Sie sind auf Seite 1von 27

Seismic Collapse Safety of Reinforced Concrete Buildings:

II. Comparative Assessment of Non-Ductile and Ductile Moment Frames

Abbie B. Liel1, Curt B. Haselton2, Gregory G. Deierlein3

ABSTRACT

This study is the second of two companion papers to examine the seismic collapse safety of
reinforced concrete frame buildings, and examines non-ductile moment frames that are
representative of those built prior to the mid-1970s in California. The probabilistic assessment
relies on nonlinear dynamic simulation of structural response to calculate the collapse risk,
accounting for uncertainties in ground motion characteristics and structural modeling. The
evaluation considers a set of archetypical non-ductile RC frame structures of varying height that
are designed according to the seismic provisions of the 1967 Uniform Building Code. The results
indicate that non-ductile RC frame structures have a mean annual frequency of collapse of
roughly 5 to 14 x 10-3 at a typical high seismic California site, which is approximately 40 times
higher than corresponding results for modern code-conforming special RC moment frames.
These metrics demonstrate the effectiveness of ductile detailing and capacity design
requirements, which have been introduced over the past thirty years to improve the safety of RC
buildings. Data on comparative safety between non-ductile and ductile frames may also inform

Assistant Professor, Dept. of Civil, Environmental and Architectural Engineering, University of Colorado,
Boulder, CO, 80309, U.S.A. Email: abbie.liel@colorado.edu
2
Assistant Professor, Department of Civil Engineering, California State University, Chico, CA, 95929, U.S.A.
3
Professor, Department of Civil and Environmental Engineering, Stanford University, Stanford, CA, 94305, U.S.A.

the development of policies for appraising and mitigating seismic collapse risk of existing RC
frame buildings.

CE Database Subject Headings: Collapse; earthquake engineering; structural reliability;


reinforced concrete structures; buildings, commercial; seismic effects.

INTRODUCTION

Reinforced concrete (RC) frame structures constructed in California prior to the mid-1970s lack

important features of good seismic design, such as strong columns and ductile detailing of
reinforcement, making them potentially vulnerable to earthquake-induced collapse. These nonductile RC frame structures have incurred significant damage in the 1971 San Fernando, 1979
Imperial Valley, 1987 Whittier Narrows, and 1994 Northridge Earthquakes in California, and
many other earthquakes worldwide. These factors raise concerns that some of Californias
approximately 40,000 non-ductile RC structures may present a significant hazard to life safety in
future earthquakes. However, data is lacking to gauge the significance of this risk, in relation to
either the building population at large or to specific buildings. The collapse risk of an individual
building depends not only on the building code provisions employed in its original design, but
also structural configuration, construction quality, building location and site-specific seismic
hazard information. Apart from the challenges of accurately evaluating the collapse risk is the
question of risk tolerance and the minimum level of safety that is appropriate for buildings. In
this regard, comparative assessment of buildings designed according to old versus modern
building codes provides a means of evaluating the level of acceptable risk implied by current
design practice.
2

Building code requirements for seismic design and detailing of reinforced concrete have
changed significantly since the mid-1970s, in response to observed earthquake damage and an
increased understanding of the importance of ductile detailing of reinforcement. In contrast to
older non-ductile RC frames, modern code-conforming special moment frames for high seismic
regions employ a variety of capacity design provisions that prevent or delay unfavorable failure
modes such as column shear failure, beam-column joint failure, and soft-story mechanisms.
While there is general agreement that these changes to building code requirements are
appropriate, there is little data to quantify the associated improvements in seismic safety.
Performance-based earthquake engineering methods are applied in this study to assess the
likelihood of earthquake-induced collapse in archetypical non-ductile RC frame structures.
Performance-based earthquake engineering provides a probabilistic framework for relating
ground motion intensity to structural response and building performance through nonlinear timehistory simulation (Deierlein 2004). The evaluation of non-ductile RC frame structures is based
on a set of archetypical structures designed according to the provisions of the 1967 Uniform
Building Code (UBC) (ICBO 1967). These archetype structures are representative of regular
well-designed RC frame structures constructed in California between approximately 1950 and
1975. Collapse is predicted through nonlinear dynamic analysis of the archetype non-ductile RC
frames, using simulation models capable of capturing the critical aspects of strength and stiffness
deterioration as the structure collapses. The outcome of the collapse performance assessment is
a set of measures of building safety, relating seismic collapse resistance to seismic hazard. These
results are compared to the metrics for ductile RC frames reported in a companion paper
(Haselton et al. 2010a).

ARCHETYPICAL REINFORCED CONCRETE FRAME STRUCTURES

The archetype non-ductile RC frame structures represent the expected range in design and
performance in Californias older RC frame buildings, considering variations in structural height,
configuration and design details. The archetype configurations explore key design parameters for
RC components and frames, which were identified through previous analytical and experimental
studies reviewed by Haselton et al. (2008). The complete set of archetype non-ductile RC frame
buildings developed for this study includes 26 designs (Liel and Deierlein 2008). This paper
focuses primarily on 12 of these designs, varying in height from 2 to 12 stories, and including
both perimeter (P) and space (S) frame lateral resisting systems with alternative design details.
All archetype buildings are designed for office occupancies with an 8-inch flat-slab floor
system and 25-ft. (7.6 m.) column spacing. The 2- and 4-story buildings have a footprint of 125
ft. by 175 ft. (38.1 m. by 53.3 m.), and the 8- and 12-story buildings measure 125 ft (38.1 m)
square in plan. Story heights are 15 ft. (4.6 m.) in the first story and 13 ft. (4.0 m.) in all other
stories. Original structural drawings for RC frame buildings constructed in California in the
1960s were used to establish typical structural configurations and geometry for archetype
structures (Liel and Deierlein 2008). The archetypes are limited to RC moment frames without
infill walls, and are regular in elevation and plan, without major strength or stiffness
irregularities.
The non-ductile RC archetype structures are designed for the highest seismic zone in the
1967 UBC, Zone 3, which at that time included most of California. Structural designs of twodimensional frames are governed by the required strength and stiffness to satisfy gravity and
seismic loading combinations. The designs also satisfy all relevant building code requirements,

including maximum and minimum reinforcement ratios, maximum stirrup spacing, etc. The 1967
UBC permitted an optional reduction in the design base shear if ductile detailing requirements
were employed, however, this reduction is not applied and only standard levels of detailing are
considered in this study. Design details for each structure are summarized in Table 1, and
complete documentation of the non-ductile RC archetypes is available in Liel and Deierlein
(2008). Four of the 4- and 12-story designs have enhanced detailing, as described later.
The collapse performance of archetypical non-ductile RC frame structures is compared to the
set of ductile RC frame archetypes presented in the companion paper (Haselton et al. 2010a). As
summarized in Table 2, these ductile frames are designed according to the provisions of the
International Building Code (ICC 2003), ASCE 7 (ASCE 2002), and ACI 318 (ACI 2005), and
meet all governing code requirements for strength, stiffness, capacity design, and detailing for
special moment frames. The structures benefit from the provisions that have been incorporated
into seismic design codes for reinforced concrete since the 1970s, including an assortment of
capacity design provisions (e.g. strong column-weak beam (SCWB) ratios, beam-column and
joint shear capacity design, etc.) and detailing improvements (e.g. transverse confinement in
beam-column hinge regions, increased lap splice requirements, closed hooks etc.). The ductile
RC frames are designed for a typical high seismic Los Angeles site with soil class Sd that is
located in the transition region of the 2003 IBC design maps (Haselton and Deierlein 2007).
A comparison of the structures described in Tables 1 and 2 reflects four decades of changes
to seismic design provisions for RC moment frames. Despite modifications to the period-based
equation for design base shear, the resulting base shear coefficient is relatively similar for nonductile and ductile RC frames of the same height, except in the shortest structures. More
significant differences between the two sets of buildings are apparent in member design and

detailing, especially in the quantity, distribution and detailing of transverse reinforcement.


Modern RC frames are subject to shear capacity design provisions and more stringent limitations
on stirrup spacing, such that transverse reinforcement is spaced 2 to 4 times more closely in
ductile RC beams and columns. The SCWB ratio enforces minimum column strengths to delay
the formation of story mechanisms. As a result, the ratio of column to beam strength at each joint
is approximately 30% higher (on average) in the ductile RC frames than the non-ductile RC
frames. Non-ductile RC frames also have no special provision for design or reinforcement of the
beam-column joint region, whereas columns in ductile RC frames are sized to meet joint shear
demands with transverse reinforcement in the joints. Joint shear strength requirements in special
moment frames tend to increase the column size, thereby reducing axial load ratios in columns.

NONLINEAR SIMULATION MODELS

Nonlinear analysis models for each archetype non-ductile RC frame consist of a two-dimensional
three-bay representation of the lateral resisting system, as shown in Figure 1Error! Reference
source not found.. The analytical model represents material nonlinearities in beams, columns,
and beam-column joints, as well as large deformation (P-) effects that are important for
simulating collapse of frames. Beam and column ends and the beam-column joint regions are
modeled with member end hinges that are kinematically constrained to represent finite joint size
effects and connected to a joint shear spring (Lowes and Altoontash 2003). The structural models
do not include any contribution from non-structural components or from gravity-load resisting
structural elements that are not part of the lateral resisting system. The model is implemented in
OpenSees with robust convergence algorithms (OpenSees 2009).

As in the companion paper, inelastic beams, columns, and joints are modeled with
concentrated springs idealized by a tri-linear backbone curve and associated hysteretic rules
developed by Ibarra et al. (2005). Properties of the nonlinear springs representing beam and
column elements are predicted from a series of empirical relationships relating column design
characteristics to modeling parameters, calibrated to experimental data for RC columns
(Haselton et al. 2008). Tests used to develop empirical relationships include a large number of
RC columns with non-ductile detailing, and predicted model parameters reflect the observed
differences in moment-rotation behavior between non-ductile and ductile RC elements. As in the
companion paper, calibration of model parameters for RC beams is based on columns tested
with low axial load levels because of the sparse available beam data. Figure 2a shows column
monotonic backbone curve properties for a ductile and non-ductile column (each from a 4-story
building). The plastic rotation capacity, cap,pl, which is known to have an important influence on
collapse prediction, is a function of the amount of column confinement reinforcement and axial
load levels, and is approximately 2.7 times greater for the ductile RC column. The ductile RC
column also has a larger post-capping rotation capacity (pc), which affects the rate of post-peak
strength degradation. Figure 2b illustrates cyclic deterioration of column strength and stiffness
under a typical loading protocol. Cyclic degradation of the initial backbone curve is controlled
by the deterioration parameter, , which is a measure of the energy dissipation capacity and is
smaller in non-ductile columns due to poor confinement and higher axial loads. Model
parameters are calibrated to the expected level of axial compression in columns due to gravity
loads and do not account for axial-flexure-shear interaction during the analysis, which may be
significant in taller buildings.

Modeling parameters for typical RC columns in non-ductile and ductile archetypes are
summarized in Tables 3 and 4. Properties for RC beams are similar and reported elsewhere (Liel
and Deierlein 2008; Haselton and Deierlein 2007). All element model properties are calibrated to
median values of test data. While the hysteretic beam and column spring parameters incorporate
bond slip at the member ends, they do not account for significant degradations that may occur
due to anchorage or splice failure in non-ductile frames.
Unlike ductile RC frames, in which capacity design requirements limit joint shear
deformations, non-ductile RC frames may experience significant joint shear damage contributing
to collapse (Liel and Deierlein 2008). Joint shear behavior is modeled with an inelastic spring, as
illustrated in Figure 1 and defined by a monotonic backbone and hysteretic rules (similar to those
shown in Figure 2 for columns). The properties of the joint shear spring are based on selected
sub-assembly data of joints with minimal amounts of transverse reinforcement and other nonductile characteristics. Unfortunately, available data on non-conforming joints is limited.
Joint shear strength is computed using a modified version of the ACI 318 equation (ACI
2005), and depends on joint size (bj joint width, h height), concrete compressive strength
( f ' c , units: psi) and confinement (, which is 12 to 20 depending on the configuration of
confining beams) such that V 0.7

f 'c b j h . The 0.7 modification factor is based on empirical

data from Mitra and Lowes (2007) and reflects differences in shear strength between seismically
detailed joints (as assumed in ACI 318 and Chp. 21) and joints without transverse reinforcement,
of the type considered in this study. Unlike conforming RC joints, which are assumed to behave
linear elastically, non-ductile RC joints have limited ductility and shear plastic deformation
capacity is assumed to be 0.015 and 0.010 radians for interior and exterior joints, respectively
(based on Moehle et al. 2006). For joints with axial load levels below 0.095, data from Pantelides

et al. (2002) is used as the basis for a linear increase in deformation capacity (to a maximum of
0.025 at zero axial load). Limited available data suggests a negative post-capping slope of
approximately 10% of the effective initial stiffness is appropriate. Due to lack of data, cyclic
deterioration properties are assumed to be the same as for RC beams and columns.
The calculated elastic fundamental periods of the RC frame models, reported in Tables 3 and
4, reflect the effective cracked stiffness of the beams and columns (35% of EIg for RC beams;
35% to 80% of EIg for columns), finite joint sizes, and panel zone flexibility. The effective
member stiffness properties are determined based on deformations at 40% of the yield strength
and include bond-slip at the member ends. The computed periods are significantly larger than
values calculated from simplified formulas in ASCE 7 (2005) and other standards, owing to the
structural modeling assumptions (specifically, the assumed effective stiffness and the exclusion
of the gravity-resisting system from the analysis model) and intentional conservatism in codebased formulae for building period.
Nonlinear static (pushover) analysis of archetype analysis models shows that the modern RC
frames are stronger and have higher deformation capacities than their non-ductile counterparts,
as illustrated in Figure 3. The ASCE 7-05 equivalent seismic load distribution is applied in the
analysis. Lateral strength is compared based on the overstrength ratio, , defined as the ratio
between the ultimate strength and the design base shear. The ductility is compared based on the
ultimate roof drift ratio (RDRult), defined as the roof drift ratio at which 20% of the lateral
strength of the structure has been lost. As summarized in Tables 5 and 6, for the archetype
designs in this study, the ductile RC frames have approximately 40% more overstrength and
ultimate roof drift ratios 3 times larger than the non-ductile RC frames. The larger structural
deformation capacity and overstrength in the ductile frames results from (1) greater deformation

capacity in ductile versus non-ductile RC components, (e.g. compare column cap,pl and pc in
Tables 3 and 4), (2) the SCWB requirements that promote more distributed yielding over
multiple stories in the ductile frames, (3) the larger column strengths in ductile frames that result
from the SCWB and joint shear strength requirements, and (4) the required ratios of positive and
negative bending strength of the beams in the ductile frames. Figure 3b illustrates the damage
concentration in lower stories, especially in the non-ductile archetype structures. While nonlinear
static methods are not integral to the dynamic collapse analyses, the pushover results help to
relate the dynamic collapse analysis results, described later, and codified nonlinear static
assessment procedures.

COLLAPSE PERFORMANCE ASSESSMENT PROCEDURE

Seismic collapse performance assessment for archetype non-ductile RC frame structures follows
the same procedure as in the companion study of ductile RC frames (Haselton et al. 2010a). The
collapse assessment is organized using incremental dynamic analysis (IDA) of nonlinear
simulation models, where each RC frame model is subjected to analysis under multiple ground
motions that are scaled to increasing amplitudes. For each ground motion, collapse is defined
based on the intensity (spectral acceleration at the first-mode period of the analysis model) of the
input ground motion that results in structural collapse, as identified in the analysis by excessive
interstory drifts. The IDA is repeated for each record in a suite of 80 ground motions, whose
properties along with selection and scaling procedures are described by Haselton et al. (2010a).
The outcome of this assessment is a lognormal distribution (median, standard deviation) relating
that structures probability of collapse to the ground motion intensity, representing a structural

10

collapse fragility function. Uncertainty in prediction of the intensity at which collapse occurs,
termed record-to-record uncertainty (ln,RTR), is associated with variation in frequency content
and other characteristics of ground motion records.
While the nonlinear analysis model for RC frames can simulate sidesway collapse associated
with strength and stiffness degradation in the flexural hinges of the beams and columns and
beam-column joint shear deformations, the analysis model does not directly capture column
shear failure. The columns in the archetype buildings in this study are expected to yield first in
flexure, followed by shear failure (Elwood and Moehle 2005), rather than direct shear failure, as
may be experienced by short, squat non-ductile RC columns. However, observed earthquake
damage and laboratory studies have shown that shear failure and subsequent loss of gravity-load
bearing capacity in one column could lead to progressive collapse in non-ductile RC frames.
Column shear failure is not incorporated directly because of the difficulties in accurately
simulating shear or flexure-shear failure and subsequent loss of axial load-carrying capacity
(Elwood 2004).
Collapse modes related to column shear failure are therefore detected by post-processing
dynamic analysis results using component limit state checks. Component limit state functions are
developed from experimental data on non-ductile beam-columns, and predict the median column
drift ratio (CDR) at which shear failure, and the subsequent loss of vertical-load carrying
capacity, will occur. Here, column drift ratio is defined similarly to interstory drift ratio, but
excludes the contribution of beam rotation and joint deformation to the total drift, because the
functions are based on data from column component tests.
Component fragility relationships for columns failing in flexure-shear developed by Aslani
and Miranda (2005), building on work by Elwood (2004), are employed in this study. For

11

columns with non-ductile shear design and detailing in this study and axial load ratios of P/Agfc
between 0.03 and 0.35, Aslani and Miranda (2005) predict that shear failure occurs at a median
CDR between 0.017 and 0.032 radians, depending on the properties of the column, and the
deformation capacity decreases with increasing axial load. Subsequent loss of vertical-carrying
capacity in a column is predicted to occur at a median CDR between 0.032 and 0.10 radians,
again depending on the properties of the column.
Since the loss of vertical load-carrying capacity of a column may precipitate progressive
structure collapse, this damage state is defined as collapse in this assessment. In post-processing
dynamic analysis results, the vertical collapse limit state is reached if, during the analysis, the
drift in any column exceeds the median value of that columns component fragility function. If
the vertical collapse mode is predicted to occur at a smaller ground motion intensity than the
sidesway collapse mode (for a particular record) the collapse statistics are updated. This
simplified approach can be shown to give comparable median results to the convolving the
probability distribution of column drifts experienced as a function of ground motion intensity
(engineering demands), with the component fragility curve (capacity). The total uncertainty in
the collapse fragility is assumed to be similar in the sidesway-only case and the sidesway/axial
collapse case, as it is driven by modeling and record-to-record uncertainties rather than
uncertainty in the component fragilities.
Incorporating this vertical collapse limit state has the effect of reducing the predicted
collapse capacity of the structure. Figure 4 illustrates the collapse fragility curves for the 8-story
RC space frame, with and without consideration of shear failure and axial failure following
shear. As shown, if one considers collapse to occur with column shear failure, then the collapse
fragility can reduce considerably compared to the sidesway collapse mode. However, if one

12

assumes that shear failure of one column does not constitute collapse and that collapse is instead
associated with the loss in column axial capacity, then the resulting collapse capacity is only
slightly less than calculations for sidesway alone. For the non-ductile RC frame structures
considered in this study, the limit state check for loss of vertical carrying capacity reduces the
median collapse capacity by 2% to 30% as compared to the sidesway collapse statistics that are
computed without this check (Liel and Deierlein 2008).
Typical computed values of the record-to-record variability (ln,RTR) are between 0.35 and
0.45. Uncertainty in how well the nonlinear simulation model represents the behavior of the real
building (the so-called modeling uncertainty) is incorporated in the collapse assessment by
increasing the dispersion in the collapse fragility. Liel et al. (2009) have previously demonstrated
the importance of accounting for modeling uncertainty in assessment of seismic collapse risk
because of the large underlying uncertainties in parameters related to component deformation
capacity and post-capping (softening) behavior and the highly nonlinear behavior as the building
collapses. The total uncertainty is computed by combining record-to-record and modeling
uncertainty using the square-root-of-the-sum-of-the squares: ln,Total2 = ln,RTR2 + ln,Modeling2. This
simplified approach provides reasonable estimates of the probability of collapse in the lower tail
of the distribution and the mean annual frequency of collapse (Liel et al. 2009). Assuming a
modeling uncertainty of ln,Modeling = 0.50, based on studies by Haselton and Deierlein (2007),
the total uncertainty in the collapse fragility is on the order of ln,Total = 0.61 to 0.67 for the
archetype non-ductile RC frames.
As described by Haselton et al. (2010a, 2010b), parameters defining the collapse fragility
function are also adjusted to account for differences in spectral shape between rare ground
motions in California and the general ground motion set used in this study. On average, this

13

adjustment increases the median collapse capacity of a non-ductile RC frame by 25%. This
increase is less than the comparable increase of 60% in the predicted collapse capacity of ductile
RC frames because (1) the non-ductile RC frames tend to experience less period elongation
before collapse, which is necessary to benefit from the spectral shape effect, and (2) because the
non-ductile frames tend to collapse at smaller spectral accelerations which correspond to more
frequent (less rare) ground motions that are represented more directly by the general ground
motion record set. Interested readers are referred to Haselton et al. (2010b) for the details of this
procedure. The final collapse fragility curves for the archetype buildings, discussed later in this
paper, incorporate modeling uncertainty and spectral shape adjustments.
Key metrics of seismic collapse performance include the median collapse capacity, the
collapse margin ratio, and the mean annual frequency of collapse. The median collapse capacity
is expressed in terms of ground motion intensity, Sa(T1) [g], which depends on the fundamental
period of the archetype model. Another metric, which can be compared more easily between
archetype buildings, is the collapse margin ratio, defined as the median collapse capacity divided
by the ground motion intensity with 2% chance of exceedance in 50 years, Sa2/50(T1). Sa2/50 is
selected as the normalization parameter since it is the basis for the code-defined maximum
considered earthquake at most sites (ASCE 2005). A third metric is the mean annual frequency
of collapse (collapse), which is obtained by integrating the collapse fragility function with a sitespecific hazard curve. The hazard curve represents the likelihood that a ground motion with
specified intensity is exceeded at a particular site. For the purpose of this study, the hazard curve
is defined for a high seismic site in the Los Angeles basin that is not affected by near-field
directivity and is generally representative of high-seismic regions (Goulet et al. 2007). Site-

14

specific seismic hazard analysis would change the assessment at other California sites, but
archetype building designs are the same throughout the Zone 3 region in the 1967 UBC.

RESULTS OF SEISMIC COLLAPSE ASSESSMENT

Collapse performance assessments for all of the archetypical RC frames are reported in Tables 5
and 6. Recall that the collapse assessments account for both sidesway (simulated) and vertical
(non-simulated) collapse modes and incorporate uncertainties in structural modeling. In addition
to the ground motion intensity measures of collapse performance, the roof drift ratio (RDRcollapse)
and interstory drift ratio (IDRcollapse) at collapse are tabulated. These are the drifts recorded from
the highest earthquake intensity just prior to sidesway collapse of the structure.

The

corresponding collapse fragility curves are shown in Figure 5, where, for comparison purposes,
the intensity on the horizontal axis is normalized by Sa2/50(T1) for each archetype structure. The
mean annual frequencies of collapse are compared in Figure 6.

Comparative Assessment of Non-Ductile and Ductile RC Frames


As summarized in Tables 5 and 6, the analyses show that nonductile RC frames have
collapse margins ranging from 0.49 to 0.85, which indicate that the median collapse intensities
are less than the Sa2/50(T1) intensities. In contrast, the ductile RC frames have collapse margin
ratios that are approximately three times greater, ranging from 1.84 to 3.07. The non-ductile RC
frame buildings have calculated mean annual frequencies of collapse between 47 and 135 x 10-4,
corresponding to a 21 to 50% probability of collapse in 50 years at the high seismic California
site. For ductile RC frame buildings, the mean annual frequencies of collapse range from 1 to 6

15

x 10-4, corresponding to probabilities of collapse in 50 years between 0.5 and 2.9%. In general,
the collapse margins are larger for the space frames as compared to the perimeter frames, which
is largely due to higher overstrength in the space frames where tributary gravity loads are more
significant in design.
Mean annual frequencies of collapse (collapse rate) are compared further in Figure 6, which
illustrates the large difference in collapse risk between the ductile (2003) and nonductile (1967)
frame buildings. Non-ductile RC perimeter frames are approximately 20 times more likely to
collapse compared with ductile RC perimeter frames, and the non-ductile RC space frames are
60 times more likely to collapse than comparable ductile RC space frames. The plot in Figure 6
also illustrates how the collapse risk can vary significantly within each set of buildings where,
for example, there is a difference of about 3 times between the estimated risk metrics for the 2story and 8-story non-ductile space frames. To examine possible reasons for the differences in
collapse metrics, the mean annual frequencies of collapse are plotted against the static
overstrength for each frame in Figure 7. For both the non-ductile and ductile RC frames, the
mean annual frequency of collapse tends to increase with a reduction in static overstrength, but
the trend is not as well-defined for the non-ductile frames where the variation in mean annual
frequencies is much larger than for the ductile frames. Since the data set is small and the
variations reflect the peculiarities of each archetype design, one cannot draw firm conclusions
from these comparisons except to note the larger scatter for the non-ductile frames. Further
discussion of differences in the behavior of non-ductile frames is included below. Overall, the
improvement in predicted collapse capacity, ductility and consistency between the 1967 designs
and the 2003 designs reflects the aggregate effect of the design and detailing changes made in
seismic design codes for reinforced concrete in the intervening years.

16

Collapse performance assessments further reflect how the collapse behavior is related to the
overall system ductility. Referring to Tables 5 and 6, the ductile RC frames experience roof drift
ratios just before collapse (RDRcollapse) of 1.6% to 7.5%. These values are approximately 2.4
times larger than the comparable non-ductile RC frames, which have RDRcollapse = 0.6% to 2.8%.
The associated interstory drift ratios preceding collapse (IDRcollapse) are 1.8 times larger in the
ductile RC frame buildings. These differences result in part from increases in member-level
deformation capacity associated with improved detailing requirements as shown in Tables 3 and
4. The increase in RDRcollapse between the 1967 and 2003 designs also demonstrates that the
required strong column-weak beam ratio in the modern buildings helps to spread damage among
a larger number of stories, increasing energy dissipated in the structure and improving system
level ductility. Collapse mechanisms for the 8- and 12-story archetype frames are compared in
Figure 8. Regardless of height, the non-ductile RC frame buildings typically fail in the 1st and
2nd stories only, whereas the ductile modern RC frames experience more distributed yielding and
collapse mechanisms.

Variation in Collapse Performance of Non-Ductile RC Frames


Design differences among the set of archetypical non-ductile RC frames lead to variation in
predicted collapse performance, indicating the range of possible performance of non-ductile RC
frames with regular building geometries and design features. Tall perimeter frame structures and
short space frame structures exhibit the worst collapse performance. The collapse safety of
perimeter frame structures worsens with height, since the lateral load increasingly dominates the
design with increase in building height, tending to decrease lateral overstrength. In contrast,
gravity loads are more important in design of space frames and shorter buildings, indirectly

17

providing additional lateral overstrength. Thus, the cumulative effects of relatively smaller
normalized base shear capacity and stiffness in the taller perimeter frames makes them more
sensitive to P- effects. Space frames are shown to be more likely than perimeter frame
structures to experience the non-simulated failure mode, loss of gravity-load bearing capacity in
a column due to shear failure, because of the much higher vertical loads in these columns.
However, differences in performance between perimeter and space frame structures are
exaggerated in this study because simulation models for perimeter frame systems neglect the
contribution of gravity system elements (i.e., flat slab and interior gravity columns) to lateral
strength and stiffness. Haselton et al. (2008) found that including the gravity system in collapse
assessment increased the median collapse capacity of a 4-story modern RC frame by
approximately 10%.
In addition to the variation in height and framing system already considered, non-ductile RC
frame structures with superior levels of detailing, compared to typical (code-minimum)
structures, were also evaluated as part of this study. By the 1960s, engineers had begun to
recognize the importance of seismic detailing and, in some cases, chose to provide more
transverse reinforcement and better anchorage of reinforcement in beams, columns or joints.
These ideas were described in the landmark report by Blume, Newmark and Corning (1961) and
implemented in what engineer Henry Degenkolb referred to as California practice (Degenkolb
1994). The effect of better-than-average detailing is investigated by considering two design
variants. In the first, it is assumed that column stirrups are continued through the joint (designs
4S+ and 12S+ in Table 1. The provision of transverse reinforcement in joints is now required
standard practice, but was not specified in the 1967 UBC. In nonlinear analysis models, it is
assumed that this additional joint transverse reinforcement results in a 20% increase in strength

18

of the joint shear panel (based on Mitra and Lowes, 2007). The second design variant assumes
that spacing of stirrups in beams and columns is reduced by 33%, representing a 50% increase in
the provided amount of transverse reinforcement (design variants 4S* and 12S* in Table 1). This
change increases the predicted plastic and post-capping rotation capacities of RC elements and
decreases cyclic strength and stiffness deterioration, as shown in Table 3.
The assessed collapse performance of the better-detailed archetype structures is reported in
Table 5, and compared to the results obtained for the 4 and 12-story archetype RC space frames.
The provision of transverse reinforcement in the joints increases the collapse margin of the 4story frame by 11%, whereas it increases the collapse margin of the 12-story frame by 57%. The
increase is large in the 12-story frame because the increase in joint strength is sufficient to move
the damage out of the joints and into the more ductile beams and columns. In addition to
increasing the ductility locally, this change causes the damage to distribute more over the height
of the structure (as indicated by the increase in RDRcollapse between 12S and 12S+ in Table 5).
The increase in the amount of transverse reinforcement in beams and columns leads to a 33%
increase in the collapse margin of the 4-story frame and a 12% increase for the 12-story frame.
The better detailing in beams and columns (improved confinement) produces a modest increase
in the sidesway collapse capacity by increasing increased element deformation capacity and
eliminating the likelihood of experiencing column collapse due to shear failure (the nonsimulated collapse mode). Still, the analysis of design and detailing decisions show that large
improvements are needed to obtain a significant increase in collapse capacity.

CONCLUSIONS

19

This study evaluates the collapse safety of RC frames in seismic areas by designing, simulating
and assessing two sets of archetype RC frame structures, including modern (ductile) and older
(non-ductile) RC frame buildings. Archetype structures vary in height from 2 to 12 stories and
framing system (perimeter and space frames) and are designed according to the 1967 UBC and
2003 IBC building code provisions to be representative of seismically-designed moment frames
from those eras. Nonlinear models are implemented to capture differences in performance
associated with design and detailing characteristics.
As expected, the ductile (2003) RC frames demonstrate markedly superior seismic collapse
performance for all heights and framing systems when compared to the non-ductile (1967) RC
frames. Modern RC frame structures are able to withstand higher intensity ground motions and
are capable of undergoing more significant deformations before collapse. Collapse margin ratios
for the ductile RC frames are approximately three times larger than those of non-ductile frames.
In terms of the mean annual frequency of collapse, non-ductile RC frame structures have a 40
times higher risk of collapse at a typical California high seismic site.

These comparative

collapse assessments of modern code-conforming systems and existing non-ductile systems


provide a basis to gauge the relative level of safety of seismically vulnerable older structures.
Improvements in seismic performance in RC frames over the last 40 years are the result of
both better detailing in individual members reduced spacing of transverse ties, use of closed
seismic hooks, transverse ties in joints and system-level design requirements strong columnweak beam ratios and other capacity design provisions. Reinforcement detailing in beams,
columns and joints in modern RC frames improves element deformation capacity and reduces
strength and stiffness deterioration as the structure deforms. Capacity design prevents column
shear failure and joint shear deterioration and promotes yielding in beams, spreading damage and

20

energy dissipation more over the height of the structure in the ductile RC frames. These
improvements in component and system-level performance lead to the differences in collapse
safety quantified in this study.
Collapse performance assessments also serve to illustrate the variability in seismic

performance expected for non-ductile RC frame structures. Among the regular set of structures
evaluated, tall perimeter non-ductile RC frame structures are most susceptible to sidesway
collapse due to their low lateral overstrength and flexibility. Space frame structures, which have
much higher axial load levels in columns, may experience column shear failure and subsequent
loss of column load-bearing ability, potentially leading to progressive structural collapse. Other
highly irregular structures or those with design or construction flaws, not considered here, may
have even worse seismic performance. Modest detailing improvements in beams, columns and
joints, such as those that might have been employed in California design practice in the 1960s,
improve the seismic performance of non-ductile RC frames in some cases, but still fall
significantly short of modern code levels.
A variety of approximations are made in generating these results, in characterizing future
ground motions, in identifying a set of representative RC frame structures and in developing
simplified analysis models for RC frame structures. Despite the detailed analyses performed
here, there are still contributors to the collapse resistance of the structure for which we do not
account. For example, the additional lateral resistance provided by flat-plate gravity systems in
perimeter lateral-resistance framing systems is not included in simulations. Moreover, the
archetype designs do not consider enhancements that engineers of older buildings may have
implemented to improve on code-minimum design requirements. Conversely, there are other
failure modes anchorage failure in lap-splices, overturning effects on columns, punching shear

21

failure in slab-column joints that are neglected from the analysis. Conclusions may vary for
different California sites, depending on seismic hazard characteristics. The accuracy of the
collapse performance evaluations generated is a function of the simplifying approximations
(Krawinkler and Zareian 2007). The large uncertainties in structural collapse modeling and
ground motion characteristics contribute to the relatively high (absolute) measures of collapse
risk reported in this study. However, relative values and comparisons of patterns in collapse
capacity provide important metrics for seismic safety.
The collapse performance assessments conducted here confirm the expectation that nonductile RC structures are potentially vulnerable and, for the first time, systematically quantifies
differences in safety for modern (ductile) and older (non-ductile) RC structures. Metrics for
seismic collapse risk of non-ductile RC frame structures obtained in this study can be used to
inform the discussion about seismic safety in California. The state already has regulations
requiring evaluation and, in some municipalities, retrofit or demolition of unreinforced masonry
buildings. Data from this study may serve to evaluate the effectiveness of policies for mitigating
the potential collapse hazard of non-ductile RC frame structures through cost-benefit assessment
of retrofit or replacement and to identify the most vulnerable structures. Still missing in the
public policy process is a transparent discussion of desired safety goals or acceptable risks,
which can be informed by performance-based engineering assessments using technologies such
as those described in this study.

ACKNOWLEDGMENTS

22

This research has been supported by PEER Center through the Earthquake Engineering
Research Centers Program of the NSF (under award number EEC-9701568) and the Applied
Technology Council through the Federal Emergency Management Agencys P695 (ATC-63)
project. Additional support for the first author was provided by a Stanford Graduate Fellowship
and the NSF Graduate Research Fellowship Program. The authors appreciate the participation
and contributions of collaborators from PEER, ATC and FEMA in this study, and would like to
acknowledge the constructive input of Jack Baker, Brian Dean, Charles Kircher, Helmut
Krawinkler, Eduardo Miranda and C. Marc Ramirez and three anonymous reviewers.

REFERENCES
ACI. (2005). Building Code Requirements for Structural Concrete (ACI 318). American Concrete Institute,
Farmington Hills, MI.
ASCE. (2002). Minimum Design Loads for Buildings and Other Structures (ASCE 7-02). American Society of Civil
Engineers, Reston, VA.
Aslani, H., Miranda, E.M. (2005). "Probabilistic Earthquake Loss Estimation and Loss Disaggregation in
Buildings." Blume TR 157, Stanford University.
Blume, J.A., Newmark, N.M., Corning, L.H (1961), Design of multistory reinforced concrete buildings for
earthquake motions, Portland Cement Association, Chicago, 318 pgs.
Degenkolb, H. (1994). Connections (Interview by Stanley Scott). EERI Oral History Series, EERI, Oakland.
Deierlein, G. G. (2004). "Overview of a Comprehensive Framework for Earthquake Performance Assessment."
Performance-Based Seismic Design Concepts and Implementation, PEER Report 2004/05.
Elwood, K. (2004). "Modeling failures in existing reinforced concrete columns." Canadian Journal of Civil
Engineering, 31(5), 846-859.
Elwood, K. and J. Moehle (2005). "Drift Capacity of Reinforced Concrete Columns with Light Transverse
Reinforcement." Earthquake Spectra, 21(1), 71-89.

23

Goulet, C. A., Haselton, C. B., Mitrani-Reiser, J., Beck, J. L., Deierlein, G. G., Porter, K. A., and Stewart, J. P.
(2007). "Evaluation of the Seismic Performance of a Code-Conforming Reinforced-Concrete Frame Building from Seismic Hazard to Collapse Safety and Economic Losses." Earthquake Engineering and Structural
Dynamics, 36(13), 1973-1997.
Haselton, C. B., Baker, J. W., Liel, A. B., and Deierlein, G. G. (2010b). "Accounting for Ground Motion Spectral
Shape Characteristics in Structural Collapse Assessment through an Adjustment for Epsilon." Journal of
Structural Engineering, In Press.
Haselton, C. B., Mitrani-Reiser, J., Goulet, C., Deierlein, G. G., Beck, J., Porter, K. A., Stewart, J., and Taciroglu, E.
(2008). "An Assessment to Benchmark the Seismic Performance of a Code-Conforming Reinforced-Concrete
Moment-Frame Building." Pacific Earthquake Engineering Research Center 2007/12, University of California
at Berkeley.
Haselton, C. B., and Deierlein, G. G. (2007). "Assessing Seismic Collapse Safety of Modern Reinforced Concrete
Frame Buildings." Blume Center Technical Report No. 156.
Haselton, C. B., Liel, A. B., Deierlein, G. G., Dean, B.S. and Chou, J. (2010a). "Seismic Collapse Safety of
Reinforced Concrete Buildings: I. Assessment of Ductile Moment Frames." Journal of Structural Engineering,
(companion paper).
Haselton, C., Liel, A., Taylor Lange, S., and Deierlein, G. G. (2008). "Beam-Column Element Model Calibrated for
Predicting Flexural Response Leading to Global Collapse of RC Frame Buildings." Pacific Earthquake
Engineering Research Center 2007/03, University of California at Berkeley.
Ibarra, L. F., Medina, R. A., and Krawinkler, H. (2005). "Hysteretic Models that Incorporate Strength and Stiffness
Deterioration." Earthquake Engineering and Structural Dynamics, 34 1489-1511.
ICBO. (1967). Uniform Building Code. Pasadena, CA.
ICC. (2003). International Building Code. Falls Church, VA.
Krawinkler, H., and Zareian, F. (2007). "Prediction of Collapse - How Realistic and Practical Is It- And What Can
We Learn From It." Structural Design of Tall Buildings, 15 633-653.
Liel, A. B., and Deierlein, G. G. (2008). "Assessing the Collapse Risk of Californias Existing Reinforced Concrete
Frame Structures: Metrics for Seismic Safety Decisions." Blume Center Technical Report No. 166.

24

Liel, A. B., Haselton, C. B., Deierlein, G. G., and Baker, J. W. (2009). "Incorporating Modeling Uncertainties in the
Assessment of Seismic Collapse Risk of Buildings." Structural Safety, 31(2), 197-211.
Lowes, L. N., and Altoontash, A. (2003). "Modeling of Reinforced-Concrete Beam-Column Joints Subjected to
Cyclic Loading." Journal of Structural Engineering, 129(12), 1686-1697.
Mitra, N., and Lowes, L. N. (2007). "Evaluation, calibration and verification of a reinforced concrete beam-column
joint model." Journal of Structural Engineering, 133(1), 105-120.
Moehle, J., Lehman, D., and Lowes, L. (2006). "Beam-Column Connections." New Information on the Seismic
Performance of Existing Buildings, EERI Technical Seminar.
Open System for Earthquake Engineering Simulation (OpenSEES) (2009). Pacific Earthquake Engineering
Research Center, University of California, Berkeley, http://opensees.berkeley.edu/.
Pantelides, C. P., Clyde, C., and Reaveley, L. D. (2002). "Performance-Based Evaluation of Reinforced Concrete
Building Exterior Joints for Seismic Excitation." Earthquake Spectra, 18(3), 449-480.

25

List of Tables
Table 1. Design characteristics of archetype non-ductile RC frames.
Table 2. Design characteristics of archetype ductile RC frames.
Table 3. Representative modeling parameters in archetype non-ductile RC frame structures. Column model
properties are based on Haselton et al. (2008).
Table 4. Representative modeling parameters in archetype ductile RC frame structures .
Table 5. Results of collapse performance assessment for archetype non-ductile RC frame structures.
Table 6. Results of collapse performance assessment for archetype ductile RC frame structures.

26

List of Figures
Figure 1. Schematic of the RC frame structural analysis model.
Figure 2. Properties of inelastic springs used to model ductile and non-ductile RC columns in the 1st story of a
typical 4-story space frame, illustrating differences in (a) monotonic and (b) cyclic behavior.
Figure 3. Pushover analysis of (a) ductile and non-ductile archetype 12-story RC perimeter frames and (b) showing
distribution of interstory drifts at the end of the analysis.
Figure 4. IDA results for the 8-story space frame structure, illustrating (a) the effect of vertical collapse mode on
collapse capacity for a selected earthquake record in incremental dynamic analysis and (b) comparison of collapse
fragilities for sidesway collapse only and combined sidesway and vertical collapse.
Figure 5. Collapse fragility curves for archetype RC (a) perimeter frames and (b) space frames.
Figure 6. Comparison of assessed collapse risk for archetype non-ductile and ductile RC frames, in terms of mean
annual frequency of collapse.
Figure 7. Relationship between static overstrength and mean annual frequency of collapse for (a) non-ductile and (b)
ductile RC frames. Space frames are indicated by diamond markers.
Figure 8 . Expected sidesway collapse mechanisms in selected archetype space frame structures.

27

Das könnte Ihnen auch gefallen