Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Abstract
This paper presents an analytical method to calculate the buckling
stress of a rectangular thin plate under nonuniform applied axial
stresses. Two cases are considered, buckling of a plate simply
supported on all four sides and buckling of a plate simply supported on
three sides with one unloaded edge free and the opposite unloaded edge
rotationally restrained. These two cases illustrate the influence of stress
(moment) gradient on stiffened and unstiffened elements, respectively.
The axial stress gradient is equilibrated by shear forces along the
supported edges. A Rayleigh-Ritz solution with an assumed deflection
function as a combination of a polynomial and trigonometric series is
employed. Finite element analysis using ABAQUS validates the
analytical model derived herein. The results help establish a better
understanding of the stress gradient effect on typical thin plates and are
intended to lead to the development of design provisions to account for
the influence of moment gradient on local and distortional buckling of
thin-walled beams.
fix
ss ss
m=1 m=2 m=3 m=4 m=5 m=6 m=7 m=8 m=9 m=10 fix
ss
k ss ss
ss
m=1 m=2 m=3 m=4 m=5 m=6 m=7
fix
ss ss
free
m=1 m=2 m=3 m=4 ss
m=2 m=3 ss ss
m=1
free
ß
Figure 1 Buckling of uniformly compressed rectangular plates
Figure 1 indicates that when the unloaded edges are supported, the
length of the buckled wave is quite short and many buckled waves
(high m) can form in even relatively short lengths. When one of the
unloaded edges is unsupported (e.g. fix-free) the behavior is modified
from the supported case and now even at relatively large β values the
number of expected half-waves (m) are small. The behavior of the ss-
free case is particularly interesting. Instead of the distinct garland
curves of the earlier cases, now for higher β, the single half-wave case
(m=1) asymptotes to k=0.425 instead of increasing for large β. For the
ss-free case multiple wavelengths (m) all yield similar solutions for
large β.
To connect the plate solutions of Figure 1 to actual beams, consider the
top flange of the beams of Figure 2. Local buckling of the compression
flange of the hat of Figure 2(a) is somewhere between the ss-ss and fix-
fix case of Figure 1. What would be the influence of moment gradient
on the local buckling of the compression flange in this solution? One
anticipates that even a sharply varying moment gradient is unlikely to
change the stress demands on the flange significantly. Therefore,
traditional local buckling of a stiffened element – such as the
compression flange of the hat – is assumed to not require modification
due to the moment gradient. However, now consider the channel of
Figure 2(b). As the compression flange buckles the web/flange juncture
provides support somewhere between the fix-free and ss-free case of
Figure 1. The potential that moment gradient may have influence on the
buckling results is real. Further, and somewhat counter-intuitively, the
weaker the support on the unloaded edge, the greater the potential for
an increase in the buckling load due to the moment gradient.
σ min
osimply supported b
σ max
τ
Y
Figure 4 Stiffened element subject to a stress gradient
τ
o X
τ τ
σ min
o
elastic restrained edge
σ max
b
a
free edge
Y
Figure 5 Unstiffened element subject to a stress gradient
Via (4) the equilibrium equations are derived as Equation (6) (same as
Equation A11 in Libove, Ferdman and Reusch 1949).
π2 1 i2
2
+ βj 2 − i 2 +
1−r
M
(
ip i 2 + p 2 )
wij
8 k av β
1 + r p=1 ∑ w pj
(i 2
−p )
2 2
p ±i =odd
1− r
M N
∑ ∑
ijpq
− 4 =0
1 + r p=1 q=1,3,5,...
w pq 2
( )(
q − j2 i2 − p2 ) (6)
p ± i =odd
q≠ j
i = 1, 2, 3, ... M ; j = 1, 3, 5, ... N
σ min
where r = (7)
σ max
Revisiting their results we can provide Figure 6 and Table 2 to
quantitatively demonstrate the influence of stress gradient on stiffened
elements. It can be seen that the stress gradient increases the buckling
load at the maximum loaded edge and the influence is vanishing when
the aspect ratio of the plate (β) becomes large.
σ max
ss
ss
σ max
ss
π D
2
(σ max ) cr = k max
b2t
ss
σ
σ max
ss
σ
ss
Figure 6 k max vs. plate aspect ratio (ß) for stiffened element
Consistent with our intuition from Figure 1, we can conclude that the
influence of stress gradient diminishes quickly for local buckling of a
stiffened element, but perhaps not as quickly as is generally assumed in
design. The k max values of Table 1 may be used to predict the increased
local buckling stress due to the influence of a stress gradient, and for
continuous beams where sharp moment gradients are more likely to
persist, the boost may be significant. We now turn to the more
complicated and interesting problem of the stability of unstiffened
elements under a stress gradient.
Table 1 Numerical results of k max values for ss-ss stiffened elements
r
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
β
1.0 18.01815.06612.582 10.528 8.863 7.533 6.480 5.645 4.979 4.441 4.000
1.2 14.72112.75011.063 9.628 8.410 7.375 6.497 5.751 5.119 4.585 4.134
1.4 12.51911.103 9.877 8.818 7.905 7.117 6.436 5.846 5.333 4.884 4.470
1.6 10.963 9.884 8.935 8.100 7.364 6.712 6.128 5.597 5.105 4.640 4.202
1.8 9.833 8.975 8.210 7.525 6.910 6.353 5.841 5.363 4.907 4.465 4.045
2.0 8.993 8.291 7.656 7.082 6.558 6.079 5.635 5.215 4.810 4.405 4.000
2.5 7.655 7.187 6.754 6.355 5.984 5.638 5.314 5.009 4.718 4.437 4.134
3.0 6.893 6.549 6.227 5.924 5.639 5.368 5.109 4.858 4.608 4.341 4.000
3.5 6.409 6.139 5.882 5.639 5.406 5.182 4.965 4.752 4.540 4.321 4.072
4.0 6.073 5.851 5.639 5.434 5.237 5.045 4.858 4.673 4.486 4.284 4.000
4.5 5.827 5.639 5.456 5.280 5.109 4.941 4.776 4.611 4.443 4.264 4.045
5.0 5.638 5.474 5.315 5.160 5.008 4.858 4.710 4.561 4.408 4.243 4.000
5.5 5.489 5.344 5.202 5.063 4.926 4.791 4.656 4.520 4.380 4.227 4.030
6.0 5.367 5.237 5.109 4.983 4.858 4.735 4.611 4.486 4.355 4.213 4.000
6.5 5.267 5.148 5.031 4.915 4.801 4.687 4.573 4.456 4.335 4.201 4.022
7.0 5.181 5.072 4.965 4.858 4.752 4.646 4.540 4.431 4.317 4.191 4.000
7.5 5.108 5.007 4.908 4.809 4.710 4.611 4.511 4.408 4.301 4.182 4.017
8.0 5.045 4.951 4.858 4.765 4.673 4.580 4.486 4.389 4.287 4.174 4.000
8.5 4.990 4.902 4.814 4.727 4.640 4.552 4.463 4.371 4.274 4.166 4.013
9.0 4.941 4.858 4.776 4.694 4.611 4.528 4.443 4.355 4.263 4.160 4.000
9.5 4.898 4.820 4.742 4.663 4.585 4.506 4.425 4.341 4.253 4.154 4.011
10.0 4.860 4.785 4.711 4.636 4.561 4.486 4.408 4.328 4.243 4.148 4.000
11.0 4.791 4.723 4.656 4.588 4.520 4.451 4.380 4.306 4.227 4.139 4.000
12.0 4.735 4.673 4.611 4.549 4.486 4.421 4.355 4.287 4.213 4.130 4.000
13.0 4.687 4.630 4.573 4.515 4.456 4.396 4.335 4.270 4.201 4.123 4.000
14.0 4.646 4.593 4.540 4.486 4.431 4.375 4.317 4.256 4.191 4.117 4.000
15.0 4.611 4.561 4.511 4.460 4.408 4.355 4.301 4.243 4.182 4.112 4.000
16.0 4.580 4.533 4.486 4.438 4.389 4.339 4.287 4.232 4.174 4.107 4.000
17.0 4.552 4.508 4.463 4.418 4.371 4.324 4.274 4.222 4.166 4.102 4.000
18.0 4.528 4.486 4.443 4.400 4.355 4.310 4.263 4.213 4.160 4.098 4.000
19.0 4.506 4.465 4.425 4.383 4.341 4.298 4.253 4.205 4.154 4.095 4.000
20.0 4.486 4.447 4.408 4.369 4.328 4.287 4.243 4.198 4.148 4.091 4.000
21.0 4.467 4.431 4.393 4.355 4.317 4.277 4.235 4.191 4.143 4.088 4.000
22.0 4.451 4.415 4.380 4.343 4.306 4.267 4.227 4.185 4.139 4.086 4.000
23.0 4.435 4.402 4.367 4.332 4.296 4.259 4.220 4.179 4.134 4.083 4.000
24.0 4.421 4.389 4.355 4.322 4.287 4.251 4.213 4.174 4.130 4.081 4.000
25.0 4.408 4.377 4.345 4.312 4.278 4.243 4.207 4.169 4.127 4.078 4.000
26.0 4.396 4.366 4.335 4.303 4.270 4.237 4.201 4.164 4.123 4.076 4.000
27.0 4.385 4.355 4.325 4.295 4.263 4.230 4.196 4.160 4.120 4.074 4.000
28.0 4.375 4.346 4.317 4.287 4.256 4.224 4.191 4.156 4.117 4.073 4.000
29.0 4.365 4.337 4.308 4.279 4.250 4.219 4.186 4.152 4.114 4.071 4.000
30.0 4.355 4.328 4.301 4.273 4.243 4.213 4.182 4.148 4.112 4.069 4.000
Stress Gradient on Unstiffened Elements
MA MB
MB
MA
Double Curvature Bending
MB
MA
∫
S
U2 = dx (9)
2 ∂y y =0
0
The uniform shear stress at the edge y=0 can be determined by force
equilibrium in the x direction, Equation (9) is the result. In addition,
along the free edge no stress exists, Equation (10).
(σ max − σ min )b
τ xy | y= 0 = (10)
a
τ xy | y=b =0 (11)
where the body force has been neglected. By substituting Equation (11)
into Equations (12) and (13) along with the fact that
σ x |x =0 = σ min = r σ max , the compressive stresses can be obtained in
functional form as:
(1 − r ) x
σ x = σ max + r (15)
a
σy =0 (16)
Deflection function and boundary conditions
The Rayleigh-Ritz method is an approximate approach and the
accuracy of the results depends on how closely the assumed deflection
function w(x,y) describes the true deflection. In general, the selected
deflection function should satisfy the boundary conditions of the plate.
For the model of Figure 5, a total of six boundary conditions are
observed and stated below.
Simply supported at the transverse, loaded edges:
( w) x =0 = 0 (17)
( w) x =a = 0 (18)
Elastic restraint against rotation along one supported longitudinal edge:
( w) y= 0 = 0 (19)
∂ w
2
∂ w 2
∂w
D 2 + µ 2 = S (20)
∂y ∂x y =0 ∂y y = 0
One free longitudinal edge:
∂2w ∂ 2 w
D 2 + µ =0 (21)
∂y ∂x 2 y=b
∂ 3w ∂3w
D 3 + ( 2 − µ ) 2 =0 (22)
∂ y ∂ x ∂y y=b
∑ A ( y ) B ( x) ,
i =1
i i where Ai is a function of y alone and Bi is a
function of x alone. If, for every i, Ai (y)Bi (x) is compatible with all the
boundary conditions, then the linear summation of each term Ai (y)Bi (x)
will satisfy the boundary conditions. In this paper, three different
approximate deflection functions are proposed and examined.
Deflection function 1
The first deflection function considered, Equation (22), is motivated by
the work of Lundquist and Stowell (1942) who explored the buckling
of unstiffened element subject to uniform compressive stresses.
Lundquist and Stowell employed a trigonometric term in the
longitudinal direction and a polynomial term in the transverse direction
to establish the deflection function, as below:
y y 5 y
4
y
3
y π x
2
w = A + B + a1 + a2 + a3 sin (23)
b b b b b a
(26)
The equilibrium equations can be constructed by substituting the
deflection function into the total potential energy expression Equations
(7) and taking the derivative of the function as Equation (4). The
buckling stress s max is the minimum eigenvalue of the resulting matrix
of equilibrium expressions. The size of matrix is N × N .
It should be noted that deflection function 1 is not compatible with
boundary the conditions, Equations (20) and (21). However, it was
found to give reasonable results in the research done by Lundquist and
Stowell (1942) and was thus considered here.
Deflection function 2
Unlike deflection function 1 which is based on a physical
representation of the expected shape, deflection function 2 is a more
general combination of polynomials and trigonometric functions. A
fourth order polynomial is assumed for the transverse deflection and a
sin function for the longitudinal deflection as given in Equation (26).
∑ w (c ) iπx
N
w= i i1 y + ci2 y 2 + ci3 y 3 + c i4 y 4 sin (27)
i =1 a
The deflection function must satisfy all the boundary conditions.
Therefore, the parameters c i1 , ci 2 , c i 3 , ci 4 are determined by
substituting Equation 26 into the boundary conditions of Equations (16)
to (21), resulting in:
c i1 = 1 (28)
S
ci 2 = (29)
2D
i 2π 2b i π b S
2 2 2
i π
2 2
i π bS
2 2
ci 3 = µ 2 2 + µ 2 2 − (2 − µ ) 2 − (2 − µ ) 2
a k k
1 3 2a k1 k3 D a k1 k 2 k3 a k1 k 2 k3 D
S i 2π 2 b 4 S i 2π 2b 2
− 2 × µ − 12 b 2
+ µ
− 2
k1 k3 D 2 2 k D 2
a 1 a
(30)
i 2π 2b i 2π 2b2 S i2π 2 i 2π 2bS S
ci 4 = µ 2 +µ 2 − (2 − µ) 2 − ( 2 − µ) 2 −
a k1k3 2a k1k 3D a k2 k3 a k2 k3D k1k3D
(31)
i = 1, 2, K N
where
i 2π 2 b 3
k 1 = 6b − µ (32)
a2
3i 2π 2b 2
k 2 = 6 − (2 − µ ) (33)
a2
12b 2 i 2π 2 b 4 24b 4i 2π 2 b 3
k3 = −µ − + (2 − µ ) (34)
k1 a 2 k1 k2 a 2k2
The equilibrium equations can be constructed by the same method as
described for deflection function 1, the buckling stress (s max)cr is the
minimum eigenvalue of the resulting N × N matrix of equilibrium
expressions.
Deflection function 3
As an extension of the second deflection function a third deflection
function using a 5th order polynomial in the transverse direction was
also considered:
∑[w (p ) ( )] iπx
N
w= i1 i1 y + pi 2 y
2
+ pi3 y3 + pi4 y 4 + wi2 qi1y 3 + qi2 y 4 + qi3 y 5 sin
i =1 a
(35)
where wi 1 , wi 2 are arbitrary deflection amplitudes. The parameters p
and q were determined by the boundary conditions.
pi1 = 1 (36)
S
p i2 = (37)
2D
i 2π 2b S i 2π 2b 2 i 2π 2b 4 M 1
pi 3 = µ + µ − 2 +µ (38)
a 2 k1 2k1 D a2
a 2 k1
pi 4 = M 1 (39)
i 2π 2b 5 b3 i 2π 2b 4 M 2 12b 2 M 2
q i1 = µ 2
− 20 +µ − (40)
a k1 k1 a 2 k1 k1
qi2 = M 2 (41)
q i3 = 1 (42)
where
i 2π 2 b3
k1 = 6b − µ (43)
a2
3i 2π 2b 2
k 2 = 6 − (2 − µ ) (44)
a2
i 2π 2b 4
2 1
4i 2π 2b 3 1
k3 = µ − 12 b − ( 2 − v) − 24 b (45)
2 2 k
a k1 a 2
i 2π 2 i 2π 2 bS i 2π 2 b i 2π 2b 2 S S
M 1 = (2 − µ ) + ( 2 − µ ) − µ − µ +
a 2 k 2 k3 k2 k 3 D a 2 k1 k 3 2k1k 3 D Dk 1k3
(46)
5i 2π 2 b 4 60 b2 i 2π 2 b5 20b3
M 2 = (2 − µ ) 2
− −µ 2 + (47)
a k 2 k3 k2k3 a k1 k 3 k1 k 3
max ss
free
σ max ss
π 2D
(σ max ) cr = k max
b2 t
free
max ss
free
Figure 9 k max vs. plate aspect ratio (ß) for ss-free unstiffened element
Figure 11 shows the relation between the plate buckling coefficient k max
and the stress gradient factor r for a ss-free unstiffened element.
Figure 11 indicates that the increase in the plate buckling coefficient
k max can be assumed a linear function with r except for approximately
square (β =1) plates. For design, we intend to provide simple empirical
relations for k max(r, β, S)
max
fix
free
max fix
π 2D
(σ max ) cr = k max
b2 t
free
max
fix
free
Figure 10 k max vs. plate aspect ratio (ß) for fix-free unstiffened element
Figure 11 k max vs. stress gradient factor r for ss-free unstiffened element
Discussion
σmax