Professional Documents
Culture Documents
1043
__________________________________________________________________________
STEEL CONSTRUCTION:
APPLIED STABILITY
1. INTRODUCTION
The influence of imperfections, material plasticity and strain hardening on ultimate
resistance, has been reviewed in detail in the previous Lecture 6.6.1 with reference to
column flexural buckling. This is the form of member instability that has been most
studied and an almost complete understanding of the phenomenon has been achieved.
2. BUCKLING OF PLATES
Plate buckling occurs where thin-walled sections experience compressive stresses. This
phenomenon is quite similar to column buckling; it involves, however, elements of the
member cross-section rather than displacement of the section as a whole. Compressive
stresses may arise not only from compressive axial load but also from bending of the
member, and even from patch loads (Figures 1a, 1b, and 1c). Local buckling may also
occur in plates subject to shear (Figure 1d), because shear results in compressive and
tensile principal stresses.
Cold-formed sections and thin plated sections are the most sensitive to local buckling.
The elastic buckling modes and corresponding critical stresses of compressed plates are
derived from the small deflected plate equations; their analysis has been carried out in
Lecture 6.3.
It has long been recognised that the critical buckling load of a plate is not a satisfactory
measure of the actual plate resistance. Indeed, as soon as the plate tends to buckle, out-ofplane deflections result in membrane forces due to the stretching of the mid-plane of the
plate. To understand this behaviour let us consider an initially flat rectangular plate,
having all edges either simply supported, or clamped, and loaded in uniform uniaxial
compression. Let us assume, in addition, that the aspect ratio is such that the plate will
buckle in a single half-wave mode (Figure 2). For the purpose of this analysis the plate can
be replaced by a system of strips in both orthogonal directions; the longitudinal strips are
compressed and are thus liable to buckle. Below the elastic critical plate buckling load, all
the longitudinal strips are straight because the plate remains flat; they maintain their initial
stiffness and experience equal compression stress. Once the plate buckles, the longitudinal
strips close to the edge are constrained to remain straight while those away from these
edges are more prone to buckling; in other words the first ones retain their axial stiffness
while the latter lose a part of their initial stiffness.
The plate buckling causes an increase in length of the transverse strips because the
longitudinal edges are restrained against free shortening of the plate in the transverse
direction; this causes tensile membrane forces in the transverse strips, that exert a
stabilising effect on the longitudinal strips (Figure 3). This explains why stable
equilibrium states can be reached by the plate beyond the elastic critical plate buckling
load, with the result that large post-buckling resistance may be possible provided the
material does not yield prematurely. This is the fundamental difference between plate
behaviour and column buckling, which is not able to exhibit such a reserve of resistance.
(1)
where k is the plate buckling coefficient and the Poisson ratio. Due to material
plasticity, the squash load provides a limit to plate resistance and constitutes a cut-off limit
to the elastic buckling curve. As explained above, however, the ultimate average stress of
a slender plate, i.e. a plate for which cr is lower than the yield stress fy, exceeds the elastic
critical stress as a result of the post-buckling resistance; the ultimate stress curve thus lies
(2)
For a simply supported long plate, subject to uniform compression, the buckling
coefficient is roughly k =4; the corresponding value of 1 = 1,9
amounts to
between 56 and 46 for the usual steel grades S235 and S355. The value of the buckling
coefficient k, for other cases, depends on the plate aspect ratio, the type of loading and the
boundary conditions.
The existence of post-buckling resistance in a plate subject to shear can be physically
explained as follows: a square element, the edges of which are oriented at 45 to the plate
edges, experiences tensile stresses, 1, on two opposite edges and compressive stresses,
2, on the other two (Figure 7a); these compressive stresses can, therefore, induce plate
buckling in the form of elongated bulges oriented in the direction of the tensile stresses
(Figure 7b). The elastic critical shear stress, cr, is given by an expression similar to cr;
again cr depends on the aspect ratio and the edge support conditions. Once the plate
buckles in shear, the increase in compressive stresses is no longer possible while the initial
= (fy/cr)
(3)
= (b/t)(1,05/k)fy/E
(4)
Because torsion is involved in such a buckling process, the elastic critical buckling stress
depends partly on the Saint-Venant torsional rigidity and on the warping rigidity. In
addition, the end torsional and warping restraints must be reflected in the numerical
factors used.
Torsional buckling can only properly occur when the shear centre and centroid of the
cross-section are coincident, i.e, only in symmetrical sections; thin-walled components
will also obviously favour this form of buckling. Because flexural buckling (usually
termed column buckling) is still possible, each form must be examined to determine which
gives the lower critical stress.
As for flexural buckling, the elastic critical torsional buckling load is not an adequate
measure of the carrying resistance. One must again allow for geometric imperfections,
residual stresses and for the effects of material plasticity and strain hardening. Regarding
= (fy/cr,T)
(5)
4. FLEXURAL-TORSIONAL BUCKLING
Flexural and torsional displacements are combined when the shear centre and centroid of
the section are not coincident. This results in a reduction of the critical buckling load
compared to that associated with flexural buckling. This phenomenon is particularly
important for angles and channel sections.
To illustrate this form of buckling, a channel section is considered subject to a uniaxial
compressive load at the centroid, so that the end sections experience a uniform applied
stress (Figure 9a). The section is assumed to be proportioned so that the second moment of
inertia, Iyy, about yy axis, is larger than that about zz axis; the flexural buckling about the
zz axis is thus governing (Figure 9b). As soon as the strut tends to buckle in the direction
yy, i.e. by bending about the zz axis, its deflection v(x) produces a bending moment M(x)
= Nv(x), giving rise to an associated shear, V = dM/dx = N(dv(x)/dx). Everything happens
as if the strut was subject to transverse distributed forces q=-dV(x)/dx =-N(d2v(x)/dx2),
acting in the plane yy containing the centroid G, and not in the plane zz containing the
shear centre C (Figure 9c). Therefore bending about the zz axis and the corresponding
shear induce a torque moment dMt = qzo dx, where zo is the distance between the centroid
and the shear centre. This means that purely flexural buckling in the yy plane is
impossible, because bending and torque deformations are inevitably coupled. This form of
buckling is termed "flexural-torsional buckling".
Flexural-torsional buckling is governed by three equilibrium differential equations
containing the derivatives of the three functions: displacements v and w in the yy and zz
planes and the twist . The flexural-torsional buckling load thus depends on the usual
boundary conditions for v and w at the ends (pin-ended or clamped) and on the twist and
warping restraints on .
In the usual case of a pin-ended strut, which is fully free to warp, i.e.v=v"=0,w=w"=0 and
= " = 0 at the ends, the critical flexural-torsional buckling stress is as follows:
cr,FT={cr,zz + cr,T {(cr,T - cr,zz)2 + cr,zz . cr,T (zo/ro)2 ]} / {2[1 - (zo/ro)2]}
(6)
= (fy/cr,FT)
(7)
. C1
(8)
where C1 and C2 are coefficients allowing for the influence of the aforementioned factors
and zg is the distance from the shear centre to the point of application of the transverse
loads. The elastic critical lateral-torsional buckling stress is as follows:
cr,LT = Mcr,LT/ Wy
(9)
where Wy - is the elastic section modulus for bending about the strong axis.
In addition to material plasticity and residual stresses, geometric imperfections cause a
loss in resistance compared to the elastic critical load. Any imperfection which is liable to
trigger off torsion and/or lateral bending is obviously of concern. A recent statistical
evaluation of test results, carried out when preparing background documents for Eurocode
3 [1], has demonstrated that the format of the ultimate resistance curves for flexural
= (fy/cr,LT)
(10)
6. BUCKLING OF SHELLS
Shell buckling is probably the most complex instability phenomenon. Similar to plate
buckling, it involves the cross-sectional shape rather than displacement of the section as a
whole. Compressive stresses, as before, can arise from compressive axial load, bending
and locally applied concentrated loads. In addition, however, shells can also be subject to
internal or external pressures (pipes, silos, tanks, off-shore jacket legs, etc.).
The simplest case to analyse is the axially compressed cylindrical shell (Figure 11); the
elastic critical buckling stress is as follows:
cr = [1/
] Et/r
(11)
where t is the wall thickness, and r the radius of the cylinder. This formula is well known
in the format, cr = 0,605 Et/r, valid for steel cylinders.
The load-shortening behaviour shows a drastic loss of carrying resistance as soon as the
critical buckling load is reached (Figure 12); when it buckles, the shell tends to a buckled
shape, which is in equilibrium with an external compression load usually much lower than
the critical load. This is a general characteristic of shells, whatever the type of shell and
the type of loading; unlike plates, they are not able to exhibit any post-buckling reserve of
resistance.
Even very small geometric imperfections cause a premature deviation from the load displacement curve of the perfect shell usually giving a significant reduction in the load
carrying resistance (Figure 12). Elastic critical loads cannot, therefore, be attained by real
shells. The influence of imperfections on the ultimate load of shells is a very complex
problem which has been extensively researched.
According to classical theory of shell buckling, the perfect cylindrical shell can buckle
either axisymmetrically - with a succession of ring buckles - or in the form of a chessboard
(Figure 13a, b); depending on the buckling circumstances, the shell may achieve a
diamond-like post-buckling configuration. To account for imperfections, design rules
traditionally use a knock-down factor, , affecting the elastic critical shell buckling stress;
values of are derived from tests conducted on large scale models.
According to ECCS Recommendations [2], the value of this factor depends only on the
slenderness, r/t, of the shell when imperfections are kept below a specified level. An
additional partial safety factor, , is applied in the case of cylindrical shells subject to
meridional compression because of their especially unfavourable post-buckling behaviour.
The influence of the different types of imperfections on the ultimate resistance may vary
greatly depending on the type of shell and on the type of loading. A normalised resistance
curve is then defined by plotting the ratio /fy against the normalised slenderness s ( s
= (fy/cr), similar to the approach used for column and plate resistance curves (Figure
14).
8. FRAME INSTABILITY
When designing columns as parts of a frame, it would be convenient to isolate the
columns from the remainder of the frame and to treat their design as separate problems.
However most columns belonging to building frames experience bending actions applied
at their ends, in addition to axial loads. These bending actions depend on the interaction
between the column and its adjacent members -beams and columns. In some cases, the
column will support the beams at failure; such situations occur when the beams are
designed plastically and collapse while the column remains stable and elastic. There are
other situations where the beams will support the column; that is especially the case when
the beams are designed elastically and exhibit an elastic behaviour up to collapse, the
latter being initiated by column instability.
9. CONCLUDING SUMMARY
The basic features of the buckling phenomena of the members which have been
examined are summarised in the following table.
The most effective ways to improve buckling resistance are to increase the
dimensions of the cross-section and/or to use adequate bracing and restraints to
modify the buckling mode.
The stability of a no-sway frame is controlled by the stability of its individual
columns.
The stability of a sway frame is controlled by the bending stiffness of the columns
and beams and the rigidity of the beam/column connections.
Type of
member
Loading
Stiffness
parameters
Slenderness Normalised
ratio
slenderness
Flexural
Columns
Axial
compression
EIy or EIz
L/ix or L/iy
Torsional
Columns
(open sections
with double or
point
symmetry)
Axial
compression
GJ, EI
Flexural
torsional
Columns
(open sections
without
double or
point
symmetry)
Axial
compression
GJ, EI
Beams
Major axis
bending
Lateral
torsional
EIy and/or
EIz
GJ, EI ,
EIz
Plate
Thin plate or
member
component
Compression
or shear
Shell
Cylindrical
shell
Axial
compression
b/t
r/t
10. REFERENCES
[1] Eurocode 3: "Design of Steel Structures": ENV 1993-1-1: Part 1.1: General rules and
rules for buildings, CEN, 1992.
[2] European Convention for Constructional Steelwork, Recommendations of Steel Shells,
Publication 56, ECCS, 1988.