Você está na página 1de 17

Journal of Constructional Steel Research 61 (2005) 808824

www.elsevier.com/locate/jcsr

Wind pressures and buckling of cylindrical steel


tanks with a dome roof
G. Portelaa, L.A. Godoyb,
a Department of General Engineering, University of Puerto Rico, Mayagez 00681-9044, Puerto Rico
b Civil Infrastructure Research Center, Department of Civil Engineering and Surveying,

University of Puerto Rico, Mayagez 00681-9041, Puerto Rico


Received 5 July 2004; accepted 19 November 2004

Abstract
An experimental/computational strategy is used in this paper to evaluate the buckling behavior
of steel tanks with a dome roof under exposure to wind. First, wind tunnel experiments using small
scale rigid models were carried out, from which pressure distributions due to wind on the cylindrical
part and on the roof were obtained. Second, a computational model of the structure (using the
pressures obtained in the experiments) was used to evaluate buckling loads and modes and to study
the imperfection sensitivity of the tanks. The computational tools used were bifurcation buckling
analysis (eigenvalue analysis) and geometrical nonlinear analysis (step-by-step incremental analysis).
Geometric imperfections and changes in the buckling results due to reductions in the thickness were
also included in the study to investigate reductions in the buckling strength of the shell. For the
geometries considered, the results show low imperfection sensitivity of the tanks and buckling loads
associated with wind speeds 45% higher than those specified by the ASCE 7-02 standard.
2004 Elsevier Ltd. All rights reserved.
Keywords: Buckling; Dome roof; Finite element analysis; Imperfections; Steel tanks; Wind pressures; Wind
tunnel

Corresponding author. Tel.: +1787 265 3815; fax: +1787 833 8260.

E-mail addresses: gportela@uprm.edu (G. Portela), lgodoy@uprm.edu (L.A. Godoy).


0143-974X/$ - see front matter 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jcsr.2004.11.001

G. Portela, L.A. Godoy / Journal of Constructional Steel Research 61 (2005) 808824

809

1. Introduction
Short above-ground steel tanks are constructed in industrial plants to store oil,
petroleum, fuel, kerosene, water, and other fluids. Because such tanks are thin walled
structures, buckling under wind loads is a major concern for the designer. This paper
reports buckling results for tanks with shallow and deep dome roofs, while tanks with a
conical roof are addressed in a companion paper [17].
Damage and collapse of short tanks have been reported in the Caribbean Region after
hurricanes Hugo in 1989, Marilyn in 1995, and Georges in 1998 [5,7]. Failures of tanks
have many consequences, including loss of the structure, environmental problems due to
loss of the contents, and economic problems because the storage capacity of the plant is
reduced. Research in this field started in the 1960s with evaluations of critical loads via
eigenvalue analysis for simple tank models, open at the top and with uniform thickness
[4,11,12,19]. Some authors noticed that for tanks with a roof there was a lack of
information about the actual pressures that should be used to represent wind. In the United
States, ASCE [2] and Uniform Building Code [20] indicate pressures for the design of
tank-like structures under exposure to wind, but the recommendations handle tanks in the
category of other structures and provide extremely simplified wind pressures which are
constant in the circumferential direction. Other national building codes have adopted nonuniform wind pressure distributions, but they cover geometric ranges for tanks taller than
those considered in this paper.
Research regarding wind pressure distributions in silos has been reported by Esslinger
et al. [3], Gretler and Pflgel [10], and Pircher et al. [15]. For the present interest, the main
limitation of such valuable studies is that mainly tall structures have been investigated in the
past, notably with relations between height and diameter larger than 1. Wind tunnel tests
on shorter structures were performed by Maher [13] on cylindrical tanks with spherical
roofs, and by Purdy et al. [18] on cylinders with a flat roof. For short tanks with height
to diameter ratio smaller than 1, as it would be in most tanks in the Caribbean region,
the information on wind pressures is extremely scarce. This motivated new wind tunnel
experiments carried out by the authors, which are reported in the following sections.
The strategy in the research reported in this paper has been experimental/computational,
in which wind tunnel experiments were carried out on rigid models to obtain
pressure distributions, and the pressures were next used in a finite element model to
evaluate buckling loads, buckling modes, and imperfection sensitivity. Studies concerning
imperfection sensitivity of tanks subjected to wind loads have been reported in Refs. [6,9].
Miller [14] investigated the influence of initial imperfections on the buckling of cylinders
under uniform external pressure and developed the basis for the ASME code.
2. Experimental set-up used to evaluate wind pressures in tank models
There have been a surprisingly small number of wind tunnel experiments performed
on cantilever cylindrical shells with a dome roof, as evidenced by a review paper by
Greiner [8]. The first set of studies seems to be that due to Maher [13] in the 1960s,
while a second study in the following decade is due to Esslinger and co-workers [3].

810

G. Portela, L.A. Godoy / Journal of Constructional Steel Research 61 (2005) 808824

Fig. 1. Schematic side view of the wind tunnel laboratory.

Both investigations were mainly interested in geometries which represent silo structures,
with H /D > 1.
This motivated research to evaluate wind pressures on empty tanks using wind tunnel
modeling. The geometries of the tanks of interest were identified following field trips to
industrial plants in Puerto Rico [21]. One of the conclusions of the survey was that more
than 80% of the tanks with a dome roof have aspect ratios between 0.25 and 0.60. On
the basis of that observation, it was decided to investigate pressures on two tank models
having H /D = 0.48, one with a shallow dome roof (Hr /H = 0.318) and the other one
with a deep dome roof (Hr /H = 0.475), where Hr is the rise of the roof at its center. The
dimensions considered for the small scale models in the experiments were H = 115.5 mm,
D = 238.1 mm, Hr = 36.7 mm (shallow dome), and Hr = 54.9 mm (deep dome). As
a reference, Maher [13] studied models with H /D = 0.25 (Hr /H = 0.39 and 0.68)
and H /D = 0.5 (hemispherical roof with Hr /H = 1.0). The study of Maher [13]
for spherical and hemispherical roofs used different wind velocity profiles and different
heights of reference with respect to the present investigation, which makes it difficult to
make comparisons with the pressures reported in this paper.
The relative dimensions chosen for the models were influenced by the typical
geometries found in real tanks, and also by limitations of the experimental facility available
at the University of Puerto Rico at Mayagez (UPRM), such as the dimensions of the wind
tunnel, longitudinal roughness length, wind profile, and the need to avoid blockage effects.
The wind tunnel laboratory at UPRM is 12 m long. The tunnel was constructed in 1985
and has an open circuit, classified as a short tunnel because the experimental section is
smaller than 5 m. A side view of the tunnel is shown in Fig. 1. Full details of the wind
tunnel and the experiments are given in Ref. [16] and only a brief summary is provided in
this section.
The dynamic pressures at different points of the models were referenced to pressure
coefficients defined as
PL
(1)
Cp =
Pd
where PL is the local dynamic pressure at a point, and Pd is the reference dynamic
pressure at an equivalent height of 10 m [16]. The wind pressure distributions on the tanks
were measured assuming rigid models and simulating an atmospheric surface layer for an
open-terrain aerodynamic roughness condition based on a full-scale value of 0.02 m, as
recommended by Wieringa [22]. The wind velocity profile, shown in Fig. 2, was simulated

G. Portela, L.A. Godoy / Journal of Constructional Steel Research 61 (2005) 808824

Fig. 2. Wind velocity profile used for experiments.

Logarithmic model.

811

Experimental profile.

Fig. 3. (a) Model with a shallow dome roof. (b) Model with a deep dome roof.

using the logarithmic law


 
z
1
V (z) = u ln

z0

(2)

where V is the wind speed, z is the elevation from the surface terrain, z 0 is the aerodynamic
roughness length, is the von Karman constant (taken as 0.4), and u is the friction
velocity.
The models were made of PVC, and discontinuities in the surface were removed with
the application of a resin mix. The shell was rubbed, painted, and polished to get a final
smooth external surface as shown in Fig. 3. Twenty-one pressure taps were placed on the
roof (five for each meridian at 90 and one at the center), and twenty were located on the
cylinder.
The wind velocity in the wind tunnel at a height of 116 mm was 19.8 m/s, and represents
a full-scale wind speed of 64.8 m/s at a height of 10 m. This is the design wind speed in
Puerto Rico, the Caribbean region, and the south-eastern coast of United States, according
to ASCE [2]. Temperature, humidity, and barometric pressure were monitored during the

812

G. Portela, L.A. Godoy / Journal of Constructional Steel Research 61 (2005) 808824

Fig. 4. Angles at which the models were rotated in order to measure pressures.

tests in order to identify changes in the air density which could affect the local pressures
measured in the wind tunnel.
The models were instrumented with pressure taps separated at angles of 90 . However,
in order to account for local pressures on the roof and the cylinder, the tank was rotated
from 0 to 67.5 at intervals of 22.5, as shown in Fig. 4. The data acquired per instrument
was the average of 800 samples in a period of 10 s. All recorded values were corrected
due to environmental factors using psychometric relations. The experiments were repeated
several times and three tests were selected to average the results.
3. Wind tunnel results
Results for pressure contours are shown in Fig. 5 for the shallow dome (Hr /D = 0.15)
and in Fig. 6 for the deep dome tanks (Hr /D = 0.23). Dome roofs have a variable slope,
and the transition between the cylinder and the spherical cap is not so pronounced as it
would be, for example, in tanks with a conical or flat roof.
Due to the symmetry of the pressures measured on the roof with respect to the windward
meridian, the values on either side of the axis were averaged. The highest pressures on the
roof were measured at the center, and there was a 30% increase in the dome roof with
respect to the shallow dome roof. At the junction between the cylinder and the dome,
the pressures were lower than the values at the center. This is similar to the pressure
pattern found by Maher [13] in his study on hemispherical roofs. An approximate value of
C p = 0.15 was measured at the leeward region for both tanks. Maher [13] obtained a
value of C p = 0.20 for hemispherical roofs with H /D = 0.5.

G. Portela, L.A. Godoy / Journal of Constructional Steel Research 61 (2005) 808824

813

Fig. 5. Mean pressure coefficients in model CMT1; (a) roof, (b) cylinder.

Unlike pressure distributions found on the roof of the tanks, the cylindrical wall patterns
do not present very considerable changes between shallow and deep dome configurations,
and only the magnitudes have differences. Contours with pressure coefficients were
plotted along the circumference of the cylinders and are shown in Figs. 5(b) and 6(b).

814

G. Portela, L.A. Godoy / Journal of Constructional Steel Research 61 (2005) 808824

Fig. 6. Mean pressure coefficients in model CMT2; (a) roof, (b) cylinder.

The peak pressures (positive values) were obtained on the windward meridian, while the
peak suctions (negative values) were found at an angle near 90. For both shallow and
deep cases, relatively similar distributions were found about the windward meridian. The
maximum values on the cylinder were measured at elevations between 0.5 H and 0.9 H ,
referenced from the base.

G. Portela, L.A. Godoy / Journal of Constructional Steel Research 61 (2005) 808824

815

Fig. 7. Plate thickness and dimensions of models CMT1 and CMT2.

The pressure coefficients found for each of the three experimental tests around the wall
were next fitted with Fourier series. Seven coefficients provided a good approximation
of the distribution, and dispersion was found only near to the end of the period of the
function, where higher participation factors begin to contribute. The coefficients proposed
in this paper for approximating the pressure distribution at different angles around the
circumference for tests CMT1 and CMT2 are
C p = 0.1925 + 0.3708 cos + 0.5612 cos 2 + 0.2297 cos 3 0.0263 cos 4
0.0168 cos 5 0.0062 cos 6 0.0288 cos 7
(1)
C p = 0.1917 + 0.3161 cos + 0.5001 cos 2 + 0.2475 cos 3 0.0404 cos 4
0.0114 cos 5 + 0.0599 cos 6 + 0.0266 cos 7
(2)
where is measured from the windward meridian.
4. Computational buckling analysis
The pressure patterns obtained experimentally from the rigid models were used in a
computational analysis. The finite element package ABAQUS [1] was used to model tanks
with geometric properties similarly to the scale-reduced models tested in the wind tunnel.
The geometries of the tanks are shown in Fig. 7. In all computational models, the material
was assumed as elastic and isotropic, with modulus of elasticity E = 2 108 kPa, yield
strength y = 2.48 105 kPa, Poisson ratio = 0.3, and mass density = 7850 kg/m3 .
The diameter was 20.12 m and the height of the cylinder was taken as 9.66 m in both
cases, but the shallow dome had roof height Hr = 2.41 m, while the deep dome roof was
modeled with Hr = 4.60 m. The boundary conditions at the bottom of the tanks were
assumed fixed.
P and H type convergence analyses were carried out in order to select the type
of element and the mesh density to be used in the static buckling analysis. Linear and

816

G. Portela, L.A. Godoy / Journal of Constructional Steel Research 61 (2005) 808824

quadratic shell elements provided by ABAQUS were tested, but it was decided to use
quadratic elements for a faster convergence in the solution.
The quadratic rectangular shell element designated by ABAQUS as S8R5 was used to
model the cylinder, while the quadratic triangular element STRI65 was used to model the
roof. Both elements use five degrees of freedom per node: the displacements in each of the
three spatial directions and two rotations with respect to the in-plane axis. The S8R5 is a
doubly curved thin shell element with reduced integration and eight nodes and the STRI65
is a six-node triangular element.
The pressures due to fluids stored were not included in the models because empty tanks
are the worst case scenario under wind pressures. All the pressure values are referenced
to the average air properties measured during the experiment. The air density used was
air = 1.14 kg/m3 and the reference velocity of air was Vair = 64.8 m/s, corresponding
to a reference pressure of P = 2.39 kPa.
The computational models were loaded with pressures that represent a given percentage
of the reference wind pressure, based on experimental wind tunnel pressure coefficients.
The buckling load factors () multiply the reference pressure such that c is a critical load
of the tank for a wind profile that is assumed constant during the load process.
5. Computational buckling results
The computer models are designated as CMT1 for the tank with a shallow dome roof
and CMT2 for the tank with a deep dome roof. The lowest critical pressures obtained
from the eigenvalue analysis are c = 2.83 for CMT1 and c = 2.91 for CMT2, which
are associated with a critical wind velocity in the order of 110 m/s. Those are high
values; however, a bifurcation analysis represents an upper limit of the load reached by
a structural system. Furthermore, bifurcation buckling does not provide information about
the postbuckling behavior of the structure and its imperfection sensitivity. The roof of
the tanks is rigid enough to sustain the wind pressures developed on the tank with a low
thickness of 9.5 mm, as shown in Fig. 8. From these results it seems that the roof shape
(either shallow or deep dome) does not significantly influence the critical load of the tanks.
Figs. 9 and 10 show the nonlinear responses of the models CMT1 and CMT2, for
an ideal or perfect geometry and for imperfect configurations. Loaddisplacement
curves were constructed using a node in the cylindrical shell representative of the
maximum displacements experienced by the tank. The node where maximum deflections
were computed in the wind direction was used as the degree of freedom to plot
loaddisplacement curves. Positive displacements are related to outward directions, while
negative displacements are directed toward the inside of the tank.
Geometric imperfections were assumed with the same shape as the buckling mode
obtained in the eigenvalue analysis, with a maximum amplitude identified by the scalar
parameter . Imperfection amplitudes = 0.1 t, 0.25 t, 0.5 t, 1.0 t, and 2.0 t, were
considered in the analysis, where t is the smallest shell thickness in the cylinder. For the
case of the tank with perfect geometry, limit point load factors of c = 2.51 (CMT1) and
c = 2.48 (CMT2) were obtained, representing a difference with respect to the eigenvalue
results in the order of 12% (shallow dome) and 15% (deep dome). These load factors are

G. Portela, L.A. Godoy / Journal of Constructional Steel Research 61 (2005) 808824

817

Fig. 8. Modes of deformation measured using an eigenvalue analysis; (a) CMT1 and (b) CMT2.

Fig. 9. Geometric nonlinear equilibrium path computed for perfect and imperfect geometries of model CMT1.

associated with wind velocities close to V = 102 m/s. These are high values in comparison
with the design wind speed specified by ASCE [2].
Fig. 11 depicts the deformations of the models at the critical state. The initial buckling
modes reached by both models show additional vertical deformations at the region between
the cylinder and the roof. The vertical displacements in the cylinder are ten times smaller
than the radial displacements. The modes of deformation observed in the cylinder are
similar to those found in the eigenvalue analysis, but more circumferential waves develop.
The deformation shapes as well as the magnitudes of positive and negative
imperfections with values up to 0.25 t were practically the same for shallow and
deep domes. Small negative imperfections (0.1 t to 0.25 t) tend to stiffen the
loaddisplacement behavior of the tank, but the behavior is still very similar to that of
the ideal tank. In other words, the equilibrium paths show that the tanks overcome the

818

G. Portela, L.A. Godoy / Journal of Constructional Steel Research 61 (2005) 808824

Fig. 10. Geometric nonlinear equilibrium path computed for perfect and imperfect geometries of model CMT2.

Fig. 11. Deformed shape at the first critical load; (a) CMT1 and (b) CMT2.

initial negative imperfection without abrupt changes either in shape or in direction of the
displacement. The low sensitivity to imperfections is also noted in the small reduction
in the critical loads reached for = 0 to = 0.25 t. Table 1 shows that the largest
difference is less than 2.8% for CMT1 and 2.02% for CMT2, with respect to the ideal
geometry.
For higher imperfection amplitudes, the modes presented different shapes depending
on whether they were related to positive or negative imperfections. For model CMT1 and
= 0.5 t, the reduction in load capacity was also negligible with a factor c = 2.48,
which represents 1.2% of the value in the ideal tank. But when the imperfection was applied
in the same direction as the eigenvector ( = +0.5 t), the load capacity was reduced by

G. Portela, L.A. Godoy / Journal of Constructional Steel Research 61 (2005) 808824

819

Table 1
Critical load values computed for different imperfection levels
Imperfection
0.10 t
+0.10 t
0.25 t
+0.25 t
0.50 t
+0.50 t
1.0 t
+1.0 t
2.0 t
+2.0 t

CMT1
Load factor (c )

Difference (%)

CMT2
Load factor (c )

Difference (%)

2.54
2.49
2.56
2.44
2.48
2.32
2.18
2.12

1.20
0.80
1.99
2.79
1.20
7.57
13.15
15.54

2.49
2.47
2.49
2.43
2.44
2.32

0.40
0.40
0.40
2.02
1.61
6.45

A negative sign means that the load factor obtained using the corresponding imperfection is higher than the ideal.
Blanks represent that a critical load was not identified in the analysis.

Fig. 12. Deformed shape computed in model CMT1 for an imperfection = +2.0 t.

almost 8%. Larger positive imperfections ( = +1.0 t and = +2.0 t) also present
similar deformed shapes, but even when the maximum buckling load is reduced by more
than 15%, this value represents a wind speed 45% higher than the ASCE 7-02 design wind
speed. From Fig. 12 ( = +2.0 t), it may be seen that for the selected region in the
cylinder, the deformed shape at the windward meridian cannot be defined by a simple sine
function. On the other hand, higher negative amplitudes of imperfect geometries produced
mode shapes similar to those observed for imperfections with smaller magnitudes.
For model CMT2, the mode shapes at the first critical load reached for = 0.5 were
very similar to those computed for small imperfections, but the load was reduced by 6.5%,
as shown in Table 1. Critical points could not be identified for imperfections of = 1.0
and = 2.0, because the loaddisplacement curve showed a tendency to converge

820

G. Portela, L.A. Godoy / Journal of Constructional Steel Research 61 (2005) 808824

Fig. 13. Deformed shape computed in model CMT2 for imperfections = +2.0 t and = 2.0 t.

towards the lowest postbuckling capacity reached by the perfect tank. For imperfections
of = 1.0 t and = 2.0 t, positive displacements occur in the initial equilibrium path
followed by the tank (Fig. 10). Combinations of vertical waves in the meridian direction
were observed for imperfections with = 1.0 t and higher, especially for = 2.0 t,
as shown in Fig. 13.
For imperfections with maximum amplitude smaller than 1.0 t, the shell presents an
unstable postcritical behavior. For larger imperfections, the maximum in the primary path
and the minimum in the postcritical path approach each other and they coalesce in the limit,
so an inflection point is obtained. For imperfections larger than those considered in Figs. 9
and 10 ( > 2 t) the buckling loads are not reduced but the displacements increase in a
stable form.

G. Portela, L.A. Godoy / Journal of Constructional Steel Research 61 (2005) 808824

821

Fig. 14. Imperfection sensitivity computed in models CMT1 and CMT2.

In general, it seems that the buckling load factors of models CMT1 and CMT2
show little sensitivity to small imperfections (Fig. 14). This behavior is characteristic
of limit points found in the equilibrium path of a structure. Furthermore, due to the
differences in critical load factors computed by eigenvalue and nonlinear analyses,
the shape of the loaddisplacement paths, and the low sensitivity with respect to
geometrical imperfections, it seems that this case displays an unstable limit point
behavior.
6. Influence of thickness reduction on the buckling loads of tanks
The computational models CMT1 and CMT2 were also analyzed considering a
reduction in the effective thickness of the shell. For these cases, the thickness of the
cylinder was reduced to t = 6.34 mm and the roof to t = 7.9 mm. For the tank CMT1
with a shallow dome roof, a value of c = 1.59 results from a bifurcation analysis with
reduced thickness, while for a nonlinear analysis the limit load was c = 1.43, which
is 10% lower than the bifurcation value. For the deep dome tank CMT2, the bifurcation
load was c = 1.64 and the nonlinear maximum load was c = 1.39. An imperfection
sensitivity analysis was not repeated for tanks with thickness reduction, but it seems that
the tank still has a limit point behavior.
If a lower limit capacity of the tanks is identified with the lowest load in the unstable
postcritical path (Fig. 15), then the capacity of the tanks is reduced to c = 1.11 for both
models. This value represents a velocity of V = 68.3 m/s at an elevation of 10 m. Such
high buckling capacity explains why such tanks with a dome roof located in Puerto Rico
were not damaged following hurricane Georges in 1998.
7. Conclusions
The main conclusions of this research may be summarized as follows:

822

G. Portela, L.A. Godoy / Journal of Constructional Steel Research 61 (2005) 808824

Fig. 15. Equilibrium path including the effects of thickness reductions in models CMT1 and CMT2.

1. The wind pressures on the cylindrical part of a tank with a dome roof are positive on
the windward area (leading to a maximum C p = 0.87), and negative at about 90
from the windward meridian (maximum negative pressures C p = 0.83). Negative
pressures are also detected on the leeward area, with C p = 0.15. ASCE [2] suggests
that pressures on a cylindrical structure would be constant in the circumferential
direction. As shown in Section 3 this is clearly inadequate for tank structures, in
which there are changes in values and even in the sign of the pressures around the
circumference.
The details of the geometric transition between the cylindrical body and the roof are
crucial in the evaluation of pressures on the roof, since this transition changes the main
features of the flow separation. Smooth transitions, such as in dome configurations, lead
to lower pressure levels than abrupt transitions, such as in shallow conical roofs [17].
Therefore, results from one case cannot be freely applied to another case with a different
transition.
2. The tanks investigated in this paper displayed an initially stable equilibrium path
followed by an unstable nonlinear postcritical path. The postbuckling equilibrium
paths computed with both the perfect geometry of the tank and the case with
initial geometric imperfections converge for large displacements; however, for such
large displacements the structure may display plasticity, which was neglected in this
research. Tanks with dome roofs show small imperfection sensitivity in buckling
loads.
3. The influence of the roof on the buckling response is manifested in two opposing
ways: on the one hand, there is an additional stiffness provided by the roof, and
on the other hand, there is an additional surface of the structure on which suctions
occurs. For the dimensions considered in the cases of studies, the tanks can sustain
high wind pressures before buckling, and would have a stable response for the most
stringent ASCE provisions [2]. Reductions in the shell thickness reduce the buckling
loads, but these loads are still for wind velocities higher than those required by ASCE.

G. Portela, L.A. Godoy / Journal of Constructional Steel Research 61 (2005) 808824

823

The computational results seem to confirm the high buckling strength of tanks with
dome roofs, for which buckling was not observed after the main hurricanes in the
Caribbean islands.

Acknowledgements
The authors thank Dr. Ral E. Zapata for his valuable contribution to the wind tunnel
experiments. This research was supported by grants from FEMA (PR00660-A) and from
NSF (CMS-9907440).
References
[1] Abaqus user manualversion 5.6. Pawtucket (RI, USA): Hibbit, Karlsson and Sorensen; 1996.
[2] ASCE 7 standard. Minimum design loads for buildings and other structures. Reston (Virginia): American
Society of Civil Engineers; 2002.
[3] Esslinger M, Ahmed S, Schroeder H. Stationary wind loads of open topped and roof-topped cylindrical
silos. Der Stalbau 1971;18 [in German].
[4] Flores FG, Godoy LA. Buckling of short tanks due to hurricanes. Engineering Structures 1998;20(8):
75260.
[5] Godoy LA. Catastrofes Produzidas por Furacoes no Mar do Caribe. In: da Cunha AJP, Lima NA, de
Souza VCM, editors. Acidentes Estruturais na Copnstrucao Civil, vol. 2. Sao Paulo (Brazil): Editora Pini;
2000. p. 25562 [Chapter 26] [in Portuguese].
[6] Godoy LA, Flores FG. Imperfection sensitivity to elastic buckling of wind loaded open cylindrical tanks.
Structural Engineering and Mechanics 2002;13(5):53342.
[7] Godoy LA, Portela G, Sosa E, Surez LE, Virella JC, Zapata RE. Damage due to buckling in aboveground
storage tanks. In: P Castro, editor. Proc. int. conf. on the behavior of structures with damage. Rio de Janeiro
(Brazil): Universidade Federal Fluminense; 2002.
[8] Greiner R. Cylindrical shells: Wind loading. In: Brown CJ, Nilssen L, editors. Silos. London: EFN Spon;
1998. p. 37899 [Chapter 17].
[9] Greiner R, Derler P. Effect of imperfections on wind loaded cylindrical shells. Technical report. Austria:
Institute for Steel and Shell Structures, Technical University of Graz; 1993.
[10] Gretler W, Pflgel M. Wind tunnel tests of cylindrical shells at the TU Graz. Technical report. Austria:
Technical University of Graz; 1978.
[11] Jaca R, Godoy LA. Colapso de un tanque metlico en construccin bajo la accin del viento. Revista
Internacional de Desastres Naturales, Accidentes e Infraestructura Civil 2003;3(1):7383 [in Spanish].
[12] Kundurpi PS, Savamedam G, Johns DJ. Stability of cantilever shells under wind loads. Journal of the
Engineering Mechanics Division, ASCE 1975;101(5):51730.
[13] Maher FJ. Wind loads on dome-cylinders and dome-cone shapes. Journal of the Structural Division, ASCE
1966;91(3):7996.
[14] Miller CD. The effect of initial imperfections on the buckling of cylinders subjected to external pressure,
PVRC Grant 94-28, Final report. Plainfield (IL): Chicago Bridge & Iron Technical Services Company; 1995.
[15] Pircher M, Guggenberger W, Greiner R, Bridge R. Stresses in elastic cylindrical shells under wind load.
Nepean: University of Western Sydney; 1998. p. 6639.
[16] Portela G. Wind pressures and buckling of metal cantilever tanks. Ph.D. dissertation. Puerto Rico: University
of Puerto Rico at Mayagez; 2004.
[17] Portela G, Godoy LA. Wind pressures and buckling of cylindrical steel tanks with conical roof. Journal of
Constructional Steel Research [in press].
[18] Purdy DM, Maher PE, Frederick D. Model studies of wind loads on flat-top cylinders. Journal of the
Structural Division, ASCE 1967;93:37995.

824

G. Portela, L.A. Godoy / Journal of Constructional Steel Research 61 (2005) 808824

[19] Schmidt H, Binder B, Lange H. Postbuckling strength design of open thin walled cylindrical tanks under
wind load. In: Shanmugam S et al. editors. Thin Walled Structures. Oxford: Elsevier Science; 1998.
p. 20320.
[20] Uniform building code. Structural engineering design provisions. vol. 2, International conference of building
officials. 1997.
[21] Virella JC. Buckling of tanks subject to earthquake loadings. Ph.D. dissertation. Puerto Rico: University of
Puerto Rico at Mayagez; 2004.
[22] Wieringa J. Updating the Davenport roughness classification. Journal of Wind Engineering and Industrial
Aerodynamics 1992;4244:35768.

Você também pode gostar