Professional Documents
Culture Documents
Buckling of Shallow Spherical Concrete Domes Under Gravity and Earthquake Loads
Buckling of Shallow Spherical Concrete Domes Under Gravity and Earthquake Loads
Abstract: Buckling analysis of spherical concrete domes constructed over prestressed concrete tanks subjected to gravity and earthquake
load combinations is complex due to the sensitivity of spherical dome buckling to geometric imperfection, the effect of creep and shrinkage
Downloaded from ascelibrary.org by Columbia University on 03/11/20. Copyright ASCE. For personal use only; all rights reserved.
on amplifying the deflection of the dome (increasing the radius of curvature of the deformed dome and reducing the buckling resistance), and
the geometric and material nonlinearity of the concrete dome. Study of imperfections has shown that imperfections that increase the average
radius of curvature in an area the size of an elastic buckle have the highest impact in reducing buckling strength; thus, the problem of buckling
of the concrete dome is reducible to the snap-through buckling of a shallow cap equal in size to an elastic buckle of the dome. In this paper, the
buckling of a shallow cap subjected to gravity and seismic load combinations is determined using nonlinear geometry, material nonlinearity of
concrete (accounting for softening and cracking in tension and microcracking and crushing in compression), and time-dependent creep and
shrinkage of concrete. The analysis is performed in three steps: in the first step, the cap is analyzed for gravity loads; in the second step, the
cap is analyzed for the effects of creep and shrinkage strains; and in the third step, the earthquake load is applied until the snap-through
buckling occurs. The analysis was carried out on a set of domes designed for gravity load combinations alone. These domes cover the
extremes of size and rise of the domes that are currently constructed over prestressed concrete tanks. The results show that the majority
of existing domes that are not in high-seismicity zones and are designed for gravity load combinations have capacity for earthquake loading.
The results are also of value in the design of new domes for load combinations that include earthquake loading. DOI: 10.1061/(ASCE)
ST.1943-541X.0002595. © 2020 American Society of Civil Engineers.
Author keywords: Buckling; Concrete; Creep; Imperfection; Earthquake; Nonlinear analysis; Spherical domes.
Introduction buckling, the design criteria include the ACI 318 (ACI 2014) re-
quirements for stress limits in shear and bending in the edge region
Spherical concrete domes are used as covers over prestressed of the dome. (The design of the domes includes a prestressed ring
concrete tanks. They are typically cast in place or precast using dou- beam with enough prestressing to counteract the horizontal com-
bly curved panels. These domes range in diameter up to 91.4 m ponent of the meridional membrane stress resultant and a taper in
(300 ft), with a rise-to-diameter ratio from 1∶8 to 1∶16 with 1∶10 the thickness of the dome in the edge zone to handle the remaining
being typical, and in thickness from a minimum of 76 mm (3 in.) shear and bending in the dome.) Load combinations that include
to about 508 mm (20 in.). The thickness of these domes increases earthquake have often been ignored or have been considered with-
near the base of the dome, where the dome ends in a ring beam that is out the ability to predict the ultimate capacity of the dome account-
prestressed (Fig. 1). The prestress of the ring beam is designed to ing for material nonlinearity (including cracking, compression
produce a radial force equal to the horizontal component of the nonlinearity, and crushing of concrete) and the significant changes
meridional membrane stress resultant at the edge of the dome when in stress distribution in the dome that result from time-dependent
the dome is subjected to dead load. These domes are lightly rein- shrinkage and creep deformations.
forced with not less than 0.25% of the concrete cross section in both In this paper, the initial designs selected meet all design criteria
circumferential and meridional directions; the reinforcement is when subjected to load combinations that do not include earth-
placed at the middepth of the dome wall. Additional reinforcement quake. The paper’s focus is placed on predicting the ultimate
exists in the thickened part of the dome near the base. strength of the dome when subjected to the design earthquake load-
The spherical domes are currently designed against buckling by ing after time-dependent effects of creep and shrinkage under the
selecting the thickness of the dome of a given geometry, D and H, gravity loads of the design load combinations that include earth-
for gravity loads alone, i.e., dead and live or snow loads, consid- quake. Although an earthquake can occur at any time during the
ering both imperfection and creep effects, both of which influence service life of the structure, in this paper, the earthquake load is
the state of stress in the dome. In addition to design against conservatively applied after allowing the dome to undergo creep
and shrinkage deformations that tend to reduce the curvature and
1
Senior Principal, Simpson Gumpertz & Heger Inc., 480 Totten Pond increase the stresses in the dome.
Rd., Waltham, MA 02451. Email: mszarghamee@sgh.com The buckling strength of spherical domes is very sensitive to
2
Senior Project Manager, Simpson Gumpertz & Heger Inc., 480 Totten imperfections, and the strength of the dome cannot be evaluated
Pond Rd., Waltham, MA 02451 (corresponding author). ORCID: https:// without consideration of imperfections that are likely to occur in
orcid.org/0000-0003-4411-6956. Email: atsarawit@sgh.com
construction. Hutchinson and Koiter (1970) published a review of
Note. This manuscript was submitted on October 24, 2018; approved
on October 7, 2019; published online on February 22, 2020. Discussion extensive research performed on buckling of spherical shells with
period open until July 22, 2020; separate discussions must be submitted imperfections. Huang (1964) performed a geometrically nonlinear
for individual papers. This paper is part of the Journal of Structural En- analytical study of symmetric and asymmetric elastic buckling of
gineering, © ASCE, ISSN 0733-9445. thin shallow spherical shells. The results of snap-through buckling
the dome. jected to earthquake loading. The scope of the work is limited
For design of spherical concrete domes subjected to the gravity to cast-in-place and precast spherical concrete domes with pre-
(dead and live) load combinations, the Zarghamee and Heger stressed ring beams constructed on prestressed circular tanks.
(1983) design approach was based on the assumption that the dome
includes a shallow cap of imperfection with reduced curvature that
can snap through and initiate the buckling process of the dome. Method of Approach
These authors presented a design equation for buckling strength The method of approach for analysis of buckling of spherical
of the dome subjected to static loading, creep, and imperfection concrete domes subjected to combined effects of gravity and earth-
in the following form: quake loads accounts for geometric imperfection, time-dependent
2 creep and shrinkage of concrete, as well as nonlinearity of materials
t and accounts for cracking of concrete in tension and microcracking
ðpcr Þdesign ¼ 0.66Ec ϕβ i β c ð1Þ and crushing of concrete in compression. The analysis is performed
Ro
in three steps. In the first step, the cap is analyzed for gravity loads,
in the second step, the cap is analyzed for the effects of creep and
where ðpcr Þdesign = critical buckling pressure for design accounting shrinkage strains, and in the third step, the earthquake load is ap-
for creep of concrete and imperfection; ϕ = resistance factor; β i = plied until the snap-through buckling occurs.
geometric imperfection reduction factor; and β c ¼ 0.44 þ 0.063S Because snap-through buckling of the imperfection region ini-
is the creep factor for gravity load combinations per the current tiates and conservatively estimates the buckling strength of the
design procedure, where S is snow load (kPa) and β c is limited dome, the analysis is performed on a shallow cap of critical size
pffiffiffiffiffiffiffi
pffiffiffi diameter equal to d ¼ 4.3 Ro t with a radius of curvature of
to a maximum of 0.53. Based on a survey of constructed domes with
with imperfections, strict control of the thickness of domes, the 2Ro subjected to long-term gravity loading followed by short-
quality of concrete, and the geometry of the dome, an imperfection term earthquake loading. The earthquake loading is consistent with
factor of β i ¼ 0.5 was selected for design. the requirements of ASCE 7 (ASCE 2016) for combined vertical
Hamed et al. (2011a, b) performed analytical and experimental and horizontal seismic loading, expressed in the form of maximum
studies of failure behavior of thin-walled shallow unreinforced con- acceleration normal to the plane of the shallow cap of critical size
crete domes, including the long-term effects of creep and shrink- centered at the point of maximum normal acceleration. The maxi-
age. The experimental study was performed on a dome that had a mum acceleration represents the combined horizontal and vertical
3 m diameter with a rise of 100 mm (resulting in a radius of cur- response spectral acceleration of the spherical cap. Additional dis-
vature of 11.3 m), a nominal thickness of 30 mm, and a concrete cussion on the horizontal and vertical seismic loading is provided
strength of 58 MPa and modulus of elasticity of 31.4 GPa at the in Appendix II. The shallow cap acceleration is increased until
time of loading on the 20th day after casting of the dome. The snap-through buckling occurs. The analysis is repeated for different
actual casting proved to have an imperfection in thickness that var- tank diameters, rise-to-diameter ratios, snow loads, and concrete
ied between 25 mm at the base of the dome and 32 mm at its center. strength values, including extremes of existing tank designs as well
A short-term test was first conducted (Hamed et al. 2011a), result- as some more typical designs.
ing in buckling pressure of 85 kPa that was between 33.7% (for
32 mm thickness) and 55.2% (for 25 mm thickness) of Pcr . This
knockdown factor is due to imperfections in the test specimen Analysis
(consisting of thickness nonuniformity and curvature imperfections
that are most likely present in most constructed concrete domes). A nonlinear finite-element model is developed for analysis of
In addition, two tests were conducted, presumably with the same snap-through buckling of the shallow cap of the concrete dome.
imperfection: Dome D1 was subjected to 49% of the short-term The finite-element software ABAQUS version 6.14-1 is used to
pffiffiffiffiffiffiffi
buckling pressure, 42 kPa, and failed after 62 days (Hamed et al. perform the analyses. The shallowpffiffiffi cap has a diameter of 4.3 Ro t,
2011a); Dome D2 was subjected to 43.5% of the short-term buck- thickness t, radius of curvature 2Ro , and clamped boundary con-
ling pressure, 37.5 kPa, and failed after 240 days (Hamed et al. ditions. The finite-element model is axially symmetric, consisting
2011b). During the testing process, the domes were subjected to of multilayer axisymmetric shell elements, with each layer having
the standard laboratory environment. The analytical model used a stress–strain relationship of concrete. This uniaxial stress–strain
by Hamed et al. (2011b) accounted for the creep and shrinkage relationship of concrete material model is the Darwin and Pecknold
using direct measurements taken on samples of concrete, as well (1974) model, shown in Fig. 4.
as for changes in the stress state during the creep and shrinkage The material model is orthotropic to capture both radial- and
process. hoop-direction material responses. Poisson’s ratio is initially set
MPa
20 stored and tracked throughout the analysis.
The shrinkage strain is calculated and imposed as a temperature
Stress
15 change, along with creep during each increment of analysis. After
Compression imposing strain increments corresponding to the combined incre-
ment of creep and shrinkage, the new deformed geometry and the
10
stresses in the concrete dome are computed from mechanical
strains, and the total strain is the sum of creep, shrinkage, and
5 mechanical strains. Both creep and shrinkage strains result in addi-
tional deformations that increase the radius of curvature, cracking,
0 and maximum compressive stress in the shell, which in turn result
Downloaded from ascelibrary.org by Columbia University on 03/11/20. Copyright ASCE. For personal use only; all rights reserved.
to avoid material nonlinearity, and the snap-through buckling was The results show that in general, the domes that are not in high-
nearly elastic. This is the same conclusion arrived at by Hamed et al. seismicity zones and are designed according to the current design
(2011b). The difference in the time to failure is minor and is related procedure have capacity for earthquake loads. However, the calcu-
to the approximations in the calculation of creep and shrinkage, lated capacities for seismic spectral response need to be reduced by
the accuracy of the thickness variation formula, possible geometric the material resistance factor, ϕ. Obviously, there are concrete
imperfections that were not noted in the test specimen, and concrete domes that need to be thickened to meet the design spectral re-
maturing with time, which was neglected in this paper. The results sponse acceleration in high-seismicity zones.
of this comparison validate the model of geometric nonlinearity and The load-deflection diagrams and contour plots of the radial
creep and shrinkage, but not concrete nonlinearity due to cracking strains through the thickness of the shallow caps for Case 2 are
in tension or crushing in compression.
shown in Figs. 6 and 7, for Case 7 in Figs. 8 and 9, and for Case
9 in Figs. 10 and 11, respectively. The load is expressed as the ratio
Results of pressure and the factored gravity load, pgs .
For each case, the deformation versus pseudotime t is also
Using the aforementioned model, a parametric study is performed plotted over three pseudotime intervals: in the first time interval
by calculating the buckling strength for a range of dome geometries [0, 1], the shallow caps are subjected to gravity loads; in the
and snow loads S, as indicated in Table 1, where D = tank diameter second time interval [1, 2], they are subject to ultimate creep and
and H = dome rise. The dome geometries selected for this study ultimate shrinkage; and in the third time interval [2, 3], they are
included typical tank diameters in the range of D ¼ 30.5, 76.2, and subject to increasing pressure until snap-through buckling occurs.
Table 2. Earthquake pressure causing snap-through buckling for domes designed for gravity load combinations alone
Case Snow load,
number D (m) D=H t (mm) S (kPa) f c0 (MPa) ðpcr Þgse (kPa) pgs (kPa) pe (kPa) av (g)
1 30.48 8 76 2.87 27.6 21.5 2.7 18.8 7.1
2 30.48 16 76 0.57 27.6 4.1 2.2 1.9 0.8
3 91.44 8 140 0.57 27.6 7.6 3.9 3.7 0.9
4 91.44 8 152 2.87 27.6 8.5 4.7 3.8 0.8
5 91.44 16 445 0.57 27.6 27.6 12.3 15.3 1.2
6 91.44 16 445 2.87 27.6 27.4 12.7 14.6 1.2
7 91.44 10 203 2.87 27.6 12.3 6.1 6.2 1.0
8 91.44 10 165 0.57 37.9 8.6 4.6 4.0 0.9
9 91.44 10 178 2.87 37.9 10.6 5.4 5.2 0.9
10 76.2 16 324 0.57 27.6 18.0 8.4 9.5 1.1
Note: 1 kPa ¼ 0.145 psi.
(b)
Fig. 6. (Color) (a) Deflection versus pseudotime; and (b) load versus deflection for Case 2 [D ¼ 30.5 m (100 ft) and H=D ¼ 1=16] with shallow cap
diameter d ¼ 11.11 m (36.44 ft), rise h ¼ 0.18 m (6.94 in.), thickness t ¼ 76 mm (3 in.), snow load S ¼ 0.57 kPa (12 lb=sq ft), and concrete
strength fc0 ¼ 27.6 MPa (4,000 psi), subjected to gravity during pseudotime interval [0, 1], creep and shrinkage during pseudotime interval
[1, 2], and earthquake load during pseudotime interval [2, 3].
Fig. 7. (Color) Radial strain contours plot just before snap-through buckling of Case 2 [D ¼ 30.5 m (100 ft) and H=D ¼ 1=16] with shallow cap
diameter d ¼ 11.11 m (36.44 ft), rise h ¼ 0.18 m (6.94 in.), thickness t ¼ 76 mm (3 in.), snow load S ¼ 0.57 kPa (12 lb=sq ft), and concrete
strength f c0 ¼ 27.6 MPa (4,000 psi). There is visible cracking strain on the outer surface and maximum compressive strain on the inner surface
at the boundary of the cap.
The results show monotonic increasing deformation of the shallow reaches the ultimate strength of compression. The final snap-
cap during the creep and shrinkage time period and the increase in through buckling is accompanied by crushing of the small part
the radius of curvature and stress resulting from such deformation. of the boundary that is in compression. The stresses in the interior
The increase in stress results in increase in the rate of creep. At the of the cap do not reach the capacity of concrete.
point of instability in the third time interval, the rate of increase of
deformation with time becomes quite large, and the center deflec-
tion of the cap is very roughly equal to its thickness. Conclusions
The contour plots of radial strains through the thickness of shal-
low caps at the point of instability show that deep circumferential Snap-through buckling of shallow concrete caps subjected to load
cracks emanating from the top surface are formed all around the combinations that involve gravity (dead and live or snow) and
boundary of the cap, and stress at the bottom surface of the cap earthquake loads has been calculated, accounting for imperfection
(b)
Fig. 8. (Color) (a) Deflection versus pseudotime; and (b) load versus deflection for the shallow cap of Case 7 [D ¼ 91.4 m (300 ft) and
H=D ¼ 1=10], with a shallow cap of diameter d ¼ 25.13 m (82.45 ft), rise h ¼ 0.47 m (18.52 in), thickness t ¼ 203 mm (8 in.), snow load S ¼
0.57 kPa (12 lb=sq ft), and concrete strength fc0 ¼ 27.6 MPa (4,000 psi), subjected to gravity during pseudotime interval [0, 1], creep and shrinkage
during pseudotime interval [1, 2], and earthquake load during pseudotime interval [2, 3].
Fig. 9. (Color) Radial strain contours plot just before snap-through buckling of Case 7 [D ¼ 91.4 m (300 ft) and H=D ¼ 1=10] with shallow cap
diameter d ¼ 25.13 m (82.45 ft), rise h ¼ 0.47 m (18.52 in.), thickness t ¼ 203 mm (8 in.), snow load S ¼ 0.57 kPa (12 lb=sq ft), and concrete
strength f c0 ¼ 27.6 MPa (4,000 psi). There is visible cracking strain on the outer surface and maximum compressive strain on the inner surface at the
boundary of the cap.
effects, geometric and material nonlinearities, and time-dependent constructed concrete domes are consistent with this imperfection;
creep and shrinkage of concrete. The method of analysis developed the geometric imperfection considered in this study was in the
in this paper has been partly validated by comparing the calculated form of a shallow cap with λ p ¼ffiffi4ffi over an area with diameter
creep buckling performance of a shallow concrete dome with pub- d and a radius of curvature of 2Ro . In this way, the buckling
lished test results. strength of spherical concrete domes subjected to gravity and
Bearing in mind that (1) buckling strength of spherical shells earthquake load combinations was governed by the problem of
is very sensitive to imperfections in their constructed geometry, snap-through buckling analysis of a shallow spherical cap of
(2) the minimum snap-through buckling strength of shallow caps specific size and radius of curvature that represent the design
occurs with shallowness
pffiffiffiffiffiffiffi parameter λ ¼ 4 over an area with diam- imperfection geometry. The imperfection was centered at a point
eter d ¼ 4.3 Ro t, and (3) the imperfections observed in the on the dome where the normal pressure from the combined effects
(b)
Fig. 10. (Color) (a) Deflection versus pseudotime; and (b) load versus deflection for the shallow cap of Case 9 [D ¼ 91.4 m (300 ft) and
H=D ¼ 1=10] with a shallow cap of diameter d ¼ 23.51 m (77.13 ft), rise h ¼ 0.41 m (16.2 in.), thickness t ¼ 178 mm (7 in.), snow load S ¼
2.89 kPa (60 lb=sq ft), and concrete strength f c0 ¼ 37.9 MPa (5,500 psi), subjected to gravity during pseudotime interval [0,1], creep and shrinkage
during pseudotime interval [1, 2], and earthquake load during pseudotime interval [2,3].
Fig. 11. (Color) Radial strain contours plot just before snap-through buckling of Case 9 [D ¼ 91.4 m (300 ft) and H=D ¼ 1=10] with a shallow cap
of diameter d ¼ 23.51 m (77.13 ft), rise h ¼ 0.41 m (16.2 in.), thickness t ¼ 178 mm (7 in.), snow load S ¼ 2.89 kPa (60 lb=sq ft), and concrete
strength f c0 ¼ 37.9 MPa (5,500 psi). There is visible cracking strain on the outer surface and maximum compressive strain on the inner surface at the
boundary of the cap.
of horizontal and vertical earthquake acceleration response prestressed concrete tanks that are not very large in diameter and
spectra is maximum. are not in high-seismicity zones when designed for static loading
Snap-through buckling analysis was performed for a set of have the capacity needed for earthquake loading. The concrete
shallow spherical domes derived from a range of typical spherical domes with large diameter, e.g., much larger than 30 m (100 ft)
concrete domes with diameters ranging from 30.48 to 91.44 m (100 in diameter in high-seismicity zones may need to be thickened
to 300 ft) and rise-to-diameter ratios in the range of 1=8 to 1=16; for to meet the design spectral response acceleration. The results also
each dome, the seismic capacity of the dome designed for static showed that the use of higher concrete strength can significantly
loading only is determined. reduce the thickness of the dome for the same earthquake load;
The results of this study have shown that the majority of inversely, the earthquake capacity of a dome can be increased sig-
the spherical concrete domes with a prestressed ring beam over nificantly by using higher concrete strength.
(3.6 × 106 psi), and a Poisson’s ratio of 0.17. It is subjected to Location of Imperfection
gravity load, a snow load of 0.57 kPa (12 lb=sqft), and a horizontal
earthquake load of increasing magnitude until buckling occurs. The ASCE 7 (ASCE 2016) defines earthquake load as the sum of the
buckling occurs when the horizontal seismic acceleration reaches effects of horizontal and vertical seismic forces but implicitly limits
18.5g and the buckle is near the base of the dome, as shown the vertical seismic forces to only those that occur concurrently
in Fig. 12. pffiffiffiffiffiffiffi with the maximum horizontal seismic forces. This appendix shows
The size of the buckle is nearly equal to d ¼ 4.3 Ro t, i.e., equal how the critical pressure demand from the combined gravity and
to the diameter of a shallow cap with shallowness parameter λ ¼ 4. earthquake loads can be determined from load combinations that
The pressure at buckling of 18.5g seismic acceleration is approx- include both horizontal and vertical earthquake loads.
imately equal to the classical buckling strength of a shallow cap of ASCE 7 provides load combinations that involve gravity and
pcr ¼ 0.66Ec ðt=Ro Þ2 for λ ¼ 4. Obviously, if there were an imper- earthquake loads in the form of 1.2D þ 0.2S þ E, where D is dead
fection in the concrete dome where the average radius of curvature load, S is snow load (no other live load is expected to exist on
of Rimp over an area that is nearly equal to the buckle size is greater such domes), and E is earthquake load. Earthquake ground motion
than Ro , one would expect the buckling strength to be reduced by has three orthogonal components. For structures with nonparallel
the ratio of ðRo =Rimp Þ2 . In other words, the buckling strength of a system irregularities, ASCE 7 requires simultaneous application of
spherical concrete dome with imperfection can be approximated by orthogonal ground motions with 100% of the force applied in one
Fig. 12. (Color) Vertical deflection of a buckled deep spherical dome with D ¼ 91.44 m (300 ft), D=H ¼ 8, t ¼ 146 mm (5.75 in.), Ec ¼ 24.8 GPa,
and υ ¼ 0.17, subjected to gravity dead load and snow load with S ¼ 0.57 kPa (12 lb=sq ft), and earthquake with primary seismic force being
horizontal.