applied
sciences
Article
Modal Parameter Identification of a Structure Under Earthquake
via a Wavelet-Based Subspace Approach
Wei-Chih Su 1 , Liane-Jye Chen 2 and Chiung-Shiann Huang 2, *
Citation: Su, W.-C.; Chen, L.-J.;
Huang, C.-S. Modal Parameter
Identification of a Structure Under
Earthquake via a Wavelet-Based
Subspace Approach. Appl. Sci. 2024,
14, 2503. https://doi.org/10.3390/
app14062503
Academic Editors: Nan Wu,
Hossein Bisheh and Yen-Fang Su
Received: 19 February 2024
Revised: 9 March 2024
Accepted: 12 March 2024
Published: 15 March 2024
Copyright: © 2024 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Appl. Sci. 2024, 14, 2503. https://doi.org/10.3390/app14062503
1 AI Application & Development Center, Mitac Advanced Technology Corp., Taipei 11493, Taiwan;
wichcv86@mitc.com.tw
2 Department of Civil Engineering, National Yang Ming Chiao Tung University, Hsinchu 30050, Taiwan;
linkchen.en03@nycu.edu.tw
* Correspondence: cshuang@nycu.edu.tw
Abstract: This paper introduces a novel wavelet-based methodology for identifying the modal
parameters of a structure in the aftermath of an earthquake. Our proposed approach seamlessly
combines a subspace method with a stationary wavelet packet transform. By relocating the subspace
method into the wavelet domain and introducing a weighting function, complemented by a moving
window technique, the efficiency of our approach is significantly augmented. This enhancement
ensures the precise identification of the time-varying modal parameters of a structure. The capacity of
the stationary wavelet packet transform for rich signal decomposition and exceptional time-frequency
localization is harnessed in our approach. Different subspaces within the stationary wavelet packet
transform encapsulate signals with distinct frequency sub-bands, leveraging the fine filtering property
to not only discern modes with pronounced modal interference, but also identify numerous modes
from the responses of a limited number of measured degrees of freedom. To validate our methodology,
we processed numerically simulated responses of both time-invariant and time-varying six-floor
shear buildings, accounting for noise and incomplete measurements. Additionally, our approach was
applied to the seismic responses of a cable-stayed bridge and the nonlinear responses of a five-story
steel frame during a shaking table test. The identified modal parameters were meticulously compared
with published results, underscoring the applicability and reliability of our approach for processing
real measured data.
Keywords: modal parameter identification; stationary wavelet packet transform; subspace method;
time-varying system; earthquake responses
1. Introduction
The effective health monitoring of critical infrastructure, such as buildings and bridges,
is of paramount importance to ensure their longevity, provide high-quality service over
their life cycle, and mitigate the potential loss of human lives and economic assets resulting
from structural failures, especially during seismic events [1,2]. The utilization of measured
modal parameters, encompassing natural frequencies, modal damping ratios, and mode
shapes, has become a widely adopted and straightforward approach in structural health
monitoring [3–5]. These parameters not only facilitate damage detection, but also find
applications in tasks such as updating finite element models to match real behavior and
implementing vibration control strategies.
The seismic response monitoring of critical infrastructure in earthquake-prone regions
poses significant challenges. The accurate identification of modal parameters from mea-
sured seismic responses is essential for understanding structural behavior and its potential
vulnerabilities. A substantial body of the literature has explored the modal parameter
identification of structures under seismic excitation [6–8]. In this context, two primary
categories of approaches, time series methods and subspace methods, have been frequently
employed in the time domain or time–frequency domain.
https://www.mdpi.com/journal/applsci
Appl. Sci. 2024, 14, 2503 2 of 22

The time series models include autoregressive models with exogenous input (ARX)
and autoregressive moving average models with exogenous input (ARMAX). For instance,
Safak and Celebi [9] and Loh and Lin [10] employed ARX models to extract modal pa-
rameters of buildings through least-squares estimations of the coefficient matrices of ARX
models in the time domain. In the time–frequency domain, Huang et al. [11] and Huang
and Su [12] introduced discrete and continuous wavelet transforms into ARX models, re-
spectively. Using an ARMAX model, Loh and Lin [10] and Satio and Yokota [13] identified
modal parameters of buildings through the estimation of coefficient matrices of ARMAX
models using the maximum likelihood method and the differential iteration method in the
time domain, respectively, while Moore et al. [14,15] determined the modal parameters of a
structure with incompletely measured base excitations through an iterative multi-stage esti-
mation of coefficient matrices. Dziedziech et al. [16] proposed recursive ARMAX models for
identifying instantaneous modal parameters of a time-variant system. Various approaches
based on time-varying ARX [17–19] have been employed to determine instantaneous modal
parameters of time-variant or nonlinear systems. Notably, ARMAX modeling is more time
consuming than ARX modeling due to the involvement of unmeasured white noise.
Subspace methods are classified as deterministic, stochastic, and combined deterministic–
stochastic subspace identification techniques [20,21]. Stochastic subspace identification (SSI)
approaches, which include covariance-driven [22,23] and data-driven [24,25] SSI algorithms
according to the type of data matrices used in the algorithms, have been often employed
to process output-only stationary responses, like ambient vibrations. Comprehensive
literature reviews on SSI approaches in terms of methodology and engineering applications
can be found in the review papers by Perez-Ramirez et al. [7] and Shokraviu et al. [8].
Although earthquake responses, theoretically, do not fit stochastic state-space models,
Liu et al. [26] applied a recursive covariance-driven SSI approach to identify the natural
frequencies and mode shapes of a tower from its earthquake responses pre-processed
using recursive singular spectrum analysis. Pioldi and Rizzi [27] presented an improved
data-driven SSI approach to process the earthquake responses of structures. Their identified
modal parameters for a three-story shear building through the processing of numerically
simulated earthquake responses without noise yielded the maximum errors of up to 14%
and 70% in frequencies and modal damping ratios, respectively, and the values of the
modal assurance criterion (MAC) [28] decreased to 0.4.
While output-only approaches offer practical benefits by not requiring input dynamic
loadings and saving costs on monitoring input loadings, input–output methods gener-
ally yield more accurate modal parameter estimates [29–31]. Consequently, deterministic
subspace identification (DSI) methods and combined deterministic–stochastic subspace
identification (CDSSI) methods are much preferred to SSI approaches for processing the
earthquake responses of a structure. These two classes of methods were first proposed
by Moonen et al. [32] and Van Overschee and De Moor [33], respectively. To enhance the
efficiency of these original subspace approaches, an instrumental variable [34], a gener-
alized Schur algorithm [35], empirical mode decomposition [36,37], a covariance-driven
algorithm [38], and nuclear norm optimization [39] have been incorporated into subspace
methods. To expand the application fields of subspace approaches from a time-invariant
system to a time-variant system, the moving window technique [40,41] and recursive
update algorithms [41–45] have been adopted.
Capitalizing on the advantages of wavelet transformations, this study presents a
novel time–frequency subspace approach by integrating deterministic subspace method
with the stationary wavelet packet transform (SWPT). Additionally, a Gaussian weighting
function was introduced to collaborate with the moving window technique, ensuring the
high-precision identification of modal parameters in time-varying systems. The proposed
approach excels in extracting modal parameters mode-by-mode with precision, leveraging
signals in various subspaces within the stationary wavelet packet transform. This capa-
bility extends to identifying weakly excited modes. Validation was carried out through
simulated earthquake responses of three six-story shear buildings, considering noise and
Appl. Sci. 2024, 14, 2503 3 of 22

incomplete measurements. The shear buildings were modelled as a time-invariant system,

a periodically varying system, and a sharply varying system, respectively. Two real-world
examples are presented to demonstrate the approach’s effectiveness in processing mea-
sured responses. The first example concerns the earthquake responses of a two-span
cable-stayed 510 m long bridge. The second example concerns the experimental accelera-
tion responses of a five-story steel frame, of which its columns were yielded under a shaking
table test with a large excitation input. The identified modal parameters were compared
with published results and underwent finite element analysis, affirming the accuracy of the
proposed approach.

2. Methodology
2.1. Equations of Motion for a Structure under Multiple Support Excitation
When a structure undergoes prescribed motions at its different supports, the total
displacement vector xt of the degrees of freedom within the structure, excluding the
supports, is decomposed into two parts [46]:

xt =xs +x. (1)

According to Chopra [46], x and xs satisfy the following equations, respectively,

.. . .. g
Mx+Cx+Kx= − MΓx , (2)

and
xs =Γxg , (3)
where M, C, and K are the mass, damping, and stiffness matrices, respectively; Γ is the
influence matrix that describes the impact of prescribed support displacement vector xg on
the structural displacement.
Equation (2) can be discretized to the following state equation in terms of state variable
.T
z=(x x )T [47]:
T

zk+1 = Azk + B fk , (4)

.. g
where zk =z( k∆t ), fk = −MΓx ( k∆t ), and ∆t is a time increment. In employing
Equations (1)–(3), the observation equation for the total acceleration responses at mea-
sured degrees of freedom is expressed as

yk = Ezk , (5)

where E = L −M−1 K −M−1 C , and L is an observation matrix of measured degrees of

 

freedom, with components equal to 0 or 1.

In the case of uniform base excitation, Γ becomes a column vector with all components
equal to 1, and xg has only one component. In such instances, relative acceleration responses
are commonly employed. As a result, Equation (5) is adjusted to

yk = Ezk + Dfk , (6)

where D = LM−1 .
Equations (4) and (6) form the discrete-time deterministic state-space model that serves
as the foundation for the proposed wavelet-based subspace identification approach in this
study. It is assumed that the structure has a total of N degrees of freedom, with m of them
being measured. As a result, vectors zk and yk have 2N and m components, respectively.

2.2. Stationary Wavelet Packet Transform

It is widely recognized that the wavelet packet transform offers a superior time–
frequency localization of signals compared to the discrete wavelet transform. The SWPT
exhibits advantages over the wavelet packet transform in two key aspects. Firstly, the SWPT
Appl. Sci. 2024, 14, 2503 4 of 22

is a translation-invariant transformation, providing excellent computational efficiency for

functions with time shifts. Additionally, the SWPT eliminates the need for down-samplers
and up-samplers present in the wavelet packet transform [48].
(0)
The SWPT decomposes a measured signal g(t) in space U0 into its subspaces as
follows (Figure 1):

(0) (0) (1) (0) (1) (2) (3)

U0 = U1 ⊕ U1 = U2 ⊕ U2 ⊕ U2 ⊕ U2 =
(7)
(0) (1) (2) (2k −2) (2k −1)
· · · = Uk ⊕ Uk ⊕ Uk ⊕ · · · ⊕ Uk ⊕ Uk ,

where ⊕ represents the union of two subspaces; subscript k denotes the level of decomposi-
(i ) ( j) (2( k −1) − n ) (2k −2n+1) (2k −2n)
tion, and Uk ⊥Uk for i ̸= j. Space Uk−1 is decomposed into Uk and Uk
(m)
with n ≤ k. Thesubspace m at level k, Uk , is expanded through the set of orthonormal
basis functions µm,k,l (t), l ∈ Z , where

t − l∆t
 
1
µm,k,l (t) = √ µm , (8)
2k 2k
1
and ∆t is the sampling rate for the signal under consideration. The basis functions are con-
structed by selecting a mother wavelet function and its corresponding scale function [49,50].
The measured signal g(t) at the kth level of decomposition, which corresponds to the sub-
(m)
spaces Uk with m = 0, 1, 2, . . ., 2k − 1, is expressed as

2k −1
g(t) = ∑ ∑ a(m, k, l )µm,k,l (t), (9)
m =0 l =0

(m)
where a(m, k, l ) are the stationary wavelet packet coefficients in subspace Uk . Using the
property of translation invariance yields

Appl. Sci. 2024, 14, x FOR PEER REVIEW 2k −1 5 of 24

g(t − τ∆t) = ∑ ∑ a(m, k, l − τ )µm,k,l (t). (10)
m =0 l =0

Figure 1.
Figure Decompositionofofsubspace.
1. Decomposition subspace.

The frequency range of the signal decomposed into subspace U k( m ) is dictated by the
frequency range of the basis functions of U k( m ) . In this study, the Meyer mother wavelet
ψ ( t ) and its corresponding scale function φ ( t ) [51], as depicted in Figure 2a, with their
Fourier spectra ψˆ (ω) and φˆ(ω ) , shown in Figure 2b, are utilized to construct the basis
 functions of the subspaces. As illustrated in Figure 2b, ψ ( t ) and φ ( t ) function as a
band-pass and low-pass filter, respectively, with their frequency ranges partly overlap-
ping.
It is advantageous to choose a suitable subspace for modal parameter identification
Appl. Sci. 2024, 14, 2503 if distinct, main, and preserved frequency ranges are designated for each subspace at the
5 of 22
same level of decomposition, ensuring non-overlapping frequency ranges between differ-
ent subspaces. To facilitate this, the main preserved frequency ranges for ψ ( t ) and φ ( t )
(m)
The frequency
are defined range
as [0.5, 1] of the
Hz and [0,signal decomposed
0.5] Hz, into(Figure
respectively 2b).UThe
subspace k is of ψ
dictated
values ˆ (ωthe
by )
(m)
frequency range of the basis functions of Uk . In this study, the Meyer mother wavelet ψ(t)
and φˆ (corresponding
and its ω ) exhibit largerscalevalues
function ϕ(t50
than 2 %as within
) [51], depictedthese two frequency
in Figure ranges,
2a, with their re-
Fourier
spectra ψ̂(As
spectively. ω ) aand ϕ̂(ω
result, ) , shown
Figure in Figurethe
3 illustrates 2b,main
are utilized
preserved to construct
frequencythe basis functions
ranges in differ-
of the subspaces. As illustrated
ent subspaces applicable in this context.in Figure 2b, ψ ( t ) and ϕ ( t ) function as a band-pass and
low-pass filter, respectively, with their frequency ranges partly overlapping.

(a) (b)

Figure 2.
Figure (a) Plots
2. (a) Plots for
for Meyer
Meyer mother wavelet ϕφ(t)t and
mother wavelet and ()
corresponding scale
corresponding function
scale ψ(tψ
function ); (b)
t ;Fourier
(b) ()
spectra of Meyer mother wavelet and scale function.
Fourier spectra of Meyer mother wavelet and scale function.
It is advantageous to choose a suitable subspace for modal parameter identification
if distinct, main, and preserved frequency ranges are designated for each subspace at the
same level of decomposition, ensuring non-overlapping frequency ranges between different
subspaces. To facilitate this, the main preserved frequency ranges for ψ(t) and ϕ(t) are
defined as [0.5, 1] Hz and [0, 0.5] Hz, respectively (Figure 2b). The values of ψ̂(ω ) and
√
Appl. Sci. 2024, 14, x FOR PEER REVIEWϕ̂ (ω ) exhibit larger values than 50 2% within these two frequency ranges, respectively.
6 of 24
As a result, Figure 3 illustrates the main preserved frequency ranges in different subspaces
applicable in this context.

Figure
Figure 3. 3. Main
Main preserved
preserved frequency
frequency ranges
ranges in in different
different subspaces.
subspaces.

2.3. Formulation for Modal Parameter Identification

The subspace-based method with instrumental variables, initially proposed by
Viberg et al. [34], was restructured to collaborate with the SWPT. The application of the
SWPT to Equations (4) and (6) results in
2k −1 2k −1 2k −1

m , k ,l + 1 m , k ,l ( )
 Z μ t = A   Z m , k ,l μ m , k , l ( t )+ B   Fm , k , l μ m , k , l ( t ) , (11)
m=0 l=0 m=0 l=0 m=0 l=0
Appl. Sci. 2024, 14, 2503 6 of 22

2.3. Formulation for Modal Parameter Identification

The subspace-based method with instrumental variables, initially proposed by Viberg
et al. [34], was restructured to collaborate with the SWPT. The application of the SWPT to
Equations (4) and (6) results in

2k −1 2k −1 2k −1
∑ ∑ Zm,k,l+1 µm,k,l (t) = A ∑ ∑ Zm,k,l µm,k,l (t) + B ∑ ∑ Fm,k,l µm,k,l (t), (11)
m =0 l =0 m =0 l =0 m =0 l =0

2k −1 2k −1 2k −1
∑ ∑ Ym,k,l µm,k,l (t) = E ∑ ∑ Zm,k,l µm,k,l (t) + D ∑ ∑ Fm,k,l µm,k,l (t), (12)
m =0 l =0 m =0 l =0 m =0 l =0

where Z ∞
(Zm,k,l , Ym,k,l , Fm,k,l ) = (z(t), y(t), f(t))µm,k,l (t) dt. (13)
−∞
It is noteworthy that in scenarios involving multiple support excitations, matrix D
should be nullified. Due to the orthonormality of the basis functions, Equations (11) and (12)
can be further simplified to Equations (14) and (15), respectively.

Zm,k,l +1 = AZm,k,l + BFm,k,l . (14)

Ym,k,l = EZm,k,l + DFm,k,l . (15)

To conduct modal identification in the wavelet domain, all the measured responses
undergo transformation into the wavelet
h domain.
 When utilizing the responses of a
structure within the time interval t ∈ l∆t, l + 2L ∆t] for modal parameter identification,
h
Ym,k,l and Fm,k,l with l ∈ l, l + 2L] are employed to construct Equations (14) and (15). The
parameters identified are then hassociated
 with
 a linear system equivalent to the actual
system within the window t ∈ l∆t, l + 2L ∆t] when dealing with a nonlinear or time-
varying structural system.
The following equation is obtained from Equation (14) and (15):
s
Ym,k,l +s = EAs Zm,k,l + DFm,k,l +s + ∑ EAi−1 BFm,k,l +s−i−1 . (16)
i =1

Applying Equation (16) and following the methodology outlined by Viberg [52] for time-
domain signal processing leads to the derivation of Equation (17).

Ŷl = Γs Ẑl + Hs F̂l , (17)

where Γs is the extended observability matrix, and

  T T
Γs = ET (EA) T
··· EA s −1 , (18)

where superscript T denotes the transpose operation of a matrix, Hs is the Toeplitz matrix,
and
D 0 0 ··· 0
 
 EB
 D 0 ··· 0 
Hs = 
 EAB EB D ··· 0 , (19)
 .. 
 . 
EAs−1 B EAs−2 B · · · D
 
Ŷl = Yl Yl +1 · · · Yl + N −1 , (20)
 
Ẑl = Zl Zl +1 · · · Zl + N −1 , (21)
Appl. Sci. 2024, 14, 2503 7 of 22

 
F̂l = Fl Fl +1 · · · Fl + N −1 , (22)
 T
Yl = YTm,k,l YTm,k,l +1 · · · YTm,k,l +s−1 , (23)

¯  T
F l = FTm,k,l FTm,k,l +1 · · · FTm,k,l +s−1 . (24)

To deal with the time-varying system, the present approach introduced a weight-
ing function, as expressed in Equation (25), to precisely
h identify
 the
 modal parameters
corresponding to the midpoint of the time window t ∈ l∆t, l + 2L ∆t] :

2
w(el ) = e−((lm −l )∆t/d) , (25)
e

h h
l ∈ l, l + 2L] , and d is a specified parameter.
where lm is the middle point of l, l + 2L] , e
The following are now defined:
⌣ h i
Y l = w ( l )Yl w ( l + 1 )Yl +1 · · · w ( l + N − 1 )Yl + N −1 , (26)

⌣ h i
Z l = w ( l )Zl w ( l + 1 )Zl +1 · · · w ( l + N − 1 )Zl + N −1 , (27)
⌣ h i
F l = w ( l )Fl w ( l + 1 )Fl +1 · · · w ( l + N − 1 )Fl + N −1 . (28)

Equation (29) can be readily derived from Equation (17) through elementary matrix multi-
plication operations.
⌣ ⌣ ⌣
Y l = Γ s Z l + Hs F l . (29)
⌣
The measured Ylˆ and F lˆ carry greater weights in formulating Equation (29) as lˆ
approaches the midpoint of the considered window. Multiplying Equation (29) by an
⌣T ⌣ ⌣T −1 ⌣
 
⌣
⊥
orthogonal projection matrix O = I − F l F l F l F l onto the null space of F l leads to

⌣ ⌣
Y l O⊥ = Γ s Z l O⊥ (30)

To mitigate the impact of noise on the estimation of coefficient matrices in

Equations (4) and (6), instrumental variables P [34] and a weighting matrix Wc [53] are
introduced and applied to Equation (30), leading to the following equation:

1⌣ ⊥ T 1 ⌣
Y l O P W c = Γ s Z l O ⊥ PT W c , (31)
N N
where "⌣ #  − 1
Fp 1 2
P= ⌣ , Wc = PO⊥ PT , (32)
Yp N

and p is smaller than l. Performing singular value decomposition on the left-hand side of
Equation (31) yields
1⌣ ⊥ T
Y O P Wc ≈ Un Sn VTn , (33)
N l
where Sn is a diagonal matrix with the n largest singular values, and the columns of
Un and Vn are the corresponding left and right singular vectors, respectively. Notably, in
this study, the smallest diagonal component in Sn was set to be larger than 10−6 times the
largest one. Using Equations (31) and (33) yields

Γ s = Un T
en, (34)
Appl. Sci. 2024, 14, 2503 8 of 22

where  +
e n = Sn VTn 1⌣ ⊥ T
T Z O P Wc , (35)
N l
and the superscript “+” indicates the generalized inverse operation.
e m,k,l = T
Z e n Zm,k,l is introduced, and Equations (14) and (15) are rewritten as

e m,k,l +1 = A
Z eZe m,k,l + BF
e m,k,l , (36)

Ym,k,l = E
eZe m,k,l + DFm,k,l , (37)
where
e =T
A e n−1 , B
e n AT e=T e−
e = ET
e n B, and E 1
n . (38)
Importantly, Equation (38) highlights the similarity between A
e and A. In progress-
ing from the derivations of Equations (17) and (18) based on Equations (14) and (15),
Equation (39) can be derived from Equations (34), (36), and (37).
 T T T
eT e s −1
 
Un = E E
eAe ··· E
eA . (39)

Un is obtained through Equation (33) by utilizing the measured responses and base
excitations. Subsequently, A
e and Ee can be derived through the following procedure,
employing Equation (39). Two submatrices of Un are defined as follows:
 T T T
eT e s −2
 
Un1 = E E
eAe ··· E
eA , (40)

 T T T T
e2 e s −1
 
Un2 = E
eAe E
eA ··· E
eA . (41)

Equations (40) and (41) result in

e = U+ Un2 .
A (42)
n1

To estimate E,
e the following matrix from Un is constructed:
h i
Ωk = E e E e ··· E
eA ek ,
eA (43)

where k is much smaller than s. Then,

h i+
e=Ω I
E k A
e e2
A ··· ek
A . (44)

In this study, k is set equal to 5.

The modal parameters of the structure under consideration are determined from A e
and Ee [47]. When the equations of motion of a structure are discretized and expressed as
Equation (4), the modal parameters of the structure are obtained from the eigenvalues and
eigenvectors of A. As A and A e are similar, they possess the same eigenvalues. Let λ j and φj
be the jth eigenvalue and the corresponding eigenvector of A. e Since λ j is a complex number,
let eλ j ∆t = a j + ib j . Consequently, the pseudo-undamped circular natural frequency and
modal damping ratio of the jth mode are, respectively,
q 
ωj = α2j + β2j and ξ j = −α j ωj , (45)
Appl. Sci. 2024, 14, 2503 9 of 22

      
where α j = 1 2∆t ln a2j + b2j and β j = 1 ∆t tan−1 b j a j . Equations (36) and (37)
indicate that the jth modal shape corresponding to the measured degrees of freedom is
ej = Eφ
φ e j.

2.4.
Appl. Sci. 2024, 14, x FOR PEER REVIEW Procedure of Analysis 10 of 24

A flowchart outlining the procedure of the proposed approach for estimating the
modal parameters of a structure is presented in Figure 4. Given that the proposed method
isisapplicable
applicabletotoboth bothtime-invariant
time-invariant andandtime-varying
time-varying systems, the initial
systems, the initial step involves
step involves
generating
generatinga ascalogram
scalogram forfor
thethe
firstfirst
mode modeof the
of system under
the system consideration
under consideration to determine
to determine
whether
whetherthe thesystem
system exhibits
exhibitstime-varying
time-varying behavior
behavioror not.
or To
not.establish
To establish the level theof decom-
level of decom-
positionk,k,it itis is
position recommended
recommended to set it asitthe
to set integer
as the value
integer of ( ln(
value / f(1 )f s /
off s(ln − 1 ),(2where
ln(f 12))/ln ) − 1),fswhere
isfsthe
is the sample
sample frequency
frequency fordata,
for the the data,
and f1and f 1 approximately
is the is the approximately estimated estimated
frequency frequency
of
the first mode obtained from the scalogram of measured responses.
of the first mode obtained from the scalogram of measured responses. As a result, the As a result, the fun-
damental
fundamentalfrequency resides
frequency withinwithin
resides the preserved frequency
the preserved range of
frequency subspace
range U k( 0 ) , fa-
of subspace
(0)
Uk , fa-
cilitating
cilitatingthe straightforward identification
the straightforward identification of
of modal
modalparameters
parametersfor
forhigher
higher modes
modes through
through the utilization of signals in other
the utilization of signals in other subspaces.subspaces.

Figure 4. A flowchart outlining the procedure of the approach for estimating modal parameters.
Figure 4. A flowchart outlining the procedure of the approach for estimating modal parameters.

The identified modal parameters are associated with the eigenvalues and eigenvec-
tors of A , where A
 is an n×n matrix. Many of these eigenvalues and eigenvectors cor-
respond to spurious modes. The genuine mechanical modes are discerned using the sta-
bilization diagram technique. Stable modal parameters are determined by comparing
Appl. Sci. 2024, 14, 2503 10 of 22

Appl. Sci. 2024, 14, x FOR PEER REVIEW The identified modal parameters are associated with the eigenvalues and eigenvectors 11 of 24
of A, where A is an n × n matrix. Many of these eigenvalues and eigenvectors correspond
e e
to spurious modes. The genuine mechanical modes are discerned using the stabilization
diagram
results technique.
obtained fromStable modal parameters
five consecutive valuesareof sdetermined
(in Equations by comparing results
(23) and (24)), obtained
correspond-
from five consecutive values of s (in Equations (23) and (24)), corresponding
ing to the order of a time series model [54]. The outcomes are considered stable to the order of
if the
a time series model [54]. The outcomes are considered stable if the relative
relative differences in the frequency and damping ratio of the desired mode are less than differences in the
frequency
1% and 10%, and damping ratio
respectively, andof the
the MACdesired mode
values are less
exceed 0.95.than 1% and 10%, respectively,
and the MAC values exceed 0.95.
In the proposed procedure, the range of s (denoted as [smin , smax ] ) needs to be defined
In the proposed procedure, the range of s (denoted as [smin , smax ] ) needs to be defined
before establishing Y l and¯ Fl . The value of N in constructing Equations (17) and (29) is
before establishing
recommended to setYlN and F l . The
Δ t equal to value
2 / f1 ,of N in constructing
ensuring that sufficient Equations (17) are
data points andused
(29) to
is
recommended to set N∆t equal to 2/ f 1 , ensuring that sufficient data pointsh are used to
identify both the first mode and other modes. The value of 2 L in the window [ l , l + 2 L ]
identify both the first mode and other modes. The value of 2 L in the window l, l + 2L] is
is set equal to N + smax .
set equal to N + smax .
3. Numerical Verification
3. Numerical Verification
To
To validate
validate the
the proposed
proposed modal
modal parameter
parameter identification
identification approach
approach forfor structures,
structures, wewe
applied
applied it it to
to process
process thethe acceleration
accelerationresponses
responsesofofthree
threesix-story
six-storyshear
shearbuildings
buildings (Figure
(Figure 5)
5)
in in a numerical
a numerical simulation.
simulation. TheThe first
first building
building waswas characterized
characterized by constant
by constant story
story stiff-
stiffness
ness and modal
and modal damping
damping ratios.
ratios. In contrast,
In contrast, the second
the second building
building featured
featured periodic
periodic varia-
variations
tions in stiffness within the first story and modal damping ratios. The
in stiffness within the first story and modal damping ratios. The third shear building third shear building
differed from the
differed from thefirst
firstbybyincorporating
incorporating sharply
sharply time-varying
time-varying stiffnesses
stiffnesses within
within the first
the first and
and third stories and modal damping ratios. All three structures experienced
third stories and modal damping ratios. All three structures experienced identical base identical
base excitations,
excitations, andequations
and the the equations of motion
of motion werewere
solvedsolved
usingusing the Runge–Kutta
the Runge–Kutta methodmethod
with
with a time step of
a time step of 0.005 s. 0.005 s.

Figure 5. Schematic
Schematic of
of aa six-story
six-story shear building.

Modal parameter
Modal parameter identification
identification becomes
becomes more
more intricate
intricate when
when confronted
confronted with
with chal-
chal-
lenges such
suchasasincomplete
incompletemeasurements
measurements and noise.
and In our
noise. analysis,
In our the proposed
analysis, approach
the proposed ap-
was employed
proach to process
was employed the basethe
to process excitations and acceleration
base excitations responses
and acceleration of the first
responses of and
the
sixth floors, while introducing white noise with a noise-to-signal ratio (NSR) of 10%.
first and sixth floors, while introducing white noise with a noise-to-signal ratio (NSR) of
10%.
 Appl. Sci. 2024, 14, x FOR PEER REVIEW
Appl. Sci.2024,
2024,14,
14,2503
x FOR PEER REVIEW 12ofof24
12 24
Appl. Sci. 11 of 22

3.1.Time-Invariant
3.1. Time-InvariantSystem
System
3.1. Time-Invariant System
The six-story shear buildingunder
The six-story shear building underconsideration
considerationisischaracterized
characterizedby
bymodal
modaldamping
damping
The six-story shear building under consideration is characterized by modal damping
ratios ( ξ( ξ )i ) ofof 5%, 5%, mm 1==44ton, ton, m 2==33tonton, ,mmi(i(i==3~6)
3~6)==22ton,
ton, 1==18180000 kkNN/ /mm, ,
ratios
ratios (ξ i i) of 5%, m 1 = 14 ton, m2m 2= 3 ton, mi i (i = 3 ∼ 6) = kk2 1 ton, k 1 = 1800 kN/m,
=12
kkk222== 12 00 kN
1200
00 kN and
/ /mm, , and
kN/m, andkki ki(ii(i=(=i33= 6)=∼
~~6)3 =600
600
6)= kN/
kN/ mm,kN/m,
600 ,where
wherewhere𝑚𝑚 and
andm𝑘i𝑘and represent themass
k i represent
represent the mass
the mass and
and
and stiffness
stiffness
stiffness ofof ofthe
the the
ithith
ith story,respectively.
story,
story, respectively.The
respectively. modal
Themodal
modal frequencies
frequencies
frequencies ofofthe
thestructure
of structure
the are0.801,
structure
are 0.801,
are 0.801, 2.14,
2.14,
2.14, 3.15,
3.15,3.15,4.25, 4.25,
5.04,
4.25, 5.04,
5.04,and and
and5.375.37
5.37 Hz. Hz.Notably,
Hz. Notably,the
Notably, thefrequency
the frequencydifference
frequency difference
difference between
between
between thefifth
the fifthfifth
the andand sixth
and
sixth
sixth
modes modes
modes isisless
is less less
than than
than 10%,
10%,
10%, indicatingaaasignificant
indicating
indicating significantdegree
significant degree
degreeofofofmode
mode
mode interference
interference
interference between
between
between these
these
these
two two twomodes.
modes. modes.
In
InIn Figure
Figure 6,6,the
thebase
baseexcitations
excitationsand andacceleration
accelerationresponses
responseswith
withadded
addednoisenoiseforforthe
the
Figure 6, the base excitations and acceleration responses with added noise for the
first
first and
and sixth
sixth floors
floors are illustrated.
are illustrated. The first
The firstThe mode
mode scalogram
scalogram of the responses of the sixth
first and sixth floors are illustrated. first mode of the responses
scalogram of theof the sixth
responses of the
floor,presented
floor, presentedininFigure Figure7a, 7a,indicates
indicatesaatime-invariant
time-invariantsystem.system.The
Thescalogram
scalogramwas wascom-
com-
sixth
putedusing floor,
usingthe presented
theMorlet in
Morletwavelet Figure
wavelettransform. 7a,
transform.indicates a time-invariant system. The scalogram was
puted
computed using the Morlet wavelet transform.

Figure
Figure
Figure6.6.6.
Time
Time histories
histories
Time andand
and
histories Fourier
Fourier spectra
spectra
Fourier ofofthe
spectra thesimulated
simulated
of acceleration
acceleration
the simulated responsesresponses
responses
acceleration ofofthe
thesix-story
six-story
of the six-story
shearbuilding.
shear building.
shear building.

Figure
Figure
Figure 7.7.7.
Scalograms
Scalograms ofofthe
Scalograms the responses
ofresponses
the ofofthe
responses theof
six
six floors
the sixof
floors ofshear
shear
floors ofbuildings:
buildings: (a)time-invariant
(a) time-invariant
shear buildings: system; system;
(a) time-invariant
system;
(b)
(b) periodically
(b)periodically
periodically varying
varying
varying system; (c)sharply
system;
system; (c) sharply varying
(c) sharply
varying system.
varying
system. system.

The
The
The data within
data
data within
within thethe
the time
time interval
time t t==10
interval
interval 10ttoto 20s,to
= 20
10 s,as
as depicted
20depicted ininFigure
s, as depicted Figure in6,6,Figure
underwent
6, underwent
underwent
processing
processing using
processingusing the
usingthe proposed
theproposed
proposed approach.
approach.
approach. Modal
Modal
Modal parameters
parameters
parameters for
for for the
thethe first
firstfirstand
andand second
second modes
second
(1 )
modeswere
modes wereidentified
identifiedusing
usingresponses
responsesininsubspaces
subspaces U (U and (U2U)( 26( )2 ), ,respectively.
1)(16) and respectively.The The
were identified using responses in subspaces U6 6 and U6 6, respectively. The third and
(4)
fourth modes were identified through responses in subspace U6 , while the fifth and
(3)
sixth modes were obtained using responses in subspace U6 . Stabilization diagrams
for the identified frequencies and modal damping ratios, as shown in Figure 8, reveal
very few spurious modes denoted by blue symbols in the preserved frequency ranges of
these subspaces.
 third and fourth modes were identified through responses in subspace U 6( 4 ) , while the
fifth and sixth modes were obtained using responses in subspace U 6( 3 ) . Stabilization dia-
grams for the identified frequencies and modal damping ratios, as shown in Figure 8, re-
Appl. Sci. 2024, 14, 2503 veal very few spurious modes denoted by blue symbols in the preserved frequency ranges 12 of 22
of these subspaces.

Figure
Figure8. 8.
The stabilization
The diagrams
stabilization diagrams of theofidentified frequencies
the identified and damping
frequencies ratio usingratio
and damping different
using different
values of s: (a) using signals in U6(1) ; (b)(1using
) signals in U 6( 2 ) ; (c) using
(2) signals in U 6( 4 ) ; (d) using
(4)
values of s: (a) using signals in U6 ; (b) using signals in U6 ; (c) using signals in U6 ; (d) using
signals in U 6( 3()3).
signals in U6 .
Table 1 summarizes the identified modal parameters, with parenthesized data repre-
Table 1 summarizes the identified modal parameters, with parenthesized data repre-
senting analytical (true) results. The identified results demonstrate excellent agreement
senting analytical (true) results. The identified results demonstrate excellent agreement
with the true values. Frequencies and modal damping ratios exhibit maximum differences
with
of 0.9%the true
and values.
5%, Frequencies
respectively. and modal
The values of MAC damping ratios
exceed 0.97 forexhibit maximum
all identified differences
mode
of 0.9% and 5%, respectively. The values of
shapes in comparison to theoretical mode shapes. MAC exceed 0.97 for all identified mode shapes
in comparison to theoretical mode shapes.
Table 1. Identified results for the time-invariant six-story shear building.
Table 1. Identified results for the time-invariant six-story shear building.
Mode No. 1 2 3 4 5 6
Mode No. 1 0.80 2.14
2 3.15 3 4.24 4 5.00 5 5.44 6
ƒn (Hz)
(0.80) (2.14) (3.15) (4.25) (5.04) (5.37)
0.80 2.14 3.15 4.24 5.00 5.44
ƒn (Hz) 5.19 5.27 5.20 5.02 4.65
ξ (%) (0.80) (2.14) (3.15) (4.25) (5.04)5.22 (5.37)
(5.00) (5.00) (5.00) (5.00) (5.00) (5.00)
5.19 5.27 5.20 5.02 4.65 5.22
(%)
ξMAC 1.00 1.00 1.00 0.99 0.98 0.98
(5.00) (5.00) (5.00) (5.00) (5.00) (5.00)
MAC
3.2. Periodically 1.00System
Varying 1.00 1.00 0.99 0.98 0.98
A shear building under consideration features periodically varying stiffness in the
3.2.story,
first Periodically Varying
and its modal System ratios are described as follows:
damping
A shear building under consideration features periodically varying stiffness in the
first story, and its modal damping ratios are described as follows:

  1800
   0≤t≤8
k1 = and
1800 1 − 0.3 sin (t − 8)π 5 t≥8


  0.05
   0≤t≤8
ξi = . (for i = 1, 2, . . . , 6).
0.05 1 + 0.4 sin (t − 8)π 5 t≥8

The first mode scalogram of the responses from the sixth floor (Figure 7b) illustrates that
the building exhibits characteristics indicative of a time-varying system.
Appl. Sci. 2024, 14, 2503 13 of 22

To process the noisy base excitations and responses of the first and sixth floors, the
SWPT was applied. Responses in different subspaces at decomposition level 6 were utilized
to ascertain the time-varying modal parameters. Modal parameter identification involved
windowed responses with a window length of 2 L, equal to 4 s. In Equation (25), the
parameter ‘d’ in the weighting function was set to 0.06 s, and adjacent windows had an
overlay of 3.9 s.
The identified frequencies and modal damping ratios for all modes are presented in
Figure 9, where results without employing the weighting function (denoted by d = ∞) are
also displayed. Notably, the identified results obtained using the weighting function with
‘d’ set to 0.06 s exhibit excellent agreement with the analytical values. Maximum differences
in frequency were less than 1% for all six modes. Regarding modal damping ratios, the
maximum differences were less than 15% for the first five modes and 20% for the sixth
Appl. Sci. 2024, 14, x FOR PEER REVIEWmode. The results obtained using the weighting function with ‘d’ set to 0.06 s significantly
15 of 24

outperformed those obtained without its use.

Figure
Figure 9. Identified
9. Identified modal
modal parameters
parameters for thefor the periodically
periodically varying varying
system. system.

3.3.Sharply
3.3. Sharply Varying
Varying System
System
The
The six-story
six-story shear
shear building
building examinedexamined
in thisin this section
section is identical
is identical to the onetodiscussed
the one discussed
ininthe
the preceding
preceding section,
section, withwith the exception
the exception of the of the parameters
parameters described
described as follows:
as follows:
 1800 0≤t≤8


1800 0≤t≤8
k1 = 1800 1 − 0.2 ( t − 8 ) 0.005  8 < t < 8.005 ,

k1 = 1800[1 − 0.2(t − 8)/0.005] 8 < t < 8.005 ,
  1152 1152 8.005 ≤ t 8.005 ≤ t
 600 0 ≤ t ≤ 15

k3 = 600 1 − 0.2 ( t − 15 ) 0.005  15 < t < 15.005 ,

 384 15.005 ≤ t
 0.05 0≤t≤8

 0.05 1 + 0.3 ( t − 8 ) 0.005  8 < t < 8.005
 0.065 8.005 ≤ t ≤ 15
ξi =  (for i = 1, 2, …, 6).
  2 
0.065 1 + ( t − 15 ) 0.005  15 < t < 15.005
Appl. Sci. 2024, 14, 2503 14 of 22


 600 0 ≤ t ≤ 15
k3 = 600[1 − 0.2(t − 15)/0.005] 15 < t < 15.005 ,
384 15.005 ≤ t


0≤t≤8


 0.05
 0.05[1 + 0.3(t − 8)/0.005] 8 < t < 8.005


ξi = 0.065 8.005 ≤ t ≤ 15 (for i = 1, 2, . . . , 6).
2
  
0.065 1 + (t − 15) 0.005 15 < t < 15.005



 13
Appl. Sci. 2024, 14, x FOR PEER REVIEW 
0.075 15.005 ≤ t 16 of 24

The first mode scalogram of the sixth-floor responses (Figure 7c) indicates sharp
changes in frequency.
Utilizing the same windows and weighting function outlined in Section 3.2, we pro-
Utilizing the same windows and weighting function outlined in Section 3.2, we pro-
cessed the noisy base excitation and acceleration responses from the first and sixth floors
cessed the noisy base excitation and acceleration responses from the first and sixth floors
within subspaces at level 6 of the SWPT. This methodology yielded time-varying identi-
within subspaces
fied frequencies and at modal
level 6 damping
of the SWPT. ratiosThis methodology
for all six modes, asyielded time-varying
depicted identified
in Figure 10. In
frequenciesthe
employing and modal damping
weighting function withratios
‘d’ for all0.06,
set to six modes, as depicted
the identified results in Figureusing
obtained 10. In em-
ploying
the the weighting
weighting function
function with ‘d’ setwith ‘d’ sset
to 0.06 to 0.06, excellent
exhibited the identified results
agreement withobtained using the
the analyt-
weighting function with ‘d’ set to 0.06 s exhibited excellent agreement
ical values. The maximum differences of the first three modes in frequency were less than with the analytical
values.
4%, whileThe
themaximum
differencesdifferences
were less than of the
2% forfirst three
the fourmodes in modes.
and sixth frequency were less
Regarding than 4%,
modal
damping ratios, the maximum differences were less than 15% for the first sixth mode.modal
while the differences were less than 2% for the four and sixth modes. Regarding
damping
These ratios, the
identified maximum
results markedly differences
outperformedwere less than
those 15% forwithout
obtained the firstthe
sixth mode. These
weighting
identified results markedly outperformed those obtained without the
function, especially in precisely capturing the abrupt changes in the frequency and damp- weighting function,
especially
ing ratio. in precisely capturing the abrupt changes in the frequency and damping ratio.

Figure10.
Figure Identifiedmodal
10.Identified modal parameters
parameters forfor
thethe sharply
sharply varying
varying system.
system.

4. Applications to Real Measured Responses

Having demonstrated the effectiveness of the current approach in handling numeri-
cally simulated responses, we extend its application to process actual measured responses
from a two-span cable-stayed bridge subjected to an earthquake. Additionally, we em-
ployed the approach in the analysis of nonlinear responses from a five-story steel frame
under a shaking table test.
Appl. Sci. 2024, 14, 2503 15 of 22

4. Applications to Real Measured Responses

Having demonstrated the effectiveness of the current approach in handling numeri-
cally simulated responses, we extend its application to process actual measured responses
from a two-span cable-stayed bridge subjected to an earthquake. Additionally, we
Appl. Sci. 2024, 14, x FOR PEER REVIEW em-
17 of 24
ployed the approach in the analysis of nonlinear responses from a five-story steel frame
under a shaking table test.

4.1. Application
4.1. Application to
to aa Two-Span
Two-Span Cable-Stayed
Cable-Stayed Bridge
Bridge
Theconsidered
The consideredtwo-span
two-spanbridge,
bridge,illustrated
illustratedininFigure
Figure11,11,isissituated
situatedononFreeway
Freeway33in in
Taiwanand
Taiwan andfeatures
featuresaa34.5
34.5m mwide
widedeck.
deck. Comprising
Comprising aa 330330 mmlong
longsteel
steelmain
mainspanspanand
andaa
180m
180 mlong
longprestressed
prestressedconcrete
concretesidesidespan,
span,the
thebridge
bridgeisissupported
supportedby byaa183.5
183.5m mtall
tallpylon
pylon
withan
with aninverted
invertedY-shaped
Y-shapedreinforced
reinforced concrete
concrete structure
structure andand a diaphragm
a diaphragm wallwall founda-
foundation.
tion.pylon’s
The The pylon’s
two legs two
arelegs are interconnected
interconnected by a transverse
by a transverse beam. The beam. Theis bridge
bridge is sup-
supported by
60 steel by
ported cables in total,
60 steel ensuring
cables in total,stability
ensuring and structural
stability andintegrity.
structural integrity.

Figure11.
Figure 11. Arrangement
Arrangementof
ofmonitoring
monitoringinstruments
instrumentson
onthe
thecable-stayed
cable-stayedbridge.
bridge.

A
Acomprehensive
comprehensivemonitoring
monitoring system
systemwaswasinstalled to measure
installed the dynamic
to measure responses
the dynamic re-
of the bridge
sponses in the
of the facein
bridge of the
wind and
face ofearthquake forces. A portion
wind and earthquake forces.ofAthis monitoring
portion of thissystem
moni-
istoring
presented
systeminisFigure
presented 11, with other
in Figure 11,sensors placed
with other along
sensors the cables
placed along omitted
the cables from the
omitted
illustration, as they were not utilized in this study. Three-axis accelerometers
from the illustration, as they were not utilized in this study. Three-axis accelerometers (ACC) and
one-axis
(ACC) and velocity meters
one-axis (VEL)
velocity were strategically
meters placed, including
(VEL) were strategically placed,atincluding
locations such as the
at locations
abutment
such as the (ACC 02), within
abutment (ACCthe 02),box girder
within theofbox
the girder
deck (VELs
of the04–15), near the
deck (VELs cablenear
04–15), on the
the
deck
cableslab (VELs
on the deck16–18), atop a
slab (VELs pylonatop
16–18), (VELs 01–03),
a pylon on the
(VELs foundation
01–03), on the (ACCs 03 and
foundation 04),
(ACCs
and at pier
03 and 04),P2
and(ACC 05).P2The
at pier positions
(ACC 05). Theof these sensors
positions are marked
of these sensorsby arered dots inbyFigure
marked 11.
red dots
The proposed
in Figure 11. approach was employed to identify the modal parameters of the bridge
in theThetransverse
proposed direction
approach using
wasearthquake
employed to measurements.
identify the modal Transverse acceleration
parameters mea-
of the bridge
surements from ACCs 03 and 04 were averaged and treated
in the transverse direction using earthquake measurements. Transverse accelerationas the transverse base exci-
tations at P1. Additionally, the transverse acceleration measurements
measurements from ACCs 03 and 04 were averaged and treated as the transverse base of ACCs 02 and 05
were considered
excitations at P1.base excitation the
Additionally, inputs. Transverse
transverse velocitymeasurements
acceleration measurementsof ofACCs
VELs 0202,and
04,
07, 10, and 17 were used as responses. Figure 12 illustrates the measurements
05 were considered base excitation inputs. Transverse velocity measurements of VELs 02, from some of
these sensors along with the corresponding Fourier spectra. The
04, 07, 10, and 17 were used as responses. Figure 12 illustrates the measurements from Fourier spectra of base
excitations clearly indicate that the bridge was subjected to multiple base excitations.
some of these sensors along with the corresponding Fourier spectra. The Fourier spectra
of base excitations clearly indicate that the bridge was subjected to multiple base excita-
tions.
Appl.
Appl.Sci.
Sci.2024,
2024,14,
14,x 2503
FOR PEER REVIEW 18 of 24
16 of 22

Figure12.
Figure Thetime
12.The timehistories
historiesand
andcorresponding
corresponding Fourier
Fourier spectra
spectra of
of transverse
transverse acceleration
acceleration or
or velocity
veloc-
responses
ity responsesat at
different locations
different of of
locations thethe
cable-stayed bridge
cable-stayed under
bridge underananearthquake.
earthquake.

Althoughnot
Although notpresented
presentedhere,
here,the
thefirst
firstmode
modescalogram
scalogramfrom
fromthe
themeasurement
measurementofofVEL VEL
02 suggests that the bridge exhibited linear behavior during the considered
02 suggests that the bridge exhibited linear behavior during the considered earthquake. earthquake.
Datawithin
Data withinthethetime
timeinterval
intervalofof2020toto3030s swere
wereprocessed
processedusing
usingthe
theproposed
proposedapproach,
approach,
identifyingeight
identifying eight modes.
modes. The
The identified
identifiedmodalmodalfrequencies
frequencies and damping
and damping ratios are are
ratios provided
pro-
in Table 2, while the identified modal shapes for the bridge deck are depicted
vided in Table 2, while the identified modal shapes for the bridge deck are depicted in in Figure 13.
Table 2 and Figure 13 also include results obtained from a commercial
Figure 13. Table 2 and Figure 13 also include results obtained from a commercial finite finite element
packagepackage
element based on design
based on data
designand from
data and anfrom
autoregressive (AR) time
an autoregressive series
(AR) timemodel
series applied
model
to the ambient vibration responses of the bridge [55]. The shown
applied to the ambient vibration responses of the bridge [55]. The shown MAC values MAC values forfor
the
obtained modal shapes were computed with respect to those obtained from the finite
the obtained modal shapes were computed with respect to those obtained from the finite
element analysis. The present results align well with those from the finite element analysis
element analysis. The present results align well with those from the finite element analysis
and Su et al. [55]. Notably, the second and sixth modes identified from the present approach
and Su et al. [55]. Notably, the second and sixth modes identified from the present ap-
were not found by Su et al. [55] processing ambient vibration measurements along the
proach were not found by Su et al. [55] processing ambient vibration measurements along
bridge deck. This discrepancy is attributed to these two modes being dominated by the
the bridge deck. This discrepancy is attributed to these two modes being dominated by
pylon, with the ratios of the identified modal component at the top of the pylon to the
the pylon, with the ratios of the identified modal component at the top of the pylon to the
maximum modal component along the deck being approximately 41 and 6 for the second
maximum modal component along the deck being approximately 41 and 6 for the second
and sixth modes, respectively.
and sixth modes, respectively.
Table 2. Identified modal parameters of the cable-stay bridge.
Table 2. Identified modal parameters of the cable-stay bridge.
Mode
Mode 1 1 2 2 3 3 4 4 5 5 6 6 77 88
ƒn (Hz) 0.64
ƒn (Hz) 0.64 1.19 1.19 1.71 1.71 2.21 2.21 2.532.53 2.83
2.83 3.14
3.14 3.82
3.82
Present
Present ξ ξ0.038 0.0380.032 0.032 0.033 0.033 0.0340.034 0.029
0.029 0.036
0.036 0.029
0.029 0.028
0.028
MAC MAC 0.98 0.98 0.97 0.97 0.99 0.99 0.99 0.99 0.970.97 0.99
0.99 0.97
0.97 0.98
0.98
ƒn (Hz) ƒn (Hz)
0.65 0.65 / / 1.64 1.64 2.19 2.19 2.572.57 / / 3.17
3.17 3.89
3.89
Ambient
Ambientvibrations
vibrations ξ ξ0.026 0.026 / / 0.027 0.027 0.0240.024 0.029
0.029 / / 0.026
0.026 0.028
0.028
[55]
[55] MAC 0.97 / 0.98 0.97 0.97 / 0.96 0.98
MAC 0.97 / 0.98 0.97 0.97 / 0.96 0.98
ƒn (Hz) 0.65 1.17 1.68 2.15 2.49 2.81 3.15 3.95
FEM
MAC 1 1 1 1 1 1 1 1
 Appl. Sci. 2024, 14, 2503 17 of 22

Table 2. Cont.

Mode 1 2 3 4 5 6 7 8
ƒn (Hz) 0.65 1.17 1.68 2.15 2.49 2.81 3.15 3.95
Appl. Sci.
FEM2024, 14, x FOR PEER REVIEW 19 of 24
MAC 1 1 1 1 1 1 1 1

Figure 13.Mode
Figure 13. Modeshapes
shapesof of
thethe bridge
bridge in transverse
in the the transverse direction.
direction.

4.2. Application to a Five-Story Steel Frame

4.2. Application to a Five-Story Steel Frame
The investigated frame, illustrated in Figure 14, has dimensions of 3 m in length,
2 m inThe investigated
width, frame,
and a height of 13illustrated in Figure
m. Each floor 14, haswith
was loaded dimensions of 3tomachieve
lead blocks in length, 2 m
in
an approximate mass of 3664 kg per floor. In subjecting the frame to a base excitation an ap-
width, and a height of 13 m. Each floor was loaded with lead blocks to achieve
proximate
equivalent to mass
60%ofof3664
the kg per floor.
intensity In Kobe
of the subjecting the frame
earthquake, to a base
as shown excitation
in Figure equivalent
15, the
to 60% of the
acceleration intensity
responses of first
of the the and
Kobe earthquake,
fifth floors in the as shown direction
long-span in Figure 15, theasacceleration
(denoted the
x-direction in
responses of Figure 14)and
the first were recorded
fifth floorsat inathe
sampling rate of
long-span 250 Hz. These
direction dataas
(denoted served as
the x-direction
input
in for identifying
Figure the modal at
14) were recorded parameters
a sampling of the frame.
rate of 250 Hz. These data served as input for
Huang et al. [19] observed clear permanent strains in the measured strains at a column
identifying the modal parameters of the frame.
of the first story, indicating the yielding of the steel columns under the considered base
excitation. Although not presented here, the first mode scalogram from the measurement of
the fifth floor illustrates the variation in the fundamental frequency over time, suggesting
nonlinear behavior in the frame under the considered base excitations. It is worth noting
that in a nonlinear system, the stiffness and damping matrices can be functions of displace-
ment or velocity responses, which in turn depend on time. Consequently, the instantaneous
modal parameters of the system exhibit time variation, and our proposed approach is
capable of identifying these time-varying modal parameters, as demonstrated below.
Assuming rigid floors, which are commonly employed in building designs, the frame
can be simplified into a five-degrees-of-freedom system in each of its long-span and short-
span directions. Modal parameter identification involved windowed responses with a
window length of 2 L, equal to 2 s. In Equation (25), the parameter ‘d’ in the weight-
ing function was set to 0.06 s, and adjacent windows had an overlay of 1.9 s. The pro-
posed approach yielded time-varying modal frequencies and modal damping ratios for
 Figure 13. Mode shapes of the bridge in the transverse direction.

4.2. Application to a Five-Story Steel Frame

The investigated frame, illustrated in Figure 14, has dimensions of 3 m in length, 2 m
Appl. Sci. 2024, 14, 2503
in width, and a height of 13 m. Each floor was loaded with lead blocks to achieve 18 of 22an ap-

proximate mass of 3664 kg per floor. In subjecting the frame to a base excitation equivalent
to 60% of the intensity of the Kobe earthquake, as shown in Figure 15, the acceleration
five modes, as depicted in Figure 16. Figure 16 also includes results from Huang et al. [19],
responses of the first and fifth floors in the long-span direction (denoted as the x-direction
who utilized a time-varying autoregressive model with exogenous input (TVARX) with the
incontinuous
Figure 14)Cauchy
were recorded at a sampling
wavelet transform rateresponses
to process of 250 Hz. These
from data
all five served
floors. The as input for
identi-
identifying the modal parameters of the frame.
fied results from the present study exhibit favorable agreement with the published findings.

Appl. Sci. 2024, 14, x FOR PEER REVIEW 20 of 24

Figure14.
Figure Schematic of
14.Schematic of the
the five-story
five-storysteel frame.
steel frame.

Figure
Figure 15.15.
TheThe relativeacceleration
relative acceleration responses
responses ofofthe
thefirst and
first fifth
and floors
fifth of the
floors of steel frameframe
the steel underunder
shaking
shaking table
table tests.
tests.

Figure 16 reveals significant variations in natural frequencies and modal damping

Huang et al. [19] observed clear permanent strains in the measured strains at a col-
ratios over time when 4.5 s < t < 7 s. Such variations indicate that the frame exhibited
umn of the first
nonlinear story,under
behaviors indicating the yielding
the considered of the steelwithin
base excitations columns under
this time the considered
range. These
base
nonlinear behaviors are indicative of earthquake-induced damage to the frame. the
excitation. Although not presented here, the first mode scalogram from measure-
Beyond
ment
thisof the fifth
specific floor the
interval, illustrates the variation
frame behaved linearly,in
andthethe
fundamental
variations in frequency over time,
identified modal
suggesting
parametersnonlinear
with timebehavior
are not as in the frame
significant under
as the the considered
published ones. base excitations. It is
worth noting that in a nonlinear system, the stiffness and damping matrices can be func-
tions of displacement or velocity responses, which in turn depend on time. Consequently,
the instantaneous modal parameters of the system exhibit time variation, and our pro-
posed approach is capable of identifying these time-varying modal parameters, as demon-
strated below.
Assuming rigid floors, which are commonly employed in building designs, the frame
can be simplified into a five-degrees-of-freedom system in each of its long-span and short-
Appl. Sci. 2024, 14, x FOR PEER REVIEW 21 of 24
Appl. Sci. 2024, 14, 2503 19 of 22

Figure16.
Figure Identified time-varying
16.Identified time-varyingmodal
modalfrequencies andand
frequencies damping ratiosratios
damping for thefor
steel
theframe.
steel frame.
5. Concluding Remarks
Figure 16 reveals significant variations in natural frequencies and modal damping
A novel approach that integrates the subspace method with the SWPT was developed
ratios over time
for identifying thewhen
modal4.5 s < t < 7in s.
parameters Such variations
structures, indicate
encompassing both that
linearthe
andframe exhibited
nonlinear
nonlinear
behaviors. The introduction of a Gaussian weighting function, coupled with a moving These
behaviors under the considered base excitations within this time range.
nonlinear behaviorsenhances
window technique, are indicative of earthquake-induced
the approach’s damage
accuracy in identifying to the frame.
time-varying modalBeyond
this specific Leveraging
parameters. interval, the
theframe
Meyer behaved
wavelet aslinearly, andfilter
a band-pass theandvariations in identified
its corresponding scale modal
function as a low-pass filter constructs the basis functions
parameters with time are not as significant as the published ones.for SWPT subspaces. The fine
filtering properties inherent in the SWPT not only facilitate mode discrimination but also
5.enable the identification
Concluding Remarksof numerous modes from a limited number of measured degrees
of freedom.
AThenovel approach
validation that
of the integrates
proposed the subspace
approach commenced methodwithwith the SWPTofwas
the processing developed
numeri-
for identifying
cally simulatedthe modal parameters
acceleration responses from in structures, encompassing
six-story shear both linear
buildings, exhibiting and nonlin-
constant,
ear behaviors.time-varying,
periodically The introduction of a Gaussian
or sharply time-varyingweighting
stiffnessfunction,
and modal coupled
damping with a moving
ratios.
The results underscore the effectiveness and robustness of the method
window technique, enhances the approach’s accuracy in identifying time-varying modal in accurately deter-
mining time-varying
parameters. Leveragingmodal theparameters,
Meyer waveleteven when
as a base excitation
band-pass andand
filter the responses of
its corresponding
only function
scale two floors,aswith a 10% NSR,
a low-pass were
filter utilized. the
constructs Compared with true values,
basis functions for SWPTthe identified
subspaces. The
frequencies and modal damping ratios for all six modes in the time-invariant shear building
fine filtering properties inherent in the SWPT not only facilitate mode discrimination but
exhibited maximum differences of 0.9% and 5%, respectively. For the periodically time-
also enable
varying thebuilding,
shear identification of numerous
the maximum modes
differences in from a limited
frequency were number
less thanof1%measured
for all de-
grees of freedom.
six modes, with maximum modal damping ratio differences less than 15% for the first five
modesTheand validation of the
20% for the sixthproposed
mode. In theapproach commenced
case of the with the processing
sharply time-varying of numer-
shear building,
ically simulated
the identified acceleration
frequencies responses
for the first fourfrom
modes six-story
deviatedshearby lessbuildings, exhibiting
than 1%, while the con-
stant, periodically time-varying, or sharply time-varying stiffness and modal damping ra-
tios. The results underscore the effectiveness and robustness of the method in accurately
determining time-varying modal parameters, even when base excitation and the re-
sponses of only two floors, with a 10% NSR, were utilized. Compared with true values,
the identified frequencies and modal damping ratios for all six modes in the time-invari-
Appl. Sci. 2024, 14, 2503 20 of 22

differences were less than 2% for the fifth and sixth modes. Correspondingly, the identified
modal damping ratios differed from the true values by less than 15%.
In extending the application of the proposed approach, it was employed to process
measured linear transverse velocity responses and support acceleration excitations from
a two-span cable-stayed bridge during an earthquake, as well as nonlinear acceleration
responses from the two floors of a five-story steel frame under a shaking table test. The
identified modal parameters for the first eight transverse modes of the bridge demon-
strated admirable agreement with those obtained from finite element analysis and ambient
vibration measurements. Moreover, the identified time-varying modal parameters for
the steel frame exhibited favorable agreement with published findings obtained using a
time-varying autoregressive model with exogenous input with the continuous Cauchy
wavelet transform. Significant changes in the natural frequencies and damping ratios
indicated earthquake-induced damage to the frame. Overall, the proposed approach show-
cased its versatility and efficacy in accurately identifying modal parameters for structures
undergoing both linear and nonlinear dynamic conditions.

Author Contributions: Conceptualization, W.-C.S. and C.-S.H.; methodology, W.-C.S. and C.-S.H.;
software, W.-C.S.; validation, W.-C.S., L.-J.C. and C.-S.H.; formal analysis, L.-J.C. and C.-S.H.; investi-
gation, W.-C.S., L.-J.C. and C.-S.H.; resources, C.-S.H.; data curation, L.-J.C.; writing—original draft
preparation, C.-S.H.; writing—review and editing, C.-S.H.; supervision, C.-S.H.; project administra-
tion, C.-S.H.; funding acquisition, C.-S.H. All authors have read and agreed to the published version
of the manuscript.
Funding: This research was funded by the National Science and Technology Council, Taiwan, through
research grant number MOST 111-2221-E-A49-015-MY2.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Data is contained within the article.
Acknowledgments: The work reported herein was supported by the National Science and Technology
Council, Taiwan, through research grant number MOST 111-2221-E-A49-015-MY2. This support is
gratefully acknowledged. The appreciation is also extended to C. H. Chen at the National University
of Kaohsiung, Taiwan, for providing the measured responses of the bridge subjected to an earthquake.
Conflicts of Interest: Author Wei-Chih Su was employed by the company AI Application & Develop-
ment Center, Mitac Advanced Technology Corp. The remaining authors declare that the research was
conducted in the absence of any commercial or financial relationships that could be construed as a
potential conflict of interest.

References
1. Collura, D.; Nascimbene, R. Comparative assessment of variable loads and seismic actions on bridges: A case study in Italy using
a multimodal approach. Appl. Sci. 2023, 13, 2771. [CrossRef]
2. Zhang, H.; Reuland, Y.; Shan, J.; Chatzi, E. Post-earthquake structural damage assessment and damage state evaluation for RC
structures with experimental validation. Eng. Struct. 2024, 304, 117591. [CrossRef]
3. Farrar, C.R.; Doebling, S.W.; Nix, D.A. Vibration-based structural damage identification. Philos. Trans. R. Soc. A Math. Phys. Eng.
Sci. 2001, 359, 131–150. [CrossRef]
4. Qarib, H.; Adeli, H. Recent advances in health monitoring of civil structures. Sci. Iran. Trans. A Civ. Eng. 2014, 21, 1733–1742.
5. Sony, S.; Laventure, S.; Sadhu, A. A literature review of next-generation smart sensing technology in structural health monitoring.
Struct. Control. Health Monit. 2019, 26, e2321. [CrossRef]
6. Sirca, G.F.; Adeli, H. System identification in structural engineering. Sci. Iran. Trans. A Civ. Eng. 2012, 19, 1355–1364. [CrossRef]
7. Perez-Ramirez, C.A.; Amezquita-Sanchez, J.P.; Adeli, H.; Valtierra-Rodriguez, M.; Romero-Troncoso, R.J.; Dominguez-Gonzalez,
A.; Osornio-Rios, R.A. Time-frequency techniques for modal parameters identification of civil structures from acquired dynamic
signals. J. Vibroenginerring 2016, 18, 3164–3185. [CrossRef]
8. Shokravi, H.; Shokravi, H.; Bakhary, N.; Rahimian Koloor, S.S.; Petru, M. Health monitoring of civil infrastructures by subspace
system identification method: An overview. Appl. Sci. 2020, 10, 2786. [CrossRef]
9. Safak, E.; Celebi, M. Seismic response of Transamerica building, II: System identification. J. Struct. Eng. ASCE 1991, 117, 2405–2425.
[CrossRef]
Appl. Sci. 2024, 14, 2503 21 of 22

10. Loh, C.H.; Lin, H.M. Application of off-line and on-line identification techniques to building seismic response data. Earthq. Eng.
Struct. Dyn. 1996, 25, 269–290. [CrossRef]
11. Huang, C.S.; Hung, S.L.; Lin, C.I.; Su, W.C. A wavelet-based approach to identifying structural modal parameters from seismic
response and free vibration data. Comput. Aided Civil Infrastruct. Eng. 2005, 20, 408–423. [CrossRef]
12. Huang, C.S.; Su, W.C. Identification of modal parameters of a time invariant linear system by continuous wavelet transformation.
Mech. Syst. Signal Process. 2007, 21, 1642–1664. [CrossRef]
13. Satio, T.; Yokota, H. Evaluation of dynamic characteristics of high-rise buildings using system identification. J. Wind Eng. Ind.
Aerodyn. 1996, 59, 299–307. [CrossRef]
14. Moore, S.M.; Lai, J.C.S.; Shankar, K. ARMAX modal parameter identification in the presence of unmeasured excitation-I:
Theoretical background. Mech. Syst. Signal Process. 2007, 21, 1601–1615. [CrossRef]
15. Moore, S.M.; Lai, J.C.S.; Shankar, K. ARMAX modal parameter identification in the presence of unmeasured excitation-II:
Numerical and experimental verification. Mech. Syst. Signal Process. 2007, 21, 1616–1641. [CrossRef]
16. Dziedziech, K.; Czop, P.; Staszewski, W.J.; Uhl, T. Combined non-parametric and parametric approach for identification of
time-variant systems. Mech. Syst. Signal Process. 2018, 103, 295–311. [CrossRef]
17. Niedźwiecki, M. Identification of Time-Varying Processes; John Wiley & Sons: New York, NY, USA, 2000.
18. Loh, C.H.; Lin, C.Y.; Huang, C.C. Time domain identification of frames under earthquake loadings. J. Eng. Mech-ASCE 2000, 126,
693–703. [CrossRef]
19. Huang, C.S.; Liu, C.Y.; Su, W.C. Application of Cauchy wavelet transformation to identify time-variant modal parameters of
structures. Mech. Syst. Signal Process. 2016, 80, 302–323. [CrossRef]
20. Van Overschee, P.; De Moor, B. Subspace Identification of Linear Systems: Theory, Implementation, Applications; Kluwer Academic
Publishers: Dordrecht, The Netherlands, 1996.
21. Shokravi, H.; Shokravi, H.; Bakhary, N.; Heidarrezaei, M.; Rahimian Koloor, S.S.; Petru, M. Application of the subspace-based
methods in health monitoring of civil structures: A systematic review and meta-analysis. Appl. Sci. 2020, 10, 3607. [CrossRef]
22. Li, Z.; Fu, J.Y.; Liang, Q.S.; Mao, H.J.; He, Y.C. Modal identification of civil structures via covariance-driven stochastic subspace
method. Math. Biosci. Eng. 2019, 16, 5709–5728. [CrossRef] [PubMed]
23. Wang, C.; Hu, M.H.; Jiang, M.H.; Zuo, Z.N.; Zhu, Y.F.; Zhu, Z.Q. A modal parameter identification method based on improves
covariance-driven stochastic subspace identification. ASME J. Eng. Gas Turbines Power. 2020, 142, 061005. [CrossRef]
24. Cheynet, E.; Daniotti, N.; Jakobsen, J.B.; Snaebjörnsson, J. Improved long-span bridge modeling using data-driven identification
of vehicle-induced vibrations. Struct. Control. Health Monit. 2020, 27, e2574. [CrossRef]
25. Niu, Y.B.; Ye, Y.; Zhao, W.J.; Shu, J.P. Dynamic monitoring and data analysis of a long-span arch bridge based on high-rate
GNSS-RTK measurement combining CF-CEEMD method. J. Civ. Struct. Health Monit. 2020, 11, 35–48. [CrossRef]
26. Liu, Y.K.; Loh, C.S.; Ni, Y.Q. Stochastic subspace identification for output-only modal analysis: Application to super high-rise
tower under abnormal loading condition. Earthq. Eng. Struct. Dyn. 2013, 42, 477–498. [CrossRef]
27. Pioldi, F.; Rizzi, E. Assessment of frequency versus time domain enhanced technique for response-only modal dynamic identifica-
tion under seismic excitation. Bull. Earthq. Eng. 2018, 16, 1547–1570. [CrossRef]
28. Allemang, R.L.; Brown, D.L. A correlation coefficient for modal vector analysis. In Proceedings of the First International Modal
Analysis Conference, Bethel, CT, USA, 6–9 April 1983; pp. 110–116.
29. Kim, J.; Lynch, J.P. Subspace system identification of support-excited structures—Part I: Theory and black-box system identifica-
tion. Earthq. Eng. Struct. Dyn. 2012, 41, 2235–2251. [CrossRef]
30. Kim, J.; Lynch, J.P. Subspace system identification of support excited structures—Part II: Gray-box interpretations and damage
detection. Earthq. Eng. Struct. Dyn. 2012, 41, 2253–2271. [CrossRef]
31. Mellinger, P.; Döhler, M.; Mevel, L. Variance estimation of modal parameters from output-only and input/output subspace-based
system identification. J. Sound Vib. 2016, 379, 1–27. [CrossRef]
32. Moonen, M.; De Moor, B.; Vandenberghe, L.; Vandewalle, J. On and off-line identification of linear state space models. Int. J.
Control 1989, 49, 219–232. [CrossRef]
33. Van Overschee, P.; De Moor, B. N4SID: Subspace algorithms for identification of combined deterministic-stochastic systems.
Automatica 1994, 30, 75–93. [CrossRef]
34. Viberg, M.; Wahlbergs, B.; Otterstens, B. Analysis of state space system identification methods based on instrumental variables
and subspace fitting. Auromatica 1997, 33, 1603–1616. [CrossRef]
35. Mastronardi, N.; Kressner, D.; Sima, V.; Van Dooren, P.; Van Huffel, S. A fast algorithm for subspace state-space system
identification via exploitation of the displacement structure. J. Comput. Appl. Math. 2001, 132, 71–81. [CrossRef]
36. Yu, D.J.; Ren, W.X. EMD-based stochastic subspace identification of structures from operational vibration measurements. Eng.
Struct. 2005, 27, 1741–1751. [CrossRef]
37. Altunisik, A.C.; Sunca, F.; Sevim, F. Modal parameter identification and seismic assessment of historical timber structures under
near-fault and far-fault ground motions. Structures 2023, 49, 1624–1651. [CrossRef]
38. Gandino, E.; Garibaldi, L.; Marchesiello, S. Covariance-driven subspace identification: A complete input–output approach. J.
Sound Vib. 2013, 332, 7000–7017. [CrossRef]
39. Verhaegen, M.; Hansson, A. N2SID: Nuclear norm subspace identification of innovation models. Automatica 2016, 72, 57–63.
[CrossRef]
Appl. Sci. 2024, 14, 2503 22 of 22

40. Juang, J.N. System realization using information matrix. J. Guid. Control Dyn. 1997, 20, 492–500. [CrossRef]
41. Siringoringo, D.M.; Fujino, Y. Influence of movable bearings performance on the dynamic characteristics of a cable-stayed bridge:
Insights from seismic monitoring records. Bull. Earthq. Eng. 2022, 20, 4639–4671. [CrossRef]
42. Oku, H.; Kimura, H. A recursive 4SID from the input-output point of view. Asian J. Control 1999, 1, 258–269. [CrossRef]
43. Oku, H.; Kimura, H. Recursive 4SID algorithms using gradient type subspace tracking. Automatica 2002, 38, 1035–1043. [CrossRef]
44. Loh, C.H.; Chen, J.D. Tracking modal parameters from building seismic response data using recursive subspace identification
algorithm. Earthq. Eng. Struct. Dyn. 2017, 46, 2163–2183. [CrossRef]
45. Huang, S.K.; Chen, J.D.; Loh, K.J.; Loh, C.H. Discussion of user-defined parameters for recursive subspace identification:
Application to seismic response of building structures. Earthq. Eng. Struct. Dyn. 2020, 49, 1738–1757. [CrossRef]
46. Chopra, A.K. Dynamics of Structures. In Theory and Applications to Earthquake Engineering; Prentice Hall: Hoboken, NJ, USA, 2014.
47. Huang, C.S.; Lin, H.L. Modal identification of structures from ambient vibration, free vibration, and seismic response data via a
subspace approach. Earthq. Eng. Struct. Dyn. 2001, 30, 1857–1878. [CrossRef]
48. Pesquet, J.C.; Krim, H.; Carfantan, H. Time-invariant orthonormal wavelet representations. IEEE Trans. Signal Process. 1996, 44,
1964–1970. [CrossRef]
49. Fowler, J. The redundant discrete wavelet transform and additive noise. IEEE Signal Process. Lett. 2005, 12, 629–632. [CrossRef]
50. Perez-Solano, J.J.; Felici-Castell, S.; Rodriguez-Hernandez, M.A. Narrowband interference suppression in frequency-hopping
spread spectrum using undecimated wavelet packet transform. IEEE Trans. Veh. Technol. 2008, 57, 1620–1629. [CrossRef]
51. Mallat, S. A Wavelet Tour of Signal Processing, 2nd ed.; Academic: San Diego, CA, USA, 1999.
52. Viberg, M. Subspace-based methods for the identification of linear time-invariant systems. Automatica 1995, 31, 1835–1851.
[CrossRef]
53. Verhaegen, M. Identification of the deterministic part of MIMO state space models given in innovations form from input-output
data. Automatica 1994, 30, 61–74. [CrossRef]
54. Yang, Q.J.; Zhang, P.Q.; Liand, C.Q.; Wu, X.P. System theory approach to multi-input multi-output modal parameters identification
method. Mech. Syst. Signal Process. 1994, 8, 159–174. [CrossRef]
55. Su, W.C.; Huang, C.S.; Chen, C.H.; Liu, C.Y.; Huang, H.C.; Le, Q.T. Identifying the modal parameters of a structure from ambient
vibration data via the stationary wavelet packet. Comput. Aided Civil Infrastruct. Eng. 2014, 29, 738–757. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.