Professional Documents
Culture Documents
Advances in Marine Structures: MARSTRUCT - Indb I MARSTRUCT - Indb I 2/18/2011 5:39:02 PM 2/18/2011 5:39:02 PM
Advances in Marine Structures: MARSTRUCT - Indb I MARSTRUCT - Indb I 2/18/2011 5:39:02 PM 2/18/2011 5:39:02 PM
Editors
C. Guedes Soares
Instituto Superior Técnico, Technical University of Lisbon, Portugal
W. Fricke
Hamburg University of Technology (TUHH), Germany
All rights reserved. No part of this publication or the information contained herein may be reproduced,
stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, by
photocopying, recording or otherwise, without written prior permission from the publisher.
Although all care is taken to ensure integrity and the quality of this publication and the information
herein, no responsibility is assumed by the publishers nor the author for any damage to the property or
persons as a result of operation or use of this publication and/or the information contained herein.
Table of contents
Preface xi
Organisation xiii
Fatigue strength
Fatigue of high-speed aluminium ships: A master curve formulation 267
J.H. den Besten & R.H.M. Huijsmans
Stress and strain-based approaches for fatigue life evaluation of complex structural details 277
M. Biot & L. Moro
Different finite element refinement strategies for the computation of the strain energy density
in a welded joint 289
C. Fischer, A. Düster & W. Fricke
Round-robin on local stress determination and fatigue assessment of load-carrying
fillet-welded joints 295
W. Fricke, M. Codda, O. Feltz, Y. Garbatov, H. Remes, G. Risso, C. Rizzo & J. Romanoff
Fatigue estimation of a ship structural detail 303
S. Giuglea, A. Chirica, I. Chirica & E.F. Beznea
vi
Dynamic analysis
Methods for hull structure strength analysis and ships service life evaluation, under
extreme hydroelastic wave loads, for a large oil-tanker 459
L. Domnisoru & A. Chirica
Dynamic analysis in the marine environment considering FSI—Ship-like structure case 473
A. Martínez Cimadevilla
vii
viii
ix
Preface
This book collects the papers presented at the 3rd International Conference on Marine Structures,
MARSTRUCT 2011, which was held in Hamburg, Germany 28 to 30 March. This Conference follows
up from the initial ones that were held in Glasgow, Scotland and in Lisbon, Portugal, respectively in 2007
and 2009. These conferences aim at bringing together researchers and industrial participants specially
concerned with structural analysis and design of marine structures. Despite the availability of several con-
ferences about ships and offshore structures, it was felt that there was still no conference series specially
dedicated to marine structures, which would be the niche for these conferences.
The initial impetus and support has been given by the Network of Excellence on Marine Structures
(MARSTRUCT), which was funded by the European Union from 2005 to 2010, bringing together
33 European research groups from Universities, research institutions, classification societies and indus-
trial companies that are dedicated to research in the area of marine structures. With the end of this
EU project, a new organisation was created to maintain the cooperation ties among the groups that work
in this general area. The MARSTRUCT Virtual Institute was created in 2010, with the aim of being an
association of research groups interested in cooperating in the field of marine structures. It started with
the same members as the participants in the Network of Excellence but in the near future it is open to
accept the membership of other European groups that have the same aims.
The Virtual Institute is organised in the following Technical Committees:
• Methods and tools for establishing loads and load effects
• Methods and tools for strength assessment
• Experimental analysis of structures
• Materials and fabrication of structures
• Methods and tools for structural design and optimisation
• Structural reliability, safety and environmental protection
which in turn are divided in subcommittees. The aim is to promote the exchange of information and the
cooperation in these subject areas. This can take the form of promoting comparative or benchmark stud-
ies in various subjects, promoting joint research activities and joint research projects, organising short
specialised courses, workshops and conferences.
In particular the Virtual Institute will take the responsibility of organising the MARSTRUCT Confer-
ence biannually in different countries, starting from the present one that is already organised in a coop-
eration between the MARSTRUCT Virtual Institute and the Hamburg University of Technology in a
scheme that is planned to be continued in the future: the Virtual Institute will be responsible for the prepa-
ration of the technical programme and processing of the papers and the host country organization will be
responsible for the conference organization and management.
Despite being organised in Europe, this Conference is not meant to be restricted to European attendees
and a serious effort has been made to involve in the planning of the Conference participants from other
continents that could ensure a wider participation, which is slowly happening.
The conference reflects the advances that have been made in the last years within its domain including
the full range of methods and modelling procedures for the structural assessment of marine structures.
Various assessment methods are incorporated in the methods used to analyze and design efficient ship
structures, as well as in the methods of structural reliability to be used to ensure the safety and environ-
mental behaviour of the ships. This book deals also with some aspects of fabrication of ship structures.
This book includes almost 80 papers, which are organised into the themes that correspond to the Virtual
Institute Technical Committees, as listed above. The papers were accepted after a review process, based on
the full text of the papers. Thanks are due to the Technical Programme Committee and to the Advisory
xi
xii
Organisation
Conference Chairmen
Prof. Wolfgang Fricke, Hamburg University of Technology (TUHH), Germany
Prof. Carlos Guedes Soares, IST, Technical University of Lisbon, Portugal
Advisory Committee
Prof. F. Brennan, Cranfield University, UK
Prof. A. Campanile, University of Naples, Italy
Prof. G. Chen, Shanghai Jiaotong University, P.R. China
Dr. F. Cheng, Lloyd’s Register, UK
Prof. S.-R. Cho, University of Ulsan, Korea
Prof. Y.S. Choo, Nat. Univ. Singapore, Singapore
Prof. W.C. Cui, CSSRC, P.R. China
Prof. C. Daley, Memorial University, Canada
Dr. M. Dogliani, Registro Italiano Navale, Italy
Prof. A. Ergin, ITU, Turkey
Prof. S. Estefen, COPPE/UFRJ, Brazil
Prof. M. Fujikubo, Osaka University, Japan
Prof. T. Fukasawa, Osaka Prefecture University, Japan
Prof. C.-F. Hung, National Taiwan University, Taiwan ROC
xiii
xiv
Thomas Gazzola, Nikita Tryaskin, Nicolas Moirod, Jérôme De Lauzon & Louis Diebold
Bureau Veritas, Marine Division, Research Department, Neuilly-Sur-Seine Cdx—France
ABSTRACT: This paper presents a numerical investigation of the influence of raised invar edges on
wet drop tests pressures. The aim of this study is to evaluate the capabilities of the CFD software Open-
FOAM ([6]) to deal with fluid structure impact problems in the context of sloshing inside membrane
tanks. More precisely, the objective is to evaluate OpenFOAM capabilities for water drop tests simulations
for a smooth wedge on one hand and a wedge equipped with invar edges (like these which equip the Cargo
Containment System NO96 produced by GTT) on the other hand, both falling into calm water without
inclination. The obtained numerical results are then compared to Wagner’s solution for the smooth wedge
and to experimental measurements for the wedge with edges presented in ([2]). Numerical simulations are
in very good agreement with Wagner’s solution and experimental results ([2]) showing the OpenFOAM
capability to deal with fluid structure impact problems in the context of sloshing inside membrane tanks.
The agreement between numerical and experimental results confirms also that raised invar edges tend to
enhance the magnitude of sloshing pressures. This confirmation emphasizes the importance of consider-
ing the physics of invar edge effects in defining the design pressure to be used in assessing the integrity of
membrane LNG tanks.
Figure 1. Drop test scheme. Figure 2. Wedge configuration for scale 1:6.
700000
p10
600000 p11
p12
p13
500000
p14
Figure 3. Mesh used for the smooth wedge simulation 400000
p15
p16
p17
300000
p18
N°Sensor Before 1st edge (mm) Between edges (mm) N°Sensor 100000
1 151.611 234.611 10 0
2 160.833 243.833 11
3 170.056 253.056 12 –100000
p1 p10
1.2e+006 p2 p11
p3 p12
p4 p13
1e+006
p5 p14
p6 p15
800000
pressure (Pa)
p7 p16
p8 p17
600000 p9 p18
400000
200000
1e+006 p7 p16
p8 p17
800000
p9 p18
600000
400000
200000
–200000
0 0.005 0.01 0.015 0.02 0.025
time (s)
p1 p10
1.6e+006 p2 p11
1.4e+006
p3 p12
Figure 11. Experimental drop test with invar edges (0).
p4 p13
1.2e+006 p5 p14
p6 p15
1e+006
that the first edge does not have any influence on
pressure (Pa)
p7 p16
p8 p17
800000
600000
p9 p18
upstream pressure sensors.
400000
OpenFOAM well predicts the over peak impact
200000
pressure (in comparison with smooth wedge) before
0 the first edge (see sensor #8 & #9).
-200000 OpenFOAM well predicts the under peak impact
0.004 0.005 0.006 0.007 0.008 0.009 0.01 0.011
time (s)
pressure (in comparison with smooth wedge) just
after the first edge (see sensor #10 & #11).
Figure 8. Pressure time histories for all pressure sensors OpenFOAM well predicts the over peak impact
calculated by OpenFOAM. Drop test simulation of the pressure (in comparison with smooth wedge)
wedge with invar edges (zoom for the right figure). before the second edge (see sensor #17 & #18).
1:6, 0.6m
1:6, 0.8m
and the obtained time pressure histories were
1.5 1:20, 0.12m
1:20, 0.18m
analysed. As these time pressure histories signals
1
1:20, 0.24m
1:35, 0.069m
are noisy, it was decided to filter these signals with
1:35, 0.086m
1:35, 0.103m
a low-pass filter (2.5 kHz, 8th order Butterworth).
0.5
Experimental
In order to quantify the influence of this low-pass
0
9 10 11 12 13 14 15 16 17 18 19
filtering on peak impact pressures of interest, this
Number of the Pressure Sensor filter was applied on the time pressure histories
obtained for the smooth wedge. It was then shown
Figure 12. OpenFOAM calculations (symbols) and that this low-pass filtering has a limited influence
experimental (0) (black line) maximum impact pres- on peak impact pressures of interest.
sures ratios between wedge with invar edges and the Then, the time pressure histories were com-
smooth one. pared qualitatively and quantitatively with the
experimental ones obtained in (0). The agreement
is very good considering the complexity of such
3.3 Results for the ratio (invar case/Wagner) simulation.
As a conclusion, it can be said that OpenFOAM
All these qualitative observations can be quanti- can be used to simulate drop tests simulations not
fied by comparing the ratio between the filtered only for smooth wedge but also for wedge equipped
maximum pressures for the wedge with invar edges with invar edges which represent a much more
and the Wagner’s solution as it is done in (0). complicated case. The next step is to perform drop
The agreement between OpenFOAM and Exxon tests simulations for wedges equipped with corru-
experiments (0) curves is satisfactory. Furthermore, gations (like these which equip the Cargo Contain-
precise information on the experimental procedure ment System MarkIII produced by GTT).
and its post-processing lack to perform clearer
quantitative conclusions. Hence, we can conclude
that OpenFOAM well predicts qualitatively and REFERENCES
quantitatively the peak impact pressures for the
drop test simulation of a wedge equipped with Gazzola, T. “Contribution au problème d’impact non
invar edges. linéaire: le problème de Wagner couplé”, Ecole Cen-
trale Paris 2007 (written in English).
He, H., Kuo, J.F., Rinehart, A. & Yung, T.W. “Influ-
ence of Raised Invar Edges on Sloshing Impact
4 CONCLUSIONS
Pressures”, 1st Sloshing Dynamics Symposium,
ISOPE 2009, Vol. 3, www.isope.org
The aim of this study was to check/confirm the Molin, B. “Hydrodynamique des Structures Offshore”,
findings presented in (0) from one hand and to Editions Technip.
evaluate the capabilities of the CFD software OpenFOAM, v.1.7.1, www.openfoam.com
OpenFOAM to deal with fluid structure impact Viviani, M., Brizzolara, S. & Savio, L. “Evaluation
problems in the context of sloshing inside mem- of slamming loads using smoothed particle hydro-
brane tanks from the other hand. dynamics and Reynolds-averaged Navier–Stokes
More precisely, the objective was to evaluate methods”. Journal of engineering for the maritime
environment, 2009, 223:17–31, ISSN: 1475-0902, doi:
OpenFOAM capabilities for drop tests simulations
10.1243/14750902 JEME131.
for a smooth wedge on one hand and a wedge Wagner, H. “Über Stoss und Gleitvorgänge an der Ober-
equipped with invar edges (like these which equip fläche von Flüssigkeite ”, ZAMM, Vol. 12, 193–215.
the Cargo Containment System NO96 produced
by GTT) on the other hand, both falling into calm
water without inclination.
In section 3-1, the drop test simulation of the
smooth wedge was performed and the obtained
time pressure histories were then compared to the
Wagner’s analytical solution (valid for this studied
configuration). The agreement between the two
solutions is excellent and thus OpenFOAM drop
test simulation for smooth wedge is validated.
ABSTRACT: We present a modal approach to the coupled computation of the fluid flow and the
motion and elastic deformation of a floating body. For the fluid part, i.e. the transient viscous free sur-
face flow around the structure, the open souce code OpenFOAM is adopted. For the structure part, i.e.
the fully non-linear six degree of freedom rigid body motion and the linear elastic deformations due to
the forces exerted by the fluid on the structure, an in-house code has been developed and linked to the
OpenFOAM solver. An algorithm maps fluid forces to the finite element model nodes of the structure
and maps the structure motion/deformations to the vertices of the fluid grid. In our modal approach,
the structure deformation is represented by a linear combination of a small number of precomputed
modes (in the current study, the eigenmodes of the structure), resulting in an ODE for the “modal coef-
ficients”. This approach proves to be computationally efficient: on the one hand, the size of the equation
system is much smaller than the full finite element equations, on the other hand, the restriction to low-
frequency eigenmodes suppresses high-frequency oscillations encountered in direct FE computations,
allowing for larger time step sizes and in general a more stable coupling. Also, suitable selection of the
modes allows for focussing on “critical” structure deformations e.g. for fatigue assessment. To this end,
deformations corresponding to the most relevant load cases according to classification rules can be used
as modes, too.
The first six modes φi, i = 1, …, 6 will be assumed the angular acceleration stiffness
to represent the rigid body motion of the ship. The
time dependent modal coefficients ri, i = 1, …, n are H ∫V ρ
T
( )dV , (8)
assembled in the vector
the centripetal forces
r = [ri, …, rn]T (3)
A good choice for representing the global defor- Gr = 2∫ ρ ΦT (Ω × Φ )dV r, (10)
mation of the ship typically are the lowest eigen- V
modes which result from the modal decomposition
of the homogeneous, undamped and unsupported and the mass matrix
system
∫V ρ (11)
T
M dV .
Mü + Ku = 0. (5)
All terms of the equation of motion in the mov-
This leads to the “dry free vibration” mode ing coordinate system have been given above for
shapes φi, i =1‥n and the corresponding circular completeness. In ship applications, however, the
eigenfrequencies ωi, i = 1‥n. Note that since the forces resulting from the terms containing G, H,
ship is assumed to be unsupported ωi = 0, i = 1‥6 and Z are typically negligibly small whereas R
corresponding to six rigid body modes. While con- results in a significant contribution.
ceptually similar to “generalized” or “principal Now adding the stiffness
coodinates” in classical mechanics, we prefer the
term “modal coefficients” in order to stress the
∫V ρ (12)
T
K dV
approximation by a small number of modes.
10
) Z (Ω )]r by requiring
Mr + [C G (Ω)]
)]r [ K + H (Ω
(13)
= ∫ φT FdA
d R( , , U ). det (K − ω2M) = 0,
A
If the modes used for the projection are nor- resulting in one rigid body eigenvalue/eigenvector
malized and orthogonal with respect to the mass
matrix, M becomes the unit matrix. In case moreo- ω 12 0 φ1 (1/ 2 )(1, 1)T
ver global vibration eigenmodes are chosen, K is
the diagonal matrix of the squared circular eigen- and one elastic eigenvalue/eigenvector
frequencies ω i2. Damping is then typically chosen
as applied for a global ship vibration analysis, i.e. ω 22 2k φ2 (1/ 2 )( 1, 1)T .
as a percentage of critical damping applied to the
respective mode (see e.g. (Asmussen, Menzel, and The modes
Mumm 2001)).
Introducing the abbreviations 1 ⎛1 −1⎞
Φ = (φ1, φ2 ) =
C* (Ω) = C + G(Ω), (14) 2 m ⎜⎝1 1⎟⎠
*
(Ω , Ω)
K +H Ω ( ) Z (Ω ) , (15) are ortho-normal with respect to the mass
matrix, i.e.
, Ω, U ) , U )
R* (t, Ω ∫A φ (16)
T
F (t )dA
d R(Ω, Ω φiT Mφ j = δ iijj ,
equation (13) reads where δij denotes the Kronecker delta. Substitut-
ing u = Φr and multiplying by ΦT, the equation of
Mr + C ( )r + K ( , Ω))r = R (t, Ω, Ω
* * , U ).
* (17) motion (1) becomes the modal equation of motion
Note that the projected matrices can be com-
ΦT MΦr + ΦT KΦr
Φ = ΦT F ,
puted efficiently before the time integration, i.e. no
matrix vector products with the full system matri-
ces are required during time integration. with the modal mass matrix
Solution of equation (17) requires a nonlinear
M = ΦTMΦ = diag (1,1)
solver, because of the constraints ri = 0, i = 1, …, 6
and the fact that Ü, Ω, and Ω are unknown for
and the modal stiffness matrix
the specific timestep. The solution procedure is
described in more detail in section 2.4.
K = ΦT KΦ = diag ( 1,
2
2 ).
2
2.1 Example: Two-mass oscillator Considering the inertia force due to rigid body
As a very simple example, consider two nodes, each acceleration Ü, i.e. x = u + (1,1)TU, where u denotes
of mass m, connected by a spring of stiffness k, the nodal degrees of freedom of the finite element
resulting in a two degree-of-freedom (each node’s model and U denotes the rigid body translation,
translation in the direction of the spring) system the nodal force vector becomes
represented by the mass matrix
F F − M( , )T U.
⎛m 0⎞
=⎜ With r = (r1, r2)T etc., the final modal equations
⎝ 0 m⎟⎠ of motion in the body fixed reference frame are:
and the stiffness matrix 2m r1 1 2 2 mU,
− 2mU (18)
⎛ k −k⎞ 2k 1 (19)
K=⎜ .
r2 + r2 = ( F2 F1 ).
⎝ −k k ⎟⎠ m 2m
11
rt Δt = a1rt Δt + a4 rt Δt + a5 rt Δt , (21) In order to solve these equations the kinetic vari-
ables Ω, Ω and U have to be known. Furthermore
= a0rt + a2 rt the time-variant rotation matrix S transforming
rt Δt Δt Δt + a3 rt Δt (22)
local to global coordinates is required due to the
fact that the hydrodynamic forces are determined
leads to the temporal discretization of the equation in the global reference frame; hence there is the
of motion need to transform into the body fixed coordinate
system. Determining the kinetic variables is a non-
A( )r = R∗ ( U ) linear problem which can be solved by exploit-
t t t t t t t
∗
+ M
rt
t + C ( t )r t Δt , (23) ing equation (26). Here we take into account that
the rigid body motion is affected by the elastic
deformation represented by the coupling terms
where a0, …, a5 are the Newmark parameters that denoted by subscript “elas−rbm” in equation (26).
depend on the time step size Δt. According to this, the solution procedure will be
As a result taking the rigid body modes nested.
into account equation (23) can be subdivided
according to rigid body motion and elastic
deformations 2.3 Determining the rotation matrix
A popular method to determine the rotation matrix
⎛ A Aelas − rbm ⎞ in body dynamics is the successive execution of
⎜ rbm ⎟r
⎜ AT Aelas ⎟⎠ t three elementary rotations where the rotation
⎝ elas − rbm
⎛R ⎞ ⎛ M matrix is described by three parameters. The most
Meelas − rbm ⎞
= ⎜⎜
rb
bm ⎟
+ ⎜⎜ T
rrbm ⎟ popular parametrizations are the Euler-Angles
R
⎝ elas ⎠
⎟
⎝ M elas − rbm Meelas ⎟⎠ r t Δt and the Cardan-Angles. Here, a three parameter
⎛ C Ceelas − rbm ⎞ approach is applied with elementary rotations
+ ⎜⎜ T
rbm ⎟
. (24)
⎝ Celas − rbm Celas ⎟⎠ r t Δt about z, y' and x" axes corresponding to yaw, pitch
and roll motion. The primes denote the axes of the
one time rotated system and the two time rotated
To enforce suppression of rigid body motion system, respectively. Summarizing the three rota-
in the local reference frame the modal coefficients tion angles in the vector of generalized coordinates
r1..6 corresponding to the rigid body modes have to α [ψ , θ , φ ]T the rotation matrix S reads
be 0. Dropping the first six columns of the matrices
reduces equation (24) to S(α) = Sz(α1)Sy'(α2)Sx"(α3). (28)
12
13
The coupling of fluid and structure solver imposes Here, αi is the normal vector of the cell face
information exchange between the FVM and scaled by its area and pi is the corresponding
FEM at the interface. To maintain kinematic and pressure. pi as well as αi are located at the face
dynamic consistency at the interface the Dirichlet center xi. The task is to determine FE nodes and
distribute the pressure forces onto those nodes
υ( ξ (t, ξ ))) u( ξ ) ξ Γ 0 ,t ∈[ 0 ] (34) such that the sum of output forces equals the FE
input forces while conserving the pressure induced
and Neumann moment acting on the FE model.
On the other hand fluid cells at the interface have
σ f (t )n σ S (t y )n y Γ t ∈[ 0 T ] (35) to be adapted according to the structure deforma-
tion. Due to the fact that the hydrodynamic forces
boundary conditions have to be fulfilled. Gener- are mapped according to the face center, a direct
ally the Dirichlet boundary condition (34) will be inversion of this mapping isn’t applicable to deter-
gained by determining the structural displacements mine the face defining node displacements at the
which will be derived and adopted in the CFD sim- interface. Furthermore a fluid mesh smoothing
ulation. Whereas integrating the left hand side of strategy is needed to maintain grid quality and
the Neumann boundary conditions (35) over the prevent cell deterioration.
fluid interface results in the hydrodynamic forces
acting on the structure. The organization of the 4.1.1 Mapping of hydrodynamic forces
information exchange at the boundary influences Since typically the fluid mesh is finer than the FE
the accuracy of the results and affects the stabil- mesh, only single finite elements will receive forces
ity of the coupled simulation as well. Keeping in related to a particular force result point. The con-
mind the non-matching grids at the interface and dition to conserve hydrodynamic force fi at location
the need to interpolate variables from one domain xi when mapping on nodes xi,n of an FE element
into the other these conditions can’t be fulfilled reads as follows
exactly.
In the partioned fluid structure interaction
approach two major coupling strategies are
fi ∑ fi ,n (37)
n
applied namely the weak and strong coupling.
They differ in the information exchange frequency and for the moment
per time step which influences the fulfillment of
kinematic and dynamic consistency at the inter-
face (34) and (35). Weak (explict) coupling has
∑ fi ,n xi ,n − xi ) = 0. (38)
n
only one boundary condition exchange per time
step whereas strong (implicit) coupling has several The task is to find appropriate weights wi,n that
exchanges a time step depending on the number of
iteration steps per time step performed for solving fi,n = wi,n fi (39)
14
15
6 CONCLUSIONS
Figure 4. Comparison of rigid body motion in head A method for the solution of the fluid structure
waves with and without consideration of elastic defor- interaction problem in ship design has been pre-
mation. Rigid body motion is hardly affected by elastic
sented. It is characterized by the approximation of
deformation.
the structural deformation by a number of modes.
This serves three purposes,
that vertical bending is the only mode that is sig-
nificantly excited. This is the expected behavior for • the significant reduction of the number of
head waves. Also, this modal coefficient closely fol- degrees of freedom of the distretized structural
lows the pitch angle as the driving mechanism for equation system,
vertical bending. On the other hand, for oblique • the suppression of high frequency contents of
waves, other modes become significant. In Fig. 6, the structural deformation which might lead to
the modal coefficients for the elastic modes are overly small limits on the time step size, and
16
17
A. Ferrari
D’Appolonia S.p.A., Genova, Italy
ABSTRACT: The paper presents the design of two prototypes of a hull monitoring system which
employs temperature compensated laser based optical sensors. The prototypes have been developed for
a double side bulk carrier and an ice class tug boat. A set of finite element calculations has been carried
out for defining the optimum sensor’s installation areas in each ship. Preliminary tests on the algorithms
developed for the software onboard and a first analysis of a set of data measured on the tug boat are
reported.
19
20
0
and known issues, a preliminary list of points of
interest for measuring strains was defined, which
–2 Target included sensors on deck and bottom for bending
moments, sensors on the side for shear stresses and
–4
Steel weight sensors at the hot spots reported in the previous
–6
Local loads
Adjustment loads
section. From this preliminary list, a subset of loca-
VBM predicted tions has been selected, with the aim of optimizing
–8
VBM calculated
the number of sensors considering the constraints
90 100 110 120 130 140
Longitudinal location [m]
150 160 170
related to junction box specifications. Moreover,
there was the need for developing a system actu-
Figure 1. Example of load distributions and compari- ally affordable in the market. The following layout
son of the predicted total bending moment with the result appeared to be a good balance of technical and
of FE analysis. economical requirements.
21
22
23
4
MV ∑ KV iσ i (1)
i =1
4
MH ∑ K H iσ i (2)
i =1
24
1.0E+06
ing all support and stiffening members of the deck.
0.0E+00
–1.0E+06
Again, the aim of the analysis was to identify areas
–2.0E+06
with increased load levels during winch operation
–3.0E+06 for sensor placement.
–4.0E+06 Figure 15 presents the bow section area of the
–5.0E+06
FL-A FL-A FL-B FL-C FL-D BL-A BL-A BL-B BL-C BL-D
vessel modeled for the ice-loading study. All hull
Crest Trough Crest
Load Case
Trough
bottom and side plating including their respective
transverse and longitudinal reinforcing members
Figure 13. Comparison of vertical bending moments. between frames 46 and 59 have been included in
the FE model, and the platform deck (at 3200 mm
from baseline) and bulkheads at frames 50 and 55
2.0E+06
FEM have been fully modeled as well (Figure 16). The
Estimated
1.5E+06
structures at the fore and aft ends (Fr. 46 & 59)
1.0E+06 and the main deck itself have not been modeled
5.0E+05 directly, but they were taken into account by using
HBM [kN*m]
25
26
27
28
29
5 CONCLUSIONS
Figure 31. Tug bow FE model, typical test panel.
The paper describes a hull stress monitoring sys-
tem which employs temperature compensated
neighboring panels is about 2.5, and two of sec- laser based optical sensors, developed within the
tion 3.4 analysis load cases were used to evaluate MOSES project. The design of two prototypes are
the results. described, one installed on a double side bulk car-
The numerical strains calculated from the FE rier and the other on an ice class tug boat. The eval-
model at the “middle” and “quarter” positions of uation of hull girder bending moments along the
the panel shown in Figure 26 were used as input ship length and local stresses in selected hot spots
in the corresponding mathematical expressions of has been selected as main features for the system
section 3.4, thus resulting in the analytical calcula- on the bulk carrier. The measurement of stresses in
tion of the maximum panel von Mises stress. This the bow area due to ice pressure loads and stresses
value was then compared to the maximum von on the forward winch deck support structure have
Mises stress calculated by the FE analysis and the been considered instead more important for the
difference was found to be between 2.5 and 6%. tug boat.
This difference is very small and fully acceptable The optimum sensor’s installation areas have
for the purposes of the present study and may be been selected by means of FE analyses, taking into
justified because of the curvature of the actual account also constraints related to junction box
bow panels and the hydrostatic sea pressure on the specifications and looking for a good balance of
hull sides, which were both not considered in the technical and economical requirements.
flat rectangular panel analysis. The paper describes also the algorithms for
the software provided onboard. Particularly, on
the bulk carrier the software correlates measured
4 PRELIMINARY RESULTS strains with hull girder bending moments, while
on the tug it estimates maximum stress levels on
The system installed in the tug boat is currently selected areas of the hull structure when the tug
under test. At present, the ship operators have operates as an icebreaker.
provided for evaluation purposes only one set of Preliminary tests on the software, performed
runs recorded during a week in the beginning of employing FE analyses, show that the two proto-
February 2010, when the tug operated in 20 cm types should be capable of providing results with
thick floating ice. accuracy fully acceptable for the purposes of each
Sample plots of strain data for some 24-hour system.
periods verified the presence of increased strain At the present time, no measured data are avail-
levels for those time frames the Tug was actually able for the bulk carrier, while initial readings from
operational. Figure 32 shows a 3-hour strain plot the tug boat operating in the gulf of Tallinn have
of two strain sensors (P1P is on a port side plate been obtained and currently investigated. The ini-
and P1S is on the symmetric panel on starboard tial readings indicate that the structure of the tug
side, see Figure 22) where the Tug was operational does not suffer high stresses under normal ice
between 09:35 am and 11:00 am. The scale between breaking operating conditions. In any case, further
two horizontal grid lines is 10 με, which illustrates analyses for both the vessels are expected in the
that the structure is not heavily stressed and the future on a wider set of data.
30
31
L. Kaydıhan
Delta Marine, Istanbul, Turkey
ABSTRACT: This paper presents a hydroelastic investigation into the dynamic response characteristics
of a group of bulkers with different load carrying capacities, e.g., two handysize vessels with carry-
ing capacities of 20 000 and 32 000 dwt, respectively; one handymax vessel with a carrying capacity of
53 000 dwt; one panamax vessel with a carrying capacity of 76 000 dwt; two capesize type vessels with
carrying capacities of 140 000 and 180 000 dwt, respectively. For all the bulk carriers adopted in this study,
detailed three dimensional finite element structural models are prepared, separately, by using commercial
finite element software. The calculations are carried out for two different loading conditions, namely, fully
loaded and ballast conditions. The dry and wet frequencies are computed by using the finite element, and
they are compared with those calculated by using a higher-order 3-D hydroelasticity theory.
33
34
c( ξ ) ( ξ ) ∫∫ {G( , ξ)q
ξ ) q( )
ξ) G n ( , ξ ) φ ( )} dS
d . For time-harmonic free-surface flows without
SW (13) forward speed, the radiation potential is complex
in general, and Zkr may be expressed in terms of
Here, ξ = (ξ, η, ζ) and x respectively represent the generalized added mass coefficient, Akr, and
the source and field points on the mean wetted hydrodynamic damping coefficient, Bkr, that are
surface of the structure, Sw, q = ∂φ/∂n refers to the in phase with the acceleration and velocity, respec-
flux, the free term c(ξ) identifies the fraction of tively. Namely,
35
36
37
38
39
40
ABSTRACT: An explicit finite element code is applied to study the impact loads on one two-dimensional
rigid wedge subjected to water entry. This wedge with deadrise angle 30 degree impacting the calm water
is modeled. The numerical results are compared and validated against published experimental slamming
force, pressure distributions at different time instances, and pressure histories at different points on the
wetted surface, obtaining very good comparisons. A convergence study for parameters, such as mesh
density and penalty factor, is carried out. The computational efficiency and accuracy of the results is
discussed.
41
42
43
44
45
(b) t2 = 0.0158s
46
47
48
ABSTRACT: In this paper a simple method is proposed to estimate extreme ship response, defined as
the return values of the responses. Real ship responses are often non-Gaussian, hence a transformation,
defined by the cubic Hermite polynomials, of a Gaussian process is employed to model the responses.
The transformation is a function of the standard deviation, skewness, kurtosis and zero up-crossing fre-
quency of a response. The parameters vary with changing sea conditions and operation conditions and
are easily available from measured responses. In the case when measurements are not available the param-
eters are estimated by an empirical model from the significant wave height and operation conditions.
The model is derived from the measurements, but can be also estimated by means of a theoretical analysis.
The proposed method is compared with the typical engineering approach to estimate the return values of
a response. The full-scale measurements of a 2800TEU container ship during the first six months of 2008
are used in the comparisons.
49
50
μ + (x |W )
1 − FX m ( x | W ) ≤ , (6) where f(W) is the pdf of the vector of parameters
μ + (0 | W ) defining the sea states encountered in the period t.
Further the fraction n(t)/t in Eq. (5) is approxi-
see Rychlik and Leadbetter [12] for a proof. The
mated by
upcrossing frequency μ+(x|W) need to be estimated
or computed from the model.
n(t)/t ≈ Ûμ+(0|W)f(W)dW . (8)
In Figure 1 (a), empirical cdf of Xm is compared
with the approximated bound when μ+(x|W) is Combining Eqs. (6–8) gives the narrow-band
estimation of the return value.
2.3 Examples
In the following, ship response from half year’s
full-scale measurements of a 2800TEU container
vessel is taken to compare different estimates of
the long term cdf of Xm. For the region there are
well established models (Weibull) long term pdf of
encountered significant wave height f(W). How-
ever the measuring period is rather short and we
do not expect that the variability of the encoun-
tered significant wave heights is well described by
the model. Hence the empirical cdf of Xm could dif-
fer from that estimated by Eq. (7). The alternative
could be to use the encountered seas to estimate
the long term pdf of W. This would still require
numerical computations of the integral in Eq. (7).
Since in this section we only wish to compare dif-
ferent means to estimate the long term cdf of Xm,
a related method, not requiring the integration,
will be used instead in the following.
As before, let Nt+ ( x ) denote the number of up-
crossings of level x by the response in a period of
length t, then the empirical distribution of encoun-
tered response maximums Xm satisfies
Nt+ ( x )
FX m ( x ) ≥ 1 − . (9)
Figure 1. (a): short term distribution FXm(x|W), Nt+ ( )
bounded by Eq. (6) (solid line) and empirical distribution
(dotted line) from full-scale measurements at a stationary Again the bound is close to 1–FXm(x) for high
sea state; (b): long term (t = 1 month) empirical cdf FXm(x) values of x as can be seen in Figure 1(b) where one
(dotted line) and the bound given in (9) (solid line). presents the empirical cdf of encountered local
51
1
E ( N1+ ( xT )) = , (10)
T
+
where E ( N1 ( x )) is the expected number of
upcrossings of the level x by the responses in
one year. Obviously the expected value has to be
estimated.
Let assume that one can choose a minimal
length of the stationary periods (sea states) Δt,
e.g. 20 or 30 minutes. Furthermore, one estimate
that in average the ship will encounter K sea states
during a year, i.e. the sailing time is KΔt. Then
the expected number of upcrossings at level x,
E ( N1+ ( x )), is given by
52
1 λ2 ⎛ x2 ⎞ (13)
μ + ( |W ) = exp ⎜ − ⎟, Note that fz(W) is defined for the Gaussian
2π λ0 ⎝ λ0 ⎠ process u(t).
53
54
⎛ ⎛ x⎞k⎞
( N1+ ( x )) ( N1+ (0 ))exp ⎜
E(N
E(
E (N ⎟, (17)
⎝ ⎝ λ⎠ ⎠
Figure 6. Upcrossing spectrum in time t = 0.5 year
where E(N1+(0)) is estimated by means of Eq. (11) estimated by Eqs. (15–16) (referred to Method 2), by
fitted Weibull distribution as Eq. (17) (referred to as
or from the actual measured responses. Method 1), and computed by Eq. (9) assuming Gaussian
The methods will be compared using the time response, together with the observed upcrossing for both
history from the full-scale measurements at two after section (a) and mid section (b).
places of the 2800TEU container ship. The meas-
ured places are located at the 1/4 ship length for-
ward of after perpendicular (denoted as after upcrossings. It can be seen that the Gaussian
section), and amidships (denoted by mid section), assumption of ship response largely underesti-
respectively. The detailed information can be mates the upcrossing spectrum at high levels x. For
referred to Storhaug et al. [19]. the real non-Gaussian ship response, fitted Weibull
Firstly, assuming the ship response under each approach (Method 1) and transformed Gaussian
sea state to be Gaussian processes, the correspond- approach (Method 2) give almost identical results,
ing upcrossing spectrums are computed by Rice’s and both are close to the observed upcrossings.
formula as Eq. (13). The upcrossing spectrum at all Further, the upcrossing spectrums computed by
sea states are then integrated as Eq. (11) to get the Method 1 and 2 converge when extrapolating to
long term upcrossing spectrum in time t. Secondly, even higher levels. Therefore, both methods are
instead of Gaussian assumption, the upcrossing able to estimate the extreme response for the given
spectrum at each sea states is estimated by the trans- set of data.
formed Gaussian approach as Eqs. (15–16) (also
refer to the Method 2 in this paper). In addition,
when the time history of ship responses is avail- 4 EXAMPLE OF EXTREME PREDICTION
able, the fitted Weibull distribution is then used to
approximate the long term upcrossings as Eq. (17). In this section, the Method 2 will be applied for
Figure 6 presents the upcrossings computed by the the extreme response prediction using the full-
above 3 approaches, together with the observed scale measurements through an practical example.
55
56
5 CONCLUSIONS
57
58
ABSTRACT: The authors developed a whole ship finite element analysis system from wave loads to
structural strength at real sea state. In the present study, methodology for the rational analysis of struc-
tural strength by means of such a whole ship analysis particularly from the viewpoint of loads is dis-
cussed. Consequently, for the rational evaluation of strength in waves, the effect of operation particularly
on wave loads should be considered. It is verified that the evaluation without the effect of operation may
overestimate the stress induced by waves.
59
60
0.001
Exp. (Fn = 0.11)
200000
Exp. (Fn = 0.164)
0.0008
100000
(ton-m)
Tx/ρgζL B
0.0006
2
0
300 320 340 360 380 400 0.0004
-100000
0.0002
-200000
0
-300000 0 0.5 1 1.5
Time (sec.)
λ/L
Figure 2. Time history of the wave vertical bend- Figure 4. The wave torsional moment response ampli-
ing moment of the post-panamax container ship at tude operator of the mega container ship at S.S.5.5
S.S.7.5 and S.S.5 (χ = 180 deg., Fn = 0.219, Hw = 9 m, (χ = 150 deg., wave height: Hw = 6 m).
λ/L = 1.0).
0.02
MV/ρgζL2B
0.015
0.01
0.005
0
0 0.5 1 1.5
λ/L
61
Figure 6 shows an example of the wave pressure Figure 7. The whole ship finite element analysis in the
distribution on finite element model. It is found regular head seas (Stress distribution under the hogging
condition) (N = 12, Head seas, wave height:12 m, wave
that wave pressure is loaded only on the outer length ratio to ship length:0.8.).
shell.
The NASTRAN software is used for the present
finite element analysis. Prior to the computa-
tion, density of container was tuned to adjust the
weight, center of gravity and radius of inertia of a
finite element model. Dynamic forces and inertia
of a finite element model is adjusted by means of
the inertia relief function of the NASTRAN.
3 EXAMINATION OF RESULTS
OF FNITE ELEMENT ANALYSIS
62
Mv/ρgζBL2
0
0.0008 600 1100 1600 2100 2600
-0.02
0.0004
-0.04
0
-8 -7 -6 -5 -4 -3 -2 -1 0 -0.06
log10Q t(sec.)
Mv/ρgζBL2
the voyage records. However, it is considered that it 0
1400 1410 1420 1430 1440 1450 1460 1470 1480
is difficult for new type of ship to compensate such -0.02
a discrepancy based on a voyage re-cord. There-
-0.04
fore, it is important to evaluate wave loads ration-
ally and to assess the effect of operation on it by -0.06
t(sec.)
means of a direct computation.
Figure 10. The example of time history of wave ver-
tical bending moment (χ = 180 deg., significant wave
4 EVALUATION OF THE EFFECT height:10 m, mean wave period:15.0 sec.) (Below: magni-
OF OPERATION ON WAVE LOADS fication of above time history of 2000 second within one
hour duration).
4.1 The effect of the change of ship speed
on the short term probability of wave loads
Sagging (T02 = 15sec., H1/3 = 10m)
Figure 10 shows the example of the computed time 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
1 Without speed reduction (Fn = 0.182)
history of vertical bending moment. It is found With speed reduction
that the hull girder vibration owing to the whip- Wave component only (without hull girder vibration)
ping occurs frequently in such a severe sea state. 0.1
It is confirmed that two-node vibration occurs with
natural frequency of the present container ship.
P
63
0.01 0.01
0.001 0.001
Mv/ρgζBL^2 Mv/ρgζBL2
Figure 12. The effect of ship speed on the probability Figure 13. The relation of wave encounter angle with
of occurrence of wave vertical bending moment at mid- the probability of occurrence of wave vertical bend-
ship (χ = 180 deg., significant wave height:5 m, mean ing moment at midship (χ = 180 deg., significant wave
wave period:15.0 sec.). height:10 m, mean wave period:15.0 sec.).
64
REFERENCES
65
ABSTRACT: The purpose of the paper is to investigate long-term distribution of slamming loads of
containerships accounting for different types of environmental and operational uncertainties. Following
uncertainties are studied: the choice of the wave scatter diagram, the effect of the avoidance of heavy
weather, the effect of the maneuvering in heavy weather and the uncertainty of the method for predic-
tion of the long-term extreme slamming pressure. Long term distributions of bottom slamming pres-
sures are computed using different combinations of the aforementioned uncertainties. The purpose of
the study is the uncertainty assessment for the application in the reliability based design of ultra-large
containerships.
67
1 σV
υ sl = ⋅ ⋅ Psslam (9)
2 CALCULATION METHODS OF LONG- 2π σ M
TERM EXTREME BOTTOM SLAMMING
PRESSURE while the mean slamming frequency in all sea states
is given as:
2.1 The lifetime weighted sea method for long- nβ
term distribution of slamming pressure Δβ
υ sl = ∑
*
f ( psl ) = λ ⋅ e ⎣
(
⎡ − λ ⋅ p0 psp ⎤
⎦) (4) Model tests and full-scale trials show that nec-
essary and sufficient conditions for occurrence of
psp slam impact are the bottom emergence, i.e. rela-
(
F psl HS TZ , β = ) ∫ f ( p) dp = 1 − e ⎡⎣−λ ⋅ ( p
sl p )⎤⎦ tive bow motion being larger than local draught d,
and impact velocity being higher than the critical
p0
(5) velocity v0.
68
69
In order to assess the influence of the shipping Table 2. The most probable bottom slamming pressures
route on the extreme bottom slamming pressures, in 20 years according to the lifetime weighted sea
long-term calculations are performed for following method (k = 1).
scatter diagrams:
The most probable extreme slamming pressure
• Standard IACS North Atlantic scatter diagram
Floxtra S175
(IACS)
• North Pacific scatter diagram (N-P) Wave scatter
• Scatter diagram for shipping route North diagram Ship speed FL FL BL
Europe–Far East through Suez Chanel (Suez).
IACS vn 237 242 278
More details on the wave zones considered and 0.8 vn 201 207 236
the resulting extreme slamming pressures are pre- zero 60 40 61
sented in Section 4. N-P vn 179 223 275
The next issue covered in the paper is the effect 0.8 vn 152 191 236
of the heavy weather avoidance. The problem is zero 49 37 64
approached by the truncating probability density Suez vn 170 203 257
function of a significant wave height, as explained 0.8 vn 142 172 221
in Section 5. The truncation of the scatter diagrams zero 48 39 69
results in two additional scatter diagrams for which
long-term calculations are performed: N-P modified vn 170 214 267
0.8 vn 145 185 230
• Modified N-P scatter diagram, zero 56 47 79
• Modified Suez scatter diagram. Suez modified vn 147 199 254
0.8 vn 128 175 223
Two main maneuver that master can undertake zero 59 53 89
when the ship is in heavy seas are course changing
from beam seas to head or following seas and vol- * FL – full load condition
untary speed reduction (Guedes Soares 1990). BL – ballast condition
70
0.80
IACS vn 277 327 423
0.8 vn 227 265 337 0.70
zero 58 38 60
Figure 2. Influence of the avoidance of the heavy weather
N-P vn 181 248 333 on the slamming pressure for full load condition.
0.8 vn 167 208 273
zero 48 36 64
Suez vn 169 217 293 10.0
zero 47 38 69 8.0
2.0
1.60
1.10
1.40 1.05
1.00
1.20
0.95
0.90
1.00
IACS/N-P (full load) 0.85
Results are elaborated in Section 6 of the paper. Figure 4. Influence of the calculation method on the
Furthermore, for each scatter diagram, calcula- slamming pressure (3 hours assumed duration of short-
tions are performed for two long-term calculation term seas states in Poisson outcrossing methods).
methods:
• The lifetime weighted sea method and all load conditions respectively. In all cases,
• Poisson outcrossing method. unitary pressures (k = 1) are calculated at section at
the distance of 10% from the fore perpendicular.
Detailed results of that analysis are presented in
Section 7.
Summary results of the long-term slamming 4 CHOICE OF THE SHIPPING ROUTE
pressure analysis for two ships are presented in
Tables 2 and 3 for the lifetime weighted sea method One of the main issues analyzed within the present
and Poisson outcrossing method respectively. study is the influence of the shipping route on the
Also, these results are graphically presented in extreme slamming pressure. The North Atlantic
Figures 1–3 and Figure 4 for full load condition environment, proposed by IACS for calculation of
71
72
73
74
M. Wilken
Germanischer Lloyd SE, Hamburg, Germany
A. Menk
Robert Bosch GmbH, CR/APJ3, Stuttgart-Schwieberdingen, Germany
H. Voss
Hamburg University of Technology, Hamburg, Germany
C. Cabos
Germanischer Lloyd SE, Hamburg, Germany
ABSTRACT: Simulating global ship vibration can be split into three steps: firstly, the computation
of the dry elastic vibration of the ship structure, secondly determination of the hydrodynamic pressures
caused by a given time harmonic velocity distribution on the outer shell and thirdly, the solution of the
coupled vibration problem by considering the interaction of fluid and structure. In this paper various
approaches for the solution of the third problem for large models are compared and discussed. They are
based on reduction methods for the hydrodynamic mass matrix and make use of fast solution methods
for the exterior fluid problem for given velocity distributions of the shell. A numerical example is used to
assess the accuracy and the speed of the solution procedures.
1 INTRODUCTION
75
76
77
78
79
Figure 9. Similarity of MODAL_HYM eigenvectors Figure 12. Relative error of PROJECTION eigenvalues
and reference eigenvectors according to MAC. and reference eigenvalues ordered by MAC value.
80
REFERENCES
81
82
Ultimate strength
ABSTRACT: In this paper postbuckling behaviour and estimation of global buckling ultimate strength
of the delaminated rectangular plates are presented. The influence of the position and geometry of
elliptical delamination on the changes in the buckling behaviour of ship deck plates made of composite
materials is considering. The composite plates models were analyzed using COSMOS/M and database
is prepared for different diameter ratios and position of delamination. A delamination model, describ-
ing delaminating mode, by using COSMOS/M soft package, is applied, so that the damaged part of the
structures and the undamaged part have been represented by layered shell elements. The influence of the
position and the ellipse’s diameters ratio of delaminated zone on the critical buckling force is investigated.
The applied methods have been improved in the Marstruct Project.
85
k
Q23 = (C ′ ′
11 − 2C33 − C12
′
)cs + ( 3 ′
C12 ′
− C222 ′
+ 2C33 )c s.
3
s = sin θ ; c = cos θ .
2 SHEAR BUCKLING EQUATIONS
OF ORTHOTROPIC PLATES For the orthotropic plane material
86
– linear behaviour;
Figure 1. Delamination model. – nonlinear behaviour (Tsai-Wu failure criterion).
87
88
The buckling load determination may use the stresses. The strength of the laminated composite
Tsai-Wu failure criterion in the case of the general can be based on the strength of individual plies
buckling does not occur till the first-ply failure within the laminate. In addition, the failure of plies
occurring. In this case, the buckling load is con- can be successive as the applied load increases.
sidered as the in-plane load corresponding to the There may be a first ply failure followed by other
first-ply failure occurring. ply failures until the last ply fails, denoting the ulti-
The Tsai-Wu failure criterion provides the math- mate failure of the laminate. Progressive failure
ematical relation for strength under combined description is therefore quite complex for lami-
stresses. Unlike the conventional isotropic materi- nated composite structures. A simpler approach
als where one constant will suffice for failure stress for establishing failure consists of determining the
level and location, laminated composite materials structural integrity which depends on the defini-
require more elaborate methods to establish failure tion of an allowable stress field. This stress field is
89
usually characterized by a set of allowable stresses commonly used for orthotropic materials with une-
in the material principal directions. qual tensile and compressive strengths. The failure
The failure criterion is used to calculate a fail- index according to this theory is computed using
ure index (F.I.) from the computed stresses and the following equation (Thurley & Marshall 1995)
user-supplied material strengths. A failure index
of 1 denotes the onset of failure, and a value less
than 1 denotes no failure. The failure indices are F.I. = F1 σ1 F2 ⋅ σ 2 + F1111 ⋅ σ12 (8)
computed for all layers in each element of your
+ F22 σ 22 F66 ⋅ σ 62 + 2FF1122 ⋅ σ1 ⋅ σ 2
model. During post processing, it is possible to
plot failure indices of the mesh for any layer. 1 1 1
The Tsai-Wu failure criterion (also known F1 = − ; F11 = T C ;
as the Tsai-Wu tensor polynomial theory) is R1T R1C R1 R1
90
Position Type of
of delamination degradation Dx/Dy = 0.5 Dx/Dy = 1 Dx/Dy = 2
Macro-layer 1 Tension 25 20 20
– Macro-layer 2 Compression 90 – –
Macro-layer 2 Tension 25 20 20
– Macro-layer 3 Compression 90 – –
Macro-layer 3 Tension 25 20 20
– Macro-layer 4 Compression 90 – –
Macro-layer 4 Tension 25 20 20
– Macro-layer 5 Compression 90 – –
1 1 1 4 CONCLUSIONS
F2 = T
− C ; F22 = T C ;
R2 R2 R2 R2
Delamination is a phenomenon that is of critical
1 importance to the composite industry. It involves a
F66 = 2 . (9) breakdown in the bond between the reinforcement
R12
and the matrix material of the composite. Under-
standing delamination is essential for preventing
The coefficient F12 which represents the catastrophic failures. Due to the geometry and
parameter of interaction between σ1 and σ2 is to be general load of the ship deck, buckling is one of
obtained by a mechanical biaxial test. In the equa- the most important failure criteria.
tions (9), the parameters RCi , RTi are the compres- The FEM based methodology was success-
sive strength and tensile strength in the material in fully developed for the investigation of buck-
longitudinal direction (i = 1) and trasversal direc- ling problems of composite plates having central
tion (i = 2). The parameter R12 is in-plane shear delamination.
strength in the material 1–2 plane. Two hypotheses regarding the type of material
According to the Tsai-Wu failure criterion, the modeling were used (linear and nonlinear). The
failure of a lamina occurs if FEM model is robust in that it can be used to predict
the global buckling loads of composite plates either
F.I > 1 (10) on one side or both sides. Finite-elements analysis
was carried out to assess the reliability of the meth-
In COSMOS, nonlinear material is considered odology. The two-sublaminate model developed in
as a material with nonlinear behaviour (the non- this work provides a convenient method to model
linear material curve) or case of introducing the delaminated composite panels.
material strength components for Failure criteria For the values of in-plane loads lower than
using for composites. This latest case is the case 80 MPa, the displacement values are increasing
analyzed in the paper. since the diameters ratio is decreasing. This trend is
The failure index in calculated in each ply of each due to the fact the transversal diameter is decreas-
element. In the ply where failure index is greater ing (since the delaminated area remains constant).
than 1, the first-ply failure occurs, according to the Smaller transversal diameter means increasing of
Tsai-Wu criterion. In the next steps, the tensile and shear stiffness.
compressive properties of this element are reduced In the case of the in-plane loading values
by the failure index. If the buckling did not bigger than 80 MPa, the displacement values are
appeared until the moment of the first-ply failure increasing since the diameters ratio is increasing.
occurring, the in-plane load corresponding to this This trend is due to the contact pressure between
moment is considered as the buckling load. the layers in contact in the delamination, which is
In the nonlinear calculus, for the buckling load, increasing since the loading force is increasing.
the graphical method and Tsai-Wu failure criterion For an in-plane loading of about 130–300 MPa
were used. The values obtained for buckling load in each case, a small instant jumping of transversal
were placed in the range specified in each case. displacement is observed.
In Table 3, the variations of the buckling load This means that what is recover in plate stiffness
corresponding to the fails in the tension and after the increasing of contact pression in the dela-
compressive cases, versus diameters ratios are mination area, is lost due to the lamina damage
presented. occurring.
91
92
ABSTRACT: Shear buckling and postbuckling behaviour of the square plates, made of composite
materials, with central cut-outs is treated in this paper. In the analysis, finite element method (FEM) was
applied to perform parametric studies on various plates based on the shape and position of the ellipti-
cal hole. This study addressed the effects of an elliptical/circular cutout on the buckling load of square
composite plates. The laminated composite plates were arranged as symmetric cross-ply. The cutouts
are either circular holes or elliptical holes. The shear-buckling strengths of the plates could be increased
considerably only under aspect ratios. The plate-buckling mode can be symmetrical or anti-symmetrical,
depending on the plate boundary conditions, aspect ratio, and the hole size. In this paper, the analysis
has been performed only for the plate clamped on sides. The results and illustrations provide important
information for the efficient design of ship structural panels made of composite materials, having cut-
outs. The aim of the work presented in this paper is to analyze the influence of cut-out on the changes
in the buckling behaviour of ship deck plates made of composite materials. For each diameters ratio
there are plotted variation of the transversal displacement of the point placed in the middle of the plate,
according to the pressure that has been applied. Buckling load determination for the general buckling of
the plate has been made by graphical method. The post-buckling calculus has been performed to explain
the complete behaviour of the plate.
93
Dx/Dy 0.5 1 2
pcr [MPa] 154.52 158.87 160.159
Dx
x
Dy
y
Figure 1. Shear loaded plate with cut-out. Figure 2. Results obtained in the convergence study.
94
Pcr [MPa]
158
157
156
155
154
0 0.5 1 1.5 2 2.5
Dx/Dy
Figure 3. The deformed plate after buckling, for Figure 6. Variation of buckling load (linear calculus)
Dx/Dy = 0.5. versus diameters ratio Dx/Dy.
(K + KI)i−1Δdi = ΔQ (1)
where
K is the linear stiffness matrix,
KI is an incremental stiffness matrix based upon
displacements at load step i−1,
Δdi is the increment of displacement due to the
Figure 5. The deformed plate after buckling, for i–th load increment,
Dx/Dy = 2. ΔQ is the increment of load applied.
The correct form of the incremental stiffness
matrix has been a point of some controversy. The
The results of the linear analysis are presented incremental approach is quite popular (this is the
in Figure 6, where the variation of the buckling procedure applied in this study). This is due to
load function of the ratio Dx/Dy is plotted. the ease with which the procedure may be applied
As it is seen, the buckling load is increasing since and the almost guaranteed convergence if small
the ratio Dx/Dy is increasing. enough load increments is used.
Opposite, in the case of uniaxial compressive Buckling and post-buckling analysis has been
buckling the variation is so as the buckling load performed for all types of panels.
is decreasing since the ratio Dx/Dy is increasing In certain cases, the general buckling does not
(Chirica, Beznea & Chirica 2009). occur till the occurring of the first-ply failure.
95
where
1 1 1
F1 = − ; F11 = T C ;
R1T R1C R1 R1
1 1 1 1
F2 = T − C ; F22 = T C ; F66 = 2 . (3)
R2 R2 R2 R2 R12
1
96
In the nonlinear calculus, for the buckling load, − as the ratio Dx/Dy is increasing, the value of
the graphical method and Tsai-Wu failure criterion the buckling load corresponding to an element
were used. The values obtained for buckling load failing which is damaged by compression is also
were placed in the range specified in each case in increasing.
Figure 7.
When the hole size becomes considerably large
4 CONCLUSIONS relative to the plate width, most of the compres-
sive load is carried by the narrow side strips of
In the paper, the results of the FEM based meth- material along the plate boundaries. As it is well
odology that was successfully developed for the known, a stronger plate boundary condition (e.g.,
investigation of buckling problems of composite clampedrather than simply-supported bounda-
plates with central elliptical cut-out is presented. ries) increases the buckling strength, while the
Two hypotheses regarding the type of material higher stress concentration decreases the buckling
modeling is used (linear and nonlinear). strength.
The buckling behavior of plates with cen- Thus, which effects become dominant will
tral holes as presented in figure 8 is quite pecu- determine the increase or decrease of the buckling
liar because, under certain boundary conditions strengths of the perforated plates.
(clamped edges) and cut-out aspect ratios, the For the circular-hole cases, the narrow shear
mechanical-buckling strengths of the perforated side strips are under stress concentration, which
plates, contrary to expectation, increase rather reduces the buckling strengths.
than decrease as the hole sizes grow larger. The The unusual buckling characteristics of the per-
conventional wisdom is that, as the hole sizes forated plates offer important applications in ship
increase, the plates lose more materials and structural panel design. Namely, by opening holes
become weaker. Therefore, the buckling strengths of proper sizes in ship structural panels for weight
were expected to decrease as the hole sizes saving, their buckling strengths can be boosted
increase. This was not the case. Such peculiar simultaneously.
buckling phenomenon of the perforated plates The buckling load determination is too difficult
may be explained as follows. without applying a graphical method. In certain
Certain conclusions after the shear buck- cases, the using of Tsai-Wu criterion may predict
ling analysis of the perforated plates may be so named “buckling load”, if general buckling of
performed: the plate does not occurred before first-ply failure
occurring.
− the buckling load is increasing since the ratio The first failure occurring in an element is based
Dx/Dy is decreasing; on the Tsai-Wu failure criterion, which provides
− the buckling load corresponding to an element the mathematical relation for strength under com-
failing which is damaged by tension, has the bined stresses may be used.
same value that is not depending on the ratio The lack of the criterion is referring to the antici-
Dx/Dy; pation of the real mode to occurring the cracking.
97
98
Sang-Rai Cho
University of Ulsan, Ulsan, Korea
Hyun-Soo Kim
Hyundai Heavy Industries, Ulsan, Korea
Jeong-Bon Koo
Samsung Heavy Industries, Geoje, Korea
ABSTRACT: A robust ultimate strength formulation is proposed for stiffened plates subjected to com-
bined axial compression, transverse compression, shear force and lateral pressure loadings. Before the
formulation was derived, a simplified numerical method was developed to trace the structural behavior
of stiffened plates under combined loadings. A rigorous parametric study was, then, performed using the
developed numerical method to predict the ultimate strength of various stiffened plates under various
combinations of loadings. The formulation was derived through a regression study using the parametric
study results. The accuracy and reliability of the proposed formulation were compared with those of a
commercial package, ABAQUA and DNV PULS and with the experimental results.
99
100
∫ [B ] { o } d (vol
T
2.2.2 Internal forces re v ) (6)
The nodal displacements at the (n + 1)th time step vol
can be calculated using Equation 4 together with
the nodal displacements at the (n − 1)th and nth where [B] is the strain-displacement matrix.
time step and {Rint} and {Rext} of the nth time 2. Normal internal forces
step. Accordingly, the velocity and accelerations at In order to satisfy the moment equilibrium
the (n + 1)th time step can then be obtained from conditions two adjacent element normal forces
Equations 3a and 3b. We assume in this study are required at node i and l due to the torsional
that the characteristics of the applied loading are moment at the common edge j–k (see Figure 2).
known. Therefore the problem to be solved is how These normal forces are then balanced by the
to calculate the internal force vectors, {Rint}n using normal forces at node j and k. The first step in
the nodal displacements at the nth time step. deriving the relationship for the out-of-plane
For the simplicity of equations the subscript ‘n’ internal force is to describe the folding angle in
will be omitted when deriving the equations for the terms of the nodal translational displacements.
internal forces, {Rint}n when the displacement vec- The nodal internal forces for each element
tors at the nth time step are provided. As mentioned should be calculated in the local coordinate and
101
102
be higher than those assumed in the theoretical 10% less than those of both PULS and ABAQUS.
analyses. Therefore, further study seems necessary Similar to the trend of Model-I, the strength
to investigate the effects of imperfection level on predicted by SPUSA is 10% greater than that of
the ultimate strength of stiffened plates. PULS for transverse compression, but is 10% less
for shear force.
2.3.2 Shear force and transverse compression
For the case of combined transverse compression
and shear force loadings the ultimate strengths of 3 PARAMETRIC STUDY
two models provided by DNV (Byklum 2003) were
analyzed using SPUSA and the results were com- 3.1 Stiffened plates for parametric study
pared with those of PULS and ABAQUS. Table 3 Prior to deriving the ultimate strength formula-
provides the dimensions and material properties tion proposed in this study, a rigorous parametric
of the two models. The models with thicknesses of study was performed. The stiffened plates consid-
12.0 mm and 14.5 mm are denoted as Model-I and ered in the parametric study were similar to those
Model-II, respectively. provided in ISSC 2000 committee VI.2 (Yao et al.,
The ultimate strength analysis results of the two 2000). The size of the local panel between stiffen-
models subjected to pure shear force and pure ers was taken as:
transverse compression are summarized in Table 4
together with those of ABAQUS and PULS. When a × b = 2,400 × 800; 4,000 × 800 (mm)
the pure transverse compression is applied to
Model-I, the SPUSA prediction is approximately tp = 10; 13; 15; 20; 25 (mm)
10% greater than that of PULS but is similar to
that of ABAQUS. Under shear force, the ultimate Three types of stiffeners were considered;
strength predicted by SPUSA is approximately a flat-bar, an angle-bar and a tee-bar, and three
103
Location Constraints
x=0 Tx = 0, Ry = 0
x = a/2, 3a/2 Tz = 0
x = 2a Ty = 0, Ry = 0
y = 0, 4b Tz = 0
w po = Ao si mπ x
sin π y
a πx
+ Be sin (7)
b a
wso Bo sin π x vso i πx
Co sin
, (8)
a a
104
105
due to Meq; σec = Euler column buckling the, local plate buckling stress under transverse
stress of stiffener including associate plating; compression; ρ y λc1 34 } is the knock-
σY = σYp Ap σYs As Apps , is the mean yield stress;
down factor for local plate buckling under trans-
σ et 1 I o (G
GJJ 4π 2 L2 ( ECw )), is the elastic verse compression;
tripping stress of a stiffener; σ xb = σY × M eq M p , Cbc p{ . β1 65 (b a ) } is the strengthen-
03
is the equivalent bending stress due to the ing coefficient of the boundary condition for local
end bending moment and lateral pressure; plate buckling.
M eq M e + pbl 2 16 ; J hsf tsf 3 hswtssww3 3 , is the
St. Venant torsion constant; I 0 I w + As es2 + I f , is
the moment of inertia of the stiffener; Iw = polar 4.4 Shear force alone
moment of inertia of stiffener web; If is the polar When pure shear force is applied the mean ultimate
moment of inertia of the stiffener flange; es = dis- strength approaches the yield stress. The formula-
tance between the stiffener centroid (plate excluded) tion for pure shear force derived through regres-
and its toe; Cw I f (h ( hw t f )2 , is the torsional sion analysis of the parametric study results is as
warping constant;
onst follows:
σ eoax = n π Dy ax B 2 [ Dx B 2 Dy L2 + 2 m 2 Dxxy n2
Dy m 4 L2 n 4 B 2 ], is the overall grillage buck- 2
ling stress; L,B are overall length and breadth, ⎛ τ xy ⎞
2
⎛ 3τ xy ⎞
respectively; ax is the average cross-section area per ⎜⎝ ρ τ ⎟⎠ + ⎜ σ ⎟ = 1 (12)
τ e ⎝ Y ⎠
unit width of plating and longitudinal stiffeners; Dx,
Dy are effective flexural rigidity per unit width of
where τ e = { 806 + 3.. (b′ a )2 }E (tt p b )2 is the
stiffeners with attached plating in the longitudinal
elastic shear buckling stress of local plate; and
(x) and transverse (y) directions, respectively; Dxy is
ρτ = 4.0 is the knock-down factor for plate shear
the twisting rigidity per unit width
buckling
ρc is the knock-down factor for column buckling
of stiffener; ρt is the knock-down factor for trip-
4.5 Combined loads
ping of stiffener; and ρoa is the knock-down factor
for overall buckling of stiffened plate under axial Finally, the formulation to predict the ultimate
compression. strength of stiffened plates subjected to combined
106
⎛ σ xa σ xa + σ xbs σ xa
⎜⎝ ρ σ + ρ σ +
ρoaσ eoax
c ec t et
2
τ xy ⎞
2
⎛ σy ⎞
+⎜ ⎟ + ⎟
⎝ ρ yCbcσ ey ⎠ ρτ τ e ⎟
⎠
2
⎛
(σ xa + σ xb )2 − (σ xa + σ xb )σ y + σ y2 + 3τ xy2 ⎞⎟
+⎜ =1
⎜ σY ⎟
⎝ ⎠
(13)
5 CONCLUSIONS
107
REFERENCES
108
M.D. Collette
University of Michigan, Ann Arbor, Michigan, USA
ABSTRACT: A series of rapid semi-analytical methods for predicting the collapse of aluminum
structures is presented, including methods for tensile and compressive limit states. The methods presented
have been designed to be extensible to a wide range of structural topologies, including both conventional
stiffened-panel topologies and more advanced extrusion topologies. Unlike existing steel ultimate strength
methodologies, particular attention is paid to capturing aluminum-specific response features, such as
alloy-dependent material stress-strain curve shapes and the weakening effect of fusion welds. The methods
are validated against finite element analysis and previously published experimental results.
109
110
σU _ IOOP
σ IOOP σ PERFECT (9) 2.2 Plate element verification
σU _ PERFECT
Three different sources of experimental validation
This correction includes the slight reduction in data are available for simply-supported aluminum
stiffness observed in plates with large IOOPD. plates. The total available database consisted of
111
1.5
16 un-welded 5456 plates were used in the current Experimental Failure Stess/0.2% Proof Stress
dation study, it was assumed that the 6061-T6 alloys Stowell, No OOP
Stowell, With OOP
had a knee exponent, n, of 25 while both tempers of Stowell, With OOP and Weld
1
the 5456 alloy had an exponent of 16. The experi-
mental program addressed quite slender plates, with
b/t ranging from 48 to 144. IOOPD were noted to 0.5
be as-supplied, and believed to be typical of what is
present in the shipyard, but were not measured.
The third and final test program is a series of 0
aerospace alloy plates tested by NACA (Anderson 0 0.2 0.4 0.6 0.8 1
Experimental Failure Stess/0.2% Proof Stress
1.2 1.4
112
113
114
115
116
117
ABSTRACT: Advanced design methods and procedures are getting published every now and then from
all corners of the world. But the practicing codes could not always append all the refined or essential
recommendations timely. Stiffened cylinders are one among those structural components. Most of the
offshore floating platform components are made as stiffened cylinders and an improved model in the
design process can affect the total construction cost and schedule to a great extent. Reliability based
design approach is now seems to be advantageous over the deterministic type of structural designing
process as it addresses uncertainties in the design variables and leads to consistent level of safety. The Reli-
ability based approach still needs a robust strength model to predict the capacity with respect to random
design variables. Numerical analysis methods are suitable for this purpose but the time and effort involved
are quite high and hence a robust analytical approach is preferred for reliability analysis. DNV and API
are the most widely used design codes which offer strength models for stiffened cylindrical shell under
different loading conditions. This paper establishes a strength model for ring, stringer and orthogonally
stiffened cylindrical shells which is actually a modified version of a strength model proposed earlier. The
proposed model shows better agreement with the experimental results compared to the practicing DNV
and API design codes. The model uncertainty factor and the strength model can be utilised for the reli-
ability analysis of similar structures.
119
2 BUCKLING OF STIFFENED
CYLINDERS
120
Z = XmR−S ⎧ ⎛ R ⎞
⎪0.75 − 0.142(Zl 1)
04
0 003Zl 1 − ;
⎝ 300t ⎟⎠
⎪
Xm is the model uncertainty factor associated ρn = ⎨ 1 ≤ Zl < 20
with strength R ⋅ S is the load and Z is the g(.) ⎪ R
function in the first order Second order reliabil- ⎪0.35 − 0.0003
0
t
; Zl ≥ 20
ity analysis. Z represents the safety margin in the ⎩
{
structural component.
1.3
Xm can be calculated for various theoretical B=
1 + 0.3λ n λ n ≥ 1
strength models comparing with the experimental
data. A good analytical strength model will have A quadratic interaction of λσy and σe can be
mean of Xm tends to unity and the coefficient of used to predict the inelastic collapse stress.
variation will be small. The strength models can be
σc = φσy
compared based on these values for different load-
ing cases as axial, radial and combined. where
1
φ=
4 STRENGTH MODELLING OF RING (1 + λ e4 )
STIFFENED SHELLS
σy
The design strength of the ring stiffened cylin- λe =
ders under axial, radial and combined loading are σe
computed based on the limit state approach. This
formulation focuses on the shell collapse between Now the Model uncertainty factor for the axial
ring stiffeners. The recommended formulation is load,
similar the one suggested earlier, Das et al. (2003) σ
with some modification on the knockdown factors Xm =
σc
so that the experimental results are getting closer
to the prediction. 4.2 Under hydrostatic pressure
For hydrostatic pressure, the proposed formula-
4.1 Under axial compression tion, Faulkner et al. (1983) is identical with the
The limit state approach estimate the elastic buck- approach in BS5500. The inelastic hydrostatic col-
ling strength of a ring stiffened cylinder subjected lapse pressure is estimated as,
to axial compression as,
⎧0.5 phm ; py ≥ phm
σe = BρnCσcr ⎪
phc = ⎨ ⎛ py ⎞
⎪ py ⎜⎝1 − 0.5 p ⎟⎠ ; py < phm
where ⎩ hm
Et
Classical elastic axial buckling stress,
phm = R
Timoshenko & Gere, (1961) 2
1 ⎛ πR ⎞
n2 − 1 +
L2 2⎝ L ⎠
Zl = 1 − ν2 ⎧ ⎫
Rt
⎪ 2
⎪
⎪⎪ t2 ⎡ 2 ⎛ π R ⎞ ⎤ ⎪⎪
2
σy ×⎨
1
+ ⎢n − 1 + ⎥
λn = 2 2 2
⎪ ⎡ 2 ⎛ L ⎞ 2 ⎤ 12R 1 − υ ⎢⎣ ( ) ⎝ L ⎠ ⎥ ⎬⎪
ρnCσ cr ⎦
⎪ ⎢ n ⎝ ⎠ + 1⎥ ⎪
⎪⎩ ⎢⎣ πR ⎥⎦ ⎪⎭
C is a Length dependent coefficient,
121
⎛ υ⎞
J 1− 4.4 Under combined axial compression
⎝ 2⎠ and radial pressure
γ =
2 Nt
J + trht + Most of the design codes handle the combined load-
α ing based on an interaction approach. The general
2 interaction expression is in the following form.
⎛ R⎞
J Ar ⎜
⎝ Rcr ⎟⎠ m n
⎛σ⎞ ⎛ p ⎞
⎜⎝ σ ⎟⎠ + ⎜⎝ p ⎟⎠ = 1
1.285 c rc
α=
Rt
The above expression demonstrates the limiting
cosh ( L ) cos ( L ) criteria for the structure stability under combined
N= ,
sinh ( L ) sii ( L ) loading. Hence, there exist a number of load-
ing combinations which can cause the collapse
⎡ ⎛ αL ⎞ ⎛ αL ⎞ ⎛ αL ⎞ ⎛ αL ⎞ ⎤ of the structure. Eventually, the above expression
2 ⎢sinh cos + cosh siin
⎣ ⎝ 2 ⎠ ⎝ 2 ⎠ ⎝ 2 ⎠ ⎝ 2 ⎠ ⎥⎦ provides the model uncertainty factor for com-
G= bined loads at which a structural collapse occurs.
sinh ( ) + sin
si ( )
The best results with the available data is noticed
G and N are transcendental functions of with m = 1 and n = 2.
αL, i.e., which cannot be expressed in terms of Therefore,
algebraic operations or satisfy a polynomial
equation. 2
⎛σ⎞ ⎛ p ⎞
The Model uncertainty factor for the hydrostatic Xm = ⎜ ⎟ + ⎜
pressure load, ⎝ σ c ⎠ ⎝ prc ⎟⎠
122
σ pps
vi. The weld induced residual stress is incorporated ps =
using the width of the tension block η. For σy
continuous structural fillet welds, η = 4.5. For The structural proportional limit ps is 0.75
light fillets or for significant shake down situation, for stress relieved structures and 0.5 for all other
η = 3. For stress relieved structures, η = 0. cases.
123
σc σ es ⎧ 0.85
λ re = = λr
σ es σy ⎪0.25 + 500 g g ≤ 500
Kp = ⎨
0.12
⎪0.98 + g 500 < g < 2500
⎧⎛ 1.05 0.28 ⎞ ⎩ 500
se ⎪⎜ − 2 ⎟ K r ; λ re
re ≥ 0.53
=⎨ λ λ re ⎠
s ⎪⎝ re g=
M x Mθ LtAs
, Mθ =
s
⎩1; λ re < 0.53 Is Rt
xiii. Average ultimate collapse stress
iv. Bay instability stress
⎛A set ⎞
σu σc ⎜ s (R t)
⎝A s st ⎟⎠ σθ =p Kθ L
t
The Model uncertainty factor for the axial where
load,
Xm =
σ
σu
Kθ L = { 1; M x ≥ 3.42
1 − ξε ; M x < 3.42
⎧1; x ≤ 1.26
5.2 Under radial pressure ⎪
ξ = ⎨1.58 − 0.46M x ;1.266 < x < 3.42
The proposed strength formulation for radial pres- ⎪⎩0; x ≥ 3.42
sure is according to API Bul 2U with some changes
in the effective pressure correction factor.
ε=
( − )
The steps to calculate the axial strength are as ⎛ Let ⎞
follows. ⎜⎝1 + ⎟
J ⎠
i. Local buckling pressure of un-stiffened shell
Nφ
⎧⎛ 1.27 ⎞ ⎛ t ⎞ 2 k2 =
Nθ
⎪⎜ 1.18 ⎟ E ⎜ ⎟ ; M x 1.5; H < 2.5
⎪⎝ H + 0.5 ⎠ ⎝ R ⎠
⎪ 0.92 ⎛ t ⎞ 2 N
⎛ R⎞ Nφ =
⎪ E ⎜ ⎟ ; 2.5 < H < 0.208 ⎜ ⎟ 2π R
⎪ A ⎝ R⎠ ⎝t⎠
peL =⎨ 3
⎪0.836C −11.061E ⎛ ⎞ ; 0.208 < C < 2.85
t Nθ = p(R + 0.5t)
⎪ p ⎜⎝ ⎟⎠ p
R
⎪ 3 ⎛ R ⎞
2
⎪0.275E ⎛ t ⎞ ;C > 2.85 J Ar ⎜
⎪⎩ ⎝ R⎠ p ⎝ Rcr ⎟⎠
124
2 ( − x )2 ( − )2
Cc = −1 7 STRENGTH ANALYSIS AND RESULTS
φxφθ
where The data collected are carefully arranged and tab-
ulated with all the necessary inputs for the code
σ based design. The data is then pushed through the
Rx = analytical relations of DNV, API and the Recom-
σy mended Models for stiffened cylinders. The strength
predicted by each of the models is then compared
p (R + t) with the experimental results to evaluate the model
Rθ =
tσ y uncertainty factor for each set of data. The mean
and COV of the model uncertainty factor is then
σu computed for each codes for ring stiffened and ring-
φx = stringer stiffened cases based on three loading con-
σy
ditions, axial, radial and combined. The predicted
and experimental strength are then represented in
σ θcB
φθ = a graphical form which is normalised in terms of
σy yield strength. For combined loading cases, the
model uncertainty is plotted against L/R ratio as it
Since the above expression for combined loading is not straight forward to represent the strength.
characterises the limiting structural stability crite-
ria, it represents the model uncertainty factor of 7.1 Ring stiffened cylinders
the structure for the combined loading condition.
The model uncertainty factor for combined The ring stiffened cylinders are basically checked
loading is, against the local shell buckling which is the domi-
nant failure mode in this type of structures. Other
2 2 modes of failure and its interactions also have been
⎛R ⎞ ⎛R ⎞ taken into account.
X m = ⎜ x ⎟ + Cc Rx Rθ + ⎜ θ ⎟
⎝ φx ⎠ ⎝ φθ ⎠
7.1.1 Under axial compression
Table 1 shows the statistical results of the ring stiff-
6 EXPERIMENTAL DATA ened cylinders under axial compression for a popu-
lation of 40 for DNV, API and the recommended
The experimental test results are collected from a strength model. Figure 2 to Figure 4 shows the
wide literature survey over the last century. It is comparison of predicted and experimental data
observed that majority of the experimental works for the different approaches. The strength predic-
on stiffened cylinders are being undertaken during tion of the recommended model is more accurate
1960’s to 1980’s and there is not much experimen- compared to the other two approaches in terms of
tal works available recently as the researches are its statistical measures. Figure 4 shows the spread
comfortable with the numerical results with the of the results about its mean line which is having
increased capabilities and accuracy. In this paper, a better bias to the unity with low COV and it is
only the valid experimental data collected from evident visually also.
the documentation mentioned in the references
are considered. The data collected from various Table 1. Statistical results of ring stiffened cylinder
technical documents mentioned in the reference under axial loading.
are subjected to critical examination to avoid any
unreliable data. The data collected can be classified Recommended
based on various factors like geometrical proper- DNV API model
ties, material properties, Method of production, Mean 1.28 1.15 1.05
Test conditions etc. The proposed formulation COV 17.94% 11.84% 10.01%
considered here is applicable to structures with Population 40
parameters of the following range.
125
0.60
Table 2. Statistical results of ring stiffened cylinder
0.40
under radial loading.
Mean - 1.28
COV - 17.94% Recommended
0.20 DNV API model
1.25
1.00
1.00
0.80
φ - Test
0.75
φ - Test
0.60
0.50
0.40
Mean - 1.15
COV - 11.84% 0.25
Mean - 0.98
0.20 COV - 19.43%
0.00
0.00 0.50 1.00 1.50
0.00
0.00 0.20 0.40 0.60 0.80 1.00 φ - Predicted
φ - Predicted
Figure 5. DNV prediction and test results of ring stiffened
cylinders under radial pressure.
Figure 3. API prediction and test results of ring stiff-
ened cylinders under axial compression.
1.50
1.00
1.25
0.80
1.00
φ - Test
φ - Test
0.60 0.75
0.40 0.50
Mean - 1.05
COV - 10.01%
0.20 0.25
Mean - 1.35
COV - 19.09%
0.00 0.00
0.00 0.20 0.40 0.60 0.80 1.00 0.00 0.50 1.00 1.50
φ - Predicted φ - Predicted
Figure 4. Prediction of recommended model and test Figure 6. API prediction and test results of ring stiffened
results of ring stiffened cylinders under axial compression. cylinders under radial pressure.
126
0.50
2.50
0.25
Mean - 1.01
COV - 17.51%
2.00
0.00
0.00 0.50 1.00 1.50
φ - Predicted 1.50
Xm
Figure 7. Prediction of recommended model and test
results of ring stiffened cylinders under radial pressure. 1.00
0.50
comparison of predicted and experimental data
Mean - 1.45
for the different approaches. The average and COV - 20.79%
spread of the population is much better for the 0.00
recommended model. The recommended model 0.00 1.00 2.00 3.00 4.00 5.00
shows better central tendency compared to the L/R
other two approaches.
Figure 8. DNV prediction and test results of ring
7.1.3 Under combined axial compression stiffened cylinders under combined loading.
and radial pressure
Table 3 shows the statistical results of the ring
stiffened cylinders under combined axial com- 1.80
7.2.1 Under axial compression Figure 13 shows the comparison of predicted and
Table 4 shows the statistical results of the ring and experimental data for the different approaches.
stringer stiffened cylinders under Axial compres- The recommended model predicts the strength
sion for a population of 32 for DNV, API and almost similar to that of the API model and which
the recommended strength model. Figure 11 to is better when compared to DNV model.
127
1.60
1.20
1.40
1.00
1.20
1.00 0.80
φ - Test
Xm
0.80
0.60
0.60
0.40 0.40
Recommended 1.20
DNV API model
1.00
Mean 1.00 1.06 1.00
COV 23.18% 14.92% 14.99%
0.80
φ - Test
Population 32
0.60
1.40
0.40
1.20
0.20 Mean - 1.00
COV - 14.99%
1.00
0.00
0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40
0.80
φ - Test
φ - Predicted
0.60
Figure 13. Prediction of recommended model and test
0.40
results of ring and stringer stiffened cylinders under axial
compression.
0.20 Mean - 1.00
COV - 23.18%
Table 5. statistical results of ring and stringer
0.00 stiffened cylinder under radial pressure.
0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40
φ - Predicted Recommended
DNV API model
Figure 11. DNV prediction and test results of ring and
stringer stiffened cylinders under axial compression. Mean 1.33 1.12 1.06
COV 47.38% 21.54% 18.38%
Population 9
7.2.2 Under radial pressure
Table 5 shows the statistical results of the ring
and stringer stiffened cylinders under Radial
pressure for a population of 9 for DNV, API and experimental data for the different approaches. The
the recommended strength model. Figure 14 to recommended model has low bias and low COV
Figure 16 shows the comparison of predicted and compared to the other two models.
128
Mean - 1.33
Table 6. Statistical results of ring and stringer
0.25 COV - 47.38%
stiffened cylinder under combined loading.
Recommended
0.00 DNV API model
0.00 0.25 0.50 0.75 1.00
φ - Predicted Mean 1.84 1.33 1.26
COV 43.82% 22.19% 20.12%
Figure 14. DNV prediction and test results of ring and Population 25
stringer stiffened cylinders under radial pressure.
4.00
Xm
2.00
0.50
1.50
Mean - 1.12
0.25 COV - 21.54%
1.00
0.50
0.00
0.00
0.00 0.25 0.50 0.75 1.00
0.00 0.20 0.40 0.60 0.80 1.00 1.20
φ - Predicted
L/R
1.80
1.00
1.60
1.40
0.75
1.20
φ - Test
1.00
Xm
0.50
0.80
Mean - 1.06
COV - 18.38% 0.60 Mean - 1.33
0.25
COV - 22.19%
0.40
0.20
0.00
0.00 0.25 0.50 0.75 1.00 0.00
φ - Predicted 0.00 0.20 0.40 0.60 0.80 1.00 1.20
L/R
Figure 16. Prediction of recommended model and test
results of ring and stringer stiffened cylinders under Figure 18. API prediction and test results of ring and
radial pressure. stringer stiffened cylinders under combined loading.
129
130
131
ABSTRACT: The application of composite materials in many structures poses to engineers the problem
to create reliable and relatively simple methods, able to estimate the strength of multilayer composite struc-
tures. Multilayer composites, like other laminated materials, suffer from layer separation, i.e., delamina-
tions, which may affect the stiffness and stability of structural components. Especially deep delaminations
in the mid surface of laminates are expected to reduce the effective flexural stiffness and lead to collapse,
often due to buckling behaviour. This paper deals with the numerical modelling of the buckling strength
of composite laminates containing delaminations. Namely, non-linear buckling and post-buckling analy-
ses are carried out to predict the critical buckling load of elementary composite laminates affected by
rectangular delaminations of different sizes and locations, which are modelled by finite elements using
different techniques. Results obtained with different finite element models are compared and discussed.
133
B
2 DIFFERENT MODELLING
APPROACHES
134
135
4 CONSIDERATIONS FOR 3D SOLID Buckling Mode Map, panel aspect ratio A/B=1.36, UD, (flat)
1
MODELS FE − Global
0.9
Non dimensional delamination size, b/B
FE − Local
FE − Global/Local I
A panel with a delamination subjected to uni-axial 0.8
FE − Global/Local II
compression can buckle in local or global modes as 0.7 Global limit
Sub−mode limit
described in the introduction. A typical local buck- 0.6
ling mode is shown in Fig. 6 for solid models. 0.5
However, results from the 3D solid models also 0.4
show that combinations of the global and local
0.3
modes can appear as so-called combined modes
0.2
or sub-modes. The sub-modes are found to appear
for special combination of delamination size and 0.1
136
137
138
139
Y. Liu
Dalhousie University, Nova Scotia, Canada
ABSTRACT: Numerical simulation is used to study the influence of welding-induced residual stress
in welded, tee-stiffened plates focusing on the effect of shakedown. Residual stresses are simulated using
3D thermo-elasto-plastic finite element analysis. The influence of strain hardening and number of load
cycles on residual stress shakedown is then investigated. Load versus end-shortening curves are used to
characterize the strength and behaviour of stiffened plates under axial compression both before and after
shakedown. Results show that the reduction in residual stress due to shakedown occurs entirely during the
first load cycle provided that the magnitude of that load is not subsequently exceeded. Both the tensile
and compressive welding residual stresses are reduced by as much as 40% when the applied load causes an
average stress equal to 50% of the yield stress. This level of shakedown increased the ultimate strength of
tee-stiffened plates by as much as 6%.
141
142
143
144
on the effects of shakedown. It is also noted that Longitudinal residual stress (MPa)
due to the low magnitude of compressive residual
stress and the amplitude of the applied load, no Figure 8. Longitudinal residual stress in web of
plastic straining occurred during the compressive model T2.
portion of the load cycle. Shakedown in this case
occurred only during the tensile part of the load. 0
The residual stress distribution in the plate of -60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60
model T2 before shakedown is shown in Figure 7.
Longitudinal residual stress (MPa)
T1 1.00 0.20 0.79 0.17 0.57 0.13 Figure 9. Longitudinal residual stress in flange of
T2 1.01 0.14 0.79 0.12 0.57 0.09 model T2.
T3 1.02 0.10 0.79 0.08 0.57 0.06
T4 1.02 0.08 0.79 0.07 0.57 0.05
T5 1.01 0.13 0.79 0.11 0.57 0.08 welding, the width of the equivalent tensile stress
T6 1.02 0.14 0.79 0.11 0.56 0.08 block decreased so that η = 2.17 and η = 1.57 for
T7 1.01 0.13 0.79 0.11 0.56 0.09 shakedown stresses of 0.25 σy and 0.5 σy, respec-
tively. This is consistent with Faulkner’s (1975)
* Normalized with respect to yield stress of 360 MPa statement that the range of η decreases from 4.5–6
145
2.5
Table 4. Normalized ultimate strengths.
2
0.3
1
No shakedown (1)
0.9 (3)
(2)
Normalized axial stress, σ/σy
Vertical distortion of plate at
0.7
146
200
A comparison of ultimate strengths after shake-
down at a stress of 0.5 σy against values calculated 100
without considering residual stress reveals that the 3, 5, 7
Node 2
welding simulation. In order to understand how
those results might differ from those of a shake- 100
3,5,7
down analysis where kinematic strain hardening
0
is considered, shakedown analyses of a stiffened -400 -300 -200 -100 0 100 200 300 400
plate with no strain hardening and with a kin- -100
1 8
147
5
0.25
0.2
6 CONCLUSIONS
3 7
0.15
Applied stress, σ/σy
Node 1
Node 2 0.5 σy, longitudinal residual stresses were decreased
100 by around 20% and 40%, respectively. Residual
3 7
5 stresses are relieved entirely during the first load
0
-400 -300 -200 -100 0 100 200 300 400
cycle of a constant amplitude cyclic load. When
-100 1 8 the load amplitude is varied, the magnitude of the
4 2 6 highest load is the primary factor controlling the
-200 amount of residual stress shakedown achieved.
Strain (με) After partial stress relief by shakedown under
average axial stresses of 0.25 σy and 0.5 σy, the
Figure 18. Stress-strain history for variable amplitude ultimate strengths of tee-stiffened plates increased
loading. by 1.5–3.5% and 4.5–7%, respectively. When
148
149
Yoichi Sumi
Faculty of Engineering, Yokohama National University, Yokohama, Japan
ABSTRACT: Strength and deformability of steel plates for marine use are studied from the viewpoint
of geometry of corrosion pits and the size effect of corroded plates. The actual shape of corrosion pit
depends on the surrounding environment, which may result in a variety of pitting shapes such as coni-
cal and ellipsoidal shapes. In the present study, the effect of the two shapes has been investigated by
the non-linear, large deformation and three-dimensional finite element analyses for simulated corrosion
surfaces generated by a probabilistic model of a corrosion process. The strength of corroded plates with
semi-ellipsoidal pits is found to be estimated by the empirical formula obtained from that with conical
pits, where the estimation is based on the minimum cross sectional area of the plate. The deformability
and energy absorption of the corroded plates could be estimated by the surface roughness represented by
the difference of averaged plate thickness and that at the section of minimum cross sectional area. Having
investigated the size effect of corroded plates probabilistically, it has been quantitatively shown that the
strength and deformability reduce with increasing plate length, while they may increase with increasing
plate width. The size effect is more pronounced for deformability with the change of width.
151
1 ⎧⎪ ( t − )2 ⎫⎪
fT0 (t ) = exp ⎨ − 0
⎬. (1)
2πσ 0t 2σ 02 ⎪⎭
⎩
152
153
154
DOP (%) 0 20 50 78 93 99
A0 AP
Dm = , (9)
A0
and
In equation (10) and (11) Rd is the ratio of total Figure 13. Strength reduction at different damages for
elongation of pitted plate to that of an intact plate specimens with ellipsoidal pits.
155
and the variables representing surface roughness where σ is the standard deviation of corrosion
are given by diminution of the whole domain. The distribution
function for the maximum Z M among N sections
zavg Pmax zavgg can be given by:
RS Dm − ; RP = , (12), (13)
T T
{ }
N −1
respectively, where zavg is the average corrosion gN max ( z ) = N FZ M ( z ) fZ M ( z ) , (16)
diminution, T is the thickness of the intact plate
and Pmax is the depth of the deepest pit. where fZ M ( z ) is the probability density function
As illustrated in Figure 14, current study shows of the normal distribution of Z M . The mode
that equation (10) is also applicable for ellipsoidal value of the above distribution can be expressed by
pits with the specified range, whereas equation (11) using the inverse of standard normal distribution
gives good estimation in a very limited range of function Φ:
application, say surface roughness (RP) up to 0.1
for ellipsoidal pits (Figure 15). With increasing RP ⎛ 1⎞
value, the scatter of Rd increases, so that equation z MN mod = FZ−M1 1 − ⎟
⎝ N⎠
(10) is recommended for use in the estimation of
σ ⎛ 1⎞
deformability. Deformability is reduced more than =μ+ Φ −1 ⎜1 − ⎟ , (17)
60% due to ellipsoidal pit for a corroded surface M ⎝ N⎠
with DOP = 99%.
where μ is the mean corrosion diminution of
the whole domain. Equation (17) shows that the
5 SIZE EFFECT maximum average sectional corrosion diminu-
tion decreases and converges to μ with increasing
Yamamoto (2008) studied size effect of pitted plate representative width, M, while it monotonically
in the evaluation of average thickness diminution increases with increasing representative length, N.
156
1
N= . (18)
(
⎛ M z MN
1− Φ⎜
− )⎞⎟
⎜⎝ σ ⎟⎠
Re = Ru ⋅ Rd (20)
Figure 16. Linear approximation for relation between where Ru and Rd represent strength reduction
actual and representative width. (equation (8)) and deformability reduction (equa-
tion (10)), respectively. Estimation of Ru solely
depends on damage of surface, whereas Rd can
be estimated from damage and average corrosion
diminution. In order to investigate the size effect
on Ru, Rd and Re, damage can be redefined as:
z MN mod
Dm = (21)
T
157
REFERENCES
Figure 21. Size effect on energy absorption reduction
factor. Ahmmad Md. M. & Sumi, Y. (2010). Strength and
deformability of corroded steel plates under quasi-
static tensile load. J Mar Sci Technol vol. 15(1),
the size effect on strength is not so significant for pp. 1–15.
plates of wider than 500 mm, while deformability Bridgman, P.W. (1964). Studies in Large Plastic Flow and
reduction and energy absorption reduction are Fracture. Cambridge, Massachusetts.
influenced in a wider range. Hallquist, J.O. (1998). LS-DYNA Theoretical Manual.
The results also show that strength and deform- Livermore Software Technology Corporation.
ability reduce with increasing plate length, while Nakai, T., Matsushita, H. & Yamamoto, N. (2004). Effect
they increase with increasing plate width. This of pitting corrosion on local strength of hold frames
of bulk carriers (2nd report)—Lateral-distortional
phenomenon may be interpreted in the following buckling and local face buckling. Marine Structures
way; the parallel load-bearing paths strengthen vol. 17, pp. 612–641.
the structural redundancy with increasing a plate Nakai, T., Matsushita, H. & Yamamoto, N. (2005).
width. On the contrary, a least durable section may Pitting corrosion and its influence on local strength
appear by increasing the specimen length. of hull structural members. Proceedings of the ASME
158
159
ABSTRACT: In Mexico, offshore pipelines are installed inside trenches. This has the effect of restraining
their movements to some extent, due to the weight of the backfill and the friction of the soil around the
pipe. When high temperature hydrocarbons start flowing through the pipelines, the friction around them
prevents the elongation and an axial force builds up. This force reaches a point where the restraining forces
are not enough to hold the pipe and buckling occurs. For buried pipelines the buckling is upwards, thus
the name Up Heaval Buckling (UHB). The mitigation measure for this problem is to add weight in the
form of rock or concrete mattresses at the places where UHB might occur. In Mexico a group of factors
have resulted in very expensive solutions, thus, PEMEX’s engineering department is working on finding
the most adequate methods to cope with UHB.
161
2 BASIC DEFINITIONS FOR UHB where H is the residual tension from the instal-
ANALYSIS lation of the pipe in N, which is normally taken
as null to be conservative; Δpi is the difference in
As in most of the structural analysis and design internal pressure compared to as laid in Pa; Ai is
situations, UHB design is about making sure that the internal transverse area in m2; v is the steel’s
162
2.1 Prop type imperfection Figure 2. Flexural moment diagram and deformed
shape showing the convex zone and wave-length.
Imagine that while laying a pipeline on a flat sea-
floor, a segment of it rests on an imperfection on
the trench bottom that is in contact with the pipe-
line just at one point, as depicted in Figure 1.
As seen in Figure 1, the soil is modelled as
Winckler springs, and the imperfection is modelled
as a simple support at a certain node, that has an
upwards ground displacement of δ m. Note that
the theoretical configuration is symmetrical with
respect to the simple support. For constructing
the model shown in Figure 1, springs in tension
are taken out, to have a better representation of
Figure 3. Transverse section of a trenched pipeline
the pipe/soil interaction. There is a direct relation- mitigated with rock dump.
ship between the magnitude of the curvature and
the propensity to UHB, the higher the curvature,
the lesser is the restraining force preventing UHB. Manipulating Expression 1 and taking the limit,
Besides the height of the theoretical imperfection Equation 3 below is obtained:
δ, other important parameters affecting the prop
shape and the curvature are the submerged weight 4 δ wo
wo, the stiffness E ⋅ I of the pipeline, and the stiff- R max | Sefffff | − w p 11 w o (3)
ness of the soil. It should be clear now that in real- E I
ity, when laying a pipeline inside a trench on the
when Rmax is null or negative means that there is
bottom of the sea, the occurrence of a theoretical
no requirement for additional weight to prevent
imperfection configuration is very rare.
UHB, the own weight and stiffness of the pipe are
enough to keep the pipeline stable. If Rmax is posi-
2.2 Mitigation measures against UHB tive, means that in order to prevent UHB, a down-
ward force of Rmax magnitude has to be applied on
Two important definitions are derived from the
top of the pipeline in the convex zone. The most
flexural moment diagram of the prop imperfection
common way to add weight on top of the pipeline
configuration. The so-called wave-length, which is
to prevent UHB is by dumping gravel inside the
two times the horizontal distance between the max-
trench, on top of the pipeline. Figure 3 depicts a
imum and the minimum elevations of the pipeline,
transverse section of a trenched pipeline mitigated
which are the points with horizontal tangents; and
with rock dump.
the convex zone which can be determined as the
When the length of the pipeline to be mitigated
horizontal distance between the inflection points
against UHB is relatively short, other solutions
of the longitudinal axis of the pipeline. At those
to achieve additional weight are used, as concrete
inflection points the bending moment is null.
mattresses or geotextile bags filled with sand or
In Figure 2 below, the flexural moment diagram of
other material.
the theoretical imperfection configuration is pre-
Analyzing Equations 2 and 3, it can be seen that
sented with the prop shape. The convex zone and
the parameters inducing UHB are:
wave-length are indicated.
− High internal operating pressure Δpi
− steel transverse area As
− high internal operating temperature ΔT
− height of the imperfection δ.
The parameters preventing UHB are:
− Submerged weight of the pipeline wo
− submerged weight of the fluids being trans-
Figure 1. Theoretical configuration for an imperfection ported, included in wp
(prop) of δ m of height. − bending stiffness E ⋅ I.
163
164
165
166
5 CONCLUSIONS
4 RESULTS
The analysis and design of HT/HP pipelines
For the technical merit of the saddlebags or other
prone to buckling is a new challenge for the oth-
solutions, an analysis has been made of the tradi-
erwise offshore experienced PEMEX’s engineering
tional rock-dumping.
department.
PEMEX designed and installed a 30″ pipeline
Allowing the design of exposed offshore pipe-
that is 21 km long, and will be transporting oil
lines could be a means of addressing HT/HP pipe-
at around 120°C. During the design phase of the
lines, but it requires changing the Mexican design
project (PEMEX 2010), the downward force Rmax
code for offshore pipelines (PEMEX, 2009), and
required to keep the pipeline stable, thus preventing
solving hydrodynamic stability in hurricane condi-
UHB was determined. In table 1 in the next page,
tions in different ways than trenching it.
the values of Rmax for different heights of the theo-
Laying, trenching, surveying and smoothing
retical imperfection δ, and 120°C are presented:
practices have a very high impact on the amount
The values presented in the second column of
of mitigating measures needed for a specific pipe-
table 1 are then converted to either rock quanti-
line prone to UHB.
ties or number of concrete mattresses. Using as
Commonly used mitigating measures against
example rock-dump, for each value of downward
UHB in other parts of the world might not be the
force, a rock cover height is determined. From the
most adequate for Mexico’s Bay of Campeche. The
rock cover height the amount of rock is computed
lack of operators, other than PEMEX, in Mexico
(PEMEX 2010). In table 2 below the values of rock
makes mobilization and demobilization costs
cover height and amount of rock are presented:
extremely high. The use of alternative solutions, as
Comparing the downward force required to
the saddlebags presented here, could be adopted by
prevent UHB from happening (second column of
countries where the traditional solutions are not as
table 1) against the weight of rock needed to provide
convenient as they are for the countries that origi-
such downward force (third column of table 2), it is
nally developed them.
found that the amount rock not directly acting on
top of the pipeline, thus not “working” in the UHB
solution ranges from 60% to 85% of the total. REFERENCES
The saddlebags solution, by design, applies all the
mobilized weight on top of the pipeline, thus mak- DNV 2007. Global buckling of submarine pipelines.
ing it a more efficient solution in this aspect also. Structural design due to high temperature/high pressure
DNV-RP-F110. Norway: Det Norske Veritas.
Table 1. Downward forces Hooper, J., Maschner, E. & Farrant, T. 2004. HT/HP
required to prevent UHB for Pipe-in-Pipe Snaked Lay Technology—Industry Chal-
different imperfection heights lenges. OTC 16379. Houston: OTC.
and 120°C. Palmer, A.C., Ellinas, C.P., Richards, D.M. & Guijt, J.
1990. Design of Submarine Pipelines Against Upheaval
Imperfection Buckling. OTC 6335. Houston: OTC.
height (M) Rmax (kN/m) PEMEX 2009. Diseño de líneas submarinas en el Golfo de
México. Mexico: PEMEX.
0.1 4.17 PEMEX 2010. 30″ Pipeline Between—and—Predictive
0.2 11.21 Upheaval Buckling Report. London: KW LTD.
0.3 15.64
0.4 19.07
0.5 21.71
167
O.F. Hughes
Virginia Tech, Blacksburg, USA
ABSTRACT: The objective of this paper is to check the accuracy of the ALPS/ULSAP (Analysis of
Large Plated Structures/Ultimate Limit State Assessment Program) method’s use to determine the ulti-
mate strength of plates and stiffened panels. The details of the ALPS/ULSAP method and theory are
presented in both Ultimate Limit State Design of Steel-plated Structures, co-authored by J.K. Paik and
A.K. Thayamballi, and Ship Structural Analysis and Design, co-authored by O.F. Hughes and J.K. Paik.
In this benchmark study, the accuracy of the plate and stiffened panel ultimate strength obtained with
the ALPS/ULSAP method is ascertained through comparison with that obtained using nonlinear finite
element methods and the DNV/PULS method.
169
tp tp tp
N. A. N. A. N. A.
hw hw hw
tw tw tw
tf tf
bf bf
Figure 1(b). Stiffened plate structure—nomenclature (c) After buckling, the unloaded edges remain straight.
of the stiffeners.
Figure 3. Membrane stress distribution inside a plate
subject to uniaxial compressive loads in the case of one
bulge in the middle of the plate.
170
171
172
Long. stiffeners
the smallest value among those computed taken as b/2
its real ultimate strength.
b
b/2
b/2
3 BENCHMARK STUDY
a/2 a/2 a a/2 a/2
3.1 Study methods
The three following methods were employed for Figure 6(a). Nonlinear finite element method
comparison purposes in this benchmark study. modelling—extent of the analysis for the plate.
FEM
ported. No welding residual stresses are consid- ALPS/ULSAP
DNV/PULS
ered, although the plates have initial deflection wopl, 0.4
which corresponds to the plate buckling mode, as
follows. 0.2
( )
0 0.2 0.4 0.6 0.8 1.0
173
FEM
ALPS/ULSAP • Elastic modulus, E = 205800 MPa
0.4
DNV/PULS
• Poisson’s ratio, v = 0.3
No welding residual stresses are considered. The
0.2
following condition is applied for the plate initial
deflection.
0.0
• wopl = 0.1β 2tp for FEA and ALPS/ULSAP;
0 0.2 0.4 0.6 0.8 1.0
• wopl = b/200 for DNV/PULS,
σxu/ σY where wopl = the plate initial deflection ampli-
tude corresponding to the buckling mode,
Figure 8. Ultimate strength interaction relationship
between biaxial compressive loads for the plate with
β p (
σYp E . )
tp = 16 mm. For the initial distortions of the stiffeners,
the following condition is applied for all of the
1.0
methods.
a ×bb==2550×
a× 2550 ×850(mm)
850 (mm)
tp= 22
22mm
mm • Column-type initial distortion, woc = 0.0015a
0.8
• Sideways initial distortion, wos = 0.0015a
Three types of stiffeners, namely, flat-bar, angle-
0.6
bar, and T-bar stiffeners, are considered. The four
σyu/ σY
0.6
174
175
σxu/ σYeq
σxu/ σYeq
III IV IV
0.6 III 0.6
III
Mode III
III III
0.4 0.4
III
III
Mode III III
0.2 0.2
0.0 0.0
0 1 2 3 4 0 1 2 3 4
(b / t p ) σYp / E (b / t p ) σYp / E
Figure 12(b). Ultimate strength of the panels under Figure 13(a). Ultimate strength of the panels under lon-
longitudinal compression for flat-bar stiffeners (size 2). gitudinal compression for angle-bar stiffeners (size 1).
1.2
1.2 Panel C: hw× bf× tw/tf = 235 × 90 × 10/15 (mm) (A)
Panel C: hw × tw = 350 × 35 (mm) (F)
FEA (ANSYS)
1.0 FEA (MSC/MARC)
1.0 Design Formula (ALPS/ULSAP)
Design Formula (DNV/PULS)
0.8
0.8
σxu/ σYeq
III IV
σxu/ σYeq
1.2 1.2
Panel C: hw × tw = 550 × 35 (mm) (F) Panel C: hw × bf × tw/tf = 383 × 100 × 12/17 (mm) (A)
1.0 1.0
Mode III IV
0.8 0.8 III
σxu/ σYeq
σxu/ σYeq
IV Mode III
V
IV IV V
II V V
0.6 0.6
0.4 0.4
FEA (ANSYS)
FEA (ABAQUS) FEA (ANSYS)
0.2 0.2 FEA (MSC/MARC)
FEA (MSC/MARC)
Design Formula (ALPS/ULSAP) Design Formula (ALPS/ULSAP)
Design Formula (DNV/PULS) Design Formula (DNV/PULS)
0.0 0.0
0 1 2 3 4 0 1 2 3 4
(b / t p ) σYp / E (b / t p ) σYp / E
Figure 12(d). Ultimate strength of the panels under Figure 13(c). Ultimate strength of the panels under lon-
longitudinal compression for flat-bar stiffeners (size 4). gitudinal compression for angle-bar stiffeners (size 3).
176
1.0 1.0
Mode III V
0.8 0.8
V V
σxu/ σYeq
σxu/ σYeq
Mode III
II V
II V
II V V
0.6 0.6
0.4 0.4
FEA (ANSYS)
FEA (ABAQUS) FEA (ANSYS)
0.2 0.2 FEA (MSC/MARC)
FEA (MSC/MARC)
Design Formula (ALPS/ULSAP) Design Formula (ALPS/ULSAP)
Design Formula (DNV/PULS) Design Formula (DNV/PULS)
0.0 0.0
0 1 2 3 4 0 1 2 3 4
( b / t p ) σYp / E ( b / t p ) σYp / E
Figure 13(d). Ultimate strength of the panels under lon- Figure 14(c). Ultimate strength of the panels under
gitudinal compression for angle-bar stiffeners (size 4). longitudinal compression for T-bar stiffeners (size 3).
1.2 1.2
Panel C: hw× bf× tw/tf = 138 × 90 × 9/12 (mm) (T) Panel C: hw × bf × tw/tf = 580 × 150 × 15/20 (mm) (T)
FEA (ANSYS)
1.0 FEA (ABAQUS) 1.0
FEA (MSC/MARC)
Design Formula (ALPS/ULSAP)
Mode V V
Design Formula (DNV/PULS) 0.8
0.8 V
σxu/ σYeq
σxu/ σYeq
II
II
II
0.6 0.6
III 0.4
0.4
III
III FEA (ANSYS)
III
Mode III III FEA (ABAQUS)
0.2 0.2
FEA (MSC/MARC)
Design Formula (ALPS/ULSAP)
Design Formula (DNV/PULS)
0.0 0.0
0 1 2 3 4 0 1 2 3 4
(b / t p ) σYp / E ( b / tp ) σYp / E
Figure 14(a). Ultimate strength of the panels under Figure 14(d). Ultimate strength of the panels under
longitudinal compression for T-bar stiffeners (size 1). longitudinal compression for T-bar stiffeners (size 4).
0.8
σxu/ σYeq
III
V
0.6
V V
0.6
III
Mode III 0.4
0.4
0.2
0.2
0.0
0.0 0.0 0.5 1.0 1.5 2.0 2.5
0 1 2 3 4 (a / π r ) σYeq / E
( b / t p ) σYp / E
Figure 15(a). Variation in normalized panel ultimate
Figure 14(b). Ultimate strength of the panels under strength under longitudinal compression as a function of
longitudinal compression for T-bar stiffeners (size 2). the column slenderness ratio with flat-bar stiffeners.
177
0.8 0.8
σxu/ σYeq
Mode I
σyu/ σYeq
0.6 0.6
III
0.4 0.4 III
III
III III
0.2 0.2
0.0 0.0
0.0 0.5 1.0 1.5 2.0 2.5 0 1 2 3 4
(a / π r ) σYeq / E ( b / t p ) σYp / E
Figure 15(b). Variation in normalized panel ultimate Figure 16(a). Ultimate strength of the panels under
strength under longitudinal compression as a function of transverse compression for flat-bar stiffeners (size 1).
the column slenderness ratio with angle-bar stiffeners.
1.2
1.2 Panel C: h w× tw = 250× 25 (mm) (F)
Tee bar under longitudinal compressive loads
FEA (ANSYS)
FEA (ANSYS) Design Formula (ALPS/ULSAP)
Design formula (ALPS/ULSAP) 1.0
1.0
0.8
0.8
σyu/ σYeq
Mode III
σxu/ σYeq
0.6
0.6
III
0.4
0.4 III
III
IV IV
0.2 0.2
0.0 0.0
0.0 0.5 1.0 1.5 2.0 2.5 0 1 2 3 4
(a / π r ) σYeq / E ( b / t p ) σYp / E
Figure 15(c). Variation in normalized panel ultimate Figure 16(b). Ultimate strength of the panels under
strength under longitudinal compression as a function of transverse compression for flat-bar stiffeners (size 2).
the column slenderness ratio with T-bar stiffeners.
1.2
Panel C: h w× tw = 350× 35 (mm) (F)
Through the wide range of panel dimensions, the FEA (ANSYS)
Design Formula (ALPS/ULSAP)
1.0
comparisons show that the ALPS/ULSAP method
solutions are in very good agreement with those
of the nonlinear FEA, though slightly on the con- 0.8
σyu/ σYeq
Mode III
servative side, whereas the ALPS/ULSAP method
tends to underestimate the ultimate strength of the 0.6
panels with large flat-bar stiffeners, as shown in III
Figs. 12(c) and 12(d). 0.4 IV
IV
IV IV
3.3.2 Stiffened panels under transverse 0.2
compression
Figures 16 to 18 show the normalized panel ultimate 0.0
strength under transverse compression as a function 0 1 2 3 4
of the slenderness coefficient β of the plate between (b / t p ) σYp / E
the longitudinal stiffeners by comparison with
nonlinear FEA and ALPS/ULSAP for flat-bar, Figure 16(c). Ultimate strength of the panels under
angle-bar, and T-bar stiffeners, respectively. transverse compression for flat-bar stiffeners (size 3).
178
0.8 0.8
σyu/ σYeq
σyu/ σYeq
0.6 0.6
IV V
0.4 IV
0.4
V
IV IV
IV IV IV IV
0.2 0.2
0.0 0.0
0 1 2 3 4 0 1 2 3 4
(b / t p ) σYp / E ( b / t p ) σYp / E
Figure 16(d). Ultimate strength of the panels under Figure 17(c). Ultimate strength of the panels under
transverse compression for flat-bar stiffeners (size 4). transverse compression for angle-bar stiffeners (size 3).
1.2 1.2
Panel C: hw × bf × tw/tf = 138 × 90 × 9/12 (mm) (A) Panel C: hw × bf × tw/tf = 580 × 150 × 12/17 (mm) (A)
FEA (ANSYS) FEA (ANSYS)
Design Formula (ALPS/ULSAP) Design Formula (ALPS/ULSAP)
1.0 1.0
0.8 0.8
Mode V
σyu/ σYeq
σyu/ σYeq
Mode III
0.6 0.6
IV
III
0.4 0.4 IV
III
IV
III IV IV
III III
0.2 0.2
0.0 0.0
0 1 2 3 4 0 1 2 3 4
(b / t p ) σYp / E ( b / t p ) σYp / E
Figure 17(a). Ultimate strength of the panels under Figure 17(d). Ultimate strength of the panels under
transverse compression for angle-bar stiffeners (size 1). transverse compression for angle-bar stiffeners (size 4).
1.2 1.2
Panel C: hw× bf× tw/tf = 235× 90× 10/15 (mm) (A) Panel C: hw× bf× tw/tf = 138× 90× 9/12 (mm) (T)
FEA (ANSYS) FEA (ANSYS)
Design Formula (ALPS/ULSAP) Design Formula (ALPS/ULSAP)
1.0 1.0
0.8 0.8
σyu/ σYeq
σyu/ σYeq
0.6 0.6
III III
0.4 III
0.4
III
V III
V V III III
0.2 0.2
0.0 0.0
0 1 2 3 4 0 1 2 3 4
( b / t p ) σYp / E ( b / t p ) σYp / E
Figure 17(b). Ultimate strength of the panels under Figure 18(a). Ultimate strength of the panels under
transverse compression for angle-bar stiffeners (size 2). transverse compression for T-bar stiffeners (size 1).
179
Mode III
sive strength tends to remain unchanged for the
0.6 column slenderness ratio of the longitudinal stiff-
III
eners as long as the plate is identical. It is interest-
0.4 ing to note that a clear relationship between the
III
V
V
panel ultimate transverse compressive strength
V
0.2 and the column slenderness ratio of the longitu-
dinal stiffeners exists, representing that the panel
0.0
ultimate transverse compressive strength increases
0 1 2 3 4 as the column slenderness ratio of the longitudinal
( b / t p ) σYp / E stiffeners with attached plating increases when the
0.6
0.8
σyu/ σYeq
Mode III
0.4
0.6
V
0.2
0.4
V
IV
IV IV 0.0
0.2 0.0 0.5 1.0 1.5 2.0 2.5
( a / π r ) σYeq / E
0.0
0 1 2 3 4
(b / t p ) σYp / E Figure 19(a). Variation in normalized panel ultimate
strength under transverse compression as a function of
the column slenderness ratio with flat-bar stiffeners.
Figure 18(c). Ultimate strength of the panels under
transverse compression for T-bar stiffeners (size 3).
0.8
Mode V
σyu/ σYeq
0.6
0.6
IV 0.4
0.4 IV
IV
IV IV
0.2
0.2
0.0
0.0 0.0 0.5 1.0 1.5 2.0 2.5
0 1 2 3 4 (a / π r ) σYeq / E
( b / t p ) σYp / E
Figure 19(b). Variation in normalized panel ultimate
Figure 18(d). Ultimate strength of the panels under strength under transverse compression as a function of
transverse compression for T-bar stiffeners (size 4). the column slenderness ratio with angle-bar stiffeners.
180
σyu/ σYeq
0.6
0.6
0.4
0.4 Mode III
0.2 0.2
0.0 0.0
0.0 0.5 1.0 1.5 2.0 2.5
0.0 0.2 0.4 0.6 0.8 1.0
( a / π r ) σYeq / E σxu/ σYeq
Figure 19(c). Variation in normalized panel ultimate Figure 20(a). Ultimate strength interaction relationship
strength under transverse compression as a function of between biaxial compressive loads for tp = 18.5 mm and
the column slenderness ratio with T-bar stiffeners. flat-bar stiffeners (size 1).
1.0
Panel C: tp= 18.5 mm
plate thickness is changed but with identical longi- hw× tw = 250× 25 (mm) (F)
tudinal stiffeners. 0.8
FEA (ANSYS)
Design Formula (ALPS/ULSAP)
The comparisons show that the ALPS/ULSAP
method solutions for the panel ultimate strength
σyu/ σYeq
0.6
The comparisons show that the ALPS/ULSAP
method computations are in very good agreement
0.4
with those of the nonlinear FEA for the wide Mode III
Mode II
range of plate and stiffener dimensions with differ-
ent types of stiffeners. 0.2
Mode IV
181
σyu/ σYeq
0.6 0.6
0.4 Mode IV
0.4 Mode III
Mode II Mode II
Mode III
0.2 0.2
Mode IV Mode V
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
σxu/ σYeq σxu/ σYeq
Figure 20(d). Ultimate strength interaction relationship Figure 21(c). Ultimate strength interaction relationship
between biaxial compressive loads for tp = 18.5 mm and between biaxial compressive loads for tp = 18.5 mm and
flat-bar stiffeners (size 4). angle-bar stiffeners (size 3).
1.0
1.0 Panel C: tp= 18.5 mm
Panel C: tp= 18.5 mm
hw× bf× tw/tf = 580× 150× 15/20 (mm) (A)
hw× bf× tw/tf = 138× 90× 9/12 (mm) (A) FEA (ANSYS)
FEA (ANSYS) 0.8 Design Formula (ALPS/ULSAP)
0.8 Design Formula (ALPS/ULSAP)
σyu/ σYeq
0.6
σyu/ σYeq
0.6
0.2 0.2
Mode IV
Mode V
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
σxu/ σYeq σxu/ σYeq
Figure 21(a). Ultimate strength interaction relationship Figure 21(d). Ultimate strength interaction relationship
between biaxial compressive loads for tp = 18.5 mm and between biaxial compressive loads for tp = 18.5 mm and
angle-bar stiffeners (size 1). angle-bar stiffeners (size 4).
1.0 1.0
Panel C: tp= 18.5 mm Panel C: tp= 18.5 mm
hw× bf× tw/tf = 235× 90× 10/15 (mm) (A) hw× bf× tw/tf = 138× 90× 9/12 (mm) (T)
FEA (ANSYS) FEA (ANSYS)
0.8 Design Formula (ALPS/ULSAP) 0.8
FEA (MSC/MARC)
Design Formula (ALPS/ULSAP)
σyu/ σYeq
σyu/ σYeq
0.6 0.6
0.2 0.2
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
σxu/ σYeq σxu/ σYeq
Figure 21(b). Ultimate strength interaction relationship Figure 22(a). Ultimate strength interaction relationship
between biaxial compressive loads for tp = 18.5 mm and between biaxial compressive loads for tp = 18.5 mm and
angle-bar stiffeners (size 2). T-bar stiffeners (size 1).
182
σyu/ σYeq
0.6 0.6
0.2 0.2
Mode II
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
σxu/ σYeq σxu/ σYeq
Figure 22(b). Ultimate strength interaction relationship Figure 23(a). Ultimate strength interaction relationship
between biaxial compressive loads for tp = 18.5 mm and between biaxial compressive loads for tp = 11 mm and
T-bar stiffeners (size 2). T-bar stiffeners (size 4).
1.0 1.0
Panel C: tp = 18.5 mm Panel C: tp= 12.5 mm
hw × bf × tw/tf = 383 × 100 × 12/17 (mm) (T)
hw× bf× tw/tf = 580× 150× 15/20 (mm) (T)
FEA (ANSYS) FEA (ANSYS)
0.8 FEA (MSC/MARC) 0.8 FEA (MSC/MARC)
Design Formula (ALPS/ULSAP) Design Formula (ALPS/ULSAP)
σyu/ σYeq
σyu/ σYeq
0.6 0.6
0.2 0.2
Mode V Mode II
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
σxu/ σYeq σxu/ σYeq
Figure 22(c). Ultimate strength interaction relationship Figure 23(b). Ultimate strength interaction relationship
between biaxial compressive loads for tp = 18.5 mm and between biaxial compressive loads for tp = 12.5 mm and
T-bar stiffeners (size 3). T-bar stiffeners (size 4).
1.0 1.0
Panel C: tp= 18.5 mm Panel C: tp= 15 mm
hw× bf× tw/tf = 580× 150× 15/20 (mm) (T) hw× bf× tw/tf = 580× 150× 15/20 (mm) (T)
FEA (ANSYS) FEA (ANSYS)
0.8 FEA (MSC/MARC) 0.8 FEA (MSC/MARC)
Design Formula (ALPS/ULSAP) Design Formula (ALPS/ULSAP)
σyu/ σYeq
σyu/ σYeq
0.6 0.6
0.4 Mode IV
0.4
Mode IV Mode II
Mode II
0.2 0.2
Mode V
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
σxu/ σYeq σxu/ σYeq
Figure 22(d). Ultimate strength interaction relationship Figure 23(c). Ultimate strength interaction relationship
between biaxial compressive loads for tp = 18.5 mm and between biaxial compressive loads for tp = 15 mm and
T-bar stiffeners (size 4). T-bar stiffeners (size 4).
183
0.6
0.4 Mode IV
Mode II
0.2
Mode V
1.0
Panel C: tp= 25 mm
hw× bf× tw/tf = 580× 150× 15/20 (mm) (T)
FEA (ANSYS)
0.8 FEA (MSC/MARC)
Design Formula (ALPS/ULSAP)
σyu/ σYeq
0.6
Mode III
Figure 25(a). Shape of the initial deflection before lateral
pressure loading (with an amplification factor of 20).
0.4 Mode II
0.2
Mode V
0.0
0.0 0.2 0.4 0.6 0.8 1.0
σxu/ σYeq
Mode V
0.6
0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
σxu/ σYeq
Figure 23(f). Ultimate strength interaction relationship Figure 25(c). Shape of the initial deflection after lateral
between biaxial compressive loads for tp = 37 mm and pressure loading in the stiffener-sided pressure loading
T-bar stiffeners (size 4). (with an amplification factor of 20).
184
σxu/ σYeq
ing order changes the shape of the initial deflec- 0.6
tions before longitudinal compression is applied,
as shown in Fig. 25. Plate-sided Pressure
Figure 26 shows the deformed shapes of the 0.4
panel with T-bar stiffeners (size 3) and tp = 15 mm
at the ultimate limit state under combined longi-
tudinal compression and lateral pressure loads, 0.2
where p = 0.25 MPa. Stiffener-sided Pressure
0.6
this paper was to check the accuracy of the ALPS/
ULSAP method’s use to determine plate and stiff-
0.4 ened panel ultimate strength, compared with non-
Plate-sided Pressure
linear FEA and the DNV/PULS method.
Mode III
The dimensions and material properties of
0.2 a real ship panel were selected as the standard
Stiffener-sided Pressure panel for testing purposes, and a wider range of
0.0
tp=15 mm plating and stiffener dimensions was considered
0 0.1 0.2 0.3 0.4 by varying the panel’s properties. Three types
p (MPa) of stiffeners, namely, flat-bar, angle-bar, and T-bar
stiffeners, were considered, and different loading
Figure 27. Ultimate strength interaction relationship conditions, including longitudinal compression,
between longitudinal compression and lateral pressure transverse compression, biaxial compression, and
loads for the panel with T-bar stiffeners (size 3) and combined longitudinal compression and lateral
tp = 15 mm. pressure loads, were applied.
185
REFERENCES
186
A.E. Mansour
University of California, Berkeley, USA
J.B. Caldwell
Emeritus Professor, Newcastle University, Newcastle upon Tyne, UK
ABSTRACT: The objective of this paper is to develop a modified Paik-Mansour formula for the
ultimate strength calculations of ship hulls subject to vertical bending moments. The method is based on
a credible bending stress distribution over the hull cross-section presumed at the ultimate limit state. The
accuracy of this method is demonstrated through comparison with computations obtained using more
refined methods, such as nonlinear finite element method, intelligent super-size finite element method,
and idealized structural unit method. Statistical analysis of the hull girder ultimate strength based on
comparisons among the various computations is carried out in terms of their mean values and coefficients
of variation. The original Paik-Mansour method is found to be inapplicable to the case of a pure vertical
bending moment depending on the ship’s hull type and/or vertical bending direction, but the modified
Paik-Mansour method is more general and is able to resolve this issue.
187
188
+ –
x
+ – denoted by Mus (negative value) for the sagging
σ xE condition and by Muh (positive value) for the hog-
σ xE + Tens.
– Comp.
ging condition.
gus guh
189
190
191
equivalent to the material yield stress, whereas the ⎧ 032β 4 β 2 + 1.00 f β ≤ 1.5
ultimate strength of those under axial compression σ xu ⎪ −0.03
= ⎨1.274 / β β ≤ 3.0 (5)
is assumed to be as follows. σYp ⎪
⎩1.248 / β 0.283 β > 3.0
2
192
ing that only the tension flange, i.e., the deck panel 6
in the hogging condition or the outer bottom panel 5
in the sagging condition, has yielded until the 4
ultimate strength is reached. It can be seen that the 3
original and modified Paik-Mansour methods ANSYS
ALPS/HULL
2
produce identical results for this case. CSR
1 Modified Paik-Mansour formula method
0
4.2 Container ship hull 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50
Curvature (1/km)
Figure 8 shows the results of the ultimate strength
behavior comparison for the container ship hull Figure 8(a). Ultimate strength behavior of the con-
tainer ship hull under a hogging moment.
12
9
Hogging bending moment (MNm)
10
Sagging bending moment (GNm)
7
8
6
6 5
4
4
ANSYS 3
ALPS/HULL ANSYS
CSR 2 ALPS/HULL
2 CSR
Modified Paik-Mansour formula method
1 Modified Paik-Mansour formula method
0 0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50
Curvature (1/km) Curvature (1/km)
Figure 7(a). Ultimate strength behavior of the Dow’s Figure 8(b). Ultimate strength behavior of the con-
frigate test hull under a hogging moment. tainer ship hull under a sagging moment.
193
0
subject to a vertical bending moment. Table 2 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
presents the heights of the collapsed and yielded Curvature (1/km)
Figure 10 shows the results of the ultimate Hogging (mm) Sagging (mm)
strength behavior comparison for the Suezmax-
class double-hull tanker hull subject to a vertical Method hC hY hC hY
bending moment.
Original P-M – – 17935.0 0.0
Table 4 presents the heights of the collapsed and Modified P-M 1654.1 13.7 17935.0 0.0
yielded parts of the Suezmax double-hull tanker
hull. It can be seen that, in the hogging condition,
the height of the yielded part is larger than that of 16
the collapsed part because the double bottom struc-
Hogging bending moment (GNm)
14
tures are much heavier than the deck structures
in this type of ship. The modified Paik-Mansour 12
6
ANSYS
4.5 Single-hull tanker hull 4 ALPS/HULL
CSR
Figure 11 shows the results of the ultimate strength 2 Modified Paik-Mansour formula method
hull subject to a vertical bending moment. Table 5 0.00 0.05 0.10 0.15
Curvature (1/km)
0.20 0.25 0.30
194
12
ANSYS
Original P-M 7035.2 0.0 15225.5 0.0
4
ALPS/HULL Modified P-M 7035.2 0.0 15225.5 0.0
CSR
2 Modified Paik-Mansour formula method
0
0.00 0.05 0.10 0.15 0.20 0.25 0.30
Curvature (1/km) 30
Method hC hY hC hY 0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40
Original P-M – – 16078.5 0.0 Curvature (1/km)
Modified P-M 12.1 2210.6 16078.5 0.0
Figure 12(a). Ultimate strength behavior of the double-
hull tanker hull under a hogging moment.
21
Hogging bending moment (GNm)
18
15 30
Sagging bending moment (GNm)
12
25
9
ANSYS 20
6
ALPS/HULL
CSR
3 Modified Paik-Mansour formula method 15
0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 10
Curvature (1/km) ANSYS
ALPS/HULL
5 CSR
Figure 11(a). Ultimate strength behavior of the single- Modified Paik-Mansour formula method
hull tanker hull under a hogging moment.
0
16
Figure 12(b). Ultimate strength behavior of the double-
hull tanker hull under a sagging moment.
12
ANSYS
4.6 Double-hull tanker hull
ALPS/HULL
4 CSR Figure 12 shows the results of the ultimate strength
Modified Paik-Mansour formula method
behavior comparison for the double-hull tanker
0
hull subject to a vertical bending moment.
0.00 0.05 0.10 0.15 0.20 0.25 0.30
Curvature (1/km) Table 6 presents the heights of the collapsed and
yielded parts of the double-hull tanker hull, which
Figure 11(b). Ultimate strength behavior of the single- has similar characteristics to the Suezmax-class
hull tanker hull under a sagging moment. double-hull tanker hull (see Table 4).
195
(Mu)Formula (GNm)
5.2 Formula method versus ALPS/HULL 20
intelligent super-size finite element method
Table 8 presents the mean values and coefficients of 15
Original P-M – – 20240.7 0.0 Figure 13. Trend of the deviation in ultimate hull
Modified P-M 15.9 3816.0 20240.7 0.0 strength between the formula method and the ANSYS
nonlinear finite element method.
Table 7. Comparison between developed formula method and ANSYS nonlinear finite element method.
Hogging Sagging
Dow’s test 0.013 0.010 0.772 0.011 0.840 0.920 0.009 0.697 0.011 0.793 0.879
hull
Container 9.220 6.400 0.694 6.969 0.756 0.918 7.077 0.768 6.951 0.754 1.018
ship
Bulk carrier 20.394 16.576 0.813 17.500 0.858 0.947 14.798 0.726 15.800 0.775 0.937
D/H 17.677 13.965 0.790 14.066 0.796 0.993 12.213 0.691 11.151 0.631 1.095
Suezmax
S/H VLCC 22.578 18.701 0.828 17.355 0.769 1.078 17.825 0.789 16.179 0.717 1.102
D/H VLCC 32.667 25.667 0.786 27.335 0.837 0.939 22.390 0.685 22.495 0.689 0.995
Mean 0.966 1.004
S-D 0.061 0.088
COV 0.063 0.087
Note: MP = fully plastic bending capacity, Muh = ultimate hogging moment, Mus = ultimate sagging moment,
S-D = standard deviation, COV = coefficient of variation.
196
Hogging Sagging
Dow’s test 0.013 0.010 0.772 0.011 0.799 0.966 0.009 0.697 0.010 0.743 0.939
hull
Container 9.220 6.400 0.694 6.916 0.750 0.925 7.077 0.768 6.635 0.720 1.067
ship
Bulk carrier 20.394 16.576 0.813 16.602 0.814 0.998 14.798 0.726 15.380 0.754 0.962
D/H 17.677 13.965 0.790 13.308 0.753 1.049 12.213 0.691 11.097 0.628 1.101
Suezmax
S/H VLCC 22.578 18.701 0.828 17.335 0.768 1.079 17.825 0.789 17.263 0.765 1.033
D/H VLCC 32.667 25.667 0.786 25.600 0.784 1.003 22.390 0.685 22.000 0.673 1.018
Mean 1.003 1.020
S-D 0.055 0.061
COV 0.055 0.060
Note: Mp = fully plastic bending capacity, Muh = ultimate hogging moment, Mus = ultimate sagging moment,
S-D = standard deviation, COV = coefficient of variation.
20
element method for all six types of ship hull.
Figure 17 shows the trend of the deviation in ultimate
15 hull strength between these two methods.
10
Hog / Sag
5.6 CSR idealized structural unit method versus
: Dow’s frigate test hull ALPS/HULL intelligent super-size finite
: Container ship
: Bulk carrier element method
0.0125 : D/H Suezmax
: S/H VLCC
: D/H VLCC
Table 12 presents the mean values and coeffi-
0
cients of variation for the CSR idealized structural
0 0.0125 10 15 20 25 30
unit method versus the ALPS/HULL intelligent
(Mu)ALPS/HULL (GNm) super-size finite element method for all six types of
ship hull. Figure 18 shows the trend of the devia-
Figure 14. Trend of the deviation in ultimate hull tion in ultimate hull strength between these two
strength between the formula method and the ALPS/ methods.
HULL intelligent super-size finite element method.
197
Hogging Sagging
Dow’s test 0.013 0.010 0.772 0.012 0.888 0.870 0.009 0.697 0.010 0.764 0.912
hull
Container 9.220 6.400 0.694 8.040 0.872 0.796 7.077 0.768 7.843 0.851 0.902
ship
Bulk carrier 20.394 16.576 0.813 17.941 0.880 0.924 14.798 0.726 14.475 0.710 1.022
D/H 17.677 13.965 0.790 15.714 0.889 0.889 12.213 0.691 12.420 0.703 0.983
Suezmax
S/H VLCC 22.578 18.701 0.828 19.889 0.881 0.940 17.825 0.789 17.868 0.791 0.998
D/H VLCC 32.667 25.667 0.786 28.352 0.868 0.905 22.390 0.685 24.798 0.759 0.903
Mean 0.887 0.953
S-D 0.051 0.054
COV 0.058 0.056
Note: Mp = fully plastic bending capacity, Muh = ultimate hogging moment, Mus = ultimate sagging moment,
S-D = standard deviation, COV = coefficient of variation.
198
Hogging Sagging
Dow’s test 0.013 0.011 0.840 0.011 0.799 0.952 0.011 0.793 0.010 0.743 0.936
hull
Container 9.220 6.969 0.756 6.916 0.750 0.992 6.951 0.754 6.635 0.720 0.955
ship
Bulk carrier 20.394 17.500 0.858 16.602 0.814 0.949 15.800 0.775 15.380 0.754 0.973
D/H 17.677 14.066 0.796 13.308 0.753 0.946 11.151 0.631 11.097 0.628 0.995
Suezmax
S/H VLCC 22.578 17.355 0.769 17.335 0.768 0.999 16.179 0.717 17.263 0.765 1.067
D/H VLCC 32.667 27.335 0.837 25.600 0.784 0.937 22.495 0.689 22.000 0.673 0.978
Mean 0.962 0.984
S-D 0.026 0.045
COV 0.027 0.046
Note: Mp = fully plastic bending capacity, Muh = ultimate hogging moment, Mus = ultimate sagging moment,
S-D = standard deviation, COV = coefficient of variation.
30
side structures and longitudinal bulkheads) to the
ultimate hull bending strength was recognised in
25 the original paper as being less straightforward;
and the assumption of a modified “strength
(Mu)ALPS/HULL (GNm)
199
Hogging Sagging
Dow’s test 0.013 0.011 0.840 0.012 0.888 1.058 0.011 0.793 0.010 0.764 0.963
hull
Container 9.220 6.969 0.756 8.040 0.872 1.154 6.951 0.754 7.843 0.851 1.128
ship
Bulk carrier 20.394 17.500 0.858 17.941 0.880 1.025 15.800 0.775 14.475 0.710 0.916
D/H 17.677 14.066 0.796 15.714 0.889 1.117 11.151 0.631 12.420 0.703 1.114
Suezmax
S/H VLCC 22.578 17.355 0.769 19.889 0.881 1.146 16.179 0.717 17.868 0.791 1.104
D/H VLCC 32.667 27.335 0.837 28.352 0.868 1.037 22.495 0.689 24.798 0.759 1.102
Mean 1.090 1.055
S-D 0.056 0.091
COV 0.052 0.086
Note: Mp = fully plastic bending capacity, Muh = ultimate hogging moment, Mus = ultimate sagging moment,
S-D = standard deviation, COV = coefficient of variation.
30
30
25
25
20
(Mu)CSR (GNm)
20
(Mu)CSR (GNm)
15
15
10
Hog / Sag
10
Hog / Sag : Dow’s frigate test hull
: Dow’s frigate test hull : Container ship
: Container ship : Bulk carrier
: Bulk carrier 0.0125 : D/H Suezmax
0.0125 : D/H Suezmax : S/H VLCC
: S/H VLCC : D/H VLCC
: D/H VLCC
0
0 0 0.0125 10 15 20 25 30
0 0.0125 10 15 20 25 30 (Mu)ALPS/HULL (GNm)
(Mu)ANSYS (GNm)
Figure 18. Trend of the deviation in ultimate hull
Figure 17. Trend of the deviation in ultimate hull strength between the CSR idealized structural unit
strength between the CSR idealized structural unit method method and the ALPS/HULL intelligent super-size finite
and the ANSYS nonlinear finite element method. element method.
on the tension side of this web, is postulated to In the Applied Examples summarised in Sec-
have reached the material yield stress (see Figure 4) tion 4, it is evident that in certain cases this yielding
when the ultimate bending strength is reached. It of the material in the “webs” of ship hull girders
is noted that the modified Paik-Mansour method might indeed extend for significant distances ver-
presented in the paper covers both the above two tically from the yielding “flange”. The amount
‘extreme’ models. Furthermore, the modified of this contribution to the ultimate longitudinal
Paik-Mansour method is more general and is able strength of a ship from the tensioned material in
to accommodate a pure bending moment regard- the cross-section will, of course, depend on the
less of the geometrical properties of ship hulls and/ proportion and disposition of “vertical” material
or bending moment direction. in the hull cross-section.
200
Hogging Sagging
Dow’s test 0.013 0.011 0.799 0.012 0.888 1.111 0.010 0.743 0.010 0.764 1.029
hull
Container 9.220 6.916 0.750 8.040 0.872 1.163 6.635 0.720 7.843 0.851 1.182
ship
Bulk carrier 20.394 16.602 0.814 17.941 0.880 1.081 15.380 0.754 14.475 0.710 0.941
D/H 17.677 13.308 0.753 15.714 0.889 1.181 11.097 0.628 12.420 0.703 1.119
Suezmax
S/H VLCC 22.578 17.335 0.768 19.889 0.881 1.147 17.263 0.765 17.868 0.791 1.035
D/H VLCC 32.667 25.600 0.784 28.352 0.868 1.108 22.000 0.673 24.798 0.759 1.127
Mean 1.132 1.072
S-D 0.038 0.087
COV 0.034 0.081
Note: Mp = fully plastic bending capacity, Muh = ultimate hogging moment, Mus = ultimate sagging moment, S-D = stand-
ard deviation, COV = coefficient of variation.
Formula/ANSYS Formula/ ALPS Formula/ CSR ALPS/ ANSYS CSR/ ANSYS CSR/ ALPS
Ship Hog Sag Hog Sag Hog Sag Hog Sag Hog Sag Hog Sag
Dow’s test 0.920 0.879 0.966 0.939 0.870 0.912 0.952 0.936 1.058 0.963 1.111 1.029
hull
Container 0.918 1.018 0.925 1.067 0.796 0.902 0.992 0.955 1.154 1.128 1.163 1.182
ship
Bulk carrier 0.947 0.937 0.998 0.962 0.924 1.022 0.949 0.973 1.025 0.916 1.081 0.941
D/H 0.993 1.095 1.049 1.101 0.889 0.983 0.946 0.995 1.117 1.114 1.181 1.119
Suezmax
S/H VLCC 1.078 1.102 1.079 1.033 0.940 0.998 0.999 1.067 1.146 1.104 1.147 1.035
D/H VLCC 0.939 0.995 1.003 1.018 0.905 0.903 0.937 0.978 1.037 1.102 1.108 1.127
Mean 0.966 1.004 1.003 1.020 0.887 0.953 0.962 0.984 1.090 1.055 1.132 1.072
S-D 0.061 0.088 0.055 0.061 0.051 0.054 0.026 0.045 0.056 0.091 0.038 0.087
COV 0.063 0.087 0.055 0.060 0.058 0.056 0.027 0.046 0.052 0.086 0.034 0.081
While useful methods have been developed in limiting resistance. Any margin above this “critical
the literature, it is however emphasized that the bending moment” provided by progressive yielding
scarcity of full-scale data from experiments or ship of the remaining structure is likely to be small, and
failures makes it difficult to judge which method can be seen as a small margin of safety reflecting
might provide the best basis for ship design or our uncertain knowledge of the anatomy of final
approval calculations. Back-breaking hull failure failure of ships’ hulls in bending.
involves so much buckled and fractured metal, that
it must be doubted whether the reality can ever be
modelled with precision. It might therefore be pru- ACKNOWLEDGEMENTS
dent to take as the design “critical limit state” the
longitudinal bending moment at which the tension The present study was undertaken at The Lloyd’s
“flange” of the hull structure first reaches the yield Register Educational Trust (The LRET) Research
stress after the compressed “flange” has reached its Centre of Excellence at Pusan National University,
201
202
ABSTRACT: During the past decades a number of studies have been conducted in terms of evaluation
of the nonlinear buckling characteristics of curved plates. However any explicit expression or formula
for calculating buckling/ultimate strength of curved plates is not available yet. Therefore it is impera-
tive to develop practical formulas for estimating buckling/ultimate strength of curved plates. The aim
of this study is to analyze nonlinear buckling characteristics of curved plates. A series of FEM analyses
are performed on curved plates varying several parameters such as flank angle (curvature), plate thick-
ness, loading conditions, etc. According to the various conditions applied to the curved plates, buckling
strength and ultimate strength are calculated. And also the stress-strain curve is drawn for each set of
applied conditions from the numerical calculations. It is shown that buckling/ultimate strength formula
developed for a curved plate can give a reasonable estimate of strength for curved plate, when the newly
defined curvature correction parameter considering the increase of the buckling strength due to curvature
is applied. It is our hope that the obtained buckling/ultimate strength characteristics would be used as
practical design guide for estimating the nonlinear buckling strength of curved plates in the field of ships
and offshore structures.
203
204
3 NUMERICAL SIMULATION
AND RESULTS
(a) Longi.compression (b) Trans.compression
3.1 Eigen buckling mode
A series of eigen-value buckling analysis is carried Figure 2. Typical buckling modes obtained by elastic
out to evaluate buckling strength and to examine eigen-value analysis.
the significant buckling mode. The latter result is
used to produce initial deflection. Figure 2 shows
typical buckling modes varying flank angle and
slenderness ratio. When a flank angle is zero, that
is in case of a flat thinner plate, the plate buckles
into four half-waves in the longitudinal direction.
However, as the flank angle increases to 5 degrees,
the buckling mode changes so that the deflection
flattened out towards loading direction and deflec-
tion locally grow at the loading edges as shown in
Fig. 2(a)-flank angle 5 degrees, 10 mm. When a
flank angle is greater than these values, buckling
takes place in an irregular mode as indicated in
Fig. 2(a). For larger thickness of plate, five or four
Figure 3. Elastic buckling mode varying flank angle.
half-waves mode appears in the loading direction
as shown in Fig. 2(a). In general, buckling mode of
plate under transverse compressive loading takes
place one half-wave, however, plate with curvature the outward deflection develops from the begin-
appears in the circumferential direction at the local ning of compressive loading because of Poisson’s
parts as shown in Fig. 2(b)-flank angle 30 degrees, effect. It is noticed that the curved plates can have
20 mm. This is considered as a part of shell buck- a primary buckling mode that is different from the
ling of a cylinder wall in a so-called diamond buck- buckling mode of flat plate and that the buckling
ling mode [2, 3]. strength is generally larger than that of the flat
plate.
3.2 Fundamental buckling mechanism
3.3 Benchmark study
Figure 3 shows the corresponding elastic buck-
ling modes varying flank angles. The solid red line Figure 4 shows the results of critical buckling
represents compressive load, and solid blue line is strength for curved plate which represent compari-
tension. When the flank angle is less than or equal son between class rule and finite element analy-
to 2 degrees, the buckling takes place with three sis using commercial code ABAQUS. As shown
longitudinal half-waves as in the case of a rectan- in Fig. 4, a comparison of two results is made in
gular flat plate having the same aspect ratio. With which the critical buckling strength from both
further increase in the flank angles, the buckling FEM and class rule calculation is relatively unani-
mode changes in one longitudinal half wave with mous in case of thicker plate. But in case of thinner
additional swelled component of deflection near plate, the result of critical buckling strength shows
the transverse edges. In the case of curved plates, significant errors because curvature reduction
205
0.6 RINA
0.6
σxu/σy
KR
A C
0.4 0.4
B
0.2 0.2
A : Secondary Buckling
B : After 2nd Buckling
C : Ultimate Strength
0 0
0 0.2 0.4 0.6 0.8 1 0 1 2 3 4
FEA(ABAQUS, σcr /σy) ε xu /ε y
Figure 4. Verification results of buckling strength for Figure 5. The relationships of between average stress
curved plate under axial compression. and average stain of plate with/without curvature under
longi.compression.
factor used in class’s equations currently is not in
the reflected buckling phenomenon.
206
0.6 0.6
σyu/σy
σXU/σY
0.4 0.4
Longitudinal
Compression
5 Deg.
0.2
0.2 10 Deg.
20 Deg.
30 Deg.
0
0
0 1 2 3 4 5 0 1 2 3 4
ε yu /εy Plate Slenderness Ratio, β
length of the plate as 1,000 mm and 3,700 mm, b = 1000 mm, a/b = 3.7, ω0 = 0.05β2t
16t
20t
respectively. The flank angle is taken as 5, 10, 20 and Frank Angle = 10 deg.
24t
30t
0.6
30 degrees. The calculated ultimate strength is plot-
σYU/σY
35t
207
0.4
4 DEVELOPMENT OF FORMULATION
Correlation : 0.991
Standard Deviation : 0.163
• Transverse compressive load
0.6
2
π 2E ⎛ t ⎞
σE =k ⎜ ⎟ × CB (6)
12(1 − v 2 ) ⎝ b ⎠
TC
0.4
β
0.2
CB ( 1 v2 × θ × ) σY / E
× CR (7)
−1.514
⎛ R⎞
CR = 0.003 ⎜ ⎟ (8)
0
0 0.2 0.4 0.6 0.8 1 ⎝t⎠
FEA (σxcr /σy)
208
eral form as the one due to Faulkner to Guedes 0.6 Standard Deviation : 0.08
=⎜ − 2 ⎟ × CF
Correlation : 0.998
Standard Deviation : 0.133 CP (17)
0.6
σY ⎝ β β ⎠
Ca Cb
CF = + + Cc (18)
0.4 β2 β
∴ Range ( ) = 10 ~ 30
0.2
⎡ ⎛ b⎞
2
⎛ b⎞ ⎤
⎢ Ca = 3.434 ⎜ ⎟ − 1.989 ⎜ ⎟ + 0.646 ⎥
⎢ ⎝ R ⎠ ⎝ R ⎠ ⎥
0
0 0.2 0.4 0.6 0.8 1 ⎢ 2 ⎥ (19)
⎢C = −4.138 ⎛ ⎞
b ⎛ ⎞
b
FEA( σxcr /σy)
1 ⎜ ⎟ + 1.934934 ⎜ ⎟ − 1.023⎥
⎢ b
⎝ R⎠ ⎝ R⎠ ⎥
Figure 13. Correlation of the empirical formula with
⎢ 2
⎥
⎢ ⎛ b⎞ ⎛ b⎞ ⎥
ultimate strength obtained by FEM of curved plate ⎢ Cc = 1.001
001⎜ ⎟ − 0.181⎜ ⎟ + 1.382 ⎥
⎝ R⎠ ⎝ R⎠
under longi. compressive load. ⎢⎣ ⎥⎦
209
Formula (σXU/σY)
ω0 = 0.05β2 t, P = 0.2 MPa
0
Ca Cb
CF = + + Cc (23) 0 0.2 0.4 0.6 0.8 1
β2 β FEA( σXU/σY)
⎡ ⎛ b⎞
2
⎛ b⎞ ⎤ Figure 16. Correlation of the empirical formula with
⎢ Ca = 2.596 ⎜ ⎟ − 1.712 ⎜ ⎟ + 0.415 ⎥
⎢ ⎝ R ⎠ ⎝ R ⎠ ⎥
ultimate strength obtained by FEM of curved plate
under transverse compression and lateral pressure.
⎢ 2 ⎥
⎢C = −2.096 ⎛ b ⎞ ⎛ b ⎞
⎜ ⎟ 9 ⎜ ⎟ − 0.136 ⎥
+ 0.929 (24)
⎢ b
⎝ R⎠ ⎝ R⎠ ⎥
⎢ 2
⎥ formula plotted against slenderness ratio is checked
⎢ ⎛ b⎞ ⎛ b⎞ ⎥
⎢ Cc = 1.009
009 ⎜ ⎟ − 0.724 ⎜ ⎟ + 0.322 ⎥
⎝ R⎠ ⎝ R⎠
by a comparison with finite element solutions as
⎢⎣ ⎥⎦ shown in Figures 15 and 16. It is noted that cor-
relation ratio and standard deviation of the error
CP = −0.196β 2 + 0.893β + 0.314 (25) in the empirical formula against FEA is 0.99 and
0.27, of the curved plate subjected to combined
longitudinal compression and lateral pressure,
The coefficient CP assumed as function of slen- respectively.
derness ratio, and the other factors (Ca, Cb and Cc) The normalized ultimate strength by FEA is
represents effect of curvature with the change in plotted against the normalized ultimate strength
the thickness of plate. The accuracy of the present by empirical formula of curved plate under com-
bined transverse compression and lateral pressure
as shown in Figure 16. Good agreements are also
1 observed in both results.
Model1
LC+LP
FEM-Formula
Linear
0.8 5 CONCLUSION
b = 1000 mm, a/b = 3.7
Formula (σXU/σY)
ω0 = 0.05β2 t, P = 0.2 MPa The objective of the present paper is to clarify and
0.6 examine the fundamental buckling/plastic collapse
behaviour and ultimate strength of cylindrically
curved plate under a variety of loading conditions
0.4 (compression and combined compression and lat-
eral pressure).
On the basis of the calculated results, the effects
0.2
Correlation : 0.995
of curvature (R), slenderness ratio and loading
Standard Deviation : 0.271 effect on the buckling and ultimate strength have
0
been discussed. A simple formulation is developed
as an efficient method to predict the critical buck-
0 0.2 0.4 0.6 0.8 1
ling strength and ultimate strength. The following
FEA( σXU/σY) points can be concluded:
Figure 15. Correlation of the empirical formula with 1. Cylindrically curved plate under longitudinal
ultimate strength obtained by FEM of curved plate compression takes place the buckling mode
under longitudinal compression and lateral pressure. with several half waves for loading direction
210
ACKNOWLEDGMENTS
211
ABSTRACT: The aim of this study is to investigate the effect of non-linear randomly distributed
nonuniform corrosion on the ultimate strength of unstiffened rectangular plate subjected to axial
compressive loading. A series of 570 plate surface geometries are generated by Monte Carlo simulation
for different degree of corrosion, location and ages and nonlinear finite element analyses are carried out,
using a commercial finite element code. Based on a regression analysis, empirical formulae to predict
strength reduction because of corrosion have been derived demonstrating a good accuracy.
213
Corrosion depth, mm
The third approach, which is the one that is
adopted here, is to consider that a model should 1.4
provide the trend that is derived from for the domi- 1.2
nating mechanism and then it should be fit to the
field data. Guedes Soares and Garbatov, (1999) pro- 1.0
posed a model for the non-linear time-dependent 0.8
function of general corrosion wastage. This time- 0.6
dependent model separates corrosion degrada-
tion into three phases. In the first one there is no 0.4
corrosion because the protection of the metal sur- 0.2
face works properly. The second phase is initiated
0.0
when the corrosion protection is damaged and
0 4 8 12 16 20 24 28 32 36
corresponds really to the start of corrosion, which
decreases the thickness of the plate. The third t, years
phase corresponds to a stop in the corrosion proc-
Figure 1. Corrosion depth of deck plates of ballast
ess and the corrosion rate becomes zero.
tanks.
The model is based on the solution of a differen-
tial equation of the corrosion wastage:
1
⎧ ⎛ −t −τ c ⎞
⎪d ⎜ e τt
⎟ , t ≥ τc 0.9
d (t ) = ⎨ ∞ ⎜ (1)
⎟⎠
⎪ ⎝
0.8
Standard Deviation, mm
214
Figure 3. Modelled plate surface with an average thick- Figure 5. Modelled plate surface with an average thick-
ness of 9.8 mm at the 15th year. ness of 9.2 mm at the 25th year.
215
Ux Uy Uz rotx
y=0 F C C C
y=L F F C C
x=0 F F C F
x=b F F C F
x = b/2 and y = 0 C C C C
w0 h0β20 (4)
⎛ b⎞ σ
y
β0 = ⎜ ⎟ (5)
⎜⎝ h ⎟⎠ E
0
216
0.7
217
⎧ σ u ,0
⎪ , t ≤ τC
σ u (t ) ⎪ σ y (8)
=⎨ ⎧ ⎛ ⎞ t⎫
n
σy ⎪ σ u,0 exp ⎪⎨− t − τC ⎪⎬ , t ≥ τC
⎪ σy ⎜ ⎟
⎩ ⎪⎩ ⎝ τt ,U ⎠ ⎪⎭
218
REFERENCES
219
220
ABSTRACT: Short stiffened panels are simulated and compared with test results under axial compression
until collapse to investigate the influences of the stiffener’s geometry. The stiffened panels with different
combinations of mechanical material properties and geometric configurations are considered. Four type
stiffeners are made of mild or high tensile steel for bar stiffeners and mild steel for ‘L’ and ‘U’ stiffeners.
To produce adequate boundary conditions at the loaded edges in the experiments, three bays stiffened
panel are used in the test and in the FEM analysis. The influence of the stiffener’s geometry on the ulti-
mate strength of the stiffened panels under compression is analyzed.
221
3 EXPERIMENTAL WORK
222
Figure 3. The FE model of ‘U’ Series stiffened panel. Figure 4. The coordinate and model of stiffened panel.
223
224
Stress (Mpa)
200 300
150
200
100
100
50
0 0
0 1 2 3 4 5 6 0 5 10
3 3
dL/L(10 ) dL/L(10 )
Figure 5. Average stress-shortening curve of experi- Figure 8. Average stress-shortening curve of experi-
ment and FEM for FS3-B3. ment and FEM for BS3-A3.
350
200
Stress (Mpa)
150 300
100
200
50
100
0
0 2 4 6 8 10 12
3
dL/L(10 ) 0
0 5 10
3
Figure 6. Average stress-shortening curve of FEM for dL/L(10 )
FS3-A3.
Figure 9. Average stress-shortening curve of experi-
ment and FEM for BS3-B3.
225
400
Stress (Mpa)
300
200
100
0
0 5 10
3
Figure 10. Von Mises stress distributions at the ultimate dL/L(10 )
limit state for BS3-A3.
Figure 13. Average stress-shortening curve of experi-
ment and FEM for LS3-B3.
400
test. The main reason may be the residual stresses
Stress (Mpa)
300
in stiffened panel, and this aspect is not consid-
ered in FEM analysis. The ultimate strength of LS
200 series specimen is bigger than the BS series.
The buckling of plate occurs in the middle span
100 of the panels, and then induces stiffener failure
in the ‘LS’ series specimen from Figure 14 and
0 Figure 15. The series of panels reinforced with L
0 5 10 stiffeners allowed obtaining results as expected
dL/L(10 )
3
that the ‘L’ series have good capability to avoid
buckling.
Figure 12. Average stress-shortening curve of experi-
ment and FEM for LS3-A3.
5.4 US series results
Figure 16 and Figure 17 show the average stress
analysis for LS series specimen. The slope of the and shortening curve of the test and the FEM
average stress-shortening curve is different between analysis for the ‘US’ series specimen.
FEM analysis and the test. The stiffness of the The slopes of the average stress-shortening
FEM analysis is bigger than the stiffness of the curve are also different between FEM analysis and
226
200 6 CONCLUSIONS
200
The ultimate strength of ‘U’ series stiffeners
drops down very quickly in both test and FEM
analysis. The ultimate strength stress distribution
100 pictures of the ‘U’ series show that buckling occur
in the flange occur very early. ‘U’ series stiffeners
have high ultimate strength value, but it’s necessary
0 to further research to determine their suitable
0 2 4 6 8 10 dimension.
3
dL/L(10 )
227
228
ABSTRACT: A numerical study is conducted to analyze the influences of the stiffener’s geometry and
panel boundary conditions on the ultimate strength of stiffened panels under compression. The four types
of short stiffened panels analyzed are made of mild or high tensile steel and have bar, ‘L’ and ‘U’ stiffeners.
To understand the effect of finite element modeling on the ultimate strength of the stiffened panels, four
types of models with different geometry are investigated in the FE analysis. Moreover, different boundary
conditions for the same FE model are also investigated. From the results of the FE analysis, two of the
four models studied can produce adequate boundary conditions at the loaded edges. The stiffened panels
with different combinations of mechanical material properties and geometric configurations are consid-
ered. The initial geometric imperfection affects the collapse behaviour of stiffened panel and is analyzed
in FE simulation.
229
Stiffener
– Dim (mm) σY
FS3-I3 I 20 × 4 690
FS3-I21 I 20 × 4 690
FS3-I22 I 20 × 4 690
FS3-I1 I 20 × 4 690
Figure 1. Geometry of 1/2 + 1 + 1/2 and 3 bays. BS3-I3 I 30 × 8 343
BS3-I21 I 30 × 8 343
BS3-I22 I 30 × 8 343
BS3-I1 I 30 × 8 343
LS3-I3 L38 × 19 × 4 296
LS3-I21 L38 × 19 × 4 296
LS3-I22 L38 × 19 × 4 296
LS3-I1 L38 × 19 × 4 296
US3-I3 U (40 × 150 × 40) × 2 200
US3-I21 U (40 × 150 × 40) × 2 200
US3-I22 U (40 × 150 × 40) × 2 200
US3-I1 U (40 × 150 × 40) × 2 200
230
and higher strength steel (ABS 2006). (b) The FE model of ‘L’ Series
The following are the material properties: 1
ELEMENTS
FEB 4 2010
1
ELEMENTS
Z X Z X
model shape (also after deformation). Therefore, (c) The FE model of ‘U’ Series
a balance between required accuracy and efforts
is needed. It is considered that the element size Figure 4. The FE model of continuous stiffened panel.
2 bays
3 bays 2 bays (1 + 1) 1bay
Boundary condition C1 C2 C3 C4 C5 C6 C7 C8 C9
Note: Different model and boundary condition correspond to different location of stiffener.
231
232
233
Stress (MPa)
400 450
350 400
300 350
250 300
100 200 300 400 500 600 700 800 100 200 300 400 500 600 700 800
Width (mm) Width (mm)
Figure 9. Ultimate strength of three bays with C2. Figure 11. Ultimate strength for three bays with C3.
400
C2/C1
1.5
350
1.0
300
100 200 300 400 500 600 700 800
0.5
100 200 300 400 500 600 700 800 Width (mm)
Width (mm)
Figure 12. Ultimate strength for three bays with C4.
Figure 10. C2 divided by C1 for three bays.
Table 3. The ultimate strength of FEM analysis
for three bays (MPa).
either C3 or C4, as shown in Figure 11, Figure 12
and Table 3. That illustrates that the width wouldn’t Case C1 C2 C3 C4
affect the ultimate strength of the stiffened panel in
this circumstance. So the ‘A’ series stiffened panel FS3-A 308 302 326 325
with C3 and C4 for three bays can get reasonable FS3-B 303 322 355 327
result. FS3-C 277 319 356 332
The following calculations will focus on the FS3-D 264 334 356 332
geometry of the model and the ‘A’ and ‘B’ series is FS3-E 220 340 356 332
further studied. Table 3 shows the ultimate strength BS3-A 399 404 450 441
of the stiffened panel for ‘A’ and ‘B’ series. The BS3-B 391 414 448 443
results are almost the same between the ‘A’ and ‘B’ BS3-C 348 412 448 442
series. Due to symmetry the boundary condition BS3-D 366 413 448 442
is applied on the AB and A1B1 edges shown in BS3-E 358 447 451 442
Table 3. The mean value of ratio between three LS3-A 460 441 495 487
LS3-B 486 481 492 488
and two bays with C5 /C3 and C6 /C4 is 1.13 and
LS3-C 479 486 492 490
1.16, namely the result of 1/2 + 1 + 1/2 bays big-
LS3-D 439 487 492 490
ger than three bays with same boundary condition.
LS3-E 422 489 486 490
The biggest ratio is 1.43 and 1.46 for clamp and
US3-A 356 364 416 415
simply supported boundary condition. Figure 13
US3-B 366 415 415 414
presents a comparison of the ultimate strength US3-C 325 412 414 416
for one/two bays panel with clamped and sim- US3-D 303 412 416 414
ply supported boundary conditions at the end US3-E 197 410 411 413
edges of the panel. The ultimate strength of the
234
edges of the panels affects not only the ultimate SMN =.652E+07
SMX =.690E+09
MN
3Bay-C1
600 3Bay-C2
3Bay-C3
3Bay-C4
550 2Bay1-C5
2Bay1-C6
2Bay2-C7
500 2Bay2-C8
Stress (MPa)
1Bay-C9
1
NODAL SOLUTION
Y
STEP=1 APR 5 2010
SUB =8 MX 17:43:55
TIME=.390886 Z X
SEQV (AVG) MN
DMX =.008786 MX
SMN =.943E+07
SMX =.690E+09
BS3-B3
.851E+08 .236E+09 .388E+09 .539E+09 .690E+09
strengths but also the collapse mode indicated in
Figure 15 and Figure 16. The collapse modes of
Figure 14. Von Mises stress distributions of the stiffened three and 1/2 + 1 + 1/2 bays are similar, but they are
panel at the ultimate limit state for 3bay-C1. both different from the 1 + 1 bays.
235
MN
MN
MX
.285E+07 .156E+09 .308E+09 .461E+09 .614E+09
.792E+08 .232E+09 .385E+09 .537E+09 .690E+09
BS3-B2
.150E+07 .154E+09 .307E+09 .460E+09 .613E+09
.780E+08 .231E+09 .384E+09 .537E+09 .690E+09
BS3-B2
MN
MN
MX
5 CONCLUSIONS
.601E+07 .158E+09 .310E+09 .462E+09 .614E+09
.820E+08 .234E+09 .386E+09 .538E+09 .690E+09
BS3-B2
Nine geometric configurations of stiffened panels
are investigated, including 3 bays, two (1/2 + 1 + 1/2)
Figure 20. Von Mises stress distributions of the stiffened
panel at the ultimate limit state for 2bay2-C7. bays, 1 + 1 bays and 3 bay with different boundary
conditions. From the results of the 120 calculations,
some conclusions can be obtained. The frame on
The unsymmetrical collapse modes occur for the boundaries of the plate is also important in
1 + 1 bays in Figures 20 and 21. This phenomenon the simulation and should be included in the FEM
will increase the uncertainty of the result. So 3 and analysis when the width of panel or frame is not
1/2 + 1 + 1/2 bays stiffened panels are more robust strong enough, as resulted from the comparisons
than 1 + 1 bays in this circumstance. between C1 and C2.
236
REFERENCES
237
ABSTRACT: A new approach to predicting the dynamic collapse behavior of a ship’s hull girder in
waves is proposed. The work is to evaluate the consequence of the collapse, which is important in evaluat-
ing the associated risk of the hull girder collapse. The progressive collapse analysis of the ship’s hull could
be followed by a hydro-elastoplastic response analysis. In this research, the whole ship hull is modeled as a
two-rigid-bodies system connected by a rotational spring, which represents the nonlinear relation between
the displacement and the moment. Nonlinear strip theory is used to solve the force equilibrium of the two
bodies. The approach is validated against tank tests, which show that the collapse increases rapidly after
the ultimate strength is reached, and the plastic deformation grows until unloading starts and the bending
rigidity recovers. It is concluded that the fundamental dynamic collapse behavior can be followed by the
numerical method.
239
240
⎪ ⎪ ⎢ ⎥⎪ ⎪
u1 u2 (4) = ⎨ M1w ⎬ − ⎢ m21 1 1
m22 0 ⎥ ⎨θ1 ⎬
⎪ M ⎪ ⎢ ⎥⎪ ⎪
⎩ 2w ⎭
⎣⎢ m21
2
0 m222 ⎥⎦ ⎩θ2 ⎭
where,
⎡C11
1
+ C111
2 1
C12 2 ⎤
C112 ⎧u ⎫
⎢ ⎥⎪ 2⎪
C222 0 ⎥ ⎨θ1 ⎬
{} {}
u1 u2
− ⎢ C221 1 1
{U1} {U 2 } = ⎢ 2 2 ⎥ ⎪θ ⎪
1 2 ⎢⎣ C21 0 C222 ⎥⎦ ⎩ 2 ⎭
⎡ m1 m12 ⎤
1
⎡ m2 m2 ⎤ ⎡ K11 K111 K12
1 2 1
K112 ⎤ ⎧u2 ⎫
2
[M1 ] = ⎢ 111 ⎥ [M 2 ] = ⎢ 112 122 ⎥ ⎢
1
⎥⎪ ⎪
⎣ m21 m22 ⎦ ⎣ m21 m22 ⎦
1
− ⎢ K 21 1 1
K 222 0 ⎥ ⎨θ1 ⎬
⎡C111 C12 ⎤
1
⎡C112 C122 ⎤ ⎢ 2 ⎥ ⎪θ ⎪
[C1 ] = ⎢ 1 ⎥ [C2 ] = ⎢ 2 2 ⎥ ⎢⎣ K 21
2
0 K 222 ⎥⎦ ⎩ 2 ⎭ (5)
⎣C21 C 22 ⎦ ⎣C21 C22 ⎦
1
⎡K1 K112 ⎤
1
⎡ K112 K1122 ⎤
[K1 ] = ⎢ 111 ⎥ [K 2 ] = ⎢ 2 ⎥ Newmark β method is applied to perform the
⎣ K 21 K 22 ⎦ ⎣ K 21 K 22 ⎦
1 2
time-domain analysis in Eq. (5). The analysis
⎧F f ⎫ ⎧F f ⎫ results will be given later.
{F1 } = ⎨ 1w ⎬ {F2 } = ⎨ 2 w ⎬
⎩M1w M⎭ ⎩M 2 w M ⎭
3 EXPERIMENT
{U} = displacement vector for the respective
bodies; [M] = Mass matrixes, which consist of
3.1 Model requirements and design idea
hull mass and added mass for the respective
bodies; [C] = damping matrixes, which includ- A box type ship is selected as a generic ship for
ing structural damping and wave induced damp- simplicity and clarity, and also for validation by the
ing; [K] = restoring force matrixes; {F} = external experiments. The box-type ship are length 300 m,
forces and moments. Subscript 1, 2 in equations breadth 40 m, depth 30 m and draft 10 m. Tank
(2) and (3) correspond to fore part and aft part, tests using a scaled model need to be performed for
respectively. In the mass matrixes, damping validation of the numerical method addressed in
matrix and restoring force matrix, the former the previous section.
1, 2 of the subscript corresponding to force and The scaled model should follow the law of simil-
moment, respectively. The later 1, 2 correspond- itude in terms of strength as well as geometry and
ing to vertical displacement and rotational angle, stiffness. Its scale ratio is assumed to be 1/100 as
respectively. conventional scaled models for tank tests. The main
For the evaluation of the external forces and particulars of model are length 3 m, breadth 0.4 m,
moments, a nonlinear strip theory proposed by depth 0.3 m and draught 0.1 m. It is completely a
Fujino (1984) is employed. box-ship consisting of two rigid box bodies.
From equations (1)–(4), a system of equa- The requirements for the scaled model are, a) to
tions of motion may be written in incremental realize the strength model of the hull girder col-
form as, lapse behavior in loading and unloading as shown
241
242
243
244
Figure 13. Time histories of rotational angle for the Figure 15. Time histories of rotational angle with
three cases. different amplitude.
245
246
247
Y.J. Lee
Department of Engineering Science and Ocean Engineering, National Taiwan University,
Taipei, Taiwan, R.O.C.
ABSTRACT: The objective of this paper is to analyze the structural dynamic buckling strength of
containership bow structures subjected to impact force using finite element method. In this study, a con-
tainership bow structure designed and constructed by CSBC Corporation, Taiwan is modeled and solved
by geometric nonlinear and transient dynamic programs; and the wave impact pressures are obtained
through applying the semi-empirical formula established by Lloyd’s Register. Furthermore, one-time
thickness deformation method and buckling failure method are employed here to determine the initial
buckling load of bow structure. The results show that the impact force integration is the dominant factor
for structure buckling strength assessment. This conclusion offers the designer an effective and efficient
way to predict the structure dynamic buckling strength; moreover, it is can help the designer realize the
moment when the structure happens buckling under any different impact force case.
249
250
Particular Value
251
4 SUMMARY
Pbf 0.5( K bfVbf2 K rv H rvVrv2 ) kN / m 2 (3) Figure 10. Load-transverse deflection curve.
252
253
254
255
M.R. Zareei
Chabahar Maritime University (CMU), Chabahar, Iran
ABSTRACT: In this paper, the results of an investigation into the post-buckling behavior of
high-strength aluminum alloy stiffened plates with Y-stiffener subjected to axial compression load using
non-linear finite element approach is presented. Both material and geometric non-linearities have been
taken into account. The principal variables studied are the geometries of stiffened plates. The influence
of Dimensions on the post-buckling behavior and ultimate strength of such stiffened plates has been
investigated in details.
257
258
Y-stiffener
Model ID L BT t1 B1 B2 H1 t2 H2 B3 t3 b 1 σ Ult σ Y
259
Figure 7. Test set-up, schematic view of simply supported panel in test rig.
260
f02(MPa)
(b)
Figure 10. Engineering stress-strain curves for panel
Cross section (mm) materials.
t1 t2 t3 s2 h
(c)
Imperfection (mm)
261
3.2 Discussion
From results of stiffened plates with Y-stiffener,
ultimate strength of panels increases with increas-
ing of bending rigidity. Also, we can see that, pan-
els with lower bending rigidity have the collapse
mode in a buckling mode. In these models with
increase in bending rigidity, the collapse mode
changes from buckling mode to quasi-clamped
mode. In these models, Stiffened plates with high
rigidity result in increasing post-ultimate strength.
The creations of the plastic hinges at the ends of
the stiffened plate are a common feature of the col-
lapse mode. Also the strength reserve after collapse
is significant. At the final stage of calculations, it
is observed that unloading (stress removal) takes
place in some part of the stiffened plate while in the
Figure 12. Load shortening curve for O model. remaining parts localized plastic deformations are
accumulated. We can see this phenomenon at mid-
dle of model 4 or quarter of model 6 in Figure 15.
Figure 13. Load shortening curve for P model. Figure 14. Average stress-average strain relationships.
262
263
264
ABSTRACT: Aim is to develop a design method concerning high-cycle fatigue of arc-welded joints
in high-speed aluminium ships. Accordingly, a Linear Elastic Fracture Mechanics (LEFM) based
non-similitude two-stage (micro- and macro) crack propagation model is proposed and elaborated into
a fatigue master curve formulation. First, an analytical, parametric expression of the micro-crack propa-
gation dominating Stress Concentration Factor (SCF) related weld notch stress distribution along the
(virtual) crack has been developed for welded basic joints. This expression is used to correct the macro-crack
propagation governing LEFM parameter, the Stress Intensity Factor (SIF). Subsequently, a crack prop-
agation model is developed and successfully verified using aluminium Centre Cracked Tension (CCT)
specimen experimental data. Finally, a fatigue master curve formulation is obtained and satisfactorily
validated using fatigue test data of aluminium welded basic joints.
267
2 WELD NOTCH STRESS DISTRIBUTIONS The minus sign in (Equation 2) is the result of
the coordinate system used: counter-clockwise
Analytical, parametric weld notch stress formulations— positive.
with its characteristic singularity—are developed The bending stress ratio R (Equation 3) describes
for basic welded joints, which take the involved the amount of bending relative to σs. A positive R
geometry parameters into account, e.g. the base indicates that the notch stress distribution will be
plate thickness tb, the cross plate thickness tc, the non-monotonic; a monotonic weld notch stress
weld length and height lw and hw and the notch distribution appears in case of a negative R.
radius ρ. σb
A weld notch stress distribution σw (r/tb) is R sg (M b ) ⋅ (3)
assumed to be a linear superposition of an equi- σs
librium equivalent stress part (the linear far field For Williams’ asymptotic solution (Equation 4),
stress) and a self-equilibrating stress part (consist- the normal stress component σθθ with particular
ing of a non-linear notch stress: Williams’ asymp- stress angle β has been used (Williams 1952) which
totic solution and a linear weld geometry induced denotes for arc-welded geometries:
bending contribution), as shown in (Figure 2) for
a Full Penetration (FP), Double Sided (DS) cruci- λ s −1
⎛ r⎞
form joint. σ θθ μ s λ s ⋅ (λ + ) ⋅ ⎜ ⎟
The far field stress distribution σf (Equation 1) is ⎝ tb ⎠
the linear stress distribution in the cross-section at ⋅ ⎣cos {(λ
⎡ ) } χs ⋅ cos {(λ − ) ⋅ β }⎤⎦
the weld toe. It is consistent with the far field stress λ a −1
definition in the fracture mechanics context and ⎛ r⎞
characterised by the structural stress amplitude + μa ⋅ λ a ⋅ ( + )⋅⎜ ⎟
⎝ tb ⎠
σs as already introduced by (Niemi et al., 2004) and
the bending stress ratio R. ⋅ ⎡⎣si {( } {
+ ) ⋅ β − χa ⋅ sin (λ a − ) ⋅ }⎤⎦ (4)
⎧ ⎛ r⎞ ⎫ t
σ gb σ s ⋅ C ggb ⋅ ⎨4 ⋅ ⎜ ⎟ − 1⎬ for 0 < r ≤ b (5)
⎩ ⎝ tb ⎠ ⎭ 2
Note that this stress component is not part of
Figure 2. Weld notch stress distribution. the equilibrating far field stress.
268
K σs a ⋅Ym
⋅ ⎡⎣sgn ( Fm ) Ygm
tc/t b = 1.0; hw/lw = 1.0; lw/t b = 1.0; ρ /t b = 0.0
0.0
0.3
The geometry factor Yg includes a linear super-
position of the membrane and bending contri-
0.4
bution, Ygm and Ygb, obtained from (Tada et al.,
r/t b [ - ]
269
da/dn
3.0
2.5
II
10-6
2.0
1.5
1.0
0.5 10-7
0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
a/tb [ - ] I
-8
10
Figure 5. SIF correction factors, monotonic case. 100 K th 101 102 K c 10
K
4.5
3.0 Yg
region III, the number of cycles consumed in this
Ym
2.5
YmYg
region is quite small and it is considered to be not
2.0 FEM
that important from fatigue (design) point of view.
1.5
Hence, a well-known Paris based two-stage, region
I and II, crack propagation model is introduced:
1.0
γ
⎧ ⎛ ΔKth ⎞ ⎫
0.5
da
⎟ ⎬ ⋅( K)
m
= C ⋅ ⎨1 − ⎜ (8)
⎩ ⎝ ΔK ⎠ ⎭
0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 dn
a/tb [ - ]
270
0.1
0.2
0.3
b 0.4
r/t b' [ - ]
0.5
ag 0.6
ag r sc, mc , bc m 0.7 crack notch stress σ c /σ sc
a 0.8
far field stress σ f /σ sc
FE result
b t b'
0.9
1.0
-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
σc /σsc, σf /σsc [ - ]
Figure 8. Centre Cracked Tension (CCT) specimen. Figure 9. CCT specimen crack notch stress distribution.
271
( Yn/Yl )
3.5
FEM
factor (Yn/Yl) has been moved to the left-hand side
3.0
of this equation to achieve single-slope, region II
behaviour.
2.5
1.0 10-3
40 10 -5-
rolling direction
0.5 -4
10
(da/dn) [mm/cycle]
0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
10-5
a/t b' [ - ]
10-6
Figure 10. CCT specimen correction factors.
base material; Rs = -1.0
10-7
base material; Rs = 0.0
⎝ π⎠
10
(ag + a)
2
ag −r 2
-3
10
(da/dn) . (Yl / Yn) [mm/cycle]
-4
10
⎛ 2⎞ ⎛ ag ⎞
Yl = 1 − ⎜ ⎟ ⋅ arcsin ⎜ ⎟ 10
-5
⎝ π⎠ ⎝ ag + a ⎠
-6
10
-5
10
272
-3
Δσnom
10
(da/dn).. (Yl / Yn) [mm/cycle]
-4
10 2
10
-5
10
-6
10 1
10
T-joint
-7 weld material; Rs = -1.0
10 Butt joint
weld material; Rs = 0.0 Cruciform joint
Longitudinal stiffener joint
-8
10 1 2 3 4 0
10 10 10 10 10 3 4 5 6 7 8
ΔK . (Yl / Yn) 2/3 [MPa √ mm] 10 10 10 10 10 10
N
273
1083–1092.
2-m
274
275
ABSTRACT: The present work starts from the idea to improve fatigue strength of an actual struc-
tural detail where welded joints are subject to proportional biaxial variable-amplitude loadings leading
to elastic-plastic strains. After a brief discussion on how to manage low-cycle fatigue in ship structural
design, the case study is presented. Basing on outcomes drawn from the analyses performed on the con-
sidered structural detail, a critical examination is carried out on the procedures applied to check fatigue
strength. A comparison of fatigue damages obtained from application of linear elastic approach and
elastic-plastic approach is outlined in order to analyze advantages and drawbacks of the two approaches
and their potential to be implemented in a design-oriented practical assessment procedure. Investigations
point out that nonlinear approach may be a valid alternative to linear one only if designer can control
safety factors relating to fatigue capability and has available deeper information on material behaviour
and load history experienced by the structure.
277
278
279
280
281
282
283
284
DNL(P)−B(PM) nc
e
Nonlinear damage DNL
50 % 0
2
σ
ere
DNL(P)−B(HAZ)
σm /Δ
e f
30 %
r
7%
1.5
16, %
25
1.5
1 σm/Δσ0 % = 16.7 σm/Δσ0 % = 0
0% 1
0.5
0.5
0
0 100 200 300 400 500 0
Δσ0, MPa 0 0.5 1 1.5 2 2.5
Linear damage DL
Figure 9. Comparison between fatigue damage DNL(P)
calculated for given values of the mean stress, and linear Figure 10. Damage comparison based on different data
elastic based fatigue damage DL(Δσ). for the material parameters.
285
Fatigue damage D
plane strain rule (Eq. 5)
are not very tangible, as far as the parent material
σm effect (Eq. 3)
is considered. On the other hand, changes appear
more evident if data are referred to the material 1 σm/Δσ0 % = 16.7
heat-affected by welding procedure. Moreover,
the general trend of damage parameter P is con-
firmed to have a greater sensitivity when calcula-
σm/Δσ0 % = 0
tions are carried out at lower values of σm /Δσ0,
as could also be deduced by the rate by which
DNL(P) damage changes as function of σm /Δσ0 in 0.1
Figure 9. 100 200 300 400 500 600
A further analysis has been performed about the
Δσ0, MPa
validity of the basic hypotheses made in defining
the methodology. First of all, the conversion from Figure 12. Fatigue damage DNL(P) calculated under dif-
the elastically computed to the actual elastic-plastic ferent hypotheses within the elastic-plastic procedure.
stress-strain field has been carried out under both
the hypothesis of plane stress (Equation 4) and
plane strain (Equation 5). Figure 11 shows the con- σm to Δσ0 ratios. They all agree in confirm that
version from the calculated linear stress-strain field formulation of Equation 3 appears ineffective to
pair (point K) to the elastic plastic pair according treat mean stress effect, being DNL(ε) almost the
to Neuber’s rule (point S) or according to the plane same for the different σm values. Figure 12 shows
strain hypothesis (point N). the results of the investigation: values calculated
Comparative results are shown in Figure 12 with σm /Δσ0 ranging from 0 to 50% are all aligned
with reference to the same mean stress levels above with the curve of DNL(P) calculated for zero mean
referred to explaining Figure 10. As the graph stress.
clearly illustrates, damage is almost independent By summing up, systematic investigations point
from the rule by which conversion is made. So, out that the most important factor influencing out-
conversion method appears to be ineffective as far comes in elastic-plastic approach is the mean stress
as Equation 4 and 5 are considered. value. Secondly, set of material data to be used for
The second basic hypothesis analysed is that specifying stress-strain curve (two parameters) and
referring to possibility to adopt a mean stress cor- capability curve (four parameters) has an impor-
rection embedded in the fatigue capability ε-N tant role just if distinction is made in welded joints
curve, as stated in Equation 3. Analyses have been between parent material and material lying in the
accomplished on the whole range of considered heat altered zone along the weld line. Finally, alter-
native hypotheses about mean stress correction on
ε-N curve end about conversion from elastic to
400 elastic-plastic stress-strain field have no effect to all
practical purposes.
σa,k K
S
σa S
σa N 5 CONCLUSIONS
σa, MPa
N
This work deals with the improvement of the
method for fatigue checking ship structural details
cyclic σa−εa curve during design stage. First, theory and practice
σa S εa S = (Ktσa,hs/Khs)2 /E related to a basic nonlinear procedure has been
σa /εa= E outlined, within the cumulative damage approach.
0 Then, a case study has been discussed in order to
0 εa N εa S 0.3 analyse advantages and drawbacks of the method
εa % proposed to account for large plastic zone at crack
tip, like in case of a low cycle fatigue scenario.
Figure 11. Methods to convert FE calculated stress- Finally a comparison has been made between lin-
strain field to elastic-plastic stress-strain field (E = 206.85 ear elastic and elastic-plastic approach, that shows
GPa and Δσ0 = 627 MPa). a general agreement between the two methods, on
286
287
ABSTRACT: Structural failure due to a lack of fatigue strength and crack initiation is often caused
by local stress concentrations at welded joints. By decreasing the weld notch radius to small values, the
magnitude of stress concentration no longer affects the fatigue strength but a mean value of stress deter-
mined around the notch tip. Moreover, in linear elasticity the local stress distribution becomes singular
with vanishing radius. Besides the common methods for fatigue assessment, the utilisation of an averaged
value of strain energy, computed for a small volume of material, is an alternative approach. The method
shows promise with respect to finite element analysis of complex problems. In most of the previous work
the quality of the numerical approximation was controlled by mesh refinement. In this paper, alternative
refinement techniques are investigated, accounting for the singular behaviour of the exact solution in the
vicinity of the re-entrant corner. After the introduction of the concept based on the averaged strain energy
density, different refinement strategies improving the quality of the finite element approximation are
described and applied to a selected example. The results of the different approaches related to both h and
p-refinement are presented and discussed also in view of application to more complex three-dimensional
problems.
1 INTRODUCTION
289
290
291
5 NUMERICAL ANALYSIS
OF THE DOUBLER PLATE
292
Table 2. Computed ΔW and relative error e for Table 4. Computed ΔW and relative error e for
R0 = 0.28 mm and Δσn = 150 MPa at weld toe for a uni- R0 = 0.28 mm and Δσn = 150 MPa at weld toe for a radi-
form mesh. cal mesh.
293
REFERENCES
294
Wolfgang Fricke
Hamburg University of Technology (TUHH), Hamburg, Germany
Matteo Codda
CETENA—Centro Tecnico Navale, Genova, Italy
Olav Feltz
Hamburg University of Technology (TUHH), Hamburg, Germany
Yordan Garbatov
Instituto Superior Técnico (IST), Lisbon, Portugal
Heikki Remes
Aalto University, Helsinki, Finland
Giovanni Risso
CETENA—Centro Tecnico Navale, Genova, Italy
Cesare Rizzo
Università di Genova (DINAEL), Genova, Italy
Jani Romanoff
Aalto University, Helsinki, Finland
ABSTRACT: The fatigue assessment of welded ship structures includes high uncertainties. One major
source of uncertainty is the determination of relevant local stresses based on the structural hot-spot
or notch stress approach. To quantify the uncertainties and to improve the analysis guidelines, round-
robin studies were performed within the MARSTRUCT project. One recent study concerned load-
carrying fillet welds which are treated in different ways in structural stress approaches. In total, five
partners participated in the work. Two structural configurations at 12 mm thick plates were analysed,
i.e., double-sided lap-joints and doubler plates, each with two different weld throat thicknesses (2.5 mm
and 7 mm). After shortly describing the fatigue assessment approaches, the models of the partners and
their results are described in more detail. Differences in the results are identified and conclusions drawn
with respect to modelling guidelines and typical scatter of computed fatigue lives.
295
296
No contact:
1 None (theory) 1.00 1.00 1.00 1.00
2 Very fine 1.04 1.03 0.95 0.98
(t/10xt/10)
3 Fine (0.4txt/4) 0.87 0.95 0.81 0.90
4 Fine (0.4txt/2) 1.20 1.12 – –
Contact:
1 None (theory) 1.00 1.00 1.00 1.00
2 Very fine 0.98 1.00 0.95 0.98
(t/10xt/10)
3 Fine (0.4txt/4) 0.99 1.02 0.99 1.02 Figure 3. Nominal and structural hot-spot stress in
4 Fine (0.4txt/2) 1.18 1.12 – – the lap joint (a) and meshing effects on the extrapolated
structural stress (b, c).
297
298
Table 3. Computed structural stress concentration Table 4. Computed structural stress concentration
factors acc. to the approach after Dong (2001). factors acc. to the approach after Poutiainen (2006).
299
The effective notch stress approach, using the elas- Participant Element size L-2.5 D-2.5 L-7 D-7
tic stress in the notches rounded by a reference No contact:
radius rref = 1 mm, is able to assess both, weld toes 1 0.2 mm 8.38 3.54 3.97 3.03
and weld roots (Radaj et al., 2006; Hobbacher, 2 0.2 mm 8.23 4.01 3.66 2.93
2009). Weld roots at the end of non-fused plate 3 0.05 mm 9.32 3.71 4.17 3.09
surfaces as occurring in the test specimens are usu- 4 0.2 mm 6.62 3.69 – –
ally rounded with a so-called keyhole shape. Contact:
Relatively fine-meshed finite element models 1 0.2 mm 6.62 3.77 3.35 2.82
have been created for the numerical analyses (Fig. 7) 2 0.2 mm 7.42 4.01 3.41 2.84
considering the recommendations by Fricke (2008). 3 0.05 mm 6.89 3.86 3.24 2.73
Participant 4 used again a 3D model. 5 0.2 mm 7.14 3.97 3.42 2.84
As the throat thicknesses is partly only 2.5 mm
and the rounded keyhole notch reduces the throat
thickness by approx. 0.3 mm when placed at the Table 6. Computed notch stress concentration factors
end of the slit, two participants (1 and 5) modi- for the weld root.
fied the model by shifting the circle in horizontal
and/or vertical direction so that the nominal throat Participant Element size L-2.5 D-2.5 L-7 D-7
thickness is maintained. Figure 8 illustrates the
No contact:
modification made by the participants.
1 0.2 mm 7.61 2.59 2.73 1.38
Tables 5 and 6 summarize the results at the weld
2 0.2 mm 10.38 4.43 4.17 2.23
toe and weld root, respectively. The SCFs were
3 0.05 mm 9.01 2.94 2.87 1.40
computed using the max. principal stress at the
4 0.2 mm 10.37 4.55 – –
rounded surface, according to Fricke (2008).
Contact:
Some scatter in the results is observed. For 1 0.2 mm 8.38 4.04 4.12 2.24
instance, the models without shifted keyhole notch 2 0.2 mm 9.28 4.38 4.30 2.36
3 0.05 mm 8.66 4.20 4.05 2.24
5 0.2 mm 7.89 3.90 3.89 2.13
300
301
302
ABSTRACT: A FEM analysis used for fatigue assessment, based on the rules, was done in the paper.
The approach for fatigue assessment in the early design stage has been developed. To overcome the
challenges due to limited information in the early design stage, generic structural elements and predefined
fatigue-critical details were chosen. This allows the development of a common approach for different ship
types, which is also applicable for optimization purpose. The aim of the sensitivity analysis is to study the
effect of the structural scantlings on the structural hotspot stress factor. The analysis is focused on a pillar
and girder connection of the ship hull structure of a Ropax vessel.
303
304
305
306
307
ABSTRACT: We have developed a multi-scale design sensitivity analysis method for transient dynamics
using a bridging scale method by a projection operator for scale decomposition. Employing a mass-weighted
projection operator, we can fully decouple the equations of motion into fine and coarse scales using the
orthogonal property of complimentary projector to the mass matrix. Therefore, independent solvers in
response analysis can be utilized for the fine scale analysis of molecular dynamics and the coarse scale
analysis of finite element analysis. To reduce the size of problems and to improve the computational
efficiency, a Generalized Langevin Equation (GLE) is used for a localized Molecular Dynamics (MD)
analysis. Through demonstrative numerical examples, it turns out that the derived sensitivity analysis
method is accurate and efficient compared with finite difference sensitivity.
311
{ },
concurrently using the velocity Verlet method for
= NT f z
Md f (7) MD and explicit central difference for FE. The
MD simulation is advanced by m steps of size
where the internal force f(q) can be obtained as Δtm = Δt/m while the FE simulation is advanced
Eq.(8). through a single time step of size Δt. The more
detailed process can be referred in the reference
∂U ( q ) (Wagner, G.J. and Liu, W.K., 2003).
f (q) = − (8)
∂q
3 DESIGN SENSITIVITY ANALYSIS (DSA)
2.2 Generalized Langevin equation
Most of the multi-scale problems are inter- The first order variation the total solution z can be
ested in the fine scale effects only on locally con- expressed as,
fined regions, not the whole domain so that we
introduce the bridging scale approach by reducing z′ Nd Qq ′ . (12)
the MD domain to analyze a locally confined fine
scale region in which we are interested as described We decomposed the total adjoint variable λ into
in Figure 1. fine scale ξ and coarse scale η like the total solution
To prevent fine scale wave reflection on the decomposition in the response analysis.
MD/FE boundary, a GLE force f imp is introduced.
The multi-scale equations of motion is written as, λ η ξ. (13)
Interested atoms Ωa
We can obtain the multiscale adjoint equations
fimp (GLE Force)
Eliminated atoms Ωb using the decomposed adjoint variables.
∂fa ( q a , v b , b )
ζTa Aa ζTa (t )
∂q a
Ωa tT
Ωa Ωc
+ ∫ ζTa ( )θ( t )dτ
t
∂f MD ( + a a)
Ωb + η(t )T NT Qa
∂( + a a)
Full Domanin (MD) Fine scale (GLE) Coarse scale (FE) ⎛ ∂G d ∂G ⎞
+⎜ − , (14)
Figure 1. Bridging scale method. ⎝ ∂q a dt ∂q a ⎟⎠
312
+ η( )T
∂NT f CB
C
{
(d,, ) } In the Figure 2, the red line denotes the initial
crack line. In this model, uniaxial tensions are
∂d imposed on both the top and bottom of the struc-
∂ a f ( q a ,v
, vb , ) ture with the velocity profile given in the right side
+ ζTa ( ) N(( b)
∂v b of Figure 2. The analysis results at each time step
tT are shown in Figure 3 and Figure 4. The contours
− ∫ ζTa ( )θ( , b)) (Xa ) in Figure 3 are L-J potential values and the ones in
t
⎛ ∂G d ∂G ⎞ Figure 4 are y-directional displacement.
+⎜ − . (15)
⎝ ∂d ddt ∂d ⎟⎠ Performance measure is the total displacement
on the point XP at each time step. Table 1 shows the
The adjoint system is a terminal value problem design sensitivity obtained by FDM, DDM AVM,
in transient dynamics. The corresponding terminal respectively. Compared with the results of FDM
conditions are derived as and DDM, that of AVM is very accurate.
Figure 5 shows the comparison of computational
∂g costs of the FDM, DDM, and AVM. The computa-
ζTa (t )MAa = , (16) tion cost for the FDM and DDM linearly increases
t tT
∂q a t tT as the number of design variables increases.
However, since the AVM needs only one more time
⎛ ∂g ∂G ⎞ integration for the adjoint equations regardless of
ζ Ta = −⎜ + , (17)
Aa
t tT
⎝ ∂q a ∂q a ⎟⎠ t tT
V
∂g
η( )T M = , (18) FE region
t tT ∂d t tT V
b/2
XP
Initial crack
h/7
h
h/2
4 / 7b Vmax
h/4
(19) MD+FE region
FE region
4 NUMERICAL EXAMPLES
313
V
Non-designable
thickness
t8
9
h FE 1
20 b t7
2
t6
t5
h Xd
t4
Initial crack t3
t2
MD+FE Xq t1
Non-designable
b/8 FE b/8 thickness
V
Figure 5. Comparison of computational costs. b
314
A multi-scale DSA method in a bridging scale This research was supported by Basic Science
approach is developed using the AVM. To avoid any Research Program through the National Research
iterative computations between the scales, we use a Foundation of Korea (NRF) funded by the
fully decoupled equation for each scale in original Ministry of Education, Science and Technology
response as well as sensitivity analyses. Through (Grant Number 2010-18282). The support is grate-
a GLE, a locally confined MD region instead of fully acknowledged.
whole MD system is considered for the fine scale
solution whereas a FE analysis for the coarse scale
solution is performed on the whole region. The REFERENCES
efficiency of the developed method is achieved
due to the fully decoupled multi-scale equations Choi, K.K. & Kim, N.H. 2004. Structural Sensitivity
Analysis and Optimization 1: Linear Systems. New York:
in these scales, which are derived by the use of Splinger.
identical mass-weighted projection in the response Kadowaki, H. & Liu, W.K. 2004. Bridging multi-scale
as well as adjoint analyses. Numerical implementa- method for localization problems. Computer Methods in
tions demonstrate the accuracy of the developed Applied Mechanics and Engineering. 193: 1733–1772.
DSA method for various design variables. The Liu, W.K., Karpov, E.G. & Park, H.S. 2004. An intro-
developed method turns out to work very well duction to computational nanomechanics and
for any design variable in both scales. The devel- materials. Computer Methods in Applied Mechanics
oped bridging scale DSA method can be utilized and Engineering. 193: 1529–1578.
in industrial problems if it is extended to parallel Park, H.S., Karpov, E.G. & Liu, W.K. 2005. The bridg-
ing scale for two-dimensional atomistic/continuum
computation and additional considerations. Also coupling. Philosophical Magazine. 85(1): 79–113.
realistic simulation of crack propagation problems Wagner, G.J. & Liu, W.K. 2003. Coupling of atomistic
should include thermal effects due to the breaking and continuum simulations using a bridging scale
of an atomic bond. That is a future work for our decomposition. Journal of Computational Physics.
research group. 190: 249–274.
315
Ken Takagi
The University of Tokyo, Japan
ABSTRACT: In this paper, in order to examine the wave loads for fatigue design, a tank test using
newly designed elastic model was performed, and a further detailed analysis of full scale measurement
on Post-panamax container ship was carried out. It was verified that the actual stress by the full scale
measurement is quite smaller than that by the direct computation, and the difference is mainly caused
by of environmental condition applied to the computation. Moreover, the probability of occurrence of
whipping during long term was estimated. It was found that whipping induced stress cumulates a mount
of fatigue damage, but the probability of occurrence of whipping is small in terms of the subjected ship.
Finally, the further advanced view for the fatigue design is proposed.
317
318
40 4
It is found that the short term damage includ- 20 2
ing the effect of hull girder vibration is sensitive to 0 0
the ship speed and wave direction (χ). The speed -20 280 290 300 310 320 330-2
reduction and/or course change generate lower -40 -4
damage. For the rational evaluation of the fatigue -60 -6
strength, it is necessary to evaluate wave laods tak- -80 -8
ing the real operation in rough seas, speed change -100 -10
Time (sec.}
and course change, into account.
Figure 5 shows the long-term fatigue damages Figure 6. Example of time history on full scale
based on the tank test. The IACS Recommenda- measurement.
tion No.34 was used for the wave scatter diagram.
Short term sea state excluded in the tank test, was
complemented by means of the response ampli- Long term distribution of wave height
tude operator measured in the regular wave test. 30
Full scale measurement
Under the assumption that the subjected ship is
JPN-EU (GWS)
operated at its service speed during 25 years, which 25
North Atrantic (GWS)
corresponds to a whole life, long-term fatigue
damage including the hull girder vibration (RAW) 20
Hw (m)
5
3 FULL SCALE MEASUREMENT
0
0 1 2 3 4 5 6 7 8
3.1 Subjected ship and measured data
-LOG(Q)
In order to evaluate the characteristic of the long-
term fatigue damage, the data of full scale meas- Figure 7. Probability of exceedance of encountered
urement (Okada, 2006) were analyzed. wave height.
319
*Probability of exceedance.
**Full scale measurement.
***Global wave statistics (Hogben, 1986).
1.00E-04
0.00E+00
dominant to long-term fatigue damage. The dif- 0 2000 4000 6000 8000
ferences can be quantitatively confirmed from the Number of short term measurement
Weibull shape parameter shown in Table 1. Assum-
ing that vertical bending stress and wave height Figure 8. Transition of the cumulative damages based
have linear correlation, the long-term fatigue dam- on the actual stress by the full scale measurement.
age was able to be simply estimated. Table 2 shows
the normalized long-term fatigue damages calcu-
Table 3. Fatigue damage extrapolated to 25 years based
lated based on the each wave distributions. It is
on the actual stress by the full scale measurement.
found that the long-term fatigue damage based on
the actual wave is much lower than that based on Long-term fatigue damage
the scatter diagram, and the effect of the long term
distribution of encountered wave height on fatigue DRAW 0.065
strength is significant. DLPF 0.030
DHPF 0.005
3.3 Fatigue analysis based on the full scale
measurement
damages. Rain flow method and D curve of UK-HSE
Cumulative long-term fatigue damage of deck are applied. The long-term fatigue damages based
structure was estimated based on the full scale on the actual stress is quite low, even taking the hull
measurement. As with the tank test, three kind girder vibration into account (DRAW).
of frequency component of stress response have
been prepared taking the natural frequency of the
2-node hull girder vibration of vertical bending
4 FATIGUE DESIGN ASSESSMENT
into account.
• RAW: 0.01 ∼ 1.0 Hz; 4.1 Direct calculation
• LPF: 0.01 ∼ 0.5 Hz;
A direct calculation of spectral-based fatigue
• HPF: 0.5 ∼ 1.0 Hz
analysis (ABS, 2007; DNV, 2008; LRS, 2004) was
Figure 8 shows the cumulative trend of fatigue performed in order to verify the actual stress and
damages. Extrapolation to 25 years is necessary fatigue strength. As should be noted, the direct cal-
in order to estimate the long-term fatigue damage culation which is based on liner theory does not
because the stress had been recorded in 20 minuts take the effect of hull girder vibration into account.
in every 2 hours and the period two and half years. The summery of the direct calculation in this study
Table 3 shows the predicted long-term fatigue is shown in following.
320
100
5 EFFECT OF THE VIBRATION STRESS
80
ON HULL GIRDER FATIGUE DESIGN
ASSESSMENT
60
40
5.1 Long-term fatigue damage
20
0 The effect of hull girder vibration on long-term
0 1 2 3 4 5 6 7 8 fatigue damage (CEL) is simply defined in follow-
-logQ
ing equation.
Figure 9. Long-term distribution of computed stress DRAW
(without considering hull girder vibration) and actual CE
EL = (2)
stress by the full scale measurement. DLPF
321
3000
Frequency
In order to evaluate the effect of elastic vibration
2000
on fatigue strength, the probability of occurrence 1253
770
1000 465
of whipping during long term is estimated by the 267 125
57 51 41 19 20 10 3 6 2 5 2 3 2 1 1 1
0
full scale measurement. In this study, in order to
10.5
0.5
1.5
2.5
3.5
4.5
5.5
6.5
7.5
8.5
9.5
10
11
1
9
determine whether or not the whipping occurs, the Hs(m)
peak value of high pass filtered stress was utilized. Whipping occurance
40
Maximum or minimum HPF stress during zero-
30
cross period of LPF stress is cyclically counted, and
Frequency
Stress_trh=20MPa
the long term distribution of the slamming induced 20
Stress_trh=30MPa
10.5
0.5
1.5
2.5
3.5
4.5
5.5
6.5
7.5
8.5
9.5
10
11
1
9
as encounter wave period. Figure 10 shows the
Hs(m)
probability of exceedance of LPF stress and
HPF stress. The HPF stress become larger under Figure 12. Frequency distribution of the sea states
Q = 10−3. However, for example assuming whipping during period of the full scale measurement (upper) and
over 20 N/mm2 of deck stress single amplitude, number of whipping occurrence per sea state (lower).
the probability exceedance is about 5 * 10−5 which
equal about 2,000 times during life (N = 108). It
is assumed that the probability of occurrence of the incident wave is judged as induced whipping.
whipping during a whole life is small. Figure 11 shows a case of the threshold is set to
20 N/mm2 in single amplitude. The histogram of
5.3 Frequency of whipping in each sea state whipping occurrence per sea states is obtained.
Figure 12 shows an example in case of threshold
When the peak HPF stress during the encounter setting to 20 N/mm2 and 30 N/mm2 in single
wave period exceeds the threshold value of stress, amplitude. Any whipping was not occurred in
less than of equal to in 2 m significant wave
height. Moreover, the sea state lower than 2 m
1.E+00 in significant wave height accounts for 95% of
1.E-01 the total number of sea states, as known in upper
Probability of exceedance
1.E-07
0 10 20 30 40 50 60
Stress Amplitude (MPa)
6 CONCLUSIONS
Figure 10. Probability of exceedance of deck stress The analysis of fatigue strength based on tank test
amplitude induced by the wave loads (LPF) and by the and full scale measurement were performed. The
hull girder vibration (HPF). conclusions are as follows.
322
REFERENCES
323
325
1 λ4
υ+ = (5)
2π λ2
1 λ2
υ0 = (6)
2π λ0
υ0 λ2
IF = = , (7)
υ+ λ0 λ 4
where, υ+ is peak period, υ0 is zero up-crossing
period, IF is irregularity factor and λm is spectral
moment defined by:
∞
λm ∫ω S (ω )dω
m
m = 1, 2,... (8) Figure 2. Random responses of time domain corre-
0 sponding with sampling time step.
326
λm
αm = (12)
4 NUMERICAL SIMULATIONS λ0 λ2 m
Spectrum examples performed in this study were Resultant time history and rain-flow range
used from Tove-Benasciutti (2005) like Figure 4. distribution are shown in Figure 5.
327
∫z
m
pRR ( z )ddz
M RR ( m ) = 0
(13)
∞ z2
−
∫z
m 2
ze ( z )dz
0
328
⎢ x1
⎥
⎢ m⎞ ⎥
m
m
⎛ m⎞ ⎛1 +
⎢⎣ + C2 ( x 2 ) 2 ⎝ 1 + 2 ⎠ + C3 (
2
2 ⎠ ⎥⎦
)
JB ⎝
M RR ( m ) = m
( )
2 ⎛1 + m⎞
⎝ 2⎠
Γ
⎛ m + 1⎞
C1 ⎝ 2 ⎠
= ( σ x1 ) m
+ C2 ( )m + C 3 (19)
⎛ m⎞
x2
π
Γ 1+
⎝ 2⎠
4 equations to be used were made as below.
C1 C2 + C3 = 1 (20)
V1C1σ x1 + C x2 C3 = α 2 for m = 1 (21)
V2C1σ x12 + C x2
2
C3 = A fo m = 2 (22)
V3C1σ x13 + C x2
3
C3 = B fo m = 3 , (23)
where,
1 (1)
V1 =
π Γ (1.5)
1 (1.5)
V2 =
π Γ (2)
1 (2)
V3 =
π Γ (2.5)
329
In case of two-slope S-N curve, the associated The damage ratio between the spectra in Figure 4
damage equation is as follows: using the developed model and damage from
⎡
⎢ ⎜ (
⎛ 2 2λ m
0
σ
) ⎛
m ⎜ m +1
Γ
⎛ S0 ⎞ ⎞
2
⎟
⎞ ⎤
⎟ ⎥
⎢ ⎜ c x1 ; ⎜ ⎟ ⎟ ⎥
⎜
⎝
2 ⎝ 2 2 λ σ ⎠ ⎟
⎠
⎢ C1 ⎜ 0 x1
⎟ ⎥
⎢ π⎜ 2⎞⎟ ⎥
( )
m′
⎛
⎢ ⎜ + 2 2 λ0 σ x1m ′γ ⎜
m′ + 1 ⎛ S0 ⎞ ⎟ ⎥
⎢ ⎜ ;⎜ ⎟ ⎟⎟ ⎥
⎢ c′ ⎜⎝ 2 ⎝ 2 2 λ0 σ 1 ⎠ ⎟⎠ ⎠ ⎥
⎝
⎢ ⎥
⎢
⎢
⎛ 2 2λ m
⎜ (0 ) ⎛
σ x 2 Γ ⎜1 + ; ⎜
m ⎛ S ⎞ ⎞
2 ⎞⎥
⎟⎥
⎟ ⎟
m 0
⎢ ⎜ c ⎜⎝ 2 ⎝ 2 2 λ0 σ x 2 ⎠ ⎟ ⎟⎥
⎠
DRFC ( m ) = Tdν p ⎢ + C2 ⎜
JB ⎟⎥ (33)
⎢ ⎜ m′ ⎛ ⎛ ⎞
2 ⎞⎟ ⎥
⎜+ ( 2 2 λ ) m ′ S ⎟
⎢ 0
σ x 2 m ′γ ⎜1 + ; ⎜ 0
⎟ ⎟ ⎥
⎢ ⎜ c′ ⎜⎝ 2 ⎝ 2 2 λ0 σ 2 ⎠ ⎟ ⎟ ⎥
⎢ ⎝ ⎠⎠ ⎥
⎢
⎢
⎛ 2 2λ m ⎛
⎜ (0
Γ ⎜ +
) m ⎛ S0 ⎞
2⎞
⎟
⎞
⎟
⎥
⎥
⎢ ⎜ c 1 ; ⎜ ⎟ ⎥
⎜ 2 ⎝ 2 2 λ0 ⎠ ⎟ ⎟
⎢ ⎜ ⎝ ⎠ ⎟ ⎥
⎢ + C ⎥
( )
3⎜ m′ 2⎞⎟
⎛
⎢ ⎜ + 2 2 λ0 γ ⎜1 + ; ⎜
m′ ⎛ S0 ⎞ ⎟ ⎥
⎢ ⎜ ⎟ ⎟⎟ ⎥
⎢ c′ ⎜⎝ 2 ⎝ 2 2 λ0 ⎠ ⎟ ⎥
⎣ ⎝ ⎠ ⎠ ⎦
330
331
Figure 13. Fatigue damage ratio for Dirlik model Figure 15. Fatigue damage ratio for Jiao-Moan model
according to Vanmarcke parameter. according to Vanmarcke parameter.
332
REFERENCES
333
S. Tanaka
Graduate School of Engineering, Hiroshima University, Japan
H. Okada
Faculty of Science and Technology, Tokyo University of Science, Japan
S. Okazawa
Graduate School of Engineering, Hiroshima University, Japan
ABSTRACT: The purpose of the present study is to solve surface crack in welded joint of ship structure.
To know the mechanical properties and the crack growth predictions of the surface crack are very
important from the point of view of fatigue strength evaluation. In this study, shell-solid mixed analysis
using Rigid Body Element 3 (RBE3) in MSC.Nastran and Virtual Crack Closure Method (VCCM) for
quadratic tetrahedral finite elements adopted to evaluate Stress Intensity Factors (SIFs) in the welded
joint of a ship. As a numerical example, the calculation of SIFs to the surface crack in welded joint of
three cargo hold carrier are demonstrated. The accuracy and effectiveness of the proposed method are
discussed.
335
336
G IIIVCCM
K III = 2μ G III
VCCM
. (5)
G IIIVCCM
4 APPLICATION TO A SHIP’S
STRUCTURAL MEMBERS
337
Nodes Elements
338
339
340
ABSTRACT: In this paper the results of the study on the protective capacity of ship hull structures
made of composite materials subjected to an explosion of a spherical charge are presented. The main
application of these structures may be followed by design of ship structures of great importance, which
should also be protected against exceptional loads of this kind. In this study, a nonlinear analysis with the
finite-elements was done. The methodology to apply the blast pressure and the mechanism of the blast
wave in free air is given. The space pressure variation is determined by using Friedlander exponential
decay equation. Various parametric calculus to evaluating the behaviour of the ship structure laminated
plate to blast loading: explosive magnitude, distance from source of explosion, plate thickness.
343
344
345
346
347
Figure 7. Variation of the maximum transversal dis- Figure 10. Variation of the transversal displacement of
placement in the case h = 0.2 m. the plate’s central point, without damping.
348
349
350
S. Ehlers
Aalto University, School of Science and Technology, Marine Technology, Espoo, Finland
ABSTRACT: A grounding accident of a Suezmax tanker, built in compliance with IACS Common
Structural Rules (CSR), is presented and analysed from the structural point of view. The analysis focuses
on the simulation of structural deformations caused by the accidental event. The analysis is performed
using the non-linear finite element method. The validation of the simulation results is performed through
the comparison of calculated deformations with those measured on the actual damaged structure
of the vessel. The measurements were performed in the dry dock prior to the steel replacement. Thereby,
the validity of the finite element simulations is secured and the conclusions on the state of deformation
can be made with confidence. Furthermore, critical structural arrangements can be outlined, in order to
contribute to an improvement of the similar new buildings. The validity of the simulation procedure raises
the belief in the analysis of future grounding events.
351
352
353
354
Energy [MJ]
967 MJ
800
Force [kN]/100
600
400
200
0
0 5 10 15 20 25 30
Damage length [m]
Figure 14. Comparison of the simulated and measured Figure 16. Simulation a) versus photograph b) of
(black) frame deformations. frame 90F.
355
356
ABSTRACT: Most of the present semi-submersibles adopt single hull design for their columns.
The anti-collision capability of the single hull column could fail to prevent the water pouring into the
column in case of the collision accident. A double-hull design concept is proposed in the paper. The ring
frame, the inner shell and the outer shell create an integrated system to undertake the collision. Three sets
of finite element models are built. The nonlinear numerical simulation code, MSC.Dytran, is employed
to do the simulation. Three collision scenarios are defined. The structural damages, the curves of
relationship between collision force vs. penetration, and the curves of energy dissipation are obtained, to
evaluate the anti-collision characteristics for the different column design and different collision scenarios.
Consequently, comparisons are made, to evaluate the improvement of the anti-collision characteristic for
the double-hull design, and some suggestions are also provided.
357
The double hull design concept takes advantage of A semi-submersible is chosen as the struck object,
the free end of the ring frame, which provides a a containership is chosen as the striking vessel and
support of the inner shell. In the original design, their scantling data is listed in Table 1 and Table 2,
one end of the ring frame is connected with outer respectively.
shell, and the other end is free. In the double hull
design, an inner shell is designed in the column and
4 FINITE ELEMENT MODELS
connected to the other end of ring frame. The inner
shell is designed between the two decks in vertical
Three sets of finite element models are made,
direction, and parallel to the outer shell. The inner
which are:
shell, the ring frame and the outer shell build an
integrated protection system, which can increase − Set1, whole semi-submersible and whole striking
the anti-collision capability. In the original design, vessel
the free end of the ring frame always deform with- − Set2, a single hull column and a rigid striking bow
out constrain. In the double hull design, the ring − Set3, a double hull column and a rigid strik-
frame will deform with restriction. Besides, the ing bow
358
Length m 114.07
Width m 78.68
Column height m 17.385
Column width m 15.86
Draft m 19.0
Displacement ton 51751.3
Figure 3. Finite element models set2, single hull column
with original design and the striking bow.
Table 2. Scantling data of striking vessel.
Length m 201
Displacement ton 36198.9
Draft m 12
359
Oblique
Lateral
360
3.00E+07
energy dissipation, the numerical simulation for
2.50E+07 a column can represent that for the whole vessels.
Collision force (N)
2.00E+07 101
201
It can also be pointed out from Figure 7 that the
1.50E+07
1.00E+07
differences of curves of collision force vs. penetra-
5.00E+06 tion in each comparison are not large. In the com-
0.00E+00
0 1 2 3 4 5 6 7 parison between 101 and 201, the value of collision
penetration (m)
a. Case 101 and 201 force for single hull is higher than that of whole
Collision force comparison vessels, proving the case 101 brings conservative
3.50E+07
results than that of case 201. In the comparison
3.00E+07
2.50E+07
between 102 and 202, and the comparison between
Collision force (N)
2.00E+07 102
103 and 203, the curves of collision force cross
202
1.50E+07
each other. But the difference is not large.
1.00E+07
5.00E+06
On the basis of these comparisons, it can be
0.00E+00
0 1 2 3 4 5 6
summarized that the simulation of a rigid bow
penetration (m)
striking the column is able to bring a conservative,
b. Case 102 and 202 at least a similar result for that of the simulation
2.50E+07
Collision force comparison
of the whole vessels collision. Thus, in the follow-
ing analysis, the structural performance and anti-
2.00E+07
collision characteristic of the double hull design is
Collision force (N)
1.50E+07
103
203
assessed by the simulations for a single column and
1.00E+07
rigid bow.
5.00E+06
0.00E+00
0 1 2 3
penetration (m)
4 5 6
6.2 Assessment of anti-collision characteristic
c. Case 103 and 203 for the double hull design
Figure 7. Comparison of curves of collision force vs. As introduced in section 2, fatal moment is a
penetration for the first step analysis. factor of crucial importance for the anti-collision
capability of semi-submersible. Values of fatal
moment for simulation cases from 201 to 303
Energy dissipation comparison are picked out from simulation results, and the
9.00E+07
8.00E+07
values of corresponding penetration can also be
7.00E+07
determined. These data is listed in Table 4. In
Distortion energy (J)
6.00E+07
5.00E+07 101
201
Table 4, rupture time means the moment when the
4.00E+07
3.00E+07
outer shell or inner shell ruptures. For original shell
2.00E+07
1.00E+07
design, rupture time of the outer shell is the fatal
0.00E+00
0 1 2 3 4 5 6 7
moment. Nevertheless, for double hull design, the
Penetration (m)
fatal moment is the rupture time of the inner shell.
a. Case 101 and 201 For the comparison between single hull design
8.00E+07
Energy dissipation comparison
and double hull design, the comparisons for col-
7.00E+07 lision force and distortion energy are presented in
6.00E+07
Figure 9 and Figure 10. In Figure 9, the curves of
Distortion energy (J)
5.00E+07
4.00E+07
102
202 collision force vs. penetration for the corresponding
3.00E+07
0.00E+00
curves of distortion energy vs. penetration for the
0 1 2 3 4
Penetration (m)
5 6 7
corresponding simulation cases are presented.
b. Case 102 and 202 It has been emphasized in the paper that the fatal
Energy dissipation comparsion
moment is the crucial important time to assess the
6.00E+07
5.00E+07
anti-collision characteristic of the semi’s column,
the comparison is made on the results at the fatal
Distortion energy (J)
4.00E+07
3.00E+07
103
203
time. Although the simulations continue to a large
2.00E+07 penetration state, the assessment only adopts the
1.00E+07
values at the fatal time and the rest of the results
0.00E+00
0 1 2 3
Penetration (m)
4 5 6 7 are used for reference. Because the distortion
c. Case 103 and 203 energy is easy to compare, the energy dissipation
by structural components in each simulation case
Figure 8. Comparison of curves of distortion energy vs. is listed in Table 5, and a summary comparison is
penetration for the first step analysis. presented in Table 6. The structural deformation
361
Simulation case Shell Rupture time (s) Fatal moment (s) Penetration (m)
5.00E+07 1.60E+08
4.50E+07
1.40E+08
4.00E+07
3.50E+07 1.20E+08
3.00E+07 1.00E+08
201
2.50E+07 201
301 8.00E+07
301
2.00E+07
6.00E+07
1.50E+07
1.00E+07 4.00E+07
5.00E+06
2.00E+07
0.00E+00
0 1 2 3 4 5 6 0.00E+00
Penetration (m) 0 1 2 3 4 5 6
Penetration (m)
3.50E+07 1.20E+08
3.00E+07 1.00E+08
2.50E+07
Distortion energy (J)
8.00E+07
Collision force (N)
5.00E+06 2.00E+07
0.00E+00 0.00E+00
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
Penetration (m) Penetration (m)
3.00E+07 9.00E+07
8.00E+07
2.50E+07
7.00E+07
Distortion energy (J)
Collision force (m)
2.00E+07 6.00E+07
1.00E+07 3.00E+07
2.00E+07
5.00E+06
1.00E+07
0.00E+00 0.00E+00
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
Penetration (m) Penetration (m)
Figure 9. Comparison of curves of collision force vs. Figure 10. Comparison of curves of distortion energy
penetration for the first step analysis. vs. penetration for the first step analysis.
modes of the structural components of semi’s semi’s column very much, which means the semi’s
column for double hull design are presented in column can improve its survivability in case of a
Figure 11. serious collision accident. Besides, from the values
in Table 6, it can also be indicated that the scenario
Percentage = (E1−E0)/E0 × 100% (1) when the striking vessel hits the semi’s column in
the longitudinal direction is the most vulnerable
It is indicated from the Figure 9 that the colli- situation, for the energy dissipated by the semi-
sion force of double hull design is bigger than that submersible is the smallest at the fatal moment
of original design from the beginning of the colli- among all of the three scenarios.
sion course. Form Figure 10, it can also be pointed In the original design, the outer shell and the ring
out that the double hull design absorbs more dis- frame of column deform separately, though one
tortion energy than that of the original design. The side of the ring frame is connected with the outer
amount of energy dissipation increase presented in shell. The outer shell has a large space to deform,
Table 6 demonstrates that the double hull design and its major deformation mode is membrane
benefits the energy dissipation capability of the elongation and rupture. The major deformation
362
Table 6. Energy comparison at fatal moment. inner shell are membrane elongation and rupture.
The deformation modes of the ring frame and
E0 (MJ) E1 (MJ) Percentage vertical bulkhead are folding and buckling, which
are identical to those of original design. However,
Longitudinal collision
201 vs. 301 14.485 24.612 69.9
the incorporation of the inner shell changes the
Lateral collision
original design’s free deformation mode. The inner
202 vs. 302 18.511 30.082 62.5 shell not only dissipates part of the energy, but
Oblique collision also helps the integrated system to dissipate more
203 vs. 303 15.210 33.653 121.3 energy, for it delays the semi-submersible’s fatal
moment. From the values in Table 5, it is clearly
Note: In Table 6, E0 is total distortion energy by original show that the outer shell, the ring frame and the
design, E1 is total distortion energy by double hull vertical bulkhead dissipate more energy than
design, and the percentage is calculated by equation (1). those in original design, especially the ring frame.
The percentage value demonstrates the improvement The energy dissipation capability of different
effect of the double hull design, compared with that of structural components is described as following:
the original design.
− The energy dissipated by the ring frame increases
quite a while from 2.519MJ to 5.146MJ in case
201 to case 301, from 2.104MJ to 5.577MJ in case
202 to 302, and from 4.127MJ to 10.412MJ in
case 203 to case 303. In all of the three scenarios,
the energy dissipated by the ring frame increase
more than 100 percent.
− The outer shell also enhances its energy
dissipation capability for the double hull design
cases, but the augmentation is not as obvious
as that by the ring frame. For the three pairs of
comparisons, the enhanced energy dissipation
capability of the outer shell is 16.6% in case 201
and 301, 50% in case 202 and 302, and 50.3% in
case 203 and 303. The reason for the outer shell
can dissipate more energy is the delay of the fatal
moment and the increase of penetration.
− The inner shell contributes the energy dissipation
of 3.392MJ in case 301, 3.741MJ in case 302,
Figure 11. Structural deformations of the column struc- and 5.406MJ in case 303. The amount of energy
tural components in case 301 (from left to right: whole
column, inner shell, ring frame, and vertical bulkhead).
dissipated by inner shell is similar to augmented
amount of energy dissipated by outer shell.
− The vertical bulkhead plays an important role in
mode of the ring frame and the vertical bulkhead the energy dissipation for the collision scenario in
is folding and buckling. In double hull design, the longitudinal direction and lateral direction. The
outer shell, the inner shell and the ring frame build double hull design helps the vertical bulkhead
an integrated system to undertake the impact. increases its energy dissipation capability from
The deformation modes of the outer shell and the 3.591MJ to 6.358MJ in case 201 and 301, and it
363
364
Y.T. Huang, K.P. Wu, H.L. Chien, C.M. Chou & K.C. Tseng
Department of Design, CSBC Corporation, Taiwan, Kaohsiung, Taiwan, R.O.C.
ABSTRACT: Many researches on damage of bow structure and hull structures in ship collision have
been carried out. Considering the failure conditions of striking and struck vessels depends on the overall
longitudinal stiffness of bow structures and the crash resistance of struck hull structures. This study inves-
tigated the effects of damage conditions of different bow structures. The vessels designed and constructed
by CSBC Corporation, Taiwan were selected as simulation scenario, i.e. a 203,000 tons bulk carrier was
struck by an 8,240 TEU container ship. The critical speed is defined as the minimum initial velocity of
striking vessel to cause the rupture of inner hull of struck double hulls. Different bulbous bow structure
with same shape but different longitudinal and transverse stiffness were considered. The nonlinear tran-
sient dynamics analysis of both ships in collision was analyzed. The energy dissipation and reaction force
of different type of bow structures were compared.
365
366
367
Figure 6. Deformed shape of different damaged bow Figure 7. Total reaction force of double hull struck by
structures. different bow vs. indentation depth.
368
3.1 FE model
The FE model of bow structure of container is
similar to Figure 5. A cargo hold space was selected
as analysis domain for bulk carrier, and the dimen-
sions of its FE-model are shown in Figure 12.
The model of a striking bow and the struck side
hull is shown in Figure 9 and Figure 10. The mass
of container ship excluding model mass was distrib-
uted on rear bulkhead of model; the mass of bulk
carrier excluding the model was distributed on the
two side transverse bulkheads. The material con-
stants of the FE-model are the same as Table 1.
369
370
Energy
Structure Indentation Reaction dissipation
type depth (m) force (kN) (kJ)
371
372
ABSTRACT: This study introduces a generalized model for the shape of the sea floor which is relevant
for ships during grounding. The sea floor is characterized as a paraboloid that allows a parametric modeling
of the sea floor topology. Nonlinear explicit Finite Element (FE) code LS-DYNA was used to simulate the
ship grounding scenarios. The simplified formulae to estimate the average grounding forces are proposed.
The analytical formulae show reasonable agreement with the results from FE simulations. A simple
computer program was used to simulate the ship response, which is assumed to be two-dimensional rigid
motion, i.e. surge, heave and pitch. The results show that the sea floor topology is indeed a key parameter
in the response analysis of a ship bottom during grounding. The penetration induced by the ship motion
is also a function of ship motion. The envelopes of shear force and bending moment which are induced
during grounding are checked against the DNV’s still water requirements. The ultimate goal of the analysis
is to allow prediction of the risk of penetration into cargo tanks with oil spill as a result, the risk of hull
girder failure and estimation of the required tug force for pulling the vessel off the ground.
373
374
70
Figures 4 and 5 show the results of the horizontal
μ = 0.6
and vertical components of the grounding forces 65
for the obstruction (α = 10, β = 10) at different
levels of initial penetration. It is seen that the 60
horizontal grounding force is affected very much
by μ. On the other hand, the vertical grounding 55
375
100 35 (7,1)
μ = 0.6
25
60
20
40
15
20
10
0
0 5 10 15 20 25
Grounding length [m] 5
0 5 10 15 20
Displacement [m]
Figure 5a. Longitudinal grounding force (Fx) for
(α = 10, β = 10) at δ = 2.0 m. Figure 6a. Longitudinal grounding force (Fx) for
asymmetric seabeds at δ = 1.0 m and μ = 0.3.
130
μ = 0.0
70
μ = 0.3
Vertical grounding force [MN]
(7,7)
120 μ = 0.6 (1,7)
(7,1)
Vertical grounding force [MN]
60
110
50
100
40
90
30
80
70 20
0 5 10 15 20 25
Grounding length [m]
10
0 5 10 15 20
Figure 5b. Vertical grounding force (Fz) for (α = 10, Displacement [m]
β = 10) at δ = 2.0 m.
Figure 6b. Vertical grounding force (Fz) for asymmetric
In this respect, the present work also considers the seabeds at δ = 1.0 m and μ = 0.3.
effect of asymmetric geometries, e.g. (α = 1, β = 7)
and (α = 7, β = 1). Figure 6 shows the results of seabed surface: this pressure is assumed to be
the grounding forces for the asymmetric seabed equivalent to the pressure under vertical punching
topologies together with those of the symmetric for the same level of penetration. In this respect, the
case (α = 7, β = 7). It is observed that this effect is simplified formulae to estimate the average ground-
considerable. Particularly, the horizontal grounding ing forces (vertical and horizontal components) can
force of (α = 1, β = 7) is highest in correspondence be derived. For the sake of simplicity, only circular
of the bulkhead (Fig. 6a). This result is due to the paraboloid is considered, i.e. a = b in Equation 1.
fact that obstructions with larger size in the trans- The resultant vertical force under vertical punch-
verse direction (y-axis) oppose more resistance in ing which is induced by the uniform pressure p is
the horizontal direction (x-axis). Considering the computed as (refer to (Johnson 1986)):
vertical component, (α = 7, β = 7) features the high-
est forces (Fig. 6b). This is due to the fact that the
Fv
vertical force component depends on the projected Fv π δ a 2 p or p = (2)
contact area which is normal to the z-axis. π δ a2
376
100 simplified FX
sliding FZ
80
Grounding force [MN]
simplified FZ
simplified FX
80 simplified FZ
60
60
40
40
20
20
0
0 5 10 15 20
0 Grounding distance [m]
0 5 10 15 20
Grounding distance [m]
Figure 8b. Resistance—penetration and the grounding
forces for (α = 10, β = 10), δ = 1.0 m and μ = 0.6.
Figure 7a. Resistance—penetration and the grounding
forces for (α = 10, β = 10), μ = 0 and δ = 1.0 m.
simplified FX
140 the total horizontal grounding force:
simplified FZ
120
Fxtot = Fx + μFz (4)
100
0
0 5 10 15 20 3 DYNAMIC GROUNDING
Grounding distance [m]
Figure 7b. Resistance—penetration and the grounding The tanker previously illustrated is used also for the
forces for (α = 10, β = 10), μ = 0 and δ = 2.0 m. analysis of dynamic grounding. At the present state,
377
FZ - static
35
It is realized that the impact force at bow is
difficult to estimate because the formula obtained 30
in Equation 3 is for the area of the cargo tank, and 25
not for bow area. This leads to an unreasonable
20
large impact force at bow due to the large contact
area. In order to circumvent this, a scale factor is 15
essential to scale down the impact force at bow 10
during the first passage of impact. In this respect,
5
it is reasonable to assume that the impact load
is increasing linearly with time. Following Alsos 0
0 50 100 150 200 250
(2008), the excitation force is scaled down when the Grounding distance [m]
width of obstruction (BI) is greater than the dam-
age width (BD) as expressed in Equation 6. Figure 9a. Grounding forces for (α = 7, β = 7) at
δ = 1.0 m and speed of 12 knots.
⎧ BD ( x )
⎪ F ( x, ) BD ( x,
x, ) < BI ( )
Fg ( x,δ ) = ⎨ BI ( ) g (6) 90
⎪F ( x, ) otherwise Fx dynamic
⎩ g 80 Fz dynamic
Fx static
70
Grounding force [MN]
50
4.1 Grounding forces
40
The horizontal and vertical components of the
grounding force are computed for two obstruc- 30
378
40
0.8
20
0.6
0
0.4
–20
0.2 –40
0 –60
0 50 100 150 200 250 0 50 100 150 200 250
Grounding distance [m] Distance from AP [m]
Figure 10a. Penetration of ship bottom, δ = 1.0 m and Figure 11a. Envelopes of shear force at δ = 1.0 m and
speed of 12 knots. speed of 12 knots.
1.4
12 knots 5000
15 knots Still water requirement
1.2 20 knots 4000 (7,7)
(10,10)
3000 (20,20)
Bending moment [MNm]
1
Penetration [m]
2000
0.8
1000
0.6 0
–1000
0.4
–2000
0.2
–3000
0 –4000
0 50 100 150 200 250 0 50 100 150 200 250
Grounding distance[m] Distance from AP [m]
Figure 10b. Penetration of ship bottom for (α = 20, Figure 11b. Envelopes of bending moment at δ = 1.0 m
β = 20), δ = 1.0 m and speeds of 12, 15 and 20 knots. and speed of 12 knots.
379
3000 (20,20)
the residual strength of the damaged hull girder
2000
should be reassessed in case of grounding on a
1000 very large shoal, in order to ensure that the ship
0 can fulfill the safety requirement of seaworthiness.
–1000
–2000
ACKNOWLEDGMENTS
–3000
5 CONCLUSIONS REFERENCES
In this study a generalized seabed topology is used Alsos, H.S. & Amdahl, J. 2007. On the resistance of
to simulate the different grounding scenarios. The tanker bottom structures during stranding. Marine
results revealed that the seabed geometry is a key Structures 20, 218–237.
parameter involved in ship grounding. Large shoals DNV. 2009. Part 3- Chapter 1- Hull structural design
give a high resistance to penetration. The dynamic ships with length 100 metres and above. Det Norske
grounding was considered by means of nonlinear Veritas.
finite element analysis with various seabed topolo- EMSA. 2009. Maritime accident review. European Mari-
gies, friction coefficients, “initial” levels of penetra- time Safety Agency. Available for download from
www.emsa.europa.eu.
tion over the keel line and ship speeds. The horizontal Faltinsen, O.M. 1999. Sea loads on ships and offshore
grounding force is much affected by friction while structures. Cambridge University Press, Cambridge,
the vertical component is not. The grounding forces United Kingdom.
are increasing virtually twice as the “initial” level of Hong, L. & Amdahl, J. 2008. Plastic mechanism analysis
penetration is doubled. The effect of asymmetric of the grounding resistance of ship bottom longitudinal
seabeds is also considerable, especially when the girders. Ships and Offshore Structures 3, 159–171.
obstruction has a larger extent in the direction per- Johnson, D. 1986. Advanced structural mechanics. William
pendicular to the grounding direction of the ship. Collins Son & Co. Ltd, London.
Moreover, it is realized that there are uncertainties Nguyen, T.H., Garrè, L. Amdahl, J. & Leira, B.J. 2010.
Benchmark study on the assessment of ship damage
related to the actual seabeds which are arbitrary conditions during stranding. Accepted in Ships and
and unknown. More studies are therefore essential Offshore Structures.
in order to describe the effect of actual seabed on Paik, J.K. & Wierzbicki, T. 1997. A benchmark study on
ship grounding correctly. crushing and cutting of plated structures. Journal of
The simplified formulae to estimate the average Ship Research 41, 147–160.
grounding forces are proposed for paraboloid Pedersen, P.T. 1997. Ship grounding and hull girder
seabeds. The results of the simplified formulae strength. Marine Structures 44, 140–154.
show reasonable agreement with the results from Samuelides, M.S., Amdahl, J. & Dow, R. 2007. Studies on
FE simulations. It is still compromising to use the behavior of bottom structures during grounding:
Advancements in Marine Structures, C. Guedes Soares
these formulae for quick estimate of the average and P. K. Das (Eds), Taylor & Francis Group; London,
grounding forces for each seabed topology. The U.K: pp. 273–282.
difference between static and dynamic grounding Simonsen, B.C. 1997. Mechanics of ship grounding. Ph. D.
is significant. The envelopes of shear force and thesis, Department of Naval Architecture and Offshore
bending moment induced by each shape of seabed Engineering, Technical University of Denmark (DTU).
380
ABSTRACT: The authors are currently engaged in a collaborative research project which—amongst
others—deals with the comparison of different bulbous bows and side structures. Their corresponding
effects on collision will be investigated. This paper will give a short overview of the background and start-
ing phase of the project. It will focus on developing one determined test model geometry of a bulbous
bow for the next two prospective collision experiments. This geometry remains unchanged for a rigid
bulbous bow as well as for a deformable bulbous bow structure. The requirement of the designed bulbous
bow is to be able to create a significant damage of the ship side structure and itself. This pre-experimental
study shows that minors parametrical changes of bulbous bow geometry will significantly change the
collision behaviour. For this study the finite element method has been used.
381
382
Stress [N/mm2]
600
500
400
300
200 Existing Curve
100
0
Figure 3. Test-plant, tensile-test. 0.00 0.20 0.40 0.60 0.80 1.00
Strain[-]
350
Figure 5. Power law hardening based material relation
300 of existing and scaled curve.
250
Stress [N/mm2]
383
⎧ σH 1
⎪ 0 for ≤−
σ eq 3
⎪
⎪ 2
⎪ σH ⎛σH ⎞
⎪ 1+ σ 12 − 27 ⎜ ⎟
⎛σ ⎞ ⎪ eq ⎝ σ eq
e ⎠ 1 σ 1
f⎜ H⎟ = ⎨ 2 for
o − < H <
⎝ σ eq ⎠ RTCL 3 σ eq 3
⎪
2
σH ⎛σH ⎞
⎪ 3 σ + 12 − 27 ⎜ σ ⎟
⎪ eq ⎝ eq ⎠
⎪
⎪ 1 exp ⎛ 3σ H ⎞ for
σH 1
≥
⎪ 1.65 ⎜ 2σ ⎟ σ eq 3
⎩ ⎝ eq ⎠
384
2.5
/ =1 +
2.0
/
1.5
1.1
1.0
/ =1 +
0.5
0.0
Experiments
Mild steel: High tensile steel:
= 189.6 Mpa = 522.9 MPa
Manjoine (1944) Fujii et al. (1985)
= 217.5 MPa = 373.2 Mpa
Campbell & Cooper (1966) Toyosada et al. (1987)
= 283.0 MPa = 313.8 Mpa
Paik et al. (1999) Toyosada & Goto (1991)
4 NUMERICAL SIMULATION
10
Asymptote
4.1 General
5 This investigation concentrates on the worst
case for the struck vessel. This includes that the
impact is located in the middle of two web frames
0 of the side structure with a collision angle of
0.0 4.0·10-4 . 8.0·10-4 1.2·10-3 90 degrees.
ε [sec-1] The examination will distinguish between three
sizes of bulbous bow cross sections. The small-
Figure 8. Strain rate of point A moving with a constant
velocity of 0.5 mm/sec.
est bulbous bow will not hit the web frames, the
intermediate bulbous bow will touch the web
frame and the largest bulbous bow will totally
plate of length L = aR/2 will elongate as shown in collide with the web frames. The initial designs of
Figure 8. the investigated geometries are basically axisym-
It is easy to see that the strain rate will not super- metric. The cross section will have an increasing
sede the asymptotic value of 1.22 ⋅ 10−3 sec−1. elliptical shape along the global vertical axis of
The outer shell of the side structure fails at a the ship.
displacement of 400 mm by using the evaluated
bulbous bow. The simplified model has shown a
4.2 Definition of the local coordinate system
strain rate value of 5.20 ⋅ 10−4 sec−1 at a displace-
ment of 400 mm. Considering Figure 9 it is obvi- For the simulation of the different geometries of
ous that both values are located near the yield the bulbous bows the axes of the coordinate system
stress ratio 1. The result of this simplified geomet- are defined as shown in the Figures 10 and 11.
ric model presents a comparable small influence As shown in Figure 10 the x-axis is located par-
of the strain rate at a velocity of 0.5 mm/sec as allel to the longitudinal girder in the middle of the
shown in the investigation made by Ehlers et al. test-model; the y-axis proceeds from the middle of
(2010). Hence the strain rate will be neglected for the side structure to the longitudinal girder and
the calculations. the z-axis is the vertical one which points into the
In reality, the collision velocity will be much bulbous bow’s penetration direction (Fig. 11). The
larger. Considering additional scale effects, the web frame spacing aR is 800 mm.
385
Longitudinal Girder
Connection Plates
Area of
Impact X-Z Plane Y-Z Plane
Side Structure End Section End Section
1800 [mm]
aR aR aR/2
Z
Figure 12. Investigated bulbous bows, left: Geometries
Bulbous for α1/βi = 1.00; right: Geometries for β1/aR = 0.80, see
Bow also Table 2.
Connecting
Rod Table 1. Definition of the parameters α and β [mm]
Side Structure describing the end section of the bulbous bows.
Support
aR aR aR/2
X α1/βi = 1.00 α2/βi = 1.50 α3/βi = 2.00
386
Longitudinal 2500
Bulkhead 11
2000 12
Reaction Force [kN]
Stringer 13
1500
Web 500
Frame
0
0 250 500 750 1000 1250 1500
Connection Rods Bulb Profile
Tension Rods Displacement [mm]
Figure 13. Finite element model of geometry 43 (1/4- Figure 14. Comparison of the reaction force of the three
model). geometries of the upper row of Table 2 (β1/aR = 0.80).
387
3500 3000
x 41 β2 = 0.816 [m]
3000 2500
2500 2000
2000 1500 β3 = 1.000 [m]
1500 1000
1000 500 β4 = 1.200 [m]
500 0
0 0.4 1.4 2.4
0 250 500 750 1000 1250 1500 Parameter αj [m]
Displacement [mm]
Figure 16. Calculated reaction force of the geometries
Figure 15. Comparison of reaction force of the four
at the first load peak.
geometries of the left column of Table 2 (α1/βi = 1.00).
4500
Permissible Collision Force β1= 0.640 [m] EQ
Table 3. Exclusion of the geometries exceeding the 4000
maximum load capacity of the test-plant. β1= 0.640 [m] DY
Reaction Force [kN] 3500
β2 = 0.816 [m] EQ
α1/βi = 1.00 α2/βi = 1.50 α3/βi = 2.00 3000
β2 = 0.816 [m] DY
2500
β1/aR = 0.80 11 12 13 β3 = 1.000 [m] EQ
2000
β2/aR = 1.02 21 22 23 β3 = 1.000 [m] DY
1500
β3/aR = 1.25 31 32 33
β4 = 1.200 [m] EQ
β4/aR = 1.50 41 42 43 1000
β4 = 1.200 [m] DY
500
0.5 1.5 2.5
Parameter αj [m]
In contrast to Figure 14 the penetration of Figure 17. Comparison of the numerical and approxi-
the outer and inner hull is influenced by the mated reaction force at the first load peak.
growing cross section. This leads to a later fail-
ure of both hulls. For example the geometry with
α1 = β1 = 640 mm destroys the outer hull of the side amounts to 3.8% for the geometries of small
structure at a displacement of about 250 mm, the breadth.
one with α1 = β4 = 1,200 mm by more than twice of A similar relationship of the parameters is found
the above mentioned displacement. Furthermore for the second peak (Fig. 18), approximated by the
the geometry 41 will exceed the maximum load following equation:
capacity of the test-plant as well as geometries 42,
43 and 33 do. These geometries are excluded in FΙΙ = 2417β 2 − 1735β − 822
Table 3. + (2099 − 1044β)(α − β ) (6)
Due to the above mentioned influence of the
bulbous bow geometry one can assume that there Here the maximum deviation is 4.7% between
might be a relationship between the parameters of the numerical value and the value given by equa-
the bulbous bow end section and the load peak. tion 6 for the reaction force of the second load
For this reason the value of the geometries are peak. The graphs show a decreasing gradient with
plotted to Figure 16, connecting the values with an increasing β. An explanation for this phenom-
the same β. enon can be given by the numerical simulation.
Obviously all graphs have an increasing gradi- The damage of the side structure caused by bul-
ent. Therefore an equation has been developed bous bows with large α and β values includes a
which offers the possibility to describe the influ- significant area of the outer shell of the side struc-
ence of the geometry without a simulation: ture by loosing its structural resistance against the
penetrating bulbous bow (Fig. 19).
FΙ = 3387β 2 − 1986β + 1087
+ (328β − 83)(α − β ) (5) 5.3 Consideration of deformable bulbous bows
The regression in equation 5 shows a good with As mentioned above eight remaining geometries
numerical results (Fig. 17). The highest deviation have to be investigated. The aim is to find the case
388
389
390
K. Tabri
Department of Mechanics, Tallinn University of Technology, Estonia
J. Broekhuijsen
Damen Schelde Naval Shipbuilding, The Netherlands
ABSTRACT: The paper studies the influence of coupling between the ship motions and the structural
resistance in the prediction of ship collision damage. Several collision scenarios are simulated numerically
using coupled and decoupled approaches. A coupled approach implies a time-domain simulation, where a
precise description of the whole collision process together with the full time histories of the motions and
forces involved, is achieved. There, the ships’ motions are evaluated in parallel to the structural deforma-
tions i.e. the coupling between external dynamic and inner mechanics is preserved. A decoupled approach
is based on the conservation of momentum and allows faster estimation of the deformation energy with-
out providing exact ship motions. The energy to be absorbed in collision is evaluated based on the ship
masses and velocities, and in predicting this energy there is no coupling from the structural behavior. The
ensuing deformations are evaluated in so-called displacement controlled manner—the struck ship is kept
motionless and the striking ship penetrates it along the direction of its initial velocity. The comparison of
the outcomes of two methods reveals that these two methods result in significantly different predictions
of both the penetration path and the deformation energy required to breach the inner hull.
391
392
393
394
395
Figure 6. Damage length in the outer plating at the Figure 7. Difference in deformation energy as a func-
moment of breaching of the inner hull. tions collision angle.
396
ACKNOWLEDGEMENTS
REFERENCES
397
ABSTRACT: The paper presents finite element simulations examining the dynamic response of small
rectangular panels struck transversely by a mass. Three panel widths with different stiffener distribution
are considered in the analysis. The influence of the impact velocity and impact location on the final results
is discussed. The material true stress-strain curve, used in the numerical model, is obtained from tensile
tests on the actual material used to fabricate the panels. The mesh size, the true material curve definition
and the critical failure strain estimation are evaluated from the numerical simulation of the tensile test.
A brief study of the definition of the boundary conditions is conducted, in order to determine the
simplest, but accurate assumption that will be used in the finite element model. The force-displacement
curves and the shape of the deformation from the impact simulations are used to compare the plastic
response of the impacted panels.
399
400
401
Stress [MPa]
in press). Engineering
The selected material from the library of Curve
300
LS-DYNA ‘Mat.024-Piecewice Lineal Plasticity’
defines the plate and the stiffeners, and provides a 200 Steel ASTM A36
Tensile strength 400-550MPa
definition of the true stress-strain curve as an offset Yield point, min 250MPa
100
table (for details, see Section 4). The striking mass Elongation in 200 mm, min 20%
was modelled as a rigid undeformable material.
For this propose, ‘Mat.020-Rigid’ was selected 0.00 0.05 0.10 0.15 0.20 0.25 0.30
Strain [-]
from the material library of LS-DYNA assigning
mild steel mechanical properties. The mechanical
Figure 4. Engineering and true stress-strain curves.
properties are used for determining the sliding
interface parameters in the contact definition
when the rigid body interacts. It is necessary to
give realistic values of these properties, since unre- simulations of the tensile test is summarized as
alistic values may contribute to numerical prob- follows (Villavicencio et al. in press): Only the
lems in the contact definition (Hallquist 2005). length between the clamping edges of the tensile
Since the falling weight assembly was modelled as test specimen was modelled (Fig. 5). For initiating
a simple shell elements semi-sphere, an artificially necking, the width of the specimen at the centre
large density was used to give the same mass as the was gradually reduced by 0.1 mm. Shell elements
one to use in the experiments. with default LS-DYNA formulation were used.
The strain-rate sensitivity behaviour of mate- Default hourglass control was specified. How-
rials in the FE model can be included using the ever, the mesh was designed to avoid hourglassing
coefficients of the Cowper-Simonds constitutive by orientating the elements diagonally (Fig. 5). The
equation (Jones 1989). However, the theoretical translational degrees of freedom were restricted at
behaviour of strain-rate-sensitivity structures one end and at the other end a constant displace-
summarized by Jones (1989) shows that the mate- ment of 100 times the experimental speed was pre-
rial strain rate had a small effect at relatively low scribed (Ehlers and Varsta 2009). The force of the
incident energies when structures such as beams displaced nodes at the free end is obtained from the
and plates are subjected to uniform impulsive calculations. This force, plotted versus the applied
velocities. Consequently, the strain rate sensitivity prescribed displacement, gives the engineering
was omitted from the finite element model. stress-strain behaviour.
Two mesh sizes of 5.0 and 2.0 mm (denoted by
Shell5 and Shell2 respectively) were analyzed, and
4 MATERIAL AND MESH SIZE two true stress-strain curves were used to define
DEFINITION the material. The first true material curve is the one
proposed by Zhang et al. (2004), where the true
The mechanical properties of the material used in stress σt and the true strain εt are expressed by:
the FE models were obtained from the ‘Standard
Specification for Carbon Structural Steel ASTM σt C εt n (1)
A36’ (ASTM Standards 1989). The engineering
where
stress-strain curve is plotted in Figure 4.
Since collision calculations involve extreme n = 1n(1 + Ag) (2)
structural behaviour with both geometrical and
nonlinear effects, the input of material properties and
up to the ultimate tensile stress has a significant
influence on the extent of critical deformation C = Rm (e/n)n (3)
energy. Thus, it is generally recommended to
use true stress-strain relationship, which can be Ag is the maximal uniform strain related to
obtained from a tensile test. the ultimate tensile stress Rm and e is the natural
The experimental tensile test (Fig. 4) was mod- logarithmic constant. If only the ultimate stress
elled in LS-DYNA, in order to verify the plastic Rm (MPa) is available, the following approximation
response of the material and to explore the stress can be used to obtain the proper Ag:
and strain distribution in the neck at failure.
The general procedure to perform the numerical Ag = 1/(0.24 + 0.01395Rm) (4)
402
403
120
Shell2_Support Plates
100 Shell2_Bolt Constrained
Shell2_Boundary Conditions
80
Force [kN]
60
40
20
0
0 5 10 15 20 25 30 35
Displacement [mm]
Figure 7. Models of boundary conditions: a) support Figure 8. Boundary condition analysis: Panel A1; impact
plates; b) bolt constrained; c) boundary conditions. at mid-span; initial velocity 6.0 m/s.
404
140
V6
120
Panel A2
100
Force [kN]
80 V4
Panel B2
60
V2 V6
40
V4
20
V2
0
0 10 20 30 40
Displacement [mm]
405
406
60 C5 C4
differences are observed, for example Panel C4
40 reaches 35.9 and 27.8 mm at Locations 1 and
20
3, respectively. Figure 16 represent the time
necessary to absorb the initial impact energy.
0 It is observed that, the impact response is shorter
0 10 20 30 40 when the impact is above the stiffener (Location
a) Displacement [mm]
100
1 and 3). However, when the impact is in the
B4_2
B5 unstiffened zone of the panel (Location 2) the
B5_2
80 C4_2
B4 impact event takes more time.
C5_2
Force [kN]
60
40
20 C5
C4
0
0 10 20 30 40
b) Displacement [mm]
100 B5 B4
B4_3
B5_3 C4
80 C4_3
C5_3 C5
Force [kN]
60
40
20
0
0 10 20 30 40
c) Displacement [mm]
407
1.0
B5_1
0.8 B4_2
Impact Point 2 B5_2
0.6 B4_3
ACKNOWLEDGEMENTS
0.4 B5_3
The first author has been financed by the Portuguese
0.2
Foundation for Science and Technology, under
0.0 contract SFRH/BD/46369/2008.
0 2 4 6 8 10 12 14 This work contributes to the activities of
time [ms]
MARSTRUCT VIRTUAL INSTITUTE, (www.
Figure 16. Absorbed energy with time: Panels B4 marstruct-vi.com) in particular its Technical
and B5; impact at Locations 1, 2 and 3; initial velocity Subcommittee 2.5 on Crashworthiness and Impact
6.0 m/s. Strength.
7 CONCLUSIONS REFERENCES
Detailed information of the impact response of Alsos HS, Amdahl J, Hopperstad O. 2009a. On the
rectangular panels has been obtained by non-linear resistance of stiffened plates, Part I: Experiments. Inter-
explicit dynamic simulations. The results give a national Journal of Impact Engineering. 36: 799–807.
good understanding of the force-displacement Alsos HS, Amdahl J, Hopperstad O. 2009b. On the resist-
response and shape of the deformation in the ance of stiffened plates, Part II: Numerical analysis.
impacted panels. International Journal of Impact Engineering. 36:
875–887.
The performed numerical simulations of the ten- ASTM (American Society for Testing and Materials).
sile test defined well the mesh size, the true stress- 1989. Standard Specification for Carbon Structural
strain curve and the critical failure strain which Steel.
were subsequently applied in the panel FE model. Bedair OK. 1997. Analysis of stiffened plates under lateral
The definition of the true stress-strain curve has loading using sequential quadratic programming
a small influence on the results when only plastic (SQP). Computer & Structures. 62 (1): 63–80.
deformation is observed. However, when fracture Boot JC, Moore DB. 1988. Stiffened plates subjected to
occurs, an accurate definition and its validation by transverse loading. International Journal of Solid and
simulation of tensile tests are necessary. The criti- Structures. 24 (1): 89–104.
Caridis PA, Samuelides E, Frieze PA. 1994. On the
cal failure strain proposed by Zhang et al. (2004) dynamic response of ship plating under lateral impact.
gives good agreement with the tensile test simula- International Journal of Impact Engineering. 15 (2):
tions of the steel material. 149–164.
The representation of the boundary conditions Cho S-R, Lee H-S. 2009. Experimental and analytical
by nodal restriction at the supported perimeter investigations on the response of stiffened plates
gives similar results to the more real representation subjected lo lateral collisions. Marine Structures. 22:
with support plates and longitudinal restriction of 84–95.
the bolts. Deb A, Bootom M. 1988. Finite element models for stiff-
The maximum transverse displacement is larger ened plates under transverse loading. Computer &
Structures. 28 (3): 361–372.
than the permanent transverse displacement in Dieter GE. 1986. Mechanical behavior under tensile and
wider panels. The difference decreases in shorter compressive loads. ASM Handbook. 8: 99–10.
panels. Ehlers S, Broekhuijsen J, Alsos HS, Biehl F, Tabri K.
The local indentation plays an important role in 2008. Simulating the collision response of ship
the total deformation of the stiffened plate, being side structures: A failure criteria benchmark study.
more than one-quarter of the total deformation. International Shipbuilding Progress. 55: 127–144.
For impacts at one-quarter of the support, Ehlers S, Varsta P. 2009. Strain and stress relation for
the stiffener type has a small influence on the non-linear finite element simulations. Thin-Walled
force-displacement response, showing similar Structures. 47 (11): 1203–1217.
Ehlers S. 2010. Strain and stress relation relation until
deformation. fracture for finite element simulations of a thin circu-
The reinforcement of the free edges of the panel lar plate. Thin-Walled Structures. 48 (1): 1–8.
has a small influence on the plastic response. It is Hagiwara K, Takanabe H, Kawano H. 1983. A proposed
observed that, the maximum transverse displace- method of predicting ship collision damage. Interna-
ments decrease by only few millimetres. tional Journal of Impact Engineering. 1 (3): 257–279.
408
409
ABSTRACT: The paper presents a method to estimate the displacement of the neutral axis of a damaged
ship subjected to bending moment. The method uses the load-end shortening relationship obtained from
previous numerical double bottom residual strength calculations establishing force equilibrium over the
whole transverse section of the ship. The set of estimated neutral axes allows the definition of a new load
condition to induce a higher bending moment on the double bottom structure. The new residual strength
is compared with previous numerical calculations for the intact and minor damaged double bottom in
which the bending moment was induced considering the neutral axis fixed during the entire simulation.
The comparison showed a reduction in the capacity of the double bottom structure when the bending
moment is induced by the new load condition.
411
412
100
50 "inner"
"outer"
Figure 2. Finite element model (fore floor is hidden). 0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
mean strain [%]
defined initial imperfections corresponded to fabri-
cation tolerances and the shape was defined using Figure 3. Ultimate strength of double bottom. Previous
intact and damaged cases.
sinusoidal expressions.
The true stress-strain relationship of the mate-
rial was defined by an isotropic plasticity model axis of the double bottom (1340 at BL; BL is the
using a power law hardening rule (‘Mat.18 Power base line) denoted in this study by “outer” and
Law Plasticity’ in LS-DYNA). The inputs of the “inner” elements, respectively. Due to the rotation
material were: yield stress 275 MPa, strength coef- applied on the fore and aft floor to induce a bend-
ficient 740 and hardening exponent 0.24; strain ing moment, the inner and outer elements buckled
rate effects were ignored. Since only large plastic at different induced strains, see Figure 3.
deformations were studied, failure was not required The damage on the bottom panel reduced its
in the material definition. capacity to sustain a compressive load. This reduc-
Symmetry boundary conditions were defined at tion is more evident in the outer than in the inner
the center line and the fore and aft floors were fully elements. The post-buckling response of the inner
clamped. Both boundary conditions were prop- elements is less influenced by the indentation of the
erly modified to induce a bending moment. As central bottom panel. The intact ultimate strength
the tanker was considered in full load condition, of the double bottom occurred when the outer ele-
water and oil pressure were applied on the bottom ments buckled, whereas in the damaged case it was
and tank top, respectively. In the intact case the observed at the buckling of the inner elements.
sequence of the loads was pressure and a bending
moment whereas in the damaged case the sequence
was pressure, indentation damage and a bending 3 NEUTRAL AXIS DISPLACEMENT
moment. In both intact and damaged cases the
bending moment was induced by defining pre- The method used to estimate the displacement of
scribed displacements and rotations on the fore the neutral axis, due to the nonlinear response of
and aft floors. a ship subjected to bending moment, is based on
The hogging bending moment applied on the the criterion provided by IACS (2006). According
double bottom structure induced axial strains and to this criterion, the neutral axis is determined
stresses which were represented by “mean strains” by establishing force equilibrium over the whole
and “resistance forces”. The “mean strain” was transverse section of the ship.
obtained by the ratio between the longitudinal The double bottom was idealized for its repre-
displacement of a node located in the central fore sentation in the “previous” finite element model.
floor and the span of the stiffeners. The “resistance The structure of the ship above the tank top was
forces” were obtained from elements located for- used to estimate the neutral axis and the stress-
ward the central aft floor. It must be noted that the strain distribution of the structural elements.
resistance forces were given for only one side (star- Some simplifications of the structure above the
board) of the double bottom, which represented all tank top were adopted. However, the scantlings
the thickness of the centre vertical keel. of the plates and the longitudinal were the same
Figure 3 shows the “resistance force-mean strain” as the ones used in the original design. Figure 4-a
relationships of the intact and damaged cases. The shows this simplified structure where the phantom
selected indenter to induce damage has triangular lines represent the camber of the upper deck and
shape, the depth is 150 mm and causes damage the diagonal inner plating of the bilge tank used in
to five stiffeners located between the second and the original design. In the simplification, they were
third side girder. The “total” resistance force is a considered parallel to the base and the centre line,
summation of the resistance force of the structural respectively. The midship section above the tank
elements located below and above the local neutral top was divided into ordinary stiffener elements
413
414
5.0
3 0.7
1 4
4.0 5 0.6 2
6 0.5
6
3.0 0.4 5
0.3 1 3 4
2.0
0.2
1.0 0.1 Mean Strain [%]
0.0
0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Mean Strain [%]
Figure 9. Estimation of the neutral axis displacement.
Figure 8. Bending moment-mean strain response for
the set of neutral axes.
415
416
200 031406.
150
100
REFERENCES
50
Alsos HS, & Amdahl J. 2007. On the resistance of
0 tanker bottom structures during stranding. Marine
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 Structures; 20: 218–237.
-3
Curvature*10 Amlashi H, & Moan T. 2008. Ultimate strength analysis
of a bulk carrier hull girder under alternate hold load-
Figure 13. Load-end shortening curves. Damaged case. ing condition—A case study Part 1: Nonlinear finite
element modelling and ultimate hull girder capacity.
Marine Structures; 21: 327–352.
new damaged simulations” (Fig. 13) are similar to Bai Y, Bendiksen E, & Pedersen PT. 1993. Collapse anal-
those observed for the intact case. In particular, the ysis of ship hulls. Marine Structures; 6: 485–507.
buckling at the outer and inner elements is slightly Gordo JM, Guedes Soares C, & Faulkner D. 1996.
overestimated by the new models and the distance Approximate assessment of the ultimate longitudinal
between the buckling peaks is shortening. strength of the hull girder. Journal of Ship Research;
4 (1): 60–69.
Guedes Soares C, & Gordo JM. 1997. Design methods
for stiffened plates under predominantly uniaxial
5 CONCLUSIONS compression. Marine Structures; 10 (6): 465–497.
Guedes Soares C, Luís RM, Teixeira AP, Quesnel T,
A method to estimate the displacement of the neu- Nikolov PI, Steen E, Khan IA, Toderan C, Olaru VD,
tral axis of a damaged ship subjected to bending Bollero A, & Taczala M. 2008a. Parametric Study
moment is described. The method uses the load- on the Collapse Strength of Rectangular Plates with
end shortening relationship obtained from previ- Localized Imperfections under Inplane Compression.
International Shipbuilding Progress; 55: 63–85.
ous numerical residual strength calculations of
Guedes Soares C, Luís RM, Nikolov P, Downes J,
a double bottom establishing force equilibrium Taczala M, Modiga M, Quesnel T, Toderan C, &
over the whole transverse section of the ship. The Samuelides M. 2008b. Benchmark study on the use
method allows the definition of a new load con- of simplified structural codes to predict the ulti-
dition to induce a higher bending moment on the mate strength of a damaged ship hull. International
double bottom structure. Shipbuilding Progress; 55: 87–107.
The ultimate strength of the structural elements Hallquist JO. 2005. LS-DYNA Keyword user’s man-
below and above the local neutral axis of the dou- ual, Version 971. Livermore Software Technology
ble bottom structure are slightly overestimated Corporation, Livermore, California.
IACS (International Association of Classification
in the new calculations, demonstrating that the
Societies). 2006. Common Structural Rules for Double
double bottom structure is subjected to a higher Hull Oil Tankers.
compressive load at smaller induced strains, and, Liu Z, & Amdahl J, unpub. Numerical and simplified
consequently experiences smaller contribution of methods for analysis of the residual strength of ship
induced bending moment. double bottoms.
The most important difference between the pre- Luís RM, Hussein AW, & Guedes Soares C. 2007. On
vious and the new finite element calculations is the the Effect of Damage to the Ultimate Longitudi-
shortening in the distance between the buckling of nal Strength of Double Hull Tankers. Proc. 10th
the elements below and above the local neutral axis Int. Symp. on Practical Design of Ships and Other
Floating Structures (PRADS’07); ABS, Houston,
of the double bottom (outer and inner elements).
Texas, USA.
This difference is due to the increment in the rate Naar H, Kujala P, Simonsen BC, & Ludolphy H. 2002.
of the compressive load. The conclusions are valid Comparison of the crashworthiness of various bot-
for intact and damaged cases. tom and side structures. Marine Structures; 15:
443–460.
Paik JK, & Kim BJ. 2002. Ultimate strength formula-
tions for stiffened panels under combined axial load,
ACKNOWLEDGEMENTS in-plane bending and lateral pressure: a benchmark
study. Thin-Walled Structures; 40 (1): 45–83.
This work has been performed within the project Pedersen P. 1994. Ship grounding and hull-girder
FLAGSHIP (European Framework for Safe, strength. Marine Structures; 7: 1–29.
417
418
ABSTRACT: The paper presents numerical simulations of previously reported drop weight impact tests
examining the dynamic response of fully clamped steel circular plates struck transversely at the centre
by a mass with a spherical indenter. The impact velocity varied from 0.5 to 6.0 m/s. The plates showed
no visible damage at the very low incident energies, but suffered both permanent indentation and global
deformation as incident energy was increased. The numerical modelling was performed using the LS-
DYNA nonlinear, dynamic finite element software and is based on a previous finite element analysis
of aluminium plates under lateral impact. The numerical calculations used can accurately predict the
response of deflections, forces and absorbed energies, using both shell and solid element types. However,
solid elements were required to obtain a satisfactorily accurate prediction of the deformed shape includ-
ing the indentation into the thickness. The steel plastic response is compared and discussed with that of
similar aluminium plates.
419
Stress [MPa]
400
350MPa
analyzed using both shell and solid elements types 365MPa
and different mesh sizes, and additional com- 300
ments on the stress and strain distributions, and 230MPa
200
on the energy partitioning, are then given. Finally,
the behaviour of the current steel plates is com- 100
pared to those of the aluminium plates studied in
0
Villavicencio et al. (2010).
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40
Strain [-]
3 NUMERICAL MODEL
Figure 1. Clamped circular plate struck laterally at the The computations were carried out using the
centre by a mass. finite element package LS-DYNA Version 971
(Hallquist 2005) which is appropriate for non-
linear explicit dynamic simulations with large
deformations. The FE model is based on the
numerical analysis on circular aluminium plates
conducted by Villavicencio et al. 2010, where a
sensitivity analysis of the mesh size and the model-
ling of support plates to represent the experimen-
tal boundary condition were completed. In the
present section the main characteristics of the FE
model and improvements with respect to the previ-
ous model are summarized (Fig. 4).
The specimen plates were modelled with 4-node
Figure 2. Experimental clamped condition. shell elements with 4-integration points through the
420
15
strain εt are expressed in terms of the engineering
stress σe and engineering strain εe (obtained from 10
the tensile tests, and shown in Fig. 3) by:
5
σt = σe (εe + 1) (1)
0
εt = 1n (εe + 1) (2)
0 1 2 3 4 5 6
Displacement [mm]
Since for the impact tests considered here only
plastic deformation was observed, failure strain Figure 5. Force-displacement response. Thick plate
was not required for the material definition. (4.0 mm) impacted at 5.9 m/s.
421
Force [kN]
Force [kN]
15
0.50
10
0.25
5
Figure 6. Force-displacement response thick steel plates. Figure 8. Force-displacement response. Thin plate
Experimental (dashed line); Numerical (continuous line). (1.4 mm) impacted at 0.6 m/s.
6
i. Due to the inertia of the plate, at the initiation
2.55m/s of contact the impact mass and plate vibrate
4
around the contact stiffness as the plate begins
0.60m/s
2 to move.
ii. Since in the experiments the impact mass is not
0 a sphere, but a hemispherical-ended bar con-
0 2 4 6 8 10 12 nected via a load cell to a large mass, vibrations
Displacement [mm] may be set up in this (and/or other) part(s) of
the impact machine.
Figure 7. Force-displacement response thin steel plates.
Experimental (dashed line); Numerical (continuous line). Phenomena (i) would be apparent in both
experimental and numerical responses. However,
phenomena (ii) would only be apparent in the
velocities considered for the thick and thin plates experimental response.
respectively. Good agreement between numerical The fact that vibrations of the same frequency
and experimental results is seen for the whole range were noted in the experimental force measure-
of impact velocities, and the overlapping parts of ments after separation of the impact mass and
the response for different velocities are coincident. plate at the end of the impact event suggest that
However, for the thin plates deflection and force vibrations in the test machine (ii) were present in
are slightly overestimated by the numerical model the experimental results. Also, the presence of (less
at the higher impact velocities (4.39 and 5.90 m/s). severe) vibrations in the numerical results suggest
For both thin and thick plates discrepancies were that oscillations due to the contact stiffness-plate
seen between the experimental and numerical inertia phenomena (i) are also present in the exper-
force-displacement responses for impacts at the imental results, hence the larger vibrations seen
lowest velocities (0.88 and 0.60 m/s respectively) experimentally.
as illustrated in Figure 8. These are thought to be Since only oscillations due to (i) are valid mate-
due to experimental measurement errors which rial impact responses, it would be pertinent to filter
become significant when compared to the very low out any experimental oscillations due to machine
forces and displacements seen at these low incident vibrations. However, in practice this filtering may
energies. well affect the impact response of interest, and it is
All of the experimental and numerical responses extremely difficult to ensure that only the machine
in Figures 5 to 8 show oscillations during the initial vibrations are filtered out. These results show
stages of the impact, which are more pronounced that the filtering and interpretation of impact test
in the experimental results, becoming more so as results is by no means a straightforward, clear-cut
the severity of the impact test was increased. These process.
vibrations are also more apparent for the thick than Since no significant increase in accuracy was
for the thin plates. However, since the oscillations seen in the results described above through using
422
12 1.4 mm
plastic deformations at higher velocities. It is sug-
Numerical
gested that this over-prediction of the experimen-
10
tal impact force could be due to the way in which
8 Experimental the local contact is modelled numerically, since this
6 4.0 mm
would only become significant in impacts where
contact damage is dominant, as is the case for the
4 thick plates at lower incident velocities. This could
2 be because the striking mass was modelled as a
rigid body, when in fact deformation at the impac-
0
tor nose could have occurred experimentally.
0 1 2 3 4 5 6 7
Impact Velocity [m/s]
423
Energy [J]
60
tic strain throughout the plate thickness at high Kinetic
Internal
impact velocities. However, at lower impact ener- 40 Sliding
gies, elements near the neutral axis did not undergo Internal+Sliding
Total
plastic strain. 20
Effective stresses were also observed to decrease
gradually from the impact point to the supports. 0
Figure 12 shows the maximum von Mises stress dis- 0 1 2 3 4
tribution on the plate lower surface when impacted time [ms]
at the highest velocity. For the thick plates only that 100 Steel 1.4 mm
part of the specimen within the annular cut-out in
the support plates is affected by the impact. How- 80
ever, for the thin plates the stresses in the clamped
Energy [J]
424
Force [kN]
6
4.3 Comparisons with aluminium impact response
4
The impact tests on the present steel plates were
part of a wider experimental study by Sutherland 2
and Guedes Soares (2009), which also considered
the aluminium plates numerically simulated in 0
0 2 4 6 8 10 12 14
Villavicencio et al. (2010). Since impact response Displacement [mm]
is a structural and not a material property, the
plate thicknesses of the different materials were Figure 14. Force-displacement response thin alumin-
chosen to give plate bending equivalence so that ium plate. Impact velocity 5.9 m/s.
valid comparisons between the different materials
were possible. Hence, the thicknesses of the alu-
minium plates were 6.0 and 2.0 mm (thick and thin similar discrepancies between the numerical and
respectively). Here, the experimental and numeri- experimental force-deflection responses were seen
cal impact responses of the steel and aluminium for the aluminium plates at the very lowest inci-
plates are compared. dent velocities as were seen for the steel specimens
The improvements included in the FE model of (Fig. 8). Again, these are thought to be due to the
the steel plates have been retrospectively considered relatively high significance of experimental errors
in the aluminium FE model, and the numerical when compared to the very small force and dis-
results of the latter updated accordingly. In the placement measurements that must be made for
new aluminium models, the shell elements have 6 these very low-energy impact events.
and 4 integration points for the 6.0 and 2.0 mm As for the steel plates, since there was seen to
thick plates, respectively. The thin plate solid model be no advantage in using the more computation-
used 4 elements through the thickness instead of ally expensive solid elements, the Shell2 model
the 2 defined in the previous model. However, for was selected to calculate the maximum deflections
the thick plates the through-thickness number and forces for all experimental impact velocities,
of elements of the solid model was maintained as shown in Figures 15 and 16. These plots show
at 6. These modifications were included in order very good, and similar, agreement between experi-
to define the same number of integration points mental and numerical results as seen for the steel
through the thickness for both shell and solid plates in Figures 9 and 10. However, the agree-
models. The time interval between outputs was ment between experimental and numerical maxi-
also decreased ten-fold. mum deflection for the thin aluminium plates seen
The modified aluminium numerical models were in Figure 15 continues up to the highest impact
again initially evaluated and optimised in terms of velocities (as opposed to the numerical overestima-
element type, using the maximum and minimum tion of maximum deflection at high impact ener-
experimental impact velocities for each plate thick- gies for the steel plates seen in Fig. 9), although the
ness in this case. However, it was assumed that, experimental curve does also dip at higher veloci-
as found previously for the aluminium plates in ties. The slight overestimation of maximum force
Villavicencio et al. (2010), and for the steel plate for thin steel plates at higher velocities (Fig. 10) is
impacts of this paper, mesh size was not signifi- also seen in Figure 16 for the aluminium tests, and
cant and hence only the Solid1 and Shell2 elements again more impact energy is irreversibly absorbed
were considered. As for the steel plates, the alumin- experimentally than numerically predicted.
ium shell and solid models give very similar force- The overestimation of maximum impact force
displacement results, and good agreement with by the numerical model seen for thick steel plates
the experimental results is seen. As an example, (Fig. 10) is seen to be even more significant in
Figure 14 gives a comparison of the experimental Figure 16 for the thick aluminium plates. Again,
and numerical results for a thin aluminium plate it is thought that this is due to the way in which the
impacted at 5.9 m/s. local impactor-plate contact/indentation is modelled.
For both thicknesses, the improvements to The inflection observed in the force-velocity curve
the aluminium numerical model gave better of the thick steel plates at around 2.5 m/s, thought
agreement between the shell and solid models, to be due to the initiation of significant global
which has been illustrated by the inclusion of plastic deformation, is also seen in that of the thick
the ‘old’ Shell2 model results in Figure 14. Very aluminium plates.
425
12 2.0 mm
impact energies elements near the neutral axis were
10 Experimental not predicted to undergo plastic strain.
8 Numerical Although generally very good agreement was
6.0 mm obtained, two discrepancies between numerical
6
and experimental were observed, that were very
4 similar for both steel and aluminium impacts:
2 i. For thin plates at higher velocities, numerical
0 predictions overestimated the experimental
0 1 2 3 4 5 6 7 maximum forces, and underestimated the energy
Impact Velocity [m/s] irreversibly absorbed. The numerical simulations
also overestimated the thin steel plate maximum
Figure 15. Maximum deflection-velocity. experimental deflection for higher velocity
impacts. A slightly stiffer response was also seen
25 experimentally. It is postulated that this could be
due to more slippage, more plastic deformation
6.0 mm
20 generated heat dissipation, or strain-rate depend-
Maximum Force [kN]
Numerical
ant effects, or a combination of these effects.
15 ii. For thick plates the maximum forces seen exper-
Experimental imentally were overestimated by the numerical
10 2.0 mm model. It is thought that this is due to the way
in which the impactor-plate local contact, and
5 plate-support contact has been modelled.
Further work would be beneficial to refine the
0
model in terms of the static coefficients of fric-
0 1 2 3 4 5 6 7
Impact Velocity [m/s] tion used, the modelling of the striking mass,
possible material property strain-rate effects, and
Figure 16. Maximum force-velocity. additional energy absorption mechanisms. Since
it is highly probable that more than one of these
phenomena could be influential, and more impor-
5 CONCLUSIONS tantly that there may well be interactions between
these effects, any further work must be carefully
Detailed information of the impact response of designed to give a numerical model optimised in
clamped steel and aluminium circular plates has terms of all such effects.
been obtained through non-linear explicit dynamic The investigation of the impact of thicker plates
simulations using the LS-DYNA software package. would also be worthwhile. These could be geomet-
The results obtained were in good agreement with rically scaled-up versions of the present plates, or
those of previous experimental tests, indicating thicker plates of similar size to the present ones. In
that relatively coarse meshes using shell elements the former case, a dimensional analysis approach
are sufficient to predict the maximum deflections to develop scaling laws, such as used with suc-
and forces for the entire plastic response of both cess in Sutherland and Guedes Soares (2007) for
steel and aluminium. laminated glass-polyester plates, could be used.
The support plates representing the experimen- This would also identify any unexpected scaling
tal boundary conditions were capable of represent- model shortcomings, or ‘size effects’ that could
ing some small longitudinal displacement of the lead to inaccurate prediction of full-scale in-service
specimen plate between the supports, especially for behaviour from laboratory scale tests.
the thin plates at higher velocities. For the latter case of plates of lower length
The numerical simulations gave a good under- to thickness ratio, a far more contact/indenta-
standing of the shape of the deformation in plates tion dominated behaviour, with little global plate
subjected to impact loads. In terms of indentation, deformation or influence of boundary conditions,
a fine meshed solid model gave a better definition would be expected. Hence, the use of refined solid
of the shape of the deformation since this was able element models would be necessary to represent
to model both the deformation around the indenter the indentation into the plate thickness. The study
and also the change in plate thickness at the impact of lower length to thickness ratio plates could also
point. The maximum plastic strain occurred on be used to approximate the behaviour expected for
the lower surface of the plate, and the presence thinner plates at, or near to, stiffeners.
426
427
ABSTRACT: The purpose of this study is to investigate residual Ultimate Longitudinal Strength (ULS)
of a panamax bulk carrier being subjected to sagging moment. ULS of panamax bulk carrier in intact
condition is estimated using large-scale nonlinear FEM analysis code LS-DYNA as well as simplified
analysis method based on the Smith method. ULS in intact condition estimated by FEA is compared
with that estimated by simplified method. It can be found that ULS estimated by FEA and that by
simplified analysis show a fairly good agreement although small discrepancy can be seen. Moreover ULS
in damaged condition is estimated assuming collision at midship region of the side shell, where damaged
area is artificially created as a first phase of the research project.
1 INTRODUCTION
Tier I: Goals
1.1 Goal Based ship construction
Standard (GBS)
A Goal Based ship construction Standards (GBS) Tier II: Functional Requirements
for oil tankers and bulk carriers was adopted as
a new SOLAS convention at the 87th session of
Tier III: Verification of conformity
Maritime Safety Committee (MSC87) in IMO
(2010a). Figure 1 shows the illustration of an
overall framework of GBS, where 5 tiers can be Tier IV:Rules and regulations for ship design
seen. Functional requirement of tier II among and construction
5 tiers are intended to allow the performance
based criteria other than the conventional pre- Tier V: Industry practices and standards
scriptive criteria.
429
430
431
Moment [GNm]
20 Reaction Moment
10
0
0.0 0.5 1.0 1.5 2.0 2.5
Time[s]
Figure 3. Collapse of the ship due to vertical bending hold for Δt = 2.0 [s] and Δt = 0.3 [s] respectively.
moment (Intact Condition).
It is confirmed by carrying out sensitivity analysis
that the effect of load increasing time on ULS is
significant, and that Mu decreases gradually as Δt
increases mainly due to reduction of dynamic effect.
Therefore at least around Δt > = 2.0 [s] might be
required to obtain quasi-static Mu as far as present
FE model and analysis conditions. Therefore in the
present study Δt = 2.0 is used hereafter. It can be
seen in Figure 7 that reaction moment increases as
the applied moment increase, and that after reach-
ing to ULS at around t = 2.0 [s] reaction moment
decrease considerably with some fluctuations.
Figure 4. Collapse of the ship due to vertical bending
moment (Intact Condition).
Figure 8 and Figure 10 show relation between
reaction moment and curvature. Assuming the
beam theory for bending of global hull girder
(See Fig. 11), curvature (1/ρ) is calculated by rotation
angles of two T.BHDs by the following formula
1 θ1 θ 2
ρ= L
(2)
432
20
ing ULS. It is obtained from the present analysis that
ULS in intact condition (Mu,intact) is 15.22 [GNm].
10
Mu,intact = 15.22 [GNm] (3)
0
0.0E+00 2.0E-04 4.0E-04 6.0E-04 8.0E-04 Moreover fully plastic bending moment (Mp)
1/ρ[1/m] at the center of the target cargo hold is manually
calculated with using excel sheet where axial
Figure 8. Relation between reaction bending moment yielding force of all members on the transverse
and curvature (Δt = 2.0 s). section are calculated. Mp can be obtained as:
Reaction Moment
20
Mu
M u* = (5)
10 Mp
20
For verification and comparison purpose,
ULS of the ship in intact condition is estimated
10 using Smith type simplified method (Smith,
1977) using the software MARS2000 (BV, 2000).
0 Figure 12 shows an example figure of the half
0.0E+00 2.0E-04 4.0E-04 6.0E-04 8.0E-04 transverse section of the ship in intact condition
1/ρ[1/m] obtained by MARS2000. ULS in net scantling
is calculated in MARS2000. Figure 13 shows
Figure 10. Relation between reaction bending moment comparison of ULS between Simplified analysis
and curvature (Δt = 0.3 s).
curvature center
θ1 θ2
UP.DK.
Neutral Axis
BTM. ρ
T.BHD. L T.BHD.
Figure 11. Relation between curvature radius and Figure 12. Transverse section of the ship (Intact condi-
rotation angle of T.BHD. tion, MARS2000).
433
20
14.44 15.22
Mu [GNm]
15
10
434
435
436
M.R. Zareei
Chabahar Maritime University, Chabahar, Iran
ABSTRACT: Non-linear finite element method is a powerful tool for analyzing ship collision problem
and has been seen more and more applications in the recent years. Ship collision simulations still face
some difficulties in terms of providing reliable results. The accuracy of numerical modeling results sig-
nificantly depends on the proper definition of the phenomenon and careful control of the some critical
parameters such as rupture criteria. A series of FEM indentation analysis have been carried out quasi-
statically on various configurations of stiffened steel panels with T- and Y- stiffeners. The analysis method
presented has been implemented into explicit finite element code LS-DYNA, being linked with ANSYS
for pre- and post-processing of the related computations.
437
438
Table 1. Material model properties. A recent years, some research was done on colli-
sion and grounding of ship structure and stiffened
Density [kg/m3] ρ 7850 panels. One of them is penetration of stiffened
Young modulus [GPa] E 210 panel both experimentally and numerically by
Poisson ratio [−] ν 0.303 Alsos et al. (2009a, b).
Tangent modulus [GPa] Etan 1 Verification studies were conducted in order to
Cowper-Symonds strain C 40.4 check present analysis for using in panels ground-
rate parameter [Hz] ing damage studies. To achieve this verification,
Cowper-Symonds strain P 5 stiffened panels of Alsos et al. (2009a) was selected
rate parameter and analyzed numerically to compare present anal-
ysis with experimental analysis.
Alsos et al. papers deal with hull damage in
ships which were subjected to grounding actions.
Table 2. Material failure properties. A ship was assumed to settle vertically on a rock.
It was further assumed that contact actions were
Plate Stiffener
local and restricted to one plate section. The sce-
σy 285 340 nario was analyzed by conducting a series of panel
εn 0.71 0.71 indentation experiments. Various configurations
n 0.24 0.225 of stiffened panels were loaded laterally by a cone
shaped indenter until fracture occurred.
439
440
441
b σy
β= (4)
t E
a σy (5)
λ=
Figure 7. Force-indentation behavior of panel 1-FB. πt E
where: b = BT; t = t1; a = L; E = Young modulus;
σy = Yield stress and r = Panel radius of gyration.
442
Y stiffener
Model ID L BT t1 B1 B2 H1 t2 H2 B3 t3
Model ID L BT t1 Hw Bf t2
443
6 CONCLUSIONS
444
445
ABSTRACT: If a ship strands it is important to estimate the acting forces and moments in respect to its
ultimate hull girder strength. The influence of the shear force in combination with the bending moment
needs to be taken into account. Therefore a procedure to calculate the moment-shear force interaction
curve of a ship’s cross section is introduced. Three specimens of an experiment (carried out by Reckling
et al.) are recalculated with the FE method to validate the procedure. In the experiment longitudinal
stiffened girders are loaded with pure bending. After the FE results are satisfactory equal to those of the
experiment and the computing time is optimised, the shear force is additionally introduced by different
load cases. Enough nodes for the moment-shear force interaction curve are found. Three normalised
curves are presented- one for each of the three specimens. Various cross sections and lengths of the
specimen result in different curves.
447
448
800
Model 31, 22
700
600
Stress [N/mm2]
500
400
300
200
100
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Strain
Figure 3. Stress-stain curve of model 31 and 22. Figure 4. Areas on top plate.
449
450
2.5E+08
No Pre-Deformation
2.4E+08
Cicular Pre-Deformation
2.2E+08
2.1E+08
M [Nmm]
2.0E+08
1.9E+08
1.8E+08
1.7E+08
1.6E+08
1.5E+08
9.0 11.0 13.0 15.0 17.0 19.0 21.0 23.0
f [mm]
Figure 8. Model 31 with Original Pre-deformation:
Z-displacement of top plate at and after ultimate bend-
ing moment. Figure 9. Ultimate bending moment: No, circular and
original pre-deformation.
2.4E+08
For a circular pre-deformation buckling is Normal Timestep
displacement can be found in the middle (area 3). 11.0 13.0 15.0 17.0
f [mm]
19.0 21.0 23.0
451
Ideal 31
Specimen 31
ent is seen in the moment-displacement curve. To
1.0E+08 FEM 31
gain satisfactory results in the FE simulation it is
Initial Stress 31 necessary to include the initial stress.
To demonstrate the effect of initial stress, the
5.0E+07
model with the original pre-deformation is taken.
The stiffeners of the top plate are loaded with a
0.0E+00
0.0 5.0 10.0 15.0 20.0 25.0
temperature of 600°C. The model is solved until the
f [mm]
temperature is reached and not cooled down. The
procedure introduces initial stress in x-direction
Figure 11. Ultimate bending moments of model 31.
close to the yield stress (see also Lehmann 2004).
After the temperature solution the displacement is
decreases. Its largest buckle deepens 13 mm from applied onto the model as described above.
timestep 0.404 s to 0.424 s, which is also the In Figure 11 four moment-displacement curves
largest change in the displacement of the three are plotted. The curve ‘Ideal 31’ is the ideal case,
calculations. ‘Specimen 31’ shows the result of the experiment,
The maximum moment of the model with the ‘FEM 31’ is the same as the curve with the title
circular pre-deformation is closer to the moment ‘Original Pre-Deformation’ given in Figure 9 and
of specimen 31. But the collapse mode of the the curve ‘Inital Stress 31’ belongs to the model
model with the original pre-deformation is more which is loaded with initial stress.
similar to the experiment, although the buck- The gradient of the model ‘Inital Stress 31’
les between the stiffeners develop on the left is very close to the gradient of ‘Specimen 31’.
side of the pre-deformation in the experiment The total course of the moments is not identical
(see Reckling et al. 1984). but the maximum moment is exactly the same
For the following calculations the original pre- (231.7 kNm), whereas the moment of the calcu-
deformation is taken. lation without initial stress has as a deviation of
1.38%. The sudden decrease of the curve from the
3.2.3 Timestep size FE calculation is still existent.
In the moment-displacement curve a sudden decline The buckling picture shows a good conformance
is seen and buckling in the top plate occurs from with this of the experiment. The particular buckles
one timestep to the next. To reduce this phenom- are as well on the left side of the pre-deformation
enon the timestep is refined around the ultimate (see Fig. 12) and the deepest deformation occurs in
bending moment capacity of the model. Between area 3. The pre-deformation is still seen after the
calculation time 0.3 s and 0.45 s the timestep size is ultimate strength is reached.
set from 5.0 E − 03 to 5.0 E − 05. The calculation The results of the FE simulations are satisfac-
needs eight hours instead of one. tory equal to those of the experiment. So the pro-
The maximum moment, that are reached in cedure in the FE method is validated by reality and
Figure 10, are almost the same. The decline of further calculations with the model will give rea-
the curve is smoother for finer timestepping, but sonable results. The influence of the initial stress
the sudden decline cannot be reduced. The buck- on the moment-shear force interaction can be
ling picture are for both cases the same. So the neglected. There is no difference for the relation of
results are similar, but the computational time is moment and shear force if the simulation is with
not acceptable. Thus all following simulations are or without initial stress. To save computational
calculated with the ‘normal’ timestep size. time the initial stress is not applied in the following
calculations.
3.2.4 Initial stress
The gradient of the moment-displacement curves 3.2.5 Model 22 and 23
in Figure 9 is identical to the gradient of the ideal Model 22 and 23 are calculated with the original
case. In the ideal case the material model is ideal pre-deformation, the ‘normal’ timestep size and
elastic and no pre-deformation is applied. The without initial stress. The moment-displacement
moment-displacement curve in the ideal case is curves of all three models are given in Figure 13.
similar to the curve of the hypothetical girder until Model 31 is shown in Figure 13 to see the influ-
the girder starts to plasticise. In the FE method ence of the different cross sections and lengths of
a hypothetical girder which can be loaded by the the specimen. The difference in the three graphs is
452
1.5E+08 FEM 31
Specimen 31
FEM 22
1.0E+08 Specimen 22
FEM 23
Specimen 23
5.0E+07
0.0E+00
0.0 5.0 10.0 15.0 20.0 25.0 30.0
f [mm]
453
Case 1a
0.8
Case 1b
3.0E+05
Case 1 0.7 Case 2
Case 1aa Case 3b
Case 1a 0.6
Case 3a
2.5E+05 Case 1b
M/Mpl
0.5 Case 3
Case 2
Case 3b
0.4
2.0E+05 Case 3a
Case 3 0.3
Q [N]
1.5E+05 0.2
0.1
1.0E+05 0.0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1
Q/Qpl
5.0E+04
Figure 16. Model 31: Shear forces of load cases 1–3. 1.0
Model 31
0.9 Model 22
Model 23
0.8
systems are additionally given in Figure 15. The
0.7
cases reverses from the moment-displacement
curves to the shear force- displacement curves, e.g. 0.6
M/Mpl
equivalent shear force is read out for the moment- 0 0.1 0.2 0.3 0.4 0.5 0.6
Q/Qpl
0.7 0.8 0.9 1 1.1 1.2
454
455
ABSTRACT: This paper focuses on the numerical methods for the ship hull structure strengths and
fatigue analyses, in order to estimate the initial design ship service life period. The applied methods have
been improved in the Marstruct Project. The numerical analyses are divided in three-interlinked parts.
The first part includes the hull strength analysis method, based on 3D/1D-FEM models, under equivalent
quasi-static head wave loads. The second part presents the ship hull dynamic response analysis method,
based on non-linear hydroelasticity theory with second order wave spectrum. The third part includes the
fatigue analysis method for the initial ship hull structure, based on the long-term prediction ship dynamic
response, the cumulative damage ratio and the design S-N material curves. The numerical analyses are
carried out for a large double hull oil-tanker with 275 m length between perpendiculars. The full and bal-
last loading cases are analysed. The numerical results outline the extreme hydroelastic wave loads and the
ships initial service life evaluation.
459
460
140
0.80
0.60
0.40
0.20
0.00
-0.20
-0.40
-0.60
-0.80
-1.00
0.00 28.50 57.00 85.50 114.00 142.50 171.00 199.50 228.00 256.50 285.00
Mode 0 / 0.088Hz Mode 1 / 0.097 Hz Mode 2 / 0.629 Hz Mode 3 / 1.478 Hz Mode 4 / 2.536 Hz x [m]
461
462
Table 8. The maximum TK 167000 tdw oil-tanker stresses, for a) the full cargo load case and b) the ballast load
case, under sagging and hogging quasi-static equivalent head wave conditions, with the reference wave height
hw = 10.692 m.
Sagging Hogging
463
0.00 200.00
-25.00
175.00
-50.00
150.00
-75.00
125.00
-100.00
100.00
-125.00
75.00
-150.00
50.00
-175.00
-200.00 25.00
-225.00 0.00
0.00 28.50 57.00 85.50 114.00 142.50 171.00 199.50 228.00 256.50 285.00 0.00 28.50 57.00 85.50 114.00 142.50 171.00 199.50 228.00 256.50 285.00
hw=0m hw=1m hw=2m hw=3m hw=4m hw=5m hw=6m hw=7m x [m] hw=0m hw=1m hw=2m hw=3m hw=4m hw=5m hw=6m hw=7m x [m]
hw=8m hw=9m hw=10m hw=10.69 hw=11m hw=12m adm hw=8m hw=9m hw=10m hw=10.69 hw=11m hw=12m adm
Figure 5a. 1D-girder model, σxD deck normal stress Figure 7a. 3D-FEM model, maximum von Mises deck
[N/mm2], head wave sagging condition (hw = 0–12 m), full stress [N/mm2], head wave sagging condition, ballast.
cargo.
2
σVON max [N/mm ] DECK 3D-FEM Model Hogging / Quasi-static Wave / TK Oil-tanker Ballast
225.00
2
σx max [N/mm ] DECK 3D-FEM Model Sagging / Quasi-static Wave / TK Oil-tanker Full Cargo 200.00
25.00
175.00
0.00
-25.00 150.00
-50.00 125.00
-75.00 100.00
-100.00
75.00
-125.00
50.00
-150.00
25.00
-175.00
0.00
-200.00 0.00 28.50 57.00 85.50 114.00 142.50 171.00 199.50 228.00 256.50 285.00
-225.00 hw=0m hw=1m hw=2m hw=3m hw=4m hw=5m hw=6m hw=7m x [m]
0.00 28.50 57.00 85.50 114.00 142.50 171.00 199.50 228.00 256.50 285.00 hw=8m hw=9m hw=10m hw=10.69 hw=11m hw=12m adm
2
σVON max [N/mm ] DECK 3D-FEM Model Sagging / Quasi-static Wave / TK Oil-tanker Full Cargo
225.00
200.00
175.00
150.00
125.00
100.00
75.00
50.00
25.00
0.00
0.00 28.50 57.00 85.50 114.00 142.50 171.00 199.50 228.00 256.50 285.00
2
σVON max [N/mm ] DECK 3D-FEM Model Hogging / Quasi-static Wave / TK Oil-tanker Full Cargo
225.00
200.00
175.00
150.00
125.00
100.00
75.00
50.00
25.00
0.00
0.00 28.50 57.00 85.50 114.00 142.50 171.00 199.50 228.00 256.50 285.00
Figure 6b. 3D-FEM model, maximum von Mises deck Figure 8b. Wave pressure at hW = 10.692 m, hogging, full
stress [N/mm2], head wave hogging condition, full cargo. cargo.
464
465
466
deck 4.50
bottom 1.50
-7.50
0 10 20 30 40 50 60 70 t [s] 80
In Tables 9–11, for simulation time 80s and Figure 10a. The wave L-H time record (h1/3 = 12 m,
8000 records, the following statistical results are x/L = 0.5).
included:
− the ratio for the significant deformation on the η wave [m] Amplitude Spectrum FFT Analysis : Wave
2.75
1.75
1.25
0.50
σ13
1/ , M13
1/ W ,B τ1/13
1/3 = T/ max
a Kτ n − nS
σ13
1/ ,B + σ swD,B τ1///3maxSn − n + τ swSn
S −n
swSn (5)
≤1 ≤1
σ adm τ adm
467
200.00
6.0E+06
175.00
4.0E+06
150.00
2.0E+06 125.00
0.0E+00 100.00
75.00
-2.0E+06
50.00
-4.0E+06
25.00
-6.0E+06
0.00
0.00 28.50 57.00 85.50 114.00 142.50 171.00 199.50 228.00 256.50 285.00
-8.0E+06 h1/3=0m h1/3=1m h1/3=2m h1/3=3m h1/3=4m h1/3=5m h1/3=6m x[m]
0 10 20 30 40 50 60 70 t [s] 80
h1/3=7m h1/3=8m h1/3=9m h1/3=10m h1/3=11m h1/3=12m Sig_adm
Figure 11a. Oscillation bending moment time record, Figure 13a. Maximum significant normal deck stress
non-linear analysis, wave h1/3 = 12 m, vs = 15 Knots, [N/mm2], non-linear analysis + still water, h1/3 = 0–12 m,
x/L = 0.5, full cargo case. vs = 15 Knots, full cargo case.
2
Mosc_nl [KNm] Oil-tanker Full Cargo Section x/L=0.5 vs=15Knots Wave LH & ITTC h1/3=12m σx1/3nl+|ac| [N/mm ] DECK Oil-tanker Ballast vs=16 Knots & Wave LH & ITTC
Non-Linear Amplitude Spectrum FFT Analysis : Bending Moment - Oscillation Response 225.00
2.5E+06
200.00
175.00
2.0E+06
150.00
125.00
1.5E+06
100.00
75.00
1.0E+06
50.00
25.00
5.0E+05
0.00
0.00 28.50 57.00 85.50 114.00 142.50 171.00 199.50 228.00 256.50 285.00
Figure 11b. Oscillation bending moment amplitude Figure 13b. Maximum significant normal deck stress
spectrum FFT, non-linear analysis, wave h1/3 = 12 m, [N/mm2], non-linear analysis + still water, h1/3 = 0–12 m,
vs = 15 Knots, x/L = 0.5, full cargo. vs = 16 Knots, ballast case.
M_nl [KNm]
In Figures 10–13 are included the following
Oil-tanker Full Cargo Section x/L=0.5 vs=15Knots Wave LH & ITTC h1/3=12m
8.0E+06
Non-Linear Analysis Time Record : Bending Moment - Hydroelastic Response
results:
− Figures 10a, b present the time record and the
6.0E+06
4.0E+06
amplitude spectrum for the Longuet-Higgins
2.0E+06
-4.0E+06
-6.0E+06
amplitude spectrum for the non-linear oscilla-
-8.0E+06
tion bending moment (non-linear seakeeping
0 10 20 30 40 50 60 70 t [s] 80
response on heave and pitch modes), head wave
Figure 12a. Hydroelastic bending moment time record, h1/3 = 12 m, at amidships section x/L = 0.5, full
non-linear analysis, wave h1/3 = 12 m, vs = 15 Knots, cargo load case;
x/L = 0.5, full cargo. − Figures 12a, b present the time record and the
amplitude spectrum FFT, for the non-linear
hydro-elastic bending moment, head wave
M_nl [KNm]
2.5E+06
Oil-tanker Full Cargo Section x/L=0.5 vs=15Knots Wave LH & ITTC h1/3=12m
Non-Linear Amplitude Spectrum FFT Analysis : Bending Moment - Hydroelastic Response h1/3 = 12 m, at amidships section x/L = 0.5, full
cargo load case;
2.0E+06
− Figures 13a, b present the maximum hydroelas-
1.5E+06
tic significant deck normal stresses, both loading
cases.
1.0E+06
5.0E+05
468
0.35
Analyses Welding DSN_ full DSN_ball DSN L [years]
0.05
3D/1D model, σvonM1/3 deck maximum equivalent
h1/3 [m] von Mises stress
0.00
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 10 10.5 11 11.5 12
Linear Standard 0.623 0.835 0.729 >20
Nonlinear Standard 0.821 0.916 0.869 >20
Figure 14. The World Wide Trade wave height h1/3 Nonlinear Very good 0.343 0.390 0.367 >20
histogram.
469
470
471
A. Martínez Cimadevilla
Technical Division, Navantia S.A., Madrid, Spain
ABSTRACT: The vibratory behaviour of marine structures like the typical case of a vessel is a complex
phenomenon that involves the evaluation of different aspects as the assessment of the external added
mass water effect and its influence on the vibration characteristics. This added mass water effect can be
properly established by considering the fluid-structure interaction due to the surrounding water.
From that evaluation, reliable dynamic parameters were derived. Two analyses types were applied: a
forced response linear dynamic analysis (harmonic acoustic finite element method) and an unsymmetric
modal analysis, both considering the Fluid-Structure Interaction (FSI).
A Lagrangian Eulerian finite element formulation was adopted for the FSI simulation.
The model employed for such analysis, was a simplified one: a model corresponding to a ship-like
structure (rectangular section hull girder). This model was considered representative enough to highlight
the main features of a marine structure vibration analysis.
In addition, a comparison between analytical values and Finite Element results related to frequencies
data was carried out. The compared values showed a good agreement.
473
474
475
2N 0.448
3N 1.057
4N 1.724
5N 2.42
476
2N horizontal 0.645
3N horizontal 1.476
4N horizontal 2.423
5N horizontal 3.416
Axial 0.665
Torsional 2.429
Figure 11. 3N horizontal mode shape displacements. Figure 15. Torsional mode shape displacements.
477
Figure 16. Response curves location. Figure 20. Response curve “D”.
478
Ratio a/b α1
0.1 2.23
0.2 1.98
0.5 1.70
1.0 1.51
2.0 1.36
5.0 1.21
479
fa = 0.811 Hz
fS
1
0 fw = 0.5
fa (14)
2 2.5 3 3.5 4 4.5 5
⎛ Amh ⎞
⎜⎝1 + ⎟
Nd
m ⎠
Node Number
R ANSYSS
1.02
where:
N2h is the hull girder 2N horizontal natural fre-
quency, n is the number of nodes.
1.01
480
(Frequencies Ratio)
case), the derived μ value is 1.20. R ANSYSS
1
N2v = 0.498 Hz
3 And the Eq. (12), for the Kumai values becomes:
f ANSYS
Nnv = 0.498 (n − 1)1.20 (19)
(Hz)
fS 2
481
2.5 7 CONCLUSIONS
482
483
S.H. Kim, K.S. Kim, J.H. Lee, C.H. Yu & W.H. You
Department of Naval Architecture and Ocean Engineering, Inha University, Incheon, Korea
ABSTRACT: Fatigue crack propagation life assessment of a cargo hold is an overriding consideration
in the design of marine vessel such as crude oil, LNG, and LPG carriers. This paper studies the crack
propagation life of T-joint welded structure on which typical fatigue cracks are frequently initiated when
the marine vessels experience the storm load. Welding residual stresses are calculated to investigate its
effects on the fatigue life. Thereafter the residual stress distribution is applied to the AFGROW life
prediction program which incorporates the loading, the welding residual stress, and the geometric shape
of the structure. The fatigue tests of the T-joint welded specimen under storm loading show the beach
mark clearly generated on the cut section of the specimen. The crack propagation life estimated based
on the beach mark are compared with that of AFGROW to validate the life prediction. Evaluation of
the remaining service is discussed in order to allow the remaining fatigue lifetime to be estimated for the
marine vessel’s cargo hold if the random load or the storm load is introduced to the vessel.
487
488
C Si Mn P S Ni Cu Mo
489
490
491
daX
( )
m
= C ( K )m = C α S π × aX (4)
dN
NX
ΔK α ΔS π a (5)
⎡ 1 ⎤ 1
2 ⋅ ⎢ (m − 2) / 2 ⎥ − (m − 2) / 2
⎣ ai ⎦ ain
ΔN = (6)
( m 2 )C ( ΔS )m ( )m / 2 ( )m
492
ACKNOWLEDGEMENTS
493
Weiguo Wu
Wuhan University of Technology, Wuhan, Hubei, China
Key Laboratory of High-speed Ship Engineering, Ministry of Education, Wuhan, Hubei, China
ABSTRACT: In order to investigate slamming load distribution and its relation to the impact velocity,
a small scale trimaran model test was performed. During the water entry from a series of heights, the
peak pressure of the main hull, side hull and wet-deck were obtained. Based on the model test and the
similarity theory, this paper proposed a reasonable and effective prediction method of the trimaran
slamming loads.
495
Equation (1) can be written as the π exponential CP C ρCV 2 CL means PS /Pm = CL (7)
term as follow:
PS: the slamming pressure of prototype trimaran
α1 α2 Pm: the slamming pressure of model trimaran
⎛M⎞ ⎛ P⎞
π =⎜σ ⎟ ⎜⎝ σ ⎟⎠ E α 3 μα 4 α5
i
α6 α7
0
α8
δ 0α 9
⎝ ⎠ That is guaranteed to derive the slamming pres-
sure P of the prototype trimaran using this scale
×g α 10 α 11 α 12 α 13
ρ t γ Fr α 114
(2) based on meeting the similarity criteria:
496
3 SLAMMING EXPERIMENT
Figure 4. Experimental set-up arrangements.
3.1 Experimental set-up
Test setup includes the water tank, test tower, carry-
Table 1. Experimental and calculated impact speed.
ing car and a winch with manual clutch. The external
dimension of the pool is 8.90 m × 4.90 m × 3.15 m, Drop height (m) 0.75 1.00 1.25 1.50 1.75
while the inner size is 8.00 m × 4.00 m × 3.15 m. Test V (m/s) 2.264 2.452 2.668 2.836 3.123
tower, 6 meters heights with four parallel tracks, is Drop height (m) 2.00 2.25 2.50 2.75 3.00
fixed on the top of the test tank. The carrying car V (m/s) 3.344 3.619 3.968 4.236 4.432
which connected with the model is fixed through
eight guide wheel mounted within the track, which Where V stands for the relative impact speed.
497
4 EXPERIMENTAL PHENOMENA
AND RESULTS’ ANALYSIS
498
499
REFERENCES
the effect on the slamming pressure coursed by
a variety of factors on different water impact. Anders Rosén, Karl Garme, & Jakob Kuttenkeuler.
Previous studies suggest that various factors effect- 2007. Full-Scale Design Evaluation of the Visby Class
ing on slamming pressure is constant, reflecting the Corvette. Ninth International Conference on Fast Sea
k value is constant in the calculation, which would Transportation, Shanghai:583–588.
cause some errors in the prediction of slamming Cao Zhenglin. 2008. Research on the slamming load of
trimaran [PhD thesis]. Wuhan: Wuhan University of
pressure on different impact speed. So it must be Technology.
extremely careful to use this method to predict the Chuang, S.L. 1966. Experiment on flat-bottom slam-
slamming pressure. ming. Journal of Ship Research. 10(1):10–27.
The pressure of the cross structure is lower Michael R, Davis, James R. 2007. Computation of wet
than the main hull, which is in contradic- deck bow slam loads for catamaran arched cross
tion to the experimental result of Cao (2008). sections. Ocean Engineering. 34:C2265–C2276.
Figure 14 shows the typical cross section of the Ochi MK, Motter LE. 1973. Prediction of slamming
Cao’s model, and Table 2 compares the dimen- characteristics and hull responses for ship design.
sions and results of two models. From the scenes SNAME.
Ochi MD, Bonilla-Norat J. 1970. Pressure-velocity rela-
at the moment of water entry, it may cause by the tionship in impact of a ship model dropped onto the
cancellation of the spray induced by the main hull water surface and in slamming in waves. AD-709071.
and the side hull interaction, which to function Zhao R, Falinsen O.M. 1996. Water entry of arbitrary two-
as the bulbous bow in reducing the wave-making dimensional sections with and without flow separation.
resistance. Proc. 21st Symp. on Naval Hydrodynamics.
500
501
ABSTRACT: This paper analyses the failure modes based on experimental test results of multispan
corroded box-girders, representing amidships ship hull section and subjected to four-point loading.
Comparative study for three tested corroded box girder specimens with respect to collapse modes, strain
measurements, residual stresses, load-displacement and moment-curvature relationships and the effect of
different corrosion levels on structural integrity are performed.
503
504
avoid the effect of the boundary conditions for thicknesses diminution, which reduces the ultimate
the central plates, which is due to the possible strength.
eccentricity of load and the interference between
adjacent panels.
3.1 Modes of collapse
The deck panel is stiffened with five longitudinal
flat bars with a spacing of 150 mm. The side panel The collapse modes of the plate panels under pre-
is stiffened with two stiffeners on a distance of 300 dominately compressive loads usually undergo
and 500 mm respectively and the bottom panel three main types, namely plate buckling, inter-
was stiffened with one stiffener in the middle, as frame flexural buckling (stiffener tripping) and
may be seen in Figure 1. The box girders are made overall grillage collapse (plate and stiffener), see
of normal ship building steel with yield stress of Figure 2. In some cases there is an interaction
235 MPa. between two collapse modes, this interaction gen-
The geometry configurations of the analysed erally reduces the load-carrying capacity corre-
box girders are given in Table 1. The thickness sponding to a single buckling mode.
measurement survey has been performed in the In the pre-buckling stage, the plate’s response
way to find out the remaining thicknesses of the to load follows the Hooke’s Law where its load
box girders structural elements. The as-built thick- displacement relationship is linear. When com-
ness of the plates and stiffeners are 4.5 mm. The pressive loads reach the critical buckling load,
corrosion thickness is not uniformly distributed buckling occurs and the buckling profile of the
which affects the response of the box girders. plate usually in the form of half-waves of approxi-
The box girder has been mounted between two mately equal length.
stiff supporting arms, using bolt connections. The Flat bar stiffeners provide very little torsional
box girder was subjected to four-point vertical load- restraint, therefore, if the plate between stiffen-
ing. The bottom part is subjected to tension and the ers buckles, the tripping starts to take place in
upper part, deck, is under compression. The bend- the stiffener. Finally, the overall collapse mode
ing moment is kept constant along the box girder, involves the global buckling of plating, stiffeners
between the bolt connections. There are four points and transverses.
for transmitting the load, two are located at the The first collapse mode observed for three boxes
supports of the arms and two are on the boundary was overall pre-buckling of the plating and stiffen-
between the box-girder and the supporting arms. ers as one unit (see Figure 3), however, in this mode
To avoid shear effects, the load is applied on each box behaves in different shape, due to the dif-
a certain distance away from the ends of the box ferent level of corrosion deterioration present in
girder (the connections between the box girder and structures.
the supporting arms). The load is applied using As can be seen from Figure 3(a), for initially
a 700 kN hydraulic jack and it is transmitted to corroded box, only the bay no. 2 and the bay no. 3
the box girder through a horizontal beam, as are are affected. However, the bay no. 3 in which the
shown in Figure 1. achieved lateral buckling displacement in the
direction of the plate is bigger than the one in
the second bay in the direction of the stiffener.
3 EXPERIMENTAL RESULTS Therefore, the global deformation was asym-
metrical with respect to the middle section of the
During the service life of the ship, the struc- box girder, and this can be explained with the fact
ture deteriorates resulting in structural elements that the corrosion depth distribution was non
505
Figure 3c. Overall pre-buckling mode, severely cor- Figure 5a. Interframe buckling mode (stiffeners trip-
roded box. ping), moderately corroded box.
506
507
508
loading carrying capacity, it is evident that in the Failure mode assessment based on the test results
elastic range (see Figure 13) at any certain load, the of three corroded box girders has been presented.
displacement increases by increasing the corrosion The effect of corrosion deterioration on the col-
degradation level especially for severely corroded lapse modes and the overall behaviour of the box
box. Also by increasing the corrosion level the ulti- girders have been analysed. Strain gauges reading
mate carrying capacity that the box can withstand on several locations have been analysed and com-
decreases with 32.86% and 64.67% for moder- pared for the three corroded boxes. The effect of
ate and severe corrosion respectively as presented corrosion deterioration on reducing the existing
in Table 2. residual stresses has been observed and analysed.
It may be observed from Figure 13 that there With the presence of corrosion, the ultimate bend-
is an initial drop in the load-displacement curve ing moment as well as the load carrying capac-
for initially corroded box, which is following by ity of the box girders decreased with 32.86% and
the collapse, the second drop down appears to be 64.67% for moderately and severely corroded box
associated with the spread of plasticity through girders respectively.
the second bay and the third bay. On the contrary,
the behaviour of the other boxes is different due
to the corrosion degradation. Furthermore, it is ACKNOWLEDGEMENTS
noticeable that, the higher the corrosion level the
lower the elastic range. The work reported here is a contribution to
the activities of the MARSTRUCT VIRTUAL
INSTITUTE, (www.marstruct-vi.com) in particu-
3.5 Moment-curvature relationship
lar its Technical Subcommittee 2.3 on Ultimate
The experimental results of the moment-curvature Strength.
relationship are given in Table 2. In general, The first author has been funded by the Portu-
before buckling and yielding occur, the moment- guese Foundation for Science and Technology
curvature relationship is linear until the occur- (Fundação para a Ciência e Tecnologia—FCT)
rence of any of them. For the tested box girders it under contract SFRH/BD/46790/2008.
509
510
ABSTRACT: Within a joint development project between the Korean Shipyard Daewoo Shipbuilding &
Marine Engineering Co., Ltd (DSME) and Germanischer Lloyd (GL) fatigue tests of 80 mm thick plates
have been performed. The objective of the fatigue tests was to investigate the effect of different post
treatment methods on the fatigue performance. Test specimens were butt welds as well as plate edges in
untreated and post treated conditions respectively. While for the butt weld specimens ultrasonic impact
treatment (UP) was applied as post treatment method the plate edge specimens were treated by chamfering
and different combinations of edge grinding, surface grinding and UP surface treatment. Material was
higher tensile steel HT40 for all specimens. The results of the fatigue tests are discussed and compared
with the values and recommendations from the Rules and corresponding guidelines.
511
The main objective of the butt weld fatigue tests Figure 2. Geometry of the butt weld specimens.
was to investigate the effect of ultrasonic impact
peening (UP) as a post weld treatment method on
the fatigue strength. In order to avoid horizontal vibrations during
Within the JDP altogether 16 butt weld specimens the tests in the resonance testing machine and to
were tested, 8 specimens in as-welded condition prevent crack start at the edge notches mentioned
and 8 specimens in UP treated conditions. The tests above, the specimen edges were milled. This meas-
were carried out under constant amplitude axial ure reduced the specimen breadth from originally
load on the 12 MN resonance testing machine of 120 mm to approx. 100 mm. The welded and UP
the Department of Ferrous Metallurgy (IEHK) at treated specimens were rounded in the welded area
the Rheinisch Westfälische Technische Hochschule to force the crack initiation location towards the
(RWTH) in Aachen (Germany). The test frequency centre of the specimens.
was approximately 33 Hz. The axial and the angular misalignment of all
To ensure test conditions that correspond to butt weld specimens were measured by laser scan-
the actual loading conditions of the coaming ning at the Institute of Ship Structural Design and
and coaming top plate of large container vessels Analysis of the Hamburg University of Technol-
the tests were carried out under axial loading and ogy (TUHH) before testing. Figures 2 and 3 show
stress ratio R = +0.1. The specimen thickness of typical scans of the weld surface for a specimen in
80 mm as well as the higher tensile strength steel as-welded condition and in UP treated condition
E40 and the electro gas welding (EGW) used for respectively. As a result of the UP treatment at the
the specimen fabrication are also typical for the weld toes rounded channels have been generated.
upper hatch coaming area of large container As expected for the plate thickness of 80 mm the
vessels. Figure 2 shows a sketch of the butt weld measured misalignments were comparatively low.
specimens. The data for the base material is given The maximum of the measured axial misalign-
in Table 1. Table 2 shows the parameters of the ments of the specimens is 1% of the plate thickness
welding process, consumables and condition. (mean value 0.4%) and the maximum angular dis-
In delivery conditions some specimen showed a tortion is 0.5° (mean value 0.14°). More details and
horizontal deflection and notches at the specimen conclusions regarding the misalignment influence
edges near the weld bead due to thermal cutting. are given in the following chapters 2.1–2.3.
512
0.35
Ceq
additional stresses in the specimens when axially
loaded and they are superimposed by additional
0.012
static mean stress due to clamping of the specimen
TI
in the testing machine. To account for all stress
components during tests, all specimens were fitted
–
with strain gauges.
For all fatigue tests of the butt welded specimens
S-AL
0.045
the final rupture of the specimen was employed as
the failure criterion. In this way the comparability
0.018 between the test results in as-welded conditions
and the test results in UP treated conditions is
Nb
0.0
conditions
The fatigue tests of the specimens in as-welded
0.08
28%
408 MPa
Name
513
514
515
516
517
and the associated FAT classes (value at 2 ⋅ 106 load Table 3. Results of the fatigue tests.
cycles) are presented in addition. To enable a direct
comparison of the test results with the actual fatigue FAT Class detail ΔσR [N/mm2]
assessment S-N curve for grinded cut edges from
the Rules of Germanischer Lloyd (2010) these are Series m = 3.5 m = 4.0 m from test
also included in Figure 12 (dotted curve). Here the
1C 169 182 180 (m = 3.9)
increase of the fatigue strength due to the application
2R 176 186 143 (m = 2.5)
of higher tensile steel has been considered by means
of the material influence factor fm. According to 2R + G 196 207 190 (m = 3.3)
the Rules of Germanischer Lloyd (2010) fm = 1.13 2R + UP 143 160 207 (m = 6.2)
for HT40 steel resulting in modified FAT class of All 168 181 176 (m = 3.8)
FAT 158 (m = 4.0) instead of FAT 140 (m = 4.0) for
mild steel.
As a summarization of the cut edge fatigue test
the fixed slopes at this test series. In this respect it
results the influence of the different investigated
has to be mentioned that the tests of the 2R + UP
post treatment procedures as well as the effect of
series shows a poor correlation and a relatively
different slopes assumed for the fatigue strength
large scatter even for the slope calculated directly
assessment are shown in Figure 13 allegorized by
from the tests (see Figure 13 TN = 1:6.73).
the appertaining FAT classes. Complementary to
Under consideration of the limitations described
Figure 13 the results of the different evaluations of
before the following conclusions can be drawn from
the fatigue tests are given numerically in Table 2. With
the fatigue tests of the cut edge specimens:
respect to the Rules of Germanischer Lloyd (2010)
the fatigue tests results have been evaluated with a • The results of the fatigue tests showed that the
fixed slope of m = 3.5 and m = 4.0. Furthermore S-N curves from the Rules lead to a somewhat
results for slopes directly obtained from the tests are conservative assessment of the fatigue strength
presented. It should be noted that the latter values of cut edges (see Fig. 12). Even if the already
are not very meaningful as the load levels are close described limitations of the test results are taken
together on the one hand and the number of speci- into account, a modified FAT class for high qual-
men is not very high on the other hand. ity treated cut edges seems to be reasonable.
Figure 13 shows that the fatigue tests results are • Series 1C and 2R show a very similar fatigue
very sensitive to the slope m of the S-N curve that performance—no positive effect of the more
has been used for the assessment. Especially for the time-consuming 2R treatment could be observed
test series 2R and 2R + UP. For the series 2R the from the tests.
(doubtful) slope of m = 2.5 obtained directly from • Surface grinding (series 2R + G) has a positive
the tests leads to a much lower detail category com- effect on the fatigue strength.
pared to the assumption of fixed slopes m = 3.5 and • UP surface treatment influences the fatigue
m = 4.0 according to the Rules. Contrary to this behaviour—especially the slope of the apper-
the application of the fixed slopes from the Rules taining S-N curve.
leads to significantly lower detail categories than • In general the additional positive effect of com-
obtained by applying the slope from the tests for plete surface treatment on the fatigue strength
the 2R + UP series. This is mainly due to the large was found to be relatively small compared to the
scatter bands that result from the enforcement of beneficial basic effect of corner treatment.
518
519
ABSTRACT: In this paper, structural model test method was introduced to study the ultimate strength
of a model structure of high speed trimaran. Firstly, similarity theory with finite element method was
applied to instruct the structural model design. Then, to simulate the peak wave bending moment of
the high speed trimaran, three point bending was applied on the model by hydraulic jacks in the ultimate
strength test. Finally, The ultimate strength of the trimaran model was successfully determined from
the model test. The load-stress and load-displacement curves of the model, and the trimaran failure mode
were obtained.
Keywords: high speed trimaran; ultimate strength; structure model test; similarity theory
521
Ir er
= CL3Ct = CL (3)
Im em
Errors
2.3 Similarity verification
Sections Neutral axis Inertia moment To examine the similarity of the test model, the
FE analysis for both the actual trimaran and its
Calculated section 1 0.6% 0.9%
scale model are carried out. Only linear static
Calculated section 2 0.1% 1.1%
Calculated section 3 2% 0.04%
analysis was performed. Half of FEA models
Middle section 1 0.2% 0.1%
were applied in the calculation of model structure
Middle section 2 0.1% 1.8% and trimaran, symmetry constraints were adopted
in the longitudinal section. The load condition
522
4 MODEL TEST
523
524
Figure 10. Failure mode of deck. Figure 13. Longitudinal displacement curve of main hull.
525
526
C.S. Shim
Mokpo national university, Mokpo, Republic of Korea
J.K. Kang
Daewoo Shipbuilding & Marine Engineering Co. Ltd, Goeje, Republic of Korea
D.H. Kim
Hyundai Heavy Industries Co. Ltd, Ulsan, Republic of Korea
Y.S. Suh
Samsung Heavy Industries Co. Ltd, Goeje, Republic of Korea
Y.L. Shim
ABS, Houston, USA
H.S. Urm
DNV Korea Ltd, Busan, Republic of Korea
M.S. Kim
Lloyd’s Register Asia, Busan, Republic of Korea
G.B. An
POSCO, Pohang, Republic of Korea
ABSTRACT: IMO type B CCS (cargo containment system) is considered for LNG FPSOs. For the
three materials of SUS304, 9% Ni steel and Al 5083-O alloy, regarded as possible candidates for IMO type B
CCS, extensive tensile, fatigue, fatigue crack growth rate and CTOD tests were carried out at room, inter-
mediate low (−100°C) and cryogenic (−163°C) temperatures in this paper. All materials show a tendency
for initial yield and tensile strengths to increase with decrease in the temperature, while change in elastic
moduli is not as remarkable. The largest and smallest increase ratios of the initial yield strengths due to
temperature reduction are seen in SUS304 and Al 5083-O alloy, respectively. In fatigue tests, fatigue
strengths increase due to the temperature decrease for three materials. The highest increase of fatigue
strength due to the temperature decrease is observed in Al 5083-O alloy, while 9% Ni steel shows lowest
increase. In fatigue crack growth rate test, SUS304 and Al 5083-O alloy show reduction in crack propaga-
tion rate, due to the reduction in temperature, while any visible improvement of da/dN is not witnessed in
case of 9% Ni steel. In CTOD test, CTODs tend to decrease in case of SUS304 and CTODs for Al 5083-O
alloy increase with reduction in temperature.
527
528
ΔP ( + )
ΔK = 3
B W
( − )2
× (0. 86 + 4.664α − 13 32α 2 + 14.772α 3− 5 6α 4 )
(4)
529
⎛ 1 ⎞
δ =⎜
⎝ mσY ⎟⎠
⎧ ⎫
⎪ (
⎪K2 − υ )+ η Ap ⎪
⎪
⎨ ⎬
⎪ E ⎡ ⎛ α +z ⎞⎤⎪
⎪⎩ ⎢B ( − ) 1+
⎝ ⎥
0.88a0 + 0 2W ⎟⎠ ⎦ ⎪
⎣ ⎭
(a) SUS304
3 RESULTS AND DISCUSSION
530
531
532
533
4 CONCLUSIONS
534
535
ABSTRACT: A comprehensive model was developed for the calculation of the thermal profile based
on frictional heating from FSW tool and absorption of laser energy on the plate surface. The calculations
showed that using a suitable laser source 75% of the desired temperature level was attained through con-
duction mode and the rest was attained through frictional heat. Thus achieving desired plasticizing of the
plate material, the forces required to move the FSW tool through the butt will be considerably less result-
ing in less wear and tear and longer life of the FSW tools. The computed temperature profile due to laser
source agreed well with the corresponding experimental data. The good agreement demonstrates that the
comprehensive model can serve as a basis for development of a feasible FSW process for shipbuilding
quality normal strength and higher strength steels.
539
540
d?
2.1.2 Tool pin—work piece interface
Heat generated from a cylindrical tool pin con-
dr sists of two contributions; Q2 from the side surface
rd? and Q3 from the tip surface of the tool pin. Heat
generated due to friction of an elemental area on
the cylindrical pin surface is given by,
Q3
Rp dQV = ω r dFV
dz
Q1
Rs Therefore dQV = ω r2 μp dθ dz.
Integrating dQV over the tool pin vertical side
surface one obtains,
a
2 π HP
Q2 ∫ ∫ μ ppR
RP2 ddzdθ
0 0
= 2 π μ p RP2 H P (10)
From Eq. (7) one can observe that the concave Qtotal
2
3
p (R 3
S 3RP2 H P ) (13)
shoulder surface actually contributes to increased
frictional area by a factor of tan α.
Therefore combining Eq. (6) and Eq. (7) the heat 2.1.3 Heat source model for FSW tool
generation from the elemental shoulder surface is The heat source modeling for FSW tool was done
given by, based on the following assumptions,
i. the heat input is linearly proportional to the
dQS = ω r dFS = ω r2 μ p dθ dr(1 + tan α ) (8) distance from the center of the tool,
ii. the plunging force applied to the plate surface
The heat generated through friction of tool by the tool creates a uniform pressure over the
shoulder with the plate surface is obtained by inte- shoulder surface,
grating Eq. (8) from pin root radius to the outer iii. the heat is generated from the work done by the
radius of shoulder surface. friction force only.
541
5
3
5.8
10
35
10
25
542
100
Laser beam radius at plate surface (mm), rb 2.0
Laser power (W), Q 1000 50
Laser beam speed (mm/min), Vb 200
0
Absorption coefficient, η 0.63 0 50 100 150 200 250 300
Time (s)
543
∂ ⎡ ∂T ⎤ ∂ ⎡ ∂T ⎤ ∂ ⎡ ∂T ⎤ ∂T
K + ⎢K ⎥+ K = ρC
∂x ⎢⎣ ∂x ⎥⎦ ∂y ⎣ ∂y ⎦ ∂z ⎢⎣ ∂z ⎥⎦ ∂t
544
800 Temperature rise temperatures achieved with the dual heat source
due to laser heating
600 for different tool rpm is shown in Table 8.
The combination of weld parameters as shown
400 in Sl. No. 3 of Table 8 will lead to a temperature
200 level of 1227°C. At this temperature it will be very
much feasible to carryout FSW of steel without
0
0 25 50 75 100 125 150 175 200 excessive wear and tear of the FSW tool.
Time (s)
300 800
o
Yield Stress (MPa)
250
600
200
150 400
100 200
50
0
0 0 50 100 150 200
0 200 400 600 800 1000 Time (s)
Temperature (C)
Figure 9. Temperature distribution for different
Figure 8. Variation of yield stress with temperature. position of laser beam ahead of FSW tool.
545
FSW tool
Laser beam Maximum
Plate thickness Rotational Traverse speed Plunging position ahead temperature
Sl. No (mm) speed (rpm) (mm/min) force (N) of FSW tool (mm) rise (οC)
FSW tool
Tool rpm 1000 iii. The laser source would heat up the plate to
1400
Tool rpm 1400 about 900°C reducing the yield stress of the
1200 Tool rpm 2000 material to around 10 MPa. It would result in
lesser tool plunging force with less tool wear.
1000 iv. Laser heating with a laser power source of
Temperature ( C)
800
would result in a temperature level of about
600 1200°C at the plate surface.
v. The time, Δt8−5 taken to cool over the range
400 800–500°C as observed from above analysis was
about 25 s. It is almost double to that of the
200
manual metal arc process with a heat input of
0 about 1.3 kJ/mm.
0 50 100 150 200 250 300 vi. The above study indicates a strong possibility of
Time (s) successfully carrying out laser assisted FSW of
steel plates with reduced tool wear and tear.
Figure 10. Temperature distribution for dual source
with varying tool rpm at a welding speed of 80 mm/min.
ACKNOWLEDGEMENTS
6 CONCLUSIONS
The results reported in this paper are from a
The combined heating due to a laser source and research project carried out under funding from
the frictional effect of FSW tool was investigated Naval Research Board vide their sanction number
through experiments and computer modeling. The DNRD/05/4003/NRB/169 dated 08/10/2008. The
major findings are as follows: authors gratefully acknowledge NRB’s financial
support to successfully carry out the investigation.
i. The major part of the total heat was generated
through the laser source in the proposed laser
assisted FSW process of C-Mn steel plates. Thus NOMENCLATURE
sufficient plasticizing of the plate material could
be achieved through laser preheating which α shoulder concavity angle,
would reduce the wear and tear of FSW tool. β half angle of beam divergence,
ii. The laser energy absorption coefficient for δ extent of slip,
shot blasted steel plates coated with zinc based λ laser beam wavelength,
primer was found to be around 0.63. ρ density of plate material,
546
547
548
ABSTRACT: Concrete is commonly used in many civil engineering structures, often re-enforced in some
way with steel, either as tensioning rods or as part of a composite structure. Composite steel-concrete
structures have been used in bridge building for many years, with rules for design freely available (British
Standards Institution, 2008a,b), and has been a topic of research for maritime structures, e.g. for arctic
structures (Matsuishi et al. 1985) or impact resistance (Iwata & Hattori 1994).
This paper explores the potential of steel-concrete-steel sandwich in floating structures and exposes
the advantages, disadvantages and challenges of using this material through two design studies. The use
of light weight concrete (with a density as low as 1000 kg/m3) is compared with more traditional higher
strength concretes (density 2500 kg/m3) and it is found that the lighter weight concrete is desirable if the
sandwich is to be competitive with steel designs on weight. The paper concludes by proposing where this
technology may have potential and suggests areas for further exploration.
549
550
Characteristic cylindrical
Density compression strength Mean tensile strength Youngs modulus
little to the bending stiffness. This will cause a shift designer, these include composite steel and concrete
of the neutral axis towards the compressive face of structures for civil applications (British Standards
the panel which should be accounted for in design. Institution 2008a,b). In addition, Eurocode 2 on
the design of concrete structures (British Stand-
3.2 Transverse shear failure ards Institution 2007) has sections on lightweight
concrete and on design of unreinforced concrete.
Transverse shear is the force acting to transit the
The Steel Construction Institute published the
load from the face plate through the bonds into
“Design Guide for Steel-Concrete-Steel Sandwich
the core. The bond between the steel and concrete
Construction—Volume 1: General principles and
must be strong enough to take this load otherwise
rules for basic elements” (Narayanan et al. 1994).
debonding will occur. Such debonding can lead to
This is the only document which was found to
a catastrophic failure of the sandwich and is akin
address the design of elements of an SCS panel.
to loosing the flange of a steel I-beam.
It provides rules for the design of sandwich ele-
ments considering ultimate and service limit states.
3.3 Vertical shear cracks
It does not cover the fatigue limit state.
These cracks occur when the shear force transferred These codes can be used with Class Society Rules,
from the face plate through the bond into the core guidance documents and standards (Det Norske
is greater than the tensile capacity of the concrete. Veritas 2003, 2007, 2008 and 2010a,b) to assess an
These cracks will eventually extend through the SCS structure for use in the marine environment.
core between the faces following the direction of To use the above combination of standards
the shear forces at 45 degrees. connectors must be provided to improve the bond
between the concrete and the steel face plates,
3.4 Summary see Figure 2. These standards effectively forbid
adhesive bonding between layers, for example in
The sandwich structure can be designed using
section 6, paragraph A502 of DNV-OSS- C502
traditional sandwich theory suited to concrete
(Det Norske Veritas 2007).
sandwich structures, modified to take account of
This rule set does not consider in any detail
cracks in the concrete core.
how to join structural elements together, or how
This cracking is clearly undesirable and should
to join structural elements to traditional stiffened
be avoided for ‘normal’ expected load conditions.
steel structures. The civil engineering standards
However, it is important to understand the crack-
assume a homogeneous structure is built in
ing behaviour of the core material for any ultimate
‘one shot’ rather than by assembling prefabricated
limit state analysis.
SCS panels. Thus, joining of structural elements
remains an area for further detailed investigation.
4 DESIGNING A SCS SANDWICH
STRUCTURE 4.2 Connecting stud design
Connecting studs provide the important bonding
4.1 Design codes
between the face plates and the concrete core and
Because steel and concrete are widely used engi- transfer the shear loads into the core. The shear
neering materials many standards exist to guide the stress transferred into the core should be limited
551
Cylindrical
studded connector
Figure 2. SCS beam with shear studs. Maximum spacing indicated to avoid complete cracking of the core between
the steel faces.
Figure 3. SCS beam with shear studs. Maximum spacing to avoid wrinkling of compression plate when subject to
in-plane loads.
552
16 Capacity curve
Applied load
12
Increase in bending
10 capacity of ≈ 55%
8
6 Pre-stressing force
4
0
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7
-2
-4
-6
BENDING MOMENT [MNm]
Figure 4. Capacity curve for a SCS beam with a core of high strength concrete. Core thickness is 200 mm sandwiched
between two steel face plates each 10 mm thick.
Floating Point
Pre-stressing force
4
Increase in bending
capacity of ≈ 9%
2
0
0 0,05 0,1 0,15 0,2 0,25 0,3 0,35 0,4 0,45 0,5
-2
-4
-6
BENDING MOMENT [MNm]
Figure 5. Capacity curve for a SCS beam with a core of light weight concrete. Core thickness is 200 mm sandwiched
between two steel face plates each 10 mm thick.
bending capacity can be achieved for the highest Knowledge gained from these studies could there-
strength concrete. fore indicate what span was feasible to cover as well
However, the addition of light weight aggre- as the expected self weight of the panels. Based on
gates to make light weight cements reduces the this knowledge two typical marine applications
advantages of pre or post stressing considerably as were selected to compare original steel design with
shown in Figure 5, where the additional bending an SCS sandwich solution.
capacity is reduced to only 9%. It can therefore be
concluded that the additional effort of pre or post 5.1 RoRo ship
stressing the concrete core is not worthwhile for
light weight concretes. The main deck above the machinery space of a
large RoRo car carrier was chosen. This deck was
designed to carry a range of vehicles and for a uni-
5 DESIGN STUDIES form load of 1.0 tonnes/m2. The deck is supported
fore and aft by bulkheads (spaced over 100 metres
Parametric studies were conducted to give an apart), with longitudinal and transverse girders
indication of SCS panel performance for differ- supported periodically by pillars.
ent cross sectional parameters, loads and bound- Two different options are considered for the
ary conditions (Weitzenböck & Grafton 2010). comparison, the first, number 1 in Figure 6, is
553
• Transverse Girder
• Longitudinal Girder
• Pillar
Figure 6. RoRo deck structural girder system. Shaded areas show where SCS panels would replace existing steel
structure.
an 11.4 metre simply supported panel between and 700 USD pr ton steel. The material costs used
longitudinal girders. The second and third aim to in the comparison are based on discussion with
optimise the design by using SCS panels over three shipyards in Asia and are only used only to give
spans as a continuous beam (of approximately an indication of the cost. The material costs for the
30 metres span). The continuous beam will reduce existing steel structure, plates, stiffeners and trans-
the bending moment and may allow a lighter verse girders, is 115 USD/m2.
panel to carry the load. This option is number 2 in The study revealed the following advantages and
Figure 6. The second longer span panel is designed challenges for replacing the existing steel structure
with both thick steel plates and thin core (2a) and with a continuous SCS panel.
thin steel plates and thick core (2b). The light
weight concrete described previously is used to 5.1.1 Advantages
obtain the lightest weight SCS structure. − Total depth of deck is decreased from 600 to
SCS panel design is controlled by ultimate 170 mm. This 72% reduction in deck height
limit state, i.e. bending and shear caused by the gives a potential to install an extra deck, with
local static loads from cargo and self weight. economical benefits, if the same reduction
The fatigue design life is checked for dynamic applies for several decks.
axial forces, acting in the longitudinal direction − Elimination of stiffeners and some girders,
of the vessel, caused by global hogging/sagging reducing fabrication complexity and area of deck
moment. As a reference, the maximum deflection for coating and fire protection (where needed).
at midspan is calculated as a serviceability limit − With a well-established welding procedure the
state. The maximum deflection was found to be welding of the studs will be very efficient. The
75 mm, which was considered acceptable. Note total amount of welding will also be reduced
that racking loads, which can influence the size of compared to the existing steel structure.
the transverse girders for this type of ship, have − The panel is very robust and has a large potential
not been considered. The concrete core and con- to absorb energy from impact loading or rough
necting studs are assumed sufficient to withstand handling. This is especially useful for decks sub-
the local loads acting through the tyres of the jected to wheel loading.
vehicles. − The panel will be easier to maintain due to the
Table 2 shows the SCS panels proposed to replace flat surfaces.
the existing steel structure. Options 2a and 2b, with
a continuous panel over three spans, are the most 5.1.2 Challenges
advantageous with respect to weight and material − Total weight of the panel is increased by 101%,
costs. Option 2a, with thick steel plates and thin from 162 to at best 326 kg/m2.
core, has the lightest weight and option 2b, with thin − Estimated material costs are increased by 33%,
steel plates and thick core, has the lowest material from 113.4 to 150.7 USD/m2.
costs. The material costs in Table 2 are based on a − Assembly, connection and production of the
material cost estimate of 300 USD pr ton concrete panels have to be investigated further.
554
(*) Normal steel with 235 MPa yield strength; (**) Per metre squared of deck area.
555
Figure 7. Structural arrangement of the process deck, showing underdeck stiffeners and girders. SCS panel shown
shaded. Total SCS panel assumed to be constructed of interconnected panels placed atop the existing girder system.
Figure 8. Static model of the FPSO process deck, showing point loads for heavy equipment.
(*) High strength steel with 355 MPa yield strength; (**) Per metre squared of deck area.
556
5.3 Limitations
7 FURTHER WORK
The limitation of these comparisons is that the
main structural system used for both the conven- Before competitive SCS structures can be built
tional steel structure and the SCS panels is the same. some areas of further research are needed:
This system is the result of an optimization process
meant for a steel structure. In the studies SCS − Studies need to be completed to demonstrate
panels are used in place of parts of the structure cost effectiveness, for both initial cost and
and only the SCS panels are optimised, no changes through life cost.
are made to the existing steel structure. − Joints between structural elements need to be
carefully designed, qualified and tested, using
for example the DNV technology qualification
process (Det Norske Veritas 2001 and 2006).
6 HOW TO EXPLOIT SCS STRUCTURES
557
558
ABSTRACT: When welding block joints of ships it may occur that the gaps between them are relatively
large. Furthermore high stresses due to the constraint conditions may result in high residual stresses.
Therefore an investigation was performed to determine if gap widths up to 30 mm fulfil the requirements
regarding the fatigue and fracture strength. The welding of butt joints at 250 mm wide plates was per-
formed under definite constraints presenting the surrounding ship structure. Among other parameters
the welding method (string-bead, weaving technique) was varied. During the welding the reaction force
was measured and it was noticed that it was remarkably higher for the weaving technique compared to the
string-bead technique. Both welding techniques lead to different fracture and fatigue results. The weaving
technique seems to be more adequate when welding large gaps. A numerical welding simulation using the
weaving technique showed compressive plastic strains in the root layer that maybe reasonable for the good
results. Generally the investigations indicated a clear influence of welding technique on the fatigue and
fracture strength. The gap width did not have a significant influence.
1 INTRODUCTION
559
600
Base Filler
material material
Steel grade D36 FILARC PZ 6113 Figure 5. Finite element model for the string-bead tech-
nique and computed stresses.
Yield stress fy 355 MPa For protective gas
Actual yield stress 390 MPa MISON 18:
460 MPa Table 2. Load types for the fatigue tests.
Tensile stress 520 MPa For protective gas
Load types I II
MISON 18:
460 MPa
Δσ [MPa] 270 270
Ultimate tensile strain 29% Min. 22% σupper [MPa] 270 Max. 400
Welding procedure Automatic MAG Stress ratio R [−] 0 Max. 0.325
560
Ps=10%
R=0,27
140
R=0 obtained. The figure shows that the scatter bands
R=0
are very different for the welding techniques. For
Cycles [MPa]
R=0,27 Ps=50%
R=0,32 R=0
111 Ps=90%
the weaving technique the influence of the raised
100 107 R=0,32
99 96
Ps=97,7% stress ratio R clearly leads to a reduced fatigue
85
78
85
76
strength. For the string-bead technique, however,
60 almost no influence regarding the stress ratio could
8 mm gap width 30 mm gap width
be found. Generally, for small gaps the weaving
technique seems to be adequate, for larger gaps the
Figure 6. Scatter bands of the fatigue strength.
string-bead technique has a better fatigue strength.
Regarding the influence of the gap width itself no
Generally one can predict or compute the clear tendency could be found. It can be concluded
crack initiation point due the specimen’s bend- that the influence of the increased gap width can
ing distortion resulting from the weld process. be neglected.
For the specimens welded with the weaving tech- The local approaches include geometrical
nique the predictions matched with the tests, for and distortion effects. The two mainly applied
the string-bead welded specimens, however, they approaches are the structural and the notch
did not. The expected crack initiation point is at stress approach. Distortion effects as the angular
the top layer, but the crack started from the root misalignment are involved in both procedures. The
notch. There are several possible reasons for that, notch stress approach additionally includes the
e.g. a worse notch geometry at the root notch. local notch geometry. For the local approaches as
Due to this the notch stresses were computed with well as for the nominal stress approach the clas-
the finite element program ANSYS, taking into sification societies require a minimum fatigue
account the actual notch geometry as well as the class (FAT) to fulfil (Germanischer Lloyd, 2009)
bending distortions that were measured in terms (Hobbacher, 2009):
of angular misalignments. − nominal stress approach—FAT90
When comparing the maximum stresses in − structural stress approach—FAT100
Figure 6 for the root and the top layer it can be − notch stress approach—FAT225
seen that both reach a similar stress level. The
unsymmetrical stress distribution on the left and Figure 7 compares the resulting FAT classes
right side results from the misalignment. The flank from the nominal and local stress approaches for a
angles of the string-bead technique specimens gap width of 30 mm.
were often smaller than 135° and therefore critical. For each approach and welding technique a rec-
But there were also specimens having larger flank tangle is illustrated representing as upper or lower
angle but they, too, cracked from the root. That boundary the result for the different stress ratios
means the large flank angles are a possible reason that were tested. Referring to this it can be seen
for the unexpected crack initiation location, but that for the string-bead technique the FAT classes
due to the fact that all specimens independently for R > 0 are higher than for R = 0. The weaving
from the flank angle cracked from the root a fur- technique has opposite results considering the
ther influence is assumed to be reasonable, i.e. the stress ratio’s influence, as expected.
material properties in the heat affected zone can be With regard to the rule’s requirements for the
an important issue. three concepts it can be stated that the results of
the weaving technique fulfill the requirements.
For the string-bead technique specimens, however,
4 EVALUATION OF THE FATIGUE TESTS
300
R=0
For the evaluation the nominal stress as well as two 30 mm gap width
local stress approaches were used (Niemi, 2006),
FAT classes [MPa]
S=String-bead technique S
200 P=Weaving technique
(Radaj, 2006). The evaluation based on the nomi- R>0 R>0
nal stress approach is independent of the local R>0
P R=0 R=0
P
R=0 S
weld geometry and the imperfections of the speci- 100 R>0
R=0 R>0
mens, the result only refers to the applied load and S
R>0
R=0 P
the number of cycles to failure. Therefore all local
0
effects affecting the failure of the specimens are nominal
1 stress
2 structural
3 stress
4 notch5 stress r = 16 mm
implicitly included in this procedure.
When extrapolating the fatigue strength for Figure 7. Comparison of the FAT classes derived from
the survival probabilities Ps = 10%, 50%, 90% and the nominal and local approaches for 30 mm gap width.
561
8 mm 30 mm
562
0,8
Weld Material Weaving Technique
0,6
0,4
0,2
Weld Material String-beadTechnique
0
8 mm gap width 30 mm gap width
563
564
565
0,02 ACKNOWLEDGEMENTS
Transverse Plastic Strain [-]
566
ABSTRACT: Fatigue aspects are leading criterion for the scantling optimization of ships structures.
LBR-5 software is an integrated package which performs, at the early design stage, cost and/or weight opti-
mization of ships structures (scantling). This software takes into account structural constraints (yielding,
buckling, displacement, ultimate strength). Until now, the fatigue failure issue was not implicitly included in
the optimization loop. At the early design stage, limited information and details are available. Therefore sim-
plified methods have to be applied. LBR-5 software provides the nominal stress, so chosen procedure uses the
nominal stress and the Miner’s rule. This approach requires a library of stress concentration factors for pre-
defined typical structural details. The goal of the present work is to develop a tool for fatigue assessment that
can be integrated on LBR5 optimization process, to realize global optimization taking into account the fatigue
issues. Until now, the integration of the fatigue tool is not finished. Only fatigue verification on the optimized
structure is possible. On the present paper, the procedure adopted is presented and a longitudinal scantling of
the mid-ship section of a LNG is optimized with LBR-5. Production cost is considered as objective functions.
The optimized scantling is checked by the fatigue tool. The fatigue criterion is not respected on some panels.
Corrections are performed on these panels to have an optimized scantling without fatigue problems.
569
objective function (not empirical). For least Crack propagation and fatigue aspects are leading
weight, these unitary costs are not used and the criterion for the scantling optimization of ships
objective function depends only on the geometri- and FPSO structures, which are submitted to
cal parameters. Using all these data (constraints, cycling loading-unloading operations. Fatigue
objective function and sensitivity analysis), the damage is an extremely localized phenomenon.
optimum solution is found using an optimization In particular, the welded joints are regions of
algorithm CONLIN based on a convex lineariza- weakness in a structure for their low fatigue
tion of the non-linear functions (constraints and strength, due to the presence of crack-like defects,
objective functions) and on a dual approach, high stress concentration effects, and tensile resid-
(Fleury & Braibant 1986, Rigo & Fleury 2001). ual stress fields caused by thermal expansion.
Independent of the number of design variables In general, there are several approaches for fatigue
and constraints, the number of iterations requir- assessment. In Figure 2, (Radaj et al., 2006), the
ing a complete structural re-analysis is limited to most important basic variants of the methods are
10 or 15. plotted. The simplest approach is the Nominal stress
For each panel one can associate up to 9 design approach. It necessitates only the knowledge of the
variables (XI). These 9 design variables are main dimension of the structure. But it’s not very
respectively: efficient when we consider real complex structures.
On those situations, other methods more relevant
• Plate thickness,
can be used. These methods are structural hot-spot
• For longitudinal members (stiffeners, crossbars,
stress and notch stress approaches. The fracture
longitudinals, girders, etc.):
mechanics approaches with J-integral or stress
− web height and thickness,
− flange width,
− spacing between 2 longitudinal members.
• For transverse members (frames, transverse
stiffeners, etc.):
− web height and thickness,
− flange width,
− spacing between 2 transverse members
(frames).
Since 1980 the FEM has become a standard to
evaluate constraints on stress, displacement and Figure 2. Method for fatigue assessment, (Radaj et al.,
ultimate strength at each iteration. With FEM, 2006).
570
Due to the complexity of the ship structural details, 3.2.1 Nominal stress
the notch stress approach is chosen. The notch For fatigue-critical structural details nominal stress
stress method can cover different structural to be considered are the principal stresses at the
geometries and details defining fatigue strength
with one design curve.
The fatigue module has to be linked with
LBR-5 software to calculate, at the early stage
design, the fatigue damage on critical connections
of the ship structures. LBR-5 software provides
only the nominal stress. The hot spot stress will be
calculated by multiplying the nominal stress by a
hot spot stress concentration factor. In similar way,
the notch stress can be calculated by multiplying
the hot spot stress by a notch factor.
The S-N curve considered is derived from
Bureau VERITAS rules (1998), for non-tubular
joints in the as-rolled condition with no flame cut
edges and correspond to the curve B provided by
the U.K. department of Energy (1993). This basic
curve is corrected to take into account the influ- Figure 3. Nominal, hot spot and notch stresses, Bureau
ence of static and residual stress, the effect of VERITAS (2009a).
571
σ nominal ( σ1 , σ 2 ) (4)
Figure 5. Biaxial stress at weld toe, Hobbacher (2007). Figure 6. Stresses on plate/frames connections.
572
(11)
573
distribution. The two slope S-N curve is defined by The fatigue loadings (moments and pressures)
the Constant Kp. were defined in LBR5 software by an automatic
Ship structures are subjected to various types of transfer of the fatigue loads calculated by Bureau
fluctuating loads. For fatigue assessment, the main Veritas Rules from MARS software (BV) to LBR5
critical loads are induced by the wave action. optimization tool.
574
575
576
Panel 8
Plate thickness 17.61 13.17 13.2
HP320 × 12
(Continued)
577
Panel 25
Plate thickness 19 16.6 20.0
HP320 × 12
Web height 305 182.8 287.2
Web thickness 10 6.8 10.0
Flange breath 59 80.6 55.8
Flange thickness 35 10.3 32.8
Panel 26
Plate thickness 22 16.6 20.0
HP320 × 12
Web height 305 182.8 287.2
Web thickness 10 6.8 10.0
Flange breath 59 80.6 55.8
Flange thickness 35 10.3 32.8
Table 6. Initial, optimized without fatigue constraints stress concentration factors. The difference of the
and fatigue corrected scantling comparison. results compared to VERISTAR can be explained
by the fact that LBR-5 does not have the capability
Mass Gain in Cost Gain of a finite element analysis and is restricted to pris-
Scantling [tons] mass [M€] in cost
matic structures and linear 3D analysis. For the
Initial 1840.44 3.16 early design stage, the results of LBR5 are accept-
Optimization 1694.99 7.90% 3.00 5.16% able and help the designer to choose the best
(without fatigue solution (scantling). The remain work is finishing
constraints) the integration of the fatigue tool inside LBR5
Optimization 1714.13 6.86% 3.02 4.58% optimization loop.
(after fatigue
corrections)
ACKNOWLEDGEMENTS
Table 7. Optimized without fatigue constraints and The authors thank the IMPROVE PROJECT
fatigue corrected scantling comparison. (FP6) and the associated partners as part of the
present paper has been developed in the framework
Scantling Mass [tons] Cost [M€] of this project.
Optimized (without 1694.99 3
fatigue constraints) REFERENCES
Optimized & Corrected 1714.12 3.02
fatigue Bureau VERITAS, 1998. Fatigue strength of welded ship
Difference 19.14 0.02 structures. Paris, Bureau Veritas.
% 1.13% 0.51% Bureau VERITAS, 2009a. Bureau VERITAS rules, Pt B,
Ch 7, Sec 4, Paris, Bureau Veritas.
Bureau VERITAS, 2009b. Bureau VERITAS rules Pt B,
Ch 12, Sec 2, Paris, Bureau Veritas.
5 CONCLUSIONS D.N.V., 2005. Fatigue assessment of ship structures.
Classification notes No. 30.7. Hovik, Norway.
A fatigue tool has been developed in the frame- Fleury, C. & Braibant, V., 1986. Structural Optimization:
work of the IMPROVE Project. A methodology is A New Dual Method using Mixed Variables. Int
chosen by taking into account of the capabilities J Numer Methods Eng; 23:28.
of the LBR5 software and the need to be used early GERMANISCHER Lloyd, 2007. GERMANISCHER
deigns stage. Some tests were performed by com- Lloyd rules.
paring the results to the results obtained with Hobbacher, A., 2007, IIW Joint Working Group XIII–XV.
Recommendations for fatigue design of welded joints
VERISTAR software provided by Bureau VER- and components. IIW document XIII-2151-07/
ITAS. This comparison shows that the LBR5 XV-1254-07. Paris, France.
results are not perfect and further tests must be Hughes, O.F., 1983. Ship Structural Design: A Rationally-
performed to calibrate more the fatigue module Based, Computer-Aided, Optimization Approach,
and to add more critical details on a library of John Wiley & Sons, NY.
578
579
M. André
Technical University of Catalonia, Barcelona, Spain
ABSTRACT: The issue of the environmental impact of shipping activities is treated in this work from
the specific viewpoint of underwater noise emissions from the vessels. There has been recently a rising con-
cern about the negative effects that this kind of emissions can have on the marine wildlife in general and
in particular on marine mammals. In these animals, acoustic communication and perception has acquired
a privileged role compared with other senses and other zoological groups. The paper examines the role in
this scenario of noise emissions from shipping and explores how their impact can be quantified, in order
to establish a target for the control of such emissions. The standards recently issued for the characterisa-
tion of underwater noise signature by commercial vessels are also reviewed. The paper builds on the first
results of the project SILENV, funded with the 7th Framework programme of the European Union.
581
582
583
584
2500
4 MEASUREMENT OF SHIP
UNDERWATER NOISE
585
586
587
10
3
tions of sources and receivers are needed, too.
In a simple case, one ship, one emitter and one
receiver can be considered, while in a real case mul-
tiple ships, multiple receivers and multiple emitters
2
can interact with each others.
10
A further simplification consists in omitting the
characterisation of both the emitting and receiving
individuals and concentrating on the disturbance
10
2
10
3
10
4 (ship radiated noise spectrum). In this case it can
Frequency [Hz]
be set as a target that the ship at a certain distance
does not contribute more than the (natural) back-
Figure 8. Critical bandwidth vs. 1/3 octave bandwidth
(for humans).
ground noise (or, similarly, the maximum distance
from the ship at which this occurs can be set). This
approach is effective for preventing both com-
Accordingly, the power contents featured by a munication and behavioural changes, but is quite
masker within the critical bandwidth sum up in conservative; in fact it is plausible that noise levels
affecting the capabilities of perceiving single tonal higher than the background noise could even not
components of sound in the same range. On the be heard in a certain range by animals or, if heard,
contrary, masker signal components acting out of could not interfere with communication.
critical bandwidth of the signal do not affect the Improving the accuracy of the model, further
perception of the signal itself. A typical situation elements could be introduced in the analysis in
is represented by a tonal signal corresponding to a addition to the ship radiated noise.
marine mammal vocalisation and a masker effect As regards communication aspects, if the char-
due to the broadband noise emitted by a ship. acteristics of the emitting individual are available
Critical bandwidths vs. the signal frequency are (vocalisation spectrum), the target can be set in
shown in Figure 8 for humans. The figure shows maximising the area where the signal emitted over-
that the critical bandwidth follows quite closely comes the ship-dominated background noise. The
the line corresponding to a third-octave bandwidth emitted vocalisation results to be ‘perceivable’ in
in the frequency region above 400–500 Hz, while, such area. This condition, however, is necessary
below that region, the critical bandwidth tends to but not sufficient to ensure that the sound is actu-
be constant (and larger than a 1/3 octave). In pro- ally perceived by a receiving individual placed in
portion, the masking effect appears therefore to be that area, because this depends also on the receiv-
potentially larger for low frequencies. ing capabilities.
What above, however, refers to the human ear If the focus is on the prevention of behavioural
and is not applicable for other species. Further, it is changes, in addition to the ship radiated noise, the
noted that a direct quantification of critical band- hearing sensibilities (audiograms) of the receiver
widths is, for large cetaceans, probably even more could be characterised in order to evaluate the har-
difficult than a survey of audiograms. assment due to noise.
588
1 Behavioural x
Communication x
2 Behavioural x x
Communication x x
3 Communication x x x
589
7 CONCLUSIONS AND FUTURE This work was developed in the frame of the col-
PERSPECTIVES laborative project SILENV—Ships oriented Inno-
vative soLutions to rEduce Noise & Vibrations,
The noise emitted by ships in water has certainly funded by the E.U. within the Call FP7-SST-2008-
a responsibility in raising the background noise in RTD-1 Grant Agreement SCP8-GA-2009-234182.
590
591
ABSTRACT: The work addresses the problem of the ship noise emissions towards the inner part of
the vessel and towards the external space. In the former case, the focus is on the air-borne and structure-
borne transmission to living and working spaces on board. Limits and regulations are devoted primarily
to the preservation of the worker’s health and integrity and to the attainment of comfortable conditions in
resting and recreational spaces for crew and passengers. As regards airborne noise emissions from the ship,
the main concern is for the inhabitants of the living areas near ports and along the coast in case of heavy
ship traffic. Finally, noise emissions in water are considered; in this case, the concern is about their negative
impact on the marine wildlife. The present situation in terms of national, regional and international
requirements and the trends in the normative framework for the various emissions are highlighted.
593
594
595
Insulation index
Space dB
Cabin to cabin la = 30
Mess rooms, recreation rooms la = 40
to cabins and hospitals
Reduction
Type of ear
protection dB(A)
596
597
For what concerns noise inside the ship, limit levels None of these standards, however, contain limits
are mainly fixed, for the various locations, on the or requirements to be fulfilled.
perceived power [instantaneous levels in dB(A)].
By instantaneous is here meant a value that is 3.2 European regulations
obtained from a short term average [of the order of As regards the European legislative framework,
minutes]. In the documents devoted to workers’ health the acoustic indicators currently adopted for
(in particular IMO 1981), limits are fixed on the total assessing environmental noise are mainly focused
amount of perceived energy during the working shift. on noise classified in two large categories: mobile
In fact, requirements are set in terms of both sound sources (like road-traffic, rail-traffic and aircraft
pressure levels and exposure time. Limits are fixed around airport) and stationary sources, like
in order not to overcome reference equivalent levels those represented by industrial plants or whole
corresponding to a continuous exposure to a given settlements.
level for a given duration [f.i. in IMO (1981): 80 dB(A) Ships are not specifically mentioned; in princi-
for 24 hours (LAeq(24) = 80 dB(A)]. Problems in apply- ple, however, they belong to both categories as said
ing this approach are represented by the difficulty in before. Two main European directives deal with the
evaluating precisely exposure times and levels for the problem of airborne noise: the European Directive
various categories of workers. 2002/49/EC (EU 2002) and the European Directive
Both methods described above do not capture 2003/44/EC (EU 2003b). The first one applies to
possible concentrations of energy in time or fre- environmental noise to which humans are exposed
quency. Therefore, the annoyance due to inter- in particular built-up areas, in public parks or other
mittent sounds or tonal components may not be quiet areas in urban agglomerations, in quiet areas
adequately reflected in the indicators so far used in open country, near schools, hospitals and other
in the rules. noise-sensitive buildings and areas. The indicators
Moreover, the general trend towards a reduction used are Lden and Lnight (day-evening-night levels),
of broad-band noise levels has probably enhanced that are equivalent levels defined as follow:
the effect of these components concentrated in
time and/or in frequency.
1 ⎛ ⎞
Lday Lev . + 5 Lnight
i + 10
Lden l
log ⎜ 12 ⋅ 10 10 + 4 ⋅ 10 10 + 8 ⋅ 10 10
⎟
24 ⎝ ⎠
3 NOISE EMISSIONS IN AIR
(3)
Air noise pollution from ships affects people living
near ports or channels interested by intense ship- where Lday, Levening, Lnight are A-weighted long-term
ping traffic. When the ship is characterised as a average sound level as defined in ISO (1987).
whole source, it must be considered as a moving The aim of these indicators is to correlate the
source during navigation in ports or channels and a measurement of sound pressure level and the per-
fixed source at quay. Within the ship, main sources centage of people who have negative effects on
of airborne noise are: funnels, air intakes, air dis- their health due to a prolonged exposure to an
chargers and in general all those openings which put examined noise source.
the internal sources (ventilation systems, engines) in In EU (2003b) limits for airborne noise are given
communication with the external environment. In for recreational crafts in function of the engine
the following a brief review of the normative frame- power (see Table 4). Such limits are to be measured
work existing at different levels will be provided. according to, ISO (2007), ISO (2009).
598
599
600
601
602
ABSTRACT: The present study has been performed by the University of Trieste within the BESST
European Project with the aim to design an operational procedure for predicting ship’s outdoor noise
in harbour. Procedure is here outlined in all its steps from the pre- to the post-processing phase and its
effectiveness is proved. Method is implemented by application to a comprehensive case study. Numerical
simulations are performed by exploiting a selected commercial software suited for acoustic mapping on
noise emission in open field conditions. Results are then discussed with the aim to check operational
procedure and computational tool. Basing on the same results, suggestions on the proper technique to
approach the topical subject of ship’s outdoor noise in harbour are given. Paper starts with an introduction
covering the matter of ship’s noise pollution in harbour and how it is today managed.
603
604
605
606
607
608
609
REFERENCES
610
ABSTRACT: Fluid-Structure Interaction problems are solved by applying a smoothed particle hydro-
dynamics method to a weakly compressible Navier-Stokes equation as well as an equilibrium equation
for geometrically nonlinear structures in updated Lagrangian formulation. The geometrically exact
interface, consisting of B-spline basis functions and the corresponding control points, includes the high
order geometric information such as tangent, normal, and curvature. The exactness of interface is kept
by updating the control points according to the kinematics obtained from response analysis. Under the
scheme of explicit time integration and updated Lagrangian formulation, the required shape design
velocity should be updated at every single step. The update scheme of design velocity is developed
using the sensitivity of physical velocity. The developed sensitivity analysis method is further utilized in
gradient-based shape optimization problems and turns out to be very efficient since the interaction pairs
of particles determined in the response analysis can be directly utilized.
611
⎧ 3 ⎛ | ab | ⎞ 2 3 ⎛ | ab | ⎞ 3 |x |
⎪1 − ⎝ ⎠
+
⎝ ⎠
, 0 ≤ ab ≤ 1,
⎪ 2 h h h
= α ddim ⎪⎪ 1 3
⎛ | x ab | ⎞ |x |
dim ⎨ ⎝2 − ⎠
, 1 ≤ ab ≤ 2,
h ⎪ 4 h h
⎪ | x ab |
⎪0, > 2,
⎪⎩ h
(1)
where (•)ab = (•)a − (•)b. For two-dimensional sys-
tem, the dimensional parameter αdim is defined as
α2 = 10/7π. Figure 2. Geometrically exact initial interface.
612
613
⎛ n σa n σb n ⎞
+ ∑ mb ⎜
n n n
+ + Π ab l + Θ abb ⋅ (∇
( ∇ a Wabb )′ + ( f ′ a )
b ⎝ ρa
n 2 n 2
ρb ⎠
(17)
5 NUMERICAL EXAMPLES
614
615
Ωf
Ωs Ωs Ωs
616
The FSI problems are solved by applying a smoothed This research was supported by Basic Science
particle hydrodynamics method to a weakly com- Research Program through the National Research
pressible Navier-Stokes equation as well as an Foundation of Korea (NRF) funded by the
equilibrium equation for geometrically nonlinear Ministry of Education, Science and Technology
structures in updated Lagrangian formulation. The (Grant Number 2010-18282). The support is grate-
CAD-based geometrically exact interface includes fully acknowledged.
the high order geometric information such as tan-
gent, normal, and curvature. The exactness of inter-
face is kept by updating the control points according REFERENCES
to the kinematics directly obtained from the solid
particles. In the SPH framework, the analytical Antoci, C.A., Gallati, M., & Sibilla, S. 2007, Numerical
shape sensitivity based on the direct differentiation simulation of fluid-structure interaction by SPH.
is derived from the strong form, considering the Computers and Structures, 85: 879–890.
design dependency of the kernel function and the Haug, E.J., Choi, K.K., & Komkov, V. 1986, Design
Sensitivity Analysis of Structural Systems. Orlando:
interface conditions. Under the scheme of explicit Academic.
time integration and updated Lagrangian formula- Ha, Y.D., & Cho, S. 2010, Efficient design sensitivity
tion, the required shape design velocity should be analysis of incompressible fluids using SPH projection
updated at every single step. The update scheme of method. Structural and Multidisciplinary Optimization,
design velocity is developed using the sensitivity of 40(1): 307–318.
physical velocity. Through some numerical exam- Monaghan, J.J., & Lattanzio, J.C. 1985, A refined parti-
ples, the necessity and efficiency of the analytical cle method for astrophysical problems. Astronomy and
shape DSA method is discussed. The developed Astrophysics, 149: 135–143.
sensitivity analysis method is further utilized in Piegl, L., & Tiller, W. 1997, The NURBS Book. 2nd edn.
New York: Springer.
gradient-based shape optimization problems and
turns out to be very efficient since the interaction
pairs of particles determined in the response analy-
sis can be directly utilized.
617
S. Ehlers
Aalto University, School of Science and Technology, Marine Technology, Espoo, Finland
ABSTRACT: The continuous demand for efficient structures contributes to the development of light-
weight steel structures, such as steel sandwich panels. However, these thin steel sandwich panels need to
be joined to one-another and to conventional structures. Hence, an appropriate joint shape needs to be
identified with an as low as possible equivalent stress level due to the joints’ sensitivity to fatigue. There-
fore, the aim of this paper is to identify the optimum joint shape using a finite element-based optimisation
procedure together with a B-spline based parametric joint shape. The optimization is performed with two
objectives, i.e. to minimize the stress and mass of the joint. As a result, a series of optimum joint shapes
and their characteristics will be presented and discussed, followed by an outlook of future work aspects
related to the joint fabrication and welding.
619
L
s3x
s2x
t1+t3
t1+t3
y
x
s3y
3
s2y
s1y
fr
t3
H
2
t1
1
t2
s1x
Figure 2. The parametric joint shape (grey section is Figure 3. Discretized joint and applied boundary
optional). conditions.
620
the MS EXCEL-file and the optimum solution can 0.7 mass(min stress)
stress
0.6
be obtained. 0.5
0.4
0.3
0.1
The optimization procedure is validated with a 0
621
0.8
mass However, due to the asymmetric support second-
stress (min mass)
0.7 mass (min stress) ary bending is invoked resulting in the ‘swan-neck’
0.6
stress
type of joint shape.
0.5
0.4
Considering the very simple boundary/loading
0.3
condition, the B-spline n°2 is probably almost
0.2 completely unaffected by the optimization process.
0.1
0
The similarity of the shape for both minima alter-
0 10 20 30 40 50 60
Generation
70 80 90 100 110
natives in each case confirmed this. Thus indi-
cating that case 2, without the additional grey
Figure 7. Optimization progress through generation section, is the preferred joint shape. Additionally,
(case 1). the results clearly indicate that the joint shape with
622
0.9
4 CONCLUSIONS
mass
0.8 stress(min mass)
0.5
optimisation procedure for steel sandwich panel
0.4 joints. Therefore, the optimisation procedure was
0.3
described, validated and used to identify joint
0.2
0.1
shapes having a reduced stress and mass. As a
0
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140
result, the obtained joint shapes are presented and
Generation technical feasibility is discussed.
Furthermore, the identification of the optimised
Figure 10. Optimization progress through generation joint shape illustrates the basic principles of the pro-
(case 2). cedure. In the future, the procedure can be applied
to determine also the principal joint dimensions
in practical applications, namely the joint length
and height, considering shape limitations and vari-
ous boundary conditions. Additionally, further
research should concern the production complex-
ity and production cost to obtain optimized and
industrially feasible steel sandwich panel joints.
623
A. Niemelä
STX Europe, Turku, Finland
S. Bralic
Brze Vise Bolje d.o.o, Zagreb, Croatia
H. Naar
MEC Insenerilahendused, Tallinn, Estonia
ABSTRACT: The paper investigates the interaction between the hull and the superstructure in opti-
mized passenger ships when exposed to bending loads. The investigation is done using the Coupled Beams
theory which extends the basic beam theory to account vertical and shear stiffness between various decks.
Optimization of passenger ship with respect to weight and Vertical Centre of Gravity (VCG) is carried
out to create a set of Pareto-optimal solutions. The responses of these designs are compared in detail. The
investigation shows that the vertical and shear coupling between different decks significantly affects the
response of the passenger ships and changes load carrying mechanism of the hull girder. In the weight
optimal design the vertical bending moment is shared equally by the hull and the superstructure, while
in VCG optimal design the neutral axis approaches bottom plating of the ship considerably increasing
the share of load carrying of the superstructure. This means that the global response evaluation needs
to include vertical and shear coupling along whole length of the ship, and thus the simplified 2D section
models are not adequate for the conceptual design of passenger ship structures.
625
626
627
Table 1. The weight and VCG optimal designs. First line indicates the plate thickness, stiffener spacing
and yield strength of the material, while the second line indicates the bulb flat dimensions and material. The
units are mm and MPa.
628
d 2w
ε κ = −z M
(2)
dx 2
dw
γ =
Q
(3)
dx
629
630
631
632
633
3 INPUT DATA
3.1 Environment
Environmental data summarizes the characteris-
tics of the site, data related to fatigue analysis and
load cases considered for the ultimate limit state
analysis.
Firstly, site data contains the values for water
depth d, the power law exponent α characterizing
the vertical distribution of wind speeds over the
tower height, the densities of the air ρa and of the
sea water ρw.
Secondly, for fatigue concern a distinction is
made between waves and wind actions. On one
hand, waves participation is presented under the
form of a list of sea states (or scatter diagram), each
one being characterized by a significant wave height Figure 3. Screen shot of the optimization tool.
HS, a mean zero up-crossing period TZ and a per-
centage of occurrence of the sea state PSS. On the
other hand, spectrums of punctual tower top loads monopile structure: the hub height Hhub, the
are used to describe fluctuating wind loadings on number of blades n, the rotor diameter drotor, the
the structure. weight of the rotor-nacelle assembly mTop, the tech-
Finally, data related to ultimate limit states are nical design lifetime of the turbine and its range of
listed under the form of a series of environmen- rotational speed.
tal situations and their associated wind and waves The interface of the optimization tool is pre-
conditions: average wind speed at hub height Vhub, sented on Figure 3.
water level elevation Δd compared to the mean still
water level MSL (elevation due to tide or storm for 4 ASSESSMENT OF CONSTRAINTS
example), wave height Hw and period Tw and a set
of punctual tower top loads. 4.1 Generals
The constraints implemented in the optimization
3.2 Support structure process are typically divided in two categories: geo-
metrical and structural constraints.
The monopile offshore wind tower considered in the
The first type of constraints refers to geomet-
study is an assembly of several conical or cylindri-
rical requirements such as equality of shell rings
cal tubular segments. The segments themselves are
diameters in one segment or decrease of shell
made of shell rings linked together with butt welds.
thicknesses and diameters while progressing to the
Each shell ring is characterized by a shell thickness,
top of the monopile structure.
upper and lower diameters, a height, a steel grade
The second type represents structural constraints
and a category of detail for the butt weld.
related to the verification of the structural integrity
Four framing systems are envisaged for the
of the offshore wind turbine towards fatigue, ultimate
scantling of shells belonging to the same tower
limit states and resonance phenomena. The verifica-
segment: unstiffened, longitudinally stiffened, ring
tion of those constraints requires the use of either
stiffened or orthogonally stiffened shells.
quasi-static or dynamic analysis of the structure.
634
Figure 4. The 2D dynamic model of the offshore wind Figure 5. Campbell diagram for a 3-bladed offshore
turbine studied in the optimization process. wind turbine.
635
⎛ πD 2 ⎞
Ai Cm ρw ⎜ i ⎟ LF ,i (4)
⎝ 4 ⎠
636
ΔMwiind ,i
Swiind ,i = (7)
Wi
637
638
639
Wind turbine power 5 [MW] Reference wind speed Vref 37.5 [m/s]
Number of blades 3 [−] Average wind speed Vave 10 [m/s]
Rotor diameter 118 [m] Intensity of turbulences I 0.18 [−]
Nacelle mass (incl. rotor 390 [tons]
blades)
Speed range 4−14.2 [min−1] Table 4. Values of cumulative fatigue damages for the
initial scantling S235.
640
641
642
643
ξ ξi ξi ξ and
Ni p (ξ ) Ni p −1 (ξ ) + p +1
ξi p ξi ξi p +1 ξi +1 z(Ξ) = ΣWI (Ξ)zI. (11)
× Ni 1 p −1 (ξ ) , p = 1, 2, 3, … (3)
Note that the NURBS basis functions are not
interpolatory. The variational equation (7) can be
Next, a B-spline curve is obtained from the written as
linear combination of B-spline basis functions
and corresponding control points, which are the
∫Ω ∑ J ∇ΞWJ
T
coefficients for each B-spline basis. For the given a( , ) : C : z I ∇TΞWI Ω d ΩΞ (12)
Ξ
n pairs of p-th order B-spline basis functions Ni,p I ,J
∑ Ni p (ξ )Bi .
J
C (ξ ) (4)
∫Γ ∑ zJ WJ JΓ d Γ Ξ.
^ −1
i =1 + N
T
(13)
Ξ
J
Similarly, the B-spline surfaces and solids are
defined as a tensor product of coordinates as Note that the domain and boundary are still in
a continuum form and constructed by the com-
n m
bination of NURBS basis functions and control
S (ξ η ) ∑∑ i p (ξ M j q η )Bi j , (5) points.
i =1 j =1
n m l
S (ξ η,ζ ) ∑ ∑ ∑ Ni p ξ )M j q (η Lk ,rr ζ )Bi j ,k, (6) 3 ISOGEOMETRIC SHAPE
i =1 j =1 k =1
OPTIMIZATION FOR
respectively, where Ni,p(ξ), Mj,q(η), and Lk,r(ζ) are PRESSURE LOADING
B-spline basis functions in the ξ, η, and ζ coordi-
nates, respectively, with given knot vectors Ξ = {ξ1, In the gradient-based shape design optimization
ξ2, …, ξn+p+1}, Η = {η1, η2, …, ηm+q+1} and Ζ = {ζ1, problem, the exact evaluation of shape design
ζ2, …, ζ l+r+1}. sensitivity is very important, since it determines
one-dimensional search direction during optimiza-
tion. Therefore, inaccurate sensitivity information
2.2 Elasticity problems causes slow convergence rate or even yields the
Using the principle of virtual work, an equilibrium completely wrong shape.
equation for elasticity problems is expressed as For the fixed loading case, the shape design sen-
sitivity equation is derived as
a( , ) ( )), ∀ Z, (7)
where the bilinear strain energy and linear load ∫Ω cijkl zi, j zk ,l d Ω
forms are defined, respectively, as
= ∫ cijkl zi ,mVm, j zk ,l d ∫Ω cijkl zi, j zk ,mVm,l d Ω
Ω
a( , z ) ∫Ω cijkl zi, j zk ,l d Ω (8) −∫
Ω
Ω ∫ fi ,m ziVm d Ω
Ω
and + ∫ fi zV V
i m,m d Ω + ∫Γ i,m i md Γ
t z
zVVt
Ω
( ) = ∫ + ∫ t ti zi d Γ. (9) + ∫ κ (ti zi )Vm nm d Γ, (14)
Ω Γ Γ t
644
∫Ω cijkl zi, j zk ,l d Ω
= ∫ cijkl zi ,mVm, j zk ,l d ∫Ω cijkl zi, j zk ,mVm,l d Ω
Ω
− ∫ cijkl zi , j zk ,lVm,m d Ω
Ω
+∫ Ω ∫ fi zzV iVm ,m d Ω
Ω Ω
+ ∫ ( p, jV j ni zi + pV
t
V j j ni zi − pVi j n j zi )d Ω . (15)
Γ
As we can see in the last integral term in Eq. (15), Figure 2. Non-normal design perturbation.
the exact evaluation of normal vector is very cru-
cial for the precise sensitivity calculation. In the
isogeometric-based shape sensitivity analysis, nor- Table 1. Shape design sensitivity for non-normal design
mal vectors can be precisely calculated from the perturbation.
geometrically exact geometry using NURBS basis DOF FDM (A) % (B) %
functions. The limitation of Eq. (14) is that it can
only be applied in normal perturbation case, which 1x 2.935E-5 7.839E-6 374.495 2.938E-5 99.895
preserves normal vectors along the design boundary. 2x 1.283E-5 1.721E-5 74.575 1.284E-5 99.971
However, in a real engineering case, loading direc- 3x 1.140E-5 8.634E-6 132.037 1.140E-5 99.969
tions can be changed according to the design modi- 4x 1.866E-5 5.778E-5 32.304 1.868E-5 99.924
fication. In this sense, Eq. (15) can be useful in the 5x 8.162E-6 1.681E-6 485.470 8.181E-6 99.769
design-dependent pressure loading problems. 6x 1.409E-5 1.995E-5 70.629 1.411E-5 99.870
In order to verify shape design sensitivity
expressions for pressure loading cases, the quarter
model of a circular pipe is introduced as shown in
In order to generate non-normal shape design
Figure 1. Internal pressure of P is applied inside
velocity field, only three control points on the
the inner hole, and symmetric boundary condition
oblique line are perturbed as shown in Figure 2.
is given. Due to the symmetry, only the displace-
When pressure loading is applied along the curved
ment sensitivities in x-direction for control points
boundary, the direction of loading is simultane-
1 to 6 are considered. For the sensitivity verifica-
ously changed under this non-normal perturbation.
tion, analytic sensitivities from Eqs. (14) and (15)
In Table 1. analytic sensitivities are also compared
are compared with finite difference sensitivities.
with finite difference sensitivity. In the third column,
analytic sensitivity for a fixed traction loading case
is obtained, which shows large disagreement with
finite difference as shown in the fourth column.
This is because Eq. (14) does not express the direc-
tion change of loading. However, analytic sensi-
tivity for a pressure loading case still shows very
good agreement compared with finite difference
sensitivity as shown in the last column.
4 NUMERICAL EXAMPLE
645
∫Ω d Ω,
T
minimize C (16)
Figure 3. Rectangular plate subject to non-uniform We presented some demonstrative numerical exam-
pressure load. ples for shape optimization, where the accuracy
646
647
Nian-Zhong Chen
ABS, Houston, USA
Ge Wang
ABS, Shanghai, China
ABSTRACT: A case study was conducted to investigate the extreme value of the combination of
Still-Water Bending Moment (SWBM) and Vertical Wave-Induced Bending Moment (VWBM).
Three wave scatter diagrams, namely ABS Base 1, Base 2, and one from IACS Rec No. 34, were used for
the prediction of the extreme value of the VWBM based on the extreme value theory and Response
Amplitude Operators (RAOs). The SWBM is modeled as an alternating renewal pulse process. The point-
crossing method, load coincidence method, and Ferry Borges-Castanheta method are utilized for the
prediction of the extreme value of the combination of SWBM and VWBM. A sample tanker under sag-
ging condition is utilized for the case study and results show that for the predicted extreme values of the
combination of SWBM and VWBM: 1) the results predicted based on the wave scatter diagram of ABS
Base 1 are higher than those predicted based on ABS Base 2 or IACS; 2) the results based on the wave
scatter diagrams of ABS Base 2 and IACS are very close; 3) the point-crossing method and load coinci-
dence method yield almost identical results; 4) The Ferry Borges-Castanheta method yields higher values
than those predicted by the point-crossing method or the load coincidence method; 5) the highest value
with a design period less than 20 years is 93.3% of the total design bending moment, which is predicted
by the Ferry Borges-Castanheta method based on the wave scatter diagram of ABS Base 1.
651
652
Three wave scatter diagrams are utilized herein for The integral shown in Eq.(12) is usually not easy
prediction of the VWBM. Two of the wave scatter to calculate analytically. However, the upper bound
diagrams are from ABS, namely ABS Base 1 and of v+(a) can be derived as (Larrabee, 1981):
ABS Base 2 (ABS, 2010). Another one is from IACS
Rec No. 34 (2010), which is called ‘IACS’ herein. ∞ +
The wave data from ABS Base 1 and Base 2 v + (a ) ≤ ∫ v ( y ) fX 2 a − y dy
d
−∞ 1
are determined based on the unrestricted service ∞ +
wave environment. ABS Base 1 is represented by +∫ v ( y ) fX1 a − y ddy (13)
−∞ 2
the North Atlantic wave data developed by Walden
(1964) for site-specific analysis. ABS Base 2 is rep- where vi+ is the upcrossing rate of the process
resented by the North Atlantic route wave data X i(t). fX i (.) is the probability density function
from ABS wave database for route-specific analysis. of Xi(t).
The wave data from IACS Rec No. 34 is represented If X1(t) or X2(t) has a discrete distribution, this
by the North Atlantic data derived from BMT’s upper bound formula provides an exact solution.
global wave statistics. In general, the solution of the upper bound formula
is exact, if:
3.5 Vertical wave-induced bending moment P [ X i (t ) 0 d X j (t ) < 0 ] = 0 (14)
Based on the sea states provided in the wave scatter
diagrams and the dynamic responses of the ship
4.2 Load coincidence method
from RAOs, the extreme value of VWBM is then
calculated according to the formulae of the extreme For two non-negative impulse type processes X1(t)
value given in Section 3.2. and X2(t) with the pluses returning to zero, Wen (1977)
653
where τmi is the mean pulse arrival rate of the proc- Mc,T
ψ= (19)
ess Xi(t), and: Md
χ i (a ) 1 FX i (a )
a 5 CASE STUDY
χ12 ( ) χ 2 (a ) + ∫ χ1(a u ) fX 2 u )du (16)
0
A sample tanker is utilized for this case study.
where fX i (.) and FX i (.) are the probability density
In this study, only the sagging condition is consid-
function and the cumulative distribution function
ered and the SWBM is modeled as an alternating
of Xi(t).
renewal pulse process in which the average dura-
Studies of Wen (1977, 1990) show that the load
tion of one voyage and the average time in port of
coincidence method usually provides conservative
the tanker are assumed to be 12 and 1.8 days,
results and the method can be applied for the com-
respectively.
bination of Poisson processes, intermittent proc-
The VWBM is calculated based on wave scatter
esses, and pulse and intermittent process, etc.
diagrams, RAOs and the extreme value theory.
There are three wave scatter diagrams utilized for
4.3 Ferry Borges–Castanheta model the calculation. Two are from ABS (Base 1 and
Ferry Borges and Castanheta (1971) proposed a Base 2) and the other is from IACS Rec No. 34.
simple but practically accurate load combination The point-crossing method, load coincidence
model in which the combined stochastic process is method, and Ferry Borges-Castanheta method are
represented by a sequence of rectangular pulses of used for prediction of the extreme value of the
fixed duration and the intensities of pulses are combination of SWBM and VWBM.
modeled as random variables. The principal dimensions of the tanker and
Therefore, for a combined process X(t) = X1(t) + the stochastic properties of SWBM and VWBM
X2(t), if there are m occurrences of process X2(t) are shown in Tables 1–3, respectively. Note that the
during each occurrence of X1(t), the cumulative mean value of the SWBM in Table 2 is normalized
distribution function of the maximum combined by the total design bending moment Md, which is
value of X(t) when there are n occurrences of X1(t) defined as 100 units.
during the time period T is given by:
{ }
m n
⎡ FX ( x − x ) ⎤ fX ( x ) ddx
x Table 1. The principal dimensions of the tanker.
Fmax,T ( x ) = ⎣ ⎦ (17)
Nomenclature Value
The Ferry Borges–Castanheta model has been Length btw perpendiculars (m) 230.0
shown to be a sufficiently good estimate of the Beam (m) 32.3
cumulative distribution function of a long-duration Depth (m) 18.3
combined stochastic load process (Turkstra and
Madsen, 1980).
Table 2. Stochastic properties of SWBM.
4.4 Load combination factors
The load combination factors for SWBM and Distribution type Normal
VWBM are usually defined as follows:
Mean value 23.4
Mc,T Mw,T Coefficient of variation 0.3
ψ sw =
M ssw,T
Mc,T M sw s ,T
ψw = (18) Table 3. Stochastic properties of VWBM.
Mw,T
Mc,T Distribution type Weibull
ψc = Shape parameter 0.809
M sw,T Mw,T
654
(a) 0.85
Point Crossing Method
0.95
0.8 Ferry Borges - Castanheta Model
Point Crossing Method
0.9 Load Coincidence Method
ψ 0.75
0 5 10 15 20
0.85
Design period (year)
(b)
Figure 1. Relationship between the ratio ψ and the design bending moment. Figure 2 shows the rela-
design period T for wave scatter diagrams (ψ = Mc,T/Md. tionship between the ratio ψ predicted using differ-
Mc,T —the combined extreme value of SWBM and VWBM ent load combination methods as a function of
with a design period T; Md —the total design bending design period T. Table 4 shows the ratios ψ with a
moment). design period of 20 years.
655
656
657
ABSTRACT: The objective of this work is to analyze various uncertainties related with the prediction of
the ultimate strength of a stiffened panel. The effect of different structural parameters on the uncertainty
is evaluated based on Monte Carlo simulation and an ANOVA methodology is used to determine the
most relevant parameters. The ultimate strength is predicted by the finite element method and the influ-
ence of plate thickness, Young modulus, yield and ultimate tensile strain of material, the shape of the
initial geometry imperfection and slenderness ratios are accounted for.
659
660
Load, [N]
tions was defined to achieve acceptable output. 2.000E+06
However, due to the fact that the objective is
to analyze to the contribution of different fac- 1.500E+06
predictions. 3.000E+06
Six sets of initial geometric imperfections are
2.500E+06
defined with distorted profile in the plate and out-
of-straightness in the stiffener. Faulkner, (1975) and
Load ,[N]
2.000E+06
Smith et al. (1988) have reported that the imperfec-
1.500E+06
tions in a plate can be assumed to be proportional
to β2. They suggest that the initial deformation 1.000E+06
Young Modulus=1.80E11 Pa
w0 for the average imperfection can be calculated as 5.000E+05 Young Modulus=2.10E11 Pa
w0 = 0.1β2t. The maximum permissible camber tol- Young Modulus=2.40E11 Pa
erance, wmax, for a standard shape is usually assumed 0.000E+00
0.000 0.002 0.004 0.006 0.008 0.010 0.012
to be 0.2% of the length of the length. The initial Displacement, [m]
3.500E+06
where l is the length of the panel and b is the breath
of the panel. x and y are the Cartesian coordinates 3.000E+06
of the any location over the plate. m and n are wave 2.500E+06
numbers, and w0 represents the quantity of initial
Load, [N]
2.000E+06
deflection. The shape of the initial imperfections
is defined by the coefficients m and n, where n is 1.500E+06
a constant which has a value of one and m ranges 1.000E+06 Yield Stress=3.00E+08
from one to six. Yield Stress=3.15E+08
5.000E+05
Yield Stress=3.30E+08
0.000E+00
3 ULTIMATE STRENGTH ASSESSMENT 0.000 0.002 0.004 0.006 0.008 0.010 0.012
Displacement, [m]
661
2.00E+06
4 R R R A A A
1.50E+06
0.65
5 R R R A A A
1.00E+06 Ultimate tensile strain=4.800E+08
0.60
6 R R R A A A
Ultimate tensile strain=5.00E+08
5.00E+05 1 2 3 4 5 6
Ultimate tensile strain=5.600E+08 Ranked yield stress
0.00E+00
,0.000 ,0.002 ,0.004 ,0.006 ,0.008 ,0.010 ,0.012
Dsiplacement,[m]
Figure 6. Box plot and null hypothesis acceptance/
rejection of the yield stress.
Figure 4. Effect of ultimate tensile strain on load—
displacement (t = 15 mm, E = 210 GPa, σy = 315 MPa,
w0 = 4.9 mm, m = 1 and n = 1). considered include waves longer than the funda-
mental buckling mode. The plate ultimate strength
may be larger than that of a perfect one, when the
3.500E+06
initial geometric imperfection shape is much differ-
3.000E+06 ent from the original collapse mode, having thus a
2.500E+06 strengthening effect. This may be explained with
the fact that such an initial imperfection shape will
Load, [N]
2.000E+06
disturb occurrence of the plate collapse mode.
1.500E+06 However, one of the parameters most strongly
1.000E+06 m=1, n=1
influencing ultimate strength, in addition to the
m=2, n=1 ones already discussed is the slenderness (Faulkner,
5.000E+05 m=3, n=1
m=4, n=1 1975, Guedes Soares, 1988 a,b). The effect of slen-
0.000E+00 derness is significantly important for the varia-
0.000 0.003 0.006 0.009 0.012
Displacement,[m] tion of ultimate strength and for the uncertainty
induced into the structural response and is included
Figure 5. Effect of initial imperfection shape on load— into the study presented here.
displacement (t = 15 mm, E = 210 GPa, σy = 315 MPa,
σT = 500 MPa and w0 = 4.9).
4 MONTE CARLO SIMULATIONS
662
store the results as yi; repeat the previous two steps where DFW is the within group degree of freedom,
for i = 1 to n; analyze the results using histograms, MSE is the within group mean square and ni, nj is the
summary statistics, confidence intervals, etc. number observation in groups i and j respectively.
In the study presented here, uniformly distrib- The results for null test for pair wise compari-
uted random numbers are used as an input for sons of the means are equal or not are given in
a normal distribution function to represent the Figures 6 to 12 (right), where “R” is for rejection
uncertainty in a ultimate strength model originat- and “A” is for acceptance of the null hypothesis
ing from the uncertainty in plate thickness A, the and Figures 6 to 12 (left) shown the box plot of the
Young modulus B, the yield stress C, the ultimate studied input parameters vs. normalized ultimate
tensile strain of material D, and column slender- strength.
ness E. The shape of the initial geometric imper- It has to be pointed out that the plate thick-
fection F is modelled as a uniform distribution ness and plate slenderness ratio ranks are fully
function. The statistical descriptors of the param- rejected by the null hypothesis for any par i ≠ j
eters are given in Table 1. demonstrating a significant effect on ultimate
In order to obtain an error less than 2% in the strength. Figure 6 shows a mixed behaviour of
Monte Carlo simulations, 550 samples have been
produced for each parameter involved in the ulti-
mate strength calculation. 1 2 3 4 5 6
1 A A R R R R
5 FACTORIAL ANALYSIS 2 A A A R R A
3 R A A A A A
At this point it is interesting to make comparisons of
4 R R A A A A
the means (Bailey, 2008) of the plate ultimate strength
as a function of the input random parameters the 5 R R A A A A
0.75
2 R A R R R R
when i = −3 then Mean = min and when i = 2 then
Mean = max. 0.70 3 R R A R R R
A null hypothesis, H0:τ1 = τ2, H0:τ1 = τ3 etc., 4 R R R A R R
where τ1 is the parameter effect, is tested by the use
0.65
5 R R R R A R
of the analysis of variance, knowing that some of 0.60
6 R R R R R A
the parameter levels means are different. When the 1 2 3 4 5
Ranked plate thickness
6
663
0.75
2 A A A A A A
0.70 3 A A A A R A
4 A A A A A A
0.65
5 A A R A A A
0.60
6 A A A A A A
1 2 3 4 5 6
Ranked Young modulus
0.80
1 2 3 4 5 6 Figure 13. Interaction plot.
1 A A A A R A
Normalized ultimate strength
0.75
2 A A R A R A
0.70 3 A R A A R A The statistical importance of the ultimate
4 A A A A R A tensile strain and the shape of initial geometry
0.65
5 R R R R A R imperfection in shown Figures 7 and 10 revealed
0.60
6 A A A A R A similar behaviour. Figure 9 shows that the ranks
1 2 3 4 5
Shape of initial imperfection
6
3–5 demonstrate statistical importance for the var-
iation of Young modulus on ultimate strength.
Figure 10. Box plot and null hypothesis acceptance/ Figure 11 shows that the null hypothesis fully
rejection of the shape of initial geometry imperfection. accepted any par combination of column slen-
derness ratio ranks demonstrating that the varia-
0.80
tion of this parameter, in the range that has been
1 2 3 4 5 6 defined here, statistically is not producing different
1 A A A A A A results.
Normalized ultimate strength
0.75
0.75
Figure 12. Box plot and null hypothesis acceptance/ Uncertainty is associated with the range of values
rejection of the plate slenderness ratio. that can reasonably be attributed to the analyzed
quantity. When uncertainty is evaluated, the level
of confidence has to be indicated. It is expected
the effect of the yield stress, demonstrating that that the values of the variable actually lie within
for any rank par i ≠ j where i ∈ [1, 3] ∩ j ∈ [3, 6] the range defined by the uncertainty interval.
(due to the symmetry of the acceptance/rejection The uncertainty is also a consequence of
matrix only the upper triangular one is discussed) the unknown sign of the random effects and
the null hypothesis is rejected revealing the statisti- is therefore expressed as an interval about the
cal importance of the specified ranks. result. It is evaluated by combining a number of
664
0.75 0.75
the normal distribution may be used to construct
0.70 0.70
a confidence interval. Using MSE as an estima-
0.65 0.65
tor of σ2. The square root of MSE is the “Pooled
0.60
1 2 3 4 5 6
0.60
1 2 3 4 5 6
StDev”, based on the confidence interval and the
Ranked plate thickness Ranked Young modulus
t-distribution, since
Figure 14. Marginal plot of plate thickness (left) and
Young modulus (right) vs. normalized ultimate strength.
T (Yi − μi ) MS
SE
n (6)
0.75
interval on the mean of the ith treatment μi, which
0.70
leads to:
0.65
0.60
1 2 3 4 5 6
MS
SE
yi − tα / ≤ μi ≤ yi
Ranked yield stress
,a ( n − ) n
Figure 15. Marginal plot of yield stress (left) and (7)
MS
SE
ultimate tensile strain (right) vs. normalized ultimate + tα / ,a ( n−
n )
strength. n
665
0.7462
99
in Figures 18 to 24, where the ultimate strength is 95 95 95
90
presented as a function of different parameters
and their ranks. 50
10
0.7377
0.7384
1
Percent
0.6 0.7 0.8 0.1
4 5 6
99.9 99.9
1 2 3
0.7506
99
0.7603
99
95 95 95 95 95 95
90 90
50
50
10
10
0.7946
0.7481
1
Percent
0.7493
0.7625
0.1 1
99.9
4 5 6 0.1
0.7300
10
Figure 21. Probability plot of normalized ultimate
strength conditional to ranked Young modulus.
0.7348
0.7295
1
0.1
0.6 0.7 0.8 0.6 0.7 0.8
0.7522
99
strength conditional to ranked yield stress. 95 95 95
90
50
0.6 0.7 0.8
1 2 3 10
99.9
0.7479
0.7400
Percent
0.7332
99 1
95 95 95 0.1
90 4 5 6
99.9
0.7331
50 99
95 95 95
10 90
0.7160
0.7437
1
Percent
50
0.1
4 5 6
99.9
10
0.7524
99
0.7519
0.7421
95 95 95 1
90
0.1
50 0.6 0.7 0.8 0.6 0.7 0.8
Normalized ultimate strength
10
0.7488
0.7536
1
0.1 Figure 22. Probability plot of normalized ulti-
0.6 0.7 0.8
Normalized ultimate strength
0.6 0.7 0.8
mate strength conditional to the shape of initial
imperfections.
Figure 19. Probability plot of normalized ultimate
strength conditional to ranked ultimate tensile strain.
0.50 0.75 1.00
1 2 3
99.9
0.7541
99
95 95 95 50
90
10
50
0.8202
0.7458
Percent
1
10
0.1
4 5 6
0.6661
0.7186
1 99.9
Percent
0.7396
0.1 99
4 5 6
99.9 95 95 95
90
0.7558
99
95 95 95
90 50
50 10
0.7394
0.7506
1
10
0.1
0.7368
0.7758
666
0.7556
99
95 95 95 uncertainties is.
90
The analysis performed here may also be used
50
to define a nominal or characteristic measure of
10 capacity that corresponds to a specified percentage
0.7770
0.7358
below the probability curve for the corresponding
Percent
1
0.1
99.9
4 5
0.6992
6
random variable. For instance, one may base char-
99
95 95 95
acteristic strength or capacity on a lower bound or
90
95% exceedence value, while a characteristic load
50 or demands may be based on an upper bound or a
10 5% exceedence value.
0.7172
0.6816
1
0.1
0.6 0.7 0.8 0.6 0.7 0.8
Normalized ultimate strength 7 CONCLUSIONS
Figure 24. Probability plot of normalized ultimate This work analyzed the influence of plate thick-
strength conditional to the ranked plate slenderness ratio.
ness, Young modulus, yield and ultimate tensile
strain of material, column and plate slenderness
It can be observed form Figures 18 to 24 and and the shape of the initial geometric imperfection
Figure 2 that the standard deviations of ultimate of stiffened plate.
strength as a function of studied parameters along The analysis pointed out that the plate thickness
ranks show relatively small variation. If the stand- and plate slenderness ratio ranks are fully rejected
ard deviations are averaged for all ranks, the small- by the null hypothesis for any par i ≠ j demon-
est one is calculated for plate thickness followed strating a significant effect on ultimate strength.
by plate slenderness, shape of initial imperfection, A mixed behaviour of the effect of the yield stress
ultimate strain, column slenderness, the Young on ultimate strength demonstrated that for some
modulus and yield stress. ranks the null hypothesis is rejected revealing
However the standard deviation shows more the statistical importance of the specified ranks.
than just possible range around the estimated The statistical importance of the ultimate tensile
Table 2. Statistical descriptors of normalized ultimate strength conditional to the ranks of different parameters.
Rank 1 2 3 4 5 6
667
668
Safety analyses for bulk carriers using metamodels of still water loads
P. Georgiev
Technical University of Varna, Bulgaria
ABSTRACT: This paper presents an approach that combines metamodeling technique and Monte-
Carlo simulation for safety analyses of still water loads for bulk carriers. The approach is applied to study
the influence of deviations from cargo loading plan on net load of double bottom and still water bending
moments. As example a Handymax BC-A type ship is used. The metamodels approximate the work of
installed on board mandatory loading instrument and give the relation between the distribution of cargo
and the trim and still water bending moments in controlled sections. The Monte-Carlo simulation uses
the fitted metamodels to obtain a rich set of statistical data that permit preparing event tree analysis for
possible overloading of double bottom and the evaluation of likelihood for exceeding of permissible
bending moments.
669
670
671
Frequency
400
200
200
0 0
11200 11600 12000 12400 12800 13200 12400 12800 13200 13600 14000 14400 14800 15200
Cargo mass, t Cargo mass, t
a) b)
1200
-T1
Hold No 5
1000 -T2
-T3
800 -T4
600
Frequency
400
200
0
12000 12400 12800 13200 13600 14000 14400
Cargo mass, t
c)
Figure 2. a, b, c) Cargo distribution for different trim ranges; d) Bubble plot for cargo distribution in the holds.
-T2 -T4
0.6
⎛ σ ⎞ wr=19.255N-0.4917
⎜⎝ μ ± z1−α / 2 ⎟ (5) R2=0.997
N⎠ 0.4
wr=14.686N-0.4921
where μ is sample mean,σ is standard deviation, 0.2
R2=0.997
N is number of runs, and z1−α/2 is the number that 0 1000 2000 3000 4000
Φ(zγ) = γ (Φ denotes the standard normal c.d.f). It Numberofruns, N
is common practice in simulation to use and report
the absolute (wa) and relative (wr) widths of the Figure 3. Relative width for 95% and 99% confidence
confidence interval (CI), defined as for cargo mass in Hold No 3.
σ w
wa 2 z1 α /2 and wr = μa (6) Table 3. Descriptive statistics for mass of cargo in holds
N at T1 trim range.
In Figure 3 the relative width in percent (wr) for Item Hold 1 Hold 3 Hold 5
α = 5% and = 1% significance levels is shown. The
results include all trim ranges and as one can see Plan P0, t 12536 13884 13186
from the trend lines, the accuracy of estimation MaxLoad, t 12536 13884 13.835
of the mean is proportional to N−0.5. The maxi- μ, t 12446.0 13818.4 13141.5
mum relative width of CI is less than 1% and goes 95% CI, t ±9.6 ±18.3 +11.2
down to less than 0.4% at 50,000 starting runs. 99% CI, t ±12.6 ±24.0 ±14.7
The descriptive statistics for mass of cargo dis- σ, t 291.9 556.8 341.7
tribution in the holds at T1 range for sample size CoV,% 2.3 4.0 2.6
N = 3565 are shown in Table 3. The table presents Skewness −0.0001 0.0314 −0.0299
the planned and maximum permissible (Max- Ex. Kurtosis −0.1283 −0.4916 −0.2440
Load) amount of cargo.
672
673
Ti B1 B3 B5 Outcome
80
39.4 43.6
6.5
Probability, %
42.9
60
4.4
40 18.2 27.4
30.1
20 42.0
32.7 27.1
22.2
0
T1 T2 T3 T4
Figure 5. Probability (%) of overloading of every cargo Figure 6. Probability for overloading of separate holds
hold and combination of holds. at given trim range T1–T4.
interval for the mean at all controlled sections for sections and corresponding trend line are shown
trim range T1 are shown in Table 5. The last two in Figure 7. The least accuracy of estimation of
rows include values for SWBM for the planned the mean is obtained for frame 129. The frame is
loading condition and location of the section. The located at the middle of the ship and the standard
relative width of 95% confidence interval for three deviation of SWBM for this frame is greatest.
674
Frames
Statistic Fr. 75 Fr. 86 Fr. 93 Fr.111 Fr.129 Fr.139 Fr.147 Fr.165 Fr.183
3.0 16
%
wr = 42.475N-0.5013 ; R2=0.9956
wr = 64.005N-0.5038 ; R2=0.9946 14
2.5
wr = 24.328N -0.4971 2
; R =0.9966
12
Relativewidth, wr,%
2.0
10
1.5
8
1.0 6
- CoV
0.5 4
-σ
x/L
Fr. 111 Fr. 129 Fr. 165 2
0.0 0.3 0.4 0.5 0.6 0.7 0.8
0 1000 2000 3000 4000
Number of runs, N Figure 8. Standard deviation and CoV along ship length.
675
676
677
A.W. Hussein
Department of Naval Architecture & Marine Engineering, Suez Canal University, Port Said, Egypt
C. Guedes Soares
Centre for Marine Technology and Engineering, Technical University of Lisbon, Instituto Superior Técnico,
Lisboa, Portugal
ABSTRACT: This paper studies the reliability of two single hull bulk carriers subjected to damage.
Two damage scenarios are assumed, collision at side and grounding at the bottom. The ultimate strength
is calculated using the Progressive Collapse Method. The reliability is assessed for both damage scenarios
and compared with the intact ship reliability. In the damaged condition, the ship should not operate with
a very high speed, and she will avoid very rough sea. Therefore, the related wave-induced bending moment
may be smaller than that for the normal design extreme condition. On the other hand, damage to the vessel’s
hull may result in the ingress of water and for fluid cargoes cargo out-flow may occur which will result in a
change of the loading condition and a variation to the still water bending moment. The reliability is assessed
and sensitivity analysis is performed to study the importance of the variables after and before damage.
679
4 ULTIMATE STRENGTH
OF DAMAGED SHIPS
680
the keel area, the bottom below the hopper tank or 0.85
0.80
0.65
Smith’s method (Smith 1977) has been exten- 0.60
sively used to determine the hull girder ultimate keel bottom below hopper
tank
bilge sheer strake and side at hopper tank
upper hopper tank joint
strength of the intact ship. Many authors calcu- Damage location side
681
0.80
0.75
conditions and a Gaussian distribution at a ran-
0.70 dom point in time. The mean and standard deri-
0.65 vations of this distribution are the average of the
0.60
keel bottom below hopper bilge sheer strake and side at hopper tank respective values at departure and arrival.
tank
Damage Location
upper hopper tank
side
joint
For the bulk carrier three different loading
conditions are defined, namely: alternate hold
Figure 5. Ultimate strength in hogging for damaged loading condition (LC1), homogeneous hold
ships. loading condition (LC2) and ballast condition
(LC3). In each loading condition, a percentage
Table 4. Stochastic modeling of UBM. of ship life can be identified according to an esti-
mate of the operational profile for the ship type.
Sagging Hogging A homogeneous hold loading condition refers
to the carriage of cargo, evenly distributed in all
μ σ μ σ cargo holds. This condition is usually adopted for
low density cargoes such as coal and grain, but may
Bulk 1 3142 251.36 4311 344.88 also be permitted for high-density cargoes under
Bulk 2 2524 202 3895 312 certain conditions. However, heavy cargo, such as
iron ore, is often carried in alternate holds. This
type of cargo distribution raises the ship’s centre
From the figures one can see that in sagging of gravity, which eases the ship’s rolling motion.
condition, the worst scenario which causes the In this paper the SWBM distribution is calcu-
worst ultimate strength after damage is when the lated based on studying the data of the loading
sheer strake is damaged. The loss in the ultimate manual. The mean value and the standard devia-
strength due to side damage at the hopper tank tion are calculated as a function of the maximum
joint in Bulk 1 is greater than that in Bulk 2. This allowed value defined in the loading manual.
can be attributed to the transverse stiffening of Table 5 shows the stochastic model of the still
Bulk 1. Although the loss in the ultimate strength water bending moment based on analyzing the
in hogging is more than that in sagging, never-
theless the sagging conditions still represents the Table 5. Stochastic model of SWBM (% max allowed).
critical condition.
Bulk 1 Bulk 2
4.1 Stochastic model of UBM μ σ μ σ
The ultimate bending moment is calculated using
the progressive collapse method. The ultimate LC1: Ballast 0.74 0.09 0.79 0.12
capacity distribution is lognormal with a mean LC2: Homogeneous 0.15 0.2 0.21 0.09
value as calculated by the code and the COV is 0.08 LC3: Alternate 0.85 0.04 0.85 0.09
the mean value. Table 4 shows the stochastic model
of all the ultimate capacity for the two ships.
Table 6. SWBM of Bulk 1.
682
1500
1000
Table 9. Stochastic model of WBM in MN. Bulk 2.
683
g(x) = Mu ⋅ XR − [Mwn ⋅ Xst ⋅ Xnl + Msw ⋅ Xsw] (6) One can notice from the sensitivity analyses
done that the ultimate capacity importance is the
where, MU is the ultimate capacity with a model highest among all the variables in LC1 and LC3.
uncertainty factor XR. MWV is the wave bending while the WBM is the most important variable in
moment with model uncertainty factors; Xst for LC2. The SWBM importance increases in the bal-
the linear response calculation and Xnl for nonlin- last condition while the WBM importance increases
ear effects. MSW is the random still water bending in the homogenous loading condition while the
moment with a model uncertainty factor XSW. The SWBM showed very low importance. The results
values of the above mentioned uncertainties are are presented in Tables 12 and 13. Figures 7 and 8
taken as presented in Table 10. show the results of Bulk 1.
The reliability calculations were carried out using
the computer program COMREL (Gollwitzer
et al. 1988), considering the distribution of the Table 12. Bulk 1.
extreme values of the load effects in the three load
conditions. The reliability indices for the different LC1 LC2 LC3
load conditions are presented in Table 11 for the
Sagging
intact condition.
SWBM −0.32 −0.28 −0.55
It is clear from the results that sagging is more
WBM −0.33 −0.61 −0.30
critical than hogging. The homogenous hold loading
UBM 0.53 0.41 0.45
condition has the highest reliability, while the alter-
nate and ballast loading give very low reliability. LC1 LC2 h LC3
Hogging
Table 10. Summary of uncertainties of the models.
SWBM −0.27 −0.17 −0.49
Variable Distribution μ δ WBM −0.44 −0.71 −0.36
UBM 0.51 0.37 0.47
XR Normal 0.85 0.1
Xst Normal 1 0.1
Xnl Normal 1 0.1 Table 13. Bulk 2.
XSW Normal 1 0.1
LC1 LC2 LC3
684
0.60
Sagging
DAMAGED CONDITION
0.40
Sensitivity variables
0.20
0.00
In the damaged condition, the ship should not
-0.20
SWBM WBM UB M
operate with a very high speed, and it will avoid
-0.40
very rough seas. Therefore, the related wave-
-0.60
-0.80
induced bending moment may be smaller than that
for the normal design extreme condition. On the
Figure 7. Sensitivity variables of intact ship in sagging. other hand, damage to the vessel’s hull may result
in the ingress of water and, for fluid cargoes cargo
LC 1 h LC 2 h LC 3 h out-flow can occur, which will result in a change
0.60
Hogging of the loading condition and a variation to the
0.40
still water bending moment (Santos and Guedes
Sensitivity variables
0.20
0.00
Soares 2008).
-0.20
SWBM WBM UBM Regarding the still water load coefficient, this is
-0.40
usually considered equal to 1.0 for the intact ship.
-0.60 The ABS (ABS 1995) guide for damaged ships rec-
-0.80 ommends that for hogging the value used should
be 1.1, while for sagging should be 0.9 as shown
Figure 8. Sensitivity variables of intact ship in hogging. in Table 14.
Thus, the bending moment according to the
7
lc1 s lc2 s lc3 s Bulk 1 Sagging ABS guide is equal to:
6
5
4
Mt = Kus ⋅ Msw + Kuw ⋅ Mw (8)
β
3
2
1
where Kus and Kuw are load combination coeffi-
0
intact keel bottom keel sheertrake hopper side
cients due to damage. The reliability is assessed
considering the ultimate bending moment after
Figure 9. Reliability indices Bulk 1 with damage. damage presented in Table 3 and the load combi-
nation factors presented in Table 14. Results are
lc1 h lc2 h lc3h Bulk 1 Hogging
presented in Tables 15 and 16.
8
7
6
5
4
Table 14. Bending moment combination factors
β
3
2
1
(ABS 1995).
0
intact keel bottom keel sheertrake hopper side
Intact Grounding Collision
Figure 10. Reliability indices Bulk 1 with damage.
Hogging Kus 1 1.1 1
Kuw 1 0.5 0.7
lc1s lc2 s lc3 s Sagging
6.00 Sagging Kus 1 0.9 1
5.00
Kuw 1 0.5 0.7
4.00
3.00
β
2.00
1.00
0.00
Intact keel bottom bilge sheertrake hopper Table 15. Reliability indices of damaged ship (Bulk 1).
side
3.00
2.00
1.00
Hogging
0.00
Intact keel bottom bilge sheertrake hopper
LC1 2.57 4.10 4.13 4.29 3.22 2.51
side LC2 4.36 7.00 7.02 7.13 5.45 4.97
LC3 1.98 3.33 3.35 3.49 2.58 1.96
Figure 12. Reliability indices Bulk 2 with damage.
685
Sagging
LC1 0.62 3.159 3.199 3.279 1.055 1.712
LC2 2.72 5.574 5.603 5.66 3.634 4.127
LC3 0.62 3.103 3.142 3.218 1.078 1.688
Hogging
LC1 3.51 4.927 5.036 5.388 4.562 4.214
LC2 4.90 7.205 7.281 7.528 6.096 6.251
LC3 3.42 4.793 4.898 5.24 4.226 4.421
LC1 s
SWBM −0.32 −0.39 −0.39 −0.39 −0.38 −0.37
WBM −0.33 −0.31 −0.31 −0.31 −0.30 −0.32
UBM 0.53 0.54 0.54 0.54 0.54 0.54
xsw −0.29 −0.36 −0.36 −0.36 −0.35 −0.34
xnl −0.36 −0.26 −0.26 −0.26 −0.30 −0.30
Xst −0.36 −0.26 −0.26 −0.26 −0.30 −0.30
xr 0.40 0.44 0.44 0.44 0.41 0.42
LC2 s
SWBM −0.28 −0.27 −0.27 −0.27 −0.32 −0.29
WBM −0.61 −0.74 −0.74 −0.74 −0.69 −0.70
UBM 0.41 0.36 0.36 0.36 0.39 0.38
xsw −0.04 −0.04 −0.04 −0.04 −0.05 −0.04
xnl −0.42 −0.34 −0.34 −0.34 −0.38 −0.37
Xst −0.42 −0.34 −0.34 −0.34 −0.38 −0.37
xr 0.32 0.30 0.30 0.30 0.31 0.30
LC3 s
SWBM −0.55 −0.63 −0.63 −0.63 −0.63 −0.61
WBM −0.30 −0.23 −0.23 −0.23 −0.25 −0.26
UBM 0.45 0.46 0.46 0.46 0.45 0.45
xsw −0.21 −0.31 −0.31 −0.31 −0.27 −0.27
xnl −0.35 −0.23 −0.23 −0.23 −0.28 −0.28
Xst −0.35 −0.23 −0.23 −0.23 −0.28 −0.28
xr 0.33 0.36 0.36 0.36 0.33 0.34
One can see from the figures that the homog- included in the limit state function and compare its
enous loading conditions represents the safest importance before and after damage. The results
condition, this is because of the uniform cargo are presented in Tables 17 to 20.
distribution inside the compartments. While the One can conclude from the calculations the
ballast loading condition and the alternative load- following:
ing condition which give high SWBM give low
• When the ship is damaged at her bottom, the
reliability. The worst reliability occurred when
importance of the variables after damage does
the side is damaged, whether at the sheer strake
not depend on the position of the damage.
or at the joint of the hopper tank with the side.
• When the ship is damaged at her side, the impor-
tance of the variables after damage does not
9 SENSITIVITY OF THE VARIABLES depend on the location of the damage.
AFTER DAMAGE • The change in the importance of the ultimate
bending moment after damage is very small.
Sensitivity analysis is made for the damaged con- • The importance of the SWBM increased after
ditions to study the importance of the variable damage.
686
LC1 h
SWBM −0.27 −0.41 −0.41 −0.41 −0.33 −0.35
WBM −0.44 −0.28 −0.29 −0.29 −0.39 −0.35
UBM 0.51 0.55 0.55 0.55 0.53 0.54
xsw −0.25 −0.38 −0.38 −0.38 −0.31 −0.33
xnl −0.35 −0.22 −0.22 −0.22 −0.30 −0.30
Xst −0.35 −0.22 −0.22 −0.22 −0.30 −0.30
xr 0.40 0.45 0.45 0.46 0.43 0.42
LC2 h
SWBM −0.17 −0.25 −0.25 −0.25 −0.19 −0.19
WBM −0.71 −0.74 −0.74 −0.74 −0.73 −0.72
UBM 0.37 0.36 0.36 0.36 0.36 0.37
xsw −0.03 −0.04 −0.04 −0.04 −0.03 −0.04
xnl −0.37 −0.33 −0.33 −0.33 −0.35 −0.36
Xst −0.37 −0.33 −0.33 −0.33 −0.35 −0.36
xr 0.29 0.30 0.30 0.30 0.29 0.30
LC3 h
SWBM −0.49 −0.65 −0.65 −0.64 −0.57 −0.59
WBM −0.36 −0.20 −0.20 −0.20 −0.30 −0.28
UBM 0.47 0.47 0.47 0.47 0.47 0.46
xsw −0.22 −0.35 −0.35 −0.35 −0.28 −0.28
xnl −0.33 −0.19 −0.19 −0.19 −0.27 −0.27
Xst −0.33 −0.19 −0.19 −0.19 −0.27 −0.27
xr 0.36 0.37 0.37 0.37 0.37 0.36
LC1 s
SWBM −0.31 −0.39 −0.38 −0.38 −0.38 −0.37
WBM −0.34 −0.33 −0.34 −0.34 −0.31 −0.33
UBM 0.55 0.56 0.56 0.56 0.56 0.56
xsw −0.30 −0.37 −0.37 −0.37 −0.36 −0.35
xnl −0.37 −0.27 −0.27 −0.27 −0.32 −0.31
Xst −0.37 −0.27 −0.27 −0.27 −0.32 −0.31
xr 0.34 0.37 0.37 0.38 0.36 0.36
LC2 s
SWBM −0.28 −0.27 −0.27 −0.27 −0.32 −0.29
WBM −0.56 −0.68 −0.68 −0.69 −0.59 −0.63
UBM 0.43 0.40 0.40 0.40 0.42 0.41
xsw −0.08 −0.10 −0.10 −0.10 −0.11 −0.10
xnl −0.41 −0.33 −0.33 −0.33 −0.37 −0.36
Xst −0.41 −0.33 −0.33 −0.33 −0.37 −0.36
xr 0.28 0.27 0.27 0.27 0.28 0.27
LC3 s
SWBM −0.40 −0.48 −0.48 −0.48 −0.46 −0.45
WBM −0.33 −0.30 −0.31 −0.31 −0.35 −0.37
UBM 0.52 0.54 0.54 0.54 0.52 0.52
xsw −0.27 −0.35 −0.35 −0.35 −0.32 −0.31
xnl −0.37 −0.26 −0.26 −0.26 −0.31 −0.30
Xst −0.37 −0.26 −0.26 −0.26 −0.31 −0.30
xr 0.33 0.36 0.36 0.36 0.33 0.33
687
LC1 h
SWBM −0.23 −0.39 −0.39 −0.37 −0.28 −0.27
WBM −0.53 −0.36 −0.37 −0.39 −0.53 −0.55
UBM 0.51 0.56 0.56 0.56 0.51 0.51
xsw −0.22 −0.37 −0.37 −0.36 −0.27 −0.26
xnl −0.35 −0.24 −0.24 −0.24 −0.30 −0.30
Xst −0.35 −0.24 −0.24 −0.24 −0.30 −0.30
xr 0.33 0.39 0.39 0.39 0.35 0.35
LC2 h
SWBM −0.17 −0.25 −0.25 −0.25 −0.19 −0.19
WBM −0.71 −0.70 −0.71 −0.71 −0.72 −0.73
UBM 0.39 0.39 0.39 0.39 0.38 0.38
xsw −0.05 −0.10 −0.10 −0.10 −0.07 −0.06
xnl −0.36 −0.32 −0.32 −0.32 −0.34 −0.34
Xst −0.36 −0.32 −0.32 −0.32 −0.34 −0.34
xr 0.25 0.27 0.27 0.27 0.26 0.26
LC3 h
SWBM −0.31 −0.49 −0.49 −0.49 −0.33 −0.32
WBM −0.50 −0.30 −0.30 −0.30 −0.57 −0.59
UBM 0.51 0.54 0.54 0.54 0.48 0.47
xsw −0.22 −0.37 −0.37 −0.37 −0.24 −0.23
xnl −0.35 −0.22 −0.22 −0.22 −0.30 −0.30
Xst −0.35 −0.22 −0.22 −0.22 −0.30 −0.30
xr 0.33 0.37 0.37 0.37 0.32 0.32
Table 21. General SWBM stochastic model. The damage is simulated by removing the dam-
aged elements from the mid-ship section and
μ σ re-calculating the ultimate strength of the section.
The reliability is calculated considering the
LC1: Ballast 0.77 0.11
ultimate limit state function; the SWBM statistic
LC2: Homogeneous 0.18 0.15
model is calculated based on studying the data
LC3: Alternative 0.85 0.07
of the loading manual. The mean value and the
standard deviation are calculated as a function of
the maximum allowed value defined in the loading
• The importance of the WBM decreased after manual.
damage. A final conclusion can be given from that study
• In sagging condition, there was no big difference telling that the mean value and the COV values
between the importance of the variables when of a bulk carrier of the same size can be taken as
the ship is damaged at her side or at her bottom. shown in Table 21.
• In hogging condition, the importance of the The stochastic model of wave induced load
WBM is higher when the side of the ship is dam- effects is defined based on the evaluation of the
aged than that when the bottom of the ship is wave induced load effects that occur during long-
damaged. term operation of the ships in a seaway.
• In hogging condition, the importance of the Calculation of the reliability of the intact ships
SWBM decreased when the ship is damaged at showed that the sagging is more critical than
her side that when she is damaged at her bottom. hogging. The homogenous loading condition has
the highest reliability, while the alternate and ballast
loading give very low reliability.
10 CONCLUSIONS Sensitivity analyses showed that the ultimate
capacity importance is the highest among all the
This paper studies the ultimate strength and reli- variables in LC1 and LC3, while the WBM is
ability of two single side bulk carriers. The ulti- the most important variable in LC2. The SWBM
mate strength is calculated using Smith’s method. importance increases in the ballast condition while
688
689
690
Y. Kawamura
Yokohama National University, Yokohama-shi, Kanagawa-ken, Japan
M. Miyazaki
Yokohama National University, Yokohama-shi, Kanagawa-ken, Japan (Now in Imabari Shipbuilding Co. Ltd.,
Marugame-shi, Kagawa-ken, Japan)
ABSTRACT: In this study, we propose a new strategy of lifecycle structural optimization by using
structural reliability analysis and risk evaluation. As a simple example, optimization of the hold frame of
a bulk carrier is carried out. In the first optimization, construction cost is used as an objective function
and the failure probability is used as a design constraint. In the second optimization, the lifecycle cost
including the risk of failure is regarded as an objective function, while lifecycle benefit is used as an objective
function in the last optimization. It is noted that not only the dimensions of the structural members but
also a number of repairs (re-coating) are used as design variables in the lifecycle optimization.
691
692
693
where μ is a factor which value differs by the type 3.2 Optimization by genetic algorithm
of the structure, tL is the life of the structure, and In this study, we developed a computer program of
n is a number of people who uses the structure. Genetic Algorithm (GA) for optimization proce-
Also, Bhattacharya et al. (2001) compared the dure. Because not only continuous design variables
target reliability of marine structures obtained but also a few discrete variables are included in the
from the past researches with the CIRIA’s criteria objective function, GA seems to be more suitable
and etc. However, these criteria is generally related than the general optimization methods that require
to the failure (accident) of the whole ship, while computation of derivatives. In this GA, 4 or 5 bit
our structural reliability analysis is just for a local length in a gene is assigned to each design variable.
part of the ship structure, so that the acceptable And it is assumed that each variable can take dis-
limit for this structural reliability analysis may crete values between the lower limit value and the
be less than the criteria for the whole ship. It can upper limit value as shown in Table 2. The number
be concluded that, it is very difficult at present to of genes in one generation is set as 40. Also in this
define an reasonable acceptable limit of the failure GA, rate of mutation is set as 0.05, and multiple-
probability for this structural reliability analysis. point crossover method at two points is used.
For this reason, we just assumed in this paper It is noted that a lot of structural reliability anal-
that the acceptable limit (PfN) of the structural ysis should be carried out during the optimization
reliability analysis can be defined based on the process, so that structural reliability analysis should
CIRIA’s criteria as follows, be called directly from the optimization program.
For this purpose, we use a numerical library named
PfN = kPfCIRIA, (10) NESSUS-PA (Southwest Research Institute (SRI),
2009) which is the library of commercial reliability
where tL, μ and n in the Equation (9) are assumed
software, NESSUS (Thacker et al., 2006).
as 30, 5.0 and 25, respectively. Also k is assumed
as 0.5. Then, the acceptable limit of the failure
probability (PfN) is calculated as PfN = 3.0 × 10−4. 3.3 Optimization result
And the constraint condition of this optimization
can be represented by the following equation. Figure 5 shows the result of the optimization.
It can be observed that the thickness and height of
PfL < PFN (11)
Table 2. Gene coding of design variables.
In this equation, PfL is the failure probability
in the lifecycle of the bulk carrier, and can be Design Lower Upper Gene
computed from the annual probability of failure variable limit limit length Interval
obtained from the structural reliability analysis.
N 27 42 4 1
In this paper, we use the following simple formula ts (mm) 13.0 20.5 4 0.5
to calculate PfL; tw (mm) 12.0 27.0 5 0.5
tf (mm) 10.0 41.0 5 0.5
tL i
∑∏( (12) hw (mm) 200.0 820.0 5 10.0
PfL Pf i − ) )Pfi bf (ww) 10.0 320.0 5 5.0
i =1 j =1
694
695
S=1087(mm)
¥30
LCC(×1.0E+07[Japanese Yen])
696
(DWTT )
281.2
ROPEi reef + Wdifff RC × NR , (17) 281.0
280.8
where RC is defined as the revenue for the unit- 280.6
weight cargo per one voyage, and NR is the number 280.4
of voyages in one ship-year. DWTref is the dead- 280.2
weight of a reference ship (Table 5), which is 280.0
697
698
Eeteng Khoo
American Bureau of Shipping, Houston, Texas, USA
ABSTRACT: Corrosion is one of the most commonly found degradation mechanisms on marine
vessels. It is inevitable and can compromise the structural integrity if not properly addressed and
monitored. In extreme cases, it may have an impact on the safety of the vessel. Hence, proper and
wise corrosion management is essential. Many corrosion wastage models have been proposed by
researchers. One of the most important challenges to develop wastage models is consideration of
uncertainties within corrosion data. These uncertainties are generally dependent upon the environ-
ment, effectiveness of corrosion protection, or accuracy of the thickness measurement. The objective
of this study is to investigate these uncertainties involved and develop a process of generating corro-
sion wastage model. Characteristics of existing approaches are also reviewed and assessed while
developing a new model.
699
Ship type Single hull tanker Single hull tanker Single hull tanker Single hull tanker
Vessels 140 197 230 54
Measurement records 110,082 >250,000 33,820 Not known
Ship size (m or DWT) 168–401 m 100–400 m Not known 150,000–300,000 DWT
Ship age (years) 12–26 −23 12–26 −25
Build years Mostly 1970’s Not known Not known 1960’s (16)
1970–1975 (38)
Measured 1992–2000 Not known Not known Not known
Data source SafeHull condition Gauging records Gauging records Owner, class
assessment
Gauging reports 157 346 Not known Not known
700
d (t ) = a (t − t )b t > t0
(2)
d (t ) = 0 t ≤ t0
⎡ ⎛ t − τc ⎞ ⎤
d (t ) = d∞ ⎢1 − exp ⎜ − ⎥, t > τc
⎣ ⎝ τ t ⎟⎠ ⎦ (3a)
d (t ) = 0 t ≤ τC
d∞
Figure 3. Scatter plot of corrosion wastage on deck τt = , d∞ d0 (3b)
plate in cargo oil tanks of the tanker. tan α
701
702
⎛ k⎞ x
k −1 ⎡ ⎛ x⎞k⎤
f x ( x, k , λ ) = ⎜ ⎟ exp ⎢ ⎥ , x, k , λ > 0
⎝ λ⎠ λ ⎢⎣ ⎝ β ⎠ ⎥⎦
(5)
Table 3. Summary of existing corrosion models for upper deck plate in cargo oil tank of tankers.
Corrosion model Paik et al. (2005) Garvatov et al. (2007) Guo et al. (2008)
704
705
E. Johnson
Department of Building Technology and Mechanics, SP Technical Research Institute of Sweden, Borås, Sweden
ABSTRACT: Composite materials have been used successfully in vessels since the 1960s. Changes
during the last decade in international regulations, fuel prices and environmental concerns have dramati-
cally increased the interest of the marine industry in lightweight composite materials. There are many
questions regarding the material’s properties and behaviour that still remain, leading, in some instances to
too robustly designed crafts. Therefore, the new generation of composite vessels requires a finer tuning of
safety and weight, which can be achieved by analyzing thoroughly the behaviour of the material and the
effects of uncertainties and approximations found in the design process. This paper presents the study of
two design methodologies with different degrees of complexity, through deterministic and probabilistic
analyses, in order to determine their benefits, drawbacks and challenges.
707
708
709
3 ANALYSIS CASE
Figure 3. Stress-strain relation of the AS4/3501-6
carbon epoxy unidirectional lamina properties.
The behaviour of a composite laminate plate
subjected to a biaxial in-plane loading is used as
the mean of evaluating the methodologies. This compressive (σ2, ε2) loading. However, in this
simple geometry can be considered as a representa- investigation the behaviour of the material was
tive part of a ship structure such as a plate in a free assumed to be always linear elastic, and there-
sheet field. Figure 2 depicts the analysis case: a fore, to guarantee congruence in the stress-strain
(square) plate with side dimensions a × a, subjected calculations, the ultimate strains and stresses fol-
to distributed loads (PX and PY) around the plate’s low the relation shown in Figure 3. The interac-
edges. tion of σ2 and τ21 on their stress-strain diagrams
For the global coordinates, as depicted in (Puck & Schürmann 1998) due to micro cracking is
Figure 2, utilize the sub-indices x and y, while lam- neglected. The expectancy values of the mechani-
ina coordinates employ the sub-indices 1 and 2, 1 cal material properties of the unidirectional lamina
being the aligned with the orientation of the fibers. are presented in Table 2.
The boundary conditions are set as freely supported. The analyzed laminate is a quasi-isotropic
The influences of other boundary conditions, such one with a [90°/±45°/0°]S stacking sequence of
as fixed supported edges, are left for a future work. equally thick laminas. The overall thickness of the
The material utilized for the modelling of the modelled laminate is of 1.1 mm.
laminate is AS4/3501-6 carbon/epoxy unidirectional
lamina. This material is selected because it is well
documented in the literature, and hence, there is 4 FINITE ELEMENT ANALYSIS
access to previous numerical and experimental
results to compare with; see Soden et al. (1998, The response of the laminate is calculated with the
2002). The unidirectional lamina exhibits different finite element software ANSYS (ANSYS 2009).
degrees of non-linearity, a slight convex one under The maximum strain, FiReCo AS’ modified
axial loading (σ1, ε1) and a strong concave one Tsai-Wu and Puck’s failure criteria were imple-
under in-plane shear (τ21, γ21) and transverse mented in the software through a set of macros in
the post processing. This is done in order to start
developing tools and methodologies that can be
easily incorporated to the current design practice
of the maritime industry.
710
711
712
713
714
ABSTRACT: Germanischer Lloyd currently update rules for stowage and lashing of containers to
ensure safe and reliable container loadings and adequate dimensioning of lashing equipment. Various
research and development works, focusing mainly on performance of fully automatic locks and reliable
calculation of container and lashing loads, were performed. Essentials of the corresponding findings and
their impact on classification rules are presented. Particular attention is given to the new test procedure
for fully automatic locks.
715
716
717
718
719
720
Figure 13. Container top corner castings in benchmark Figure 14. Effect of container flexibility (left) and the
tests. equivalent test setup modification for the pull test (right).
721
722
ABSTRACT: The statistical data for ship steel strength suggest that the best-fit probability density
functions vary depending upon the type of steel and the type of failure mode. Various studies suggest
that lognormal, Weibull, or Gumbel distributions may offer the best descriptions for yield stress, ultimate
stress and failure strain. In this paper, we select several distribution functions for yield limit states and
particular steels and study the resulting reliability of steel beam components. The beams are subjected to
compressive loading conditions as commonly encountered for marine structures and the loading resultants
are thus also treated as random variables. The influence of the use of various probability density func-
tions is studied. The results of including uncertainty in the structural geometry are also discussed. It is
found that the probability of failure can vary by orders of magnitude for similar nominal safety factors
depending upon the type of steel or the yield function employed.
723
σC = a + bw1 (9)
Therefore, the pdf of σC becomes
( a )2
− C
2 2 2
pdf ( C) = e 2b v
, a ≤ σC < ∞ (10)
Figure 1. Initially bent beam. b π v12
724
2.2 Loading, external force uncertainty Note that in this model the maximum compres-
sive stress will reach the yield stress before the
Generally, the external forces and moments that
external force reaches the critical buckling force.
marine structures encounter should also be treated
Therefore the failure mode of this probabilistic
as random variables. In this section, we assume the
model of compressive stress generated from load-
external force P has a normal distribution while the
ing uncertainty is yielding. To avoid yielding for
geometric parameters are assumed deterministic.
the nominal conditions, we set the mean value of
The probability density function of P is assumed
the compressive stress of each material to be 2/3
of the form:
of the corresponding mean value of yield stress to
have a nominal safety factor of approximately 1.5.
(P μP )2 In order to generate such stochastic characteristics
−
1 2v 2
pdf ( P ) = e , 0 P<∞ (11) of compressive stress, the mean value and standard
2π v p2 deviation of the external forces P for the analysis
of each material are listed in Table 1.
where μp and νp are the mean value and standard
deviation of P. 3 YIELD STRESS OF THE MARINE
Since STRUCTURE COMPONENT
b( C b2 ( C)− PE σ C
4 AP 1 ⎡ 1n(σ y ) μ ⎤
P= = g( C) (13) F1( y) f ⎢−
erfc ⎥ (16)
2 2 ⎣ 2V 2 ⎦
725
Gumbel maximum:
⎡ ⎛α σy ⎞⎤
F3 ( y) = exp ⎢ exp ⎜ ⎥ (19)
⎣ ⎝ β ⎟⎠ ⎦
where
v 6
a = μ − γβ β = , γ = 0.5772 (20)
π
Gumbel minimum:
Figure 2. Compressive and yield stress distributions of
⎡ ⎛ αy σ ⎞ ⎤ A part.
F4 ( y ) = 1 − exp ⎢ exp ⎜ ⎥ (21)
⎣ ⎝ β ⎟⎠ ⎦
where
v 6
α = μ + γβ , β = , γ = 0.5772 (22)
π
Gumbel mix:
1
F5 ( y)
⎡ F ( y )Gumbel max
2⎣ (23)
+ F ( y )Gumbel min ⎤⎦
Figure 3. Compressive and yield stress distributions of
A part 2.
In Table 2, the mean values and standard devia-
tions of the yield stresses and compressive stresses
generated from geometric uncertainty and force
uncertainty are compared. The nominal safety fac-
tors SF = μ(σy)/μ(σc), for both geometric and load-
ing uncertainty cases are also provided in the table.
Comparisons of the distribution of compressive
stress and yield stress are shown in Figures 2 to 9.
The chi-square test method was used in our
former study (Yu et al. 2010) to evaluate the good-
ness of fit of each distribution of yield stress.
Chi-square values of all, first half and first quar-
ter of the intervals in the histogram of yield stress
were calculated. Since the overlap of compressive
stress and the yield stress is mainly located at the Figure 4. Compressive and yield stress distributions of
lower quarter of the yield stress distribution, both AH32 part 1.
726
Figure 7. Compressive and yield stress distributions of Table 4. Chi-square value for first quarter histogram of
AH36 part 2. yield stress.
727
728
REFERENCES
729