Thin-Walled Structures 74 (2014) 59–69

Contents lists available at ScienceDirect

Thin-Walled Structures
journal homepage: www.elsevier.com/locate/tws

Experimental investigation on the effect of geometric imperfections

on the buckling and post-buckling behavior of steel tanks under
hydrostatic pressure
Amin Niloufari a, Hossein Showkati b, Mahyar Maali a,n, Seied Mahdi Fatemi a
a
Department of Civil Engineering, Islamic Azad University, Maragheh Branch, Iran
b
Department of Civil Engineering, Urmia University, Urmia, Iran

art ic l e i nf o a b s t r a c t

Article history: Weld-induced geometric imperfections have been reported to have especially detrimental effects on the
Received 22 October 2012 buckling resistance of shells under hydrostatic pressure. The effect of circumferential imperfections
Received in revised form caused by continuous welding on the joined areas between the curved panel edges of the cylindrical and
30 August 2013
conical shells of steel storage tanks with fixed conical roofs is the most important case in this context.
Accepted 5 September 2013
The present paper discusses 12 laboratory specimens in three groups, labeled SP200 (S ¼Specimen,
Available online 16 October 2013
P ¼Perfect, 200 ¼ height (mm) and radius of Cylinder), SP250 (S¼Specimen, P ¼ Perfect, 250¼ height
Keywords: (mm) and radius of Cylinder) and SP300 (S¼Specimen, P¼ Perfect, 300 ¼height (mm) and radius of
Imperfection Cylinder) loaded under uniform hydrostatic pressure. The samples were modified to include circumfer-
Buckling
ential imperfections at the junctions between the curved edges of the panels of the cylindrical and
Test
conical shells, with amplitudes of 2t, 4t and 8t in depth (where t is the thickness of the conical or
Welding
Steel tanks cylindrical shell). The results of testing under different codes are compared. This study shows that
geometrical imperfections at different ratios of t/R (where R the radius of the tanks) may have decreasing,
neutral or increasing effects on buckling resistance and can result in softening or stiffening behaviors of
the shells.
& 2013 Elsevier Ltd. All rights reserved.

1. Introduction in eccentricity, ovality and welding; they concluded that imperfec-

Currently, thin-walled shell structures are widely applied. The
reason for this growing trend is the lightweight and high-strength
characteristics of such shells. Geometric imperfections are inher-
ent in thin-walled structures and occur during construction and
assembly. Because buckling capacity can differ due to the slightest
imperfection, defect investigation becomes extremely important.
Buckling is a crucial failure phenomenon for tanks. In recent years,
the buckling of conical shells under varying loads, including
hydrostatic loading, has drawn a great deal of attention, especially
in tanks for liquids. However, it is important to mention that
the literature devoted to the analysis of geometrically imperfect
tanks shells is lacking. In 1995, H. Showkati and P. Ansourian [1]
investigated the influence of primary boundary conditions on the
buckling of shallow cylindrical shells under uniform external
pressure. In 2008, G. Forasassi et al. [2] considered the ratio of
diameter to thickness in investigating the effects of imperfections
n
Corresponding author. Tel.: þ 98 9143227636; fax: þ 98 4127224870.
E-mail addresses: h.showkati@mail.urmia.ac.ir (H. Showkati),
Mahyar.Maali@gmail.com, mahyar.maali@yahoo.com (M. Maali),
Mahdi.fatemi@gmail.com (S. Mahdi Fatemi).
tions caused by welding have the greatest effects on the buckling
of cylindrical shells. In 2008, Golzan and Showkati [3] studied the
buckling behavior of thin-walled conical shells under uniform
external pressure and showed that construction-induced imper-
fections have a significant effect on the buckling strength of
conical structures. In 2012, Maali et al. [4] studied the buckling
behavior of conical shells with weld-induced imperfections and
showed that weld-induced imperfections have a stiffening effect
on the buckling strength of conical structures. Donnell [1] calcu-
lated the buckling load for a cylindrical shell and obtained a
theoretical load on the cylindrical shell under hydrostatic pressure.
Jawad [5] calculated the buckling load for a conical shell by
analyzing the cylindrical shell and obtaining a theoretical load
on the conical shell under hydrostatic pressure. However, these
authors ignored the manufacturing processes. Additionally, Euro
Code 3, ECCS and DINI18800 [6–8] all contain set limitations for
rolling- and welding-induced imperfections.
Generally speaking, most of these experiments are conducted
on manufactured specimens. The present study investigated 4
specimens, labeled SP200 (S ¼Specimen, P¼Perfect, 200 ¼
height and radius of Cylinder), SP250 (S¼ Specimen, P¼ Perfect,
250 ¼height and radius of Cylinder) and SP300 (S ¼Specimen,

0263-8231/$ - see front matter & 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.tws.2013.09.005
60 A. Niloufari et al. / Thin-Walled Structures 74 (2014) 59–69

Nomenclature n number of half-wave imperfections

δmax imperfection's maximum depth
R radius of specimens Pcr critical buckling pressure
H height of cylinder E Young's modulus
H height of conical roof t thickness of specimens
δ0 imperfection's initial depth α semi-vertex angle of the cone
s length of imperfection's curvature

P¼Perfect, 300 ¼height and radius of Cylinder) under uniform
hydrostatic pressure. Each group contained one perfect specimen,
with the remaining specimens having circumferential imperfec-
tions with amplitudes of 2t, 4t or 8t (where t is the thickness of the
tank shell) induced by continuous welding on the curved junction
at the edge of the cylindrical and conical shells of steel storage
tanks with fixed cones.
The shell material consisted of mild steel with a yield stress of
194.238 MPa. All specimens were considered to be simply supported,
such that a radial constraint at the edge of the cone was the only
boundary condition. Hydrostatic pressure loading was applied by a
gauged vacuum pump, which applied suction to the shell. The results
of the initial and overall buckling and collapse pressures were
compared to the results reported in previous papers and in interna-
tional codes. The imperfect specimens were also compared to the
perfect specimen in each group. All of the tests were conducted by
the authors in the structural lab of Urmia University, Urmia, Iran.
2. Testing process
2.1. Test specimens and properties
The number and geometric dimensions of the specimens were
selected based on the available laboratory facilities and the experi-
mental goals. Primarily, the laboratory equipment and fabrication
processes limited the model geometry; one such important para-
meter was the vacuum pump capacity. Although the nominal avail-
able vacuum pump capacity was 100 kPa, the practical capacity was
70 kPa at most. Another significant factor was the size of the testing
system. In the present study, the system could accommodate a
sample up to 600 mm in diameter. Second, the thickness-to-radius
ratios (t/R) and imperfection amplitudes of the models were kept
within realistic boundaries. Typical values for the R/t ratio are in the
range of 300–1000 [3].
In this study, a total of twelve shell specimens were manufactured
and tested. In the majority of these specimens, a specific-amplitude
geometric imperfection was created. Among these twelve models,
three specimens, labeled SP200, SP250 and SP300, did not contain
any imperfections and were assumed to be reference models. All
specimens were tested with three different t/R ratios.
Welding two plates causes their edges to warp. One of the most
important parts of these tanks is the joint between the conical roof
and the cylindrical wall because of the high possibility of warping
and the great influence of the junction on the buckling and post-
buckling capacity. Because of this issue, the imperfections in the joint
between the roof and the body are believed to take a sinusoidal form.
Because the two imperfection types (dimples and lumps) cannot
possibly occur simultaneously in the joints, dimples are created, and
lumps are considered to occur in perfect edges. The number of half-
wave imperfections, n, is found continuously around the joint. In all
three groups, there was an attempt to keep the maximum depth of
the imperfections consistent (δmax). These depths are coefficients of
thickness, and they are labeled 2t, 4t and 8t. The thickness of the
specimens was 0.6 mm, and the height to radius ratio was 1(H/R¼1)
for the cylindrical part and 0.25(h/R¼0.25) for the conical roof. The
details of the specimens are presented in Table 1.
The specimen names have the form (AtB), where At is the
maximum depth of the imperfection (δmax), t is the thickness of
the shell, and B is the radius of the cylindrical shell.
Three tensile coupon tests were performed to obtain the prop-
erties of the shell material. The average yield and failure stresses of
the steel were found to be 194.238 MPa and 325.495 MPa, respec-
tively. The Young's modulus and Poisson's factor were found to be
200 GPa and 0.28, respectively.
2.2. Manufacturing process
An important factor in shell experiments is the specimen
fabrication quality, including both the choice of material and the
welding method. Several fabrication techniques have been devel-
oped for conical shells [9–10]. One of the methods is soldering the
seams, which results in improvements in both manufacturing and
testing processes and allows for accurate predictions of weld
performance. The present study uses this method.

Table 1
Aspect ratios and initial geometries of the specimens.

Specimen label Imperfection δmax Half-wave imperfections (n) Thickness (t) (mm) Radius (R) (mm) Height cylinder (mm) Height cone (mm) t/R

SP200 Perfect 0 0.6 200 200 50 0.003

S2t200 2.08t 20 0.6 200 200 50 0.003
S4t200 4.27t 14 0.6 200 200 50 0.003
S8t200 8.45t 10 0.6 200 200 50 0.003

SP250 Perfect 0 0.6 250 250 62.5 0.0024

S2t250 2.15t 22 0.6 250 250 62.5 0.0024
S4t250 4.08t 16 0.6 250 250 62.5 0.0024
S8t250 7.29t 12 0.6 250 250 62.5 0.0024

SP300 Perfect 0 0.6 300 300 75 0.002

S2t300 2.16t 24 0.6 300 300 75 0.002
S4t300 3.87t 18 0.6 300 300 75 0.002
S8t300 8.76t 12 0.6 300 300 75 0.002
 A. Niloufari et al. / Thin-Walled Structures 74 (2014) 59–69 61

After defining the accurate dimensions of the specimens, steel
plates were cut and rolled into cylinders and cones using a rolling
machine. In the present study, the specimen plates were joined by
soldering.
To join the conical roofs of the specimens (which included an
imperfection) to the cylindrical body, the perfect edges of the
cones were first soldered to the edge of the cylindrical wall. Then,
the imperfect edges were pushed by hand and joined together,
leading to the desired imperfection. Fig. 1 shows the details of the
conical roofs, the cutting of the plates and the final construction of
the specimens. It is important to note that soldering was per-
formed from both sides of the sample. It was also important that

c
δ0 R- δmax

h
X= s/2 a
X=δ0 b
s
a

b X= s/2 R H

Fig. 1. The process of construction.

Table 2
Detailed descriptions of specimens.

Specimen R (mm) t/R δ0 n s (mm) a (mm) b (mm) c (mm) δmax

SP200 200 0.003 Perfect – – – – – Perfect

S2t200 200 0.003 1.99t 20 31.414 4.099 15.708 183.603 2.08t
S4t200 200 0.003 4.06t 14 44.88 6.033 22.44 175.867 4.27t
S8t200 200 0.003 7.69t 10 62.832 8.778 31.416 164.889 8.45t

SP250 250 0.0024 Perfect – – – – – Perfect

S2t250 250 0.0024 2.06t 22 35.7 4.629 17.85 231.484 2.15t
S4t250 250 0.0024 3.89t 16 49.088 6.519 24.544 223.925 4.08t
S8t250 250 0.0024 6.91t 12 65.45 8.943 32.725 214.23 7.29t

SP300 300 0.002 Perfect 0 – – – – Perfect

S2t300 300 0.002 2.07t 24 39.27 5.063 19.635 279.746 2.16t
S4t300 300 0.002 3.7t 18 52.352 6.887 26.176 272.452 3.87t
S8t300 300 0.002 8.3t 12 78.538 10.732 39.269 257.072 8.76t

Grooved plate

Strain geuge

Vacuum pump
Circular groove

Cord-oriented weiding

Transducer
Silicon Glue

Fig. 2. Experimental test system.

62 A. Niloufari et al. / Thin-Walled Structures 74 (2014) 59–69
Strain
Testing specimen
Supporting base
Load cell
Vacuum pump
Fig. 3. Experimental test system.
any sudden opening of the joints be prevented during testing.
Table 2 shows detailed descriptions of the specimens.
2.3. Testing system
The test apparatus for all specimens consisted of two parts (Figs. 2
and 3), which were invented by H. Showkati in Urmia University,
Urmia, Iran. The first part, which was designed to hold the test speci-
men in the desired position, comprised two rigid, circular, grooved
plates. These grooves held both sides of the specimen. Four long
threaded bars were provided to adjust the plates to the height of the
specimen (truncated conical shells or cylindrical shells). The second
part of the apparatus consisted of a small platform that held a vacuum
pump [3]. This pump was employed to generate a uniform hydrostatic
pressure over the shell surface. Accurate measurements of the test
results were obtained via two circumferentially and longitudinally
mounted strain gauges (YEFLA-5), a manometer and five transducers.
All of the collected data were processed using a data logger and the
UCAM-20PC software.
For all specimens, the relevant edge was simply supported by
only a radial restraint. This is a classic boundary condition in the
literature on shells. For this purpose, a grooved steel supporting
plate was used for all specimens (Fig. 2a, b). Silicon glue was then
applied at the supported edge of the shell to stop air leakage
during vacuum loading (Fig. 2c). In this rotation-free edge condi-
tion, we could provide a radial restraint; therefore, it was exactly
the same as the classic boundary condition. It is well known that
the most effective edge restraint is in the radial direction for thin-
walled conical and cylindrical shells. Other conditions, such as
rotational or axial restraints, have smaller effects on the buckling
behavior of shells under hydrostatic pressure. The investigation of
different boundary conditions is beyond the scope of this paper.
Each specimen was fabricated using cord-oriented welding
along the rolled sheet fragment edges, as seen in Fig. 2c.
As seen in Fig. 2b, a uniform hydrostatic pressure was applied
to the shells via a gauged vacuum pump. The vacuum pump had a
hydrostatic pressure capacity of 70 kPa.
3. Measurement of imperfections
Because both the buckling and post-buckling capacities of thin-
walled shells are extremely sensitive to imperfections, a precise
measurement of both the imperfections due to the construction
Computer
Data
logger
process and the local imperfections created in this study were of
crucial importance to the accuracy of the tests.
In addition, precisely measuring the deformed shapes of the
shells during loading helped to determine and compare the buck-
ling/post-buckling/collapse modes with theoretical predictions.
Many techniques for measuring shell imperfections have been
proposed [11]. LVDTs (Linear Variable Differential Transformer)
and other contact probes were used in most of the earlier mea-
surement systems [12–13]. A simpler method was used in the
present measurement system for appraising both the initial and
local imperfections induced by the construction of the specimen.
Because complete measurements of a conical surface require a
three-dimensional investigation of the radial, circumferential and
meridional coordinates (Mesh), manual scanning was implemented
as the measurement technique in this study. First, meridional
lines were drawn on the test shells at 5.23-cm increments. Then,
circumferential lines were added to these lines at 2-cm increments.
Finally, all of the intersections of the meridional and circumferential
lines were recorded. To record the intersection locations, a machine
was installed on the platform, and a meridional ruler was placed
on the shell at a 901 angle. This ruler was set on the side of
the specimen, and the distance between the horizontal lines was
measured with 0.01-mm precision.
For all specimens and for each node of the obtained mesh, a
three dimensional coordinate (r, θ, z) was measured carefully. The
recorded data revealed that the size of the local longitudinal
imperfections was very large; therefore, initial buckling occurred
at these imperfections, and other imperfections on the shell were
less significant. It should be noted that, because there were one or
two large specific imperfections, the loads on other parts of the
local imperfections were recorded. This measurement confirmed
the uniformity of the imperfections created by the machines in the
specimens built for the test.
The exact positions of the circumferential (horizontal) and merid-
ional (vertical) strain gauges and transducers are presented in Table 3.
4. Test results
4.1. Comparison between the results of the first group of samples
(t\R ratio equal to 0.003)
During the initial loading stages, the circumferential and
meridional strains were similar and approximately relative to the
load magnitude (Figs. 4,6 and 8), indicating that the linear and
 A. Niloufari et al. / Thin-Walled Structures 74 (2014) 59–69 63

Table 3
Layout of strain gauge and transducers for all specimens.

Specimenlabel Transducers Strain gauges

T3 T4 T5 T7 T8 Horizontal gauge Vertical gauge

SP200 Top 59.51 59.51 239.51 59.51 129.51 127.51

100 mm Edge 175 mm 175 mm 175 mm 175 mm
S2t200 Top 3301 3301 3301 1501 381 401
110 mm Edge 190 mm 170 mm 130 mm 130 mm
S4t200 1201 1201 Top1 3001 – 1901 1931
175 mm 70 mm Edge – 68 mm 70 mm
S8t200 601 Top 601 2401 601 781 801
185 mm Edge 140 mm 95 mm 170 mm 170 mm
SP250 1001 Top – 1001 1001 1201 1201
225 mm – 145 mm Edge 225 mm 215 mm
S2t250 Top 2101 301 301 301 1281 1301
180 mm Edge 160 mm 240 mm 180 mm 180 mm
S4t250 – 2101 2101 Top 2101 1321 1301
– 140 mm 140 mm Edge 160 mm 160 mm
S8t250 – Top 1401 1401 1401 2051 2101
– Edge 230 mm 130 mm 130 mm 130 mm
SP300 – 201 Top 201 201 3221 3201
– 130 mm 280 mm Edge mm 190 mm 190 mm
S2t300 301 Top 1501 301 301 2721 2701
160 mm Edge Edge 280 mm 160 mm 160 mm
S4t300 – Top 2101 2101 2101 1521 1501
– 160 mm Edge 280 mm 190 mm 190 mm
S8t300 – 2101 2101 Top 2101 1421 1401
– 160 mm 280 mm Edge 190 mm 190 mm

All distances are measured from the top along the slant length.
Edge: joint between cone and cylinder shells.
Top: apex of conical shell.

SP200
25 25

20 20
Pressure (kPa)
Pressure (kPa)

15 15

10 10

T3
T4 5 5
T5
T7 Meridional
T8 0 Circumferential 0
-20 -15 -10 -5 0 -0.0012 -0.001 -0.0008 -0.0006 -0.0004 -0.0002 0
Total Displacement (mm) Strain

30 30
S2t200
25 25
Pressure (kPa)

Pressure (kPa)

20 20

15 15

T3 10 10
T4
T5 5 Meridional 5
T7 Circumferential
T8 0 0
-21 -16 -11 -6 -1 -0.0009 -0.0007 -0.0005 -0.0003 -0.0001 0.0001
Total Displacement (mm) Strain

24 20
S4t200
20
15
Pressure (kPa)
Pressure (kPa)

16

12 10

8
T3 5
T4 4
T5
T7
0 0
-22 -20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 -0.0005 0.0005 0.0015 0.0025 0.0035 0.0045
Total Displacement (mm) strain

Fig. 4. Load–displacement and load–strain graphs for the SP200, S2t200 and S4t200 specimens.
64 A. Niloufari et al. / Thin-Walled Structures 74 (2014) 59–69

Sole-fish

Piscivorous bird beak

Fig. 5. Post-buckling images displaying the sole-fish and piscivorous bird beak phenomenon and generated waves for (a) SP200 (b, c) S2t200 (d) S4t200 and (e, f) S8t200.

25 25
S2t250

20 20
Pressure (mm)

pressure(kPa)

15 15

10 10
T3
T4
5 5
T7 Meridional
T8 Circumferential
0 0
-22 -17 -12 -7 -2 -0.0045 -0.0035 -0.0025 -0.0015 -0.0005 0.0005
Total Displacement (mm) strain

30 30
S8t250
25 25
Pressure (kPa)

20 20
Pressure(kPa)

15 15

10 10

T5 5 5
Meridional
T7
T8 Circumferential
0 0
-23 -18 -13 -8 -3 -0.01 -0.008 -0.006 -0.004 -0.002 0
Totall Displecement(mm) Strain

Fig. 6. Load–displacement and load–strain graphs for the S4t250 and S8t250 specimens.

dominant axes displayed symmetric behaviors. Due to local
imperfections in the test specimens, the curves became nonlinear
immediately following the onset of loading. In other words, after
reaching the initial buckling load, the specimens experienced both
a slight load decrease and nonlinear deformations, which may
have resulted from one imperfection causing failure.
As seen in Fig. 5, buckling in all samples occurred in the conical
shell, and the number of generated waves in them is almost the
four same size waves.
After the occurrence of four waves on the conical shell,
the failure lines created by these waves were broken from two
thirds, or half of their lengths (except for the perfect geometry of
four-fifths the length from base-apex of the cone); then, the yield
location moved toward the apex of the cone shell. The conical
shells are going to be sole-fish shaped [3]. All samples in this
group failed via tearing or crumpling of the junction between the
conical and cylindrical shell; the Piscivorous bird beak phenom-
enon was generated. Table 4 and Figs. 4 and 5 show that the 2t and
4t imperfections reduced the buckling capacity by 39.91 and
30.98%, respectively, and the 8t imperfection increased the capa-
city by 59.15% with respect to the initial buckling loads of the
perfect samples. For the overall buckling load, the 2t and 4t
imperfections decreased by 20.53 and 11.31%, respectively, and
the 8t imperfection increased by 27.89% compared to the perfect
 A. Niloufari et al. / Thin-Walled Structures 74 (2014) 59–69 65

Fig. 7. Post-buckling images displaying the sole-fish and piscivorous bird beak phenomenon and generated waves for (a, b) SP250 (c, d) S4t250 and (e, f) S8t250.

20 20
SP300
16 Pressure(kPa) 16
Pressure(kPa)

12 12

8 8
T4
T5 4 4
T7 Meridional
T8 Circumferential
0 0
-23 -18 -13 -8 -3 2 -0.0049 -0.0039 -0.0029 -0.0019 -0.0009 0.0001
Totall displacement(mm) Strain

16 16
S4t300

12 12
Pressure(mm)
Pressure(kPa)

8 8

T4
T5 4 4
T7 Meridional
T8 Circumferential
0 0
-20 -15 -10 -5 0 -0.004 -0.003 -0.002 -0.001 0 0.001
Totall displacement(mm) strain

S8t300 16
16
Pressure (kPa)

12
12
Pressure (kPa)

8 8

T4 4 4
T5 Meridional
T7 Circumferential
T8 0 0
-20 -15 -10 -5 0 -0.004 -0.0035 -0.003 -0.0025 -0.002 -0.0015 -0.001 -0.0005 0
Total displacement (mm) Strain

Fig. 8. Load–displacement and load–strain graphs for the SP300, S4t300 and S8t300 specimens.

samples. It should be noted that, despite the reduction in the 4.2. Comparison between the results of the first group of samples
initial and overall buckling capacity relative to the perfect state, with the ratio of t\R equal to 0.0024
we observed post-buckling capacities of 178 to 307%. Thus, it can
be concluded that 4t, 2t and 8t imperfections produced softening As seen in Fig. 7, the buckling in all samples occurred in
behavior in the samples compared to the geometrically perfect the conical shell, and four same-size waves were consistently
state (Table 4). generated.
66 A. Niloufari et al. / Thin-Walled Structures 74 (2014) 59–69

After the occurrence of four waves on the conical shell, the respect to the geometrically perfect samples. However, the 8t
failure lines were broken at two thirds or half of their lengths imperfection caused stiffening behavior relative to the samples in
(base-apex cone). Then, the yield location moved toward the apex the geometrically perfect state.
of the cone shell. The conical shells assumed a sole-fish shape [3].
All samples of this group failed via tearing or crumpling of the
4.3. Comparison between the results of the first group of samples
junction between the conical and cylindrical shells. The piscivor-
with the ratio of t\R equal to 0.002
ous bird beak phenomenon occurred, except for the perfect
specimens, in which the formation of ten approximately similar
As seen in Fig. 9, the buckling in all samples of these groups
waves and ejection from the groove prevented the test from being
occurred in their conical shell, and four same-size waves were
continued. As reported in Table 5 and Figs. 6 and 7, the 8t and 4t
generated.
imperfections reduced the buckling capacity by 27.78 and 21.37%,
After creating four waves on the conical shell, the failure lines
respectively, and the 2t imperfection increased the initial buckling
were broken from two thirds or half of their lengths (base-apex
load by 5.98% compared to the perfect samples. The 2t imperfec-
cone). Then, the yield location moved toward the apex of the cone
tion decreased the overall buckling load by 14.69%, and the 4t and
shell in all samples. The conical shell samples assumed a sole-fish
8t imperfections increased it by 18.48 and 5.88%, respectively,
shape [3]. The failure mode of all samples is this group in tearing
compared to that of the perfect samples. We can see the post-
or crumpling at the junction between the conical and cylindrical
buckling capacity of 233 to 521%. Table 5 shows that the 2t and 4t
shells; all samples generated the Piscivorous bird beak phenom-
imperfections produced softening behavior in the samples with
enon. According to Table 6 and Figs. 8 and 9, the 2t, 4t and 8t
imperfections increase by 36.91, 20.8 and 0.67%, respectively, over
Table 4 the initial buckling loads of the perfect samples. The 2t, 4t and 8t
Comparison between the results of the first group of samples with t\R equal imperfections increase 63.96, 45.68 and 41.62%, respectively, over
to 0.003.
the overall buckling loads of the perfect samples.
Initial % to Overall % to Collapse % to Post- It should be noted that, despite the reduction in the initial and
buckling perfect buckling perfect (kPa) perfect buckling overall buckling capacities with respect to the perfect state, the
(kPa) (kPa) capacity post-buckling capacity increased by 215 to 345%. From Table 6, it
(%) can be concluded that the 2t and 4t imperfections produced
SP200 10.65 – 19 – 23.7 – 122.53
softening behavior compared to perfect samples, though the 8t
S2t200 7.35  30.98 16.85  11.31 29.95 26.37 307.48 imperfection produced stiffening behavior. A close inspection of
S4t200 6.4  39.91 15.1  20.53 20.02  15.51 215.89 Figs. 4,6 and 8, reveals that the straight lines have broken in some
S8t200 16.95 59.15 24.3 27.89 47.12 98.84 178.02 regions. Each of these breaks represents a buckling wave appears
on the shell. In some cases (e.g., the imperfect specimens), each

Table 5 Table 6
Comparison between the results of the first group of samples with the ratio of t\R Comparison between the results of the first group of samples with the ratio of t/R
equal to 0.0024. equal to 0.002.

Initial % to Overall % to Collapse % to Post- Initial % to Overall % to Collapse % to Post-

buckling perfect buckling perfect (kPa) perfect buckling buckling perfect buckling perfect (kPa) perfect buckling
(kPa) (kPa) capacity (kPa) (kPa) capacity
(%) (%)

SP250 5.85 – 10.55 – 33.75 – 476.92 SP300 3.725 – 4.92 – 15.45 – 314.76
S2t250 6.2 5.98 9  14.69 20.67  38.74 233.47 S2t300 5.1 36.91 8.07 63.96 19.57 26.7 283.82
S4t250 4.6  21.37 12.5 18.48 24.47  27.48 432.06 S4t300 4.5 20.8 7.17 45.68 14.17  8.252 215
S8t250 4.22  27.78 11.17 5.88 26.25  22.22 521.3 S8t300 3.75 0.67 6.67 41.62 16.7 8.09 345.33

Fig. 9. Post-buckling images displaying the sole-fish and piscivorous bird beak phenomenon and generated waves for (a, b) S2t300 (c, d) S4t300 (e, f) S8t300.
 A. Niloufari et al. / Thin-Walled Structures 74 (2014) 59–69 67

break represents two or even more symmetrical waves. The for- There is also a correlation between the load-total displacement
mation of these symmetrical concurrent waves proves how care- and load–strain plots of these two completely different tools, which
fully the specimens were constructed. can be taken as further proof of the installation and measurement

Table 7
The results of the Jawad equation for the laboratory specimens.

Jawad equation (kPa) Initial buckling (kPa) % to initial Overall buckling (kPa) % to overall Collapse (kPa) % to collapse

SP200 21.01 10.65  49.31 19  9.57 23.7 12.80

SP250 12.03 5.85  51.37 10.55  12.30 33.75 180.55
SP300 7.62 3.72  51.11 4.92  35.37 15.45 102.76

Table 8
Geometric tolerances according to different standards.

Definition of the depth of initial dimples Tolerances according to different standards

Depth of initial dimples in plates measured in vertical direction DIN 18800-4

tr o 1% of measuring length Lmx

ESSCRec{2008}
L ¼ Ir:tr ¼ w
tr o 1% of measuring length Lmx
pr ENV 1993-1-6 {2007}
Lmx ¼ Lg:tr ¼ Δw0
class A:tr r0.006Lmx
class B:tr r 0.010Lmx
class C:tr r 0.016Lmx
Initial dimple of circumferential weld seam DIN 18800-4

tr o 1% of measuring length Lmx

ESSCRec{2008}
Lmx ¼ Ir:tr ¼w
tr o 1% of measuring length Lmx
pr ENV 1993-1-6 {2007}
Lmx ¼ L:tr ¼Δw
class A:tr r0.006Lmx
class B:tr r 0.010Lmx
class C:tr r 0.016Lmx
68 A. Niloufari et al. / Thin-Walled Structures 74 (2014) 59–69

precisions. This can be observed in Figs. 5, 7 and 9, all transducers Table 9

and strain gages were installed at locations of high buckling Main parameters of the studies compared to standards.
potential. However, in a nearly perfect specimen, it is not easy to
δmax s s/t δmax/s Initial buckling to Overall buckling to
predict the initial buckling location exactly. Table 3 shows the precise (mm) (%) perfect (%) perfect (%)
location of the gages installed on all of the specimens. The reason
why strain gauge graph moves left or right is due to random direc- S2t200 2.08t 31.416 52.36 3.97  30.98 181.22
tion of buckled panels. S4t200 4.27t 44.88 74.8 5.71  39.91 88.03
S8t200 8.45t 62.832 104.72 8.07 59.15 342.49

S2t250 2.15t 35.7 59.5 3.61 5.98 253.42

S4t250 4.08t 49.088 81.81 4.99  21.37 403.85
5. Approximate theoretical prediction S8t250 7.29t 65.45 109.08 6.68  27.78 348.72

S2t300 2.16t 39.27 65.45 3.30 36.91 425.50

An equation for the critical buckling load on thin-shelled cones
was developed in 1980 by Jawad [5] and is shown for an ideal
geometric form in Eq. (1):
"
 #
R t e 2:5
P cr ¼ 0:92E ð1Þ
Le R
where E is the Young's modulus; te is the effective thickness of the
cone and is equivalent to t cos α, where α is the semi-vertex angle
of the cone and t is the thickness of the cone wall; and Le is the
effective length of the cone, found by L/2(1 þr/R), in which r is
the upper radius, R is the lower radius and L is the height of the
cone.
Table 7 presents the results of the Jawad equation for the labo-
ratory specimens. With t/R equal to 0.003 and 0.0024, the overall
buckling loads sustained by the testing samples were closer to the
buckling capacities determined by the Jawad approximate relation
and differed by  9.57% and  12.3%, respectively. However, with t/R
equal with 0.002, the difference increased and became  35.37%.
Table 7 shows that the initial buckling of perfect specimens
differed by  51% from the buckling capacity computed by approx-
imate Jawad equation. In Table 7, it is seen by increasing height
of shell, the shell buckling is easier and faster than the shorter
shell. In other words the shorter shell shows more stability and
resistance against buckling. Jawad equation predicts buckling load
of short shell, which is completely different from the buckling of
long one.
6. Geometrical imperfections tolerated by the standards
International codes for the design of thin-walled shells such as
Euro Code 3, ECCS and DIN 18800-4 simply limit the permitted
deformations and disturbances during the welding and construc-
tion of conical shells (Table 8).
The various methods proposed for measuring the magnitude
of the initial imperfection depth on the shell's skin in the long-
itudinal direction include the following:
lg ¼4(r/t)0.5.
lg ¼25tr500 mm at welds for prENV 1993-1-6 (2007): Euro-
code 3 [6].
lr ¼4(r/t)0.5 r0.95 times the distance of the welds.
lr ¼25t at welds for ECCS (2008) [7].
lmx ¼ 4(r/t)0.5 r 2000 mm.
lmx r 500 mm at welds for DIN 18800 [8].
In these equations, r is the middle surface of the conical shell; t
is the thickness of the cone; and lmx, lr and lg are all of the
longitudinal imperfections created circumferentially. The mx, r
and g indexes in different codes represent the same length in
various forms. Table 8 demonstrates the exact location of this
dimension. In the present study, the size of the dent tv and the
length of the curve lmQ were designed and created with precise
S4t300 3.87t 52.352 87.25 4.43 20.80 280.54
S8t300 8.76t 78.538 131.39 6.69 0.67 384.32
tools, so there is no need to compute the limitations of the above-
mentioned formulas.
The ratio of the length of the imperfection curvature to the
thickness of the specimens (s/t) and the ratio of the imperfection's
maximum depth to the length of its curvature (δmax/s) are
presented in Table 9. It is worth mentioning that the parameters
mentioned here are the same parameters tolerated by the stan-
dards [6–8]. In practice, the initial buckling loads are important
because they reveal the beginning of failure; therefore, the rates of
difference between the initial buckling and post-buckling capa-
cities of the specimens in contrast to the perfect forms are shown
in the tables as well to provide a better comparison.
Although the imperfections of the specimens in this test were
greater than those tolerated by the standards, all specimens
showed a large post-buckling capacity compared to the initial
buckling of the perfect specimens. Moreover, as the t/R ratio
increased, the 2t and 4t imperfections moved toward increased
initial buckling compared to perfect specimens. In the case of the
8t specimen, when the t/R ratio was 0.003, the initial buckling
capacity increased, but in the case of the t/R of 0.0024, the initial
buckling capacity decreased. However, as the t/R ratio reached
0.002, no effect was observed. It should be noted that, to provide
more accurate results, more data must be collected. In the joints
between the roof and the body, under uniform vacuum load, the
standard load dictated by codes could be increased according to
the t/R ratio and the maximum depth of the imperfection.
7. Conclusions
This study discusses the buckling behavior of thin-walled tanks
shells. Twelve simply supported specimens with various imperfec-
tion amplitudes were loaded under hydrostatic pressure. The
following results were noted:
1. In all laboratory tanks with the same thickness of wall and roof
under hydrostatic pressure, buckling occurred in the conical
shells, and there were generally four nearly equal waves.
2. In all laboratory samples, after the occurrence of four waves on
the conical shell, the submission lines were broken from two-
thirds or half their lengths (base-apex cone). Then, the yield
location moved toward the apex of the cone shell. The conical
shell samples assumed a flounder shape.
3. The failure mode of all laboratory samples, except SP250
sample, was tearing or crumpling of the junction between
the conical and cylindrical shells. In the SP250 sample, the
creation of a wave in the cylindrical shell and ejection of the
sample from the slot of machine stopped the test.
 A. Niloufari et al. / Thin-Walled Structures 74 (2014) 59–69
4. With t/R equal to 0.003, the 2t and 4t imperfections reduced
the initial buckling loads by 39.91 and 30.98%, respectively,
and the 8t imperfection produced an increase of 59.15%
relative to the perfect samples.
5. With t/R equal to 0.003, the 2t and 4t imperfections reduced
the failure buckling loads by 20.53 and 11.31%, respectively,
and the 8t imperfection produced an increase of 27.89%
relative to the perfect samples.
6. With t/R equal to 0.0024, the 2t imperfection increased the
initial buckling load by 5.98%, and the 4t and 8t imperfections
produced decreases of 27.78 and 21.37%, respectively, com-
pared to the perfect samples.
7. With t/R equal to 0.0024, the 2t imperfection reduced the
failure buckling load by 14.69%, and the 4t and 8t imperfec-
tions produced increases of 18.48 and 5.88%, respectively,
compared to the perfect samples.
8. With t/R equal to 0.002, the 2t, 4t and 8t imperfections
produced increases of 36.91, 20.8 and 0.67%, respectively, of
the initial buckling loads of the perfect samples.
9. With t/R equal to 0.002, the 2t, 4t and 8t imperfections
produced increases of 63.96, 45.68 and 41.62%, respectively,
of the failure buckling loads of perfect samples.
10. With t/R equal to 0.003 and 0.0024, the overall buckling loads
from testing samples were closer to the buckling capacity
resulting from Jawad's approximate relation and differed by
 9.57% and  12.3%, respectively. With t/R equal to 0.002, the
difference increased and became  35.37%.
11. The rates tolerated by the standards for the geometric imper-
fections in the joints between the roof and the cylindrical part
of steel tanks under hydrostatic pressures could be increased
based on the R/t ratio of the tanks and the maximum depth of
the imperfection.
12. Initial buckling of perfect specimens differed by  51% from
the buckling capacity computed by the approximate Jawad
equation.
69
References
[1] Showkati H, Ansourian P. Influence of primary boundary conditions on the
buckling of Shallow cylindrical shells. Constructional steel Research 1995;36
(No. 1):53–75.
[2] Lo Frano R, Forasassi. G. Experimental evidence of imperfection influence on
the buckling of thin cylindrical shell under uniform external pressure. Journal
of Nuclear Engineering and Design 2008:239–193-200.
[3] Golzan BS, Showkati H. Buckling of thin-walled conical shells under uniform
external pressure. Thin-Walled Structures 2008;46:516–29.
[4] Maali M, Showkati H, Fatemi SM. Investigation of the buckling behavior of
conical shells under weld-induced imperfections. Thin-Walled Structures
2012;57:13–24.
[5] Jawad Maan H. Theory and design of plate and shell structures. . Chapman &
Hall, Inc; 344.
[6] EN 1993-1-6: Eurocode 3. Design of steel structures. Part 1.6: General rules—
Strength and stability of shell structures. Eurocode 3 Part 1.6, CEN, Brussels;
2007.
[7] ECCS EDR5. European Recommendations for steel construction. Buckling of
shells, 5th ed. In: Rotter JM, Schmidt H, editors. European convention for
constructional steelwork. Brussels; 2008 384 pp.
[8] DIN 18800. Stahlbauten.Teil 4: Stabilitatsfalle, Schalenbeulen 1990.
[9] Babcock CD. Experiments in shell buckling. In: Fung YC, Sechler EE, editors.
Thin-shell structures: theory, experiment and design. Englewood Cliffs, NJ:
Prentice-Hall; 1974. p. 345–69.
[10] Singer J. Buckling experiments on shells- a review of recent developments.
Solid Mech Arch 1982;7:213–313.
[11] Singer J, Abramovich H. The development of shell imperfection measurement
techniques. Thin-Walled Structures 1995;23:379–98.
[12] Dowling PJ, Harding JE Experimental behavior of ring and stringer stiffened
shells. In: Harding JE, Dowling PJ, Agelidis N, editors. Buckling of shells in
offshore structures. London, Granada; 1982, pp. 73-107.
[13] Walker AC, Andronicou A, Shridharan S. Experimental investigation of the
buckling of stiffened shells using small scale models. In: Harding JE, Dowling
PJ, editors. Buckling of shells in offshore structures. London, Granada; 1982,
pp. 25–43.