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

net/publication/228342077

Bracing systems for seismic retrofitting of steel frames

Article  in  Journal of Constructional Steel Research · February 2009

DOI: 10.1016/j.jcsr.2008.02.013

CITATIONS READS

70 1,513

2 authors:

Luigi Di Sarno Amr S. Elnashai

Università degli Studi del Sannio University of Houston
120 PUBLICATIONS   888 CITATIONS    304 PUBLICATIONS   6,624 CITATIONS   

SEE PROFILE SEE PROFILE

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

Experimental & Computational Hybrid Assessment of Natural Gas Pipelines Exposed to Seismic Risk View project

EXTREME LOADING ANALYSIS OF PETROCHEMICAL PLANTS AND DESIGN OF METAMATERIAL-BASED SHIELDS FOR ENHANCED RESILIENCE View project

All content following this page was uploaded by Luigi Di Sarno on 21 October 2017.

The user has requested enhancement of the downloaded file.

 Journal of Constructional Steel Research 65 (2009) 452–465
www.elsevier.com/locate/jcsr

Bracing systems for seismic retrofitting of steel frames

L. Di Sarno a,∗ , A.S. Elnashai b
a Department of Engineering, University of Sannio, Benevento, Italy
b Department of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, USA

Received 14 August 2007; accepted 15 February 2008

Abstract

The present study assesses the seismic performance of steel moment resisting frames (MRFs) retrofitted with different bracing systems. Three
structural configurations were utilized: special concentrically braces (SCBFs), buckling-restrained braces (BRBFs) and mega-braces (MBFs). A
9-storey steel perimeter MRF was designed with lateral stiffness insufficient to satisfy code drift limitations in zones with high seismic hazard. The
frame was then retrofitted with SCBFs, BRBFs and MBFs. Inelastic time-history analyses were carried out to assess the structural performance
under earthquake ground motions. Local (member rotations) and global (interstorey and roof drifts) deformations were employed to compare
the inelastic response of the retrofitted frames. It is shown that MBFs are the most cost-effective bracing systems. Maximum storey drifts of
MBFs are 70% lower than MRFs and about 50% lower than SCBFs. The lateral drift reductions are, however, function of the characteristics
of earthquake ground motions, especially frequency content. Configurations with buckling-restrained mega-braces possess seismic performance
marginally superior to MBFs despite their greater weight. The amount of steel for structural elements and their connections in configurations with
mega-braces is 20% lower than in SCBFs. This reduces the cost of construction and renders MBFs attractive for seismic retrofitting applications.
c 2008 Elsevier Ltd. All rights reserved.

Keywords: Bracing; Buckling restrained braces; Steel frames; Concentrically braced frames; Moment resisting frames; Ductility; Seismic retrofitting; Performance
assessment; Time history analyses

1. Introduction buildings are classified as having unbraced (UFs) or braced

Damage experienced during past earthquakes worldwide
demonstrates that steel multi-storey building structures gener-
ally exhibit adequate seismic response (e.g. [1]). This is due
to the favourable mass-to-stiffness ratio of base metal and the
enhanced energy absorption of structural ductile systems em-
ployed. Nonetheless, relatively recent earthquakes, e.g. those in
the 1994 Northridge (California), 1995 Kobe (Japan) and 1999
Chi-Chi (Taiwan), have shown that poor detailing of connec-
tions (e.g. beam-to-column, brace-to-beam, brace-to-column
and column-to-base) and buckling of diagonal braces can un-
dermine the seismic performance of the structure as whole (see,
for example, [2–6]). Fig. 1 shows the distribution of damage
level and the damage to structural members and connections
with respect to structural type as surveyed in the aftermath of
the 1995 Hyogoken-Nanbu (Kobe) earthquake [7]. Damaged
∗ Corresponding address: Department of Engineering, University of Sannio,
Piazza Roma, 21, 82100Benevento, Italy. Tel.: +39 0824305566; fax: +39
(0)824 325246.
E-mail address: disarno@unina.it (L. Di Sarno).
(BFs) frames. Thus, considering the two principal framing ori-
entations of a building, the surveyed structures include the
following designations: UF-UF (unbraced frames in two hori-
zontal directions), UB-BF (unbraced frames in one horizontal
direction and braced frames in the other direction), and BF-
BF (braced frames in both horizontal directions). Beams con-
sisted almost exclusively of wide-flange sections, either rolled
or built-up. For columns, wide-flange (H) sections were used
most extensively; square-tube (S) sections were also utilized in
some structural systems. Considering the 988 damaged steel
buildings, 432 (43.7%) are UF-UF, 134 (13.6%) are UF-BF
and 34 (3.4%) are BF-BF, with 388 (39.3%) having uniden-
tified framing systems. These statistics indicate that the ma-
jority of damaged buildings had unbraced moment resisting
frames (MRFs) as earthquake-resistant system. Fig. 1 also dis-
plays the location of damage, namely columns, beams, beam-
to-column connections, braces and column bases, as a function
of frame type. Major observations from the collected data are
as follows [8]: (i) columns in UFs suffered the most damage
relative to other frame elements (in terms of the number of

c 2008 Elsevier Ltd. All rights reserved.

0143-974X/$ - see front matter

doi:10.1016/j.jcsr.2008.02.013
 L. Di Sarno, A.S. Elnashai / Journal of Constructional Steel Research 65 (2009) 452–465
Fig. 1. Distribution of damage level (left) and damage to structural members and connections (right) with respect to structural type. Key: UF = Unbraced frame;
BF = Braced frame; H = Wide flange sections; S = Square tube sections.
Fig. 2. Fracture in beam-to-column connections in the Northridge earthquake
(top) and web tear-out in bolted brace-to-column connections during the 1995
Kobe earthquake (bottom).
buildings), while braces in BFs were the most frequently dam-
aged structural element; (ii) damage to beam-to-column con-
nections and column bases was also significant in UFs; (iii)
damage to beam-to-column connections was most significant
for UFs employing hollow section (square-tube) columns; and
(iv) damage to columns was most significant for UFs with wide-
flange members. The discussion of the above surveyed data is
representative of typical structural response of steel buildings
damaged by moderate-to-severe earthquake ground motions.
The occurrence of buckling, often in the plastic range in
multi-storey buildings, erodes the capacity of the structure and
453
Fig. 3. Characteristics of global intervention approaches in seismic retrofitting
of structures.
may lead to sudden changes in the dynamic characteristics
of the lateral resisting structure system. Brittle fractures, as
for example, those depicted in Fig. 2 for beam-to-column and
brace–column connections, impair the global ductile response
of frames and hence their energy dissipation capacity under
earthquake loads. As a result, beam-to-column connections and
braces may be inefficient in ductile MRFs or concentrically
braced frames (CBFs) if they are not adequately capacity-
designed (e.g. [9–12] among many others).
Bracing is a very effective global upgrading strategy to
enhance the global stiffness and strength of steel UFs. It can
increase the energy absorption of structures and/or decrease
the demand imposed by earthquake loads whenever hysteretic
dampers are utilized. Structures with augmented energy
dissipation may safely resist forces and deformations caused by
strong ground motions. Generally, global modifications to the
structural system are conceived such that the design demands
on the existing structural and nonstructural components are
less than their capacities (Fig. 3). Lower demands may reduce
the risk of brittle failures in the structure and/or avoid the
interruption of its functionality and, in turn, the downtime due
to the retrofitting, which are key features in the earthquake
loss assessment [13,14]. The attainment of global structural
ductility is achieved within the capacity design framework by
forcing inelasticity to occur within dissipative zones (plastic
hinges in MRFs and braces in CBFs) and ensuring that all other
members and connections behave linearly. Diagonal braces
can be aesthetically unpleasant where they change the original
454 L. Di Sarno, A.S. Elnashai / Journal of Constructional Steel Research 65 (2009) 452–465
Fig. 4. Typical unbonded brace: layout (left) and response curve (right).
architectural features of the building (e.g. [15] and [16]).
Additionally, ordinary braces may transmit very high actions
to connections and foundations and these frequently need to be
strengthened.
The present work investigates the seismic performance of
steel MRFs retrofitted with different bracing systems. The
latter include special concentrically-braced frames (SCBFs),
buckling-restrained braces (BRBFs) and mega-braces (MBFs).
The inelastic structural response has been expressed in terms
of local and global deformation parameters, namely storey and
roof drifts, derived by means of nonlinear time-history analyses.
2. Bracing systems
Several configurations of braced frames may be used for
seismic rehabilitation of existing steel, composite steel–concrete
and reinforced concrete building structures, e.g. [17,18]
among others. The most frequently used systems include
concentrically-braced frames (CBFs), eccentrically-braced
frames (EBFs) and the knee-brace frames (KBFs). Common
configurations for CBFs encompass V and inverted-V bracings,
K, X and diagonal bracings [9]. However, V bracings are not ad-
vised for seismic retrofitting because of the likelihood of dam-
age in the beam mid-span. Under horizontal forces the com-
pressed braces may buckle, thus reducing their load-bearing
capacity abruptly. Conversely, the force in the tension braces
increases monotonically reaching yield strength and eventually
strain-hardening. The net result is an unbalanced force concen-
trated at the brace-to-beam connection [19,20]. The effects in
the beam, e.g. additional bending and shear, should be added
to those due to gravity loads [21]. Alternatively, the unbalanced
force in the beams may be eliminated through ad hoc bracing
configurations such as macro-bracings, e.g. two, three storey
X-bracings or V-bracings with a zipper column [22–24].
Macro-bracings can be utilized for strengthening and
stiffening of steel-framed buildings. They are often employed
to form MBFs, which exhibit high stiffness and enhanced
ductility. Brace configurations with MBFs were utilized in
this analytical work to retrofit a medium-rise steel MRF
with inadequate lateral stiffness. Alternative systems, such
as traditional SCBFs and frames with unbonded braces
were also assessed. Unbonded or buckling-restrained braces,
as, for example, those shown in Fig. 4, are hysteretic
dampers very popular in Japan, the US and Italy for seismic
retrofitting of steel and reinforced concrete structures (e.g.
[25–29,35]). Such braces are based on the same metallic
yielding principle of added-damping-added-stiffness devices,
i.e. tension/compression yielding brace. They consist of a core
steel plate encased in a concrete-filled steel tube, as pictorially
shown in Fig. 4. Yielding of the interior component under
reversal axial loads provides stable energy dissipation; the
exterior concrete-filled steel tube prevents local and member
buckling.
In this study a lightweight concrete with specific weight
γ = 1800 kg/m3 was assumed as filler material. A special
coating is applied to reduce friction between steel and concrete.
Since lateral and local buckling are prevented in unbonded
braces, high energy absorption and dissipation is attainable.
Several experimental tests have been carried out on this type
of braces in the last decade [30–35]. Buckling-restrained
braces show compressive strength which is about 15%–20%
greater than tensile capacity, as also displayed in the axial-
deformation response curve provided in Fig. 4. Such increase
in the compression capacity is due to the confinement effect
of the inner steel plate generated by the concrete filler and
outer tube. Moreover, unbonded braces can reach cumulative
cyclic inelastic deformations exceeding 300 times the initial
yield deformation of the brace before failure. The latter depends
on several factors, including material properties, local detailing,
loading conditions and history. Buckling-restrained braces are
often fabricated with low-yield steels, e.g. LYP100 and LYP235
with yield strengths ( f y ) equal to 100 MPa and 235 MPa,
respectively. Configurations with low-yield buckling-braces
were also considered in this study for comparisons; both SCBFs
and BRBFs employ LYP235 diagonal members.
Unbonded braces are a viable alternative with respect to
other types of passive protection devices used for seismic
structural retrofitting, e.g. high damping rubber bearings used
for base isolation and/or viscous dampers [36,37]. The former
devices are generally more cost-effective than the latter. The
cost of hysteretic damper is given as a function of the maximum
axial loads and maximum displacements of the brace. In
Europe, for example, the manufacturers provide generally
devices with maximum displacements of about ±20 mm with
three levels of axial loads: 200–300 kN, 400–500 kN and
 L. Di Sarno, A.S. Elnashai / Journal of Constructional Steel Research 65 (2009) 452–465 455

600–800 kN. For these devices the cost may vary between Table 1
1500e (200 kN) and 3000e (600 kN). Frames assessed in the present study

Frame label Brace configuration Brace type Brace steel grade

3. Case studies
SCBF-N Concentric Ordinary A36
BRBF-CN Concentric Unbonded A36
3.1. Description of the sample frames SCBF-L Concentric Ordinary LYS235
BRBF-CL Concentric Unbonded LYS235
A 9-storey steel MRF building was designed with lateral MBF-N Mega Ordinary A36
stiffness that does not comply with drift limitations imposed BRBF-MN Mega Unbonded A36
MBF-L Mega Ordinary LYS235
for structural systems in earthquake-prone regions. The multi-
BRBF-ML Mega Unbonded LYS235
storey frame has five 9144 mm-long bays. The interstorey
height is 3962 mm for all but the first floor which is 5486 mm Key: For A36: f y = 248 MPa; for LYS235: f y = 235 MPa.
high. At the first and second storeys, beam span loads are equal
to 14.88 kN/m and joint vertical loads are equal to 158 kN Table 2
and 107 kN at interior and perimeter joints, respectively. Beam Modal properties of the frames assessed in the present study
loads of 12.65 kN/m and joint vertical loads of 140 kN (interior Frame system Period (s) Effective modal
joints) and 92 kN (exterior joints) were used at the roof. The mass(fraction of the
total vertical load of the frame for seismic combinations is total mass)
1st 2nd 3rd 1st 2nd 3rd
Wtot = 45 070 kN. Nominal yield strength equal to 345 MPa
(50 ksi) was used for columns while girders have strength MRF 2.05 0.76 0.44 0.84 0.11 0.36
SCBFs & BRBFs 1.08 0.31 0.16 0.77 0.16 0.04
equal to 248 MPa (36 ksi). The prototype MRF represents a MBFs 1.01 0.34 0.21 0.79 0.12 0.04
typical perimeter frame of multi-storey residential buildings in
areas with high earthquake hazard, e.g. Los Angeles (seismic
zone 4), in California. The value of 0.40g was assumed for brace types were employed. The set of sample frames assessed
the peak ground acceleration at the bedrock; the soil type for in this study are summarized in Table 1. Two layouts were
the construction site is rock. The seismic base shear is equal selected to stiffen the MRF; these are SCBFs and MBFs. Both
to VB = 5803 kN and was estimated through the design braced configurations allow a reduction of 77% in the first
provisions in [38]. The assumed response modification factor storey to be achieved. Their fundamental periods of vibration
is R = 8.50; the equivalent viscous damping is ξ = 5%. are very similar: 1.08 s (SCBF) versus 1.01 s (MBF). The
The natural period of vibration used to evaluate VB is 1.28 period of the MRF is thus halved by using bracing systems.
seconds; it was computed as a function of the frame total The modal properties of the sample frames are summarized in
height, i.e. H = 37,182 mm. The fundamental period derived Table 2; the computed values of the effective modal masses
by eigenvalue analysis is 2.05 s, about 60% higher than the show that configurations with braces exhibit similar dynamic
value derived by utilizing the simplified code expression. It response.
is instructive to note, however, that seismic codes of practice In SCBFs, X-braces were placed in two central bays, while
tend to underestimate the fundamental period of vibration to MBFs employ braces which extend over four storeys. The
account for the stiffening effects of nonstructural elements, typical layout of SCBFs and MBFs is provided in Fig. 5.
e.g. partition walls and infills. Such effects are of primary Circular hollow sections were used in both configurations for
importance for steel structures, which exhibit relatively low the diagonals; their strength is equal to 248 MPa (36 ksi).
horizontal stiffness. The underestimation of the natural period In SCBFs, sections with diameters D of 400 mm and wall
leads to conservative design assumptions, e.g. higher seismic thickness tw of 20 mm were used between the first and fourth
base shear and, in turn, larger storey lateral displacement drifts. floors, while hollow sections with D = 350 mm and tw =
The reliability/redundancy factor ρ was assumed equal to 1.25 17.5 mm were used at the fifth, sixth and seventh floors.
to account for the perimeter configuration of the MRF (low Braces at the eighth and ninth floors have d = 300 mm
redundancy). The maximum lateral displacement d of the frame and tw = 15.0 mm. Two sections were utilized for diagonals
was found at the first storey; the estimated maximum interstorey in the MBF: at the first four storeys, the hollow sections
drift d/ h is 3.96%. The latter value exceeds the recommended have D = 350 mm and tw = 17.5 mm, while in the
drift provided by [38] for limit state of ‘near collapse’, remaining D = 300 mm and tw = 15.0 mm. The design
i.e. 3.8%. On the other hand, data collected from several of the braces was carried out in compliance with the seismic
post-earthquake reconnaissance reports have demonstrated that provisions in [40]. Local slenderness ratios D/tw are 20;
values of d/ h equal to 3.0%, or even smaller, should be utilized this value is nearly half of the limiting width-to-thickness
in seismic design to prevent structural collapse of framed ratio λ p = 35 recommended for grade A36 steel. Bracing
systems [39]. global slenderness ratios (kl/r ) vary between 80 (lower floors)
Bracing was utilized as a means to retrofit the MRF with and 106 (upper floors); intermediate storeys have kl/r =
inadequate lateral stiffness. The design target was the reduction 85. Consequently, braces possess intermediate slenderness.
of the interstorey drift d/ h at first storey of the prototype Comparisons between the above ratios and the limitations
unbraced structure. In so doing, different configurations and in the European standards [41,42] show that the selected
456 L. Di Sarno, A.S. Elnashai / Journal of Constructional Steel Research 65 (2009) 452–465
Fig. 5. Layout of braced frames: concentrically- (top) and mega-braced (bottom) frames.
brace cross-sections are compliant with strength and stiffness
design requirements in these codes. For example, the maximum
slenderness λ is 1.16, about 25% lower than the allowable value
λ = 1.50 prescribed by [42].
Unbonded braces were also used to retrofit the substandard
sample MRF; these braces were employed for both SCBFs
and MBFs. They are made of two steel grades: A36 ( f y =
248 MPa) and LYS235 ( f y = 235 MPa). The combination
of the different system layouts (SCBFs and MBFs) and braces
(ordinary and unbonded) gives rise to eight different structural
configurations as summarized in Table 1.
3.2. Structural modelling
Elastic and inelastic (static and dynamic) analyses were
carried out through DRAIN-2DX [43], a general purpose
finite element (FE) computer program for the two-dimensional
structures. Bare frames, employing centreline dimensions,
were modelled by means of inelastic beam–column elements
employing lumped plasticity-based models. Beam-to-column
connections are modelled as rigid joints and the column-to-base
connections are assumed fully restrained. All seismic masses
and gravity loads were applied at the beam-to-column joints,
which is sufficiently accurate for models dealing with the lateral
response of frames.
Bilinear elasto-plastic behaviour with strain hardening of 1%
was adopted to model plastic hinges. A linearized biaxial plastic
domain was utilized to account for bending-axial interaction.
Shear behaviour of beams and columns was assumed to
remain linearly elastic. Inelastic truss FEs were employed for
diagonal braces. These are assumed to buckle elastically in
conventional SCBFs while buckling is restrained for unbonded
braces in BRBFs. Geometric nonlinearities, i.e. P-∆ effects,
were included in the elastic and inelastic analyses.
The numerical direct integration scheme implemented in
DRAIN-2DX [43] to solve the equations of motion is the
Newmark method [44]. Unconditional stability is obtained by
setting the integration parameters β = 0.25 and γ = 0.25.
A traditional proportional or Rayleigh damping was assumed
in the dynamic analyses; a 2% constant equivalent viscous
damping was utilized for the first and the third modes of
vibration. The adopted modal frequencies are listed in Table 2.
 L. Di Sarno, A.S. Elnashai / Journal of Constructional Steel Research 65 (2009) 452–465 457

Table 3
Characteristics of earthquake ground motions

Record Pr. of exc. (% in 50 yrs) Magnitude (Mw ) Source distance (km) PGA (g) PGV (m/s) PGD (cm) Arias intensity (m/s)
Morgan Hill 50 6.2 15 0.32 0.32 6.14 1.71
Whittier 50 7.3 17 0.77 0.92 11.32 5.42
Loma Prieta 10 7.0 12.4 0.66 0.70 18.41 4.24
Landers 10 7.3 36 0.42 0.36 16.08 2.10
Northridge 2 6.7 7.5 0.43 0.65 12.21 2.03
Kobe 2 6.9 3.4 1.28 1.46 30.31 14.61

Fig. 6. Spectral accelerations (left) and spectral velocity (right) for the earthquake ground motions used (damping = 5%).

The damping value assumed for the dynamic analyses is not

essential; the reason is twofold. First, the damping value does
not greatly affect frames responding in the inelastic range.
Second, it is primarily of interest to compare the behaviour
of each of the sample structures; therefore consistency is
paramount to correctly interpret the results. In the present study
the results of the inelastic time history analyses are discussed.

3.3. Ground motion records

Response-history analyses were carried out by employing

suites of ground motions developed for the FEMA-SAC steel
project in the USA [45]. These earthquakes are spectrum-
compatible; they include horizontal records matching the
1997 NERHP design spectrum [46]. The selected ground
accelerations correspond to the 1997 USGS hazard level for Los
Angeles downtown area, i.e. seismic zone 4. The seismological
properties of the records used for this study are summarized
in Table 3. Three levels of seismic hazard were employed:
50%, 10% and 2% probability of exceedence in a 50-year
period. The selected earthquake scenario is compliant with
the performance-based engineering framework encompassing
multi-level limit states (e.g. [13,38]).
The distances from the sources for the earthquake records
range between 3.4 km (Kobe, in Japan) and 36 km (Landers,
in California). Therefore, the above suite of strong motions
covers a well-defined design scenarios, i.e. near and far-field
conditions.
The plots in Fig. 6 show the 5% damped acceleration and
velocity response spectra of the ensemble of accelerograms.
Large amplification factors of the spectral accelerations are
evaluated for Kobe (factors vary between 3.1 and 3.5) and
Whittier (factors vary between 2.2 and 2.4) in the range of the
natural periods of vibrations of the prototype SCBFs, BRBFs
and MBFs.
Fig. 7. Duration of the sample earthquake records.
The duration of the ground motion and the corresponding
number of cycles are of paramount importance for the assess-
ment of low-cycle fatigue damage in structural earthquake en-
gineering applications (e.g. [47]). The selected records exhibit
different durations as displayed in Fig. 7 in which the estimated
of the bracketed, significant and uniform durations are summa-
rized. It is noted that the definitions of significant and uniform
durations lead to similar values for all earthquakes. This out-
come depends on the engineering seismology properties of the
sample accelerograms (e.g. [48]).
The energy content, expressed in terms of ARIAS intensity
A I , was also computed and the values are listed in Table 3.
It is found that Kobe has the shortest duration (6.86 and
9.38 s, for significant and uniform durations, respectively) but
it has the highest seismic demand in terms of input energy
(A I = 14.61 m/s). This observation is further confirmed by
the velocity response spectra provided in Fig. 6.
3.4. Structural performance criteria
Performance levels are generally set depending on the
construction material and structural type, the importance of
View publication stats