ATC-114 Next-Generation Hysteretic Relationships

for Performance-based Modeling and Analysis

Ronald O. Hamburger, SE
Simpson Gumpertz & Heger Inc.
San Francisco, CA

Greg Deierlein,
Stanford University
Palo Alto, CA

Dawn Lehman, Laura Lowes

University of Washington
Seattle, WA

Benson Shing
University of California
San Diego, CA

John van de Lindt

Colorado State University
Fort Collins, CO

Dimitrios Lignos
McGill University
Montreal, Canada

Ayse Hortacsu
Applied Technology Council
Redwood City, CA

1
 demand deformations, considering the nonlinear behavior of
Abstract the overall structural system.

Nonlinear analysis has become an increasingly useful and The ATC-33 project team recognized that nonlinear dynamic
important tool for evaluation, upgrade and design of structures analysis was not yet practical as a design tool, given the
for seismic resistance. However, despite steady improvements previously described limitations. Therefore, the team
in analysis capability, most practice remains anchored to developed the FEMA 273/274 guidelines around the concept
guidelines developed more than 20 years ago. Under its ATC- of nonlinear static, or pushover, analysis. Recognizing that
114 project, the Applied Technology Council is developing even pushover analysis would be a large step for the practicing
updated hysteretic envelope models for use in seismic analysis engineer of the time, many of whom still relied on equivalent
of new and existing buildings. The intent of this project is to lateral force analysis in their design of new structures, the
support the development of updated building code criteria ATC-33 project included linear analysis procedures for
contained in such standards as ASCE 7 and ASCE 41. Project structures with good regularity and limited ductility demand.
support is provided by the National Institute of Standards and The FEMA 273/274 guidelines and commentary permitted the
Technology. use of nonlinear dynamic analysis, recognizing that technique
might someday become practical, but included many warnings
Introduction to the engineer about the potential pitfalls associated with this
technique and the need for both caution and external, expert
Although researchers have been performing nonlinear analysis review.
for many years, the first direct, practical applications of this
technique in seismic design occurred in the 1980s, as engineers Following publication of FEMA 273/274 and the closely
began to design base isolated structures. In these early related ATC-40 (ATC, 1996) methodology, structural
applications, nonlinearity was limited to the base isolation engineering applications developers recognized the power of
system itself, with the superstructure modeled elastically, and the performance-based design approach embodied in these
often as a simple stick system having appropriate mass and documents and began to incorporate nonlinear analysis, both
modal properties. There are a number of reasons nonlinear static and dynamic, into their suites of software. Importantly,
analysis was not commonly used by engineers at this time first RAM PERFORM, then CSI ETABs and SAP
including limited availability of software to perform such incorporated the FEMA 273/274 hysteretic backbone data and
analysis and the fact that typical computers available to nonlinear acceptance criteria directly into their element
engineers did not have the required speed or processing libraries. Armed with these powerful new tools, structural
capability necessary to make this technique practical. Another engineers rapidly embraced nonlinear analysis as a design tool,
major impediment was the lack of consensus guidance or code- not only for evaluation and retrofit of existing structures, but
specified criteria on how to use nonlinear analysis in design. also design of new structures. In 2002 the American Society
The first such guidance was developed by the SEAOC of Civil Engineers (ASCE) incorporated the FEMA 273
Seismology Committee in an Appendix to its Blue Book nonlinear dynamic analysis guidelines into its ASCE 7
(SEAOC, 1990). This guidance, adopted in the 1991 Uniform Standard (ASCE, 2002). Then, in 2006, ASCE updated FEMA
Building Code (ICBO, 1991) addressed only analysis of base 273/274 and published it as the ASCE 41 Standard (ASCE,
isolated systems, as described above. 2006).

More general nonlinear analysis guidelines for use in seismic Today, many engineers use nonlinear dynamic analysis not
analysis and design were first developed as part of the ATC- only for evaluation and retrofit of existing buildings but also
33 project and appeared in the FEMA 273/274 Guidelines and for design of major new structures and it has become the
Commentary (ATC, 1997). The ATC-33 project team preferred technique for design of high-rise buildings in regions
recognized that a key problem faced in the design of seismic of high seismic risk. However, despite the extensive
upgrades is to assure deformation compatibility between the laboratory investigations into nonlinear behavior of wood,
existing building elements, often having limited ductility, and masonry, concrete and steel elements that have occurred in the
new retrofit elements. It was clear that the linear approach past 20 years, with few exceptions, engineers performing
used for the design of new buildings, where deformation nonlinear dynamic analysis continue to use the basic hysteretic
compatibility could be assured by detailing requirements, relationships and acceptance criteria derived from those
could not be relied upon for existing buildings, as detailing of relationships, first developed by the ATC-33 project.
the existing structure could not be controlled or modified.
Instead analysis conducted in support of retrofit design would The National Institute of Standards and Technology (NIST), in
have to directly account for the nonlinear deformation capacity its role as lead NEHRP agency identified the need to support
of the individual elements, old and new, and also compute further development of performance-based design as a priority.

2
The ATC-114 project is one of a series of projects sponsored backbones were intended to account for cyclic strength and
by NIST to further this goal. The specific objectives of the stiffness degradation in an approximate manner, recognizing
ATC-114 project are to provide updated recommendations on that static analysis could never accurately capture these effects.
hysteretic characterization of nonlinear structural behavior for Figure 2 illustrates the construction of these second-cycle
use in performance-based seismic design. Like FEMA backbones from test data as well as the 1st cycle backbone later
273/274, it addresses all four major structural materials. The adopted by ASCE 41.
eventual ATC-114 report will be informational in nature and
is primarily intended for use by members of the ASCE 7 and
ASCE 41 committees as well as other groups developing
seismic analysis guidelines and standards.

ASCE 41 Hysteretic Backbones

The nonlinear analysis procedures first developed in FEMA

273/274 and carried forward into the ASCE 7 and ASCE 41
standards parse structural elements into two primary types:
deformation-controlled and force-controlled. Force-controlled
elements, by definition, have no appreciable ductility and are
represented in analysis as elastic elements. Acceptable
behavior of force-controlled elements is judged based on the
amount of force-demand computed from analysis, and the
margin provided against exceedance of a lower-bound
estimate of strength. Figure 2 Development of cyclic backbones

Deformation-controlled elements have non-negligible FEMA 273/274 never intended the use of these backbones in
ductility. For nonlinear static analysis, ASCE 41 specifies that nonlinear dynamic analysis and noted that hysteretic models
element nonlinear modeling conform to one of two idealized for such analysis should be derived to match appropriate cyclic
relationships, the most general of which is shown in Figure 1. test data. However, lacking such data, engineers often applied
In the figure, Q represents a generalized force quantity and  the FEMA 273/274 nonlinear representations, many of which
a generalized deformation. were developed without specific data using judgment, for both
static and dynamic analysis.

In 1997 FEMA entered into a cooperative effort with the

American Society of Civil Engineers to develop FEMA
273/274 into a pre-standard, FEMA 356 (ASCE, 2000). FEMA
356 carried forward nearly all of the FEMA 273/274 hysteretic
models with the exception that modeling recommendations for
steel moment frames were updated using data available from
the recently completed FEMA/SAC program to address
seismic hazards in welded steel moment frames. ASCE 41,
published in 2006, retained nearly all of the hysteretic models
contained in FEMA 356, however, supplement 1 to the new
standard, published in 2007 contained updates to the models
for reinforced concrete columns. The 2013 edition of the
standard (ASCE, 2013) revised the definition of the standard
Figure 1 ASCE 41 Nonlinear Representation hysteretic model from a second cycle to first cycle backbone.
As illustrated in Figure 2, the first cycle backbone typically
The nonlinear behavior illustrated in Figure 1 was intended to exhibits less degradation than does the second cycle backbone,
represent, in an approximate manner, a second cycle backbone and as a result provides both greater deformation capacity and
of data obtained from typical ramped cyclic component tests. strength retention at large ductility demand. The argument
The second cycle backbones, constructed as a series of secants used to support this change is that structures responding to
drawn on hysteretic plots by connecting the crossings of the ground motions with the low probability of exceedance
second cycle of loading to a given deformation increment with associated with structural collapse often experience only a few
the first cycle for the next increment of loading. Second cycle large nonlinear excursions, and that the many cycles

3
represented by second cycle curves over-stated the degradation Loading Effect on Hysteretic Response
experienced by such structures. Regardless of this definition
change, the values of the control points (A, B, C, D, and E in Starting in the 1970s most laboratory testing of components
Figure 1) specified by the standard for most elements did not conducted to characterize likely seismic behavior used ramped
change. cyclic protocols in which the specimen was subjected to
repeated fully reversed cycles of motion to increasing
In addition to defining the force-deformation relations used to deformation amplitude such as that illustrated in Figure 4.
model nonlinear behavior in analysis the ASCE 41 backbones Protocols such as that contained in ATC-24 (ATC, 1992) used
also form the basis for acceptance criteria for deformation so-called rain flow analysis to attempt to balance both the
controlled components. Figure 3 illustrates the relationship of hysteretic energy and number and size of inelastic excursions
nonlinear deformation acceptance criteria to points on the contained in such protocols against that likely to be
backbone for primary and secondary components. Acceptance experienced by a structural component in a severe earthquake,
criteria used for linear analysis are generally take as 75% of however, response plots for elements obtained from nonlinear
those permitted for nonlinear analysis, accounting in analysis seldom look much like these testing protocols. Often,
approximate manner for the greater inherent uncertainty as illustrated in Figure 5, response to real earthquakes entails
associated with demands predicted by linear analysis. relatively few large cycles of motion, early in the record,
followed by a large number of smaller cycles, often about a
permanently displaced position. Hysteretic backbones derived
from ramped cyclic protocols may not accurately represent
behavior in real earthquakes.

Figure 3 – Deformation acceptance criteria

The ASCE 41 hysteretic models are largely based on laboratory

testing that was available in the mid-1990s when the ATC-33
project was conducted. For many of the component types
covered by the ASCE 41 standard there was no such data
available and therefore, the ATC-33 project team relied on
judgement to develop these models. In the time since a large
body of additional laboratory testing has become available.
One important purpose of the ATC-114 project is to provide Figure 4 – Ramped cyclic testing protocol
improved recommendations for hysteretic models based on
this updated data. A second purpose is to update the basic
ASCE 41 model based on the improved current understanding
of nonlinear analysis and behavior now available.

The ASCE 41 backbones, even when based on laboratory data,

are lacking in two primary ways. First, the backbones are
inherently tied to the cyclic protocol on which they are based.
Second, the backbones, though intended to represent best
estimate, or median relationships provide no indication of
potential variability or uncertainty. Figure 5 – Displacement History, 1 second structure,
LGP000, 1989 Loma Prieta earthquake

The hysteretic response of some structural elements is

significantly affected by the loading protocol used. Figure 6,
taken from Suzuki and Lignos (2015) illustrates the difference

4
in nonlinear response for a steel column loaded using the
typical ATC-24 protocol, a monotonic loading and an
alternative cyclic loading termed a collapse protocol that
entails relatively few cycles and an increasing directional bias
associated with onset of P-delta instability.

Figure 7 ATC 114 hysteretic models

Key parameters on the monotonic backbone include the yield

point Qy, y, which is typically protocol independent; the peak
strength point Qmax, m; the post-capping strength point QR,
Figure 6 – Response of steel beam to alternate pc; the point of loss of lateral force resistance QR, ult; and the
loading protocols loss of gravity (vertical) load resistance, LVCC. Control points
for the cyclic backbone have similar designations with a prime
As can be seen in the figure, the monotonic response is (‘) denotation. It should be noted that for a number of reasons,
characterized by increased strength, delayed onset of few laboratory tests explore specimen behavior out to LVCC or
degradation and substantially larger ultimate deformation even ult. Therefore, each of the ATC-114 recommended
capacity than either of the cyclic protocols. Response to the hysteretic models includes a recommended limit on the valid
so-called “collapse protocol” though it incorporates substantial range of modeling, i.e. the deformation level at which behavior
cyclic degradation, demonstrates substantially increased is not known simply because no test data is available to show
deformation capacity than does response to the ATC-24 behavior at larger deformation levels.
protocol loading. This response is consistent with findings
from recent shake table collapse experiments on steel frame For both cyclic and monotonic backbones, two paths are
buildings (Lignos et al. 2011; Suita et al. 2008). shown beyond the peak load strength points, Qmax, Q’max. One
path includes a residual strength plateau, QR, Q’R, and the other
Since the response induced in structures by each earthquake does not. The path with residual strength is used only for those
ground motion is unique, no one loading protocol, nor any component types that have a specific post-initial failure
hysteretic model based on a single protocol can accurately behavior that corresponds to the residual strength. An example
capture a structure’s likely response to the wide range of is a simple shear plate connection attaching a beam to a column
motions it may be subject to. Ideally, hysteretic behaviors and supporting a slab. When loaded with the slab in
incorporated in structural elements used for nonlinear analysis compression initial behavior will be a composite of the steel
should be adaptive, and be able to accurately reproduce the clip plate and connecting bolts in tension and the slab in
response of elements to any testing protocol applied. A few compression. An initial failure will consist of crushing of the
such adaptive hysteretic models presently exist, though none slab, resulting in loss of the composite action with behavior
are compatible with software commonly employed in design dominated by the flexural capacity of the shear plate and bolts.
offices. However, the ATC-114 project team anticipates that This is illustrated in Figure 8, taken from Liu and Astaneh
such elements will become increasingly available. To (2008) in which the positive quadrant is characterized by initial
facilitate the future development of such models, the ATC-114 composite behavior, lost after crushing while the negative
team elected to present both cyclic and monotonic hysteretic quadrant is limited to the flexural behavior of the shear plate.
models, for each component type, where sufficient monotonic In many other component types, no such physical behavior,
data is available to permit this. The resulting data is presented associated with development of a residual strength exists, and
in the format illustrated in Figure 7. In the figure, the blue the apparent residual strength observed in plots of test data is
curve represents monotonic response while the red curve really a consequence of the large increments in loading taken
response to a typical cyclic protocol, such as that of ATC-24. at large deformations. The steep decline found in ASCE 41
backbones for many component types is not believed to be real
for most component types and can result in analytical

5
instability. Therefore, the more gradual degrading phase is More recently, design procedures incorporating nonlinear
adopted by ATC 114 for many such component types. analysis have used more rigorous statistical methods to achieve
desired reliability. The design procedures recommended by
FEMA 350 (SAC, 2000) included rigorous incorporation of
uncertainties in demand prediction, element capacity and
global structural capacity in a demand and resistance factor
format. This procedure parsed uncertainty into epistemic
(uncertain, or reducible) and aleatory (random) parts to enable
targeting of both FEMA 350 a probability of failure and a
confidence level associated with achieving that failure. Later,
the FEMA P695 (ATC, 2009) procedure simplified this
computation by combining aleatory and epistemic
Figure 8 Hysteretic behavior of steel clip plate uncertainties into a single quantity and establishing that the
connection with composite slab (Liu and Astaneh, reliability goal for ordinary occupancy structures is to provide
2008) less than a 10% probability of collapse, given MCE shaking.
The Pacific Earthquake Engineering Research Center applied
Since monotonic backbones are always going to be an upper a load factor approach in their procedures for performance-
bound on an element’s strength and deformation capacity they based design of tall buildings (PEER, 2009) to account for
should never be used directly for seismic analysis. Rather the uncertainties in demand prediction, while relying on the use of
intent is that these monotonic backbones can be used by lower bound values for acceptance criteria, in order to achieve
element developers to calibrate their adaptive models. the target 10% probability of collapse suggested by FEMA
Adaptive hysteretic models are the preferred approach for P695. ASCE 7-16 (ASCE, 2016) expanded the PEER
nonlinear dynamic analysis as they have the potential to most methodology to address acceptance criteria for both force and
accurately portray true response, regardless of the particular deformation-controlled components considering uncertainty in
ground motion’s loading path. However adaptive models demand prediction and capacity.
should not be used for nonlinear static analysis as it is
impossible for static analyses to replicate any specific loading To facilitate the development of appropriate demand and
history, and therefore allow an adaptive model to capture resistance factors in future editions of ASCE 7, ASCE 41 and
correct stiffness, strength or deformation capacity. Rather, the other design procedures, the ATC-114 project elected to
cyclic model should continue to be used for nonlinear static provide coefficients of variation, to represent uncertainty in
analysis. both the strength and deformation values assigned to the
control points in the cyclic and monotonic backbones.
Modeling Uncertainty
Acceptance Criteria
Since the introduction of Load and Resistance Factor design
methods, design procedures for new construction have been As noted earlier, and illustrated in Figure 3, ASCE 41 uses the
formulated to achieve target reliabilities with both load and hysteretic backbones as its basis for acceptance criteria for
resistance factors established considering uncertainties in load deformation controlled behaviors for both nonlinear and linear
intensity, analytical methods and capacities. FEMA 274 procedures. The nonlinear static procedure attempts to
commentary describes an intended low probability, suggested produce mean estimates of demand while the acceptance
as on the order of 10%, that a structure upgraded to meet criteria themselves are mean values. Assuming that the
specific performance objectives would fail to do so when uncertainty distribution around capacity and demand are both
actually subjected to a design event. However, ASCE 41 does represented by lognormal distributions, and assuming
not have design procedures or criteria established to dispersions in demand and capacity respectively on the order
specifically achieve such reliability. Rather, the ASCE 41 of 0.5 and 0.2, there is roughly a 40% chance that any element
procedures handle reliability in a qualitative manner. evaluated as just meeting the acceptance criteria under
Specifically, the backbone control points are intended to be predicted demands would actually exceed the deformation
median values. When definition of the structure’s construction associated with the acceptance criteria if the structure were
obtained from drawings, specifications and field investigation subjected to design earthquake loading. This is not particularly
is limited, a knowledge factor is applied to discount acceptance compatible with the FEMA 274 stated goal of approximately a
criteria. Lower bound values, estimated at 5th percentile, are 90% reliability level, i.e. 10% chance of failure. Fortunately,
used for acceptance criteria for force-controlled behaviors. only the Collapse Prevention performance level in ASCE 41
has real physical meaning- that is, the structure does not
collapse. Most of the acceptance criteria for deformation-

6
controlled behaviors associated with Collapse Prevention in associated with acceptance criteria, for primary and secondary
ASCE 41 do not represent actual loss of gravity load carrying deformation-controlled components. The collapse prevention
capacity, but instead, in some cases, loss of most of the acceptance criteria for deformation-controlled components at
element’s lateral force carrying resistance, and in many cases, LVCC. The blue curve is the monotonic envelope, for which
the end of predictable behavior, simply because testing upon no acceptance criteria are shown because all earthquakes
which the acceptance criteria are based did not proceed beyond include some inelastic cycles and will result in some cyclic
that deformation level. degradation.

The probability of designs conforming to ASCE 41 not

experiencing failure of force-controlled behaviors is similarly
poor. The acceptance criteria for force controlled behaviors in
ASCE 41 are intended to be lower bound values, specifically,
5th percentile. The demands computed from analysis however,
are mean values. Assuming uncertainties in the capacity
equation on the order of 15% and uncertainties in demand
prediction on the order of 40%, it can be seen that force-
controlled components designed with demand to capacity
ratios of unity will have approximately a 25% chance of
failure, given the occurrence of design ground shaking, not a
particularly desirable outcome.

ASCE 7-16 (ASCE, 2016) Chapter 16 recognizes this

unacceptable failure probability inherent in the ASCE 41 Figure 9 Acceptance criteria and hysteretic
approach and applies load and resistance factors to the backbones
acceptance evaluation equations so as to provide the desired
90% reliability for elements having computed demand to In the figure, the green curve represents the envelope of
capacity ratios of unity, and superior reliability for elements adaptive hysteretic response for an element responding to a
with lower design demand to capacity ratios. hypothetical earthquake motion. As will typically be the case,
the adaptive backbone shows less degradation than does the
By providing uncertainty values for the hysteretic backbone cyclic backbone and more than the monotonic envelope.
control points, ATC-114 makes it possible to apply a similar However, the exact shape of these adaptive response envelopes
approach to the ASCE 41 evaluation procedures. However, the will be unique to the ground motion and structure. Further, the
problem remains that most laboratory tests stop at deformation end point for the green adaptive envelope does not represent a
levels below those that actually cause loss of gravity load capacity, but is simply the maximum deformation experienced
carrying capacity, and, as a result, acceptance checks are based in response to the particular ground motion.
on presumed failure at the limit of available laboratory data
and will be more conservative than if actual failure data were Since it is not possible to develop collapse prevention
available. Further, to the extent that a real earthquake loading acceptance criteria based on the adaptive hysteretic models, it
may not result in as much degradation as that experienced in is proposed instead that when adaptive models are used, a run
ramped cyclic laboratory testing, the acceptance criteria will be considered to produce acceptable response as long as the
have greater conservatism. While this may be a reasonable analysis converges, and demands on force-controlled
situation, given that failure is undesirable, the use of components remain within acceptable limits. In order to gain
excessively conservative acceptance criteria will result in confidence that failure will not result when this approach is
retrofit designs that are needlessly costly, and may result in taken, it will be necessary to perform analyses using a
some structures remaining vulnerable because the cost of sufficiently large suite of motions to obtain statistics on the
retrofit was excessive. probability of collapse. For example, ASCE 7-16 requires a
suite of 11 motions so as to gain moderate confidence (75%)
The use of models incorporating elements with adaptive that collapse probability if less than 10%. Larger suites of
hysteretic relationships can reduce part of this conservatism by analyses are necessary to gain higher confidence.
providing demands that have not been biased by excessive
degradation in the response model. However, as illustrated in When models are constructed using elements tuned to cyclic
Figure 9 it is not clear how to develop acceptance criteria based backbones only, or when nonlinear static analysis is
on such adaptive models. In the figure, the standard cyclic performed, acceptance criteria for deformation-controlled
backbone is shown in red, together with the control points behaviors should continue to be based on the backbone control

7
points illustrated in Figure 9, with the exception that value λhd specified by AISC 241-10 for highly ductile
uncertainty should be considered in judging acceptance, as elements.
done in ASCE 7.

The sections below present summary level recommendations

for steel braced frames and moment frames, flexure-governed
concrete walls, reinforced masonry walls and wood walls.
Work on shear-controlled concrete walls and concrete moment
frames is under development and will be included in the final
ATC-114 report.

Steel Concentric Braced Frames

For special concentric braced frames (SCBFs) complying with

AISC 341-10 (AISC, 2010) explicit modeling of the braces and
gussets in a manner that will directly capture cyclic
degradation due to buckling is recommended. To accomplish
this, individual braces should be modeled with a minimum of Figure 11 Comparison of actual and analytically
four elements per brace extending between connections. The predicted hysteresis for adaptive braced frame
geometry of the brace elements should be configured such that model
an out-of-plane off set at brace mid-point with an amplitude of
1/500 of brace length is captured. Beam column joints should
be modeled with rigid end offsets for the brace, beam and
column, and a rotational spring representing the gusset plate
stiffness. Figure 10 shows the recommended modeling
approach at a typical brace-beam-column connection and
Figure 11 illustrates the hysteresis obtained from typical
models of this type.

Figure 12 Limiting deformation of SCBF brace as a

function of section slenderness

For nonlinear static analysis, it is recommended that pairs of

Figure 10 – Model of brace to beam and column
braces (one in tension and one in compression) be represented
connections
by a single element having the simplified backbone illustrated
For braces conforming to the criteria for SCBFs failure and
in Figure 13. In the figure Qave is the average of the brace yield
loss of strength will result from brace fracture due to low-cycle
and compressive strengths, ave is the average of the
fatigue and the large plastic strains that occur at the point of
deformation at initiation of yield and buckling and f is the
plastic hinge formation during buckling. For cyclic loading
residual strength at fracture given by the equation shown in the
protocols, the value of the brace section slenderness ratio,
figure. Deformation at failure is obtained from Figure 12, just
either h/d or b/t, is the best predictive parameter for failure.
as for nonlinear dynamic analysis.
Figure 12 presents the recommended cyclic deformation limit,
measured as frame story drift ratio, as a function of the ratio of
the critical section slenderness parameter, λ to the limiting

8
Figure 13 SCBF cyclic backbone showing residual
strength Figure 16 Measured and analytically predicted
response of planar wall using three alternative
Flexure-controlled Concrete Walls models

Planar reinforced concrete shear walls controlled by flexure

can be adequately modeled using fiber-type beam-column
elements. Figure 14 illustrates the cross section of such a
beam-column element and Figure 15 the constitutive
properties for the concrete and steel fibers. Figure 16
illustrates a comparison of three different such models with the
cyclic test for the modeled wall. Modeling accuracy can be
improved with the addition of nonlinear shear models such as
that shown in Figure 17.

Figure 14 Cross section of fiber-type beam column

element
Figure 17 – Nonlinear shear model for use with
flexural beam-column wall models

Based on work by Whitman (2015) flexural behavior of

concrete shear walls can be viewed as controlled either by
shear-compressive interaction, buckling compression failure,
or buckling-induced rupture of bars. As illustrated in Figure
18, the controlling mode can be predicted as function of shear
stress and cross-sectional aspect ratio. Limiting plastic hinge
Figure 15 Constitutive models for concrete and rotations of 1.6% and 0.9% are recommended respectively for
steel materials walls controlled by compression buckling or buckling rupture,
and compressive shear interaction failure.

9
 Figure 20 illustrates the recommended cyclic backbone for
reinforced masonry walls. Key control points are the point at
which 80% of peak strength is achieved upon initial loading,
the point of peak strength, and points at which strength
degrades to 75% and 50% of peak values, the latter
representing the end of the valid range of modeling.
Displacements at these key control points are presented as a
function of axial compression ratio and reinforcing ratio.

Figure 18 Concrete wall flexural failure as a function

of shear stress and cross section aspect ratio

Reinforced Masonry Walls

Reinforced masonry walls should be modeled as assemblages

of beam-column elements calibrated to reflect the excepted Figure 20 Cyclic backbone for flexure-controlled
plastic-hinge length of the wall components. The shear reinforced masonry walls
behavior of a wall component should be represented by a linear
or nonlinear shear spring. Cyclic backbone curves for flexural Figure 21 presents the recommended cyclic backbone for
behavior are presented based on detailed analytical models reinforced masonry walls with shear nonlinearity. Peak shear
calibrated to 21 cyclic tests of walls employing different aspect strength is assumed to occur at a shear drift ratio of 0.5%,
ratios, reinforcement ratios and axial loading. Similarly, cyclic degradation to residual strength at twice that deformation and
backbone curves for shear behavior are presented based on the valid range of modeling is assumed at 2% drift. The
similar analytical modeling calibrated to 16 cyclic tests. These residual strength is a function of the shear reinforcing ratio as
walls were all fully grouted. Monotonic tests are not available given by equation (1). These values are based on test data for
for reinforced masonry elements. full grouted walls. Partially grouted walls will exhibit more
brittle behavior. When nonlinear dynamic analysis is
Depending on the aspect ratio, axial loading and reinforcement performed hysteretic behavior should include significant
ratio, masonry walls dominated by nonlinear flexural response pinching, similar to that indicated in Figure 19 for flexural-
exhibit ductile hysteresis, such as that shown in Figure 19. The controlled walls.
pinching and eventual load degradation exhibited in the
hysteresis are normally caused by masonry crushing, rebar Av
buckling and fracture, or the failure of the lap splices in the Qr  Vns  0.5 f y dv Eq. 1
vertical reinforcement. s

Figure 21 Cyclic backbone for shear-controlled

reinforced masonry walls
Figure 19 Typical flexural hysteresis for reinforced
masonry wall (Sherman, 2011)

10
Wood Shear Walls

Wood shear wall hysteretic response is characterized by a lack Steel Moment Resisting Frames
of a well-defined yield point and continuous softening from
initial loading through achievement of peak resistance The extensive testing of steel frame components of various
followed by degradation. When nonlinear dynamic analysis is types conducted following the 1994 Northridge earthquake
performed the CUREE element shown in Figure 22 is have enabled the development of robust, adaptive models.
recommended, however, any element with significant pinching These models have been used to develop recommended cyclic
and the ability to capture stiffness and strength degradation can and monotonic backbones together with the key control
be used. parameters for a variety of beams, beam-columns and beam
column connections, considering section compactness, lateral
bracing and connection type. As illustrated in Figure 24, steel
beams with RBS (Ricles et al. 2004) subjected to monotonic
loading exhibit substantially reduced degradation and
additional ductility.

Figure 24 – Cyclic and monotonic loading behavior

Figure 22 CUREE (Folz and Filliatrualt, 2001) wood for RBS beam to column connection
wall model
For steel elements, cyclic and monotonic backbones like those
The standard cyclic backbone shown in Figure 23 is shown in Figure 7 are presented. Depending on the steel
recommended. The parameters, F0, Fu, r1, r2, and u,max are structural element, control points on their backbones are a
defined in tabular form dependent on the sheathing material, function of axial load ratio, h/tw, bf/tf, L/r, and L/d. Equation
nailing and wall aspect ratio. 2, below illustrates the form of equations provided to define
these control points, this one presented for a fully-restrained
beam to column connection, without reduced beam section. In
the equation, the parameter cunit is a conversion factor between
metric and English units. Coefficients of variation are
provided for all parameters.

0.1
h
0.3
b  0.7
 L   cunit d 
0.3

 cap, pl  0.07      f 
     Eq. 2
 tw   2t f   d   533 
Importantly, based on recent testing (Suzuki and Lignos 2015;
Elkady and Lignos 2016) permissible ductility for steel
columns with substantial axially load is greatly increased
relative to the values contained in ASCE 41. Further,
recommendation is made to base the axial load ratio on
sustained, gravity load, rather than peak transient load.
Figure 23 Cyclic backbone for wood walls with
various sheathing materials Summary

11
Upon completion, forecast for the spring of 2017, the ATC- ATC. 1992, ‘‘Guidelines for cyclic seismic testing of
114 project will be a valuable resource to code and standards components of steel structures for buildings.’’ Rep. No. ATC-
committees engaged in the development of criteria for using 24, Applied Technology Council, Redwood City, Calif.
nonlinear analysis in design and retrofit, as well as individual
engineers engaged on projects. The completed document will ATC, 1996, Seismic Evaluation and Retrofit of Concrete
provide greatly improved backbone relationships for structures Buildings ATC-40, California Seismic Safety Commission,
of concrete, masonry, steel and wood construction. These Sacramento, CA
backbones will permit improved modeling and also, in many
cases, less restrictive acceptance criteria. In the ASCE 41 ATC, 1997, Guidelines and Commentary for Seismic
standard, where acceptance criteria for linear procedures are Rehabilitation of Buildings, FEMA 273/274, Federal
derived from the backbones this will also provide benefit for Emergency Management Agency, Washington D.C.
the linear procedures.
ATC, 2009, Quantification of Building Seismic Performance
For steel frame buildings, adaptive models that are capable of Factors, Report No. FEMA P-695, Federal Emergency
representing hysteretic response without calibration to specific Management Agency, Washington, D.C.
loading protocols are available. As these models find their way
into software commonly used in the design office, and as other Elkady, A. and Lignos, D.G., 2016, “Dynamic stability of deep
similar models are developed for other structural systems, and slender wide-flange steel columns – full scale
nonlinear dynamic analysis will become more attractive on experiments,” Proceedings, Annual Stability Conference,
design projects as the behavior of structures subjected to Structural Stability Research Council (SSRC), Orlando,
extreme loadings will be better predicted, allowing Florida, USA.
liberalization of the acceptance criteria presently available.
Folz, B. and Filiatrault, A. (2001) “Cyclic analysis of wood
Though not specifically addressed herein, two companion shear walls.” Journal of Structural Engineering, 127(4), 433-
projects being conducted under the ATC-114 contract address 441
detailed modeling and analysis criteria for structural steel and
reinforced concrete moment frames. These companion ICBO, 1991, Uniform Building Code, International
publications will be immediately useful to engineers Conference of Building Officials, Whittier, California.
performing nonlinear analysis of these structures in design and
evaluation projects. Liu, J. and Astaneh, A., 2000, “Cyclic testing of simple
connections including effects of slab,” Journal of Structural
References Engineering, ASCE, Vol. 126, No.1, pp. 32-39

AISC, 2010, Seismic Provisions for Steel Structures, AISC Lignos, D.G., Krawinkler, H. and Whittaker, A.S., 2011,
341-10, American Institute of Steel Construction, Chicago, IL “Prediction and validation of sidesway collapse of two scale
models of a 4-story steel moment frame.” Earthquake
ASCE, 2000, Prestandard and Commentary for Seismic Engineering and Structural Dynamics, 137(10), 807-825.
Rehabilitation of Buildings FEMA 356, Federal Emergency
Management Agency PEER 2010, TBI Guidelines for Performance-based Design of
Tall Buildings, Pacific Earthquake Engineering Research
ASCE, 2002, Minimum Design Loads for Buildings and Other Center, Berkeley, CA
Structures, ASCE 7, American Society of Civil Engineers,
Reston, VA SAC Joint Venture, 2000, Recommended Seismic Design
Criteria for Welded Steel Moment Resisting Frames, Report
ASCE, 2006, Seismic Rehabilitation of Buildings, ASCE 41, No. FEMA 350, Federal Emergency Management Agency,
American Society of Civil Engineers, Reston, VA Washington, D.C.

ASCE, 2013, Seismic Rehabilitation of Buildings, ASCE 41, SEAOC. 1990, Recommended Lateral Force Requirements
American Society of Civil Engineers, Reston, VA and Commentary, Seismology Committee, Structural
Engineers Association of California, Sacramento, CA.
ASCE, 2016, Minimum Design Loads and Associated Criteria
for Buildings and Other Structures, ASCE 7, American Sherman, J.D., 2011, “Effects of key parameters on the
Society of Civil Engineers, Reston, VA performance of concrete masonry shear walls under in-plane

12
loading,” Master Thesis, Washington State University,
Pullman, WA.

Suita, K., Yamada, S., Tada, M., Kasai, K., Matsuoka, Y. and
Shimada, Y., 2008, “Collapse experiment on 4-story steel
moment frame part 2 detail of collapse behavior,”
Proceedings, the 14th World Conference on Earthquake
Engineering, October, 12-17, 2008, Beijing, China.

Suzuki, Y. and Lignos, D.G., 2015, “Large-scale collapse

experiments of wide flange steel beam-columns,”
Proceedings, 8th International Conference on the Behavior of
Steel Structures in Seismic Areas, Shanghai, China, July 1-3,
2015.

Whitman, 2015 “Investigation of Seismic Failure Modes in

Flexural Concrete Walls Using Finite Element Analysis “,
Thesis, University of Washington,

13