Sei sulla pagina 1di 25

Journal of Constructional Steel Research 50 (1999) 151–175

Buckling of stiffened steel plates—a parametric


study
G.Y. Grondin*, A.E. Elwi, J.J.R. Cheng
Department of Civil and Environmental Engineering, University of Alberta, Edmonton, Alberta,
Canada, T6G 2G7

Received 18 August 1997; received in revised form 28 July 1998; accepted 2 August 1998

Abstract

The stability of plates stiffened with tee-shape stiffeners was investigated using a finite
element model. Four series of stiffened plate panels were modeled using a finite strain four-
node shell element. The model was validated using the results of tests on full-size stiffened
plate specimens and was subsequently used to perform the study of various parameters
presented in this paper. The parameters investigated are: the shape and magnitude of initial
imperfections in the plate; residual stress magnitude and direction of applied uniform bending;
plate slenderness ratio; plate aspect ratio; and plate to stiffener cross-sectional area ratio. The
effect of the investigated parameters on the axial load carrying capacity and the mode of
failure of stiffened plates is investigated both in the elastic and inelastic ranges. A comparison
of these results with design guidelines formulated by Det norske Veritas and the American
Petroleum Institute indicates that the guidelines are generally conservative for cases where
initial imperfection magnitudes do not exceed the guidelines’ prescribed maximum.  1999
Elsevier Science Ltd. All rights reserved.

Keywords: Steel plate; Buckling; Effective width; Finite element; Initial imperfections; Residual stresses;
Tee stiffener; Tripping

NOMENCLATURE
Ap area of the plate (B ⫻ t)
Ast area of the stiffener

* Corresponding author. Tel: ⫹ 1-403-492-2794; fax: ⫹ 1-403-492-0249; e-mail: gygrondin@civil.ual-


berta.ca

0143-974X/99/$ - see front matter  1999 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 4 3 - 9 7 4 X ( 9 8 ) 0 0 2 4 2 - 9
152 G.Y. Grondin et al. / Journal of Constructional Steel Research 50 (1999) 151–175

B longitudinal stiffeners spacing, also taken as the width of a stiffened


plate panel
Be plate panel effective width
bf stiffener flange width
E modulus of elasticity
h height of the stiffener stem
L length of a stiffened plate panel
np number of half sine waves in the deformed shape of the plate along
the length of the plate
r radius of gyration of stiffener
t plate thickness
tf stiffener flange thickness
W stiffener stem thickness
w0 magnitude of the maximum out-of-plane deflection
wp magnitude of out-of-plane deflection
x distance along the length of the plate
y distance across the width of the plate
␤ plate slenderness ratio defined by Eq. (2)
␴Y yield strength

1. Introduction
Thin steel plates that are stabilized in one direction by stiffeners are used exten-
sively for plating of ship decks and hulls, components of offshore structures, box-
girder bridges, bridge decks and other structures in which a high strength-to-weight
ratio is important. Flexure of the entire hull of a ship or box girder of a bridge
will induce longitudinal compressive stresses in the stiffened panels that form these
elements. This may be coupled with local bending moments arising from transverse
loads acting directly on the stiffened panels, e.g. wheel loads acting on a bridge
deck. Because of the presence of compressive axial forces and bending moments,
stiffened plates are susceptible to failure by instability. Instability failure can take
various forms: buckling of the plate between stiffeners; lateral torsional buckling of
the stiffeners, also called tripping; overall buckling of the stiffened panel as a column,
referred to in the following as Euler buckling, or; local buckling of the stiffener.
This last mode of failure is not considered in this study.
Test results [1,2] have indicated that failure by tripping of the stiffener is more
critical than failure by buckling of the plate because it is associated with a sudden
collapse. Although plates stiffened on one side have considerable ability to carry
transverse loads that put the flange of the stiffener in tension, tripping of the stiffener
must also be considered in the analysis where the structure is such that bending will
cause compression to develop in the flange of the stiffener. For example, initial
imperfections favoring tripping in the lower deck of a box-girder cannot be avoided.
Similarly, wheel loads acting on alternate panels of the upper deck will induce nega-
tive moments (which put the flange of the stiffener in compression) in the unloaded
panels, again predisposing the stiffeners to tripping failure.
G.Y. Grondin et al. / Journal of Constructional Steel Research 50 (1999) 151–175 153

Collapse analysis of stiffened plates subjected to axial compression has received


considerable attention [2–5]. A common simplified approach to predict the capacity
of stiffened plates has been to treat one stiffener with the associated plate width as
a simply supported beam-column [6,7]. Assuming that the ultimate load is reached as
the outer fibers reach yield under compression and bending, a simple Perry-Robertson
formula can be derived. Using this approach, Euler buckling is obtained using a
column formed of one stiffener and an attached plate width equal to the stiffener
spacing. Similarly, the plate buckling mode of failure is taken into account by reduc-
ing the width of the plate associated with each stiffener, from the stiffener spacing
to an effective width. The effective width of the plate has been determined either
analytically or experimentally [7]. The main advantage of this method is its sim-
plicity. This model, however, is not normally suitable for tripping of stiffeners,
though it could if satisfactory predictions could be obtained for the limiting stress
in the stiffeners. Since the model cannot properly account for interaction between
the stiffener and plate in the pre- and post-buckling ranges it has limited applicability.
With current analysis tools and computing power, however, more precise modeling
of stiffened steel plates can be achieved. Factors such as residual stresses, initial
imperfections and yielding of significant parts of the cross-section can be explicitly
incorporated into numerical models. The finite difference method used by Smith and
co-workers [8] included material yielding, plate and stiffener imperfections and local
buckling. That model, however, assumed that tripping of the stiffeners is suppressed
by suitable proportioning of the stiffeners. The authors of the current work have
obtained excellent correlation [1] between results of tests on full-size stiffened plates
and a numerical model using the finite strain four node shell element S4R from
ABAQUS [9]. The same finite element model was used to perform an extensive
study of the behaviour and strength of stiffened plates under axial in-plane loading
and bending moment. This paper presents the results of the numerical investigation.
The parameters that have been investigated are the effects of the magnitude and
shape of initial imperfections, residual stresses, plate slenderness, interaction between
axial load and out-of-plane bending, plate aspect ratio, and stiffener to plate area
ratio.

2. Finite element model

A stiffened plate is typically fabricated from a flat plate with equally spaced longi-
tudinal stiffeners and may span between girders. A typical cross-section of the type
of stiffened plate panel considered in this investigation is shown in Fig. 1. Because
of the symmetry of the stiffened plate, only one panel, namely a portion of the plate
of width B with a stiffener centered on the plate strip, was modeled. The stiffened
plate panel was modeled and analyzed using the commercial finite element code
ABAQUS. A total of 576 plate bending S4R elements were used to model the stiff-
ened panel. The flange and stem of the stiffener were each modeled with 96 elements
and the plate was modeled with 384 elements. This mesh size was found to yield
satisfactory convergence. The S4R element is a four node shell element that allows
154 G.Y. Grondin et al. / Journal of Constructional Steel Research 50 (1999) 151–175

Fig. 1. Stiffened plate panel.

for changes in the thickness as well as finite membrane strains. The model invoked
large displacement using a Total Lagrangian formulation. The plate material behav-
iour was modeled by an elastic-plastic constitutive model incorporating a von Mises
yield surface and an isotropic strain-hardening flow rule. The rotation about the longi-
tudinal axis was suppressed at all the nodes along the unloaded edges to simulate
full continuity. The loaded ends of the plate panel were constrained so that the plane
end sections would remain plane. This type of boundary conditions simulates the
framing of the stiffeners into transverse beams or bulkheads.
In all cases studied the residual stresses were first introduced using a procedure
to be described in detail later. If required, an initial bending moment was then intro-
duced. Finally, in-plane axial compressive forces were applied and increased gradu-
ally to failure.
In order to model the full range of behaviour of the panel, including both the pre-
buckling and the post-buckling regimes, the solution strategy started with a load
control standard Newton-Raphson iterative procedure in the initial stage of loading,
then shifted to a modified Riks procedure as the ultimate load was approached. This
procedure permits tracing the behaviour in the softening post-buckling regime [10].
The cross-sectional parameters outlined in Fig. 1 can all vary independently of
each other over a certain range. In order to keep the various possible combinations
of dimensions to a manageable number, simplifying assumptions had to be made.
The panel width, B, the stiffener flange width, bf, and the overall depth of the stiff-
ened panel (t ⫹ h ⫹ tf) were taken as 500 mm, 100 mm and 125 mm, respectively.
The ratio of the stiffener stem thickness to flange thickness was taken as 0.75. For
all the cases investigated, except where indicated otherwise, the ratio of plate area
to stiffener area, (Ap/Ast), was taken as 3.0, the length to width aspect ratio, L/B,
was taken as 4.0, the compressive residual stress in the plate was taken as 15% of
the yield strength, ␴y, and the initial imperfection pattern was taken to be a three
half wave sinusoidal pattern of average magnitude. Factors investigated such as
Ap/Ast, residual stresses, initial imperfection and transverse loads will be discussed in
detail later. That leaves the plate width to thickness ratio B/t as a primary parameter.
G.Y. Grondin et al. / Journal of Constructional Steel Research 50 (1999) 151–175 155

3. Modes of failure of stiffened plates

Stiffened plates in real structures can be loaded under a combination of in-plane


and out-of-plane loading. In-plane loads can include axial or biaxial compression
and in-plane shear. Out-of-plane loading would include lateral pressure or bending
about the transverse and longitudinal axes of the stiffened plate. Although a stiffened
plate panel can be loaded in a number of ways, the work presented in the following
is limited to two different loading conditions. In one, uniaxial compression only was
introduced in the direction of the stiffeners. In the other, an initial bending about
the transverse axis was introduced followed by uniaxial compression. When axial
loading is superimposed on bending, an initial bending moment of either 30 percent
or 60 percent of the plastic bending moment capacity of the stiffened plate panel
cross-section is applied following the application of the residual stresses. These load-
ing conditions are common in several civil engineering structures such as the bottom
flange of box girders and bridge decks.
Instability of stiffened plates under such loading conditions can take one of three
forms: buckling of the plate between the stiffeners (Fig. 2(a)); lateral torsional buck-
ling of the stiffener with rotation of the stiffener about the junction between the plate
and the stiffener (this mode of failure is also referred to as tripping of the stiffener)
(Fig. 2(b)) and; buckling of the plate between the stiffeners followed by an overall,
or “Euler”, buckling (Fig. 2(c)).
For all the cases investigated a tripping failure mode was triggered only when a
bending moment was applied to induce initial compressive stresses in the flange of
the stiffener. Tripping of the stiffener was not observed with other loading conditions.
Danielson and co-workers [11] demonstrated that tripping of the stiffener is depen-
dent on the torsional stiffness of the stiffener. The torsional stiffness of the stiffener
relative to the stiffness of the plate in bending is believed to be a determining factor
in deciding whether, under axial compression, the failure mode will take the form
of plate buckling or stiffener tripping. In the work presented here the ratio of stiffener
torsional stiffness to plate bending stiffness is probably too high to make stiffener
tripping the governing buckling mode, except when the compression in the plate was
decreased by the application of bending moment.
The load deformation response of stiffened plates is strongly affected by the mode
in which the stiffened plate fails. The load versus plate shortening response for stiff-
ened plates failing by various failure modes are displayed in Fig. 3. The three stiff-
ened plate specimens failing by plate buckling and Euler buckling were loaded in
axial compression whereas the plate that failed by stiffener tripping was loaded under
combined end moments and axial compression. Each of the stiffened plates illustrated
in Fig. 3 had a different slenderness, B/t, ratio so that different failure modes could
be triggered. A comparison of the load deformation behaviour for the modes of
failure observed in the analysis indicates that failure by Euler buckling is the most
favorable mode of failure since it has a more stable post-buckling behaviour. On the
other hand, failure by tripping of the stiffener is the least desirable since it is charac-
terized by a sudden drop in load capacity just after the peak load is reached.
Although the average applied peak stress in several of the cases investigated was
156 G.Y. Grondin et al. / Journal of Constructional Steel Research 50 (1999) 151–175

Fig. 2. Failure modes of stiffened plates.


G.Y. Grondin et al. / Journal of Constructional Steel Research 50 (1999) 151–175 157

Fig. 3. Typical load vs. deformation behaviour of stiffened steel plates.

well below the yield stress of the material, the analysis showed that in all cases there
was always some degree of yielding in the cross-section before the peak load was
reached. Localized high strains were reached in slender plates as a result of the
superposition of residual stresses, applied axial stresses, and local bending stresses
introduced by the axial load acting on the initially deformed shape. Details of the
magnitude and distribution of the residual stresses and initial imperfections are
presented in the following sections.
The effective width concept has frequently been used in the design of stiffened
plates to account for the post-buckling strength reduction. After buckling of the plate
between the stiffeners, a portion of the load in the plate shifts towards the stiffener.
As a result, a non-uniform stress distribution is formed in the plate with the maximum
stress magnitude present in the plate strip immediately adjacent to the stiffener. At
the time of failure it is assumed that the load is resisted by a strip of material on
either side of the stiffener stressed to the yield level, whereas a central portion of
the plate between the stiffeners takes no load. This concept has been shown to work
well when applied to plate failure alone. However, in the case of stiffener failure
the behaviour is more complicated because tripping failure is essentially torsional
in nature [11] and will result in interaction between the stiffener and the plate.
The membrane stress distribution across the width of a stiffened plate panel at
the third point along the length of the plate is plotted in Fig. 4 at various stages of
loading. Although residual stresses were included in the model, only the applied
stresses are shown here. In the following the term “applied membrane stresses” will
be used to reflect the removal of residual stresses from the stress picture. Fig. 4
illustrates the case where the plate was loaded under combined axial load and an
initial bending moment of 0.3 Mp that placed the flange of the stiffener initially in
compression. One observes that the applied bending stresses are initially almost uni-
158 G.Y. Grondin et al. / Journal of Constructional Steel Research 50 (1999) 151–175

Fig. 4. Stress distribution in the plate (failure by stiffener tripping).

form across the width of the plate. As the compressive load is increased, the applied
membrane stresses become gradually non-uniform but remain symmetric about the
stiffener. At the peak load the stresses are significantly higher in the vicinity of the
stiffener. This phenomenon results from shear lag in the plate and from the partial
loss of plate stiffness as the out-of-plane deformations increase. In the post buckling
range, at an axial shortening of about 1.0 percent of the plate length, the applied
membrane stresses in the plate are reversed and skewed. This is caused by the sudden
drop in axial load at tripping accompanied by large lateral displacements as well as
twisting of the stiffener. The flexural component becomes dominant causing stress
reversal and the twist introduces warping stresses in the system.
Fig. 5 illustrates the applied membrane stress distribution at various stages of
loading of a panel failing by plate buckling. At the initial stage of loading (designated
as the pre-buckling stage in Fig. 5) the applied stresses are uniform across the plate
panel. As the peak load level is approached, however, the portion of the plate near
the stiffener carries more load. Since the stiffener in this type of failure retains its
effectiveness, as evidenced by the stress distribution in the post-buckling range, the
post-buckling behavior after plate buckling is much more ductile than in the case of
tripping failure. The stress distribution in the plate, therefore, does not change charac-
ter in the post-buckling range.
G.Y. Grondin et al. / Journal of Constructional Steel Research 50 (1999) 151–175 159

Fig. 5. Stress distribution in the plate (failure by plate buckling).

4. Parametric study

4.1. Effect of initial imperfections

Irrespective of the degree of sophistication of the numerical method used, the main
drawback of any procedure attempting to use numerical results as a basis for deriving
design data remains the uncertainty related to the magnitudes and distributions of
the initial deformations and residual stresses. Initial imperfections in structures that
have not been subjected to damage usually result from the fabrication process. A
large number of measurements of post-welding distortions have been reported in the
literature [12]. However, the majority are made on panels which were specifically
produced for experiments. Most of the panels were not made to full scale, and it is
doubtful whether the welding conditions for the stiffeners resemble those in pro-
duction workshops. Although a number of studies on plate distortion were made on
actual ship hulls and decks only in very few cases have the distortions been referred
to the dimensions of the plates. Furthermore, often only the maximum or central
deflections have been reported.
In this study the effect of initial imperfections and of residual stresses are con-
sidered in turn. Only one type of initial out-of-straightness is accounted for, namely,
out-of-plane imperfections in the plate. The effect of stiffener imperfections is not
considered in this study.
From measurements on 196 plates, Carlsen and Czujko [12] determined that the
160 G.Y. Grondin et al. / Journal of Constructional Steel Research 50 (1999) 151–175

deformed shape of welded stiffened plates used in ship structures can be expressed
by a double trigonometric series. For simplification in this study, only one term of
the series is used as follows:
wp(x,y) ⫽ w0 sin (np␲x/L)sin(␲y/B) (1)
where x and y are distances along the length and across the width of a plate, respect-
ively, L is the length of the plate, B is the width of the plate, w0 is the magnitude
of the maximum out-of-plane deflection, and np is the number of half sine waves in
the deformed shape of the plate along its length. Carlsen and Czujko [12] found that
the most critical shape was obtained when only one term of the series was used with
np ⫽ 3 for a plate with an aspect ratio of 3.0.
Using the results of extensive surveys of actual ship deformations, Smith et al.
[13] defined w0 as slight (w0 ⬍ 0.025␤2t), average (w0 ⬍ 0.10␤2t), and severe (w0
⬍ 0.30␤2t) where ␤ is the plate slenderness ratio:

␴Y
冪E
B
␤⫽ (2)
t

In this study the initial imperfections are defined according to Eq. (1) with B, the
stiffener spacing, taken as 500 mm. The analysis is performed for a range of slender-
ness ratios (␤ ⫽ 0.63, 1.67 and 2.72) and for initial imperfection magnitudes equal to
the upper limits defined by Smith et al. for slight, average and severe imperfections.
The constant term np in Eq. (1), the number of half waves along the length of the
plate, is varied from 1 to 12. The deformed shape of a stiffened plate panel for np
equal to 10 is shown in Fig. 6(a). A limited number of other initial imperfections
were also investigated. In two models (␤ ⫽ 1.67 and 2.72) randomly distributed out-
of-plane ripples were introduced using a random number generator but constrained
to zero at the plate to stiffener intersection as shown in Fig. 6(b). In two other cases
(␤ ⫽ 1.67 and 2.72) the initial imperfection pattern consisted of one single wave
placed diagonally across the plate as shown in Fig. 6(c). The crest of the sine wave
follows a straight line intersecting the stiffener at zero imperfection.
The magnitude of the compressive residual stresses in the plate was taken as 15%
of the yield strength. The distribution of residual stresses in the specimen and the
method used to incorporate both the residual stresses and the initial imperfections
in the model are discussed in the following section. It should be noted also that all
the models investigated above were loaded only with an axial force. Therefore, the
tripping mode was excluded.
Given the very large number of analyses performed, only some of those results
will be discussed in detail. The effect of the magnitude of initial imperfections for
one and four half sinusoidal waves along the length of the stiffened plate is presented
in Figs. 7 and 8, respectively, for different plate slenderness ratios. Fig. 7 indicates
that the magnitude of the initial imperfections has little effect on the stiffened plate
capacity when initial imperfections take the form of one half wave. Fig. 8, however,
shows that the magnitude of initial imperfections has a very significant effect on the
G.Y. Grondin et al. / Journal of Constructional Steel Research 50 (1999) 151–175 161

Fig. 6. Initial imperfections.


162 G.Y. Grondin et al. / Journal of Constructional Steel Research 50 (1999) 151–175

Fig. 7. Effect of the magnitude of initial imperfections (one half sine wave).

Fig. 8. Effect of the magnitude of initial imperfections (four half sine waves).

stiffened plate ability to carry load when the initial imperfections take the form of
four half waves. A conservative estimate of plate capacities for initial imperfections
consisting of two and three half waves can be obtained by linear interpolation
between the results obtained for one half wave and four half waves.
The mode of failure changes from Euler buckling (Fig. 2(c)) to plate buckling
G.Y. Grondin et al. / Journal of Constructional Steel Research 50 (1999) 151–175 163

(Fig. 2(a)) as the plate slenderness increases. The plate slenderness ratio where this
change in mode of failure is observed is indicated in Figs. 7 and 8.
Examination of these figures indicates that the load carrying capacity of the stiff-
ened plate decreases as the number of half-waves in the initial imperfection pattern
increases. This is observed more readily in Fig. 9 where the ratio of peak load to
yield load is plotted as a function of the number of half waves along the length of
a plate panel. Three curves are presented: one for a plate slenderness ratio of 2.72
(failure by buckling of the plate with little yielding), one for a plate slenderness ratio
of 1.67 (failure by bucking of the plate with significant yielding of the section), and
a curve corresponding to a plate slenderness ratio of 0.63 (failure by Euler buckling).
Average magnitude initial imperfections were assumed in all cases. Fig. 9 indicates
that, for plates with a slenderness ratio of 2.72 and 1.67 the critical load is strongly
dependent on the shape of the initial imperfections. The minimum critical load is
reached when eight half-waves are used for the configuration of the initial imperfec-
tions. A reduction of capacity of about 25 percent is observed as the number of half-
waves in the initial shapes increases from one to eight. For initial imperfections with
one to 12 half-waves the analysis showed that the initial imperfections are amplified
until the peak load is reached. After the peak load is reached, typically only two
half waves amplify while the other half-waves gradually become smaller as the load
decreases in the post-buckling range. Fig. 2(a) shows the final configuration of a
stiffened plate that started out with 10 half-waves in its initial configuration. The
effect of initial imperfections was found to be negligible for a stiffened plate that
fails by Euler buckling (this is the case for the plate with a slenderness ratio, ␤,
of 0.63).
The peak loads obtained for the random initial imperfections and the skewed wave

Fig. 9. Effect of the number of half waves on the load carrying capacity of a stiffened plate.
164 G.Y. Grondin et al. / Journal of Constructional Steel Research 50 (1999) 151–175

were found to be somewhere between the peak loads obtained for a plate with four
half-waves and a plate with two half waves. It is apparent from Fig. 9 that the
prediction of plate capacity based on initial imperfections consisting of three to four
half sine waves may be sufficiently conservative. Measurements of post-weld distor-
tions of plates of different aspect ratios have indicated that the deformed shape con-
sists of two to three half waves along the length of a panel [12]. As the number of half
waves in the initial deflected shape increases from four to eight a further reduction of
capacity of as much as 15% is observed. For all the initially deformed shapes investi-
gated in this study the post-buckling configuration of the plate consists of two half-
waves located near the centre portion of the plate as shown in Fig. 2(a).
In general, this study clearly shows the sensitivity of stiffened plates to initial
imperfections. Measurements of initial imperfections by others have indicated that
two to three half waves along the length of a plate are quite possible. As a conserva-
tive estimate, if the actual shape of the initial imperfections is not known, four half
sine waves can be assumed for the shape of the initial imperfections. One should
keep in mind that eight half sine waves is highly unlikely and indicate a very severe
condition. As a less conservative assumption for the initial imperfections, the follow-
ing study uses three half waves to describe the shape of the initial imperfections.
The reader can accurately make the necessary corrections for other numbers of half
waves using Figs. 7–9. When using these figures to estimate the capacity of a stiff-
ened plate with a number of half waves other than one or four, Figs. 7 and 8 are
used to obtain the capacity for one and four half sine waves, respectively. These
values can then be plotted in Fig. 9 where, using the trend established in the figure
for plate slenderness ratios of 0.63, 1.67, and 2.72, the buckling capacity for the
given number of half waves can be readily obtained. Alternatively, lines similar to
those shown in Figs. 7 and 8 can be reconstructed from data points obtained along
a vertical line drawn in Fig. 9. The reconstructed curve can be used to estimate the
failure load for slenderness ratios, ␤, other than those presented in Fig. 9.

4.2. Effect of residual stresses

The presence of residual stresses in stiffened plates is mainly attributable to the


welding of stiffening members to the plate. The residual stresses in the weld and in
the plate and stiffener material in the vicinity of the weld are close to the yield limit
as a result of the contraction of the welds. Residual stress measurements in stiffened
plates have shown that the tension zone around the weld extends three to six times
the plate thickness [1,14,15].
The magnitude and distribution of the residual stresses are governed by welding
parameters such as heat input and cooling rate; both are governed by the welding
procedure adopted for fabrication. Most test specimens would be fabricated using
manual welding, such as shielded metal arc welding. In production workshops, how-
ever, semi-automatic or automatic welding techniques are more likely to be used.
Hence the information on residual stresses in real structures is scarce. For this reason,
measurements of longitudinal residual stresses were performed by Grondin et al. [1]
on test specimens fabricated using automatic submerged arc welding to weld the
G.Y. Grondin et al. / Journal of Constructional Steel Research 50 (1999) 151–175 165

stiffener to the plate simultaneously on both sides of the stem. This welding technique
is representative of large scale fabrication processes.
A typical residual stress pattern used for the analysis is shown in Fig. 10. The
tension existing at the plate to stiffener junction is balanced by residual compression
which exists largely in the plate. It has been reported that the magnitude of the
compressive residual stresses increases as the plate slenderness becomes smaller.
Values of up to 75 percent of the yield strength have been measured, although 25
percent of the yield strength is believed to be more common for typical ship struc-
tures [15]. Because of the relatively high magnitude of residual stresses in stiffened
plates, it is important that residual stresses be included in strength calculations.
The residual stress picture of Fig. 10 was introduced in the finite element model
by imposing initial strains in the form of a temperature distribution. Only longitudinal
residual stresses were investigated in this study. In order to introduce initial strains
only in the longitudinal direction, an orthotropic temperature material property was
used that had zero thermal expansion coefficients in directions transverse to the axis
of the stiffened plate. The initial strains introduce initial stresses, upon which an
iteration is carried out in order to establish equilibrium. A complication arises
because the introduction of residual stresses in the model creates a distortion of the
specimen which must be considered when investigating the effect of initial imperfec-
tions. The application of residual stresses was therefore performed in two stages.
First, a set of strains, equal in magnitude but opposite to the desired residual strains,
was introduced in the model with the desired initial imperfections. The resulting
displacements were superimposed on the initial imperfections and a new deformed,
stress-free, mesh was thus generated. Initial strains corresponding to the residual
strains were then applied on the newly generated model. At this stage the model
incorporates the desired magnitude and distribution of the initial imperfections and
residual stresses.
The magnitude of the residual stress at the plate to stiffener junction (Fig. 10)
was taken as the yield strength of the material in tension. The residual stresses in
the stiffener were assumed to be similar to those measured in the full-size specimen
reported by Grondin et al. [1]. The exact distribution, however, varied slightly,
depending on the changes required during an equilibrium iteration. As mentioned

Fig. 10. Typical residual stress pattern in a stiffened plate.


166 G.Y. Grondin et al. / Journal of Constructional Steel Research 50 (1999) 151–175

before, the compressive residual stresses in the plate were taken to be 15 percent of
the yield strength of the plate material in the models used to investigate the effect
of initial imperfections, out of plane loading, plate aspect ratio and plate to stiffener
area ratio. The 15 percent level is consistent with the level measured in a full-size
panel fabricated using automatic submerged dual arc welding with the fillets on either
side of the stiffener stem laid simultaneously [1].
To investigate the effect of the level of residual stresses, three cases were exam-
ined, namely, residual stresses in the plate of zero, 15 and 30 percent of the yield
strength. Fig. 11 shows the effect of residual stresses on the capacity of stiffened
plates. At plate slenderness ratios larger than about 1.7 the residual stresses in the
plate decreases the strength of stiffened plates roughly in direct proportion to the
magnitude of the compressive residual stresses in the plate. As the plate slenderness
ratio where the mode of failure changes from plate buckling to Euler buckling is
approached, the capacity does not decrease proportionally with the magnitude of the
residual stresses in the plate. This is due to a number of reasons. For cases where
failure is by plate buckling, the uniform residual compressive stresses, present over
most of the plate, govern the behaviour of the stiffened plate. For this later condition,
the buckling stress in the presence of residual stresses can be approximated as the
difference between the critical stress when the plate is free of residual stresses and
the magnitude of residual stresses. When Euler buckling is the governing mode of
failure, the residual stresses in the plate become less influential and the overall distri-
bution rather than the residual stresses in the plate only becomes important. It should
also be noted that at plate slenderness ratios, ␤, smaller than about 1.7, the combined
effect of the average buckling stress and the residual stress in the plate causes the
plate to yield before buckling.

Fig. 11. Effect of residual stresses.


G.Y. Grondin et al. / Journal of Constructional Steel Research 50 (1999) 151–175 167

4.3. Effect of out-of-plane loading

The superposition of out-of-plane loads to in-plane loads was investigated by


superposing equal end moments to the in-plane axial load. Five loading combinations
were investigated, namely, no applied end moments, end moments of 30 percent of
the plastic moment capacity of the cross-section to place the stiffener initially either
in tension or in compression, and end moments of 60 percent of the plastic moment
capacity placing the stiffener initially either in tension or in compression. For each
loading condition the plate area to stiffener area ratio, Ap/Ast, was taken as 3.0, the
residual compressive residual stresses in the plate were fixed at 15 percent of the
yield strength, and initial imperfections of average amplitude with three half sine
waves along the length of the plate were assumed. The plate slenderness ratio,
B/t√␴Y/E, was varied from about 0.6 to about 2.8. Fig. 12 shows that superimposed
bending has two distinct effects, namely, to reduce the load carrying capacity of
stiffened plates and to change the failure mode from one of plate buckling when the
bending moment places the flange of the stiffener initially in tension, to stiffener
tripping when the bending moment places the flange of the stiffener initially in com-
pression.
As might be expected, when tripping is the mode of failure, plate slenderness has
little to no effect on the load carrying capacity of the stiffened plate. However, when
plate buckling is the mode of failure, the slenderness of the plate has obviously a
strong effect on the capacity of the stiffened plate. As the failure mode becomes
Euler buckling, the effect of plate slenderness on the ratio of buckling load to yield
load becomes insignificant.
When stiffened plates are subjected to a bending moment that places the flange
of the stiffener in compression, the analysis shows that an increase in the bending

Fig. 12. Effect of initial bending moment and plate aspect ratio.
168 G.Y. Grondin et al. / Journal of Constructional Steel Research 50 (1999) 151–175

moment from 0.3 Mp to 0.6 Mp, results in a decrease in capacity of about 0.3 PY.
A comparison of the buckling curve for stiffened plates with no bending moment
with that for plates subjected to 0.3 Mp with the stiffener initially in compression
indicates a similar reduction in capacity of 0.3 PY when Euler buckling is the govern-
ing failure mode (these are plates with small slenderness ratios). However, when
failure occurs by plate buckling the change in load carrying capacity as the applied
moment is increased from zero to 0.3 Mp varies with the plate slenderness ratio and
becomes very small as the plate slenderness approaches 3.0.
Fig. 12 also indicates that stiffened plates are not as sensitive to bending moments
when the applied bending moment places the flange of the stiffener into tension. An
increase in moment of 0.3 Mp generally results in a decrease in capacity of 0.12 PY.
This strength reduction is observed for all values of plate slenderness.

4.4. Effect of plate aspect ratio

Plates of two aspect ratios, L/B, were investigated, namely, 3.0 and 4.0. The aspect
ratio was varied by changing the length of the plate. The plates with an aspect ratio
of 3.0 were investigated with a superimposed bending moment of 0.3 Mp and 0.6
Mp applied to cause initial compression or initial tension in the flange of the stiffener.
The initial imperfection shape was taken as three half waves of average magnitude.
The results for plate slenderness, ␤, from 0.6 to 2.8 are presented in Fig. 12. The
figure indicates only a small change in capacity as the plate aspect ratio changes
from 4.0 to 3.0. The figure also indicates that, in general, a decrease of plate aspect
ratio increases the plate capacity when the failure mode is either Euler buckling or
stiffener tripping. Insignificant change in capacity is observed when failure occurs
by plate buckling.

4.5. Effect of plate to stiffener area ratio

The parametric study presented in the preceding sections was performed with a
plate to stiffener area ratio of 3.0. Additional cases were investigated where the plate
to stiffener area ratio was varied by changing the stiffener area while keeping the
overall depth of the section and its slenderness ratio, L/r, constant. Two load cases
were investigated, namely one where the stiffened plate is subjected to axial loading
and the other where a bending moment is superimposed to the axial load. The super-
imposed moment corresponds to 30 percent of the plastic moment capacity and
causes compression in the flange of the stiffener. The plate slenderness ratio, ␤, used
for this investigation is 2.09. The initial imperfections are taken as three half waves
of average magnitude. The residual compressive stresses in the plate are assumed
to be 15 percent of the yield strength.
Fig. 13 presents a plot of the ratio of peak to yield load versus plate to stiffener
area ratio. Fig. 13 indicates that the mode of failure of stiffened plates under pure
axial load is plate buckling for all the plate to stiffener area ratios investigated. In
addition, it is seen that the area ratio has little effect on the load carrying capacity
for area ratios less than about 1.5 and greater than about 4.0. However, for area
G.Y. Grondin et al. / Journal of Constructional Steel Research 50 (1999) 151–175 169

Fig. 13. Effect of plate to stiffener area ratio.

ratios between 1.5 and 4.0, the effect of plate to stiffener area ratio is significant
and a reduction in capacity of 20 percent is observed over that range.
When stiffened plates are subjected to a bending moment that places the flange
of the stiffener in compression, the mode of failure is either tripping of the stiffener
(for plate to stiffener area ratios greater than about 2.5) or Euler buckling. The load
carrying capacity is found to decrease almost linearly with increasing plate to stiff-
ener area ratio. The sensitivity of the stiffened plates to the area ratio is more signifi-
cant for this loading case than for the case where no moment is applied to the
stiffened plate. This is likely a reflection of the fact that the area ratio was varied
by varying the area of the stiffener while keeping the area of the plate constant.
Since the tripping and Euler buckling failure modes are controlled by the stiffener,
the effect of varying the stiffener area is probably more influential on these two
modes of failure than it is when plate buckling is the governing failure mode.
The above observations appear at first to be in contrast to the observations made
by Carlsen [16]. In the work of Carlsen the plate to stiffener cross-sectional area
ratio was varied by changing the area of the plate while keeping its slenderness ratio
constant. Two failure conditions were detected, namely, a plate induced failure and
a stiffener induced failure. When the failure mode was stiffener tripping, little effect
of the area ratio was detected. This is likely because the area of the stiffener was
kept constant. Carlsen also found that the plate to stiffener area ratio was not influen-
tial when the mode of failure was plate buckling. This, again is expected since the
plate slenderness ratio was kept constant when its area was changed. Consequently,
Carlsen concluded that the plate to stiffener cross-sectional area ratio was not sig-
nificant.
The work performed by the authors and the work of Carlsen are reconciled if it
can be stated that the cross-sectional area of the stiffened plate component that causes
170 G.Y. Grondin et al. / Journal of Constructional Steel Research 50 (1999) 151–175

instability of the stiffened plate is the important parameter rather than the ratio of
the area of the plate to the area of the stiffener.

5. Comparison with design practice

A large number of numerical analyses were carried out to compare with widely
accepted design guidelines [17,18]. The numerical investigation included the effect
of initial imperfections and residual stresses, which are usually not specifically
included in the design procedures. The effect of residual stresses and initial imperfec-
tions are normally accounted for through the selection of a plate effective width,
obtained from examination of test results. Furthermore, the bulk of the research con-
ducted on the behaviour and strength of stiffened steel plates has concentrated on
the elastic behaviour. The numerical analysis presented here has considered both
elastic and inelastic behaviours. An assessment of current design practice is therefore
performed in light of this new body of data.
The two design guidelines selected for this assessment are Classification note No.
30.1 by Det norske Veritas [17] and the American Petroleum Institute Bulletin 2V
[18]. Both guidelines present a comprehensive procedure for computation of the
buckling strength of stiffened steel plates whereby the stiffened plate capacity is
evaluated based on the various failure modes identified in this paper.
The procedure adopted in the two guidelines uses significantly different plate panel
effective widths. The plate panel effective width adopted by DnV [17] depends on
whether failure of the stiffened plate in induced by the stiffener or by the plate. API
RP 2V [18] does not make this distinction. It does reduce, however, the plate effec-
tive width significantly when stiffener failure occurs before the plate material around
the stiffeners reaches yield. This reduction was supported by test results and was
recommended by Faulkner [14]. Fig. 14 illustrates the difference in the ratio of plate
effective width, Be, to full width, B, predicted using API RP 2V and DnV design
approaches. It can be observed that API RP 2V is significantly more conservative
than DnV when stiffener failure occurs before the plate panel yields near the stiff-
eners.
A comparison between API, DnV and the numerical analysis results is presented
in Fig. 15. The results of the numerical analysis are presented for two different
shapes and two different magnitudes of initial imperfection, namely, one half and
four half sine waves along the length of the plate and for small and severe magnitudes
of initial imperfections. The cross-sectional parameters outlined in Section 2 of this
paper were used. The plate length and width used here are 2000 mm and 500 mm,
respectively. Since the current design methods [17,18] are semi-empirical in nature,
they account for some level of geometrical imperfections and residual stresses. The
predictions of stiffened plate capacity all follow the same general trend across the
plate slenderness range investigated. As expected from the comparison of plate panel
effective width presented in Fig. 14, API RP 2V provides a more conservative esti-
mate of the plate buckling capacity than DnV. In fact, when compared to the results
of the numerical analysis, API RP 2V provides a conservative strength prediction
G.Y. Grondin et al. / Journal of Constructional Steel Research 50 (1999) 151–175 171

Fig. 14. Plate panel effective width.

Fig. 15. Effect of plate slenderness—assessment of design practice.


172 G.Y. Grondin et al. / Journal of Constructional Steel Research 50 (1999) 151–175

for the entire range of plate slenderness investigated. On the other hand, strength
predictions based on DnV exceed those obtained using the numerical analysis in the
range of slenderness from 1.1 to 2.5 for plates with severe imperfections and four
half waves. It should be noted, however, that the maximum initial imperfections for
which DnV guidelines are applicable fall in the range of small to medium amplitude
imperfections as defined earlier.
Fig. 16 presents a comparison of the predicted plate capacity under combined axial
load and uniform bending. Initial imperfections are assumed to consist of three half
waves of average magnitude. The uniform bending moments used for this assessment
are 0.3 and 0.6 times the plastic moment capacity of the cross-section. The formu-
lation presented in DnV limits the maximum strain in the cross-section to the yield
strain. API RP 2V allows the section to reach its ultimate capacity, namely the plastic
moment capacity of the section. This has the effect of reducing somewhat the gap
in predicted capacity between DnV and API that was observed above. Fig. 16(a)
shows the predicted axial load carrying capacity for the case where the bending
moment is applied to place the plate into compression. A very good agreement is
observed between the DnV guidelines and the numerical analysis results. On the
other hand, API RP 2V still shows a very conservative strength prediction at both
moment levels.
Fig. 16(b) shows the predicted strength when the applied moment places the plate
panel initially in tension. For this case, the DnV guidelines are found to be more
conservative than the API guidelines, especially at higher applied moment. It can
be seen that, with an applied moment of 60 percent of the plastic moment capacity,
the DnV design procedure would allow little to no axial load. For this case, however,
API RP 2V is in excellent agreement with the finite element predictions.

6. Summary and conclusions


The capacity of stiffened plates subjected to combinations of axial compression
and bending was investigated using a finite element model which was demonstrated
to be able to predict accurately the capacity of full-size stiffened plates tested under
various loading conditions. A large displacement formulation with an elastic-plastic
isotropic strain-hardening material model was used for the analysis. The parameters
that were investigated are initial imperfections in the plate, residual stresses, end
bending moments and plate to stiffener cross-sectional area ratio. The stiffener spac-
ing, B, the stiffener flange width, bf, and the overall depth of the stiffened panel (t
⫹ h ⫹ tf) were taken as 500 mm, 100 mm and 125 mm, respectively. This was
done to keep the number of variables in the investigation to a reasonable size. The
ratio of the stiffener stem thickness to flange thickness was taken as 0.75. For all
the cases investigated, except where indicated otherwise, the ratio of plate area to
stiffener area, (Ap/Ast), was taken as 3.0, the plate aspect ratio, L/B, was taken as
4.0, the compressive residual stress in the plate was taken as 15% of the yield
strength, and the initial imperfection pattern was taken to be a three half sine waves
of average magnitude. The plate width to thickness ratio B/t was taken as the primary
parameter and its value was varied to include elastic as well as inelastic behaviour.
G.Y. Grondin et al. / Journal of Constructional Steel Research 50 (1999) 151–175 173

Fig. 16. Effect of initial bending—assessment of design practice.


174 G.Y. Grondin et al. / Journal of Constructional Steel Research 50 (1999) 151–175

The load versus deformation response of stiffened plates in the post-buckling range
was found to be strongly affected by the mode of failure of the stiffened plate, with
overall buckling being the most desirable mode of failure and stiffener tripping the
least desirable due to its abrupt reduction in capacity past the peak load.
Both the magnitude and the shape of the initial imperfections were shown to have
a significant influence on the capacity of stiffened plates failing by plate buckling.
Although a minimum plate buckling capacity is reached when the initial imperfec-
tion configuration consists of eight half sine waves along the length of a panel, it is
recommended that three or four half sine waves be assumed when the exact shape
of the initial imperfections is unknown. Initial imperfections in the plate were found
to have negligible effect on the capacity of plates failure by overall Euler buckling.
Three levels of residual stresses, namely, compressive residual stresses in the plate
of zero, 15 and 30 percent of the yield strength were investigated. At plate slender-
ness ratios, ␤, larger than about 1.7 the residual stresses in the plate decreases the
strength roughly in direct proportion to the magnitude of the compressive residual
stresses in the plate. However, when yielding sets in before buckling, the effect of
residual stresses is diminished. When failure of stiffened plates is by overall Euler
buckling the effect of the residual stresses in the plate is less pronounced.
Out-of-plane loading, applied here in the form of end moments, has a significant
impact on the mode of failure. For the stiffened plate proportions used in this investi-
gation it was found that out-of-plane loading causing compression in the flange of
the stiffener was necessary to trigger stiffener tripping. The sensitivity of stiffened
plates to out-of-plane loading was found to be twice as severe when the out-of-plane
loading causes compression in the flange of the stiffener, compared to tension in the
stiffener flange.
An investigation of the plate to stiffener area ratio indicates this parameter is not
influential. Rather, the cross-sectional area of the stiffened plate element that initiates
instability of the stiffened plate is the important parameter. This observation is con-
sistent with earlier observations made by Carlsen [16].
An assessment of current design practice outlined in DnV [17] and API RP 2V [18]
indicated that current practice is generally conservative. Although a large number of
cases were investigated in this study, further work is necessary to investigate a
broader range of parameters. More specifically, the effect of initial imperfections in
the stiffener and its influence on the tripping behaviour of stiffened plates need to
be investigated.

References
[1] Grondin GY, Chen Q, Elwi AE, Cheng JJR. Stiffened steel plates under compression and bending.
Journal of Constructional Steel Research 1998;45(2):125–48.
[2] Murray NW. Buckling of stiffened panels loaded axially and in bending. The Structural Engineer
1973;51(8):285–301.
[3] Timoshenko SP, Gere JM. Theory of elastic stability, McGraw-Hill, 1961.
[4] Sherbourne AN, Liaw CY, Marsh C. Stiffened plates in uniaxial compression. IABSE 1971;31:145.
[5] Balaz I, Murray NW. A comparison of some design rules with results from tests on longitudinally
stiffened deck plates of box-girders. Journal of Constructional Steel Research 1992;23:31–54.
G.Y. Grondin et al. / Journal of Constructional Steel Research 50 (1999) 151–175 175

[6] Bonello MA, Chryssanthopoulos MK, Dowling PJ. Ultimate strength design of stiffened plates under
axial compression and bending. Marine Structures 1993;6:533–52.
[7] Carlsen CA. Simplified collapse analysis of stiffened plates. Norwegian Maritime Research
1977;4:20–36.
[8] Smith CS. Strength of stiffened plating under combined compression and lateral pressure. Trans-
actions of the Royal Institution of Naval Architects 1991;133:131–47.
[9] Hibbit et al. ABAQUS/Standard. Version 5.4, Hibbitt, Karlsson and Sorensen, Inc: Pawtucket, RI,
1994.
[10] Riks E. An incremental approach to the solution of snapping and buckling problems. International
Journal of Solids and Structures 1979;15:529–51.
[11] Danielson DA, Kihl DP, Hodges DH. Tripping of thin-walled plating stiffeners in axial compression.
Thin-Walled Structures 1990;10:121–42.
[12] Carlsen CA, Czujko J. The specification of post-welding distortion tolerances for stiffened plates in
compression. The Structural Engineer 1978;56(5):133–41.
[13] Smith CS, et al. Strength and stiffness of ship plating under in-plane compression and tension. The
Royal Institute of Naval Architects, 1987.
[14] Faulkner D. A review of effective plating for use in the analysis of stiffened plating in bending and
compression. Journal of Ship Research 1975;19(1):1–17.
[15] Faulkner D, Adamchak JC, Snyder GJ, Vetter MF. Synthesis of welded grillages to withstand com-
pression and normal loads. Computers and Structures 1973;3:221–46.
[16] Carlsen CA. A parametric study of collapse of stiffened plates in compression. The Structural Engin-
eer 1980;58(2):33–40.
[17] Det norske Veritas. Buckling strength analysis of mobile offshore units. Classification Note 30.1,
October, 1987.
[18] American Petroleum Institute. Bulletin on design of flat plate structures. API Bulleting 2V, 1st Ed.,
May 1, 1987.

Potrebbero piacerti anche