Professional Documents
Culture Documents
Effectiveness of A 2D TLD and Its Numerical Modeling: M. J. Tait, A.M.ASCE N. Isyumov, F.ASCE and A. A. El Damatty
Effectiveness of A 2D TLD and Its Numerical Modeling: M. J. Tait, A.M.ASCE N. Isyumov, F.ASCE and A. A. El Damatty
Effectiveness of A 2D TLD and Its Numerical Modeling: M. J. Tait, A.M.ASCE N. Isyumov, F.ASCE and A. A. El Damatty
Abstract: The tuned liquid damper 共TLD兲 is increasingly being used as an economical and effective dynamic vibration absorber to
mitigate the dynamic response of structures. The ability to design a TLD to operate in two directions simultaneously is particularly
attractive. In this paper the results of bidirectional 共2D兲 structure-TLD tests are reported on. These include the free-surface motion, the
Downloaded from ascelibrary.org by WALTER SERIALS PROCESS on 06/12/13. Copyright ASCE. For personal use only; all rights reserved.
resulting base shear forces, and the work done by bidirectional tuned liquid dampers 共2D TLD兲 attached to simple structures and response
displacements and accelerations of 2D structure-TLD systems. The importance of this experimental study is that it examines 2D structure-
TLD behavior over a range of excitation amplitude values covering the practical range of serviceability accelerations for buildings
subjected to wind loads. Experimental results are subsequently used to verify the applicability of a unidirectional structure-TLD numerical
model to 2D structure-TLD analysis. Findings indicate that the structure-TLD model is capable of describing the structure-TLD response
within the range of system response amplitudes experimentally tested.
DOI: 10.1061/共ASCE兲0733-9445共2007兲133:2共251兲
CE Database subject headings: Structural control; Damping; Passive control; Vibration; Numerical models.
冋冕 册
mated by
⬁ −1
where M *⫽generalized mass of the primary structure correspond-
eff = f s 兩H共f兲兩2df − s 共6兲
ing to the vibration mode being suppressed; and ⫽normalized 4 0
modal deflection value of the structure at the TLD location. The
absorber mass, mTLD, for a TLD with damping screens can be where 兩H共f兲兩⫽frequency response function of the structure with
approximated using potential flow theory 共i.e., mTLD ⬇ m1兲, where an attached vibration absorber; and s⫽the damping ratio of the
m1 is calculated using linear potential flow theory as 共Graham and primary structure.
Rodriguez 1952兲 Optimal tuning ratio, ⍀opt, and inherent damping ratio, opt,
冋 册
values expressed in terms of the mass ratio , have been deter-
h
Downloaded from ascelibrary.org by WALTER SERIALS PROCESS on 06/12/13. Copyright ASCE. For personal use only; all rights reserved.
冑
quency of the generalized structure having generalized mass, excited by a random white noise force to be
stiffness, and damping values of M *, K*, and C*, respectively. 1 + 2
The fundamental sloshing frequency of a TLD with damping eff-opt = 共7兲
4 3
screens designed to mitigate wind-induced motions to acceptable 1+
levels has been found to be weakly nonlinear for L / h values 4
ranging from 5 to 8 共Tait 2004兲. As a result an estimate of The efficiency, ⌿, of a TLD as a dynamic absorber is denoted
f TLD can be made using linear wave theory 共Warnitchai and Pin- here as the ratio of the experimentally determined effective damp-
kaew 1998; Ju et al. 2004; Tait et al. 2004a兲 if it is assumed that ing eff to that of an optimized linear TMD vibration absorber,
f TLD ⬇ f w, where f w is given by 共Lamb 1932兲 eff-opt, with a mass ratio value based on the total mass of the
fw =
1
2
冑 g
L
tanh
L
冉 冊
h
共4兲
water
⌿=
eff
eff-opt
⫻ 100 共8兲
achieve the same dynamic response as the structure-TLD system. The structure’s natural sway frequencies f sx and f sy in the two
Vickery and Davenport 共1970兲 calculated the value of eff by principal directions are given by
冑
The normalized probe locations are referenced from the center of
1 K*x the tank and are listed in Table 1. The motion of the structure was
f sx = 共11兲
2 M* recorded using four laser transducers, as shown in Fig. 2, allow-
ing both the sway and torsional motions to be measured. The
f sy =
1
2
冑 K*Y
M*
共12兲
accelerations of the structure were measured using two acceler-
ometers, one oriented along each principal direction of motion.
The actuator motion in each direction was recorded using a laser
Fig. 3共a兲 shows a 2D TLD located in the upper portion of a transducer.
rigid frame and a ballast mass in the lower portion of the rigid The properties of the 2D TLDs tested in this study are given in
frame. The TLD and ballast mass were suspended from the frame Table 2. Columns 2 and 3 describe the length of the tank in the x-
Fig. 3. 共a兲 TLD rigid test frame and load cells; 共b兲 TLD tank dimensions and damping screen configuration; and 共c兲 TLD damping screens
ties were applied to each system with the RMS values of the mental results to assess the validity of combining two orthogonal
applied actuator force, denoted as F, summarized in Table 4. A 1D numerical structure-TLD simulations to predict the response
total of four different tests were conducted for each load intensity of a 2D structure-TLD system.
level. For all load intensities tested the structure-TLD system
being investigated was first subjected to a 1D excitation in each Frequency Response
Downloaded from ascelibrary.org by WALTER SERIALS PROCESS on 06/12/13. Copyright ASCE. For personal use only; all rights reserved.
principal direction. Subsequently, the structure-TLD system was Structure-TLD system frequency response functions, 兩H共f兲兩, cal-
excited by applying the same excitation force in both directions culated using the measured applied excitation and system re-
simultaneously. This excitation case was fully correlated and de- sponse motions for Test Series ST1 for all load types 共X, Y, C,
noted as “C” in this paper. Finally, the 2D structure-TLD system and U兲 are shown in Figs. 4 and 5. Fig. 4共a兲 provides a compari-
test was performed using uncorrelated excitations, and is denoted son of the frequency response functions of the system in the
as “U.” In uncorrelated test cases the random forces applied in the x-direction, 兩Hx共f兲兩, for both 1D and 2D excitation tests. The fre-
principal axes were orthogonal and uncorrelated. All data were quency response functions from the 1D and 2D correlated and 2D
sampled at 15 Hz and an analog low-pass filter, with a cutoff limit uncorrelated tests are found to be in excellent agreement, indicat-
of 5.0 Hz, was used to suppress noise from the recorded signals ing the system’s x-direction dynamic response characteristics re-
and prevent aliasing. The sampling rate used in this study allows main effectively unchanged under bidirectional excitation. The
the response signal to be examined in both time and frequency same behavior is found for the y-direction dynamic response
domains. characteristics of the system by comparing 兩Hy共f兲兩 in Fig. 4共b兲.
For identification purposes, each test is described using three The same dynamic response behavior was observed for System
symbols. The first symbol, either ST1 or ST2, is used to identify ST2. This dynamic response behavior is expected to occur in a
the tested system. The second symbol 共LC1–LC5兲 describes the system having orthogonal 共uncoupled兲 modes.
load intensity 共excitation amplitude兲 values shown in Table 4. The As a consequence of the inherently nonlinear behavior of a
third symbol is either X, Y, C, or U and represents a 1D TLD, the effectively uncoupled behavior found for Load Intensity
x-direction, 1D y-direction, 2D correlated or 2D uncorrelated LC3 is examined over the range of excitation amplitudes consid-
loading type, respectively. For example, ST1-LC3-U represents a ered in this study. Figs. 5共a and b兲 show the frequency response
2D uncorrelated test conducted on System ST1 for Load Intensity functions of System ST1 in the x-direction for the minimum and
LC3. maximum load intensities tested. The average peak hourly accel-
eration values recorded for ST1-LC1 and ST1-LC5 test cases
were approximately 5 and 32 g⫻10−3, respectively. These re-
Experimental Results sponse accelerations adequately cover the serviceability limit cri-
teria range. The frequency response functions corresponding to
Normalized measured time histories of various measured param- the x-direction for ST1 are shown in Fig. 5共a兲 for Load Intensity
eters, including excitation forces, structural displacements, struc- LC1 and Fig. 5共b兲 for Load Intensity LC5, respectively.
tural accelerations, TLD free-surface motions and TLD base shear Figs. 4 and 5 show negligible difference in the frequency re-
forces, were recorded for both 1D and 2D excitations in both sponse behavior of the 2D structure-TLD system when subjected
principal directions. Frequency response functions of the to either 1D or 2D excitation. This type of system response be-
structural-TLD systems for both 1D and 2D excitation tests are havior indicates both perpendicular structural sway modes and
compared to determine if any significant changes in the response perpendicular fundamental TLD sloshing modes are effectively
behavior of the structure-TLD systems tested occurred. The TLD uncoupled. It should be recognized that coupling does occur be-
base shear forces are examined using a statistical approach to tween the structure and the TLD in both principal directions.
determine if the fundamental sloshing modes of the TLD are or- However, the structure-TLD system response in the x- and
thogonal or uncoupled under uncorrelated excitation. An energy y-directions is found to be effectively uncoupled. As the response
approach is employed to further investigate if the fundamental of the system has been found to remain uncoupled, the total re-
sway modes and fundamental sloshing modes of a structure-TLD sponse of the system under 2D excitation can be considered to be
system can be treated as effectively uncoupled. The effective the superposition of the response in each principal direction.
damping provided by a 2D TLD is calculated and its efficiency
level examined. Numerical simulations using a 1D nonlinear TLD Shear Forces
numerical model are carried out and results combined using su- The frequency response functions for the 2D structure-TLD sys-
perpositioning. Numerical solutions are then compared to experi- tem examined above, having two fundamental structural sway
remain constant under all loading combinations considered in this TLD Shear Forces Resulting from Bidirectional Excitation
study. To validate the assumption of effectively uncoupled funda- The normalized time histories of FTLDx in Fig. 6共a兲, FTLDy in
mental sloshing modes the TLD shear forces, f TLDx and f TLDy are Fig. 6共b兲, and those found in Fig. 6共c兲, corresponding to Test Case
examined for different excitation cases. ST1-LC5-C, are in excellent agreement. This shows negligible
Time history segments of the normalized shear forces, change occurred in the TLD shear forces when the system was
measured between the structure and the TLD, are shown in excited bidirectionally. The normalized time history segment of
Figs. 6共a–d兲 for ST1-LC5 for all four loading cases. Figs. 6共a–d兲 FTLDx shown in Fig. 6共d兲 for Test Case ST1-LC5-U is found to be
allow comparisons to be made between the shear forces that de- in excellent agreement with that of Test Cases ST1-LC5-C and
velop in the x- and y-directions corresponding to the fundamental ST1-LC5-X.
sloshing modes FTLDx and FTLDy under the four different excita- Two random variables are linearly independent if the correla-
tion cases considered. tion coefficient, xy is equal to zero. Furthermore, two uncorre-
Fig. 5. ST1 frequency response functions in x-direction for 共a兲 LC1; 共b兲 LC5
Wxx = 兺
i=1
FTLDx共ti兲⌬x共ti兲 共13兲
n
Wyy = 兺
i=1
FTLDy共ti兲⌬y共ti兲 共14兲
n
Wxy = 兺
i=1
FTLDx共ti兲⌬y共ti兲 共15兲
n
Wyx = 兺
i=1
FTLDy共ti兲⌬x共ti兲 共16兲
W⬘yy. Figs. 8共a and b兲 show the normalized work done by the shear 2D-TLD Efficiency
forces for the ST1-LC2-U and ST1-LC5-U test cases. The work
done by the terms W⬘xy and W⬘yx was also found to be less than 6% TLD efficiency in the x- and y-directions, calculated using Eq.
of the work done by the terms W⬘xx and W⬘yy. The negative work 共8兲, are shown in Figs. 9共a and b兲 as a function of the average
done by terms W⬘xy and W⬘yx indicates that there is a negligible hourly peak acceleration, âhourly. The average hourly peak accel-
transfer of energy into the “orthogonal” direction of the sloshing eration value is used in wind serviceability criteria as a measure
motion. of the acceptable wind-induced motion limits 共Isyumov 1995兲.
The findings from this study indicate that the orthogonal fun- The maximum efficiency is found to occur near the 20 g⫻10−3
damental sway modes and the orthogonal fundamental sloshing response level, which corresponds to the target value selected in
modes remain effectively uncoupled. This uncoupled response be- the design of the tested structure-TLD systems. Since TLD is
havior was observed for both ST1 and ST2 structural-TLDs under proportional to the square of the fluid velocity at the screens, the
all load intensities tested. No wave breaking was observed to TLD will have insufficient inherent damping at small system re-
occur for the load intensities considered in this study. From the sponse amplitudes resulting in decreased efficiency. When the
above findings, along with previous shake table studies 共Tait et al. structural response equals the target design response level, an
2005兲, a 2D TLD can be designed assuming the fundamental optimal value of TLD will be achieved resulting in maximum
sloshing modes in the principal directions are effectively un- TLD efficiency. For larger structural response amplitudes TLD
coupled. This has been found for both square and rectangular will exceed the optimal value and the efficiency of the over-
tanks where the free surface of the TLD does not experience wave damped TLD decreases. However, the onset of wave breaking
breaking. will be significantly suppressed.
Fig. 10. Normalized free surface response motions under 1D excitation: 共a兲 x-direction WP-7; 共b兲 y-direction WP-5
Fig. 11. Normalized free surface response motions at WP-9: 共a兲 correlated loading; 共b兲 uncorrelated loading
Additionally, Figs. 9共a and b兲 show that the 2D correlated and next time step. This is accomplished using a fourth-order Runge–
uncorrelated tests lead to similar efficiency values as a result of Kutta method. The equation of motion for the structure is ex-
the uncoupled response behavior in the two principal directions. pressed as
The TLD exceeds an efficiency level of 80% over a peak struc-
tural acceleration response range of 10–30 g⫻10−3, which is in M *ẍs共t兲 + C*ẋs共t兲 + K*xs共t兲 = Fe共t兲 + FTLD共t兲 共18兲
the 10-year range for offices and apartments 共Isyumov 1995兲, in
the x- and y-directions for both ST1 and ST2 structure-TLD sys- where M *, C*, and K*⫽mass, damping, and stiffness of the pri-
tems. The TLD design can be changed to target a different range mary structure; and xs⫽structural displacement. Fe共t兲⫽applied
of accelerations by changing the solidity ratio of the screens excitation; and FTLD共t兲⫽TLD shear force generated by the fluid
and/or the number of damping screens installed inside the tank. motion. The value of FTLD共t兲 is determined by calculating the
momentum change associated with the sloshing fluid over each
time step, as suggested by Keneko and Ishikawa 共1999兲. The
Numerical Simulation input excitation, Fe共t兲, used in the numerical model matched the
experimentally applied excitation forces allowing direct compari-
Results from the experimental study are compared to simulations sons to be made between the numerical simulations and experi-
of a TLD using a nonlinear numerical model based on shallow mental results. The structure-TLD system properties used in the
water wave theory 共Lepelletier and Raichlen 1988兲 with damping model were taken from values given in Tables 2 and 3.
screens 共Kaneko and Ishikawa 1999兲. The model simulates the The numerical structure-TLD system model was analyzed in
sloshing behavior of a TLD subjected to unidirectional excitation. each direction using 1D excitation, independently. The numerical
The interaction between the TLD and the structure is captured model is used to simulate the free-surface response at WP-6 and
by applying the resulting TLD shear forces, calculated by the WP-7 in the x-direction and WP-1 and WP-5 in the y-direction,
nonlinear fluid model, on the structure at each time step in whose locations are shown in Fig. 2. Additionally, the TLD forces
addition to the applied excitation force. The resulting structural resulting from the sloshing fluid motion and the structural dis-
acceleration is then used as the input excitation for the TLD in the placement and acceleration are calculated. Results obtained by
Fig. 12. Normalized TLD shear forces: 共a兲 x-direction 共correlated loading兲; 共b兲 y-direction 共correlated loading兲; and 共c兲 y-direction 共uncorrelated
loading兲
summing 1D numerical response time histories are used to predict and uncorrelated bidirectional excitation cases by summing the
various 2D response parameter time histories for different excita- 1D numerical time history results for WP-7 and WP-5 shown in
tion configurations applied in the tests. Fig. 10. Fig. 11 shows the estimated free-surface time histories
for both the correlated 关Fig. 11共a兲兴 and uncorrelated 关Fig. 11共b兲兴
Model and Experimental Response Comparisons excitation forces for ST1-LC3. The results are in good agreement
with experimental measured values, which confirms a 1D nonlin-
TLD Free-Surface Motions ear model can be used to predict free-surface response motions of
The 1D numerical model is first validated using the measured a 2D excited bidirectionally.
results from the 1D experimental tests. Fig. 10 compares the mea-
sured and simulated free-surface profiles at WP-7 and WP-5 lo- TLD Shear Forces
cated near the end walls of the tank for the ST1-LC3-X The measured shear forces that develop from the sloshing fluid
and ST1-LC3-Y test cases. The simulations are found to be in are compared with values calculated using the 1D numerical
good agreement with the experimental results, however, the nu- model. Fig. 12共a兲 shows the predicted normalized shear force
merical model overestimates the maximum free-surface response time history is in good agreement with the measured one in the
amplitude. x-direction for ST1-LC3-C. Figs. 12共b and c兲 compare the mea-
To confirm that the TLD sloshing modes are effectively un- sured and predicted shear force time histories in the y-direction
coupled, the free-surface response is summed at each time step for both the ST-LC3-C and ST1-LC3-U cases. For all excitation
to simulate the free-surface motion under 2D excitation. The amplitudes studied, good agreement is found between the pre-
free-surface response is calculated at WP-9 for both the correlated dicted and measured base shear forces.
Fig. 13. Normalized structural accelerations: 共a兲 x-direction 共correlated loading兲; 共b兲 y-direction 共correlated loading兲; and 共c兲 y-direction 共uncor-
related loading兲
U.K.
Lepelletier, T. G., and Raichlen, F. 共1988兲. “Nonlinear oscillations in
rectangular tanks.” J. Eng. Mech., 114共1兲, 1–23.
Fig. 15. Normalized resultant structural accelerations Preumont, A. 共1994兲. Random vibration and spectral analysis, Kluwer
Academic, Dordrecht, The Netherlands.
the response of a 2D structure-TLD system. Additionally, it can Read, D., Yu, J., Yeh, H., and Gardarsson, S. 共1998兲. “Investigation of
be used to accurately predict the free-surface motions and TLD tuned liquid dampers under large amplitude excitation.” J. Eng.
shear forces. Mech., 124共4兲, 405–413.
Shimizu, K., and Teramura, A. 共1994兲. “Development of vibration control
system using U-shaped water tank.” Proc., 1st Int. Workshop and
Conclusions
Seminar on Behavior of Steel Structures in Seismic Areas, Timisoara,
The performance of a 2D structure-TLD system has been inves- Romania, 25–34.
Soong, T. T., and Dargush, G. F. 共1997兲. Passive energy dissipation sys-
tigated in this paper. Experiments have been conducted to in-
tems in structural engineering, Wiley, New York.
vestigate the response characteristics of 2D structure-TLD sys-
Sun, L. M., Fujino, Y., Pacheco, B. M., and Isobe, M. 共1989兲. “Nonlinear
tems under random white noise excitation. An experimental ap-
waves and dynamic pressures in rectangular tuned liquid dampers:
paratus capable of modeling the fundamental sway modes of a
Simulation and experimental verification.” Struct. Eng./Earthquake
structure is presented. A 2D TLD is attached to the structure and
Eng., 6共2兲, 251s–262s.
subsequently the system is subjected to both 1D excitation and
Sun, L. M., Fujino, Y., Chaiseri, P., and Pacheco, B. M. 共1995兲. “The
2D correlated and uncorrelated excitation. properties of tuned liquid dampers using a TMD analogy.” Earth-
The response of a 2D structure-TLD system excited bidirec- quake Eng. Struct. Dyn., 24共9兲, 967–976.
tionally was found to correspond to the linear superposition of the Tait, M. J. 共2004兲. “The performance of 1D and 2D tuned liquid damp-
responses of two 1D structure-TLD systems. Findings from this ers.” Ph.D. thesis, Univ. of Western Ontario, London, Canada.
study indicate that by choosing the appropriate aspect ratio for the Tait, M. J., E1 Damatty, A. A., and Isyumov, N. 共2004a兲. “Testing of
TLD it can be used to reduce structural responses in two modes of tuned liquid damper with screens and development of equivalent
vibration simultaneously with no penalty on its performance. The TMD model.” Wind Struct., 7共4兲, 215–234.
efficiency of a 2D TLD was compared to that of an optimal linear Tait, M. J., Isyumov, N., and E1 Damatty, A. A. 共2004b兲. “The efficiency
TMD. Findings indicate that a 2D TLD can operate at near to and robustness of a unidirectional tuned liquid damper and modelling
90% efficiency at a target building acceleration of 20 g⫻10−3. In with an equivalent TMD.” Wind Struct., 7共4兲, 235–250.
addition, the efficiency of a 2D structure-TLD was found to re- Tait, M. J., El Damatty, A. A., and Isyumov, N. 共2005兲. “An investigation
main greater than 80% over a peak response acceleration range of of tuned liquid dampers equipped with damping screens subjected to
10–30 g⫻10−3. 2D excitation.” Earthquake Eng. Struct. Dyn., 34共7兲, 719–735.
A 1D nonlinear numerical model was used to predict the Tamura, Y., Fujii, K., Ohtsuki, T., Wakahara, T., and Kohsaka, R. 共1995兲.
response of a 2D structure-TLD system by assuming that the “Effectiveness of tuned liquid dampers under wind excitation.” Eng.
system can be represented by two uncoupled 1D structure-TLD Struct., 17共9兲, 609–621.
systems. The model is used to predict the free-surface response, Vickery, B. J., and Davenport, A. G. 共1970兲. “An investigation of the
the TLD shear forces, and the structural accelerations for both 2D behavior in wind of the proposed CentrePoint Tower, in Sydney, Aus-
correlated and uncorrelated loading cases. The 1D model is found tralia.” Research Rep. BLWT-1-70, Boundary Layer Wind Tunnel
to be capable of predicting 2D structure-TLD system response by Laboratory, Univ. of Western Ontario, London, Canada.
linear superpositioning. This greatly simplifies the design of a 2D Warburton, G. B. 共1982兲. “Optimum absorber parameters for various
structure-TLD system as it allows it to be designed as two effec- combinations of response and excitation parameters.” Earthquake
tively uncoupled 1D structure-TLD systems. Eng. Struct. Dyn., 10共3兲, 381–401.
Warnitchai, P., and Pinkaew, T. 共1998兲. “Modeling of liquid sloshing in
rectangular tanks with flow-dampening devices.” Eng. Struct., 20共7兲,
References 593–600.
Welt, F., and Modi, V. J. 共1992兲. “Vibration damping through liquid slosh-
Graham, E. W., and Rodriguez, A. M. 共1952兲. “The characteristics of fuel ing. Part 1: A nonlinear analysis.” Trans. ASME, J. Vib. Acoust.,
motion which affect airplane dynamics.” J. Appl. Mech., 19共3兲, 114共1兲, 10–16.