Engineering Structures 87 (2015) 32–46

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Seismic response factors of reinforced concrete pedestal in elevated

water tanks
R. Ghateh a, M.R. Kianoush a,⇑, W. Pogorzelski b
a
Department of Civil Engineering, Ryerson University, 350 Victoria Street, Toronto, Ontario M5B 2K3, Canada
b
WP Engineering Inc., 1593 Ellesmere Rd., Suite 106, Scarborough, Ontario M1P 2Y3, Canada

a r t i c l e i n f o a b s t r a c t

Article history: Reinforced concrete (RC) elevated water tanks are critical structures that are expected to remain func-
Received 19 October 2013 tional after severe earthquakes in order to serve the water system networks. Despite this significant role,
Revised 30 December 2014 the number of research studies which investigated the nonlinear seismic response of RC pedestals in ele-
Accepted 9 January 2015
vated water tanks is very limited. In the current codes and standards, the seismic response factors are
Available online 30 January 2015
mainly based on engineering judgement. In this paper, a systematic approach is employed to establish
the seismic response factors for a wide range of elevate water tank sizes and RC pedestal dimensions
Keywords:
commonly built in industry. In total, forty-eight model configurations (prototypes) are selected and
Elevated water tanks
Seismic
designed based on current codes and standards. The finite element (FE) method is then used for nonlinear
Reinforced concrete pedestal static (pushover) analysis of the prototypes. The pushover curve of each prototype is developed and the
Finite elements seismic response factors are determined accordingly. The effect of various parameters such as fundamen-
Pushover analysis tal period, height to diameter ratio, seismic design category, and tank size on the seismic response factors
Cracking pattern of elevated water tanks is evaluated. Furthermore, the cracking propagation pattern in RC pedestal is
Seismic response factors studied. The result of the study shows that the tank size is a critical parameter affecting the seismic
response of elevated water tanks. It is recommended not to use the same seismic response factors for
all RC elevated water tanks regardless of the tank size. In addition, two different patterns of cracking
depending on the height to diameter ratio of the pedestal are detected and discussed.
Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction Being considered as an important element of lifelines, elevated

Elevated water tank is a water storage facility supported by a
tower and constructed at an elevation to provide useful storage
and pressure for a water distribution system. These structures rely
on hydrostatic pressure produced by elevation of water and hence
are able to supply water even during power outages. This feature of
elevated water tanks becomes more critical in case of power out-
age after severe earthquakes.
In general, the tower structure of the elevated water tanks could
be classified as four types of reinforced concrete frame, steel frame,
masonry pedestal and reinforced concrete (RC) pedestal. This study
focuses on the last group in which the tank is mounted on top of a
RC pedestal. The tank may be constructed from steel or concrete.
As this study only focuses on the nonlinear seismic response
behavior of RC pedestals, the type of tank does not affect the
results.
⇑ Corresponding author. Tel.: +1 416 979 5000x6455; fax: +1 416 979 5122.
E-mail addresses: rghateh@ryerson.ca (R. Ghateh), kianoush@ryerson.ca (M.R.
Kianoush), wpeng@bellnet.ca (W. Pogorzelski).
water tanks are expected to remain functional after severe ground
motions to serve as a provider of potable water as well as firefight-
ing operations. Failure or malfunction of these infrastructures dis-
rupts the emergency response and recovery after earthquakes.
Elevated water tanks have not performed up to expectations in
many earthquakes in the past. The poor performance of these
structures in the past earthquakes such as Chile 1960 [1] (due to
design deficiency), Manjil-Roudbar 1990 (out of date design stan-
dards) [2], Jabalpur 1997 and Gujarat 2001 [3] has been reported
in the literature. Extent of damages has been ranging from minor
cracks in the pedestal up to complete collapse of the entire
structure.
There are many grounds that could explain this undesirable
performance. Configuration of these structures which resembles
an inverse pendulum, lack of redundancy in RC pedestal, very
heavy gravity load (comparing to conventional structures) and
poor construction detailing are among the major contributors. Cur-
rently ACI 371R-08 [4] is the only guideline in North America that
specifically addresses the structural design aspects of elevated
water tanks with RC pedestal. This guideline refers to ACI 318-08

http://dx.doi.org/10.1016/j.engstruct.2015.01.017
0141-0296/Ó 2015 Elsevier Ltd. All rights reserved.
 R. Ghateh et al. / Engineering Structures 87 (2015) 32–46
[5] in many occasions for the design and construction of RC pedes-
tal and foundation.
The seismic response of both concrete and steel tanks has been
extensively investigated by means of experimental and numerical
methods. Such studies date back to as early as 1940s and later by
works of Housner [6] and other researchers such as Kianoush and
Ghaemmaghami [7], Moslemi and Kianoush [8], El Damatty et al.
[9] and Kianoush and Chen [10]. On the other hand, although the
RC pedestal is an important part of the elevated water tank struc-
ture, their seismic response has been the subject of only a handful
of research studies.
In one of the earliest studies on seismic response of elevated
water tanks, Shepherd [11] validated the accuracy of the two mass
representation of the water tower structures by comparing the
theoretical results to the results of a dynamic test on a prestressed
concrete elevated water tank. The comparison of the theoretical
and experimental tests proved the efficiency and acceptable accu-
racy of the theoretical two mass modeling of elevated water tanks.
Steinbrugge and Rodrigo [1] investigated the performance of
elevated water tanks during the 1960 Chile earthquake. A
4000 m3 elevated water tank which was empty at the time of
earthquake received vertical cracks all over the height of pedestals
midway between the fins as shown in Fig. 1(d). The pedestal height
and diameter were 30 m and 14.5 m respectively.
Memari and Ahmadi [2] investigated the behavior of two con-
crete elevated water tanks during the 1990 earthquake of Manjil-
Roudbar. They concluded that although the tanks were designed
based on the standards of the construction time, the design loads
were almost one fifth the design loads of the current standards.
They also concluded that the sloshing and P–D effect was very
minor in elevated water tanks. Fig. 1(c) shows a 1500 m3 RC ele-
vated water tank that collapsed during this earthquake. There were
also two 2500 m3 elevated water tanks which were empty during
this earthquake and only received minor cracks at base of the ped-
estal above the openings.
Rai [3] studied the performance of elevated tanks in the 1997
Jabalpur and Bhuj earthquake of 2001. It was concluded that with
low axial load, small longitudinal steel ratio, and a thick wall,
acceptable ductility is gained for RC pedestals. In addition, the con-
crete jacketing was performed on an RC pedestal as a retrofitting
strategy. The RC jacket was shown to enhance lateral strength
and ductility of the pedestal by changing the failure mode from
the concrete crushing to tension yielding.
Dutta et al. [12] studied the dynamic behavior of elevated water
tanks (both RC pedestal and frame) with soil structure interaction
by means of finite element analysis and small scale experimenta-
tions. This study concluded that generation of axial tension in
the tower structure should be commonly expected in the empty-
tank condition, while base shear is principally governed by full
tank condition. Furthermore, the effect of soil–structure interac-
tion was shown to produce considerable increase in tension at
one side of the staging in comparison to fixed support condition.
Moslemi et al. [13] employed the finite element technique to
investigate the seismic response of liquid-filled tanks. The free
vibration analyses in addition to transient analysis using modal
superposition technique were carried out to investigate the
fluid–structure interaction problem in elevated water tanks. The
computed FE time history results were compared with current
practice and very good agreement was observed.
In this study, the finite element (FE) method is employed to
investigate the nonlinear seismic response of RC pedestals in ele-
vated water tanks. Multiple prototypes (models of elevated water
tanks designed based on provisions of code) in accordance with a
number of selection criteria are developed. The prototypes’ dimen-
sions and sizes are selected based on the most widely constructed
tank sizes and pedestal heights. In total, 48 prototypes are
33
designed and analyzed based on the requirements of ACI371R-08
[4], ASCE/SEI 7-2010 [14] and ACI 350.3-06 [15]. Each prototype
is designed for two levels of high and low seismicity. A finite ele-
ment model is developed for each prototype. Subsequently, push-
over analysis is conducted on each FE model. By extracting the
load–deformation results of the pushover analysis, the pushover
curves are generated and the seismic response factors are deter-
mined. Furthermore, the cracking propagation patterns which are
developed in the process of pushover analysis will be presented.
These patterns are compared and categorized based on the geom-
etry and dimensions of the elevated water tanks.
The main objective of this study is to provide a better under-
standing of the nonlinear seismic response of RC pedestals. All
practical tank sizes and pedestal height and diameters are included
in this research in order to develop a comprehensive database for
the seismic response factors of RC pedestals in elevated water
tanks. Furthermore, studying the cracking patterns helps detect
the location of major damages of RC pedestal when subjected to
seismic loads. The results of the study show that the tank size
has a significant effect on seismic response factors of elevated
tanks. In addition, height to diameter ratio of RC pedestal is an
important parameter that affects the seismic response behavior
of elevated water tanks.
2. Defining the study group
The main criteria for selecting the study group are pedestal
height, tank capacity, site seismicity and response modification
factor. Generally, the effect of structural plan configuration must
also be included as a criterion. This is not required in the case of
the elevated water tanks as the plans of all structures are identical
in shape (circular RC wall). In addition, due to the symmetrical plan
of pedestal, only one direction of applying lateral load is adequate.
According to ACI371R-08 [4] common tank sizes in elevated
water tanks range from 0.5 to 3 Mega gallon (Mgal) and RC pedes-
tal heights range from 8 to 60 m (1 Mgal = 3800 m3). Four pedestal
heights of 15, 25, 35 and 45 are determined to be investigated.
Majority of the pedestals constructed in industry are in this range.
Four tank sizes of 0.5, 1, 2 and 3 Mgal which are considered to be
the most widely built tank sizes are selected as well. The pedestal
wall thickness and diameter are mainly functions of the tank size.
The site seismicity affects the design response spectrum and
therefore the seismic design base shear (Vd). The study group will
be investigated for two levels of high and low seismicity. The
mapped risk-targeted maximum considered earthquake (MCER)
spectral response acceleration parameter at short period (Ss) and
1-s period (S1) are selected based on the upper and lower bounds
as required by ASCE/SEI 7-2010 [14] and are employed for deter-
mining the design spectral response acceleration (Sa). Accordingly,
the selected values are Ss = 1.25 and S1 = 0.5 for high seismicity and
Ss = 0.25 and S1 = 0.1 for low seismicity. The site class ‘‘C’’ is chosen
for designing all prototypes. The design earthquake spectral
response acceleration parameter at short period (SDS) and 1 s per-
iod (SD1) are calculated to be 0.84 and 0.44 respectively for the high
seismicity category. These values are determined as SDS = 0.2 and
SD1 = 0.11 for the low seismicity. Subsequently, the design base
shear (Vd) is calculated based on the equivalent lateral force proce-
dure of ACI371R-08 [4].
According to ASCE/SEI 7-2010 [14], response modification fac-
tor (R) of elevated water tanks is either R = 2 or R = 3 depending
on the special seismic detailing provided in the construction of
RC pedestals. For the R = 2 prototypes no special detailing is
required. On the other hand, in the R = 3 prototypes, the special
detailing according to provisions of ACI 318-08 [5] must be pro-
vided which results in more concrete confinement. The prototypes
34 R. Ghateh et al. / Engineering Structures 87 (2015) 32–46

Steel tank

RC Pedestal Center of gravity

of the tank

Pedestal diameter

Fixed base

(a) (b)

(c)

(d)
Fig. 1. Configuration of elevated water tank (a) RC pedestal elevation, (b) FE model, (c) collapse of 1500 m3 RC elevated water thank in Manjil-Roudbar earthquake (photo
credit: NOAA/NGDC, M. Mehrain, Dames and Moore), (d) 4000 m3 elevated water tank which developed vertical cracks in 1960 Chile earthquake, adapted from Steinbrugge
[1].

are designed for both seismic response factors and effect of each
value is studied separately.
By combining the above selection criteria, 48 prototypes are
developed as shown in Table 1. For each prototype, the specific
design aspect and dimensions are given in this table. The proto-
types are categorized into two main groups. The first group is
designed for R = 2 and consists of 24 prototypes. The second group
is identical to the first one except all the prototypes are designed
with the assumption of R = 3. The seismic base shear is calculated
by employing equivalent lateral force method as described in
ACI371R-08. The single mass approximation as prescribed in
Appendix A of ACI371R-08 [4] is implemented for finding the nat-
ural period of the structure.
Each prototype has been assigned a finite element model iden-
tification number (FE model ID). The first term represents the ped-
estal height and the second and third terms stand for the seismicity
 R. Ghateh et al. / Engineering Structures 87 (2015) 32–46 35

Table 1
Dimensions and specifications of prototypes.

Prototype Pedestal Design load level FE model ID dw hr R=2 R=3 R=2 R=3
no. height (m) (m) (mm)
Seismic Design Gravity qv (Pedestal qv (Pedestal Design base Design base
Category (SDC) load base) (%) base) (%) shear (MN) shear (MN)
P-1 15 SDC high Low 15-H-0.5 8.6 300 1.99 0.87 14.7 9.8
P-2 Medium 15-H-1 12 350 1.52 0.72 29.3 19.6
P-3 High 15-H-2 14.5 380 1.51 0.73 53.7 35.8
P-4 SDC low Low 15-L-0.5 8.6 200 0.50 0.50 3.3 2.2
P-5 Medium 15-L-1 12 250 0.50 0.50 6.7 4.5
P-6 High 15-L-2 14.5 250 0.50 0.50 12.4 8.3
P-7 25 SDC high Low 25-H-0.5 8.6 300 2.36 1.06 11.2 7.5
P-8 Medium 25-H-2 14.5 380 2.62 1.08 48.6 32.4
P-9 High 25-H-3 20 400 2.20 0.92 84.9 56.6
P-10 SDC low Low 25-L-0.5 8.6 200 0.50 0.50 2.2 1.4
P-11 Medium 25-L-2 14.5 250 0.50 0.50 9.5 6.3
P-12 High 25-L-3 20 300 0.50 0.50 19.5 13.0
P-13 35 SDC high Low 35-H-0.5 8.6 300 2.11 0.93 8.0 5.4
P-14 Medium 35-H-1 12 350 2.29 1.11 19.8 13.2
P-15 High 35-H-3 20 400 2.40 0.92 67.5 45.0
P-16 SDC low Low 35-L-0.5 8.6 200 0.50 0.50 1.5 1.0
P-17 Medium 35-L-1 12 250 0.50 0.50 4.0 2.7
P-18 High 35-L-3 20 300 0.50 0.50 14.1 9.4
P-19 45 SDC high Low 45-H-0.5 8.6 300 1.86 0.81 6.2 4.1
P-20 Medium 45-H-1 12 350 2.06 0.99 15.2 10.2
P-21 High 45-H-3 20 400 2.08 0.88 51.7 34.5
P-22 SDC low Low 45-L-0.5 8.6 200 0.50 0.50 1.2 0.8
P-23 Medium 45-L-1 12 250 0.50 0.50 3.0 2.0
P-24 High 45-L-3 20 300 0.50 0.50 10.8 7.2

category and tank size respectively. Therefore the FE model ID 35-
H-1 represents an elevated water tank with a pedestal height of
35 m and tank size of 1 Mgal which is located in a high seismicity
zone.
According to Table 1, the calculated pedestal wall thickness (hr)
ranges from 200 mm to 400 mm. The minimum wall thickness is
restricted to 200 mm based on requirements of ACI371R-08 [4].
The pedestal diameter (dw), is generally determined based on the
tank size and ranges from 8.6 m to 20 m. Vertical reinforcement
ratio at the base of RC pedestal (qv) is provided in Table 1 as well.
3. Finite element model
3.1. Theory and assumptions
The general purpose finite element program ANSYS is employed
for finite element modeling of pedestals. ANSYS is capable of mod-
eling reinforced concrete using SOLID 65 element. The reinforced
concrete element is nonlinear by nature due to the cracking of con-
crete under tension and requires an iterative solution. The SOLID
65 element can take into account the nonlinear behavior of con-
crete and steel such as plastic deformation, cracking in three
orthogonal directions and nonlinear stress–strain response under
different loading stages. Failure of concrete material could be due
to cracking or crushing. ANSYS employs William and Warnke fail-
ure criterion [16] in order to detect failure of reinforced concrete
elements.
The elevated water tank structure could be divided into three
substructures including the tank, pedestal and foundation. This
study is focused on the nonlinear response of the RC pedestal sub-
structure and therefore a number of simplifications are made for
modeling the other two substructures. The tank itself consists of
the vessel (either steel or reinforced concrete) and the liquid
inside. This system is replaced with a mass element located at
the center of gravity of the tank as shown in Fig. 1. The mass is cou-
pled with the RC pedestal structure.
A rigid reinforced concrete slab is constructed at the bottom
layer of the tank and supports the vessel contents inside the
pedestal. This slab is placed on top of the ring beam. The slab might
also have a dome, flat or cone shape configuration. The ring beam
and the RC slab together constrain the top layer of the RC pedestal
as a rigid diaphragm. This is reflected in the finite element model
by coupling horizontal degrees of freedom at the top of RC pedes-
tal’s FE model. In all of the FE models, the foundation is assumed to
be rigid and shaft wall is fixed at the level of foundation. Despite
the differences between the dynamic characteristics of empty
and full tank conditions, this study shows that the empty tank con-
dition is not critical for determining overstrength and ductility fac-
tors. This is due to the ACI371R-08 [4] minimum reinforcement as
well as shaft wall thickness requirements. These minimum design
requirements provide the RC pedestal structure with a lower
bound for maximum base shear (Vmax) and displacement ductility
(l). For this reason, while in the empty tank condition the design
base shear is significantly smaller than full tank condition, the
maximum base shear and displacement ductility do not decrease
with the same proportion. Consequently, the overstrength and
ductility factor are higher and therefore do not govern for the
empty tank condition. As a result, the full tank condition is the
most critical state when the structure is subjected to seismic loads
and the FE models are considered to be in full tank condition for
conducting the pushover analysis.
3.2. Reinforced concrete material nonlinearity
A number of mathematical models for simulating the stress–
strain curve of concrete are proposed and exist in the literature
[17,18]. The stress–strain model proposed by Mander et al. [19]
is selected for modeling concrete in this study. The model is dis-
played in Fig. 2(a) for two conditions of confined and unconfined
concrete. The unconfined model is adjusted by concrete compres-
sive strength ( f 0c) and concrete strain at f 0c (e0 c). If the confined
model is employed then three other parameters f 0cc (compressive
strength of confined concrete), e0 cc (concrete strain at f 0cc) and ecu
(ultimate concrete strain for confined concrete) must be defined
as well. f 0cc, e0 cc and ecu are determined based on the type, size
and spacing of confinement.
36 R. Ghateh et al. / Engineering Structures 87 (2015) 32–46

In all the FE models of the study group, the following values for FV (kN)
concrete compressive strength and corresponding strain (depend-
ing on the confinement ratio and horizontal reinforcement spac- FH (kN)
ing) are employed: f 0c = 35 MPa, e0 c = 0.002, f 0cc = 1.2 f 0c to 1.43 f 0c
and e0 cc = 0.004–0.0063 respectively. Shear transfer coefficient for
an open and closed crack as well as stiffness multiplier for cracked
tensile condition are taken as 0.2, 0.5 and 0.6 respectively. These A A
values are required for adjusting SOLID65 element stiffness matrix Section A-A
in ANSYS. They are determined by comparing the FE modeling
results to a number of experimental test results one of which is 1500 mm
125 mm
presented in the following section.
The stress–strain curve of the steel rebar is determined based
750(mm)
on steel grade and rebar sizes. Fig. 2(b) depicts the stress–strain
model for steel rebar proposed by Holzer et al. [20] which is
implemented in this study. This model consists of three branches
of linear, perfectly plastic and a strain hardening portion. The
calibration of the curve is based on the grade and size of steel rein-
forcement. The stress–strain curve could be calibrated by defining
the four strain stages and corresponding stresses. In Fig. 2(b), the Fig. 3. Geometry, loading and section of RC wall samples, Mickleborough et al. [21].
terms esh, eu and er represent strain at the onset of strain hardening,
strain at maximum steel strength and strain at steel failure respec- Fig. 3 shows the geometry, loading and the typical section of the
tively. The values of eu = 0.09, esh = 0.0125 and fy = 400 MPa are shear wall samples. In all tests the walls were first loaded with a
employed for modeling steel rebars in the FE models of the study constant vertical load (Fv). Next, the lateral load (FH) was applied
group. progressively until the failure occurred. The two selected walls
are samples SH-L and SH-H. Table 2 provides material properties,
wall dimensions and vertical and horizontal reinforcement ratios
4. Verification of the reinforced concrete finite element model
(qver and qhor respectively) of the two specimens.
The comparison between the finite element and experimental
The capability of the proposed FE model in predicting the non-
results is shown in Fig. 4. The graphs demonstrate excellent agree-
linear response, load–deformation properties and ultimate strength
ment between the load–deflection patterns of FE and experimental
prior to failure of the reinforced concrete material is validated in
results. In addition, the SOLID65 FE model is capable to estimate
this section. The geometry of the RC pedestal resembles a shear wall
accurately the maximum lateral strength of the RC shear walls.
with circular plan thus makes the shear wall a good choice for ver-
The difference between the actual test result and FE estimation
ifying the finite element model. Mickleborough et al. [21] con-
of ultimate lateral strength is limited to less than 5% for both tests.
ducted an extensive experimental and theoretical study on RC
Moreover the FE model is satisfactorily predicting the final drift of
shear walls with a variety of height to width ratios. Two of the RC
the walls before failure. The comparison of the finite element and
wall specimens with a height to width ratio of two are selected
laboratory results indicated the capability of the proposed finite
for further investigation and validation of FE model.
element method in modeling the load–deflection behavior at dif-
In the selected experimental tests the load is applied until the
ferent stages of loading procedure.
structure reaches the failure stage and partial or global collapse
occurs. This is very similar to the procedure of pushover analysis
of RC shaft walls. The entire loading procedure and corresponding 5. Pushover analysis of FE models
response of the finite element model is recorded and compared to
the actual load–deflection results of the tests. The vertical load 5.1. Pushover analysis method
which is first applied resembles the weight of the tank and the
gradually increasing lateral load simulates the force function in In general there are two main categories of pushover
the pushover analysis. analysis known as ‘‘Conventional pushover analysis’’ and ‘‘adaptive

(a) Confined concrete model

(b)
f ’cc
fu
Unconfined concrete model

f ’c
fy Strain hardening
Concrete stress, f

and softening
Steel stress, f s

Linear part
Curve part

ε’c 2ε’c εu ε’cc εcu εy εsh εu εr

Concrete strain, ε Steel strain, εs

Fig. 2. Idealized stress–strain curve models (a) confined and unconfined concrete and (b) steel rebar.
 R. Ghateh et al. / Engineering Structures 87 (2015) 32–46 37

Table 2
Properties of shear wall specimens SH-L and SH-H.

Wall ID Height (mm) Width (mm) Thickness (mm) f 0c (MPa) fy (MPa) Ec (MPa) Es (MPa) qver (%) qhor (%)
SH-L 1500 750 125 44.7 460 23,788 2e5 1.17 0.39
SH-H 1500 750 125 56 460 28,990 2e5 1.17 0.39

Fig. 4. Comparison between the finite element and experimental results (a) wall SH-L specimen and (b) wall SH-H specimen.

pushover analysis’’ [22]. Selection of the proper method of push-
over analysis highly depends on the configuration of the structures.
In an extensive investigation, Papanikolaou et al. [23] concluded
that in general the adaptive pushover was not providing consider-
able advantages over the conventional pushover analysis. Although
adaptive analysis can demonstrate better performance comparing
to conventional analysis for irregular structures, this advantage is
not valid for all cases. In the case of symmetrical structures with
no specific irregularities in configuration, conventional pushover
analysis will result in adequate accuracy. For the purpose of this
study, since elevated water tanks are symmetrical structures and
have no irregularities in the plan, the conventional method is
selected over adaptive.
The procedure of performing the pushover analysis in this study
is adapted from FEMA 273 [24]. First, the gravity loads including
weights of tank, stored water, pedestal wall and other equipments
is applied to the FE model. Next a gradually increasing lateral load
is applied to the model until the structure collapses. Since most of
the weight in an elevated water tank is concentrated in the tank
and that the modal mass participation factor based on modal anal-
ysis of the first mode is above 90%, the lateral load is applied with a
load pattern similar to the first (fundamental) mode shape.
5.2. Bilinear approximation of pushover curves
structure. This would suggest a more conservative and yet realistic
definition of Dmax for elevated water tanks. In this study, the max-
imum displacement Dmax is defined at the onset of stiffness reduc-
tion when base shear reaches its peak value. In fact, Dmax is defined
at Vmax as this point represents the onset of stiffness reduction.
Due to the nonlinear characteristics of reinforced concrete
structures, which involves cracking and crushing of concrete as
well as yielding of steel, determining the global yield displacement
(Dy) is a complicated task. In this study three requirements are
combined and implemented simultaneously for finding the effec-
tive yield point. The first one is the relationship suggested by FEMA
P695 [26] as shown in Eq. (1):
V max h g i
Dy ¼ C 0 ðmaxðT; T 1 ÞÞ2 ð1Þ
W 4p2
where W = weight of structure, T = fundamental period of structure
calculated based on code, T1 = fundamental period of structure cal-
culated using eigenvalue analysis and C0 = a coefficient that
accounts for the difference between roof displacement of a MDOF
structure and displacement of the equivalent SDOF system and is
calculated with Eq. (2):
PN
mx /1;x
C 0 ¼ /1;r P1N 2
ð2Þ
1 mx /1;x

In order to extract meaningful and practical information, it is

often required to develop an equivalent bilinear approximation
of pushover curve. The maximum base shear (Vmax) is defined as
the maximum base shear developed in the structure prior to onset Vmax
of stiffness degradation as shown in Fig. 5. Unlike Vmax, defining
Base shear

Dmax requires judgment and depends on the structure type and

A2 (The area above curve)
its occupancy. Generally, Dmax might be defined in a way to
account for post-peak deformation. This is shown in Fig. 5 as Dulti-
mate which denotes the deformation of the structure after a certain
Vfirst yield
reduction in the stiffness.
A1 (The area below curve)
Some studies suggest maximum displacement (Dmax) to be
defined at a certain level of buckling or fracture in structure [25].
However, most definitions are appropriate for structures with high ∆y ∆max ∆ultimate
level of redundancy such as RC frames. RC pedestals of elevated Lateral displacement at control
water tanks have very low level of redundancy. Reduction in the
stiffness could progressively lead to the global collapse of Fig. 5. Bilinear idealization of pushover curves.
38 R. Ghateh et al. / Engineering Structures 87 (2015) 32–46

25 20
(a)
20 15

Base shear (MN)

Base shear (MN)

15
R=2 10 R=3
10
Bilinear 5 Bilinear
5
approximation approximation
0 0
0 50 100 150 200 0 50 100 150 200
Top lateral deflection (mm) Top lateral deflection (mm)

80 60
(b)
50

Base shear (MN)

Base shear (MN)

60 40
R=2 30 R=3
40
20
20 Bilinear Bilinear
approximation 10 approximation
0 0
0 30 60 90 120 150 0 30 60 90 120 150 180
Top lateral deflection (mm) Top lateral deflection (mm)

(c) 120 100

100 80
Base shear (MN)

Base shear (MN)

80
60 R=3
60 R=2
40
40
Bilinear Bilinear
20 20
approximation approximation
0 0
0 30 60 90 120 0 30 60 90 120
Top lateral deflection (mm) Top lateral deflection (mm)

(d) 50 40

40
Base shear (MN)
Base shear (MN)

30
30
R=2 20 R=3
20
Bilinear 10 Bilinear
10
approximation approximation
0 0
0 25 50 75 0 25 50 75 100
Top lateral deflection (mm) Top lateral deflection (mm)

(e) 30 25
25 20
Base shear (MN)

Base shear (MN)

20
15
15 R=2 R=3
10
10
Bilinear Bilinear
5 5
approximation approximation
0 0
0 100 200 300 400 0 100 200 300 400
Top lateral deflection (mm) Top lateral deflection (mm)

Fig. 6. Pushover curves and corresponding bilinear approximation (a) 25-H-0.5, (b) 25-H-2, (c) 25-H-3, (d) 15-H-1 and (e) 45-H-1.

where mx = the mass at level x, /1,x = ordinate of fundamental mode
at level x and N = number of levels.
The second approach, as depicted in Fig. 5, defines Dy as the
yield point of an equivalent elasto-plastic system with reduced
stiffness at 75% of Vmax [25]. Finally, in the third method, Dy is
determined based on the principle of equal energy. According to
this last method, the elasto-plastic equivalent system absorbs the
same energy as the original structure and as a result the area
enclosed between the curve and bilinear approximation must be
equal below and above the curve which means A1 = A2 (see
Fig. 5). While developing the idealized bilinears, effort was made
to satisfy all above three criterions.
5.3. Results of pushover analysis
As discussed earlier, each prototype is designed for two values
of R = 2 and R = 3 as for the response modification factor. In all of
the pushover analysis results, prototypes with response modifica-
tion factor of R = 3 present less maximum base shear than the same
prototype designed for R = 2. This effect is more considerable for
 R. Ghateh et al. / Engineering Structures 87 (2015) 32–46 39

(a) 30
15-H-0.5 (b) 100 35-H-0.5
25
Base shear (MN)
25-H-0.5 80 35-H-1

Base shear (MN)

35-H-0.5
20 35-H-3
45-H-0.5 60
15
10 40

5 20
0 0
0 50 100 150 200 250 300 350 0 50 100 150 200 250 300
Top lateral deflection (mm) Top lateral deflection (mm)

(c)

Fig. 7. The effect of various parameters on the pushover curves (a) height effect, (b) tank size effect and (c) seismicity effect.

35-H-3 35-H-3
35-H-1 35-H-1

35(m)

(a) (b)

V1 V2 V1 V2

(c) (d)

Fig. 8. Cracking propagation of RC pedestals subjected to pushover analysis (a) model 35-H-1 elevation, (b) model 35-H-3 elevation, (c) location of initial cracks for model 35-
H-1 and (d) location of initial cracks for model 35-H-3.

models with lower capacity tanks. For example, in case of model
25-H-0.5 the ratio of Vmax(R = 2) to Vmax(R = 3) is approximately
1.5. This ratio decreases to 1.1 for model 25-H-3. In addition,
response modification factor does not appear to have a consider-
able effect on the maximum displacement (Dmax) of the proto-
types. The difference between the Dmax for prototypes designed
with R = 2 and R = 3 is not significant. The pushover curves of FE
models 25-H-0.5, 25-H-2, 25-H-3, 15-H-1 and 45-H-1 along with
the idealized bilinear approximations are depicted in Fig. 6.
The effect of pedestal height on pushover curves is illustrated in
Fig. 7(a). All four pushover curves belong to prototypes with a tank
capacity of 0.5 but with different heights. The first observation is
that prototypes with shorter RC pedestal heights demonstrate
higher maximum base shear comparing to taller ones. On the other
hand, models with short pedestals are not able to demonstrate as
much lateral displacement capacity as the tall pedestals do. The
maximum developed base shear in FE model 15-H-0.5 is approxi-
mately 2.3 times larger than FE model 45-H-0.5.
40 R. Ghateh et al. / Engineering Structures 87 (2015) 32–46

Table 3 Table 4
Cracking pattern and h/dw ratio. Definition of parameters used in Fig. 9 and related descriptions.

FE model ID h (m) dw (m) h/dw Cracking mechanism Parameter Description

15-H-0.5 15 8.6 1.7 Web-shear cracking R = Ve/Vd Response modification factor (R factor)
15-H-1 15 12 1.3 Web-shear cracking Rl = Ve/Vmax Ductility factor
15-H-2 15 14.5 1.0 Web-shear cracking X0 = Vmax/Vd Overstrength factor
25-H-0.5 25 8.6 2.9 Flexure–shear cracking Ve Maximum base shear in an equivalent completely
25-H-2 25 14.5 1.7 Web-shear cracking elastic structure
25-H-3 25 20 1.3 Web-shear cracking Vmax Maximum base shear developed in structure
35-H-0.5 35 8.6 4.1 Flexure–shear cracking Vd Design base shear (according to pertinent code)
35-H-1 35 12 2.9 Flexure–shear cracking Dy Global yield displacement
35-H-3 35 20 1.8 Web-shear cracking Dmax Maximum displacement prior to onset of stiffness reduction
45-H-0.5 45 8.6 5.2 Flexure–shear cracking l = Dmax/Dy Displacement ductility ratio
45-H-1 45 12 3.8 Flexure–shear cracking
45-H-3 45 20 2.3 Flexure–shear cracking
flexural cracks toward the sides of pedestal and parallel to the lat-
eral load direction. These are flexure–shear cracks which are the
Another noticeable trend is the effect of tank sizes on the push-
result of combined effects of flexure and shear at the base of the
over curves. This pattern is shown in Fig. 7(b) for pedestal height of
pedestal.
35 m. It is concluded that for the same pedestal height, prototypes
The cracking pattern for FE model 35-H-3 differs from model
with smaller tank sizes are providing more lateral displacement
35-H-1. In this model, initial cracks are inclined as displayed in
capacity comparing to models with larger tank sizes. The effect
Fig. 8(d). These cracks are classified as web-shear cracks. Unlike
of site seismicity for models 35-H-1 and 35-L-1 is depicted in
the flexure–shear cracks, web-shear cracks develop first only near
Fig. 7(c) as well. The FE model 35-L-1 is presenting lowest maxi-
to the base on the sides parallel to the lateral load direction as
mum base shear capacity. Response modification factor does not
depicted in Fig. 8(b). By increasing the lateral load, the cracks prop-
have an effect on the seismic response of 35-L-1 as this structure
agate throughout the height.
is designed for the minimum reinforcement requirements.
The cracking pattern in RC pedestal structures is related to the
height of pedestal (h) and the tank size. Basically flexure–shear
6. Cracking pattern in pedestals cracking is more likely to occur in taller pedestals and web-shear
cracking is observed in shorter pedestals. The tank size also indi-
The results of pushover analysis indicate two categories of rectly influences the cracking pattern by changing the diameter
cracking patterns in the RC pedestal structures. These two catego- of the pedestals. Elevated water tanks with larger tank size have
ries are classified with respect to the height to mean diameter (h/ higher pedestal diameter comparing to smaller tank sizes.
dw) ratios of pedestals. Fig. 8 demonstrates two stages of the push- Further investigation of the results of pushover analysis reveals
over analysis of FE models 35-H-1 and 35-H-3. Stage one, which is that the cracking propagation pattern could be related to the ratio
denoted by base shear V1, is when the lateral loading reaches to the of pedestal height (h) to mean diameter of the pedestal (dw). Table 3
level that cracking has just begun and the structure enters the non- presents a summary of h/dw ratios for FE models located in high
linear response region. The base shear V2 represents the second seismicity region. This ratio ranges from a minimum of 1 which
stage of loading in which the cracks are considerably propagated belongs to FE model 15-H-2 to maximum of 5.2 for FE model 45-
across the pedestal and structure has experienced substantial lat- H-0.5.
eral deflection. The results in Table 3 suggest that for all of the prototypes with
For model 35-H-1, as shown in Fig. 8(c), the cracking develop- height to diameter ratio below 1.8, the web-shear cracking pattern
ment begins with flexural tension cracks at the base of pedestal. is observed. On the other hand, prototypes with height to diameter
These cracks are horizontal and located at the pedestal side per- ratio of above 2.3, exhibit a flexure–shear cracking pattern. By per-
pendicular to the direction of lateral loading. By further increasing forming a linear interpolation, the approximate threshold to deter-
the lateral loads, inclined cracks will develop around the initial mine the cracking pattern is calculated to be ‘‘h/dw = 2’’.

Maximum force in fully elastic system

Ve

Base shear

Rμ=Ve/Vm
Bilinear approximation R=V e/Vd

Vmax

Ω0= V max/Vd Pushover curve

Vd

μ= ∆max/∆y

∆y ∆max Roof displacement

Fig. 9. Definition of seismic response factors on a typical pushover curve, adapted from FEMA P695 [26].
 R. Ghateh et al. / Engineering Structures 87 (2015) 32–46 41

Accordingly, the elevated tanks which are mounted on an RC
pedestal with a height to diameter ratio of 2 and above are
expected to have flexure–shear cracking pattern. On the other
hand, if the height to width ratio is below 2, then a web-shear
cracking is more likely to occur in the RC pedestals.
It should be mentioned that the above recommended threshold
value is valid for RC pedestals which are designed according the
provisions of current codes and guidelines.
7. Establishing seismic response factors
7.1. Seismic response factors
The seismic response factors for a typical structure are dis-
played in Fig. 9. The parameters shown on this figure are explained
in Table 4. Overstrength factor has significant effect on the seismic
response of structures. Studies have shown that higher values of
overstrength factor provide more resistance to the collapse of
structures [27]. Currently ASCE/SEI 7-2010 [14] proposes over-
strength factor value of ‘‘2’’ for RC pedestals.
Ductility factor of structures has been the subject of many
research studies. One of the first investigations which addressed
ductility factor is the one carried out by Newmark and Hall [28].
Krawinkler and Nassar [29] developed a relationship for SDOF
systems on rock or stiff soil sites. They used the results of a sta-
tistical study based on 15 western U.S. ground motion records
from earthquakes with magnitude of 5.7 to 7.7. Miranda and
Bertero [30] introduced another equation for ductility factor.
This relationship was developed for rock, alluvium, and soft soil
sites by implementing 124 ground motions. In this study, the
Newmark and Hall [28] relationship is employed mainly because
it offers a conservative lower bound [22] which seems more rea-
sonable for essential infrastructures such as water storage
facilities.

Table 5
Seismic response factors for high and low seismicity design.

FE model ID High seismicity FE model ID Low seismicity

X0 Rl X0 Rl
R=2 R=3 R=2 R=3 R=2 R=3 R=2 R=3
15-H-0.5 1.9 2.3 1.7 2.2 15-L-0.5 3.8 5.6 2.3 2.3
15-H-1 1.5 1.9 2.0 2.5 15-L-1 3.4 5.1 2.8 3.0
15-H-2 1.2 1.6 2.1 2.5 15-L-2 2.8 4.0 2.7 3.0
25-H-0.5 1.9 2.0 1.5 2.0 25-L-0.5 4.8 7.2 2.2 2.2
25-H-2 1.3 1.6 1.9 2.3 25-L-2 3.0 4.2 2.9 3.2
25-H-3 1.2 1.6 2.1 2.4 25-L-3 2.5 3.4 2.7 3.1
35-H-0.5 2.0 2.3 1.5 2.1 35-L-0.5 5.4 8.1 1.5 1.5
35-H-1 1.8 2.0 1.9 2.6 35-L-1 4.2 7.0 2.7 2.5
35-H-3 1.3 1.6 2.0 2.5 35-L-3 3.5 4.8 2.8 4.0
45-H-0.5 2.1 2.4 1.4 1.8 45-L-0.5 6.3 9.4 1.2 1.2
45-H-1 1.8 2.0 1.8 2.5 45-L-1 5.0 7.5 2.0 2.0
45-H-3 1.4 1.7 2.0 2.7 45-L-3 4.2 6.3 2.4 2.4

10 5
9
8 4
7
Overstrength factor

Ductility factor

6 3
5
4 2
3
2 1
1
0 0
0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0
Fundamental period (sec) Fundamental period (sec)

10 5
9
8 4
Overstrength factor

7
Ductility factor

6 3
5
4 2
3
2 1
1
0 0
0.0 2.0 4.0 6.0 0.0 2.0 4.0 6.0
h/dw h/dw

Fig. 10. Effect of fundamental period and height to diameter ratio on overstrength and ductility factor of RC pedestals.
42 R. Ghateh et al. / Engineering Structures 87 (2015) 32–46

7.2. Calculated seismic response factors of RC pedestals
Based on the information provided in the last section and the
pushover curves that were developed for the 48 prototypes, over-
strength factor and ductility factor are determined and demon-
strated in Tables 5.
8. Investigation of seismic response factors of RC pedestals
8.1. Effect of fundamental period and height to diameter ratio
The effect of fundamental period and height to diameter ratio on
seismic response factors is demonstrated in Fig. 10. An exponential
trend line is added to the graphs as well. The scattered data in the
graph is the result of including a wide variety of FE models with dif-
ferent RC pedestal heights and diameters, tank sizes and seismicity
level. Generally, according to trend lines, overstrength factor
increases by increasing the fundamental period. On the other hand,
ductility factor declines when fundamental period increases. Other
than the general trend lines, no specific relation is found to relate
the fundamental period to either of the seismic response factors.
A similar pattern is observed for height to diameter ratio (h/dw)
versus seismic response factors. It could be observed that over-
strength factor increases when the h/dw ratio increases. For h/dw
ratios below 2, overstrength factor ranges from 1.3 to 5. This range
is wider for h/dw ratio above 2. On the other hand, ductility factor
has inverse relationship with h/dw ratio. The range of variations of
ductility factor is not as wide as overstrength factor and fluctuates
from 1 to 3.
8.2. Effect of tank size
Fig. 11 shows graphs of tank size versus overstrength and duc-
tility factor. This figure indicates that elevated water tanks with

(a) 10 (b) 4.5

9
4
8
3.5
Overstrength factor

Ductility factor
3
6
2.5
5
2
4
3 1.5
2 1
1 0.5
0 0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Tank size (mega gallon) Tank size (mega gallon)

Fig. 11. Effect of tank size on (a) overstrength factor and (b) ductility factor.

(a) 20 15
R=2
Base shear (MN)

R=3
Base shear (MN)

15
10
10
5
5 Vd = 8 MN Vd = 5.3 MN
0 0
0 100 200 300 0 100 200 300
Top lateral deflection (mm) Top lateral deflection (mm)

(b) 40 30
R=3
Base shear (MN)

R=2 25
Base shear (MN)

30
20
20 15

Vd = 20 MN 10
10 Vd = 13 MN
5
0 0
0 100 200 300 0 100 200 300
Top lateral deflection (mm) Top lateral deflection (mm)

(c) 100 80
Base shear (MN)
Base shear (MN)

80
60
60
Vd = 67 MN 40
40 Vd = 45 MN
20
20
R=2 R=3
0 0
0 50 100 150 200 0 50 100 150 200
Top lateral deflection (mm) Top lateral deflection (mm)

Fig. 12. Pushover curves and corresponding seismic design base shear (a) 35-H-0.5, (b) 35-H-1 and (c) 35-H-3.
 R. Ghateh et al. / Engineering Structures 87 (2015) 32–46 43

higher tank sizes will demonstrate lower overstrength factor. The
ductility factor however increases as the tank size gets larger.
The effect of tank size is further explained in Fig. 12 which dis-
plays pushover curves of models 35-H-0.5, 35-H-1 and 35-H-3 and
related design base shear. It could be concluded from the graphs
that by increasing the tank size, the design base shear further
extends to the nonlinear response branch of the pushover curves.
For the models designed with R = 3, the design base shear falls in
the linear branch of pushover curve since the level of seismic forces
is reduced by 2/3 in this case (comparing to R = 2).
As the weight of the tank increases, the design base shear inter-
sects the pushover curves in higher deflection stages. For FE model
35-H-0.5, Vd (design seismic base shear) crosses the pushover
curve at a deflection of 20 mm which is far before the final deflec-
tion of approximately 280 mm. The maximum developed base
shear prior to failure of this model is nearly twice the Vd which
indicates a significant reserved strength in the structure.
For FE model 35-H-1, the lateral deformation corresponding to
design seismic base shear is not significant and is limited to less
than 50 mm for both models (R = 2 and R = 3). Moreover, the model
can demonstrate considerable reserved strength above the design
seismic base shear.
As the tank size is further increased, the lateral deflection corre-
sponding to Vd becomes larger. FE model 35-H-3(R = 2), demon-
strates a lateral deflection of nearly 100 mm at Vd which is more
than half of the ultimate lateral deflection of 175 mm. Although
in model 35-H-3(R = 3), Vd intersects pushover curve at a lower lat-
eral deflection, both models are presenting much lower over-
strength comparing to smaller tank size models. In summary,
model 35-H-3 which has the largest tank size of the three models,
will potentially experience the highest structural damages due to
the design seismic base shear Vd. This trend will be employed in
later sections for classifying the seismic performance of elevated
water tanks based on the rank sizes.
8.3. Effect of site seismicity
In this research, the seismic category effect is taken into account
both explicitly and implicitly. The explicit effect of seismic cate-
gory was explained before by designing the prototypes for two lev-
els of high and low seismicity. On the other hand, each prototype is
designed for two R factors of 2 and 3. The R = 3 group prototypes
are designed for 2/3 of the seismic load for which the R = 2 group
prototypes (with the same site seismicity) are designed. As a result,
the FE models designed for R = 2 in the high seismic category are
experiencing the highest seismic loads and FE models designed
for R = 3 in the low seismic region are subjected to the lowest seis-
mic loads. This is further illustrated in Fig. 13 by comparing the
35 m pedestal with a tank size of 1 Mgal designed for four levels
of seismicity.
The design seismic base shear is highest for model 35-H-
1(R = 2) and is 20 MN. Vd decreases by a factor of 2/3 for the second
model (35-H-1(R = 3)) and is as low as 13 MN. The third model (35-
L-1(R = 2)) is designed for a very low design base shear of Vd = 4
MN. Accordingly four seismic levels as demonstrated in Table 6
are introduced and analyzed.
Fig. 14 depicts overstrength factor for level one and two seis-
micity. Level two seismicity group models demonstrate higher
overstrength factors comparing to level one models. In the lower
seismicity levels three and four, the gravity load is governing the
design and the models are mainly designed for the minimum
reinforcement.
In addition, for all four levels of seismicity, 0.5 Mgal elevated
water tanks are giving the highest value of overstrength indepen-

(a) 40 (b) 30
R=3
R=2 25
Base shear (MN)
Base shear (MN)

30
20
20 15
Vd = 20 MN 10
10 Vd = 13 MN
5
0 0
0 100 200 300 0 100 200 300
Top lateral deflection (mm) Top lateral deflection (mm)

(c) 25 (d) 25
R=2 R=3
20 20
Base shear (MN)
Base shear (MN)

15 15
10 10
Vd = 4 MN Vd = 3 MN
5 5
0 0
0 50 100 150 200 250 0 50 100 150 200 250
Top lateral deflection (mm) Top lateral deflection (mm)

Fig. 13. Comparing pushover curves for four levels of seismicity (a) 35-H-1(R = 2)/level one, (b) 35-H-1(R = 3)/level two, (c) 35-L-1(R = 2)/level three and (d) 35-L-1(R = 3)/
level four.

Table 6
Four levels of seismicity for designing RC pedestals.

Level ‘‘one’’ seismicity Level ‘‘two’’ seismicity Level ‘‘three’’ seismicity Level ‘‘four’’ seismicity
SDS = 0.8 SDS = 0.8  2/3 = 0.53 SDS = 0.2 SDS = 0.2  2/3 = 0.13
SD1 = 0.4 SD1 = 0.4  2/3 = 0.26 SD1 = 0.1 SD1 = 0.1  2/3 = 0.06
‘‘High’’ seismic region, R = 2 ‘‘High’’ seismic region, R = 3 ‘‘Low’’ seismic region, R = 2 ‘‘Low’’ seismic region, R = 3
SDC ‘‘D’’ SDC ‘‘D’’ SDC ‘‘C’’ SDC ‘‘A’’
44 R. Ghateh et al. / Engineering Structures 87 (2015) 32–46

2.5 3.0

2.5
2.0

Overstrength factor
Overstrength factor
2.0
1.5
1.5
1.0
1.0

0.5 0.5
"level one "Level two
seismicity" seismicity"
0.0 0.0
0 0.5 1 1.5 2 2.5 3 3.5 0 0.5 1 1.5 2 2.5 3 3.5
Tank size Tank size

2.5 3.0

2.5
2.0
2.0

Ductility factor
Ductility factor

1.5
1.5
1.0
1.0

0.5 0.5
"Level one "Level two
seismicity" seismicity"
0.0 0.0
0 0.5 1 1.5 2 2.5 3 3.5 0 0.5 1 1.5 2 2.5 3 3.5
Tank size Tank size

Fig. 14. Overstrength factor and ductility factor for level one and two seismicity.

Table 7
Categories of tanks based on tank size.
Tank size (mega gallon)
Tank size 6 0.5
0.5 < Tank size < 1.5
1.5 6 Tank size
dent of the height of RC pedestal. According to Fig. 14, as discussed
before, by increasing the tank size, overstrength factor gradually
decreases. However, increasing the tank size from 2 Mgal to 3 Mgal
does not appear to have a considerable effect on the overstrength
factor.
Fig. 14 displays the ductility factor versus tank size for the four
seismicity levels as well. It could be observed that the range of var-
iation of ductility factor is not as wide as overstrength factor.
Moreover, in an opposite trend in comparison to overstrength fac-
tor, the ductility factor is increasing as the tank size increases.
However, for the models designed for high seismicity region, the
ductility factor is not very sensitive to the changes in the tank size
as shown in Fig. 14. Other than the 0.5 Mgal tank, the other tank
sizes are demonstrating nearly equal ductility factor. The observed
Tank size category
patterns suggest dividing the elevated water tanks into three cate-
Light
gories based on the size of the tank. Table 7 illustrates these three
Medium
Heavy categories of light, medium and heavy size tanks. It should be
noted that the tank size (as shown in Table 1) is directly correlated
with RC pedestal diameter and wall thickness. Consequently, heavy
and light size tanks have the highest and lowest RC pedestal diam-
eter and pedestal wall thickness respectively. In other words, this
categorization in not only based on the tank size, but also indicates
three structural groups of RC pedestals with distinct nonlinear
response behavior properties.
9. Establishing seismic response factors for RC pedestals
By employing the graphs presented in Fig. 14 combined with
the tank size classification of Table 7 and seismic levels provided
in Table 6, the average seismic response factors are established.
Table 8 demonstrates average overstrength factor for three catego-
ries of tank sizes under the defined levels of seismicity.

Table 8
Average overstrength and ductility factor of RC pedestals.

Seismicity level
Level ‘‘one’’ seismicity Level ‘‘two’’ seismicity Level ‘‘three’’ seismicity Level ‘‘four’’ seismicity
Overstrength factor Tank size Light 2 2.3 4.5 7
Medium 1.6 1.8 4 6
Heavy 1.3 1.5 3 4
Ductility factor Tank size Light 1.5 2 1.6 1.6
Medium 1.8 2.5 2.5 2.5
Heavy 2 2.5 2.5 3
 R. Ghateh et al. / Engineering Structures 87 (2015) 32–46
In cases where the overstrength factor of a specific group is
more scattered, a linear interpolation is employed. According to
Table 8, medium and heavy tank size structures located in high
seismicity zone are exhibiting an overstrength factor of below
the code-recommended (ASCE/SEI 7-2010) value of ‘‘2’’. As dis-
cussed before the range of variation of ductility factor is not wide
and it is fluctuating between 1.5 (lightest tank in highest seismicity
zone) and 3 (heaviest tank in lowest seismicity zone).
10. Conclusions
In this study, a finite element method was employed to investi-
gate the nonlinear seismic response of RC pedestals in elevated
water tanks. Since the elevated water tanks are built in various
tank capacities and pedestal heights, a combination of the most
commonly constructed tank sizes and pedestal heights in industry
were selected for investigation. 3D finite element models were
developed for all prototypes and the pushover curves and corre-
sponding bilinear approximations were constructed accordingly
by conducting pushover analysis. In addition, the cracking propa-
gation pattern was analyzed for several prototypes. The over-
strength and ductility factor were calculated based on pushover
curves and the effect of various parameters such as fundamental
period, height to diameter ratio, seismic design category, and tank
size on the seismic response factors of elevated water tanks was
studied.
The results of the study showed that for the same tank size, tal-
ler tanks demonstrate much lower maximum base shear (Vmax)
comparing to shorter tanks. Accordingly, two types of cracking
propagation were observed during the pushover analysis. Elevated
water tanks with a pedestal height to mean diameter ratio (h/dw) of
above 2 demonstrated flexure–shear cracking pattern which initi-
ates at the opposite top and bottom corners of RC pedestal. How-
ever, for h/dw ratio less than 2, the cracking propagation was due
to web-shear cracking which starts near the base, parallel to the
lateral load direction and gradually extends to the top of pedestal.
These patterns could be employed for seismic rehabilitation and
strengthening of existing elevated water tanks which are located
in high seismicity regions and do not comply with current codes
and standards.
The overstrength factors of heavy, medium and light tank size
groups are calculated to be 1.3, 1.6 and 2 respectively for elevated
water tanks located in high seismicity region. As the seismicity
level decreases, the overstrength factor increases. Consequently,
for the lowest seismicity (level four) the overstrength factors of
heavy, medium and light tank size groups increase to 4, 6 and 7
respectively. On the other hand, the range of variation of ductility
factor is not wide and fluctuates between 1.5 (lightest tank in high-
est seismicity zone) and 3 (heaviest tank in lowest seismicity
zone). The ductility factors for elevated water tanks located in high
seismicity region are determined as 2, 1.8 and 1.5 for heavy, med-
ium and light tank size groups respectively. Ductility factor is not
significantly influenced by the seismicity level as it is mainly a
function of geometry and material properties of the structure.
For elevated water tanks with the same height but with differ-
ent tank sizes which are located in the same seismicity zone, the
heavier tank sizes undergo more lateral deformation and compar-
atively experience more damages. Furthermore, increasing the fun-
damental period and h/dw ratio result in higher overstrength factor
and lower ductility factor. In the current codes and standards, all
elevated water tanks, regardless of the tank size and pedestal
dimensions are considered in the same category for seismic design.
This study shows that the variation of the tank size and pedestal
height can significantly affect the seismic response behavior of
RC pedestals in elevated water tanks and consequently result in
45
different seismic response factors. It is suggested that for seismic
design, the elevated water tanks should be divided into three
groups of light, medium and heavy based on their tank sizes.
It should be mentioned that, in this study, the base of the ped-
estal was assumed to be rigid; other restraining conditions at the
base level could be investigated and the effect of soil–structure
interaction may be taken into account as well. This study did not
evaluate and verify the response modification factor of elevated
water tanks which may be the subject of future research studies.
Acknowledgement
The work described in this paper was supported by Mitacs
Canada under the Elevate Program and the American Concrete
Institute, ACI Foundation.
References
[1] Steinbrugge KV, Rodrigo FA. The Chilean earthquakes of May 1960: a structural
engineering viewpoint. Bull Seismol Soc Am 1963;53(2):225–307.
[2] Memari AM, Ahmadi MM, Rezaee B. Behaviour of reinforced concrete water
towers during Manjil-Roudbar earthquake of June 1990. Proceedings of the
tenth world conference on earthquake engineering; Madrid, 19–24 July 1992,
vol. 9. Rotterdam: A.A. Balkema; 1992. p. 4953–9.
[3] Rai DC. Seismic retrofitting of R/C shaft support of elevated tanks. Earthq
Spectra 2002;18:745–60.
[4] ACI 371R-08. Guide for the analysis, design and construction of elevated
concrete and composite steel–concrete water storage tanks. American
Concrete Institute, ACI Committee 371, Farmington Hills, MI, USA; 2008.
[5] ACI 318-08. American Concrete Institute Committee 318. Building code
requirements for structural concrete and commentary. Farmington Hills, MI,
USA; 2008.
[6] Housner GW. The dynamic behavior of water tanks. Bull Seismol Soc Am
1963;53(2):381–9.
[7] Kianoush MR, Ghaemmaghami AR. The effect of earthquake frequency content
on the seismic behavior of concrete rectangular liquid tanks using the finite
element method incorporating soil–structure interaction. Eng Struct
2011;33(7):2186–200.
[8] Moslemi M, Kianoush MR. Parametric study on dynamic behavior of cylindrical
ground-supported tanks. Eng Struct 2012;42(September):214–30 [Complete].
[9] El Damatty AA, Saafan MS, Sweedan AMI. Experimental study conducted on a
liquid-filled combined conical tank model. J Thin Walled Struct
2005;43:1398–417.
[10] Kianoush MR, Chen JZ. Effect of vertical acceleration on response of concrete
rectangular liquid storage tanks. Eng Struct 2006;28(5):704–15.
[11] Shepherd R. Two mass representation of a water tower structure. J Sound Vib
1972;23(3):391–6.
[12] Dutta SC, Dutta S, Roy R. Dynamic behavior of R/C elevated tanks with soil–
structure interaction. Eng Struct 2009;31(11):2617–29.
[13] Moslemi M, Kianoush MR, Pogorzelski W. Seismic response of liquid-filled
elevated tanks. Eng Struct 2011;33:2074–84.
[14] American Society of Civil Engineering (ASCE). Minimum design loads for
buildings and other structures. ASCE standard ASCE/SEI 7-10; 2010.
[15] ACI 350.3-06. Seismic design of liquid-containing concrete structures (ACI
350.3-06) and commentary (350.3R-06). American Concrete Institute (ACI)
Committee 350, Environmental Engineering Concrete Structures, Farmington
Hills, Mich; 2006.
[16] Willam KJ, Warnke ED. Constitutive model for the triaxial behavior of concrete.
Proceedings of the international association for bridge and structural
engineering, vol. 19. ISMES, Bergamo, Italy; 1975. p. 174.
[17] Hognestad E. A study of combined bending and axial load in reinforced
concrete members, bulletin 399. University of Illinois Engineering Experiment
Station, Urbana, Ill., November 1951, 128 pp.
[18] Todeschini CE, Bianchini AC, Kesler CE. Behavior of concrete columns
reinforced with high strength steels. ACI J Proc 1964;61(6):701–16.
[19] Mander JB, Priestley MJN, Park R. Theoretical stress–strain model for confined
concrete. J Struct Eng ASCE 1988;114(3):1804–26.
[20] Holzer et al. 1975. SINDER. A computer code for general analysis of two-
dimensional reinforced concrete structures. Report. AFWL-TR-74-228, vol. 1.
Air Force Weapons Laboratory, Kirt-land, AFB, New Mexico.
[21] Mickleborough NC, Ning F, Chan CM. Prediction of the stiffness of reinforced
concrete shear walls under service loads. ACI Struct J 1999;96(6):1018–26.
[22] Elnashai AS, Di Sarno L. Fundamentals of earthquake engineering. John Wiley
& Sons Ltd; 2008.
[23] Papanikolaou VK, Elnashai AS, Pareva JF. Evaluation of conventional and
adaptive pushover analysis II: Comparative results. J Earthq Eng
2006;10(1):127–51.
[24] FEMA, NEHRP guidelines for the seismic rehabilitation of buildings. FEMA 273,
Federal Emergency Management Agency; 1996.
46 R. Ghateh et al. / Engineering Structures 87 (2015) 32–46

[25] Park R. Ductility evaluation from laboratory and analytical testing. In:
Proceedings of the 9th world conference on earthquake engineering, vol.
VIII. Tokyo – Kyoto, Japan; 1988. p. 605–16.
[26] FEMA P695. Quantification of building seismic performance factors. Federal
Emergency Management Agency; 2009.
[27] Elnashai AS, Mwafy AM. Overstrength and force reduction factors of
multistorey reinforced-concrete buildings. Struct Des Tall Build
2002;11(5):329–51.
[28] Newmark NM, Hall WJ. Earthquake spectra and design. Earthquake
Engineering Research Institute, Oakland, Calif.; 1982.
[29] Nassar H, Krawinkler AA. Seismic design based on ductility and cumulative
damage demands and capacities. In: Fajfar P, Krawinkler H, editors. Nonlinear
seismic analysis and design of reinforced concrete buildings. New
York: Elsevier Applied Science; 1992.
[30] Miranda E, Bertero VV. ‘Evaluation of strength reduction factors for
earthquake-resistant design. Earthq Spectra 1994;10(2):357–79.