In Elastic Buckling of Geometrically Imperfect Tubes

NOT FOR REPRODUCTION
Designing Under Pressure

Little, Ross, Short and Brown introduce a new
tool to better design pressure vessels.
Who should read this paper?
This paper will be of primary interest to structural designers involved in
designing pressure vessels, particularly submarine structures. At present, there
is very little non-military information on designing deep-diving submarine
pressure hulls.
Andrew P.F. Little
Why is it important?
Under external pressure, pressure vessels can suffer catastrophic collapse. The
purpose of the paper is to present easy-to-use design charts for use by
structural designers who design pressure vessels. The design chart is
innovative because it simplifies the design of complex structural failure modes,
particularly those for deep-sea applications.
The oceans cover some 71% of the Earth's surface, but only about 0.1% of the
oceans' bottoms have been explored. The charts will allow deep-diving
Carl T.F. Ross submarines to be designed to greater advantage for commercial exploitation and
for military purposes. The work described in this paper may ultimately improve
our ability for retrieving deep-sea methane and for the burial of greenhouse
gases, including carbon dioxide.
About the Authors

Andrew P.F. Little is the Principal Lecturer in Mechanical and Design Engineering
at the University of Portsmouth, UK. His main expertise is in the statics and
dynamics of submarine pressure hulls.
Daniel Short Carl T.F. Ross is a Professor of Structural Dynamics at the University of
Portsmouth in the UK. He is an expert in statics and dynamics of submarine
pressure hulls.
Daniel Short is a Mechanical Engineering student at the University of Portsmouth

in the UK. His expertise is in the strength of submarine pressure hulls.
Graham X. Brown is the Chief Mechanical Engineer at Sonardyne Ltd., in Yately,

Hampshire, UK. His expertise is in the research, design and construction of
deep-sea pressure vessels.
Graham X. Brown
Copyright Journal of Ocean Technology 2008 Ocean Sovereignty, Vol. 3, No. 1, 2008 75
INELASTIC BUCKLING OF GEOMETRICALLY IMPERFECT TUBES UNDER
EXTERNAL HYDROSTATIC PRESSURE
ANDREW P.F. LITTLE1, CARL T.F. ROSS1, DANIEL SHORT1 & GRAHAM X. BROWN2
1 Dept. of Mechanical & Design Engineering, University of Portsmouth, Portsmouth, United Kingdom.
2 Sonardyne International Ltd, Yateley, Hants. United Kingdom.
ABSTRACT
The paper reports on the buckling of 12 thin-walled geometrically imperfect tubes, which were tested to destruction
under uniform external hydrostatic pressure. The paper also reports on other similar tests to destruction, carried out on
quite a large number of geometrically imperfect tubes.
Theoretical studies were also carried out with well-known analytical solutions, together with a numerical solution using
the famous finite element computer package, namely ANSYS.
Whereas the theoretical analyses agreed with each other, they did not agree with the experimental data for the shorter
tubes; this was because the shorter tubes collapsed by inelastic instability due to initial geometrical imperfections of the
tubes. Exact analysis of slightly geometrically imperfect tubes, with random distribution, has so far defied reliable
theoretical solutions. However, this paper presents a design chart, which can cater to these geometrical imperfections.
The design chart may also be suitable for large vessels such as submarines, off-shore drilling rigs, silos, etc.
KEYWORDS.
Geometrically imperfect tubes, initial out-of-roundness, inelastic buckling, external pressure, von Mises,
finite elements, ANSYS.
NOMENCLATURE Pexp experimental buckling pressure

Ppred predicted buckling pressure
A mean shell radius
PKD Plastic Knockdown Factor
d mean shell diameter
SF Safety Factor
E Youngs modulus of elasticity
t shell wall thickness
L unsupported length of cylinder
Windenburg thinness ratio
L0 overall length of a cylinder
Poissons ratio
n number of circumferential
yp yield stress
lobes formed
Pcr critical (theoretical) buckling
pressure
Pdesign design buckling pressure = Ppred / SF
76 THE JOURNAL OF OCEAN TECHNOLOGY Reviews & Papers Copyright Journal of Ocean Technology 2008
1. INTRODUCTION Another mode of failure is known as axisymmetric
Circular cylinders under external pressure often appear in deformation, where the cylinder implodes
the form of submarine pressure hulls, torpedoes, off- axisymmetrically, so that its cross-section keeps its
shore drilling rigs, silos, tunnels, immersed tubes, circular form while collapsing, as shown in Figure 4.
rockets, medical equipment, food cans, etc. Such
vessels are good for resisting internal or external
pressure, however under uniform external pressure they
can collapse at a
fraction of the pressure
that will cause failure
under internal
pressure. Failure of
these vessels under
Figure 1: Shell instability. uniform external
pressure is called non-symmetric bifurcation buckling or
shell instability [1 to 3] and is shown in Figure. 1. Figure 4: Axisymmetric collapse.
To improve the In this study, we will be concerned with elastic and

resistance of these inelastic shell instability; as such vessels can collapse at
vessels to the effects pressures of a fraction of that to cause the vessels to fail
of uniform external under internal pressure. The resistance to external
pressure, the vessels pressure is further worsened if the vessel suffers from
are usually stiffened by initial out-of circularity. If the initial out-of-circularity is
Figure 2: Ring-stiffened circular cylinders.
ring stiffeners spaced large and regular, such as that considered by Bosman et
at near equal distances apart, as shown in Figure. 2. al [8], then analysis by non-linear numerical methods is
satisfactory. If, however, the initial out-of-circularity is
If, however, the ring stiffeners are not strong enough, the small and random, then exact or near exact theoretical
entire flank of the vessel can collapse bodily by a mode analyses have so far been defied. Ross, however, has
called general instability and as shown in Figure 3 [3 to 7]. shown that such vessels can be analysed by his design
chart of Figure 3.4 [3]. This design chart, however, was
for near perfect vessels and is not suitable for vessels
with small but significant initial geometrical imperfections,
such as considered in the present paper. The process
therefore is to calculate the theoretical buckling pressure
for a perfect vessel by the von Mises formula [1 to 3],
together with the Windenburg thinness ratio [2,3]. Then,
using the thinness ratio, a plastic knockdown factor (PKD)
can be determined from the design chart and divided into
the theoretical von Mises buckling pressure to give the
predicted buckling pressure, where.
Figure 3: General instability.
= Windenburg thinness ratio.

Pcr = theoretical von Mises elastic buckling pressure.
Windenburg thinness ratio.
= Predicted buckling pressure = Pcr/PKD.
gn = Ppred/SF
P = Predicted buckling pressure = Pcr/PKD. The predicted pressure Pcr used in this paper is that of
a safety factor. pred
Pdesign = Ppred/SF von Mises.
Pdesign=Pcr/(PKD*SF)
SF = a safety factor.
i.e. Pdesign=Pcr/(PKD*SF) Pcr1 = von Mises calculations for the 2006 investigation
Pcr3 = von Mises calculations for the current investigation.
1.3 WINDENBURG AND TRILLINGS THINNESS RATIO

1.1. VON MISES BUCKLING PRESSURE
von MisesThis states that
buckling the elastic instability pressure for a thin
pressure.
walled circular cylindrical shell simply supported at both
states that the elastic instability pressure for a thin walled circular Windenburg and Trilling obtained their thinness ratio [1-3]
ends and subjected to combined actions of uniform by the following approach:
drical shell simply supported at both ends and subjected to combined
lateral and axial pressure [1 to 3] is given by equation (1).
ns of uniform lateral and axial pressure [1 to 3] is given by equation (1).
They noted that experimental tests on short circular
E t / a section tubes under external hydrostatic pressure had
Pcr
2
>
n 1 0.5Sa / l
2
@ found that they fail when their circumferential stress
reaches yield, according to the well-known Boiler formula
2 2
[3,6], as follows.
1 t2 2 Sa
n 1
yp = pd/(2t)
u
> @ 12a 2 1 v l
2 2
n 2 l / Sa 2 1

or p=yp *(2t)/d , (3)
(1)
(1) where p= pressure to cause yield.
yp =Yield Stress
Where, d= 2a
re, Pcr = buckling pressure;
Pcr = t buckling
= wallpressure;
thickness of circular cylinder;
t = awall=thickness
mean radius of circular
of circular cylindrical shell;
cylinder; They further noted that experiments had shown that
/ = unsupported length of cylinder; when long thin-walled circular tubes are subjected to
va == mean radius
Poissons of circular cylindrical shell;
ratio;
E = Youngs Modulus; external hydrostatic pressure, they can buckle elastically
nl == unsupported length of cylinder;
vNo. of
= circumferential
Poissons ratio;lobes. according to the von Mises or the DTMB formula of
E = Youngs Modulus;
n = No. of circumferential lobes. equations (1) & (2). Experiments on circular section
Windenburg and Trillings buckling pressure.
thin, ratio, namely [ 2,3], has a value of less than
tubes of intermediate and shorter lengths, when a
enburg and1.2.Trillings paper AND
WINDENBURG states the buckling
TRILLINGS equation
BUCKLING for a long,
PRESSURE thinness
5
ctly circular cylinder, under uniform external pressure, is given by
Windenburg and Trillings paper states the buckling equation 0.4, have shown that they fail somewhere in-between
for formula
tion (2). This a long, thin, perfectly
is also knowncircular
as thecylinder,
David under uniform
Taylor the pressures of equations (1) and (3). Windenburg and
Model Basin
MB) [3]. external pressure, is given by equation (2). This formula is Trilling [2, 3] argued that if we equated equations (2) and
also known as the David Taylor Model Basin (DTMB) [3]. (3), we can get a thinness ratio relating these two modes
of failure, which will enable us to precisely predict the
2.42 E t / 2a
5/ 2
collapse pressures for intermediate circular cylinders;
Pcr

1 v >
2 0.75
l / 2a 0.447 t / 2a 1/ 2
@ they called this their thinness ratio . Now if we
examine equation (2), we can see that in the denominator
(2) on the right(2) hand side of equation (2), that l/d is much
Where, larger than 0.45*(t/d)0.5, thus if we neglect 0.45*(t/d)0.5

re, Pcr = buckling pressure; and assume that v = 0.3, we can simplify
Pcr = tbuckling
= wall pressure;
thickness of circular cylinder; equation (2) to the form:
t = a = mean radius of circular
wall thickness of circular cylinder; cylindrical shell;
a = Emean = radius
Youngs Modulus;cylindrical shell;
of circular Pcr = 2.6*E*(t/d)2.5 / (l/d) (4)
E =
v = Poissons
Youngs Modulus;
ratio.
v = Poissons ratio.
Equating (3) and (4), we get yp*(2t) / d = 2.6E*(t/d)2.5 / (l/d),
Or yp*t/d = 2 * E*(t/d)2.5 / (l/d),
unattached to it. That is, the boundary conditions for
Or 2=(l/d) / (t/d)-1.5 * ( yp / E)
each specimen were assumed to be simply-supported
Or = [(l/d)2 / (t/d)3]0.25 * ( yp / E)0.5

between the O rings in the closure discs. The ends of
the specimen were free to rotate during the collapse of
N.B. Windenburg and Trilling squared in the above

each specimen.
the value of would be approximately one.

calculation, so that for most intermediate length vessels, The tank lid was fitted and screwed down firmly.
The bleed valve at the top of the tank was opened and
2. EXPERIMENTAL the trapped air expelled from the tank by gently pumping
TESTING in water.
In order to obtain the
required chart to After the trapped air was expelled, the bleed valve was
enable the theoretical sealed to make the system pressure-tight.
predictions to be
made, experimental The hydraulic pressure in the tank was increased via the
work had to be hydraulic pump in small increments.
preformed. This was
done by using a high- The pressure gauge was carefully monitored until failure
pressure test tank (see occurred. Failure occurred with a bang which could easily
Figure 5). Specimens Figure 5: Pressure Test Rig. be heard, together with a large fall in pressure.
were tested to destruction and the failure pressures
recorded. The collapse pressure was recorded and the pressure
drop noted, as well.
Test Equipment:
High-Pressure Vessel The hydraulic pressure was released and then the tank
Hydraulic Pump lid was removed to retrieve and examine the collapsed
Pressure Gauge specimen.
2.1 THE TESTING PROCEDURE

The pressure pump was a hand-operated hydraulic one
that could exert a maximum pressure of 6,000 psi (414
bar), and as it was hand operated, line losses were
negligible. Additionally, as it was hand-operated, the
applied pressure could be increased in increments of
about 1psi (0.07 bar); thus, the experimental buckling
pressures were precisely determined. The tank was
capable of sustaining a pressure of 3,000 psi ( 207 bar).
The closure discs were push-fitted into the ends of each

specimen to seal each specimen and to make it
watertight. A photograph of the end bungs is shown in
Figures 6 to 8. The specimen was then placed into the
pressure tank, just resting in the tank itself and Figure 6: The end bungs or closure discs.
The two mild steel end caps were machined for push fit
connections into the tube ends; these can be seen in
Figures 6 to 8. Sealing was achieved by the use of a size
As568A-233 O-ring, manufactured in Nitrile.
3. ANALYSES
3.1 USING MISESNP FOR THE RESULTS OF NAGOPPAN [9]
Using the computer program MisesNP [3] the vessels of
Nagoppan [9] were first analysed; the results are
reproduced here because this data will be used to
produce a more heavily populated design chart in the
present paper than that provided by Nagoppan. It is
necessary to produce a more heavily populated design
chart, because many such vessels collapse at lower or
Figure 7: Some of Nagoppans models, with end bungs. higher buckling pressures than expected. Such vessels
are said to give haphazard or rogue results.
2.2. THE TEST SPECIMENS
The test specimens used for this experiment were MisesNP [3] is a DOS based program devised to
aluminum alloy tubes. The tubes were supplied by calculate the shell instability buckling pressures of
Sonardyne for research into the buckling effect of cylinders; the program was written by Ross [3]. MisesNP
Aluminum 6082-T6 seamless tubes. The two supplied uses the von Mises formula [1], together with the formula
tubes came in lengths, which were machined to the of Windenburg and Trillings [2]) to calculate the buckling
desired lengths for each test specimens, Figure 8; their pressures of circular cylinders under uniform external
details were as follows: pressure, together with their thinness ratios; these were
for an isotropic material (see Table 1).
When using MisesNP the following parameters had to be

inputted:
1. Unsupported Length in mm = L
2. Mean Radius in mm
3. Wall Thickness in mm
4. Youngs Modulus in MPa
5. Poissons Ratio
6. Yield Stress in MPa
Figure 8: Test Specimens and End Caps.
Outer Diameter = 50mm

Wall Thickness = 1.7mm
Youngs Modulus = 70,000 MPa
Poissons Ratio = 0.3
Yield Stress = 250 MPa
Density = 2,620 kg/m3
capabilities and complex finite elements,
together with inelastic material models. This
aids the designer to simulate an accurate
prediction of how a structure behaves when a
load has been applied. For every simulation
that is performed, certain parameters have to
be set, as follows:
Model dimensions; unsupported length,

mean radius and wall thickness;
Material properties i.e. Youngs modulus,
Poissons ratio and density;

Boundary conditions; these were assumed
to be simply-supported, similar to Portsmouths

Table 1: MisesNP results for calculating properties. in-house program namely MisesNP;
Structural conditions, which were Eigen
3.2. PLASTIC KNOCKDOWN FACTOR (PKD) buckling in this case.

Knowing the experimental buckling pressures for Tubes 7
to 18, the PKDs were determined for these tubes, as 3.3.1 ANSYS SHELL 93 METHODOLOGY
shown in Table 2. Theoretical calculations using Windenburg
and Trillings formulae gave us the buckling pressure Pcr Shell 93 is an eight node isoparametric rectangular
and thinness ratios ( ). The experimental buckling element; it is a very popular and a well-tried and reliable
pressure for the cylinders were denoted by the symbol element. A brief description of how each vessel was
Pexp. From these results, it was possible to calculate the analysed is as follows:
plastic knockdown factor, namely (PKD) [3], where:
1. Model dimensions and properties
PKD= Pcr/Pexp The first step was to create a solid cylinder using the
models dimensions. Once this was entered, the model
Note:
Experimental work not carried out in the 2006 Table 2: Results for calculating PKD
investigation was represented by the symbol (-);
this was because of the height of the pressure
tank wasnt long enough for some of the models.
To generate a design chart, the calculated PKD

was plotted against 1/ this was successfully
achieved by Ross [3]; but his design chart could
not cope with shorter and thicker models.
3.3. USING ANSYS

ANSYS is a finite element software package that
addresses many problems in engineering
science, but especially problems in structural
mechanics. It also provides nonlinear [8]
now had to be converted into a tube. Using the element applied on the outer surface of the cylinder, the value of
type command, the Shell 93 8-node element was 1MPa was theoretical and represented a pre-stress
selected; this converted the solid cylinder into a tube and external pressure. The next stage was to apply a pre-
the material properties were then applied. stress option and to perform a static analysis.
2. Meshing 5. Eigen buckling

Once the tube had been generated, it was possible to After the static analysis was carried out, the buckling
increase the precision of the analysis by using the mesh analysis could be done. This called for a new analysis to
tool command, namely automesh. One can define how be performed using the Eigen buckling option. The
many elements one would like to analyse and for this buckling mode was set to 5 using the Block Lanczos
investigation all analyses had an average of about 1,000 and the load step option set to the same mode value of
elements. To judge the precision of this procedure, 5; this would give the 5 lowest buckling pressures.
several of the tubes were analysed with less than half
this number of elements and the difference in buckling
pressures predicted when using about 1,000 elements
compared with less than about 500 elements was less
than 1%. Thus, it was decided that all models could be
auto-meshed with about 1,000 elements.
3. The boundary conditions Figure 10: Buckling of a 189 mm tube.

These relate to how the tube was constrained and how
the loading was applied. For this investigation the left 6. Results
hand side of the tubes was constrained along all three Using the result summary command, it was possible to
translational axes, namely X, Y and Z and the right hand read the buckling pressure for the first five buckling
side was constrained along the translational X and Y modes, the lowest of these was recorded and tabulated
directions only; this enabled the tube to move along the Z in Table 3. Additionally it was possible to run the
axis (axially) and to rotate about both ends. Thus, the simulation property by using the animate command.
boundary conditions represented simply-supported ends Figure 10 shows a snap shot of the buckling mode of a
and it was believed that this combination of constraints typical cylindrical tube.
best represented the effects of the end caps (see Figure 9).
The required buckling pressure was the lowest of the 5
4. Structural analysis values generated for each vessel and the value of n, the
For the structural analysis a pre-stress of 1MPa was number of lobes that the vessel buckled into; this was
obtained by graphically plotting the eigenmode,
corresponding to the appropriate value of
buckling pressure on the screen; i.e. n
was counted. The buckling pressures
could be obtained either from a table or
from the graphical plot of the
eigenmode, which was plotted on the
screen; where it was referred as a frequency value.
Figure 9: The mesh and the boundary conditions that were applied to the cylinder.
Tube lengths were chosen to correspond to their predicte
Table 4).
Model Unsupported lambda Pcr1 Ansys Shell 93 populate the chart and length PKD
Length L O 1/ lobes pressure
(m) (n) (MPa) continue the study, it was
tube1 0.378 2.058 0.486 2 9.80
tube2 0.378 2.058 0.486 2 9.80 decided to select tube mm
tube3 0.378 2.058 0.486 2 9.80
tube4 0.315 1.879 0.532 2 10.36 lengths for further testing 230 1
tube5 0.315 1.879 0.532 2 10.36
tube6 0.315 1.879 0.532 2 10.36 that would enable more 160 1.5
tube7 0.252 1.681 0.595 2 11.73
tube8 0.252 1.681 0.595 2 11.73 population of the sparse 103 2
tube9 0.252 1.681 0.595 2 11.73
tube10 0.189 1.456 0.595 2 16.45 areas, thereby improving 83 3
tube11 0.189 1.456 0.687 2 16.45
tube12 0.189 1.456 0.687 2 16.45 the uniformity of the 63 4
tube13 0.126 1.188 0.842 2 24.08
tube14 0.126 1.188 0.842 2 24.08 graphic display. 50 4
30 5
tube15 0.126 1.188 0.842 2 24.08
tube16 0.063 0.840 1.190 3 48.45
tube17 0.063 0.840 1.190 3 48.45
tube18 0.063 0.840 1.190 3 48.45 3.5.2 TUBE LENGTHS Table 4.
Tube lengths were chosen
Table 3: Ansys Shell 93 Results for Aluminium 6082-T-6 seamless tubes.
Table 3: Ansys Shell 93 Results for Aluminium 6082-T-6 seamless tubes. Table 4.
to correspond to their predicted PKD values (see Table 4).
3.4 NAGOPPANS RESULTS A tolerance of 0.05mm on the length of each tube was us
Naggopans results [9] appeared to be very successful, A tolerance of 0.05mm on the length of each tube was
adequate for parting off on a standard lathe and it amoun
but they were sparse for shorter and thicker tubes and used; this was adequate for parting off on a standard
because of this they were not completely reliable; the itshortest
was tubes
lathe and it of about
amounted to a0.17%
toleranceandfor thefor the longest tub
shortest
because of this that the current work was carried out. As tubes of about 0.17% and for the longest tubes of about
17
The effects of these length tolerances were negligible on
many such vessels collapse at unexpectedly higher or 0.02%. The effects of these length tolerances were
lower pressures than predicted, it was necessary results
to andnegligible
the theoretical
on the effects predictions.
of the results and the theoretical
obtain a more densely populated design chart. A predictions.
photograph of one of Nagoppans buckled models,
together with an end bung, is shown in Figure 11.
3.6 Meteorology
3.6 METEOROLOGY
To gain an Toaccurate representation
gain an accurate representation of theof geometry
the geometry of the of th
3.5 THE CURRENT WORK tubes, they were measured using a Computer Controlled
measured using a Computer Controlled Measuring (CMM
Measuring (CMM)
3.5.1 JUSTIFICATION FOR TUBE LENGTH SELECTION Specimen was machine.sent to Solent Mould Tools Ltd. in Waterlo
A 63mm
Nagoppans design chart was sparsely populated in the Specimen was sent
measured. They took points around the circumference of
PKD regions of 0.1 to 1.0 and from 0.8 to 3 and as many to Solent Mould
locations
large vessels will fall into this region, it was necessary to (see
Tools Figure
Ltd. in 12 for the results of a cross-section
extend his design chart. Thus, in order to more sensibly Waterlooville,
HANTS to be
Figure 11: A closure disc with a buckled specimen. measured. They
Figure 12: Cross section of tubing.
took points around
the circumference of the tube at 4 locations (see Figure
12 for the results of a cross-section of the tubing).
A 230mm long specimen was sent to the metrology

laboratory at the University of Portsmouth. The aim of the
measuring process was to see what the profile of the
tube looked like, internal and externally. 100 Points were
taken around the circumference in two areas 1) Z -
Figure Then
5.0mm, 2) Z -224.99mm. 12: theCross
resultingsection
data was of tubing
plotted on to a graph, which magnified the profile.
Copyright Journal of Ocean Technology 2008 193, No. 1, 2008 83

Ocean Sovereignty, Vol.
very long,
very long, its buckling
its buckling resistance
resistance will bewill
verybe small,
very small, the vessel
the vessel suffering
suffering failurefailure
inwas
a sent
flattening
in a flattening
A 230mm long specimen to mode mode 13. (i.e. at
(i.e.laboratory
the metrology ovalling),
ovalling), as shown
the as shown in Fig.in13.
Fig.
University of Portsmouth. The aim of the measuring process was to see what
the profile of the tube looked like, internal and externally. 100 Points were
The resulting data confirmed that the tubes were
taken around the circumference in two areas 1) Z -5.0mm, 2) Z -224.99mm.
eccentric and that their profiles were consistently similar
Then the resulting data was plotted on to a graph, which magnified the profile.
throughout (see Table 5). The reason for this was

The resulting data confirmed that the tubes were eccentric and that their
because the aluminium was drawn through a die; if the
profiles were consistently similar throughout (see Table 5). The reason for this
die because
were notthepositioned correctly, the tube would Figure
Figure13: 13:Ovalling
Ovallingofofcylinder.
cylinder.
was aluminium was drawn through a die; if thehave
die were not
had the same
positioned fault
correctly, thethrough
tube would itshave
length.
had For the purposes
the same ofits length.
fault through
For
thisthe purposes
study, of this
we had tostudy,
accept we that
had toallaccept that all specimens
specimens had had similar
From
From experimental
experimental results carried out in the 2006 investiga
cross-sections throughout.
similar cross-sections throughout. Figure 13: Ovalling of cylinder.results carried out in the 2006 investig
present
presentpaper, paper,it itwas wasclearclearthat
thatthisthiswaswastrue.
true.AllAllcylinders
cylinderso
13: in this manner,
true. Allat a alower
lowerpressure
pressurethan thanthat
this was cylinders over 189mm failed in this predicted by th
External Internal Figure
Figure 13:
in
Ovalling Ovalling
this manner, ofatcylinder.
of cylinder. that predicted by
manner, at a lower pressure than that predicted by the
Z-5.0mm Z-224.9mm Z-5.0mm Z-224.9mm
Windenburg
Windenburg and Trillings calculations (see Table
Table6).
No. of Points 100 100 100 100
von Mises and and Trillings
Windenburg calculations
and Trillings (see(see
calculations 6).
0.0416 From0.0423 experimental results0.047 carried out in the 2006 investigation and in the
Variance (mm)
Circularity (mm)
From0.155
experimental
0.1578
results
0.0465
0.1893
carried
0.1902
outTablein the6). 2006 investigation and in the
Diameter (mm) 50.805
present present
paper,paper,
50.806
it wasitclear
47.514
was clear 47.516
that this Figure
thatwas
Figure 14
thistrue.14shows
was true.
shows
All the
All cross-sections
cylinders
the
cylinders over 189mm
cross-sections
over 189mm offailed
ofthethetheoretical
theoreticalcircum
failed circ
Figure 14
patterns of shows
the the cross-sections
buckling modes ofdue
the theoretical
to elastic andinstability,unu
instability,
Table 5: CMM results data for results
the 230 data
mm tube.
in thisinmanner,
Table 5: CMM
this manner, at a lower
for the 230 mm tube.
at a lower pressure pressure
patterns
than thanof
that that
thepredicted
predicted buckling
by the byvon
modestheMises
von
due Mises
toand
elastic
circumferential wave patterns of the buckling modes due
Windenburg and Trillings pressure
calculations
pressure andandFigs.
(see Table
Figs. 1515
6).&&1616show the collapsed vessels.
3.7
3.7Experimental
EXPERIMENTAL
Windenburg
Results.
RESULTS
and Trillings calculations to (see
elasticTable 6).under
instability, external show
pressurethe andcollapsed
Figures vessels.
The
Theexperimental
experimentalresultsresults
are givenare
in Table
given6. in Table 6. 15 & 16 show the collapsed vessels.
FigureFigure
14 shows14 shows the cross-sections
the cross-sections of theoftheoretical
the theoretical circumferential
circumferential wave wave
MODESpatterns
3.8 FAILURE patterns of theofbuckling
the buckling
modes modes
due todue to elastic
elastic instability,
instability, underunder external
external
==22vessels. nn==33 nn==44

Ross states in his book [3], that under uniform external
pressure
pressure and Figs.
and Figs. 15 & 16 15 show
& 16 show
pressure, a thin-walled circular cylinder may buckle in the nnvessels.
the collapsed
the collapsed
manner shown in Figure 1; usually at a fraction of that

pressure required causing axisymmetric yield. If the
cylinder is very long, its buckling resistance will be very
small, the vessel suffering failure
20 in a flattening mode (i.e. Figure 14: Circumferential wave patterns for buc
ovalling), as shown in Figuren13.= n 2= 2 n = n3 = 3Figure n =14:
n4 Circumferential
= 4 n = n8 = wave 8 patterns for bu
From experimental results carried out in the 2006

investigation and in the present paper, it was clear that 19
19
Figure 14: Circumferential wave patterns for buckling modes.
Figure
Figure 14: Circumferential
14: Circumferential wavewave patterns
patterns for buckling
for buckling modes.
modes.
19 19
Table 6: Experimental Results.

Table 6: Experimental Results.
3.9 PICTORIAL RESULTS OF EXPERIMENTS CARRIED 3.10 USING MISESNP
OUT IN 2007 When using MisesNP (Figure 17) certain parameters had
to be set for this analysis; these were described in
Section 3. A screen shot when using MisesNP is shown
in Figure 17.
Figure 15: Picture of all specimens that were tested in 2007. Figure 17: Screen shot of MisesNP when calculating Pcr3 for a
Model Tube.
Models 2-3: Length 160mm, buckled due to inelastic

shell instability. 3.11. PLASTIC KNOCKDOWN FACTOR FOR THE
PRESENT SERIES
Theoretical calculations using Windenburg and Trillings
theorem produced the buckling pressures Pcr, together
with the thinness ratios ( ). Experimental results gave
values for the buckling pressure for the cylinders, namely
Pexp. From these results, it was possible to calculate the
plastic knockdown factor, namely PKD [3]; see Table 7,
Models 7-9: Length 83mm, buckled due to inelastic shell where
instability; all rupturing. PKD= Pcr/Pexp
Note:
The symbol (-) represents missing data that was not carried
out during the experiments of 2007; this was due to the
maximum pressure constraints of the pressure tank.
3.12. USING ANSYS SHELL 93

ANSYS predictions were carried out for each of the
Models 10-12: Length 63mm, all ruptured. specimen lengths, so that direct comparison could be
made to the other prediction methods, the results of the
analysis are tabulated in Table 8.
The following parameters had to be fed in:

Model dimensions; unsupported length, mean
radius and wall thickness;

Material properties i.e. Youngs modulus, Poissons
Figure 16: The collapsed vessels.
ratio and Density;
Boundary conditions;
Structural conditions (i.e. Eigen buckling).
length lambda Pcr3 von Mises Pcr3 DTMB PKD Pexp Pexp was decided to geometrically model
Model L 1/ Lobes Pressure Pressure Pressure Pressure 2007 2007
m MPa bar Psi MPa MPa Psi
the eccentric tube in Pro Engineer and
Tube 1 0.211 1.538 0.650 2 8.69 86.9 1260 9.84 0.869 10.00 1450 then import this model into ANSYS to
Tube 2 0.141 1.257 0.800 2 14.83 148.3 2150 14.88 1.144 12.96 1880
Tube 3 0.141 1.257 0.800 2 14.83 148.3 2150 14.88 1.034 14.34 2080 carry out a feasibility study to see if a
Tube 4 0.084 0.970 1.031 3 24.62 246.2 3570 25.48 1.552 15.86 2300
246.2
plastic buckling analysis of an
Tube 5 0.084 0.970 1.031 3 24.62 3570 25.48 1.464 16.82 2440
Tube 6 0.084 0.970 1,.031 3 24.62 246.2 3570 25.48 1.552 15.86 2300 eccentric tube could be carried by
Tube 7 0.064 0.847 1.181 3 32.99 329.8 4784 33.99 1.899 17.37 2520
Tube 8 0.064 0.847 1.181 3 32.99 329.9 4784 33.99 1.914 17.24 2500
ANSYS. Also to see the static stress
Tube 9 0.064 0.847 1.181 3 32.99 329.9 4784 33.99 2.010 16.41 2380 distribution for a tube that was
Tube 10 0.044 0.702 1.425 4 50.96 509.6 7389 51.02 2.532 20.13 2920
Tube 11 0.044 0.702 1.425 4 50.96 509.6 7389 51.02 2.639 19.31 2800 eccentric about its longitudinal axis.
Tube 12 0.044 0.702 1.425 4 50.96 509.6 7389 51.02 2.621 19.44 2820
748.8
From the CMM data obtained from
Tube 13 0.031 0.589 1.700 4 74.88 10858 75.66 - - -
Tube 14 0.031 0.589 1.700 4 74.88 748.8 10858 75.66 - - - Solent Mould Tools, two three-
Tube 15 0.031 0.589 1.700 4 74.88 748.8 10858 75.66 - - -
Tube 16 0.011 0.351 2.849 5 283.28 2832.8 41076 294.4 - - -
dimensional models were generated.
Tube 17 0.011 0.351 2.849 5 283.28 2832.8 41076 294.4 - - -
Tube 18 0.011 0.351 2.849 5 283.28 2832.8 41076 294.4 - - -
Table 7: Results for calculating PKD

Table 7: Results for calculating PKD
3.13.1 MODEL 1:
3.13
UNote: USING PRO ENGINEERS MECHANICA ECCENTRIC TUBE
The
Onesymbol
part of (-)
therepresents
project was missing data
to carry outthat was not carried
a feasibility study out during
Using the X and Y
experiments of 2007; this was due to the maximum
of a plastic buckling analysis by ANSYS of a tube that pressure constraints of the a
co-ordinates,
pressure tank. about its longitudinal axis. Although
was eccentric consistent eccentric
ANSYS was capable of analysing such a vessel, its tube was generated.
3.12. Using ANSYS Shell 93 Figure 18: Eccentric tube with
geometric modeller was not as good as that of Pro Then a uniform
pressure loading.
Engineer. However, although Pro Engineers geometric pressure was applied
ANSYS predictions were carried out for each of the specimen lengths, so that
modeller was very good, its finite element capabilities did to its exterior surface using Mechanica. The pressure
direct comparison could be made to the other prediction methods, the results
not allow plastic buckling and were therefore not as good value was that of Pcr3 for model 4 (see Figure 18).
of the analysis are tabulated in Table 8.
as ANSYS. However, Pro Engineers model could be
imported
The intoparameters
following ANSYS andhad alsotoPro Engineers
be fed in: Mechanica Results from Mechanica showed a maximum stress of
was1.capable of statically analysing a tube that was
Model dimensions; unsupported length, mean radius and 450MPa was acting at both ends of the tube, together
wall thickness;
eccentric aboutproperties
2. Material its longitudinal axis. modulus,
i.e. Youngs For thesePoissons
reasons itratio and
withDensity;
a maximum displacement of 0.17mm (see Figure 19);
3. Boundary conditions; this tied in with the positions of
4. Structural
Model conditions (i.e. Eigen
Unsupported buckling).
lambda P Ansys 93
cr3 experimental failure of the tubes and
L O
3 1/3 lobes pressure
(m) (n) (MPa)
showed that Mechanica was of use for
Tube 1 0.211 1.538 0.650 2 14.16 this analysis.
22
Tube 2 0.141 1.257 0.800 3 23.13
Tube 3 0.141 1.257 0.800 3 23.13
Tube 4 0.084 0.970 1.031 3 32.1
Tube 5 0.084 0.970 1.031 3 32.1
3.13.2 MODEL 2: ECCENTRIC TUBE
Tube 6 0.084 0.970 1.031 3 32.1 The CMM data acquired from Solent
Tube 7 0.064 0.847 1.181 3 47.52
Tube 8 0.064 0.847 1.181 3 47.52 Mould Tools produced a three
Tube 9 0.064 0.847 1.181 3 47.52
Tube 10 0.044 0.702 1.425 4 60.66
dimensional representation of the
Tube 11 0.044 0.702 1.425 4 60.66 63mm tube. During the measuring
Tube 12 0.044 0.702 1.425 4 60.66
Tube 13 0.031 0.589 1.700 5 89.9 process the CMM machine took points
Tube 14 0.031 0.589 1.700 5 89.9
Tube 15 0.031 0.589 1.700 5 89.9 around the circumference at 4 points.
Tube 16 0.011 0.351 2.849 8 681.3
Tube 17 0.011 0.351 2.849 8 681.3
Tube 18 0.011 0.351 2.849 8 681.3 Using Pro Engineer, these points were
Table 8: ANSYS Shell 93 Results for Aluminium 6082-T-6 seamless tubes. generated using the co-ordinate
Table 8: ANSYS Shell 93 Results for Aluminium 6082-T-6 seamless
tubes.
3.13 Using Pro Engineers Mechanica
One part of the project was to carry out a feasibility study of a plastic buckling
an eccentric tube under uniform external pressure. The
purpose of the present study, however, was to provide an
alternative but simpler method of predicting the plastic
buckling pressure of slightly geometrically imperfect
tubes. In any case, the above Pro Engineer/ANSYS
analysis would not be of much use for the design of a
large vessel, such as
a submarine
pressure hull, as
metrological
measurements would
Figure 19: Screen Shot From Mechanica showing the maximum Von
Mises stress and displacement. be required prior to
its manufacture,
references that were provided. Then using the sweep which is impossible.
command the rings were joined together (see Figure 20). This emphasises the
need for the design
Figure 21: Pressure loading
A uniform pressure was applied to the exterior of the chart approach
cylinder using Pro Engineers Mechanica. The pressure adopted in the present paper, where the maximum
value was that of Pcr3 for model 4 (see Figure 21). permissible out-of-roundness for a full-scale vessel, such
as a submarine pressure hull, can be given to the
constructors of the vessel, prior to its manufacture.
4. DESIGN CHART
Once all the theoretical and experimental results had
been calculated, it was possible to generate a design
chart. This was done by plotting 1/ against PKD.
Figure 23 shows the design chart for Aluminum 6082-T6
seamless tubes form data obtained from experiments
carried out in 2006 and 2007, together with those of
References [2 and 7], which used other metals. Initial
Figure 20: Screen shot of sweep.
Results from the Mechanica analysis clearly showed the

significant areas of stress concentration and maximum
displacement (see Figure 22). From studying the deformed
test pieces, this was exactly where each tube failed due
to inelastic shell instability. Unfortunately, Mechanica can
only perform static pressure loading and not plastic
buckling. Nevertheless, the present study showed that
Pro Engineer could be imported into ANSYS and a
separate study (not reported here) showed that ANSYS Figure 22: Screen Shot From Mechanica showing maximum von
could carry out a successful plastic buckling analysis of Mises stress and displacement.
the wall thickness of their vessels.
imperfections of the aluminium alloy tubes

1/ AGAINST PKD
1/Lam bda against PKD
of the present paper were between 0.104t 1.600
to 0.13t, where t was the wall thickness 1.400
of the vessels, and the corresponding

1.200
values for those of Windenburg and Trilling

1.000
were between 0.11t & 0.16t, where t was Experimental 2006

Experimental Data 2006
Experimental 2007
1/Lambda
1/ LAMDA
Experimental Data 2007
the wall thickness of their vessels. 0.800
ReynoldsReynolds Welded
Windenburg Soldered
Windenburg
0.600 SAFE
SIDE
This design chart can now be used to 0.400
calculate the predicted (experimental) 0.200
buckling pressure Ppred for a pressure

0.000
vessel made out of a similar material. 0.000 0.500 1.000 1.500
PKD
PKD
2.000 2.500 3.000
to be calculated. Now, as we have a design chart, it is possible to obtain the

During the process of obtaining the design
PKD from the Design Chart (Figure 23), whichFigure then23:can be used
Updated DesigntoChart.
calculate
charts, the factors PKD and had to be
the predicted buckling pressure, namely Ppred [3]; where
calculated. Now, as we have a design
chart, it is possible to obtain the PKD from the Design thoseFigure
of 200623: [9], resultsDesign
Updated in much higher predicted
Chart.
Chart (Figure 23), which then can be used toPcalculate buckling pressures than the experimental results,
Ppred PpredcrThis design chartespecially
can now be forused to calculate the predicted (experimental)
the predicted buckling pressure, namely PKD [3]; where the shorter vessels. The results for Pro
buckling pressure Ppred for a pressure vessel made out of a similar material.
Engineers Mechanica are not shown, as it was not
During the process of obtaining the design charts, the factors PKD and had
Pcr intended to use Mechanica for buckling analysis.
4.1 Design
Ppred Charts
= & Comparisons
PKD
The graph of Figure 24 compares the theories of von Mises, Windenburg and 27
Results from experimental data acquired in 2006 - 2007
Trilling, ANSYS Shell 93 and Experimental results obtained from the present
4.1 DESIGN CHARTS & COMPARISONS and ANSYS Shell 93 have been plotted in Figure 25 (1/
study. It is evident from this graph that the prediction models, like those of
The graph of Figure 24 compares the theories of von against PKD), where in Figure 25:
2006 [9], results in much higher predicted buckling pressures than the
Mises, Windenburg and Trilling, ANSYS Shell 93 and
experimental results, especially for the shorter vessels. The results for Pro
Experimental results obtained from the present study. It is PKD = Pcr(ANSYS)/Pexp.
Engineers Mechanica are not shown, as it was not intended to use
evident from this graph that the prediction models, like
Mechanica for buckling analysis.
5. EVALUATION
It was apparent from studying
the buckling pressures obtained
from the theoretical and
experimental results that the
length and initial out-of-
circularity of the tubes had a
marked influence on the buckling
resistance, particularly for the
shorter tubes. The experimental
study showed that stresses due
to manufacturing should be
considered in establishing the
ultimate buckling resistance, in
addition to the pipe thickness,
diameter, length and ovality.
Figure 24: Graph of Predicted Buckling Pressures against tube lengths.
Figure 24: Graph of Predicted Buckling Pressures against tube lengths.
28
6. CONCLUSIONS The plastic buckling method of using ANSYS via the
The experimental and theoretical investigations were Pro Engineer/Mechanica route, as described here is
performed successfully on all test samples, with the not practical for the design of large vessels, such as
exception of the 50 and 30mm lengths because their submarine pressure hulls, as their metrological data
experimental buckling pressures would have exceeded during design is required prior to manufacture,
the maximum permitted design pressure of the tank. which is impossible.
The detrimental effects of initial built-in stresses
All specimens tested failed by shell instability. due to manufacture have not been taken into
All specimens suffered from manufacturing account.
imperfections. They were not concentric and precise
diametric measurements clearly showed variations REFERENCES
in tube wall thickness.
Initial imperfections of the aluminium alloy tubes of [1] R. von Mises, Der Kritische Aussendruck fr Allseits
the present paper were between 0.104t to 0.13t, Belastete Zylindrische Rohre, Fest Zum 70.
where t was the wall thickness of the vessels, and
the corresponding values for those of Windenburg Geburtstag von Prof. Dr. A. Stodola, Zrich, pp.
and Trilling were between 0.11t and 0.16t, where t 418-30. Translated and annotated by D.F.
was the wall thickness of their vessels.
Results from experimental data acquired
Windenburg, 1936,inReport
2006 No.- 366,
2007 and ANSYS Shel
DTMB,
Failures occurred in the areas of thinner wall
havetobeen
thickness, due, it is thought plotted in Figure 25 (1/ against PKD), where in Figure 25:
higher stress Washington D.C, USA, 1929.
concentrations at these points; this was predicted
by Mechanica. [2] D.F. Windenburg and C. Trilling, Collapse by
PKD = Pcr(ANSYS)/Pexp.
The analyses carried out with the three methods Instability of Thin Cylindrical Shells Under External
resulted in small differences between the theoretical
buckling pressures. Pressure, Trans., ASME, 11, pp 819-825, 1934.
Theoretical buckling pressures were far higher than [3] C.T.F. Ross, Pressure Vessels: External Pressure
the actual buckling pressures recorded during
pressure testing, especially for shorter vessels; this Technology. Horwood Publishing Ltd., Chichester,
was due to initial values of out-of-circularity. UK, 2001. (http://www.mech.port.ac.uk/
The paper shows that the
design charts appear to be sdalby/mbm/CTFRProg2.htm)
suitable for designing such
vessels; although there may be Design Chart (ANSYS)
some scale effect.
1.6
The design charts should 1.4
only be applied to 1.2
1/Lambda
circular cylinders under 1

uniform external 0.8
0.6
pressure and whose initial
0.4
out-of-circularity 0.2
does not exceed 0.16t, 0
where t is the wall 0 0.5 1 1.5 2 2.5 3 3.5
thickness of such vessels. PKD
Figure 25: ANSYS Shell 93 for 2006 & 2007 results.
Figure 25: ANSYS Shell 93 for 2006 & 2007 results.

[4] T. Tokugawa, Model Experiments on the Elastic
Stability of Closed and Cross-Stiffened Circular
Cylinders under Uniform External Pressure, Proc.
World Engineering Congress, Tokyo, Vol.29, Paper
No.651, pp.249-79, 1929.
[5] S. Kendrick, The Buckling under External Pressure of
Circular Cylindrical Shells with Evenly Spaced Equal
Strength Circular Ring Frames Part 1, NCRE
Report No. R.211, 1953.
[6] C. T. F. Ross, Mechanics of Solids, Horwood
Publishing Ltd., Chichester, UK, 1999.

.
[7] T. E. Reynolds, Inelastic Lobar Buckling of Cylindrical
Shells under External Hydrostatic Pressure, DTMB
Report No. 1392, Aug., 1960.
[8] T.G.Bosman, N.G. Pegg & P.J. Kenning, Experimental
and Numerical Determination of Non-Linear Overall
Collapse of Imperfect Pressure Full Compartments,
Int. Symp on Naval Submarines 4, RINA, 11-13
May, 1993,London.
[9] A. Nagoppan, Buckling of Aluminium Alloy tubes
under external water pressure. Mechanical
Engineering Student Project Report, University of
Portsmouth, Portsmouth, UK, June 2006.

In Elastic Buckling of Geometrically Imperfect Tubes

Uploaded by

Copyright:

Available Formats

In Elastic Buckling of Geometrically Imperfect Tubes

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

In Elastic Buckling of Geometrically Imperfect Tubes

Uploaded by

Copyright:

Available Formats

NOT FOR REPRODUCTION

Designing Under Pressure

About the Authors

Daniel Short is a Mechanical Engineering student at the University of Portsmouth

Graham X. Brown is the Chief Mechanical Engineer at Sonardyne Ltd., in Yately,

NOMENCLATURE Pexp experimental buckling pressure

To improve the In this study, we will be concerned with elastic and

Figure 3: General instability.

= Windenburg thinness ratio.

1.3 WINDENBURG AND TRILLINGS THINNESS RATIO

Where, larger than 0.45*(t/d)0.5, thus if we neglect 0.45*(t/d)0.5

Or = [(l/d)2 / (t/d)3]0.25 * ( yp / E)0.5

N.B. Windenburg and Trilling squared in the above

the value of would be approximately one.

2.1 THE TESTING PROCEDURE

The closure discs were push-fitted into the ends of each

When using MisesNP the following parameters had to be

Outer Diameter = 50mm

Model dimensions; unsupported length,

Poissons ratio and density;

to be simply-supported, similar to Portsmouths

3.2. PLASTIC KNOCKDOWN FACTOR (PKD) buckling in this case.

To generate a design chart, the calculated PKD

3.3. USING ANSYS

2. Meshing 5. Eigen buckling

3. The boundary conditions Figure 10: Buckling of a 189 mm tube.

A 230mm long specimen was sent to the metrology

Copyright Journal of Ocean Technology 2008 193, No. 1, 2008 83

throughout (see Table 5). The reason for this was

==22vessels. nn==33 nn==44

manner shown in Figure 1; usually at a fraction of that

From experimental results carried out in the 2006

Table 6: Experimental Results.

Models 2-3: Length 160mm, buckled due to inelastic

3.12. USING ANSYS SHELL 93

The following parameters had to be fed in:

radius and wall thickness;

Structural conditions (i.e. Eigen buckling).

Table 7: Results for calculating PKD

Figure 20: Screen shot of sweep.

Results from the Mechanica analysis clearly showed the

imperfections of the aluminium alloy tubes

of the present paper were between 0.104t 1.600

to 0.13t, where t was the wall thickness 1.400

of the vessels, and the corresponding

values for those of Windenburg and Trilling

were between 0.11t & 0.16t, where t was Experimental 2006

the wall thickness of their vessels. 0.800

calculate the predicted (experimental) 0.200

buckling pressure Ppred for a pressure

vessel made out of a similar material. 0.000 0.500 1.000 1.500

to be calculated. Now, as we have a design chart, it is possible to obtain the

circular cylinders under 1

Figure 25: ANSYS Shell 93 for 2006 & 2007 results.

Stability of Closed and Cross-Stiffened Circular

Cylinders under Uniform External Pressure, Proc.

World Engineering Congress, Tokyo, Vol.29, Paper

No.651, pp.249-79, 1929.

[5] S. Kendrick, The Buckling under External Pressure of

Circular Cylindrical Shells with Evenly Spaced Equal

Where, larger than 0.45(t/d)0.5, thus if we neglect 0.45(t/d)0.5