Indian Journal of Engineering & Materials Sciences

Vol. 14, April 2007, pp. 133-145

Seismic behaviour of cylindrical elevated tanks with a frame supporting

system on various subsoil
Ramazan Livaoglua* & Adem Dogangunb
a
Department of Civil Engineering, Karadeniz Technical University, 29000 Gümüşhane, Turkey
b
Department of Civil Engineering, Karadeniz Technical University, 61080 Trabzon, Turkey
Received 7 November 2005; accepted 3 October 2006
The seismic responses of an elevated tank with a frame supporting system on various subsoils are investigated. The
finite element method is used for the modelling of the elevated tank and subsoil system. Fixed-base and elastic media
assumptions are made for subsoil. The tank fluid has been modelled as lumped masses such as impulsive and convective
masses proposed by Housner. An added mass approach is used for fluid-structure interaction by representing both the
impulsive mass and the convective mass. For estimating the seismic response of elevated tanks, response spectrum analysis
with mode superposition is used. The results obtained from modelling the elevated tanks on a fixed base and on an elastic
medium are compared. It has been observed that the seismic response of the elevated tank is altered significantly depending
on the properties of the subsoil. The soil-structure interaction especially affects the impulsive modes and lateral
displacement of elevated tanks rather than the torsional modes. It is proposed that the coupling effect for perpendicular
directions in the axisymmetrical plan geometry should be considered in the seismic design of the elevated tanks.
IPC Code: E03B11/12

Elevated tanks are the structures frequently used in Upsetting circumstances were experienced through
order to store fluid for not only drinking but also for damage to the staging of elevated tanks in some
fire fighting. There has been intensive focus on the earthquakes which occurred in different regions of the
seismic safety of lifeline structures; i.e., tanks in world3-5.
seismic regions; particularly those critical structures Numerous studies have been done for dynamic
that may fail during earthquakes and can potentially
behaviour of liquid storage tanks, most of which are
endanger drinking water, fail to stop large fires and
concerned with ground level cylindrical tanks.
results in substantial economical losses. The 1999
Kocaeli earthquake in Turkey caused significant Contrary to this, very few studies exist related to
structural damage to the Tubras refinery itself and an underground and elevated tanks. It is generally
associated tank farm with crude oil and product assumed that the elevated tanks are fixed to the
jetties. Of 112 tanks on the farm six of varying sizes ground. So, attention is focused on the dynamic
were damaged due to ground shaking and fire. The behaviour of the fluid and/or supporting structure.
consequent fire in the refinery and on the tank farm In the 1950s the concept of analyzing probability of
caused extensive additional damage. Fire started in elevated water tanks as a single degree of freedom
one of the naphtha tanks lasting for three days and system was suggested6. Some codes still recommend
endangered the safety of the whole region. During the this approach7. If the tank is completely full of water,
earthquake this fire initiated as a result of sparks vertical motion of the sloshing is prevented, and
created by bouncing of the floating roof in one of the supporting structure has a uniform rigidity along the
tanks. The sparks ignited the naphtha1. The height. In this regards, the elevated tank may be
Production Index dropped by 12.1% and 9%,
treated as a single degree of freedom system or in
respectively, during August and September of 1999,
resulting in an annual drop of 5% from 1998 level. other words as a normal inverted pendulum. When the
This is largely attributed to the slowdown in fluid in the vessel oscillates, this changes dynamic
production at the TUBRAŞ2. behaviour of the elevated tanks. It is indicated that
All the reasons mentioned above show that with observations of real elevated tanks large errors
this type of structure and its reliability against are involved in using a single-degree-of-freedom
failure under seismic load are of critical concern. system model. Therefore, the methods which consider
______________ sloshing in the elevated fluid storage tanks are
*For correspondence (E-mail: [email protected]) preferred and used in present study.
134 INDIAN J. ENG. MATER. SCI., APRIL 2007

Housner8 proposed the equivalent impulsive mass seismic response of isolated elevated tanks. These
(mass moving firmly with the walls) and convective researchers concluded that the base shear of elevated
mass (sloshing mass) to represent the dynamic liquid storage tank was significantly reduced due to
behaviour of fluid. The impulsive mass is connected isolation and proposed approximate methods which
to tank walls by rigid links, whereas the convective accurately predict the peak response of the isolated
one by springs. A two-mass model is developed by elevated steel tank with significantly less
using this equivalent masses and springs. In this computational efforts. It is clear from the literature;
model, walls were assumed as rigid and the rigidity of very few researchers have presented the seismic
supporting structure characterized by k1 rigidity which behaviour of elevated tanks when compared to tens of
equals to that of the supporting structure for a the studies for ground-supported cylindrical tanks23.
horizontal force applied at the same height as the But Livaoğlu and Doğangün24 proposed a simple
mass. Further, applications of two-mass model for analytical procedure for the seismic analysis of fluid-
elevated tanks have been reported by Sonobe and elevated tank-foundation/soil systems and they used
Nishikawa9 and Shepherd10. Housner’s two-mass this approximation in selected tanks considering fluid-
model has been commonly used for seismic analysis elevated tank-soil/foundation system. Livaoğlu25
of elevated tanks11. performed a comparative study of seismic behaviour
Haroun and Ellaithy3 developed a model which of the elevated tanks by taking both fluid and soil
included an analysis of a variety of elevated rigid interaction effects on the elevated tanks into account.
tanks undergoing translation and rotation; the model Livaoğlu and Doğangün26 summarized simplified
includes sloshing modes; and it assesses the effect of techniques which are able to determine seismic
tank wall flexibility on the earthquake response of response of the fluid-elevated tanks-soil/foundation
elevated tanks. Resheidat and Sunna12 investigated the system.
behaviour of rectangular elevated tank during Soil-structure interaction may be more important in
earthquakes considering soil-foundation-structure elevated tanks due to the fact that most of the masses
interaction. They neglected the sloshing effects on the lumped above the ground and foundation supported on
seismic behaviour of the elevated tanks. Haroun and relatively small area. The problem of soil-structure
Temraz13 analyzed models of two-dimensional X- interaction for ground supported cylindrical tanks was
braced elevated tanks supported on isolated footings addressed by many investigators27,28. But the same
to investigate the effects of dynamic interaction condition is not relevant to the elevated tanks. So, there
between the tower and the supporting soil-foundation is a need to study the fluid-structure-foundation/soil
system. While doing this, they also neglected the interaction for elevated tanks. Therefore, the purpose
sloshing effects. Marashi and Shakib14 carried out an of this study is to investigate the seismic behaviour of
ambient vibration test in order to evaluate the elevated tanks with frame supporting system on
dynamic characteristics of elevated tanks. Dutta et different subsoils.
al.15,16 studied on the comparisons of the supporting
system of elevated tank with reduced torsional Fluid-Structure-Soil/Foundation Interaction
vulnerability and they suggested approximate The seismic analysis and design of elevated tanks are
empirical equations to determine the values of lateral, complicated by fluid, structure, foundation and soil
horizontal and torsional stiffness for different frame interaction. The model shown in Fig. 1 is used to
supporting systems. They also investigated how the investigate these interaction problems by using general
inelastic torsional behaviour of the tank system with purpose structural analysis programs.
accidental eccentricity varied in accordance with the
increasing number of panel and column17. Housner’s equivalent spring-mass model for fluid
18
Asthana and Sridhar added fluid mass to the wall The equivalent spring-mass model proposed by
to investigate seismic behaviour of elevated tank. Housner regarding dynamic behaviour of fluid inside
Dogangün et al.19 studied the efficiency of added a vessel is shown in Fig. 2. It is important to note that
mass approach for fluid-structure interaction in the factors in Housner’s equations were modified by
elevated tanks neglecting subsoil effects. Rai5 Epstein29. The fluid is replaced by an impulsive mass
searched the seismic behaviour and retrofitting of mi and this mass is rigidly attached to the tank wall
reinforced concrete elevated tanks. Shenton & and a convective mass mo to the tank wall by linear
Hampton20 and Shrimali & Jangid21,22 investigated springs whose total stiffness is k2. This fluid-structure
 LIVAOGLU & DOGANGUN: SEISMIC BEHAVIOUR OF CYLINDRICAL ELEVATED TANKS 135

Fig. 1—Considered fluid-structure-foundation/soil interaction model

Fig. 2—Equivalent spring-mass model for fluid

model is used to account for fluid-structure interaction even if the fundamental frequency of the structure is
effects for elevated water tanks in this study. in the vicinity of one of the natural frequencies of
The impulsive and convective masses for a sloshing3.
cylindrical tank are determined from; Sloshing frequency of ω and the stiffness of k2 for a
cylindrical tank are given by;
tanh ‚ (1.74 R h )
mi = mt ; g 1.84 ⋅ h
(1.74 R h ) ω2 = 1.84 tanh ; k2 = mo ω2 …(2)
R R
R
mo = mt ⋅ 0.318 tanh ‚ (1.84 h R ) …(1)
h where g is the ground acceleration. The impulsive and
convective masses are located from the bottom of the
where mt is the total mass of the fluid, R is the inner vessel at the distances hi and ho as shown in Fig. 2,
radius of the vessel and h is the depth of the fluid. respectively; the heights are given by,
Convective masses of additional higher-mode may
also be included if desired. A single convective mass 3 ⎡ cosh (1.84 h R ) − 1 ⎤
is generally used for practical design of elevated hi = h ; ho = ⎢1 − ⎥h
tanks21,30 and higher modes of sloshing have 8 ⎣⎢ 1.84 ⋅ h R ⋅ sinh (1.84 ⋅h R ) ⎦⎥
negligible influence on the forces exerted on the tank … (3)
136 INDIAN J. ENG. MATER. SCI., APRIL 2007

Soil/foundation-structure interaction this figure, the absolute displacement (U) is estimated

The simulation of the infinite medium in the out of the sum of the free field displacement (v) and
numerical method is a very important topic in the added displacement (u).
dynamic soil-structure interaction problems. The From the direct stiffness approach in structural
general method for treating this problem is to divide analysis, the dynamic force equilibrium of the system
the infinite medium into the near field (truncated is given in terms of the absolute displacements, U, by
layer), which includes the geometric irregularity as the following sub-matrix equation:
well as the non-homogeneity of the foundation, and
the far field, which is simplified as an isotropic 0 ⎤ ⎡U& &⎤ ⎡ K
⎡M ss 0 K sc 0 ⎤ ⎡U s ⎤ ⎡ 0 ⎤
homogeneous elastic medium31. The near field is ⎢
s ss
⎢ 0 M cc 0 ⎥⎥ ⎢U& &⎥ + ⎢K K cc K cf ⎥⎥ ⎢⎢U c ⎥⎥ = ⎢⎢ 0 ⎥⎥
modelled using finite elements and the far field is ⎢ c⎥ ⎢ cs

treated by adding some special artificial boundaries or ⎣⎢ 0 0 M ff ⎦⎥ ⎢⎣U&

&⎥ ⎢ 0
f ⎦ ⎣ K fc K ff ⎦⎥ ⎣⎢U f ⎦⎥ ⎣⎢ 0 ⎦⎥
connecting some special elements. The soil is in most … (4)
cases a semi-infinite medium, an unbounded domain,
or so large in extent that the simultaneous modelling where the mass and the stiffness at the contact nodes
together with the structure may be impractical. In a are the sum of the contribution from the structure (s)
dynamic problem, it may be insufficient to prescribe a and foundation (f), and are given by;
zero displacement at a large distance from the
structure, as is routinely done in static32. But
M cc = M scc + M fcc K cc =K (s)
cc + K cc
(f)
… (5)
sufficiently a large soil model can define the soil
structure interaction as was performed in this study
(Fig. 3). Furthermore, reflecting and radiation effects The three-dimensional free-field solutions are
of the propagating waves from the structure- designated by the free field displacements v and
foundation layer may be avoided by an adequately accelerations v&
&. By a simple change of variables it is
large model. It is important to note that as others used now possible to express the absolute displacements U
approximations, it may be said; the artificial and/or and accelerations U& & in terms of displacements u
transmitting boundaries with numerical method like relative to the free-field displacements v as given
finite element or boundary element. Furthermore, below:
reflecting and radiation effects of the propagating
waves from the structure-foundation layer may be ⎡U s ⎤ ⎡us ⎤ ⎡vs ⎤ ⎡U&
&⎤ ⎡u&
s &s⎤ ⎡v&&
s⎤
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢& ⎥ ⎢ ⎥ ⎢ &⎥
avoided by means of these types of boundaries. There &
are different types in frequency or time domain with ⎢U c ⎥ ≡ ⎢uc ⎥ + ⎢vc ⎥ and ⎢U c ⎥ ≡ ⎢u&
&c ⎥ + ⎢ vc ⎥
& …(6)
⎢⎣U f ⎥⎦ ⎢⎣uf ⎥⎦ ⎢⎣vf ⎥⎦ ⎢&&⎥ ⎢u& &⎥ ⎢v& &⎥
different sensitivities.
33
Firstly, Lysmer and ⎣U f ⎦ ⎣ f ⎦ ⎣ f ⎦
Kuhlmeyer developed viscous boundaries using
one-dimensional beam theory and this theory has been
commonly used with the FEM. Then more complex
boundary types are used and developed like damping-
solvent extraction method and doubly-asymptotic
multi directional transmitting boundary.
The most common soil structure interaction (SSI)
approach used for three-dimensional soil structure
systems is based on the added motion formulation.
This formulation is mathematically simple,
theoretically correct, and is easy to automate and use
within a general linear structural analysis program.
Soil/foundation-structure interaction model given in
Fig. 1 is considered in this study. The model is divided
into three sets of node points34. The common nodes at
the interface of the structure and foundation are
identified as “c”; some within the structure as “s”;
and the others within the foundation as “f” nodes. In Fig. 3—Vertical cross-section of the elevated tank.
 LIVAOGLU & DOGANGUN: SEISMIC BEHAVIOUR OF CYLINDRICAL ELEVATED TANKS 137

Eq. (4) can now be rewritten as &

foundation. The terms of v& && &&
x , vy and vz are the free-

field components of the acceleration if the structure is

⎡ M ss 0 0 ⎤ ⎡ u& &s⎤ ⎡ K ss K sc 0 ⎤ ⎡ us ⎤ not present. The column matrices, mi, are the
⎢ 0 ⎥ ⎢ ⎥ ⎢ K cf ⎥⎥ ⎢⎢uc ⎥⎥
0 ⎥ ⎢u& c ⎥ + ⎢ K cs
M cc & K cc directional masses for the added structure only.
⎢
In addition, the formulation is valid for free-field
⎣⎢ 0 0 M ff ⎦⎥ ⎣⎢u&
& ⎥ ⎣⎢ 0
f ⎦ K fc K ff ⎦⎥ ⎣⎢uf ⎦⎥
motions caused by earthquake waves generated from
⎡ M ss 0 0 ⎤ ⎡ v& &s⎤ ⎡ K ss K sc 0 ⎤ ⎡ vs ⎤ all sources. The method requires that the free-field
= − ⎢⎢ 0 M cc ⎥
0 ⎥ ⎢ v&⎢ & ⎥ ⎢
c ⎥ − ⎢ K cs K cc K cf ⎥⎥ ⎢⎢ vc ⎥⎥ motions at the base of the structure be calculated
⎢⎣ 0 0 M ff ⎥⎦ ⎢⎣v&
& ⎥ ⎢⎣ 0 K ff ⎥⎦ ⎢⎣ vf ⎥⎦ before the soil structure interaction analysis. Most
f ⎦ K fc
structural analysis computer programs automatically
= {R} … (7) apply the seismic loading to all mass degrees-of-
freedom within the computer model and cannot solve
If the free-field displacement vc is constant over the the soil-structure interaction problem. This lack of
base of the structure, the term vs is the rigid body capability has motivated the development of the
motion of the structure. Therefore, Eq. (7) can be massless foundation model. This allows the correct
further simplified by the fact that the static rigid body seismic forces to be applied to the structure; however,
motion of the structure is: the inertia forces within the foundation material are
neglected34. To activate the soil-structure interaction
⎡ K ss K sc ⎤ ⎡ vs ⎤ ⎡ 0 ⎤ within a general purpose structural analysis program it
⎢K = … (8)
K ccs ⎥⎦ ⎢⎣vc ⎥⎦ ⎢⎣ 0 ⎥⎦
is only necessary to identify the foundation mass in
⎣ cs order that the loading is not applied to that part of the
structure. The SAP200036 general purpose structural
Also, the dynamic free-field motion of the foundation analysis program has been selected to consider not
requires that; only soil/foundation-structure interaction but also
fluid-structure interaction for the sample elevated
⎡ M fcc 0 ⎤ ⎡ v& &c⎤ ⎡ K fcc K cf ⎤ ⎡ vc ⎤ ⎡ 0 ⎤ tanks. In the model that takes into soil structure and
⎢ ⎥ ⎢ &⎥ ⎢ + ⎥⎢ ⎥ = ⎢ ⎥ … (9) fluid structure interaction effect account, the soil
⎣ 0 M ff ⎦ ⎣ v&f ⎦ ⎣ K fc K ff ⎦ ⎣vf ⎦ ⎣ 0 ⎦
beneath the structure can be modelled with massless-
solid elements. For fluids, it can be modelled by using
Therefore, the right-hand side of the Eq. (7) can be two mass approximation of Housner8 that was
written as modified by Epstein29. So, it would be possible to
model fluid structure-foundation/soil model with
⎡M ss 0 0 ⎤ ⎡ v&&s⎤ general purpose structural analysis programs.
⎢ ⎥⎢ ⎥
R=⎢ 0 M scc 0 ⎥ ⎢v& &c⎥ ...(10) Finite Element Models And System Data
⎢ 0 0 0 ⎥⎦ ⎢⎣ 0 ⎥⎦ R/C elevated tank that has frame supporting system
⎣
with a vessel capacity of approximately 900 m3 in Fig.
Hence, the right-hand side of the Eq. (7) does not 4 is considered. This tank has been accepted as a
contain the mass of the foundation. Therefore, three- standard project by Ministry of Public Works and
dimensional dynamic equilibrium equations for the Settlement of Turkey. The tank vessel is intze type,
complete soil-structure system with damping added, (i.e., below the cylindrical vessel is a conical shell
are of the following form for a lumped mass system35. with a dome shaped tank floor that provides an
economical advantage if not thick floor slabs in
&+ C u&+ K u = − mx v& elevated tanks have to be used). Since it has an
x − my vy − mz vz
&
Mu & && && … (11)
optimal load balancing shape the intze-type vessels
where M, C and K are the mass, damping and are widely preferred5.
stiffness matrices of the soil/foundation-structure Finite element method is used for modelling the
model, respectively, the added, relative elevated tanks as shown in Fig. 3. Degrees of freedom
displacements, u, exist for the soil-structure system were fixed at the base nodes and left free at the others
and must be set to zero at the sides and bottom of the for the one called fixed-base system. Finite element
meshes were used to model the subsoil for the other
138 INDIAN J. ENG. MATER. SCI., APRIL 2007

Fig. 5—Finite element Model of test systems that have three

Fig. 4—Finite element model of soil/foundation-structure-fluid different mesh type of subsoil medium
system.
quadrilateral shell elements. Because of lack of
systems. In the finite element models, massless rotational freedom capability of brick element on the
foundation approach (Fig. 1) was applied to represent soil-structure interaction surface, foundation is
the soil-structure interaction. 3D finite element modelled using the shell element and soil with the
meshes (Fig. 3) are generated and intended to model isoparametric 8 node-brick element that has three
the influence of fluid-structure and soil-structure translational degrees of freedom per node. The
effects on the seismic behaviour of elevated tank. A rotational DOF of the shell elements at the common
parametric study is carried out to determine the nodes with the brick elements are neglected. It should
distance of the boundary from the tank. For this be noted that plate element also can be used easily.
purpose three different models with three different Fluid-structure interaction problems can be
mesh type of soil medium are used. These models and investigated by using added mass, Lagrangian,
mesh types are illustrated in Fig. 5. A parametric Eulerian and Lagrangian-Eulerian approaches in the
study is performed by using transient analysis with finite element method. Added mass approach was
direct integration technique to decide the dimension selected for fluid-tank interaction. Equivalent
of the soil medium of the model and to select the impulsive and convective masses proposed by
mesh type of the soil medium. Firstly, the models Housner were used for the determination of added
were analyzed with 25 m diameter of the soil medium masses. These masses for fluid were calculated as mi =
and then this value was increased as 5 m for each test 591.870 kg and mo = 210.300 kg via Eq. (1). By using
model. Finally, it is seen that the subsoil beyond this Eq. (3), their heights from vessel ground level hi and
boundary does not affect the seismic behaviour of the ho were determined as 3.0 m and 5.26 m, respectively.
tank. In other words, displacements at the nodes on These parameters were obtained for intze-type tank
the lateral boundaries are almost zero. The common vessel by considering it as an equivalent cylindrical
nodes on the interface of the structure and foundation tank. It was shown that errors associated with such an
are free but the soil nodes on the bottom boundaries approximation were small and model parameters
are fixed. It is important to note here that all test corresponding to equivalent cylindrical tank for intze
model illustrated in Fig. 5 give almost same result. tanks could be used for the design purposes5.
So, there is not a disadvantage of them and can be Impulsive mass is added to the unit mass of finite
easily used for the analyses carried out in this study. elements for vessel wall and truncated invert cone
Columns and beams were modelled as frame suitable for the height level of impulsive mass.
elements and vessel walls and truncated cone as Convective mass was connected to the finite element
 LIVAOGLU & DOGANGUN: SEISMIC BEHAVIOUR OF CYLINDRICAL ELEVATED TANKS 139

joints at the ho level by the stiffness of k2 (=623.44 35 0.4 90 D

kN/m) determined by Eq. (2). Medium clay 20 0.4 68 D
Young’s modulus and unit weight of concrete were Response modification factor was selected as 2.
taken to be 32.000 MPa and 25 kN/m3, respectively. Commonly building systems have an average value of
The vessel was filled with the water density of 1.000 6, which is reduced to one-third for elevated tank
kg/m3. Soil conditions, Young’s modulus and systems. This means that the design forces determined
Poisson’s ratios accounted for this study were given for these systems are about three times larger than
in Table 132. Types of subsoil defined in EC8-Part 133 those of a building system having similar dynamic
were considered and the values of Young’s modules for properties.
soil are selected to be suitable for these classes. The damping values for the reinforced concrete
The analyses were carried out taking four different tanks are taken as 5% for impulsive mode and 0.5%
subsoil classes as subgrade medium. Subsoil classes for convective mode as recommended in most
regarded in this study can be classified as subsoil literature29,38. The 0.5% damped spectra for
class of A, B, C, and D defined in the Eurocode-8. convective mode is available in the EC8-Part 139. In
Soil properties accounted for this paper are given in this study, 5% of damping ratio is considered for the
Table 1. The soil characteristics in Table 1 can be models. The design response spectrums are drawn in
calculated with well-known equations Figs 6a and 6b for 5% and 0.5% damping, for subsoil
classes defined in EC-8 Part 1 that is A, B, C and D,
( G = E / 2(1 + ν ) and vs = G / ρ ; where, ρ is the
with structure performance factor of 2. It should be
mass density and G is the shear modulus). noted for convective mode that the elastic spectrum
may be used as the convective mass which mostly
Table 1 — Properties of subsoil considered37,39 does not excite the structure into inelastic range. It
should be explained that a performance factor of 2 is
Subsoil Young’s Poisson Shear Subsoil class
modulus E, ratio, ν wave (from EC 8) considered in the study as it is recommended for
[MPa] velocity elevated tanks by EC-8 Part-139.
(vs) [m/s]
Seismic analysis of the elevated tanks
Limestone 30,000 0.2 2476 A A total of twelve soil-elevated tank-fluid systems
Granite partially 7,000 0.2 1196 A
decomposed 2,000 0.2 639 B
were taken into account for seismic analysis. These
systems were given a name from System I through
Very dense glacial till 500 0.3 307 C
Sandy clay 250 0.3 229 C System XII to symbolize the models. Influence of the
150 0.3 177 D subsoil is ignored for the first system, System-I and a
Hard clay 100 0.3 145 D fixed support is assumed. And for the rest from
75 0.4 132 D System II to System XII, the influence of the subsoil
50 0.4 108 D

Fig. 6—Design response spectrums of Type-1 recommended by Eurocode-8 for subsoil classes of A ( ), B ( ), C ( )
and D (………). (peak ground acceleration also called “design ground acceleration” ag is taken as 0.4 g and response modification factor
is taken as 2)
140 INDIAN J. ENG. MATER. SCI., APRIL 2007

is taken into account, and subsoil is also divided into Response spectrum analyses were carried out using
finite elements having eleven different Young’s SAP200036 package program. First twelve modes are
modulus, and three different Poisson ratios given in taken into account in the modal analysis. SRSS and
Table 1. CQC methods were used for mode superposition and
In this study, to judge the validity of the applied the practically same results were determined from both
model which is named as massless foundation methods as similarly mentioned in the study by
methods, result obtained analysis of 3D finite element Shrimali and Jangid40. From these results estimated
model with artificial boundary or viscous boundary period values, order of the modes and contribution of
are used. For comparison, transient analyses are the modal mass for selected systems are presented in
performed by using the North-South component of Table 2. Because it was aimed for using the response
1999 Kocaeli Earthquake (Yarımca record) for the spectrum to investigate how soil-structure and fluid-
considered elevated tank models. As can be seen from structure interactions change the dynamic behaviour of
Figs 7 and 8, both approximations gave almost same structure with known modal characteristics, it was
result for roof displacements and base shear assumed free field motion at the bottom of foundation
deviations, i.e., while the maximum displacement level. All the models were subjected to the same
obtained from the massless foundation model in A seismic effect. But it is important to note that for design
subsoil class are calculated as 0.087 m, from the other process of such structures considering interaction
more rigorous approximation is evaluated as 0.089, effects researcher or designer must consider wave
similarly for D subsoil class this results are obtained propagation effects by comparison with other free field
0,111 and 0,118 respectively. When the time history analysis results.
of the base shear is evaluated, same comparison can Periods for two modes given in Table 2 have the
be made easily, almost coincided deviation is seen in same value for the 1st and the 2nd sloshing modes and
the illustrations for both A and D subsoil classes. for the 3rd and the 4th impulsive modes with
These illustrations show that using the massless perpendicular directions in the axisymmetrical plan
foundation approximation cannot cause the large geometry of the elevated tank. As can be seen from
errors or misleading result for the scope of Table 2 although the values of the period for sloshing
investigated model in this research. modes increase from 3.68 s to 3.79 s, the values of the

Fig. 7— For two different approximation, variations of the roof displacement in time for two different subsoil classes

Fig. 8— For two different approximations, variations of the base shear force in time for two different subsoil classes
 LIVAOGLU & DOGANGUN: SEISMIC BEHAVIOUR OF CYLINDRICAL ELEVATED TANKS 141

Table 2—Modal properties such as periods, mode orders and % contributions of the model masses

Systems Mod properties Sloshing Horizontal mode for structure Torsional mode HM for CSS * Vertical
Mode mode

System-I: Period, (s) 3.68 1.03 0.19 --- 1.01 0.14 0.14 ---
Fixed-base Mode order 1st & 2nd 3rd & 4th 6th & 7th --- 5th 10th 8 th & 9 th ---
Contribution % 14 2 76 1 0 0 --- 1 0 0 0 ---
System-II: Period, (s) 3.68 1.04 0.19 --- 1.01 0.14 0.15 ---
Limestone- Mode order 1st & 2nd 3rd & 4th 6th & 7th --- 5th 10th 8 th & 9 th ---
E=30 GPa, ν =0.2 Contribution % 14 2 75 1 1 0 --- 0 0 0 0 ---
System-III Period, (s) 3.68 1.04 0.19 --- 1.01 0.14 0.15 ---
GPD2 Mode order 1st & 2nd 3rd & 4th 6th & 7th --- 5th 10th 8 th & 9 th ---
E=7GPa, ν =0.2 Contribution % 14 2 75 1 1 0 --- 0 0 0 0 ---
System-IV: Period, (s) 3.68 1.05 0.19 --- 1.01 0.14 0.15 ---
GPD1 Mode order 1st & 2nd 3rd & 4th 6th & 7th --- 5th 10th 8 th & 9 th ---
E=2 GPa, ν =0.2 Contribution % 14 2 75 1 1 0 --- 0 0 0 0 ---
System-V: Period, (s) 3.68 1.08 0.20 0.14 1.01 --- 0.15 ---
Very dense glacial till Mode order 1st & 2nd 3rd & 4th 6th & 7th 10th 5th --- 8 th & 9 th ---
E=500 MPa,ν =0.3 Contribution % 15 1 76 1 1 0 3 0 --- 0 0 ---
System-VI: Period, (s) 3.69 1.13 0.22 0.15 1.01 --- 0.15 ---
Sandy clay- Mode order 1st & 2nd 3rd & 4th 6th & 7th 10th 5th --- 8 th & 9 th ---
E=250 MPa,ν =0.3 Contribution % 17 0 71 6 0 0 2 0 --- 0 0 ---
System-VII: Period, (s) 3.69 1.19 0.23 0.15 1.01 --- 0.15 ---
Sandy clay)- Mode order 1st & 2nd 3rd & 4th 8th 5th --- ---
E=150 MPa,ν =0.3 Contribution % 17 0 50 26 2 0 2 0 --- 0 0 ---
System-VIII: Period, (sec.) 3.70 1.26 0.25 0.16 1.01 --- --- 0.17
Hard clay- Mode order 1st & 2nd 3rd & 4th 6th & 7th 9 th & 10 th 5th --- --- 8th
E=100 MPa,ν =0.3 Contribution % 15 3 50 26 1 0 2 0 0 --- --- --- 1
System-IX: Period, (sec.) 3.71 1.31 0.26 0.16 1.01 --- --- 0.19
Hard clay- Mode order 1st & 2nd 3rd & 4th 6th & 7th 9 th & 10 th 5th --- --- 8th
E=75 MPa, ν =0.4 Contribution % 15 4 54 22 1 0 3 0 0 --- --- --- 0
System-X: Period, (sec.) 3.72 1.42 0.27 0.17 1.01 --- --- 0.22
Hard clay- Mode order 1st & 2nd 3rd & 4th 6th & 7th 9 th & 10 th 5th --- --- 8th
E=50 MPa, ν =0.4 Contribution % 16 3 58 16 1 0 3 0 0 --- --- --- 0
System-XI: Period, (sec.) 3.74 1.56 0.29 0.17 1.01 --- --- 0.29
Hard clay- Mode order 1st & 2nd 3rd & 4th 6th & 7th 9 th & 10 th 5th --- --- 8th
E=35 MPa, ν =0.4 Contribution % 20 1 69 3 2 0 3 0 0 --- --- --- 0
System-XII: Period, (sec.) 3.79 1.84 0.32 0.18 1.01 --- --- 0.36
Medium clay- Mode order 1st & 2nd 3rd & 4th 7th & 8th 9 th & 10 th 5th --- --- 6th
E=20 MPa, ν =0.4 Contribution % 17 9 44 24 2 1 3 0 0 --- --- --- 0
1: HM for CSS: Horizontal mode for vessel scope shaft.
2: GPD: Granite partially decomposed.

periods for the 3rd and the 4th impulsive modes classes like C and D). In other words, while
significantly increase from 1.03 s to 1.84 s. This contribution of the 3rd and 4th impulsive modal masses
significant increase for the period values may cause for fixed-base system are 76% for excitation direction
different seismic response of elevated tank. and 1% for perpendicular direction to the excitation,
When the contribution of modal masses for the these values for System VIII were 50% and 26%,
analyzed systems are investigated, it can be seen from respectively. So, the coupling effects must be taken into
Table 2 that the contribution of the convective mass for consideration in dynamic analysis of soil-elevated tank-
excited direction of the systems reaches 20%, and the fluid systems, especially, for soft soils.
contributions for impulsive modes range 44-76%. As Subsoil classes were generally defined in earthquake
seen from Table 2, one of the most important result is codes with similar subsoil class definition. But as can be
that the coupling effects for the 3rd and the 4th impulsive seen from Table 1, different Young’s modulus between
modes of the systems in two directions (excitation the intervals from 35 MPa to 100 MPa were proposed
direction and perpendicular to the excitation direction) for hard clay. Furthermore, massive volcanic rocks and
in the axisymmetrical plan geometry can be more hard rock are in the same soil group of A in the Turkish
effective with increasing softness of the soil (subsoil Earthquake Code41. When the soil is considered as an
142 INDIAN J. ENG. MATER. SCI., APRIL 2007

elastic media, changes in properties of the soil like For this reason coupling effect was considered in this
Young’s modules and Poisson ratio affect not only the paper and the results calculated for the two directions is
values of periods but also deterioration on the row
regularities for vibration modes except for torsional
mode. For example, period value of the 8th mode for
Systems of VIII, IX, X and XI in which soil may be
assumed soft is for vertical vibration mode. But this
period value for the other systems is of horizontal mode
of the vessel scope shaft. When first seven systems are
investigated (Table 2.) it can be seen that the 8th and the
9th modes relating to the vessel scope shaft do not
efficiently affect the dynamic behaviour of systems
whereas the 10th mode is more important than the 8th
and 9th modes for design. Furthermore, vertical mode Fig. 9— Variations of the systems periods with the Young’s
only occurs at the 8th and 6th vibration modes of the last modulus of subsoil for sloshing (◊), impulsive (□) and torsional
five systems (defined as soft soil). So, this indicates that modes (∆).
choosing the actual values for the properties of the
subsoil is very important for seismic design of elevated
tanks. Besides, the results may include some important
deficiencies, if only first few modes are included in the
design.
The relationship between periods in the first five
modes and Young’s modulus of soil for the systems can
be seen from Fig. 9 in the logarithmic scale. The
periods for convective and impulsive modes decrease
with increasing Young’s modulus for soil. This
decrease was dramatic in the interval of 35 MPa~500
MPa values of Young’s modulus for impulsive mode.
When the Young’s modulus increase on the interval of
500 MPa to 7 GPa, the same decrease occurs gradually.
For the larger values of Young’s modulus the periods of
Fig. 10—Lateral displacements of elevated tank systems (a) for
all modes have practically the same value as the fixed- excited direction (b) for perpendicular direction to excited
base system. But the decrease of the sloshing periods direction; System I ( ), System III ( ), System VI
(1st & 2nd) is relatively small compared to the period for ( ), System VII ( ), System XI ( ),
impulsive mode as in Fig. 9. Here, the period value for System XII ( )
torsional modes of the elevated tank is not considerably
affected by the properties of the subsoil.
Lateral displacements of the systems I, III, VI, VII,
XI and XII are shown in Figs 10 and 11 to avoid
confusion. Fig. 10a presents the lateral displacements of
the selected soil-elevated tank-fluid systems in excited
direction. Here, the lateral displacement of the fixed-
base system has the smallest value of 59 mm and
System-XI yields the largest value of 173 mm.
Similarly, Fig. 10b presents the lateral displacement
occurred in the perpendicular direction to the excitation
direction. Again the smallest value is obtained for the
fixed-base system as 11 mm, System-XII gives the
largest value of 142 mm in this direction. Thus,
misleading results can be determined as seen from
Figs 10a and 10b, if one rules out the coupling effects.
 LIVAOGLU & DOGANGUN: SEISMIC BEHAVIOUR OF CYLINDRICAL ELEVATED TANKS 143

Fig. 11—Final lateral displacements of the elevated tank systems; times larger than that of fixed-base system and
System I ( ), System III ( ), System VI ( ), allowable displacement was exceeded for System-IV to
System VII ( ), System XI ( ), System XII
( ) XII. In some cases these large displacements may cause
superposed with the both SRSS and CQC techniques. instability of the elevated tanks, although internal forces
Fig. 11 presents one of the final lateral displacements have small values. The first three systems (I to III) have
because SRSS and CQC techniques give the same practically the same lateral displacement.
results in the studied models. As would be expected; the Maximum roof displacements for excitation (x) and
smallest value of 60 mm is recounted for the fixed-base perpendicular to the excitation (y) directions are shown
system and System-XII gives the largest value of 210 in Fig. 12a. Final roof displacement of the elevated tank
mm. The deviation between analysis including the is shown in Fig. 12b. As can be seen from Fig. 12 the
coupling effects and ignoring the coupling effect is coupling effect is more pronounced with the value of
1.7% for the fixed base system, while it is 35% for the Young's modulus less than 500 MPa. Contrary, the
System XII. base shear and overturning moment decreased and
Following comments may be drawn from Figs 10 larger for the values than 7000 MPa, they practically
and 11. Maximum lateral displacement occurs at the top had the same value as was obtained for fixed-base
of the elevated tank for System-XII whereas it occurs at system (Figs 13 and 14).
the intersection of truncated invert cone and vessel wall Maximum base shear forces for excitation (x) and
for fixed-base system. Therefore, deformed shapes of perpendicular to the excitation (y) direction at the
these systems are significantly different from each supporting structure of the elevated tank are presented
other. Maximum displacement for System-XII is 3.5 on normalized form according to the fluid weight in

Fig. 12—Roof displacements of the elevated tank (a) for excitation (x) (◊) and perpendicular to the excitation (y) directions (□) directions,
(b) final roof displacement of the elevated tank.

Fig. 13—Normalized base shears for the elevated tank (a) for excitation (x) (◊) and perpendicular to the excitation (y) directions (□),
(b) normalized superposed base shears of the elevated tank.
144 INDIAN J. ENG. MATER. SCI., APRIL 2007

Fig. 14—Overturning moments for the elevated tank (a) for excitation (x) and perpendicular to the excitation (y) directions (□),
(b) superposed base shears of the elevated tank.

Fig. 13a and superposed values of normalized base values. It may be considered for the roof design
shears are shown in Fig. 14b. Responses for overturning and details.
moments are given in Figs 14a and 4b. (iv) The periods for impulsive modes were
As can be seen from Figs 13 and 14, the values of the significantly influenced by the rigidity of the soil.
base shear and the overturning moment are the highest But effect was getting smaller as Young’s
of all when value of Young’s modulus is equal to 500 modulus increases. The periods of impulsive
MPa. Coupling effect is also important for this type of modes were not influenced by the Young’s
structures. For example the base shear of the system VII modulus if it was higher than 2000 MPa.
increases from 3068 kN to 4135 kN in case of the Additionally, another mode that could be called as
coupling effects are considered and the deviation is torsional mode was not considerably affected by
almost 35%. the properties of the subsoil in all cases.
(v) Any change in Young’s modulus and Poisson
Conclusions ratios of the soil affects not only the values of the
The following conclusions are drawn from the study: periods but also the mode orders. This indicates
(i) Variations of the displacements along the height that considering the first few modes only may
of the elevated tank are rather different from each cause important deficiencies in the design. In fact,
other for the fixed base system and for the systems vertical modes tend to be included within the first
with low Young’s modulus. In fact, it is observed ten modes if the value of the Young’s modulus is
that the elevated tanks supported on an elastic low.
medium having relatively low value of Young’s (vi) The seismic behaviour of elevated tanks supported
modulus may have displacement larger than the on soft soil, especially if the Young’s modulus is
allowable limits. In some cases, the displacements less than 50 MPa, was sensitive to the contribution
are so large that the elevated tanks can loose of modal masses and mode order of the systems.
stability even though the internal forces are small. In this interval truncation of the soil domain may
(ii) The base shear and bending moment at the base of affect the behaviour of the elevated tanks. Effects
supporting structure increases as the Young’s of horizontal, rocking and vertical motion increase
modulus increases and Poisson’s ratio decreases. as the stiffness decreases. It is observed that the
(iii) The periods for convective (sloshing) modes were horizontal and rocking motion are the most
typically long and were less influenced by the significant factors for the elevated tanks. For this
foundation displacement. But this does not mean reason, further studies should be performed in the
that sloshing effects can be ignored for seismic inelastic range to reveal actual seismic response of
design of elevated tanks. The contribution of the elevated tanks supported on such soils.
convective modes may reach to the significant (vii) When the perpendicular modes are evaluated, it is
true that the modes at perpendicular directions are
 LIVAOGLU & DOGANGUN: SEISMIC BEHAVIOUR OF CYLINDRICAL ELEVATED TANKS 145

not taking place in the same time. But earthquake 17 Dutta S C, Jain S K & Murty C V R, J Sound Vib, 242(1)
motion has 3D-components, so if perpendicular (2001) 151-167.
18 Asthana A & Sridhar P, Earthquake analysis of elevated
modes a system have ~15% mass participation as water tanks using SESAM, Proc 4th Int Conf Civil Eng,
being in this study, the perpendicular modes have Tehran, Iran, I 1997, pp 449-457.
critical importance and during the design process 19 Dogangün A, Ayvaz Y & Durmuş A, Earthquake analysis of
this must be considered. This may not be needed water towers, Proc 4th Int Conf Civil Eng, Tehran, Iran, I pp
439-448) 1997.
for a regular building system. But for the system 20 Shenton W H & Hampton P H, J Struct Eng, 125(9) (1999)
investigated in this study. The results show that 965-976.
the coupling effect may be critical. 21 Shrimali M K & Jangid R S, Bull New Zealand Soc Earthquake
The coupling effects for excitation and perpendicular Eng, 36(3) (2003) 141-164.
to excitation direction should be considered in the 22 Shrimali M K & Jangid R S, J Construct Steel Res, 59 (2003)
1267-1288.
seismic design of the soil/foundation-elevated tank-fluid 23 Rammerstorfer F G, Scharf K & Fischer F D, Appl Mech
systems. Otherwise the design may include significant Rev, 43 (1990) 261-281.
deficiencies. 24 Livaoğlu R & Doğangün A, Struct Eng Mech Int J, 21(1),
(2005) 101-119.
25 R Livaoğlu, Investigation of the Earthquake Behavior of
References Elevated Tanks Considering Fluid-Structure-Soil
1 Scawthorn C, Preliminary Report, Kocaeli (Izmit) Interactions, Ph.D Thesis, (in Turkish), Dept of Civil Eng,
Earthquake of 17 August 1999, Multidisciplinary Center for Karadeniz Tech Univ, Trabzon, 2005.
Earthquake Engineering, State University of New York at 26 Livaoğlu R & Doğangün A, J Fluids Struct, 22(3), (2006)
Buffalo, 1999. 421-439.
2 Akgiray V, Barbarosoglu G & Erdik M, The 1999 Marmara 27 Haroun M A & Abdel-Hafiz E A, Int J Soil Dyn Earthquake
Earthquakes in Turkey, Research Report Bogazici Eng, 5(4) (1986) 217-225.
University, 2003, Istanbul Turkey. 28 Haroun M A & Abou-Izzeddine W, J Struct Eng, 118(3)
3 Haroun M A & Ellaithy M H, J Tech Topics Civil Eng, 111 (1992) 783-812.
(1) (1985) 1-15 29 Epstein H I, ASCE J Struct Div, 102 (1976) 1659-1673.
4 Jain S K & Sameer U S, Bridge Struct. Engr, 23 (1) (1993) 1- 30 Eurocode-8, Design of structures for earthquake resistance-
16 Part 4: silos, tanks and pipelines, European Committee for
5 Rai D C, Earthquake Spectrav, 18 (2002) 745-760. Standardization, Final PT Draft, 2004.
6 Chandrasekaran A R & Krishna J, Water towers in seismic 31 Wolf J P & Song C H, Finite-Element Modelling of
zones, Proc.Third World Conf Earthquake Eng, New Unbounded Media (John Wiley & Sons, Chichester), 1996.
Zealand, :IV, pp 161-171, 1954. 32 Nofal E M H, Analysis of Non-linear Soil-Pile Interaction
7 Guide to the analysis design and construction of oncrete- Under Dynamic Lateral Loading, Ph.D. Thesis, University of
pedestal water tower, American Concrete Institute ACI California, Irvine, USA,1998
371R, 1995. 33 Lysmer J & Kuhlmeyer R L, ASCE Eng Mech Div J, 95
8 Housner G W, Bull Seismol Soc Am, 53 (1963) 381-387. (1969) 859-877.
9 Sonobe Y & Nishikawa T, Study on the earthquake proof 34 Wilson E L, Three-dimensional static and dynamic analysis
design of elevated tanks, Proc Fourth World Conf of structures — A physical approach with emphasis on
Earthquake Eng, Chile, pp 11-24, 1969. earthquake engineering, 3rd Ed (Computers and Structures,
10 Shepherd R, J Sound Vib, 24 (4) (1972) 391-396. Inc. Berkeley, California, USA), 2002.
35 Clough R & Penzien J, Dynamics of structures, 2nd Ed
11 Priestley M J N, Davidson B J, Honey G D, Hopkins D C,
(McGraw-Hill, Inc., ISBN 0-07-011394-7), 1993.
Martin R J, Ramsey G, Vessey J V & Wood J H, Seismic
36 SAP2000, Integrated software for structural analysis &
Design of Storage Tanks, Recommendation of a Study Group
design (Computers and Structures Inc, Berkeley, California),
the New Zealand Society for Earthquake Engineering, New
2003.
Zealand, 1986.
37 Bardet P J, Experimental soil mechanics (Prentice Hall,
12 Resheidat R M & Sunna H, Behavior of elevated storage
Upper Saddle River, New Jersey 07458), 1997.
tanks during earthquakes, Proc 3th U.S. Nat Conf Earthquake
38 Dogangün A & Livaoglu R, Struct Eng Mech 17:2 (2004)
Eng, pp 2143-2154. 1986.
203-214.
13 Haroun M A & Temraz M K, Soil Dyn Earthquake Eng, 11 39 Eurocode-8. Design of structures for earthquake resistance-
(2) (1992) 73-86. Part 1.1:General rules — Seismic action and general
14 Marashi, E S & Shakib H, Evaluations of dynamic requirements for structures, European Committee for
characteristics of elevated water tanks by ambient vibration Standardization, 2001.
tests, Proc 4th Int Conf Civil Eng, Tehran, Iran, I, pp 367- 40 Shrimali M K & Jangid R S, J Sound Vib, 275, (1-2) (2003)
373), 1997. 59-75.
15 Dutta S C, Jain S K & Murty C V R, Soil Dyn Earthquake 41 TEC, Specification for Structures to be Built in Disaster
Eng, 19 (2000) 199-215. Areas, Ministry of Public Works and Settlement Government
16 Dutta S C, Jain S K & Murty C V R, Soil Dyn Earthquake of Republic of Turkey, 1998.
Eng, 19 (2000) 183-197.