See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/284122545

Passive energy dissipation-based retrofit strategies for R/C frame water

towers

Article  in  Engineering Structures · January 2016

DOI: 10.1016/j.engstruct.2015.10.038

CITATIONS READS

30 769

3 authors:

Stefano Sorace Gloria Terenzi

University of Udine University of Florence
70 PUBLICATIONS   819 CITATIONS    68 PUBLICATIONS   848 CITATIONS   

SEE PROFILE SEE PROFILE

Claudia Mori
University of Florence
3 PUBLICATIONS   48 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Advanced seismic protection technologies for new and existing structures View project

Research Topic on "Innovative Approaches in Computational Structural Engineering" View project

All content following this page was uploaded by Stefano Sorace on 08 January 2018.

The user has requested enhancement of the downloaded file.

 Engineering Structures 106 (2016) 385–398

Contents lists available at ScienceDirect

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

Passive energy dissipation-based retrofit strategies for R/C frame water

towers
Stefano Sorace a,⇑, Gloria Terenzi b, Claudia Mori b
a
Department of Civil Engineering and Architecture, University of Udine, Via delle Scienze 206, 33100 Udine, Italy
b
Department of Civil and Environmental Engineering, University of Florence, Via S. Marta 3, 50139 Florence, Italy

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

Article history: Pre-normative elevated liquid storage tanks with reinforced concrete (R/C) frame staging are among the
Received 30 March 2015 most seismically vulnerable structures in urban areas, due to their unfavourable earthquake-resistant
Revised 20 October 2015 configuration. A representative case study, i.e. a R/C water tower designed by the world-famous Italian
Accepted 22 October 2015
engineer Pier Luigi Nervi and built in the early 1930s, is examined herein. The assessment of the structure
was developed with a detailed finite element model, which includes a multi spring-mass assembly to
reproduce the fluid–tank dynamic interaction. The time-history evaluation analysis, initially carried
Keywords:
out by assuming an elastic behavior of the staging members, showed general unsafe response conditions
Water towers
Pre-normative R/C structures
of beams and columns under seismic action scaled at the maximum considered earthquake (MCE) level.
Seismic assessment These results were confirmed at a second step of the analysis, where the plastic behavior of beams and
Seismic retrofit columns was investigated after incorporating lumped plastic hinges at their end sections, which resulted
Dissipative braces in the collapse of the model. Based on these data, two passive energy dissipation-based retrofit measures
Fluid viscous spring-dampers were proposed, for which dissipative bracing systems incorporating either pressurized fluid viscous
Steel yielding dissipators spring-dampers or added damping and stiffness steel yielding devices, respectively, were installed. The
two systems were designed with a mutual energy-based criterion and for the same target performance
objectives, that is, reaching elastic response up to the MCE. The benefits induced by the two protective
systems to the seismic response of the tank structure are discussed, and the geometrical dimensions
and relevant architectural impact are compared with the ones of a conventional non-dissipative bracing
retrofit solution developed for the same performance.
Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction to existing frame buildings have been accomplished worldwide

Passive energy dissipation (PED) technologies are being com-
monly adopted in earthquake-prone regions as seismic retrofit
strategies for building structures. They are in fact capable of guar-
anteeing remarkable improvements of seismic response with low
architectural impact, short or null interruption in use during instal-
lation, and lower costs as compared to traditional rehabilitation
solutions designed for the same objectives [1–3]. As is known,
the dissipators incorporated in PED systems are either rate-
dependent, when their resisting force is a function of the relative
velocity across the device, or rate-independent. Fluid viscous and
viscoelastic solid devices belong to the former class; metallic yield-
ing and friction devices to the latter. Extensive studies have been
dedicated to all these types of dissipators, and several applications
⇑ Corresponding author. Tel.: +39 432 558050; fax: +39 432 558052.
E-mail addresses: stefano.sorace@uniud.it (S. Sorace), gloria.terenzi@unifi.it
(G. Terenzi), cla.mori87@gmail.com (C. Mori).
[4–9].
On the other hand, very few applications are noticed for civil
infrastructures with reinforced concrete (R/C) or steel frame stag-
ing, such as elevated liquid storage tanks, gas and oil pipelines,
aqueducts, and transmission and observatory towers [10–12].
Indeed, conventional rehabilitation techniques are normally
adopted for this wide stock of facilities, for which well-
established protocols concerning structural maintenance and
strengthening interventions have been developed during time,
and are still implemented. In view of this, the effective design
and installation procedures currently available for PED systems
in frame building structures, as well as their competitive costs,
suggest to explore the possibility of extending their field of appli-
cation to the above-mentioned facilities too.
A study on this topic is presented in this paper, where attention
is focused on pre-normative (i.e. designed before the release of
Seismic Standards) R/C elevated waters, which are among the most
seismically vulnerable lifeline units. This is a consequence of the

http://dx.doi.org/10.1016/j.engstruct.2015.10.038
0141-0296/Ó 2015 Elsevier Ltd. All rights reserved.
386 S. Sorace et al. / Engineering Structures 106 (2016) 385–398

Nomenclature

List of symbols fy,min reinforcing steel minimum yielding stress

Ag gross R/C section area g acceleration of gravity
0
D tank diameter hc height of the convective liquid mass from the bottom of
ED energy dissipated by a PED system the tank wall
0
ED,f ED value computed at the end of the analysis hi height of the impulsive liquid mass from the bottom of
EtD tentatively estimated ED value the tank wall
EtD;pl;m energy dissipated by a T-ADAS plate in a cycle bounded hL liquid height
by dmax,e k1, k2 stiffness values of the response branches of a FV spring-
EtD;pl;tot total energy dissipated by a T-ADAS plate damper situated below and beyond F0
Ee elastic strain energy kc/2 stiffness of each one of the two identical elastic springs
EI absolute input energy connecting the convective mass to the tank wall
En nominal energy dissipation capacity of a FV spring- kc/4 stiffness of each one of the four identical elastic springs
damper connecting the convective masses to the tank wall in the
Ep,h energy dissipated by the plastic hinges of the finite ele- presence of a manhole
ment model kcj/4 kc/4 value for the j-th liquid volume fraction
Es steel Young modulus ke, kp stiffness values of the elastic and post-elastic response
Fd, Fne damping and non-linear elastic reaction forces of a FV branches of a T-ADAS plate
spring-damper m seismic mass of the structure
Fmax maximum reaction force of a FV spring-damper mc convective liquid mass
F0 static pre-load force of a FV spring-damper mc/2 convective mass of each one of the two liquid portions
FC confidence coefficient determined by the tank manhole
H, T, B height, thickness and base of a T-ADAS plate mcj/2 mc/2 value for the j-th liquid volume fraction
Mlc,1, Mlc,2 mi impulsive liquid mass
bending moments around the local axes 1 and 2 of col- mi/2 impulsive mass of each one of the two liquid portions
umns determined by the tank manhole
Mu ultimate resisting moment of a R/C section mij/2 mi/2 value for the j-th liquid volume fraction
My yielding resisting moment of a R/C section mL total liquid mass (mL = mi + mc)
N axial force in columns n number of water volume fractions
Np number of plates of a T-ADAS damper rN normalized axial force in columns
t time variable
Ntp tentative total number of plates of a T-ADAS damper
s displacement of a FV spring-damper
Ns axial force in columns calculated for the basic combina- s_ velocity of a FV spring-damper
tion of gravitational loads smax stroke of a FV spring-damper
Py, Pu yielding and ultimate force of a T-ADAS plate tc instant at which the maximum inter-level drift is
T1 topographic category reached
V maximum computed shear ug ground displacement
VN nominal structural life €t
u absolute acceleration
VR reference time period for the evaluation of the probabil- Dt integration time-step
ity of exceedance for a seismic level a fractional exponent of velocity
bw web width b preliminary fixed energy ratio
c damping coefficient of a FV spring-damper c strain hardening ratio
cu coefficient of use of a structure hn circumferential angle (hn = 180°/n) determining the
d displacement of a T-ADAS damper water volume fractions
dc distance from extreme compression fiber to centroid of hpl plastic chord rotation
tension reinforcement in a R/C section hpl,lim,CP limit plastic chord rotation at CP performance level
dmax,e estimated peak displacement of a T-ADAS damper hpl,u ultimate plastic chord rotation
dy, du yielding and ultimate displacement of a T-ADAS plate hu ultimate chord rotation
fc concrete compressive cylinder strength hy yielding chord rotation
fc,cube concrete compressive cube strength n linear viscous damping ratio
ft steel ultimate stress ne equivalent linear viscous damping ratio
fy steel yielding stress

tall and slender geometry of staging, of the little redundancy and
low ductility of the constituting members, as well as of an unfa-
vourable structural configuration with respect to seismic action,
i.e. with the highest portion of masses (vessel plus contained liq-
uid) concentrated on top. Another hazard is that water towers
are often situated in urban areas, sometimes in city centers, rather
than in suburban areas like industrial liquid tanks. Therefore, their
partial or global failure may cause heavy damage to the surround-
ing buildings and infrastructures, with serious consequences for
the safety of a great number of inhabitants [13].
A representative case study, i.e. a R/C frame-supported elevated
tank designed by the world-famous Italian engineer Pier Luigi
Nervi in the early 1930s as a water supply for Santa Maria Novella
Station in Florence, still in service, is examined herein. The seismic
performance of the tower is assessed via time-history analysis,
with the dynamic behavior of the fluid simulated by a three-
dimensional assembly of the traditional Housner’s two-mass plan
model. The results of the numerical analysis, carried out initially
by assuming an elastic behavior of staging members, and subse-
quently by incorporating a set of plastic hinges at their end sec-
tions, show that collapse conditions are reached under seismic
action scaled at the maximum considered earthquake (MCE) level.
Based on these data, two PED-based retrofit hypotheses are pro-
posed, consisting in the installation of dissipative bracing systems
 S. Sorace et al. / Engineering Structures 106 (2016) 385–398
including either pressurized fluid viscous (FV) spring-dampers or
added-damping-and-stiffness (ADAS) steel yielding devices,
respectively. The two systems are designed with a mutual
energy-based criterion, as well as for the same target performance
objectives, that is, reaching elastic undamaged response up to the
MCE. The mechanical parameters and technical implementation
details of the two protective systems, and the benefits induced in
the seismic response of the tank structure are discussed herein,
showing the attainment of target design objectives. The geometri-
cal dimensions and the architectural impact of both solutions are
also compared with the ones of a conventional non-dissipative
steel bracing retrofit solution developed for the same performance.
2. Geometrical and structural characteristics of the water tower
A view of the case study water tower in 1935, a few months
after the completion of the construction works, is displayed in
the left image of Fig. 1. The cross section and base plan are shown
in the right drawing of the same figure. The structure is 21.45 m
high, with staging height of 14.8 m (3.58 m being the height of
the first staging level, and 3.74 m the height of the second through
fourth levels), and vessel height of 6.65 m. The maximum available
water volume is 100 m3. The internal diameters of the vessel and
coaxial cylindrical manhole are equal to 6 m and 1 m, respectively;
the external diameter of the staging base is equal to 4.5 m. The
external wall, manhole wall, bottom slab and floor slab of the ves-
sel are 140 mm, 80 mm, 150 mm and 90 mm thick, respectively.
The R/C frame structure comprises six columns with mutual
6000
6650
3740
3740
3740
4500
3580
C3 C2
4500
C4
C5 C6
Fig. 1. Photographic view of the water tower taken in 1935, vertical and plan cross
section.
387
section of (300  300) mm  mm (column alignments are num-
bered in the plan of Fig. 1 with symbols C1 through C6); three
intermediate ring beams with section of (300  300) mm  mm,
situated at a mutual distance along the height of about 3.7 m;
and a top ring beam sized (300  500) mm  mm, constituting
the annular support of the vessel. The foundation consists of a ring
beam with rectangular section of (450  650) mm  mm, enlarged
at the feet of the columns by six blocks with the same height of the
beam and base section of (1400  1100) mm  mm. These mem-
bers lie on a several-meter-deep compressed mixed mortar–stone
substrate, specially built to significantly increase the soil bearing
capacity and prevent any ground settlement under the design sta-
tic loads, according to a typical building protocol used for elevated
tanks at the time.
The mechanical properties of concrete and steel and reinforce-
ment details have also been deducted from the original design
drawings and characterization test reports. These documents high-
light that the compressive cube strength of concrete, fc,cube, is not
less than 40 MPa for the vessel and supporting structures, and
30 MPa for the foundation members. Reinforcing steel is in smooth
bars with a minimum yield stress, fy,min, equal to 220 MPa. Consid-
ering that no diagnostic activities were carried out on the R/C
members to check design data, the values of concrete and steel
mechanical parameters were divided by a confidence coefficient,
FC, equal to 1.2 in stress state verifications, according to the pre-
scriptions of the Commentary on the Italian Technical Standards
[14] with regards to existing structures, in addition to the basic
safety coefficients of the two materials.
3. Finite element model of the water tower
The finite element model of the water tower structure was gen-
erated by SAP2000NL program [15]. A perspective view of the
model is displayed in the left image of Fig. 2, where the global ref-
erence coordinate system, constituted by the Cartesian axes x, y
and z, is visualized too. The mesh of the vessel is made of shell-
type elements, and the staging structure of frame-type elements.
The computational simulation of fluid–tank interaction was carried
out according to the classical Housner’s two-mass equivalent
model [16,17], which is being widely used in most international
Seismic Standards and Design Guidelines on liquid storage tanks
[18–20]. The model splits the total liquid mass in two fractions,
i.e. an impulsive mass mi, which oscillates synchronously with
the tank wall, and a convective mass, which undergoes sloshing
motion. According to this schematization, mi is rigidly connected
to the tank wall, whereas mc is linked by two elastic springs with
14800
identical stiffness kc/2. Both masses are rigidly joined to the vessel
0 0
wall in vertical direction, Named hi ; hc the heights of the two
0 0
masses from the bottom of the tank wall, the mi, mc, hi ; hc and kc
values assumed in the analysis were computed by means of the fol-
lowing normative expressions for circular tanks [18–20]:


D
mi tanh 0:866 hL
¼ ð1Þ
mL 0:866 hD L


mc tanh 3:68 hDL
¼ 0:23 ð2Þ
mL hL
D
0
C1 hi D
¼ 0:45 for < 0:75 ð3aÞ
hL hL
0
hi 0:866 hDL D
¼
  0:125 for P 0:75 ð3bÞ
hL 2 tanh 0:866 D hL
hL
388 S. Sorace et al. / Engineering Structures 106 (2016) 385–398

kcj/4
mcj/2
mij/2
y h'
c h'
i

x
θn

'
hi (m) 2.38
' (m) 2.84
hc
2
mij (kN⋅s /m) 5564
2
mij/2 (kN⋅s /m) 2782
z mcj (kN⋅s2/m) 2770
2
mcj/2 (kN⋅s /m) 1385
z
y x kcj (kN/m) 16.4
kcj/4 (kN/m) 4.1 x
Fig. 2. Perspective view, plan and vertical cross section of the finite element model, and parameter values relevant to the impulsive and convective masses.


0
hc cosh 3:68 hDL  2:01 4. Performance assessment analysis in current conditions
¼1
 ð4Þ
hL 3:68 hL  sinh 3:68 hL
D D 4.1. Modal analysis

  The modal analysis carried out by the finite element model

mc  g hL
kc ¼ 3:68 tanh 3:68
D D
where mL = mi + mc is the total liquid mass, and g is the acceleration
of gravity.
In order to geometrically reproduce the inner wall in the finite
element model, mc and mi were first split in two identical portions,
with masses equal to mc/2, mi/2, and convective mass spring stiff-
ness equal to kc/4 for each spring. Secondly, a 3-D implementation
of the basic 2-D layout of the two-mass model was adopted, so that
the hydrodynamic pressure effects are evenly spread across the cir-
cumference of the vessel walls. This was obtained by subdividing
the water volume into n equal fractions, each one being identified
in plan with a circumferential angle hn equal to 180°/n, which
determines mcj convective and mij impulsive masses, and kcj con-
vective mass spring stiffness of the j-th volume fraction equal to
mcj = mc/n, mij = mi/n and kcj = kc/n (j = 1, . . ., n). Lastly, mcj, mij and
kcj were split in the two sub-portions of each volume fraction.
The parameters relevant to each sub-portion, i.e. mcj/2 for the
two split convective masses, mij/2 for the two split impulsive
masses, and kcj/4 for the four convective mass springs, are visual-
ized in the vertical cross section of the finite element model shown
in the right image of Fig. 2.
The influence of n on the finite element model response was
evaluated in [21] by varying it from 6 (i.e. hn = 30°) to 18
(hn = 10°), showing an acceptable reproduction of the analytical
hydrodynamic pressure distribution on the vessel walls provided
in all cases, and totally satisfactory starting from n = 12 [21]. The
results of modal and time-history analyses are presented in the
next sections for this value. The corresponding mij = mi/12, mcj =
0 0
mc/12, mij/2, mcj/2, kcj/4 values, along with hi and hc , are listed in
Fig. 2 too.
Soil-structure interaction effects, which can significantly affect
the seismic response of elevated tanks situated on soft soils, were
not simulated in this study because of the presence of the hard
substrate underlying the foundations of both tanks, described in
Section 2, which can be assumed as a rigid soil in the dynamic
problem.
described above shows a first translational mode in x and y, deter-
ð5Þ
mined by the convective water masses, with vibration period of
2.58 s and effective modal mass (EMM) equal to 21.9%. The second
mode (translational too in x and y), is related to the impulsive
water masses plus the masses of the tank structure, with vibration
period of 1.29 s and EMM equal to 70.4%. The first and second
mode are sufficient to activate a summed effective modal mass
(SEMM) greater than 90% (namely 92.3%) of the total seismic mass
of the tank. The first and second vertical translational modes have
periods of 0.05 s and 0.02 s, and EMMs of 88.6% and 2.1%, respec-
tively, with SEMM of 90.7%. The first and second rotational modes
around the vertical axis z have periods of 0.9 s and 0.01 s, with
EMMs of 92.4% and 4.1%, and SEMM of 96.5%. 80 modes are needed
to activate about 100% of seismic masses along and around all axes.
4.2. Time-history verification and performance assessment analysis
The performance evaluation analysis was carried out for the
four reference seismic levels fixed in the Italian Standards [22],
that is, Frequent Design Earthquake (FDE, with 81% probability of
being exceeded over the reference time period VR); Serviceability
Design Earthquake (SDE, with 50%/VR probability); Basic Design
Earthquake (BDE, with 10%/VR probability); and Maximum Consid-
ered Earthquake (MCE, with 5%/VR probability). The VR period is
fixed at 75 years, which is obtained by multiplying the nominal
structural life VN of 50 years by a coefficient of use cu equal to
1.5, imposed to structures whose seismic resistance is of impor-
tance in view of the consequences associated with their possible
collapse. By referring to topographic category T1 (flat surface),
and C-type soil (deep deposits of dense or medium-dense sand,
gravel or stiff clay from several ten to several hundred meters
thick), the resulting peak ground accelerations for the four seismic
levels referred to the city of Florence are as follows: 0.082 g (FDE),
0.098 g (SDE), 0.223 g (BDE), and 0.27 g (MCE), for the horizontal
motion components; and 0.017 g (FDE), 0.022 g (SDE), 0.079 g
(BDE), and 0.111 g (MCE), for the vertical component. Relevant
elastic pseudo-acceleration response spectra at linear viscous
damping ratio n = 5% are plotted in Fig. 3. The time-history
 S. Sorace et al. / Engineering Structures 106 (2016) 385–398
1 1
0.9 Florence
Pseudo-Acceleration [g]
0.8 MCE Horizontal Component
0.7 ξ=5%
BDE Vn=50 years
0.6 cu=1.5; T1
0.5 C-Type Soil
SDE
0.4
FDE
0.3
0.2
0.1
0 0
0 0.5 1 1.5 2 2.5 3 3.5
Period [s]
Fig. 3. Normative pseudo-acceleration elastic response spectra for Florence city — horizontal and vertical components.
analyses were developed by assuming artificial ground motions as
inputs, generated in families of seven by SIMQKE-II software [23]
from the spectra above, both for the horizontal components (two
families) and the vertical one (one family). In each time-history
analysis the accelerograms were applied in groups of three simul-
taneous components, i.e. two horizontal components, with the first
one selected from the first generated family of seven motions, and
the second one selected from the second family, plus the vertical
component. The solution of the equations of motion was carried
out with the average acceleration method of Newmark-beta fam-
ily. A 0.005 s value of the integration time-step, Dt, was adopted
throughout the analyses, so as to meet accuracy conditions for
the numerical solution with wide margins.
At a first step of the assessment study, the staging members
were modeled as elastic elements. The results of the analyses at
FDE and SDE levels show safe response conditions of columns
and beams, with maximum values of the inter-level drift ratio
(i.e. the ratio of inter-level drift to inter-level staging height) rang-
ing from 0.2% on the first level to 0.26% on the third level (FDE),
and from 0.24% to 0.32% (SDE). These values are below the 0.33%
limitation adopted by [22] at the Operational (OP) performance
level for frame structures interacting with drift-sensitive non-
structural elements. This limitation was assumed as a reference
for the examined water tower too, as no performance threshold
of the inter-level drift for elevated tanks with a frame staging
structure is postulated in codes [18–20], nor in other Standards
on liquid storage tanks. The resulting FDE–OP and SDE–OP correla-
tions assess a satisfactory performance, ensuring adequate protec-
tion to the vertical water pipes and the electrical equipment of the
auxiliary pumping plant, which are the non-structural elements
housed in this structure.
The time-history response to the input motions scaled at the
BDE and MCE levels highlight that bending-compression checks
are not passed by 50% (BDE) and 90% (MCE) of columns, and shear
verifications by 60% (BDE) and 85% (MCE) of beams. As way of
example of the results of the analysis at the MCE, the Mlc,1–Mlc,2
biaxial moment interaction curves (being Mlc,1, Mlc,2 the bending
moments around the local axes 1 and 2 of columns in plan)
obtained in the most stressed sections of columns C1 and C2
(according to column numbering in the plan of Fig. 1), i.e. the base
sections on the first level, from the most demanding among the
seven groups of MCE-scaled accelerograms, are plotted in Fig. 4.
The local axes 1 and 2 are rotated by 45° with respect to the x
and y axes of the global coordinate system for C1, and are parallel
to them (axis 1 to x, and axis 2 to y) for C2. The boundary of the
Mlc,1–Mlc,2 safe interaction domain of columns traced out for the
value of the axial force N referred to the basic combination of grav-
ity loads, named Ns in the following (equal to 330 kN for the base
sections on first level), is also displayed as a reference in the two
389
0.9 Florence
Pseudo-Acceleration [g]
0.8 Vertical Component
0.7 ξ=5%
Vn=50 years
0.6 cu=1.5; T1
0.5 C-Type Soil
MCE
0.4
BDE
0.3
0.2 SDE
FDE
0.1
4 0 0.5 1 1.5 2 2.5 3 3.5 4
Period [s]
graphs of Fig. 4. The response curves highlight remarkably unsafe
conditions, with maximum Mlc,1–Mlc,2 combined values about
67% greater than the corresponding values situated on the safe
domain boundary, for column C1, and about 74% greater, for C2.
In view of this very high stress demand, a second step of the
time-history assessment study was developed at the BDE and
MCE levels, where the plastic behavior of beams and columns
was investigated by incorporating lumped plastic hinges at the
end sections of these members. This allowed evaluating the inelas-
tic demand and the attainment of possible collapse response con-
ditions of the frame staging for the two highest normative
earthquake levels. The plastic hinges included in SAP2000NL basic
library adopted for this analysis are one-component type, and take
into account the influence of the axial force N on the response
cycles of members at each step of the analysis, starting from the
initial value in static conditions, i.e. Ns. The non-linear response
of the hinges is governed by a bilinear skeleton curve, built by
assigning the yielding and ultimate values of the resisting
moments, My and Mu, and chord rotations, hy and hu, of the member
end sections. In particular, the two chord rotations were computed
by expressions (8.7.2.1a) and (C8A.6.1) provided by the Commen-
tary to the Italian Technical Standards [14], not reported here for
brevity’s sake, substantially coinciding with formulas (A.10a) and
(A.1) of Eurocode 8–Part 3 [24]. Since the response of the plastic
hinge elements is expressed in terms of plastic chord rotation,
hpl, rather than total (elastic plus plastic) rotation, the response
capacity of columns and beams was evaluated by comparison with
the ultimate value of hpl, hpl,u, obtained by subtracting hy from hu.
For columns, hpl,u = 0.406  102 radians was computed for the
base sections on the first level, with Ns = 330 kN, as mentioned
above.
This hpl,u value was preliminarily compared to the limit plastic
chord rotation angle, hpl,lim,CP, suggested by ASCE 41-06 Recom-
mendations for the structural rehabilitation of existing buildings
[25] with regard to the inelastic response of R/C beams and col-
umns at the CP performance level. The hpl,lim,CP value was selected
from Table 6.8 of ASCE 41-06 document according to the accep-
tance criteria for columns controlled by flexure, by referring to
two further evaluation criteria, as follows: transverse reinforce-
ment not meeting ductility requirements in the flexural hinge
Vpffiffiffi
region; and shear ratio 6 3, with V = maximum computed
bw dc fc
shear, bw = web width (coinciding with the base section side for
non-flanged columns), dc = distance from extreme compression
fiber to centroid of tension reinforcement, and fc = concrete com-
pressive cylinder strength, to be divided by the confidence factor
FC (equal to 1.2 in this case, as observed in Section 2). Based on
the value of the normalized axial force r N ¼ AgNf c , with Ag = gross sec-
tion area, calculated for N = Ns, the following CP-related plastic
390
150
C1 Column
100
50
Mlc,2 [kNm]
0 0
-50
-100
-150
-150 -100 -50
Fig. 4. Mlc,1–Mlc,2 interaction curves for columns C1 and C2 of the elastic model obtained from the most demanding MCE-scaled group of input accelerograms.
rotation limit was derived for the first level base sections by inter-
polating the data in the above-mentioned Table 6.8 of ASCE 41-06:
hpl,lim,CP = 0.54  102 radians, which is greater than hpl,u. The same
also occurs for the first level top sections, as well as for the base
and top column sections on all other levels, for which hpl,lim,CP
ranges from 0.55  102 to 0.6  102 radians. Hence, a compar-
ison between hpl,u and hpl,lim,CP points out that the ultimate plastic
rotation capacity of columns computed according to [14,24] is
lower than the CP-related ASCE 41-06 limitation. In order to assess
whether the attainment of hpl,u actually corresponds to the numer-
ical collapse of the finite element model, and by considering also
the beneficial effect of the axial force increase in seismic response
conditions on the ultimate plastic rotation of columns, the latter
was not limited to hpl,u in the non-linear time-history analysis,
but it was left free to reach a possible numerical ultimate state cor-
responding to the algorithmic divergence of the solution.
The analyses carried out with the input motions scaled at the
BDE level highlight plastic rotations in 12 out of 24 columns, with
maximum hpl values equal to 0.267  102 radians, attained in the
first level base sections. Furthermore, shear verifications are not
met by 60% of beams. These data assess severely damaged
response conditions, but with some residual margins toward struc-
tural collapse, which allows meeting the basic requirement of the
Collapse Prevention (CP) performance level.
The results of the analyses at the MCE show that divergence of
the solution is reached, due to the collapse of the six columns on
the first level, and thus of the model, for all seven groups of input
ground motions. As way of example of the pre-collapse hysteretic
response obtained, the Mlc,2–hpl bending moment–plastic rotation
response cycles in the base section of a first level column (namely
C2) induced by the most severe among the seven groups of input
150
100
Moment [kNm]
50
0 and a comparative analysis with an alternative PED retrofit mea-
-50
-100
-150
- 0.6 -0.4 -0.2 0 0.2
Plastic Rotation [radians]
Fig. 5. Response cycles of the base plastic hinge of first-level C2 column obtained
from the most demanding MCE-scaled group of input accelerograms.
S. Sorace et al. / Engineering Structures 106 (2016) 385–398
150
C2 Column
100
50
Mlc,2 [kNm]
-50
Safe Safe
-100
Domain Domain
-150
0 50 100 150 -150 -100 -50 0 50 100 150
Mlc,1 [kNm] Mlc,1 [kNm]
motions are plotted in Fig. 5, showing a divergence of the numer-
ical solution after three plastic cycles. Similar results are observed
for the base and top sections of all other first level columns. The
corresponding maximum pre-collapse inter-level drift ratios are
equal to 1.08%, 1.46%, 1.4%, and 1.02% on the first through fourth
level. This confirms a more accentuated drift demand on the two
intermediate levels, already observed in elastic response condi-
tions at the FDE and SDE, also in the presence of the severe plastic
activity induced by the MCE-scaled input actions. The 30–40%
greater drift demand on intermediate staging levels suggests to
tentatively assign a greater damping capacity to the dissipators
placed on these two levels in the design of the two dissipative
bracing retrofit solutions, as discussed in Sections 5 and 6.
The response of the vessel wall results to be safe up to the MCE.
As a consequence, a seismic rehabilitation intervention is needed
only for the frame staging structure. The checks on the foundation,
which was examined separately in order to limit the dimensions of
the finite element model in the non-linear time-history computa-
tions, were passed by all members. Furthermore, the stress states
transferred by the foundation to the hard substrate reach peak
local values far from the estimated ultimate bearing capacity of
the latter.
5. Dissipative bracing retrofit solution incorporating FV spring-
dampers
5.1. Characteristics of the protective system
Fluid viscous dampers are among the most widely used types of
rate-dependent PED devices installed in dissipative bracing tech-
nologies worldwide. This is owed to their high damping capacities,
limited visual impact, competitive costs and, in the case of pressur-
ized elements, inherent self-centering qualities [26–28]. Within
the class of dissipative braces incorporating FV devices, a special
system has been studied for several years by the first two authors,
focusing attention on its application to frame buildings [29–33]. A
new research study was recently started to extend the use of the
system to the seismic retrofit of elevated tanks with frame staging
[21]. The study is developed here, focusing on the design aspects
sure, discussed in the next Section.
A typical cross section of a FV spring-damper mounted in the
First level
fluid viscous dissipative bracing system, named FV-DB system in
C2 Column the following, is shown in Fig. 6. The time-dependent Fd damping
x10-2 and Fne non-linear elastic reaction forces corresponding to the
0.4 0.6
damper and spring functions are effectively simulated by the
following analytical expressions [27,34]:
F d ðtÞ ¼ c sgnðs_ ðtÞÞjs_ ðtÞja
ð6Þ
 S. Sorace et al. / Engineering Structures 106 (2016) 385–398
Shear connectors
IPE 100 profile
Internal casing
Locknuts
External casing
Threaded steel bar
Interfacing plate Piston
Seal Stop-block
Connection flange
Silicone fluid
Fig. 6. FV-DB retrofit solution. Cross section of a fluid-viscous spring-damper, and installation details.
ðk1  k2 ÞsðtÞ
F ne ðtÞ ¼ k2 sðtÞ þ   5 1=5
 
1 þ k1FsðtÞ 
0
where t = time variable; c = damping coefficient; sgn() = signum
function; s_ ðtÞ = device velocity; || = absolute value; a = fractional
exponent, ranging from 0.1 to 0.2 [34]; F0 = static pre-load force;
k1, k2 = stiffness of the response branches situated below and
beyond F0; and s(t) = device displacement.
The installation layout of the FV springs-dampers, illustrated in
the elevation views of Fig. 6, is similar to the basic configuration
adopted for frame buildings. The only difference is represented
by the fact that, in consideration of the relatively small cross sec-
tion of the staging R/C beams, the FV devices are in this case con-
nected to a steel profile placed at the bottom side of the beams,
rather than directly to them. Consistently with the axial symmetry
of the staging structure, the dissipators are installed in all bays, as
well as on all levels, which determines a total of 48 (24 pairs)
devices.
For the development of the numerical analyses, the finite ele-
ment model of FV spring-dampers is obtained by combining in par-
allel a non-linear dashpot element and a non-linear spring element
with reaction forces given by (6) and (7), respectively [29]. Both
types of elements are currently incorporated in commercial struc-
tural analysis programs, such as the SAP2000NL code used in this
study.
5.2. Design of the FV spring-dampers
The design of the FV devices was developed by adapting to the
frame staging of the tank the general criterion formulated in [29]
with reference to multi-story buildings, where the set of devices
to be installed on a story is assigned the capability of dissipating
a prefixed fraction of the maximum seismic input energy com-
puted by the numerical model of the structure on that story. Adap-
tation is imposed by the absence of intermediate floors in the
frame staging, and thus of relevant subdivision in stories along
the height, which causes the three ring beams situated on the first,
second and third level to act essentially as tie-beams to reduce the
effective length of the columns, while transmitting bending
moments to them (only the fourth level beam bears the vessel
floor). Consistently, in this case the energy balance is more suitably
391
R/C beam Out-of-plane constraint
Cap
Hinge bar
computed for the overall structure rather than story-by-story, as
ð7Þ
summed up in the following expression:
EtD ¼ bEI
ð8Þ
where EtD is the tentatively estimated value of ED energy dissipated
by the whole set of installed spring-dampers, b is the preliminary
fixed energy ratio, and
Z tc
EI ¼ € t dug
mu ð9Þ
0
is the ‘‘absolute” input energy [35], being tc the instant at which the
maximum drift — intended here as inter-level drift of the staging
structure — is reached, m the seismic mass of the structure, u € t the
absolute acceleration, and ug the ground displacement.
The target design objectives of the retrofit intervention consist
in attaining a completely elastic response of the staging members
and restraining the inter-level drifts below the 0.5% drift limitation
assumed in [22] for R/C frame structures at the Immediate Occu-
pancy (IO) performance level, up to the MCE. The drift-related
objective corresponds to a reduction of a rounded factor 3 of
the above-mentioned 1.46% and 1.4% peak values surveyed at the
pre-collapse instant — coinciding with instant tc in (9) — on the
second and third level in current conditions. In order to properly
fix b value in (8), since the examined case study represents the first
application of the energy-based design criterion to a frame water
tower, reference is tentatively made to the suggestions offered in
[29,31] for multi-story building structures, where drift reductions
of 3 are met for b approximately equal to 0.8. By assuming this ten-
tative b choice, and considering that the EI value computed via
time-history analysis for the most demanding group of input
motions is equal to 119.5 kN m, EtD = 95.6 kN m comes out from
(8). It is noted that the contribution to EI of the energy dissipated
by the plastic hinges of the finite element model, Ep,h, amounts to
97.4 kN m, i.e. near the tentatively estimated EtD value. Hence, the
drift-related b choice of 0.8 also allows transforming the inelastic
energy demand calculated for the staging members in supplemen-
tal damping energy dissipated by the protective system, thus
achieving a completely elastic response of the structure, as
contextually targeted at the design stage.
The EtD demand can be met by a proper combination of two of
the smaller FV device types in current production (named A and
B-types in the following), characterized by a nominal energy
392 S. Sorace et al. / Engineering Structures 106 (2016) 385–398
dissipation capacity, En, equal to 1.1 kN m (A-type) and 1.5 kN m
(B-type) [36]. Consistently with the pre-collapse peak drift distri-
bution along the height discussed in Section 4.2, A-type spring-
dampers are tentatively placed on the first and fourth level, and
B types on the second and third level, so as to supply a damping
capacity approximately proportional to the maximum inter-level
drift demand. The results of the time-history analyses carried out
for this first design hypothesis, not detailed here for brevity’s sake,
show that the performance objectives are practically met (with
only the second level drift, equal to 0.52%, slightly exceeding the
assumed 0.5% drift limitation), but with maximum displacements
of A-type devices very near to the available stroke, smax, equal to
±15 mm. This suggests installing model B spring-dampers on the
first and fourth level too, as final design solution. The remaining
values of the mechanical parameters of these devices are as fol-
lows: c = 2.94 kN(s/mm)a; a = 0.2; k2 = 910 N/mm; k1 = 15;
k2 = 13,650 N/mm; F0 = 5 kN; smax = ±20 mm; and maximum reac-
tion force, Fmax, equal to ±30 kN. The small value of Fmax also deter-
mines small dimensions of the supporting diagonal trusses, for
which tubular profiles sized 48.3 mm (diameter)  3.2 mm (thick-
ness), made of European S 275 H steel, are adopted.
Based on the outcome of the design process, a new modal anal-
ysis was carried out with the finite element model including the
FV-DB system components, which shows a first translational mode
in x and y associated to the convective water masses, like for the
original tank. The vibration period is coincident (2.58 s), whereas
the effective modal mass is slightly reduced (19.2% instead of
21.9%), owed to the small increase of the frame staging masses
caused by the incorporation of the steel trusses of the protective
system. The second mode is related to the impulsive water masses
plus the masses of the tank structure in protected configuration
too, with vibration period of 1.03 s (i.e. 20.1% lower than the period
of the corresponding mode of the as-built structure) and EMM
equal to 73.9%, yielding a SEMM of the first and second horizontal
translational modes equal to 92.8%. The first and second vertical
translational modes have periods of 0.05 s and 0.015 s, and EMMs
of 91.2% and 1.8%, respectively, with SEMM of 93%. The first and
second rotational modes around z have periods of 0.72 s and
0.01 s, with EMMs of 93.7% and 3.2%, and SEMM of 96.9%. 43
modes are needed in this case to activate about 100% of seismic
masses along and around the three axes.
5.3. Time-history verification and performance assessment analysis in
retrofitted conditions
The results of the time-history verification analyses in rehabil-
itated conditions are synthesized in Figs. 7–10, all referred to the
response induced by the most demanding group of input ground
motions scaled at the MCE level. It can be observed that the
150
C1 Column
100
50
Mlc,2 [kNm]
0 0
-50
-100 Safe
Domain
-150
-150 -100 -50 0 50 100
Mlc,1 [kNm]
Fig. 7. FV-DB retrofit solution. Mlc,1–Mlc,2 interaction curves for columns C1 and C2 of the elastic model obtained from the most demanding MCE-scaled group of input
accelerograms.
maximum values displayed in these Figures are about 5% (drifts
and moments) to about 9% (FV device forces and dissipated energy)
greater than the corresponding mean values computed from the
response to the seven groups of accelerograms.
The Mlc,1–Mlc,2 interaction curves of the base sections of col-
umns C1 and C2, plotted in Fig. 4 above for the original structure,
are duplicated in Fig. 7 in retrofitted configuration, highlighting
that the protective action of the FV-DB system allows confining
the curves within the biaxial moment safe domain. This substantial
performance enhancement, reached in all remaining columns too,
is confirmed by the results of the additional analysis carried out
with the finite element model incorporating the plastic hinges in
columns and beams, whose response becomes totally elastic. At
the same time, all beams meet shear verifications. The base shear
demand/capacity ratio of the frame staging structure is equal to
0.94, that is, 20% lower than the response in current conditions.
The drift ratio diagram displayed in Fig. 8, in superimposition to
the corresponding curve in original conditions, shows peak values
reduced to 0.42%, 0.48%, 0.49% and 0.45%, for the first through
fourth levels of the frame staging, meeting the targeted 0.5% IO-
related drift limitation. The damping action of the FV-DB system
also determines a more uniform drift profile along the height, with
maximum 20% differences among the peak drifts on the four levels,
as compared to about 40% differences in current configuration
commented above.
The total reaction force–displacement [(Fd(t) + Fne(t)]  s(t)
cycles of the four pairs of FV dissipators situated along one of the
six vertical alignments (specifically the one enclosed between col-
umns C1 and C2) are visualized in Fig. 9. The cycles exhibit peak
displacements equal to 10.1 mm (second level), i.e. half the avail-
able stroke limit of ±20 mm mentioned above. Similar results are
obtained for the other vertical alignments.
The time-histories of the energy dissipated by the FV-DB system,
plotted in Fig. 10 for x direction (similar to the response histories
along all directions in plan, due to the axial symmetry of the struc-
ture), show that the value computed at the end of the analysis, ED,f,
equal to 98.6 kN m, is only 3% greater than the EtD value of
95.6 kN m assumed to initialize the design process. The contribu-
tions of the sets of spring-dampers installed on the four levels, plot-
ted in the same graph, are equal to 23.9 kN m, 27.6 kN m, 26.6 kN m
and 20.5 kN m, i.e. 24.2%, 28%, 27% and 20.8% of the FV-DB dissi-
pated energy, on the first through fourth level, respectively.
The results of the energy response histories were elaborated to
estimate the equivalent linear viscous damping ratio ne, so as to
obtain a conventional quantification of the dissipative capacity of
the FV-DB system. The classical expression
ED
ne ¼ ð10Þ
4p  Ee
150
C2 Column
100
50
Mlc,2 [kNm]
-50
Safe
-100
Domain
-150
150 -150 -100 -50 0 50 100 150
Mlc,1 [kNm]
 S. Sorace et al. / Engineering Structures 106 (2016) 385–398 393

4 structures since the early 1970s [37]. In the following decades, sev-
eral different types of devices belonging to this class, dissipating
energy by flexural, shear or extensional plastic deformation, have
3
FV-DB Solution been designed and implemented [38–45]. The most widely applied
dampers in building structures are flexural Added Damping and
Level

Current Conditions Stiffness (ADAS) devices, consisting in the assembly of multiple

2
1 section linearly varying along the span, similarly to the bending
0 simple geometry, which implies lower manufacturing and installa-
0 0.5 1 1.5
Inter-level drift ratio [%]
Fig. 8. FV-DB retrofit solution. Maximum inter-level drift ratio profile obtained
from the most demanding MCE-scaled group of input accelerograms, and compar-
ison with the corresponding profile in current conditions.
was applied to this purpose, where Ee is the elastic strain energy,
which provided an average ne value of 31.6% for the MCE-scaled
input motions.
FDE, SDE and BDE-related performance too, not detailed here
for brevity’s sake, results to be significantly improved by the retro-
fit intervention. This is assessed by peak inter-level drift ratios
reduced by about 55% at FDE and SDE, as compared to current con-
ditions, and equal to 0.31% at BDE, which allows meeting the 0.33%
limitation relevant to the OP level for this earthquake level too, in
addition to FDE and SDE.
6. Dissipative bracing retrofit solution incorporating steel
T-ADAS dampers
6.1. Characteristics of the protective system
Displacement-dependent steel hysteretic dampers have been
used as PED devices for the seismic protection of bridges and
X-shaped (X-ADAS) or triangle-shaped (T-ADAS) steel plates
mounted in parallel, which act as short cantilever beams with cross
moment diagram. The success of these dampers is owed to their
tion costs as compared to any other type of steel yielding technolo-
gies, among which the Buckling Restrained Bracing (BRB) systems
growingly applied in retrofit and new designs, but still more
expensive. Based on these observations, ADAS devices were
adopted for the second retrofit solution of the case study tank,
which allowed developing a more direct comparison with the
intervention based on the incorporation of the FV-DB system. T-
ADAS types were selected, because they involve the simplest con-
nections to the supporting trusses, thanks to the hinged bottom
section of the constituting plates. The typical installation layout
of ADAS and T-ADAS dampers within the frame skeleton is the
same as for the fluid viscous dissipators in the FV-DB system, that
is, on top of inverse-chevron trusses.
6.2. Design of the T-ADAS dampers
By referring to the geometrical parameters highlighted in the
left drawing of Fig. 11, the design of T-ADAS devices was carried
out for the following standard dimensions of the isosceles triangu-
lar plates [40,41]: H = height = 150 mm, T = thickness = 15 mm, and
B = base = 50, 75 or 100 mm. The constituting steel is European S
275 type, like for the supporting diagonal trusses, with yielding
stress, fy, of 275 MPa, ultimate stress, ft, of 430 MPa, and Young
modulus, Es, equal to 200,000 MPa. The only explicit design vari-
ables are B and the number of plates per device, Np. Based on the

50 50
40 st
1 Level 40 nd
2 Level
30 30
Reaction Force [kN]

Reaction Force [kN]

20 20
10 10
0 0
-10 -10
-20 -20
-30 -30
-40 -40
-50 -50
-15 -10 -5 0 5 10 15 -15 -10 -5 0 5 10 15
Displacement [mm] Displacement [mm]
50 50
40 rd
3 Level 40 th
4 Level
30 30
Reaction Force [kN]

Reaction Force [kN]

20 20
10 10
0 0
-10 -10
-20 -20
-30 -30
-40 -40
-50 -50
-15 -10 -5 0 5 10 15 -15 -10 -5 0 5 10 15
Displacement [mm] Displacement [mm]

Fig. 9. FV-DB retrofit solution. Response cycles of the spring-damper pairs situated on the four levels in the bay enclosed between columns C1 and C2 obtained from the most
demanding MCE-scaled group of input accelerograms.
394 S. Sorace et al. / Engineering Structures 106 (2016) 385–398

120 BT 2
Pu ¼ f y ð15Þ
4H
100 FV-Dissipated Energy

Pu  Py
Energy [kNm]

80
60
2nd Level
40 3rd Level
20
st
1 Level
0 4th Level
0 5 10 15 20 25 30
Time [s]
Fig. 10. FV-DB retrofit solution. Energy time-histories in x direction obtained from
the most demanding MCE-scaled group of input accelerograms.
dimensions of the staging beams, the intermediate value of the B
range above, i.e. 75 mm, was selected, so as to obtain the best geo-
metrical proportions of the dissipator-diagonal truss-R/C beam
joints. In order to properly establish the design value of Np, the pre-
liminary sizing criterion applied to FV spring-dampers was
extended to T-ADAS devices too. Therefore, the b energy ratio
and relevant EtD energy value were fixed at 0.8 and 95.6 kN m also
for this second retrofit measure. Then, to meet design targets, the
energy dissipation capacity of each plate was estimated for device
displacements no greater than the values corresponding to the
assumed IO-related inter-level drift ratio limitation of 0.5%. Since
the greatest contribution to inter-level displacements is given by
the plastic deformation of the dissipators, about 75% of the drift
ratio was preliminarily assumed to be concentrated in them at this
stage. Thus, the estimated peak displacement of the dampers (and
thus of each constituting plate), dmax,e, was tentatively assumed
equal to (0.75  0.5%) = 0.375% of inter-level heights, i.e. 14 mm.
The parameters of the P(t)–d(t) force–displacement response
cycle of each plate, sketched with idealized bilinear shape in the
right drawing of Fig. 11, are: Py = yielding force, dy = yielding dis-
placement, ke = stiffness of the elastic branch, kp = stiffness of the
post-elastic branch, Pu = ultimate force, and du = ultimate displace-
ment. Relevant expressions are summarized below:
BT 2
Py ¼ f y
6H
Py
dy ¼
ke
Es BT 3
ke ¼
6H3
kp ¼ cke
du ¼ dy þ ð16Þ
kp
where c is the strain hardening ratio characterizing the slope of the
post-elastic branch, normally assumed as equal to 0.03. Based on
the fy, Es, B, T, H and c values above, the following response param-
eters are obtained by means of relations (11)–(16): Py = 5156 N,
dy = 2.1 mm, ke = 2500 N/mm, kp = 75 N/mm, Pu = 7734 N, and
du = 36.4 mm. The energy dissipated by a plate in a cycle bounded
by the maximum estimated (positive and negative) displacement
dmax,e, EtD;pl;m , given by the geometrically subtended area, is equal
to EtD;pl;m ¼ 2P y  2ðdmax;e  dy Þ ¼ 245 Nm ¼ 0:245 kN m. By observ-
ing that the energy dissipated in the remaining cycles, characterized
by smaller amplitudes, can be estimated to be about 4 times EtD;pl;m ,
the total energy dissipated by a plate, EtD;pl;tot , is tentatively calcu-
lated as follows: EtD;pl;tot ¼ ð1 þ 4Þ  0:245 ¼ 1:23 kN m. In order to
suitably fix Np, it must also be noted that, for any given input direc-
tion of seismic action, the plastic energy demand is not uniformly
spread among the devices placed on the six vertical alignments,
as a consequence of their different orientations within the hexago-
nal staging plan. This causes to incorporate a total number of plates
about 50% greater than the number computed as the ratio of EtD to
EtD;pl;tot . Hence, the following tentative total number of plates, N tp ,
is obtained: Ntp ¼ 1:5  EtD =EtD;pl;tot ¼ 1:5  ð95:6=1:23Þ ¼ 117,
rounded to 120 for the inherent approximations of the preliminary
sizing process.
Like for the FV-DB system, a first distribution of the set of 120
plates along the four staging levels is formulated by referring to
the pre-collapse peak drifts in current conditions. The following
proportions are assumed accordingly: 20% of N tp on the first and
fourth level, and 30% on the second and third level. This correspond
to 24 plates, i.e. 4 plates per damper, on the first and fourth level,
and 36 plates, i.e. 6 plates per damper, on the second and third
ones. Drawings of the two types of devices, and relevant installa-
tion details, are displayed in Fig. 12. The tubular profiles of sup-
porting trusses have sections of 88.9 mm  2.6 mm.
ð11Þ
A new modal analysis was carried out for the ADAS-DB solution
too, showing again a first translational mode in x and y associated
ð12Þ to the convective water masses, with vibration period of 2.58 s, and
effective modal mass of 16.3%. The latter is reduced further as com-
pared to the 19.2% fraction obtained for the FV-DB solution, owing
to the slightly greater own weights of the ADAS-DB system. The
ð13Þ
second mode is related to the impulsive water masses plus the
masses of the tank structure in this case too, with period of
ð14Þ 0.95 s (i.e. 26.4% lower than the period of the corresponding mode

Fig. 11. Geometrical and mechanical parameters of the steel plates constituting the T-ADAS dampers.
 S. Sorace et al. / Engineering Structures 106 (2016) 385–398 395

R/C beam R/C beam

Shear connectors Shear connectors
IPE 100 profile IPE 100 profile

6-plate T-ADAS damper 4-plate T-ADAS damper

Fig. 12. ADAS-DB retrofit solution. Views of the 6 and 4-plate T-ADAS dampers, and installation details.

150 150
C1 Column C2 Column
100 100

50 50
Mlc,2 [kNm]

Mlc,2 [kNm]

0 0

-50 -50
Safe
Safe Safe
Safe
-100 -100
Domain
Domain Domain
Domain
-150 -150
-150 -100 -50 0 50 100 150 -150 -100 -50 0 50 100 150
Mlc,1 [kNm] Mlc,1 [kNm]

Fig. 13. ADAS-DB retrofit solution. Mlc,1–Mlc,2 interaction curves for columns C1 and C2 of the elastic model obtained from the most demanding MCE-scaled group of input
accelerograms.

case to activate about 100% of seismic masses along and around

4
the three axes.

3 6.3. Time-history verification and performance assessment analysis in

ADAS-DB Solution
retrofitted conditions
Level

2 Current Conditions
1 sure are summarized in Figs. 13–16. The Mlc,1–Mlc,2 interaction
0
0 0.5 1 1.5
Inter-level drift ratio [%]
Fig. 14. ADAS-DB retrofit solution. Maximum inter-level drift ratio profile obtained
from the most demanding MCE-scaled group of input accelerograms, and compar-
ison with the corresponding profile in current conditions.
of the as-built structure) and EMM equal to 75.3%, which provides
a SEMM of the first and second horizontal translational modes
equal to 91.6%. The first and second vertical translational modes
have periods of 0.05 s and 0.015 s, and EMMs of 87.9% and 2.7%,
respectively, with SEMM of 90.6%. The first and second rotational
modes around z have periods of 0.68 s and 0.12 s, with EMMs of
92.2% and 4.2%, and SEMM of 96.4%. 49 modes are needed in this
Similarly to Figs. 7–10 for the FV-DB solution, the results of the
time-history verification analyses for the ADAS-DB retrofit mea-
curves of the base sections of columns C1 and C2, plotted in
Fig. 13, are restrained within the biaxial moment safe domain in
this case too. Safe conditions are assessed as well by the elastic
response of the finite element model including the plastic hinges.
The base shear demand/capacity ratio of the structure is equal to
0.97, i.e. similar to the value obtained for the FV-DB retrofit, and
17% lower than the value in unprotected conditions.
Peak drift ratios are equal to 0.37%, 0.48%, 0.49% and 0.4%, for
the first through fourth levels of the frame staging (Fig. 14). The
assumed IO-related threshold of 0.5% is met in this case too, with
drift values coinciding with the ones obtained in the FV-DB case,
on the second and third level, and slightly lower, on the first and
fourth level.
The P(t)–d(t) cycles of the four T-ADAS devices placed along the
same vertical alignment referred to in the graphs of Fig. 9 relevant
to the FV-DB solution, are plotted in Fig. 15. The maximum
396 S. Sorace et al. / Engineering Structures 106 (2016) 385–398

50 50
st nd
40 1 Level 40 2 Level
30 30

Reaction Force [kN]

Reaction Force [kN]

20 20
10 10
0 0
-10 -10
-20 -20
-30 -30
-40 -40
-50 -50
-15 -10 -5 0 5 10 15 -15 -10 -5 0 5 10 15
Displacement [mm] Displacement [mm]

50 50
rd th
40 3 Level 40 4 Level
30 30
Reaction Force [kN]

Reaction Force [kN]

20 20
10 10
0 0
-10 -10
-20 -20
-30 -30
-40 -40
-50 -50
-15 -10 -5 0 5 10 15 -15 -10 -5 0 5 10 15
Displacement [mm] Displacement [mm]

Fig. 15. ADAS-DB retrofit solution. Response cycles of the dampers situated on the four levels in the bay enclosed between columns C1 and C2 obtained from the most
demanding MCE-scaled group of input accelerograms.

120 devices installed on the four levels, equal to 18.3 kN m (16.3%),

34.5 kN m (34.9%), 29.2 kN m (29.6%), and 17 kN m (17.2%), are
100 ADAS-Dissipated Energy more differentiated than those computed for the FV spring-
dampers. This is a consequence of the different number of plates
Energy [kNm]

80
60
40
20
0 th
0 5 10 15 20 25
Time [s]
Fig. 16. ADAS-DB retrofit solution. Energy time-histories in x direction obtained
from the most demanding MCE-scaled group of input accelerograms.
displacement, reached on the second and third level, is equal to
about 13.9 mm in both cases, i.e. approximately equal to dmax,e.
This confirms that about 75% of drift ratio is actually produced
by the plastic deformation of the dampers, as predicted at design
stage. The same proportion is also kept for the devices installed
on the first and fourth level. The areas subtended by the cycles
visualized in the four graphs are equal to 5.28 kN m (first level),
7.18 kN m (second), 7.03 kN m (third), and 5.04 kN m (fourth). By
dividing these values by the respective numbers of constituting
plates, the following values of the total dissipated energy per plate
come out: 1.32 kN m, 1.2 kN m, 1.17 kN m, and 1.26 kN m, for the
devices mounted on the first through fourth level, all differing by
less than 10% from EtD;pl;tot .
The energy time-histories graphed in Fig. 16 for x direction
show a ED,f value equal to 98.8 kN m, nearly coincident with the
corresponding value computed for the FV-DB retrofit solution,
and 3.7% greater than EtD . The contributions of the sets of T-ADAS
incorporated in the two intermediate levels, and thus of the greater
damping capacity of relevant steel dissipators. The equivalent vis-
2nd Level cous damping ratio ne is equal to 31.8%, nearly coincident with the
3rd Level value computed for the FV-DB solution at the MCE. The response to
the FDE, SDE and BDE-scaled input actions is similarly improved
too, which allows meeting the OP performance level up to the
1st Level BDE also for the ADAS-DB retrofit intervention.
30 4 Level The study was completed by varying the distribution and num-
ber of plates with respect to the 24 + 36 + 36 + 24 design solution,
so as to assess the influence of both parameters on the system
performance. Various different distribution hypotheses were
considered by fixing the number of plates at 120, among which
the 30 + 30 + 30 + 30 uniform installation along the height
showed the best performance. Peak drifts of 0.34%, 0.51%, 0.53%
and 0.38% resulted in this case, differing by no more than 8%
from the 24 + 36 + 36 + 24 basic design layout, but slightly exceed-
ing the OP-related 0.5% limitation on the second and third level.
The solutions based on the incorporation of a greater number of
plates (30 + 36 + 36 + 30, 30 + 42 + 42 + 30, etc.) do not provide
appreciable performance improvements, because the proportion-
ally increased stiffness of ADAS devices prevents practically
exploiting their additional plastic capacity. This does not motivate
the installation of a greater number of plates as compared to the
basic design choice.
A final dimensional evaluation was carried out by comparison
with a conventional non-dissipative bracing retrofit solution devel-
oped for the same performance objectives met by the PED systems,
i.e. maximum inter-level drifts below the 0.33% OP-related (up to
the BDE) and 0.5% IO-related (MCE) drift limitations, and elastic
response of all members up to the MCE, bracing trusses included.
Based on these assumptions, the design of the bracing system
 S. Sorace et al. / Engineering Structures 106 (2016) 385–398
150
C1 Column
100
50
Mlc,2 [kNm]
0 0
-50
-100
-150
-150 -100 -50 0 50
Mlc,1 [k]Nm
Fig. 17. Non-dissipative bracing retrofit solution. Mlc,1–Mlc,2 interaction curves for columns C1 and C2 of the elastic model obtained from the most demanding MCE-scaled
group of input accelerograms.
4
3 at the BDE and MCE levels, respectively.
Non-Dissipative Solution
Level
2 Current Conditions
1
0
0 0.5 1 1.5
Inter-level drift ratio [%]
Fig. 18. Non-dissipative bracing retrofit solution. Maximum inter-level drift ratio
profile obtained from the most demanding MCE-scaled group of input accelero-
grams, and comparison with the corresponding profile in current conditions.
was developed by adopting a behavior factor equal to 1. The
resulting dimensions of the tubular trusses are as follows:
168.3 mm  4 mm (first and fourth level), and 219.4 mm  4 mm
(second and third level). By way of example of the response
obtained from the time-history verification analyses at the MCE,
the Mlc,1–Mlc,2 interaction curves of the base sections of columns
C1 and C2, and the inter-level drift ratio peak profile are replicated
in Figs. 17 and 18, highlighting the attainment of the design objec-
tives. The main difference with the performance of the two dissipa-
tive designs is represented by the axial forces in the bracing
trusses, which are about 3 times greater for the non-dissipative
solution. This determines proportionally higher stress states in
the column–beam–brace joints, which requires a steel jacketing
strengthening intervention of the end zones of columns and
beams. Furthermore, the resulting base shear is 2.65, 3.28 and
3.18 times the values computed in original, FV-DB retrofitted and
ADAS-DB retrofitted conditions, respectively, which determines
unsafe stress states in the foundation blocks. A strengthening
intervention is needed for these elements too, which consists in
adding a 150 mm-thick continuous R/C cap slab to the top and
the sides of the six members. The side additions also determine
an enlargement of 300 mm per side of the base section of the
blocks, which allows reaching safe conditions on the foundation
soil too.
7. Conclusions
The multi-stage assessment analysis developed on the water
tower designed in 1930s by Pier Luigi Nervi for Santa Maria Novella
Station in Florence allowed evaluating the poor seismic
397
150
C2 Column
100
50
Mlc,2 [kNm]
-50
Safe
Safe Safe
Safe
-100
Domain
Domain Domain
Domain
-150
100 150 -150 -100 -50 0 50 100 150
Mlc,1 [kNm]
performance of the structure, highlighted by the attainment of
near-collapse and collapse conditions under seismic action scaled
The two PED-based retrofit measures of the frame structure
offer excellent growth in performance, guaranteeing an elastic
response of the R/C members up to the MCE. This is obtained with
much smaller dimensions of the bracing elements as compared to a
non-dissipative solution designed for the same objectives.
Further specific remarks, arising from the results of the study,
are reported below.
 Severely unsafe conditions are noticed from the response of the
elastic model with input motions scaled at the BDE and MCE
intensities. These data are corroborated by the results of the
non-linear analysis, which shows plastic rotations in 12 out of
24 columns at the BDE, and the numerical collapse of the six
columns on the first level at the MCE.
 The mutual target design objectives of the two retrofit interven-
tions are met for a b energy ratio choice of 0.8, tentatively fixed
according to the suggestions reported in previous studies on
building structures retrofitted with the FV-DB technology to
obtain a drift reduction of 3. Moreover, the selected b value
helps completely transform the inelastic energy demand calcu-
lated for the staging members in supplemental damping energy
dissipated by the protective system, thus achieving their con-
textually targeted elastic response.
 The preliminary sizing criterion applied to the FV spring-
dampers, extended for the first time to T-ADAS devices within
this study, establishes an effective starting point for the design
of the ADAS-DB system too, by-passing a time-expensive itera-
tive search process of the most suitable total number of consti-
tuting steel plates.
 The equivalent viscous damping ratio estimated from the
response to the MCE-scaled input motions is equal to about
32% for both rehabilitation strategies, which is nearly the max-
imum damping level that allows using equivalent elastic code-
based analyses. This value belongs to the typical ne ranges iden-
tified for several FV-DB and ADAS-DB systems incorporated in
building structures, when subjected to earthquake motions of
similar amplitudes to the MCE level considered for this case
study.
 The main differences between the two protective technologies
lie in the undamaged response of the FV devices and their
self-centering capacities, as opposed to the plastic response
and the presence of residual offset displacements in the T-
ADAS dissipators, which causes an extra cost arising from the
need to replace them after the most severe earthquake shaking
considered in the analysis.
398 S. Sorace et al. / Engineering Structures 106 (2016) 385–398
 The about 3 times greater axial forces in the diagonal trusses
computed for the non-dissipative retrofit in comparison to the
two PED systems, imposes to adopt sizeable strengthening
interventions in the end zones of R/C members, as well as in
the foundation, for the former. As a consequence, the conven-
tional bracing solution results to be more intrusive, both from
a structural and an architectural viewpoint, as well as more
expensive, as stressed by approximately doubled costs as com-
pared to the FV-DB and ADAS-DB solutions.
Acknowledgements
The study reported in this paper was sponsored by the Italian
Department of Civil Protection within the ReLUIS-DPC Project
2014/2016 – Research Line Nr. 6: Isolation and Dissipation. The
authors gratefully acknowledge this financial support.
References
[1] Constantinou MC, Soong TT, Dargush GF. Passive energy dissipation systems
for structural design and retrofit. Monograph series no. 1. Buffalo
(NY): MCEER; 1998.
[2] Christopoulos C, Filiatrault A. Principles of passive supplemental damping and
seismic isolation. Pavia, Italy: IUSS Press; 2006.
[3] Symans MD, Charney FA, Whittaker AS, Constantinou MC, Kircher CA, Johnson
MW, McNamara RJ. Energy dissipation systems for seismic applications:
current practice and recent developments. ASCE J Struct Eng 2008;134
(1):3–21.
[4] Di Sarno L, Elnashai AS. Innovative strategies for seismic retrofitting of steel
and composite frames. J Prog Struct Eng Mater 2005;7(3):115–35.
[5] Chan RWK, Albermani F, Williams MS. Evaluation of yielding shear panel
device for passive energy dissipation. J Constr Steel Res 2009;65(2):260–8.
[6] Mazza F, Vulcano A. Control of the earthquake and wind dynamic response of
steel-framed buildings by using additional braces and/or viscoelastic dampers.
Earthq Eng Struct Dynam 2011;40(2):155–74.
[7] Di Sarno L, Manfredi G. Experimental tests on full-scale RC unretrofitted frame
and retrofitted with buckling restrained braces. Earthq Eng Struct Dynam
2012;41(2):315–33.
[8] Weng DG, Zhang C, Lu XL, Zeng S, Zhang SM. A simplified design procedure for
seismic retrofit of earthquake-damaged RC frames with viscous dampers.
Struct Eng Mech 2012;44(5):611–31.
[9] Whittle JK, Williams MS, Karavasilis TL, Blakeborough A. A comparison of
viscous damper placement methods for improving seismic building design. J
Earthq Eng 2012;16(4):540–60.
[10] Martelli A, Forni M, Clemente P. Recent worldwide applications of seismic
isolation and energy dissipation and conditions for their correct use. In:
Proceedings of the 15th world conference on earthquake engineering. Lisbon,
Portugal; 2012, Paper 397.
[11] Drosos JC, Tsinopoulos SV, Karabalis DL. Seismic response of spherical liquid
storage tanks with a dissipative bracing system. In: Proceedings of the GRACM
international congress on computational mechanics. Limassol, Cyprus; 2005.
p. 313–9.
[12] Summers PB, Castellano MG, Bergamo G, Gatti F, Marti J, Poggianti A. Seismic
risk reduction at petrochemical and LNG facilities: main results from INDEPTH
project. In: Proceedings of the 14th world conference on earthquake
engineering. Beijing, China; 2008, Paper 06-0069.
[13] Rai DC. Elevated tanks. In: Jain, Lettis, Bardet, Murty (editors). 2001 Bhuj, India
earthquake reconnaissance report. Earthquake Spectra 2002; Supplement A to
18. p. 279–95.
[14] Commentary on the Technical Standards on constructions. Rome, Italy: Italian
Council of Public Works; 2009 [in Italian].
[15] SAP2000NL. Theoretical and users’ manual. Release 17.03. Berkeley (USA):
Computers & Structures Inc.; 2015.
[16] Housner GW. Dynamic behavior of water tanks. Bull Seismol Soc Am 1963;53
(2):381–7.
View publication stats
[17] Haroun MA, Housner GW. Seismic design of liquid storage tanks. ASCE J Tech
Counc 1981;107(1):191–207.
[18] ACI 350.3-06. Seismic design of liquid-containing concrete
structures. Farmington Hills, MI (USA): ACI, American Concrete Institute;
2006.
[19] IITK-GSDMA. Guidelines for seismic design of liquid storage tanks. Kanpur,
India: National Information Center of Earthquake Engineering; 2007.
[20] EN 1998-4. Eurocode 8: design of structures for earthquake resistance – Part 4:
Silos, tanks and pipelines. Kanpur, India: National Information Center of
Earthquake Engineering; 2007.
[21] Sorace S, Terenzi G, Mori C. Analysis of an elevated water storage tank with R/C
frame staging structure. In: Proceedings of the 14th world conference on
seismic isolation, energy dissipation and active vibration control of structures.
San Diego, CA; 2015.
[22] Technical Standards on constructions. Rome, Italy: Italian Council of Public
Works; 2008 [in Italian].
[23] Vanmarcke EH, Fenton GA, Heredia-Zavoni E. SIMQKE-II – conditioned
earthquake ground motion simulator: User’s manual, version 2.1. Princeton
(USA): Princeton University; 1999.
[24] EN 1998-3. Eurocode 8: Design of structures for earthquake resistance – Part
3: Assessment and retrofitting of buildings. Bruxelles, Belgium: CEN, European
Committee for Standardization; 2005.
[25] ASCE/SEI 41-06. Seismic rehabilitation of existing buildings. Reston, VA:
American Society of Civil Engineers – Structural Engineering Institute; 2006.
[26] Constantinou MC, Symans MD. Experimental study of seismic response of
buildings with supplemental fluid dampers. J Struct Des Tall Build 1993;2
(1):93–132.
[27] Pekcan G, Mander JB, Chen SS. The seismic response of a 1:3 scale model R.C.
structure with elastomeric spring dampers. Earthq. Spectra 1995;11
(2):249–67.
[28] Ponzo FC, Di Cesare A, Nigro D. Visco-re-centring energy dissipating system for
seismic protection of framed buildings. Int J Mech 2013;7:370–8.
[29] Sorace S, Terenzi G. Seismic protection of frame structures by fluid viscous
damped braces. ASCE J Struct Eng 2008;134(1):45–55.
[30] Sorace S, Terenzi G. Fluid viscous damper-based seismic retrofit strategies of
steel structures: general concepts and design application. Adv Steel Constr
2009;5(3):322–39.
[31] Sorace S, Terenzi G, Fadi F. Shaking table and numerical seismic performance
evaluation of a fluid viscous-dissipative bracing system. Earthq Spectra
2012;28(4):1619–42.
[32] Sorace S, Terenzi G. Dissipative bracing-based seismic retrofit of R/C school
buildings. Open J Constr Build Technol 2012;6:334–45.
[33] Sorace S, Terenzi G. Motion control-based seismic retrofit solutions for a R/C
school building designed with earlier Technical Standards. Bull Earthq Eng
2014;12(6):2723–44.
[34] Sorace S, Terenzi G. Non-linear dynamic modelling and design procedure of FV
spring-dampers for base isolation. Eng Struct 2001;23(12):1556–67.
[35] Uang CM, Bertero VV. Use of energy as a design criterion in earthquake-
resistant design. Report No. UCB-EERC 88/18. Berkeley, CA: University of
California at Berkeley; 1988.
[36] Jarret SL. Shock-control technologies; 2014. <http://www.introini.info>.
[37] Kelly JM, Skinner RI, Heine AJ. Mechanisms of energy absorption in special
devices for use in earthquake resistant structures. Bull NZ Nat Soc Earthq Eng
1972;5(3):63–88.
[38] Whittaker AS, Bertero VV, Thompson CL, Alonso LJ. Seismic testing of steel
plate energy dissipation devices. Earthq Spectra 1991;7(4):563–604.
[39] Tsai KC, Chen HW, Hong CP, Su YF. Design of steel triangular plate energy
absorbers for seismic-resistant construction. Earthq Spectra 1993;9
(3):505–28.
[40] Martinez-Romero E. Experiences on the use of supplemental energy
dissipators on building structures. Earthq Spectra 1993;9(3):581–624.
[41] Dargush GF, Soong TT. Behavior of metallic plate dampers in seismic passive
energy dissipation systems. Earthq Spectra 1995;11(4):545–68.
[42] Foti D, Bozzo L, Lopez-Almansa F. Numerical efficiency assessment of energy
dissipators for seismic protection of buildings. Earthq Eng Struct Dynam
1998;27(6):543–56.
[43] Ribakov Y, Gluck J. Optimal design of ADAS damped MDOF structures. Earthq
Spectra 1999;15(2):317–30.
[44] Moreschi LM, Singh MP. Design of yielding metallic and friction dampers for
optimal seismic performance. Earthq Eng Struct Dynam 2003;32(8):1291–311.
[45] Di Sarno L, Elnashai AS. Bracing systems for seismic retrofitting of steel frames.
J Constr Steel Res 2009;65(2):452–65.