Marine Structures 59 (2018) 510–535

Contents lists available at ScienceDirect

Marine Structures
journal homepage: www.elsevier.com/locate/marstruc

A new method to estimate ultimate strength of stiffened panels

T
under longitudinal thrust based on analytical formulas
Murat Ozdemira,b,∗, Ahmet Ergina, Daisuke Yanagiharac, Satoyuki Tanakad,
Tetsuya Yaoe
a
Faculty of Naval Architecture and Ocean Engineering, Istanbul Technical University, Turkey
b
Faculty of Marine Sciences, Ordu University, Turkey
c
Graduate School of Engineering, Kyushu University, Japan
d
Graduate School of Engineering, Hiroshima University, Japan
e
Hiroshima University, Osaka University, Japan

AR TI CLE I NF O AB S T R A CT

Keywords: In this paper, a new approximate method based on analytical formulas is proposed to estimate the
Stiffened panels ultimate strength of stiffened panels by investigating collapse mechanisms of the stiffened panels.
Ultimate strength In this respect, firstly, a series of detailed elastoplastic large deflection Finite Element Analysis
Nonlinear FEA (FEA) is performed to investigate the collapse behavior and the ultimate strength of the stiffened
Simple method
panels. In the analysis, the initial deflections are considered in the form of thin-horse mode plus
overall buckling mode for the plates, and flexural buckling mode plus tripping mode for the
stiffeners. On the basis of the observed FEA results, an efficient method is proposed to estimate
the ultimate strength of stiffened panels based on Elastic Large Deflection Analysis (ELDA) with
the initial yielding concept. In the ELDA, the deflection modes are defined as the sum of overall
buckling mode plus local plate buckling mode. The welding residual stresses are not taken into
account in this study. The ultimate strength is predicted by examining yielding at several critical
points. The calculated results by the proposed method and the nonlinear FEA are compared, and
a very good agreement is obtained for all collapse scenarios investigated.

1. Introduction

Stiffened panels are main structural components of ship hulls. The ship hull is subjected to overall and local bending during
operation, and it has to resist the forces and bending moments caused by self-weights, cargo loads, buoyancy forces, wave forces etc.
[1–4]. Therefore, the ultimate bending moment capacity of the ship hull girder has been studied by many researchers, e.g. Refs. [5–8].
Due to the overall bending of the hull girder, the stiffened panels on the deck and bottom structures are subjected to longitudinal
thrust loads, respectively, under sagging and hogging conditions. Hence, many research works have also been performed on the
ultimate strength assessment of the stiffened panels by nonlinear FEA [9–15], simplified methods [16–32], and Idealized Structural
Unit Method (ISUM) [33–37].
The FEA is one of the most powerful tools to evaluate the ultimate strength as well as to simulate the collapse behavior. Owing to
the development of computer and software technologies, it has become possible to perform analysis for large-scale structures by
considering the nonlinearities arising from large deflection and plasticity. The recent works on the ultimate strength analysis of
stiffened panels by Finite Element Method (FEM) mainly focus on the effects of imperfections, modeling extent and boundary

∗
Corresponding author. Faculty of Naval Architecture and Ocean Engineering, Istanbul Technical University, Turkey.
E-mail address: mozdemir@itu.edu.tr (M. Ozdemir).

https://doi.org/10.1016/j.marstruc.2018.01.001
Received 9 May 2017; Received in revised form 3 January 2018; Accepted 3 January 2018
0951-8339/ © 2018 Elsevier Ltd. All rights reserved.
M. Ozdemir et al. Marine Structures 59 (2018) 510–535

conditions [9–13]. Despite the popularity of FEM, the ultimate strength analysis by FEM may not be economical for large-scale
structures.
Simple formulas have been proposed, in the open literature, to estimate the ultimate strength of stiffened panels. For instance, two
of the authors [16,17] proposed a simplified method for stiffened panels, with many attached stiffeners, under longitudinal thrust. In
these studies, the stiffened panels are idealized as a stiffener and attached plating combination, and the axial stress is computed at
several locations of the plate-stiffener structure. The initial yielding is assumed as the ultimate strength state of the structure. It is also
assumed that the local plate buckling takes place first. This method was then extended for the stiffened panels under combined biaxial
thrust and lateral loads [38–42].
Semi-analytical methods are also available in the open literature to obtain the ultimate strength of stiffened panels. For instance,
Paik et al. [18] proposed a simple analytical method for calculating the ultimate strength of the stiffened panels by considering initial
deflections and welding residual stresses. Three collapse modes, namely plate induced, stiffener induced and local buckling of the
stiffener web, are considered. A plate-stiffener combination model is used as a representative of the stiffened panel. Paik et al. [19]
introduced a large deflection orthotropic plate approach for estimation of the ultimate strength of stiffened panels considering overall
grillage buckling. In the study, axial compression/tension and lateral loading, as well as initial imperfections, are considered. In the
last decade, PULS code [20] was released by DNV for the ultimate strength analysis of stiffened panels. The fundamental theory
behind the code was presented in Refs. [21–23]. Byklum et al. [21] proposed a new computational semi-analytical method for
calculating the ultimate strength of stiffened panels. In the proposed method, the membrane stresses at certain critical locations are
examined using the Von Mises yield criterion, and the onset of yielding at those points is assumed as the ultimate strength state of the
structure. In-plane compression-tension, shear loading, and lateral load effects are also taken into consideration. Later, Byklum et al.
[22] improved the previous concept, allowing one-way interaction between global and local mode deflections. It was assumed that
the overall deflections do not influence the local mode of deflections. This is to say that local plate buckling is the primary buckling,
and the overall buckling of the stiffened panel takes place as secondary buckling. Zhang and Khan [30] performed nonlinear FE
computations and they defined the yielding modes at the ultimate state. A simple semi-analytical formula to estimate the ultimate
strength of stiffened panels under axial compression was also proposed in Ref. [30]. The application limits of this formula was then
revised in Ref. [31]. In 2006, Common Structural Rules were released for bulk carriers (CSR-B) [43] and double hull oil tankers (CSR-
T) [44], respectively, as the structural design standards. The possible failure modes for the stiffened panels are defined as local plate
buckling, overall panel buckling, and torsional stiffener buckling. The lowest failure strength of the possible failure modes is assumed
as the ultimate strength of the stiffened panel. In 2015, Harmonized Common Structural Rules [45] came into effect to supersede the
CSR-B and CSR-T. Recently, Kim et al. [32] proposed a simple empirical formula to obtain the ultimate strength of stiffened panels by
curve fitting based on the results of nonlinear FE computations.
In the present study, a detailed elastoplastic large deflection FEA is performed to investigate the collapse behavior and the
ultimate strength of stiffened panels. In the analysis, the initial deflections are considered in the form of thin-horse mode plus overall
buckling mode for the plates, and flexural buckling mode plus tripping mode for the stiffeners. Observing the FE simulations, a new
efficient and simple method, based on ELDA with the initial yielding concept, is proposed to estimate the ultimate strength of
stiffened panels under longitudinal thrust load. The initial and the total deflections for ELDA are assumed as the sum of deflections of
an overall stiffened panel buckling mode and a local plate buckling mode. The welding residual stresses are not taken into account in
this study.
Rest of the present paper is organized as follows: In Chapter 2, FEA is performed to evaluate the ultimate strength, as well as to
examine the collapse behavior of stiffened panels, then formulation and verification for ELDA and concept for ultimate strength
estimation are presented in Chapter 3. The estimated ultimate strength results and discussions are documented in Chapter 4. Main
conclusions are drawn in Chapter 5.

2. Finite element analysis of stiffened panels

2.1. Stiffened panels for analysis

In the nonlinear FEA, two different panels, namely bottom panel of a Bulk Carrier (BC) and deck panel of a Very Large Crude oil
Carrier (VLCC) in Ref. [1], are considered. The aspect ratios of the local plates are taken as a/ b = 3.0 for the BC panel, and as
a/ b = 5.0 for the VLCC panel. The local plate (a × b ) is considered with six different thicknesses in the analysis. In Fig. 1, a is the
distance between two successive transverse frames, while b denotes the longitudinal stiffener spacing.
The properties of the stiffened panels, which are adopted from Ref. [1] and used in the FEA, are given as follows.

• BC model:
a × b = 2,550 × 850 mm (a/ b=3.0)
tp = 33,22,16,13,11,9.5 mm
β = 1.01, 1.51, 2.07, 2.55, 3.02, 3.49,
• VLCC model:
a × b = 4,750 × 950 mm (a/ b =5.0)
tp = 37,25,18.5, 15,12.5, 11 mm
β = 1.00, 1.48, 2.00, 2.47, 2.97, 3.37 ,

511
M. Ozdemir et al. Marine Structures 59 (2018) 510–535

Fig. 1. Modeling extent of stiffened panels in the FEA (for the case of two longitudinal stiffeners between adjacent longitudinal girders).

where tp is the local plate thickness, and β is the slenderness ratio of the local plate, which is defined as:

b σY
β= .
tp E (1)

In Eq. (1), σY and E are the yield stress and Young's modulus of the material, respectively. As the stiffeners, flat-bar and tee-bar are
employed. The types and sizes of the stiffeners are presented in Table 1. The stiffener parameters are shown in Fig. 2.
Three different stiffener sizes, namely S1, S2, and S3, are adopted for all the panels under consideration (see Table 1). The size S1
enables the overall buckling of the stiffened panels. While, the size S2 has intermediate flexural stiffness, which enables both the
overall buckling and the local plate buckling, depending on the plate thickness as well as the number of stiffeners. The size S3 is the
largest one, and it prevents the overall buckling of the stiffened panel. However, the stiffeners of size S3 are vulnerable to torsional
buckling. The notation used for the stiffened panels are similar to that of Ref. [1]. For example, ”F5S2N8B20″ stands for the stiffened
panel with eight flat-bar stiffeners of the size S2, of which local plate has an aspect ratio of 5.0 and the slenderness ratio of 2.0. In the
computations, the number of stiffeners is varied as one, two, three, four and eight, respectively, for each local plate size (a × b × tp ).

2.2. Modeling extent and boundary conditions

The transverse frames and longitudinal girders are not modeled explicitly, but simply supported boundary condition is imposed
along the longitudinal girder and transverse frame lines. When the number of half-waves in local buckling mode in the longitudinal
direction is odd, the symmetry condition is imposed along pr and qs edges (see Fig. 1), which is expressed as:

⎧ u: uniform,
⎨ θ = θz = 0. (2)
⎩ y

In order to deal with even number of buckling half-waves in longitudinal direction, periodical boundary conditions as given below,
have to be adopted along pr and qs .

Table 1
Types and sizes of stiffeners.

Size Flat-bar Tee-bar

S1 150 × 17 150 × 9 +90 × 12

S2 250 × 25 250 × 10 +90 × 15
S3 350 × 35 400 × 12 +100 × 17

512
M. Ozdemir et al. Marine Structures 59 (2018) 510–535

Fig. 2. Stiffener types and parameters.

⎧u: uniform,
⎪ wpr = wqs at the same y− coordinate,
⎨ θy pr = θyqs at the same y− coordinate,
⎪ θ = θ = 0 at the same y−coordinate.
⎩ z pr z qs (3)

In this paper, however, the numbers of buckling half-waves are three and five, and therefore Eq. (3) is not used. As for the modeling
extent in the transverse direction, 1 + 1 full bays modeling is considered when the number of stiffeners is one. A full breadth between
adjacent longitudinal girders is considered, as the extent of modeling, when the number of stiffeners is three. In those cases, the
simply supported boundary conditions with straight edges is adopted along the longitudinal girders. On the other hand, when number
of stiffeners with symmetric cross-section is even, the symmetry boundary condition can be imposed along pdq and res edges, which
are the cross-sections at the center of two central stiffeners (see Fig. 1) as follows:

⎧ v: uniform,
⎨ θ = θz = 0. (4)
⎩ x

In Eqs. (2)–(4), u and v are the displacements in the x and y directions, respectively. While, θx , θy and θz are the rotations about x, y
and z axes, respectively.

2.3. Finite element model

ANSYS [46] finite element solver is employed for the FEA computations using bilinear general purpose finite strain shell element,
which has four nodes and six degrees of freedom (DOF) per node. The adopted shell element simulates the large strains and rotations
in the nonlinear calculations, on the basis of Mindlin-Reissner shear deformation theory for thin to moderately thick shells. As for the
material modeling, the elastic-perfectly plastic assumption is made. The material properties adopted in the analysis are as follows:
The yield stress of the material is σY = 313.6 MPa, Young's modulus is E = 205,800 MPa, and Poisson's ratio is ν = 0.3.
As for the meshing of the stiffened panel, one local buckling half-wave is divided by 10×10 elements regardless of the plate aspect
ratio. Moreover, the stiffener webs and flanges are divided into six elements towards the depth and breadth directions, respectively.
This discretization is independent of the stiffener size. The finite element model of a typical stiffened panel is depicted in Fig. 3.
The initial deflections are assumed as in the form of overall plus thin horse mode for the plate, and overall plus tripping mode for
the stiffeners, see Refs. [1,4], for details. As for the evaluation of the ultimate strength, forced displacement is imposed in the
longitudinal direction, and then the reaction forces are obtained at each node along the forced edge. The calculated total reaction
force in the longitudinal direction along the forced edge is divided by the total cross-sectional area to evaluate average stress, σavg . The
average stress–average strain curves are then obtained.

Fig. 3. Finite element model of a typical stiffened panel (double span/double bay model with four stiffeners).

513
M. Ozdemir et al. Marine Structures 59 (2018) 510–535

Fig. 4. Comparison of present results with those of Tanaka et al. [1].

2.4. Results of finite element analysis and comparisons

The ultimate strengths of the stiffened panels are evaluated by FEA and compared with the results calculated by MSC Marc in Ref.
[1]. The comparison of the present calculations with the reference results is given in Fig. 4.
For the results presented in Fig. 4, the ultimate stress is normalized by the yield stress, and N1, N2, N3, N4 and N8 stand for the
cases of one, two, three, four and eight attached stiffeners, respectively. It can be seen from Fig. 4 that a very good agreement is
obtained with the reference results. Furthermore, the collapse behavior of the stiffened panels is examined. In Fig. 5(a), the FEA
results are given for the stiffened panels, which undergo overall collapse. In the given figure, the ultimate stress, σu , is normalized by
the yield stress in the vertical axis, while the horizontal axis stands for the plate slenderness ratio. It is clearly seen from Fig. 5(a) that
the ultimate strength of the stiffened panels can be directly related to number of attached stiffeners (and so the aspect ratio of whole
stiffened panel), and to the plate slenderness ratio when weak stiffeners are provided. Those panels mainly collapse due to the
occurrence of overall buckling, which causes a rapid increase in deformations and sudden loss of load carrying capacity due to the
occurrence of stiffener yielding.
The Von Mises stress distribution and collapse deformation of T5S1N8B35 stiffened panel at the ultimate strength state are shown
in Fig. 6, and the average stress–average strain relationship is shown in Fig. 7(a). Fig. 6 shows that, when the ultimate state is
attained, a slight level of local plate buckling deformation exists. In Fig. 7, the vertical axis is the average stress normalized by yield
stress, while εavg stands for the average longitudinal strain and εY for the yield strain. In the average stress–average strain curves, the
onsets of plastic deformation for the stiffeners and the plate are indicated by ”□ ” and ”△ ” marks, respectively. Fig. 7(a) indicates
that the occurrence of secondary buckling of an overall mode causes drastic reduction of the load carrying capacity of the stiffened
panel.
In case of the stiffened panels with the stiffeners of the largest size, the number of stiffeners has a negligible effect on the ultimate
strength of the panels, and the collapse mechanism is related to the local buckling and collapse of plating and/or stiffeners, as seen in
Figs. 5(b) and 8. Furthermore, the ultimate strength might be almost same with the yield strength of the material, depending on plate
thickness.
In addition to the ultimate strength, the initial yielding mechanism is also examined. It is observed that, when the collapse is due
to overall panel buckling, the initial yielding mostly takes place at the top of stiffeners, and the reserve strength up to the ultimate

Fig. 5. FEA results: (a) overall collapse case (F5S1-series), (b) local collapse case (T3S3-series).

514
M. Ozdemir et al. Marine Structures 59 (2018) 510–535

Fig. 6. Von Mises stress distribution (MPa) and deflection mode of T5S1N8B35 at ultimate state.

Fig. 7. Average stress-average strain curves: (a) T5S1N8B35 panel, (b) T3S3N8B35 panel.

Fig. 8. Von Mises stress distribution (MPa) and deflection mode of T3S3N8B35.

strength is very small when the slender plates are accompanied by a large number of weak stiffeners. As seen in Fig. 7(a), the ultimate
state is attained shortly after the initial yielding at the top of stiffeners, while the plate is still fully elastic at the ultimate state.
On the other hand, in the case of slender plates with the stiffeners of the largest size, the initial yielding is mostly observed in
plating, and soon after the initial yielding in plating along the mid-span line, the plastic deformation is observed at the stiffener
bottoms, see Fig. 7(b). In case of the local collapse, the reserve strength up to the ultimate strength beyond the initial yielding is
relatively large compared to the overall collapse case, see Fig. 7(b). This phenomenon may be due to that the deformation in the
stiffeners is very small compared to that of the plate, and therefore the stiffeners can carry additional thrust load beyond the initial
yielding.
In concrete, collapse occurs due to the overall stiffened panel buckling when weak stiffeners are provided. In this case, the
ultimate strength is usually higher than the elastic overall buckling strength of the stiffened panels. When the stiffeners of

515
M. Ozdemir et al. Marine Structures 59 (2018) 510–535

Fig. 9. Column type collapse of stiffener with attached plating (Case 1): (a) load-deflection curve, (b) load-shortening curve, (c) collapse mode.

intermediate flexural stiffness are provided, the column type collapse is usually observed, and the ultimate strength is lower than the
Euler column buckling strength of the stiffener with attached plating. On the other hand, local plate buckling collapse is usually
observed when heavy stiffeners are provided. In addition, the ultimate strength becomes very close to the yield strength of the panel
material when heavy stiffeners are accompanied by thick plate.
Horizontal bending of the stiffeners takes place when the ratio of the stiffener web's height to thickness is relatively large.
However, the horizontal bending of the stiffeners usually becomes significant after the ultimate strength is attained. It should be
noticed that the ultimate strength either decreases or remains same even if the plate thickness is increased, when the plate slen-
derness ratio is lower than a critical value (β ≤ 1.80 ) in case of heavy stiffeners are provided. Such arrangement with too heavy
stiffeners and thick plates is thus not practical in real ship structures.

3. Ultimate strength estimation by proposed new method

3.1. Possible collapse modes

As observed from the results of nonlinear FEA in the previous chapter, the mostly observed collapse behaviors of the stiffened
panels could be classified into four cases. In Case 1, the collapse is due to the column buckling of stiffeners with attached plating,
prior to the local plate buckling, as schematically indicated in Fig. 9(a) and (b). Furthermore, collapse mode is also given in Fig. 9(c).
Case 1 is observed for relatively thick plates with the stiffeners of intermediate flexural stiffness. The range of occurrence of the
column type collapse prior to the local plate buckling (Case 1) may be defined based on observations from the FE computations as:
0.60 < λ ≤ 1.30 and β ≤ 1.80, (5)
where λ denotes the slenderness ratio of the stiffener with the attached plating (column slenderness ratio), which is defined as:
σY
λ= .
σcrcl (6)
In Eq. (6), σcrcl
denotes the Euler column buckling stress for simply supported column(stiffener) with attached plate considering full
breadth. Case 2 is the column type collapse after local plate buckling takes place as shown in Fig. 10. This type of collapse is generally
observed for thin plates with the stiffeners of intermediate flexural stiffness. In this case, the effect of the number of stiffeners on the
ultimate strength is relatively small. The column type collapse after local plate buckling (Case 2) usually occurs in the following

Fig. 10. Column type collapse after local plate buckling (Case 2): (a) load-deflection curve, (b) load-shortening curve, (c) collapse mode.

516
M. Ozdemir et al. Marine Structures 59 (2018) 510–535

Fig. 11. Local collapse after local plate buckling (Case 3): (a) load-deflection curve, (b) load-shortening curve, (c) collapse mode.

range.

0.60 < λ ≤ 1.30 and β > 1.80. (7)

In Case 3, the local collapse is observed after local plate buckling. The Load–deflection/displacement behaviors for this case are
schematically illustrated in Fig. 11. This type of collapse is observed for thin plates with deep stiffeners, and the column buckling or
the overall buckling never happens before the ultimate strength is attained. The effect of the number of stiffeners on the collapse
behavior and ultimate strength is small. The local plate buckling collapse is usually observed in the following range:

λ ≤ 0.60, (8)

Case 4 is the overall collapse of the stiffened panel observed after the overall buckling. In this case, deformations increase
drastically after the occurrence of overall buckling. This collapse is mainly observed for thick plates with many weak stiffeners. The
load–deflection/displacement behaviors for Case 4 are given in Fig. 12. In this case, the local plate buckling never takes place before
the ultimate strength is attained. Overall buckling collapse mostly takes place in the following range:

λ > 1.30. (9)

Overall buckling collapse (Case 4) is not considered in many simplified methods including those proposed in Refs. [16] and [17]. This
is because this type of collapse may be rare in actual ship structures.
In Figs. 9–12, Pcrcl , Pcrpt and Pcrstpl denote the critical buckling loads for the stiffeners with attached plating, local plate between the
longitudinal stiffeners and overall buckling of stiffened panels, respectively. On the other hand, u and w are end–shortening and out-
of-plane displacements of the stiffened panel, respectively.

3.2. Formulation for ELDA

The finite element simulations show that the ultimate strength is not only related to buckling behaviors, but also post-buckling
behaviors. The ultimate strength state is attained due to the occurrence of secondary buckling of an overall mode, when slender plates
are accompanied by many weak stiffeners. In this case, the ultimate strength is related to the secondary buckling.
The ELDA of stiffened panels is performed to simulate both primary and secondary buckling behaviors as well as to estimate the
ultimate strength. The present formulation takes into account the overall buckling of a stiffened panel and the local buckling of the
plate between the longitudinal stiffeners. The deflection components are given in Eq. (10).

Fig. 12. Overall collapse after overall buckling (Case 4): (a) load-deflection curve, (b) load-shortening curve, (c) collapse mode.

517
M. Ozdemir et al. Marine Structures 59 (2018) 510–535

Fig. 13. Modeling extent for ELDA (for the case of stiffened panel with two stiffeners).

πx πy mπx nπy
w = wov + wloc = C sin sin + Amn sin sin .
a B a B (10)

Hence, initial deflections are imposed in the form of assumed deflection modes as:
πx πy mπx nπy
w0 = w0ov + w0loc = C0 sin sin + A0mn sin sin ,
a B a B (11)

where m and n are the numbers of the local buckling half-waves in longitudinal and transverse directions, respectively. n = N + 1,
and N is the number of stiffeners.

3.2.1. Extent of modeling

The problem domain is indicated in Fig. 13. When the initial deflections do not exist, the thrust load can be applied at the centroid
of the cross-section of the stiffened panel, without load eccentricity, until buckling takes place. However, when the initial deflections
exist, the deflections will increase with increasing load, and the centroid of the cross-section will move with increasing applied load.
In this case, initially non-eccentric load becomes eccentric load with increasing load. To avoid such eccentricity, symmetry lines
( pq −rs ) and full breadth model ( pr −qs ) in Fig. 13 are considered in the longitudinal and transverse directions, respectively. This
modeling is referred as double-span/single-bay model.

3.2.2. Stresses and strains in plating

Assuming plane stress condition, the compatibility equation of the plate is expressed as:

F , xxxx + 2F , xxyy + F , yyyy = E (w ,2xy − w , xx w , yy − w0,2 xy + w0, xx w0, yy ), (12)

where F is Airy's stress function, and subscript comma denotes partial differentiation. Substituting Eqs. (10) and (11) into Eq. (12), the
solution of the compatibility equation is derived as follows, see Ref. [4].

F= ( Eα 2
32
2πx
cos a +
E
32α2
2πy
cos B ) (C 2 − C02)

(m − n)2Eα2
−⎡
(1 − m) πx (1 − n) πy
cos cos B
⎢ 4{(1 − m)2 + (1 − n)2α2}2 a
⎣
(m + n)2Eα2 (1 + m) πx (1 − n) πy
− 2 cos a
cos B
4{(1 + m)2 + (1 − n)2α2}
(m + n)2Eα2 (1 − m) πx (1 + n) πy
− 2 cos a
cos B
4{(1 − m)2 + (1 + n)2α2}

(m − n)2Eα2
cos B ⎤ (CAmn
(1 + m) πx (1 + n) πy
+ 2 cos a ⎥
− C0 A0mn )
4{(1 + m)2 + (1 + n)2α2}
⎦
+ ( n2Eα2
32m2
2mπx
cos a +
m2E
32n2α2
2nπy
cos B ) (A 2
mn − A02mn ).
(13)

Here, α = a/ B is aspect ratio of the whole stiffened panel. In-plane stress components can be easily derived from Eq. (13) using the
theory of elasticity as:

σxp = F , yy , σyp = F , xx , τxyp = −F , xy. (14)

Corresponding in-plane strain components are derived as:

518
M. Ozdemir et al. Marine Structures 59 (2018) 510–535

1
εxp = (σ
E xp
− νσyp),
1
εyp = (σ
E yp
− νσxp),
2(1 + ν )
γxyp = E
τxyp. (15)

Bending strain components are expressed as:

εxb = −z (w , xx − w0, xx ),
εyb = −z (w , yy − w0, yy ),
γxyb = −2z (w , xy − w0, xy ), (16)

where z is from − tp/2 to tp/2 . Applying the stress-strain relationships of the plane stress condition, bending stresses are expressed as
follows:
E
σxb = (ε + νεyb),
1 − ν 2 xb
E
σyb = (ε + νεxb),
1 − ν 2 yb
E
τxyb = γ .
2(1 + ν ) xyb (17)

It should be noticed that the overall bending takes place with respect to neutral axis of the whole cross-section, and this causes
additional in-plane strain components in plating as:
a
εxp = −(−ep)(wov, xx − w0ov, xx ),
a
ε yp = −(−ep)(wov, yy − w0ov, yy ),
a
γxyp = −2(−ep)(wov, xy − wov0, xy ). (18)

Here, ep represents the distance between mid-thickness plane of the plate and neutral axis of the whole cross-section, and it is
depicted in Fig. 14. The corresponding additional in-plane stresses are expressed as:
a E
σxp = (ε a a
+ νε yp ),
1 − ν 2 xp
a E
σ yp = (ε a a
+ νεxp ),
1 − ν 2 yp
a E
τxyp = γa
2(1 + ν ) xyp (19)

3.2.3. Stress and strain in stiffeners

In addition to in-plane strains in plating, the overall bending produces bending strains in the stiffeners. Hence, the strain of the i-
th stiffener located at y = yi is expressed as:
a
εsi = (εxp + εxp ) y = yi − z ′ (wov, xx − w0ov, xx ) y = yi . (20)

In Eq. (20), z ′ is the coordinate of a certain point in the stiffener measured from the mid-thickness plane of the plate. Finally, the axial
stress in the stiffeners is given as:
σsi = Eεsi. (21)

Fig. 14. Cross-section of a typical stiffener.

519
M. Ozdemir et al. Marine Structures 59 (2018) 510–535

3.2.4. Application of principle of virtual work

A stiffened plate is in equilibrium state under average longitudinal compressive load, producing average stress, σx , as well as stress
a a a a a a
components σxp , σyp , τxyp , σxb , σyb , τxyb , σxp , σ yp , τxyp , σsi and corresponding strain components, εxp , εyp , γxyp , εxb , εyb , τxyb , εxp , ε yp , γxp and εsi ,
respectively. At this state, a small virtual displacement increment is applied, and then the virtual work done by internal forces are
expressed as:
a /2 B tp/2
δWi = ∫−a/2 ∫0 ∫−t /2 [(σxp + σxpa + Δσxp + Δσxpa)(δ Δεxp + δ Δεxpa) + (σyp + σypa + Δσyp + Δσypa)(δ Δεyp + δ Δεypa)
p

a a ⎛ a ⎞
+ (τxyp + τxyp + Δτxyp + Δτxyp ) ⎜δ Δγxyp + δ Δγxyp ⎟ + (σxb + Δσxb ) δ Δεxb + (σyb + Δσyb ) δ Δε yb + (τxyb + Δτxyb ) δ Δγxyb] dxdydz

⎝ ⎠
N a /2
+∑ ∫−a/2 ∫A
(σsi + Δσsi ) δ Δεsi dAsi dx .
si
i=1 (22)

In Eq. (22), Asi is the cross-sectional area of the i-th stiffener. In case of the collapse by local buckling, the influence of interaction
between the rotation of the plate and the stiffener along stiffeners' lines may become significant. The interaction between the plate
and the stiffeners is not considered explicitly, but artificial torsional springs are introduced along stiffeners' lines. Then, the internal
virtual work done by the torsional springs, which is expressed as below, is added to the total internal virtual work in Eq. (22).
N 1 a /2
δWϕi = ∑ a ∫−a/2 (Mi + ΔMi) δ Δϕi dx ,
i=1 (23)

where ϕi is angle of rotation of the plate, which is defined as:

ϕi = (w , y − w0, y ) y = yi . (24)

The torsional moment developed in spring by rotation of the plate can be written in incremental form as:
Mi + ΔMi = kt (ϕi + Δϕi ), (25)

where kt is torsional constant of the artificial springs, and it is taken as a function of the local plate dimensions and torsional stiffness
of the stiffener cross-section as:
3
⎛ t p ⎞ C′
kt = ⎜ ⎟ 2 .
⎝b⎠a (26)

C ′ is provided in Appendix 1. a is distance between two adjacent frames. On the other hand, the virtual work done by the external
force is expressed as follow.
N
⎡
δWe = (σx + Δσx ) ⎢Btp + ∑ Asi ⎤⎥ δ Δu .
⎣ i=1 ⎦ (27)

In Eq. (27), u is end shortening in the longitudinal direction, and is expressed for the 1/2 + 1/2 double-span model as:
a 1 a /2 B 1 a π2 2 m2π 2 2
u = − σx +
E B
∫−a/2 ∫0 2
[(w , xx )2 − (w0, xx )2] dxdy = − σx −
E 8a
(C − C02) −
8a
(Amn − A02mn ).
(28)

From the equality of the virtual works done by internal and external forces, incremental equation system is derived as:

⎡ a11 a12 ⎤ ⎧ ΔC ⎫ = ⎧ b1 + c1 Δσx ⎫.

⎢ a21 a22 ⎥ ⎨ ΔAmn ⎬ ⎨ b2 + c2 Δσx ⎬ (29)
⎣ ⎦⎩ ⎭ ⎩ ⎭
Solution of the incremental equations for the increment of deflection coefficients, ΔC and ΔAmn , is derived as:

⎧ ΔC ⎫ = 1 ⎡ a22 − a12 ⎤ ⎧ b1 + c1 Δσx ⎫.

⎨ ΔAmn ⎬ a11 a22 − a12 a21
⎢− a21 a11 ⎥ ⎨ b2 + c2 Δσx ⎬ (30)
⎩ ⎭ ⎣ ⎦⎩ ⎭
Parameters in Eq. (30) are provided in Appendix 2. Finally, the average stress-average strain relationship is derived as:
1 π2 m2π 2 2
εavg = − σx − 2 (C 2 − C02) − (Amn − A02mn ).
E 8a 8a2 (31)

Here, the minus sign denotes the compressive average strain.

3.3. Accuracy of proposed ELDA method

As for the verification of the proposed analytical ELDA method, ELDAs are also performed by FEM. The initial deflections are
assumed to have the same magnitude and mode for FEM and analytical ELDA models as:

520
M. Ozdemir et al. Marine Structures 59 (2018) 510–535

Fig. 15. Average stress–average strain curves for square panels (α = 1.0): (a) h/ tw = 4.0, (b) h/ tw = 10.0.

C0 = 0.001a, A0mn = 0.01tp. (32)

For the panels under consideration, the stiffener spacing is b = 750 mm, and the thicknesses of plate and stiffener web are the
same, which are tp = tw = 16.0 mm. The aspect ratio of the local plate is taken as a/ b = 3.0. The number of stiffeners adopted in the
calculations is two (α = 1.0) and four (α = 0.6), respectively. Moreover, the stiffener web height is taken as h/ tw = 4.0 and 10.0,
respectively, to simulate the overall buckling of stiffened panel and the local plate buckling. Young's modulus is assumed as
E = 205,800 MPa. Poisson's ratio, ν is 0.3.
The ELDAs by FEM are performed considering the double-span/double-bay model, which is the same with the FEA model of the
ultimate strength calculations in Chapter 2, see Fig. 1. The average stress–average strain curves for the square stiffened panels are
given in Fig. 15. In the given figures, ”Analytical” denotes the ELDA by the developed method.
The buckling modes are the overall buckling and the local plate buckling, respectively, for the cases of h/ tw = 4.0 and
h/ tw = 10.0, as shown in Fig. 15(a) and (b). In the given figure, σcrov and σcrpt denote the overall buckling stress of the stiffened panel
and the local plate buckling stress, respectively, and they are provided in Appendix 1. εcrov and εcrpt are the corresponding strains for the
overall buckling stress and the local plate buckling stress, respectively. In the case of overall buckling, the stiffnesses obtained by the
proposed analytical method and FEM beyond the buckling coincide well. However, the differences increase with increasing load. This
is because the deflection components other than assumed buckling modes increase with increasing load, and, as a result, the de-
flection mode changes from a simple sinusoidal mode to a flattened mode. This can be observed by comparing the deflection modes
soon after buckling and in post-buckling range, for the case of h/ tw = 4.0 as shown in Fig. 16(a) and (b). The flattened mode can be
observed more clearly in Fig. 19(b).
As for the local buckling mode, such flattening is not observed in the strain range considered in the analysis, as seen in Fig. 17. A
slight difference is observed between the buckling stresses by the proposed analytical method and FEM in Fig. 15(b). This difference
is attributed to the difference in assumed resistance of the stiffeners against plate rotation along the stiffeners' lines in the proposed
analytical method and the FEA.
The average stress–average strain curves are presented for the wide stiffened panels in Fig. 18. It can be observed from Fig. 18(a)
that the difference between the results by the proposed analytical method and FEM starts to increase shortly after the occurrence of the
overall buckling. This difference may be explained by the earlier flattening of the overall deflection mode in case of small aspect ratios.
In this case, the post-buckling behavior cannot be accurately simulated by adopting simple two sinusoidal deflection components.
Deflection modes for overall buckling case are given in Fig. 19. This figure indicates that the flattening of deflection mode beyond
the buckling is more significant when the aspect ratio of a whole stiffened plate is small. The deflection modes for the local buckling
are given in Fig. 20. It is clear from the figure that the post-buckling behavior is simulated well by the proposed analytical method in

Fig. 16. Deflection modes for square panels (α = 1.0), h/ tw = 4.0: (a) buckling deformation mode, (b) post-buckling deformation mode.

521
M. Ozdemir et al. Marine Structures 59 (2018) 510–535

Fig. 17. Deflection modes for square panels (α = 1.0), h/ tw = 10.0: (a) buckling deformation mode, (b) post-buckling deformation mode.

Fig. 18. Average stress–average strain curves for wide panels (α = 0.6): (a) h/ tw = 4.0, (b) h/ tw = 10.0.

Fig. 19. Deflection modes for wide panels (α = 0.6), h/ tw = 4.0: (a) buckling deformation mode, (b) post-buckling deformation mode.

Fig. 20. Deflection modes for wide panels (α = 0.6), h/ tw = 10.0: (a) buckling deformation mode, (b) post-buckling deformation mode.

522
M. Ozdemir et al. Marine Structures 59 (2018) 510–535

Fig. 21. Stiffened panel model and presumed yielding control points.

the analyzed range of strains. With a further increase in applied average stress, flattening shall take place. Flattening observed in
Figs. 16(b) and 19(b) is attributed to the development of deflection components other than the assumed buckling modes. This is
explained in sections 4.5.2.2 and 4.5.4.2 of Ref. [4].

3.4. Concept for the ultimate strength estimation by ELDA

As explained in Section 3.1 based on the results of nonlinear FEAs, the collapse of stiffened panels is usually induced by local plate
buckling, column type buckling and overall buckling. In the case of collapse by local plate buckling, the initial yielding takes place in
the plate part, then plastic deformation spreads towards the stiffener's bottom. The reserve strength beyond initial yielding is rela-
tively larger for the local plate buckling collapse. Thus, yielding examination at the bottom of the stiffener becomes the most
reasonable approximation for the local plate buckling collapse. In the cases of column type collapse and overall buckling collapse, the
initial yielding usually takes place at the stiffener's top. In those cases, the yielding condition may be examined at the stiffener's cross-
section above the neutral axis of the whole panel section. On the basis of these observations, two different points are considered for
the stress examination in the ELDA model as shown in Fig. 21. When the number of stiffeners is odd, the yielding condition is
examined at the cross-section of central stiffener. On the other hand, the yielding condition has to be examined at the cross-section of
either two central stiffeners, when the number of stiffeners is even.
Point 1 is considered for the overall buckling collapse and the column type collapse, in which cases the onset of yielding usually
occurs at the stiffener's top. On the other hand, point 2 is considered for the local plate buckling collapse. Examination of yielding
condition at point 2 also reveals very good ultimate strength estimations when the ultimate strength is very close to the yield strength
of the material.
The column slenderness ratio is taken as a parameter to decide the location of yielding examination points for the aforementioned
collapse cases. Then, the conditions for the yielding examination points are defined based on the results of FE computations and given
in Table 2.
Both stresses at point 1 and point 2 are computed at the stiffener's cross-section. The axial stress at any point of the stiffener's
cross-section can be computed by Eq. (33), which is the explicit form of Eq. (21). When the axial stress given in Eq. (33) is equal to the
yield stress of the material, it is considered that the ultimate strength is attained. The related parameters in Eq. (33) are given in
Appendix 2. When the yielding condition is examined at point 2, z′ is taken as tp/2 , describing the location of the stiffener's bottom
from the mid-thickness plane of the plate. As for the point 1, that is at the top of the stiffener, z′ is set as h + tp/2.

2πyi 2πx ⎤ 2 (1 − m) πx (1 − n) πyi

σsi = −⎡px1 cos − νpy1 cos (C − C02) − σx − ⎡ (px 2 − νpy2 )cos cos
⎣ B a ⎦ ⎢ a B
⎣
(1 + m) πx (1 − n) πyi (1 − m) πx (1 + n) πyi
− (px 3 − νpy3 )cos cos − (px 4 − νpy 4 )cos cos
a B a B
(1 + m) πx (1 + n) πyi ⎤ 2nπyi 2mπx ⎤ 2
+(px5 − νpy5 )cos cos (CAmn − C0 A0mn ) − ⎡px 6 cos − νpy6 cos (Amn − A02mn )
a B ⎥ ⎣ B a ⎦
⎦
πx πy
+ (z′ − ep) Eebx1sin sin i (C − C0).
a B (33)
When λ = λ cr = 0.60, the ultimate strength values calculated at point 1 and point 2 do not coincide in general. In this respect, if λ
is around the critical value, the ultimate strengths are obtained at both point 1 and point 2. Then, the new ultimate strength is

Table 2
The conditions for yielding examination.

Collapse case Condition Stress examination point

Column type collapse 0.60 < λ ≤ 1.30 Point 1

Overall buckling collapse λ > 1.30 Point 1
Local plate buckling collapse λ ≤ 0.60 Point 2

523
M. Ozdemir et al. Marine Structures 59 (2018) 510–535

Fig. 22. Schematic illustration of ultimate strength interpolation near λ cr .

obtained by the linear interpolation using those values. If the following condition in Eq. (34) is satisfied, the linear interpolation is
performed between two ultimate strengths for a smooth ultimate strength estimation as schematically illustrated in Fig. 22.
λ cr − Δλ < λ < λ cr + Δλ, (34)

where Δλ = 0.02 .
In case of the yielding examination at point 1, the location of point 1 is shifted using Eq. (35) towards the neutral axis of the whole
cross-section. This modification is needed to approximately take into account the reserve strength beyond the initial yielding. The
reserve strength is related to the column slenderness ratio (λ) and aspect ratio of the whole panel (α), when the overall buckling
collapse and column type collapse take place. The distance, z t , between point 1 and the neutral axis of the whole cross-section is given
as:
z t = ηz (z′ − ep), (35)

where ηz is a parameter to modify the location of point 1, and it is defined as:

⎧ 1 0.60 < λ ≤ 0.65

⎪ 20(λ − 0.65) 1
2 + 20(0.7 − λ ) 0.65 < λ < 0.7
ηz = (R σ λ0.7α )
⎨ 1
⎪ 2
λ ≥ 0.7.
⎩ (R σ λ0.7α ) (36)

When ηz >1.0, ηz is set as 1.0. Rσ is the ratio of buckling stresses and expressed as:
σcrov
Rσ = ,
σcrcl (37)

where is the overall buckling stress of the stiffened panel, and it is provided in Appendix 1. Eq. (36) is derived by curve fitting
σcrov
assuming that relationship between the ultimate strength of finite and infinite number of stiffeners cases is similar to the ratio of
overall buckling stress to the column buckling stress (Rσ ). When the number of stiffeners is few, Rσ becomes higher than unity. On the
other hand, as for the cases with many stiffeners, Rσ close to unity is obtained.

4. Estimated results by proposed method and discussions

The ultimate strength of stiffened panels is estimated by ELDA with the initial yielding concept. Detailed comparison of the
ultimate strength estimations by present ELDA method and FEA is carried out in Fig. 23 through 26. In those figures, the ultimate
strengths obtained by FEA are indicated by different marks with different colors. On the other hand, the ultimate strength estimations
by present ELDA method are represented by smooth lines. The different colors for marks and lines stand for different number of
stiffeners. The calculated results are plotted against slenderness of the local plate, β, given by Eq. (1).
In the case of F3-series, the agreement is mostly well between the results of FEA and proposed method. However, as for the cases
with few weak stiffeners and lower plate slenderness ratios (F3S1 and T3S1 series with thick plates and one stiffener), the proposed
method underestimates the ultimate strength slightly. The maximum differences between FEA and ELDA estimations are %7.23 and
%10.58 of FEA results for F3S1 and T3S1 series, respectively. The collapse modes of F3S1-series may be classified as the column type
collapse by column buckling in case of thick plate and column type collapse after local plate buckling in the case of thin plate,
respectively. In those cases, the yielding condition is examined at point 1, then the location of point 1 is shifted using Eq. (35) towards
the neutral axis to consider the reserve strength beyond the initial yielding. A slight difference between the estimated results and
those of FEA could be attributed to the difference in the estimated reserve strength by shifting the location of point 1, and the FEA
beyond the initial yielding. On the other hand, the local collapse is observed in the cases of F3S2 and F3S3-series for larger plate
slenderness ratios. In case of local buckling collapse a good agreement is achieved.

524
M. Ozdemir et al. Marine Structures 59 (2018) 510–535

Fig. 23. Comparison of the ultimate strength estimation by ELDA with results of FEA for F3-series.

Fig. 24. Comparison of the ultimate strength estimation by ELDA with results of FEA for T3-series.

In the case of T3-series, mostly good agreement is observed except for the cases of few weak stiffeners with thick plate. The
possible reason for the difference in the estimated ultimate strengths and FEA results of T3S1-series is the same with that of F3S1-
series as explained above. When S2 and S3 size of stiffeners are provided with slender plates, the local plate buckling collapse is
observed, and the agreement between the results of the proposed method and FEA is excellent. In the case of T3S2 and T3S3-series
with lower plate slenderness ratios, the ultimate strength is very close to yield strength and the yielding examination at point 2
revealed very good estimations for those cases.

525
M. Ozdemir et al. Marine Structures 59 (2018) 510–535

Fig. 25. Comparison of the ultimate strength estimation by ELDA with results of FEA for F5-series.

Fig. 26. Comparison of the ultimate strength estimation by ELDA with results of FEA for T5-series.

In the case of VLCC panels, collapse mechanism is various, but a good agreement is achieved between the estimated and the
calculated results. When weak stiffeners are provided, the overall buckling collapse is observed. In those cases, point 1 is adopted for
the examination of the yielding condition. It is seen from the results of F5S1 and T5S1-series that shifting of the location of yielding
examination point revealed relatively good estimation of the ultimate strength. When heavy stiffeners are accompanied by thick
plates, the ultimate strength is usually estimated well. However, the horizontal bending of the stiffeners is slightly significant in the

526
M. Ozdemir et al. Marine Structures 59 (2018) 510–535

Fig. 27. Comparison of the ultimate strength estimations by: (a) ELDA, (b) PULS, (c) FY method with FEA.

case of T5S3-series with lower plate slenderness ratios, which is not considered in the present method. The ultimate strength is thus a
little overestimated.
All the estimated ultimate strengths are compared with the FEA results in Fig. 27(a). On the other hand, the estimated results by
PULS are also compared with the FEA results in Fig. 27(b). These figures show that proposed ELDA method estimates the ultimate
strength much better than the PULS code. Although the assumed deflection components of PULS are superior to those of proposed
ELDA method. PULS considers local buckling as the combination of simply supported and clamped modes of plate deflections, while
present ELDA method considers only simply supported modes of local plate buckling. Furthermore, sideway deflections of the stif-
feners are not taken into account in the present ELDA method, whereas PULS considers both sideway and vertical bending of the
stiffeners. Fundamental theory behind the PULS method is summarized by Ref. [4].
The ultimate strength is sometimes underestimated by the PULS, this may be attributed to the collapse criterion of PULS code,
which is based on minimum yield strength criteria; so the reserve strength beyond the initial yielding is not taken into account in
PULS. In F5S1N2B15 and F5S1N2B20 cases, PULS estimated ultimate strength of those panels almost half of the FEA. The collapse
pattern of those panels is Case 4 (overall buckling collapse). In those cases, initial yielding takes place at the stiffener's top, however
the panels can resist additional load owing to the thick plates. Estimated ηz values for F5S1N2B15 and F5S1N2B20 panels to consider
reserve strength beyond initial yielding are 0.078 and 0.156, respectively. In other words, yielding examination is performed near the
neutral axis of whole cross-section for those cases. The difference between FEA and ELDA estimations for F5S1N2B15 and F5S1N2B20
cases are %9.32 and %6.56 of FEA results, respectively.
Present method is also compared with other estimation methods, i.e., FY method [1,16,17], Zhang's formula [31] and a
simple formula given by Kim et al. [32]. The recommended application limits in Refs. [31] and [32] are λ < 1.0 and 0.5 ≤ λ < 5.0 ,
which do not cover the overall and local buckling collapse cases, respectively. Therefore, only the results of FY method are
provided here.
FY method provides good estimation when the collapse behavior is associated with local plate buckling or column buckling,
which are the specified cases in the FY method. On the other hand, it is obvious that FY method underestimates the ultimate
strength of the stiffened panels mostly for the overall buckling collapse cases. This is because FY method assumes that initial
yielding takes place before reaching Euler column buckling load of the stiffener with attached plating, which is the ultimate
strength condition. However, the ultimate strength becomes higher than Euler column buckling load in case of overall buckling
collapse. It should be noticed that FY method was developed for local buckling and column type buckling collapse cases, while the
overall buckling collapse case was out of consideration in FY method. The starting point of the present study is to improve accuracy
of the estimated results for local buckling and column type collapse cases as well as to cover wide range of collapse scenarios. To
conclude, it is clear to say that the present method can simulate wide range of collapse scenarios including overall collapse case
with high accuracy.

527
M. Ozdemir et al. Marine Structures 59 (2018) 510–535

5. Conclusions

A new simple and efficient method is proposed to estimate the ultimate strength of the stiffened panels by observing collapse me-
chanisms of the stiffened panels. Accordingly, at first, a series of detailed elastoplastic large deflection FEA is performed to investigate the
ultimate strength and collapse behavior of the stiffened panels under longitudinal thrust. In the analysis, the initial deflections are con-
sidered in the form of thin-horse mode plus overall buckling mode for the plates, and flexural buckling mode plus tripping mode for the
stiffeners. The welding residual stress is not taken into account. The results of present FEAs and the reference results are compared, and a
very good agreement is obtained. Observations obtained from the elastoplastic large deflection FEAs are summarized as follows.
It has been found that, for the same stiffener spacing, the number of stiffeners (and so aspect ratio of the whole stiffened panel)
has considerable effects if the collapse occurs due to the overall stiffened panel buckling, while the effect of number of stiffeners can
be negligible when local plate buckling dominates the collapse behavior. The latter is also the case when the ultimate strength is very
close to yield strength, which is the case of stockier plate is accompanied by heavy stiffeners.
The ultimate strength is reduced with the increase of the plate slenderness ratio when the local plate buckling collapse takes place. In
the case of thick plates with a large number of weak stiffeners, the ultimate strength is reduced as the plate thickness increases. This is
because the overall bending takes place with respect to the neutral axis of the whole cross-section, and an increase of the plate thickness
results in the larger distance between the stiffener's top and the neutral axis, which increases stress at the top of stiffener. In the case of thin
plates with a large number of weak stiffeners, the ultimate strength is attained by the occurrence of secondary buckling of overall mode.
To simulate elastic large deflection behavior of the stiffened plate in a simple manner, the analytical formulation is derived with
two deflection components, which are overall stiffened panel buckling and local plate buckling modes. The developed ELDA method
simulates elastic large deflection behavior of the stiffened panels well. However, when the aspect ratio of whole stiffened panel is
small, there is a clear difference in the post-buckling stiffness obtained by the analytical method and FEM in case of the overall
stiffened panel buckling. This could be explained by the development of deflection components other than the assumed buckling
modes in post-buckling range when FEA is performed.
Then, on the basis of the observed results, a new simple method, based on ELDA with the initial yielding concept, is proposed to
estimate the ultimate strength of stiffened panels under longitudinal thrust loads. The welding residual stresses are not taken into
account. The ultimate strength is estimated by examining yielding condition at two critical points in the ELDA model. The proposed
simple method estimates ultimate strength mostly well. In addition, the ultimate strength estimations obtained by applying PULS and
FY method are also compared with the FEA results. The proposed method shows much better agreement with the FEA results than
PULS and FY method.

Acknowledgements

The first author was supported by The Scientific and Technological Research Council of Turkey (TUBITAK) under 2214-A
International Doctoral Research Fellowship Program (1059B141500898), and it is gratefully acknowledged. Financial support of the
Japan Society of Naval Architects and Ocean Engineers (JASNAOE) to the international collaborative research for young researchers
is also gratefully acknowledged. This research was partially supported by the JSPS Grants-in-Aid for Scientific Research (A)
(15H02328) and (C) (15K06632).

Appendix 1. Buckling Stress Formulas and Parameters for Stiffened Panel under Longitudinal Thrust

• Overall elastic buckling stress for the stiffened panels:

1 Eπ 2 t 2 2 Eπ 2ep2 be
σcrov =
⎡ ⎛ p⎞ ⎛a + B⎞ + ⎛ ⎞
⎢ 2
1 + NAs /(Btp) ⎣ 12(1 − ν ) ⎝ B ⎠ ⎝ B a⎠ a2 ⎝ b ⎠
N N
2π 2 ⎧ EIy πyi C′ π 4Γ′ EIz π 2z s2 ⎞ πyi ⎫ ⎤
+ ∑ sin2 ⎛ ⎞ + ⎜⎛ + 2 2 + ⎟ ∑ cos2 ⎛ ⎞ ⎥.
Btp ⎨ a2 ⎝ B ⎠ ⎝ B 2 a B a2B2 ⎠ ⎝ B ⎠⎬ (A1.1)
⎩ i=1 i=1 ⎭⎦
Iy is the second moment of inertia of stiffener's cross-section without attached plate with respect to neutral axis of the whole cross-
section. Iz is the second moment of inertia of stiffener's cross-section for horizontal bending.

• Elastic local buckling stress for the plate [4]:

π 2E t 2 a 2
σcrpt = 2
⎛ p⎞
12(1 − ν ) ⎝ b ⎠ mb
+
mb
a { } ,
(A1.2)
where m is integer satisfying m (m − 1) ≤ a/ b ≤ m (m + 1) .

• Neutral axis of whole cross-section:

N
∑i = 1 Asi esi
ep = N
.
Ap + ∑i = 1 Asi (A1.3)

528
M. Ozdemir et al. Marine Structures 59 (2018) 510–535

• Effective breadth for plating [17]:

σp
be = b ⎧ (1 − αp) cr + αp ⎫.
⎨
⎩ σY ⎬
⎭ (A1.4)

• Effective thickness for flat-bar stiffeners [17]:

σp
te = tw ⎧ (1 − αs ) cr + αs ⎫.
⎨
⎩ σY ⎬
⎭ (A1.5)
When local buckling takes place first, stiffener cross-section parameters, Iy , Γ′ and C ′ are recalculated considering effective thickness
for flat-bar stiffeners.

• Elastoplastic buckling strength [1]:

2
σcrp p
1 ⎧ σecr p
⎛ σecr − 1⎞ + 0.25 ⎫,
= +1− ⎜ ⎟

σY 2 ⎨ σY ⎝ σY ⎠ ⎬
⎩ ⎭ (A1.6)
p
where σecr is elastic buckling stress of the local plate considering plate-stiffener interaction, and is calculated as in Ref. [47].

• Stiffness parameters beyond local buckling [4]:

1 + (a/ mb) 4
αp = .
3 + (a/ mb) 4 (A1.7)

1+ (32/45)(mπh/ a) 4
αs = .
1 + (8/45)(4 + 5η2)(mπh/ a) 4 (A1.8)
η is a parameter to get better agreement with FEA results.
3 3
h tp
η = 12 ⎧10 ⎛ ⎞ − 0.3⎫ + 0.324 ≤ 1.0.
⎜ ⎟

⎨ b ⎝ tw ⎠ ⎬ (A1.9)
⎩ ⎭

• Torsional and warping constants for the stiffeners:

- Flat-bar stiffeners:
G 3 E 3 3
C′ = ht w, Γ′ = h t w.
3 144 (A1.10)
- Tee-bar stiffeners:

G ⎧⎛ tf 1 tf 3 1 3 3⎫
C′ = h − ⎞ t w3 + bf t f3 ⎫ , Γ′ = E ⎧ ⎛h − ⎞ t w3 + bf t f .
3⎨⎩⎝ 2⎠ ⎬
⎭ ⎨ 36
⎩ ⎝ 2 ⎠ 144 ⎬
⎭ (A1.11)

Appendix 2. Parameters for ELDA

a11 = (α1 + α5)(3C 2 − C02) + α2 Amn

2 2
+ α3 + α4 + 2α6 CAmn + (β6 + α 7)(2CAmn − C0 A0mn ) + α8 (Amn − A02mn ) + α12 + αTS
+ (α10 − α13) σx . (A2.1)

a12 = α2 (2CAmn − C0 A0mn ) + α6 (C 2 − C02) + (β6 + α 7 )C2 + 2α8 CAmn + 2

α 9 (3Amn − A02mn ) + α11 σx . (A2.2)

a21 = (β2 + β6)(2CAmn − C0 A0mn ) + β4 (3C 2 − C02) + 2β5 CAmn + β7 Amn

2 2
+ β8 (Amn − A02mn ) + β10 σx . (A2.3)
2
a22 = (β1 + β9)(3Amn − A02mn ) + (β2 + β6) C 2 + β3 + βTS + β5 (C 2 − C02) + β7 (2CAmn − C0 A0mn ) + 2β8 CAmn + (β11 − β13) σx .
(A2.4)

b1 = −{(α1 + α5) C + α6 Amn }(C 2 − C02) − {(α2 + β6) Amn + α 7 C }(CAmn − C0 A0mn ) − (α3 + α4 + α12 + αTS )(C − C0 )
2
− (α8 C + α 9 Amn )(Amn − A02mn ) − {α11 Amn + (α10 − α13) C } σx . (A2.5)
2
b2 = −{(β1 + β9) Amn + β8 C }(Amn − A02mn ) − {(β3 + βTS )(Amn − Aomn ) − (β4 C + β5 Amn )(C 2 − C02)
− {(β2 + β6) C + β7 Amn }(CAmn − C0 A0mn ) − {β10 C + (β11 − β13) Amn } σx . (A2.6)

529
M. Ozdemir et al. Marine Structures 59 (2018) 510–535

c1 = −(α10 − α13) C − α11 Amn . (A2.7)

c2 = −β10 C − (β11 − β13) Amn . (A2.8)

1
α1 = aBtp (px1 epx1 + py1 epy1).
2 (A2.9)
1
α2 = aBtp (px 2 epx 3 + px 3 epx 4 + px 4 epx5 + px5 epx 6 + py2 epy3 + py3 epy 4 + py 4 epy5 + py5 epy6 + pxy1 epxy1 + pxy2 epxy2 + pxy3 epxy3
4
+ pxy 4 epxy 4 ). (A2.10)
1
α3 = aBtp (sax eax + say eay + saxy eaxy ).
4 (A2.11)
1
α4 = aBtp3 (bx1 ebx1 + by1 eby1 + bxy1 ebxy1).
48 (A2.12)
N 1
α5 = ∑ ⎧⎨aAsi ⎛si1 ei1 + si2 ei2 ⎞ ⎫.
i=1 ⎩ ⎝ 2 ⎠⎬⎭ (A2.13)
N
α6 = ∑ aAsi (si2 ei3 Sx1 + si2 ei4 Sx 2).
i=1 (A2.14)
N
α7 = ∑ aAsi (si3 ei2 Sx1 + si4 ei2 Sx 2).
i=1 (A2.15)
N
α8 = ∑ aAsi si5 ei1.
i=1 (A2.16)
N
α9 = ∑ aAsi (si6 ei3 Sx3 + si6 ei4 Sx 4).
i=1 (A2.17)
N
α10 = ∑ aAsi ei1.
i=1 (A2.18)
α11 = 0. (A2.19)
N 1
α12 = ∑ 2 (aIz1 si7 ei7).
i=1 (A2.20)

π2
α13 = Atotal .
4a (A2.21)
Atotal is area of whole cross-section

kt π 2 N πyi
αTS = ∑ cos2 ⎛ B ⎞.
2B i = 1 ⎝ ⎠ (A2.22)
1
β1 = aBtp (px 6 epx 7 + py6 epy7).
2 (A2.23)

β2 = α2. (A2.24)
1
β3 = aBtp3 (bx 2 ebx 2 + by2 eby2 + bxy2 ebxy2).
48 (A2.25)

β4 = α6. (A2.26)
N
β5 = ∑ aAsi si1 ei5.
i=1 (A2.27)
N
β6 = ∑
i=1
{ 12 aA (s e
si i3 i3 }
+ si 4 ei 4 ) .
(A2.28)

530
M. Ozdemir et al. Marine Structures 59 (2018) 510–535

N
β7 = ∑ aAsi (si3 ei6 Sx3 + si4 ei6 Sx 4).
i=1 (A2.29)

β8 = α 9. (A2.30)
N 1
β9 = ∑ ⎧⎨aAsi ⎛si5 ei5 + si6 ei6 ⎞ ⎫.
i=1 ⎩ ⎝ 2 ⎠⎬⎭ (A2.31)

β10 = 0. (A2.32)
N
β11 = ∑ aAsi ei5.
i=1 (A2.33)

β12 = 0. (A2.34)

m2π 2
β13 = Atotal .
4a (A2.35)

kt n2π 2N
βTS = .
2B (A2.36)
1
⎧2 m = 3,
Sx1 =
⎨ 0 m ≠ 3. (A2.37)
⎩
1
⎧ m = 1,
⎪ 2
Sx 2 = 2(m + 1)cos mπ
⎨ (2)
⎪ π (m − 1)(m + 3) m ≠ 1. (A2.38)
⎩
Sx3 = 0. (A2.39)
1
⎧ m = 1,
⎪ 2
Sx 4 = 2cos mπ
⎨ (2) {(m − 1)cos(mπ ) − 2m }
m ≠ 1.
⎪ π (m − 1)(3m + 1) (A2.40)
⎩
π 2E
px1 = .
8a2 (A2.41)

(1 − n)2 (m − n)2π 2Eα 2

px2 =
4{(1 − m)2 + (1 − n)2α 2} 2B2 (A2.42)

(1 − n)2 (m + n)2π 2Eα 2

px3 = .
4{(1 + m)2 + (1 − n)2α 2} 2B2 (A2.43)

(1 + n)2 (m + n)2π 2Eα 2

px4 = .
4{(1 − m)2 + (1 + n)2α 2} 2B2 (A2.44)

(1 + −n)2 (m n)2π 2Eα 2

px5 = .
4{(1 + m)2 + (1 + n)2α 2} 2B2 (A2.45)

m2π 2E
px6 = .
8a2 (A2.46)

π 2E
py1 = .
8B2 (A2.47)

(1 − m)2 (m − n)2π 2Eα 2

py2 = .
4{(1 − m)2 + (1 − n)2α 2} 2a2 (A2.48)

(1 + m)2 (m + n)2π 2Eα 2

py3 = .
4{(1 + m)2 + (1 − n)2α 2}2a2 (A2.49)

(1 − m)2 (m + n)2π 2Eα 2

py4 = .
4{(1 − m)2 + (1 + n)2α 2} 2a2 (A2.50)

531
M. Ozdemir et al. Marine Structures 59 (2018) 510–535

(1 + m)2 (m − n)2π 2Eα 2

py5 = .
4{(1 + m)2 + (1 + n)2α 2} 2a2 (A2.51)

n2π 2E
py6 = .
8B2 (A2.52)

(1 − m)(1 − n)(m − n)2π 2Eα 2

pxy1 = .
4{(1 − m)2 + (1 − n)2α 2} 2aB (A2.53)

(1 + m)(1 − n)(m + n)2π 2Eα 2

pxy2 = .
4{(1 + m)2 + (1 − n)2α 2} 2aB (A2.54)

(1 − m)(1 + n)(m + n)2π 2Eα 2

pxy3 = .
4{(1 − m)2 + (1 + n)2α 2} 2aB (A2.55)

(1 + m)(1 + n)(m − n)2π 2Eα 2

pxy4 = .
4{(1 + m)2 + (1 + n)2α 2}2aB (A2.56)

ep π 2E 1 ν
sax = − ⎛ + 2 ⎞.
(1 − ν 2) ⎝ a2 B ⎠ (A2.57)

ep π 2E ν 1
say = − ⎛ + 2 ⎞.
(1 − ν 2) ⎝ a2 B ⎠ (A2.58)

ep π 2E
saxy = .
(1 + ν ) aB (A2.59)

π 2E 1 ν
bx1 = ⎛ + 2 ⎞.
(1 − ν 2) ⎝ a2 B ⎠ (A2.60)

π 2E ⎛ m2 νn2
bx2 = + 2 ⎞.
⎜ ⎟

(1 − ν 2) ⎝ a2 B ⎠ (A2.61)

π 2E 1 ν
by1 = ⎛ + 2 ⎞.
(1 − ν 2) ⎝ B2 a ⎠ (A2.62)

π 2E ⎛ n2 νm2
by2 = + 2 ⎞.
⎜ ⎟

(1 − ν 2) ⎝ B2 a ⎠ (A2.63)

π 2E
bxy1 = − .
(1 + ν ) aB (A2.64)

mnπ 2E
bxy2 = − .
(1 + ν ) aB (A2.65)

π2
epx1 = .
4a2 (A2.66)

νπ 2
epx2 = .
4B2 (A2.67)

(m − n)2π 2α 2 2 2
⎧ (1 − n) − ν (1 − m) ⎫ .
epx3 =
4{(1 − m)2 + (1 − n)2α 2} 2 ⎨
⎩ B 2 a2 ⎬
⎭ (A2.68)

(m + n)2π 2α 2
⎧ (1 − n)2 ν (1 + m)2 ⎫.
epx4 = −
4{(1 + m)2 + (1 − n)2α 2}2 ⎨
⎩ B
2 a2 ⎬
⎭ (A2.69)

(m + n)2π 2α 2 2 2
⎧ (1 + n) − ν (1 − m) ⎫ .
epx5 =
4{(1 − m)2 + (1 + n)2α 2} 2 ⎨
⎩ B 2 a2 ⎬
⎭ (A2.70)

(m − n)2π 2α 2
⎧ (1 + n)2 ν (1 + m)2 ⎫.
epx6 = −
4{(1 + m)2 + (1 + n)2α 2} 2 ⎨
⎩ B
2 a2 ⎬
⎭ (A2.71)

m2π 2
epx7 = .
4a2 (A2.72)

532
M. Ozdemir et al. Marine Structures 59 (2018) 510–535

νn2π 2
epx8 = .
4B2 (A2.73)

π2
epy1 = .
4B2 (A2.74)

νπ 2
epy2 = .
4a2 (A2.75)

(m − n)2π 2α 2 2 2
⎧ (1 − m) − ν (1 − n) ⎫ .
epy3 =
4{(1 − m)2 + (1 − n)2α 2} 2 ⎨
⎩ a2 B 2 ⎬
⎭ (A2.76)

(m + n)2π 2α 2 2 2
⎧ (1 + m) − ν (1 − n) ⎫ .
epy4 = 2 2 2 2 2 2
4{(1 + m) + (1 − n) α } ⎨ ⎩ a B ⎬
⎭ (A2.77)

(m + n)2π 2α 2 2 2
⎧ (1 − m) − ν (1 + n) ⎫ .
epy5 =
4{(1 − m)2 + (1 + n)2α 2} 2 ⎨
⎩ a2 B 2 ⎬
⎭ (A2.78)

(m − n)2π 2α 2 2 2
⎧ (1 + m) − ν (1 + n) ⎫ .
epy6 = 2 2 2 2 2 2
4{(1 + m) + (1 + n) α } ⎩ ⎨ a B ⎬
⎭ (A2.79)

n2π 2
epy7 = .
4B2 (A2.80)

νm2π 2
epy8 = .
4a2 (A2.81)

(1 − m)(1 − n)(m − n)2π 2α 2

epxy1 = 2(1 + ν ) .
4{(1 − m)2 + (1 − n)2α 2} 2aB (A2.82)

(1 + m)(1 − n)(m + n)2π 2α 2

epxy2 = 2(1 + ν ) .
4{(1 + m)2 + (1 − n)2α 2}2aB (A2.83)

(1 − m)(1 + n)(m + n)2π 2α 2

epxy3 = 2(1 + ν ) .
4{(1 − m)2 + (1 + n)2α 2} 2aB (A2.84)

(1 + m)(1 + n)(m − n)2π 2α 2

epxy4 = 2(1 + ν ) .
4{(1 + m)2 + (1 + n)2α 2} 2aB (A2.85)

π2
eax = −ep .
a2 (A2.86)

π2
eay = −ep .
B2 (A2.87)

π2
eaxy = 2ep .
aB (A2.88)

π2
ebx1 = .
a2 (A2.89)

m2π 2
ebx2 = .
a2 (A2.90)

π2
eby1 = .
B2 (A2.91)

n2π 2
eby2 = .
B2 (A2.92)

2π 2
ebxy1 = − .
aB (A2.93)

2mnπ 2
ebxy2 = − .
aB (A2.94)

π 2E 2πyi
si1 = cos .
8a2 B (A2.95)

533
M. Ozdemir et al. Marine Structures 59 (2018) 510–535

νπ 2E
si2 = .
8B2 (A2.96)

si3 =
(m − n)2π 2Eα2
4{(1 − m)2 + (1 − n)2α2}
2 { (1 − n)2
B2
−
ν (1 − m)2
a2 } cos (1 − n) πyi
B

+
(m + n)2π 2Eα2
4{(1 − m)2 + (1 + n)2α2}
2 { (1 + n)2
B2
−
ν (1 − m)2
a2 } cos (1 + n) πyi
B
.
(A2.97)

si 4 =
(m + n)2π 2Eα2
4{(1 + m)2 + (1 − n)2α2}
2 { (1 − n)2
B2
−
ν (1 + m)2
a2 } cos (1 − n) πyi
B

+
(m − n)2π 2Eα2
4{(1 + m)2 + (1 + n)2α2}
2 { (1 + n)2
B2
−
ν (1 + m)2
a2 } cos (1 + n) πyi
B
.
(A2.98)

m2π 2E 2nπyi
si5 = cos .
8a2 B (A2.99)

νn2π 2E
si6 = .
8B2 (A2.100)

π 2E πyi
si7 = sin .
a2 B (A2.101)
si8 = 0. (A2.102)
s
ei1 = 2 i1 .
E (A2.103)
s
ei2 = 2 i2 .
E (A2.104)
s
ei3 = i3 .
E (A2.105)
s
ei 4 = i 4 .
E (A2.106)
si5
ei5 = 2 .
E (A2.107)
si6
ei6 = 2 .
E (A2.108)
si7
ei7 = .
E (A2.109)
si8
ei8 = = 0.
E (A2.110)

• Flat-bar stiffeners:
Iz1 = tw h { 14 t + 12 t h + 13 h − e (t + h) + e }.
2
p p
2
p p
2
p
(A2.111a)

• Tee-bar stiffeners:
{t+
Iz1 = tw (h − t f )
1 2
4 p
1
t (h
2 p
1
− t f ) + 3 (h − t f )2 − ep (tp + h − t f ) + ep2 }
2
+ b t { ( t + h) − ( ) ( ) }.
1 1 1 2 1
f f 2 p
t
2 p
+ h tf + t
3 f
− 2ep 2 tp + h + ep t f + ep2
(A2.111b)

References

[1] Tanaka S, Yanagihara D, Yasuoka A, Harada M, Okazawa S, Fujikubo M, et al. Evaluation of ultimate strength of stiffened panels under longitudinal thrust. Mar
Struct 2014;36:21–50.
[2] Yao T, Astrup OC, Caridis P, Chen YN, Cho S-R, Dow RS, et al. Ultimate hull girder strength, report of STC IV.2. Proceedings of 14th International Ship and
Offshore Structures Congress (ISSC), 2. 2000. p. 321–91.
[3] The Japan Society of Naval Architects and Ocean Engineers. Report of research committee for verification of ISO formulas to evaluate ultimate strength. 2011.
[4] Yao T, Fujikubo M. Buckling and ultimate strength of ship and ship-like floating structures. first ed. Butterworth-Heinemann, Elsevier; 2016.
[5] Yao T. Hull girder strength. Mar Struct 2003;16:1–13.
[6] Tayyar GT, Nam J-M, Choung J. Prediction of hull girder moment-carrying capacity using kinematic displacement theory. Mar Struct 2014;39:157–73.

534
M. Ozdemir et al. Marine Structures 59 (2018) 510–535

[7] Olmez H, Bayraktarkatal E. Effects of key factors on hull girder ultimate strength estimation by progressive collapse calculations. Lat Am J Solid Struct
2016;13:2371–92.
[8] Olmez H, Bayraktarkatal E. Maximum load carrying capacity estimation of the ship and offshore structures by progressive collapse approach. Pol Marit Res
2016;23:28–38.
[9] Yao T, Fujikubo M, Yanagihara D. On loading and boundary conditions for buckling/plastic collapse analysis of continuous stiffened plate by FEM. Proceedings of
12th Asian Pacific technical exchange and advisory meeting on marine structures (TEAM98). 1998. p. 305–14.
[10] Yao T, Fujikubo M, Yangihara D, Varghese B, Niho O. Influences of welding imperfections on buckling/ultimate strength of ship bottom plating subjected to
combined bi-axial thrust and lateral pressure. Proc. Int. Symp. Singapore: Thin-Walled Structures; 1998. p. pp.425–432.
[11] Paik JK, Seo JK. Nonlinear finite element method models for ultimate strength analysis of steel stiffened-plate structures under combined biaxial compression
and lateral pressure actions–Part II: stiffened panels. Thin-Walled Struct 2009;47:998–1007.
[12] Xu MC, Yanagihara D, Fujikubo M, Soares CG. Influence of boundary conditions on the collapse behaviour of stiffened panels under combined loads. Mar Struct
2013;34:205–25.
[13] Estefen SF, Chujutalli JH, Soares CG. Influence of geometric imperfections on the ultimate strength of the double bottom of a Suezmax tanker. Eng Struct
2016;127:287–303.
[14] Tekgoz M, Garbatov Y, Soares CG. Ultimate strength assessment of welded stiffened plates. Eng Struct 2015;84:325–39.
[15] Xu MC, Song ZJ, Pan J, Soares CG. Ultimate strength assessment of continuous stiffened panels under combined longitudinal compressive load and lateral
pressure. Ocean Eng 2017;139:39–53.
[16] Fujikubo M, Yanagiara D, Yao T. Estimation of ultimate strength of continuous stiffened plates under thrust. J Soc Nav Archit Jpn 1999;185:203–14. [in
Japanese].
[17] Fujikubo M, Yanagiara D, Yao T. Estimation of ultimate strength of continuous stiffened plates under thrust (2nd Report). J Soc Nav Archit Jpn 1999;186:631–8.
[in Japanese].
[18] Paik JK, Thayamballi AK, Kim DH. An analytical method for the ultimate compressive strength and effective plating of stiffened panels. J Constr Steel Res
1999;49:43–68.
[19] Paik JK, Thayamballi AK, Kim BJ. Large deflection orthotropic plate approach to develop ultimate strength formulations for stiffened panels under combined
biaxial compression/tension and lateral pressure. Thin-Walled Struct 2001;39:215–46.
[20] Det Norske Veritas. Nauticus Hull user manual. 2005. PULS.
[21] Byklum E, Amdahl J. A simplified method for elastic large deflection analysis of plates and stiffened panels due to local buckling. Thin-Walled Struct
2002;40:925–53.
[22] Byklum E, Steen A, Amdahl J. A semi-analytical model for global buckling and post buckling analysis of stiffened panels. Thin-Walled Struct 2004;42:701–17.
[23] Steen A, Byklum E. Ultimate strength and post buckling stiffness of plate panels subjected to combined loads using semi-analytical models. Proceedings of
international conference on marine research and transportation (ICMRT). 2005. p. 1–8.
[24] Fujikubo M, Yao T, Khedmati MR, Harada M, Yanagihara D. Estimation of ultimate strength of continuous stiffened panel under combined transverse thrust and
lateral pressure, Part 1: continuous plate. Mar Struct 2005;18:383–410.
[25] Fujikubo M, Harada M, Yao T, Khedmati MR, Yanagihara D. Estimation of ultimate strength of continuous stiffened panel under combined transverse thrust and
lateral pressure, Part 2: continuous stiffened panel. Mar Struct 2005;18:411–27.
[26] Ozguc O, Das PK, Baltrop N. The new simple design equations for the ultimate compressive strength of imperfect stiffened plates. Ocean Eng 2007;34:970–86.
[27] Brubak L, Hellesland J. Semi-analytical postbuckling and strength analysis of arbitrary stiffened plates in local and global bending. Thin-Walled Struct
2007;45:620–33.
[28] Brubak L, Hellesland J. Strength criteria in semi-analytical, large deflection analysis of stiffened plates in local and global bending. Thin-Walled Struct
2008;46:1382–90.
[29] Masaoka K, Mansour A. Compressive strength of stiffened plates with imperfections: simple design equations. J Ship Res 2008;52:227–37.
[30] Zhang S, Khan I. Buckling and ultimate capability of plates and stiffened panels in axial compression. Mar Struct 2009;22:791–808.
[31] Zhang S. A review and study on ultimate strength of steel plates and stiffened panels in axial compression. Ships Offshore Struct 2016;11:81–91.
[32] Kim DK, Lim HL, Kim MS, Hwang OJ, Park KS. An empirical formulation for predicting the ultimate strength of stiffened panels subjected to longitudinal
compression. Ocean Eng 2017;140:270–80.
[33] Ueda Y, Rashed SMH. The idealized structural unit method and its application to deep girder structures. Comput Struct 1984;18:277–93.
[34] Ueda Y, Rashed SMH. ISUM (idealized structural unit method) applied to marine structures. Trans JWRI 1991;20:123–36.
[35] Ueda Y, Rashed SMH. Modern method of ultimate strength analysis of offshore structures. Int J Offshore Polar 1991;1:27–41.
[36] Ueda Y, Rashed SMH, Nasser AY. An improved ISUM rectangular plate element taking account of post-ultimate strength behavior. Mar Struct 1993;6:139–72.
[37] Fujikubo M, Kaeding P. New simplified approach to collapse analysis of stiffened plates. Mar Struct 2002;15:251–83.
[38] Harada M. Practical method to estimate ultimate strength of members in ship structure [Dr Thesis]. Hiroshima University; 2004. [in Japanese].
[39] Harada M, Fujikubo M, Yanagihara D. Development of simple formulae of ultimate strength of continuous stiffened plates under combined thrust and lateral
pressure. J Kansai Soc Nav Archit Jpn 2004;241:159–68. [in Japanese].
[40] Harada M, Fujikubo M, Yanagihara D. Estimation of ultimate strength of continuous stiffened plate under combined biaxial thrust and lateral pressure. J Soc Nav
Archit Jpn 2004;196:189–98. [in Japanese].
[41] Harada M, Fujikubo M, Yanagihara D. Development of a set of simple formulae for estimation of ultimate strength of a continuous stiffened panel under
combined loads. J Jpn Soc Nav Archit Ocean Eng 2005;2:387–95. [in Japanese].
[42] Harada M, Fujikubo M, Yanagihara D. Development of a set of closed-form formulae for estimation of ultimate strength of a continuous stiffened panel under
combined in-plane loads and lateral pressure. Class NK Tech Bull 2007;25:11–21.
[43] IACS. Common structural rules for bulk carriers. 2006.
[44] IACS. Common structural rules for double hull oil tankers. 2006.
[45] IACS. Harmonized common structural rules. 2015.
[46] ANSYS 14.5, user’s manual.
[47] Fujikubo M, Yao T. Elastic local buckling strength of stiffened plate considering plate/stiffener interaction and welding residual stress. Mar Struct
1999;12:543–64.

535