Thin-Walled Structures 79 (2014) 83–94

Contents lists available at ScienceDirect

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

Buckling of cylindrical open-topped steel tanks under wind load

Yang Zhao n, Yin Lin
Spatial Structures Research Center, Zhejiang University, Hangzhou 310058, China

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

Article history: Vertical cylindrical welded steel tanks are typical thin-walled structures which are very susceptible to
Received 7 December 2013 buckling under wind load. This paper investigates the buckling behavior of open-topped steel tanks
Received in revised form under wind load by finite element simulation. The analyses cover six common practical tanks with
6 February 2014
volumes of 2
103 m3 to 100
103 m3 and height-to-diameter ratios H/Do1. The linear elastic
Accepted 11 February 2014
bifurcation analyses are first carried out to examine the general buckling behavior of tanks under wind
load, together with comparison to that of tanks under uniform pressure and windward positive pressure
Keywords: (only loaded by positive wind pressure in the windward region). The results show that for larger tanks in
Buckling practical engineering, the stability carrying capacity of wind load is relatively lower. It is also indicated
Steel tanks
that the buckling behavior of tanks under wind load is governed by the windward positive pressure
Wind load
while wind pressure in other region of tank essentially has no influence on the buckling performance.
Open-topped
Imperfection sensitivity The geometrically nonlinear analyses are then conducted to investigate the more realistic buckling
Wind girder behavior of tanks under wind load. It is found that the buckling behaviors of perfect tanks and imperfect
tanks are much different. The weld induced imperfection only has little influence on the wind buckling
behavior while the classical buckling mode imperfection has significant influence, leading to a
considerable reduction of wind buckling resistance. The influences of thickness reduction of cylindrical
wall, liquid stored in the tank and wind girder on the buckling behavior are also examined. It shows that
the thickness reduction of cylindrical wall considerably reduces the wind buckling resistance while
sufficient liquid stored in the tank and wind girder significantly increase the wind buckling resistance.
& 2014 Elsevier Ltd. All rights reserved.

1. Introduction Because of serious economic losses and environmental problems due

Vertical cylindrical welded steel tanks are widely used for fluid
and bulk storage in industrial and agricultural plants. They usually
consist of a thin bottom plate, a cylindrical shell with uniform or
stepped thickness and a closed-roof or open-top (with or without
floating roof) [1]. The typical structure of cylindrical open-topped
steel tanks is shown in Fig. 1. With the development of economy
and oil industry, more and more oil storage tanks are put into
service in recent decades, especially large tanks. Some tanks even
have diameter larger than 100 m, with volumes of more than
100
103 m3. Fig. 2 shows steel tanks in practical engineering.
As typical thin-walled structures, tanks are very susceptible to
buckling under wind load especially when they are empty or
partially filled. Over the past few decades, buckling failures of
cylindrical steel tanks and silos during windstorm have occurred
in many countries and regions [2–4]. Buckling of tanks sometimes
even occurs under moderate wind load during their construction [5].
n
Corresponding author.
E-mail address: ceyzhao@zju.edu.cn (Y. Zhao).
to the destruction of storage tanks, studies about buckling of tanks
under wind load have been conducted extensively over the past few
decades.
Kundurpi et al. [6] studied the buckling behavior of open-topped
cylindrical tanks due to wind load based on the energy theory and
compared with experimental evidence. It was shown that the
engineering practice of determining buckling load was conservative.
Uematsu and Uchiyama [7] investigated the buckling behavior of
closed-ended, thin cylindrical shells by wind-tunnel tests. They found
that the buckling load was not sensitive to the wind pressure
distribution and the buckling behavior exhibited a pronounced non-
linearity. An empirical formula for the buckling pressure considering
the height/radius ratio and the radius/thickness ratio was also
proposed by them. Greiner and Derler [8] performed numerical
analyses on the wind buckling of cylindrical shells considering several
different imperfection shapes. It was found that the longer cylinders
were most sensitive to local rectangular or ring shape imperfections
while stocky cylinders were more sensitive to global eigenmode-
shaped imperfections. Schmidt et al. [9] carried out tests on both PVC
and steel cylinder models under a “wind-like” load to study the post-
buckling phenomena and provided some design recommendations

http://dx.doi.org/10.1016/j.tws.2014.02.010
0263-8231 & 2014 Elsevier Ltd. All rights reserved.
84 Y. Zhao, Y. Lin / Thin-Walled Structures 79 (2014) 83–94

From literature reviewed above, it can be found that few

studies have been conducted on the buckling behavior of practical
open-topped steel tanks with stepped wall when they are subject
to wind load including internal pressure.
This paper covers six common practical open-topped tanks
with volumes of 2
103 m3 to 100
103 m3 and aspect ratios
(height-to-diameter) H/D o1. The internal pressure due to wind
is included in present study. The following three types of buckling
analyses recommended by EN 1993-1-6 [18] are performed for
numerical investigation: (a) LBA—linear elastic bifurcation analysis
of the perfect tank; (b) GNA—geometrically nonlinear elastic
analysis of the perfect tank; (c) GNIA—geometrically nonlinear
elastic analysis of the imperfect tank.
Fig. 1. Typical structure of open-topped steel tanks.
The layout is organized as follows: Section 2 expatiates upon
the structure prototypes of six tanks, the finite element models
used for analyses and the wind pressures on both external and
internal wall of tank. Section 3 presents the linear elastic buckling
behavior of tanks under wind load, together with comparison to
that of tanks under uniform pressure and windward positive
pressure (only loaded by positive wind pressure in the windward
region). Geometrically nonlinear buckling behavior is investigated
in Section 4, using perfect and imperfect models. The influences of
the thickness reduction of cylindrical wall, the liquid stored in tank
and the wind girder on the buckling behavior are examined
and discussed in Section 5. Finally, some valuable conclusions
are drawn in Section 6.

2. Computational models

2.1. Structure prototypes

Fig. 2. Steel tanks in practical engineering.
Table 1 summarizes the geometries of tanks TK1–TK6, which
are representative of six common practical tanks with volumes of
based on the experimental results. Flores and Godoy investigated the 100
103 m3, 50
103 m3, 20
103 m3, 10
103 m3, 5
103 m3,
buckling behavior of short cylindrical shells subject to strong wind 2
103 m3, correspondingly. These six types of tanks are con-
load by numerical approach, in which models were representative of structed extensively not only in China but also in other countries
tanks devastated during hurricane Marilyn in the Caribbean islands. [19]. The structure of cylindrical shell and the dimensions (in mm)
Portela and Godoy [10,11] addressed the buckling of steel tanks with a of shell courses for each tank are summarized in Table 2. Data of
conical or dome roof due to wind load, in which the imperfection these tanks refers to practical engineering and some reasonable
sensitivity was examined through geometrically nonlinear analyses. simplifications are made for analyses. It can be seen that as the
They also investigated open-topped tanks using the same cylindrical volume increases, the aspect ratio H/D decreases while the average
wall model and wind load to confirm the roof effect on the buckling thickness of cylindrical shell increases. Fig. 3 shows the relative
behavior of cylindrical tanks. Jaca et al. [12] adopted a reduced sizes of different tanks.
stiffness approach for evaluating the lower bound of buckling of
open-top cylindrical tanks under wind load based on reduced energy 2.2. Finite element model
model. They found that results obtained from reduced stiffness
approach constituted a lower bound to those obtained from numerical The finite element package ABAQUS is employed to carry out
analyses as well as experiments. Further work had been performed by the analyses. The 8-node, quadrilateral, first-order interpolation,
Sosa and Godoy [13] and Jaca and Godoy [14], which put the method stress/displacement continuum shell element with reduced inte-
into a wider application. Jaca and Godoy [5] noticed that wind- gration S8R5 is chosen to discretize the cylindrical wall. This
induced damage of tanks occurred not only during windstorm but element has 5 degrees of freedom per node: translations in the
also under moderate wind load during their construction. They tried nodal x, y, and z directions and rotations about the nodal x, and
to reveal the mechanism of collapse through geometrically nonlinear y-axes, which is more economical in comparison to that using six
analyses. This work was also carried out by Borgersen and Yazdani degrees of freedom [20]. The fixed boundary condition (Ux, Uy, Uz,
[15], in which some design recommendations and construction Rotx, Roty, Rotz ¼0) is applied on the bottom of the FE model. The
methods to prevent failure during construction were given. Chen material of cylindrical shell is assumed to be isotropic elastic with
and Rotter [16] performed analyses on the buckling of anchor
cylindrical shells due to wind pressure with focus on the uniform Table 1
Geometries of tanks.
thickness cylinders, covering a wide range of tanks with different
aspect ratios. Zhao et al. [17] investigated the buckling behavior of TK1 TK2 TK3 TK4 TK5 TK6
steel silos subject to wind pressure through a great deal of numerical
analyses, which aimed to improve the understanding of buckling H (m) 21.8 19.35 15.85 15.85 14.27 12.69
behavior of large circular steel silos subject to wind pressure. It was D (m) 80 60 40.5 28.5 22 14.5
H/D 0.27 0.32 0.39 0.56 0.65 0.88
indicated that the buckling resistance of steel silo was closely V (m3) 100,000 50,000 20,000 10,000 5000 2000
correlative with loading conditions as well as geometrical parameters.
 Y. Zhao, Y. Lin / Thin-Walled Structures 79 (2014) 83–94 85

Table 2
Structure of cylindrical shell for each tank (mm).

Course (hi
ti) TK1 TK2 TK3 TK4 TK5 TK6

1 (Bottom) 2420
32 2410
26 2000
24 2000
18 1800
14 1800
8

2 2420
28 2410
22 2000
20 2000
16 1800
12 1800
6
3 2420
24 2410
20 2000
18 2000
14 1800
10 1800
5
4 2420
20 2410
16 2000
16 2000
12 1800
8 1800
5
5 2420
16 2410
14 2000
12 2000
8 1800
6 1800
5
6 2420
12 2410
10 2000
8 2000
8 1800
6 1800
5
7 2420
12 2400
10 1885
8 1885
6 1800
6 1820
5
8 2380
12 2400
10 1885
6 1885
6 1590
6 70
5 (Angle)
9 2380
12 90
10 (Angle) 80
6 (Angle) 80
6 (Angle) 80
6 (Angle)
10 100
10 (Angle)

TK1 TK2 TK3 TK4 TK5 TK6

Fig. 3. Relative sizes of different tanks.

expression as follows:
m
cp ðθÞ ¼ ∑ ai cos ðiθÞ ð2Þ
i¼0

where θ is the longitude measured from the windward, and ai is

the Fourier coefficient. The representative Fourier coefficients
obtained from typical studies and design codes are summarized
in Table 3.
Fig. 5 shows the distributions of wind pressure coefficient along
the circumference proposed by different authors and design codes.
It should be noted that the Fourier coefficients given above
relate only to cylinders with closed-top, not including the internal
wind pressure. For cylinders with open-top, an additional uniform
negative wind pressure coefficient should be subtracted to take
account of the internal suction [25]:
Ux, Uy, Uz, Rotx, Roty, Rotz=0 (
 0:8; H=D Z2
cp ¼ : ð3Þ
Fig. 4. Finite element model of tank (TK1).  0:5; H=D r1

In order to reflect the distinction of wind pressure distribution

an elastic modulus E of 2.06
105 N/mm2 and a Poisson's ratio μ of
0.3. The finite element model of tank is shown in Fig. 4.
2.3. Wind loads on tanks
Wind pressures p acting on the structure surfaces can be
defined as [21]
p ¼ cp qðzÞ
where cp is the wind pressure coefficient and qðzÞ is the velocity
pressure of incoming wind which varies with the height.
The wind pressure coefficients on cylinders usually vary along
both the circumference and the height. As the variation along the
height is not pronounced compared with circumferential variation,
the wind pressure coefficient is generally assumed to be constant
along the height and only dependent on the longitude. The
distribution of wind pressure coefficient along the circumference
of the cylinder has been proposed by several authors or specified
by design codes [8,22–24], using the Fourier series decomposition
for tanks with different aspect ratios, the pressure distributions
specified by EN 1993-4-1 [23] are used for analyses. For different
tanks, the circumferential distributions of wind pressure coeffi-
cient considering internal suction are shown in Fig. 6. Notice that
for tanks with aspect ratios H/Do0.5, the wind pressure distribu-
tion on the external wall should be calculated adopting the aspect
ratio H/D¼ 0.5 according to EN 1993-4-1 [23]. Thus tanks TK1–TK3
have same wind pressure distribution along the circumference.
The velocity pressure of incoming wind should be calculated as
follows:
ð1Þ
qðzÞ ¼ w0
ðh=h0 Þ2α ð4Þ
where w0 is the basic wind pressure at the reference height, h is the
loading height above ground level, h0 is the reference height and α is
the coefficient of terrain category. As storage tanks are usually
constructed in coastal area exposed to the open sea, the terrain
category A [26] is assumed and applied to wind load in the following
numerical analyses, i.e., α ¼0.12. In this paper, wind load applied on
each tank is defined by a factor λ, representing a given multiplier for
the reference pressure of 1 kPa at the height of 10 m (i.e., w0 ¼1000
N/m2), which represents a wind velocity of 40.8 m/s at height of 10 m.
86 Y. Zhao, Y. Lin / Thin-Walled Structures 79 (2014) 83–94

Table 3
Fourier coefficients obtained from different sources.

Source a0 a1 a2 a3 a4 a5 a6 a7

Greiner [8]  0.65 0.37 0.84 0.54  0.03  0.07

Rish [21]  0.387 0.338 0.533 0.471 0.166  0.066  0.055
ACI 334 [22]  0.2636 0.3419 0.5418 0.3872 0.0525  0.0771  0.0039 0.0341
EN 1993-4-1 [23]  0.54 þ 0.16 (D/H) 0.28 þ0.04 (D/H) 1.04–0.20 (D/H) 0.36–0.05 (D/H)  0.14þ 0.05 (D/H)

1.5 Table 4
Rish
Buckling load factors obtained from LBA.
1.0 Greiner
EN 1993-4-1,H/D=1.0 TK1 TK2 TK3 TK4 TK5 TK6
0.5 EN 1993-4-1,H/D=0.5
ACI 334 λcl,1 0.655 0.900 1.152 1.313 1.424 1.683
0.0 λcl,2 0.655 0.900 1.152 1.313 1.424 1.683
Cp

-0.5
stability carrying capacity of wind load is relatively higher, while
-1.0 for larger tanks, the stability carrying capacity of wind load is
relatively lower.
-1.5 Although the first two eigenvalues are same in magnitude, the
buckling modes attach to them are different, as shown in Fig. 7.
-2.0
0 45 90 135 180 The symmetric and anti-symmetric modes come in pairs by
θ/° alternating for all tanks, which can be regarded as phase differ-
ence. For tanks TK1, TK4 and TK6, the first buckling modes are
Fig. 5. External wind pressure coefficients along the circumference of cylinders.
anti-symmetric and the second modes are symmetric, while for
other tanks, the reverse is true. It is also found that the buckling
2.0 deformation mainly occurs in the windward region, where the
length of predominant wave increases as the aspect ratio increases
1.5 TK1~TK3 in general.
TK4 The buckling deformations vary up the height for different
1.0
TK5 tanks as shown in Fig. 8. It is observed that the patterns of
0.5 TK6 cylindrical shell deformation are not like those of common
cantilever structures such as cantilever beams. This is probably
CP

0.0 because the top angle iron has strengthened the stiffness of upper
shell, leading to shifting downward of the largest displacement.
-0.5
Although the first two circumferential modes are different, the
-1.0 first two vertical modes are in substantial agreement. The vertical
modes fall into two groups: (1) for tanks with lower aspect ratios
-1.5 (H/Do 0.5) and thick walls relatively, the deformation develops
0 45 90 135 180
upward as the aspect ratio increases and (2) for tanks with upper
θ/° aspect ratios (H/D40.5) and thin walls relatively (H/D40.5), the
Fig. 6. Applied wind pressure coefficients for different tanks. deformation develops downward as the aspect ratio increases.

3.2. Buckling under uniform pressure

3. Linear elastic bifurcation analyses (LBA)
It is meaningful to compare the buckling behavior of tanks
The linear elastic bifurcation analysis ignoring change in geome- under wind pressure with that of tanks under uniform pressure. In
try and degradation in material will provide a preliminary evaluation this section, the buckling behavior of tanks under uniform pres-
of buckling behavior, from which the obtained lowest buckling load sure is investigated through Donnell theory as well as numerical
usually constitutes the upper bound of stability carrying capacity and simulation.
the buckling mode can be introduced to the structure as an initial For short and intermediate length cylindrical shells, the critical
geometric imperfection for nonlinear analysis. In this section, the uniform pressure of buckling determined from Donnell theory is
bifurcation buckling analyses of tanks under wind load are carried calculated by
out and the bifurcation buckling behavior of tanks under uniform
2:59Et 2
pressure and windward positive pressure is also investigated for pcr ¼ pffiffiffiffiffiffiffiffi ð5Þ
comparison. HD D=t
where E is the elastic modulus and t is the uniform thickness of the
3.1. Buckling under wind load cylinder. For stepped shells, the equivalent height of the cylinder H
and equivalent thickness t can be calculated by the following
The first and second critical buckling load factors λcl,1 and λcl,2 equations [27]:
obtained from eigenvalue analyses for different tanks under wind sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 
load are listed in Table 4. It can be found that the first critical load t min 5
H ¼ ∑hi ð6Þ
factor is equal to the second critical load factor for each tank. The ti
buckling load increases from TK1 to TK6, whose aspect ratio
increases while volume decreases. That is, for smaller tanks, the t ¼ t min ð7Þ
 Y. Zhao, Y. Lin / Thin-Walled Structures 79 (2014) 83–94 87

1 TK1 1 TK1
Normalized displacement 0 0

Normalized displacement
-1 TK2 -1 TK2

TK3 TK3

TK4 TK4

TK5 TK5

1 TK6 1 TK6
0 0
-1 -1
-60 -40 -20 0 20 40 60 -60 -40 -20 0 20 40 60
θ/° θ/°
Fig. 7. Circumferential buckling modes for different tanks: (a) first modes and (b) second modes.

1.0 1.0

TK3 TK3
TK4 TK4
TK5 TK5
Z/H
Z/H

0.5 0.5
TK1 TK1
TK2 TK2

TK6 TK6

0.0 0.0
0.0 1.0/0.0 1.0 0.0 1.0/0.0 1.0
Normalized displacement Normalized displacement
Fig. 8. Vertical buckling modes for different tanks: (a) first modes and (b) second modes.

Table 5 In fact, the wind load can be regarded as sum of a uniform

Critical buckling pressures for different loading conditions (
103 N/m2). pressure with several additional cosine pressures, leading to
some similarities of buckling behavior between tanks under wind
Critical pressure TK1 TK2 TK3 TK4 TK5 TK6
load and uniform pressure, including buckling load and buckling
Uniform pressure 0.950 1.151 1.257 1.479 1.569 2.118 deformation.
Stagnation wind pressure 1.185 1.582 1.930 2.199 2.326 2.673
3.3. Buckling under windward positive pressure
The critical stagnation wind pressure is calculated by: λcl,1
w0
(H/10)0.24.

It is mentioned above that the predominant buckling deforma-

where hi, ti is the height and thickness of shell course, respectively,
and tmin is the thickness of thinnest shell course.
The critical uniform pressures of buckling calculated by Donnell
theory for different tanks are summarized in Table 5.
It could be seen that the critical uniform pressure of buckling is
larger than critical stagnation wind pressure of buckling for each
tank, with deviation of about 25–50%. The non-uniform distribu-
tion of wind pressure leading to pressure reduction in most
regions may be one of the reasons. Another aspect is that the
angle iron (see Fig. 1) strengthens the stiffness of resistance to
wind pressure which is not considered in the calculation using
Donnell theory. For a preliminary evaluation of the wind buckling
critical load, the Donnell theory may work although its result is
relatively conservative.
Fig. 9 shows the buckling modes of tanks under uniform
pressure. It can be seen that the length of predominant wave
increases as the aspect ratio increases which is similar to buckling
phenomenon of tanks under wind load.
tion is only registered in the windward region for tanks under
wind load (see Fig. 7). However, the buckling deformation is all
along the circumference for tanks under uniform pressure (see
Fig. 9). It is thus natural to surmise that the buckling behavior of
tanks under wind load may be governed by positive pressure in
the windward region which is referred to windward positive
pressure in this paper. To check this assumption, calculations are
carried out on tanks only loaded by positive pressure in the
windward region which has a same distribution as that of wind
load. Fig. 10 shows the two different load cases: wind load and
windward positive pressure, with inward pressure positive, which
are referred to load case 1 (LC1) and load case 2 (LC2), respectively.
The buckling load factors of tanks under windward positive
pressure are summarized in Table 6, in comparison to those of
tanks under wind load. It can be found that the buckling loads of
tanks under windward positive pressure are in substantial agree-
ment with those of tanks under wind load, with deviations of less
than 1‰. Examination the buckling modes of tanks under wind-
ward positive pressure reveals that the buckling modes are very
88 Y. Zhao, Y. Lin / Thin-Walled Structures 79 (2014) 83–94

Fig. 9. Buckling modes of tanks under uniform pressure: (a)TK1; (b)TK2; (c)TK3; (d)TK4; (e)TK5; and (f)TK6.

Fig. 10. Wind load and windward positive pressure: (a) wind load (LC1) and (b) windward positive pressure (LC2).

consistent with those of tanks under wind load, as shown in Table 6

Fig. 11. It is thus well-founded to believe that the buckling Comparison of buckling load factor between different load cases.
behavior of tanks under wind load is governed by windward
TK1 TK2 TK3 TK4 TK5 TK6
positive pressure while wind pressure of other region practically
has no substantive influence. For wind engineers, it seems not LC1 0.655 0.900 1.152 1.313 1.424 1.683
feasible to neglect the negative wind pressure of tanks. However, LC2 0.655 0.900 1.152 1.312 1.424 1.682
for structural engineers, it is perfectly acceptable for estimate of Deviation 0 0 0 2.3
10  4 2.8
10  4 5.3
10  4
buckling behavior.

level and the plasticity is not detected, the material nonlinearity is

4. Geometrically nonlinear analyses thus not included in the simulation.

In this section, a series of geometrically nonlinear elastic
analyses on both perfect tanks and imperfect tanks (GNA and
GNIA) are performed to investigate the more realistic buckling
behavior [10] using the Riks method (arc-length solution techni-
que). As is indicated in preliminary analysis, the stress is on low
4.1. Perfect tanks (GNA)
The geometrically nonlinear buckling behavior of perfect tanks
is considered herein. The load–displacement curves for tanks TK1–
TK6 are plotted in Fig. 12, in which the symbol Δ denotes radical
 Y. Zhao, Y. Lin / Thin-Walled Structures 79 (2014) 83–94 89

1.0
1 TK1 LC1
0
-1 LC2
Normalized displacement
TK2 TK1 TK2 TK3 TK4 TK5 TK6
TK3

Z/H
TK4 0.5

TK5
LC1
1 TK6 LC2
0
-1
0.0
-60 -40 -20 0 20 40 60 0 1 0 1
θ/° Normalized displacement
Fig. 11. Comparison of buckling mode between LC1 and LC2: (a) circumferential modes and (b) vertical modes.

2.0
1 TK1
TK6 0

Normalized displacement
1.5 -1 TK2
TK5
TK4 TK3
1.0
λ

TK4
TK3
TK2
TK5
0.5 TK1
1 TK6
0
-1
0.0
0.00 0.03 0.06 0.09 0.12 0.15 -180 -135 -90 -45 0 45 90 135 180
Δ/m θ/°
Fig. 12. Geometrically nonlinear equilibrium paths for perfect tanks. Fig. 13. Geometrically nonlinear buckling deformations at maximum loads for
perfect tanks.

Table 7
Buckling load factors obtained from GNA.
TK1 TK2 TK3
GNA 0.653 0.897 1.146
λcg/λcl,1 99.5% 99.6% 99.4%
displacement with inward positive. It is found that for tanks with
perfect geometries, the buckling behavior is typical branching
instability. The load–displacement curve is almost linear in the
stage of prebuckling. Deflection in this stage is so small that the
effect of change in geometry is not significant. The critical buckling
load factor λcg is very near to that of LBA for each tank, as shown in
Table 7, with a reduction of less than 1%, which reveals that the
deformation of shell has very limited influence on the buckling
resistance of the perfect tank. When the maximum load is got
across, the tank switches to the stage of postbuckling. The load–
displacement curve maintains relatively linearity for very small
deflection at first in this stage. Then as the deflection increases, a
nonlinear feature arises gradually, and a highly nonlinear behavior
can be observed when the deflection continues increasing. After
that the curve stops falling down and goes back to incremental
segment again which reveals that the stiffness is strengthened
with large displacement, leading to the increase of postbuckling
strength.
While examining the buckling deformations at maximum
loads, as shown in Fig. 13, it can be found that the deformations
present some difference from those obtained from the LBA. It is
observed that the bifurcation buckling mode only occurs in the
windward region, not larger than 7601 (see Fig. 11). However, the
TK4 TK5 TK6
deformation on other region at the maximum load obtained from
1.309 1.421 1.678
GNA cannot be neglected in comparison to that in the windward
99.7% 99.8% 99.7% region, especially for relatively slender tanks. It is also found
that deformations obtained from GNA present symmetric feature
although some first classical buckling modes come out to be anti-
symmetrical.
Based on the comparison of results between GNA and LBA, it is
believed that the linear elastic bifurcation analysis is a good
indicator for buckling load and rough deformation at the max-
imum load point although it cannot predict the deformation
exactly. And the lowest eigenvalue load constitutes the upper
bound for geometrically nonlinear buckling load.
4.2. Imperfect tanks (GNIA)
The buckling behavior considered in Section 4.1 is only for
perfect tanks, which could be deemed as the representative of the
ideal performance. Nevertheless, practical tanks are usually fabri-
cated with initial imperfection especially geometric imperfection
which may result in change of wind buckling behavior. There are
several types of geometric imperfections, and two most common
patterns are adopted in present analyses: (a) weld induced
imperfection and (b) classical buckling mode imperfection.
90 Y. Zhao, Y. Lin / Thin-Walled Structures 79 (2014) 83–94

4.2.1. Tanks with weld induced imperfections
The shape of weld induced imperfection occurring during the
construction can be defined by [28]
 
πχ πχ
δ ¼ δ0 e  πχ =γ cos þ ξ sin
γ γ
pffiffiffiffiffi
π Rt
γ¼
½3ð1  ν2 Þ1=4
where δ0 is the amplitude of the depression, χ is the distance
measured from the weld, ξ is the bending parameter varying from
0 to 1, γ is the linear longitudinal bending half wavelength, ν is the
Poisson's ratio, and t is the equivalent wall thickness of tank.
Fig. 14 shows some shapes of weld induced imperfection with
bending parameter ξ ¼ 1, 1/2 and 0. Notice that for ξ ¼1 and 0, the
imperfection shape is defined as type A and type B, respectively,
according to Rotter and Teng [29].
It can be found that it is a longitudinal depression occurring
around the weld joint for this type of imperfection. For cylindrical
shells under axial loads, the buckling mode is mainly longitudinal
waveform deformation, which is similar to the shape of weld
induced imperfection, therefore it is an unfavorable imperfection
type [27,28]. For relatively slender cylindrical shells subject to
wind pressure, the weld induced imperfection may also affect the
buckling behavior to some extent because of their longitudinal
waveform buckling modes [30]. However, for cylinders with
relatively low aspect ratios (H/Do1), it is found that the buckling
mode is mainly a circumferential waveform deformation occurring
on the middle and upper part of cylindrical wall (see Figs. 7 and 8).
In other words, the buckling mode of a stocky cylindrical tank
under wind load does not match the shape of weld induced
imperfection. The weld induced imperfection is thus expected to
have less influence on the wind buckling resistance of a tank with
relatively low aspect ratio.
Fig. 15 shows the comparison of critical buckling load between
GNIA and GNA. The vertical axis is the normalized buckling load
factor based on the GNA buckling load factor λcg. Three tanks TK1,
ð8Þ
TK4 and TK6 with various aspect ratios of 0.27, 0.56 and 0.88 are
chosen for analyses. For imperfection of either type A or type B, it
can be found that the GNIA buckling loads by introducing weld
ð9Þ
induced imperfections (δ0/tmin r1) are very near to GNA results.
Even for the largest imperfection amplitude considered in this
section, the deviation of results between GNIA and GNA is less
than 5%. It reveals that the weld induced imperfection has very
limited influence on the wind buckling resistance of tanks with
relatively low aspect ratios.
4.2.2. Tanks with classical buckling mode imperfections
The geometrically nonlinear buckling behavior of tanks by
introducing classical buckling mode imperfections is stated in the
following. The nonlinear equilibrium paths for different imperfection
amplitudes are shown in Fig. 16(a) for tank TK1. For all imperfection
levels, the equilibrium paths are highly nonlinear. It reveals that the
linear theory is not valid for the stability design of practical tanks. For
relatively small imperfection amplitudes, δ0/tmin ¼0.1–1.0, the equili-
brium path shapes are similar, all of which have stable prebuckling
stages with relatively small deflections and unstable postbuckling
paths with quickly increasing displacements. It is typical primary
buckling behavior. The maximum load point can be observed before
large displacement is reached (Δ/tmin o3). After that, the displace-
ment increases quickly with decreasing load. For δ0/tmin ¼ 2.0, there is
no more declining segment, indicating that the tank is substantially
stiffened and the structure performance is quite different from that of
relatively smaller imperfection amplitude.
Fig. 16(b)–(f) shows the nonlinear equilibrium paths for tanks
TK2–TK6, respectively. It can be found that for each tank, the
equilibrium paths are highly nonlinear, similar to tank TK1. In general,
for slender and thin tanks, the equilibrium paths are relatively gentle.
Fig. 17 shows the sensitivities of classical buckling mode
imperfection for different tanks. It is clear that all tanks are very
sensitive to this type of imperfection. For example, the buckling
resistance has been reduced by 30% for an imperfection level of δ0/
tmin ¼ 0.5 for tank TK1. It can also be seen that the stockier and
thicker the tank is, the more sensitive it is to the classical buckling
mode imperfection in general.

5. Parametric analyses

The influences of the thickness reduction of cylindrical wall, the

liquid stored in the tank and the wind girder on the buckling
Fig. 14. Shapes of weld induced imperfection. behavior are examined and discussed in this section.

TK1 TK4 TK6 TK1 TK4 TK6

1.50 1.50
1.25 1.25
1.00 1.00
λcgi/λcg

λcgi/λcg

0.75 0.75
0.50 0.50
0.25 0.25
0.00 0.00
1/100 1/10 1/4 1/2 3/4 1 1/100 1/10 1/4 1/2 3/4 1
δ0/tmin δ0/tmin
Fig. 15. Comparison of critical buckling load between GNIA and GNA: (a) type A and (b) type B.
 Y. Zhao, Y. Lin / Thin-Walled Structures 79 (2014) 83–94 91

1.0 1.0

0.8 0.8

0.6 0.6
λ/λcg

λ/λcg
0.4 0.10tmin 0.75tmin 0.4 0.10tmin 0.75tmin
0.25tmin 1.00tmin 0.25tmin 1.00tmin
0.2 0.50tmin 2.00tmin 0.2 0.50tmin 2.00tmin

0.0 0.0
0 2 4 6 8 10 0 2 4 6 8 10
Δ/tmin Δ/tmin

1.0 1.0

0.8 0.8

0.6 0.6
λ/λcg

0.10tmin 0.75tmin λ/λcg 0.10tmin 0.75tmin

0.4 0.4
0.25tmin 1.00tmin 0.25tmin 1.00tmin
0.50tmin 2.00tmin 0.50tmin 2.00tmin
0.2 0.2

0.0 0.0
0 2 4 6 8 10 0 2 4 6 8 10
Δ/tmin Δ/tmin

1.0 1.0

0.8 0.8

0.6 0.6
λ/λcg

λ/λcg

0.4 0.10tmin 0.75tmin 0.4 0.10tmin 0.75tmin

0.25tmin 1.00tmin 0.25tmin 1.00tmin
0.2 0.50tmin 2.00tmin 0.2 0.50tmin 2.00tmin

0.0 0.0
0 2 4 6 8 10 0 2 4 6 8 10
Δ/tmin Δ/tmin
Fig. 16. Geometrically nonlinear equilibrium paths for imperfect tanks: (a) TK1; (b) TK2; (c) TK3; (d) TK4; (e) TK5; and (f) TK6.

5.1. Influence of thickness reduction
Corrosion is usually observed during the normal service of tanks,
resulting in reduction of effective thickness of shell, which may
become an unfavorable condition for stability of tanks. To determine
the influence on stability of corrosion, tank TK1 is used for analyses.
The shell may suffer different corrosion during its use, but for
simplified analysis, a uniform thickness reduction, 0.5–4 mm, is
introduced to the data of shell.
The normalized buckling load factors based on the tank with-
out corrosion are summarized in Table 8. The LBA results and GNA
results are very near to each other, indicating that for tanks with
thickness reduction in wall, the change in geometry has little
influence on the buckling resistance. It seems that the thickness
reduction of shell considerably decreases the buckling resistance
in general. For smaller thickness reduction, 0.5 mm, the buckling
load has no significant decrease. However, for a little larger
thickness reduction, 1 mm, the buckling load has been cut down
by larger than 15%, in comparison to that of the tank without
corrosion. While for larger thickness reductions, 2 mm and 3 mm,
the buckling loads decrease by 23% and 47%, respectively, com-
pared with that of the tank without corrosion. And for a thickness
reduction of 4 mm, the buckling resistance is only 41% of that of
the tank without corrosion. It is believed that cylindrical shell
92 Y. Zhao, Y. Lin / Thin-Walled Structures 79 (2014) 83–94

Table 9
Normalized buckling loads for different tanks with various liquid levels.

TK1 TK2 TK3 TK4 TK5 TK6

1/8H 1.000 1.000 1.000 1.000 1.000 1.001

1/4H 1.004 1.002 1.000 1.001 1.002 1.017
1/2H 1.132 1.050 1.012 1.055 1.067 1.313
3/4H 2.215 1.655 1.317 1.424 1.814 2.084
H 20.444 15.380 9.190 9.932 3.995 4.156

Fig. 17. Imperfection sensitivities of different tanks.

Table 8
Normalized buckling load factors.

0.5 mm 1 mm 2 mm 3 mm 4 mm

LBA 0.997 0.833 0.675 0.532 0.406

GNA 0.994 0.830 0.673 0.530 0.406

Fig. 19. Geometrically nonlinear buckling equilibrium paths for imperfect tanks
1.0 with stored liquid: δ0/tmin ¼0.1.

0.8
counteracts external inward wind pressure. Certainly, the storage
of liquid may improve the wind buckling resistance of tank. To
0.6
examine the influence of the stored liquid, several different liquid
λ/λc

0.10tmin 0.75tmin levels are used to carry out the analyses. The density of the stored
0.4 liquid is assumed to be 1000 kg/m3.
0.25tmin 1.00tmin
The normalized buckling loads based on the empty model for
0.50tmin 2.00tmin
0.2 different tanks with various liquid levels are summarized in
Table 9. The LBA and GNA results are very near, and only the
0.0 LBA results are given for discussion due to space limitation. For
0 2 4 6 8 10 lower water levels, 1/8H–1/4H, the buckling resistance is almost
Δ/tmin not raised for any tank. For medium water level, 1/2H, the buckling
loads increase for a certain degree for some tanks. For example,
Fig. 18. Geometrically nonlinear equilibrium paths for tank TK1 with 2 mm thickness
reduction. the buckling loads of tanks TK1 and TK6 increase by 13% and 31%,
respectively. While for the other tanks, the increases are still
limited, smaller than 7% in comparison to empty tanks. For larger
design should consider corrosion allowance from the perspective water level, 3/4H, the wind buckling loads of all tanks have
of structural stability and some measurements should be taken to increased universally. For full-filled water level, the wind buckling
improve the resistance of corrosion. loads of all tanks increase greatly. Even for tank TK5, which
The tank TK1 with 2 mm reduction in shell thickness is used to the smallest growth margin is registered in, the wind buckling
investigate the nonlinear buckling behavior of imperfect tank with resistance reaches up near to 4 times as that of the empty tank. It
thickness reduction. The imperfection introduced to the model is seems that it is effective to impound sufficient water for empty
same to the analyses in Section 4.2.2. The nonlinear equilibrium tanks before a windstorm to prevent tanks from buckling due to
paths for tank TK1 with different imperfection amplitudes are strong wind load.
plotted in Fig. 18. The behavior of tank with smaller imperfection Fig. 19 shows the geometrically nonlinear equilibrium paths for
amplitudes (δ0 r0.5tmin) have same pattern with tank without imperfect tanks with a liquid level of 3/4H. The classical buckling
thickness reduction under same imperfection level. However, for mode imperfection is introduced to each model with an imperfec-
those with larger imperfection amplitudes, δ0 ¼0.75–2.0tmin, the tion amplitude of δ0/tmin ¼0.1. For most tanks, the deflections in
displacement–load curve has no more declining segment, the stage of prebuckling are smaller than those of empty tanks
although the stiffness is reduced and the displacement increases under same imperfection level. It seems that the hydrostatic
fast. This reveals that thinner shells are more likely to be strength- pressure generated by liquid acts like decreasing the cantilever
ened by depressed deformation. height of tank wall.

5.2. Influence of stored liquid 5.3. Influence of wind girder

The liquid stored in tanks will bring in two effects: (a) Tanks may be provided with wind girders as additional stiffen-
the internal suction due to wind vanishes under the liquid level ing rings to maintain roundness when they are subject to wind
and (b) the hydrostatic pressure applied on the internal wall load [31]. Some tanks have only one wind girder and others may
 Y. Zhao, Y. Lin / Thin-Walled Structures 79 (2014) 83–94 93

Table 11
Comparison of buckling load between tanks with and without wind girders.

TK1 TK2 TK3 TK4 TK5 TK6

λwl 0.835 1.182 1.738 1.885 1.904 2.141

λwl/λcl,1 1.275 1.313 1.508 1.436 1.337 1.273
λwg 0.828 1.172 1.711 1.880 1.901 2.139
λwg/λcg 1.269 1.307 1.493 1.436 1.338 1.274

1.0

Fig. 20. Section of primary wind girder. TK1 TK2 TK3 TK4 TK5 TK6

Z/H
Table 10 0.5
Wind girder section parameters.

TK1 TK2 TK3 TK4 TK5 TK6

Z (cm3) 4810 2402 896 444 238
b (mm) 1350 900 600 350 200
Z0 (cm3) 5845 2851 1098 550 281
Z0 is the modulus of used section.
have more or even none, according to the wind buckling resistance
of the tank as well as the wind load it supposed to carry. The
primary wind girder should be installed near the top of top shell
course, usually 1 m away from the top of the shell. Tanks TK1–TK6
are provided with primary wind girders to investigate the influ-
ence of wind girder on the buckling behavior herein. The section of
the primary wind girder installed on the tank is shown in Fig. 20.
In order for comparison, the section of wind girder is designed to
satisfy the minimum modulus suggested by API [30] as follows:
 
D2 H V w 2
Z¼ ð10Þ
17 190
where Z is the required minimum section modulus and Vw is the
design wind speed assumed to be equal to 40.8 m/s.
The sections used for different tanks based on formula (10) are
summarized in Table 10. For each tank, the modulus of wind girder
used for analyses is about 20% larger than the minimum modulus
suggested by API.
Table 11 summarizes the buckling loads of tanks with primary
wind girders obtained from LBA and GNA, compared with those of
tanks without primary wind girders, in which λwl and λwg is the
buckling load factor of tank with primary wind girder obtained
from LBA and GNA, respectively. The GNA results is very close to
LBA results, which indicates that the displacement has little
influence on the buckling load for tank with wind girder. It is
found that the primary wind girder improves the wind buckling
resistance of tank for considerable extent. Even for tank TK1,
which the smallest growth margin is registered in, the wind
buckling resistance increases by larger than 25%, in comparison
to that of tank without wind girder. Although all tanks are
provided with primary wind girders by uniform rule, but the
increases of wind buckling strength are quite different. It seems
that some improvements may be need for design criterion of wind
girder, in order to provide more unambiguous prediction of wind
buckling resistance for tanks with wind girders.
Fig. 21 shows the first classical buckling modes of tanks with
primary wind girders in comparison with those of tanks without
wind girders. It can be clearly seen that the deformation of shell
above the wind girder is reduced considerably and the displace-
ment of shell at the height of wind girder installed is almost zero.
It is also found that the displacement shits downward for tanks
without wind girder
92
with wind girder
100
110 0.0
0 1 0 1
Normalized displacement
Fig. 21. Comparison of buckling mode between tanks with and without wind
girders.
Fig. 22. Geometrically nonlinear buckling equilibrium paths for imperfect tanks
with wind girders: δ0/tmin ¼0.1.
with primary wind girders. It is indicated that the primary wind
girder strengthens the stiffness of the upper part of shell, leading
to improvement of wind buckling resistance.
Fig. 22 shows the nonlinear equilibrium paths of imperfect
tanks with wind girders. For all tanks, the load–displacement
curves are highly nonlinear. It is confirm again that the buckling
behavior of tanks exhibits pronounced nonlinearity. The deflec-
tions of tanks with wind girders in the stage of prebuckling are
smaller than those of tanks without wind girders under same
imperfection level. For example, the displacement at the max-
imum load point is 1.5tmin for tank TK1 without wind girder
(see Fig. 16(a)) but decrease to 1tmin for the same tank without
wind girder, under an imperfection level of δ0/tmin ¼0.1. Further-
more, the postbuckling strength of tanks with wind girders
increases universally, or in other words, under same deformation
in the stage of postbuckling, tanks with wind girders have larger
load carrying capacity, compared with those of tanks without
wind girders.
In general, it is confirmed that the wind girder improves the
buckling strength and stiffness of tanks subject to wind load.
94 Y. Zhao, Y. Lin / Thin-Walled Structures 79 (2014) 83–94
6. Conclusions
This paper investigates the buckling behavior of cylindrical
open-topped steel tanks under wind load by a great deal of
numerical simulations. The main conclusions from this study can
be summarized as follows:
(1) The stability carrying capacity of wind load decreases as the
aspect ratio decreases for practical tanks. Therefore for larger
tank with a lower aspect ratio, it is expected to be more
susceptible to buckling during the wind storm.
(2) The buckling resistance of tanks under wind load is larger than
that of thanks under uniform pressure, with deviation of about
25–50%. The critical uniform pressure of buckling based on the
Donnell theory can be used for a preliminary evaluation of the
wind buckling critical load.
(3) The buckling behavior of tanks under wind load is governed by
windward positive pressure while wind pressure of other
region practically has no influence.
(4) The weld induced imperfection has no significant influence on
the wind buckling of cylindrical tanks with relatively low
aspect ratios, while the classical buckling mode imperfection
considerably decreases the nonlinear buckling resistance of
cylindrical tanks. The stockier and thicker the tank is, the more
sensitive it is to the classical buckling mode imperfection in
general.
(5) The thickness reduction of shell decreases the wind buckling
resistance of tank significantly. Design of cylindrical shells
should consider corrosion allowance and some measurements
should be taken to improve the resistance of corrosion.
(6) The stored liquid plays an important role in the wind buckling
resistance of tank. It is effective to impound sufficient water
for empty tanks before a windstorm to prevent tanks from
buckling due to strong wind load.
(7) The wind girder improves the buckling strength and stiffness
of tank wall, leading to a considerable increase of wind
buckling resistance and a pronounced decrease of buckling
deformation especially above the wind girder.
Acknowledgments
The authors gratefully acknowledge the support of the National
Natural Science Foundation of China (No. 51378459) and the Key
Science and Technology Innovation Team Program of Zhejiang
Province, China (No. 2010R50034). The authors also thank Mr. Lu
and Ms. Jin for providing some valuable data of tanks in practical
engineering.
References
[1] BS EN 14015-2004. Specification for the design and manufacture of site built,
vertical, cylindrical, flat-bottomed, above ground, welded, steel tanks for the
storage of liquids at ambient temperature and above. British Standard 14015;
2004.
[2] Pavlovic P, Folic R, Radonjanim V, Tatomirovic M. The testing and repair of
steel silo. Constr Build Mater 1997;11(5):353–63.
[3] Flores FG, Godoy LA. Buckling of short tanks due to hurricanes. Eng Struct
1998;20(8):752–60.
[4] Godoy LA. Performance of storage tanks in oil facilities following Hurricanes
Katrina and Rita. J Perform Constr Facil, ASCE 2007;21(6):441–9.
[5] Jaca RC, Godoy LA. Wind buckling of metal tanks during their construction.
Thin-Walled Struct 2010;48(6):453–9.
[6] Kundurpi PS, Savamedam G, Johns DJ. Stability of cantilever shells under wind
loads. J Eng Mech Div, ASCE 1975;101(5):517–30.
[7] Uematsu Y, Uchiyama K. Deflection and buckling behavior of thin, circular
cylindrical shells under wind loads. J Wind Eng Ind Aerodyn 1985;18(3):245–61.
[8] Greiner R, Derler P. Effect of imperfections on wind-loaded cylindrical shells.
Thin-Walled Struct 1995;23:271–81.
[9] Schmidt H, Binder B, Lange H. Postbuckling strength design of open thin-walled
cylindrical tanks under wind load. Thin-Walled Struct 1998;31(1–3):203–20.
[10] Portela G, Godoy LA. Wind pressures and buckling of cylindrical steel tanks
with a dome roof. J Constr Steel Res 2005;61(6):808–24.
[11] Portela G, Godoy LA. Wind pressures and buckling of cylindrical steel tanks
with a conical roof. J Constr Steel Res 2005;61(6):786–807.
[12] Jaca RC, Godoy LA, Flores FG, Croll JGA. A reduced stiffness approach for the
buckling of open cylindrical tanks under wind loads. Thin-Walled Struct
2007;45(9):727–36.
[13] Sosa EM, Godoy LA. Challenges in the computation of lower-bound buckling loads
for tanks under wind pressures. Thin-Walled Struct 2010;48(12):935–45.
[14] Jaca RC, Godoy LA, Croll JGA. Reduced stiffness buckling analysis of aboveground
storage tanks with thickness changes. Adv Struct Eng 2011;14(3):475–87.
[15] Borgersen S, Yazdani S. Finite element analysis of wind induced buckling of
steel tank. In: Ghafoori N, editor. Challenges, opportunities and solutions in
structural engineering and construction. London: Taylor & Francis Group;
2010.
[16] Chen L, Rotter JM. Buckling of anchored cylindrical shells of uniform thickness
under wind load. Eng Struct 2012;41:199–208.
[17] Zhao Y, Cao QS, Su L. Buckling design of large circular steel silos subject to
wind pressure. Thin-Walled Struct 2013;73:337–49.
[18] Eurocode 3. Design of steel structures—Part 1-6, strength and stability of shell
structures. European Standard EN 1993-1-6; 2007.
[19] Noreta E, Prod’homme G, Yalamas T, Reimeringer M, Hanus J-L, Duongb D-H.
Safety of atmospheric storage tanks during accidental explosions. Eur
J Environ Civ Eng 2012;16(9):998–1022.
[20] ABAQUS Element Reference. ABAQUS Inc.; 2010.
[21] Eurocode 1. Actions on structures—Part 1-4: general actions—wind actions.
European Standard EN 1991-1-4; 2005.
[22] Rish RF. Forces in cylindrical chimneys due to wind. Proc Inst Civ Eng, ASCE
1967;36:791–803.
[23] ACI-ASCE Committee 334. Reinforced concrete cooling tower shells-practice
and commentary. ACI 334-2R; 1991.
[24] Eurocode 3. Design of Steel Structures—Part 4-1: silos. European Standard EN
1993-4-1; 2007.
[25] Portela G. Wind pressure and buckling of metal cantilever tanks [Ph.D. thesis].
University of Puerto Rico, Mayagüez Campus, Puerto Rico, America; 2004.
[26] GB 5009-2012. Load code for design of building structure. Beijing: China
Architecture & Building Press; 2012 [in Chinese].
[27] BS 2654-1989. Specification for manufacture of vertical steel welded storage
tanks with butt-welded shells for the petroleum industry. British Standard
2654; 1989.
[28] Pircher M, Berry PA, Ding X, Bridge RQ. The shape of circumferential weld-
induced imperfections in thin-walled steel silos and tanks. Thin-Walled Struct
2001;39(12):999–1014.
[29] Rotter JM, Teng JG. Elastic stability of cylindrical shells with weld depressions.
J Struct Eng 1989;115(5):1244–63.
[30] Pircher M. The influence of a weld-induced axi-symmetric imperfection on
the buckling of a medium-length silo under wind loading. Int J Solids Struct
2004;41(20):5595–610.
[31] API 650. Welded tanks for oil storage. American Petroleum Institute; 2012.