Sie sind auf Seite 1von 8

Transactions, SMiRT

Transactions, 19,19,
SMiRT Toronto,
Toronto,August
August2007
2007 Paper
Paper # ????
# M06/4

. Seismic Fragility Analysis of Liquid Storage Tanks

Dileep Singh1) and A. S. Ahlawat2)

1) Department of Civil Engineering, Indian Institute of Science Bangalore-560012, India


2) Department of Civil Engineering, Indian Institute of Science Bangalore-560012, India. (anuahl@civil.iisc.ernet.in)

ABSTRACT

Elevated spherical and cylindrical tanks are commonly used to store liquids either under pressure or at
atmospheric pressure in various industrial installations, e.g., nuclear power plants, chemical plants, etc. If the stored
liquid is dangerously toxic or inflammable, failure of these tanks is catastrophic. Therefore, safety of these liquid storage
tanks is of paramount importance in the event of extreme natural loading such as earthquakes. A number of approaches
are available for seismic fragility analysis of structures but most of these consider safe shutdown earthquake (SSE)
having a particular peak ground acceleration (PGA). Even for a particular site, earthquakes characteristics vary not only
in amplitude but also in frequency, frequency band and duration. To arrive at a rational estimate of seismic fragility of a
structure, it must be evaluated over a reasonable range of all the parameters to define a realistic seismic excitation for the
particular site. Though, lower and upper limit of the parameters such as frequency, damping, etc., can be decided for a
particular site on the basis of geological conditions, the worst possible seismic excitation depends not only on the site but
also on the structure. In this study, an analytical and numerical methodology has been proposed for seismic fragility
analysis of elevated spherical liquid storage tank. Though, the earthquakes are non-stationary, for development of present
approach, input excitation has been represented by a stationary Kanai-Tazimi spectrum. Parameters (frequency and
damping of the Kanai-Tazimi spectrum) corresponding to critical (worst possible) excitation have been obtained for the
spherical liquid storage tank under consideration. The obtained critical power spectral density (psd) has been used to
obtain conditional probability of failure of the structure for different PGA. The proposed methodology has been
demonstrated for an elevated spherical liquid storage tank for two different sites one is having stiff soil and another is
having soft soil. A significant difference has been observed in the results obtained by proposed approach in comparison
to existing approach for these two types of sites.

INTRODUCTION

Since the failure of large liquid storage structures in industrial installation such as nuclear power plants, is
synonymous to catastrophe, safety of these structures is of paramount importance. However, evaluation of the safety of a
structure is very complex in nature because of uncertainties in loading, material behavior, geometry of the structure and
approximations made in the model used in the analysis. Using a probabilistic approach, uncertainties can be handled
easily and results in more rational and quantitative estimate of the risk [1].
Among all the loads, earthquake loads are typically highly uncertain and very large in magnitude. Therefore,
seismic ground motion has been a very important consideration in evaluation of safety/risk associated with it. A
significant stride has been made by the research community in developing models for realistic representation of seismic
ground motion. Thus, synthetic earthquake can be generated by adjusting the parameters of the model for geological
conditions of the site.
Over the recent developments in computational methods lead to high fidelity simulation and analysis procedures
for complex structure. However, at the same time regulatory agencies have been made more stringent performance,
safety norms and increase in evidence of seismic hazard not envisaged earlier. A number of approaches are available in
the literature for seismic fragility analysis of structures but most of these consider safe shutdown earthquake (SSE) that is
defined by amplitude of the earthquake, i.e., peak ground acceleration (PGA) with a particular frequency, damping and
duration depending on the soil type [2]. However, even for a particular site, earthquakes characteristics vary not only in
amplitude but also in frequency, frequency band and duration. In addition to this earthquake is a non-stationary process.
To arrive at a rational estimate of seismic fragility of a structure, it must be evaluated over a reasonable range of
all the parameters to define a realistic seismic excitation for the particular site. Though, the range (lower and upper limit)
of the parameters such as frequency, damping, etc., can be decided for a particular site on the basis of geological
conditions, the worst possible seismic excitation depends not only on the site but also on the structure.
In present study, an analytical and numerical methodology has been developed for seismic fragility analysis of
elevated spherical liquid storage tank. Though, the earthquakes are non-stationary, for development of present analytical
approach, input excitation has been represented by a stationary Kanai-Tazimi spectrum [3]. Parameters (frequency and
damping of the Kanai-Tazimi spectrum) corresponding to critical excitation have been obtained for the spherical liquid
storage tank under consideration. Here critical excitation is one that results in worst response of the parameter governing
the failure, e.g., maximum stress, maximum strain, etc., while meeting the constraints limits of the excitation parameters
defined for particular site. Thus, obtaining critical excitation for a given site and structure can be formulated an
optimization problem. A reduced order model of dynamics of the spherical tank under ground excitation modeled by a

1
Transactions, SMiRT
Transactions, 19,19,
SMiRT Toronto,
Toronto,August
August2007
2007 Paper
Paper # ????
# M06/4

. Kanai-Tazimi spectrum [3], has been used in this optimization problem. Thus, obtained critical psd is applied to the
model of the structure to obtain conditional probability of failure.
The proposed methodology has been demonstrated for an elevated spherical liquid storage tank given in [4]. A
finite element model of spherical tank with liquid filled to its full capacity has been developed to carryout dynamic
analysis with critical psd excitation and to obtain a reduced order model of the dynamics of the structure. Here, fluid
structure interaction has been considered but only one case where liquid is filled to its full capacity has been used.
Analysis for various partially filled cases is very similar and hence, it is not being presented here. Maximum bending
stress has been used as failure criteria. The proposed approach is equally applicable for all other failure criterion as well.
Two different sites, one is having stiff soil and another is having soft soil, have been considered to demonstrate the
difference in results obtained by proposed approach in comparison to existing approaches. Various parameters and
constraint limits have been used primarily to demonstrate the approach and results will be different for different values of
these parameters, constraint limits and failure criteria.

SEISMIC FRAGILITY ANALYSIS OF THE STRUCTURE

The seismic fragility of a structure may be defined as the conditional probability of its failure as a function of
measure of severity of earthquake, e.g., PGA. The failure could be defined in terms of a limiting value of failure criteria,
e.g., maximum stress and/or maximum strain/displacement, etc. In other words, seismic fragility with respect to a
particular limit state is defined as the conditional probability of failure Pf, that the structural response S exceeds the
structural capacity or resistance R for a given PGA.

Pf | Ap = P( R ≤ S | Ap ) (1)

Where Ap is peak ground acceleration, R is resistance or capacity function and S is response. The reliability of
the structure is given as (1 − Pf | Ap ) .
Equation of motion of a randomly excited, viscously damped linear system is given by

M&y& + Cy& + Ky = F (t ) (2)

Where, M, C and K are mass, viscous damping and stiffness matrices, respectively. y, y& and &y& are unknown
vectors of displacement, velocity and acceleration response, respectively. F(t) is a vector of jointly stationary, zero mean
Gaussian random force. The M, C and K are function of a vector of random variables x, representing uncertainties in the
structural properties. However in the present study structural properties has been assumed to be deterministic, i.e., there
is no uncertainty in the properties of the structure. Response quantities of interest, such as stresses, strains, etc. can be
represented by a vector of random processes, z(t) as

z (t ) = f ( y, y& , &y&, x) (3)

If one or more components of z(t) exceeds the corresponding permissible thresholds zi* , (i = 1,2,3,.....) , during
the time interval [t0, t0+T ]. The permissible thresholds may be modeled as a vector of random variables, which are in
general mutually dependent and non-Gaussian. Performance function may be defined as

g i = zi* − zi (t , x), (i = 1,2,3,.....) (4)

For the structure associated failure modes are generally in series, i.e., structure fails is any one of the associated
limit states is violated. Thus, the probability of structural failure can be given as

[
Pf = 1 − P I ni=1 zi (t , x ) ≤ zi* ; ∀t ∈ (t0 ,t 0 + T ) ] (5)

Here P[.] denotes the probability measure. If zi_max is another random variable defined as
[ ]
zi _ max = max ( zi (t , x)) then Pf = 1 − P I ni=1 zi _ max ≤ zi* . Let N i* ( zi* , t 0 , t 0 + T ) being the number of times
t0 ≤t ≤( t0 +T )

the level z i is crossed in interval (t 0 , t 0 + T ) for each zi (t , x) . For sufficiently high threshold z i , crossing can be
* *

assumed to be rare with zero memory. Since z (t , x) is stationary and Gaussian, N = {N i* ( zi* , t 0 , t 0 + T )}in=1 can be
taken to constitute a vector of mutually correlated Poisson random variables. If the correlation between the components

1
Transactions, SMiRT
Transactions, 19,19,
SMiRT Toronto,
Toronto,August
August2007
2007 Paper
Paper # ????
# M06/4

. of n dimensional vector N can be expressed in terms of a vector of n(n + 1) / 2 mutually independent Poisson random
variables having parameters {λi }i =1
n ( n +1) / 2
, the multivariate extreme distribution is given by

⎡ n ( n+1) / 2 ⎤
Pz1 _ max ...z n _ max x( z1*, ..., z n* | x) = exp⎢− ∑ λi ⎥ (6)
⎣ i =1 ⎦

Parameters {λi }in=(1n+1) / 2 depend on the covariance matrix of z (t | x) and marginal distribution
* *
Pz1_ max ...z n _ max x( z ..., z | x) for zi _ max shown in Eq. 6 follows a Gumbel model given by
1, n

⎡ 1
⎧⎪ 1 ⎛ z * − µ 2
⎫⎪⎤
⎛ γ 2i ⎞2 ⎞
PZi _ max |x ( z ) = exp ⎢− ⎥
T
*
⎜⎜ ⎟⎟ exp⎨− ⎜⎜ i i
⎟⎟ ⎬⎥ (7)
i
⎢ 2π ⎝ γ 0i ⎠ ⎪⎩ 2 ⎝ γ 0i ⎠ ⎪⎭⎥
⎣⎢ ⎦

Where, γ 2i and γ 0i are spectral moments given by γ ji = ∫ ω j S ii (ω | x )dω , (j=0,2), ω is frequency, µi is
−∞
mean of the process z (t | x) with Sii (ω | x ) being its corresponding power spectral density function conditioned on x .
Hence, probability of safety for given random variables conditioned on expected PGA and having zero mean can be
given as
⎡ 1
⎧⎪ 1 ⎛ z 2 ⎞⎫⎪⎤
⎛ γ 2 ⎞2
= 1 − exp ⎢− ⎟ ⎬⎥
T
Pf | Ap ⎜⎜ ⎟⎟ exp⎨− ⎜ − 2 (8)
⎢ 2π ⎝γ0 ⎠ ⎪⎩ 2 ⎜⎝ σ z| Ap ⎟ ⎪⎥
⎣⎢ ⎠⎭⎥⎦

Where, γ 0 = σ z2| A p
is variance of the random variable and γ 2 = σ& z2| A p
is variance of the time derivative of
random variable.

FLUID STRUCTURE INTERACTION

For liquid storage structure consideration of fluid structure interaction is very important. In the present study,
liquid has been considered to be incompressible without any net in/out flow. Small liquid displacement has been
considered, i.e., there is no separation of liquid from structure at the interface. Displacement constraints between liquid
and container have been imposed in the direction normal to interface. This model captures the sloshing modes of the
liquid and effects of the liquid on the dynamics of the structure. However, energy dissipation in the viscous boundary
layer at interface has not been considered.

SEISMIC EXCITATION MODEL

Though, the earthquakes are non-stationary random process, here only strong motion portion of the earthquake
having duration of 20 seconds has been considered to evaluate the seismic fragility. For a short duration strong motion,
process may be assumed to be stationary. Therefore, in the development of present approach a zero mean stationary
random process given by Kanai-Tazimi Spectrum [3] has been used to simulate the earthquake.

2
⎛ω ⎞
1 + 4ξ ⎜ ⎟ 2
⎜ω ⎟ g

S g (ω ) = ⎝ g⎠ S0 (9)
2
⎡ ⎛ω ⎞ ⎤ 2
⎛ ⎞
2

⎢1 − ⎜ ⎟ ⎥ + 4ξ g2 ⎜ ω ⎟
⎢ ⎜⎝ ω g ⎟⎠ ⎥ ⎜ω
⎝ g


⎣ ⎦

Where Sg is earthquake power spectrum, S0 is intensity of the spectrum, ω is frequency, ωg is dominant ground
frequency and ξg damping ratio of the ground. The mean peak ground acceleration Ap can be given as

1
Transactions, SMiRT
Transactions, 19,19,
SMiRT Toronto,
Toronto,August
August2007
2007 Paper
Paper # ????
# M06/4

.
Ap = η g σ g (10)
0.5772
Where η g = C1 +
C1
(11)
1
⎡ ⎡ω gT ⎤ ⎤ 2
C1 = ⎢2 ln ⎢ ⎥⎥ (12)
⎣ ⎣ π ⎦⎦

For zero mean, stationary Gaussian process, variance can be given by


σ g2 = ∫ S g (ω )dω (13)
−∞

The integral in Eq. 3 reduces to

⎡ 1 ⎤
σ g2 = πω g ⎢ + 2ξ g ⎥ S 0 (14)
⎣⎢ 2ξ g ⎦⎥
For finite band of excitation spectrum, limits for integration may be replaced with minimum and maximum
frequency of the spectrum. Using Eqs. 10-14, Intensity of spectrum S0 can be computed for specified PGA level Ap.

CRITICAL SEISMIC EXCITATION

Critical seismic excitation is defined by a psd input that leads to worst possible response of the structure in terms
of the failure criteria, e.g. maximum stress, strain, displacement, etc., subjected to a set of constrains on various
parameters that define the psd. Thus, evaluation of critical seismic excitation can be represented as following
optimization problem to find critical damping parameter (ξg_c), critical frequency parameter (ωg_c) with an objective to

Max z = f ( K , M , C ,........S g ) (2)


Subjected to

ω2
∫ω1
S g (ω ) dω = const. (3)

ξ g _ min ≤ ξ g ≤ξ g _ max (4)

and ω g _ min ≤ ω g ≤ω g _ max (5)

Where z is the vector of response quantity of interest and K, M, C,……. are the parameters of the model of the
structure and material. Sg is the psd input modeled as given in Eq. 1, ξg_min, ωg_min, ξg_max, and ωg_max are minimum and
maximum limits of the damping and frequency parameters of Kanai-Tazimi spectrum. ω1 and ω2 are the lower and upper
cutoff frequencies of the earthquake spectrum.
Since structure is linear and excitation psd model is stationary, a linear scaling can be used to obtain response
corresponding to different PGA. Hence, the optimization problem is not to be run for each and every PGA.

CASE STUDY

An elevated spherical liquid storage tank given in [4] has been used as an example problem to demonstrate the
proposed approach for seismic fragility analysis. A case where liquid filled to its full capacity has been considered in this
case study. Failure criteria used is the maximum bending stress. Permissible maximum bending stress has been adjusted
to account for prestress. Constraint limits used in optimization problem to obtain critical excitation, have been decided
for two different sites, one is having stiff soil and another is having soft soil. Complete details about geometry and

1
Transactions, SMiRT
Transactions, 19,19,
SMiRT Toronto,
Toronto,August
August2007
2007 Paper
Paper # ????
# M06/4

. material properties of the example structure are given in [4]. However, a brief description has been given here for
completeness.

Geometry of the Spherical Liquid Storage Tank


Geometry of the spherical tank has been shown in Fig. 1. Storage capacity of the liquid storage tank is 3864.1
m3 and diameter of spherical tank is 22.07m. The spherical tank is supported on a conical tower. Total height of the tank
from footing to apex of the sphere is 33.40m. Liquid stored in tank is water.

Fig. 1 Geometry of Spherical Liquid Storage Tank [4]

Material properties
Nominal values of the material properties have been taken from [4]. The spherical tank is prestressed and grade
of concrete used is M35 ( f ck = 35MPa ) . Poisson’s ratio has been taken as 0.15. Liquid in the tank is water. Unit weights
of the concrete and water have been taken as 25kN-m-3 and10 kN-m-3, respectively. Structure has been considered to be
linear. As the maximum bending stress has been taken as failure criteria and prestress has not been considered in the
dynamic analysis, maximum permissible bending stress has been reduced to account for stresses due to prestress. In this
example problem maximum permissible stress has been taken as 20 MPa.

Finite Element Model of the Structure


The structure is assumed to be fixed at the footing and seismic excitation is applied as ground acceleration at
footing level. The water inside the tank is assumed to be incompressible. Fluid structure interaction has been modeled
through displacement constraints in direction normal to structure-liquid interface surface. The finite element (FE) model
of the structure has been developed using ANSYS [5]. For spherical tank, cupola and conical supporting structure has
been modeled using shell-43 element and fluid has been modeled using fluid-80 element. Reduced order model of
dynamics of the tank has been extracted using modal reduction, i.e., only the modes with significant participation have
been considered in the reduced order model.

Idealization, Constraint Limits and Critical Excitation


Close observation of the dynamics of the example structure shows that the maximum bending stress (that occur
at the junction of footing and supporting cone) is directly proportional to maximum displacement of the tank structure.
Thanks to cantilever with tip mass type of configuration leading to maximum participation from first bending mode of

1
Transactions, SMiRT
Transactions, 19,19,
SMiRT Toronto,
Toronto,August
August2007
2007 Paper
Paper # ????
# M06/4

. the structure. This provides an opportunity to modify the objective function used to obtain the critical excitation and
decouple the fragility analysis problem into two parts. (i) Computation of excitation psd that maximize the displacement
response, using reduced order model of the dynamics of the structure (ii) Computation of conditional probability of
failure using obtained critical excitation psd input.
Two different sites one having stiff soil and other having soft soil have been considered. The constraint limits
for damping ratio (ξg) and frequency (ωg) for stiff and soft soil are given in Table 1. Constraint on total rms power of the
excitation psd has been enforced by normalizing it to unity (magnitude ‘unity’ is irrelevant as psd can be linearly scaled
for different PGA, any value instead of unity can be used). The lower and upper cutoff frequencies ω1 and ω2 have been
taken as 0 Hz and 30 Hz, respectively.

Table 1 Constraint Limits for Parameter of Kanai-Tazimi Spectrum (ξg, ωg)

Damping ratio(ξg) Frequency (ωg)


Soil Type
Lower Limit (ξg_min) Upper Limit (ξg_max) Lower Limit (ωg_min) Upper Limit (ωg_max)
Stiff Soil 0.6 0.85 1.20 Hz 4.00 Hz
Soft Soil 0.6 0.85 0.25 Hz 2.50 Hz

Parameters corresponding to critical psd obtained are damping ratio (ξg_c)=0.6 and frequency (ωg_c)=2.394 Hz
for both the soil type. A normalized psd for stiff soil, soft soil and critical case has been shown in Fig. 1.

0.4

0.35

Stiff Soil ωg=2.5Hz ξg=0.6

0.3 Critical ωg ξg for stiff and soft soil

Soft Soil ωg=1.2Hz ξg=0.85


S(ω)dω]

0.25

0.2
S(ω)/ [∫ωmax
0

0.15

0.1

0.05

0
0 5 10 15 20 25 30
Frequency (in Hertz)

Fig. 2 Normalized Power Spectral Density (psd)

Results
Results for the above described case study have been shown in Fig. 3 and 4. It is clear from the results that
conditional probability of failure for critical excitation is higher than that is for recommended excitation psd for stiff and
soft soils. Results have been plotted on semi-log scale in Fig. 4 to highlight the results in the region of interest (tail
region). However, the critical excitation psd varies with the constraint limits and in the present case study these values
have been used for demonstration purpose only. A suitable constraint limits for a particular site can easily be obtained
from its geological condition.

CONCLUSIONS

An approach for fragility analysis of liquid storage tanks has been demonstrated where conditional probability
of failure has been computed not only corresponding to various PGA but also considered a critical input spectrum instead

1
Transactions, SMiRT
Transactions, 19,19,
SMiRT Toronto,
Toronto,August
August2007
2007 Paper
Paper # ????
# M06/4

. of a recommended spectrum only. It is clear from the results that a significant difference is observed in safety evaluation
for small variation in input excitation. The proposed approach provides a better (critical) estimates of conditional
probability of failure for a range of input excitations.

1
Stiff Soil ωg=2.5Hz ξg=0.6
0.9
Critical ωg ξg for stiff and soft soil

0.8 Soft Soil ωg=1.2Hz ξg=0.85


Conditional probablity of Failure

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Peak Ground Acceleration (PGA) (in g)

Fig. 3 Seismic Fragility Curves

0
10

-2
10

-4
10
Conditional probablity of Failure

-6
10
Stiff Soil ωg=2.5Hz ξg=0.6
-8
10 Critical ωg ξg for stiff and soft soil
Soft Soil ωg=1.2Hz ξg=0.85
-10
10

-12
10

-14
10

-16
10
0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Peak Ground Acceleration (PGA) (in g)

Fig. 4 Seismic Fragility Curves

1
Transactions, SMiRT
Transactions, 19,19,
SMiRT Toronto,
Toronto,August
August2007
2007 Paper
Paper # ????
# M06/4

.
REFERENCES

1. Gupta, Sayan and Manohar, C.S. “Reliability analysis of randomly vibrating structures with parameter
uncertainties”, Journal of Sound and Vibration, Vol 297, 2006, pp. 1000-1024.
2. Singh, Dileep, Seismic Fragility Analysis of Liquid Storage Structures, Master of Engineering Dissertation, Indian
Institute of Science, Bangalore-560012, India, 2004.
3. Nigam, N.C. and Narayanan, S., Application of Random Vibrations, Narosa Publishing House, New Delhi, 1995.
4. Bhargava, K., Ghosh, A.K., Agrawal, M.K., Patnaik, R., Ramanujam, S., Kushwaha, H.S., “Evaluation of seismic
fragility of structures-a case study”, Nuclear Engineering and Design, Vol. 212, 2002, pp. 253-272.
5. ANSYS, Ansys inc., Southpointe, 275, Technology Drive, Canonsburg, PA, USA, 2005, ver. 10.0.

Das könnte Ihnen auch gefallen