Professional Documents
Culture Documents
www.elsevier.com/locate/tws
A semi-analytical model for global buckling
and postbuckling analysis of stiened panels
Eirik Byklum
a,
, Eivind Steen
a
, Jrgen Amdahl
b
a
Det Norske Veritas, Maritime Technology and Production Centre, Veritasveien 1,
N-1322 Hvik, Norway
b
Department of Marine Structures, Norwegian University of Science and Technology,
N-7491 Trondheim, Norway
Received 27 February 2003; received in revised form 24 November 2003; accepted 23 December 2003
Abstract
A computational model for global buckling and postbuckling analysis of stiened panels
is derived. The loads considered are biaxial in-plane compression or tension, shear, and lat-
eral pressure. Deections are assumed in the form of trigonometric function series, and the
principle of stationary potential energy is used for deriving the equilibrium equations. Lat-
eral pressure is accounted for by taking the deection as a combination of a clamped and a
simply supported deection mode. The global buckling model is based on Marguerres non-
linear plate theory, by deriving a set of anisotropic stiness coecients to account for the
plate stiening. Local buckling is treated in a separate local model developed previously.
The anisotropic stiness coecients used in the global model are derived from the local
analysis. Together, the two models provide a tool for buckling assessment of stiened
panels. Implemented in the computer code PULS, developed at Det Norske Veritas, local
and global stresses are combined in an incremental procedure. Ultimate limit state estimates
for design are obtained by calculating the stresses at certain critical points, and using the
onset of yielding due to membrane stress as the limiting criterion.
# 2004 Elsevier Ltd. All rights reserved.
Keywords: Stiened panel; Global deection; Buckling; Postbuckling; Ultimate strength; Analytical
model; Energy principles
_
_
C
11
C
12
C
13
Q
11
Q
12
Q
13
C
21
C
22
C
23
Q
21
Q
22
Q
23
C
31
C
32
C
33
Q
31
Q
32
Q
33
Q
11
Q
21
Q
31
D
11
D
12
D
13
Q
12
Q
22
Q
32
D
21
D
22
D
23
Q
13
Q
23
Q
33
D
31
D
32
D
33
_
_
_
_
De
1
De
2
De
3
Dj
1
Dj
2
Dj
3
_
_
_
_
2
The stiness matrix is symmetric, so that C
ij
C
ji
and D
ij
D
ji
. The stiness coef-
cients may be divided into a linear and a nonlinear part:
C
ij
C
L
ij
C
NL
ij
3
D
ij
D
L
ij
D
NL
ij
4
Q
ij
Q
L
ij
Q
NL
ij
5
The linear parts are the ones corresponding to classical orthotropic stiness coe-
cients, and these are independent of load and displacement. The nonlinear parts
are calculated using the local buckling model, and will be functions of load and
displacement. For compressive loads they will be negative, resulting in a stiness
reduction.
3. Derivation of reduced stiness
For calculation of the reduced stiness coecients to be used in the global buck-
ling model, the total forces are written as follows:
N
M
_ _
C Q
Q
T
D
_ _
L
e
j
_ _
g
N
A
mn
g
M
A
mn
_ _
6
where g
N
(A
mn
) and g
M
(A
mn
) are nonlinear terms, which are due to local buckling
eects. The reduced stiness coecients are then dened as:
C
ij
@N
i
@e
j
C
L
ij
@g
N
i
@e
j
7
Q
ij
@N
i
@j
j
Q
L
ij
@g
N
i
@j
j
8
@M
j
@e
i
Q
L
ij
@g
M
j
@e
i
9
D
ij
@M
i
@j
j
M
L
ij
@g
M
i
@j
j
10
The reduced stiness coecients are derived using the local buckling model pre-
sented in [1]. First, the resultant forces and moments are calculated by integration
705 E. Byklum et al. / Thin-Walled Structures 42 (2004) 701717
of the membrane stress:
N
i
_
h
r
ii
dz 11
M
i
_
h
zr
ii
dz 12
The neutral axis of the stiener is not known, since it is continuously changing
during buckling. The bending moment is therefore calculated about the middle
plate plane. The neutral axis and the neutral bending stiness coecients can be
calculated once the in-plane and bending stiness are determined.
For open prole stieners, the following expressions were calculated in [1] for
the internal axial and transverse force:
N
1
E
A
T
b
1
m
2
bt
A
T
1 m
2
_ _
e
1
mtE
1 m
2
e
2
EA
s
z
gs
b
j
1
g
N
1
13
N
2
mtE
1 m
2
e
1
tE
1 m
2
e
2
g
N
2
14
where A
T
is the total cross-sectional area, A
s
is the stiener area, and z
gs
is the dis-
tance from the plate plane to the centroid of the stiener. The curvature j
1
is due
to global deection. The shear force and bending moment are:
N
3
Gte
3
g
N
3
15
M
1
EA
s
b
e
1
EI
b
j
1
g
M
1
16
where I is the moment of inertia of the whole cross-section. The stiness coe-
cients are found by dierentiation of the above expressions. The linear parts are
given directly as:
C
L
11
E
A
T
b
1
m
2
bt
A
T
1 m
2
_ _
17
C
L
12
C
L
21
mtE
1 m
2
18
C
L
22
tE
1 m
2
e
2
19
C
L
33
Gt 20
Q
L
11
EA
s
z
gs
b
21
D
L
11
EI
b
22
E. Byklum et al. / Thin-Walled Structures 42 (2004) 701717 706
All other linear coecients are zero for open prole stieners. The nonlinear
parts of the stiness coecients are calculated as:
C
NL
ij
@g
N
i
@w
L
@w
L
@e
j
23
Q
NL
ij
@g
N
i
@w
L
@w
L
@j
j
24
D
NL
ij
@g
M
i
@w
L
@w
L
@j
j
25
where w
L
is the local deection. The rst part may be found directly by dieren-
tiation once g
N
and g
M
are known. They can be calculated from expressions for N
i
and M
i
derived for the local model in [1]. The second part is calculated using the
equilibrium equations for the local buckling problem. By applying partial dieren-
tiation to the stationary potential energy, we get:
@@P=@w
L
@e
i
@
2
P
@w
2
L
@w
L
@e
i
@
2
P
@w
L
@e
i
0 26
@@P=@w
L
@j
1
@
2
P
@w
2
L
@w
L
@j
1
@
2
P
@w
L
@j
1
0 27
By introducing the incremental stiness matrix K and load vectors G, dened as
K
@
2
P
@w
2
L
28
G
e
i
@
2
P
@w
L
@e
i
29
G
j
1
@
2
P
@w
L
@j
1
30
we can write:
K
@w
L
@e
i
G
e
i
0 31
K
@w
L
@j
1
G
j
1
0 32
This means that
@w
L
@e
i
K
1
G
e
i
33
and
@w
L
@j
1
K
1
G
j
1
34
707 E. Byklum et al. / Thin-Walled Structures 42 (2004) 701717
An example of how the stiness coecients may change during local deformation
is given in Fig. 4. The calculations are for a steel angle bar (Table 1). The load is
axial, and the imperfection is 1 mm in the local eigenmode.
The values plotted are the ratio between the nonlinear stiness coecients and
the corresponding initial values. Q
12
is divided by Q
11
since Q
L
12
is zero. The sti-
ness ratios are slightly smaller than 1.0 at the start of the analysis due to the imper-
fection. If the imperfection were larger, the stiness ratios would also have smaller
initial values.
It is seen that the stiness reduction is signicant for C
11
and C
22
, but the most
drastic change is for C
12
, which even changes sign. The reason is that C
12
is posi-
tive for a at plate due to the Poisson eect, while it becomes negative for large
deection due to membrane stretching. The change in Q
11
, Q
12
, and D
11
is small.
There is some reduction in D
n11
, which is the neutral bending stiness. This coef-
cient is dened in the next section.
It is seen that the stiness reduction is quite localized. The reduction occurs
around the buckling strain, and the stiness is almost constant after this. This is a
Fig. 4. Change in stiness properties during local buckling due to axial load for steel angle bar.
E. Byklum et al. / Thin-Walled Structures 42 (2004) 701717 708
general trend found for all the stieners investigated here. For smaller imperfec-
tions the stiness reduction will be even more sudden, while for larger imperfec-
tions there will be a more gradual transition.
4. Global buckling model
For the following derivations, the stieners are assumed to be in the longitudinal
direction, but transverse stiening can be analyzed simply by switching panel
length and breadth. The stiened panel is supported on all edges by transverse and
longitudinal girders. The length of the panel is a, while the width is B. The loads
considered are in-plane compression or tension, shear force, and lateral pressure.
The edge loads are assumed to be constant in magnitude.
Two stiener spans and panel widths are included in the model. The intention is
to account properly for the eect of lateral pressure on the panel. The pressure
must be carried by the stieners, and the deection of the stieners may therefore
be a combination of the simply supported mode and the clamped mode (Fig. 5).
The deection shape is therefore taken as a combination of sine terms and cosine
terms:
w w
s
w
c
35
w
0
w
s
0
w
c
0
36
where s and c denote sine and cosine mode deection, respectively:
w
s
M
s
m1
N
s
n1
A
s
mn
sin
mpx
a
_ _
sin
npy
B
_ _
37
Fig. 5. Global deection of stiener in simply supported mode and clamped mode.
Table 1
Dimensions for stiened steel plate
Stiener a (m) b (m) t (m) h (m) t
w
(m) b
f
(m) t
f
(m) r
f
(MPa)
Angle bar 2.73 0.85 0.0165 0.350 0.012 0.100 0.017 355
709 E. Byklum et al. / Thin-Walled Structures 42 (2004) 701717
w
c
M
c
m1
N
c
n1
A
c
mn
2
1 cos
2mpx
a
_ _ _ _
sin
npy
B
_ _
38
w
s
0
M
s
m1
N
s
n1
B
s
mn
sin
mpx
a
_ _
sin
npy
B
_ _
39
w
c
0
M
c
m1
N
c
n1
B
c
mn
2
1 cos
2mpx
a
_ _ _ _
sin
npy
B
_ _
40
The anisotropic material law for plane stress, using the stress resultants N
1
, N
2
,
and N
3
, is:
N
1
N
2
N
3
_
_
_
_
C
11
C
12
C
13
C
21
C
22
C
23
C
31
C
32
C
33
_
_
_
_
e
1
e
2
c
3
_
_
_
_
41
The corresponding exibility relation is needed for derivation of the stress func-
tion. It is written as:
e
1
e
2
c
3
_
_
_
_
M
11
M
12
M
13
M
21
M
22
M
23
M
31
M
32
M
33
_
_
_
_
N
1
N
2
N
3
_
_
_
_
42
The stiness relation for the resultant bending moments is:
M
1
M
2
M
3
_
_
_
_
D
11
D
12
D
13
D
21
D
22
D
23
D
31
D
32
D
33
_
_
_
_
j
1
j
2
j
3
_
_
_
_
43
It is assumed that there is no coupling between resultant forces and moments, i.e.
all Q
ij
terms are zero. This can be done by performing a neutralization of the sti-
ness coecients, as explained in [2]. This means that the bending stiness coe-
cients are redened so that no coupling occurs. The neutral bending stiness matrix
2M
s
0
2N
s
0
f
s1
mn
cos
mpx
a
_ _
cos
npy
B
_ _
51
F
s2
2M
s
0
2N
s
0
f
s2
mn
sin
mpx
a
_ _
sin
npy
B
_ _
52
F
c1
2M
c
0
2N
c
0
f
c1
mn
cos
2mpx
a
_ _
cos
npy
B
_ _
53
F
c2
2M
c
0
2N
c
0
f
c2
mn
sin
2mpx
a
_ _
sin
npy
B
_ _
54
F
sc1
M
s
M
c
0
N
s
2N
c
0
f
sc1
mn
sin
mpx
a
_ _
cos
npy
B
_ _
55
F
sc2
M
s
M
c
0
N
s
2N
c
0
f
sc2
mn
cos
mpx
a
_ _
sin
npy
B
_ _
56
711 E. Byklum et al. / Thin-Walled Structures 42 (2004) 701717
By substitution of the assumed stress function into the compatibility equation, it
is found that the coecients f
1
mn
and f
2
mn
must be:
f
s1
mn
1
4a
2
B
2
K1s K2s
2
=K1s
rspq
b
s
rspq
A
s
rs
A
s
pq
A
s
rs
B
s
pq
A
s
pq
B
s
rs
_ _
57
f
s2
mn
K2s
K1s
f
s1
mn
58
f
c1
mn
1
4a
2
B
2
K1c K2c
2
=K1c
rspq
b
c
rspq
A
c
rs
A
c
pq
A
c
rs
B
c
pq
A
c
pq
B
c
rs
_ _
59
f
c2
mn
K2c
K1c
f
c1
mn
60
f
sc1
mn
1
4a
2
B
2
K1sc K2sc
2
=K1sc
rspq
b
sc
rspq
A
s
rs
A
c
pq
A
s
rs
B
c
pq
A
c
pq
B
s
rs
_ _
61
f
sc2
mn
K2sc
K1sc
f
sc1
mn
62
where
K1s
m
4
a
4
M
2222
m
2
n
2
a
2
B
2
2M
1122
M
1212
n
4
B
4
M
1111
63
K2s 2
m
3
n
a
3
B
M
2221
2
mn
3
aB
3
M
1112
64
K1c 16
m
4
a
4
M
2222
4
m
2
n
2
a
2
B
2
2M
1122
M
1212
n
4
B
4
M
1111
65
K2c 16
m
3
n
a
3
B
M
2221
4
mn
3
aB
3
M
1112
66
K1sc
m
4
a
4
M
2222
m
2
n
2
a
2
B
2
2M
1122
M
1212
n
4
B
4
M
1111
67
K2sc 2
m
3
n
a
3
B
M
2221
2
mn
3
aB
3
M
1112
68
and f
0,0
is dened as zero. The coecients b
s
rspq
, b
c
rspq
, and b
sc
rspq
can be found in [12].
The potential of internal energy is generally written as
U
1
2
_
V
re dV 69
The strain is rst divided into a constant membrane part e
m
and a linearly varying
bending part e
b
zj. Integration is then performed over the thickness in order to
E. Byklum et al. / Thin-Walled Structures 42 (2004) 701717 712
express the potential energy as a function of stress resultants:
U
1
2
_
V
re
m
e
b
dV 70
1
2
_
V
re
m
dV
1
2
_
V
zrj dV 71
1
2
_
A
Ne
m
dV
1
2
_
V
Mj dV 72
U
m
U
b
73
By substitution of the material law, the membrane energy is written as:
U
m
1
2
M
11
_
A
N
2
1
dA M
22
_
A
N
2
2
dA M
33
_
A
N
2
3
dA
_
2M
12
_
A
N
1
N
2
dA 2M
13
_
A
N
1
N
3
dA 2M
23
_
A
N
2
N
3
dA
_
74
The membrane energy is calculated by substitution of the stress function and inte-
grating over the plate area. The nal expression can be found in [12]. The bending
energy is:
U
b
1
2
D
11
_
A
j
2
1
dA D
22
_
A
j
2
2
dA D
33
_
A
j
2
3
dA
_
2D
12
_
A
j
1
j
2
dA 2D
13
_
A
j
1
j
3
dA 2D
23
_
A
j
2
j
3
dA
_
75
The resulting expression is found by substituting j
1
w
;xx
, j
2
w
;yy
, and
j
3
2w
;xy
, and performing the integration. The result is given in [12].
The energy due to in-plane tension or compression load is:
T
c
_
2a
0
_
2B
0
N
1
u
;x
dy dx
_
2a
0
_
2B
0
N
2
v
;y
dy dx 76
The shear energy is:
T
s
N
3
_
2a
0
_
2B
0
u
;y
v
;x
dy dx 77
The energy due to lateral pressure is:
T
lp
_
2a
0
_
2B
0
pw dy dx 78
It is seen that contribution from the sine deection to the lateral pressure energy
vanishes upon integration. This is due to the anti-symmetry of the sine deection.
Physically, this means that lateral pressure will only give rise to deection in the
713 E. Byklum et al. / Thin-Walled Structures 42 (2004) 701717
cosine mode. For combined loads, the deection will be a combination of the two.
All nal expressions can be found in [12].
5. Results
For verication of the computational model developed, analyses were performed
using the nonlinear nite element code ABAQUS.
In the rst case, the anisotropic material option was applied for the case of an
unstiened plate. This means that all the in-plane stiness coecients may be given
independently, while the bending stiness is given directly by integration over the
thickness.
Fig. 6 shows the nondimensional load-shortening response for an 840 980
11 mm aluminium plate with elastic modulus E 70 000 MPa and yield stress
r
f
240 MPa. A combination of lateral pressure p 0:2 MPa, corresponding to a
20 m water column, and transverse compression, S
y
240 MPa, is applied pro-
portionally. The combination of in-plane and out-of-plane loads gives a deection
mode in between simply supported and clamped. In order to check the model with
anisotropic stiness, the stiness was arbitrarily chosen so that C
1112
C
2221
C
iso
1212
. The imperfection is 10 mm in the eigenmode.
It is seen that this load combination results in a snap-back response. The reason
is that the imperfection and the transverse loading gives a deection in the simply
supported mode in the rst part of the analysis, while the lateral pressure eventu-
ally forces the deection into a clamped mode. It is seen that the agreement
between the results from the model and from ABAQUS is very good. This shows
that the rst order perturbation expansion works very well even for such a complex
Fig. 6. Loadaverage strain response for anisotropic plate subjected to combined transverse load and
lateral pressure.
E. Byklum et al. / Thin-Walled Structures 42 (2004) 701717 714
response. Also, the lateral pressure eect, which gives a resulting deection mode
in between simply supported and clamped, is well taken care of.
In order to check the bending stiness formulations, analyses are performed
using a stiened panel in ABAQUS. It is desirable to compare the model with a
stiened panel deecting in a pure global mode, i.e. without local buckling of plate
and stieners. One way to achieve this is to model a panel with very large plate
thickness compared to the panel dimensions. Such a panel is likely to deect glo-
bally without local deformations.
A stiened panel consisting of three aluminium proles (Table 2) is modelled in
ABAQUS. The thickness of the plate, the web, and the ange is increased from the
original dimensions given in the table to 50 mm. This geometry is so stocky that
the resulting deformation is purely in the global mode. Analyses are performed on
the stiened panel in ABAQUS for axial and transverse loading, with an imperfec-
tion in the global mode equal to 3.6 mm. Analyses are then performed with the
global buckling model using linear anisotropic stiness coecients according to the
stiener dimensions, as explained previously. The results are shown in Fig. 7. It is
seen that the agreement is very good.
In Fig. 8, the response of this panel under axial loading, calculated using linear
stiness coecients, is compared with the response calculated using reduced sti-
ness resulting from local buckling analysis. The imperfection is 3.2 mm in the local
mode and 3.6 mm in the global mode. It is seen that the global deection is larger
when the reduced stiness is applied, as expected. In this analysis, the stiness
Table 2
Dimensions for aluminium stieners
Stiener a (m) b (m) t (m) h (m) t
w
(m) b
f
(m) t
f
(m) r
f
(MPa)
Tee bar 2.4 0.32 0.005 0.075 0.005 0.040 0.005 240
Fig. 7. Loadaverage strain response due to pure global buckling for stiened panel subjected to axial
compression (left) and transverse compression (right).
715 E. Byklum et al. / Thin-Walled Structures 42 (2004) 701717
coecients input to the global model are gradually reduced, corresponding to the
current load factor.
6. Concluding remarks
A computational model for the analysis of global buckling and postbuckling of
stiened panels has been derived. The model was developed as part of a tool for
buckling assessment of stiened panels. It is formulated using large deection plate
theory and energy principles. Any combination of biaxial in-plane compression or
tension, shear, and lateral pressure may be analyzed. The procedure is semi-ana-
lytical in the sense that all energy formulations are derived analytically, while a
numerical method is used for solving the resulting set of equations, and for incre-
menting the solution. The loaddeection curves produced by the proposed model
are compared with results from nonlinear FEM. Good correspondence is achieved,
and the eciency of the calculations is high.
The global model is combined with a local buckling model in the DNV com-
puter code PULS [13]. In this program, the ultimate strength of panels is estimated
by checking the stress at certain critical points at each increment. Using the von
Mises yield criterion, the onset of yielding is taken as the collapse load for design
purposes. This is conservative, and a sound, design approach, since yielding will
give unwanted permanent deformations in the structure.
Acknowledgements
This study has been performed with support from the Norwegian Research
Council and Det Norske Veritas.
Fig. 8. Loadaverage strain response during axial load calculated from global buckling model using lin-
ear and nonlinear stiness coecients.
E. Byklum et al. / Thin-Walled Structures 42 (2004) 701717 716
References
[1] Byklum E, Amdahl J. A simplied method for elastic large deection analysis of plates and stif-
fened panels due to local buckling. Thin-Walled Structures 2002;40(11):92351.
[2] Det Norske Veritas. PULS 1.5 theory manual. 2003.
[3] Byklum E, Amdahl J. Nonlinear buckling analysis and ultimate strength prediction of stiened steel
and aluminium panels. The Second International Conference on Advances in Structural Engineer-
ing and Mechanics, Pusan. 2002.
[4] Huseyin K. Nonlinear theory of elastic stability. Leyden: Nordho International Publishing; 1975.
[5] Steen E. Application of the perturbation method to plate buckling problems. Research report in
mechanics 98-1, University of Oslo, 1998.
[6] Ueda Y, Rashed S, Paik J. An incremental Galerkin method for plates and stiened plates. Compu-
ters and Structures 1987;27(1):14756.
[7] Paik J, Thayamballi A, Lee S, Kang S. A semi-analytical method for the elasticplastic large deec-
tion analysis of welded steel or aluminium plating under combined in-plane and lateral pressure
loads. Thin-Walled Structures 2001;39:12552.
[8] Steen E. Elastic buckling and postbuckling of eccentrically stiened plates. International Journal of
Solids and Structures 1989;25(7):75168.
[9] Steen E. Buckling of stiened plates using a Shanley model approach. Research report in mechanics
99-1, University of Oslo, 1999.
[10] Marguerre K. Die mittragende Breite der gedruckten Platte. Luftfahrtsforschung 1937;14(3):1218.
[11] Levy S. Bending of rectangular plates with large deection. Report 737, NACA, 1942.
[12] Byklum E. Ultimate strength analysis of stiened steel and aluminium panels using semi-analytical
methods. PhD thesis, Norwegian University of Science and Technology, 2002.
[13] Det Norske Veritas. PULS 1.5 users manual. 2003.
717 E. Byklum et al. / Thin-Walled Structures 42 (2004) 701717