Professional Documents
Culture Documents
DOI 10.1007/978-3-642-36197-5_136-1
# Springer-Verlag Berlin Heidelberg 2014
Introduction
An important issue in performance-based earthquake engineering is the estimation of structural
performance under seismic loads. In particular, one is interested in estimating the probabilistic
distribution of structural response in the form of engineering demand parameters (EDPs) such as
peak interstory drift, peak floor acceleration, moment, or shear, given the level of seismic intensity
represented by a (typically scalar) intensity measure (IM). Incremental dynamic analysis (IDA) has
been developed to provide such information by employing nonlinear dynamic analyses of the
structural model under a suite of ground motion records, each scaled to several intensity levels
designed to force the structure all the way from elasticity to final global dynamic instability.
The use of ground motion scaling for determining the response of a structure at increasing levels
of intensity is an old technique. Still, it had not been used systematically to quantify the probabilistic
nature of structural response until the start of the SAC/FEMA (2000a, b) project that was conceived
in the aftermath of the Northridge 1994 earthquake. Therein, a precursor to IDA was proposed in the
form of the “dynamic pushover,” a method to determine the global collapse capacity of structures
(Luco and Cornell 1998). Subsequently, this was recast and formalized by Vamvatsikos and Cornell
(2002) as a comprehensive procedure to assess the statistics of EDP demand given the IM as well as
the required (“capacity”) IM to achieve given values of EDP at any level of structural behavior. It is
this format that is known as IDA and will be described in the following sections, discussing the
necessary steps and concepts in executing, postprocessing, and applying IDA to solve problems of
engineering significance.
*Email: divamva@mail.ntua.gr
Page 1 of 8
Encyclopedia of Earthquake Engineering
DOI 10.1007/978-3-642-36197-5_136-1
# Springer-Verlag Berlin Heidelberg 2014
hazard curve can be practically estimated will never be fully sufficient; thus, excessive scaling may
introduce biased estimates of response (Luco and Bazzurro 2007).
A standard choice used in the literature is the 5 % damped first-mode (pseudo)spectral acceler-
ation Sa(T1) (Shome et al 1998; Shome and Cornell 1999). This is generally adequate for first-mode-
dominated structures that do not displace far into the nonlinear region, as is the case of most existing
brittle or moderately ductile low/mid-rise buildings. For taller or asymmetric structures where higher
modes become important or modern buildings that exhibit significant ductility, improved IM
alternatives should be sought. One particularly attractive option is the geometric mean of spectral
acceleration values at several periods (Cordova et al 2000; Vamvatsikos and Cornell 2005; Bianchini
et al 2009) or an inelastic spectral displacement-based IM (Luco and Cornell 2007). Both can largely
alleviate the effect of spectral shape, being able to capture the period elongation characterizing
ductile structures and the effect of higher modes.
In terms of selecting ground motion records, the general idea is to reduce the bias that a less than
ideal IM may introduce. A general rule of thumb is to utilize accelerograms that can significantly
damage the investigated structure without much scaling. For modern ductile structures, this gener-
ally means records having naturally high Sa(T1) values. Still, IDA is structured in such a way that the
same set of records is used throughout the entire range of structural behavior (elastic–inelastic
collapse). Whenever site conditions are deemed to have different impact at different levels of
intensity, thus fundamentally changing the nature of records that one would use at low versus
high levels of the IM, then methods of analysis other than IDA should probably be employed.
Structural Model
For performing IDA, a (a) realistic, (b) low-to-medium-complexity, (c) robust nonlinear structural
model should be formed. Each of these three requirements comes with its own reasons.
For one, realism means incorporating all pertinent sources of material and geometric nonlinearity
that are expected to arise. This should include, for example, plastic-hinge formation zones for
moment-resisting frames, brace buckling for braced frames, and P-Delta effects. Nonsimulated
failure modes, such as the shear failure of members or the brittle failure of beam–column joints, can
be incorporated in the analysis a posteriori. Still, they essentially remove the model’s ability to track
structural behavior beyond their first occurrence. This means that whenever nonsimulated failures
are found to have occurred, one cannot trust the model to provide estimates beyond that point.
The requirement of low-to-moderate complexity is necessary for ensuring that the computation-
ally intensive IDA will generally be confined to reasonable amounts of time. Complex nonlinear
models can complicate the execution to no end, easily forcing each nonlinear analysis to last several
hours or, worse, often forcing the analyst to consider reducing the number of records or dynamic
runs per record employed. This can be a tricky situation that may lead to inaccurate or biased results.
Thus, it is best to strike a trade-off between model complexity and fidelity, striving to find a good
balance that allows using a rich set of at least 20–30 ground motion records with six or more time
history analyses for each.
Finally, robustness becomes important in tracking structural behavior in the postelastic region,
especially beyond the structure’s maximum-strength point, where it starts deteriorating and
approaching collapse. To be able to ensure such results, a highly reliable software platform should
be used. IDA by nature will drive both the model and the analysis software to their limits, making
robustness difficult to achieve. Even then, typical complex finite element models that have been used
successfully for elastic or mildly inelastic analysis tend to fare poorly when driven close to global
Page 2 of 8
Encyclopedia of Earthquake Engineering
DOI 10.1007/978-3-642-36197-5_136-1
# Springer-Verlag Berlin Heidelberg 2014
dynamic instability, being consistently less reliable than their simpler versions besides significantly
raising the computational cost. For example, it is often the case that distributed plasticity models are
numerically less stable and always more expensive than lumped-plasticity ones. Therefore, consid-
erable attention should be paid to the formation and testing of the model before IDA is performed,
checking its behavior and stability via nonlinear static pushover analyses.
Execution
Performing IDA is conceptually simple. One only needs to take one record at a time, incrementally
scale it at constant or variable IM steps, and perform a nonlinear dynamic analysis each time. Start
from a low IM value where the structure behaves elastically, and stop when global collapse is
encountered. The latter is defined as the occurrence of a nonsimulated failure mode or the appear-
ance of a global dynamic instability as a collapse mechanism showing “infinite” EDP values at a
given IM level. For a well-executed analysis and robust structural model, global dynamic instability
manifests itself as numerical nonconvergence.
Selecting constant IM steps is often the simplest but also the most wasteful approach to IDA. As
global collapse is encountered at widely varying IM levels for each record, some ground motions
will require twice or thrice the dynamic analyses than others. Using variable IM steps estimated via
the hunt & fill procedure of Vamvatsikos and Cornell (2002, 2004), offers instead consistent
accuracy at a predefined computational cost. Due to the nearly perfect computational independence
of each dynamic analysis, IDA execution is an easily parallelizable problem (Vamvatsikos 2011).
Hence, one can benefit from using a cluster of N computers to divide the total running time (almost)
by N. Such savings make possible the use of Monte Carlo–based algorithms (Vamvatsikos 2014) to
economically estimate any bias and additional variability introduced by modeling uncertainty.
Postprocessing
The results of IDA initially appear as distinct points, one per each dynamic analysis, in the EDP
versus IM plane, as observed in Fig. 1 for a nine-story steel moment-resisting frame. Linear or spline
interpolation (Vamvatsikos and Cornell 2004) is then employed to generate continuous IDA curves,
one for each individual ground motion record, shown in Fig. 2. The variability offered by such
curves is actually one of the eye-opening features of IDA, visually representing the probabilistic
nature of seismic loading and the differences introduced even by (seemingly) similar ground
motions. At the highest IM level that each IDA curve can reach, one can identify the characteristic
flatline representing global collapse. In the example of Fig. 2, this is the intensity where the
maximum (over all stories) peak interstory drift (the EDP of choice) increases without bound for
each record, leading to global system collapse.
The complex picture of structural response shown by the IDA curves can be significantly
simplified by summarizing them into the 16/50/84 % fractile curves shown in Fig. 3. This is simply
the determination of the 16/50/84 % percentile values of EDP at each level of the IM, taking
horizontal cross sections of the IDA curves or, equivalently, the 84/50/16 % percentiles of the IM
given the EDP when vertical cross sections are assumed instead. The three curves of Fig. 3 thus
provide the full characterization of the distribution of structural response via the central value
(median) and the dispersion of EDP structural demand given the IM. For the example of the nine-
story steel frame, one can observe how the median demand obeys the “equal displacement” rule,
Page 3 of 8
Encyclopedia of Earthquake Engineering
DOI 10.1007/978-3-642-36197-5_136-1
# Springer-Verlag Berlin Heidelberg 2014
1.2
0.8
0.6
0.4
0.2
0
0 0.01 0.02 0.03 0.04 0.05 0.06
maximum interstory drift ratio, θmax
Fig. 1 The resulting IM-EDP points for 30 ground motion records when performing IDA for a 9-story steel frame using
Sa(T1) as the IM and the maximum over all stories peak interstory drift ymax as the EDP
1.2
"first−mode" spectral acceleration Sa(T1,5%) (g)
0.8
0.6
0.4
0.2
0
0 0.01 0.02 0.03 0.04 0.05 0.06
maximum interstory drift ratio, θmax
Fig. 2 The 30 IDA curves derived by interpolating the IM-EDP points of Fig. 1 for the 9-story steel frame
Page 4 of 8
Encyclopedia of Earthquake Engineering
DOI 10.1007/978-3-642-36197-5_136-1
# Springer-Verlag Berlin Heidelberg 2014
1.2
0.8
0.6
0.4
0.2
0
0 0.01 0.02 0.03 0.04 0.05 0.06
maximum interstory drift ratio, θmax
Fig. 3 The 16/50/84 % fractile IDA curves obtained by summarizing the individual IDA curves of Fig. 2 for the 9-story
steel frame
1.2
"first−mode" spectral acceleration Sa(T1,5%) (g)
0.8
0.6
median IM
0.4 capacity
0.2
median EDP capacity
0
0 0.01 0.02 0.03 0.04 0.05 0.06
maximum interstory drift ratio, θmax
Fig. 4 Thirty IDA curves, 30 limit-state capacity points and the corresponding EDP and IM capacity distributions for
the 9-story steel frame (From Vamvatsikos 2013)
probabilistic, the latter obviously being the more realistic option. When such threshold values are
exceeded, the limit state is deemed to have been violated. The first (in terms of the lowest IM)
exceedance of the limit state appears as a single point on each IDA curve (Fig. 4). The resulting
distribution of IM coordinates (or IM capacities) of the limit-state points can be used to define the
so-called fragility functions, as discussed in the following section.
Page 5 of 8
Encyclopedia of Earthquake Engineering
DOI 10.1007/978-3-642-36197-5_136-1
# Springer-Verlag Berlin Heidelberg 2014
Summary
Incremental dynamic analysis is a method for assessing the distribution of structural response at
every level of structural behavior from elasticity to final global dynamic instability. It employs
numerous nonlinear dynamic analyses of the structural model subject to a suite of ground motion
records that are scaled to multiple levels of intensity. By interpolating such results and appropriately
summarizing them, one can estimate any desired statistic of response given the seismic intensity.
Additionally, using the IDA results, any number of limit states from minor damage to global collapse
may be defined, and the associated fragility functions can be effortlessly estimated. Still, to preserve
the validity of such results and ensure unbiased estimation, attention should be paid to the ground
motions used and the intensity measure selected to represent them. Then, IDA becomes a powerful
tool for performance-based assessment of structures, factually linking seismic hazard and structural
behavior to estimate losses, downtime, and casualties.
Page 6 of 8
Encyclopedia of Earthquake Engineering
DOI 10.1007/978-3-642-36197-5_136-1
# Springer-Verlag Berlin Heidelberg 2014
Cross-References
▶ Analytic Fragility and Limit States [P(EDP|IM)]: Nonlinear Dynamic Procedures
▶ Analytic Fragility and Limit States [P(EDP|IM)]: Nonlinear Static Procedures
▶ Assessment of Existing Structures using Response History Analysis
▶ Collapse Prediction of Structures under Seismic Actions
▶ Nonlinear Dynamic Seismic Analysis
▶ Performance-Based Earthquake Engineering
▶ Seismic Collapse Assessment
▶ Seismic Fragility Analysis
▶ Seismic Loss Assessment
▶ Time History Seismic Analysis
▶ Uncertainty in Structural Properties: Effects on Seismic Performance
References
Bianchini M, Diotallevi PP, Baker JW (2009) Predictions of inelastic structural response using an
average of spectral accelerations. In: Proceedings of the 10th International Conference on
Structural Safety & Reliability (ICOSSAR), Osaka
Cordova PP, Deierlein GG, Mehanny SS, Cornell CA (2000) Development of a two-parameter
seismic intensity measure and probabilistic assessment procedure. In: Proceedings of the 2nd
U.S.-Japan workshop on performance-based earthquake engineering methodology for reinforced
concrete building structures, Sapporo, pp 187–206
Cornell CA, Krawinkler H (2000) Progress and challenges in seismic performance assessment.
PEER Center News. http://peer.berkeley.edu/news/2000spring/index.html. Accessed May 2014
FEMA (2009) Quantification of seismic performance factors. FEMA P-695 report, prepared by the
Applied Technology Council for the Federal Emergency Management Agency, Washington, DC
FEMA (2012) Seismic performance assessment of buildings. Methodology, vol 1. FEMA P-58-1
report, prepared by the Applied Technology Council for the Federal Emergency Management
Agency, Washington, DC
Luco N, Bazzurro P (2007) Does amplitude scaling of ground motion records result in biased
nonlinear structural drift responses? Earthq Eng Struct Dyn 36:1813–1835
Luco N, Cornell CA (1998) Effects of random connection fractures on demands and reliability for a
3-storey pre-Northridge SMRF structure. In: Proceedings of the 6th US national conference on
earthquake engineering, Seattle
Luco N, Cornell CA (2007) Structure-specific scalar intensity measures for near-source and ordinary
earthquake ground motions. Earthq Spectra 23(2):357–392
SAC/FEMA (2000a) Recommended seismic design criteria for new steel moment-frame buildings,
FEMA-350, SAC Joint Venture, Federal Emergency Management Agency, Washington, DC
SAC/FEMA (2000b) Recommended seismic evaluation and upgrade criteria for existing welded
steel moment-frame buildings, FEMA-351, SAC Joint Venture, Federal Emergency Management
Agency, Washington, DC
Shome N, Cornell CA (1999) Probabilistic seismic demand analysis of nonlinear structures. RMS
program, report no RMS35, PhD thesis, Stanford University
Shome N, Cornell CA, Bazzurro P, Carballo JE (1998) Earthquakes, records and nonlinear
responses. Earthq Spectra 14(3):469–500
Page 7 of 8
Encyclopedia of Earthquake Engineering
DOI 10.1007/978-3-642-36197-5_136-1
# Springer-Verlag Berlin Heidelberg 2014
Page 8 of 8