Академический Документы
Профессиональный Документы
Культура Документы
ABSTRACT take into account of the strength reduction of stiffened plate elements
after their ultimate strength as well as the time lag in collapse of
New ISUM stiffened plate model that consists of large plate elements individual elements. It is assumed that following the beam theory the
for local plate panels and beam-column elements for stiffeners is cross section remains plane and perpendicular to neutral axis. The
applied to the progressive collapse analysis of a ship’s hull girder under analysis is hence reduced to one-dimensional one. Yao et al (1991 and
longitudinal bending. The employed ISUM model is characterized by 1992) developed a method of progressive collapse analysis of ship’s
the shape functions for deflection of local plate panels based on the hull girder based on Smith’s approach with a semi-analytical
collapse modes and the ability to consider the localization of plastic formulation of average-stress/average strain relationships of stiffened
deformation. High computational efficiency and sufficient accuracy of plate elements.
the new ISUM stiffened plate model are demonstrated through a series
of ultimate strength analyses of continuous stiffened plates and a The Idealized Structural Unit Method (ISUM), proposed by Ueda and
progressive collapse analysis of a 1/3-scale welded steel frigate model Rashed (1984), gives one framework for implementing a more general
under longitudinal bending. approach to progressive collapse analysis of structural systems. Though
using a matrix formulation similar to conventional FEA, ISUM
KEY WORDS: Progressive collapse analysis; Ultimate strength; employs particular definitions of elements, which are of the same scale
Buckling/plastic collapse; Longitudinal bending; Stiffened plates; Ship in size as the structural members themselves. The material and
hull girder; Idealized Structural Unit Method geometrical nonlinearities are idealized and included in the element
formulations. Ueda et al (1993) and Paik (1995) developed the ISUM
INTRODUCTION plate element employing an effective width concept for post-buckling
behaviours. These models however cannot accurately predict the
If the working longitudinal bending moment exceeds the cross- decrease in load-carrying capacity beyond ultimate strength. In addition,
sectional capacity of ship’s hull girder, the buckling/plastic collapse of rather complicated element formulation is required because nonlinear
deck or bottom structure takes place and this may lead to an ultimate behaviour is idealized by analytical expressions of effective width.
collapse. So, the ability to predict the progressive collapse behaviour of
ship’s hull girder is a very important aspect of ship structure design. To eliminate complex analytical procedures of idealizing geometrical
nonlinearity, Masaoka et al (1998) proposed another ISUM plate
There exist several approaches to carry out the progressive collapse element. Inplane and out-of-plane deformations over the plate panel are
analysis of ship’s hull girder. Accurate solutions can be obtained by the described by shape functions, and the amplitude of the deflection was
application of conventional FEA considering both material and treated explicitly as a degree of freedom, beside nodal ones. In their
geometrical nonlinearities. For the collapse analysis of large systems original formulation, the shape function for deflection was
like a ship’s hull, however, tremendously large computing resources are approximated by elastic buckling modes. This approach however tends
required as well as manpower to achieve reliable results. So, many to overpredict the post-ultimate strength particularly for a thin plate or
studies have been performed to search for computationally efficient and a plate predominantly under transverse thrust. This is because the
sufficiently accurate simplified methods to predict the ultimate capacity collapse modes for these cases significantly different from the elastic
of ship’s hull girder. Smith (1977) developed a simplified method for buckling modes. From these observations, Fujikubo and Kaeding
the analysis of ultimate hull girder strength. The fundamental idea is to (2002) employed the shape functions based on the collapse mode to
have a better accuracy for the post-ultimate strength behaviors.
Copyright ©2004 The International Society of Offshore and Polar Applying this ISUM plate element, they developed an ISUM stiffened
plate model that consists of ISUM plate elements for local plate panels
Engineers. All rights reserved.
and beam-column elements for stiffeners. A new element subdivision
a/b=3.0 a/b=3.0
0 0 0 0
1 1.5 2 2.5 3 b 3.5 1 1.5 2 2.5 3 b 3.5 1 1.5 2 2.5 3 b 3.5 1 1.5 2 2.5 3 b 3.5
0.8 0.8
0.8 0.8
0.4 0.4
0.4 0.4
Astrup(1) Yao(1) Astrup(1) Yao(1) Astrup(1) Yao(1) Astrup(1) Yao(1)
Cho Yao(2) Cho Yao(2) Cho Yao(2)
Rigo(1) NEW 0.2 Rigo(1) NEW Rigo(1) New 0.2 Cho Yao(2)
0.2 0.2 Rigo(1) NEW
Soares FEM(ULSAS) Soares FEM(ULSAS) Soares FEM(ULSAS)
Soares FEM(ULSAS)
0 0 0 0
1 1.5 2 2.5 3 b 3.5 1 1.5 2 2.5 3 b 3.5 1 1.5 2 2.5 3 b 3.5 1 1.5 2 2.5 3 b 3.5
0 0 0 0
1 1.5 2 2.5 3 b 3.5 1 1.5 2 2.5 3 b 3.5 1 1.5 2 2.5 3 b 3.5 1 1.5 2 2.5 3 b 3.5
Fig. 4 Compressive ultimate strength of stiffened plate plotted against Fig. 5 Compressive ultimate strength of stiffened plate plotted against
slenderness ratio of panel (angle-bar stiffener) slenderness ratio of panel (tee-bar stiffener)
Figures 4 and 5 show the comparison of ultimate strengths of a determine the ultimate strength. Some methods give the predictions that
stiffened plate obtained by eight different methods. The ultimate considerably deviate from the FEA results. On the other hand, the
strength is expressed in terms of the average stress for a full cross present ISUM gives the predictions which are in good correlation with
section. Astrup(1) is based on a beam-column approach and Shanley’s the FEA results. For example, for some combinations of the stiffener
model is used. Rigo(1) is based on Rahman-Hughes’s model and Perry- size and the panel aspect ratio, b, the maximum value of ultimate
Robertson formula is applied. Soares’s method is based on a beam- strength exists when plotted against b. This is because for very thin
column approach considering the load-shedding after reaching ultimate plates the panel buckling strength and associated effective width are
strength. Yao(1) is also based on a beam-column approach and decreased with increase in b, while for very thick plates the stiffening
obtained from the average-stress/average-strain relationship given in a effect of stiffeners is relatively decreased with decrease in b. The
semi-analytical method (Yao and Nikolov, 1991, 1992). Yao(2) is a present ISUM captures this behavior. It, however, tends to
new simplified method improving Carlsen’s method (Fujikubo et al, underestimate the ultimate strength when compared with FEA. This is
1999a). Cho used NASTRAN. FEM (ULSAS) was obtained by the probably because the stability of the stiffener modeled by beam-column
non-linear FEA code developed by Yao et al. NEW represents the elements is more susceptible to the initial yielding of cross section than
results by the present ISUM stiffened plate model. The parameter b is the stiffener modeled by shell elements that allow the ductile cross-
the panel slenderness parameter and given as sectional deformation and the resulting more spread of yielding.
b s Yp Although this is to be remarked, it has been generally confirmed that
b= (7) the present ISUM model can give accurate and reasonable predictions
tp E
of the ultimate strength of a stiffened plate under longitudinal thrust.
where E and sYp are Young’s modulus and yield strength of the plate
panel, respectively. COLLAPSE ANALYSIS OF SHIP’S HULL GIRDER
Significant scatter exists among the predicted ultimate strengths in Figs. A 1/3-scale welded steel frigate model was tested by Dow (1991) under
4 and 5. This is mainly due to the difference in the collapse modes four-point bending in sagging condition, which resulted in overall
assumed, the formulation of panel effective width and the criteria to collapse accompanied by buckling/plastic collapse of deck, side and
The model consists of a central mild-steel section for the test and two
outer high-strength steel sections used as loading beams. The total
dimensions of the model are 18m length, 4.1m beam and 2.8m depth.
The material properties of the test section are: elastic modulus 207
GPa, yield stress 245 MPa, Poisson’s ratio 0.3.
Figure 6 shows the mid-ship section of the tested part. The frame space
is 457.2mm. To simplify constructions, decks and major bulkheads
other than the main load carrying structures were represented by
4
equivalent heavy fabricated tee-bars. The thickness of plate panel at the 6
central part of deck is smaller than that near the corner. Extensive 5
measurements of initial deformation are carried out for deck and side
shell including the overall component of longitudinal girders and
interframe component for stiffeners and plate panels. The weld induced Fig. 6 Cross section of 1/3-scale welded steel frigate model
residual stresses were also measured. transverse frame cross-section Trans. frame
u D1 u D2
ui i
The collapse behavior observed in the test in sagging condition was as
deck
follows (Dow, 1991): uz = 0
1. An overall grillage buckling of the deck structure initially took qx = 0
zi
q q
place between two transverse bulkheads. The occurrence of this side side
uy = 0 uy = 0
overall bucking had been predicted by preliminary buckling qx = 0 qx = 0
M M
m2 u m2
analysis because of the high slenderness of deck structure. It was m1
judged that the overall grillage buckling did not have a significant
influence on the hull girder moment/curvature response due to bottom
large stiffness and strength reserves. The test was then continued uz = 0 qz = 0 qz = 0
qx = 0 q y same q y same
up to the ultimate collapse in bending.
2. Approaching the ultimate load, local stiffener and plate panel u B1 u B2
frame space
deformations developed on the deck superimposed on the overall z z
deck deformation. y x
3. The side shell clearly showed inter-frame buckling of the
longitudinal stiffeners with associated plate buckling. Then, the Fig. 7 Boundary and loading conditions
mild steel test section reached the ultimate strength.
4. During the post-collapse phase of the loading, fairly large Pure bending moment is applied to two end cross sections in the
deformation of the structure was achieved; this resulted in following manner in both ISUM and FEM: first, a master node, m, is
significant tripping deformations on both the small longitudinals defined at an arbitrary height of a cross-section as shown in Fig.7.
and the deep girders representing internal deck structure in the test Denoting the axial displacement and the rotational angle at the master
section. node by, um and qm, respectively, and assuming that the cross section
5. Significant shifting of the neutral axis of the test section also remains plane, the axial displacement, ui, and the rotational angle, qi, at
occurred during the post-collapse phase. This was apparent from the node i located at a distance of zi from the master node are given by
the amount of compressive buckling of longitudinals that occurred {d i } = [ H i ]{d m } (8)
between No. 3 and 4 decks.
where d = ìui ü , é1 zi ù ìum ü
6. Final permanent deformed shapes show the inter-frame collapse of { i} í ý [ H i ] = ê0 , {d m } = í ý
the side shell and the internal damage to the test section. îqi þ ë 1 úû îq m þ
Method of Progressive Collapse Analysis On the other hand, the axial nodal force, fi, and the bending moment, Mi,
at the node i can be transformed to the equivalent nodal forces with
Procedure of Bending Moment Application. Longitudinal hull girder respect to the master node as
{Fm } = [ H i ] {Fi }
T
segment extending a half frame spacing in fore and aft directions from (9)
a transverse frame is considered, and a forced rotational angle is
ì fm ü ì fi ü
applied to both end cross-sections as shown in Fig.7 assuming that the where {Fm } = í ý , {Fi } = í ý
cross sections remain plane. Overall grillage buckling of deck structure, îM m þ îM i þ
as observed in the test, is not simulated, and the inter-frame collapse of Applying the transformation of Eqs. 8 and 9 to all the nodes at the
deck and side structure in longitudinal bending is analyzed. both-end cross sections, the stiffness equation with respect to the nodal
force {Fi} and the nodal displacement {di} is transformed to that with
For the progressive collapse analysis, the rotational angle, qm, at the
master node is increased at the both-end cross sections as shown in
Fig.7, allowing the axial displacement um at one end cross section and
taking the corresponding axial force, fm, as zero. The shift of neutral
axis due to buckling/yielding of structural members can be
automatically considered by allowing the axial displacement at the
master node under the condition of zero axial loads. The deformation of
transverse cross section in y-z plane is constrained at the location of
transverse frame as shown in Fig.7. The deck, side and bottom
Z
structures are modeled by double-span models, as Fig.3. The out-of- Y X
plane deformation of plate panels and stiffeners is constrained at the
location of the transverse frame, which corresponding to the y-axis in
Fig.3. The symmetrical boundary conditions are imposed on the out-of- Fig. 8 ISUM plate element mesh
plane deformation of plate panels and stiffeners at the both-end cross
sections, which correspond to the mid-span cross sections (x=a/2, -a/2)
in Fig. 3.
The calculated average stress-average strain curves are shown in Fig.10 0.4
where the average stress is non-dimensionalized by the yield stress and
0.3
the average strain by the yield strain. ISUM (with localization)
ISUM (no localization)
0.2 FEA(ULSAS)
ISUM results exhibit a lower stiffness after buckling than FEA. This is FEA(ABAQUS)
because the increase in panel buckling strength due to the torsional 0.1
stiffness of the stiffener (Fujikubo and Yao, 1999b) is not considered in
0
the present model. For ISUM model, the solution can follow the FEA 0 0.5 1 ex/eY 1.5
results only when the localization of plastic deformation is considered
in the panel deflection. An approach neglecting this localization shows Fig. 10 Average stress-average strain curves for typical stiffened plate
A
6 0.6
0
0
0 0.001 0.002 0.003 0.004 0.005
curvature(1/m) 0 2 4 6 8 ex/eY 10
Fig. 11 Moment-curvature relationships Fig. 13 Average stress-average strain curves in sagging condition