Professional Documents
Culture Documents
A R T I C L E I N F O A B S T R A C T
Keywords: The subspace state-space system identification method has drawn extensive attention in structural
Subspace state-space system identification modal identification, which is generally involved with the stabilization diagram for estimating
High-rise building structural modal parameters. However, the conventional stabilization diagram has an inherent
Earthquake excitation problem, i.e., some spurious modes may be identified as stable results, leading to the adverse
Modal identification effect on structural modal identification. To address this issue, this paper proposes an improved
Field measurement
subspace algorithm, in which a Monte Carlo-based stabilization diagram is involved. The per
formance of the Monte Carlo-based stabilization diagram for discriminating the poles denoting
the physical modes from those representing spurious modes is demonstrated through a numerical
study. The simulation results further prove that the proposed method can accurately estimate the
time-varying structural modal parameters. Moreover, the proposed method is applied to field
measurements on a 218-m-tall building during the 1994 Northridge earthquake event, and the
identified results verify the applicability and effectiveness of the proposed method in field
measurements. This paper aims to provide an effective tool for modal identification of high-rise
buildings under earthquake excitations.
1. Introduction
Identification of structural modal parameters (i.e., natural frequencies, damping ratios, and mode shapes) of high-rise buildings
under earthquake excitations is drawing increasing attention, as it plays a significant role in reducing gaps between real buildings and
their design models which helps to describe the structural behaviors in a more accurate and reliable manner [1,2]. The estimated
modal parameters can provide not only the information required in the seismic-design or dynamic analysis of structures but also the
prerequisite information for damage identification and post-earthquake emergency management to ensure structural safety [3–5].
Moreover, they can also be used to perform model updating, improve vibration control techniques, verify retrofit procedures and
improve earthquake-resistant structural designs, etc. [6].
The modal identification techniques of civil structures can be generally categorized into two groups: output-only methods and
input-output approaches. The input-output approaches require measurements of both external excitations and structural responses (e.
g., acceleration, velocity, displacement, and strain). While the output-only methodology is more convenient as it does not need to
measure the external excitations acting on structures. Although the output-only identification approaches are a popular choice in the
field of civil engineering, it has several notable limitations. The foremost one among these limitations is that a mathematical evaluation
* Corresponding author. Dept. of Architecture and Civil Engineering, City University of Hong Kong, Hong Kong.
E-mail address: bcqsli@cityu.edu.hk (Q.-S. Li).
https://doi.org/10.1016/j.jobe.2022.104373
Received 25 December 2021; Received in revised form 13 March 2022; Accepted 14 March 2022
Available online 17 March 2022
2352-7102/© 2022 Elsevier Ltd. All rights reserved.
K. Zhou and Q.-S. Li Journal of Building Engineering 52 (2022) 104373
of the accuracy of the system models is difficult to perform due to the absence of the input excitation [7]. In this regard, the
input-output approaches are generally favored over the output-only ones if the excitation is known.
Since the earthquake excitations (i.e., the ground accelerations) acting on high-rise buildings can be measured, the input-output
approaches are generally employed for the modal identification of high-rise buildings in earthquake engineering [8]. In the past
decades, several input-output methods have been proposed and implemented in this field. For instance [9], proposed a system
identification technique based on the least-squares minimization between the measured seismic response of a building and that
predicted by a parametric model [10]. developed a discrete-time filter method and applied it to identify structural dynamic parameters
of five buildings under earthquake actions [11]. proposed a Bayesian statistical framework for extracting structural modal parameters
from measured seismic responses [12]. presented an adaptive fading Kalman filter method for identifying the time-varying structural
dynamic properties of frame structures under earthquake actions [13]. proposed a maximum likelihood estimator-based frequency
domain identification method and verified its effectiveness through both numerical simulation and shaking table tests [14]. developed
an objective function minimization approach to estimate the modal frequencies, damping ratios, and mode shapes of high-rise
buildings under earthquake excitations.
Recently, the subspace state-space system identification approach is attracting increasing attention in structural modal identifi
cation under earthquake excitations due to its prominent advantages (e.g., state-space models are directly estimated, no parameter
izations are needed, robust linear algebra tools are employed) [15–17]. Generally, the subspace-based approach is applied with the
stabilization diagram for conducting structural modal identification [7]. The conventional stabilization diagram is generally estab
lished by displaying the identified modal parameters with gradually increased system order and thereby observing the stable poles.
However, as presented in several studies [18–20], the conventional stabilization diagram has an inherent problem, i.e., some spurious
modes may be identified as stable results. This may introduce errors in further analysis of the conventional stabilization diagram and
thereby has adverse effects on the accuracy of modal identification results. To mitigate the adverse effects, several efforts have been
made. For instance, Magalhães et al. [21] proposed an improved hierarchical clustering algorithm for the conventional stabilization
diagram to extract the stable poles, in which a cutoff distance parameter needs to be carefully assigned. Su et al. [22] combined the
convolutional neural network with the conventional stabilization diagram for modal identification, but such a method is only suitable
for single mode. Ubertini et al. [23] proposed a noise modes elimination method-based stabilization diagram and He et al. [46]
developed a robust mixed clustering method-based stabilization diagram, while the applications of these two stabilization diagrams in
field measurements indicated that several spurious poles were still misidentified as stable results. The literature review reveals that the
inherent problem of the conventional stabilization diagram has not been well addressed yet. To this end, this paper proposes a Monte
Carlo-based stabilization diagram that can effectively discriminate the physical modes from the spurious ones. Based on this novel
diagram, an improved subspace-based algorithm is established to achieve the accurate identification of modal parameters of high-rise
buildings under earthquake excitations. The effectiveness and accuracy of the proposed algorithm are further verified and demon
strated by numerical simulation and field measurements.
The structure of the rest of this paper is as follows: Section 2 introduces the methodology of the improved subspace algorithm;
Section 3 validates the effectiveness and stability of the proposed algorithm by a numerical study on a three-degree-of-freedom (3DOF)
model; Section 4 applies the proposed method to field measurements on a 52-story high-rise building during a strong earthquake event,
followed by the conclusions summarized in Section 5.
2. Methodology
In the past decades, several algorithms for identifying the subspace state-space system have been developed, such as the canonical
variate analysis (CAV) method [24], the multivariable output-error state-space (MOESP) algorithm [25], and numerical algorithm for
subspace state-space system identification (N4SID) [26], etc. Among these algorithms, the N4SID has been widely used in civil en
gineering due to its efficiency, robustness, and convenience [27–31]. More importantly, the N4SID is quite suitable for support-excited
structures like high-rise buildings [7,30]. In this way, the N4SID is employed in this paper for the modal identification of high-rise
buildings under earthquake excitations. In this section, a brief introduction of the N4SID algorithm will be first presented, and the
proposed improved algorithm will be described accordingly.
in which X is the state sequence; U and Y are the input and output sequences, respectively; A is the dynamical system matrix, B is the
input matrix describing how the deterministic inputs influence the next state, C is the output matrix characterizing how the internal
state influences the outputs and D is the direct transition matrix; W and V are zero-mean Gaussian random sequences corresponding to
the process noise and measurement noise, respectively. The combination of Equations (1) and (2) leads to
2
K. Zhou and Q.-S. Li Journal of Building Engineering 52 (2022) 104373
[ ] [ ] [ ] [ ]
Xi+1 A B W
= Xi + Ui + (3)
Yi C D V
By employing the projection and Kalman filter, Xi+1 and Xi can be determined following [7]. Then, the system matrices A and B can
be estimated based on a linear least-squares approximation, and the input and output matrices C and D can be accordingly unraveled
by solving a set of linear equations [26]). Next, the complex eigenvalues (i.e., λ) and eigenvectors (i.e., ψ ) of the damped system can be
calculated from the system matrix A [32] and the structural modal parameters can be determined as follows [10]:
⎧
⎨ fk = |λk |/2π
ξ = Re(λk )/2πfk (4)
⎩ k
φk = |Cψ k | · sign[Re(Cψ k )]
where Re(a) is the real part of a; sign[a] denotes the algebraic sign of a; fk , ξk , and φk represent the natural frequency, damping ratio, and
mode shape of the kth structural mode, respectively. It is worth noting that the order of the state-space model (also known as the
dimension of state sequence X) is a key parameter that needs to be assigned in the analysis procedure of the N4SID method, and its
determination will be discussed in the following section.
3
K. Zhou and Q.-S. Li Journal of Building Engineering 52 (2022) 104373
stage stability check approach. In the first-stage stability check, the criterion shown in Equation (5) is employed following [18], which
is commonly used in the conventional stabilization diagram.
⎧
⎪
⎪ fa + fb
⎪
⎪ Natural frequency : Δf = ≤ 2%
⎪
⎪ fa
⎨
ξ + ξb (5)
⎪
⎪ Damping ratio : Δξ = a ≤ 15%
⎪
⎪ ξa
⎪
⎪
⎩
Mode shape : 1 − MAC(φa , φb ) ≤ 5%
in which f, ξ, and φ are the natural frequency, damping ratio, and mode shape of one structural mode, respectively; subscripts a and b
represent two different structural modes; MAC denotes the modal assurance criterion of two vectors. By comparing the estimated
modal parameters Θ in the other s − 1 groups of Ω (i.e., [Ω2 , Ω3 , …, Ωs ]) with Θ11 in Ω1 , the number of Θ satisfying the criterion in
Equation (5) can be readily determined, which is regarded as the stable index for Θ11 . While in the second-stage stability check, only if
the stable index for Θ11 is larger than 0.3s, Θ11 can be regarded as a stable pole. By applying the proposed two-stage stability check
approach to the estimated modal parameters Θ in s groups of Ω, the stable poles denoting the physical modes can be effectively
discriminated against from those representing spurious modes. Based on the above analysis, the Monte Carlo-based stabilization di
agram can be established accordingly, in which the modal parameter estimates in s groups of Ω are plotted together (x axis-natural
frequency; y axis-order of random integers [i, n] that belongs to [1, s]). Based on the Monte Carlo-based stabilization diagram, the
structural modal parameters can be readily determined via the K-means clustering method [35].
The novelty of the proposed method is the consideration of the effects of both time lag parameter i and window length parameter l
in establishing the Monte Carlo-based stabilization diagram, and the usage of the two-stage stability check approach to eliminate the
spurious poles. It is noted that only the effect of system order parameter n is considered in the establishment of the conventional
stabilization diagram. In this way, similar spurious modes may be misidentified as stable poles, since the conventional stabilization
diagram just compares the modal parameters obtained using two adjacent system orders. In the proposed method, a number of
different sets of [i, l] can be generated by the Monte Carlo simulation. Based on this, similar spurious modes can only occur occa
sionally, while the physical modes exist more frequently. Then, by using a two-stage stability check approach, the spurious poles will
be eliminated and the physical modes will remain to achieve the accurate estimation of structural modal parameters. In the next
section, the procedure of the proposed upgraded N4SID algorithm will be demonstrated through a numerical example with a 3DOF
4
K. Zhou and Q.-S. Li Journal of Building Engineering 52 (2022) 104373
model.
The damping matrix C is determined by C = α · M + β · K, in which α and β are the damping proportional constants. In this section,
m1 , m2 , and m3 are set as 2 units, 1.5 units, and 1 unit, respectively. Besides, to simulate the time-varying modal parameters of the
3DOF model, the variations of stiffness coefficients k1 , k2 , k3 and damping proportional constants α, β during an 80-s-long period are
shown in Fig. 4. Based on these parameters, the time-varying structural dynamic properties (i.e., natural frequency and damping ratio)
of the 3DOF model can be readily determined as shown in Fig. 5.
As shown in Fig. 3, an earthquake excitation is exerted on the 3DOF model, its time history is illustrated in Fig. 6(a). By employing
the Newmark − β method, the acceleration responses of the three floors of the model can be readily calculated. The time duration of the
simulated responses is 80 s with a sampling frequency of fs = 10 Hz. Based on the simulated responses of the 3DOF model shown in
Fig. 6(b–d), the modal identification using the proposed algorithm will be demonstrated in the following sections.
5
K. Zhou and Q.-S. Li Journal of Building Engineering 52 (2022) 104373
Fig. 5. Time-varying natural frequencies and damping ratios of the 3DOF model.
6
K. Zhou and Q.-S. Li Journal of Building Engineering 52 (2022) 104373
Fig. 6. Simulation of acceleration responses of the 3DOF model with time-varying modal parameters (a) ground acceleration (b) acceleration response of the 1st floor
(c) acceleration response of the 2nd floor (d) acceleration response of the 3rd floor.
can be obtained as illustrated in Fig. 8(a). The stable poles in the Monte Carlo-based stabilization diagram are represented by red dots,
while the unstable poles are denoted by blue dots. For comparison purposes, the conventional stabilization diagram for this simulation
example is also provided in Fig. 8(b). Although three modes can be identified in the conventional stabilization diagram, it encounters a
serious problem that several spurious modes are identified as stable poles, which may compromise the accuracy of the modal iden
tification results. While in the Monte Carlo-based stabilization diagram, the three structural modes are clearly identified and no
spurious modes are misidentified as stable poles. The natural frequencies and damping ratios of these identified stable poles are plotted
in Fig. 9. Based on this, the K-means clustering method is employed and the structural modal parameters are readily identified as listed
in Table 1. It is shown that the modal parameters estimated via the proposed method are in good agreement with their theoretical
values, with relative errors less than 0.4% and 8% for natural frequency and damping ratio, respectively. The results and above
discussion prove that the proposed method can provide the accurate modal parameter estimates of high-rise buildings under earth
quake excitations.
7
K. Zhou and Q.-S. Li Journal of Building Engineering 52 (2022) 104373
8
K. Zhou and Q.-S. Li Journal of Building Engineering 52 (2022) 104373
Fig. 9. Natural frequencies and damping ratios of the identified stable poles for the 3DOF model.
Table 1
Estimated natural frequencies and damping ratios of the 3DOF model.
Theoretical value Proposed method Relative errora Theoretical value Proposed method Relative error
9
K. Zhou and Q.-S. Li Journal of Building Engineering 52 (2022) 104373
Fig. 11. Identification of time-varying structural modal parameters of the 3DOF model.
during a strong earthquake event, with the aim of verifying the applicability and effectiveness of the proposed method in field
measurements.
4.1. Field measurements on a 218-m-tall high-rise building during the 1994 northridge earthquake
The 218-m-tall building (denoted by BLD hereafter) shown in Fig. 12 is a 52-story office building, which is currently the 9th tallest
building in Los Angeles. The structural system of BLD includes a braced core, twelve steel columns, and eight deep outrigger beams at
each floor connecting the inner and outer columns [36]. As illustrated in Fig. 12, BLD has a 52-story superstructure with a 47.6 m ×
47.6 m square footprint and a 5-story basement with a 83.5 m × 80.2 m square footprint. Since the building is located in Los Angeles,
an earthquake-prone region, a monitoring system including 20 accelerometers was installed on BLD by the U.S. Geological Survey to
monitor the building’s dynamic behavior during earthquakes (CESMD 2021). These 20 accelerometers were installed on 7 floors of the
building (i.e., B5 floor, B2 floor, 14th, 22nd, 35th, 49th, and roof levels) and their layouts are shown in Fig. 12.
At 4:31 a.m. on January 17, 1994, a 6.7-magnitude earthquake occurred in the San Fernando Valley region of Los Angeles. The
1994 Northridge earthquake, named after the San Fernando Valley community near its epicenter, was one of the costliest in U.S.
history, with damages estimated at more than $20 billion, and resulted in 57 deaths. The 1994 Northridge had a duration of
approximately 10–20 s, and its peak ground acceleration of 1.82 g was the highest ever instrumentally recorded in an urban area in
North America [37]. During this strong earthquake, the monitoring system instrumented on BLD was triggered to record the ground
motion and the building accelerations, and their time histories are illustrated in Fig. 13. It is shown that the floor accelerations
generally increased along building height. Noteworthily, the acceleration at the roof floor (i.e., 52nd floor) was significantly larger
than that at the 49th floor, which can be attributed to the whiplash effects of high-rise buildings. Fig. 13 also shows that the peak
building acceleration at the roof floor reached 400 cm/s2, providing a rare opportunity to investigate structural dynamic properties of
high-rise buildings under strong earthquake excitations.
10
K. Zhou and Q.-S. Li Journal of Building Engineering 52 (2022) 104373
N × 2 = 14, and the upper limit is nmax = nmin × 2 = 28. Thus, the value ranges for the window length parameter l and system order n
can be determined as [12, 24] and [14, 28], respectively. By performing the Monte Carlo simulation, s = 100 sets of [l, n] satisfying the
conditions that l ∈ [12, 24] and n ∈ [14, 28] can be generated. Then, based on the structural responses and the earthquake excitation
shown in Fig. 13, as well as the 100 sets of key parameters, 100 groups of structural modal parameters, i.e., [Ω1 , Ω2 , …, Ω100 ], can be
obtained by performing the N4SID approach.
Next, the proposed two-stage stability check approach is employed and the corresponding Monte Carlo-based stabilization diagram
is obtained as illustrated in Fig. 14(b). The conventional stabilization diagram for this case is also plotted in Fig. 14(c) for comparison
purposes. As expected, a number of spurious modes are misidentified as stable poles in the conventional stabilization diagram, which
may lead to the adverse effects on the accuracy of the modal identification results. While in the Monte Carlo-based stabilization di
agram, seven structural modes are readily identified and no spurious modes are misidentified as the stable poles. Following the same
procedure, based on the structural responses and the earthquake excitation along the Y-direction, the Monte Carlo-based stabilization
diagram is established as shown in Fig. 15. The natural frequencies and damping ratios of the first six modes of BLD can be plotted
(Fig. 16) based on the stable poles identified from Figs. 14 and 15. Based on this, the K-means clustering method is employed and the
structural modal parameters can be readily identified as provided in Fig. 16.
Next, the structural dynamic properties of BLD at different time instants are estimated, as illustrated in Fig. 17. The results show
that, after the earthquake reached its peak ground acceleration, the natural frequencies of the six modes of BLD generally showed an
increasing trend, while the damping ratios generally remained unchanged. The increasing trend of natural frequencies of BLD can be
explained by the stick-slip model proposed by [38]. According to the stick-slip model, the contact surfaces between structural members
slip with building vibration, thereby weakening structural stiffness. In general, the number of slipping contact surfaces may increase
with the amplitude of building vibration [39]. As shown in Fig. 13, the building acceleration gradually decreased after the earthquake
reached its peak ground acceleration. This may lead to a reduction in the number of slipping contact surfaces, and thereby resulting in
11
K. Zhou and Q.-S. Li Journal of Building Engineering 52 (2022) 104373
Fig. 13. Ground motion and building accelerations recorded on BLD during the 1994 Northridge earthquake (Data source: https://www.strongmotioncenter.org/
index.html, last visited on December 15, 2021).
12
K. Zhou and Q.-S. Li Journal of Building Engineering 52 (2022) 104373
13
K. Zhou and Q.-S. Li Journal of Building Engineering 52 (2022) 104373
14
K. Zhou and Q.-S. Li Journal of Building Engineering 52 (2022) 104373
Fig. 16. Natural frequencies and damping ratios of the identified stable poles (the first six modes).
Fig. 17. Time-varying structural modal parameters of BLD during the 1994 Northridge earthquake.
15
K. Zhou and Q.-S. Li Journal of Building Engineering 52 (2022) 104373
Fig. 18. Comparison between the predicted and measured accelerations at the roof floor of BLD.
response. Thus, the variation of structural modal parameters with the RMS acceleration is revealed as illustrated in Fig. 19. The results
show that the natural frequencies generally decreased with the amplitude of building accelerations, which is reasonable and can be
well explained by the stick-slip model. The increase of building acceleration amplitude led to the increase of slipping contact surfaces
inside the building, which thereby resulted in the decrease of structural natural frequencies. While for damping ratio, the results show
that it did not change significantly with the building acceleration amplitude. The observation is consistent with some previous studies
Fig. 19. Amplitude-dependency of structural modal parameters of BLD during the 1994 Northridge earthquake.
16
K. Zhou and Q.-S. Li Journal of Building Engineering 52 (2022) 104373
on tall buildings under typhoon conditions (e.g. [44,45], and earthquake excitations (e.g., [14]. In summary, the above presented
results and associated analysis illustrate that the proposed algorithm is an effective structural modal identification tool.
5. Conclusions
This paper proposed an improved methodology based on the N4SID for accurate estimation of modal parameters of high-rise
buildings under earthquake excitations. Firstly, the value ranges for two parameters in performing the N4SID (i.e., system order
parameter n and window length parameter l) were determined based on the PSD analysis of structural dynamic responses. Next, several
sets of random integers [n, l] were generated by employing the Monte Carlo simulation, and the N4SID was accordingly performed for
estimating structural modal parameters. Then, a two-stage stability check approach was employed to discriminate the estimated re
sults denoting the physical modes from those representing spurious modes, and the corresponding Monte Carlo-based stabilization
diagram can be established. Finally, based on the established diagram, the K-means clustering method was adopted to determine the
structural modal parameters.
Both the numerical simulation study and the investigation based on the field measurements on a high-rise building during a strong
earthquake event were performed to verify the efficiency and accuracy of the proposed improved methodology. From the numerical
simulation of a 3DOF model, the performance of the Monte Carlo-based stabilization diagram for discriminating the poles denoting the
physical modes from those representing spurious modes was verified. The numerical simulation results demonstrated that the pro
posed method can estimate the time-varying structural modal parameters with high accuracy and stability. As for the field mea
surement application, the time-varying structural dynamic properties of a 218-m-tall high-rise building during the 1994 Northridge
earthquake were determined by the proposed method. The analyzed results of the field measurements indicated that the natural
frequencies of the six modes of the monitored building generally showed an increasing trend after the earthquake reached its peak
ground acceleration. The amplitude-dependency of structural modal parameters was also revealed accordingly. It was found that the
natural frequencies generally decreased with the amplitude of building accelerations, while the damping ratios generally remained
unchanged. The results of this study verified that the proposed algorithm is applicable and effective for estimating structural modal
parameters of high-rise buildings under earthquake excitations.
Author statement
Kang Zhou: Conceptualization, Methodology, Software, Data Curation, Writing - Original Draft, Writing - Review & Editing,
Investigation, Visualization; Qiu-Sheng Li: Supervision; Funding acquisition; Writing - Review & Editing.
Acknowledgment
The work described in this paper was fully supported by a grant from the Research Grants Council of Hong Kong Special
Administrative Region, China (Project No. CityU 11207519) and a grant from the National Natural Science Foundation of China
(Project No. 51778554).
References
[1] Y. Minami, S. Yoshitomi, I. Takewaki, System identification of super high-rise buildings using limited vibration data during the 2011 Tohoku (Japan)
earthquake, Struct. Control Health Monit. 20 (11) (2013) 1317–1338.
[2] Y.C. Ni, F.L. Zhang, Fast Bayesian frequency domain modal identification from seismic response data, Comput. Struct. 212 (2019) 225–235.
[3] J.D. Chen, C.H. Loh, Tracking modal parameters of building structures from experimental studies and earthquake response measurements, Struct. Health Monit.
16 (5) (2017) 551–567.
[4] F. Gara, D. Arezzo, V. Nicoletti, S. Carbonari, Monitoring the modal properties of an RC school building during the 2016 Central Italy Seismic Swarm, J. Struct.
Eng. 147 (7) (2021), 05021002.
[5] Y. Li, S.T. Mau, A case study of MIMO system identification applied to building seismic records, Earthq. Eng. Struct. Dynam. 20 (11) (1991) 1045–1064.
[6] K. Zhou, Q.S. Li, Effects of time-variant modal frequencies of high-rise buildings on damping estimation, Earthq. Eng. Struct. Dynam. 50 (2) (2021) 394–414.
[7] J. Kim, J.P. Lynch, Subspace system identification of support-excited structures—part I: theory and black-box system identification, Earthq. Eng. Struct. Dynam.
41 (15) (2012) 2235–2251.
[8] D.M. Siringoringo, Y. Fujino, Observed dynamic performance of the Yokohama-Bay Bridge from system identification using seismic records, Struct. Control
Health Monit. 13 (1) (2006) 226–244.
[9] J.L. Beck, P.C. Jennings, Structural identification using linear models and earthquake records, Earthq. Eng. Struct. Dynam. 8 (2) (1980) 145–160.
[10] E. Şafak, Identification of linear structures using discrete-time filters, J. Struct. Eng. 117 (10) (1991) 3064–3085.
[11] J.L. Beck, System identification methods applied to measured seismic response, in: Proceedings of 11th World Conference on Earthquake Engineering, 1996.
[12] C.H. Loh, C.Y. Lin, C.C. Huang, Time domain identification of frames under earthquake loadings, J. Eng. Mech. 126 (7) (2000) 693–703.
[13] S.B. Kim, B.F. Spencer Jr., C.B. Yun, Frequency domain identification of multi-input, multi-output systems considering physical relationships between measured
variables, J. Eng. Mech. 131 (5) (2005) 461–472.
[14] C. Cruz, E. Miranda, Evaluation of damping ratios for the seismic analysis of tall buildings, J. Struct. Eng. 143 (1) (2017), 04016144.
[15] P. Mellinger, M. Döhler, L. Mevel, Variance estimation of modal parameters from output-only and input/output subspace-based system identification, J. Sound
Vib. 379 (2016) 1–27.
[16] H. Shokravi, H. Shokravi, N. Bakhary, M. Heidarrezaei, S.S. Rahimian Koloor, M. Petrů, Application of the subspace-based methods in health monitoring of civil
structures: a systematic review and meta-analysis, Appl. Sci. 10 (10) (2020) 3607.
[17] V. Sima, Efficient data processing for subspace-based multivariable system identification, IFAC Proceed. Vol. 37 (12) (2004) 871–876.
17
K. Zhou and Q.-S. Li Journal of Building Engineering 52 (2022) 104373
[18] R. Ceravolo, G. De Lucia, E. Matta, A. Quattrone, L. Zanotti Fragonara, Equivalent modal parameters in monitored buildings during the recent Italian seismic
events, in: Proceedings of 17th ANIDIS Conference on Earthquake Engineering, 2017.
[19] R. Dhakal, System Identification and Seismic Analysis of a Residential RC Building in Kathmandu, Nepal (Doctoral dissertation, 2020.
[20] W.H. Wu, S.W. Wang, C.C. Chen, G. Lai, Modal parameter identification for closely spaced modes of civil structures based on an upgraded stochastic subspace
methodology, Struct. Infrastruct. Eng. 15 (3) (2019) 296–313.
[21] F. Magalhaes, A. Cunha, E. Caetano, Online automatic identification of the modal parameters of a long span arch bridge, Mech. Syst. Signal Process. 23 (2)
(2009) 316–329.
[22] L. Su, J.Q. Zhang, X. Huang, J.M. LaFave, Automatic operational modal analysis of structures based on image recognition of stabilization diagrams with
uncertainty quantification, Multidimens. Syst. Signal Process. 32 (1) (2021) 335–357.
[23] F. Ubertini, C. Gentile, A.L. Materazzi, Automated modal identification in operational conditions and its application to bridges, Eng. Struct. 46 (2013) 264–278.
[24] W.E. Larimore, System identification, reduced-order filtering and modeling via canonical variate analysis, in: Proceedings of 1983 American Control
Conference, 1983.
[25] M. Verhaegen, Subspace model identification Part 3: analysis of the ordinary output-error state-space model identification algorithm, Int. J. Control 58 (3)
(1993) 555–586.
[26] P. Van Overschee, B. De Moor, N4SID: subspace algorithms for the identification of combined deterministic-stochastic systems, Automatica 30 (1) (1994) 75–93.
[27] M. Chang, S.N. Pakzad, Observer Kalman filter identification for output-only systems using interactive structural modal identification toolsuite, J. Bridge Eng.
19 (5) (2014), 04014002.
[28] K.Y. Koo, S.G. Cho, J. Cui, D. Kim, Seismic response prediction for cabinets of nuclear power plants by using impact hammer test, Nucl. Eng. Des. 240 (10)
(2010) 2500–2511.
[29] K. Nakajima, N. Giron, M. Kohiyama, K. Watanabe, M. Yoshida, M. Yamanaka, O. Yoshida, Seismic response analysis of a semi-active-controlled base-isolated
building during the 2011 great east Japan earthquake, in: Proceedings of International Symposium on Engineering Lessons Learned from the 2011 Great East
Japan Earthquake, 2012.
[30] S. Tuhta, F. Günday, H. Aydin, Numerical algorithms for system identification of benchmark steel structure, in: Proceedings of International Conference on
Innovation, Sustainability, Technology and Education in Civil Engineering, 2019.
[31] Y. Xiang, F. Naeim, F. Zareian, Evaluation of natural periods and modal damping ratios for seismic design of building structures, Earthq. Spectra 36 (2) (2020)
629–646.
[32] Y.C. Liu, C.H. Loh, Y.Q. Ni, Stochastic subspace identification for output-only modal analysis: application to super high-rise tower under abnormal loading
condition, Earthq. Eng. Struct. Dynam. 42 (4) (2013) 477–498.
[33] C.C. Lin, C.E. Wang, H.W. Wu, J.F. Wang, On-line building damage assessment based on earthquake records, Smart Mater. Struct. 14 (3) (2005) S137.
[34] A. Masuda, A. Sone, Identification of structural degradation by time-frequency system analysis, in: Proceedings of 12th World Conference on Earthquake
Engineering, 2000.
[35] A. Likas, N. Vlassis, J.J. Verbeek, The global k-means clustering algorithm, Pattern Recogn. 36 (2) (2003) 451–461.
[36] C. Ventura, B. Laverick, R. Brincker, P. Andersen, Comparison of dynamic characteristics of two instrumented tall buildings, in: Proceedings of IMAC-21:
Conference on Structural Dynamics, 2003.
[37] N. Luco, Seismic design and hazard maps: before and after, Struct. Mag. (2019) 28–30.
[38] A.G. Davenport, P. Hill-Carroll, Damping in tall buildings: its variability and treatment in design, in: Proceedings of ASCE Spring Convention, 1986.
[39] Y. Tamura, Amplitude dependency of damping in buildings and critical tip drift ratio, Int. J. High-Rise Build. 1 (1) (2012) 1–13.
[40] A.P. Jeary, Damping in structures, J. Wind Eng. Ind. Aerod. 72 (1997) 345–355.
[41] Q.S. Li, J.Q. Fang, A.P. Jeary, C.K. Wong, D.K. Liu, Evaluation of wind effects on a supertall building based on full-scale measurements, Earthq. Eng. Struct.
Dynam. 29 (12) (2000) 1845–1862.
[42] Y. Tamura, A. Yoshida, Amplitude dependency of damping in buildings, in: Proceedings of Structures Congress 2008: 18th Analysis and Computation Specialty
Conference, 2008.
[43] X. Li, Q.S. Li, Monitoring structural performance of a supertall building during 14 tropical cyclones, J. Struct. Eng. 144 (10) (2018), 04018176.
[44] Q.S. Li, K. Zhou, X. Li, Damping estimation of high-rise buildings considering structural modal directions, Earthq. Eng. Struct. Dynam. 49 (6) (2020) 543–566.
[45] K. Zhou, Q.S. Li, Investigation of time-varying structural dynamic properties of high-rise buildings under typhoon conditions, J. Build. Eng. 46 (2022), 103796.
[46] M. He, P. Liang, E. OBrien, X. Sun, Y. Zhang, Continuous modal identification and tracking of a long-span suspension bridge using a robust mixed-clustering
method, J. Bridge Eng. 27 (3) (2021), 05022001.
18