Professional Documents
Culture Documents
net/publication/323335750
CITATIONS READS
0 1,173
3 authors, including:
Some of the authors of this publication are also working on these related projects:
Short and Long Term Performance of Concrete Structures Repaired/Strengthened with FRP View project
All content following this page was uploaded by Kamyar Kildashti on 22 February 2018.
Abstract. Steel storage tanks are generally vulnerable to dynamic failure under
strong ground motion. The reason for that is partly because of small thickness of
tank walls which contribute to either elastic buckling or inelastic post-buckling
mechanisms. The tank aspect ratio plays a pivotal role in developing different
failure modes. In order to investigate different failure mechanisms, case study tanks
considering different aspect ratios that are fully anchored at the base are
introduced. Dynamic fluid-structure interaction is utilised based on so-called added-
mass method and tank walls are modelled using general purpose finite element
platform to detect potential failure modes. Seven strong ground motions are
selected from NGA database and spectrally matched with target spectrum.
Incremental dynamic analyses are conducted to identify critical failure modes and
construct fragility curves. The results demonstrate that changing tank aspect ratio
contributes to different modes of failure ranging from elephant foot and diamond
shape to secondary buckling modes.
1 INTRODUCTION
It is well recognised that steel liquid storage tanks are vulnerable to extensive damage after
major earthquakes. Thin-walled buckling is the major source of structural failures because of the
small thicknesses of tank’s wall. Many studies have confirmed that fluid hydrodynamic variations
in the course of seismic events play an important role in triggering local damage and eventual
collapse for this class of structures 1-5. Dynamic buckling phenomenon in steel tanks under strong
ground motions was generally documented in the form of elastic-plastic or elastic buckling,
confirmed through experimental or numerical studies. So called “elephant’s foot” buckling is
attributed to elastic-plastic buckling, which is identified as an outward bulge close to the tank’s
base2, 5. In this failure mode, axial compression due to earthquake overturning moment is basically
compromised by circumferential tensile stresses resulting from both hydrostatic and hydrodynamic
pressures at the bottom of the tank and eventually outward bulge occurs. Diamond-shaped buckling
mode is the second type of possible tank’s failure, which is associated with elastic buckling and is
most likely to happen at small values of hoop stresses 1. The last type of buckling mode, which
†
Woolacotts Consulting Engineers
1
Kamyar Kildashti, Neda Mirzadeh and Bijan Samali
again happens within the elastic range of material behaviour, is developed due to inward
hydrodynamic suction at upper courses of steel tanks where outward hydrostatic pressure is minor 6.
Pioneering research on seismic behaviour of anchored steel storage tanks by means of fluid-
structure interaction concept were conducted by Haroun and Housner 7, Veletsos and Yang 8. In
these studies hydrodynamic pressure is decomposed into impulsive and convective liquid mode.
The latter is related to upper portion of liquid in the container that moves with a long period sloshing
motion, while the former is associated with short period mode coincidentally vibrating with the tank.
There are considerable differences between natural period of impulsive and convective fluid
motion, therefore; these two modes are considered as uncoupled in normal cases.
This study focuses on identification of different failure modes in anchored steel storage tanks
with different aspect ratios, i.e. height to diameter ratios (H/D), under horizontal earthquake
excitations using 3D nonlinear finite element analysis (NFEA). Only impulsive modes of liquid-
structure interaction are considered herein and added-mass technique is utilized to simulate this
phenomenon5, 9. A suite of seven strong ground motions are selected from NGA database 10 and
spectrally matched to target spectrum. Incremental dynamic analyses (IDA) are perfor med to
identify the onset of critical state, when steel shells undergo excessive deformation. Eventually,
fragility curves are determined to relate probability of failure to intensity measure s (IMs).
𝑝𝑖 (𝜉, 𝜍, 𝜃, 𝑡) = 𝐶𝑖 (𝜉, 𝜍)𝜌𝑅 × (𝐴ℎ1 (𝑡) 𝑐𝑜𝑠 𝜃 + 𝐴ℎ2 (𝑡) 𝑠𝑖𝑛 𝜃) (1)
∞
(−1)𝑛 𝜈𝑛
𝐶𝑖 (𝜉, 𝜍) = 2𝛾 ∑ 𝜈 𝑐𝑜𝑠(𝜈𝑛 𝜍) 𝐼1 ( 𝜉)
𝐼1 ( 𝛾𝑛 ) 𝛾 (2)
𝑛=0 𝜈𝑛
(𝐼0 ( 𝛾 ) − 𝜈𝑛 ) 𝜈𝑛2
𝛾
2𝑛+1 𝐻
where 𝜈𝑛 = ( ) 𝜋, 𝛾 = 𝐿; 𝐼0 , 𝐼1 are the modified Bessel functions of zero and first order,
𝑛 𝑅
respectively, 𝐴ℎ1 (𝑡), 𝐴ℎ2 (𝑡) are the orthogonal horizontal ground acceleration time histories in the
𝑟 𝑧
free-field, 𝜌 is the mass density of the fluid, 𝜉 = , 𝜍 = are the non-dimensional coordinates, 𝑅 is
𝑅 𝐻𝐿
the tank’s radius, and 𝑟, 𝜃, 𝑧 are components of cylindrical coordinates with origin at the centre of
the tank and the 𝑧 axis vertical. In Figure 2, the pressure coefficient distribution defined in Eq. (2)
for each of the tank models for 𝜃 = 0 is presented.
2
Kamyar Kildashti, Neda Mirzadeh and Bijan Samali
1600
Freeboard
6× 1500+
600
THK.6
1000
15600
Freeboard
5× 1500+
600
Steel Tank
THK.6
9600
2 ×1500
THK.8
2 ×1500
Steel Tank
THK.10
THK.8
1500
(a) (b)
1800
Freeboard
7× 1500+
600
THK.6
18600
(c)
Figure 1: Tank models; THK=shell thickness (a) tank with H/D=0.62 (b) tank with H/D=1.01 (c)
tank with H/D=1.21
To incorporate fluid-structure interaction in NFEA, impulsive pressure components are normally
simulated by lumped added-mass for dynamic analysis (see Figure 3)5. Technically, added-mass
concept indicates that some portion of the stored liquid can coincidentally accelerate or decelerate
with a tank’s wall, which in turn reproduces the impulsive pressure distribution along the tank’s
height. In fact, added-mass is a comprehensive alternative to Housner’s traditional method 7. In this
study, time-independent lumped added-mass are considered for dynamic simulations.
1.20
Model i
1.00 Model ii
Model iii
0.80
ξ=z/HL
0.60
0.40
0.20
0.00
0.00 0.20 0.40 0.60 0.80 1.00
(Ci)
3
Kamyar Kildashti, Neda Mirzadeh and Bijan Samali
Tank wall
r Added-mass
Rigid spring
Radial roller
Figure 4: Finite element model for each tank type (a) model i (b) model ii (c) model iii
4
Kamyar Kildashti, Neda Mirzadeh and Bijan Samali
and scaled from individual recorded events. For each pair of horizontal ground motion components, a
square root of the sum of the squares (SRSS) spectrum shall be constructed by taking the SRSS of the
5 percent damped response spectra for the scaled components (where an identical scale factor is
applied to both components of a pair). Each pair of motions shall be scaled such that for a period
between 0.2T1 and 1.5T1, (T1= fundamental period of vibration), the average of the SRSS spectra from
all horizontal component pairs does not fall below 1.3 times the corresponding ordinate of the design
response spectrum, target spectrum, by more than 10 percent17. Design spectrum is defined in
accordance with requirements of ASCE7-0517 for site class C with Ss=0.75 and S1=0.3, in which; Ss is
the mapped Maximum Considered Earthquake (MCE) spectral response acceleration at short periods,
and S1 is the mapped MCE spectral response acceleration at a period of 1s. Since multi-directional
ground motions need to be scaled for nonlinear analyses, target spectrum is constructed by multiplying
each individual ordinate of design spectrum by the value of 1.3, satisfying the requirements of ASCE7-
05. It is worthwhile to note that design of the steel tanks is carried out based on response spectrum
analysis, where the simultaneous effects of horizontal ground motion is considered.
In this Study seven pairs of ground motion records, from FEMA P695 far field record set20, have been
selected from a subset of PEER/NGA database10, see Table 1. In Figure 5, comparison between SRSS
of pseudo spectral accelerations obtained from each pair of scaled acceleration time histories and target
spectrum within predefined period range is depicted.
* **
Record Magnitude Rrup Vs30
No. Event Year Station Mechanism
Seq. no. Mw (km) (m/s)
1 RSN125 Friuli, Italy-01 1976 Tolmezzo 6.50 Reverse 16 505
Imperial Valley- El Centro Array
2 RSN174 1979 6.53 strike slip 13 197
06, USA #11
Superstition El Centro Imp.
3 RSN721 1987 6.54 strike slip 18 192
Hills-02, USA Co. Cent
Loma Prieta, Reverse
4 RSN752 1989 Capitola 6.93 15 289
USA Oblique
Cape
5 RSN825 Mendocino, 1992 Cape Mendocino 7.01 Reverse 7 568
USA
Northridge-01, Canyon Country -
6 RSN960 1994 6.69 Reverse 12 326
USA W Lost Cany
7 RSN1111 Kobe, Japan 1995 Nishi-Akashi 6.90 strike slip 7 609
Table 1. Far field record set used for nonlinear dynamic analyses
*
Rrup: Closest distance to rupture plane
**
Vs30: Time-averaged S-wave velocity in top 30 m
Pseudo-spectral Acceleration (g)
Design Spectrum
Median Spectrum-Far Field Set
0.1 SRSS of individual pair
Target Spectrum
0.2T1 1.5T1
0.01
0.01 0.1 1 10
Period (sec)
Figure 5: SRSS between each pair of 7 pseudo spectral accelerations consistent with target
spectrum within period range
5
Kamyar Kildashti, Neda Mirzadeh and Bijan Samali
tanks can be classified as; shell buckling, roof and miscellaneous steel damage, anchorage
failure, and foundation failure. In the current study, shell buckling is defined as most governing
failure mode which is considered for fragility analyses. To recognise dynamic steel shell
buckling, incremental dynamic analyses21 (IDA) are performed to obtain the critical state
corresponding to onset of buckling. Nonlinear time history analyses according to bidirectional
ground motions are conducted, thereby accelerating the model in both x- and y- directions. The
Budiansky and Roth22 criterion is utilised herein to detect the critical state. As multidirectional
accelerations are applied to the models, collection of control nodes which are more vulnerable
to maximum deformations are selected, see Figure 6.
Y-Direction
X-Direction
Selected points
Figure 7: Different types of failure modes (a) model i; diamond-shaped buckling mode for
RSN752, PGA=0.32g (b) model ii; secondary buckling mode for RSN174, PGA=0.31g (c) model
iii; elephant’s foot buckling mode for RSN125, PGA=0.36g
6
Kamyar Kildashti, Neda Mirzadeh and Bijan Samali
PGA (g)
PGA (g)
PGA (g)
Figure 8: IDA curves (a) model i, (b) model ii, (c) model iii
Fragility curves can be obtained as lognormal cumulative distribution functions, having the
logarithmic median value, 𝜇 , and logarithmic standard deviation or dispersion, 𝜎. The mathematical
form of fragility curves is as follows:
ln(𝑖𝑚) − 𝜇
𝐹𝐼𝑀𝑐𝑟 (𝑖𝑚) = 𝛷( ) (3)
𝜎
Where 𝛷 is the conditional probability that the component will be damaged to damage state 𝐼𝑀𝑐𝑟 ,
critical IM, as a function of 𝑖𝑚. 𝜇 , 𝜎 can be calculated by the following expressions 23:
𝑚
𝜇 = ln
𝑣 (4)
√1 +
𝑚2
𝑣 (5)
𝜎 = √ln (1 + 2 )
𝑚
In which, 𝑚, 𝑣 are the median and variance of the non-logarithmized sample values.
Critical IMs are used to develop fragility curves for each tank’s model. Three different fragility
curves are depicted in Figure 9. As can be seen, analytical curves show good agreement with
discrete data points, thereby proving the accuracy of lognormal cumulative density function to
describe fragility of steel storage tanks. The PGA values for 50% probability of exceedance in this
figure for models i, ii, and iii, are 0.34g, 0.26g, and 0.22g. It simply indicates that by increasing the
tank’s aspect ratio, the tank is more vulnerable to onset of failure. More specifically, all tanks have
been initially designed in accordance with API-650 requirements for PGA=0.30g. Nevertheless,
only tank model i shows greater PGA values (0.34g) obtained from fragility analysis compared to
design-based PGA and two the other models ii, iii reveal lower PGA values (0.26, 0.22g,
respectively) according to fragility curves. As a result, the tank’s design as per API -650 might be
underestimated for tanks with higher aspect ratios, despite no clear distinction for various aspect
ratios reported in API-650 design guides.
7
Kamyar Kildashti, Neda Mirzadeh and Bijan Samali
1.00
0.90
0.80
0.70 Model i; Fragility Curve
Model ii; Fragility Curve
P(IMcr≤im)
0.60 50% Model iii; Fragility Curve
0.50 Model i; Data Points
0.40 Model ii; Data Points
0.30 Model iii; Data Points
0.20
0.22g
0.26g
0.34g
0.10
0.00
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80
PGA (g)
7 CONCLUSION
In this paper, different failure modes of tanks with various aspect ratios (model i; H/D=0.62, model
ii; H/D=1.01, and model iii, H/D=1.21) were examined. To this aim, the tanks were initially designed
in accordance with API-650 requirements. Strong ground motions were selected and scaled to
target spectrum according to the procedure reported in ASCE7 -05. IDAs were then conducted by
using NFEAs to obtain critical PGA in which incipient failure as a sudden jump in tank’s shell radial
displacement was detected. Fragility curves were constructed to reveal the dependency of critical
PGAs to probability of exceedance. The following conclusions can be drawn:
1. Three different failure modes were observed during nonlinear analyses including; elephant’s
foot buckling, diamond-shaped buckling, and secondary buckling modes. The latter two
corresponded to elastic buckling behaviour, while the former in dicated the elastic-plastic
buckling close to tanks’ base.
2. Results from fragility analyses revealed that by increasing the aspect ratio, the tanks were more
vulnerable to dynamic buckling failure.
3. Only tank’s model i showed greater PGA values (0.34g) obtained from fragility analysis
compared to design-based PGA and two other models ii, iii revealed lower PGA values (0.26,
0.22g, respectively) according to fragility curves. As a result, the tank’s design as per API-650
might be underestimated for tanks with higher aspect ratios, despite no clear distinction for
various aspect ratios reported in API-650 design guides.
REFERENCES
[1] Brunesi, E.; Nascimbene, R.; Pagani, M.; Beilic, D., Seismic Performance of Storage Steel Tanks
during the May 2012 Emilia, Italy, Earthquakes. Journal of Performance of Constructed Facilities
2015, 29, (5).
[2] Buratti, N.; Tavano, M., Dynamic buckling and seismic fragility of anchored steel tanks by the
added mass method. Earthquake Engineering & Structural Dynamics 2014, 43, (1), 1-21.
[3] Hamdan, F. H., Seismic behaviour of cylindrical steel liquid storage tanks. Journal of
Constructional Steel Research 2000, 53, (3), 307-333.
[4] K. Meskouris; B. Holtschoppen; C. Butenweg; Rosin, J., Seismic analysis of liquid storage tanks. In
2nd INQUA-IGCP-567 International Workshop on Active Tectonics, Earthquake Geology,
Archaeology and Engineering, Corinth, Greece, 2011.
[5] Virella, J. C.; Godoy, L. A.; Suarez, L. E., Dynamic buckling of anchored steel tanks subjected to
horizontal earthquake excitation. Journal of Constructional Steel Research 2006, 62, (6), 521-531.
[6] Rammerstorfer, F. G.; Scharf, K.; Fischer, F. D., Dynamic Fluid-Structure-Interaction and Stability
in Storage Tank Design. Structural Dynamics, Vols 1 and 2 1991, 113-120.
[7] Haroun, M. A.; Housner, G. W., Earthquake Response of Deformable Liquid Storage Tanks.
8
Kamyar Kildashti, Neda Mirzadeh and Bijan Samali