooding method. Further details can be founded in the cited publications. COUPLING OF CARGO, SHIP AND SEABED The simulation of the offloading operation of semisubmersible heavy lift ships requires the coupling of the different objects involved in the process. The cargo to be offloaded will start to float after a certain draught is reached and the cargo buoyancy exceeds its weight. In addition, the bottom of the ship might hit the seabed, if the water depth is restricted and the trim becomes very large. In both cases, the interaction will be modeled with spring elements. The main focus is on the correct simulation of the coupled motion, other aspects like the structural strength or the accurate modeling of the properties of the seabed are of minor importance for this application. Basic Kinematics The motions of a vessel for hydrostatic evaluations are described by three degrees of freedom in two righthanded coordinate systems. The ship fixed coordinate system moves with the ship and is located in the keel bottom (z vertical positive upward), in the centerline (y transversal positive to the portside) and at the aft perpendicular (x in longitudional direction positive forward). 4 Copyright c 213 by ASME
The movement is described by the draught T, the trim angle ϑ (positive to the bow) and the heel angle ϕ (positive to starboard). The earth fixed system is located at the same origin but does not move. Each ship fixed point x can be transformed to the earth fixed system x by the rotation matrix R. x = R x t, (8) ξ η = cosϑ sinϕ sinϑ cosϕ sinϑ cosϕ sinϕ x y. ζ sinϑ sinϕ cosϑ cosϕ cosϑ z T (9) The rotation matrix is orthogonal, its inverse is equal to its transposed matrix. An iterative algorithm is applied to the three degrees of freedom to find an hydrostatic equilibrium at which the resulting displacement force, the trim and the heel moments are vanishing. The force and momentum sum is obtained in the earth fixed coordinate system. Soft Grounding The interaction of the ship with the seabed can be modeled by a simple spring model. If a critical draught is exceeded, each point of the hydrostatic data model is checked, if it is located below the seabed floor. The resulting spring force F for one point p is defined by the spring stiffness C and the earth fixed distance dz of the point to the seabed located at the waterdepth H w F = C dz, (1) dz = H w r 3i p + T, (11) where r 3i is the third row vector of the matrix R. A grounding force is only present if the value dz is greater than zero, otherwise the value of the force is set to zero. It is further assumed that the spring force is directed in earth fixed vertical direction. The corresponding trim M η and heel moments M ξ are obtained as follows: M η = F ξ = F r 1i x 1, (12) M ξ = F η = F r 2i x 1. (13) The resulting spring forces and moments are added to the overall force and momentum balance. For the purpose of hydrostatic motion analyzes, the value of the seabed stiffness C is of minor importance. The springs are only elements to restrict the motions of the vessel during the iterative search to obtain a new floating equilibrium position in the grounded case. Cargo-Ship Interaction The interaction of the ship with its floating cargo is also modeled with the help of spring elements. The coupling is more complicated, because the relative motions between cargo and ship are required to obtain the interacting forces and moments. In addition, the cargo may float independently of the ship at the end of the offloading process. This requires six degrees of freedom to describe the motion of both objects. At the defined discrete spring element points, the resulting forces and moments are obtained, which affect both the cargo and the ship, but in opposite direction. To compute the relative motions, the earth fixed distance of the cargo fixed point x to the ship fixed plane of the deck in question have to be computed. The orthogonal distance to this plane in ship fixed direction is derived by a transformation of the cargo fixed point in the ships coordinate system x 1 : x = R x t, (14) x = R 1 x 1 t 1, (15) x = R T (R 1 x 1 t 1 + t ). (16) The ship fixed vertical displacement dz of the point together with the earth fixed value is then given as: dz = z z 1, = z [r i3, (R 1 x 1 ) + r 33, (T T 1 )], (17) dz = dz. r 33, (18) The resulting spring forces and moments are computed in the same manner as for the seabed grounding forces. This strategy allows to simulate a correct and continous simulation of the submerging process of two floating objects like a semisubmersible heavy lift ship with a drilling rig as a deck cargo. Until the weight of the cargo exceeds its buoyancy, the remaining forces and moments for an equilibrium result from the springs located on the ships deck and affect also the hydrostatics of the ship. After the cargo becomes afloat, the interacting spring forces are zero and both objects float independently of each other. Three different iterative strategies are possible to find a solution for the coupled system: 1. Inner iteration of the cargo position 2. Sequential iteration, first ship and then cargo 3. Instantaneous search for all six degrees of freedom The first search strategy iterates a new position for the cargo at each iterative step of the ship. This requires many inner iterations and makes the whole iteration unstable. The second approach requires less iterations, but the solution is never fully converged, because cargo and ship influences each other as long as the cargo does not afloat. 5 Copyright c 213 by ASME
The last method appears to be the most complicated one at a first glance, but even though the size of the system of nonlinear equations doubles, the number of required iterations is not increased very much. This is very important, because the computational most expensive part of hydrostatic calculations are the function evaluations to obtain the current resulting forces and moments for a certain combination of draft, trim and heel angle. A combined solution of the ship-cargo and the ship-seabed interaction is possible as well, as shown in the following results section of the accident investigation. INVESTIGATION OF THE MIGHTY SERVANT 3 ACCI- DENT The described simulation method is applied to the investigation of the MIGHTY SERVANT 3 accident, which occured off the coast of Angola on December 6, 26 at 7:53 hours local time. The vessel, loaded with the Aleutian Key drilling rig, is shown in Fig. 6 during the final voyage before the incident happened. The drilling rig was successfully discharged before the Mighty Servant 3 developed an excessive trim to the stern and sank to the ground. After the successful salvage by SMIT, the MIGHTY SERVANT 3 has been redelivered to the owners on 26 May 27 and returned to duty. Previous Investigations After the accident, an investigation has been carried out for the Maritime Disciplinary Court of the Netherlands. The public available investigation file has been retrieved from the court. This file provided useful information about the accident and the vessel, for example: - Hydrostatic particulars, - Stability booklet, - Operation manual for ballasting & de-ballasting, - Dockwise transport manual, - Report on the investigation of sinking. Unfortunately, no information about the hull form or more detailed GA drawings and tank plans could be made available. Any request to the owner Dockwise or the yard has been denied. Additional information is used from [8], where parts of the body plan are given. The hull form and the tank layout is reconstructed from the given information and validated by the given hydrostatic properties of the hull form and the tank particulars. Computational Data Model For the following numerical flooding simulation, the whole vessel including the rig has to be described. In addition, several conditional pump elements and the valves to the ballast water tanks are defined. The reconstructed hydrostatic data model is shown in Fig. 7. General Description of the Vessel The vessel was built in Japan at Oshima Shipbuilding Co. Ltd in 1984 for the Dutch shipping company with the current name Dockwise Shipping. The main particulars are summarized in Tab. 1. TABLE 1: Main dimensions of the MIGHTY SERVANT 3 Length over all L OA 181.23 m Length between perpendiculars L BP 165.7 m Breadth B 4. m Depth D 12. m Draft submerged T M 22. m Summer draft T 9.63 m Service Speed V S 12. kn FIGURE 7: THE HYDROSTATIC DATA MODEL OF THE MIGHTY SERVANT 3 AND THE ALEAUTIAN KEY PLATFORM Final Loading Condition The flooding simulation starts at 6:12 local time at a draught of 19 meters on even keel. This situation is affirmed by statements from the crew. The intact loading condition at this time is given in Tab. 2 together with the righting lever arm curve in Fig. 8. This loading condition and the lever arm curve is computed for the combined floating object of the MIGHTY SERVANT 3 loaded with the ALEAUTIAN KEY. The values obtained for the lever arm curve must be treated with care, because at larger heel angles the platform would probably start to move, which strongly influences the righting lever arm of the overall system. 6 Copyright c 213 by ASME
FIGURE 6: THE MIGHTY SERVANT 3 JUST BEFORE OFFLOADING THE ALEAUTIAN KEY PLATFORM [7] TABLE 2: Loading Condition of the MIGHTY SERVANT 3 at 6:12 hours just before offloading 3.5 3 UNFALL612 - port side Shell Plating Factor 1.3 - Density of Sea Water 1.25 t/m 3 Ships Weight 8185.46 t Longit. Centre of Gravity 77.919 m.b.ap Transv. Centre of Gravity -.158 m.f.cl Vertic. Centre of Gravity (Solid) 13.524 m.a.bl Free Surface Correction of V.C.G..124 m Vertic. Centre of Gravity (Corrected) 13.647 m.a.bl Draft at A.P (moulded) 19. m Draft at LBP/2 (moulded) 19.2 m Draft at F.P (moulded) 19.5 m Trim (pos. fwd).5 m Heel (pos. stbd) -.635 Deg. Volume (incl. Shell Plating) 78229.656 m 3 Longit. Centre of Buoyancy 77.919 m.b.ap Transv. Centre of Buoyancy -.111 m.f.cl Vertic. Centre of Buoyancy 9.421 m.a.bl Area of Waterline 1978.633 m 2 Longit. Centre of Waterline 111.379 m.b.ap Transv. Centre of Waterline.46 m.f.cl Metacentric Height 3.284 m Righting lever [m] 2.5 2 1.5 1.5 -.5 1 2 3 4 5 6 7 GZ [m] Requ. or Max. h: 2.178 m Progfl. or Max.: 7. Deg. Heeling angle [deg] GM at Equilib. : 3.284 m Area under GZ [mrad] FIGURE 8: LEVER ARM CURVE OF THE MIGHTY SER- VANT 3 AND THE ALEUTIAN KEY AT 6:12 HOURS Reconstruction of the Submerging Process The offloading operation by submerging the carrying vessel until the cargo floats is achieved by a combination of gravity ballasting by openings valves and the usage of pumps in the later phase to finetune the flooding sequence. The exact pump rates and the soundings in each ballast tank during the critical submerging operation is not known for all phases of the operation. However, the offical report on the investigation of sinking states several useful information from the crew recorded by the Voyage Data Recorder (VDR). This timeline of events is reconstructed in the following by using reasonable estimations for the pump rates and the states of the bottom valves used for the gravity flooding. Most Likely Sinking Sequence The chain of events of the ballast operation procedure leading to the loss of the vessel is 7 Copyright c 213 by ASME
reconstructed in the following. The frames from the simulation for selected time steps are shown in Fig. 9. In addition, the volume fillings of the involved ballast tanks are shown in Fig. 1. seabed. The vessel comes to rest on the seabed in a water depth of 6 m at around 8:5. This sequence relates to the third scenario of possible failure modes described in more detail in the following section. 6:12: 1 8.5.5 7:17: 7:31:8 Volume Filling (%) 6 Trim 4 FOREPEAK PS FOREPEAK SB 7 CENTER 7 PS WING 7 SB WING 2 8 PS WING 8 SB WING 9 PS WING 9 SB WING ENGINE ROOM 6:15 6:3 6:45 7: 7:15 7:3 7:45 Time (H:M) 1 1.5 2 2.5 3 Trim (m) FIGURE 1: VOLUME FILLINGS IN THE BALLAST TANKS DURING THE SUBMERGING OPERATION 7:58:26 8:5: FIGURE 9: TIME STEPS FROM THE ANIMATION OF SUB- MERGING The platform discharges from the deck at around 7:17. The buoyancy columns at the stern completely submerge at around 7:32. Beyond this point, the MIGHTY SERVANT 3 developes an excessive trim until at around 7:58 the vessel s stern hits the Possible Failure Modes The described submerging procedure until 7:21 matches quite well with the statements of the crew and the described ballasting procedure in the operational manual. It is now investigated in more detail, which failure mode might have caused the development of the excessive trim angle followed by the sinking down to the seabed. In the official report it is stated that at 7:24 the master realised that the situation went out of control and he made the order to pump out of the 7 wing tanks, which are located below the main deck at the stern of the vessel. The influence of this order is now investigated. In addition, a possible water ingress in the 7 center tank is simulated. This gives the following three scenarios: 1. The ballast water pump to the 7 wing tanks is reversed to the suction position. 2. The pump keeps working in the ballasting direction, more water enters the 7 wing tanks. 3. A possible water ingress in the 7 center tank through a broken bottom valve. The first scenario marked with (1) assumes that after an delay of approximately one minute, the pump to the 7 wing tanks switched to the suction position and removed water from these tanks after 7:25. In the second scenario (2), the pump is kept 8 Copyright c 213 by ASME
running and no deballasting from the 7 wing tanks where made possible. The behavior of the vessel is illustrated in Fig. 11 showing the development of the trim and the metacentric height GM for all three scenarios. The counteraction of deballasting.5 2.4 sudden trim development could be indentified, at least without a better knowledge of the ballast system of the MIGHTY SERVANT 3. The main purpose of this investigation is to show, that the numerical simulation gives reasonable and useful results to study different accident scenarios. The likelihood of these scenearios and the identification of the actual cause, which lead to the sinking of the MIGHTY SERVANT 3, is left to others. 2.2 Trim (m).5 1 1.5 2 Trim (1) Trim (2) 2.5 Trim (3) GM (1) GM (2) GM (3) 3 6:15 6:3 6:45 7: 7:15 7:3 7:45 Time (H:M) FIGURE 11: TRIM MOTION DEPENDING ON FLOODING OF 7 TANKS the 7 wing tanks influences the trim at this point. The developed large trim is decreased and the aft buoyancy columns come back to the sea surface. The vessel is actually saved by counterballast action taken in this first scenario. However, the second scenario (2) shows that the vessel would not sink even if the pumps to the 7 wing tanks would not have been switched to the suction position. The developed trim is larger compared to the first scenario, but the floating position is still stable and no further openings become submerged. This means that without any further interaction, the vessel develops a large trim but remains still in a stable floating position. The last scenario (3) assumes that the bottom valve of the 7 center tank breaks at an assumed water column height above the opening of 18 m. Until now, the 7 center tank has not been used during the ballast operation. The opening breaks at around 7:22, which leads to a very large additional heeling moment to the stern. the deballasting of the 7 wing tanks does in this case not prevent, that the trim further increases. Only a few minutes later, the stern of the MIGHTY SERVANT 3 hits the seabed. In the following, the forward decks house becomes submerged, which allows water to enter the engine room. This leads to a further progressive flooding and finally to a fast sinking of the vessel. It should be mentioned, that a water ingress to the 7 center tank as a possible failure mode was excluded in the offical investigation reports. However, almost all other ballast tanks were completely flooded at the critical point and no other cause for the 2 1.8 1.6 1.4 1.2 1 GM (m) CONCLUSIONS AND OUTLOOK The MIGHTY SERVANT 3 incident has been successfully reconstructed. The simulation of the submerging operation followed by the sinking sequence down to the seabed including the grounding reaction at the stern gives very reasonable results. The time line of events matches very well with the statements of the crew. As an example for a possible failure mode, the influence of a water ingress in the 7 center tank through a broken bottom valve is shown, which leads to the observed sudden development of an excessive trim to the stern. Possible control options to avoid such an accident are for example valves, which are automatically controlled by trim or heel dependent conditions to avoid excessive motions. This would allow a fast and safe ballast operation by using only gravity flooding of the ballast tanks. Furthermore, the presented numerical simulation method could be used for crew training to examine such ballast operations in advance. Extensions of the method are for example the determination of friction forces at the contact points of the cargo with the main deck to control and avoid unwanted transversal motions of the cargo during the ballast operation. In addition, the structural strength of the semi-submersible vessel and the involved cargo could be investigated in more detail. Docking operations of a ship in a floating dock could also be simulated in the time domain with the help of this flooding simulation method to identify critical points of such a complex operation. ACKNOWLEDGMENT I thank Prof. Krüger for supporting my work at our institute. Special thanks go out to the secretary E. Doeven of the Maritime Disciplinary Court of the Netherlands who provided me with the very useful material of the investigations carried out at the court after the accident of the MIGHTY SERVANT 3. REFERENCES [1] Dankowski, H., and Hatecke, H., 212. Stability Evaluations of Semi-Submersible Heavy-Lift Vessels by a Progressive Flooding Simulation Tool. In Proceedings of the ASME 212 31st International Conference on Ocean, Offshore and Arctic Engineering, ASME. 9 Copyright c 213 by ASME
[2] van Walree, F., and Papanikolaou, A., 27. Benchmark study of numerical codes for the prediction of time to flood of ships: Phase I. In Proceedings of the 9th International Ship Stability Workshop, Maritime Research Institute Netherlands (MARIN), National Technical University of Athens (NTUA), pp. 45 52. [3] Ruponen, P., 26. Model Tests for the Progressive Flooding of a Boxshaped Barge. Tech. rep., Helsinki University of Technology. [4] Ruponen, P., 27. Progressive Flooding of a Damaged Passenger Ship. PhD thesis, Helsinki University of Technology. [5] Dankowski, H., and Krüger, S., 212. A Fast, Direct Approach for the Simulation of Damage Scenarios in the Time Domain. In 11th International Marine Design Conference, University of Strathclyde. [6] Dankowski, H., 212. An Explicit Progressive Flooding Simulation Method. In 11th International Conference on the Stability of Ships and Ocean Vehicles, K. J. Spyrou, N. Themelis, and A. D. Papanikolaou, eds. [7] Wikipedia, 27. Mighty Servant 3 - Wikipedia, the free encyclopedia. [Online; accessed 13-January-213]. [8] Journée, J., 23. Review of the 1985 Full-Scale Calm Water Performance Tests Onboard m.v. Mighty Servant 3. TUD Report 1361, Ship Hydromechanics Laboratory, Delft University of Technology, August. 1 Copyright c 213 by ASME