Sie sind auf Seite 1von 16

Journal of Constructional Steel Research 56 (2000) 116

www.elsevier.com/locate/jcsr
Hydrostatic buckling of shells with various
boundary conditions
Rodney Pinna
a,*
, Beverley F. Ronalds
b
a
Centre for Offshore Foundation Systems, The University of Western Australia, Nedlands, Perth, W.A.
6907, Australia
b
Centre for Oil & Gas Engineering, The University of Western Australia, Nedlands, Perth, W.A. 6907,
Australia
Received 3 June 1999; accepted 24 November 1999
Abstract
Eigenvalue buckling of cylindrical shells with various boundary conditions under hydrostatic
load is examined, using an energy method. Results are compared to known solutions, where
these solutions exist. It is found that, for shells of intermediate length, buckling loads for
different end conditions may be determined by applying a simple, scalar multiplier to the pin-
ended case. This does not apply to long shells, where the circumferential wave number n3.
For n=2, the ring equation may be applied to all cases, as the boundary conditions no longer
inuence the solution. It is seen for the case of a shell with one end pinned and the other end
free that the buckling solution collapses to the long shell solution, for geometries of practical
interest. The effect of radial elastic restraint at the open end is also examined, as an intermedi-
ate case between pinned and free ends. The work has application to the design of suction
caissons, where cylinder dimensions are usually in the range of intermediate length shells.
2000 Elsevier Science Ltd. All rights reserved.
Keywords: Boundary conditions; Buckling; Elastic restraint; Shells; Suction caissons
1. Introduction
A novel application for cylindrical shells which has found increasing use is that
of the suction caisson. These have been used as a foundation system for a number
of offshore petroleum production facilities [1,2]. Such foundations are essentially
cylinders, with one end closed by an end cap, and the other end open. The upper
* Corresponding author.
0143-974X/00/$ - see front matter 2000 Elsevier Science Ltd. All rights reserved.
PII: S0143- 974X( 99) 00104- 2
2 R. Pinna, B.F. Ronalds / Journal of Constructional Steel Research 56 (2000) 116
Nomenclature
a buckling multipler for different boundary conditions

z1
,
q1
,
zq1
shell curvatures
e
z1
,e
q1
,g
zq1
shell strains
A
i
undetermined coefcient
C
Eh
1v
2
, shell membrane stiffness
D
Eh
3
12(1v
2
)
, shell bending stiffness
j number of terms in series solution
k

f
non-dimensionalised spring constant
k
f
spring foundation modulus
m, n longitudinal and circumferential wave numbers
N
z0
, N
q0
, N
zq0
pre-buckle membrane force per unit length
P

cr
non-dimensionalised shell buckling load
P
cr
shell buckling pressure
u
1
, v
1
, w
1
displacements in the r, q and z directions
V potential energy function
Z Batdorf parameter,
L
2
ah
1v
2
closed end may range from a light steel plate to a heavily stiffened structure, which
may be idealised as providing either a pinned or clamped end condition (for example,
a caisson attached to a jacket leg, as in Fig. 1), respectively, to the shell. The open
end of the cylinder is effectively sealed by the sea bed. During installation, the
cylinder is subjected to varying amounts of lateral restraint from the surrounding
soil, and suction pressures which are increasing so as to continue the installation
process against rising soil resistance. Thus, while the buckling load of the shell
increases, the applied load on the shell is also increasing. To enable the economic
design of the shell structure, use may be made of the increase in buckling load
provided by the surrounding soil mass.
To assess the increase in buckling load that may be possible, this paper examines
the buckling load of a cylindrical shell with various end conditions. Lateral end
restraint is provided by means of radial Winkler springs in the case of an open-
ended shell, with free and pinned extreme cases examined. Buckling loads are found
using energy functions, solved using a variational approach. Comparisons are also
made with existing results, and a number of nite element solutions.
3 R. Pinna, B.F. Ronalds / Journal of Constructional Steel Research 56 (2000) 116
Fig. 1. Example of an effectively clamped suction caisson top.
2. Buckling analysis
The properties of the shell being considered are shown in Fig. 2. This shows an
open-ended cylinder, with elastic restraint around its base. The elastic restraint is
provided by a set of uniform Winkler springs, with modulus k
f
. Buckling analysis
of the shell/spring system is performed using an eigenvalue solution method, based
on the method of variations [3]. To do this, the second variation in the energy func-
tional is required. This may be divided into two parts: the contribution from the
shell, and that from the elastic support. Details for deriving these expressions may
be found in a number of texts [4,5]. The second variation of the potential energy of
the shell is given by:
4 R. Pinna, B.F. Ronalds / Journal of Constructional Steel Research 56 (2000) 116
Fig. 2. Suction caisson parameters.
1
2
d
2
Va
C
2

e
2
z1
e
2
q1
2ve
z1
e
q1

1v
2
g
zq1
dz dq
a
2

N
z0
w
2
1,z
N
q0
w
2
1,q
a
2
2N
zq0
w
1,z
w
1,q
a
dz dqa
D
2

2
z1

2
q1
2v
z1

q1
2(1v)
zq1
dz dq (1)

1
2
P
cr
v
2
1
v
1
w
1,q
v
1,q
w
1
w
2
1
dz dq
where the subscript 1 denotes the post-buckled state, while 0 denotes the pre-buckled
shell state. The expression for the second variation in potential energy of the spring
restraint is:
1
2
d
2
V
s

ak
f
2

w
2
1
(zL,q) dz dq (2)
Adding Eqs. (1) and (2) leads to the nal variational expression. With this equation
it is then necessary to substitute expressions for the shell displacements, strains and
curvatures after buckling. The required expressions are [5]:
e
z1

z
u
1
(z,q) (3)
5 R. Pinna, B.F. Ronalds / Journal of Constructional Steel Research 56 (2000) 116
e
q1

q
v
1
(z,q)+w
1
(z,q)
a
(4)
g
zq1

z
v
1
(z,q)

q
u
1
(z,q)
a
(5)

z1

z,z
w
1
(z,q) (6)

q1

q
v
1
(z,q)
q,q
w
1
(z,q)
a
2
(7)

zq1

1
2

z
v
1
(z,q)2
z,q
w
1
(z,q)
a
(8)
These expressions are valid for both shallow and non-shallow shells. Substituting
these into the variational energy expressions results in an equation for the second
variation in terms of the pre-buckling membrane stresses and the post-buckled shell
displacements. Hence, to determine eigenvalues of the problem, it is necessary to
have expressions for both of these. For the following analysis, it is assumed that the
pre-buckled shell stresses are adequately described by shell membrane theory. This
leads to the membrane forces per unit length of:
N
z0

P
cr
a
2
(9)
N
q0
P
cr
a (10)
N
zq0
0 (11)
The displacement functions that describe the post-buckled shape of the shell
depend on the boundary conditions that are present (see Table 1). In this paper six
Table 1
Lateral displacement boundary conditions
Designation Boundary conditions w
1
(z=0) w
1
(z=L) w
1
(z=0)
z
w
1
(z=L)
z
P-P Both ends pinned 0 0 Free Free
C-C Both ends clamped 0 0 0 0
P-F Upper edge pinned, lower edge free 0 Free Free Free
C-F Upper edge clamped, lower edge free 0 Free 0 Free
P-E Upper edge pinned, lower edge elastically 0 Elastic Free Free
restrained
C-E Upper edge clamped, lower edge elastically 0 Elastic 0 Free
restrained
6 R. Pinna, B.F. Ronalds / Journal of Constructional Steel Research 56 (2000) 116
cases are examined, with the displacement functions shown in Table 2. With one of
these sets of functions in place, the buckling load is determined using the Rayleigh
Ritz method. The total expression for the second variation in potential energy of the
system is minimised against each of the undetermined coefcients A
i
, i.e.

A
i

1
2
d
2
V
1
2
d
2
V
s
0 for i12j (12)
This produces a set of j simultaneous equations, [C
1
]{A
1
A
2j
}+P
cr
[C
2
]
{A
1
A
2j
}=0. The eigenvalues of this equation then give the critical buckling load
in terms of n and, where applicable, m. Where m is present, the lowest buckling
load is found with m=1. Further details may be found in Pinna and Ronalds [6].
The cases of a cylinder with an open, unrestrained end (C-F and P-F) are special
cases of a cylinder with an elastically restrained end (C-E and P-E). The displacement
functions for these two end conditions are the same, with the buckling mode shapes
dependent on the values of the coefcients A
j
found by minimising the energy func-
tion, which in turn depend on the value of the spring constant k
f
. As the end restraint
becomes stiffer, the change in curvature along the length of the shell also becomes
greater. To accurately model this change in curvature, it is necessary to ensure that
a sufcient number of terms, j, are included in the displacement functions.
It should be noted that Eq. (1) and Eqs. (3)(8) are valid for all circumferential
wave numbers n. Omission of the last line of Eq. (1) would result in a set of equations
valid for shells under dead-loading. These equations would produce results that are
accurate for short to intermediate length shells under hydrostatic pressure loading,
but inaccurate for long shells (n3). The point where a shell becomes long is a
function of both its geometry and end conditions, as discussed in the next section.
The lateral spring stiffness k
f
may be non-dimensionalised by
k

f
k
f
La
Eh
2
(13)
The use of this factor is also discussed below.
Table 2
Displacement functions for various boundary conditions
Designation u
1
v
1
w
1
P-P
A
1
sin(nq)cos

mpz
L

A
2
cos(nq)cos

mpz
L

A
3
sin(nq)sin

mpz
L

C-C
A
1
sin(nq)sin

2mpz
L

A
2
cos(nq)

cos

2mpz
L

1

A
3
sin(nq)

cos

2mpz
L

1

P-E, P-E
sin(nq)

i
A
i
z
L

i1
cos(nq)

i
A
i+j
z
L

i
sin(nq)

i
A
i+2j
z
L

i
C-F, C-E
sin(nq)

i
A
i
z
L

i
cos(nq)

i
A
i+j
z
L

i+1
sin(nq)

i
A
i+2j
z
L

i+1
7 R. Pinna, B.F. Ronalds / Journal of Constructional Steel Research 56 (2000) 116
3. Results
Fig. 3 shows the effect of increasing end restraint on the buckling load of both
P-E and C-E shells, where the buckling load is non-dimensionalised by:
Fig. 3. Shell buckling load as k
f
varies (Z=500, L/a=1.61889).
8 R. Pinna, B.F. Ronalds / Journal of Constructional Steel Research 56 (2000) 116
P

cr

L
2
a
p
2
D
P
cr
(14)
It can be seen that there is a substantial increase in the buckling load in each case.
For the geometry shown in the gure, the fully restrained buckling loads are 19.8
and 2.10 times the unrestrained buckling loads for the P-E and C-E cases, respect-
ively. Also shown in this gure are a number of results produced by nite element
analysis using the ABAQUS [7] program. Details of this analysis can be found in
Pinna and Ronalds [6]. It may be seen that for both cases, there is very good agree-
ment between the results found using the method outlined above, and the nite
element analysis.
For the case of a shell with one end pinned and the other free, the analysis assumes
that buckling occurs with n2. It is found in both the variational and nite element
analyses that other very low buckling loads may occur for this combination of end
conditions, where these modes correspond to rigid body eigenmodes. These modes
can be restricted so that buckling occurs in a circumferential mode, which reects
the situation for a suction caisson. In the nite element analysis, this is done by
xing the shell at four nodes, around the top of the shell, against axial displacement.
Alternatively, an n=2 solution may be found by ignoring the rigid body modes. For
the installation of a suction caisson, it would be expected that the rigid body modes
would be restricted, and buckling would therefore occur with n2. This assumption
is carried through this analysis. In contrast, Koga and Morimatsu [8] give a buckling
load for a P-F cylinder (S3-FR in their notation) of zero, using an asymptotic method,
and with n2.
The variation in buckling load, as the geometry of the shell is changed, is shown
for various elastic end restraint conditions for P-E and C-E shells in Figs. 4 and 5.
These graphs show that the increase in buckling load shown in Fig. 3 is valid for a
range of Z values. One immediately apparent difference between these two gures
is that, for the case of a shell with a clamped top, there is a relatively large number of
circumferential waves in the buckling solution. In comparison, P-F shells of practical
geometries (that is, of intermediate length) have a buckling mode with n2. Apply-
ing the assumption that n=2 results in a solution identical to that for a long shell,
that is [5]:
P

cr

3
p
2
1v
2
Z
h
a
(15)
Any sufciently long shells critical load will converge on Eq. (15), which is
independent of the shells end conditions. For all types of boundary conditions exam-
ined, this occurs after the value of Z where n rst equals 2. A distinguishing property
of P-F shells is that they enter this long shell mode immediately for practical geo-
metries. However, as may be seen in Fig. 4, with the presence of some lateral restraint
at the base of the shell, buckling occurs with n2, and thus the series form of the
solution is required.
Fig. 6 demonstrates this, showing the eigenmode for a P-F shell with no elastic
restraint. It can be seen that there are only two circumferential waves around the
9 R. Pinna, B.F. Ronalds / Journal of Constructional Steel Research 56 (2000) 116
Fig. 4. P-E shells with various elastic end restraint conditions (a/h=200).
shell. Also, the lack of rotational restraint at the top of the shell is evident. This
circumferential mode is similar to that which would be obtained for a much longer
shell with other boundary conditions. These results are also reected in Fig. 3. The
much larger increase in buckling load for the P-E shell as k

f
increases, compared
with the C-E shell, is caused by the change in buckling mode from a long to an
intermediate shell. The C-E shell does not undergo this change, as it remains in the
intermediate shell solution regime.
Bounds for these cases are found by setting the value of k

f
to either zero or
innity. The result of doing this is shown in Fig. 7. In this gure, solid lines are
10 R. Pinna, B.F. Ronalds / Journal of Constructional Steel Research 56 (2000) 116
Fig. 5. C-E shells with various elastic end restraint conditions (a/h=200).
those determined from the variational analysis, while dotted lines are found using
the multiplier a. The critical buckling load is then given by:
P

cr
ab (16)
where
b2

1+
8Z
3p
2
(17)
where values for a are given in Table 3, and b [9] is a lower bound approximation
for the buckling value of a P-P shell. It can be seen that, for all cylinders with
11 R. Pinna, B.F. Ronalds / Journal of Constructional Steel Research 56 (2000) 116
Fig. 6. Eigenmode for a P-E shell.
solutions in the intermediate shell range, the buckling load can be arrived at by
applying a multiplier to the pin supported case. As the P-F case falls immediately
into the long shell range, it is not proportional to b, and it does not follow this
multiplier rule. Table 3 also compares the multipliers obtained here with those pre-
viously published for these bounding cases. It can be seen that there is no difference
between published results, and the results found here using exact trigonometric dis-
placement functions, that is, the P-P and C-C cases. Further, for the cases where
truncated series solutions are used (C-F and C-P), the agreement with published
results is also good. It should be noted that the solution arrived at by Malik et al. [10]
was by an approximate method, using a similar starting equation to that employed by
Koga and Morimatsu [8]. The a multiplier can also be used for intermediate restraint.
This is shown in Figs. 4 and 5, where the lateral restraint at the elastically supported
end varies from none to fully effective in each case. It can be seen that for n4,
the buckling load can be expressed in terms of a.
The effect of non-dimensionalising k
f
in accordance with Eq. (13) is shown in
Figs. 8 and 9 for P-E and C-E shells, respectively. The buckling load is given in
terms of the multiplier a. These gures provide further evidence that the nondimen-
sional restraint factor k

f
is appropriate over the complete range where the shell
12 R. Pinna, B.F. Ronalds / Journal of Constructional Steel Research 56 (2000) 116
Fig. 7. Buckling loads for various end conditions.
transitions from weak to near full lateral support. For large values of k

f
, the vari-
ations in the nondimensional buckling loads in Fig. 8 are caused by restricting the
value of n to an integer. The minimum buckling load that is predicted analytically
may occur for a non-integer value of n. Physically, however, a complete number of
waves must form around the shell when buckling occurs. Thus, the minimum found
with integer n, while corresponding to the physical minimum, may not correspond
to the absolute minimum that could be found. The spread of curves at low k

f
in
13 R. Pinna, B.F. Ronalds / Journal of Constructional Steel Research 56 (2000) 116
Table 3
Multipliers for various boundary conditions
Designation a Published results
P-F Not applicable
C-F 0.58 0.58 [10]; 0.6 [8]
P-P 1
C-P 1.22 1.25 [10]; 1.25 [8]
C-C 1.5 1.5 [8]; 1.5 [12]
Fig. 8 is caused by the restriction that n=2, as discussed above. Fig. 9 also shows
the effect of allowing n to be a non-integer. It may be seen that this gives much
better convergence over a wide range of Z values.
4. Discussion
The set of graphs provided allow for the calculation of buckling loads for shells
with various end conditions. For the case where the shell has one of the limiting
boundary conditions (C-C, C-P or C-F), the a multiplier given in Table 3 can be
applied to the P-P case to arrive at the buckling load. Where a cylinder has an
elastically supported free end of stiffness k

f
, then Fig. 8 or Fig. 9 can be used to
determine the value of a, depending on the top end condition (pinned or clamped).
The use of a multiplier factor a is advantageous as it allows existing design codes,
which generally provide formulae for the P-P case only, to be applied to other bound-
ary conditions representative of suction caissons. After applying the multiplier to the
P-P case, knock-down factors for imperfections and elasto-plastic behaviour can be
included in design calculations. Such factors are provided in, for example, the DnV
Buckling Strength Analysis code [11]. This applies to shells in the short to inter-
mediate length range.
For long shells, the effects of boundary conditions may be ignored. Thus, the ring
buckling formula, Eq. (15), can be applied directly. It should be noted, however,
that the point where a solution enters into the long shell solution depends on both
the shell geometry and the boundary conditions. For the case of a shell with P-F
end conditions, or with only small lateral restraint, then the ring solution can be
applied for all practical geometries. It is also found in this study that a P-F shell
will tend to buckle in a rigid body mode, if there is no axial restraint present. Only
a minimal axial restraint is required, however, to force the shell into a circumferential
buckling mode. During the installation of a suction caisson, it is likely that the pres-
ence of even a relatively exible end cap would be sufcient to force this mode.
14 R. Pinna, B.F. Ronalds / Journal of Constructional Steel Research 56 (2000) 116
Fig. 8. Buckling load for P-E boundary conditions, Z varying (50Z5000).
5. Conclusions
The results presented in this paper allow the hydrostatic buckling load of cylindri-
cal shells with various end conditions to be found. For cylinders with elastic lateral
restraint at one end, Figs. 8 and 9 give the non-dimensional buckling load, based
15 R. Pinna, B.F. Ronalds / Journal of Constructional Steel Research 56 (2000) 116
Fig. 9. Buckling load for C-E boundary conditions, Z varying (250Z5000).
on the boundary conditions at the other end. The non-dimensional equation for lateral
elastic restraint, Eq. (13), works well for shells with either pinned or clamped top
ends, over all geometries where the shell can be considered of intermediate length.
For all cases, a simple multiplier can be applied to the buckling load for a pinned
16 R. Pinna, B.F. Ronalds / Journal of Constructional Steel Research 56 (2000) 116
pinned shell, allowing the adaptation of existing buckling codes for offshore struc-
tures to boundary conditions relevant to suction caisson design.
For long shells, boundary conditions have no effect on the buckling load, and Eq.
(15) can be applied directly to all cases. Where a shell becomes long depends on
various parameters, and is indicated by a buckling solution with n=2. Guidance for
determining where a long shell solution can be applied is offered by existing buck-
ling codes.
Acknowledgements
The authors gratefully acknowledge the support of the Special Research Centre
for Offshore Foundation Systems, funded through the Australian Research Councils
Research Centres Program.
References
[1] Dyvik R, Anderson KH, Hansen SB, Christophersen HP. Field tests of anchors in clay I: Description.
J Geotech Eng 1993;119(10):151531.
[2] Tjelta TI. Geotechnical experience from the installation of Europipe jacket with bucket foundations.
In: Proceedings of the Offshore Technology Conference, OTC7795, Houston, 1995.
[3] Courant R. Variational methods for the solution of problems of equilibrium and vibrations. Bull Am
Math Soc 1943;49:123.
[4] Langhaar HL. Energy Methods in Applied Mechanics. New York: Wiley, 1962.
[5] Brush DO, Almroth BO. Buckling of bars, plates and shells. New York: McGraw-Hill, 1975.
[6] Pinna R, Ronalds BF. Eigenvalue buckling of elastically restrained cylinders under hydrostatic load,
G1418. Centre for Offshore Foundation Systems, University of Western Australia, 1999.
[7] ABAQUS Users Manual, version 5.7. Rhode Island (NY): Hibbit, Karlsson and Sorenson, Inc, 1997.
[8] Koga T, Morimatsu S. Bifurcation buckling of circular cylindrical shells under uniform external
pressure. AIAA J 1989;27(2):2428.
[9] Odland J. On the strength of welded ring stiffened cylindrical shells primarily subjected to axial
compression. Technical Report UR-81-15, Divsion of Marine Structures, The University of
Trondheim, The Norwegian Institute of Technology, 1981.
[10] Malik Z, Morton J, Ruiz C. Buckling under normal pressure of cylindrical shells with various end
conditions. J Press Vessel Technol 1980;102:10711.
[11] Det Norsk Veritas, Buckling strength analysis. Det Norsk Veritas Classication As, Veritasveien 1,
N-1322 Hovik, Norway, 1995.
[12] Calladine CR. Theory of shell structures. Cambridge: Cambridge University Press, 1983.

Das könnte Ihnen auch gefallen