Computers and Structures 89 (2011) 1232–1240

Contents lists available at ScienceDirect

Computers and Structures

journal homepage: www.elsevier.com/locate/compstruc

Development of nonlinear analytical model and seismic analyses of a steel

frame with self-centering devices and viscoelastic dampers
Theodore L. Karavasilis ⇑, Tony Blakeborough, Martin S. Williams
Department of Engineering Science, University of Oxford, Oxford OX1 3PJ, UK

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

Article history: This paper highlights the role of advanced structural analysis tools on the conception of high-perfor-
Available online 28 September 2010 mance earthquake-resistant structural systems. A new steel frame equipped with self-centering devices
and viscoelastic dampers is described. A prototype building using this frame is designed and a detailed
Keywords: nonlinear analytical model for seismic analysis is developed. Seismic analyses results show the effective-
Self-centering ness of the proposed frame to enhance structural and non-structural performance by significantly reduc-
Viscoelastic dampers ing residual drifts and inelastic deformations, and by reducing drifts, total floor accelerations and total
Steel MRF
floor velocities. These results are the basis for further studies aiming to develop design methods and cri-
Seismic design
High-performance
teria for the proposed high-performance frame.
Damage-free Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction are becoming very effective in reducing peak structural response

[8]. Dampers made of high-damping elastomer have been tested
Conventional structural seismic-resistant systems, such as steel and found to exhibit a modest energy dissipation capacity but less
moment resisting frames (MRFs) [1] or concentric braced frames sensitivity to frequency and temperature compared to conven-
(CBFs) [2], are currently designed [3,4] to experience significant tional VE dampers [9]. It has been though found impossible to de-
inelastic deformations and form a global plastic mechanism under sign elastomeric dampers and steel MRFs at practical sizes and cost
moderate-to-strong earthquakes. Such a design philosophy [3,4], for the building to remain elastic under strong earthquakes [10].
which results in inelastic deformations, has several advantages Both VE and elastomeric dampers transfer high forces on beams
including economy and reduced forces developed in structural and columns of the MRF. These forces cannot be used directly in
members and foundation due to inelastic softening. However, conventional capacity design rules since they are out of phase with
inelastic deformations result in damage, residual drifts and eco- the peak structural displacements. Recent studies by Karavasilis et
nomical losses such as repair costs, costly downtime while the al. [11–14] designed and tested steel MRFs with compressed elas-
building is repaired and cannot be used or occupied, and perhaps, tomer dampers providing VE damping under small-to-moderate
building demolition due to the complications associated with amplitudes of deformation, friction-based damping under high
repairing and straightening large residual drifts [5]. amplitudes of deformation and a limit on the peak damper force.
Modern resilient societies demand structural systems able to However, friction behavior results in permanent damper deforma-
achieve high performance, i.e., no damage under small and moder- tion and residual drifts in the structure following strong
ate earthquakes, and little damage which can be repaired without earthquakes.
loss of building operation under strong earthquakes [6]. Perfor- Research efforts [15–17] developed new earthquake-resistant
mance-based seismic design is expected to focus on modern systems called self-centering (SC) systems with the potential to
energy dissipation systems such as passive dampers and self-cen- eliminate residual drifts and inelastic deformations under strong
tering (SC) devices [6,7]. If carefully designed, these systems will earthquakes. SC systems exhibit a softening flag-shaped force-drift
slightly increase the initial building design cost and significantly behavior due to: (1) separations developed in structural interfaces
reduce the great life-cycle cost related to earthquake damage [6]. (e.g., beam-to-column connections); (2) elastic pretensioning ele-
Viscoelastic (VE) dampers provide supplemental stiffness and ments (e.g., high strength steel tendons); and (3) energy dissipa-
damping and when combined with flexible MRFs so that they tion elements (EDs: friction-based, yielding or viscous) which are
would carry a large fraction of the lateral dynamic forces, they activated when separation in structural interfaces initiates. A re-
cent work [18] proposed SC systems with a visco-plastic ED (sim-
⇑ Corresponding author. Tel.: +44 (0) 1865 2 73144; fax: +44 (0) 1865 2 73010. ilar to the damper tested and modelled in [11–14]) as a better
E-mail address: [email protected] (T.L. Karavasilis).
alternative to SC systems with yielding or friction-based EDs.

0045-7949/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compstruc.2010.08.013
 T.L. Karavasilis et al. / Computers and Structures 89 (2011) 1232–1240 1233

Shape memory alloys (SMA) materials also exhibit stable flag- ED

g¼ ð1Þ
shaped hysteresis and have been used to develop SC devices 2  p  ES
[19,20]. SC systems avoid structural damage since they eliminate
inelastic deformations and residual drifts. However, they exhibit where ED is the energy dissipated per cycle of sinusoidal loading
drift and acceleration demands comparable to those of conven- and ES is the maximum strain energy stored during a cycle of sinu-
tional bilinear yielding systems of the same period and strength soidal loading. ED can be determined by integrating the hysteresis
[21] and therefore, provide conventional performance in terms of loops and ES can be calculated from kd and maximum displacement.
non-structural damage. Fig. 1b shows the hysteresis of the SCVD under earthquake load-
This paper highlights the role of advanced structural analysis ing. Under small amplitudes of deformation, d, there is no separa-
tools on the conception of high-performance earthquake-resistant tion across the SC device due to the forces Ft and FED, and the SCVD
structural systems. The concept of a new high-performance steel behaves as a conventional VE damper offering VE (velocity-depen-
MRF equipped with self-centering viscoelastic damping devices dent) damping. When the force in the VE damper becomes larger
(SCVDs) is described. SCVDs result from a novel and strategic in than the sum of Ft and FED, separation (relative movement) is tak-
series combination of VE dampers and SC devices. A prototype ing place in the SC device. This separation results in significant
building using the proposed MRF is designed and a nonlinear ana- softening since the stiffness of the SC is now equal to the tendon
lytical model for seismic analysis is developed. Extensive seismic stiffness assuming a zero stiffness of the ED during sliding. The
analyses results show the effectiveness of the proposed MRF to en- SCVD now offers hysteretic (friction-based) damping. Upon
hance structural and non-structural performance by significantly unloading, the SCVD exhibits VE behavior until the ED will be acti-
reducing residual drifts and inelastic deformations, and by reduc- vated again and eventually returns back to the initial position due
ing drifts, total floor accelerations and total floor velocities. These to the re-centering force of the tendon. An optimum design of the
results are the basis for further studies aiming to develop design SCVD device includes a force Ft equal to FED in order to maintain
methods and criteria for the proposed high-performance MRF. full re-centering capability and maximum hysteretic energy dissi-
pation capacity, as well as a high value of g in order to amplify VE
damping. SCVD strategically combines the advantages of the com-
2. Concept and design of self-centering viscous damping devices
pressed elastomer damper designed and tested in [11–14] with the
(SCVD)
advantages of SC systems and offers: (1) VE damping under small-
to-moderate amplitudes of deformation; (2) hysteretic damping at
The SCVD results from a novel and strategic in series combina-
large amplitudes of deformation; (3) limit on the peak device force;
tion of a VE damper with a SC device. Fig. 1a shows a simple
and (4) full re-centering capability. In addition, the peak force of
mechanical analog of the SCVD where a generalized Maxwell
the SCVD can be directly used in conventional seismic design
(GM) model, representing a VE damper, is connected in series with
capacity rules since it is velocity independent.
a SC device consisting of a pretensioning elastic tendon and a fric-
The SC device of the SCVD can be realized in different configu-
tion-based ED. The design parameters of the SCVD are the initial
rations. Fig. 2a shows a possible implementation similar to that in
pretension force in the tendon, Ft, the stiffness of the tendon, kt,
[17] using: (1) two steel elements of the same length but with sec-
the force required to activate the friction-based ED, FED, and the
tions of different size so that the member with the smaller section
properties of the VE damper, namely, the storage stiffness, kd,
can be inserted into the member with the larger section (the exte-
and the loss factor, g. Both kd and g can be easily obtained from dis-
rior element has an inverted U-shape, while the interior element
placement-controlled sinusoidal tests, i.e., kd can be calculated as
has a hollow rectangular section); (2) end plates that are attached
the ratio of the force at maximum displacement to the maximum
but not connected to the steel elements; and (3) pretensioning ten-
displacement, while the loss factor is determined as follows:
dons anchored on the end plates so they can hold the plates and
the steel members in place. Energy dissipation is provided with
friction at the upper contact interface of the steel elements due
a ED to a normal force applied by tightening bolts. The relative displace-
ment at the contact interface is enabled with a slot cut on the inte-
f, d rior steel element.
GM Alternatively, Fig. 2b shows another possible implementation
[19] using two steel elements and two sets of SMA wires anchored
Tendon
in such a way so that relative displacement between the steel ele-
ments will produce tension in one of the two sets of SMA wires. A
slot cut on the bottom steel element permits the relative displace-
ment between the two steel elements while additional energy dis-
sipation is provided with friction at the interface.
b
Viscoelastic
damping
3. High-performance steel MRF with SCVDs

3.1. Prototype building

f

Hysteretic Fig. 3a shows the plan view of the 5-story, 3-bay by 3-bay pro-
damping totype office building used for the study. The building has two 3-
bay identical perimeter steel MRFs (one at each side) to resist lat-
eral forces in the N–S direction. The design study focuses on one
perimeter MRF. This MRF is designed either as a conventional
d
MRF or as MRF with SCVDs in order to compare their seismic re-
Fig. 1. (a) Mechanical analog of the proposed SCVD and (b) SCVD hysteresis under sponse. The SCVDs are supported by braces and connected to the
earthquake loading. bottom flange of the beam of the steel MRF as shown in Fig. 3b.
1234 T.L. Karavasilis et al. / Computers and Structures 89 (2011) 1232–1240

a Viscoelastic material
Steel plate Bolts

Tendon Interior element Exterior element

b Bolts Viscoelastic material SMA wires

Top steel element Bottom steel element

Fig. 2. Possible configurations of the SCVD: (a) pretensioning tendons and friction-based energy dissipation [17] and (b) SMA wires and friction-based energy dissipation [19].

based on the summation of the areas and flexural stiffnesses of

a Braced frame
the gravity columns. The column base and beam-to-column
Moment resisting frame

connections of the MRFs are assumed to be fully rigid. A rigid

diaphragm constraint is imposed at the nodes of each floor level.
3@8m

3.2. Design of conventional steel MRF

The perimeter MRF of the building is designed as a conventional

N MRF using the modal response spectrum analysis procedure of
EC8. The MRF satisfies Ductility Class High [3] by using compact
S Class 1 cross-sections [24] and hence, the behavior (or ‘‘strength
3@8m reduction”) factor is equal to q = 5  au/a1 = 6.5, where au/a1 is the
overstrength factor with a recommended value of 1.3 for multi-
bay multi-story steel MRF. The displacement behavior (or ‘‘dis-
b placement amplification”) factor is equal to q, i.e., EC8 adopts the
4.0 m 4 @ 3.2 m

equal-displacement rule to estimate peak inelastic drifts. These

drifts are then used to check second order (P–D) effects. Addition-
ally, EC8 imposes a serviceability limit on the peak story drift, hmax,
under the frequently occurred earthquake (FOE) with a return per-
iod equal to 95 years. The FOE has intensity equal to 40% (reduction
SCVDs BRACES factor v = 0.4 [3]) of the intensity of the DBE and the associated
hmax is equal to 0.75% assuming ductile non-structural compo-
nents. The MRF satisfies the strong column-weak beam capacity
3@8m design rule of EC8 [3], the beam-to-column connections are de-
Fig. 3. Prototype building structure: (a) plan view and (b) perimeter MRF with signed to be fully rigid, and the panel zones are strengthened with
SCVDs and supporting braces. doubler plates to avoid yielding [3]. A strength-based design with
q = 6.5 under the DBE was first performed. However, beams and
columns from strength-based design had to be increased to satisfy
The yield stress of structural steel is assumed to be equal to the serviceability drift requirements under the FOE. The final sec-
275 MPa. The dead and live gravity loads considered in the design tions were found iteratively, i.e., by decreasing the value of q,
are selected according to Eurocode 1 [22]. The design seismic ac- designing the MRF and then checking drifts under the FOE.
tion, referred to herein as design basis earthquake (DBE), has a re-
turn period equal to 475 years and is expressed by the Type 1 3.3. Design of high-performance steel MRF with SCVDs
elastic response spectrum of the Eurocode 8 (EC8) [3] with a peak
ground acceleration equal to 0.3 g and ground type B (average The response of steel MRFs with SCVDs is complex due to the
shear wave velocity between 360 and 800 m/s). The program different sources of energy dissipation (VE and hysteretic damping)
SAP2000 [23] is utilized for designing the steel MRFs according and cannot be predicted using results and conclusions applicable
to Eurocode 3 [24] and EC8 [3]. to SC systems [21] or systems with VE dampers [8]. However, a
The 2D SAP2000 model used for design is based on the center- simplified conservative design approach that neglects the supple-
line dimensions of the MRFs without accounting for the finite pa- mental damping provided by the VE material and makes use of
nel zone dimensions. A ‘‘lean-on” column is included in the modal response spectrum analysis has been adopted for this preli-
SAP2000 model to account for the P–D effects of the vertical loads minary investigation.
acting on the gravity columns in the tributary plan area (half of the The SAP2000 model used for design includes the members of
total plan area) assigned to the perimeter MRF. This column is the steel MRF (modeling details described in Section 3.1), the sup-
pinned at its base, continuous over the height of the building, porting braces (pinned connected) and a horizontal spring repre-
and carries the vertical loads acting on the gravity columns. The senting the storage stiffness of the VE damper kd. The SC device
cross-section area and flexural stiffness of the lean-on column is is assumed to be rigid before separation initiates. The q factor is
 T.L. Karavasilis et al. / Computers and Structures 89 (2011) 1232–1240 1235

equal to 6.5 to enable a direct comparison with the conventional ered in the design process (Section 3.3). Table 2 provides the prop-
steel MRF and defines the force level at which separation in the erties of the SCVD designs at each story. The storage shear modulus
SC initiates. The design is performed iteratively by selecting values G0 (x1 = 4.16 rad/s, temp = 24 °C) of the VE material is equal to
of kd at each story so that they produce a uniform distribution of 1086 kPa [25]. Table 2 shows that both components (VE material
drift demands based on modal response spectrum analysis. Braces and tendons) of the SCVDs can be designed to have practical sizes.
are sized to be stiff enough so that the story drift produces SCVD Fig. 4 compares the base shear coefficient (V/W)–roof drift (hr)
deformation rather than brace deformation. For all stories, a ratio curves of the two frames obtained from nonlinear cyclic static
of total brace horizontal stiffness per story to damper stiffness analysis using analytical models described in the next section. V
equal to 10 is adopted. Beams and columns are designed according is the base shear force and W is the seismic weight. The analysis
to capacity design rules so that they do not yield under the DBE. was performed at two cycles with roof drift amplitudes equal to
They also satisfy the strong column-weak beam capacity design 1.5% and 2.5% of the total building height. Fig. 4 shows that the
rule of EC8 [3]. Beam-to-column connections are designed to be MRF with SCVDs has less strength, smaller yield displacement
fully rigid and panel zones are strengthened with doubler plates and less energy dissipation capacity than the conventional MRF.
[3]. The strength-based design of the MRF with SCVDs under the For the MRF with SCVDs, softening begins at hr equal to 0.5% due
DBE was found to satisfy the serviceability limit on hmax under to separations in the SC. Beams and columns remain elastic for hr
the FOE. lower or equal to 1.5%; indicating that the MRF with SCVDs sus-
The VE damper area, Ad, at each story is determined by tains no damage and exhibit full re-centering capability (minimal
Ad = (kd  td)/G0 (x1, temp), where td is the thickness of the VE dam- residual drifts) under the DBE (see hmax estimates in Table 1). For
per, G0 (x1, temp) is the storage shear modulus of the VE material, hr higher than 1.5%, the MRF with SCVDs sustains inelastic defor-
x1 is the first-mode cyclic frequency of the MRF with SCVDs and mations and possible residual drifts which are though significantly
temp is the design ambient temperature (considered equal to smaller than those sustained by the MRF.
24 °C). The VE material used in this study is the ISD-110 material
studied by Fan [25].
The SCVD device is designed with the configuration shown in
Fig. 2a. The activation force in the SC device at each story (i.e., Table 2
Ft + FED) is designed equal to the force in the spring of the Properties of SCVDs.
SAP2000 model. FED is designed equal to Ft. The tendons are cables Story kd (kN/m) td (m) Ad (m2) Pt + PED (kN) At (m2) kt (kN/m)
made of composite polymers tested by Christopoulos et al. [17] and 1 29,691 0.04 1.09 371 2.08e4 3870
have cyclic modulus Et equal to 93 GPa and elongation capacity 2 34,015 0.04 1.25 350 1.51e4 2810
close to 2.5%. With the Ft known, the area of the tendons At at each 3 30,437 0.04 1.12 313 1.35e4 2510
story is determined by At = Ft/ro, where ro is the initial pretension 4 22,449 0.04 0.83 231 9.96e5 1850
5 17,103 0.04 0.63 176 7.59e5 1410
stress in the tendon material. The stiffness kt of the tendon is deter-
mined by (At  Et)/Lt. The length of the device is designed equal to
5 m by considering the tendon elongation capacity and the ex-
pected drifts under the maximum considered earthquake (MCE).
The MCE has return period equal to 2500 years and intensity equal
to 150% of the intensity of the DBE. 0.2 MRF with SCVDs
Conventional MRF

0.1
3.4. Design details
V/W

0
Table 1 compares properties of the conventional MRF and the
MRF with SCVDs. The table lists the column cross-sections, beam -0.1
cross-sections, steel weight, fundamental period of vibration (T1),
and equal-displacement rule estimates of hmax under the FOE and -0.2
the DBE. Table 1 shows that significant reductions in steel weight -0.03 -0.02 -0.01 0 0.01 0.02 0.03
and higher performance in terms of hmax can be achieved by using
SCVDs. It is emphasized that hmax for the MRF with SCVDs are
expected to be lower than those listed in Table 1 due to the Fig. 4. Comparison of the base shear coefficient–roof drift responses from nonlinear
supplemental damping of the VE material which was not consid- cyclic static analysis.

Table 1
Properties of conventional MRF and MRF with SCVDs.

Frame Columns Beams Braces Steel weight (kN) T1 (s) hmax (%)
Story Section Story Section Story Ab (m2) FOE DBE
Conventional MRF 1 HEB400 1 IPE450 – – 180 1.70 0.72 1.75
2 HEB400 2 IPE450
3 HEB400 3 IPE400
4 HEB360 4 IPE400
5 HEB360 5 IPE360
MRF with SCVDs 1 HEB280 1 IPE270 1 8.28e4 124 1.51 0.60 1.50
2 HEB280 2 IPE270 2 6.54e4
3 HEB280 3 IPE270 3 6.54e4
4 HEB240 4 IPE240 4 3.90e4
5 HEB240 5 IPE240 5 3.02e4
1236 T.L. Karavasilis et al. / Computers and Structures 89 (2011) 1232–1240

4. Nonlinear dynamic analyses 3

a
4.1. Analytical model development

G (MPa)
2
2D nonlinear analytical models of the conventional MRF and the
MRF with SCVDs were developed for nonlinear dynamic analysis
using OpenSEES [26]. A distributed plasticity force-based beam Experimental
1
column was used to model beams and columns. An elastic beam Analytical
column element was used to model the lean-on column that ac-
counts for the P–D effects of the vertical loads on the interior grav- 0
ity columns of the prototype building. An elastic truss element was 0 5 10 15 20
used to model the braces. The panel zones of the beam-column (rad/s.)
joints were modelled as proposed by Herrera et al. [27]. The SC de-
vices were modelled using a zero-length element exhibiting flag-
shaped hysteresis. This zero-length element has a high initial stiff-
b 1.5
ness representing the rigidity of the SC device before activation of
the ED. This element is suitable for the SC device shown in Fig. 2a. 1
However, a different self-centering flag-shaped model is needed
for the SMA wires of the SC device shown in Fig. 2b [19]. Experimental
The shear stress (s)–shear strain (c) behavior of the VE material 0.5 Analytical
was modelled using the GM model shown in Fig. 5. Under har-
monic loading of cyclic frequency x, the GM model provides stor-
age shear modulus equal to 0
0 5 10 15 20
0
X
n
ðxbm Þ2 (rad/sec)
G ðxÞ ¼ G0 þ  Gm ð2Þ
m¼1 1 þ ðxbm Þ2
Fig. 6. Comparison of experimental and analytical values of the mechanical
and loss factor equal to properties of the VE material at 24 °C (a) G0 and (b) g.

P ðxbm Þ
ðxb0 ÞG0 þ nm¼1 1þð xbm Þ2
 Gm
gðxÞ ¼ Pn ðxbm Þ2
ð3Þ used to model inherent 2% critical damping at the first two modes
G0 þ m¼1 1þðxb Þ2  Gm of vibration. A diaphragm constraint was imposed on the nodes of
m
each floor level. A nonlinear load control static analysis under the
These equations can be used to calibrate the GM model against gravity loads was first performed and then, the nonlinear dynamic
experimentally obtained values of G0 and g for different cyclic fre- earthquake analysis was executed. Each dynamic analysis was ex-
quencies. Fan [25] calibrated the GM model of Fig. 5 with m = 4 tended beyond the actual earthquake time to allow for damped free
for the ISD-110 VE material and the resulting values of the param- vibration decay and correct residual drift calculation.
eters are given in Table 3. Fig. 6 compares the experimentally ob-
tained values of G0 and g at 24 °C [24] with those obtained by
using the parameters of Table 3 in Eqs. (2) and (3). Acceptable 4.2. Earthquake ground motions
agreement is observed. The GM model was represented in Open-
SEES as a combination of linear springs and dashpots. The stiffness An ensemble of 20 earthquake ground motions recorded on
of the springs of the GM model was determined by multiplying G0 ground type B were used in 2D nonlinear dynamic analyses to eval-
and Gm with the ratio Ad/td of each story to transform s–c behavior uate the performance of the conventional MRF and the perfor-
to force (F = s  Fd)  deformation (d = c  td) behavior.The Newmark mance of the MRF with SCVDs. None of the ground motions
method with constant acceleration was used to integrate the equa- exhibit near-fault forward-directivity effects. The ground motions
tions of motion. The Newton method with tangent stiffness was were scaled to the DBE level using the scaling procedure of Somer-
used for the solution algorithm. A Rayleigh damping matrix was ville [28]. Table 4 provides the scale factors and information on the
20 ground motions. Fig. 7 compares the DBE elastic response spec-
trum of EC8 with the mean (l) and mean plus/minus one standard
deviation (l + r) spectra of the DBE ground motions. The ampli-
Go tudes of the DBE ground motions were further scaled by 0.4 and
Go G1 G2 Gm
1.5 to represent FOE and MCE ground motions, respectively.
1G1 2G2 mG

5. Seismic response results

The seismic performance of the MRFs is quantified in terms of

the hmax; the peak residual story drift, hrmax; the peak total floor
Fig. 5. GM model for VE material. acceleration, amax; and the peak total floor velocity, vmax. hmax
and hrmax quantify structural damage and damage of drift-sensi-
tive non-structural components, while amax and vmax quantify
Table 3
GM model parameters [29]: G in kPa and s in s (NOT t all beta). damage of rigidly attached equipment and non-rigidly attached
block-type objects, respectively [29].
G0 (kPa) G1 (kPa) G2 (kPa) G3 (kPa) G4 (kPa)
Fig. 8 compares the roof drift time histories of the conventional
191.4 2713.7 1356.3 11018.4 409.5 MRF and the MRF with SCVDs under the HSP ground motion scaled
bo b1 b2 b3 b4
to the DBE (Table 4). Near the end of the time history the MRF with
0.1225 0.040 0.1356 0.0061 1.2394
SCVDs oscillates around the origin, indicating negligible residual
 T.L. Karavasilis et al. / Computers and Structures 89 (2011) 1232–1240 1237

Table 4
Properties of the ground motions used for nonlinear dynamic analyses.

Earthquake Station Component Magnitude Distance Scale factor

(Mw) (km) FOE DBE MCE
Imperial Valley 1979 Cerro Prieto H-CPE237 6.53 15.19 0.82 2.05 3.08
Loma Prieta 1989 Hollister – S&P HSP000 6.93 27.67 0.29 0.72 1.08
Loma Prieta 1989 Woodside WDS000 6.93 33.87 1.40 3.49 5.24
Loma Prieta 1989 WAHO WAH090 6.93 17.47 0.48 1.20 1.80
Manjil 1990 Abbar ABBAR-T 7.37 12.56 0.28 0.70 1.05
Cape Mendocino 1992 Fortuna – Fortuna Blvd FOR000 7.01 15.97 0.99 2.47 3.71
Cape Mendocino 1992 Rio Del Overpass – FF RIO360 7.01 14.33 0.50 1.25 1.88
Landers 1992 Desert – Hot Springs LD-DSP000 7.30 21.78 0.95 2.37 3.56
Northridge 1994 LA – W 15th St W15090 6.69 25.60 1.14 2.86 4.29
Northridge 1994 Moorpark – Fire Sta MRP180 6.69 16.92 0.78 1.94 2.91
Northridge 1994 N Hollywood – Cw CWC270 6.69 7.89 0.53 1.33 2.00
Northridge 1994 Santa Susana Ground 5108-360 6.69 1.69 0.78 1.95 2.93
Northridge 1994 LA – Brentwood VA 0638-285 6.69 12.92 0.85 2.12 3.18
Northridge 1994 LA – Wadsworth VA 5082-235 6.69 14.55 0.62 1.54 2.31
Kobe 1995 Nishi-Akashi NIS090 6.90 7.08 0.48 1.19 1.79
Kobe 1995 Abeno ABN090 6.90 24.85 1.00 2.49 3.74
ChiChi 1999 TCU105 TCU105-E 7.62 17.18 0.96 2.39 3.59
ChiChi 1999 CHY029 CHY029-N 7.62 10.97 0.53 1.32 1.98
ChiChi 1999 CHY029 CHY041-N 7.62 19.83 0.56 1.40 2.10
Hector 1999 Hector HEC090 7.13 10.35 0.42 1.04 1.56

2 5
EC8 DBE (a)
1.5 4
Conventional MRF ( )
Sa (g)

Story

1 3 Conventional MRF ( )
2 MRF with SCVDs ( )
0.5 MRF with SCVDs ( )
1
0
0 1 2 3 4 FOE
Period (s.) 0
5
Fig. 7. Comparison of the DBE EC8 spectrum with the spectra of the DBE ground (b)
motions used for nonlinear dynamic analyses. 4

3
Story

2 2
MRF with SCVDs
Conventional MRF 1
DBE
(%)

0
0
5
r

(c)
4
Story

-2 3
0 20 40 60
Time (s.) 2

Fig. 8. Comparison of the roof drift time histories from nonlinear dynamic analysis
1
under the HSP ground motion scaled to the DBE. MCE
0
0 0.01 0.02 0.03 0.04

drift, while the conventional MRF experiences appreciable residual

drift due to inelastic deformations. Fig. 9. Statistics of peak story drifts from nonlinear dynamic analyses under the set
of 20 ground motions: (a) FOE; (b) DBE; and (c) MCE.
Fig. 9 shows l and l + r values of hmax in both frames under the
earthquake ground motions of Table 4 scaled to the FOE, DBE and
MCE. Both frames show uniform height-wise distributions of hmax.
The MRF with SCVDs exhibits significantly higher, higher and the FOE, DBE and MCE, respectively. The associated decreases in
slightly higher performance than the conventional MRF under hmax are 65%, 24% and 8%. For the MRF with SCVDs, hmax is 0.32%
1238 T.L. Karavasilis et al. / Computers and Structures 89 (2011) 1232–1240

(l) and 0.42% (l + r) under the FOE, 1.35% (l) and 1.80% (l + r) 5
under the DBE, and 2.30% (l) and 2.90% (l + r) under the MCE. (a)
The corresponding median values are 0.31%, 1.30% and 2.30%. 4 Conventinal MRF ( )
The above mean (l) and median values of hmax indicate perfor-
mance levels of immediate occupancy (IO), close to IO and life
3 Conventional MRF ( )

Story
safety (LS) under the FOE, DBE and MCE, respectively [30].
Fig. 10 shows l and l + r values of hrmax. Under the FOE, both MRF with SCVDs ( )
frames experience anticipated zero residual drifts; indicating IO
2
MRF with SCVDs ( )
[30]. Under the DBE and MCE, the MRF with SCVDs shows a signif-
icantly higher performance than the conventional steel MRF. The 1
associated decreases in hrmax are approximately 80% and 70%. FOE
For the MRF with SCVDs, hrmax is 0.05% (l) and 0.15% (l + r) un- 0
der the DBE, and 0.2% (l) and 0.36% (l + r) under the MCE. The
5
corresponding median values are 0.02% and 0.19%. The above mean (b)
(l) and median values of hrmax indicate performance levels of IO
under the FOE and close to IO under the DBE and MCE [30].
4
Fig. 11 shows l and l + r values of amax. The MRF with SCVDs
3

Story
shows significantly higher, higher and slightly higher performance
than the conventional MRF under the FOE, DBE and MCE, respec-
tively. The associated decreases in amax are approximately 49%, 2
21% and 9%; indicating less damage in rigidly attached non-struc-
tural components. For the MRF with SCVDs, amax is 1.93 (l) and 1
2.76 (l + r) m/s2 under the FOE, 5.29 (l) and 7.29 (l + r) m/s2 un- DBE
der the DBE, and 7.7 (l) and 10.37 (l + r) m/s2 under the MCE. The 0
corresponding median values are 1.63, 4.82 and 7.28 m/s2.
5
(c)
4
5
(a) 3
Story

4 Conventional MRF ( ) 2
3 Conventional MRF ( )
1
Story

MRF with SCVDs ( ) MCE

2 MRF with SCVDs ( ) 0
0 2.5 5 7.5 10 12.5 15
2
1 max
(m/s. )
FOE
0 Fig. 11. Statistics of peak total floor accelerations from nonlinear dynamic analyses
under the set of 20 ground motions: (a) FOE; (b) DBE; and (c) MCE.
5
(b)
4
Fig. 12 shows l and l + r values of vmax. The MRF with SCVDs
shows significantly higher performance under the FOE and higher
3
Story

performance under the DBE and MCE than the conventional MRF.
The associated decreases in vmax are approximately 53%, 29% and
2 22%; indicating less damage in non-rigidly attached block-type ob-
jects. For the MRF with SCVDs, vmax is 0.26 (l) and 0.34 (l + r) m/s
1 under the FOE, 0.71 (l) and 0.92 (l + r) m/s under the DBE, and 1.0
DBE (l) and 1.35 (l + r) m/s under the MCE. The corresponding median
0 values are 0.24, 0.65 and 0.97 m/s.
5 The response of the MRF with SCVDs depends on the value of
(c) q which defines the force level at which SCVDs are transformed
4 from VE dampers to SC devices. A decrease in q increases the
force at which activation in the SC device initiates, and therefore,
3
Story

increases the velocity-dependent energy dissipation capacity of

SCVDs. Such a design results in lower values of hmax than those
2
presented above for the DBE and MCE but also increases the peak
1 force developed in SCVDs and as a result, increases the sizes of
beams and columns which are designed according to capacity de-
MCE
0 sign rules. Increasing member sizes to achieve higher perfor-
0 0.005 0.01 0.015 0.02 mance seems acceptable since the MRF with SCVDs (designed
with q = 6.5) is 30% lighter in steel weight than the conventional
r-max
MRF. More work is needed in order to show how the seismic per-
Fig. 10. Statistics of peak residual story drifts from nonlinear dynamic analyses formance and weight of the MRF with SCVDs are affected by the
under the set of 20 ground motions: (a) FOE; (b) DBE; and (c) MCE. value of q.
 T.L. Karavasilis et al. / Computers and Structures 89 (2011) 1232–1240 1239

5 the proposed MRF results in significant reductions in steel weight

(a) (i.e., 30%) compared to a conventional MRF.
4 The promising results presented herein are the basis for fur-
Conventional MRF ( ) ther studies aiming to develop design procedures and criteria
for the proposed high-performance steel MRF. Design equations
3 Conventional MRF ( )
Story

able to predict peak story drifts by considering the effect of

2 MRF with SCVDs ( ) both viscoelastic (velocity-dependent) damping and hysteretic
(friction-based or yielding-based) damping are needed. Perfor-
MRF with SCVDs ( )
mance-based design of the proposed system can be optimized
1
by identifying the optimum value of the behavior factor q which
FOE defines the force level at which SCVDs are transformed from
0 viscoelastic dampers to self-centering devices and which pro-
5 duce the desired performance under different earthquake inten-
(b) sities. Emphasis should be also given to the collapse capacity of
4 the proposed system by considering P–D effects under seismic
intensities even higher than the MCE, and by investigating the
3 ultimate response of the SCVDs. Finally, an experimental re-
Story

search program is needed to validate design procedures/criteria

2 and help move the proposed high-performance MRF into
practice.
1
DBE
0
References
5
(c) [1] Karavasilis TL, Bazeos N, Beskos DE. Drift and ductility estimates in regular
4 steel MRF subjected to ordinary ground motions: a design-oriented approach.
Earthquake Spectra 2008;24(2):151–431.
[2] Karavasilis TL, Bazeos N, Beskos DE. Estimation of seismic drift and ductility
3
Story

demands in planar regular X-braced steel frames. Earthquake Eng Struct

Dynam 2008;36:2273–89.
2 [3] CEN. EN 1998-1, Eurocode 8: design of structures for earthquake resistance.
Part 1: General rules, seismic actions and rules for buildings. European
1 Committee for Standardization; 2004.
[4] Mazzolani FM. Steel and composites structures in European seismic areas:
MCE research, codification, design, and applications. Earthquake Spectra
0 2003;19(2):415–52.
0 0.25 0.5 0.75 1 1.25 1.5 1.75 [5] Ramirez SM, Miranda E. Building-specific loss estimation methods & tools for
v (m/s.) simplified performance-based earthquake engineering. In: John A, editor.
Blume earthquake engineering research center 2009. Report No. 171. Stanford
Univ.
Fig. 12. Statistics of peak total floor velocities from nonlinear dynamic analyses [6] Wada A. Seismic design for resilient society. Keynote lecture. In: Joint
under the set of 20 ground motions: (a) FOE; (b) DBE; and (c) MCE. proceedings: 7th International conference on urban earthquake engineering
and 5th int. conf. on earthquake engineering, Tokyo, 3–5 March 2010.
[7] Christopoulos C, Filiatrault A. Principles of supplemental damping and seismic
isolation. Milan, Italy: IUSS Press; 2006.
[8] Fu Y, Kasai K. Comparative study of frames using viscoelastic and viscous
6. Summary and conclusions dampers. ASCE – J Eng Mech 1998;124(5):513–22.
[9] Lee KS. Seismic behavior of structures with dampers made from ultra high
The concept of a new high-performance steel moment resisting damping natural rubber. PhD dissertation. Bethlehem, PA: Department of Civil
and Environmental Engineering, Lehigh University; 2003.
frame (MRF) equipped with self-centering viscoelastic damping
[10] Lee KS, Ricles J, Sause R. Performance based seismic design of steel MRFs with
devices (SCVDs) is described. SCVDs result from a novel and strate- elastomeric dampers. ASCE – J Struct Eng 2009;135(5):489–98.
gic in series combination of viscoelastic dampers and self-center- [11] Karavasilis TL, Sause R, Ricles JM. Design of steel buildings for earthquake
ing devices. SCVDs offer: (1) viscoelastic damping at small conditions using next-generation elastomeric dampers. In: Proceedings of the
2010 structures congress, Orlando, Florida, USA, 12–15 May 2010.
amplitudes of deformation; (2) friction-based damping at large [12] Karavasilis TL, Sause R, Ricles JM. Seismic design and evaluation of steel MRFs
amplitudes of deformation; (3) limit on the peak device force; with compressed elastomer dampers. Earthquake Eng Struct Dynam,
and (4) full re-centering capability. In addition, the peak force of submitted for publication.
[13] Karavasilis TL, Ricles JM, Sause R, Chen C. Large-scale real-time hybrid testing
the SCVD can be directly used in conventional capacity seismic de- of steel frames equipped with elastomeric dampers. In: Behavior of steel
sign rules since it is velocity independent. structures in seismic areas, Proceedings of STESSA conference, Philadelphia,
A prototype building using the proposed MRF is designed and a Pennsylvania, USA, 16–20 August 2009.
[14] Karavasilis TL, Ricles JM, Sause R, Chen C. Performance evaluation of steel
detailed nonlinear analytical model for seismic analysis is devel- MRFs equipped with compressed elastomer dampers using large-scale real-
oped. Seismic analyses results show the effectiveness of the pro- time hybrid simulation. Eng Struct, submitted for publication.
posed MRF to enhance structural and non-structural performance [15] Garlock M, Ricles JM, Sause R. Experimental studies on full-scale post-
tensioned steel connections. ASCE – J Struct Eng 2005;131(3):438–48.
by significantly reducing residual drifts and plastic deformations, [16] Kim H-J, Christopoulos C. Friction damped posttensioned self-centering steel
and by reducing drifts, total floor accelerations and total floor moment-resisting frames. ASCE – J Struct Eng 2007;134(11):1768–79.
velocities. [17] Christopoulos C, Tremblay R, Kim HJ, Lacerte M. Self-centering energy
dissipative bracing system for the seismic resistance of structures:
In terms of the peak story drifts and residual drifts, the perfor-
development and validation. ASCE – J Struct Eng 2008;134(1):96–107.
mance of the proposed steel MRF is found close to immediate occu- [18] Kam W-Y, Pampanin S, Palermo A, Carr AJ. Self-centering structural systems
pancy under the design earthquake action (return period equal to with combination of hysteretic and viscous energy dissipations. Earthquake
475 years) of Eurocode 8. Under the same seismic action, the peak Eng Struct Dynam 2010;39:1083–108.
[19] Zhu S, Zhang Y. Seismic analysis of concentrically braced frame systems with
total floor accelerations and velocities are reduced approximately self-centering friction damping braces. ASCE – J Struct Eng 2008;134(1):
20% and 30% compared to a conventional steel MRF. Moreover, 121–31.
1240 T.L. Karavasilis et al. / Computers and Structures 89 (2011) 1232–1240

[20] Yang CS, DesRoches R, Leon RT. Design and analysis of braced frames with [26] Mazzoni S, McKenna F, Scott M, Fenves G. Open system for earthquake
shape memory alloy and energy-absorbing hybrid devices. Eng Struct engineering simulation (OpenSees). User command language manual, Version
2010;32(2):498–507. 1.7.3. Berkeley: University of California, Pacific Earthquake Engineering
[21] Christopoulos C, Filiatrault A, Folz B. Seismic response of self-centering Research Center; 2006.
hysteretic SDOF systems. Earthquake Eng Struct Dynam 2002;31:1131–50. [27] Herrera R, Ricles J and Sause R. Analytical studies of steel MRFs with CFT
[22] CEN. EN 1991-1-1, Eurocode 1: Actions on structures – Part 1.1: General columns under earthquake loading conditions. In: 4th International
actions-densities, self-weight, imposed loads for buildings. European conference on behavior of steel structures in seismic areas (STESSA) 2003,
Committee for Standardization; 2002. Naples, Italy. p. 519–25.
[23] SAP2000. Static and dynamic finite element analysis of structures. Version [28] Somerville P. Development of ground motion time histories for phase 2 of the
9.1.4. Berkeley, California, USA: Computers and Structures Inc.; 2005. FEMA/SAC steel project. Report No. SAC/DB-97/04 1997. Sacramento, CA.
[24] CEN. EN 1993-1-1, Eurocode 3: design of steel structures – Part 1.1: General [29] ATC. ATC-58 task report, phase 2, task 2.3. Engineering demand parameters for
rules and rules for buildings. European Committee for Standardization; 2005. non-structural components. Redwood City, CA: Applied Technology Council;
[25] Fan CP. Seismic analysis, behavior, and retrofit of non-ductile reinforced 2004.
concrete frame buildings with viscoelastic dampers. PhD dissertation. [30] Building Seismic Safety Council, BSCC. NEHRP guidelines for the seismic
Bethlehem, PA: Department of Civil and Environmental Engineering, Lehigh rehabilitation of buildings. Report No. FEMA-273. Washington, DC: Federal
University; 1998. Emergency Management Agency; 1997.