Thin–Walled Structures 192 (2023) 111201

Contents lists available at ScienceDirect

Thin-Walled Structures
journal homepage: www.elsevier.com/locate/tws

Full length article

Corrected similarity method for designing scaled-down stiffened plate

subjected to compression load with improved processing feasibility
Zhuo Wang a, b, Xiangshao Kong a, *, Weiguo Wu a
a
Green & Smart River-Sea-Going Ship, Cruise and Yacht Research Center, Wuhan University of Technology, Wuhan, 430063, PR China
b
School of Naval Architecture, Ocean and Energy Power Engineering, Wuhan University of Technology, Wuhan, 430063, PR China

A R T I C L E I N F O A B S T R A C T

Keywords: This paper proposes a corrected similarity method for designing scaled-down stiffened plates subjected to
Stiffened plates compression loads that carefully considers conflict between the minimum welding distance and stiffener spacing
Similarity method by introducing a stiffener layout adjustment approach. Reliable plate scenarios were adopted in a numerical
Scaled-down model
parametric study to investigate the effects of different influential parameters on stiffener layout adjustment
Numerical study
design, and an empirical formula and associated design procedure for the stiffener layout adjustment approach
Stiffener layout adjustment
were proposed. The proposed approach can help engineers obtain scaled-down stiffened plates that are closer to
ideal conditions and can improve the precision of experimental studies.

1. Introduction The acquisition of precise outcomes in a scaled-down model test

necessitates the initial development of a reliable scaled-down model. To
Stiffened plates constitute indispensable and fundamental structural date, various similarity design methods have been proposed and
elements in ships. Ship safety is intricately associated with the loading extensively employed in the experimental investigations of ship struc­
capacities and collapse behaviours of stiffened plates, which have tures. The complete geometric-similarity method, which is based on
attracted research attention for many decades. Numerous theoretical dimensional analysis, is the most commonly employed similarity
[1–3] and numerical [4–6] investigations have been conducted to method for ship structures [13]. This approach is user-friendly and has a
evaluate the loading capacities and buckling behaviours of stiffened substantial positive impact on design similarity. However, the complete
plates. These theoretical and numerical techniques are robust and have geometrical similarity method requires the adoption of the same scale
significant potential for guiding practical structural designs and ratio in all dimensions of the scaled-down model, which is typically
ensuring structural safety [7]. Nevertheless, given the structural impractical for thin-walled structures such as stiffened plates, owing to
complexity and safety requirements of ship structures, experimental the constraints of the steel-plate thickness. The distortional similarity
methods are essential for assessing their ultimate loading capacities. method is an improved version of the complete geometrical similarity
Conducting full-scale experiments is the most reliable approach for method [14]. By incorporating dimensionless parameters into the design
assessing the ultimate loading capacity of ship structures. However, high process, the distortional similarity method enables scaled-down models
costs and environmental constraints impede the ability to conduct suc­ to adopt different scale ratios for length and thickness. However, the
cessive full-scale tests [8–10]. In contrast, performing scaled-down model distortional similarity method fails to provide a satisfactory level of
tests is a less complex and cost-effective approach; thus, researchers in the similarity between the prototype and scaled-down models in the struc­
field more often use scaled-down model tests than full-scale experiments tural buckling process, because of the possibility that the buckling
[11]. The fundamental concept of a scaled-down model test is to utilise the modes are altered by the scaling of the stiffened plates [15]. Given the
principle of similarity to convert the outcomes of such a test into realistic limitations of these approaches, some researchers have focused on the
structural outcomes, thereby realising economical safety assessment [12]. critical-parameter-based similarity method [15–18]. The underlying
This approach has been empirically validated in several studies as principle of this method is that the loading capacities and buckling be­
appropriate for engineering design and evaluation objectives. haviours of stiffened plates can be comprehensively described using a set

Abbreviations: FE, finite element; HAZ, heat-affected zone; ISSC, International Ship and Offshore Structures Congress; MSE, mean squared error.
* Corresponding author.
E-mail address: [email protected] (X. Kong).

https://doi.org/10.1016/j.tws.2023.111201
Received 26 April 2023; Received in revised form 4 September 2023; Accepted 13 September 2023
Available online 20 September 2023
0263-8231/© 2023 Elsevier Ltd. All rights reserved.
Z. Wang et al. Thin-Walled Structures 192 (2023) 111201

Fig. 1. Manufacturing limitations of stiffened plate.

of critical parameters. The effectiveness of this method in predicting the adjusted model with different stiffener layouts is further scaled down
entire buckling process of a stiffened plate has been demonstrated in using the previously proposed similarity criterion [18]. Finally, a
experimental studies [15,16]. The challenge associated with this scaled-down model with different stiffener layouts and improved pro­
approach pertains to the identification of dimensionless parameters that cessing quality is obtained. The corrected similarity method in this study
can comprehensively characterise the buckling behaviour of stiffened holds promise as a practical tool for use in experimental studies because
plates. Currently, the determination of these parameters relies pre­ it facilitates the welding process in scaled-down models and enhances
dominantly on the expertise and experience of the designer. the precision of model tests. This approach may serve as an important
Notably, the aforementioned similarity methods are fundamentally complement to the existing methodologies used in experimental studies
grounded on the premise that the prototype and scaled-down models of ship structures.
exhibit the same geometric configurations, particularly with regard to
the number of stiffeners. However, stringent adherence to this premise 2. Background
in the similarity-based design of most ship structures may be considered
impractical, especially when a large scale ratio is applied, and could 2.1. Similarity method of stiffened plate
undermine the welding processing feasibility of the model [19]. In
various practical instances, it is imperative for the designer to decrease In a previous study [18], a universal similarity design method for the
the number of stiffeners and simultaneously to augment the precise buckling assessment of stiffened plates under compression was
inter-stiffener distance within scaled-down models. Otherwise, insuffi­ proposed by utilising a critical-parameter-based method, as shown in
cient space between the inter-stiffeners may result in inadequate weld Eq. (1):
integrity of the scaled-down models, eventually adversely affecting the √̅̅̅̅̅ √̅̅̅̅̅ √̅̅̅̅̅
accuracy and reliability of the test results. Hence, an approach that can A σY b σy a σy
π1 = , π2 = , π3 =
adjust the stiffener layout of scaled-down models while preserving their bt E t E r E
, (1)
ultimate strengths and buckling behaviours must be proposed, thereby I r4 Izf
π4 = 3 , π5 = , π6 =
improving the welding feasibility and quality of the test scaled-down bt I I
models.
In previous studies, the equivalent sectional area-based method was where E is the Young’s modulus of the material; A is the total cross-
widely employed by researchers to facilitate the layout adjustment of sectional area of the stiffened plate; a, b, and t are the length, breadth,
stiffeners in stiffened plates [10]. This method entails maintaining and thickness of the panel; σ y is the yield stress of material; r and I are
identical total sectional areas of stiffeners in both the initial and adjusted radius of inertia and inertia moment of stiffened panel; and Izf is the
models, thereby guaranteeing consistent compressive structural rigidity. moment of inertia of stiffener about the vertical axis.
Despite the satisfactory approximation of the loading capacity in the Thus, the scaled-down model from the similarity method displays
elastic regime of the material, the prescribed method can produce behaviour similar to that of real ship structures in terms of ultimate
relatively large deviations in the ultimate strength and buckling process strengths and buckling modes, thereby enabling experimental predic­
[15]. These deviations may arise from the inherent limitations of the tion of the buckling behaviour of real ship structures under longitudinal
method related to accounting for geometric and material nonlinearities, compression loads. The underlying assumption of the similarity method
which can become increasingly pronounced beyond the elastic regime of is that the stiffener layout and number of stiffeners of the scaled-down
the material. In addition, researchers have attempted to integrate sup­ model are identical to those of the prototype. However, this situation
plementary critical parameters into a procedure for modifying the may lead to insufficient processing feasibility of the scaled-down model,
layout of stiffened plates [20]. In summary, despite considerable ad­ affecting the final results of the model tests. The manufacturing limita­
vancements in experimental technology for ship structures, reliable tions of a scaled-down model should be further considered in the simi­
approaches for adjusting the stiffeners of stiffened plates are lacking. larity method to reduce the disturbance effect of the processing factors
This deficiency could significantly affect the processing quality of the in experimental studies.
scaled-down models, thereby compromising the accuracy of the test
results. 2.2. Manufacturing limitations of stiffened plates
The objective of this study was to develop a corrected similarity
method for designing a scaled-down stiffened plate under a compressive The accurate prediction of the loading capacity and buckling mode of
load. Specifically, a stiffener layout adjustment approach was proposed stiffened plates in experimental tests depends significantly on the
and introduced into a previous critical-parameter-based similarity manufacturing quality of the scaled-down model. Strict adherence to a
method. In the corrected method, the stiffener layout adjustment pro­ reasonable welding procedure throughout the manufacturing process of
cedure is firstly applied to the prototype model. Then, the obtained the scaled-down model is a primary requirement to minimise the

2
Z. Wang et al. Thin-Walled Structures 192 (2023) 111201

Fig. 2. Stiffener layout adjustment design.

influence of external factors on the test results. To satisfy this require­ following sections.
ment, the scaled-down model must be appropriately designed and
technically feasible for the intended welding application.
The minimum spacing of the stiffeners (i.e. the breadth of panel b) is 2.3. Existing approach for stiffener layout adjustment
a critical manufacturing constraint that significantly affects the welding
quality of the stiffened plates. In instances in which adjacent stiffeners 2.3.1. Equivalent cross-sectional area-based method
are placed in close proximity, the welding heat-affected zones (HAZs) of In the present article, a stiffened plate that has been modified by
these stiffeners may overlap, which could result in repeated welding of reconfiguring the stiffener layout is referred to as an ‘adjusted model’, in
the material in that area and adversely affect the material properties and relation to the ‘original model’. One of the most commonly used ap­
loading capacity of the stiffened plates [21,22]. Research has indicated proaches for adjusting the stiffener layout is the equivalent sectional
that the width of the HAZ located away from the welding fusion area area-based method, which is widely applied in the practical similarity
ranges from 20 to 50 mm, depending on the selected material and design of ship structures [10]. This method is based on the fundamental
welding parameters [23]. For thin-walled structures, particularly stiff­ concept of maintaining the same total sectional area of the stiffeners for
ened plates, pertinent specifications [24] also suggest that the spacing of both the original model AOts and the adjusted model AAts . Specifically, to
stiffeners must be maintained greater than 100 mm to guarantee con­ increase the stiffener spacing distance b of the stiffened plate, the
formity with the fundamental welding prerequisites. It should be noted number of stiffeners must be decreased while simultaneously increasing
that this specified limit is merely the starting point. In structural models their cross-sectional sizes. The equivalent cross-sectional area-based
in which thinner plates are used, the minimum stiffener spacing may method can be mathematically represented by the following formula:
need to be increased further to ensure the manufacturing quality of the ⎧ O A O A O A O A
models, which can pose a challenge for most scaled-down stiffened ⎨ L = L , B = B , a = a , tp = tp
⎪
plates, as shown in Fig. 1. nO + 1 O A , (2)
⎪
⎩ bA = A b , Ats = AOts
A possible way to address this issue is to modify the stiffener layout n +1
of the scaled-down stiffened plates while maintaining the ultimate
loading capacity and buckling behaviour of the structure, that is, the where superscripts O and A correspond to the original and adjusted
stiffener layout adjustment design, as shown in Fig. 2. Therefore, a models, respectively; L and B are the length and breadth of the stiffened
corrected similarity method that considers stiffener layout adjustments plate, respectively; a and b are the length and breadth of the panel,
in the stiffened plates of ship structures must be developed. Several respectively; tp is the panel thickness; n is the number of stiffeners of the
stiffener layout adjustment methods are introduced and discussed in the stiffened plate; and Ats is the total stiffener sectional area of the stiffened
plate.

Fig. 3. Schematic view of original stiffened plate in the parametric investigation.

3
Z. Wang et al. Thin-Walled Structures 192 (2023) 111201

Table 1
Stiffened plate scenarios in parametric study.
Plate Stiffener

tp (mm) Layout Cross-sectional dimension of stiffener

n0 — n1 Stiffener type: hw × tw × hf × tf (mm)
n0: before adjustment
n1: after adjustment

(1) 6.0 (1) 12 — 10 (1) T-bar: 80.0 × 5.0 × 25.0 × 7.0

(2) 8.0 (2) 12 — 8 (2) T-bar: 90.0 × 6.0 × 30.0 × 8.0
(3) 10.0 (3) 12 — 6 (3) T-bar: 100.0 × 6.0 × 24.0 × 9.0
(4) 12.0 (4) 8 — 6 (4) T-bar: 112.0 × 7.0 × 25.0 × 9.0
(5) 15.0 (5) 8 — 4 (5) T-bar: 125.0 × 7.0 × 25.0 × 10.0
(6) 20.0 (6) 4 — 2 (6) T-bar: 140.0 × 8.0 × 30.0 × 12.0
(7) 2 — 1 (7) T-bar: 160.0 × 8.0 × 30.0 × 15.5
(8) T-bar: 180.0 × 9.0 × 40.0 × 16.0

*Note: for all models, a = 2900.0 mm, b = 700.0 mm, σ y = 235 MPa.

Fig. 5. Mesh size convergence analysis.

Fig. 4. FE model of stiffened plate.

Table 2
Boundary conditions of the stiffened plates in FE analysis [7].
Location Translational Rotational

ux uy uz θx θy θz

A-B Fixed — Fixed Fixed Fixed Fixed

C-D Uniform displacement — Fixed Fixed Fixed Fixed
A–C and B-D — — Fixed — Fixed Fixed
E-F and G-H — — Fixed Fixed — — Fig. 6. Validation results of FE modelling technique.

As stated in Eq. (2), the transformation relationship pertaining to the b

√̅̅̅̅̅
σY
cross-sectional dimensions of the stiffeners can be expressed as β= (4)
tp E
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅
⎧ √̅̅̅̅̅
⎪ A nO + 1 O A nO + 1 O a σY
⎪ h = h ,t = t
⎪
⎨ w nA + 1 w w nA + 1 w λ= (5)
(3) πr E
⎪ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅ ,
⎪
⎪ nO + 1 O A nO + 1 O
⎩ A EI 12(1 − v2 )I
hf = h ,t = t
nA + 1 f f nA + 1 f γ= = , (6)
Db btp3
where hw denotes the height of the web of the stiffener, tw is the thickness
where σ y is the yield stress of the material, E is the Young’s modulus of
of the web of the stiffener, hf is the width of the flange of the stiffener,
the material, r is the radius of inertia of the stiffened panel, I is the
and tf is the thickness of the flange of the stiffener.
moment of inertia of the stiffened panel, and D is the bending rigidity of
The equivalent sectional area-based method is user-friendly and has
the associated panel.
a good effect on the linear deformation stage of a stiffened plate under a
In recent years, increasing attention has been paid to the similarity
compression load, because of its ability to maintain the structural stiff­
design of stiffened plates with regard to the aforementioned critical
ness of the stiffened plate under different stiffener layouts. However, the
parameters, that is, the dimensionless parameter-based method. The
precision of this method may significantly diminish as the stiffened plate
underlying principle is that the entire buckling process and ultimate
transitions into the buckling and collapse stages.
loading capacity of a stiffened plate can be comprehensively described
by a group of suitable critical parameters. This characteristic implies
2.3.2. Dimensionless parameter-based method
that distinct stiffened plates with identical critical parameters exhibit
The ultimate strengths of stiffened plates are known to be consid­
similar buckling modes and loading characteristics, which can be further
erably influenced by the panel slenderness ratio β and the column
employed in the similarity design of stiffened plates. The effectiveness of
slenderness ratio of stiffener λ [25,26]. Furthermore, the stiffness ratio
this approach, as well as the critical parameters mentioned above, in the
of the stiffener to the panel γ is recognised as a critical parameter that is
context of the similarity design of stiffened plates under compression
associated with the buckling mode of the stiffened plate [27]. The for­
loads, has been validated through a combination of numerical and
mulas for calculating β, λ, and γ are as follows:

4
Z. Wang et al. Thin-Walled Structures 192 (2023) 111201

Fig. 7. Numerical results of existing methods for stiffener layout adjustment.

experimental investigations [28,29]. Further investigation is required to Strategy 3. Maintain a consistent panel slenderness ratio β and column
evaluate the possibility of utilising a dimensionless parameter-based slenderness ratio λ between the original and adjusted stiffened plates.
method for stiffener layout adjustment design of stiffened plates. These strategies can be expressed using the following formulas:
⎧ LA = LO , BA = BO , aA = aO forallstrategies
⎪
⎪
⎪
⎪
2.4. Research strategy ⎪
⎪
⎪
⎪ bA = A
nO + 1 O A nO + 1 O A
b , tp = A
nO + 1 O
tp , Dstiffner = A D forβA = βO , γA = γO
⎪
⎪ n +1 n +1 n + 1 stiffener
⎪
⎪
⎪ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅
The fundamental principle underlying the stiffener layout adjust­ ⎪
⎨
nO + 1 O A A
3 n + 1
ment design involves achieving a congruent length L and breadth B of ⎪ bA
= b , tp = tO , DA = DOstiffener forλA = λO , γA = γ O ,
⎪ A
n +1 n + 1 p stiffner
O
the stiffened plates between the original and adjusted models. This ⎪
⎪
⎪
⎪
⎪
congruence serves as the basic premise for the inversion of the test ⎪
⎪
⎪ A nO + 1 O A nO + 1 O A O A O A O
outcomes observed between the original and adjusted stiffened plates, as ⎪ b = nA + 1 b , tp = nA + 1 tp , Dstiffner = Dstiffener forβ = β , λ = λ
⎪
⎪
⎪
⎩
shown in Fig. 2. However, maintaining identical values of all three Dstiffener = hw ortw orhf ortf
critical parameters, namely, β, λ, and γ, between the original and (7)
adjusted models is usually infeasible due to the inherent trade-offs of
different design objectives. Thus, a compromise strategy would entail where tw and hw are the thickness and height of the web of stiffeners,
maintaining consistency in the first two critical parameters while respectively; tf, and hf are the thickness and width of the flange of
aligning the third parameter as closely as possible with its counterparts stiffeners, respectively.
in both the original and adjusted models. The final research strategy can
Section 3 provides a comparative examination of the three design
be summarised as follows:
strategies in conjunction with the cross-sectional area-based method to
Strategy 1. Maintain a consistent panel slenderness ratio β and stiff­ ascertain the effectiveness of these methods in adjusting the stiffener
ness ratio of the stiffener to panel γ between the original and adjusted layout of the stiffened plates.
stiffened plates.

Strategy 2. Maintain a consistent column slenderness ratio λ and

stiffness ratio of the stiffener to panel γ between the original and
adjusted stiffened plates.

5
Z. Wang et al. Thin-Walled Structures 192 (2023) 111201

Fig. 7. (continued).

3. Parametric study ultimate loading capacity and buckling process of a stiffened plate, as
detailed in [30]. The material in the FE analysis was assumed to be
3.1. Stiffened plate scenarios elastic-perfectly plastic, neglecting the effect of strain hardening [31].
The material properties were characterised by a Young’s modulus of
In this study, we investigated a series of stiffened plates extracted 206 GPa and Poisson’s ratio of 0.3. The FE model of the stiffened plate
from a large cruise ship. Specifically, we performed a comparative was discretised using a four-node shell element (S4R) with five inte­
analysis of the stiffened plates and their respective adjusted models from gration points in the thickness direction. To account for the effect of the
the foregoing equivalent cross-sectional area-based method and boundary region between adjacent bays, a one-span/two-bay model
dimensionless parameter-based strategy. In the comparative study, the (1/2 + 1 + 1/2) was employed. An axial compressive load was applied
parametric factors considered included the panel thickness tp, stiffener to one end of the stiffened plate in the form of displacement. Fig. 4 and
number ratio n0/n1 (n0 and n1 represent the number of stiffeners in the Table 2 provide a detailed description of the FE model of the stiffened
original and adjusted models, respectively), and stiffener cross-sectional plate, as well as the boundary conditions considered in the analysis. The
dimension (Dstiffener). This approach resulted in the creation of 336 boundary condition suggested in the International Ship and Offshore
original stiffened plates and 336 corresponding adjusted models, leading Structures Congress (ISSC) report [7] was used in this study to simulate a
to a cumulative assemblage of 2688 finite element (FE) models for the simply supported constraint for the model.
parametric analysis. The comprehensive details pertaining to the orig­
inal stiffened plates are presented in Fig. 3 and Table 1. All the FE 3.2.2. Initial imperfection
models were analysed using ABAQUS software, and the numerical re­ Initial imperfections refer to irregularities or deviations that may
sults of the adjusted stiffened plates from different design approaches arise during the welding process, resulting in localised deformations or
are compared in Section 3.2 to demonstrate the effectiveness of each residual stresses. In this study, the initial distortion effect was carefully
approach in adjusting the number of stiffeners in the stiffened plates considered for the plate and stiffeners, whereas the welding residual
while maintaining identical loading capacities and buckling modes be­ stress was neglected [32]. Three types of imperfection factors were
tween the original and adjusted models. considered: plate initial deflection (wop), column-type initial deflection
of the stiffener (woc), and sideways initial deflection of the stiffener (wos),
which can be expressed as follows [33].
3.2. Modelling technique of stiffened plate The plate initial deflection is
mπ x πy
3.2.1. Material and boundary conditions wop = A0 sin sin (8)
a b
In this study, the nonlinear FE method was used to investigate the

6
Z. Wang et al. Thin-Walled Structures 192 (2023) 111201

Fig. 8. Data processing procedure of correction factor for panel slenderness ratio.

Fig. 9. Neural network model with 2 inputs, 1 output, and 10 neurons (MATLAB Nftool).

column-type initial deflection of the stiffener is the stiffened plate, as depicted in Fig. 5. The analysis revealed that a
mesh size of 50 mm, which involved at least two meshes along the height
πx πy of the web plate of the stiffeners, represented the most cost-effective
woc = B0 sin sin (9)
a B choice and was thus adopted as the standard for FE modelling in this
and sideways initial deflection of the stiffener is study.
Subsequently, the FE method was validated. The validation involved
z πx comparing the outcomes obtained through the FE model with the values
wos = C0 sin , (10)
hs a reported in the ISSC benchmark study of Tee-bar stiffened plates on a
large crude carrier [7] (see Fig. 6). Comprehensive details of the
where m is the number of half waves of the buckling mode, defined as
boundary conditions and material properties of the benchmark study are
the minimum integer satisfying a/b ≤ [m(m + 1)]1/2; x, y, and z are the
available in the ISSC report. As shown in Fig. 6, the load-shortening
coordinates of the nodes; and A0, B0 and C0 are the coefficients of the
curve obtained through the FE model exhibited commendable consis­
initial imperfection equations. The magnitudes of B0 and C0 were taken
tency with the curve presented in the ISSC report. This finding indicates
as 0.0015a, and A0 was taken as 0.1β2t.
that the adopted FE simulation method has a high degree of accuracy in
forecasting the ultimate loading capacity of a stiffened plate subjected to
3.2.3. FE modelling validation
a compression load. The FE method was employed in the subsequent
To achieve a satisfactory balance between the computational effi­
parametric investigations. It should be noted that there will always be
ciency and numerical accuracy, mesh size convergence analysis was
some deviations between numerical and experimental results. However,
firstly conducted to determine the optimal mesh size for the FE model of

7
Z. Wang et al. Thin-Walled Structures 192 (2023) 111201

Fig. 10. Neural network training performance of Cf.

the numerically validated similarity relation between prototype and adjustment of the stiffened layout. The R-squared value of the data
scaled-down model will still exist in the experimental conditions [16]. points can be used to characterise such an alignment trend. For example,
Therefore, FE method can be utilized in the current study to obtain some in the case of n0/n1 = 12/10 in Fig. 7, the R-squared values of the β-γ, λ-γ,
helpful knowledge about the similarity design. and β-λ strategies are 0.9658, 0.9073, and 0.9289, respectively. In other
words, the β-γ strategy exhibits the most pronounced trend clustering
3.3. Numerical results around the standard line, indicating that this strategy is much more
effective in obtaining the desired reliable adjusted stiffened plate. A
Fig. 7 shows the results of the stiffener layout adjustment design with similar phenomenon is also observed in the remaining cases, as shown in
respect to the ultimate strengths of the stiffened plates, as obtained from Fig. 7. In contrast to the β-γ strategy, the utilisation of the remaining
all aforementioned methodologies. For clarity, we refer to the equiva­ design strategies may result in underestimation of the ultimate load-
lent sectional area-based method as the Ast strategy, whereas the critical carrying capacity of the original model (as most data points lie
parameter-based approaches are denoted as the β-γ, λ-γ, and β-λ strate­ beneath the standard line). Moreover, as the ratio of the number of
gies. For ease of interpretation, a standard line (solid black line) and its stiffeners n0/n1 increases progressively, a consistent increase occurs in
corresponding margins of error (dashed black lines) were incorporated the deviation between the data points and standard line across all
into the image. methods.
Several pieces of instructive information can be obtained from Fig. 7. The precision of the β-γ strategy is likely due to two factors. First, the
First, the accuracy of the stiffener layout adjustment design is directly method strictly satisfies the similarity of γ for the same buckling modes
proportional to the degree of proximity between the data points and between the original and adjusted models. Second, between the two
standard line. A closer alignment between the data point and standard factors related to the ultimate strength of the stiffened plate, β and λ, the
line corresponds to a higher level of consistency in the ultimate strengths similarity of β is satisfied, whereas λ is ignored, because β has a greater
between the original and adjusted models, indicating successful influence on the ultimate strength than λ. This concept is validated by

8
Z. Wang et al. Thin-Walled Structures 192 (2023) 111201

Fig. 11. Fitting surface of correction factors.

the correlation analysis results shown in Fig. 7, where the correlation in Eq. (11):
coefficient of β-σux is obviously larger that of λ-σux. However, in the √̅̅̅̅̅ {
numerical results of the β-γ strategy, some data points exhibit consid­ bA σY nO + 1 O A nO + 1 tpO
βA = A = Cf ⋅βO ← bA = A ⋅b , tp = A ⋅ (11)
erable imprecision due to overestimation of the ultimate strength of the tp E n +1 n + 1 Cf
adjusted models compared to the original models (as the data points are Subsequently, to satisfy the similarity of γ, the dimensions of the
significantly above the standard line). This phenomenon can be ascribed stiffeners should be corrected accordingly, as shown in Eq. (12):
to the fact that the β-γ strategy failed to maintain an identical column ⎧ A
slenderness ratio λ in the original and adjusted stiffened plates. How­ ⎪
⎪ a = aO
⎪
ever, the β-γ strategy should still be a good basis for further correction ⎪
⎪
⎪
⎪ nO + 1 O A nO + 1 tp
O
⎪
⎪ bA = A
compared to the other strategies. Proper correction needs to be per­ ⎪
⎪ n +1
⋅b , tp = A
n + 1 Cf
⋅
⎨
formed on the β-γ strategy to find a balance between β and λ to improve γA =
12(1 − v2 )I A
( )3 = γ ←
o √̅̅̅̅̅̅ . (12)
the similarity of the ultimate strength between original and adjusted ⎪ O
bA tpA ⎪
⎪ n +1 4 1 O
⎪ DAstiffener = A ⋅ D
models. ⎪
⎪
⎪
⎪
n + 1 Cf 3 stiffener
⎪
⎪
⎪
⎩ I∝D4 ,D = h ort orh ort
4. Corrected similarity method stiffener stiffener w w f f

It should be noted that when β is corrected, λ will also be corrected

4.1. Stiffener layout adjustment approach accordingly, as presented in Eq. (13). Therefore, the essence of the
correction process is to modify β and λ of the stiffened plate obtained
The findings of the parametric investigation demonstrate that the β-γ from the β-γ strategy further, thereby improving the similarity of ulti­
strategy is a viable approach for modifying the number of stiffeners of an mate load between original and adjusted models. Finally, the proposed
original stiffened plate, while concurrently maintaining the same stiffener layout adjustment approach for a stiffened plate under
loading capacity between the original and adjusted models. In this compression can be expressed using Eq. (14):
strategy, γ and β are utilised to control the similarity of the buckling aA = aO
mode and ultimate strength of the stiffened plate, respectively. How­ A aA
√̅̅̅̅ nA +1 3/4 O
√̅̅̅̅̅̅
ever, λ is another important factor affecting the ultimate strength of the λ = π rA
σY
E = C λ ←{ A
nO +1 f nO + 1 4 1
Dstiffener = A ⋅ Dstiffene
stiffened plate. The failure to consider λ in the similarity method results n + 1 C3f
in discrepancies in the ultimate strength between the original and rO r∝Dstiffener , Dstiffener = hw ortw orhf ortf (13)
adjusted models, depicted in Fig. 7, thereby necessitating further ⎧ A
refinement and correction. ⎪
⎪
⎪
L = LO , BA = BO , aA = aO
⎪
To ensure the reliable safety assessment of the practical structure, a ⎪
⎪ A
⎪ βA = C βO , γ A = γ O , λA = n + 1 C3/4 λO
⎪
⎪
good similarity both in buckling modes and ultimate strengths between ⎪
⎪
⎪
f
n +1 f
O
⎪
original and adjusted models should be made in an experiment. The ⎪
⎪
⎪
⎨ O
nA + 1 O A nA + 1 tp
buckling mode of the stiffened plate under axial compression load is bA = o b , tp = O ⋅ , (14)
n +1 n + 1 Cf
highly related to γ [27]. It has proved that γ is an essential parameter in ⎪
⎪
⎪
⎪
⎪ √ ̅̅̅̅̅
̅
the similarity design of stiffened plates [16–18]. Moreover, once the ⎪
⎪
⎪ nO + 1 4 1 O
⎪
⎪ DAstiffener = A ⋅ D
buckling mode of the stiffened plate is changed, the ultimate strength of ⎪
⎪
⎪ n + 1 Cf3 stiffener
⎪
the model will be significantly affected, which will lead to large dis­ ⎪
⎪
⎩
crepancies in experimental results. To avoid this, in the current study, γ Dstiffener = hw ortw orhf ortf
is strictly kept the same between original and adjusted models. A
correction factor will then be incorporated into β to improve the preci­ where Cf represents the correction factor for the ultimate strength of the
sion of the method. The formula for the corrected version of β is shown stiffened plate, and its specific value is yet to be determined (parametric
scope: 1.18 < β < 3.94; 0.50 < λ 1.13; 0.85 < γ < 635.09).

9
Z. Wang et al. Thin-Walled Structures 192 (2023) 111201

Fig. 12. Flowchart of design procedure for stiffener layout adjustment approach.

Step (I): The original stiffened plates were selected from real ship
Table 3 structures. The material and geometric dimensions of the original
Stiffened plate scenario with one type of stiffener for validation analysis (235 model were defined first.
MPa).
Step (II): Various correction factors, denoted as Cfi, were used in the
Plate Stiffener stiffener layout adjustment of the same original stiffened plate. The
tp (mm) Layout Cross-sectional dimension of stiffener specific value of Cf for the original stiffened plate was determined by
n0 — n1 Stiffener type: hw × tw × hf × tf (mm) linear interpolation of the numerical outcomes.
n0: before adjustment Step (III): The process of Step (II) was repeated for all original stiff­
n1: after adjustment
ened plates selected in Step (I), resulting in a dataset of Cf for each
(1) 11.0 (1) 12 — 10 (1) T-bar: 138.0 × 9.0 × 90.0 × 12.0 original model.
(2) 12.5 (2) 12 — 8 (2) T-bar: 170.0 × 9.0 × 90.0 × 13.0
Step (IV): The dataset of Cf was processed using conventional data
(3) 15.0 (3) 12 — 6 (3) T-bar: 200.0 × 10.0 × 90.0 × 14.0
(4) 18.5 (4) 8 — 6 (4) T-bar: 235.0 × 10.0 × 90.0 × 15.0 processing and artificial neural network technology, leading to the
(5) 8 — 4 development of an empirical formula of correction factors for the
(6) 4 — 2 panel slenderness ratio.
(7) 2 — 1 Step (V): Accuracy tests were performed to validate the obtained
*Note: for all models, a = 2900.0 mm, b = 700.0 mm, σ y = 235 MPa. empirical formula and neural network.
The present study employed a data processing procedure to deter­ Step (VI): Finally, a validation analysis of the proposed stiffener
mine the specific value of Cf, as illustrated in Fig. 8. This procedure layout adjustment approach was conducted using a nonlinear finite
involved the following steps. element simulation.

10
Z. Wang et al. Thin-Walled Structures 192 (2023) 111201

Table 4
Stiffened plate scenario with one type of stiffener for validation analysis (355 MPa).
Plate Stiffener

tp (mm) Layout Cross-sectional dimension of stiffener

n0 — n1 Stiffener type: hw × tw × hf × tf (mm)
n0: before adjustment
n1: after adjustment

(1) 12.5 (1) 12 — 10 (1) T-bar: 138.0 × 9.0 × 90.0 × 12.0

(2) 18.5 (2) 12 — 8 (2) T-bar: 170.0 × 9.0 × 90.0 × 13.0
(3) 12 — 6 (3) T-bar: 200.0 × 10.0 × 90.0 × 14.0
(4) 8—6 (4) T-bar: 235.0 × 10.0 × 90.0 × 15.0
(5) 8—4
(6) 4—2
(7) 2—1

*Note: for all models, a = 2900.0 mm, b = 700.0 mm, σ y = 355 MPa.

Fig. 13. Numerical results of the stiffener layout adjustment approach applied to the stiffened plate with 235 MPa yield stress.

Fig. 14. Numerical results of the stiffener layout adjustment approach applied to the stiffened plate with 355 MPa yield stress.

4.2. Correction factor determination Through the data processing procedure, more than 2000 FE models of
the stiffener plate were solved.
The aforementioned data processing procedure was implemented on In this study, the neural network toolbox in MATLAB was employed
336 original stiffened plates included in the parametric study, as pre­ to develop and simulate the networks of a correction factor dataset.
sented in Table 2. Furthermore, supplementary 144 models (n0/n1 = 10/ Artificial neural network technology has significant advantages in
6, 10/8, and 6/4) were incorporated to enhance the thoroughness of the dealing with multivariable fitting problem [34,35], facilitating the
investigation, resulting in 480 original models in the data processing creation of models for complex nonlinear systems that cannot be easily
procedure of Cf. To derive a precise correction factor Cf, 3–5 adjusted expressed using closed-form equations [36]. A two-layer feed-forward
models needed to be generated for each of the original stiffened plates. network with sigmoid hidden neurones and linear output neurones was

11
Z. Wang et al. Thin-Walled Structures 192 (2023) 111201

Fig. 15. Load–displacement curves and collapse behaviours of original and adjusted models obtained by stiffener layout adjustment approach (original stiffened
plate with 235 MPa yield stress).

employed, which can fit multidimensional mapping problems with high The investigation also presented an empirical formula pertaining to
accuracy, providing a consistent dataset and sufficient hidden neurones. Cf by utilising conventional data processing techniques to enhance the
Two parameters in the proposed similarity design formula, γ and practicality of the method. The fitting outcomes of various functional
(n0+1)/(n1+1), were designated as input parameters for the neural expressions were compared in the data processing procedure. Subse­
network training of the correction factor. The Levenberg-Marquardt quently, the functional form demonstrating the most remarkable fitting
back-propagation function was utilised to train the network with two precision was adopted (see Fig. 11). The empirical formula obtained for
inputs and Cf as the output. The resulting network model included 2 Cf can be expressed as
inputs, 1 output, and 10 neurons, as illustrated in Fig. 9. The training ⎧
data comprised 70% of the correction factor data (336 samples), ⎨ Cf = C1 γ C2
C = − 0.3679η2 + 1.6292η − 0.2552 , (15)
whereas the remaining 144 samples were used for the validation and ⎩ 1
C2 = 0.1275η2 − 0.5014η + 0.3891
testing of the neural network.
The performance of the neural network was evaluated based on the where η = (n0 + 1)/(n1 + 1) and C1 and C2 are the obtained fitting
mean-squared error (MSE) for both the training and testing data, coefficients.
regression plots of the training and testing data, and error histograms, as
shown in Fig. 10. Fig. 10(a) depicts the network training performance in
terms of MSE over 132 epochs. The training, validation, and test data 4.3. Design procedure
exhibit similar decreasing trends, indicating that the network was suc­
cessfully trained. Fig. 10(b) shows the network regression results for the This section outlines the proposed design procedure for modifying
training, testing, validation, and combined datasets, all of which exhibit the stiffener layout of existing stiffened plates using the proposed stiff­
R-values greater than 0.97. In Fig. 10(c), the training error histogram, ener layout adjustment approach. The procedure can be effectively
which consists of 20 bins, shows a normal distribution with the most applied to the optimisation and design of a scaled-down stiffened plate
data points lying near the zero-error line. comprising multiple types of stiffeners typically encountered in actual
ship structures. Furthermore, such a design procedure can be expanded

12
Z. Wang et al. Thin-Walled Structures 192 (2023) 111201

Fig. 15. (continued).

to encompass the stiffener layout modification of stiffened plates with information is available in reference [18]. This method can modify the
varying material characteristics in conjunction with the prior research material yield stress of a stiffened plate while preserving the buckling
outcomes obtained by the authors [18]. mode and ultimate strength of the model, thereby expanding the
A detailed flowchart of the procedure is shown in Fig. 12. A stiffened applicability of the proposed stiffener layout adjustment procedure.
plate featuring two different stiffeners was chosen as a representative Upon performing the material adjustment step, a stiffened plate with a
case while acknowledging that the approach can be extended to other material yield stress of 235 MPa was obtained, and the subsequent step
stiffened plates for stiffener layout modification in a commensurate of the procedure was utilised to acquire the desired test model with the
manner. The procedure initially partitions the stiffened plate into mul­ appropriate stiffener layout. Alternatively, the obtained test model can
tiple stiffened plate elements, namely, P0(S1), P0(S2), and P0(S3), based be restored to its former material yield stress σ y, depending on the
on the locations of the longitudinal girders. Subsequently, the proposed particular testing conditions and loading ability of devices.
similarity method is used to modify the stiffener layout of each stiffened
plate element to generate the corresponding adjusted elements P1(S1), 5. Application to actual stiffened plates
P1(S2), and P1(S3). These modified elements were then amalgamated to
produce the final adjusted stiffened plate, and the cross-sectional di­ 5.1. Stiffened plate with one type of stiffener
mensions of the girders were adjusted by preserving the same γ between
the original and adjusted girders. This section describes the application of the proposed stiffener layout
An additional material-adjustment step must be performed in the adjustment approach and its associated procedure to a series of stiffened
procedure if the material yield stress σy of the original stiffened plate is plates from the decks of a double-hull tanker [7]. The objective of this
not equivalent to 235 MPa. The specific material adjustment method study was to validate the effectiveness of the approach in adjusting the
utilised in the procedure is shown in Fig. 12, and further detailed stiffener layout of actual stiffened plates while preserving the loading

13
Z. Wang et al. Thin-Walled Structures 192 (2023) 111201

Fig. 15. (continued).

capacity and buckling mode of the models. Detailed information about original stiffened plates with a yield stress of 235 MPa. This character­
the selected original stiffened plates is presented in Tables 3 and 4. The istic may be due to deviations caused by the material-adjustment step in
stiffened plates in Tables 3 and 4 incorporate two commonly used ma­ the design procedure. However, the precision of the proposed similarity
terial yield stresses in ship structures, namely, 235 MPa and 355 MPa. In method remains sufficient to fulfil the fundamental requirements of
total, 168 original stiffened plates were included in the validation engineering investigation (the maximum error is 4.97%), even for a
analysis. stiffened plate with a yield stress of 355 MPa.
Figs. 13 and 14 display the results of the stiffener layout adjustment Figs. 15 and 16 compare the dimensionless load-shortening curves of
design with respect to the ultimate strengths of the stiffened plates as the typical original and corresponding adjusted stiffened plates obtained
obtained from the proposed similarity method and procedure. The re­ from the proposed method and the approach of Wang [20], respectively.
sults for the original stiffened plates with a material yield stress of 235 Fig. 15 depicts the outcomes of the original stiffened plates with 235
MPa are presented in Fig. 13. This figure shows that the data points MPa material yield stress. All the load shortening curves of original
exhibit pronounced clustering around the standard line and that all the models exhibit commendable consistency with the curves of their cor­
data points are within the ±5% deviation curves for the standard line responding adjusted models. The ultimate points of the load-shortening
(the maximum error is –3.62%). This finding indicates that the proposed curves of the adjusted models are always slightly higher than those of
similarity method is effective in adjusting the stiffener layout of the the original models, indicating that the proposed similarity method
stiffened plate while preserving the same ultimate strengths between the tends to overestimate the ultimate strengths of the stiffened plates.
original and adjusted models. The results of the original stiffened plate Furthermore, each original stiffened plate shown in Fig. 15 manifests a
with a yield stress of 355 MPa are provided in Fig. 14. It can be seen from similar buckling mode to its corresponding adjusted model, i.e. the
Fig. 14 that the proposed similarity method exhibits a slightly greater overall buckling collapse of stiffeners with their associated panels.
magnitude of overall design error when applied to the original stiffened Fig. 16 depicts the outcomes for the original stiffened plates with
plates with a yield stress of 355 MPa compared to when applied to the 355 MPa material yield stress. In the stiffener layout adjustment design

14
Z. Wang et al. Thin-Walled Structures 192 (2023) 111201

Fig. 16. Load–displacement curves and collapse behaviours of original and adjusted models obtained by stiffener layout adjustment approach (original stiffened
plate with 355 MPa yield stress).

of the original models, an additional material-adjustment step was scaled-down adjusted models also exhibited ultimate strengths close to
performed, as shown in Fig. 12, to derive an adjusted model with a those of the corresponding original models. These scaled-down adjusted
material yield stress of 235 MPa. Notably, the dimensionless ultimate models were used in the test owing to their fine similarity and improved
strength of the adjusted model (235 MPa) in Fig. 16 is close to that of the processing feasibility. The approach proposed by Wang [20] also had a
original model (355 MPa). Nevertheless, the comparative analysis re­ positive effect on the stiffener layout adjustment design for several
sults in Fig. 16 also indicate that the implementation of the proposed original stiffened plates, as shown in Figs. 15 and 16. However, large
similarity method in the optimisation of the stiffener layout may lead to discrepancies may occur between several original and adjusted models
greater dimensionless displacements of the adjusted stiffened plates at because of the lack of similarity in the buckling modes in the stiffener
the ultimate point, particularly when utilised in a model with a yield layout adjustment process.
stress of 355 MPa or greater. This discrepancy may be attributed to
additional deviations introduced by the material-adjustment step during 5.2. Stiffened plate with multiple types of stiffeners
the design process. In the realm of ultimate strength analysis, the focus is
typically on the maximal load-bearing capacity of a given structure The stiffener layout adjustment approach and its associated pro­
rather than its ultimate displacement. Thus, the proposed method re­ cedures were applied to an actual stiffened plate with two types of
tains a notable degree of precision when applied to the stiffener layout stiffeners from a practical hull tanker. Information regarding the
adjustment design of stiffened plates with a material yield stress of 355 selected original stiffened plate and its corresponding full-scale and
MPa. scaled-down adjusted models obtained by the proposed similarity
An appropriate similarity criterion [18] can be applied to full-scale method is provided in Table 5. The detailed design process is presented
adjusted models to convert them into scaled-down models. Benefiting in Appendix A.
from the accurate results of the full-scale adjusted models, the obtained Fig. 17 compares the dimensionless load–displacement curves of the

15
Z. Wang et al. Thin-Walled Structures 192 (2023) 111201

Fig. 16. (continued).

original stiffened plate and its corresponding adjusted models. The 6. Conclusion
load–displacement curves of the adjusted stiffened plates, both full-scale
and scaled-down, exhibit a high degree of similarity to those of the This paper proposed a corrected similarity method for designing a
original model throughout the buckling process. The ultimate strengths scaled-down stiffened plate subjected to a compression load. A reliable
of the full-scale and scaled-down adjusted stiffened plates were 3.17% stiffener layout adjustment approach, together with its associated design
and 0.83% higher than those of the original model, respectively. Fig. 18 procedure, was introduced into the existing similarity method to
shows the buckling collapse behaviours of the original and adjusted improve the processing feasibility of the scaled-down model, enabling
stiffened plates. The buckling modes in Fig. 18 indicates that an adjusted the precision of experimental studies to be improved. In the corrected
stiffened plate with a different number of stiffeners manifests the same method, the stiffener layout adjustment procedure is first applied to the
buckling modes as the corresponding original model. In summary, the prototype model. Then, the obtained adjusted model with different
application of the proposed corrected similarity method to actual stiff­ stiffener layouts was further scaled down using the previously proposed
ened plates demonstrated that the method has good precision in pre­ similarity criterion. Finally, a scaled-down model with different stiffener
serving the same buckling modes and ultimate compressive strengths layouts and improved processing quality was derived.
between the original and adjusted stiffened plates across varying stiff­ To investigate the effectiveness of the proposed stiffener layout
ener layouts. Therefore, it can be reliably utilised in the optimisation adjustment approach, it was applied to a set of actual stiffened plates
and design of scaled-down models of real ship structures, thereby extracted from the decks of a practical hull tanker using FE analysis.
improving the processing feasibility and test precision of the Several conclusions can be drawn from the numerical results of the
experiments. validation research.

(1) The proposed stiffener layout adjustment approach yields a pos­

itive outcome in modifying the stiffener layouts of stiffened

16
Z. Wang et al. Thin-Walled Structures 192 (2023) 111201

Table 5
Stiffened plate scenario with multiple types of stiffeners for validation analysis.
Component Original model (Full scale) Adjusted model (Full scale) Adjusted model (scaled)

P0(S1) or P1(S1) n0(S1) = 4 n1(S1) = 2 n1(S1) = 2

or P2(S1) tp = 12.5 mm; b = 950 mm tp = 20.9 mm; b = 1583 mm tp = 4.2 mm; b = 316.6 mm
T 138 × 90 × 9 × 12 (mm) T 230 × 150 × 15 × 20 (mm) T 46 × 30 × 3 × 4 (mm)
P0(S2) or P1(S2) n0(S2) = 8 n1(S2) = 4 n1(S2) = 4
or P2(S2) tp = 12.5 mm; b = 950 mm tp = 22.7 mm; b = 1710 mm tp = 4.5 mm; b = 340 mm
T 138 × 90 × 9 × 12 (mm) T 250 × 163 × 16.3 × 21.7 (mm) T 50 × 32.6 × 3.3 × 4.3 (mm)
P0(S3) or P1(S3) n0(S3) = 4 n1(S3) = 2 n1(S3) = 2
or P2(S3) tp = 12.5 mm; b = 950 mm tp = 20.9 mm; b = 1583 mm tp = 4.2 mm; b = 316.6 mm
T 138 × 90 × 9 × 12 (mm) T 230 × 150 × 15 × 20 (mm) T 46 × 30 × 3 × 4 (mm)
Girder T 235 × 90 × 10 × 15 (mm) T 410 × 157 × 17.4 × 26.1 (mm) T 82 × 31.4 × 3.5 × 5.2 (mm)
Stiffened plate a = 4750 mm a = 4750 mm a = 950 mm
P0(S1+S2+⋅⋅⋅+S3) L = 9500 mm; B = 18,050 mm L = 9500 mm; B = 18,050 mm L = 1900 mm; B = 3610 mm
or P1(S1+S2+⋅⋅⋅+S3) σy = 235 MPa σy = 35 MPa σy = 235 MPa
or P2(S1+S2+⋅⋅⋅+S3)

*Note: P0 = original model, P1 = adjusted model, and P2 = scaled model. For the symbol meanings, see Fig. 12.

Fig. 17. Load–displacement curves of original and adjusted models obtained by stiffener layout adjustment approach (stiffened plate with multiple types
of stiffeners).

plates. Specifically, the adjusted model generated by this a larger range of values for β, λ, γ, and n0/n1, in order to establish a more
approach exhibits a buckling mode and ultimate compressive comprehensive empirical formula or neural network for the corrected
strength similar to those of its corresponding original model, even similarity method. Experimental studies are also required to validate
when the two models have different numbers of stiffeners. This and improve this method. This work may serve as a significant com­
approach can significantly improve the processing feasibility and plement to existing methodologies for the experimental study of ship
quality of scaled-down models. structures.
(2) The adjusted stiffened plate generated using this approach tends
to overestimate the ultimate compressive strength of the stiffened CRediT authorship contribution statement
plates. Nevertheless, the design accuracy of the approach remains
adequate to fulfil engineering and research requirements. Zhuo Wang: Writing – original draft. Xiangshao Kong: Supervision,
(3) Although the proposed approach demonstrates relatively high Conceptualization, Writing – review & editing. Weiguo Wu:
accuracy, it is generally advisable to remain the stiffener layout in Supervision.
scaled-down models, where the spacing distance of the stiffeners
is already suitable for the manufacturing process. Any modifica­
Declaration of Competing Interest
tion to the stiffener layout inevitably affects the ultimate strength
of the scaled-down models.
We declare that we have no known competing financial interests or
personal relationships that could have appeared to influence the work
The current study had some limitations. First, the sample size of
reported in this paper. There is no professional or other personal interest
original stiffened plate scenarios needs to be expanded further, covering
of any nature or kind in any product, service and/or company that could

17
Z. Wang et al. Thin-Walled Structures 192 (2023) 111201

Fig. 18. Collapse behaviours of original and adjusted models obtained by stiffener layout adjustment approach (stiffened plate with multiply types of stiffeners).

be construed as influencing the position presented in the manuscript Data availability

entitled “Corrected similarity method for designing scale-down stiffened
plate subjected to compression load with improved processing Data will be made available on request.
feasibility”.

Appendix A

Similarity design process of practical stiffened plate in Section 5.2.

1) Determine the original stiffened plate P0(S1+S2+S3)

L = 9500mm, a = 4750mm, B = 18, 050mm, σy = 235MPa

2) Divide P0(S1+S2+S3) into design elements

Stiffened plate element:

P0 (S1 ) : n0 = 4,tp = 12.5mm, b = 950mm, T138 × 90 × 9 × 12(mm)

P0 (S2 ) : n0 = 8,tp = 12.5mm, b = 950mm, T138 × 90 × 9 × 12(mm)

P0 (S3 ) : n0 = 4,tp = 12.5mm, b = 950mm, T138 × 90 × 9 × 12(mm)

Girder between stiffened plate elements:

T235 × 90 × 10 × 15(mm)

1) Apply stiffener layout adjustment design to P0(S1), P0(S2), and P0(S3)

P1 (S1 ) : n1 = 4→2,α1 = 0.9952
/
tp = (4 + 1) (2 + 1)×12.5 ÷ α = 20.9mm

b = (4 + 1)/(2 + 1)×950 = 1583mm

/ ( )1/4
{T138 × 90 × 9 × 12} × (4 + 1) (2 + 1) × 1/α1 3 = {T230 × 150 × 15 × 20}

18
Z. Wang et al. Thin-Walled Structures 192 (2023) 111201

P1 (S2 ) : n1 = 8→4, α2 = 0.9912

/
tp = (8 + 1) (4 + 1) × 12.5 ÷ α = 22.7mm

b = (4 + 1)/(2 + 1)×950 = 1710mm

/ ( )1/4
{T138 × 90 × 9 × 12} × (8 + 1) (4 + 1) × 1/α2 3 = {T250 × 163 × 16.3 × 21.7}

P1 (S3 ) : n1 = 4→2, α3 = 0.9952

/
tp = (4 + 1) (2 + 1) × 12.5 ÷ α = 20.9mm

b = (4 + 1)/(2 + 1)×950 = 1583mm

/ ( / )1/4
{T138 × 90 × 9 × 12} × (4 + 1) (2 + 1) × 1 α3 3 = {T230 × 150 × 15 × 20}

2) Adjust girder
[ / ( )1/4 / ( )1/4 ]
{T235 × 90 × 10 × 15} × (4 + 1) (2 + 1) × 1/α1 3 + (8 + 1) (4 + 1) × 1/α2 3 ÷2 = {T410 × 157 × 17.4 × 26.1}

3) Combine all the elements and obtain adjusted stiffened plate P1(S1+S2+S3)
4) Apply the appropriate scaled-down criterion to P1(S1+S2+S3) and obtain the scaled model P0(S1+S2+S3), CL = 5.0, in the sample.

References [17] T. Yuan, Y. Yang, X.S. Kong, W.G. Wu, Similarity criteria for the buckling process
of stiffened plates subjected to compressive load, Thin-Walled Struct. 158 (2021),
107183.
[1] O.F. Hughes, B. Ghosh, Y. Chen, Improved prediction of simultaneous local and
[18] Z. Wang, T. Yuan, X.S. Kong, W.G. Wu, A universal similarity method and design
overall buckling of stiffened panels, Thin-Walled Struct. 42 (2004) 827–856.
procedure for buckling assessment of stiffened plates under compression load on
[2] J.K. Paik, B.J. Kim, Ultimate strength formulations for stiffened panels under
real ships, Thin-Walled Struct. 181 (2022), 110025.
combined axial load, in-plane bending and lateral pressure: a benchmark study,
[19] K.W. Hutchinson, Ship Construction, Elsevier, 2006.
Thin-Walled Struct. 40 (2002) 45–83.
[20] Q.H. Wang, P.Y. Wei, C.T. Li, D.Y. Wang, A unified similarity criterion and design
[3] E. Jaberzadeh, M. Azhari, Elastic and inelastic local buckling of stiffened plates
method for geometrically distorted scale models of thin-walled hull girder
subjected to non-uniform compression using the Galerkin method, Appl. Math.
structures, Thin-Walled Struct 180 (2022), 109866.
Model. 33 (2009) 1874–1885.
[21] Rules For the Manufacture, Testing and Certification of Materials, Lloyd’s Register
[4] K. Ghavami, M.R. Khedmati, Numerical and experimental investigations on the
of Shipping, London, UK, 2021.
compression behaviour of stiffened plates, J. Constr. Steel Res. 62 (2006)
[22] Classification Society of China, Rules for Material and Welding of Shipping,
1087–1100.
Beijing, CN. (2021).
[5] S. Tanaka, D. Yanagihara, A. Yasuoka, Evaluation of ultimate strength of stiffened
[23] V.T. Doan, B. Liu, Y. Garbatov, W.G. Wu, C.Guedes Soares, Strength assessment of
panels under longitudinal thrust, Mar. Struct. 36 (2014) 21–50.
aluminium and steel stiffened panels with openings on longitudinal girders, Ocean
[6] J.K. Paik, B.J. Kim, J.K. Seo, Methods for ultimate limit state assessment of ships
Eng 200 (2020), 107047.
and ship-shaped off-shore structures: part II stiffened panels, Ocean Eng. 35 (2008)
[24] BS EN ISO 4063, Welding and Allied Processes – Nomenclature of Processes and
271–280.
Reference Numbers, Geneva, CH. (2000).
[7] ISSC, Ultimate strength (committee III.1), in: Proceedings of the 18th International
[25] D. Faulkner, J.C. Adamchak, G.J. Snyder, M.F. Vetter, Synthesis of welded grillages
Ship and Offshore Structures Congress (ISSC 2012), Rostock, Germany, 2012,
to withstand compression and normal loads, Comput Struct. 3 (1973) 221–246.
pp. 9–13. September.
[26] S.P. Timoshenko, J.M. Gere, Theory of Elastic Stability, McGraw-Hill, New York,
[8] J. Caldwell, Ultimate longitudinal strength, Transp. R. Inst. Nav. Arch. 107 (1965)
1961.
411–430.
[27] Y. Ueda, T. Yao, Ultimate strength of compressed stiffened plates and minimum
[9] R.S. Dow, Testing and analysis of a 1/3-scale welded steel frigate model, in: Proc.
stiffness ratio of their stiffeners, Eng. Struct. 5 (1983) 97–107.
Int. Conf. Adv. Mar Struct, 1991.
[28] M.C. Xu, C.Guedes Soares, Experimental study on the collapse strength of wide
[10] S. Xu, B. Liu, Y. Garbatov, W. Wu, C.Guedes Soares, Experimental and numerical
stiffened panels, Mar. Struct. 30 (2013) 33–62.
analysis of ultimate strength of inland catamaran subjected to vertical bending
[29] N.E. Shanmugam, D.Q. Zhu, Y.S. Choo, M. Arockiaswamy, Experimental studies on
moment, Ocean Eng. 188 (2019), 106320.
stiffened plates under in-plane load and lateral pressure, Thin-Walled Struct. 80
[11] C.L. Wang, J.M. Wu, D.Y. Wang, Experimental and numerical investigations on the
(2014) 22–31.
ultimate longitudinal strength of an ultra large container ship, Ocean Eng. 192
[30] D.K. Kim, H.L. Lim, S.Y. Yu, A technical review on ultimate strength prediction of
(2019), 106546.
stiffened panels in axial compression, Ocean Eng. 170 (2018) 392–406.
[12] C.L. Wang, J.M. Wu, D.Y. Wang, Design similar scale model of a 10,000 TEU
[31] D.K. Kim, H.L. Lim, M.S. Kim, O.J. Hwang, K.S. Park, An empirical formulation for
container ship through combined ultimate longitudinal bending and torsion
predicting the ultimate strength of stiffened panels subjected to longitudinal
analysis, Appl. Ocean Res. 88 (2019) 1–14.
compression, Ocean Eng. 140 (2017) 270–280.
[13] A.A. Sonin, The Physical Basis of Dimensional Analysis, Department of Mechanical
[32] Q. Wang, D. Wang, Scaling characteristics of hull girder’s ultimate strength and
Engineering, MIT, Cambridge, 2001.
failure behaviors: an empirically modified scaling criterion, Ocean Eng. 212
[14] J. Araneda, Dimensional-directional analysis by a quaternionic representation of
(2020), 107595.
physical quantities, J. Franklin I. 333 (1996) 113–126.
[33] O.F. Hughes, J.K. Paik, Ship structural analysis and design, Pergamon (2010).
[15] M. Xu, Z. Song, J. Pan, Study on the similarity methods for the assessment of
[34] C. Thiel, K. Neumann, F. Ludwar, A. Rennings, J. Doose, D. Erni, Coating damage
ultimate strength of stiffened panels under axial load based on tests and numerical
localization of naval vessels using artificial neural networks, Ocean Eng. 192
simulations, Ocean Eng. 219 (2021), 108294.
(2019), 106560.
[16] X.S. Kong, Y. Yang, J. Gan, T. Yuan, L. Ao, W.G. Wu, Experimental and numerical
[35] Z.R. Tahir, P. Mandal, Artificial neural network prediction of buckling load of thin
investigation on the detailed buckling process of similar stiffened panels subjected
cylindrical shells under axial compression, Eng. Struct. 152 (2017) 843–855.
to in-plane compressive load, Thin-Walled Struct. 148 (2020), 106620.
[36] S. Sapna, A. Tamilarasi, M.P. Kumar, Backpropagation learning algorithm based on
Levenberg Marquardt algorithm, Comp. Sci. Inform. Technol. 2 (2012) 393–398.

19