Академический Документы
Профессиональный Документы
Культура Документы
https://doi.org/10.1007/s13349-020-00416-1
ORIGINAL PAPER
Abstract
This paper develops a decision making framework for post-earthquake assessment of instrumented buildings in a manner
consistent with performance-based design criteria. This framework is achieved by simultaneously combining and advancing
existing knowledge from seismic structural health monitoring and performance-based earthquake engineering paradigms.
The framework consists of (1) measurement, (2) uncertainty modeling, (3) dynamic response reconstruction, (4) damage
estimation, and (5) performance-based assessment and decision making. In particular, the main objective is to reconstruct
inter-story drifts with a probabilistic measure of exceeding performance-based acceptance limits and determine the post-
earthquake re-occupancy classification of the instrumented building of interest. Since the proposed framework is probabilistic,
the outcome can be used to obtain the probability of losses based on the defined decision variables and be integrated into a
risk-based decision making process by city officials, building owners, and emergency managers. The framework is illustrated
using data from the Van Nuys hotel testbed, a seven-story reinforced concrete building instrumented by the California Strong
Motion Instrumentation Program (CSMIP Station 24386).
Keywords Decision making · Performance-based earthquake engineering · Seismic structural health monitoring · Dynamic
response reconstruction · Instrumented buildings · Real-world validation
List of symbols G0 Constant power spectral density intensity
arg min Argument of the minimum hk Height of the kth story
𝐛1 Spatial distribution of excitation I(t) Non-negative envelope function
𝐛2 Spatial distribution of process noise 𝐊 Stiffness matrix
c2 Output location matrix 𝐌 Mass matrix
𝐂𝜉 Damping matrix n Number of degrees-of-freedom
e State error p[.] Probability
𝔼 Expected value 𝐏 State error covariance
E Viscous damping coefficient 𝐏ISD Inter-story drift error covariance
𝐄 Feedback matrix q(t) Displacement vector
𝐄opt Optimal feedback matrix q̂ (t) Displacement vector estimate
F(t) Corrective force ̇ Velocity vector
q(t)
fc′ Compressive strength of concrete ̈ Acceleration vector
q(t)
fR (.) Restoring force function Sü ∗ ü ∗ (𝜔) Kanai–Tajimi power spectral density
t Time
tr(.) Trace
* Milad Roohi ü g (t) Ground acceleration vector
mroohigh@colostate.edu
v(t) Measurement noise
Eric M. Hernandez w(t) Process noise
eric.hernandez@uvm.edu
z(t) Vector of auxiliary variables
1
Department of Civil and Environmental Engineering, y(t) Measured displacement vector
Colorado State University, Fort Collins, CO 80523, USA ̇ Measured velocity vector
y(t)
2
Department of Civil and Environmental Engineering, ÿ (t) Measured acceleration vector
University of Vermont, Burlington, VT 05405, USA 𝜉g Site dominant damping coefficient
123
Vol.:(0123456789)
Journal of Civil Structural Health Monitoring
123
Journal of Civil Structural Health Monitoring
Furthermore, efforts have been made to bridge the seis- construction information) to assess the extent of the struc-
mic structural health monitoring and performance-based tural damage accurately and, subsequently, make informed
earthquake engineering paradigms to develop probabilistic decisions. Such a procedure will lead to mitigate earth-
frameworks that can assess the performance of buildings quake losses, reduce the decision making uncertainty, and
and provide vital information to support decision mak- improve community resilience.
ing immediately following an earthquake. In [28], Çelebi This paper develops a post-earthquake assessment
et al. developed a seismic monitoring system for real-time framework consistent with criteria from performance-
monitoring of inter-story drifts. This system requires meas- based design. The proposed framework consists of the fol-
uring the displacement of adjacent floors (i.e., inter-story lowing steps: (1) measurement, (2) uncertainty modeling,
drifts) on multiple pairs of building floors and then relates (3) dynamic response reconstruction, (4) damage estima-
inter-story drifts to performance-based damage thresholds tion, and (5) performance-based assessment and decision
specified in documents such as FEMA-356 or ASCE 41. making. Since the proposed framework is developed on a
Miranda [29] and Naeim et al. [16] distinguished probabil- probabilistic basis, the outcome can be used to obtain the
istic measures based on fragility functions as the most prom- probability of various losses based on the defined deci-
ising tools for real-time damage detection and conducted sion variable and be integrated into a risk-based decision
an exhaustive investigation on automated post-earthquake making process.
damage assessment of instrumented buildings. Porter et al. This paper is organized as follows. Section 3 pre-
[30] developed a framework based on PEER PBEE meth- sents the building model of interest. This is followed by
odology for near-real-time loss estimation for instrumented Sect. 4 that presents background on performance-based
buildings. This framework performs Bayesian updating to earthquake engineering. Section 5 develops the proposed
estimate the structural response based on measured base- framework for performance-based post-earthquake deci-
ment motion and a stochastic structural model and then sion making. Finally, the paper ends with a case study of
uses fragility functions to predict structural damage. Later, the Van Nuys hotel testbed, a seven-story reinforced con-
Mitrani-Reiser et al. [31] proposed a decision-support sys- crete (RC) building instrumented by the California Strong
tem for both immediate pre-event and post-event building Motion Instrumentation Program (CSMIP Station 24386),
safety and loss assessment. This system integrated fragil- to illustrate the effectiveness of the proposed framework.
ity curves and shaking intensity predictions from seismic
network data and employs the PEER PBEE methodology
to perform probabilistic damage and loss estimation and,
subsequently, determine post-earthquake occupancy class 2 Building model
of the building according to the ATC-20 procedure. Hwang
and Lignos [32] proposed a data-driven framework to esti- The global response of typical multistory building
mate story-based seismic engineering demand parameters structures to seismic ground motions can be accurately
(EDPs) along the height of instrumented steel frame build- described for engineering purposes by
ings with moment-resisting frames using a wavelet-based
𝐌q(t)
̈ + 𝐂𝜉 q(t)
̇ + fR (q(t), q(t),
̇ z(t))
damage sensitive feature. The estimated EDPs were used for
(1)
earthquake-induced loss assessment. In [33], Cremen and = −𝐌𝐛1 ü g (t) + 𝐛2 w(t)
Baker provided a methodology for quantifying the benefits
of building instrumentation, by measuring errors in damage where the vector q(t) ∈ ℝn contains the relative displace-
and loss consequence predictions calculated from the FEMA ment (with respect to the ground) of all stories. z(t) is a
P-58. Recently, Roohi et al. [5, 6] developed a nonlinear vector of auxiliary variables dealing with material nonlinear-
model-based observer for response reconstruction (i.e., state ity and damage behavior. n denotes the number of geometric
estimation) in nonlinear hysteretic structural systems and DoF, 𝐌 = 𝐌T ∈ ℝn×n is the mass matrix, 𝐂𝜉 = 𝐂T𝜉 ∈ ℝn×n
employed the observed for performance-based assessment of is the damping matrix, fR (⋅) is the resultant global restoring
experimental and real-world large-scale instrumented steel force vector. The matrix 𝐛1 ∈ ℝn×r is the influence matrix of
(Roohi et al. [34]), reinforced concrete (Roohi et al. [12]) the r ground acceleration time histories defined by the vector
and wood-frame (Roohi et al. [5]) buildings. ü g (t) ∈ ℝr . The matrix 𝐛2 ∈ ℝn×p defines the spatial distribu-
Despite the progress described, there still exists a lack tion the vector w(t) ∈ ℝp , which in the context of this paper
of a comprehensive performance-based rapid and reliable represents the process noise generated by unmeasured exci-
decision making procedure to integrate all the available tations and (or) modeling errors.
information (such as measurements, structural drawings,
123
Journal of Civil Structural Health Monitoring
123
Journal of Civil Structural Health Monitoring
response; p[DV|DM] is the conditional probability of expe- systems (which can be described by linear models), the den-
riencing a loss of certain size, given a level of damage. The sities p[EDP|M] are Gaussian. This means they can be char-
expected loss or value of the decision variable p[DV] is acterized by mean vectors and covariance matrices; thus, the
calculated as the sum of these quantities over all levels of mathematical solution becomes trackable. This is important
intensity, response, damage, and loss. The following section because in real world application there are many cases that
aims to develop a framework for decision making regarding can be addressed using this special case. However, in the
the post-earthquake assessment of instrumented buildings in case of more complicated systems, where there is a need to
a manner consistent with criteria from performance-based solve the nonlinear filtering problem, there does not exist a
design. finite set of parameters that can characterize the densities
p[EDP|M] . Instead, we seek algorithms that can provide
4 Proposed framework estimates based on approximations of the probability den-
for performance‑based post‑earthquake sity functions using the first two statistical moments [23].
decision making In the following, each step of solving Eq. 3 is discussed in
more details to obtain approximate solution of the p[DV]
The proposed decision making framework stems from the and use the outcome for performance-based post-earthquake
performance-based design framework and consists of the fol- decision making.
lowing five steps: (1) measurement, (2) uncertainty modeling,
(3) dynamic response reconstruction, (4) damage estimation,
and (5) performance-based assessment and decision making. 4.1 Measurement
Figure 2 presents a summary of the proposed framework char-
acterized by four variables consist of Response Measurement The first step of the proposed framework is to perform seis-
(M), Engineering Demand Parameter (EDP), Damage Meas- mic instrumentation and measure dynamic response during
ure (DM), and Decision Variable (DV). Using the Total Prob- a seismic event. In practice, this process begins by determin-
ability Theorem, the decision variable can be expressed by ing the type, number, and locations of the sensors consider-
ing technical, logistical, and economic constraints. With-
out loss of generality, this paper focuses on accelerometers
∭
p[DV] = p[DV|DM] p[DM|EDP] p[EDP|M] …
(3) as the sensor of choice due to their popularity, durability,
p[M] dM . dEDP . dDM and reliability. In a typical setup, accelerations are meas-
ured horizontally in three independent and non-intersecting
where p[M] is the probability density of measurement set, directions and the vector of acceleration measurements,
and p[EDP|M] is the conditional probability of experiencing ÿ (t) ∈ ℝm , is given by
a level of response given measurement set M . Except for a
few special cases, solving the multidimensional integrals ÿ (t) =
[ ] (4)
in Eq. 3 is very complex and challenging task as it requires − 𝐜2 𝐌−1 𝐂𝜉 q(t)
̇ + fR (q(t), q(t),
̇ z(t)) − 𝐛𝟐 w(t) + v(t)
the complete probability distribution of the p[EDP|M] ,
p[DM|EDP] , and p[DV|EDP] to be estimated. For instance,
to estimate p[EDP|M] in the special case of linear structural
123
Journal of Civil Structural Health Monitoring
where 𝐜2 ∈ ℝm×n is a Boolean matrix that maps the DoFs measurements during seismic events, characterized by p[M] ,
to the measurements, and v(t) ∈ ℝm×1 is the measurement are obtained.
noise.
To determine the number and location of sensors (i.e., 4.2 Uncertainty modeling
the measurement matrix 𝐜2 in Eq. 4) an optimality crite-
rion is needed to be defined. In this paper, the aim is to The second step is to model the uncertainty in the unmeas-
place accelerometers in locations that contain maximum ured ground motion excitations and the measurement noise.
information for dynamic response reconstruction, i.e., The uncertainty modeling is explicitly performed using power
select the number and locations of sensors in a way that spectral density (PSD). The uncertainty models of this step
minimizes the uncertainty of dynamic response recon- can be expressed by 𝜱vv (𝜔) and 𝜱ww (𝜔) , defined as the PSD
struction. This minimization can be achieved by selecting of the unmeasured excitations and the measurement noise,
an optimality criterion based on the variance of a user- respectively.
defined objective function related to the state of the sys-
tem, such as displacement, internal forces, and stresses. 4.2.1 Measurement noise characterization
The proposed framework selects the objective function to
be the sum of the inter-story drift estimation variances. The measurement noise is modeled using a zero-mean Gauss-
Therefore, the optimal sensor placement can be achieved ian sequence with a noise-to-signal RMS (root-mean-square)
by solving an optimization problem to select the optimal selected based on the expected accuracy of measurements.
measurement matrix, (𝐜2 )opt , subject to maximum inter- Using the white noise PSD, the PSD of measurement noise,
story drift (ISD) estimation variance being bounded by a 𝜱vv (𝜔) , is characterized as follows
maximum allowable variance of 𝜎max 2
, which can be speci-
fied based on the expected accuracy to determine perfor-
𝜱vv (𝜔) = 𝜱0 (9)
mance-based post-earthquake re-occupation category of which implies that the power of the measurement noise is
the building of interest. This optimization problem can be distributed uniformly over all frequency components. This
formulated as follows may or may not be valid, depending on the sensor. If a differ-
(𝐜2 )opt = arg min tr(𝐏ISD ) ent PSD is specified by the manufacturer, then it can simply
𝐜2 be used instead of the white assumption.
[ 2 ] (5)
s.t. max 𝜎ISD 2
(k, k) k=1∶n < 𝜎max Additionally, in some modeling situations, one can use
the measurement noise to include the effects of unmodeled
where tr(.) is trace of a square matrix (defined as the sum of dynamics (typically induced by modeling errors). Some
elements on the diagonal), 𝐏ISD (k, k) is the inter-story drift authors [6] have proposed updating the noise PSD in order to
estimation error covariance matrix, 𝜎ISD
2
(k, k) is the kth diag- include these effects. This is an interesting proposal; however,
onal element of inter-story drift estimation error covariance it lies outside of the scope of this paper, and it will not be
matrix, k is story number, and n is total number of stories. pursued further.
Here, tr(𝐏ISD ) is given by
4.2.2 Ground motion characterization
∑
n
tr(𝐏ISD ) = 𝐏ISD (k, k) (6)
k=1 The seismic ground motions can be modeled using a
Kanai–Tajimi power spectral density corrected by an ampli-
where 𝐏ISD (k, k) is described as tude function to obtain a non-stationary ground motion
𝐏ISD (k, k) = acceleration. The PSD of ground motion, 𝜱ww (𝜔) , using the
{ Kanai–Tajimi PSD, Sü ∗ ü ∗ (𝜔) , is characterized by
𝐏(1, 1), if k = 1 (7)
𝐏(k, k) + 𝐏(k − 1, k − 1) − 2𝐏(k, k − 1), if k > 1 𝜱ww (𝜔) = Sü ∗ ü ∗ (𝜔)
1 + 4𝜉g2 ( 𝜔𝜔 )2
and 𝐏 , the displacement estimation error covariance matrix, g
(10)
= G0 [ ]2
is given by
1 − ( 𝜔𝜔 )2 + 4𝜉g2 ( 𝜔𝜔 )2
[[ ]T ] g g
123
Journal of Civil Structural Health Monitoring
filtering a Gaussian white noise through a second order lin- dynamic response reconstruction is the third step of the
ear filter (single degree of freedom oscillator) with natural proposed framework. Response reconstruction refers to the
frequency 𝜔g and viscous damping 𝜉g as follows estimation of unmeasured response quantities of interest
or engineering demand parameters (EDP) from a limited
̇ + 𝜔2g u(t) = −w(t)
̈ + 2𝜉g 𝜔g u(t)
u(t) (11) number of global dynamic response measurements, given
by p[EDP|M] . The information needed for reconstruct-
where ing dynamic response are the following: (1) the dynamic
√ response of the building at all DoF and (2) a mapping
K C
𝜔g = and 𝜉g = between the global and local DoF of every element. An
M 2M𝜔g
accurate response reconstruction in this step is vital to
M, C, and K are the mass, stiffness, and damping parameters prevent under-estimation or over-estimation of the actual
of the linear filter. w(t) is a Gaussian white noise process response of the building. Furthermore, the estimation
with spectral density Sww (𝜔) = G0 . The Kanai–Tajimi PSD uncertainty bound of the reconstructed response helps to
models the earthquake-induced base motion as a stationary develop a set of maximum, mean, and minimum seismic
stochastic process, under the premise that only the frequency demand to consider the best and worst-case scenarios in
content is considered. To take into account the amplitude assessing the performance of the instrumented building.
variability of the motion the following time-dependent enve- The estimated response parameters with their associated
lope is used uncertainties can form a demand set to perform damage
estimation.
ü gm (t) = I(t)u(t)
̈ In the literature, researchers have proposed four catego-
[ ] (12) ries of state observers to perform response reconstruction
= I(t) −𝜔2g u(t) − 2𝜉g 𝜔g u(t)
̇
based on nonlinear filters including: (a) classical nonlinear
Bayesian filters (e.g., extended Kalman filter (EKF) [41]),
where ü gm (t) is the simulated ground motion and I(t) is a
(b) modern nonlinear Bayesian (or statistically linearized)
non-negative function representing the time-dependent enve-
filters (e.g., unscented Kalman filter (UKF) [42], (c) particle-
lope. For the purpose of this study, the amplitude modulating
based nonlinear Bayesian filters (e.g. the particle filter (PF)
function I(t) is selected as
[43]), and (d) model-based state observers (e.g., nonlinear
I(t) = te−𝛼t (13) model-based observer (NMBO) [5]). From these response
reconstruction approaches, the proposed framework uses the
The corrected realization of the Kanai–Tajimi PSD in Eq. 12 NMBO for response reconstruction in instrumented build-
provides a filtered white noise stochastic time series with ings. This is mainly because the NMBO has been formu-
appropriate frequency content and amplitude modulation lated in such a way that (1) it explicitly accounts for PSDs
for ground acceleration during earthquakes. Therefore, the of the excitations and measurement noise, and (2) it can
model can be conveniently used for stochastic response anal- be implemented directly as a modified nonlinear structural
ysis of structures to ground motion excitations. Figure 3 pre- model of a system subjected to corrective forces. The latter
sents a schematic of the procedure used for seismic ground feature allows for the direct implementation of the nonlinear
motion simulation. state observer using high-fidelity finite element (FE) models.
Thus, the NMBO can take advantage of a wide range of
4.3 Dynamic response reconstruction material and element models available in advanced nonlinear
simulation software packages, which not only results in the
Once data becomes available from a seismic event and the ease of implementation but also improve the state estimation
measurement noise and excitation have been characterized, accuracy due to better modeling capabilities. In contrast, the
existing state observers such as the EKF, UKF, and PF, are
associated with computational modeling limitations, because
the models that such state observers use for state estimation
are rather simple and cannot capture the complexity of the
nonlinear structural behavior. This issue is exacerbated in
problems that require modeling of highly nonlinear behavior.
The next subsection presents a summary of the NMBO and
its implementation.
123
Journal of Civil Structural Health Monitoring
The NMBO estimate of the displacement response, q̂ (t) , is the matrices 𝜱ww (𝜔) and 𝜱vv (𝜔) are, respectively, the PSDs
given by the solution of the following set of ordinary dif- of the uncertain excitation and measurement noise defined
ferential equations in the previous step.
The optimal feedback gain matrix, 𝐄opt , can be selected
𝐌q̂̈ (t) + (𝐂𝜉 + 𝐜T2 𝐄𝐜2 )q̂̇ (t)+fR (̂q(t), q̂̇ (t), z(t))
(14) by solving an optimization problem to minimize the inter-
= 𝐜T2 𝐄y(t)
̇ story drift estimation error covariance matrix, 𝐏ISD , given by
𝐌q̂̈ (t) + (𝐂𝜉 + 𝐜T2 𝐄𝐜2 )q̂̇ (t) + 𝐊0 q̂ (t) = 𝐜T2 𝐄y(t)
̇ (15)
where 𝐇o is defined as Fig. 4 Summary of the nonlinear model-data fusion using the nonlin-
ear model-based observer (NMBO)
123
Journal of Civil Structural Health Monitoring
p[DVk ≥ PL] =
4.4 Damage estimation
�PL
p[DVk |DMk ] dDMk
The third step of the proposed framework is to estimate dam- (23)
�PL
age measure (DM) from the estimated response and compare = p(ISDk ) dISDk
⎩ p(ISD ≥ CP)
⎪ for PL = CP
defined as probability of DV exceeding specific performance for PL = C
level (PL) based on performance-based acceptance criteria.
The acceptance criteria relate engineering demand parameters
(such as inter-story drifts, inelastic element deformations, and
element forces) to qualitative performance levels of Immediate 5 Case‑study: Van Nuys hotel testbed
Occupancy (IO), Life Safety (LS), Collapse Prevention (CP),
and Collapse (C) [47]. Therefore, the probability of exceed- This section illustrates the proposed framework using seis-
ing specific performance level (PL), p[ISDk ≥ PL] , can be mic response measurements from Van Nuys hotel. The
calculated for each story as follows CSMIP instrumented this building as Station 24386, and
the recorded data of this building are available from several
123
Journal of Civil Structural Health Monitoring
123
Journal of Civil Structural Health Monitoring
123
Journal of Civil Structural Health Monitoring
123
Journal of Civil Structural Health Monitoring
adjacent stories are only measured for the first two stories, 7
and thus, the actual inter-story drifts can be obtained for
6
these stories and used for validation of the estimates using
the NMBO and the FE model. Figure 8 presents the maxi- 5
mum inter-story drift ratios obtained from response meas- 4
urements for the first two stories, along with those estimated
3
for all the stories using the NMBO and open-loop FE model
analysis during the 1992 Big Bear and the 1994 Northridge 2
earthquakes. As seen, the NMBO outperforms the FE model
1
in the estimation of inter-story drifts, and therefore, the
NMBO ISD estimates and their uncertainties are used then 0
to reconstruct PDF of ISD for each story, p(ISDk ) . Figure 9 0 0.2 0.4 0.6 0.8 1
123
Journal of Civil Structural Health Monitoring
Fig. 9 Reconstructed probability density function for maximum inter-story drift ratios of the Van Nuys building during the a 1992 Big Bear and
b 1994 Northridge earthquakes
failure of 5 of 9 columns in the fourth story. However, the and Northridge earthquakes. In [12], Roohi et al. have shown
building-level estimate showed that the building could be that if the objective is the high-resolution story- or element-
classified as CP by a 0.80 probability of exceedance. There- level damage detection and localization, other damage sen-
fore, the building-level post-earthquake assessment results sitive response parameters such as element-level dissipated
are consistent with the building’s actual performance and energy, demand-to-capacity ratios, and ductility demand
post-earthquake inspection reports following the Big Bear can provide more accurate assessment results compared to
123
Journal of Civil Structural Health Monitoring
Fig. 10 Reconstructed cumulative density function for maximum inter-story drift ratios and probability of exceeding IO, LS, and CP perfor-
mance levels of Van Nuys building during a 1992 Big Bear and b 1994 Northridge earthquakes
123
Journal of Civil Structural Health Monitoring
Table 2 Story-by-story k (story) 1 2 3 4 5 6 7
probability of exceeding specific
performance level for the Van 1992 Big Bear earthquake
Nuys building during 1992
p[ISDk < IO] 1.00 1.00 1.00 1.00 1.00 1.00 1.00
p[ISDk ≥ IO]
Big Bear and 1994 Northridge
earthquakes 0.00 0.00 0.00 0.00 0.00 0.00 0.00
p[ISDk ≥ LS] 0.00 0.00 0.00 0.00 0.00 0.00 0.00
p[ISDk ≥ CP] 0.00 0.00 0.00 0.00 0.00 0.00 0.00
1994 Northridge earthquake
p[ISDk < IO] 0.01 0.00 0.00 0.00 0.44 0.57 0.99
p[ISDk ≥ IO] 0.99 1.00 1.00 1.00 0.56 0.43 0.01
p[ISDk ≥ LS] 0.33 0.13 0.66 0.04 0.00 0.00 0.00
p[ISDk ≥ CP] 0.00 0.00 0.01 0.00 0.00 0.00 0.00
123
Journal of Civil Structural Health Monitoring
Table 3 Building-level probability of exceeding and classifying spe- 3. Bernal D (2006) Flexibility-based damage localization from sto-
cific performance level for the Van Nuys building during 1992 Big chastic realization results. J Eng Mech 132(6):651–658
Bear and 1994 Northridge earthquakes 4. Sadeghi Eshkevari S, Heydari N, Nathan Kutz J, Pakzad SN, Dip-
las P, Sadeghi Eshkevari S (2019) Operational vision-based modal
1992 Big Bear earthquake identification of structures: a novel framework. Struct Health
p[ISD < IO] 1.00 p[ISD = IO] 1.00 Monit. https://doi.org/10.12783/shm2019/32502
p[ISD ≥ IO]
5. Roohi M, Hernandez EM, Rosowsky D (2019) Nonlinear seismic
0.00 p[ISD = LS] 0.00
p[ISD ≥ LS]
response reconstruction and performance assessment of instru-
0.00 p[ISD = CP] 0.00 mented wood-frame buildings—validation using NEESWood
p[ISD ≥ CP] 0.00 p[ISD = C] 0.00 capstone full-scale tests. Struct Control Health Monit 26(9):e2373
1994 Northridge earthquake 6. Roohi M, Erazo K, Rosowsky D, Hernandez EM (2020) An
extended model-based observer for state estimation in nonlinear
p[ISD < IO] 0.00 p[ISD = IO] 0.00
p[ISD ≥ IO]
hysteretic structural systems. Mech Syst Signal Process. https://
1.00 p[ISD = LS] 0.80 doi.org/10.1016/j.ymssp.2020.107015
p[ISD ≥ LS] 0.81 p[ISD = CP] 0.19 7. Erazo K, Hernandez EM (2016) Uncertainty quantification of
p[ISD ≥ CP] 0.01 p[ISD = C] 0.01 state estimation in nonlinear structural systems with application
to seismic response in buildings. ASCE-ASME J Risk Uncertain
Eng Syst Part A: Civ Eng 2(3):B5015001
8. Ching J, Beck JL, Porter KA, Shaikhutdinov R (2006) Bayesian
state estimation method for nonlinear systems and its application
to recorded seismic response. J Eng Mech 132(4):396–410
9. Hu RP, Xu YL (2019) Shm-based seismic performance assess-
ment of high-rise buildings under long-period ground motion. J
Struct Eng 145(6):04019038
(a) 1992 Big Bear earthquake 10. Şafak E (1999) Wave-propagation formulation of seismic response
of multistory buildings. J Struct Eng 125(4):426–437
11. Todorovska MI, Trifunac MD (2010) Earthquake damage detection
in the imperial county services building II: analysis of novelties via
wavelets. Struct Control Health Monit 17(8):895–917
12. Roohi M, Hernandez EM, Rosowsky D (2020) Reconstructing
element-by-element dissipated hysteretic energy in instrumented
(b) 1994 Northridge earthquake buildings: application to the Van Nuys Hotel testbed. ASCE J Eng
Mech. arXiv preprint arXiv:2002.12426 (under review)
13. Hernandez EM, May G (2012) Dissipated energy ratio as a feature
Fig. 12 Building-level estimated probability of post-earthquake per- for earthquake-induced damage detection of instrumented structures.
formance levels of Van Nuys building during a 1992 Big Bear and b J Eng Mech 139(11):1521–1529
1994 Northridge earthquakes 14. Simoen E, De Roeck G, Lombaert G (2015) Dealing with uncer-
tainty in model updating for damage assessment: a review. Mech
Syst Signal Process 56:123–149
(collapse prevention) for the Northridge earthquake. The 15. Astroza R, Ebrahimian H, Conte JP (2019) Performance comparison
post-earthquake assessment results were consistent with the of kalman-based filters for nonlinear structural finite element model
building’s actual performance and visual inspection reports. updating. J Sound Vib 438:520–542
16. Naeim F, Hagie S, Alimoradi A, Miranda E (2006) Automated post-
The study illustrates the capability of the proposed perfor- earthquake damage assessment of instrumented buildings. In: Wasti
mance-based framework to help structural engineers make ST, Ozcebe G (eds) Advances in earthquake engineering for urban
informed and swift decisions regarding post-earthquake risk reduction. Nato science series: IV: Earth and environmental
assessment of critical instrumented building structures and sciences, vol 66. Springer, Dordrecht
17. Lenjani A, Bilionis I, Dyke SJ, Yeum CM, Monteiro R (2020) A
to improve earthquake resiliency of communities. resilience-based method for prioritizing post-event building inspec-
tions. Nat Hazards 100(2):877–896. https://doi.org/10.1007/s1106
Acknowledgements Support for this research provided, in part, by 9-019-03849-0
award No. 1453502 from the National Science Foundation is grate- 18. Mohsen A, Pekcan G (2020) Structural health monitoring using
fully acknowledged. extremely compressed data through deep learning. Comput-Aided
Civ Infrastruct Eng 35:597–614
19. Pan H, Azimi M, Gui G, Yan F, Lin Z (2017) Vibration-based sup-
port vector machine for structural health monitoring. In: Interna-
tional conference on experimental vibration analysis for civil engi-
References neering structures, pp 167–178. Springer, Dordrecht
20. Mangalathu S, Sun H, Nweke CC, Yi Z, Burton HV (2020) Classify-
1. ATC (1989) Procedures for postearthquake safety evaluations of ing earthquake damage to buildings using machine learning. Earthq
buildings, report ATC-20. Technical report, Applied Technology Spectra 36(1):183–208
Council (ATC), Redwood City, CA 21. Zhang Y, Burton HV, Sun H, Shokrabadi M (2018) A machine learn-
2. ATC (1995) Addendum to the ATC-20 postearthquake building ing framework for assessing post-earthquake structural safety. Struct
safety evaluation procedures. Technical report, Applied Technol- Saf 72:1–16
ogy Council (ATC), Redwood City, CA 22. Lenjani A, Dyke SJ, Bilionis I, Yeum CM, Kamiya K, Choi J, Liu
X, Chowdhury AG (2020) Towards fully automated post-event
123
Journal of Civil Structural Health Monitoring
data collection and analysis: pre-event and post-event informa- 39. ASCE (2013) ASCE/SEI 41-13: seismic evaluation and retrofit
tion fusion. Eng Struct 208:109884. https://doi.org/10.1016/j.engst of existing buildings. Technical report, American Society of Civil
ruct.2019.109884 Engineers, ASCE/SEI 41–13, Reston, VA
23. Roohi M (2019) Performance-based seismic monitoring of instru- 40. Porter KA (2003) An overview of PEER’s performance-based earth-
mented buildings. PhD thesis, Graduate College Dissertations and quake engineering methodology. In: Proceedings of ninth interna-
Theses. 1140. University of Vermont tional conference on applications of statistics and probability in civil
24. Wu RT, Jahanshahi MR (2020) Data fusion approaches for struc- engineering. Citeseer
tural health monitoring and system identification: past, present, and 41. Gelb A (1974) Applied optimal estimation. MIT press, Cambridge
future. Struct Health Monit 19(2):552–586 42. Julier S, Uhlmann J, Durrant-Whyte HF (2000) A new method for
25. Sohn H, Farrar CR, Hemez FM, Shunk DD, Stinemates DW, Nadler the nonlinear transformation of means and covariances in filters and
BR, Czarnecki JJ (2003) A review of structural health monitoring estimators. IEEE Trans Autom control 45(3):477–482
literature: 1996–2001. Los Alamos National Laboratory, USA, pp 43. Doucet A, Godsill S, Andrieu C (2000) On sequential monte carlo
1–7 sampling methods for bayesian filtering. Stat Comput 10(3):197–208
26. Fan W, Qiao P (2011) Vibration-based damage identification 44. Hernandez EM (2011) A natural observer for optimal state estima-
methods: a review and comparative study. Struct Health Monit tion in second order linear structural systems. Mech Syst Signal
10(1):83–111 Process 25(8):2938–2947
27. Azimi M, Eslamlou AD (2020) Data-driven structural health moni- 45. Hernandez EM (2013) Optimal model-based state estimation in
toring and damage detection through deep learning: state-of-the-art mechanical and structural systems. Struct Control Health Monit
review. Sensors 20(10):2778 20(4):532–543
28. Celebi M, Sanli A, Sinclair M, Gallant S, Radulescu D (2004) Real- 46. Roohi M, Hernandez EM, Rosowsky D (2019) Nonlinear seismic
time seismic monitoring needs of a building owner—and the solu- response reconstruction in minimally instrumented buildings—vali-
tion: a cooperative effort. Earthq Spectra 20(2):333–346 dation Using Neeswood capstone full-scale tests. Structural health
29. Miranda E (2006) Use of probability-based measures for automated monitoring 2019. Presented at the structural health monitoring 2019.
damage assessment. Struct Des Tall Spec Build 15(1):35–50 https://doi.org/10.12783/shm2019/32390
30. Porter K, Mitrani-Reiser J, Beck JL (2006) Near-real-time loss 47. FEMA-356 (2000) Prestandard and commentary for the seismic
estimation for instrumented buildings. Struct Des Tall Spec Build rehabilitation of buildings. American Society of Civil Engineers
15(1):3–20 (ASCE), Reston
31. Mitrani-Resier J, Wu S, Beck JL (2016) Virtual inspector and its 48. Krawinkler H (2005) Van Nuys hotel building testbed report: exer-
application to immediate pre-event and post-event earthquake loss cising seismic performance assessment. Pacific Earthquake Engi-
and safety assessment of buildings. Nat Hazards 81(3):1861–1878 neering Research Center, College of Engineering of California
32. Hwang SH, Lignos DG (2018) Assessment of structural damage 49. Trifunac MD, Ivanovic SS, Todorovska MI (1999) Instrumented
detection methods for steel structures using full-scale experimental 7-storey reinforced concrete building in Van Nuys, California:
data and nonlinear analysis. Bull Earthq Eng 16(7):2971–2999 description of the damage from the 1994 Northridge earthquake
33. Cremen G, Baker JW (2018) Quantifying the benefits of building and strong motion data. Report CE 99:2
instruments to FEMA p-58 rapid post-earthquake damage and loss 50. Trifunac MD, Ivanovic SS (2003) Analysis of drifts in a seven-story
predictions. Eng Struct 176:243–253 reinforced concrete structure. University of Southern California
34. Hernandez E, Roohi M, Rosowsky D (2018) Estimation of element- Report CE, pp 3–10
by-element demand-to-capacity ratios in instrumented SMRF 51. Frank M, Fenves GL, Scott MH et al (2000) Open system for earth-
buildings using measured seismic response. Earthq Eng Struct Dyn quake engineering simulation. University of California, Berkeley
47(12):2561–2578 52. Jalayer F, Ebrahimian H, Miano A, Manfredi G, Sezen H (2017)
35. SEAOC (1995) Vision 2000: performance based seismic engineer- Analytical fragility assessment using unscaled ground motion
ing of buildings. Structural Engineers Association of California, records. Earthq Eng Struct Dyn 46(15):2639–2663
Sacramento 53. Saiful Islam M (1996) Analysis of the northridge earthquake
36. ATC (1996) Seismic evaluation and retrofit of concrete buildings. response of a damaged non-ductile concrete frame building. Struct
2. Appendices. Applied Technology Council (ATC), Redwood City, Des Tall Build 5(3):151–182
CA
37. FEMA (1997) NEHRP guidelines for the seismic rehabilitation of Publisher’s Note Springer Nature remains neutral with regard to
buildings. FEMA-273, Federal Emergency Management Agency, jurisdictional claims in published maps and institutional affiliations.
Washington, DC
38. FEMA (2000) Commentary for the seismic rehabilitation of build-
ings. FEMA-356, Federal Emergency Management Agency, Wash-
ington, DC
123