See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/276100579

Numerical Evaluation on Shell Buckling of Empty Thin-Walled Steel Tanks

Under Wind Load According to Current American and European Design Codes

Article  in  Thin-Walled Structures · July 2015

DOI: 10.1016/j.tws.2015.07.007

CITATIONS READS

28 2,345

3 authors:

Chrysanthos Maraveas Georgios Balokas

Agricultural University of Athens Technische Universität Hamburg
99 PUBLICATIONS   779 CITATIONS    23 PUBLICATIONS   155 CITATIONS   

SEE PROFILE SEE PROFILE

Konstantinos Daniel Tsavdaridis

City, University of London
229 PUBLICATIONS   1,976 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Post-fire seismic response of steel structures View project

Composite Flooring Systems View project

All content following this page was uploaded by Konstantinos Daniel Tsavdaridis on 10 January 2022.

The user has requested enhancement of the downloaded file.

 Thin-Walled Structures 95 (2015) 152–160

Contents lists available at ScienceDirect

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

Numerical evaluation on shell buckling of empty thin-walled steel

tanks under wind load according to current American and European
design codes
Chrysanthos Maraveas a,b, Georgios A. Balokas b,c, Konstantinos Daniel Tsavdaridis d
a
School of Mechanical, Aerospace and Civil Engineering, University of Manchester, UK
b
C.Maraveas Partnership – Consulting Engineers, Athens, Greece
c
Department of Mechanics, National Technical University of Athens, Greece
d
Institute for Resilient Infrastructure, School of Civil Engineering, University of Leeds, UK

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

Article history: Liquid storage steel tanks are vertical above-ground cylindrical shells and as typical thin-walled struc-
Received 14 May 2015 tures, they are very sensitive to buckling under wind loads, especially when they are empty or at low
Received in revised form liquid level. Previous studies revealed discrepancies in buckling resistance of empty tanks between the
4 July 2015
design method proposed by the American Standard API 650 and the analytical formulas recommended
Accepted 6 July 2015
by the European Standard EN1993-1-6 and EN1993-4-2. This study presents a comparison between the
provisions of current design codes by performing all types of numerical buckling analyses recommended
Keywords: by Eurocodes (i.e. LBA-linear elastic bifurcation analysis, GNA-geometrically nonlinear elastic analysis of
Steel tanks the perfect tank and GNIA-geometrically nonlinear elastic analysis of the imperfect tank). Such analyses
Shell buckling
are performed in order to evaluate the buckling resistance of two existing thin-walled steel tanks, with
Finite element modeling
large diameters and variable wall thickness. In addition, a discussion is unfolded about the differences
Nonlinear analysis
Wind load between computational and analytical methods and the conservatism that the latter method imposes. An
influence study on the geometric imperfections and the boundary conditions is also conducted. In-
vestigation on the boundary conditions at the foot of the tank highlights the sensitivity to the fixation of
the vertical translational degree of freedom. Further, it is indicated that the imperfection magnitude
recommended by the EN1993-1-6 is extremely unfavorable when applied to large diameter tanks.
Comments and conclusions achieved could be helpful in order to evaluate the safety of the current
design codes and shed more light towards the most accurate one.
& 2015 Elsevier Ltd. All rights reserved.

1. Introduction parameter [2,3]. Different tank variations have been investigated,

like open-topped [4] and fixed-roof [5–7], combining computa-
Above-ground, vertical tanks of cylindrical shape are con- tional methods and experimental results. Jaca and Godoy [8] in-
structed in industrial and agricultural plants to store various fluids dicated that buckling of tanks sometimes can occur under mod-
such as petroleum, oil, fuel etc. They are welded, thin-walled erate wind load during their construction. Another subject of in-
structures with large diameters, and hence buckling may occur terest is the wind buckling behavior of grouped, arranged tanks
when they are subjected to wind loads at their empty or partially [9–11]. The simulation of wind load distribution acting on the tank
filled state. Failure of such tanks results, in most cases, in a tre- shell is an open research field [12–14]. Innovative ways of
mendous loss of financial and human resources, as well as com- strengthening and improving buckling capacity have been pro-
poses a threat to public safety and an environmental hazard. posed [15]. Sosa and Godoy [16] and Burgos et al. [17] have re-
Studies concerning wind-induced buckling of steel tanks have cently taken a turn towards analytical methods, in order to im-
been increasing over the past few decades, since structural stabi- prove buckling evaluation by proposing new methodologies.
lity becomes critical for response and a major concern for the This study aims to appraise the efficiency of current design
designer. specifications in addressing structural stability of empty, large
Early studies approached this matter based on analytical for- tanks when subjected to wind actions. Most recent codes
mulations of energy theory and tried to verify results with ex- (EN1993-1-6 and EN1993-4-2) have not yet seen many field ap-
periments [1]. Following, numerical approaches have been con- plications and their results may raise doubts. This paper offers a
ducted extensively, inserting the imperfection sensitivity comparison between API 650 and the Eurocodes, by performing

http://dx.doi.org/10.1016/j.tws.2015.07.007
0263-8231/& 2015 Elsevier Ltd. All rights reserved.
 C. Maraveas et al. / Thin-Walled Structures 95 (2015) 152–160 153

three types of buckling analysis recommended by the EN1993-1-6

for numerical investigation and relating the results with previous
studies [18] conducted with analytical methods (closed-form, ex-
plicit expressions) proposed by the aforementioned codes. Thus,
the stability of two existing large-diameter, steel tanks at empty
state is evaluated.
The study is organized as follows: Section 2 describes the de-
sign philosophy of API 650 [19] and EN1993-1-6. In Section 3 the
geometry of the two existing tanks is presented in detail and
Section 4, presents the finite element models used for analyses
and the wind pressures simulated for each code. Section 5 de-
scribes the linear bifurcation analysis (LBA) and in Sections 6 and 7
geometrically nonlinear buckling behavior is investigated for
perfect (GNA) and imperfect (GNIA) models respectively. In Sec-
tion 8 comparison results are discussed and finally in Section 9,
some helpful conclusions are reached.

2. Description of current code provisions

The most commonly used standards for assessing the structural

stability of thin-walled structures are the API 650 and EN1993-1-6.
The American Standard API 650 provides two empirical methods
(the one-foot method and the variable design point method) for
selecting the thickness of each shell course, depending on the
geometry of the tank, the operational liquid level, the material
used, the density of the contained fluid and the allowance for
corrosion. The aforementioned methods are based on the concept
of limiting the tensile stresses of the shell due to hydrostatic
pressure while they do not consider for buckling. The buckling
limit state is considered only indirectly, via an empirical design
method that mandates stiffening of the shell (with circumferential
girders at specific heights) depending on the thickness, height and
wind velocity. The lack of mathematical formulation for evaluating
the shell stability poses a major disadvantage.
On the contrary, the European standard EN1993-1-6 [20] con-
tains the theoretical background and provides the state-of-the-art
methodologies for evaluating explicitly the buckling resistance of
shell structures. Provisions include analytical expressions for cal- Fig. 1. On-site pictures of tank T-776 (i) and tank T-761 (ii).
culating the buckling capacity in terms of stresses and also pro-
pose several numerical methods, like linear bifurcation analysis for
Table 1
obtaining the critical elastic buckling load as well as analyses that
Geometrical characteristics of tanks T-776 and T-761.
include geometrical and material nonlinearities and imperfections.
Even though its provisions are limited to axisymmetric geome- Tank ID Shell Roof Inside Dia- 1st Wind 2nd Wind
tries, the European Standard has a wide range of applications with Height Height meter (m) Girder Girder
(m) (m) Height* (m) Height* (m)
regard to cylindrical tanks. It is of paramount importance that the
code quantifies the buckling resistance in terms of critical stresses T-776 20.032 3.911 46.939 14.860 -
or critical loads. An analytical procedure for evaluating the buck- T-761 19.500 - 88.430 15.350 18.400
ling resistance of shells with variable wall thickness has also been
*
developed. Most of the approaches recommended by the Eur- Wind girder height is measured from the bottom of the tank.

opean Standard require the use of computational methods, such as

the finite element method, for analyzing the shell. The use of T-776 supports a conical roof with a slope equal to 1/6, while the
simplified expressions, according to basic principles of mechanics, other tank is open-topped. The conical roof is supported by a truss
for determining the design stresses is permitted only in certain structure with three groups of sections (L125x75
8, HEM1000
cases. However, it should be highlighted that the European Stan- and SHS_80x80
8). The geometrical data of both tanks, including
dard is still very recent, and its applicability to the field con- distinct locations of the ring stiffeners (wind girders) along the
struction has not been adequately confirmed up to date. circumference, are presented in Table 1. It can be seen that the
aspect ratio of tanks (H/D) is quite low (0.43 for T-776 and 0.22 for
T-761).
3. Geometry of the tanks Both tanks have variable wall thicknesses and their cylindrical
shell is divided in nine courses. The width and thickness of each
The two existing, thin-walled and large diameter steel tanks shell course along with relevant information regarding the bottom
under investigation (T-776 and T-761) are shown in Fig. 1. They are and roof (where applicable) are summarized in Table 2. The choice
located at the refinery of Motor Oil Hellas S.A. (Korinthos, Greece). for the particular representative case studies is based on the
Both tanks are cylindrical, self-supported (not anchored to the variability of the geometric characteristics (aspect ratio, stiffeners,
foundation), with flat bottoms and are considered empty. Tank roof tops etc.) covering different structural behaviours observed in
154 C. Maraveas et al. / Thin-Walled Structures 95 (2015) 152–160

Table 2
Shell courses details for tanks T-776 and T-761.

Course No.* Course Thickness (mm) Course Width (mm)

T-776 T-761 T-776 T-761

1 22.25 38.60 2438 2222

2 18.93 37.18 2438 2222
3 16.24 28.20 2438 2222
4 13.57 24.59 2438 2222
5 10.9 19.96 2438 2222
6 8.22 15.60 1940 2222
7 8.00 11.20 1940 2222
8 8.00 9.50 1940 2222
9-top 8.00 9.50 1940 1724
Bottom Shells 6.40 6.40 2102 Variable
Roof Shells 5.00 - 1502 -

*
Courses are numbered from bottom to top (i.e. No1 refers to the bottom shell
course etc.).

a considerable range of applications for storage tanks found in

practice.

4. Computational models

A separate 3D finite element (FE) model was created for each

tank. The commercial FE package ABAQUS [21] was used to si-
mulate the tanks with geometric and material properties similar to
the existing structures and to perform the required analyses. The
S8R5 element type was used for the cylindrical shell and the
bottom of each tank (but also for the roof shell for tank T-776). It is
a rectangular, doubly curved, thin, continuum shell element with
reduced integration and 8 nodes. Each node has 5 degrees of
freedom: translations in each spatial coordinate and two rotations
with respect to the in-plane axis. Such characteristics satisfy the
modeling requirements of EN1993-1-6 [20]. The remaining struc-
tural parts (wind girders, top curb angles, roof trusses etc.) were
simulated using beam elements. Element B31 was used, which is a
3D Timoshenko beam element with linear interpolation, which Fig. 2. Simulation of T-776 (i) and T-761 (ii) with the FEM.
allows for transverse shear deformation and is suitable for thick as
well as slender beams. Discretization was selected to account for
shell thickness gradual change and the location of stiffeners. Re- qw = c p q¯ (1)
garding the boundary conditions at the bottom, the most common
assumption in similar studies is the fully fixed state. In this study Where cp is the wind pressure coefficient and q̄ is the pressure
tanks are simply supported (unanchored) so linear elastic, com- value at a specific height, on the incidence of the wind (wind-
pression-only, translational springs were used to model the ward). Wind pressure coefficient is specified using Fourier series
foundation of the tanks at the vertical (meridional) direction, in decomposition:
order to allow the wind-induced uplift of the tank. The constants m
for the support springs were determined from the soil factor. A c p (θ ) = ∑ ai cos (iθ )
direct comparison is made between the two assumptions for the i=0 (2)
boundary conditions (fully fixed and compression-only vertical
Where θ is the angle measured from the windward direction
springs) for all analyses performed. In all computational models,
(θ ¼ 0° for windward and θ ¼ 180° for leeward) and ai is the Fourier
the material was modeled as elastic and isotropic for all structural coefficients. Several proposals have been made for these coeffi-
members, with the modulus of elasticity equal to E¼ 2.1
105 cients [13]. EN1991-1-4 [22] includes formulations for calculating
N/mm2 and Poisson’s ratio equal to ν ¼0.3. The FE models of both specific values of the distribution based on various parameters, but
tanks are presented in Fig. 2. does not provide the Fourier coefficients used for obtaining these
The wind load is simulated as pressure distribution acting on formulations. This study followed the proposal of Greiner [23] as it
the circumferential shell. According to current code provisions for seems to more accurately approximate the shape of the EN1991-1-
cylindrical shell structures, this pressure varies along both height 4 pressure distribution, using the expression:
and circumference of the shell. The height variation is not sig-
nificant for tanks [20], hence the pressure is assumed to be con- c p (θ ) = − 0.55 + 0.25 cos (θ ) + 1 cos (2θ ) + 0.45 cos (3θ )
stant along the height, as opposed to silos. It has been experi- + 0.15 cos (4θ ) (3)
mentally observed how cosine families can represent circumfer-
ential pressures on shells, so most of the formulations established EN1993-4-2 [24] allows wind simulation through an equivalent
to define circumferential patterns of pressure employ Fourier co- uniform pressure throughout the circumference of the tank, when
sine series. Wind pressure can generally be defined as: several requirements are fulfilled. API 650 assumes uniform wind
 C. Maraveas et al. / Thin-Walled Structures 95 (2015) 152–160 155

critical load than uniform pressure for T-761. As Zhao and Lin [4]
recommended, it might be acceptable for the structural designer
to neglect the negative wind pressures, as the windward positive
pressure govern the buckling capacity. At final, buckling initiated
at the thinner shell courses for both tanks. The buckling mode of
tank T-761 (Fig. 5) is located below both wind girders, as they are
set at a small distance and offer great stiffness to the upper thin
shell courses. It is also noted that fully fixed boundary conditions
(only concerns EN19993-1-6 wind pressure) might overestimate
buckling capacity for T-761.

6. Geometrically nonlinear analysis (GNA)

Fig. 3. Distribution of wind pressure on cylindrical tanks (left) and equivalent In order to obtain a more accurate buckling behavior, EN1993-
uniform pressure (right). 1-6 proposes a geometrically nonlinear analysis. This is based on
the principles of shell bending theory applied to the perfect
loading on any occasion. This assumption is investigated numeri- structure, including the nonlinear large deflection theory for the
cally and the results for both pressure distribution proposals are displacements which accounts for any change in geometry due to
compared. Fig. 3 demonstrates a representative schematic dis- the actions on the shell. The nonlinear analysis satisfies both
tribution of wind pressure based on the above formulations, as equilibrium and compatibility of the deflections under conditions
well as the uniform wind pressure. in which the change in the geometry of the structure caused by
loading is included. The resulting field of stresses matches the
definition of primary plus secondary stresses. This study did not
5. Linear bifurcation analysis (LBA) account for material nonlinearity, since stresses acting on the shell
of such thin walled structures are low while the buckling occurs
EN1993-1-6 recommends linear elastic bifurcation (eigenvalue) before the material yielding. ABAQUS software uses the incre-
analysis as a method which evaluates the linear bifurcation ei- mental and iterative Newton-Raphson method to obtain solutions
genvalue for a thin-walled shell structure on the basis of the small for nonlinear problems, by applying the specified loads gradually
deflection linear elastic shell bending theory, related to the perfect and incrementally working towards the final solution. Riks algo-
geometry of the middle surface of the shell. It is a linear pertur- rithm was chosen for the different variations of Newton-Raphson
bation procedure that can be the first step in a buckling analysis, method, which is an arc-length technique that can provide solu-
providing a preliminary evaluation of buckling behavior. It obtains tions even in cases of complex, unstable responses of the structure
the lowest eigenvalue at which the shell may buckle into a dif- investigated [21].
ferent deformation mode, assuming however no change of geo- The results of the geometrically nonlinear analysis are pre-
metry, no change in the direction of action of the loads and no sented herein. The load-displacement curves (equilibrium paths)
material degradation. Imperfections of all kind are ignored, but were computed by selecting a node on the windward meridian of
buckling mode results can be introduced as an initial geometric the cylindrical shell and in the wind direction, representative of
imperfection in the non-linear analysis. the maximum displacements experienced by the tank. This node
The bifurcation buckling analysis for both tanks under wind was used as the degree of freedom to plot load-displacement
load is presented herein. Wind pressure distribution, as proposed curves that are presented in Fig. 6 for tanks T-776 and T-761. Both
by EN1993-1-6, is applied and uniform pressure adopted by API results for fixed boundary conditions under EN1993-1-6 wind
650 is investigated for comparison. Results concerning dis- pressure and for uniform wind pressure adopted by API650 are
crepancies between fully fixed boundary conditions and com- displayed for comparison. The critical buckling loads obtained
pression-only translational elastic springs are also provided. The from GNA were slightly lower than those from LBA, as depicted in
buckling capacity is calculated through buckling load factors (λ) Table 4. It is observed that prebuckling equilibrium paths are lin-
that multiply the reference wind pressure such that λc is a critical ear, while buckling occurs suddenly for T-761 (across the max-
load of the tank, for a wind profile that is assumed constant during imum load), as opposed to T-776 that because of the roof shell and
the load process. Reference pressures for both EN1993-1-6 and the wind girder, buckling occasionally develops a slightly unstable
uniform distributions were set up to 1 kPa so that λc would di- response near the critical point. The wind pressure distribution
rectly represent the critical load. Results for critical buckling wind imposed by Eurocode leads to larger displacements. It should be
loads are summarized in Table 3. Buckling deformation modes highlighted that LBA should always precede nonlinear analysis to
corresponding to the critical load factors are presented in Fig. 4 ensure none bifurcation point failed to be detected numerically in
and Fig. 5, for tank T-776 and tank T-761 respectively. It is ob- the equilibrium path [20]. The designer should not rely on post
served that buckling occurs at the windward region when EN1993- buckling behavior, as thin-walled structures do not develop any
1-6 wind pressure distribution is applied, with a slightly greater post buckling resistance.

Table 3
Critical buckling load factors*obtained from linear bifurcation (eigenvalue) analysis.
7. Geometrically nonlinear analysis with imperfections (GNIA)
Wind pres- T-776 T-761
sure type Compression-only Fully fixed Compression-only Fully fixed Previously considered analyses describe the ideal buckling be-
springs springs havior of tanks T-776 and T-761, starting with a “perfect” FE model.
However tanks and structures in practice, contain features that
EN1993-1-6 λc ¼ 5.5456 λc ¼5.9928 λc ¼9.4464 λc ¼ 10.339
uniform λc ¼ 5.6518 λc ¼5.6518 λc ¼8.5665 λc ¼ 8.5665
cause changes in the geometry, such as minor deviations in the
shape, eccentricities, local indentations etc., which can be error-
*
All buckling load factors coincide with critical loads measured in kPa. induced or damage-induced. These features known as
156 C. Maraveas et al. / Thin-Walled Structures 95 (2015) 152–160

Fig. 4. Buckling deformation modes for tank T-776; (a) Wind pressure distribution according to EN1993-1-6; (b) Uniform wind pressure distribution.

imperfections can alter the buckling behavior radically and should dependent on the fabrication tolerance quality class. The max-
always be considered in the design process. imum deviation of the geometry of the equivalent imperfection
In order to estimate the magnitude of geometrical imperfec- from the perfect shape should be given by the above expression:
tions, EN1993-1-6 [20] defines three fabrication tolerance quality
classes (Class A: excellent, Class B: high and Class C: normal fab- {
Δw0, eq = max Δw0, eq,1; Δw0, eq,2 } (4)
rication). Class selection is based on representative sample mea-
surements conducted on the unloaded and completed structure. A Where:
clear distinction, based on the imperfection type being considered, Δw0, eq,1 = l g Un1 (5)
is made from fabrication quality tolerance measurements. More
specifically they are categorized as: a) out-of-roundness mea-
Δw0, eq,2 = ni tUn2 (6)
surements, which are associated with the internal diameter of the
shell, b) non-intended eccentricity measurements at the joints of Where lg is the corresponding gauge length, t is the local shell wall
the connected plates, c) dimple measurements, in the meridional thickness, ni is a multiplier to achieve an appropriate tolerance
direction and along the circumference of the shell, including level (recommended value: 25) and Un1, Un2 are the dimple im-
measurements across the welds and d) flatness measurements at perfection amplitude parameters for the corresponding fabrication
the interface of the shell and its bottom. The fabrication quality tolerance quality class. Recommended values for Un1 and Un2 are
class is assessed separately for each measurement type, according given in Table 5. The gauge length parameter included in Eq. (5)
to the tolerances specified in EN1993-1-6 [20]. The lowest quality should be taken for both the meridional and the circumferential
class is then assigned to the shell structure. directions as the maximum of the following:
It is clear that when the imperfection cannot be measured or
otherwise be made known (e.g. the structure has not been build l g x = 4 rt (7)
and the contractor does not have accurate tolerance data), the
imperfect shape with the most unfavorable effect should be as- 0.5

sumed and applied to the perfect model geometry. The eigen-

( )
l gθ = 2.3 l2rt , butl gθ ≤ r (8)
mode-affine pattern should be used unless a different unfavorable
l g w = 25t , butl g w ≤ 500mm (9)
pattern can be justified [20]. The amplitude of the adopted
equivalent geometric imperfection form should be taken as Where r is the radius of the middle surface normal to the axis of
 C. Maraveas et al. / Thin-Walled Structures 95 (2015) 152–160 157

Fig. 5. Buckling deformation modes for tank T-761; (a) Wind pressure distribution according to EN1993-1-6; (b) Uniform wind pressure distribution.

revolution, t is the shell thickness (in case of variable wall thick- taken as d/tmin ¼ 3 for practical tanks [4]. It is highlighted that the
ness the minimum is selected) and l is the meridional length of the decrease in the buckling capacity due to imperfections can reach
shell segment. up to -60%, as demonstrated in Fig. 7.
Imperfection measurements have not been carried out for
tanks T-776 and T-761 up to day, thus relevant results for all fab-
rication quality classes are presented. Fig. 7 displays load-dis- 8. Comparison between the current design Standards
placement curves for both tanks and for both wind pressures. Ei-
genmode results obtained from LBA were used as the imperfection An attempt to compare and investigate discrepancies between
shapes. Imperfection amplitudes for each quality class were cal- the shell buckling evaluation methods offered by current code
culated by implementing geometrical characteristics of the tanks provisions is made. The results obtained numerically by linear
(Tables 1 and 2) in the formulas presented in this section. The (LBA) and nonlinear (GNA and GNIA) analyses are contrasted to
results of the amplitudes Δw0,eq are summarized in Table 5. It is the results emerged from the analytical formulations included in
safe to deduce that imperfections cause an impressive decrease in standards, and presented in a previous study [18]. Comparison is
the critical buckling load and essentially change the shape of the performed through von Mises equivalent stress, and constitutes a
equilibrium path, providing a highly nonlinear behavior. It is ob- representation of the resulting two-dimensional field of primary
served that as the imperfection amplitude increases, the curves stresses, thus accounts for in-plane stress interaction on the cy-
become smoother and it is unfeasible to derive a critical load va- lindrical shell of the tank [20]. The well-known mathematical
lue, while there is no more a maximum point. Particularly for tank expression for von Mises criterion is as follows:
T-776 (Fig. 7a), the equilibrium paths for all quality classes indicate
σeq = σ x2 − σ x σθ + σθ2 + 3τx2θ (10)
that the tank is actually stiffened. This is a result of the large values
of imperfection amplitudes Δw0,eq calculated by the EN1993-1-6 Where sx, sθ and τxθ represent the meridional, circumferential and
formulas, for the large diameter tanks of this study in-plane shear stresses respectively. A main feature of this rule’s
(Δw0,eq normalized by thickness reaches 6.8 t). Critical load esti- behavior of metal components in a multi-axial stress state is that
mations for equilibrium paths with such rapid increase of dis- the stress components with the same sign (e.g. biaxial compres-
placements are made through the largest tolerable deformation sion) support each other while stresses with different signs or
according to EN1993-1-6 [20], which occurs when the maximum additional shear stress decrease the capacity in each direction.
local rotation of the shell surface attains the value of 0.1 rad. An- As it was aforementioned, API 650 does not quantify buckling
other criterion is that the largest tolerable deformation can be resistance by means of analytical expressions and it is not possible
158 C. Maraveas et al. / Thin-Walled Structures 95 (2015) 152–160

Table 5
Recommended values for parameters Un1 and Un2 and imperfection amplitudes
Δw0,eq.

Fabrication Description Recommended Recommended Δw0,eq Δw0,eq

tolerance value of Un1 value of Un2 for for
quality class T-776 T-761
(mm) (mm)

Class A Excellent 0.010 0.010 21.09 25.92

Class B High 0.016 0.016 33.74 41.47
Class C Normal 0.025 0.025 52.73 64.80

work deemed necessary in order to decide the methods to limit

the structure’s deformation.
The results in Table 6 exhibit satisfactory convergence between
the stress design analytical method and the nonlinear numerical
method for the perfect model under wind pressure distribution
proposed by EN1993-1-6. Consequently, the proposed approach
for variable thickness tanks in Annex D of the code is considered
safe. Nevertheless, the safety factors implemented by the stress
design method regarding imperfections found to be very con-
servative as opposed to the numerical results for the buckling
capacity. It should also be highlighted that models under uniform
wind pressure may not display significant discrepancies in terms
of critical loads (Fig. 7) in comparison with actual wind pressure at
the circumference of the shell proposed by EN1993-1-6; however
they present a different behavior regarding displacements and
unfavourable results regarding von Mises stresses.
Fig. 6. GNA load-displacement curves; (a) T-776; (b) T-761. Regarding API 650, due to the different problem approach a
previous study [18] showed that both tanks satisfy its require-
to evaluate the safety level provided through the equivalent von ments optimally; hence this suggests the use of an earlier version
Mises stresses. Methods proposed are well-suited for design pur- of the code for their design. This study revealed that although the
poses due to their simplicity; however there is a lack of a specific unfavourable uniform wind pressure is adopted while higher de-
procedure for determining the critical stress-state pertaining to sign wind velocity is used [19], the entire lack of imperfections in
buckling of the tank shell. Oppositely, EN1993-1-6 provides the the process is dominating.
“stress design” method that includes sophisticated mathematical
formulas for calculating critical meridional, circumferential and
9. Concluding remarks
shear stresses on every shell course of the tank and eventually
attaining the design stresses by implementing limiting safety
This study focuses on shell buckling evaluation of two empty
factors. The stresses of the perfect models, as well as the imperfect
and large diameter storage tanks under wind load as demon-
ones for every fabrication quality class, obtained from the study of
strated by current design code provisions used in a great number
Maraveas and Miamis [18] were implemented in Eq. (10) for every
of world-wide applications. The main conclusions of this research
shell course. Results are displayed for comparison with the ana- may be summarized as follows:
logous equivalent stress obtained from numerical analysis at the
shell segment that buckled (Table 6).  Linear bifurcation analysis (LBA) is a good indicator for buckling
It is noted that von Mises stresses for the imperfect model of capacity as the critical loads obtained converge with those of
tank T-776 are not defined as the critical loads cannot be identified geometrically nonlinear analysis. It should always precede a
by the nonlinear force-displacement curves. Large imperfection geometrically nonlinear analysis as it offers the bifurcation
amplitudes provide a progressively stiffening response to the shell points and buckling modes that can be used as imperfection
and large displacements (Fig. 7a). The largest tolerable deforma- shapes.
tion criterion provided by EN1993-1-6 is rather arbitrary for T-776,  Fully fixed boundary conditions may overestimate the buckling
since the stresses corresponding to 0.1 rad of local rotation are capacity of tanks. Unanchored structures’ foundation could be
extremely high while the shell has already buckled. Consideration simulated with linear compression-only springs in the vertical
of maximum tolerable displacement as d ¼3tmin is more realistic, direction and thus display a more sensible behavior, allowing
but still specifies very high stresses. Instructions for the corre- the tank to partly elevate due to horizontal wind load.
sponding code for such cases are not satisfactory, hence further  Uniform wind pressure distribution is evaluated as the more

Table 4
Critical buckling load factors*obtained from GNA and comparison with LBA.

Analysis type T-776 T-761

EN1993-1-6 EN1993-1-6 (fully fixed) Uniform pressure EN1993-1-6 EN1993-1-6 (fully fixed) Uniform pressure

GNA λc ¼ 5.5191 λc ¼5.5827 λc ¼ 5.4296 λc ¼ 9.3286 λc ¼ 10.3049 λc ¼8.5594

GNA/LBA (%) 99.5 93.2 96.1 98.8 99.7 99.9

*
All buckling load factors coincide with critical loads measured in kPa.
 C. Maraveas et al. / Thin-Walled Structures 95 (2015) 152–160 159

Fig. 7. GNIA load-displacement curves; (a) T-776 for EN1993-1-6 wind pressure (i) and uniform wind pressure (ii); (b) T-761 for EN1993-1-6 wind pressure (i) and uniform
wind pressure (ii).

Table 6 editions should be considered.

Comparison of equivalent von Mises stresses (MPa).

Tank ID Method type Perfect Class A Class B Class C

T-776 Analytical stress design EN1993- 48.78 21.21 18.07 13.84 Appendix A. Supplementary material
1-6
GNA & GNIA EN1993-1-6 52.36 - - -
Supplementary data associated with this article can be found in
GNA & GNIA uniform pressure 18.44 - - -
T-761 Analytical stress design EN1993- 51.95 22.50 18.58 14.04 the online version at http://dx.doi.org/10.1016/j.tws.2015.07.007.
1-6
GNA & GNIA EN1993-1-6 51.11 49.44 40.92 36.09
GNA & GNIA uniform pressure 28.80 26.59 24.42 23.07
References

unfavorable compared to the EN1993-1-6 distribution, which is [1] Y. Uematsu, K. Uchiyama, Deflection and buckling behavior of thin, circular
cylindrical shells under wind loads, J Wind Eng Indust Aerodyn 18 (3) (1985)
experimentally confirmed. It also provides a different behavior 245–261.
to the shell, allowing smaller displacements and stresses. [2] R. Greiner, P. Derler, Effect of imperfections on wind-loaded cylindrical shells,
 Imperfection amplitudes proposed by EN1993-1-6 decrease Thin-Walled Struct 23 (1-4) (1995) 271–281.
[3] L.A. Godoy, F.G. Flores, Imperfection sensitivity to elastic buckling of wind
considerably the nonlinear buckling resistance of cylindrical loaded open cylindrical tanks, Struct Eng Mech 13 (5) (2002) 533–542.
tanks but also cause a progressively stiffening response with [4] Y. Zhao, Y. Lin, Buckling of cylindrical open-topped steel tanks under wind
rapidly growing displacements for such large diameter struc- load. Thin-Walled Struct 79 (2014) 83–94.
[5] G. Portela, L.A. Godoy, Wind pressures and buckling of cylindrical steel tanks
tures. The standard’s “largest tolerable deformation” criterion with a conical roof, J Constr Steel Res 61 (6) (2005) 786–807.
for estimating the critical load is arbitrary and does not offer [6] G. Portela, L.A. Godoy, Wind pressures and buckling of cylindrical steel tanks
reasonable and satisfactory results. with a dome roof, J Constr Steel Res 61 (6) (2005) 808–824.
 Analytical formulations of stress design method by EN1993-1-6 [7] Y. Lin, Y. Zhao, Wind loads on fixed-roof cylindrical tanks with very low aspect
ratio, Wind Struct 18 (6) (2014) 651–668.
combined with the proposed method for variable thickness [8] R.C. Jaca, L.A. Godoy, Wind buckling of metal tanks during their construction,
tanks offer satisfactory results for perfect tank models. How- Thin-Walled Struct 48 (6) (2010) 453–459.
ever, the imperfection requirements of the aforementioned [9] G. Portela, L.A. Godoy, Shielding effects and buckling of steel tanks in tandem
arrays under wind pressures, Wind Struct 8 (5) (2005) 325–342.
methods are very conservative and limit the buckling re- [10] G. Portela, L.A. Godoy, Wind pressure and buckling of grouped steel tanks,
sistance remarkably. Wind Struct 10 (1) (2007) 23–44.
 Empirical design methods proposed by API 650 are arbitrary, as [11] L.A. Godoy, Performance of storage tanks in oil facilities damaged by hurri-
canes Katrina and Rita, J Perform Constr Facil. 21 (6) (2007) 441–449.
they do not quantify the buckling critical state and do not ac- [12] X. Hua, C. Letchford, A comparison of wind loads on circular bins, silos and
count for imperfection influence. Therefore the provided safety tanks, Proceedings of the ASCE Structures Congress (2014) 1616–1629.
level is questionable; the theoretical background of the meth- [13] Y. Uematsu, C. Koo, J. Yasunaga, Design wind force coefficients for open-topped
oil storage tanks focusing on the wind-induced buckling, J Wind Eng Indust
ods should be investigated, and improvements in future Aerodyn 130 (2014) 16–29.
160 C. Maraveas et al. / Thin-Walled Structures 95 (2015) 152–160

[14] Y. Uematsu, J. Yasunaga, C. Koo, Design wind loads for open-topped storage U.S.A; 2013.
tanks in various arrangements, J Wind Eng Indust Aerodyn 138 (2015) 77–86. [19] API, 650. Welded Tanks for Oil Storage, American Petroleum Institute (2012).
[15] M.A. Al-Kashif, H. Ramadan, A. Rashed, M.A. Haroun, Effect of FRP composites [20] Eurocode 3. Design of steel structures – Part 1-6, strength and stability of shell
on buckling capacity of anchored steel tanks, Steel Comp Struct 10 (4) (2010) structures. European Standard EN 1993-1-6; 2007.
361–371. [21] ABAQUS. Documentation. Dassault Systèmes Simulia Corp. Providence RI;
[16] E.M. Sosa, L.A. Godoy, Challenges in the computation of lower-bound buckling 2013.
loads for tanks under wind pressures, Thin-Walled Struct 48 (12) (2010) [22] Eurocode 1. Actions on structures – Part 1-4, general actions-wind actions.
935–945. European Standard EN 1991-1-4; 2005.
[17] C.A. Burgos, J.C. Batista-Abreu, H.D. Calabro, R.C. Jaca, L.A. Godoy, Buckling [23] R. Greiner, Cylindrical shells: wind loading. Chapter 17 in: Silos (Ed. C.,.J.
estimates for oil storage tanks: Effect of simplified modeling of the roof and Brown & L.Nilssen) EFN Spon, London. 378-399; 1998.
wind girder, Thin-Walled Struct 91 (2015) 29–37. [24] Eurocode 3. Design of steel structures – Part 4-2, tanks. European Standard EN
[18] C. Maraveas, K. Miamis. Shell buckling evaluation of thin-walled steel tanks 1993-4-2; 2007.
filled at low liquid level according to current design codes. Proceedings of the
Annual Stability Conference by Structural Stability Research Council, Missouri,

View publication stats