Professional Documents
Culture Documents
Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct
a r t i c l e
i n f o
Article history:
Received 15 November 2012
Revised 2 July 2013
Accepted 8 July 2013
Keywords:
Shear wall
Concrete
Earthquake
Chile
Design
Connement
Buckling
a b s t r a c t
The Mw 8.8 earthquake in 2010 in Chile provided valuable information regarding the damage and potential design code changes for reinforced concrete (RC) structures. Many modern RC buildings suffered
severe damage, mainly in the form of concrete cover spalling, followed by longitudinal boundary bar
buckling and concrete crushing. The absence of wall boundary detailing explains such behavior, but older
structures and thousands of other buildings suffered minor or no damage at all. An analysis of some
potential factors that inuenced the damage caused is carried out and the fundamental principles for
the Chilean RC design code changes are exposed. Currently, a displacement-based approach is used for
detailing of the wall boundary elements as well as for establishing damage limitation. Compressive concrete strain cannot exceed 0.008, limiting indirectly the axial load, which is one of the potential precursors of the damage. Transversal reinforcement is also provided in zones of potential yielding of the
longitudinal reinforcement such that buckling due to compression preceded by tension is minimized; this
buckling is more likely in asymmetric cross-sections, such as T-shaped sections, where usually the web
goes into larger tensile and compressive strains given the variation of location of the neutral axis for the
web under tension or compression.
2013 Elsevier Ltd. All rights reserved.
1. Introduction
The Mw 8.8 2010 earthquake in Chile impacted the country
over a coastal length of about 500 km, covering the most densely
populated regions. Within that region, the reinforced concrete
(RC) building stock increased in the last two decades, with more
than 1000 buildings with 1030 stories. During that period, the
construction of residential buildings greater than 15 stories became more common whereas earlier, construction was focused
on buildings with less than 15 stories. Modern buildings also presented a few characteristics that differ from older structures, such
as thin walls and discontinuities, among others [1]. The damage
was concentrated on newer structures, which also probably presents the largest building stock, specically taller buildings (usually over 15 stories). Most buildings presented no structural
damage or suffered very moderate damage, which required no
eviction. Some other buildings (40) presented severe damage that
was either repaired or, in rare cases, demolished. It is interesting to
note that intermediate situations were uncommon-that is, they
either presented serious damage or almost no damage at all, implying a brittle nature of the behavior. The severity of damage in modern buildings prompted the engineering community to study the
general nature of damage caused and propose an emergency code
Tel.: +56 229784984.
E-mail address: lmassone@ing.uchile.cl
0141-0296/$ - see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.engstruct.2013.07.013
to prevent current buildings from suffering similar damage in future earthquakes. Studies are currently being carried out to provide
additional information for future code considerations.
Modern RC building construction in Chile is based on shear
walls, which are designed according to ACI 318. The formal reference to ACI 318 was originally implemented though the Chilean
seismic code (NCh433.Of96 [2]), which requires the use of
ACI318-95 for RC building construction. However, based on the
good performance of RC shear wall buildings in Via del Mar in
the 1985 earthquake, the ACI 318-95 requirements for special
transverse reinforcement at wall boundaries to conne the concrete and restrain rebar bucking were eliminated. Seismic design
approaches for most RC buildings include modal response spectrum analysis, and the design spectra are based on historical
earthquakes.
As summarized by Massone et al. [1], typical buildings in Chile
have a large number of shear walls, with a total wall area over wall
plan area of roughly 3% in each principal direction of a building.
Although the wall area has been maintained over the years, newer
structures increased the number of stories, moving the average
building of 15 stories to about 25 oors. The increment of stories
has resulted in higher axial loads in newer buildings [3]. The design
code did not put a limit on the level of axial stress allowed for gravity load or combined gravity and lateral loads. The construction of
new buildings is also characterized by the use of a common layout:
longitudinal walls on each side of a central corridor that form a
1336
the capability to sustain an important axial load, given the presence of basically unconned concrete. Upon spalling of the concrete cover, 90 hooks on the transverse reinforcement opened
due to buckling of the longitudinal reinforcement. Buckling of
the longitudinal reinforcement in walls was also combined, in
some cases, with the fracture of the reinforcement, likely due to
large cyclic strains.
Even though other failure types were observed, this work focuses on the code changes and their fundamental principles
regarding the failure caused by actions of axial load combined with
bending.
2
2
an initial tension from a length of 2 10d 4d 21:5d, which
yields a total deformation of (21.5d 20d)/21.5d = 7%). This indicates that the buckled reinforcement would have almost no residual capacity (Fig. 3).
Fig. 2b shows a series of buckled bars on the right side of the
wall section, whereas on the left side last three bars also present
fracture at mid-height of the buckle length. Fracture in this case,
as can be seen in the gure, shows no indication of a reduced
cross-section or necking of the bar due to stress concentration at
the fracture zone, as commonly seen in direct tension tests. Failure
is rather associated to fracture due to fatigue, which presents an almost clean horizontal cut of the reinforcement. This type of failure
under low strain or stress is common in mechanical equipment
requiring millions or thousands of cycles. In our case, where few
1337
Fig. 2. Boundary wall reinforcement (a) general view, and (b) fracture and buckling.
fb/fs (degradation)
L/d=6
L/d=10
0.8
L/d=14
0.6
0.4
0.2
0.01
0.02
0.03
0.04
Strain (max.)
Fig. 3. Compressive degradation due to buckling for different L/d and strain values.
cycles with large wall lateral displacement demands were observed from the earthquake associated with a large strain magnitude, fracture can occur in less than 100 cycles. This is
commonly known as low cycle fatigue. Tests by previous researchers have provided an expression that allows for an estimation of
half cycles to failure (2Nf) for specic total strain amplitude (ea-half
of the range amplitude). Brown and Kunnath [7] who studied cyclic
tests under symmetric tensioncompression determined that for
25 mm-diameter bar, which is a typical bar diameter for boundary
bars in Chile, the bar life fatigue is given by ea = 0,08(2Nf)0.36, from
where, given a 3.5% amplitude strain (assuming a total strain range
of 7%, as mentioned previously), the number of half cycles that results is 10 (i.e., 5 cycles), which is consistent with the number of
cycles of large magnitude that can be observed in the earthquake
records. That expression also indicates that in order to avoid low
cycle fatigue failure for less than 50 cycles, the total strain range
should not go over 3%.
The question that remains then is-how could we observe less
strain demands in longitudinal reinforcement in order to avoid rebar fracture?, or why was not observed fracture in all cases? One of
the important aspects that caused damage in walls is the impossibility of the nonlinear or plastic deformation to distribute over a
height larger than 400500 mm. That damaged zone is consistent
with the buckled length of the reinforcement. Distribution of damage (or plastic deformation) over a larger zone would reduce the
curvature demands, and therefore the compressive strain demands
for a prescribed wall top displacement. Test results from the literature (e.g., Thomsen and Wallace [8]) have shown that the plastic
hinge could easily be in the range of 0.33lw and 0.5lw for relatively
1338
0.04
A - TW2
B - TW2
RW2
0.02
A
0
-0.02
+ tension
- compression
II
1%
Phase I
1.5%
Drift Level
Fig. 5. Strain history for extreme bers (ber model)-TW2 and RW2.
Phase II
1339
RW2 are larger than tensile strains in the ange (B), but with similar compressive strains.
Concrete crushing might not be the only concern about the
behavior of non-symmetric sections. Large tensile strains in the
web end (A) followed (upon reversal) by large compressive strains
in the same point can lead to buckling of the reinforcement, making T-shaped walls (or in general non-symmetric) more prone to
this type of damage. Moreover, even large tensile strains followed
by little or no compressive strains can lead to buckling [1214], as
shown in Fig. 6. Fig. 6 shows the analytical response (plastic hinge
model [5]-as previously described for L/d = 14) of a bar subjected to
buckling for two different loading histories, but with the same total
strain variation (6%, Fig. 6a ranging from 3% to 3%, and Fig. 6b
ranging from 2% to 4%). In both cases, onset of buckling (marked
in the gure) is observed almost immediately upon reversal from
tension.
du
du u lp hw
-0.02 to 0.04
(b)
(a)
0
-400
Buckling
No Buckling
Buckling
No Buckling
-800
-0.04
-0.02
0.02
0.04
-0.02
0.02
Strain
Strain
Fig. 6. Initiation of buckling (a) 3% to 3% strain range, and (b) 2% to 4% strain range.
-0.03 to 0.03
800
400
11
lp
2
y hw u y lp hw
40
2
0.04
1340
example, for loading phases I and II, considering that top displacements are identical. Curvature can be estimated from the gure as
the difference between the strain values at extreme ends (distance
between peaks values of A and B-shown in Fig. 5 with vertical arrows) over the length of the wall. As can be seen, in phase I, the curvature is about 25% larger than in phase II. That difference is larger
with a smaller top displacement and smaller with a larger top displacement. This comes from the fact that the yield curvature, and
therefore the yield top displacement for T-shaped walls are different in each direction. The yield curvature tends to be smaller in
T-shaped walls with anges in compression [20], resulting in larger
curvature demands in the nonlinear range when compared to the
same top displacement for the wall with the ange in tension,
where the yield curvature is larger.
7. Axial load
Damage during the 2010 earthquake in Chile was more common in modern RC wall buildings, rather than old RC wall structures. Older RC buildings were constructed usually past the
1950s. Newer structures were usually constructed around year
2000, and featured narrow wall thickness (for taller buildings)
and the presence of discontinuities, which were less common in
previously built structures. Considering that almost no building
had connement, two main, different aspects arise: (1) axial load
level, and (2) discontinuities. Both can affect the strain distribution
(e.g., Eqs. (1) and (2)). The treatment of discontinuities is a specic
and difcult task and most design tools are strength-oriented, such
as strut-and-tie models [23]. For the purpose of this work all analysis of strain distribution are related to axial load.
Few research programs have focused their attention to the axial
load effect in slender walls, especially if we consider loads larger
than 0.2fc0 Ag [21,22]. The work by Zhang and Wang [21] included
a test program that compared mid-height (effective height = 1.5 m,
length = 0.7 m) RC and steel reinforced concrete (SRC, partial
boundary longitudinal reinforcement replaced by steel sections)
walls under cyclic lateral loads. The specimens were tested in cantilever with a point top lateral load and a constant axial load applied at the wall top. The RC walls (SW7 and SW8) had similar
geometry, reinforcement conguration, and material properties,
except for the boundary reinforcement ratio (SW7: qb = 0.062,
SW8: qb = 0.046) and axial load (SW7: 0.24fc0 Ag , SW8: 0.35fc0 Ag ).
Test results for the overall load (backbone curve) vs. displacement
(at 1.5 m from base) are shown in Fig. 8. The gure shows a larger
displacement capacity for the specimen with a lower axial load,
although it had a larger amount of longitudinal boundary reinforcement. It is also important to point out that with the prescribed
250
0.02
0.015
150
u/hw
Load (kN)
200
100
SW7
50
0.01
0.005
SW8
Damaged
buildings
San Pedro record
EW dir., 2% damp.
0
0
10
20
30
40
0
5
10
15
20
N
Fig. 9. Axial load effect in drift capacity.
25
30
1341
The observed damage from the earthquake as well as information gathered from acceleration records led the Chilean civil engineering community to demand changes to existing standards-the
design of reinforced concrete (NCh430.Of2008, [26]) and the seismic design of buildings (NCh433.Of96, [2]). The changes in this
seismic design code mainly focused on developing both new design displacement spectra and a modied soil classication, which
were reected in its replacement, the DS No. 61 [27], which had a
previous version earlier in 2011.
Regarding the modications of the design standard for reinforced concrete, it focused on requirements for an improved
behavior of reinforced concrete walls, whose replacement was
the DS No. 60 [28] (also a modication of an earlier version). The
code uses the ACI 318-08 as the base code and includes modications that focused on special wall design.
1.5
SoilD
SoilC
Soil B
Soil A
0.008
(a)
(b)
1
c/lw
1.3Sd (Tcr)/hw
0.012
0.5
0.004
SoilD
SoilC
Soil B
Soil A
0
0
Tcr (s)
Tcr (s)
Fig. 10. (a) Drift demands, and (b) neutral axis depth limitation for zone 3.
1342
40
Stress (MPa)
30
0.008
20
0.003
Confined
Unconfined
10
0
0
0.005
0.01
0.015
0.02
Strain
Fig. 11. Conned and unconned concrete.
(a)
before, for larger axial loads, the plastic hinge reduces which could
be considered in Eq. (1). In both cases, it is assumed that there is a
potential plastic hinge where the inelastic deformations are concentrated, and therefore, a special detailing requirement for the
edge of the wall is provided. Considering that by exceeding a compressive strain of 0.003 in the most compressed ber in the wall
(ACI 318-08) special detailing (transversal reinforcement) needs
to be provided at wall edges, the consideration of compressive
strain limitation of 0.008 seeks to limit the potential damage that
may occur. This limitation indirectly restricts the axial load level
(through the neutral axis depth) in the walls, which is dependent
on the expected demand for the wall displacement. Thus, for buildings in seismic zones closer to the hypocenter and in soft soils, the
expected wall top displacement demand will be higher, resulting
in larger compressive strains and, thus, it will be more restrictive
with the axial load level. Since the limitation is not strictly on
the axial load, other wall characteristics can be modied (e.g.,
amount of longitudinal reinforcement, yielding of the reinforcement, and geometry).
In order to study the impact of the damage limitation requirement, the neutral axis depth limitation is estimated on the basis
of the expected drift demands provided by the DS No. 61 [27].
The drift demands are estimated for zone 3 (coastal-highest seismicity) and for the four soil types that provide spectral displacement (AD). A fundamental period for an uncracked lower bound
stiffness of T(s) = hw(m)/50 is selected. The cracked stiffness is estimated as Tcr = 1.5 T [27]. Drift demands are plotted in Fig. 10a,
which indicates that the maximum drift level yields a value of
about 1.1%. The low drift levels are promoted by the large building
stiffness. Knowledge of the drift demand and the use of Eq. (2) provide the limitation to the neutral axis depth for a compressive
strain of 0.008, as shown in Fig. 10b. Zone 3 soil D yields the smallest value (c/lw = 0.37), which is consistent with a low axial load. For
a symmetric rectangular section with yielding boundary reinforcement and no web reinforcement, it yields an axial load of P/
fc0 Ag = 0.27. The axial load reduces considerable in non-symmetric
sections such as T-shaped walls, yielding values that might be half
(b)
Fig. 12. Onset of buckling (a) rectangular, and (b) T-shaped cross-section.
lw
,
2
c
lw
to es = 0.02, where es and ec are the extreme tensile and compressive strains in the wall. A 2% tensile strain is far from causing tensile fracture of the boundary reinforcement. A stable compressive
boundary element (connement and buckling restriction might
be required) capable of equilibrating the force from the tensile longitudinal reinforcement (and axial load) that guarantees a nearly
constant plastic hinge length (here assumed as lw/2) and low drift
demands (here assumed no more than 1.1%) will be capable of preventing tensile fracture of the reinforcement. Smaller plastic hinge
length (or a hinge that reduces its length with drift) or larger drift
demands would result in larger tensile strain values in the longitudinal reinforcement. Non-symmetric sections would probably
present larger neutral axis depth, reducing even further the tensile
strain.
Another important aspect related to the limitation of damage is
the fact that in order to have the distribution of plasticity along the
height of the wall, a positive post-yield stiffness is required. Considering that concrete in compression should be conned if the
compressive strain is larger than 0.003, low strain values in unconned concrete would present just slight degradation of capacity
associated with concrete in compression, but exhibit an overall stable response (e.g., momentcurvature). If we consider the connement requirements for wall special boundary elements by ACI 318f0
08, that is, Ash 0:09bc s fc0 , for normal material properties
y
Ash fy
bc s
1343
concrete, and therefore positive post-yield is expected at the lateral top drift demand, similarly to the case of unconned concrete
with maximum compressive strain below 0.003. The stress vs.
strain response in compression is shown in Fig. 11 for conned
and unconned concrete, where both the compressive strain limits
are shown (0.003 for unconned concrete and 0.008 for conned
concrete).
In cases requiring connement, the wall thickness must be at
least 300 mm, and the conned length may not be less than the
width of the wall in the conned zone to ensure proper containment of the transverse reinforcement. In this sense, and also to ensure good concrete placement and behavior, bar diameter size
limitations are imposed on both the longitudinal reinforcement
(db_long 6 1/9hSBE, with hSBE thickness of the wall at the boundary
element) and the transversal reinforcement (db_transv P 1/3db_long)
at wall boundary elements, as well the vertical (s 6 1/2hSBE) and
horizontal (hx 6 1/2hSBE, 200 mm) separation of supported vertical
bars by transversal reinforcement. Additionally, both the stirrups
as cross-ties must have their ends bent at an angle greater than
or equal to 135.
Where connement is not required, but there is a longitudinal
reinforcement ratio greater than 2.8/fy, and it is expected that this
reinforcement might yield, the spacing of the transverse reinforcement is limited to 6db and 200 mm, with db the diameter of the longitudinal reinforcement. This limitation is consistent to the one
provided in ACI 318-08; however, here the spacing limitation to
6db prevents or delays reinforcement buckling to wall drift levels
that are expected in stiff shear wall structures in Chile. Thus, for
example, in asymmetrical walls, such as T-shaped wall sections,
where the compression strain do not exceed 0.003, they are required to have buckling restriction as these bars do not reach signicant compression deformation, but they might buckle if they
have previously endured large tensile strains. They could even
buckle while still elongated, that is, under tensile strain, but subject to compression stress. In order to quantify the impact of such
limitation, the drift capacities should be estimated based on the
initiation of bar buckling, providing a limit state for the buckling
condition.
Collection of experimental data by Rodriguez et al. [12] for steel
coupons loaded under cyclic strain history, recalling bars in columns and beams under seismic actions, show how the onset of
buckling is dependent on stirrups spacing (xed-end condition).
The onset of buckling is dened experimentally, by Rodriguez
et al., as the point where the instability of the bar was captured
with extensometers place on two sides of the bar, where one of
the extensometers presented increase of compression (concave
side of the buckle bar) and the other decrease of compression (convex side of the buckle bar). Initiation of buckling was dened at the
point where the difference between both sensors was 20% of the
compression observed in the concave side. The strain at buckling
was studied by means of a strain ep* that corresponds to the total
strain from the previous zero stress point until buckling was determined. This, for simplicity, can be approximated to be the entire
strain between the maximum tensile strain and the subsequent
compressive (or tensile) strain at the onset of buckling. If we consider a rectangular wall cross-section with symmetric reinforcement under a seismic action from left to right that results in a
strain prole described in Fig. 12a (direction 1), the extreme bers
would reach strains et and ec at both ends. Upon reversal (direction
2) and considering that the same top displacement is applied, the
maximum strains would be the same, but in reverse order, that
is, ec and et. Thus, each extreme ber would have reached a maximum strain of et + ec, which is approximately ep* once buckling begins. According to Rodriguez et al., for stirrups spacing of s = 6db,
the onset of buckling occurs, in average, at about ep* = 0.04. If we
1344
lw
2
eclwet
lw
2
0:04
0:02,
2
0:04
0:01. The drift capacity reduces to 0.01, which is
4
tension to compression causing instability even before compressive strains are observed.
Where damage was produced by large compressive or tensile
strains that resulted in degrading response (softening post-yield
response) with localization of damage within a small area, a better
distribution of strain is expected for a non-degrading response, by
providing connement and/or buckling constraint. The better
strain distribution results in smaller maximum strain values,
reducing damage and making less probable the presence of fracture of longitudinal reinforcement due to low cycle fatigue.
Other limitations were also imposed to improve the concrete
placement and effectiveness of connement, as well as prevention
of global buckling. Several analytical and experimental programs
have been carried out lately to study, in more detail, aspects such
as overall response, effective connement, effect of discontinuities,
among others, by the author and other researchers, which would
allow for the checking and improvement of the code provisions.
Acknowledgements
This work is the summary of long discussions within the subcommittee for the RC design code committee formed under the
Construction Institute (Chile). The author of this publication, coordinator of the subcommittee, would like to thank the contribution
from several persons related to the academia and industry who
served as members and collaborated by enriching the discussion
with examples, case studies, and analysis of the new code requirements and its impact on Chilean construction. The members, in
alphabetical order, are: Marcial Baeza, Patricio Bonelli, Leopoldo
Breschi, Jorge Carvallo, Rodrigo Concha, Luis Daz, Augusto Holmberg, Ivn Hrepic, Matas Hube, Denis Jequier, Rodrigo Jordn, Marianne Kpfer, Mario Lafontaine, Ren Lagos, Alfonso Larran, Carl
Lders, Rodrigo Mujica, Miguel Sndor, Hernn Santa Mara, Eduardo Santos, Rodolfo Saragoni, Carlos Seplveda, and Rodrigo Vsquez. The author would also thank the wall specimen photos
provided by prof. John Wallace at UCLA.
References
[1] Massone LM, Bonelli P, Lagos R, Lders C, Moehle J, Wallace JW. Seismic design
and construction practices for RC structural wall buildings. Earthq Spectra
2012;28(S1):S24556.
[2] NCh433.Of96. Earthquake resistant design of buildings. Instituto Nacional de
Normalizacin, INN 1996, 43pp [in Spanish].
[3] Caldern JA. Update on structural system characteristics used in RC buildings
in Chile. Civil Engineering Thesis, University of Chile; 2007. 76pp. [in Spanish],
[4] Wallace JW, Massone LM, Bonelli P, Dragovich J, Lagos R, Lders C, Moehle J.
Damage and implications for seismic design of RC structural wall buildings.
Earthq Spectra 2012;28(S1):S28199.
[5] Massone LM, Moroder D. Buckling modeling of reinforcing bars with
imperfections. Eng Struct 2009;31(3):75867.
[6] Lacaze C. Study and modeling of the impact of buckling in low cycle fatigue in
reinforcing bars for reinforced concrete. Civil Engineering Thesis, University of
Chile; 2009. 94pp [in Spanish].
[7] Brown J, Kunnath SK. Low-cycle fatigue failure of reinforcing steel bars. ACI
Mater J 2004;101(6):45766.
[8] Thomsen JH, Wallace JW. Displacement-based design of slender RC structural
walls-experimental verication. J Struct Eng, ASCE 2004;130(4):61830.
[9] Karamlou A, Kabir MZ. Experimental study of L-shaped slender R-ICF shear
walls under cyclic lateral loading. Eng Struct 2012;36:13446.
[10] Massone LM, Wallace JW. Load-deformation responses of slender reinforced
concrete walls. ACI Struct J 2004;101(1):10313.
[11] Cordero FA. Modeling of RC wall compression failure observed in the Mw 8.8
earthquake in 2010 in Chile. Civil Engineering Thesis, University of Chile;
2011. 105pp [in Spanish].
[12] Rodriguez ME, Botero JC, Villa J. Cyclic stressstrain behavior of reinforcing
steel including effect of buckling. J Struct Eng, ASCE 1999;125(6):60512.
[13] Moyer MJ, Kowalsky MJ. Inuence of tension strain on buckling of
reinforcement in concrete columns. ACI Struct J 2003;100(1):7585.
[14] Syntzirma DV, Pantazopoulou SJ, Aschheim M. Load-history effects on
deformation capacity of exural members limited by bar buckling. J Struct
Eng, ASCE 2010;136(1):111.
1345