Measurement 89 (2016) 189–196

Contents lists available at ScienceDirect

Measurement
journal homepage: www.elsevier.com/locate/measurement

Fast vision-based wave height measurement for dynamic

characterization of tuned liquid column dampers
Junhee Kim, Chan-Soo Park, Kyung-Won Min ⇑
Department of Architectural Engineering, Dankook University, Jukjeon-dong, Suji-gu, Yongin-si, Gyeonggi-do 448-701, Republic of Korea

a r t i c l e i n f o a b s t r a c t

Article history: In this study, a novel and rapid vision-based sensing strategy is developed exclusively for dynamic wave
Received 24 October 2014 height measurement of tuned liquid column dampers (TLCDs). The image processing algorithm of the
Received in revised form 7 April 2016 vision-based sensing method simply counts white pixels in a binary image and thus expedites the
Accepted 12 April 2016
vision-based wave height measurement. In addition to the experimental achievement, a practical
Available online 13 April 2016
methodology of dynamic characterization for the TLCDs is proposed combining linearized equations
for the TLCDs and experimental data measured. An experimental characterization of dynamic behaviors
Keywords:
and damping properties of the TLCDs is undertaken utilizing the vision-based sensing developed.
Vision sensing
Wave height measurement
Ó 2016 Elsevier Ltd. All rights reserved.
Dynamic characterization
Structural control
Tuned liquid column dampers

1. Introduction
Structural developments in design and construction of build-
ings and towers have rendered structures tall, slender, and flexible.
The tall and slender structures are vulnerable to dynamic loading
such as wind and earthquakes. While the structures operate within
safety limits, they may suffer from lack of serviceability due to
undesirable vibrations induced by the dynamic loading. Structural
control is needed to reduce the dynamic responses of the struc-
tures and to maintain their functional performances. A constant
flow of developing novel and effective devices incorporated in
structures for attenuation of structural vibration has been seen
over the past decades [1].
One of the pervasive strategies widely applied to attenuate
structural vibration is the installation of a secondary mass damper
on the top floor of a building, i.e., a device generating a reaction
force induced from the oscillating motion of a secondary mass.
The secondary mass is a small fraction of the entire mass of the pri-
mary structure and interfaced to inherent damping devices for
increasing the energy dissipation capability. Depending on the
oscillatory media, the secondary mass dampers are categorized
into two groups: tuned mass damper (TMD) and tuned liquid dam-
per (TLD). The TMDs are mechanical devices of a solid mass with
springs and dashpots attached to the primary building [2] and
the TLDs are liquid containers [3]. As the counterpart of the TMDs,
⇑ Corresponding author.
E-mail address: kwmin@dankook.ac.kr (K.-W. Min).
the TLDs have proven their advantages: for example, simplicity,
low cost, easy installation and maintenance, just to name a few
[4–7].
Two different configurations of the TLDs have been investigated
and adopted in construction sites [8]. Tuned Liquid Mass Damper
(TLMD) and Tuned Liquid Column Damper (TLCD) utilize energy
dissipating liquid motions of oscillation in narrow tubes and wave
braking/sloshing in free liquid surface, respectively. Referring to
numerous design parameters relating the configuration of the
TLCDs and resultant tuning feasibility to determine their dynamic
characteristics, practical advantages of the TLCDs over the TLMDs
have frequently been emphasized in the literature [4,9,10].
Prior to installation of a TLCD at a site, a factory test for verifi-
cation and tuning of dynamic characteristics must be conducted
with the pre-fabricated TLCD. Among the dynamic characteristics
of the TLCD, optimal tuning frequency and damping ratio are con-
sidered primal, since they are directly related to control perfor-
mance, i.e., vibration suppression of the primary structure. Wave
height of the oscillating liquid column of the TLCD is measured
during the factory test and the dynamic characteristics are then
estimated from wave height data measured.
To date, capacitive wavemeters immersed into liquid have been
dominantly used for measurement of varying wave heights. How-
ever, a number of intrinsic disadvantages associated with the con-
tact sensors have been constantly addressed: high price, laborious
installation, and a loss of accuracy due to interference from a liquid
medium, e.g., parasitic capacitance. While great advances have
been made in analytical studies of dynamic behaviors of the TLCDs

http://dx.doi.org/10.1016/j.measurement.2016.04.030
0263-2241/Ó 2016 Elsevier Ltd. All rights reserved.
190 J. Kim et al. / Measurement 89 (2016) 189–196

[6,10,11] and practical design formulas/guidelines [9,12,13], com- The Lagrange’s equations of motion on the basis of the Hamil-
paratively less research has been conducted toward novel mea- ton’s principle for conservative systems [20] are adopted to formu-
surement strategy and further characterization of the TLCDs late equations of motion of the TLCD [6,21,22]. Summation of the
based on experimentally measured data. kinetic energy of the oscillating liquid along the two vertical and
Recently, there has been growing interest in noncontact sensing one horizontal columns leads to
in the areas of structural monitoring and assessment [14]. The
1
precedent works of noncontact sensing, especially vision-based T ¼ qALv ðy_ 2 þ x_ 2 Þ þ qALh ðy_ 2 þ 2x_ y_ þ x_ 2 Þ ð1Þ
remote sensing, are found in the literature [15–19]. In this study, 2
a rapid, high precision, and cost-effective vision-based sensing sys- where q is the density of the liquid filled in the TLCD. Considering
tem is developed exclusively for dynamic wave height measure- wave height changes in the two vertical liquid columns, the poten-
ment of the TLCDs. A practical methodology of experimental tial energy is expressed as
estimation of dynamic characteristics of the TLCDs is also proposed
based on both dynamic equations derived and experimental data y y
V ¼ qAðyÞg þ qAyg ¼ qAgy2 ð2Þ
measured. Finally, a series of experimental investigations are con- 2 2
ducted for the verification of the vision-based sensing system and
The energy dissipated by the liquid flow passing through the
for showcasing the methodology of dynamic characterization of
orifice is often referred to as head loss in fluid dynamics. Combin-
the TLCDs.
ing the energy balance equation, i.e., the Lagrange’s equation, and
the non-conservative force of the head loss leads to



2. Formulation of dynamic behavior of tlcds d @ðT  VÞ @ðT  VÞ 1
 ¼  qAgjyj
_ y_ ð3Þ
dt @ y_ @y 2
The equations of dynamic behavior of the TLCDs are derived for
estimation of their dynamic characteristics based on the experi- where g is the head loss coefficient. The equation of motion of the
mental data measured. While natural frequencies of the TLCDs TLCD is derived by inserting Eqs. (1) and (2) into Eq. (3):
can be directly determined by measurements, damping estimation € þ c1 y_ þ k1 y ¼ m2 €x
m1 y ð4Þ
requires physics-based relationships between damping and mea-
surable quantities. Thus, a practical equation for estimating the where the total mass of the liquid column is m1 ¼ qAðLh þ 2Lv Þ; the
damping ratio of the TLCDs is derived in this section. mass of the horizontal liquid column is m2 ¼ qALh ; the nonlinear
friction damping coefficient is c1 ¼ ð1=2ÞqAgjyj;
_ the term related
to stiffness is k1 ¼ 2qAg.
2.1. Derivation of linearized equation of motion The nonlinear friction damping can be treated as linear viscous
damping by the concept of equivalent damping. Considering a har-
A simplified model of the TLCD built on the primary structure is monic excitation of the primary structure with the frequency, x
illustrated in Fig. 1. The TLCD considered here has two vertically and corresponding sinusoidal displacement response of the verti-
upright tubes connected together at the bottom orifice forming cal liquid motion with the amplitude, uy , i.e., y ¼ uy sin xt, the
sharp-edged elbows. Identical cross sectional areas, denoted as A non-conservative force of nonlinear friction damping is calculated
in Fig. 1, of the vertical and horizontal tubes are considered. The as
displacements of horizontal structural motion of the exciting pri-
1
mary structure and vertical surface motion of oscillating liquid of f nc1 ¼ c1 y_ ¼ qAgu2y x2 j cos xtj cos xt ð5Þ
the TLCD are denoted as x and y, respectively. As for the dimen- 2
sions of the TLCD, the horizontal and vertical column lengths are The energy dissipated in a full cycle is calculated by integrating
defined as Lh and Lv, respectively. the non-conservative damping force in the first quarter cycle and
quadrupling the result:
8 !2 9
Z Z < =
uy u sin xt
2qAgu2y x2 1 
y
ED1 ¼ 4 f nc1 dy ¼ dy
0 : uy ;

4
¼ qAgu3y x2 ð6Þ
3
On the contrary, the non-conservative force is composed with
the linear viscous damping as
f nc2 ¼ ceq y_ ¼ ceq uy x cos xt ð7Þ

The energy dissipated by the linear viscous damping is calcu-

lated as
Z Z 2p
x
Z 2p
x
ED2 ¼ f nc2 dy ¼ _ ¼
f nc2 ydt ceq u2y x2 cos2 xtdt
0 0

¼ ceq pu2y x ð8Þ

By equating the dissipated energy of the nonlinear friction

damping to that of linear viscous damping, the equivalent viscous
damping is derived as
4
ceq ¼ qAguy x ð9Þ
Fig. 1. A TLCD subjected to a vibrating structure.
3p
 J. Kim et al. / Measurement 89 (2016) 189–196 191

Inserting Eq. (9) into the equation of motion of the TLCD yields A video of a laterally shaking TLCD where blue dye was added to
the linearized equation of motion accentuate the liquid was preliminarily taken by a digital cam-
corder. The lateral motion of the TLCD resulted in a vertically oscil-
4
€þ
m1 y qAguy xy_ þ k1 y ¼ m2 €x ð10Þ lating liquid column. A couple of freeze-frames grabbed from the
3p
video are given in Fig 2: Fig. 2(a) and (b) are still images at the
Dividing both sides of Eq. (10) by m1 leads to instants when the free surface of the liquid column reaches the
4guy x highest and lowest positions, respectively. As seen in figures,
€þ
y y_ þ x2y y ¼ p€x ð11Þ height and horizontal location of the liquid column are changing
3pLe
simultaneously. The measurement quantity, i.e., wave height, is
where Le ¼ Lh þ 2Lv is the effective length of the TLCD; defined by distance between upper and lower corners of the liquid
pffiffiffiffiffiffiffiffiffiffiffiffi
xy ¼ 2g=Le is the natural frequency of the TLCD; p ¼ Lh =Le is the column seen in Fig. 3(a) where the corners are highlighted by two
ratio of the horizontal length to total length of the liquid column. L-shaped markers. As a result, detection of the two corners in dig-
ital images leads directly to wave height measurement.
2.2. Derivation of damping ratio To locate a moving object by analyzing digital images entails
two sequential image processing procedures [23]: edge detection
Considering a harmonic motion of the exciting primary struc- and moving object tracking. Edges are paths of rapid change of
ture and corresponding response of the liquid motion, i.e., image intensity in a digital image and outline targeted objects,
x ¼ x0 eixt and y ¼ y0 eixt , respectively, the linearized equation of i.e., the liquid column in the TLCD in this study (Fig. 3(b)). Edge
motion of Eq. (11) yields the amplitude of the liquid motion: detection is used to identify the edges in an image. Height of the
liquid column of the TLCD is measured from the edge detected,
x0 pb2 while laterally tracking the position-changing liquid column. An
y0 ¼   ð12Þ
4gu b2 image processing technique for tracking moving objects is conven-
ð1  b2 Þ þ i 3pLy e
tionally conducted by correlation analysis. In this study, the two
where b ¼ x=xy is the frequency ratio of the exciting frequency to corners are moving objects to be tracked and reference image
the natural frequency of the TLCD. The amplitudes of the liquid frame is used for the correlation analysis (Fig. 3(c)). However, the
motion in both sides of Eq. (12) can be arranged by the identity of consecutive image processing techniques, i.e., edge detection fol-
uy ¼ jy0 j. As a result, the amplitude uy is calculated as the magni- lowed by correlation analysis, are computationally expensive and
time consuming. Thus, their implementation presents a challenge
tude of the complex amplitude of liquid motion of Eq. (12):
to the use of real time image processing for the measurement
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi
 2 2 the liquid height in the TLCD.
2 2 2 4 4 4 g b
ð1  b Þ þ ð1  b Þ þ 4x20 p2 b 3pLe In this study, an expedited image processing technique tailored
uy ¼ pffiffiffi4gb2  ð13Þ to the measurement of the liquid height of the TLCD is developed
2 3pLe to skip the aforementioned retardant processes of edge detection
and correlation analysis. The suggested technique examines on
Arranging Eq. (13) with respect to the head loss leads to image portion, termed region of interest (ROI), extracted by crop-
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ping the original image. In this study, the ROI is set to be a portion
3pLe ðuy  uy b2 þ x0 pb2 Þðuy b2  uy þ x0 pb2 Þ of the image specifying the liquid column of the TLCD. The rectan-
g¼ ð14Þ
4u2y b2 gle ROI contains a true-color image of the liquid column truncated
above the lower sharp-edged elbow and is dealt as an RGB image,
Damping ratio of the TLCD is calculated from the equivalent vis- an m  n  3 data array that defines red, green, and blue color
cous damping of Eq. (9) components for m  n pixels. In general, a graphic file format of
ceq 2guy b the RGB image is 24-bit digital information, where the red, green,
ny ¼ ¼ ð15Þ and blue components are 8 bits each. Since the liquid column of
ccr 3pLe
the TLCD is blue, the information of blue color is exclusively con-
where the critical damping is given as ccr ¼ 2m1 xy . Eventually, the verted to a grayscale color map consisting of 8 bits per pixel.
equation for the damping ratio of the TLCD by inserting Eq. (14) into Finally, a binary image (1 bit per pixel) is acquired from the grays-
Eq. (15) is derived as cale image using an appropriate threshold.
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Explanatory image matrices of the binary images at two differ-
ðuy  uy b2 þ x0 pb2 Þðuy b2  uy þ x0 pb2 Þ ent instants are given at Fig. 4. Two dimensional matrices of 1’s
ny ¼ ð16Þ
2uy b (white pixels) and 0’s (black pixels) represent the liquid column
of the TLCD and the other parts, respectively. As seen, the height
As seen in Eq. (16), the damping ratio is a function of amplitude and horizontal position of the liquid column change in the two fig-
of motion of the primary structure, natural frequencies of both the ures. Since the feature aimed to be measured in the study is the
primary structure and the TLCD, and geometrical configurations of height of the liquid column regardless of its position, a simple
the TLCD. strategy of analyzing the binary image matrices are adopted. It
requires solely the counting of white pixels, i.e., the total sum of
3. Expedited digital image processing for TLCDS ones in the binary image matrix, which is directly proportional
to the height of the liquid column.
In this section, a novel and rapid video processing technique Mapping of the total sum of the white pixels to a metric system
tailed to measurement of wave height of the TLCD is presented of length needs a reference, which is the number of pixels along a
to replace the conventional wave height meter, taking advantages line of measured length. In this study, an a priori measurement of
of noncontact sensing. The technique requires solely counting of the width of the liquid column is adopted. Thus, the height of the
white pixels in the binary image matrix and is thus rapid enough liquid column at time step i, h(i) is calculated as
for near instant processing. A detailed explanation on the tech-
sðiÞ
nique will follow introduction to conventional retardant video pro- hðiÞ ¼ a ð17Þ
cessing technique. W
192 J. Kim et al. / Measurement 89 (2016) 189–196

Fig. 2. Two still images of a working TLCD at two different instants.

Fig. 3. Conventional image processing technique for wave height measurement: (a) definition of wave height; (b) edges detected; (c) two reference frames of corners.

Fig. 4. Binary information of the image matrices for a moving liquid column in the ROI.

where s(i) is the total sum of white pixels in the ROI at time step i; a
is the mapping factor (unit of m/pixel) determined by the afore-
mentioned measurement; W is the number of white pixels along
the width of the liquid column. It should be again noted that the
algorithm to measure the liquid height of the TLCD using a video
recorded does not require any time-consuming image processing
technique and implementation of the algorithm is thus near instant.
4. Experimental verification
In the experimental phases of this study, two experimental
examples are present: first, a demonstration of the performance
of the expedited vision-based wave height sensing is given with
a comparison with conventional measurements. Next, the vision-
based sensing strategy is directed to experimentally analyze
dynamic behaviors of the TLCDs with different orifices.
4.1. Performance of vision-based wave height sensing
Prior to the test, a laboratory-scaled prototype (Fig. 5) of the
TLCDs was fabricated with 1 cm thick acrylic plates. The width
and height of the TLCD are 650 and 400 mm, respectively. The
inner dimension of vertical and horizontal column, i.e., cross sec-
tional area of liquid flow, is 150  150 mm. The TLCD was mounted
on a uniaxial shake table driven by a mechanical linear actuator
where rotary motion of an AC servo motor (HC-SFS502, MITSU-
BISHI) is converted to linear motion. The AC servo motor is oper-
ated in the velocity control mode in which angular velocity of
the rotary motion is subjected to applied analog electrical signals
programmed in a function generator. A 60 s long 0.2–2.0 Hz chirp
signal with a constant amplitude of peak-to-peak ±1 V was applied
to the AC servo motor via the function generator.
To measure motion of the shake table, a laser displacement
sensor (CD4-350, OPTEX FA) was utilized. The laser displacement
 J. Kim et al. / Measurement 89 (2016) 189–196 193

Fig. 5. Experiments of the vision-based wave height sensing for a shaking TLCD – oscillating liquid column is clearly noticeable.

sensor is a precision optoelectronic transducer to determine
dynamic displacement by sensing phase differences between
transmitted laser and back scattered laser from the moving surface
of targeted objects. During the dynamic test, the laser displace-
ment sensor was set 0.35 m away from an edge of the shake table
and sensed the shake table’s displacement motion while aligning
along the axis of moving direction of the shake table. The measured
displacement time histories of the shake table are plotted in Fig. 6
(a). Displacement amplitude decaying due to the velocity control of
the AC servo motor is witnessed over the duration of the applied
excitation. The Fourier spectrum of the measured displacement is
25
Displacement (mm)
presented in Fig. 6(b) where the 0.2–2.0 Hz frequency band of
the applied excitation is confirmed.
To demonstrate the cost-effectiveness and accuracy of the pro-
posed vision-based sensing system compared to a conventional
contact sensor, both sensors were utilized record the dynamic
motion of the TLCD during the test. A commercially available dig-
ital camcorder (HMX-QF20, SAMSUNG) which costs 300 USD in the
market was deployed. The high-definition (HD) digital camcorder
is a 1920  1080 pixel full HD camcorder capturing 60 frames
per sec. The video compression format supported by the camcorder
is H.264, i.e., MPEG-4, which is the latest video coding format stan-
dardized by the ISO/IEC and compatible with the image processing
toolbox of MATLAB. The camcorder was placed on a stationary tri-
pod 2 meters away from the shake table. Accuracy of the vision-
based wave height measurement system was cross-checked with

that of a conventional wavemeter, i.e., capacitive wave height sen-

sor (CH-601, KENEK). The wavemeter costs 3000 USD and its
installation is laborious compared to the proposed system – the
0 wave probe was immersed in the liquid column of the TLCD as well
as the sensor body was clamped on top of the TLCD as seen in
Fig. 5.
After the dynamic test was completed, the video recorded by
the camcorder was transferred to a PC via USB interface for digital
-25
0 10 20 30 40 50 60 image processing as proposed by this study. The number of frames,
Time (sec) image size, video duration, and frame-rate were identified in the
MATLAB image processing toolbox – the frame rate is 60 frames
(a) per sec and thus the 60 s video is composed of a total of 3840
3 frames; the 1920  1080 pixels of the individual frames have 24
10
bit RGB information. The aforementioned image processing for
wave height measurement was consecutively conducted with
sequences of the freeze-frames grabbed from the video and the
Amplitude

results are depicted in Fig. 7. For comparison purposes, time histo-

10
0 ries of wave height measured by the wavemeter are superimposed.
Time synchronization of the two different signals was conducted
by maximizing correlation of the signals. As seen, a very close
match of two measurements is observed in Fig. 7(a). A quantitative
accuracy evaluation of the proposed methodology was conducted
-3
10 by calculating the following RMS error between the signals:
0 0.5 1 1.5 2 2.5 3
Frequency (Hz)
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
X ffi
2
RMS Error ¼ ðhðiÞ  hðiÞ Þ =N ð18Þ
(b) i video wave

Fig. 6. Frequency shift excitation of the shake table: (a) measured displacement; (b) where hðiÞvideo and hðiÞwave are measured wave heights at time step i,
corresponding Fourier spectrum of the displacement. by the vision-based sensing and wavemeter, respectively; N is the
194 J. Kim et al. / Measurement 89 (2016) 189–196
40
Wavemeter
Vision-based sensing
Displacement (mm)
0 3
-40
0 10 20 30 40 50
Time (sec)
(a)
3 0
Wavemeter
Vision-based sensing
2
Amplitude
1 Time (sec)
0
0 0.5 1 1.5 2 2.5
Frequency (Hz)
(b)
Fig. 7. Comparison of responses of the TLCD mounted on the shake table (thin:
vision-based sensing proposed, thick: wavemeter): (a) measured heights; (b)
corresponding Fourier spectra of the heights.
number of discrete data measured. In this study, the RMS error was
evaluated as 1.053 mm. Considering the fact that the wave height is
sensed at a point of the liquid surface by the wavemeter while the
vision-based sensing yields the average height of the liquid column,
the small RMS error evaluated suggests excellent accuracy of the
vision-based sensing proposed. The Fourier spectra of the measured
wave heights are presented in Fig. 7(b). Dominant peaks at 0.854 Hz
are clearly observed at the two measurements. Amplitude decay is
witnessed near 2.0 Hz, suggesting the bandwidth of the applied
excitation, i.e., 0.2–2.0 Hz.
So far, the high precision of the expedited vision-based sensing
proposed in this study has been quantitatively proven through the
comparison with the conventional capacitive wavemeter. Based on
the successful results, advantages of the expedited vision-based
sensing over the conventional image processing technique
explained in Fig. 3 were studied. Identical tests with the 60 s long
shake table driven by 0.2–2.0 Hz chirp signal were repeated three
times and then three 3840 frame video files were recorded – in
each image frame, 1920  1080 pixels contained 24 bit RGB infor-
mation. For each video file, two different video processing tech-
niques were applied to determine wave heights. As for the
conventional technique, a 250  250 pixel image frame of an L-
shaped marker was used as the reference in correlation analysis
to detect locations of the liquid corners. Since the two video pro-
cessing techniques were implemented with the MATLAB image
processing toolbox in an identical personal computer processed
by Intel Core2 Q9400 (2.66 GHz), a comparison of computing time
was carried out and the results are tabulated in Table 1. As seen in
the table, it consistently takes nearly the same amount of time to
compute the wave height in each technique. It is concluded that
the expedited technique proposed in this study is 3.55 times faster
than the conventional technique. One of the three results is given
in Fig. 8 – for a more precise examination, time histories in a range
Table 1
Computing time of digital video processing.
No. of test Conventional technique (s) Proposed technique (s)
1 47.470 13.385
2 47.346 13.353
47.315 13.354
Average 47.377 13.364
40
60 Proposed technique
Conventional technique
Displacement (mm)
-40
20 30 40
Fig. 8. Comparison of responses of the TLCD mounted on the shake table (thin:
vision-based sensing proposed, thick: wavemeter):
3
of 20–40 s are plotted. A good agreement between two wave
height measurements is confirmed; thus, it can be concluded that
performance of the proposed technique matches that of the con-
ventional technique, despite a significant drop in time
consumption.
4.2. Dynamic characterization of TLCDS
On the basis of the previous confirmation of the accuracy of the
vision-based wave height sensing, dynamic characteristics of the
TLCDs were estimated from experimentally measured data solely
by the vision-based sensing. To investigate wide ranges of dynamic
characteristics of the TLCDs, three different orifices with varying
blocking ratios were placed in the laboratory scaled prototype
(Fig. 5). The blocking ratio of an orifice is defined as the ratio of
closed to cross-sectional areas. In this study, blocking ratios of
25%, 50%, and 75% were selected as seen in Fig. 9.
First, damped natural frequencies of the TLCDs with different
orifices were estimated from the vision-based wave height mea-
surements during shake table tests with the previous chirp excita-
tion of Fig. 5, while consecutively installing one of the orifices in
the TLCD. Time histories of wave heights measured are plotted in
Fig. 10(a). As the blocking ratios increase, the vertical oscillations
of the liquid column of the TLCDs are suppressed especially in time
interval of 18–29 s. The trend witnessed is also confirmed in the
Fourier spectra seen in Fig. 10(b) – the magnitude drops of the
peaks near resonance frequencies depending on the blocking ratios
are noticeable. Since the energy dissipation, i.e., damping, of the
TLCDs is mainly induced by narrowing liquid flow and disturbance
near the orifices, the damping is proportional to the blocking ratio.
The dominant peak at 0.854 Hz observed with the 25% blocking
ratio orifice becomes blunt in the experiments for the 50% and
75% blocking ratio orifices. Flattening and multiple peaks, i.e., rip-
ples, of frequency components of liquid motion suggest the afore-
mentioned complex phenomena relating the energy dissipation
and similar experimental findings are reported in the literature
[24,25]. Regardless of the blunt peaks, migration of the damped

Fig. 9. (a) Three different orifices of 25%, 50%, and 75% blocking ratios and (b) perspective view of their installations in the TLCD.
40
Displacement (mm)
0
-40
0 10 20 30 40
Time (sec)
(a)
3 ratios.
2 The wave height of the liquid column and displacement of the
Amplitude
1
0 respectively, followed a total of 51 (=17  3) dynamic tests for
0 0.5 1 1.5 2 2.5
Frequency (Hz)
(b)
Fig. 10. Dynamic characteristics of the TLCD with 25%, 50%, and 75% blocking ratio
orifices: (a) measured heights; (b) corresponding Fourier spectra of the heights.
natural frequencies of TLCDs depending on damping ratio is also
observed in Fig. 10(b).
Next, dynamic characteristics of damping ratios of the TLCDs
with different orifices were quantitatively studied using the prac-
tical equation, i.e., Eq. (16), derived in Section 2.2. Since in the
equation, damping ratio is a function of the amplitude of harmonic
oscillation of the liquid column, a series of harmonic excitation
tests were undertaken changing the frequencies of the harmonic
excitation of the shake table. 17 consecutive harmonic excitation
tests in the frequency range of 0.4–1.3 Hz with an increment of
0.05 Hz were conducted with the TLCDs.
Excitation force applied to the liquid of the TLCDs during the
tests is the reaction force against the shake table in harmonic
motion. The excitation force is the product of acceleration of the
shake table and mass of the liquid filling the horizontal column
of the TLCDs. In the experimental phases, the harmonic excitation
motions of the shake table were set to generate a constant acceler-
ation amplitude of peak-to-peak ±0.2 m/s2. Thus, constant excita-
tion forces were assumed for the 17 consecutive harmonic
excitation tests regardless of frequencies of the harmonic motion.
J. Kim et al. / Measurement 89 (2016) 189–196 195
2
= 25% 10
= 50%
= 75%
Damping ratio (%)
= 25%
= 50%
= 75%
50 60 1
10
0.4 0.6 0.8 1 1.2 1.4 1.6
Frequency ratio ( )
Fig. 11. Damping ratios of the TLCDs with orifices of 25%, 50%, and 75% blocking
= 25%
= 50%
= 75%
shake table were measured by the vision-based sensing and laser
displacement sensor, respectively. To measure the wave height of
harmonic steady state response, 20 s tests were run.
Damping ratio calculation, by inserting the measured wave
height and shake table displacement into uy and x0 in Eq. (16),
3 the TLCDs with three different orifices and the results are summa-
rized in Fig. 11. In the figure, x axis ranging from 0.47 to 1.53 with
increments of 0.06 is the frequency ratio of the excitation frequen-
cies of harmonic motion to the natural frequency of the TLCD, i.e.,
0.854 Hz. Two notable aspects of the damping estimates can be
found: (1) proportionality of damping ratio to blocking ratio and
(2) frequency dependent equivalent damping. This is intuitive
because damping is proportional to head loss experienced by the
liquid column moving through the narrowed orifice and attenua-
tion of the oscillation of the liquid column under harmonic
dynamic tests with constant excitation forces is also proportional
to the head loss.
5. Conclusions
Dynamic characterization of the TLCDs is a primal factory task
prior to their installation at a site and is mainly undertaken by
measurement of wave height during the test. In this study, a rapid,
high precision, and cost-effective vision-based sensing system is
developed exclusively for dynamic wave height measurement of
the TLCDs. The robust image processing algorithm developed
enables fast and accurate measurement of the wave height simply
by counting binary digits in image pixels representing liquid col-
umns of the TLCDs without requiring conventional retardant image
processing techniques.
In addition to the novel wave height sensing developed in this
study, a practical methodology of experimental estimation of
dynamic characteristics of the TLCDs is also proposed: under the
196 J. Kim et al. / Measurement 89 (2016) 189–196
assumption of harmonic excitation, a linear solution of the liquid
motion of the TLCDs is derived and then utilized for damping esti-
mation with experimental data measured. In laboratory tests with
a prototype of the TLCDs, accuracy of the proposed vision-based
sensing is confirmed through comparison with conventional con-
tact measurement and dynamic characterization is successfully
conducted including estimation of damped natural frequencies
and damping ratio.
Ongoing and future work is aimed at developing a more sophis-
ticated vision-based wave height sensing system for the TLCDs and
other types of TLDs. For example, a combined approach of hetero-
geneous sensor data fusion by multi-rate Kalman filtering [26,27].
In addition, twofold efforts are underway to close a gap between
basic researches and engineering practice: (1) development of
design procedure and (2) direct consideration on the nonlinear
behavior of the TLCDs. Since the amplitude-dependent behaviors
of the TLCDs, especially damping, dynamic characterization of
the TLCDs are still needed be sophisticated to further develop a
practical design guide. The authors are interested in pursuing
developing the practical ways of damping estimation for the
TLCDs. Leveraging the fast and simple vision-based system devel-
oped in the study, a combined approach of experiments and com-
puter simulations, e.g., pseudo-dynamic testing of the TLCDs [28],
will be strategically pursued.
Acknowledgement
The present research was supported by the research funds of
Dankook University in 2015. The authors gratefully acknowledge
the financial support.
References
[1] A. Kareem, T. Kijewski, Mitigation of motions of tall buildings with specific
examples of recent applications, Wind Struct. 2 (3) (1999) 201–251.
[2] K.C.S. Kwok, B. Samali, Performance of tuned mass dampers under wind loads,
Eng. Struct. 17 (9) (1995) 655–667.
[3] Y. Fujino et al., Tuned liquid damper (TLD) for suppressing horizontal motion
of structures, J. Eng. Mech., ASCE 118 (10) (1992) 2017–2030.
[4] K.W. Min et al., Performance evaluation of tuned liquid column dampers for
response control of a 76-story benchmark building, Eng. Struct. 27 (7) (2005)
1101–1112.
[5] Y.L. Xu, B. Samali, K.C.S. Kwok, Control of along-wind response of structures by
mass and liquid dampers, ASCE J. Eng. Mech. 118 (1992) 20–39.
[6] H. Gao, K.C.S. Kwok, B. Samali, Optimization of tuned liquid column dampers,
Eng. Struct. 19 (1997) 476–486.
[7] T. Balendra, C.M. Wang, G. Rakesh, Vibration control of various types of
buildings using TLCD, J. Wind Eng. Ind. Aerodyn. 83 (1999) 197–208.
[8] K.W. Min, J. Kim, Y.W. Kim, Design and test of tuned liquid mass dampers for
attenuating wind responses of a full scale building, Smart Mater. Struct. 23
(2014) 045020.
[9] J.-C. Wu et al., Design guidelines for tuned liquid column damper for structures
responding to wind, Eng. Struct. 27 (2005) 1893–1905.
[10] S.K. Yalla, A. Kareem, Optimum absorber parameters for tuned liquid column
dampers, J. Struct. Eng., ASCE 126 (2000) 906–915.
[11] J.-C. Wu, C.-H. Chang, Y.-Y. Lin, Optimal designs for non-uniform tuned liquid
column dampers in horizontal motion, J. Sound Vib. 326 (1–2) (2009) 104–
122.
[12] C.C. Chang, W.L. Qu, Unified dynamic absorber design formulas for wind-
induced vibration control of tall buildings, Struct. Des. Tall Build. 7 (1998)
147–166.
[13] K.W. Min, J. Kim, H.R. Lee, A design procedure of two-way liquid dampers for
attenuation of wind-induced responses of tall buildings, J. Wind Eng. Ind.
Aerodyn. 129 (2014) 22–30.
[14] H. Sohn, Noncontact laser sensing technology for structural health monitoring
and nondestructive testing, in: SPIE, Sensors and Smart Structures
Technologies for Civil, Mechanical, and Aerospace Systems. San Diego, CA,
2014.
[15] H.-N. Ho et al., A synchronized multipoint vision-based system for
displacement measurement of civil infrastructures, Sci. World J. 2012 (2012)
519146.
[16] J.J. Lee, M. Shinozuka, A vision-based system for remote sensing of bridge
displacement, NDT E Int. 39 (2006) 425–431.
[17] P. Olaszek, Investigation of the dynamic characteristic of bridge structures
using a computer vision method, Measurement 25 (1999) 227–236.
[18] H.-S. Choi et al., Structural dynamic displacement vision system using digital
image processing, NDT E Int. 44 (7) (2011) 597–608.
[19] A.M. Wahbeh, J.P. Caffrey, S.F. Masri, A vision-based approach for the direct
measurement of displacements in vibrating systems, Smart Mater. Struct. 12
(2003) 785–794.
[20] L. Meirovitch, Principles and Techniques of Vibrations, Prentice-Hall, Upper
Saddle River, NJ, 1997.
[21] J.C. Wu, Experimental calibration and head loss prediction of tuned liquid
column damper, Tamkang J. Sci. Eng. 8 (4) (2005) 319–325.
[22] C.C. Chang, C.T. Hsu, Control performance of liquid column vibration
absorbers, Eng. Struct. 20 (1998) 580–586.
[23] J. Lewis, Fast normalized cross-correlation, Vision Interface 10 (1) (1995) 120–
123.
[24] S.K. Yalla, A. Kareem, Semiactive tuned liquid column dampers: experimental
study, J. Struct. Eng., ASCE 129 (2003) 960–971.
[25] S. Colwell, B. Basu, Experimental and theoretical investigations of equivalent
viscous damping of structures with TLCD for different fluids, J. Struct. Eng.,
ASCE 134 (2008) 154–163.
[26] J. Kim, K. Kim, H. Sohn, Autonomous dynamic displacement estimation from
data fusion of acceleration and intermittent displacement measurements,
Mech. Syst. Signal Process. 42 (2014) 194–205.
[27] C.C. Chang, X.H. Xiao, An integrated visual–inertial technique for structural
displacement and velocity measurement, Smart Struct. Syst. 6 (9) (2010)
1025–1039.
[28] K.-W. Min, S.-J. Jang, J. Kim, A standalone vision sensing system for pseudo-
dynamic testing of tuned liquid column dampers, J. Sens. 2016 (in press).