Você está na página 1de 8

-

Reprinted From

PVP Vol. 176. Computational Experiments


Editors: W. K. Liu, P. Smolinski, R. Ohayon,
J. Navickas, and J. Gvildys
Book No. H00491 1989

The American Society of


Mechanical Engineers

AXIAL BUCKLING OF A THIN CYLINDRICAL SHELL:


EXPERIMENTS AND CALCULATIONS
S. W. Kirkpatrick and B. S. Holmes
SRI International
Menlo Park. California

ABSTRACT
Thin cylindrical shells were tested under axial
compression beyond the critical huckling load. Both
pretest and posttest finite element calculations were
performed to calculate the huckling loads and posthuckling deformations. Finite element simulations of
the shell included pretest measured imperfections in
shell geometry and asymmetry in the axial load.
Results show that the axial collapse load is sensitive
to imperfections in both the shell geometry and the
load distribution. Careful modeling of the imparfecZions resulted in accurate predictions of the buckling
load and postbuckling deformations for the shells.
INTRODUCTION AND BACKGROUND
One of the major design criteria of thin shell
structures that experience compressive loads is the
huckling load limit. Bucking of a thin shell structure can often lead to a catastrophic failure; therefore, it is important that the huckling loads he
known. Typically, thin shell structures are designed
with large safety factors to preclude the possibility
of a buckling failure. However, large safety factors
result in thicker and heavier structures. For many
applications, such as in the aerospace industry, the
excess weight and material required for a large safety
factor can greatly increase the cost of the structure.
In these applications, it is important to be able to
accurately predict the huckling load.
This paper
discusses a numerical modeling technique using finite
element methods that can accurately predict the static
buckling load for an axially compressed cylindrical
shell.
Buckling of thin shell structures can occur under
static or dynamic loading conditions. When the load
duration is long compared to the response time of the
structure, the structure is said to buckle statically,
as in the case of an aluminum can crushed in a trash
compactor. However, even though the process is called
static huckling, the transition from the prehuckling
to the postbuckling state is definitely a dynamic
process for thin shells. Dynamic huckling occurs when

the load duration is shorter than the structural


response time. In this dynamic "pulse" buckling, the
amplitude of the load is of necessity higher than the
static buckling load and, in general, the buckling
mode is different. Examples of dynamic pulse buckling
occur frequently in impact problems such as the impact
of an arrow on a solid target or the response of a
structure to blast loading. In both the quasistatic
and dynamic huckling cases, the huckling process is
initiated by imperfections in the materials and
geometry of the structure and nonuniformities in the
load application. In general, the critical buckling
load is governed by the amplitude of the imperfections,
so that structures with large imperfections have lower
critical loads.
This is the reason for the large
scatter in data of many experimental buckling studies.
Because of the importance of imperfections on the
buckling load, a third type of buckling problem is
possible. In this problem, the structure is subjected
to a static load below the critical static load level
and also a dynamic load. The dynamic load introduces
transient shape imperfections that can trigger buckling under the action of the static load. An example
of this problem occurs when a rocket in flight is
exposed to blast loading. In this study, we solved
the problem of a thin cylindrical shell under a static
axial load; however, the methods used were chosen so
that the analysis could easily be extended to include
the effects of combined static preload and dynamic
external radial loads.
The stability limit for an axially compressed
elastic cylindrical shell was determined more than
75 years ago by Lorenz. Southwell, and Timoshenko
(1961) from linear equilibrium equations.
These
equations were derived for a long cylindrical shell
with simply supported boundaries under the action of
a uniform axial load.
However, comparison of the
theoretical huckling load with experimental data for
thin, axially compressed shells showed that buckling
occurs at load levels significantly below those
predicted by this solution. This discrepancy in the
huckling loads has been the subject of considerable
study. Possible explanations for the difference

between the experimental buckling load and the


theoretical load limit included the effects of end
conditions, imperfections in shell geometry, imperfections in shell thickness, nonuniformities in material
properties, and nonuniformities in the load.
In
general, geometric imperfections are the primary
reason for the discrepancy between experiment and
classical theory for thin shells.
The pioneering work on the effect of geometrical
imperfections on shell stability was performed by
Donnell (1934) and Donnell and Wan (1950). A number
of publications have used the nonlinear Donnell shell
equations to examine the effects of initial imperfection distributions (Arbocz, 1974) on the axial huckling load. By using imperfections in a single made or
a combination of a few modes, the nonlinear shell
equations can be solved to obtain the buckling load.
With this approach, it has been shown that an imperfection in the shape of the preferred buckling mode at
a realistic amplitude can account for the discrepancy
between the experimental and theoretical buckling
loads. However, use of this method to predict the
buckling load requires knowledge of both the preferred
buckling mode and the "effective amplitude" of the
shape imperfection in that mode.
Solution of the
Donnell equations for a more general description of
the initial imperfections is considerably more
complex.
In addition, this method does not allow
calculation of the postbuckling strength of the shell.
In the theoretical work described above, the
amplitudes of the geometric imperfections were
unknown. Typically, the imperfection amplitude was
chosen such that the analytical prediction of the
buckling load matched experimental data.
More
recently, surveys have been performed to measure the
imperfections that naturally exist in cylindrical
shell structures. These data have been collected into
"data banks," and Arbocz (1982) suggests that they be
used to improve design criteria for buckling load
limits of thin shells. The imperfection data banks
can be used in calculations to more accurately model
the characteristics of the imperfections present in
tbin cylindrical shell structures.
Recent computational analyses of dynamic buckling
problems (Kirkpatrick and Holmes, 1988a.b) have
applied this approach, incorporating in the numerical
model the actual imperfections measured in the shell
or a statistically accurate description of the imperfections. In these studies, the dynamic buckling of a
thin cylindrical shell subjected to an external radial
impulse was modeled using finite element methods. The
imperfections measured on the shells or a statistically
accurate description of the imperfections was included
in the finite element mesh for the shells. This technique was very successful for predicting the dynamic
buckling response for thin shells and was applied to
the static axial buckling problem in our study.
Many static buckling problems have been solved
numerically. Many authors prefer using static analysis
for the buckling process, while others prefer dynamic
analysis.
Each of these techniques bas advantages
depending on the information required by the analyst
(Kroplin and Dinkler, 1986). Static analysis can rely
on static equilibrium states and simplifies the
numerical analvsis. However. static solutions mav not
simulate the transition from the prebuckling to the
postbuckling state and often must account for the
existence of a large number of bifurcation paths.
Dynamic analysis leads to a physically unique post-

buckling pattern. For this study, we were interested


in the postbuckling defolmations and strength in
addition to the buckling load limit. We also wanted
to be able to extend the analysis to the case of
buckling under a static axial prestress in combination
with an external impulsive pressure.
For these
reasons, we chose to use a dynamic analysis that could
cope with the difficult dynamic phenomena between the
pre- and postbuckling states following a unique
solution path.
The following sections describe the geometry of
the problem and the measurement of the imperfections
in the shell, the experiments and the experimental
results, and the numerical methods used in the
analysis.
We then compare the predicted buckling
response from the pretest calculation with the
experimental results. Finally, we perform posttest
calculations that incorporate the load asymmetry
measured in the experiments and again compare the
calculated and experimental results.
PROBLEM GEOMETRY AND IMPERFECTION MEASUREMENT
In our experiments, two tbin 6061-T6 aluminum
shells with radius-to-thickness ratios ( a h ) of 150
were tested under an axial load (Figure 1).
The
shells used in the experiments were manufactured by
spinning 30.5-cm-diameter,22.9-cm-longcylinders on a
mandrel. A 2.5-cm length at either end of the shell
was clamped between a shrink-fit, 5.1-cm-thick end
plate and a clamping ring to produce a 17.8-cm-long
test section in the shell.
When mounted in the
testing machine, the fixtures produced clamped end
conditions.

SHELL CHARACTERISTICS

- 6061-T6Aluminum
- 30.5cm Diameter
-a/h=150
- U2a = 0.58
- Displacement Ecundary
Conditions
Fig. 1 Problem definition
Geometrical imperfections of both the shells were
measured before testing. Each shell was mounted in a
modified lathe bed so that it could he rotated freely.
A potentiometer was used to measure the angular position of the shell, and a linear voltage displacement
transducer (LVDT) was used to measure the radial
position of the shell surface. Outputs from the LVDT

and potentiometer were recorded on a digital oscilloscope while the shell was rotated. Approximately
1800 data points were obtained around the shell, and
this measurement was repeated at 19 evenly spaced
axial locations.
Projected three-dimensional images of the two
imperfection sets are shown in Figure 2 . The spatial
resolution of the points plotted in the figure corresponds to the refinement of the mesh used later in the
The amplitude of the
finite element calculations.
imperfections was multiplied by a factor of 500 in the
figure to make them visible. The imperfections seen
on the first shell (Figure 2a) are similar in character
to the imperfections measured on shells made by other
manufacturing techniques (Arbocz, 1982, and Kirkpatrick
and Holmes, 1988). In particular, the dominant imperfections appear as ridges along the axial direction.

A Fourier analysis of the radial imperfections was


performed at the axial location at midheight on the
first shell. Figure 3 shows the modal amplitude of
the radial imperfections around the circumference
expressed as a percentage of the wall thickness. Over
the range of modes from 1 to 160, the magnitude of the
modal imperfections measured on this shell is 3 to 10
times less than that of the modal imperfections
measured on rolled and welded shells used in previous
studies (Kirkpatrick and Holmes, 1988).
The imperfections on the shells in this study are smaller
because of the spinning process used to manufacture
the shells. The spinning process forms a more perfect
cylindrical shell than rolling and welding, and
because these shells are imperfection sensitive, the
spun shells will have correspondingly higher buckling
loads.

(a) Shell 1 Imperfections

'0

MODE NUMBER

Fig. 3 Modal composition of measured imperfections

&J

In contrast to the first shell, the second shell


has imperfections of a different character (Figure 2b).
The imperfections on this shell have some dominant
features that run a significant distance around the
circumference. The character of these imperfections
may be a result of the spinning process, which could
produce imperfections with different character from
shell to shell. Another possible explanation for the
differences between the first and second shells is
that strain gages were applied to the second shell.
Strain gage application requires polishing the
aluminum surface under the gage for a good adhesive
bond and warming the shell to cure the adhesive. It
is possible that this process altered the character of
the imperfections.
EXPERIMENTS
(b) Shell 2 Imperfections

Fig. 2 Measured imperfection distributions

The cylinders were tasted in e 450-kN MTS servohydraulic testing machine operated in the displacement
crmtrol mode. The top end plate was rigidly mounted
to the upper support of the MTS machine to produce a
rigid clamped end condition at the upper edge of the
cylindrical shell. A load cell was mounted between

uniformity of the strain distribution produced by the


load. When the shell was first mounted in the MTS
machine and an initial preload was applied, the
uniformity of the strains around the circumference
varied by approximately + 5 % . Shims were used between
the loading plate and the end plate of the shell to
reduce the load asymmetry to approximately +4%. When
this level of uniformity was achieved, the second
experiment war performed.

the top end plate and the support to measure the load
time history of the shell. The lower end plate of the
shell was loaded by a parallel plate mounted rigidly
on the stroke arm of the MTS machine. The load was
by the loading plate moving upward at a
uniform velocity during the experiments.

The first experiment was performed at a loading


rate of 1.26 cm/s. Shell 1 buckled at a critical load
of 205 kN, which corresponds to an axial buckling
stress of 210 MPa.
The shell buckled into a mode
number of 9 , with a buckling pattern that was not
uniform around the circumference. The postbuckling
deformations for the first shell experiment are shown
in Figu~e4. The asymmetry of the buckling pattern
around the circumference suggested that the load may
not have been perfectly uniform around the circumference of the shell. Misalignment of the lower end
plate and the loading plate on the MTS arm would
produce the asymmetry in the load.

The second experiment was performed at a loading


rate of 0.20 m/s. Shell 2 buckled at a critical load
of 228 kN, which corresponds to an axial buckling
stress of 234 MPa.
This shell buckled in a mode
number of 9, wich a uniform buckling pattern. The
postbuckling deformations of the second shell are
shown in Figure 5. The two uniform rows of buckles
are slightly offset toward the upper end of the shell.
Uniformity of the axial strains measured in the experiment was within +3.6%. The measured postbuckling
collapse load was approximately 76 kN, which corresponds to a postbuckling average axial stress of
approximately 80 MPa.

To address the question of load asymmetry, we


placed axial strain gages at five locations around the
circumference of the second shell to measure the

Fig. 4 Buckled shape of Shell 1

Fig. 5 Buckled shape of Shell 2

70

PRETEST CALCULATION
The DYNA3D computer code used in the analysis of
shell response was developed at Lawrence Livermore
National Laboratory (Hallquist and Benson, 1986).
DYNA3D is an explicit finite element code for the
nonlinear dynamic analysis of structures in three
dimensions. The equations-of-motionare integrated in
time using the central difference method.
Spatial
discretization was achieved with the Belytschko-Tsay
shell element (Belytschko and Tsay, 1981). This is a
four node element with single point integration and a
stabilization procedure for the kinematic modes
(Belytschko and Tsay, 1983).
The Belytschko-Tsay
element offers the advantage of allowing a time step
size that is insensitive to shell thickness and larger
than that for corresponding brick elements.
This
feature permits more economical solutions and is very
important in this application, where many elements are
required for a good simulation.
The material model used in the calculation was an
elastic-plastic model with combined isotropic/kinematic
hardening (Krieg and Key, 1976), a Mises yield function, and an associated flow rule. A combination of
20% isotropic and 80% kinematic hardening was used. A
yield stress of 276 MPa was used in the material model,
with an elastic modulus of 69 GPa and a hardening
modulus of 586 MPa. These values were baaed on quasistatic tensile tests of 6061-T6 aluminum sheet stock.
The mesh for the pretest calculation had 100
elements uniformly distributed around the circumference
and 20 elements along the length. This distribution
produced elements with an aspect ratio of approximately
1 in the circumferential and axial directions. Thi
mesh used in the calculations is shown in Figure 6.
Because of the measured imperfections that are superimposed on the mesh generated for a perfect cylindrical
shell, no planes of symetry exist in the mesh. The
use of symmetry planes also places constraints on both
the locations in which buckles can form and the modes
into which the shell can buckle.
The mesh used in
these calculations modeled the entire shell to avoid
these difficulties

Fig. 6 Mesh used for finite element calculations

Both rotational and translational degrees of


freedom for the nodes at the upper end of the cylindrical mesh were fixed. Similarly, at the lower end of
the mesh, all degrees of freedom except the translation
along the axis of the shell were fixed.
Nodal
velocities along the shell axis were specified for
these nodes at a uniform rate of 10.2 cm/s.
To
facilitate a smooth load application at this rate, we
applied a linear initial velocity field to all the
nodes in the mesh, with initial velocities varying from
0.0 em/= at the fixed end to 10.2 cm/s at the loaded
end. This velocity field eliminates any stress wave
reverberations along the shell that would occur if the
velocity were specified only at the end of the shall.
Additional calculations that demonstrate the effect of
loading rate are discussed below.
The postbuckling deformations predicted by the
pretest calculation are shown in Figure 7.
The
calculation predicts that two uniform rows of buckles
around the circumference, slightly offset toward the
loaded end, will be produced by the axial load. The
predicted buckling made number is 10 at a critical
buckling load of approximately 235 kN, which produces
an axial buckling stress of 240 MPa.
Comparison of the pretest calculation with the
results of the first experiment shows that the calculation overpredicted the critical buckling load by
approximately 15%.
In addition, the calculation
predicted a uniform buckling pattern with a mode number
of 10, but a nonuniform buckling pattern with a mode
number of 9 was observed in the experiment. Although
the agreement between the analysis and experiment is
reasonable, we investigated further to explain the 15%
difference observed. In the next section, we describe
how more accurate solution requires introducing the
effect of load asymmetry.
POSTTEST CALCmATIONS
The first posttest calculation was performed using
the same mesh as the pretest calculation that incorporated the imperfections measured on Shell 1.
The
objective of the calculation was to investigate the
effects of the load asymmetry on the critical buckling
load and the postbuckling deformations. On the basis
of initial measurements of the strain in the second
experiment, before adjustments were made to reduce the
load asymmetry, a side to side load variation of 15%
was used in the calculation. The imperfection of the
load was input as a fractional change in the end
velocity proportional to 15% of the uniform velocity
(10.2 cm/s) times the cosine of the angular position.
The comparison of the postbuckling deformations
predicted by this calculation with the experimental
results are shown in Figure 8a. The shell buckled in a
mode number of 9 at an axial stress of approximately
195 MPa, which corresponds to a buckling load of 185
.
This buckling load is approximately 8% lower than
the value measured in the first experiment.
In
addition, the calculation properly predicts the mode
number and the nonuniform character of the buckling
pattern observed in the experiment.
These results
suggest that the load asymmetry was responsible far the
initial disagreement between the pretest calculation
and the experimental results for Shell 1.

Fig. 7

Buckle pattern from pretest calculation of Shell 1

(a) Shell 1

(b) Shell 2

Fig. 8

Comparison of posttest caledlations with experiments

72

The second posttest calculation was performed to


duplicate the the second experiment performed on Shell
2.
The mesh used the imperfection set measured on
Shell 2 (Figure 2b) and an imperfection in the load
equal to 3.6% of the uniform end velocity (10.2 cm/s)
times the cosine of the angular position plus a phase
shift chosen to approximate the measured distribution.
A comparison of the buckled shape predicted by this
calculation with the experimental result for Shell 2 is
shown in Figure 8b. The calculation predicted that the
shell would buckle at an axial load of 232 kN, which
corresponds to an axial buckling stress of 238 MPa.
This buckling load is within 2% of the experimentally
observed buckling load for Shell 2. The calculation
predicts a buckling mode number of 9, with a nonuniform
buckling pattern around the circumference. The experimentally observed mode number was also 9; however two
uniform rows of buckles were produced, with a regular
buckling pattern around the circumference.
The final two posttest calculations were performed
to demonstrate the effects of the loading rate on the
calculated buckling response and postbuckling strength.
The mesh from Shell 2 was used with loading rates of
20.3 and 5.1 cm/s, which are, respectively, twice as
fast and twice as slow as those used in the previous
calculations. The buckling patterns predicted by the
three calculations were very similar. A comparison of
the average axial stresses for the three calculations
is shown in Figure 9. This figure shows that there is
a slight increase in the peak compressive load each
time the loading rate is doubled. An explanation for
these differences is that buckling is initiated at a
critical average axial stress of 232 MPa but the
buckling process requires approximately 150 ps before
the axial stress begins to drop off. For the slowest
loading rate, the end displacements that occur during
this time increase the peak axial stress over the
However, at loading
buckling stress by only 3 MPa.
rates of 10.2 and 20.3 cm/s, the peak axial stresses
are 6 and 12 MPa greater than the buckling stress,
respectively.
These increases correspond to the
magnitude of the differences in the peak axial stresses
seen in Figure 9.
An additional observation made from Figure 9 is
that all three calculations predict the postbuckling
average axial stress to he approximately 115 MPa, which
corresponds to a postbuckling collapse load of approximately 112 kN. This value for the calculated posthuckling strength overpredicts the measured postbuckling
strength by nearly 50%. The difference may partly be
due to fundamental differences between the experiments
and the numerical simulations. These differences can
be seen in Figure 10, which plots the average axial
stress measured in the experiment (measured load
divided by shell cross-sectional area) against the
testing machine stroke arm displacement. A comparison
of Figures 9 and 10 shows that the compliance of the
cylinder and testing machine together is approximately
2.5 times the compliance of the shell alone from the
calculated response. Therefore, when the shell buckles
in the experiments, larger dirplacsments are produced
in the shell because of the compliance of the testing
machine and the stored energy that is released
(increased shell end displacements) as the load drops.
In the calculations, the end displacements are
specified and there is no jump in the end displacements
when the shell buckles. Therefore, the buckles in the
calculation immediately after the buckling occurs are
not as large as those in the experiment. The more
developed buckling pattern in the experiments may well
produce a lower postbuckling strength.

0.02

0.04

0.06

0.08

0.10

0.12

END DISPLACEMENT (cm)

Fig. 9 Comparison of average stress histories for


different loading rates

0.05

0.10

0.15

0.20

0.25

STROKE ARM DISPLACEMENT (cm)

Fig. 10 Measured average axial stress against stroke


a m displacement for Shell 2
SUMMARY AND CONCLUSIONS
Careful imperfection surveys were performed on
thin aluminum cylindrical shells tested under axial
The
compression beyond the critical buckling load.
first shell buckled at an average axial stress of
approximately 210 MPa. The buckling mode number was 9,
and the buckling pattern was not uniform around the
circumference or along the length. The nonunifomity
of the buckling pattern suggested that the load was
asymmetric.
Strain gages were placed on the second

shell to obtain a measure of the uniformity of the load


around the circumference.
The initial side-to-side
asymmetry wan approximately f15% when the shell was
first placed in the test fixture. With careful adjustment, the load symmetry was reduced to approximately
t3.6%, and the shell was then tested.
The shell
buckled at an average axial stress of approximately
238 MPa. The shell buckled into a mode number of 9,
with two uniform rows of buckles around the circumference. The postbuckling strength produced an average
axial postbuckling stress in the shell of approximately
80 MPa.
Two methods of analysis are available for performing numerical simulations of the buckling process. The
first approach is to perform a static analysis that
treats the axial buckling as a bifurcation process.
The static analysis approach allows simplifications in
the numerical methods used hut has the difficulty that
a number of different bifurcation paths are possible.
The second approach, which was used in this study, is
to perform a dynamic analysis of the buckling process.
When the dynamics are included in the numerical
analysis, a unique buckling solution is obtained.
Finite element simulations of the axial buckling
process were performed using the explicit three dimensional code DYNA3D. Imperfections were included in the
calculations by superimposing the measured imperfections on the mesh for a perfect cylinder and including
the load asymmetry in the displacement boundary
conditions. The finite element calculation predicted
the buckling mode and overpredicted the buckling load
by approximately 2% when both the geometric and load
imperfections were modeled.
The calculations overpredicted the postbuckling strength by approximately
50%; however, this discrepancy may have resulted from
the compliance of the testing machine, which was not
accounted for in the calculations.
Both this study and previous work on dynamic
buckling problems (Kirkpatrick and Holmes, 1988) have
shown that, when the imperfections in the geometry of
the structure and the imperfections in the applied load
are accurately modeled, the buckling process can be
accurately modeled by dynamic finite element analysis.
However, in both of these buckling processes, small
changes in the imperfections can result in comparatively large changes in the mode number or buckling load.
The consequence of this imperfection sensitivity is
that, although the solution obtained from a dynamic
analysis is deterministic, it can also be considered
chaotic. If the imperfections of a given structure are
known, the buckling load for that structure can be
accurately predicted, potentially within a few percent.
However, there will be a statistical variation between
the buckling load from the first structure and that
from a second structure made by the same manufacturing
process and loaded in similar fashion. Consequently,
the numerical modeling approach used in this study is
extremely useful when the buckling load for a single
structure needs to be accurately predicted. However,
this approach can also be useful for analyzing large
groups of structures by investigating statistical
variations in imperfections and the imperfection
sensitivity of the structures.

We feel that these modeling techniques (accurately


modeling the imperfections and using dynamic analysis)
are very useful for analysis of buckling processes and
potentially useful for many different problems beyond
buckling that incorporate imperfection sensitivity,
bifurcation, localization, or softening. Some examples
include fracture of brittle materials under static
loading, ductile or brittle fracture under dynamic
loading, and development of turbulence in fluid flow.
REFERENCES
J. Arbocz, 1974. "The Effect of Initial Imperfections on Shell Stability," Thin Shell Structures,
Prentice-Hall, Englewood Cliffs, New Jersey, pp 205245.
J . Arbocz, 1982, "The Imperfection Data Bank. A
Means to Obtain Realistic ~ u c k l i nLoads,"
~
in Buckling
of
Shells, Proceedings
of
a
State-of-the-Art
Colloquium, Springer, New York, pp. 535-566.
T. B. Belytschko and C. S. Tsay, 1981, "Explicit
Algorithms for the Nonlinear Dynamics of Shells," A M I Vol. 48, ASME, pp. 209-231.
T. B. Belytschko and C. S. Tsay, 1983, "A
Stabilization Procedure for the Quadrilateral Plate
Element with One-Point Quadriture," Int. J. Num. Meth.
Eng., 19, pp. 405-419.
L. M. Donnell, 1934, "A New Theory for the Buckling of Thin Cylinders Under Axial Compression and
ending," Trans.- ~ m .Soc. Mech. Eng., 56, ip. 795-806.
L. M. Donnell and C. C. Wan, 1950, "Effect of
Imperfections on Buckling of Thin Cylinders and Columns
Under Axial Compression," J. Appl. Mech., 17, pp. 7383.
J. 0. Hallquist and D. J . Benson, 1986, "DYNA3D
Users Manual (Nonlinear Dynamic Analysis of Structures
in Three Dimensions)," Lawrence Livermore National
Laboratory, Report UCID-19592, Revision 2.
S. W. Kirkpetrick and B. S. Holmes, 1988a,
"Structural Response of Thin Cylindrical Shells to
Impulsive External Loads," AIAA J.. 26, No. 1, pp. 96. A "

I",

S. W. Kirkpatrick and B. S. Holmes, 1988b, "The


Effect of Initial Imperfections on Dynamic Buckling of
Shells," ASCE J. of Eng. Mech. (in press).
R. D. Krieg and S. W. Key, 1976, "Implementation
of a Time-Dependant Plasticity Theory into Structural
Computer Programs," in Constitutive Equations in
Viscoplasticity:
Computational and
Engineering
Aspects, ASME 20, pp. 125-137.
B. Kroplin and D. Dinkler, 1986, "Dynamic versus
Static Buckling
of Thin Walled Shell
. Analysis
S r r u c u r e s . " in F i n i t e E l o m c n c Yethpds for P l o w and
Shell Scruct_uru, Vol. 2
Fomu1at:ons-=ndplgori-chms.
T.J.R. Ilughcs and F. Hinron Fds., P i n i . r i d ~ r Press,
Swansea, U.K.
S. P. Timoshenko and J. M. Gere, 1961, Theory of
Elastic Stability, McGrau-Hill, New York.

Você também pode gostar