Вы находитесь на странице: 1из 29

D.P.Servis & M.

Samuelides National Technical University of Athens


Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 1
SHIP COLLISION ANALYSIS USING FINITE ELEMENTS
By
Dimitris P. Servis and Manolis Samuelides
Department of Naval Architecture and Marine Engineering
National Technical University of Athens
Athens, Greece
1. INTRODUCTION
Today's merchant ships confront a great number of operational requirements set both by the
shipping community and the international community. The requirements set by the shipping
community mainly relate to the ships' speed, building and operational cost. Additional
requirements set from both the shipping and international communities are the safety of the
ship as a mean of transportation as well as the safety of the natural environment and man-
built structures. These requirements set new methodologies for the design and operation of
ships. As sea routes are getting denser and speeds higher, there is a good possibility that a
ship may experience extreme loads during her lifetime. Further, higher speeds may cause
magnified operational loads or severe loads like slamming. Denser sea routes increase the
possibility of an accident involving ships or ships and shore or offshore structures. In
addition, as speeds along these sea routes are increased, the possibility of an accident also
increases.
There are two complementary ways in dealing with these problems. The first one is to
prevent the occurrence of extreme loads and accidents. This is being achieved using on board
sophisticated surveillance and monitoring equipment and well-trained crews. In addition the
surveillance of sea routes, especially in high traffic areas near harbours, channels and
offshore structures contribute a lot in minimising accident occurrence. The second aspect on
the solution of the problem is to mitigate the effects of a potential accident. This is done by
developing structures that may tolerate a damage within the limits of a required safety of the
structure, payload and environment. In order to achieve this, the loads induced on the
structure should be quantified.
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 2
During the last decades, the issue of impact loads on ship structures is of major concern, and
collision loads have been identified, at least in the cases of ships carrying hazardous cargo
initially and passenger ships more recently, as a phenomenon, which should be taken into
consideration in the design stage of a ship. The interest on the effect of collision loads on
passenger ships has become higher, since the advent of high speed ferries with conventional
or non-conventional hull shape. Since there is no previous experience on the behavior of
these new structures and of the new materials used in many cases, the investigation of
collision effects is imperative.
There are two major questions that naval engineers working on ship collisions should be able
to approach: One concerns the simulation of ship collisions and the prediction of the
damages, which occur during the incident. The other is the identification of collision scenario
or scenarios, which the ship under consideration should be checked against in order to assess
her capacity to withstand collision loads.
The paper addresses the former of the above mentioned questions. It is the main object of the
work reported herein, to examine the capabilities of explicit finite element codes for collision
simulation. In particular the paper reports on a) the simulation of impacts on ship structural
components, b) the simulation of small-scale collision tests and c) on the simulation of the
collision of a Ro-Ro vessel, the MS-DEXTRA.
The simulation of impacts on structural components allows investigating whether failure
modes, which have been observed during large-scale impact tests may be obtained by the
numerical analysis.
The small-scale collision tests that are simulated were performed in the University of
Glasgow in the early 80s. The models, which represented the parallel section of a tanker,
were either fixed or floating freely in a towing tank. The simulation of these tests could
provide necessary confidence for the numerical results.
Finally the simulation of the collision of the Ro-Ro vessel would allow examining, whether
the structural arrangement could be considered adequate to withstand the collision loads
under the defined scenario.
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 3
The FE simulation of a collision encompasses a number of individual problems, which should
be given appropriate attention. These problems are:
The selection of a mesh, which should be fine enough for the results to converge and to
reproduce the failure modes and coarse enough for executing the code in acceptable time
(for the collision analysis of the Ro-Ro presented in the paper, the CPU time is counted in
days).
The inclusion of the effect of the surrounding water.
The extent of the ship, which would be modelled as a deformable body
The modelling of the material.
The material failure criterion, which appears to be a weak point in the collision
analysis, either this is performed with simplified analytical techniques or numerically.
The work reported herein is expected to give to the investigators some guidance in order to
assist them to approach the complicated task of the collision simulation.
2. FINITE ELEMENT CODE SELECTION AND APPLICATION TO SHIP
COLLISIONS
Currently, many researchers worldwide work on the numerical simulation of ship collisions.
Kitamoura O. et al [6], [7], Lee J.W. et al [8], and Lehmann E. et al [9] carry out outstanding
work. Very large FE element models have been developed and in some cases comparisons
with experimental results were carried out. Nevertheless, the prediction of collision results
with FE methodologies has not yet been fully proven and achieved. In the field of material
failure, Lehmann [9] has introduced a sophisticated material model, which returns quite good
results.
In the framework of selecting an efficient FE code for simulating ship collisions, several
criteria should be met. These criteria relate to the modelling capabilities for both the internal
and the external collision mechanics. In this aspect, the code must be capable of modelling
the two ship motions during and after the collision (external mechanics), as well as the
deformation and collapse of the structures (internal mechanics). In general, codes like
MSC/DYTRAN [1] and ABAQUS/Explicit, using the explicit time-integration scheme
incorporate such features. Since MSC/DYTRAN is currently available in NTUA, it was
decided to use it for the simulations. This code, licensed to the Computer Centre of NTUA is
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 4
an explicit, three-dimensional finite element code for analysing the dynamic, nonlinear
behaviour of structures. MSC/DYTRAN integrates the capabilities of DYNA3D and PISCES
codes, used in the past to simulate ship impacts. Researchers currently use MSC/DYTRAN
worldwide for modelling ship collisions. The features of MSC/DYTRAN considered
essential for solving ship collisions are discussed below:
i) External Mechanics: MSC/DYTRAN being an explicit FE code, allows for the
structures to freely move in space without defining any constraints. This feature is
essential, since it means that the structures of the colliding ships may interact and
separate as they would in a zero gravity space. Thus, no boundary conditions are
needed. In addition, there is no need to prescribe the motion of the striking ship and
the collision scenario is only defined by her initial velocity and the relative position of
the two ships in space. The program outputs the motion of both striking and struck
ships during the simulation. Another critical feature is the explicit modelling of rigid
structures and point masses. Rigid structures are used for modelling striker bows and
connections between detailed and coarse areas of the struck ship. Point masses are
used to model the displacement distribution of the actual ship and the added mass
distribution due to the movement of the struck in the water.
ii) Internal Mechanics: The available material models in MSC/DYTRAN can be used to
account for the material behaviour under dynamic loading, as much as this is possible.
Such material models take into account the material yielding and strain-rate effects on
the yield point, material hardening and failure. It is also possible to define a piecewise
linear, stress-strain characteristic. The failure criterion may be the maximum plastic
strain or defined via a user subroutine. Failure means that the element, subject to such
a deformational or stress level satisfying the failure criterion, it will lose all its
stiffness.
The above mentioned characteristics briefly state the capability of the code to model the
internal mechanics of the collision phenomenon. Nevertheless, there are certain features that
are not present and would facilitate the modelling of ship structures. These features are
mainly related to the available elements in the code. Only three- and four-node shell elements
are included, while higher degree elements could exist. Higher degree elements could be used
towards a better representation of complex hull forms and more precise representation of
stress and strain results. As much as the beam elements are concerned, the cross-section
definition of the element does not include offsets for the moments of inertia, importing a
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 5
small error in the determination of the stiffness. In addition, only special cross-section beam
elements may incorporate the effect of strain-rate hardening. This excludes the general-
purpose beam elements from the modelling of ship structures, since the strain-rate effect is of
a great importance in ship collisions. It should be noted though, that based on the available
information neither ABAQUS/Explicit supports some of these features. It is likely that these
features are not present in favour of a minimal solution time. Explicit codes calculate the time
step for every cycle based on the length of the smallest element, the sound speed in the
material assigned to this element and a scale factor. This produces a very small time step and
consequently long solution times. Higher degree elements with more nodes and more
complex formulation would cause a considerable increase in solution time.
In order to evaluate MSC/DYTRAN's ability to model ship collisions, a simplified model is
developed and tested in different structural configurations. Structural description and results
can be found in [2, 3, 17]. This model consists of a simple V-shaped striker bow and a hollow
side shell structure of the struck ship. The striker bow is rigid and its mass is equal to the
mass of the whole ship (see Figure 2.1). Point masses are used to model the actual mass of
the ship and the added mass. The material model is elastic-perfectly plastic with a maximum
plastic strain failure criterion. The various structural configurations differ in the way of
modelling the remaining parts of the struck ship and the added mass.
Figure 2.1: Simplified model for ship-ship collision
This model was used to investigate and evaluate the capabilities of MSC/DYTRAN in
modelling collisions. Comparing the results for each configuration, they exhibit a consistency
in expected vs. obtained results. As a concluding remark on these benchmarks,
MSC/DYTRAN may be considered suitable for modelling ship collisions.
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 6
3. SIMULATION OF IMPACTS ON SHIP STRUCTURAL COMPONENTS
The simulation of impacts on structural components is considered essential in many ways.
Structural components may be used in to investigate the effect of an impact in terms of
structural configuration and material used. They can also be used to evaluate modelling
techniques, such as elements used, material properties and mesh densities.
In the way of simulating the structural response of deck and side shell during a collision, two
FE models of components have been developed. The purpose of these models is to achieve
certain collapse modes observed in full-scale collision tests and investigate the effect of
different mass and collision speed of the striking structure. These models include the collision
of a dihedral bow with deck and sideshell and the collision of a hemisphere bulb against a
panel.
The first model presented here consists of a dihedral bow striking part of a sideshell and deck
(see Figure 3.1). The panel consists of large web frames and longitudinal stiffeners. The
material is elastic-perfectly plastic steel and the thickness of all components is 10mm.
Figure 3.1: The dihedral bow and the sideshell-deck panel collision model.
The panel is clamped on its perimeter - except the contact side - while the bow is rigid and
free to move in space. The translational velocity of the centre of gravity of the bow is 2.57
m/sec along the z-axis. Preliminary runs have been performed, in order to check the mesh's
efficiency. In these runs, every element of the initial mesh has been split into four elements.
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 7
In that way, the original number of elements is quadrupled. As can be seen in figure 3.2 the
results for both meshes are quite alike.
0.00E+00
5.00E-02
1.00E-01
1.50E-01
2.00E-01
2.50E-01
0.00 0.02 0.04 0.06 0.08 0.10
Time [s]
D
i
s
p
l
a
c
e
m
e
n
t

[
m
]
mesh
mesh4
0.00E+00
2.00E+04
4.00E+04
6.00E+04
8.00E+04
1.00E+05
1.20E+05
1.40E+05
1.60E+05
0.00 0.02 0.04 0.06 0.08 0.10
Time [s]
D
e
f
o
r
m
a
t
i
o
n

E
n
e
r
g
y

[
J
]
mesh
mesh4
Figure 3.2: Deformation Energy vs. Time and Bow Displacement vs. Time for different mesh
densities
Mesh evaluation is a procedure that has proved critical for the collision modelling. This field
is still under investigation and considerable conclusions on how collisions should be
modelled have been reached using such techniques.
The runs performed for the above-mentioned model are set up using different bow masses.
The initial mass of the bow is equal to the total mass of the deformable structure. Runs are
performed for 1, 2, 3, 4, 6, 8 and 10 times the initial bow mass. For these runs, the
elastoplastic material's yield stress is independent to the strain rate. Additional runs are
performed for 10 times the initial bow mass, but with strain-rate dependency according to the
Cowper-Symonds rule. For these runs, the deformation energy absorbed by the deck and
sideshell structure is plotted against the displacement of the bow (figure 3.3). It is assumed
that the displacement of the bow represents the deformation of the structure. The bow is rigid
and thus non-deformable. In addition it can only move towards the structure or away from it
(that is, along the z-axis). It cannot rotate or move otherwise.
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 8
0. 00E+00
1. 00E+05
2. 00E+05
3. 00E+05
4. 00E+05
5. 00E+05
6. 00E+05
7. 00E+05
0.0 0.1 0. 2 0.3 0.4 0.5 0.6 0.7 0.8
Di spl acement [ m]
D
e
f
o
r
m
a
t
i
o
n

E
n
e
r
g
y

[
J
]
W
2 X W
3 X W
4 X W
6 X W
8 X W
1 0 X W
1 0 X W S R
Figure 3.3: Displacement vs. Deformation Energy
As can be seen in figure 1.2.2.3, in all cases except the strain-rate dependent case (10XW
SR), the displacements vs. deformation energy curves coincide. Additional runs have been
performed at this point in order to determine whether the velocity affects the solution.
Towards that direction, for the 10 times the initial bow mass case, the initial velocity of the
bow was increased and its mass was set to the initial value (1XW) in such a way that the
initial kinetic energy of the bow would remain the same. In figure 3.4, displacement vs.
deformation energy is plotted for four cases: large mass and large velocity, with or without
taking into account the strain-rate effect.
0.00E+00
2.00E+05
4.00E+05
6.00E+05
8.00E+05
1.00E+06
1.20E+06
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
Di spl acement [ m]
D
e
f
o
r
m
a
t
i
o
n

E
n
e
r
g
y

[
J
]
Large mass 10X
Large mass 10X SR
Large Vel oci ty 10X
Large Vel oci ty 10X SR
Figure 3.4: Displacement vs. Deformation Energy
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 9
Figure 3.4 shows that in the case where no strain-rate dependency is present, the results are
relatively closer. On the contrary, if strain rate is taken into account, the deformation energy
is greater in the case of greater initial velocity, for the same displacement.
The second model investigated simulates the collision of a hemisphere bulb against a panel.
In this case, the panel mentioned previously is subject to a collision with a hemisphere bulb
(figures 3.5 and 3.6). In this case, the longitudinal stiffeners and sideshell have been
removed. The diameter of the bulb is a little greater than the web frame spacing. The purpose
of this model is to examine the ability of the bulb to pierce the plate, observe the results on
the web frames and determine the conditions under which tripping of the web frames may
occur. The material of the panel is considered elastic-perfectly plastic and strain-rate
dependency of the yield stress is considered.
Figure 3.5: Hemisphere bulb in collision with panel.
Figure 3.6: Hemisphere bulb in collision with panel.
The panel is simply supported along the edges parallel to the y-axis.
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 10
The runs performed with this model can be divided in two groups: one group where the bulb
strikes the panel exactly on its geometric centre of gravity, and one where it is shifted
towards the negative x-axis. For both groups, there are two initial velocities considered for
the bulb: 7m/s and 3.5 m/s.
Figure 3.7: Strain results in time 0.5s for 7m/s, central and shifted strike
In the above figure, it can be seen that in the case where the bulb strikes on the centre of
gravity of the panel, there is no considerable bending along an axis parallel to the web
frames. On the contrary, in the case where the bulb is shifted, bending occurs in both
transverse directions of the plate. It can be also seen that in the case of the central strike, the
elements of the contact region fail and the bulb pierces the plate, whereas in the case of the
non-central the elements near the web frame fail and a flap-like part of the contact region is
cut off. This behaviour has also be observed in large-scale collision tests performed by TNO
(see fig. 3.8).
Figure 3.8: Results on large-scale plate with web frames.
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 11
4. SIMULATION OF COLLISION EXPERIMENTS
In order to validate the collision simulation capabilities and verify the extensionality of the
results, collision experiments [4] were simulated using FE. It has been observed that material
behaviour is an issue of great importance in the case of ship collisions. Material property
definition is at this time under investigation. As well as the material definition, a lot of work
is done on establishing mesh evaluation techniques and result consistency checks. The FE
model used to perform these tasks is a box-like structure with longitudinal and transverse
bulkheads (figures 4.1 to 4.4).
Figures 4.1 and 4.2 present the experimental set-up and struck ship's main dimensions. The
striking bow is a sharp rigid dihedral. The experimental set-up allows for the initial velocity
of the bow prior to the collision to be calculated. Two sets of experiments were performed:
one where the struck ship is clamped along its ends and one where the struck ship is free to
float in water. Tests are performed for different bow masses and velocities and struck ship
masses. The initial condition for this run is the striking vessel's initial velocity, equal to
3.4m/s, while it's mass is 28.6kg.
4.1 SIMULATIONS PERFORMED WITH THE STRUCK VESSEL CLAMPED
Two FE models are developed for this case. One for the whole struck vessel and one for the
area bound by two consecutive transverse bulkheads and a longitudinal bulkhead. Different
material definitions and mesh densities have been used for the several runs performed. These
materials include a linear elastic material, an elastic-perfectly plastic material, an elastoplastic
material and an elastoplastic with strain-rate dependency material. In Table 4.1 a result
summary for different materials and mesh densities is presented, along with experimental
results. The results are considered quite satisfactory. The more the material model approaches
a realistic behaviour, the more the results converge to the experimental values. For the full
model, the permanent deflection appears to be the same with the experimental result.
Deformational energy is also quite close, considering that the experimental result is the
subtraction of the final kinetic energy of the striker from the initial kinetic energy. In
addition, good convergence is achieved using different mesh densities. This is also obvious in
Figures 4.5 and 4.6. In the former, code names used to describe each curve correspond to
numbers 1-8 in Table 4.1. It can be seen that the curves for different mesh densities almost
coincide.
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 12
3800.00
1
6
9
2
.
0
0
30
Figure 4.1: Experimental set-up.
4in
1'
1'
4'
R 2in
Figure 4.2: Struck model main dimensions
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 13
Figure 4.3: Struck model: coarse and dense mesh
Figure 4.4: FE model of the whole structure
M.Samuelides & D.P.Servis
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 14
Experiment Elastic (1)* Elastic-Perfectly Plastic Elastoplastic
Coarse
Mesh(2)
Dense Mesh(3) Coarse Mesh
(4)
Dense Mesh (5)
Collision Duration [s] 0.0106 0.0132 0.0133 0.0123 0.0126
Maximum Force [N] /
Time Occ.[s]
47150.85/
0.005
13495.5/
0.0098
13158.49/
0.0101
16888.82/
0.0096
16814.88/
0.097
Final Deformational
Energy [J]
153 0.520 158.166 158.562 155.407 156.036
Percentage of the
initial Energy[%]
93% 0.31% 95.68% 96.01% 94.01% 94.57%
Deviation from
Experiments[%]
-99.66% 3.38% 3.64% 1.57% 1.98%
Permanent
Deflection [m]
0.0134 0.0003 0.0205 0.0222 0.0192 0.0204
Deviation from
Experiments[%]
52.98% 65.67% 43.28% 50.7%
Final striker's
Velocity [m/s]
-0.9 -3.391 -0.644 -0.616 -0.802 -0.783
Deviation from
Experiments[%]
-28.41% -31.6% -10.84% -13%
Table 4.1: Result summary
*Although the elastic material inappropriate for the simulation of the tests, it is included in the evaluation process to examine the energy equilibrium during collision, the
behaviour of the contact area and the determination of the deformation energy. The results for this case are presented in Table 1.2.2.1 just for completeness, though it would
be nice if ships could have such a response in actual collisions.
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 15
Force vs. Di spl acement
0.00E+00
5.00E+03
1.00E+04
1.50E+04
2.00E+04
2.50E+04
3.00E+04
3.50E+04
4.00E+04
4.50E+04
5.00E+04
0.000 0.005 0.010 0.015 0.020 0.025
Di spl acement [m]
F
o
r
c
e

[
N
]
sideresel (1)
si deresey (2)
sidereseyX2 (3)
sidereseyX3
si dereseyh (4)
sidereseyhX2 (5)
sidereseyhsr (6)
sidereseyhsrX2 (7)
dry reseyhsr (8)
Figure 4.5: Force vs. Bow Displacement curve
Force vs Bow Displacement
0.00E+00
2.00E+03
4.00E+03
6.00E+03
8.00E+03
1.00E+04
1.20E+04
1.40E+04
1.60E+04
0.00 0.01 0.01 0.02 0.02 0.03
Bow Disp. [m]
F
o
r
c
e

[
N
]
sideresey
sidereseyX2
sidereseyX3
Figure 4.6: Force vs. Bow Displacement curves example for different mesh densities
Force vs. bow displacement or indentation curves can be used as a convergence criterion
for the mesh density. This is extremely convenient, since the small solution time allows
for the impact simulation of several structural components of the struck area using
different meshes. In this way, convergence may be achieved and the appropriate mesh
density can be selected.
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 16
Figure 4.7: Von Mises stress distribution
The whole model impact simulation is also used for the estimation of the critical
modelling dimensions for a collision model. As can be seen in figure 4.7, significant Von
Mises stresses extend along three compartments of the struck ship. The side shell and
internal structure do not plastify in these regions but considerable membrane tension in
the side shell and bending of the transverse bulkheads adjacent to the impact area is
observed.
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 17
4.2 SIMULATION WITH THE STRUCK SHIP FREELY FLOATING
In this second set of experiments, the struck ship is free to float on the water surface. The
experimental setup does not differ a lot from the previous one. In table 4.2 the initial
conditions for both the experiments and corresponding simulations are defined.
Case Striking mass
[kg]
Struck mass
[kg]
Striking
velocity [m/s]
1 55.4 39.0 2.62
2 55.4 39.0 2.89
3 55.4 39.0 2.89
4 55.4 39.0 3.1
Table 4.2: Initial conditions for the tests
Additional mass was included in the form of node concentrated masses on the
longitudinal bulkheads. These masses simulate the additional weight due to measuring
devices. Such concentrated masses were also used to simulate the hydrodynamic added
mass. This mass is allocated on the side shell and boundary bulkheads up to the draught
of the floating ship in still water. A value of 40% of the struck mass was accounted for
this purpose, which is the usual case, but was also verified for the experimental model. In
tables 4.3 and 4.4 some experimental and numerical results are presented.
1
3
4
2
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 18
Experimental Numerical
Test Sway
velocity
[m/sec]
Yaw
velocity [s
-1
]
Deflection
[mm]
Sway
velocity
[m/sec]
Yaw
velocity
[s
-1
]
Deflection
[mm]
1 1.90 0.00 8.3 2.246 0.000 7.1
2 1.55 2.20 8.1 2.085 3.020 6.5
3
Not acquired due to
instrument failure
8.4 2.420 0.000 8.0
4 1.75 3.00 11.4 2.195 3.120 7.8
Table 4.3: kinematic results for the experiments and numerical simulations
Test Struck
kinetic
energy [J]
Striker
kinetic
energy [J]
Deformation
energy [J]
Collision
duration
[s]
Maximum
force [N]
Time of
occ. [s]
1 89.24 4.91 44.78 0.015 16704.26 0.0078
2 102.33 20.22 50.36 0.014 17140.55 0.0075
3 103.53 7.47 61.09 0.014 18331.94 0.0084
4 114.01 25.10 62.12 0.013 18165.36 0.0073
Table 4.4: Energy results for the numerical simulations
It can be seen in table 4.3 that the final velocities calculated by the numerical models are
greater than the experimental ones. This can be due to the approximation of the effect of
the surrounding water with concentrated masses. A concrete answer to this may be given
by combined fluid-strucutre and structure-structure interaction simulations (CFSS)
already in progress. The permanent deflection at the collision area is greater in all
experimental cases than the numerical ones. The difference though is not very significant
in the cases where a central impact is performed. This can also be due to the modelling of
added mass.
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 19
Case Initial
Energy
[J]
Maximum
Deformational
Energy [J]
Plastic
Deformational
Energy [J]
Elastic
Deformational
Energy [J]
Dry
165.308 164.238 115.818 48.420
Wet
165.308 94.075 51.173 42.903
Wet 1
190.144 91.601 44.782 46.819
Wet 2
231.353 93.924 50.364 43.559
Wet 3
231.353 111.197 61.086 50.112
Wet 4
266.197 107.884 62.121 45.764
Table 4.5: Energy results for the numerical simulations
In table 4.5 energy results from numerical simulations are presented. The two first cases
correspond to the whole model impact simulation when it is clamped and to an equivalent
simulation with the boundary conditions removed. It can be seen that the elastic
deformational energy is almost the same for all cases though the plastic deformational
energy varies significantly. It is also observed that the amount of elastic energy absorbed
in the cases where the struck ship is free to float is significant compared to the plastic
deformational energy. The elastic energy is absorbed by the side shell and bulkheads,
especially in their intersections (figure 4.8).
Figure 4.8: Von Mises stress distribution
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 20
This observation indicates that the full breadth of the test model should be modelled in
order to include all its significantly loaded areas. Combining this observation with the
previous conclusion on the modelling length, a collision FE model should at least extend
along three compartments and the full breadth of the ship. The areas apart from the struck
region absorb elastic deformational energy. If mostly the sideshell or bulkheads will
absorb this energy depends on the inertia forces exerted on the ship due to its movement.
In the case of a complicated ship structure, in order to reduce the number of elements and
nodes, equivalent formulations to the stiffened plates, such as orthotropic plates or
aggregated stiffeners, could be used for areas away from the collision area. A suitable
formulation based on orthotropic plates theory is under investigation.
5. SIMULATION OF COLLISION OF A LARGE RO-RO
Within DEXTREMEL - Design for Structural Safety under EXTREME Loads - project
the MS-DEXTRA has been modelled with finite elements. All the information and
drawings used for the development of the FE model were provided by ASTILLEROS
ESPAOLES [2], [5] as part of the deliverables of the project DEXTREMEL The
intention of this model is to use it in global collision runs. The analysis being done
includes the assessment of its structural integrity after the collision, definition of
optimum modelling of the panels at the impact region and possible substitution of other
panels with equivalent orthotropic plates. In this model, only a central part of the ship is
modelled in detail. Since the target vessel is a passenger ferry, no actual parallel body
exists. Nevertheless, at this stage it was selected to model part of the ship as it was
parallel, according to the midship section. The modelling length is about 48 structural
frames, beginning from frame 66 (bulkhead) until frame 114 (bulkhead) with the LCG
being at about frame 97.5. The corresponding length in the bodyplan begins at about
station 12.5 and ending at station 16.5. The bulkhead positions are at frames 66, 89, 102
and 114. The hull module modelled includes three compartments: the engine room and
part of the machinery room. The whole breadth of the ship is modelled in detail.
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 21
Figure 5.1: The Target Vessel FE model
Figure 5.2: Web Frame and Ordinary Frame
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 22
Figure 5.3: Bulkhead and Bow
In the previous Figures, views and sections of the target vessel are presented. In Figure
5.3 a FE of a bow is also presented. This is the bow of the target vessel and is used as a
striking bow. The target vessel FE model consists of 36391 nodes and 65811 shell and
beam elements. Shell elements were used for the sideshell, bulkheads, web frames and
stringers. Beam elements were used for all longitudinal stiffeners and upper decks'
stiffeners (figures 5.1 to 5.3). The struck ship for and aft of the three compartments,
which were modelled in detail, is modelled with elastic beams of the same area and
moments of inertia as the midship section. These beams are connected to the hull module
at the centre of gravity of the midship section using a rigid frame. In that way, the elastic
beams simulate the inertia effect of the rest of the ship on the hull module, as forces are
applied on it through the rigid frames. It was selected to use this configuration since the
standard rigid hull does not allow for horizontal bending of the ship and rotation of the
boundary bulkheads. The mass moments of inertia of the rigid bow are selected so that it
can only move along the x-axis (figure 5.3). Both ships are floating at the same draught,
but the struck vessel is in light ship condition. At this stage no added hydrodynamic mass
is used. The material used for the struck model is mild steel. The material model is
elastoplastic and includes material hardening but no strain-rate hardening. The failure
criterion is the maximum plastic strain. When the mean plastic strain of the element
reaches the value 0.18 then the element fails.
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 23
The results presented are acquired from a preliminary run. The following figures show
some equivalent stress results. These were recovered at times 0.15 sec, when the impact
force has a local maximum and 0.20 sec when there is extensive rupture of the side shell
due to bulb piercing. The images are rendered and shaded in order to make different
failure modes recognisable.
Figure 5.4: Relative position of the struck area and striking bow at 0.15secs: Equivalent
stress
Figure 5.5: Overall damage of the struck area at 0.15secs: Equivalent stress
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 24
Figure 5.6: Inside view of the damage at 0.15secs: Equivalent stress
Figure 5.7: Inside view of the damage caused by the bulb at 0.15secs: Equivalent stress
Tripping of
the stringer
Tripping of
web frame
Tripping of
Ordinary
frame
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 25
Figure 5.8: Views of the damage at 0.20secs: Equivalent stress
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 26
Figure 5.9: View of the damage caused by the bulb at 0.20secs: Equivalent stress
A first remark is that the selection of a rigid striking bow, especially of the kind used for
this analysis, may give an exessive estimation of the damage. The selected bow has a
bulb and a significant flare. In that way the contact force is not evenly distributed along
the contact area as it would with a nearly vertical bow. Of course, this is an actual bow
shape that the ship may encounter in a collision accident dyring her lifetime. The shape of
the bow and its rigidity cause very localised rupture. Further, at the upper part of the
contact area (figures 5.6 and 5.8) the rigid and sharp bulwark causes excessive horizontal
tearing of the sideshell; a damage that would had been different in case of a deformable
striking bow.
Before the rupture of the sideshell at the lower struck area, typical failure mechanisms
can be identified. Tripping of frames and stringer occurs as well as crushing of the web
frames near the main deck (figure 5.7). In figure 5.8 it can be seen that the extent of the
modelled structure seems to be sufficient, in order to obtain the stress field 0.2 sec after
contact. The stresses in areas apart from the struck one are below the first yield point and
seem to be greater along the intersection of the deck with the side shell (figure 5.8).
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 27
6. CONCLUSIONS
The paper refers to the simulation of ship collisions. The following related aspects are
considered to be essential for such a task:
(i) Selection of mesh: The selection of the mesh has been based on the convergence of
the force indentation curve, and on its adequacy to reproduce failure modes
observed in impact tests.
(ii) Effect of surrounding water: The surrounding water is modelled using concentrated
mass elements distributed over the wetted area, if a detailed model of the hull is
available. From the comparison of experimental and numerical results, it seems that
this formulation is adequate if the impact longitudinal position is close to the LCG
of the ship.
(iii) Extent of the detail model of the impacted hull: The modelling extent of the struck
ship should cover the plastic contact area and the elastically stressed areas in the
compartments, which are adjacent the heavily deformed contact area. Further there
are indications that an elastic stress field may be present over the whole breadth of
the model, and therefore it is considered appropriate to model the whole structure in
between two cross-sections. Equivalent orthotropic panel formulation should be
investigated in order to determine whether such panels can be used in areas away
from the struck region.
(iv) Failure criterion: The failure criterion, which has been used is based on the plastic
strain of an element. The value of the maximum plastic strain, which has been
selected is 18%. Such a value seems to result in realistic failure modes and in
results, which are close to experimentally obtained values of penetration. More
effort is needed in order to select a failure criterion, which would be used with
more confidence in the collision simulation.
The results obtained from the finite element simulation, will be used a) for the assesemnt
of the of collision behaviour of a ship under the defined collision scenario, b) for the
relative comparison of structural arrangments, and c) for the validation of analytical
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 28
techniques for collision analysis. Within the DEXTREMEL project the analytical
technique will be developed by DTU [18].
7. ACKNOWLEDGMENT
The work is part of the DEXTEMEL, Brite-Euram project financed by the Directorate
General XII of the European Commission. The Commission is thanked for its support.
8. REFERENCE
1. MacNeal-Schwendler Corporation, MSC/DYTRAN Version 4.0 user's manual.
2. Arias, C., DTR-1.0-AESA-03.98, Selection of design cases and operation scenarios,
Astilleros Espaoles.
3. Servis, D. P. & Samuelides M., Work done during the first six-month period of
DEXTREMEL project, Internal Report, NTUA.
4. Samuelides, M., Structural Dynamic and Rigid Body Response coupling in Ship
Collisions, PhD Thesis, Glasgow University 1984
5. Arias, C., DTR-2.0-AESA-07.98, Definition of relevant design cases and operation
scenarios, Astilleros Espaoles.
6. Kitamura, O., Comparative study on collision resistance of side structure, Marine
Technology, Vol.34, No.4, Oct 1997.
7. Kitamoura, O., Kuroiva, T., Kawamoto, Y. & Kaneko, E., A study on the improved
tanker structure against collision and grounding damage, Practical Design of Ships
and Mobile Units, 1998 Elsevier Science B.V.
8. Lee J. W., Petershagen, H., Rorup, J., Paik, H. Y. & Yoon, J. H., Collision resistance
and fatigue strength of New Oiltanker with Advanced Double Hull Structure,
Practical Design of Ships and Mobile Units, 1998 Elsevier Science B.V.
D.P.Servis & M.Samuelides National Technical University of Athens
Ship Collision Analysis Using Finite Elements
Paper Number (x) SAFER EURORO Spring Meeting, Nantes 28 April 1999 29
9. Lehmann, E. & Yu, X., On ductile rupture criteria for structural tear in the case of
ship collision and grounding, Practical Design of Ships and Mobile Units, 1998
Elsevier Science B.V.
10. Wang, G., & Ohtsubo H., Energy absorbing mechanisms involved in grounding,
MARIENV '95, p. 81-85.
11. B.C. Simonsen, B. C., Ship grounding on rock: I Theory, DCAMM, Oct.1997.
12. Lu, G. & Calladine, C. R., On the cutting of a plate by a wedge, Int. J. Mech. Sci. 32,
1990, p.p. 295-313.
13. Kuroiwa, T., Numerical Simulation of Actual Collision and Grounding Accidents,
Int. Conference on Design and Methodologies for Collision and Grounding Protection
of Ships, p.p. 7.1-7.12, San Francisco, 1996, SNAME, SNAJ.
14. Wierzbicki, T. & Thomas, P., Closed-form solution for wedge cutting force through
thin metal sheets, Int. J. Mech. Sci. 35, 1993, p.p. 209-229.
15. Pedersen, P. T. & Simonsen, B. C., Dynamics of Ships Running Aground, J. Mar.
Sci. Technol, 1995 p.p. 37-45.
16. Rodd, J. L., Observations on Conventional and Advanced Double Hull Grounding
Experiments, Int. Conference on Design and Methodologies for Collision and
Grounding Protection of Ships, p.p. 13.1-13.13, San Francisco, 1996, SNAME,
SNAJ.
17. Servis, D. P., Development of a model for the simulation of ship ship collisions,
National Technical University of Athens, Final year thesis, 1997 (in Greek), .
18. Pedersen, P. T. & Zhang, S., Collision Analysis for MS DEXTRA, Paper Number (x)
SAFER EURORO Spring Meeting, Nantes 28 April 1999. .

Вам также может понравиться