You are on page 1of 13

ARTICLE IN PRESS

International Journal of Mechanical Sciences 48 (2006) 855867


www.elsevier.com/locate/ijmecsci

Plastic buckling of tubes under axial compression and internal pressure


J.A. Paquette, S. Kyriakides
Research Center for Mechanics of Solids, Structures & Materials, WRW 110, C0600, The University of Texas at Austin, Austin, TX 78712, USA
Received 10 September 2005; received in revised form 22 February 2006; accepted 9 March 2006
Available online 11 May 2006

Abstract
The plastic buckling and collapse of long cylinders under combined internal pressure and axial compression was investigated through a
combination of experiments and analysis. Stainless-steel cylinders with diameter-to-thickness values of 28.3 and 39.8 were compressed to
failure at xed values of internal pressure up to values 75% of the yield pressure. The rst effect of internal pressure is a lowering of the
axial stressstrain response. In addition, at some plastic strain level, the cylinder develops uniform axisymmetric wrinkling. Under
continued compression, the wrinkles grow stably, gradually reducing the axial rigidity of the structure and eventually lead to a limit load
instability. All pressurized cylinders remained axisymmetric until the end of the test past the limit load.
The critical stress and wavelength were established using classical plastic bifurcation theory based on the deformation theory of
plasticity. The evolution of wrinkling, and the resultant limit state, were established by modeling a periodic domain that is one half of the
critical wavelength long. The domain was assigned an initial imperfection corresponding to the axisymmetric buckling mode calculated
through the bifurcation check. The inelastic material behavior was modeled through the ow theory of plasticity with isotropic
hardening. The variations of the axial response and of the limit strain with pressure observed in the experiments were reproduced well by
the model. Inclusion of Hill-type anisotropic yielding in all constitutive models was required for good agreement between predictions and
experiments.
r 2006 Elsevier Ltd. All rights reserved.
Keywords: Circular tubes; Circular cylindrical shells; Combined axial compression-internal pressure; Plastic buckling; Collapse

1. Introduction
A pipeline in operation is under internal pressure. The
scenarios for developing axial compression, leading to
buckling and collapse outlined in Ref. [1], apply also to a
pressurized pipeline. The general features of plastic
buckling under axial compression and internal pressure
are similar to those of pure axial loading, outlined in Fig. 1.
The cylinder rst wrinkles at an increasing load. The
wrinkle amplitude grows, leading eventually to a limit load
instability. A pipeline will fail by localized collapse at this
strain, and as a result this constitutes a limit state. The
biaxial state of stress lowers the axial stress levels of the
various critical events described in Ref. [1] but, as will be
demonstrated, has a smaller effect on the corresponding
strains. In addition, the pressure has a stabilizing effect on
the axisymmetric mode, making a switch to non-axisymCorresponding author. Tel.: +1 512 4714 167; fax: +1 512 4715 500.

E-mail address: skk@mail.utexas.edu (S. Kyriakides).


0020-7403/$ - see front matter r 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijmecsci.2006.03.003

metric modes more difcult. These features of the problem


will rst be illustrated experimentally. The formulation for
predicting the onset of plastic wrinkling is then developed,
followed by a study of how wrinkles grow, localize and lead
to collapse.

2. Experimental
Lee [2] reported results from buckling experiments on
cylindrical shells under combined internal pressure and
axial compression. The experiments were conducted on
tubes made of a soft aluminum alloy (Al-3003-0). The ends
of the tubes were clamped and, as a result, the onset of
wrinkling was masked and not reported. The stress at the
onset of collapse was measured for different loading paths.
Combined loading experiments on line pipe were
reported by Murray [3]. He used pipes of various lengthto-diameter ratios (L/D) with diameter-to-thickness ratios
(D/t) of 64.3 and 50.6. The majority of the tests involved

ARTICLE IN PRESS
J.A. Paquette, S. Kyriakides / International Journal of Mechanical Sciences 48 (2006) 855867

856

Pressure
Transducer

Upper
Grip

Signal
Generator
Span
Set Point

Linear
Encoder

Test Specimen

Strain Gages

Set Point
Press.
Control

Extensometer

LVDT
Solid Insert
Locking Assembly
Rotating Stage

Volume
Control

Seals

Spacer

Span

Pressurizing
Fluid

Servovalve

Lower Grip

Pressure
Intensifier

LVDT

Volume
Feedback

Fig. 1. Schematic of combined compression and internal pressure test setup.

combined compression, bending and internal pressure. The


tests were particularly aimed at understanding the behavior
of the pipe well past the onset of local collapse (see also
Ref. [4]).
Lees results did not allow for the establishment of the
onset of wrinkling and Murrays experiments dealt with
more complex loadings. A new set of experiments was
conducted involving SAF 2507 super-duplex tubes compressed to failure under xed levels of internal pressure.
The tests were designed to approximate this loading as
encountered in a long pipeline or shell free from end effects
or constraints.
2.1. Experimental setup
A schematic test of the experimental setup used is shown
in Fig. 1. The setup is similar to the one used in Ref. [1]
for their pure axial compression experiments (see Fig. 4,
Ref. [1]), but has been extended to include a pressurizing
system shown schematically in the gure. The cavity was
sealed with O-ring seals located on the solid end-plugs. In
addition, a solid metal insert was placed in the specimen

cavity to reduce the volume of the pressurizing uid. The


pressure was monitored by an electrical pressure transducer
and by mechanical dial gages. The cavity was lled with
hydraulic uid and pressurized by a stand-alone, closedloop control, pressure intensier. The intensier operates
on standard 3000 psi (207 bar) hydraulic power and has a
capacity of 10,000 psi (690 bar). In the present experiments
it was operated under pressure control (see Fig. 1).
The test specimens were machined out of SAF 2507
super-duplex (a type of stainless steel) stock following the
procedure outlined in Section 2 of Ref. [1]. Wrinkles
formed by compression in the presence of internal pressure
have a longer wavelength than those formed under pure
compression. For this reason, the test section of the
combined loading test specimens was 5 in (127 mmL/
RE4.65) long while their overall length was 12 in
(305 mm). The rest of the geometric characteristics were
similar to those described in Ref. [1].
In a typical experiment, the pressure was increased to the
required level and was held constant. The specimen was
then axially compressed at a displacement rate that
produced a strain rate of approximately 2  105 s1 in

ARTICLE IN PRESS

3.71
3.72
3.03
3.50
3.69
4.01
109.8 (757)
106.1 (732)
101.2 (698)
95.7 (660)
87.8 (606)
80.7 (557)
0.2800.363
0.3170.386
0.3820.446
0.4410.529
0.4550.584
0.5470.653
1.211.41
1.451.74
1.181.51
1.201.40
1.371.57
1.071.27
(664672)
(652668)
(621642)
(567581)
(519530)
(443456)
96.397.5
94.596.8
90.193.1
82.284.3
75.276.9
64.266.1
(69.7)
(138.5)
(207.6)
(275.9)
(327.2)
0
1011
2008
3010
4000
4744
(2.032.05)
(2.012.06)
(1.972.00)
(2.002.04)
(2.012.05)
(2.022.05)
0.07980.0807
0.07910.0810
0.07750.0787
0.07860.0803
0.07920.0806
0.07940.0808
28.13
28.16
28.82
28.28
28.21
28.15

eC (%)
sC ksi (MPa)
P psi (bar)
tmintmax in (mm)
D/t
t in (mm)

SAF4
SAF4
SAF4
SAF4
SAF4
SAF4
IPC5
IPC2
IPC1
IPC3
IPC4
IPC6

D in (mm)
Tube no.
Exp no.

Table 1
Specimen parameters and critical variables measured in axial buckling under internal pressure experiments (D/t 28.3)

where so is the yield stress of the tube in the axial direction,


Sy is the ratio of the yield stress in the circumferential
direction to that in the axial direction (soy/so), R is the
mid-surface radius of the tube and t is its wall thickness.
The properties of the tube stock used are listed in Table 2,
Ref. [1]. These include measures of the anisotropy
constants (Sr,Sy) established as dened in Section 4 of
Ref. [1].
A typical axial stress-shortening (sxdx/L) response for
a tube with D/t 28.3, tested at a xed pressure of
3010 psi (208 bar; P 0.439Po), is shown in Fig. 2a (dx and
L are, respectively the axial shortening and the length of
the test section). Fig. 2b shows a set of axial scans taken in
the course of the test at the positions marked on the
response with solid bullets. The response is similar to the
one presented in Fig. 5a, Ref. [1] for pure axial compression. The rst effect of pressure is that the material yields at
a lower axial stress, resulting in a lowering of the inelastic
part of the response. In the early scans, the test section
deformed nearly uniformly. Later scans revealed small
bulges forming at the two ends of the test section.
Although the linear tapers next to the test section reduced
the stress concentrations at the discontinuities, they did not
eliminate them completely. Thus, at higher compression
strains, bulges formed at these locations. Wrinkles gradually began to appear in the rest of the test section. It was
determined that the onset of wrinkling (eC) occurred
between strains of 1.20% and 1.40% (points marked with
k on response). The stress corresponding to these strains
is designated as sC. A second major effect of internal
pressure is an increase in the wrinkle wavelength. For the
case shown in Fig. 2, the wavelength (2lC) was 1.5 times
the value at zero pressure (see Table 1).
As compression progressed, the wrinkle amplitude grew,
causing a reduction in the stiffness of the specimen. At
higher strains, the growth of the two end bulges
accelerated, resulting in a limit load at an average strain
of e L 3.50%. The test was terminated soon after the limit

lC =R

Sets of tubes of two diameter-to-thickness ratios (D/t),


28.3 and 39.8, were compressed at different values of
internal pressure until a limit load instability developed.
The parameters of each set of tests and the main results are
summarized in Tables 1 and 2. Pressure is normalized by
the yield pressure dened as
t
Po S y so
,
(1)
R

(2.04)
(2.04)
(1.98)
(2.02)
(2.03)
(2.04)

2.2. Experimental results

0.0802
0.0801
0.0781
0.0797
0.0800
0.0801

sL ksi (MPa)

the test section. The loading was periodically paused for a


few seconds to perform axial scans. The scans were
subsequently used to identify the onset of wrinkling and
to establish the evolution of the wrinkles. Typically, one or
two of the wrinkles tended to localize, leading to a limit
load.

2.2555(57.29)
2.2549 (57.27)
2.2521 (57.20)
2.2552 (57.28)
2.2557 (57.29)
2.2556 (57.29)

eL (%)

J.A. Paquette, S. Kyriakides / International Journal of Mechanical Sciences 48 (2006) 855867

857

ARTICLE IN PRESS

IPC3

105.1 (725)
97.5 (672)
93.9 (648)
86.8 (599)
72.0 (497)
68.2 (470)

100
x
(ksi) 80

600

60
Axial Scans
40
20

0.2410.306
0.3200.363
0.3900.449
0.4290.474
0.5870.660
0.5660.672

eL (%)
sL ksi (MPa)
lC =R

1.89
1.31
1.52
1.59
1.49
1.89

J.A. Paquette, S. Kyriakides / International Journal of Mechanical Sciences 48 (2006) 855867

858

0.861.01
0.891.05
0.740.88
0.780.96
0.730.87
0.750.90
(672684)
(648658)
(599612)
(555568)
(456468)
(416425)
97.499.2
94.095.4
86.888.8
80.582.4
66.167.8
60.361.6
0
932 (64.3)
1852 (127.7)
2772 (191.2)
3543 (244.3)
3996 (275.6)

eC (%)
sC ksi (MPa)
P psi (bar)

D = 28.28 P = 0.439
Po
t

-3
2

4
x/R

(1.391.44)
(1.361.41)
(1.391.43)
(1.401.43)
(1.371.45)
(1.411.47)

Fig. 2. Typical results for a compression under internal pressure test: (a)
recorded axial stress-shortening response and (b) axial scans showing
evolution of wrinkles in test section.

120

800

SAF 4
D
= 28.3
t

x
(ksi)

x
(MPa)

600
80

39.84
40.62
39.82
39.65
39.85
39.01

0.05460.0567
0.05370.0553
0.05480.0564
0.05510.0564
0.05400.0572
0.05550.0579

(b)

(1.41)
(1.38)
(1.41)
(1.42)
(1.41)
(1.44)
0.0555
0.0544
0.0555
0.0558
0.0555
0.0567
(56.16)
(56.11)
(56.18)
(56.19)
(56.17)
(56.23)
SAF5
SAF5
SAF5
SAF5
SAF5
SAF5
IPC11
IPC10
IPC7
IPC8
IPC9
IPC12

SAF 4
2

-2

2.2112
2.2092
2.2117
2.2121
2.2114
2.2137

tmintmax in (mm)
D/t
t in (mm)
D in (mm)

4
x / L (%)

-1

40

Tube no.

x / L

IPC3

Exp no.

200

0
0

(a)

Table 2
Specimen parameters and critical variables measured in axial buckling under internal pressure experiments (D/t 39.8)

400
SAF 4
D
= 28.28
t
P
= 0.439
Po

2
w
R
(%) 1

x
(MPa)

P
Po

0
0.147
0.299
0.439
0.581
0.688

400

200

0
0

4
x / L (%)

Fig. 3. Axial stress-shortening responses recorded at different values of


internal pressure for tubes with D/t 28.3.

load. Interestingly, the wrinkling pattern remained axisymmetric throughout the test.
A total of six tests were performed for this D/t value.
The recorded stress-displacement responses are shown in
Fig. 3. The geometric variables and the critical parameters
of these tests appear in Table 1. Pressure lowers the
response and, as a consequence, the critical and limit
stresses are lowered also. This is demonstrated in Fig. 4a,
where the critical stress is plotted against pressure. By

ARTICLE IN PRESS
J.A. Paquette, S. Kyriakides / International Journal of Mechanical Sciences 48 (2006) 855867

859

120
800

SAF 4

D
= 28.3
t

(ksi)
100

SAF 4
D
= 28.3
t

(MPa) (%) 2.5


700

Experiments

Anisotropic

2
Anisotropic

Isotropic

600
1.5

80
500

Isotropic
60

Experiments

400
0.5
300

40
0

0.2

0.4

0.6

0
0

0.8
P /Po

(a)

0.2

0.4

0.6

0.8
P /Po

(b)

1
C

SAF 4

D
= 28.3
t

0.8

Isotropic

Anisotropic
0.6

0.4
Experiments
0.2

0
0

0.2

(c)

0.4

0.6

0.8
P /Po

Fig. 4. Critical stress (a), strain (b) and (c) axial half-wavelength (onset of wrinkling) as a function of internal pressure for D/t 28.3. Experiments and
predictions.

contrast, the critical strain is seen to remain relatively


unchanged in Fig. 4b. The corresponding wrinkle halfwavelengths are plotted against pressure in Fig. 4c. The
wavelength is seen to increase nearly linearly with pressure,
almost doubling in value when going from P 0 to
0.688P0. Fig. 5 shows photographs of two specimens at the
completion of the tests. One was tested at a relatively low
pressure (P 0.147Po), and the second at a relatively high
pressure (P 0.688Po). The wrinkles as well as the bulges
at the ends of the test sections can be seen in the
photographs. The difference in wrinkle wavelength between the two is striking.
Fig. 6a shows a plot of the limit stress against pressure.
sL decreases with pressure, and is somewhat higher than
the wrinkling stress. Fig. 6b shows a plot of the average
strain corresponding to limit stress (eL ) against pressure. eL
increases with pressure but only slightly. The corresponding eC values are included in the gure for comparison.
They are seen to be signicantly lower than eL . Finally, it is
noted that in the ve pressure experiments, the wrinkled

tubes remained axisymmetric up to collapse. By contrast,


the pure compression experiment in the set buckled into a
non-axisymmetric mode (m 2).
The results from the tests for D/t 39.8 are broadly
similar. Fig. 7 shows stress-shortening responses recorded
at six different pressure levels. The main parameters are
listed in Table 2. The critical stresses, strains and halfwavelengths are plotted against pressure in Fig. 8. The
critical stress drops with pressure, whereas the critical
strain remains relatively unaffected. The wrinkle wavelength exhibits a similar increase with pressure as in the
previous set of results. The limit stress and strain are
plotted against pressure in Fig. 9. The stress drops as
pressure increases, whereas the strain remains relatively
unaffected. eL is signicantly higher than eC, but the
difference is smaller than for the lower D/t case. It is also
worth noting that both strains have much lower values
than the thicker tubes in Fig. 6. Some scatter observed in eL
is probably caused by initial imperfections. The pressurized tubes buckled axisymmetrically for this set also,

ARTICLE IN PRESS
J.A. Paquette, S. Kyriakides / International Journal of Mechanical Sciences 48 (2006) 855867

860

Fig. 5. Wrinkled test specimens with D/t 28.3 tested at (a) P 0.147Po and (b) P 0.688Po.

whereas the one pure compression tube buckled in a nonaxisymmetric manner some time before the limit stress was
reached.
3. Analysis
3.1. Onset of axisymmetric wrinkling
The bifurcation into axially uniform wrinkling follows
along the same steps as those for pure axial compression

dex
dey

Es

given by

1=2
se s2x  sx sy s2y
where
PR
and sx is the axial stress.
sy
3a
t
Here, R is the mid-surface radius of the tube and t is its
wall thickness. [Cab] are the instantaneous deformation
theory moduli which are derived from the inverse of the
following constitutive matrix:

1 q2sx  bsy 2 ;

^ns q2sx  bsy 2asy  bsx

^ns q2sx  bsy 2asy  bsx ;

a EEs 1  a q2asy  bsx 2

[1]. Because the combined loading is stress controlled, sC


and lC can still be evaluated from


1=2  
C 11 C 22  C 212
t
sC
and
R
3
"
#1=4
C 211


Rt1=2
lC p
12 C 11 C 22  C 212

[2]. The effect of the biaxial state of stress enters the


solution through the equivalent stress, which in this case is

%(

dsx
dsy

)
,

(3b)

where
!
1
1
1
a 2; b 1 2 2 ,
Sy
Sy Sr




1 E s se
b Es
b
 1 and n^ s
n
.
q 2
4se E t se
2 E
2
Sy soy =so , Sr sor =so and fso ; sor ; soy g are the yield
stresses in the principal polar directions (x,r,y). Et and Es
are, respectively the tangent and secant moduli.
By contrast, the prebuckling deformation of the cylinder
depends on the path followed in the (sxsy) plane and, as a

ARTICLE IN PRESS
J.A. Paquette, S. Kyriakides / International Journal of Mechanical Sciences 48 (2006) 855867

result, the strains are calculated using the corresponding


ow theory given by
(

dex

dey

A sample of axial stressstrain results corresponding to


various values of internal pressures for a D/t 28.3 tube

1 Q2sx  bsy 2 ;

n Q2sx  bsy 2asy  bsx

n Q2sx  bsy 2asy  bsx ;

1 Q2asy  bsx 2

where


1
E
1 .
Q 2
4se E t se
Thus, eC is the axial strain corresponding to sC and P in
such an incremental calculation that follows the stress
history.

120
SAF 4

Experiments

L
(ksi)

800
L
700

100

(MPa)

600

Analysis
80

500
60

o = 0.5%

S = 1.15

400

D
= 28.3
t

Sr = 0.85

300

40
0

0.2

0.4

0.6

0.8
P / Po

(a)

861

%(

dsx

)
(4)

dsy

are shown in Fig. 10. The anisotropic ow theory (Eqs. (4))


was used along with the properties of SAF4 in Table 2,
Ref. [1]. Pressure reduces the yield stress in the axial
direction. Therefore, as the pressure increases, the response
traces progressively lower axial stress. The point at which
the structure bifurcates into axial wrinkling is marked on
each response with the symbol
. The bifurcation
stresses were evaluated from Eq. (2) using the anisotropic
incremental deformation theory equations (3). Once sC is
established, the corresponding strain is evaluated from the
ow theory by rst internally pressurizing and then
incrementing the axial stress to this value.
Similar calculations were also conducted for the corresponding isotropic material (Sr Sy 1). The critical states
{sC, eC, lC} calculated for D/t 28.3 are compared to the
experimental results in Fig. 4. The anisotropic material
tracks the experimental stresses, strains and wrinkle
wavelengths quite well. The critical stress decreases with
pressure, the critical strain remains relatively unaffected and
lC increases with pressure. The predictions from the isotopic
material are seen to deviate from the experimental values for
all three variables. For this biaxial loading the anisotropy
affects the whole response, not just the onset of wrinkling.
Critical state predictions for the tubes with D/t 39.8 are
included in Fig. 8. The anisotropic material results are in
good agreement with the measured variables, while the
isotropic material results are not.
3.2. Evolution of wrinkling

5
SAF 4

(%)

Experiments
4

The evolution of wrinkling and the onset of axisymmetric collapse can be established by axially compressing a
120

3
x

Analysis

800

SAF 5
D = 39.8
t

(ksi)

x
600

80

2
C
1

P
Po

40

S = 1.15
Sr = 0.85

o = 0.5%

D
= 28 .3
t

0
0
(b)

(MPa)

0.2

0.4

0.6

0.8
P / Po

Fig. 6. Limit stress (a) and strain (b) as a function of internal pressure for
D/t 28.3. Experiments and predictions.

0
0.184
0.359
0.535
0.687
0.759

400

200

0
0

0.5

1.5

2
x / L (%)

Fig. 7. Axial stress-shortening responses recorded at different values of


internal pressure for tubes with D/t 39.8.

ARTICLE IN PRESS
J.A. Paquette, S. Kyriakides / International Journal of Mechanical Sciences 48 (2006) 855867

862

120
800

SAF 5

D = 39.8
t

(ksi)

700

100

SAF 5

Isotropic

D = 39.8
t

(%)
1.5

(MPa)

Experiments

Anisotropic
600
Anisotropic

80

Isotropic

500

60

300

40
0

0.2

0.4

0.6

0
0

0.8
P / Po

(a)

Experiments

0.5

400

0.2

0.4

0.6

0.8
P / Po

(b)

C
R

SAF 5
D

0.8

Isotropic

= 39.8

t
0.6

Anisotropic

0.4
Experiments
0.2

0
0

0.2

0.4

0.6

0.8
P / Po

(c)

Fig. 8. Critical stress (a), strain (b) and (c) half-wavelength (onset of wrinkling) as a function of internal pressure for D/t 39.8. Experiments and
predictions.

uniformly wrinkled cylinder in the presence of internal


pressure. This can be achieved by a simple extension of the
formulation presented in Section 3 of Ref. [5]. The cylinder
is assigned a small initial axisymmetric imperfection, given
by
 
px
w
(5)
too cos
lC
where lC is the critical half-wavelength of axisymmetric
wrinkles yielded by Eq. (2). An external pressure work term
is added to the PVW so that it becomes
Z lC

2pR
N xx deoxx N yy deoyy M xx dkxx dx
0

w a0

dw dx

where the domain of interest is limited to lC because of the


periodicity of the geometry. (Only linear terms in w are

ai cos

 
 
Nu
X
ipx
ipx
bi sin
and u b0 x
.
lC
lC
i1
(7)

Substituting (7) into the PVW leads to

Nw
X
i1

lC

2pRP

included in the work term in Eq. (6) for consistency with


the rest of the formulation. This has an insignicant effect
on the results as the radial deformations remain small in
this problem). The problem is discretized by adopting the
following admissible expansions for the radial and axial
displacements as follows:

lC h

N xx eoxx;i N yy eoyy;i M xx kxx;i


Z

dxdqi

lC

w;i dqi dx; i 1; 2; :::::; N w N u 1,


0

ARTICLE IN PRESS
J.A. Paquette, S. Kyriakides / International Journal of Mechanical Sciences 48 (2006) 855867
120
800
C

SAF 5

;i 

700

100

80
Analysis

500

60

400

o = 1.2%
D
= 39.80
t

S = 1.11
Sr = 0.87

300

40
0.2

0.4

0.6

0.8
P / Po

(a)
3

D
= 39.8
t

(%)

SAF 5

o = 1.2%

Analysis


T
@:
and q a0 ; a1 ; . . . ; aN w ; b1 ; b2 ; :::::bN u
@qi

and b0 ( dx/lC) is the average axial strain that is


prescribed incrementally. The anisotropic ow theory
constitutive equations (4) are adopted. The incremental
solution procedure proceeds as described in Ref. [4].
A sample of results corresponding to one of the tests
with D/t of 28.15 is shown in Fig. 11a. The pressure is
P 0.69Po, while the material model includes the anisotropy parameters for SAF4 in Table 2, Ref. [1]. Results for
three imperfection amplitudes oo {0.05, 0.1, 0.2}% are
shown. The homogeneous deformation response, drawn
with a dashed line, is also included. As mentioned above,
internal pressure lowers the response, and the imperfection
causes a limit load instability. Increasing the imperfection
amplitude reduces the limit strain. Fig. 11b shows a direct
comparison of the experimental response and a calculated
response corresponding to oo 0.1%. They are seen to be
in good agreement. The predicted limit strain also agrees
very well with the experimental value. It is noteworthy that
the amplitude of oo required for this matching is rather
small.

600

where

(MPa)

Experiments

(ksi)

863

Experiments
1

80
x
(ksi)

S = 1.11

Material Response

SAF 4

60

40

0.2

0.4

0.6

0.8

= 0.69
Po
D
= 28.15
t

20

4
5
x / C (%)

(a)
80
x
(ksi)

SAF 4
D = 28.3
t

600

SAF 4

500
Experiment

60

40

P
Po

20

0.5

1.5
x (%)

400

-3

300
= 0.600

200

20

= 0.69

Po
D

200

100

= 28.15

0
0

Fig. 10. Axial stressstrain responses for a uniformly deforming cylinder


compressed at different pressure levels.

400 (MPa)

40

0
0.147
0.299
0.439
0.581
0.688

Analysis
o = 1.0x 10

(MPa)

60

100
0

0
0

x
(ksi) 80

300
200

Fig. 9. Limit stress (a) and strain (b) as a function of internal pressure for
D/t 39.8. Experiments and predictions.

100

= 0.600

400 (MPa)

P / Po

(b)

1.0 x10

500

-3

2.0 x10-3

Sr = 0.87
0

0.5 x10-3

(b)

5
x / C (%)

Fig. 11. (a) Axial stress-displacement responses of a cylinder with


axisymmetric imperfections of different amplitudes for P 0.69Po. (b)
Comparison of an axisymmetric wrinkling response with corresponding
experimental response.

ARTICLE IN PRESS
J.A. Paquette, S. Kyriakides / International Journal of Mechanical Sciences 48 (2006) 855867

864

In such idealized calculations, the amplitude of the


imperfection is a parameter that encompasses several of the
factors that affect the experimental limit strain. These
include the niteness of the tube tested, the effect of the
discontinuity at the ends (albeit reduced), small load
eccentricities, small thickness variations, etc. For this
reason, and for the purpose of comparing experiments
with predictions, the imperfection amplitude of each of the
two sets of experiments was selected based on best overall
performance. Thus, for the D/t 28.3 experiments, oo was
chosen to be 0.5%. The axial stressaxial shortening
responses calculated with this imperfection for the six test
pressures are shown in Fig. 12a. The imperfection
wavelength used in each case is the one yielded by the
bifurcation check algorithm. Pressure lowers the response
but simultaneously causes the limit stress to occur at
progressively higher strain. The limit stresses and strains
are compared to the experimental values in Fig. 6. The
predicted limit stresses are seen to be in good agreement
with the measured values. The trend of the corresponding
average strains is in agreement with the experiments for
higher values of pressure. For P 0 and 0.147Po the
predictions are lower than the measured values. The reader
is reminded that for P 0, the cylinder failed in a nonaxisymmetric buckling mode not captured by the present
analysis.

120

800

SAF 4

x
(ksi)

D
= 28.3
t

For the D/t 39.8 experiments, oo 1.2% was


adopted, along with the lC predicted for each case. The
calculated sxdx responses for the six test pressures are
shown in Fig. 12b. The trend is similar to that of the
thicker tubes but the limit point occurs at much lower
strain levels. The predicted values of sL and eL are
compared to the measured values in Fig. 9. The limit
stress follows the experimental trend quite well. eL is also in
good agreement with the experiments, except in the case of
P 0. In this test the cylinder again failed by nonaxisymmetric buckling not captured by the present
analysis.
In summary, the inelastic response and limit state of
relatively thick cylinders, compressed axially in the
presence of internal pressure, are reproduced well by an
axisymmetric model. For the present experiments, correct
representation of the plastic anisotropy in the tubes played
an important role in this successful performance of the
model.
4. Parametric study
The parametric study of the critical and limit states
conducted in Ref. [5] is extended to the problem of
combined compression and internal pressure. The material
is assumed to yield isotropically and is assigned the three
RambergOsgood stressstrain responses (Fig. 13) with the
properties listed in Table 3.

x
120

(MPa)

80

40

Po

400

40
200

x / C (%)

(a)

100

40

Po

20

0
0.184
0.359
0.535
0.687
0.759

200

0
0.5

1.5

Fig. 13. Three stressstrain responses used in the parametric study


performed.

400

o = 1.2%

4
(%)

(MPa)
600

80

SAF 5
D
= 39.8
t

60

(b)

600

30

400

0
2

15

200

o = 0.5%

80

0
0

(MPa)

(ksi)

0
0.147
0.299
0.439
0.581
0.688

x
(ksi)

800

600

2
x / C (%)

Fig. 12. Axisymmetric wrinkling axial stressdisplacement responses for


various pressures. D/t (a) 28.3 and (b) 39.8.

Table 3
Three stressstrain responses adopted in parametric study
E Msi (GPa)

sy ksi (MPa)

soa ksi (MPa)

28.2 (194.5)
28.2 (194.5)
28.2 (194.5)

80.0 (552)
85.0 (586)
88.0 (607)

76.95 (530.7)
82.70 (570.3)
86.70 (597.9)

9
15
30

so is the stress at a strain offset of 0.2%.

ARTICLE IN PRESS
J.A. Paquette, S. Kyriakides / International Journal of Mechanical Sciences 48 (2006) 855867

Tubes with D/t values ranging from 20 to 60 were


analyzed. The tubes were compressed to failure at pressure
levels of 0, 0.5Po and 0.7Po. The critical values {sC, eC, lC}
are plotted against D/t in Figs. 14ac. Nine sets of results
are presented in each plot corresponding to the three values
of n and the three pressure levels. Internal pressure lowers
the material response and results in a corresponding
reduction in sC in the same manner as seen in the
experiments (e.g., Fig. 4a). sC is highest for n 9 and
lowest for n 30. Because of the nature of the stressstrain

865

responses adopted, the critical strains follow powerlaw


relationships with D/t and appear as linear in the loglog
plot in Fig. 14b. In concert with the experimental
observations, internal pressure tends to increase eC. Even
more importantly, the results illustrate the strong inuence
of the hardening exponent on eC. The results for n 30 are
signicantly lower than the corresponding ones for n 9.
Fig. 14c shows the corresponding lC vs. D/t plots. Once
again, the results illustrate that pressure increases lC. At
the same time, increasing n results in an increase in lC.

110
n

15

3
C

(MPa)

(%)

30

(ksi)

700

0.7

0.5

Po

90
600

15

500

15

70

30
15

400

0.9

30

50

30

0.8
0.7
0.5
0

30
20

30

0.7

300
P
Po

0.6

40

50

0.5
20

60
D/ t

(a)

30

40

50
D/ t

(b)

1.2
0.7
0.5
0

C
R
1

P
Po

0.8

0.6
15

30

0.4

0.2
20
(c)

30

40

50

60
D/ t

Fig. 14. (a) Effect of hardening parameter n on the critical stress for three pressure levels, (b) effect of hardening parameter n on the critical strain for three
pressure levels, and (c) effect of hardening parameter n on the wrinkle wavelength for three pressure levels.

ARTICLE IN PRESS
J.A. Paquette, S. Kyriakides / International Journal of Mechanical Sciences 48 (2006) 855867

866

0.7

0.5

L
(%) 5

P
Po

3
9

15
n
30
2

1
0.9
o = 0.2%

0.8
0.7
20

30

40

50
D/ t

Fig. 15. Effect of hardening parameter n on the average limit strains for
three pressure levels.

The limit strain was also studied parametrically by


considering strictly axisymmetric imperfections. The halfwavelength adopted in each case corresponds to the critical
value lC. The amplitude of the imperfection used had a
constant value of oo 0.2%. The calculated values of eL
are plotted against D/t in Fig. 15 in loglog scales. As
expected, for each pressure level e L is signicantly higher
than the corresponding eC in Fig. 14b. In addition, pressure
has the effect of increasing eL for all three materials. Even
more importantly, as was the case for eC, reduction in the
hardening of the material (increase in n) reduces eL . The
limit strain is also inuenced by the material yield stress.
However, its effect is relatively secondary. Thus, for a xed
D/t and loading, material hardening is an important
variable for eL .
5. Summary and conclusions
The plastic buckling and collapse of long cylinders under
combined internal pressure and axial compression has been
investigated experimentally. The experiments involved
SAF2507 stainless-steel cylinders with D/t values of 28.3
and 39.8. The cylinders were compressed to failure at xed
values of internal pressure, in the range of 0pPo0.75Po.

Internal pressure interacts plastically with axial compression. The interaction results in a lowering of the axial
stressstrain response. J2-type plasticity with isotropic
hardening or its anisotropic counterpart can adequately
capture this interaction.
At some plastic strain level, the cylinder develops
uniform axisymmetric wrinkling. Under continued compression, the wrinkles grow stably, gradually reducing the
axial rigidity of the structure. This reduction in axial
rigidity eventually leads to a limit load instability. Beyond
the limit load, deformation localizes. Thus, the limit load
can be considered as the failure load of the structure. All
pressurized cylinders remained axisymmetric to the end of
the test past the limit load. By contrast, the pure
compression tests developed non-axisymmetric buckling
modes in concert with the observations of Ref. [1].
The stress at the onset of wrinkling was found to
decrease with internal pressure, whereas the critical strain
remained relatively unaffected. The wrinkle wavelength
increased rather signicantly with pressure. The critical
stress and wavelength were established using classical
plastic bifurcation theory based on deformation theory of
plasticity. By contrast, the corresponding strain was
evaluated using the ow theory of plasticity, due to the
non-proportionality of the stress history followed. Measured anisotropy was introduced in both models using
Hills quadratic anisotropic yield criterion. Inclusion of the
anisotropy was required for good agreement between
measured and calculated values of {sC, eC, lC}.
The evolution of wrinkling, and the resultant limit state,
were established by modeling a lC long periodic domain
assigned an initial imperfection corresponding to the
axisymmetric buckling mode calculated through the
bifurcation check. The model developed was discretized
by adopting kinematically admissible displacement functions. The inelastic material behavior was modeled through
the anisotropic ow theory of plasticity with isotropic
hardening. The variation of the limit strains with pressure
calculated for the two sets of experiments were found to be
in very good agreement with the measurements.
The limit strain is the design limit state of long cylinders
thick enough to buckle inelastically. The limit strain was
found to be strongly inuenced by the hardening of the
material, as higher hardening delays the development of
the limit load instability.
Acknowledgements
The work reported was conducted with nancial support
from a consortium of industrial sponsors. This support is
acknowledged with thanks.
References
[1] Bardi FC, Kyriakides S. Plastic buckling of tubes under axial
compressionPart I: Experiments. International Journal of Mechanical Sciences 2006; accepted for publication.

ARTICLE IN PRESS
J.A. Paquette, S. Kyriakides / International Journal of Mechanical Sciences 48 (2006) 855867
[2] Lee LHN. Inelastic buckling of cylindrical shells under axial
compression and internal pressure. Developments in Mechanics, In:
Proceedings of the Seventh Midwestern Mechanics Conference, vol. 1.
New York: Plenum Press; 1961; p. 190202.
[3] Murray DW. Local buckling, strain localization, wrinkling and
postbuckling response of a line pipe. Engineering Structures 1997;
19: 360371.

867

[4] Das S, Cheng RJJ, Murray DW, Wilkie SA, Zhou ZJ. Laboratory study
of local buckling, wrinkle development, and strains for NPS12 linepipe.
In: Proceedings of the ASME International Pipeline Conference, vol. 2.
Calgary, Alberta, Canada, October 15, 2000; p. 909915.
[5] Bardi FC, Kyriakides S, Yun HD. Plastic buckling of tubes under
axial compressionPart II: Analysis. International Journal of
Mechanical Sciences, 2006; accepted for publication.