You are on page 1of 8

Collision and Grounding of Ships and Offshore Structures – Amdahl, Ehlers & Leira (Eds)

© 2013 Taylor & Francis Group, London, ISBN 978-1-138-00059-9

Ultimate strength of damaged hulls

C. Pollalis & M.S. Samuelides


National Technical University of Athens, Athens, Greece

ABSTRACT: The Goal Based Standards define that Classification Rules for Design and Construction of
Ships must provide provisions, which guarantee a reasonable level of residual strength after damage that results
from collision, grounding and flooding. As current Classification Rules do not cover this requirement, there
are numerous recent and on-going studies addressing this problem. Finite element is a powerful tool that may
be employed for the assessment of the ultimate strength of intact and damaged ship’s hulls. The most common
practice for such an analysis is to use an explicit FE code and to apply a rotation to the hull’s cross-section(s) in
order to determine a moment-curvature relationship. However, in the case of damaged hulls such a procedure
needs to be adapted in order to account for the rotation of the neutral axis and the simultaneous action of vertical,
horizontal and torsional moments. The paper presents an investigation regarding the influence of modeling
parameters on the ultimate strength, as determined using general FE codes. The procedure is tested against
published experimental results related to intact hulls and it has been extended to predict the ultimate strength of
damaged hulls. The sensitivity of the results on boundary conditions is investigated. The rotation of the neutral
axis and the bi-axial bending are studied to identify their effect on the reduction of the ultimate strength with
respect to the intact hull.

1 INTRODUCTION

The residual strength of a ship’s hull after collision,


grounding, flooding is an issue of recent considera-
tion by the IMO (2009) and IACS (2012). Tier II.5
of Goal Based Standards (IMO 2009) state that ships
should be designed to have sufficient strength to with-
stand the wave and internal loads in specified damaged
conditions such as collision, grounding or flooding
and residual strength calculations should take into
account the ultimate reserve capacity of the hull girder,
including permanent deformation and post-buckling
behaviour. These requirements that relate to the resid-
ual strength are not covered by the rules currently in
force and the effect of structural damage on the hull
girder capacity resulting from collision or grounding
is not assessed in the present form of CSR.
In 1995, ABS published guidelines for assessing
the structural redundancy for tankers and bulk car-
Figure 1. Assumed damaged section as a result of collision
ries (ABS 1995a, 1995b). The guides provide op-tional ABS (1995b).
criteria related to the residual strength in damaged con-
dition of hulls for Oil or Fuel Oil Carriers, Bulk or Ore
Carriers, combination carriers and Container Carriers,
which when satisfied by a vessel, she is assigned with strength and the shear strength of the damaged cross-
the notation RES. The assessment is based a) on hypo- sections. The aim is to eliminate, or at least minimize
thetical damage side or bottom damages (see Figure 1); the risk of a major oil spill or loss of ship due to a post-
b) on the calculation of the applied bending moment accident collapse or disintegration of the hull during
as a linear combination of the design still water and tow or rescue operation, when a ship gets damaged as
design wave bending moment in hogging and sagging a result of a collision or grounding accident.
and c) on the elastic section modulus of the damaged Paik et al. (1998) and Wang et al. (2002) developed
cross section. The assessment covers both the bending relative fast procedures to identify the possibility of

297
hull girder failure after collision and grounding dam-
age. The procedure of Paik et al. (1998), which can be
used for the prediction of the residual strength in the
early design stage, was applied to the residual strength
assessment of a PANAMAX bulk carrier after collision
and grounding damage.
Many authors have used an iterative-incremental
method frequently mentioned Progressive Collapse
Method – to calculate the ultimate strength of a
ship’s hull at both intact and damaged state, which is
based on the principles of the Smith method. Gordo
et al. (1996) compared the accuracy of the moment-
curvature curve based on the PCM with various exper-
imental and numerical results. They also attempted
to apply PCM method on asymmetrically-sectioned Figure 2. Model of box girder, Saad-Eldeen et al. (2010).
problems by considering cross sections of tankers and
container carriers with heeling angles.
Non-linear finite element codes are recently exten- Table 1. Details of structural components of models tested
sively used for the assessment of the ultimate strength by Saad-Eldeen et al. (2010).
of both intact and damaged hulls. The method offers
many possibilities as it allows the modeling of com- Structural member Dimension (mm)
plicated geometries, it may capture various buckling
Deck plating 4,09
modes and interactions between the structural ele- Port Side plating 3,95
ments of the hull, it can represent localized or more Starboard Side plating 3,85
uniform thickness diminution and initial imperfections Bottom plating 3,75
and it may be used for the assessment of the residual Stiffeners FB 25 × 4,35
strength of non-symmetric cross sections. Web frames  50 × 50 × 6,14
However the use of a non-linear finite element code Brackets 80 × 100 × 3,91
for the assessment of the ultimate strength of intact as
well as damaged hulls involves difficulties as well as
uncertainties as to the results that are obtained. The
performed. The way to apply the rotation is by cou-
present paper reports on an on-going re-search work
pling the edge nodes fore and aft with a control point
aiming in the investigation of the effect of parameters
that is rotated around the horizontal axis. The simula-
of FE modeling and FE simulation on the results. In
tions were performed with 4-node, reduced integration
particular the work addressed the effect of boundary
mesh, consisting of 10 mm square elements. Sim-
conditions, solver and material properties. In order to
ulations with 4-node full integration, 5 mm square
obtain some confidence in the modeling procedure,
elements did not produce any significant difference
the results of simulations of relevant tests performed
in the results but required a relative higher capacity
by two different research groups were compared with
of RAM. Thus, it has been concluded that the 10 mm
the experimental measurements.
reduced integration elements provide a reliable com-
bination of precision and RAM memory requirement.
As it concerns the definition of boundary conditions,
the only degrees of freedom left free are the rotation
2 SIMULATION OF EXPERIMENTS
around the horizontal axis on both control points and
the translation along the longitudinal axis of the one
2.1 Simulation of tests of Saad-Eldeen et al. (2010)
control point to eliminate axial reaction forces. From
As a first step of the investigation of the effect of the investigation using various boundary conditions,
modeling parameters and in particular of the mesh it was found that neither the restrictions of the rota-
size, boundary conditions and solver on the simula- tion around the vertical and longitudinal axes nor the
tion, the present work simulated the tests reported by restrictions of the translation along the transverse and
Saad-Eldeen et al. (2010). A box girder with length, vertical axes affect the results. It is noted that due to
breadth and depth of 1400 mm, 800 mm and 600 mm symmetry it was expected that the restrictions of the
respectively (see Figure 2), was subjected to four point above mentioned rotations and the translation along
bending using a hydraulic jack. The dimensions of this the transverse axis not to have an influence on the
girder are presented in Table 1. response. The material of the specimen was mild steel
The aim is to determine the appropriate model- and it has been assumed to be elastic, perfectly plastic
ing parameters and to investigate if the simulation with yield stress of 245 MPa.
produces results that correlate well with the experi- In order to determine the deformation pattern and
mental results when the test is simulated through the stress field under static equilibrium, we opted for the
application of rotation to the end sections rather than use of the implicit version of ABAQUS in combination
forces as in the four point bending test that has been with the Riks solution algorithm.

298
Figure 3. Bending moment versus curvature: comparison
of measurements and numerical results. Figure 5. Sections of tested specimens by Kuo et al. (2003).

Figure 6. Side view of tested specimens by Kuo et al. (2003).

Figure 4. Neutral axis displacement of symmetric cross- Table 2. Material properties of specimens tested by Kuo
section. et al. (2003).

The numerical results and the results from the tests Plate thickness Yield stress Young’s mod. Poison’s
mm MPa GPa ratio
in terms of applied bending moment vs. curvature are
shown in Figure 3. The numerical results are obtained
3.05 293 211 0.277
in terms of bending moment vs. rotation and the 4.25 269 212 0.281
abrupt changes that appear in the respective curve 5.60 321 211 0.289
are attributed to the differentiation of the rotation to
express the bending moment versus the curvature. As
we can see from the Figure 3, we have attained almost
the same value of ultimate strength with the exper- The dimensions of the specimens are shown in Fig-
iment using the FE. However, there is also a great ures 5–6 and the material properties and thicknesses in
difference after deck collapse. This might be caused Table 2. The four different models of midship sections
due to the lack of the brackets in our model and the portray a simplified Tanker (MST), a Double bottom
difference in method of load application between the Tanker (MST), a Containership (MSC) and a Bulk Car-
experiment and the simulation. We can also notice that rier (MSB). The results of the simulations together
the curve produced by the experiment has a starting with the experimental results and the numerical pre-
point that differs from O(0, 0). This happens because dictions from Saad-Eldeen et al. (2010) are shown in
the experimental structure has an initial displacement Table 3. As we can see, the results obtained by the sim-
of u = 0.9 mm due to the weight of the equipment. ulation conducted in this present work show in most
The vertical translation of the neutral axis because cases a better correlation with the experimental results,
of the plastic region which appears at the deck plat- compared to the theoretical predictions presented in
ing and stiffeners can be shown in Figure 4. The total Kuo et al. (2003).
translation of the axis is 0.321 m − 0.142 m = 0.179 m.

3 RESIDUAL STRENGTH AFTER COLLISION


2.2 Simulation of tests of Kuo et al. (2003)
DAMAGE
The procedure of simulation that has been followed
for the tests of Saad-Eldeen et al (2010) has been also Following the simulations of the bending response
ap-plied for the simulation of the tests reported by Kuo of the intact specimen, it was attempted to perform
et al. (2003). simulations on a model of the specimen, which was

299
Table 3. Tests of Kuo et al. (2003): Measurements and theoretical predictions.

Description of test (1): Test (2): Beghin (3): Kuo (4): FE


(see Figure 5) ton-m ton-m et al. ton-m (2)/(1) (3)/(1) ton-m (4)/(1)

MST-1 sag/hog (5) 94.5 110 109.95 1.164 1.163 105.2 1.113
MST-2 sag/hog (5) 58.8 58.9 58.97 1.001 1.003 68.3 1.162
MSD- sagging 60.5 97.3 97.96 1.608 1.619 74.15 1.226
MSD- hogging 85.5 97.3 97.96 1.138 1.146 90.4 1.057
MSB- sagging 49.1 86.6 88.03 1.764 1.793 62.2 1.267
MSB- hogging 68.5 86.6 88.03 1.264 1.285 81.31 1.187
MSC- sagging 113.5 104.2 103.9 0.918 0.915 103.98 0.916
MSC- hogging 88.0 104.2 103.9 1.184 1.181 95.56 1.086

(1): test measurements.


(2) & (3): theoretical predictions from Kuo et al. (2003).
(4): from FE simulations.
(5): The plate thickness of MST-1 and MST-2 is 4.25 mm and 3.05 mm respectively (see Table 2).

Figure 7. Collision damage at end bay.

Figure 9. Moment versus rotation for damaged girder.

• Damage breadth (assuming that the ship hull col-


lided with a ship’s bow of 30 degrees angle):
h · tan θ = h · tan 30◦ ≈ 0.14 m
• Damage length = longitudinal distance between
two successive web frames = 0.400 m.
Figure 9 shows the ultimate strength for the dam-
aged box girders in comparison to the ultimate strength
in intact condition. The red color curve corresponds to
the edge damage, the blue color curve to the central
damage and finally the green color curve corresponds
Figure 8. Collision damage at central bay. to the intact state curve. As it may be seen, the residual
strength, i.e. the ultimate strength in damaged condi-
tion, is higher when the damage is at the end bay in
comparison with the damage at the center bay. This
assumed to be damaged as a result of a collision. Two
is attributed to the fact that one end section of the
damage scenarios were investigated: For the first sce-
damaged length is close to the cross section, where
nario the damage was considered at one of the end
rotations are applied and which is kept undeformed
bays (edge damage) and for the second at the central
during the application of rotations. Figures of both
bay (central damage). Illustrations of these two dif-
intact and damaged box girders can be seen in Figures
ferent scenarios can be seen in Figures 7 and 8. The
10, 11 and 12. As we can notice, the main failure mode
dimensions of the damage are as follows:
is buckling of the deck plating. Buckling and tripping
• Damage height (see Figure 1): h = 0.40 · D = 0.40 · of the upper stiffeners is also observed. Severe plas-
0.600 m = 0.24 m (from the deck) tic strain can be noted at the middle section Figure 12

300
Figure 10. Undamaged structure: Stress contour.

Figure 11. Undamaged structure: Strain contour.

Figure 13. Collision damage at central bay: translation and


rotation of Neutral Axis.

Table 4. Calculation of angles of rotation: cross-section at


the middle of the damage length.

Inc. 9 tan ϕ1 = 0.0581 ϕ1 = 3.27◦


Inc. 11 tan ϕ2 = 0.0579 ϕ2 = 3.32◦
Inc. 13 tan ϕ3 = 0.0556 ϕ3 = 3.18◦
Inc. 16 tan ϕ4 = 0.0556 ϕ4 = 3.18◦
Figure 12. Collision damage at central bay: Rotations Inc. 24 (Mmax ) tan ϕ5 = 0.0557 ϕ5 = 3.19◦
around vertical axis.

indicates that there is no bending around the vertical length respectively and at various level of rotation of
axis-horizontal bending. the cross-sections. Inc 24 corresponds to the point of
When an asymmetric section is subjected to bend- maximum bending moment. Figure 13 refers to the
ing moment, the Neutral Axis does not only translate case of central damage. As we can notice from Tables
but it is also rotated. The final position of the Neutral 4 and 5, where the rotation of the Neutral Axis is cal-
Axis is determined by the inclination of the vector of culated, the rotation at the end of the damage length is
the moment with respect to the principal axes of the lower in relation to the rotation of Neutral Axis at the
cross-section. However, as the cross sections within the middle of damaged length.
damaged length may not behave uniformly, the trans- As it concerns the boundary conditions, rotations
lation and rotation of the Neutral Axis may differ from were applied about the transverse horizontal axes, and
section to section. For the case studied, the position the longitudinal and vertical displacements were let
of the Neutral Axis can be shown in Figure 13 for the free along the respective axes. In the case of the dam-
cross section at the end and the middle of the damaged age in the center bay, runs were also performed with the

301
Table 5. Calculation of angles of rotation: cross-section at
the end of the damage length.

Inc. 9 tan ϕ1 = 0.0314 ϕ1 = 1.80◦


Inc. 11 tan ϕ2 = 0.0318 ϕ2 = 1.82◦
Inc. 13 tan ϕ3 = 0.0322 ϕ3 = 1.85◦
Inc. 16 tan ϕ4 = 0.0305 ϕ4 = 1.75◦
Inc. 24 (Mmax ) tan ϕ5 = 0.0302 ϕ5 = 1.73◦

Table 6. Values of ultimate and residual strength.

Maximum Ratio of
Bending max. BM
moment – BM Rotation at over BM Figure 14. Intact cross-section of tanker.
Case (kN-m) max BM of intact

Intact 595.5 0.0026 1


Damage mid bay 389.7 0.0020 0.655
Damage end bay 418.7 0.0024 0.703

Table 7. Main particulars of the tanker.

Length OA 264.68 m
Length BP 256.50 m
Breadth 42.50 m
Depth 22.0 m
Draught 15.0 m

Figure 15. FE model of intact tanker between wed-frames.


rotations about the vertical axis of both cross-sections
being constrained. In this particular case the difference
of the residual strength was negligible, approximately
2%. In the case of the damage at the end bay the reac-
tion forces were approximately 50 kN. A comparison
of the ultimate and residual strength can be seen in
Table 6.

4 ULTIMATE AND RESIDUAL STRENGTH


OF A TANKER

Oil transportation with tankers involves environmen-


tal risks because of contact accidents. A tanker that is
damaged as a result of a collision, must have the nec- Figure 16. Damaged cross-section of tanker.
essary residual strength to survive and not cause an
extended environmental pollution.
In this section, we present the results of the deter- During preliminary simulations employing the
mination of the ultimate strength of a tanker in intact implicit version ABAQUS, in the case of the intact
and damaged condition using FEA. The damage shape tanker, we noticed a sharp fall of the strength after
and size we simulated, is that proposed byABS (1995a) entering into the elastic-plastic region. This led to a
(see also Figure 1). The longitudinal span of the dam- disability of the implicit method to converge at a cer-
age covers the distance between two successive web tain solution. Therefore, we applied the problem using
frames. The ship modeled has main particulars pre- the explicit method. The total simulation time was
sented in Table 7. The material is A-36 steel for the hull set equal to two seconds with a time increment of
plating and deck’s and bottom and double bottom stiff- 1E-006 second. Within the simulation time the rota-
eners and mild steel elsewhere. Figures of the intact tion applied was set to 0.003 rad. The element shape
and damaged midship section at the ABAQUS.cae and size is 100 mm square element for l = 4000 mm
environment are shown in Figures 14 to 17. model longitudinal span.

302
Figure 17. FE model of damaged tanker between
wed-frames. Figure 19. Material model.

Figure 18. Bending moment versus rotation for tanker.

As it concerns the boundary conditions of the intact Figure 20. Model of damaged tanker: Displacement of
ship hull, we apply the same as those applied in the Neutral Axis.
case of the intact box girders (see sections 2.1 and
2.2), i.e. the simulation was performed by imposing
moment versus rotation curve that was obtained using
rotations around the control points, the longitudinal
an elastic–rigidly plastic material.
translation was let free at one control point and all the
From Figures 18 and 19 it is observed that when the
other degrees of freedom were restricted.
material has hardening the resistance of the ship’s hull
For the damaged hull, we conducted a study to
after yielding is higher while in case of a non-hardened
investigate how the boundary conditions affect the
material there is a sharp drop of its strength. Finally we
residual strength of the hull. Initially, we simulated the
notice that at about θ = 0.005 rad of rotation, the curves
ship’s bending with the same boundary conditions we
of damaged non-hardened midship section is found
used for the intact structure. In subsequent simulations
above the curve of the intact non-hardened curve.
a) an end section was let free to rotate about a vertical
Finally it has been checked if the ultimate and resid-
axis, and b) an end section was let free to rotate about
ual strength is higher than the design bending moment
the vertical and additionally about the horizontal axis.
of the ship. Considering that the design wave bend-
The bending moment vs. rotation curves that have
ing moment is calculated equal to 5.08 GN · m and
been obtained from the different simulations do
the maximum permissible sagging bending moment
not illustrate any significant difference. The bend-
in still water equal to 407,000 t · m or 3.99 GN · m,
ing moment versus rotation curve of the intact and
the design bending moment is 9.07 GN · m. The value
damaged hulls are shown in Figure 18. From the
is less than the ultimate strength but higher than the
simulations the ultimate strength of the intact hull
residual strength for the damage case considered.
was found to be equal to 9.55 GNm and the residual
strength of the damaged hull equal to 7.77 GNm.
A further aspect that has been investigated is the
effect of the material model on the results. The results 5 CONCLUSIONS
presented for the tanker so far were obtained with a
material having the true-stress strain curve presented The paper addresses the ultimate strength of a ship’s
in Figure 19. Figure 18 further includes a bending hull in intact and damaged condition, the latter being

303
frequently described as “residual strength”. The simu- period is set for the response. However, this leads to
lations to obtain the ultimate and residual strength were an increase on the total CPU time and needs consider-
performed using the widely used FE code ABAQUS. able more computational power. Further work is being
As far as the intact hull is concerned, the hulls were performed to investigate if explicit and implicit codes
modeled with 4 node, reduced integration shell ele- produce comparable results, or otherwise if the explicit
ments with length over thickness ratio between 2.5 code may produce a static solution in the post-ultimate
and 5. In a small scale model the implicit version of the strength region.
FE code resulted in an ultimate strength that compared
well with the ultimate strength that was measured in the
tests. Regarding the boundary conditions, the restric-
tion of rotation around the vertical and longitudinal REFERENCES
axes and the translation along the transverse axes, may
ABS. 1995a. Guide for the Assessing Hull Girder Residual
simulate the state of pure bending without generating Strength for Tankers. Houston, Texas.
any other reaction moments or reaction forces apart ABS. 1995b. Guide for the Assessing Hull Girder Residual
from the vertical bending moment. Strength for Bulk Carriers. Houston, Texas.
From the simulations of the damaged structures, Saad-Eldeen, S.; Garbatov, Y. & Guedes Soares, C. 2010.
we concluded that in the case of the box girder and the Experimental Assessment of the Ultimate Strength of a
tanker that have been investigated, it was not important Box Girder Subjected to four-point Bending Moment.
to let the rotation round vertical and longitudinal axes PRADS2010. Rio de Janeiro, Brazil.
free. The conclusion is based on the comparison of Gordo, J.M. & Guedes Soares, C. 1996. Approximate method
to evaluate the hull girder collapse strength. Marine
the bending moment vs. rotation curves obtained from
Structures 9(1): 499–70.
the simulations and the level of the ultimate strength. IACS 2012. Draft Harmonized CSR for Industry Review.
Further tests should be performed in other cases in IMO. 2009. Goal Based New Ship Construction Standards
order to investigate if such a behavior is observed with Report of the Working Group MSC 86/WP.5. International
larger damages and to assess the level of the reaction Maritime Organization.
moments about the longitudinal and transverse axes. Kuo, H.S. & Chang, J.R. 2003. A simplified approach to
A further issue that has been investigated is whether estimate the ultimate longitudinal strength of ship hull.
an implicit code may converge to a static solution. This Journal of Marine Science and Technology 11(3): 130–
was not a problem when using the implicit code for 148.
Paik, J.K.; Thayamballi, A.K. & Yang, S.H. 1998. Residual
the case of the box girders. However, in the case of a
strength assessment of ships after collision and grounding.
real ship’s hull, due to its sudden loss of strength, the Marine Technology 35(1): 38–54.
implicit code did not converge and the explicit ver- Wang, G.; Chen, Y.; Zhang, H. & Peng, H. 2002. Longitu-
sion of the code was used. An explicit method could dinal strength of ships with accidental damages. Marine
converge to the static solution if a relatively long time Structures 15: 119–138.

304

You might also like