You are on page 1of 6

Buckling and collapse of cylinders with one end simply supported and

one end open with variable axial and flexural support, under lateral load
Rodney Pinna
The School of Oil & Gas Engineering, The University of Western Australia, Crawley, Australia

ABSTRACT: Cylinders with open end simply supported and open and unsupported (SF boundary conditions)
have buckling loads which are lower than those of cylinders with more usual boundary conditions. SF cylin-
ders also have collapse loads which are above their eigenvalue buckling load. This work examines the effect
on the buckling and collapse load of varying the axial and flexural restraint at the closed end of the cylinder,
for lateral load conditions. It is shown that as the restraint at the supported end is increased, the eigenvalue
buckling load increases, while the collapse behaviour of the cylinder reverts to a more typical cylindrical shell
behaviour, with the collapse load being sensitive to the presence of imperfections in the initial geometry.

1 INTRODUCTION − The caisson is evacuated, allowing the ambient


water pressure to force the foundation into the
Cylinders with one end open and the other end sim- seabed.
ply supported have been shown (Pinna, 2003) to Due to this installation sequence, the foundation
have behaviour which is relatively unusual. The may be susceptible to buckling failure during both
buckling load of shells with these boundary condi- stages of the installation sequence; during both the
tions may be considerably lower than that of cylin- self weight penetration when loading is predomi-
ders with more typical boundary conditions, how- nately axial and during the pressure installation
ever, more unusually, the collapse load of such phase, when combined loading with axial and lateral
cylinders may be above their eigenvalue buckling components is present.
load. The combination of boundary conditions on a cyl-
One such system where such boundary conditions inder of a simply supported end and an open end has
may occur is that of the suction caisson. These are a not received much attention previously. Previous
form of foundation system used for offshore struc- work has explored cylinders with such boundary
tures (Pinna, 2001, Tjelta, 2001). These essentially conditions under axial, combined and lateral loading
consist of a cylinder, with one end open and the (Ronalds & Pinna 20003, Pinna & Ronalds, 2003)
other end closed. The closure may range from a very
stiff top plate, typically employed when the caisson 1.E+04
is used as a foundation for a jacket structure, to a
much more flexible structure, when the caisson is
used as an anchor for a catenary mooring system. 1.E+02
In the first of these cases, the boundary conditions
on the closed end of the cylinder may be approxi- SS cyl.
kh

mated by a fully built in end. In the later case, the


restraint at the open end will be somewhat less and 1.E+00 SF cyl. Plate
will tend towards a simple support as a lower bound. solution
The major attraction of the suction caisson is the
Long shell
ease of its installation. Installation is essentially a 1.E-02
three step process: 1.E-01 1.E+01 1.E+03 1.E+05 1.E+07
− The caisson touches down on the sea bed
Z
− The caisson penetrates the seabed under self Figure 1. Comparison of buckling loads for SS and SF cylin-
weight, this typically penetrates the foundation to ders against standard solutions.
between 5% and 30% of its length
Table 1. Various definitions for a simply supported end for a Two types of analysis were carried out; first, an
cylindrical shell., where u, v, and w are the displacements com- elastic eigenvalue analysis of the soil/structure sys-
ponents in the radial, circumferential and axial directions re- tem was performed to determine the critical buckling
spectively. Nxx and Nxθ are stress resultant terms. The designa- mode of the system. This eigenmode was then im-
tion follows that of Hoff and Soon (1965).
Designation Restrained degree of freedom
posed on the geometry of the perfect structure, and a
nonlinear Rik’s analysis undertaken. Nonlinear be-
S1 w = ∂ 2 w ∂x 2 = N xx = N xθ = 0 haviour was allowed for in both the material proper-
S2 w = u = ∂ 2 w ∂x 2 = N xθ = 0 ties and by accounting for the deformed geometry of
S3 w = ∂ 2 w ∂x 2 = N xx = v = 0
the system. An imperfection in the shape of the criti-
cal eigenmode generally results in the largest de-
S4 w = ∂ 2 w ∂x 2 = u = v = 0 crease in peak load (Guggenberger, 1995). The ei-
genmode was scaled such that the maximum lateral
and for the axial load case with varying amounts of deviation was equal to the shell thickness.
axial restraint at the supported end (Pinna, 2003a, b).
In the former case, it was shown that simply support
– free cylinders buckle at lower than usual loads 2.2 Boundary conditions
(Fig. 1). From this Figure, it may also be seen that
SF cylinders behave in a qualitatively different way Unlike beams, a number of definitions of a simply
to SS cylinders; while SS cylinders diverge from the supported end are possible for a cylindrical shell.
long shell solution at Z ≈ 1×105, where Z is defined These are shown in Table 1. The set of boundary
in equation (1) below. SF cylinders follow the long conditions matches the definition of a “classical”
shell solution until much lower values of Z and only simple support for a cylinder, and is the definition
diverge away for very short shells. used in this article when such a boundary condition
In Pinna (2004), it was shown that as the amount is referred to.
of axial restraint is increased, the buckling load in- To model the pinned end of the shell, the dis-
creases dramatically, but the unusual imperfection placement components in the radial and tangential
sensitivity which is present for the SF cylinder with directions are fixed, as are rotations about the radial
no additional restraint is found to disappear. and axial directions. The varying restraint at the
The present work extends this analysis by exam- supported end was modelled using either axial or ro-
ining the theoretically important case of lateral load- tational springs, with linear behaviour.
ing when the supported end has varying axial re-
straint. This paper also examines the influence of 2.3 Loading
varying flexural restraint on the buckling and col-
lapse behaviour. Results are found using nonlinear Loading was applied to the cylinder as a lateral load,
finite element analysis. using the ABAQUS DLOAD routine. This load rou-
tine in ABAQUS includes follower load effects,
which are important for buckling in modes with a
2 METHODOLOGY low number of circumferential waves (n).

2.1 Finite element model


2.4 Materials
Results in this paper were found from finite element
analysis, using ABAQUS/STANDARD (Hibbert et Typical high strength steel properties were used in
al. 2003). The cylinders were modelled using the finite element model, with E=210 GPa, ν=0.3
meshes consisting of 9 noded, thin shell elements, and σyield=340 MPa. The yield strength of the steel
employing reduced integration, i.e. S9R5 elements was modelled using a von Mises criterion.
in ABAQUS notation. These elements enforce the
Kirchoff constraints numerically, and are appropri- 2.5 Geometry
ate for small strains, but large deformations and rota-
tions. They employ strain measures which are ap- The geometry of cylindrical shells is commonly de-
proximations to Koiter–Sanders theory strains. scribed using the nondimensional Batdorf Z parame-
Eigenvalues are calculated from an initially unper- ter, where:
turbed state. To verify the results from the S9R5 ele- L2
ments, a number of eigenvalues were also found Z= 1 −ν 2 (1)
from models using solid elements. Results have also rt
been compared to those from a variational solution In the present analysis, eigenbuckling loads are
using the Donnell equations and show the S9R5 found for cylinders with Z=10, 100 and 1000, while
elements performs well (Pinna & Ronalds, 2003). nonlinear collapse analysis is carried out on the cyl-
inder with Z=100. This represents a cylinder of in- L2 h
termediate length. For all cylinders, r/t=100. kh = σh (2)
π 2D
where D is the usual expression for the bending
3 RESULTS AND DISCUSSION stiffness of the cylinder.
3.1 Bifurcation analysis It is known that for most sets of boundary condi-
tions the buckling loads of intermediate length cyl-
Bifurcation buckling results from the present analy- inders with differing boundary conditions may be re-
sis are shown in Figure 2, for varying amounts of ax- lated to the SS case through a constant multiplier
ial restraint, and in Figure 3 for varying flexural re- (Ronalds & Pinna, 2003). For a CF cylinder, the
straint . It may be seen that in both cases, increasing buckling load is 0.58 times that of an SS cylinder.
the restraint at the closed end results in a significant From the figures, it may be seen that axial re-
increase in the buckling strength of the cylinder. straint has the largest influence on the bifurcation
Both graphs show three results: the result for the load. With “fully effective” axial restraint, the bifur-
SF cylinder with varying restraint, and the solutions cation load approaches that of the CF cylinder. This
for a SS and CF cylinder. The solutions for all these behaviour is similar to that of a SF cylinder under
cylinders are expressed in terms of the buckling co- axial load, with axial restraint. For that case, it has
efficient, which is related to the buckling stress by: also been shown that the buckling load approaches
that of a semi-infinite cylinder with one free end
(Pinna, 2004).
16.00 Restraining the flexural degree of freedom has
less effect on the cylinder buckling load. This re-
SS cylinder
12.00
kh

8.00 CF cylinder

4.00 SF cylinder

0.00
1.E-01 1.E+02 1.E+05 1.E+08 1.E+11
Spring restraint (N/m)
Figure 2. Variation in eigenvalue buckling load with changing
axial spring stiffness.

16.00

SS cylinder
12.00
kh

8.00 CF cylinder

4.00
SF cylinder

0.00
1.E-01 1.E+02 1.E+05 1.E+08 1.E+11
Spring restraint (Nm/rad)

Figure 4. Effect on varying the degree of axial restraint on the


Figure 3. Variation in eigenvalue buckling load with changing cylinder buckling mode. Restraint values are 0, 1×108 and
flexural spring stiffness. 1×1013 N/m.
Table 2. The effect of varying axial and flexural restraint on 1.6
buckling load and mode.
Axial spring restraint kh n 1.4
1×10-1 0.639 2
1×100 0.639 2 1.2 0
1×101 0.639 2
1×102 0.639 2 1
1×108

P/Pcr
1×103 0.647 2 0.8
1×104 0.721 2
1×105 1.277 3 0.6
1×106 1.820 3 1×1012
1×107 2.939 4 0.4
1×108 4.699 6
1×109 5.821 6 0.2
1×1010 6.227 6 0
1×1011 6.280 6
1×1012 6.286 6 0 10 20 30
wo

Flexural spring restraint kh n


1×10-1 0.639 2 1.6
1×100 0.639 2
1.4
1×101 0.639 2
1×102 0.639 2 1.2
1×103 0.645 2 0
1×104 0.699 2 1
1×105 1.065 2
P/Pcr
1×106 1.603 3 0.8
1×107 1.910 3
1×108 2.074 3 0.6 1×108
1×109 2.124 3 0.4
1×1010 2.141 3
1×1011 2.149 3
1×1013
0.2
1×1012 2.150 3
0
flects what is know for SS cylinders under lateral 0 10 20 30
wo
load, where the largest influence on buckling is the
degree of restraint on the axial freedom. Figure 5. Behaviour of cylinders in terms of the normalised
The influence on the behaviour of the cylinder is lateral load acting on the cylinder. The upper plot shows vary-
also shown in Figure 4, where the buckling mode of ing flexural restraint, while the lower plot shows varying axial
the cylinder is shown for varying amounts of axial restraint. The value next to each curve denotes the spring con-
restraint. In particular, it may be seen that the degree stant used for the analysis.
of axial deflection decreases substantially as the Chapman, 1991). This reflects that under lateral
amount of restraint is increased. These results are load, the SF cylinder is able to resist buckling
also summarised in Table 2, where results for vary- through its torsional rigidity. In contrast, under axial
ing flexural restraint are also given. It may be seen load, SF cylinders act as a mechanism and continue
that the change in the number of circumferential to follow the long shell solution, which is much
buckling waves (n) is much less, reflecting the lower than the value that low Z SS cylinders asymp-
smaller influence that this restraint has on the buck- tote towards.
ling load.
These results show that the buckling strength of
the cylinder increase 9.83 times going from none to 3.2 Collapse results
fully effective axial restraint, and by 3.36 times for The results from the nonlinear collapse analysis
flexural restraint. This difference is not as great as conducted on the cylinders is shown in Figure 5 for
that for SF cylinder under axial load, where the dif- cylinders with both axial and flexural restraint, un-
ference may be in the order of two magnitudes. This der lateral load. This Figure shown the results in
difference may be explained by the difference in be- terms of the lateral load normalised by the buckling
haviour of SF cylinders with and without an axial load of the cylinder. The load is plotted against the
load component. maximum inwards lateral displacement of the cylin-
SF cylinders without an axial load component, der, which has been normalised by the shell wall
that is under pure lateral load, asymptote towards a thickness.
buckling load of kh=0.426, as Z tends towards 0. From these results it may be seen that the in-
This corresponds to the solution for a hinged out- fluence of varying either axial or flexural restraint
stand under uniform axial compression (Ronalds & is similar, in that for both situations the imperfec-
tion sensitivity of the system increases as the re-
straint is increased. This sensitivity is also shown
3.E+05
in Figure 6, where the P/Pcr is plotted against the
varying spring stiffness value. The typical knock- 1×1012
3.E+05
down factor of 0.6 which is used for laterally
loaded cylinders is also plotted in this Figure. 1×108
2.E+05
In these plots, it may be seen that as the stiff-

P (N)
ness increases, and the eigenbuckling mode moves
2.E+05
away from the n = 2 value, sensitivities change
from showing a collapse load above the eigen-
1.E+05 0
value buckling load to returning to more usual
values. For large restraint, where the eigenvalue
5.E+04
buckling load is closer to that of cylinders with
stiffer restraint and n > 2, the sensitivity returns to
0.E+00
values which are typically associated with cylin-
ders which have a lateral load component. 0 10 wo 20 30
The collapse strength may also be plotted in
terms of the absolute load applied to the cylinder,
as is done in Figure 7. From this Figure, it my be 7.E+05

6.E+05
1.6
5.E+05
1.4
1×1013
P (N)
1.2 4.E+05

1 3.E+05
1×108
P/Pcr

0.8 2.E+05
0.6 1.E+05
0
0.4
0.E+00
0.2 0 10 20 30
wo
0
1.0E-02 1.0E+01 1.0E+04 1.0E+07 1.0E+10 Figure 7. Behaviour of cylinders in terms of the total lat-
eral load acting on the cylinder. The upper plot shows
Spring restraint (N/m) varying flexural restraint, while the lower plot shows vary-
ing axial restraint. The value next to each curve denotes the
spring constant used for the analysis. The order of the Fig-
ures corresponds to Figure 5.
1.6
seen that in terms of absolute load SF cylinders
1.4
with stiff restraint collapse at loads which are
1.2 much greater than those of SF cylinders. This re-
1
flects that the buckling load for SF cylinders is
relatively low, so that even with a “knockdown”
P/Pcr

0.8 factor greater than 1 applied to the load, it is still


0.6
relatively small.
The behaviour of the SF cylinder with no re-
0.4 straint may be seen to be somewhat different to
0.2 that of the other two cylinders; rather than reach-
ing a clear peak and then falling away, it instead
0 steadily increases to a maximum point. While it is
0.01 10 10000 1E+07 1E+10 1E+13 difficult to see from the plot, some reduction in
load from the peak is present before the analysis
Spring restraint (Nm/rad)
terminates.
The sensitivity of a SF cylinder with no axial
or flexural restraint is studied in Figure 8. In this
Figure 6. Sensitivity of cylinder against changing spring plot the size of the initial imperfection has been
restraint value. Upper plot for varying axial restraint, lower
plot for varying flexural restraint varies from 0.1 of the wall thickness to 5 times the
wall thickness. From this result it can be seen that
sociated with a change in the location of buckling
1.6 waves around the cylinder. Collapse remains in an
1.4 n = 2 mode.
0.5
1.2
0.1
1 4 CONCLUSIONS
P/Pcr

0.8 Buckling of intermediate length SF cylinders with


1 no additional axial or flexural restraint at the simply
0.6
supported end have been shown to have a number of
0.4 5 features in common cylinders with the same bound-
ary conditions under other load conditions. In par-
0.2 ticular, these features are low bifurcation buckling
0 loads, and an insensitivity to the presence of initial
0 10 20 30 geometric imperfections. In the case of SF laterally
wo loaded cylinders, the reduction in buckling load is
Figure 8. Load deflection plots of cylinders with varying not as great as that for SF cylinders under loading
imperfection size. Numbers next to each trace indicate the with an axial component, as the lateral load is able to
imperfection as a multiple of the cylinder wall thickness. be resisted through the torsional stiffness of the cyl-
1.6 inder.
As with other loading conditions, it is shown that
1.4 increasing restraint at the simply supported end re-
sults in an increase in the buckling load. For the pre-
1.2
sent case, with “fully effective” axial restraint, the
1
buckling load approaches that of an open ended cyl-
inder with a clamped end. With increasing restraint
0.8 it is shown that the imperfection sensitivity returns
P/Pcr

to levels the same as those of shells with traditional


0.6 boundary conditions.
0.4
5 REFERENCES
0.2
Guggenberger, W. 1995. Buckling and postbuckling of imper-
0 fect cylindrical shells under external pressure. Thin Walled
0 1 2 3 4 5 6 Structures 23: 351–366.
Hibbert, Karlsson and Sorenson. 2003. ABAQUS user’s man-
Imperfection size w 0 ual, version 6.4. Hibbert, Karlsson and Sorenson: Rhode Is-
land.
Figure 9. Sensitivity of collapse load of a SF cylinder as the
Pinna, R. Ronalds, B.F. 2000. Hydrostatic Buckling of Shells
initial imperfection size is varied.
with Various Boundary Conditions, Journal of Construc-
tional Steel Research 56: 1–16.
the collapse load of an SF cylinder is not sensitive Pinna, R. 2004. Buckling of suction caissons during installa-
to the magnitude of the initial imperfection. This tion. PhD Thesis. Perth: The University of Western Austra-
result is also suggest by the shape of the load de- lia
flection curve in Figure 7. The relatively “soft” Pinna, R. & Ronalds, B.F. 2003. Buckling and Postbuckling of
collapse behaviour is typically associated with Cylindrical Shells with One End Pinned and the Other End
low sensitivity to initial geometric imperfections. Free. Thin Walled Structures 41(6): 507-527.
Ronalds, B.F. & Pinna, R. 2003. Eigen buckling of cylindrical
A plot of the load deflection behaviour for a PF shells in offshore structures: influence of geometry, loading
cylinder with no axial or flexural restraint is and end conditions. Structures and Buildings 156(SB2):
shown in Figure 8, with the maximum points 183-192.
shown in Figure 9. From this Figure, it may be Tjelta, T. I. 2001. Suction piles: Their position and application
seen that the behaviour of the cylinders is rela- today’. Proceedings of the 11th International Offshore and
Polar Engineering Conference. New York:ASME.
tively consistent; as he size of the initial imperfec-
tion is increased, the response of the cylinder be-
comes more flexible. The change in peak load is
fairly insensitive to the presence of imperfection
however.
The load-displacement curves for the low stiff-
ness cylinder show a nonuniform displacement re-
sponse. The nonuniformity of this response is as-

You might also like