A Dissertation
Doctor of Philosophy
by
________________________________
July 2001
LIQUID DAMPERS FOR MITIGATION OF
STRUCTURAL RESPONSE: THEORETICAL DEVELOPMENT AND
EXPERIMENTAL VALIDATION
Abstract
by
The current trend toward structures of increasing heights and the use of light
weight, high strength materials and advanced construction techniques has led to more ﬂex
ible and lightly damped structures. Understandably, these structures are very sensitive to
environmental excitations such as wind, ocean waves and earthquakes, leading to vibra
tions inducing possible structural failure, occupant discomfort, and malfunction of eleva
tors and equipment. Hence, it has made it critical to search for practical and effective
The most commonly used passive device is the Tuned Mass Damper (TMD),
which is based on the inertial secondary system principle. A Tuned Liquid Damper (TLD)
is a special class of TMD where the mass is replaced by liquid (usually water). Tuned liq
uid column dampers (TLCDs) are a special type of TLDs that rely on the motion of a liq
uid column in a Utubelike container to counteract the forces acting on the structure, with
The thrust of this dissertation is to study and develop the next generation of liquid
dampers for mitigation of structural response. New modeling insights into the sloshing
Swaroop Krishna Yalla
phenomenon, which incorporate the effect of the liquid slamming/impact on the container
walls, are presented through experimental and analytical studies. The mechanical model
ing of TLDs is developed using a SloshingSlamming (S2) analogy and the use of impact
characteristics functions which can describe with high ﬁdelity the phenomenological
behavior of the damper. A major focus of this study is the design and development of
semiactive control systems which maintain the optimal damping level under different
loading conditions. Experimental validation of such a system was performed in the labora
actuator and positioning system. Finally, the design, implementation, cost and riskbased
decision analysis for the implementation of liquid dampers in structural vibration control
is presented.
DEDICATION
LIST OF TABLES...............................................................................................vi
LIST OF FIGURES...........................................................................................viii
ACKNOWLEDGEMENTS................................................................................xiv
CHAPTER 1: INTRODUCTION..........................................................................1
1.1 Introduction..............................................................................................................1
1.2 Literature Review.....................................................................................................4
1.3 Applications .............................................................................................................7
1.3.1 Ship/Offshore applications...........................................................................7
1.3.2 Structural Applications ..............................................................................11
1.4 Motivation of Present Work ...................................................................................16
1.5 Organization of Dissertation ..................................................................................18
ii
2.5 Concluding Remarks..............................................................................................40
iii
5.4.2 Example 3: MDOF system under random wind loading ...........................99
5.4.3 Example 4: MDOF system under harmonic loading ...............................102
5.5 Concluding Remarks............................................................................................106
iv
8.3.1 Decision analysis framework ...................................................................159
8.3.2 Reliability Analysis..................................................................................162
8.3.3 Cost and Utility Analysis .........................................................................165
8.3.4 Riskbased Decision Analysis..................................................................166
8.4 Design of Dampers ..............................................................................................167
8.4.1 Design Guidelines....................................................................................167
8.4.2 Control Strategy .......................................................................................169
8.4.3 Design Procedure .....................................................................................170
8.4.4 Technology...............................................................................................174
8.5 Concluding Remarks............................................................................................176
APPENDIX................................................................................................................... 181
REFERENCES.............................................................................................................184
v
LIST OF TABLES
TABLE 3.2 Comparison of optimal parameters for TMD and TLCD ..........................52
TABLE 3.3 Optimum parameters for white noise excitation for different mass ratios.53
TABLE 3.4 Optimum absorber parameters for FOF for different parameter ν1 ...........54
TABLE 3.5 Optimum absorber parameters for FOF for various mass ratios................54
TABLE 3.6 Optimum absorber parameters for SOF for different values of b1 ............57
TABLE 3.7 Optimum absorber parameters for SOF for various mass ratios................57
TABLE 6.1 Time lag and impact inﬂuence factor for different sensor locations........122
vi
TABLE 8.1 Component comparison of different DVAs..............................................156
vii
LIST OF FIGURES
Figure 1.1 (a) Frahm antirolling tank (b) nutation dampers in satellite applications...5
Figure 1.4 (a) Free surface damping tanks (b) Semiactive control for structure with
open bottom tanks......................................................................................10
Figure 1.6 (a) Schematic of TLDs installed in SYPH (b) Actual installation in the
building (taken from Tamura et al. 1995)..................................................12
Figure 1.7 (a) Liquid damper with pressure adjustment concept (b) photograph of
Hotel Cosima, Tokyo.................................................................................13
Figure 1.9 (a) Shanghai Financial Trade Center (b) 7 South Dearborn Project ..........15
Figure 2.1 (a) Equivalent mechanical model of sloshing liquid in a tank (b) Impact
damper model.............................................................................................26
Figure 2.2 Variation of (a) jump frequency and (b) damping ratio of the TLD with the
base amplitude (taken from Yu et. al 1999)...............................................27
Figure 2.3 Frames from the sloshing experiments video at high amplitudes: a part of
water moves as a lumped mass and impacts the container wall. (Video
Courtesy: Dr. D.A. Reed)...........................................................................28
Figure 2.4 Schematic diagram of the proposed sloshingslamming (S2) analogy.......29
viii
Figure 2.5 Comparison of experimental results with S2 simulation results: (a), (b):
experimental results; (c), (d): simulation results for Ω = 1.0 and 0.9.......32
Figure 2.6 (a) schematic of the jump phenomenon (b)Variation of the non
dimensionalized base shear force with the frequency ratio. (experimental
results taken from Fujino et al. 1992)........................................................33
Figure 2.7 Non dimensional interaction force curves for different η..........................36
Figure 3.4 Variation of dynamic magniﬁcation factor with the headloss coefﬁcient
and frequency ratio for a TLCD.................................................................47
Figure 3.5 Comparison of optimum absorber parameters for a TLCD with varying α
and a TMD.................................................................................................51
Figure 3.6 Transfer function of the ﬁlters and the primary system: (a) ﬁrst order ﬁlters
(b) second order ﬁlters...............................................................................55
Figure 3.9 Effect of damping ratio of the dampers on the frequency response of
SDOFMTLCD system..............................................................................61
Figure 4.1 Different coupled system (a) Vibration absorber (b) Coupled penduli
system (c) Electrical system (d) Fluid coupling within two cylinders.......66
Figure 4.2 Uncontrolled and Controlled response of a structure combined with (a)
TLD (b) TLCD...........................................................................................67
ix
Figure 4.4 Phase plane portraits of the undamped coupled system.............................69
Figure 4.5 Time histories of primary system displacement for α=0 and α=0.6 .........70
Figure 4.7 Time histories of response for ζ1=0.005 and ζ1=0.05 ...............................73
Figure 4.10 Modal frequencies and modal damping ratios of combined system as a
function of the damping ratio of the TLCD...............................................76
Figure 4.11 Phaseplane 3D plots (a) uncoupled system (b) case 1: undamped system
(c) case 2: system with damping in primary system only (d) case 3: system
with damping in both primary and secondary systems..............................77
Figure 4.13 Experimental free vibration response with different oriﬁce openings (θ = 0
fully open)..................................................................................................80
Figure 5.2 Flowchart of the two algorithms (a) iterative method (b) direct method... 84
Figure 5.3 Iterative method (a) convergence of response quantities (b) optimum
headloss coefﬁcient....................................................................................85
Figure 5.4 Variation of optimum headloss coefﬁcient with loading intensity: white
noise excitation..........................................................................................86
Figure 5.6 (a) Single degree of freedom idealization of the offshore structure (b)
Concept of Liquid Dampers in TLPs.........................................................89
Figure 5.7 Optimal Absorber parameters as a function of loading conditions............91
x
Figure 5.8 (a) Variation of Optimal headloss coefﬁcient with loading conditions for
different wave spectra (b) Spectra of structural acceleration at U10=20 m/s
for different ξ.............................................................................................92
Figure 5.11 Schematic of 5DOF building with semiactive TLCD on top story.........100
Figure 5.13 Displacements and Acceleration of Top Level under various control
strategies..................................................................................................102
Figure 5.14 Variation of performance indices with maximum headloss coefﬁcient... 104
Figure 5.15 Displacement of Top Floor under various control strategies ...................104
Figure 6.1 (a) Schematic of the experimental setup (b) pressure sensor locations... 110
Figure 6.2 Sample timehistories of the shear force at Ae = 0.3 cm and 2.0 cm....... 113
Figure 6.7 Pressure time histories for various frequency ratios (Ae = 1.0 cm). ........119
Figure 6.8 Probability distribution function of the peak impact pressures ..............120
xi
Figure 6.9 (a) Anatomy of a single pressure pulse (b) wavelet scalogram of the
pressure signal..........................................................................................121
Figure 6.10 (a) Pressure pulses at different locations on the wall (b) Wavelet
coscalograms with sensor 2 as reference.................................................124
Figure 6.11 Typical sloshing wave with pressure pulse and wave mechanism schematic
for (a) shallow water (h/a =0.12) and (b) deep water (h/a = 0.25) case..125
Figure 6.12 Variation of the peak pressure coefﬁcient with height of the tank wall...126
Figure 6.14 Schematic of the experimental setup for the HIL simulation ..................129
Figure 7.1 (a) Photograph of the Electropneumatic actuator (b) Schematic diagram
of the experimental setup........................................................................134
Figure 7.2 (a) Transfer functions for different tuning ratios (b) Variation of H2 norm
with tuning ratio.......................................................................................137
Figure 7.3 Transfer functions for different valve angle openings .............................138
Figure 7.4 Variation of transfer functions for different amplitudes of excitation..... 139
Figure 7.5 (a) Optimization of H2 norm (b) lookup table for semiactive control...140
Figure 7.6 (a) Comparison of transfer functions: (a) θ =40 deg, ζf = 9 % (optimal
damping) (b) θ = 60 deg, ζf = 30% (nonoptimal damping)....................141
Figure 7.7 3D plot of transfer function as a function of effective damping and
frequency (a) experimental results (b) simulation results........................142
Figure 7.8 Excitation time histories, valve angle variations and the resulting
accelerations for uncontrolled, passive and semiactive systems for time
history 1...................................................................................................144
Figure 7.9 Excitation time histories, valve angle variations and the resulting
accelerations for uncontrolled, passive and semiactive systems for time
history 2...................................................................................................145
xii
Figure 8.1 Implementation ideas for tuned liquid dampers (a) bridge towers (b) tall
buildings...................................................................................................149
Figure 8.2 TMD system installed in the Citicorp Building, New York City (taken
from Wiesner, 1979).................................................................................151
Figure 8.3 (a) Singlestage (b) multistage Pendulumtype TMDs (c) TMDs with
laminated rubber bearings (taken from Yamazaki et al. 1992)................152
Figure 8.5 Variation of RMS accelerations of the top ﬂoor with increasing wind
velocity.....................................................................................................159
Figure 8.9 (a) Equivalent white noise concept (b) Variation of equivalent white noise
with wind velocity....................................................................................172
Figure A.1 (a) Variation of Valve Conductance (b) Variation of headloss coefﬁcient
with the angle of valve opening...............................................................183
xiii
ACKNOWLEDGEMENTS
I would like to ﬁrst thank my advisor and guru, Prof. Ahsan Kareem, who pro
vided encouragement, support and friendship throughout the length of my stay at Notre
ment. I would also like to thank my committee members, particularly Prof. Bill Spencer
and Prof. Jeff Kantor, who guided me through many concepts in dynamics and control. I
would also like to thank Prof. Yahya Kurama and Prof. Steven Skaar for their valuable
guidance and constructive comments. I would also like to thank the staff in the Depart
ment of Civil Engineering and Geological Sciences, particularly Tammy, Molly and Chris.
Our laboratory technician, Brent Bach, helped me in most stages of the experiments.
Next, I would like to thank my family, both in India and the U.S., who have con
stantly supported me during my years in graduate school. Thank you Amma, Daddy,
Kumar, Chinni and others. I don’t know what I would have done without my friends: Cass,
Vicky, Adrish and all the other long lasting friendships I made at Notre Dame. Finally,
many thanks to the wonderful campus of the University of Notre Dame whose lakes,
Grotto and Fischer graduate apartments provided a home away from home and a wonder
xiv
xv
xvi
CHAPTER 1
INTRODUCTION
If they give you ruled paper, write the other way
 Juan Ramon Jimenez
________________________________________________________________________
This chapter begins with a brief literature review in the area of liquid dampers.
Relevant literature is also referenced at appropriate places in later chapters of the disserta
tion. Some of the applications of these dampers, especially in civil engineering structures
and offshore structures, are described. The motivation of the present research is presented
in the next section. Finally, the organization of the dissertation is laid out in detail.
1.1 Introduction
The current trend toward buildings of ever increasing heights and the use of light
weight, high strength materials, and advanced construction techniques have led to increas
ingly ﬂexible and lightly damped structures. Understandably, these structures are very
sensitive to environmental excitations such as wind, ocean waves and earthquakes. This
causes unwanted vibrations inducing possible structural failure, occupant discomfort, and
malfunction of equipment. Hence it has become important to search for practical and
effective devices for suppresion of these vibrations. This has opened up a new area of
research in the last decade, aptly titled structural control (Yao, 1972).
1
The devices used for mitigating structural vibrations are divided into separate cate
gories based on their system requirements (Housner et al. 1997). Passive control devices
are systems which do not require an external power source. These devices impart forces
that are developed in response to the motion of the structure, for e.g., base isolation, vis
coelastic dampers, tuned mass dampers, etc. More details of such systems can be found in
Soong and Dargush (1997). Active control systems are driven by an externally applied
force which tends to oppose the unwanted vibrations. The control force is generated
depending on the feedback of the structural response. Examples of such systems include
active mass dampers (AMDs), active tendon systems, etc (Soong, 1990). Owing to the
uncertainty of the power supply during extreme conditions and the large power source
needed to introduce control force, passive systems are generally favored over active ones.
Semiactive systems are viewed as controllable devices, with energy requirements orders
of magnitude less than typical active control systems. These systems do not impart energy
into the system and thus maintain stability at all times, for e.g., variable oriﬁce dampers,
electrorheological dampers, etc. A recent paper by Symans and Constantinou (1999) pro
Another paper by Kareem et al. (1999) describes the control systems for mitigation of
motion of buildings under wind loading. Alternative systems are being proposed which
derive the useful characteristics of both systems. One of them is hybrid control which
implies the combined use of active and passive systems or passive and semiactive sys
tems.
The most commonly used passive device is the Tuned Mass Damper (TMD),
which is based on the inertial secondary system principle, and consists of a mass attached
2
to the building through a spring and a dashpot. In order to be effective, its parameters need
damping through modiﬁcation of the combined structural system. Such systems have been
implemented, for example, in the John Hancock tower in Boston and the Citicorp Building
ably throughout this thesis) consists of a tank partially ﬁlled with liquid. Like a TMD, it
imparts indirect damping to the structure, thereby reducing response. The energy dissipa
tion occurs through various mechanisms: viscous action of the ﬂuid, wave breaking, con
tamination of the free surface with beads, and container geometry and roughness. Unlike a
TMD, however, a TSD has an amplitude dependent transfer function which is complicated
The TLDs can be broadly classiﬁed into two categories: shallowwater and deep
water dampers. This classiﬁcation is based on the ratio of the water depth to the length of
the tank in the direction of the motion. A ratio of less than 0.15 is representative of the
shallow water case. In the shallow water case, the TLD damping originates primarily from
energy dissipation through the action of the internal ﬂuid’s viscous forces and from wave
breaking. For the deepwater damper, bafﬂes or screens are needed to enhance damping.
The damping mechanism is therefore dependent on the amplitude of the ﬂuid motion,
wave breaking patterns, and screen conﬁguration. The deepwater damper has one draw
back in the fact that a large portion of water does not participate in sloshing and adds to
the dead weight. At an intermediate level of ﬁll depth, the container can be utilized for
building water supply. If the existing water tanks are not utilized, the large space occupied
3
by water containers may, in some cases, require a part of the building roof. However, most
practical installations of TLDs use many smaller tanks so as to maximize the effective
Tuned liquid column dampers (TLCDs) are a special type of TLDs relying on the
motion of the column of liquid in a Utubelike container to counteract the forces acting
on the structure, with damping introduced through an valve/oriﬁce in the liquid passage
(Sakai et al. 1989). The damping is amplitude dependent since the valve/oriﬁce constricts
TLDs were proposed in the late 1800s where the frequency of motion in two
interconnected tanks tuned to the fundamental rolling frequency of a ship was successfully
utilized to reduce this component of motion, as shown in Fig. 1.1 (Den Hartog, 1956). Ini
tial applications of TLDs for structural applications were proposed by Kareem and Sun
(1987); Modi et al. (1987) and Fujino et al. (1988). In the area of satellite applications,
Sakai et al. (1991) proposed a new type of liquid damper which was termed as a
tuned liquid column damper (TLCD) and described an application for cablestayed bridge
towers. TLCDs were studied for wind excited structures by Honda et al. (1991); Xu et al.
(1992) and Balendra et al. (1995). Studies were also made for determining certain optimal
characteristics of these passive devices by Gao et al. (1997); Chang and Hsu (1999); and
Gao et al. (1999). The performance of TLCDs for seismic applications has been studied
4
(a) (b)
Figure 1.1 (a) Frahm Antirolling tanks (b) Nutation dampers in satellite
applications
Most of the earlier studies concerned passive versions of TLCDs. This means that
the design involves no control of the damping characteristics. The damper was designed to
excitation. In order to solve this difﬁculty, semiactive and active systems were proposed
by Kareem (1994); Haroun et al. (1994); and Abe et al. (1996). A similar active system
was proposed for TLDs by Lou et al. (1994), in which a bafﬂe was placed inside the liquid
damper. The orientation of the bafﬂe changed the effective length of the damper thereby
Most structures under the inﬂuence of environmental loads experience both lateral
and torsional motions; therefore, one option is to have separate TLCDs each oriented in
particular directions, or to simply have a bidirectional Utube (Fig. 1.2(a)). This new con
ﬁguration consists of a box container with vertical tubes like a candelabrum concept, or a
partitioned container, consisting of stacked Utube sets ranging in both directions with a
common liquid base. The design eliminates the increased weight incurred by stacking two
5
independent orthogonal Utubes. One can also have oriﬁces between the partitions
(Kareem, 1993).
Multiple Mass Dampers (MMDs) with natural frequencies distributed around the
natural frequency of the primary system requiring control have been studied extensively
by Yamaguchi and Harnpornchai (1993); Kareem and Kline, (1994); and Yalla and
Kareem (2000). Such systems lead to smaller sizes of TLCDs which would improve their
construction, installation and maintenance, and also offer a range of possible spatial distri
butions in the structure. The tuned multiple spatially distributed dampers, offer a signiﬁ
cant advantage over a single damper since multiple dampers, when strategically located,
are more effective in mitigating the motions of buildings and other structures undergoing
buildings, a new bidirectional tuned liquid damper with period adjustment equipment.
Other adjustments in shape have been proposed by researchers. To help the damper liquid
maintain its column shape, a Vshaped TLCD can be adopted as shown in Fig. 1.2(b) (Gao
et al. 1997). Another variation of TLCD is proposed, which is termed as LCVA, which
pared to that of TLCD and is found to be as or even more effective. Other advantages
include versatility and architectural adaptability, since its natural frequency is determined
not only by the length of the liquid column but also the area ratio of the horizontal and ver
tical portion of the tube (Hitchcock et al. 1997; Chang and Hsu, 1998).
6
β
(b)
1.3 Applications
The operation of a ship is affected by the motions and forces induced by rolling,
which can cause cargo damage, discomfort to passengers and reduce crew efﬁciency. The
use of devices for stabilizing motion in ships dates back to 1862 when W. Froude intro
proposed the use of a Ushaped tank as a roll stabilizer. Since early installations of such
passive anti roll tanks in the 1950s, this concept has been applied widely on commercial
vessels. The latest ship stabilizers are capable of both heel and roll control using water
computer that constantly calculates the root mean square roll, the heel and the average
apparent roll period (Honkanen, 1990) There are three basic types of passive/ controlled
passive tanks, which are used for roll stabilization in ships, as shown in Fig. 1.3, namely:
7
free surface, Utube tanks and free ﬂooding tanks. Free surface tanks are open tanks and
can have bafﬂes/nozzle plates to provide internal damping. Different rolling frequencies
can be matched by changing the liquid level in the tank. Utube tanks consist of two tanks
partially ﬁlled with liquid, with the air spaces connected by a duct and a crossover duct at
the tank bottom. Damping is provided by restricting the ﬂow of air between the tanks. Free
ﬂooding tanks are not as popular as other tank systems. It is similar to a Utube tank except
that the tanks are not connected to one another; however, there is an airduct connecting the
top of the tanks. The tank natural period is set by the size of the inlet ducts relative to the
tank’s internal free surface. It is to be noted that all these stabilizers affect only the roll
amplitude and not the roll period (Sellars and Martin, 1992).
8
The excitations acting on most offshore structures are mostly due to wind, waves
and ocean currents. The sloshing motion of the liquids in storage tanks on ﬁxed offshore
structures affects its dynamic response. By prudent selection of the tank geometry, plat
form response may be reduced by using the tanks as dynamic vibration absorbers. There
fore, no new equipment is required, but only optimum conﬁguration of tankage that is
already required for storage of water, fuel, mud or crude oil (Vandiver and Mitome, 1978).
Passive, active and semiactive motion reduction systems such as ﬁn and tank stabilizers,
variable mooring systems, controlled and uncontrolled air cushions, perforated pontoons
and columns with gasspringlike tide tanks have been researched and applied to ﬂoating
platforms and other offshore structures like semisubmersibles (Ehlers, 1987). For ﬂoating
offshore structures like TLPs (tension leg platforms), the system with controllable moor
ing tension and variable attaching position are considered. The horizontal low frequency
motions of TLPs can be reduced by active control using dynamic positioning system
thrusters. Other mechanisms include active pulse generators, open bottom tanks and pres
surized passive air cushions. Control of offshore platforms using active mass dampers,
active tendons and thrusters can be found in Suhardjo and Kareem (1997).
Patel et al. (1985) considered a passive open bottom tank system in TLPs relying
upon the oscillations of the water columns in the tanks. A platform which lies on 46 col
Huse (1987) has studied free surface damping tanks to reduce resonant heave, roll and
pitch motions of semisubmersibles and other offshore structures. The damping tanks will
be situated at the water line and will be open to the sea through suitable restrictions
(Fig.1.4(a)). As shown in the ﬁgure, the tank is open to the sea and the atmosphere through
9
two openings. As the structure undergoes vertical motion, the sea water will ﬂow in and
out of the tanks. By choosing a suitable opening size relative to the free surface area of the
tank, the water level in the tank will ﬂuctuate with a certain phase lag relative to the verti
cal motion of the structure. This will produce a damping force which would reduce the
resonant heaving motion of the structure. Ehlers (1987) considers a semiactive control
method for a structure equipped with open bottom tanks, but the valves in the upper part
can be opened or closed (Fig.1.4 (b)). The relative vertical motion between the water col
umns in the tanks and the structure is inﬂuenced by the position of the valves because of
the air which is trapped in the tank when the valve is closed. These systems however, can
be used only for reduction of vertical motions and not horizontal motions. For some appli
cations, this is very important since damping in the vertical mode is extremely small.
Damping tanks
Valve
Detail
Elevation
Plan
(a) (b)
Figure 1.4 (a) Free surface damping tanks (b) Semiactive control for structure
with open bottom tanks
10
1.3.2 Structural Applications
There have been several applications of TLDs in Japan, an example of which is the
MCC Aqua DamperTM which was installed in the Gold Tower in Chiba, Japan (Fig. 1.5).
The Aqua Damper is a cubic tank ﬁlled with water in which steel wire nets are installed
across the water movement. The TLD frequency is adjusted by changing the length of the
tank and the depth of water. The damping, on the other hand, is adjusted by manipulations
of the damping nets. The top ﬂoor of the 158 m tall Gold Tower was installed with 16 units
of the Aqua Damper totalling 10 tons of water (approximately 1% of the tower's weight)
and has witnessed a improved response of 5060% of the original structural response prior
A battery of TLDs were installed in the Shin Yokohama Prince Hotel (SYPH) in
Yokohama, Japan (Fig. 1.6). The TLD system prescribed was a multilayer stack of 9 cir
11
Details of the system can be found in Tamura et al. (1995). Before and after the installa
tion of the TLD in March of 1992, fullscale measurements were taken to document the
performance of the auxiliary damping system. It was found that the RMS accelerations in
each direction were reduced 50% to 70% by the TLD at wind speeds over 20 m/s, with the
decrease in response becoming even greater at higher wind speeds. The RMS acceleration
without the TLD for the building was over 0.01 m/s2, which was reduced to less than
0.006 m/s2, deﬁned by the ISO as the minimum perception level at 0.31 Hz. Similar instal
lations are reported for Nagasaki airport tower, Tokyo international airport tower and
(a) (b)
Figure 1.6 (a) Schematic of TLDs installed in SYPH (b) Actual installation in the
building (taken from Tamura et al. 1995)
12
A TLCD has also been installed in the Hotel Cosima in Tokyo (Fig. 1.7). The hotel
is a 26 story steel building with a height of 106.2 meters. This building has a large height
to width ratio and is therefore wind sensitive. The foundation of the building is ﬁrmly con
nected to the ground using high strength steel pretensioned grout anchors. In addition, a
super structure is adopted as the frame of the building in order to resist earthquakes and
wind loads. The 58 ton TLCD with pressure adjustment, called MOVICS, was installed in
the top ﬂoor and has been observed to reduce the maximum acceleration by 5070% and
the RMS acceleration by 50% (Shimizu and Teramura, 1994). Other MOVICS systems
have been installed in the Hyatt Hotel in Osaka and the Ichida Building in Osaka.
(a) (b)
Figure 1.7 (a) Liquid damper with pressure adjustment concept (b) Installed in
Hotel Cosima, Tokyo
13
Recently, Liquid Dampers have been planned for the proposed Millennium Tower,
Tokyo Bay, Japan. Due to this supertall building’s exposure to typhoons, external damping
sources are needed to control the wind induced vibrations. In addition to massive steel
blocks at the top, there are water tanks with ducts between them. The water would provide
passive resistance under normal conditions, but under high winds, the sensors trigger a
pumping mechanism, changing the control mode from passive to active (Sudjic, 1993).
Figure 1.8 shows the schematic of the circular TLD concept in this tower.
14
A TLD is also planned to limit the wind induced motion of the proposed Shangai
Financial Trade Center in China. This building will have a square shaft with a diagonal
face that is shaved back (Fig. 1.9(a)). An aperture is cut out of the top to relieve aerody
namic pressure (Engineering News Record, May 1996). Both the TLD and the aerody
namic aperture will ensure to keep building motion within acceptable limits. TLDs are
also being considered for the newly proposed 2000 ft building in Chicago, namely, the 7
(a) (b)
Figure 1.9 (a) Shanghai Financial Trade Center (b) 7 South Dearborn Project
Liquid tanks are being used to reduce the aerodynamic forces, in particular the
(Brancaleoni 1992; Ueda et al. 1992). Liquid vibration absorbers are also used in tall
15
chimneys. These have been proven to be economical, can be easily adjusted to the physi
cal and architectural requirements, and are extremely failsafe. They are usually designed
as a part of the circular gangway or as a coupling body for the connecting forces of a
A recent paper by Hitchcock et al. (1999) describes the full scale installation of a
bidirectional passive liquid column vibration absorber (LCVA) on a 67m steel frame
communications tower. The LCVA is a passive system with no oriﬁce to control the damp
ing. The authors observed that “At wind speeds less than approximately 10 m/s, the stan
dard deviation of the tower acceleration before and after SLCVA system installation are
essentially the same due to the motion of the SLCVA liquid being insufﬁcient to dissipate
signiﬁcant vibrational energy. At wind speeds of approximately 20 m/s, the response of the
tower is reduced by almost 50% after installation of the SLCVA system.” This shows the
inadequacy of the passive systems to perform optimally at all levels of excitation. For e.g.,
16
at low amplitudes, the liquid velocity is insufﬁcient to generate an optimal value of damp
ing to reduce the motion substantially. On the other hand, at high amplitudes of excitation,
the damping introduced at the oriﬁce may be more than the optimal and again the efﬁ
ciency of the TLCD decreases. Similar observations were made in both experimental and
fullscale studies of Tuned Sloshing Dampers (TLDs) which rely on the sloshing of the
ture.
In the proposed research, new models for TLDs and TLCDs are developed. It has
been acknowledged by researchers that the sloshing of liquid at high amplitudes is a non
linear phenomenon. This work presents a new model using sloshingslamming analogy of
TLDs based on impact characteristics. The main thrust of this research is to develop the
next generation of liquid dampers. Control concepts are introduced in order to correct
some of the problems inherent in the existing dampers, mainly the potential of liquid
dampers not being fully realized due to their damping being dependent on motion ampli
tudes or the level of excitation. TLCDs are particularly attractive, in this regard, due to the
following reasons:
1. A mathematical model is available for the TLCD, due to which the tuning of the
damper is precise, and makes it amenable for semiactive and active control.
2. The amount of damping needed to suppress a particular vibration can be easily ascer
tained and controlled through the oriﬁce. The oriﬁce opening ratio affects the headloss
coefﬁcient which in turn affects the effective damping of the liquid damper. Propor
tional valves can be actuated by a small voltage signal to obtain the required damping.
17
3. Arbitrariness of shape, giving it versatility and adaptability for housing in available
4. The TLCD can be tuned by changing its frequency of the TLCD by way of adjusting
the liquid column in the tube. This is an attractive feature should the tuning become
The advantages of liquid damper systems include low cost and maintenance
because no activation mechanism is required. The liquid damper systems are easily mobi
lized at all levels of structural motion, whereas the mechanism activating a TMD must be
set to a certain threshold of excitation. The most important advantage, however is that such
containers can be utilized for building water supply, unlike a TMD where the dead weight
of the mass has no other functional use. A more elaborate cost analysis of the two systems
is presented in Chapter 8.
The next chapter discusses new modeling efforts for TLDs. A new sloshingslam
ming (S2) damper analogy has been developed for the sloshing dampers. This is based on
impact phenomenon; and secondly, explicitly including the impact characteristics in the
linearization technique. The optimum absorber parameters for TLCDs are determined for
various loading cases. The absorber parameters for multipleTLCDs are also determined.
18
Chapter 4 presents a common phenomenon which occurs in coupled system,
namely, the beat phenomenon. The focus of this chapter is to mathematically understand
Chapter 5 discusses the development of semiactive strategies for TLCDs. The efﬁ
ciency of the semiactive algorithms is illustrated through the use of appropriate examples.
istics are derived based on experimental studies. A new type of testing method, namely the
Chapter 8 deals with cost and reliability analysis for a tall building serviceability
under wind loading. Design guidelines and practical considerations are also delineated.
Chapter 9 discusses some of the important conclusions drawn from the present research
19
CHAPTER 2
MODELING OF SLOSHING
remember, when discoursing about water,
to induce ﬁrst experience, then reason.
 Leanardo Da Vinci
“SloshingSlamming (S2)” analogy which describes the behavior of the TLD as a linear
sloshing model augmented with an impact subsystem. The second model utilizes certain
nonlinear functions known as impact characteristic functions, which clearly describe the
nonlinear behavior of TLDs in the form of a mechanical model. The models are supported
2.1 Introduction
The motion of liquids in rigid containers has been the subject of many studies in
the past few decades because of its frequent application in several engineering disciplines.
The need for accurate evaluation of the sloshing loads is required for aerospace vehicles
where violent motions of the liquid fuel in the tanks can affect the structure adversely
(Graham and Rodriguez, 1952; Abramson, 1966). Liquid sloshing in tanks has also
tant for problems relating to safety, including tank trucks on highways and liquid tank cars
20
e.g., liquid cargo or liquid fuel, can cause loss of stability of the ship as well as structural
damage (Bass et al. 1980). In structural applications, the effects of earthquake induced
loads on storage tanks need to be evaluated for design (Ibrahim et al. 1988). Recently
however, the popularity of TLDs as viable devices for structural control has prompted
study of sloshing for structural applications (Modi and Welt 1987; Kareem and Sun 1987;
schemes based on linear and/or nonlinear potential ﬂow theory. These type of models rep
resent extensions of the classical theories by Airy and Boussinesq for shallow water tanks.
Faltinson (1978) introduced a ﬁctitious term to artiﬁcially include the effect of viscous dis
sipation. For large motion amplitudes, additional studies have been conducted by Lepelle
tier and Raichlen (1988); Okamoto and Kawahara (1990); Chen et al. (1996) among
others. Numerical simulation of sloshing waves in a 3D tank has been conducted by Wu
et al. (1998).
The model presented by Lepelletier and Raichlen (1988) recognized the fact that a
sloshing at higher amplitudes. Following this approach, a semianalytical model was pre
sented by Sun and Fujino (1994) to account for wave breaking in which the linear model
was modiﬁed to account for breaking waves. Two experimentally derived empirical con
stants were included to account for the increase in liquid damping due to breaking waves
and the changes in sloshing frequency, respectively. The attenuation of the waves in the
mathematical model due to the presence of dissipation devices is also possible through a
21
combination of experimentally derived drag coefﬁcients of screens to be used in a numeri
cal model (Hsieh et al. 1988). Additional models of liquid sloshing in the presence of ﬂow
dampening devices are reported, e.g., Warnitchai and Pinkaew (1998). The main disadvan
tage of such numerical models is the intensive computational time needed to solve the sys
Numerical techniques for modeling sloshing fail to capture the nonlinear behavior
of TLDs. This is due to the inability of theoretical models to achieve long time simulations
due to numerical loss of ﬂuid mass (Faltinsen and Rognebakke, 1999). Moreover, it is very
tions of impact pressures over the walls of the tanks requires the introduction of local
physical compressibility in the governing equations. The rapid change in time and space
recent work in numerical simulation of violent sloshing ﬂows in deep water tanks are
encouraging and represent the stateoftheart in this area, e.g, Kim (2001). However, until
the numerical schemes are more developed, one has to resort to mechanical models for
predicting the sloshing behavior. The chief advantages of a mechanical model are savings
ing would be to represent it using a mechanical model. This is helpful in combining a TLD
system with a given structural system and analyzing the overall system dynamics. Some of
the earliest works in this regard are presented in Abramson (1966). Most of these are lin
ear models based on the potential formulation of the velocity ﬁeld. For shallow water
22
TLDs, various mechanisms associated with the free liquid surface come into play to cause
energy dissipation. These include hydraulic jumps, bores, breaking waves, turbulence and
impact on the walls (Lou et al. 1980). The linear models fail to address the effects of such
Sun et al. (1995) presented a tuned mass damper analogy for nonlinear sloshing
TLDs. The interface force between the damper and the structure was represented as a
force induced by a virtual mass and dashpot. The analytical values for the equivalent mass,
frequency and damping were derived from a series of experiments. The data was curveﬁt
ted and the resulting quality of the ﬁt was mixed due to the effects of higher harmonics.
Other nonlinear models have been formulated as an equivalent mass damper system with
nonlinear stiffness and damping (e.g., Yu et al. 1999). These models can compensate for
the increase in sloshing frequency with the increase in amplitude of excitation. This hard
ening effect is derived from experimental data in terms of a stiffness hardening ratio. How
ever, none of these models explain the physics behind the sloshing phenomenon at high
amplitudes.
In contrast with the preceding models, Yalla and Kareem (1999) presented an
hardening sloshing system and the observed increase in the damping currently not fully
accounted for by the empirical correction for wave breaking. At high amplitudes, the
the container walls periodically. This is similar to the impact of breaking waves on bulk
heads observed in ocean engineering. None of the existing numerical and mechanical
23
models for TLDs account for this impact effect on the walls of the container. The sloshing
The sloshingslamming (S2) analogy is a combination of two types of models: the linear
analogy using lumped masses, springs and dashpots to describe liquid sloshing. The
lumped parameters are determined from the linear wave theory (Abramson, 1966). The
equivalent mechanical model is shown schematically in Fig. 2.1(a). The two key parame
8 tanh { ( 2n – 1 )πr }
m n = M l  ; n=1, 2........ (2.1)
π r ( 2n – 1 )
3 3
g ( 2n – 1 )π tanh { ( 2n – 1 )πr }
ω n =  ; n=1, 2......
2
(2.2)
a
where n is the sloshing mode; mn is the mass of liquid acting in that mode; ωn is the fre
quency of sloshing; r = h/a where h is the height of water in the tank; a is the length of the
tank in the direction of excitation; Ml is the total mass of the water in the tank; and mo is
∞
the inactive mass which does not participate in sloshing, given by m 0 = M l – ∑ mn .
n=1
Usually, only the fundamental mode of liquid sloshing (i.e., n = 1) is used for anal
ysis. This model works well for small amplitude excitations, where the wave breaking and
24
the inﬂuence of nonlinearities do not inﬂuence the overall system response signiﬁcantly.
This model can also be used for initial design calculations of TLDs (Tokarcyzk, 1997).
mass placed in a container ﬁrmly attached to the primary system, as shown in Fig. 2.1(b)
(e.g., Masri and Caughey, 1966; Semercigil et al. 1992; Babitsky, 1998). A gap between
the container and the impact damper, denoted by d, is kept by design so that collisions take
place intermittently as soon as the displacement of the primary system exceeds this clear
ance. The collision produces energy dissipation and an exchange of momentum. The pri
mary source of attenuation of motion in the primary system is due to this exchange of
momentum. This momentum exchange reverses the direction of motion of the impacting
M ẋ˙ + C ẋ + Kx = F e ( t ) (2.3)
mż˙ = 0 (2.4)
The velocity of the primary system after collision is given as (Masri and Caughey, 1966)
( 1 – µe ) µ(1 + e)
ẋ ac =  ẋ bc +  ż bc (2.5)
(1 + µ) (1 + µ)
where e is the coefﬁcient of restitution of the materials involved in the collision, µ=m/Μ is
the mass ratio, x and z represent the displacement of the primary and secondary system,
and the subscripts ac and bc refer to the aftercollision and beforecollision state of the
25
variables. The velocity of the impact mass is reversed after each collision. The numerical
ωt)
X(t)=Aexp(i
Xn
kn cn z
mn
d/2
X2 m
k2 c2
K
m2
k1 X1
c1
M Fe(t)
m1
C
x
mo
(a) (b)
Figure 2.1 (a) Equivalent mechanical model of sloshing liquid in a tank (b) Impact
damper model
by Fujino et al. (1992); Reed et al. (1998); Yu et al. (1999), etc. The key experimental
results are summarized in Figs. 2.2 (a) and (b), where the jump frequency and the damping
ratio are shown to increase with the amplitude of excitation. The jump phenomenon is typ
ical of nonlinear systems in which the system response drops sharply beyond a certain fre
quency known as the jump frequency. These results have been taken from Yu et al. (1999)
where the increase in damping and the change in frequency have been plotted as a function
26
Figure 2.2 (a) shows that there is an increase in the jump frequency (κ ) at higher
amplitudes of excitation for the frequency ratios ( γ f = ωe/ωf) greater than 1 suggesting a
hardening effect, where ωe is the frequency of excitation and ωf is the linear sloshing fre
quency of the damper. It has been noted that as the amplitude of excitation increases, the
energy dissipation occurs over a broader range of frequencies. This feature points at the
robustness of TLDs. The coupled TLDstructure system exhibits certain nonlinear charac
teristics as the amplitude of excitation increases. Experimental studies suggest that the fre
increased damping (introduced by wave breaking and slamming) causes the frequency
response function to change from a doublepeak to a singlepeak function. This has been
1.35
1.3
20
Jump frequency ratio,
1.25
Damping ratio (%)
κ
1.2
1.15 15
1.1
1.05
10
1
0.95
0.9 5
0.02 0.04 0.06 0.08 0.1 0.02 0.04 0.06 0.08 0.1
Non−dimensional Amplitude
e
/a A Non−dimensional Amplitude
e
/a A
(a) (b)
Figure 2.2 Variation of (a) jump frequency and (b) damping ratio of the TLD with
the base amplitude (Yu et al. 1999).
27
Figure 2.3 Frames from the sloshing experiments video at high amplitudes: a part
of water moves as a lumped mass and impacts the container wall. (Video Courtesy:
Dr. D.A. Reed)
28
SLOSHINGSLAMMING DAMPER ANALOGY
Secondary system
k1 m 1 x1
c1 (linear sloshing
Secondary system
mode)
(slamming mode) z
m2 mo
X
K
M FF(t)
e(t)
C
Primary system
(structure)
the liquid motion is characterized by slamming/impacting of water mass (Fig. 2.3). This
includes wave breaking and the periodic impact of convecting lumped mass on container
walls. Some of the energy is also dissipated in upward deﬂection of liquid along the con
tainer walls. The S2 damper analogy is illustrated schematically in Fig. 2.4. Central to this
analogy is the exchange of mass between the sloshing and convective mass that impacts.
This means that at higher amplitudes, some portion of the mass m1 (the linear sloshing liq
uid), is exchanged to mass m2 (the impact mass), which results in a combined sloshing
slamming action.
The level of mass exchange is related to the change in the jump frequency as
29
of the portion of linear mass m1 acting in the linear mode. Since the total mass is con
served, this implies that the rest of the mass is acting in the impact mode. For example, Ω
=1.0 means that all of the mass m1 is acting in the linear sloshing mode. After the mass
exchange has taken place, the new masses m 1 and m 2 in the linear sloshing mode and the
˜ ˜
impact mode, respectively, are given by
m 2 = m 2 + ( 1 – Ω )m 1 (2.6)
˜
m 1 = Ωm 1 (2.7)
˜
At low amplitudes, there is almost no mass exchange, therefore, the linear theory
holds. However, as the amplitude increases, γ decreases and the slamming mass increases
explains the hardening effect. The mass exchange parameter can be related to the jump
ω1 m1
2
k
ω1
2
frequency ratio. Since = 1 = 
 , therefore using Eq. 2.7, one can obtain
˜ m1 m1
˜ ˜
κ = 1 ⁄ Ω . The empirical relations as shown in Fig. 2.2(a) for relating the mass
scheme. This scheme can be further reﬁned should it become possible to quantify more
accurately the mass exchange between the sloshing and slamming modes from theoretical
considerations. The equations of motion for the system shown in Fig. 2.4 can be written as
M Ẋ˙ + ( C + c 1 ) X˙ + ( K + k 1 ) X – c 1 ẋ 1 – k 1 x 1 = F o sin ( ω e t )
m 1 ẋ˙1 + c 1 ẋ 1 + k 1 x 1 – c 1 X˙ – k 1 X = 0
m 2 ż˙ = 0 (2.8)
30
where F o = M A e ω e . After each impact, the velocity of the convecting liquid is changed
2
in accordance with Eq. 2.5. An impact is numerically simulated at the time when the rela
tive displacement between m1 and m2 is within a prescribed error tolerance of d/2, i.e.,
Since the relative displacements have to be checked at each time step, a time domain inte
gration scheme is employed to solve the system of equations. In order to construct the fre
quency response curves, the maximum steadystate response was observed at each
excitation frequency and the entire procedure was repeated for the complete range of exci
tation frequencies.
mental study (Fujino et al. 1992). These parameters are listed in Table 2.1. It should be
noted that the initial mass ratio, prior to the mass exchange, has been assumed to take on a
very small value, i.e., m 2 ⁄ m 1 = 0.01, which is essential to realize the system in Fig. 2.4
described by Eq. 2.8. This assumption is not unjustiﬁed since experimental results show
the presence of nonlinearity in the transfer function, albeit small, even at low amplitudes
of excitation (e.g., at Ae = 0.1 cm, κ = 1.02). Figure 2.5 shows the changes that take place
in the frequency response functions as the mass exchange parameter is varied. This can
also be viewed as the amplitude dependent variation in the frequency response function. It
should be noted that the frequency response function undergoes a change from a double
peak to a singlepeak function at higher amplitudes of excitation. This model gives similar
results as Fujino et al. 1992, however, one has to note that this is a mechanical model as
31
opposed to a numerical model described in Fujino et al. 1992. These results demonstrate
that the frequency response function of the combined system derived from the sloshing
slamming model is in good agreement with the experimental data both at low and high
amplitudes of excitation. Note that uncontrolled and controlled cases in Fig. 2.5 refer to
1 1
0.9 0.9
Uncontrolled
Normalised response
Normalised response
0.8 Uncontrolled 0.8
Controlled
Experimental Results
0.6 0.6
for high amplitudes of excitation
Experimental Results
0.5 for low amplitudes of excitation 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0.85 0.9 0.95 1 1.05 1.1 0.85 0.9 0.95 1 1.05 1.1
Frequency Ratio Frequency Ratio
(a) (b)
1 1
0.9 0.9
Uncontrolled
Normalised response
Normalised response
Uncontrolled
0.8 0.8 Controlled
Controlled
0.7 0.7
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0.85 0.9 0.95 1 1.05 1.1 0.85 0.9 0.95 1 1.05 1.1
Frequency Ratio
Frequency ratio
(c) (d)
Figure 2.5 Comparison of experimental results with S2 simulation results: (a), (b):
experimental results (Fujino et al. 1992); (c), (d): simulation results for Ω = 1.0
and Ω = 0.9
32
TABLE 2.1 Parameters of the model
enon as the amplitude of excitation increases. This jump phenomenon is typical of most
nonlinear systems, for e.g., dufﬁng, vanderpol oscillators, etc. A typical transfer function
shear of TLD is plotted for various amplitudes of excitation in Fig 2.6(b) (Fujino et al.
1992). The presence of jump and hardening phenomenon can be clearly observed. Fur
thermore, the range of frequencies over which the TLD is effective increases as the base
amplitude increases.
The S2 damper analogy cannot be directly applied to the liquid damper alone due
to the way it is formulated since to determine the postimpact velocity, one requires the
knowledge of the dynamics of the primary system. Therefore, in order to formulate a sin
gle model which explains the experimental results for both damper characteristic and the
coupled structuredamper system, one can take advantage of certain impact characteristics
which describe the effects of nonlinearities imposed by the slamming mass. When repeti
tive impacts occur as part of the vibratory motion of a linear system, the problem becomes
nonlinear. Having recognized this, one can search for such impactcharacteristic functions
33
which would produce the same nonlinearities in the linear system. This is studied in the
next section.
A e=0.1cm
A e=0.25cm
A e=0.5cm
20
A e=1.0cm
15
jump
frequency
10
ω
0
0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2
ωe/ωf
(a) (b)
Figure 2.6 (a) Jump phenomenon in nonlinear systems (b) Variation of the non
dimensionalized base shear force with the frequency ratio (experimental results
taken from Fujino et al. 1992).
on the container walls was simulated using the solution of differential equations, also
known as the pointwise mapping method. The impact was modeled as a collision between
the slamming (impact) mass and the tank wall as a discontinuous function. However, from
the extensive work done in the area of vibroimpact systems, it is known that the dynamic
model studied is a limiting case of a hardening type of nonlinear system not only in terms
of structure but also function. It is well known in vibroimpact literature that one can
34
model the impact behavior by considering impact characteristics instead of simulating
impacts by numerical integration schemes (Pilipchuk and Ibrahim, 1997; Babitsky, 1998).
Hence, the basic character of the nonlinear behavior for vibroimpact systems obtained
using “exact” methods are similar to typical nonlinear hardening systems. In fact, a very
simple model can phenomenologically describe the interaction between the liquid mass
and the tank wall with a nonlinear function. Having recognized this, one can search for
such impact characteristic functions which would produce the same effect as the solution
of differential equations. This equivalence was demonstrated for harmonic as well as ran
dom excitations (Masri and Caughey, 1965). It is to be noted that in this case, we will not
distinguish the liquid mass into impact mass and sloshing mass as done in the previous
section. The nonlinear model is developed for the entire liquid mass. Consider a oscillator
where Φ ( x, ẋ ) are the impact characteristics of the system, x is the displacement of the
lumped mass; ẋ is the velocity of the lumped mass; m, c and k are the mass, damping and
2
stiffness terms of the oscillator; Fo is the excitation amplitude = mω e A e . One can assume
placement and velocity. In particular, Hunt and Crossley (1975) presented nonlinear
impact characteristics whereby one can interpret the coefﬁcient of restitution as damping
p p2
Φ ( x, ẋ ) = b 1 x 1 ẋ + b 2 x (2.10)
35
where b1, b2, p1 and p2 are parameters of the model. However, for the sake of keeping the
amplitude of excitation. Accordingly Eq. 2.9 can be expressed in the following non
where ω f is the linear sloshing frequency and ς ( A e ) is the nonlinear damping of the
TLD. In this study, we will focus exclusively on shallow water TLDs, i.e. h/a < 0.15,
where h = depth of water and a = length of the tank in the direction of the excitation.
Various functions were considered for modeling the impact characteristics, e.g.,
hyperbolic sine function, power law function, and bilinear hardening type function. Fig
ure 2.7 shows the power law function used for modeling the impact characteristics. The
power law curve is used in this study since it allows for a ﬁnite value of the impact charac
teristic function at the boundaries of the wall, i.e., x = ± a ⁄ 2 . Note that the ordinate is the
2(η – 1)
F eff ( x ) = k eff ( x )x = mω f [ 1 + ϕ ( A e )x ]x
2
(2.13)
36
2
n=1
n=2
1.5 n=10
1
(xF
)
Non−dimensional Force,
eff
0.5
x
−0.5
keff
(x)
−1
a/2 a a/2
−1.5
−2
−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1
Displacement of sloshing mass,
/(a/2)x
tion of nonlinear systems. Moreover, one can represent these systems in transfer function
or statespace form to simplify the analysis by utilizing the linear systems theory. In the
next subsections we will brieﬂy look at equivalent linear models when the external excita
Φ ( x, ẋ ) = λ + υx + ψ ẋ (2.14)
The basic idea is to ﬁrst deﬁne an error function and minimize it in the mean square sense
over an inﬁnite time interval. One can write the error function as,
37
T
1
Θ ( λ, υ, ψ ) = lim  ∫ { Φ ( x, ẋ ) – λ – υx – ψ ẋ } dt
2
(2.15)
T → ∞T
0
∂ ∂ ∂
Utilizing the fact that Θ ( λ, υ, ψ ) = 0 ; Θ ( λ, υ, ψ ) = 0 and Θ ( λ, υ, ψ ) = 0
∂λ ∂υ ∂ψ
T T T
1 1 1
lim  ∫ x ( t ) dt = 0 ; lim  ∫ ẋ ( t ) dt = 0 and lim  ∫ x ( t )ẋ ( t ) dt = 0 (2.17)
T → ∞T T → ∞T T → ∞T
0 0 0
T
1
λ = lim  ∫ Φ ( x, ẋ ) dt (2.18)
T →∞ T
0
T
1 1
υ = 2 lim  ∫ Φ ( x, ẋ )x ( t ) dt (2.19)
σ x T → ∞T 0
T
1 1
ψ = 2 lim  ∫ Φ ( x, ẋ )ẋ ( t ) dt (2.20)
σ ẋ T → ∞ 0T
Θ ( λ, υ, ψ ) ≡ E ( { Φ ( x, ẋ ) – λ – υx – ψ ẋ } )
2
(2.21)
38
where E ( g ( x, ẋ ) ) represents the expected value of the random variable function g ( x, ẋ ) .
∞
λ = ∫ Φ ( u )w ( u ) du (2.22)
–∞
∞
1
υ = 2 ∫ uΦ ( u )w ( u ) du (2.23)
σ x –∞
∞
1
ψ = 2 ∫ u̇Φ ( u̇ )w ( u̇ ) du̇ (2.24)
σ ẋ –∞
where it is assumed that x and ẋ are independent Gaussian processes with probability dis
1 –u2
w ( u ) =  exp 2 (2.25)
σ u 2π 2σ u
Φ ( x, ẋ ) = Φ ( x ) + Φ ( ẋ ) (2.26)
one can obtain the coefﬁcients of equivalent linearization (for harmonic excitation) as,
η η 2η – 2
λ =  – 1 a x ; υ =  a x and ψ = 0
2η – 1
(2.27)
2 2
39
and for random excitation, using Eqs. 2.222.24,
η
λ = 0; υ = σx ∏ ( 2η – ( 2k – 1 ) ) and ψ
2η
= 0 (2.28)
k=1
The range of validity of this equivalent linearizations is discussed in the next chap
This analogy presents insights into the underlying physics of the problem and reproduces
the dynamic features of TLDs at both low and high amplitudes of excitation. At low
higher amplitudes, the model accounts for the convection of periodically slamming
lumped mass on the container wall, thus characterizing both the hardening feature and the
Next, based on the understanding of the sloshing and impact of the liquid, explicit
impact characteristics are introduced into the equations of motion in order to derive a sim
pler mechanical model. These impact characteristics introduce the necessary nonlineari
ties into the system. Such mechanical models will be useful for design and analysis of
TLD systems. Finally, equivalent linearization technique is used to derive linear models
40
CHAPTER 3
TUNED LIQUID COLUM DAMPERS
There is nothing more practical than a good theory
 T. Von Karman
In this chapter, tuned liquid column dampers (TLCDs) are discussed. First, the
mathematical model of the TLCD is presented and the equivalent linearized model is com
pared with the nonlinear model. Next, numerical optimization studies are conducted to
determine the important parameters for optimum TLCD performance, namely, the tuning
ratio and the damping ratio. In a later section, similar values of optimal parameters have
3.1 Introduction
In the classical work on the Dynamic Vibration Absorber (also known as TMD),
Den Hartog (1956) derived expressions for the optimum damping ratio and tuning ratio
(i.e., ratio of the absorber frequency to the natural frequency of the primary system) for a
parameters which minimize the displacement response of the primary system were found
to be simple functions of the mass ratio (ratio of mass of structure and damper).
McNamara (1977) reported design of TMDs for buildings with attention to experimental
studies and design considerations. Ioi and Ikeda (1978) developed empirical expressions
to determine correction factors for optimum parameters in the case of lightly damped
structures. Randall et al. (1981) and Warburton and Ayorinde (1980) further tabulated and
41
developed design charts for the optimum parameters for speciﬁed mass ratios and different
Previous work has been done with the aim of deriving optimum parameters for
TLCDs. Abe et al. (1996) derived optimum parameters using perturbation techniques.
Gao et al. (1997) studied numerically the optimization of TLCDs for sinusoidal excita
tions. Chang and Hsu (1998) have also discussed optimal absorber parameters for TLCDs
for undamped structure attached to a TLCD. These dampers were found to be effective for
wind loading (Xu et al. 1992; Balendra et al. 1995) and earthquake loading (Won et al.
In this chapter, similar expressions have been developed and parameters have been
tabulated for undamped and damped primary systems equipped with TLCDs. Usually, in
the design of TMDs for wind and earthquake excitations, the optimum parameters are cho
sen to be those obtained by assuming a white noise random excitation. In this study, in
addition to the white noise excitation, a set of ﬁltered white noise (FWN) excitation has
dampers), where the important design parameters are the frequency range of the dampers
and the damping ratio of the dampers (Yamaguchi and Harnpornchai, 1993; Kareem and
Kline, 1995). MTLCDs are useful because the efﬁciency is higher as compared to a single
TLCD and moreover, the sensitivity to the tuning ratio is diminished. Multiple liquid
dampers have also been studied by Fujino and Sun (1993); Sadek et al. (1998) and Gao et
al. (1999).
42
3.2 Modeling of Tuned Liquid Column Dampers
Figure 3.1 shows the schematic of the TLCD mounted on a structure represented
as a SDOF system.
b
Xf
xf headloss
l ξ
coefficient
Ks
Ms
XF(t)
s
Xs
Fe(t)
Cs
The equation describing the motion of the ﬂuid in the tube is given as (Sakai et al. 1989),
1
ρAl ẋ˙f ( t ) +  ρAξ ẋ f ( t ) ẋ f ( t ) + 2ρAg x f ( t ) = – ρAb Ẋ˙s ( t ) (3.1)
2
2g
where the natural frequency of oscillations in the tube are given by ω f =  . The equa
l
where X s = response of the primary system (structure); x f = response of the liquid damper
(TLCD); Ms = mass of the primary system; Ks = stiffness of the primary system; Cs=
ω s = natural frequency of the primary system; ρ= liquid density; A = cross sectional area
43
of the tube; l = total length of the liquid column; b = horizontal length of the column; g =
gravitational constant; ξ =coefﬁcient of headloss of the oriﬁce. The two equations can be
M s + m f αm f Ẋ˙s Cs 0 X˙ s Ks 0 Xs F e(t ) (l – b)
+ + = , x f ≤  , (3.3)
αm f mf ẋ˙f 0 cf ẋ f 0 kf xf 0 2
where α = length ratio = b/l; mf = mass of ﬂuid in the tube = ρAl; cf = equivalent damp
quency of the liquid damper; kf is the stiffness of the liquid column = 2ρAg, and F e ( t ) is
the external excitation. The constraint on Eq. 3.3 is placed so as to ensure that the liquid in
the tube maintains the Ushape and the water does not spill out of the tube, thereby
Using the expressions derived in section 2.4, one can obtain equivalent linear
damping for the nonlinear TLCD damping (cf). In particular, using Eq. 2.20 one can
obtain:
4ρAξ A e ω e
c f = 
 (3.4)
3π
2
cf =  ρAξσ ẋ f (3.5)
π
44
where σ ẋ f is the standard deviation of the liquid velocity. This analytical model will be
Since the equivalent damping will be used in later studies on TLCDs, it is useful to
study the accuracy of the equivalent linearization method. The two equations, written in
Nonlinear System:
2ω s ζ s 0 F e(t )
ωs 0
2
1 + µ αµ Ẋ˙s + X˙ s Xs 

ξ ẋ f + = Ms (3.6)
α 1 ẋ˙f 0  ẋ f 0 ωf
2 xf
2l 0
F e(t )
1 + µ αµ Ẋ˙s + 2ω s ζ s ωs 0
2
0 X˙ s Xs 

+ = Ms (3.7)
α 1 ẋ˙f 0 2ω f ζ f ẋ f 0 ωf
2 xf
0
where µ is the mass ratio = m f ⁄ M s . The nonlinear equations were simulated using the
nonlinear differential equation solver in MATLABTM, while for the linear equation, an
iterative method was used to solve the equivalent linearized equations. In the second case,
one ﬁrst assumes a value for σ ẋ f , simulates the linear system, recalculates the value of
σ ẋ f and iterates till the response quantity converges to an acceptable value. In this study,
the main focus is to examine the error between the exact nonlinear and linearized equation
for variations in the parameter ξ. The excitation used is a bandlimited Gaussian white
noise with a pulse width of 0.002 seconds and a spectral intensity of 0.01 m2 /sec3/Hz.
45
Figure 3.2 shows the comparison of the response of the structure and damper for various
headloss coefﬁcients. The maximum error between the nonlinear and the equivalent linear
system is about 2%. Figure 3.3 shows the time histories of the various response quantities
for ξ = 75.
4 9
Exact (Nonlinear)
Equivalent Linear
3.9 8
3.8 7
3.7 6
σX s 3.6 5
σ ẋ f
3.5 4
3.4 3
3.3 2
3.2 1
3.1 0
0 20 40 60 80 100 0 20 40 60 80 100
(ξ)
Coeff. of headloss (ξ)
Coeff. of headloss
10 3
Exact (Nonlinear)
Equivalent Linear
8
2
6
4 1
Xs xf
f
s
0
X
X
−2 −1
−4
−2
−6
−8 −3
0 20 40 60 80 100 0 20 40 60 80 100
time (sec) time (sec)
It has been observed from numerical studies that the headloss coefﬁcient affects
the structure’s frequency response curve. As the headloss coefﬁcient (ξ) increases, the
response curve changes from a double hump curve to a single hump curve (Fig. 3.4).
Numerical studies conducted by the author indicate that an optimal damping level exists
for the TLCD which depends on the excitation level and the head loss coefﬁcient. The ﬁrst
task, however, is to obtain the optimum damping ratio and tuning ratio of the absorber.
Dynamic
Magnification
Ratio
10
4 0
20
3
1.3 40
1.2
1.1 Coefficient of
60
Frequency ratio1 Head Loss
0.9
0.8 80
Figure 3.4 Variation of dynamic magniﬁcation factor with the headloss coefﬁcient
and frequency ratio for a TLCD
The analytical model was discussed in section 3.2. One can deﬁne transfer functions in the
X s( s) x f (s)
Laplace domain, namely H X s F ( s ) = 
 and H x f F ( s ) =  , where the following
F e( s) F e( s)
47
2
∆µαω – ω + 2ζ f ω f ( iω ) + ω f
2 2
H X s F ( ω ) = 
 and
[ – ω ( 1 + µ ) + 2ζ s ω s ( iω ) + ω s ] [ – ω + 2ζ f ω f ( iω ) + ω f ] + ω α µ
2 2 2 2 4 2
αω + ∆
2
H x f F ( ω ) = 

[ – ω ( 1 + µ ) + 2ζ s ω s ( iω ) + ω s ] [ – ω + 2ζ f ω f ( iω ) + ω f ] + ω α µ
2 2 2 2 4 2
where ∆ = 1 for base excitation in which case X s is the relative displacement, and
One can compute the response quantities of interest using random vibration analysis. In
particular, we are interested in the variance of the primary system displacement and the
variance of the liquid velocity in the TLCD. The response quantities are obtained as,
∞
σX s = ∫ H X s F ( ω ) S FF ( ω ) dω
2 2
(3.8)
–∞
∞
2
σ ẋ f ∫ω H x f F ( ω ) S FF ( ω ) dω
2 2
= (3.9)
–∞
where S FF ( ω ) is the power spectral density of the forcing function. Equation 3.9 is useful
in evaluating the equivalent damping of the TLCD from Eq. 3.5. A simpliﬁed solution to
the integral for random vibration analysis has been used to evaluate Eqs. 3.8 and 3.9 (see
Appendix A.1 for details). Three representative forcing functions have been studied here,
as listed in Table 3.1. The optimal absorber parameters are derived for each individual case
of white noise and FWN excitations. It will be shown in subsequent sections that typical
wind and earthquake excitations can be approximated through the use of such ﬁlters.
48
TABLE 3.1 Example forcing functions
Based on these three excitation models, optimal parameters have been obtained for
TLCD attached to damped and undamped primary systems. It has been seen that one can
derive an explicit expression for the case of undamped structureTLCD system subjected
to white noise. However, for damped systems and/or other excitations, the development of
damper system, like invariance points, do not exist when damping is introduced in the pri
mary system (Den Hartog, 1956). Therefore, the optimal absorber parameters (i.e., ζ f and
γ = ω f ⁄ ω s ) are obtained numerically for these cases. The optimal conditions are
obtained by setting:
∂σ x s ∂σ x s
2 2
 = 0 ;  = 0 (3.10)
∂ζ f ∂γ
One can obtain ζ opt and γ opt by solving the two conditions given by Eq. 3.10
In the case of tuned mass dampers, a detailed analysis was carried out by Warbur
ton (1982) to determine optimum damper parameters for the case of random excitations
(represented by white noise), with excitation applied to the structure (as in the case of
49
wind) or as a base acceleration (as in the case of ground motion). The design of TMDs for
wind and earthquake applications, therefore, uses these design expressions for the optimal
parameters. In the next subsections, the theory to determine the optimal parameters is
The response integral in Eqs. 3.8 and 3.9 can be cast in the following form:
∞
Ξ n ( ω ) dω
σ xs = S 0 ∫ 
2
 (3.11)
Λ n ( – iω )Λ n ( iω )
–∞
Solving the two optimization conditions in Eq. 3.10 and setting ζ s = 0 yields:
2µ
µ 1 + µ – α  α
2
α 4 1 + µ 1 – 
ζ opt 2
=  
 ; γ opt =  (3.12)
α µ 1+µ
2 2
( 1 + µ ) 1 + µ – 
2
In case, one can assume the tuning ratio to be equal to one, one can obtain a sim
1 µ(µ + α )
2
ζ opt =   (3.13)
2 (1 + µ)
This is justiﬁable because for the low mass ratios of the order 12% practical for tall build
ings, the tuning ratio is close to one, and in this case the optimal damping coefﬁcient given
by Eq. 3.13 approximates Eq. 3.12 quite well. Similar expressions exist for an optimal
damping coefﬁcient and tuning ratio of a TMD given by Warburton and Ayorinde (1980),
50
µ 1 + 
3µ µ
1 4 1 + 
2
ζ opt =   ; γ opt =  (3.14)
µ 1+µ
( 1 + µ ) 1 + 
2
2
Note that in all cases considered, the optimum damping coefﬁcient is independent
of the value of S0, the intensity of white noise excitation. It is noteworthy that Eq. 3.14
optimization criteria are summarized in Table 3.2 for TMDs and TLCDs. Figure 3.5 shows
the variation of optimum parameters as a function of the mass ratio. As the length ratio
increases, the damping ratio increases because there is more mass in the horizontal portion
of the TLCD. This contributes to indirect damping, which implies that it is better to keep
the length ratio as high as possible without violating the constraints of the TLCD or the
0.16
1
α =1.0
0.14
Optimum tuning ratio of the absorber
0.12
0.98
0.1
0.97
0.08
α = 0.1
0.96
0.06
0.95
0.04
51
TABLE 3.2 Comparison of optimal parameters for TMD and TLCD
TMD TLCD
Case number and
parameter optimized
γ opt ζ opt γ opt ζ opt
1 Random µ
〈 X s〉
2 2µ
µ 1 + 
2
α µ 1 + µ – α 
3µ
Force act 1 + 
4 1 + µ 1 –  4
2
 1 2 α
ing on 1+µ   
1+µ
 
µ
( 1 + µ ) 1 + 
2 2 2
α µ
Structure 2 ( 1 + µ ) 1 + µ – 
2
2 Random µ
〈 X s〉
2 µ 2µ
µ 1 – 
2
1 – 
1 + µ 1 – 
3α µ 1 – µ + 3α 
accelera 2
 1 4 2 α 4
tion at the 1+µ   
1+µ
 
µ
( 1 + µ ) 1 – 
2 2 2
3α µ
base 2 ( 1 + µ ) 1 + µ – 
2
mum damper parameters for a damped primary system; therefore, it must be estimated
and 5% and µ= 0.5, 1, 1.5, 2 and 5% and optimum absorber parameters are presented in
Table 3.3.
Table 3.3 shows that as the mass ratio increases, ζ opt also increases. Equation 3.12
veriﬁes this for undamped case, since it is approximately proportional to the square root of
the mass ratio. The tuning ratio also decreases as the mass ratio and the damping in the
primary system increase, which is consistent with the results obtained for tuned mass
52
dampers. It is observed that for small values of ζ s , ζ opt is not affected; therefore for a
lightly damped system, the optimum absorber parameters derived for an undamped pri
mary system are valid. For higher levels of damping in the primary system, one can derive
empirical expressions for the optimum damping ratio as a function of the primary system
damping ratio.
TABLE 3.3 Optimum parameters for white noise excitation for different mass
ratios.
Undamped
primary system 1% Damping 2% Damping 5% Damping
This type of function can be used to approximate windinduced positive pressures for the
alongwind loading. Figure 3.6 (a) shows the transfer functions of the ﬁrst order ﬁlter with
different values of the parameter ν 1 . Also shown for reference is the transfer function of
the primary system. Table 3.4 gives the optimum absorber parameters for these ﬁrst order
ﬁlters. Note that when ν 1 =10, the optimum parameters are the same as those obtained for
white noise, since the ﬁlter is fairly uniform like white noise excitation around the natural
53
frequency of the primary system. However, for other cases (e.g., ν 1 = 0.1 and 1), the opti
mum parameters are slightly different. The effect is more pronounced in the case of the
tuning ratio and increases as the damping in the primary system increases. Optimum
parameters have been computed for ν1 = 1 and tabulated in Table 3.5. Though the optimal
parameters can be obtained through the simultaneous solution of the two nonlinear equa
tions resulting from Eq. 3.10, the task becomes computationally intensive for the ﬁrst and
second order ﬁlters. In this numerical study, optimal parameters were obtained by utilizing
TABLE 3.4 Optimum absorber parameters for FOF for different parameter ν1
parameter of
ﬁrst order ﬁlter γopt ζ opt
ν1 = 0.1 0.991 0.04477
ν1 = 1 0.992 0.04476
ν1 = 5 0.9925 0.04483
ν1 = 10 0.993 0.04482
(These values are computed for undamped primary system with µ =1%)
TABLE 3.5 Optimum absorber parameters for FOF for various mass ratios.
Undamped
primary system 1% Damping 2% Damping 5% Damping
54
2
10 3
Filter parameter 10 Filter parameter
Filter Parameter
_____ ν1 = v=0.10.1 b1 =6
 ν1 =v=11 b1 =10
.......... ν1 = v=5
5 b1 =15
_._._. ν1= 10
Magnitude of transfer function
0
10 Transfer function of the primary system
0
10
−1 −1
10 10
−2
10
−2 −1 0 1
10 10 10 10
−2
10
−1 0 1 frequency rad/sec
10 10 10
frequency rad/sec
Figure 3.6 Transfer function of the ﬁlters and the primary system: (a) ﬁrst order
ﬁlters (b) second order ﬁlters
A general second order ﬁlter studied here has the following spectral description,
S 0 { c1 ω + d 1 }
2 2 2
S FF ( ω ) =  (3.16)
2 2 2 2 2
[ b1 – ω ] + a1 ω
where a1, b1, c1 and d1 are the parameters of the ﬁlter. Second order ﬁlters can be used to
represent earthquake and wind excitations. For earthquake representation, the excitation
acts at the base of the structure, while for wind representation, the excitation acts on the
structure. The expression in Eq. 3.16 also describes the well known KanaiTajimi spec
2 ω 2
S 0 1 + 4ζ g 
ω g
S FF ( ω ) =  (3.17)
ω 2 2
 2 ω 2
1 – ω + 4ζ g 
ω g
g
where ω g is the dominant ground frequency and ζ g is the ground damping factor.
55
Similarly, the acrosswind excitation can be modeled as a FWN using a second
order ﬁlter. Kareem (1984) has proposed the following empirical expression for the spec
nS FF ( z, n ) nδ


 = α β  for n ≤ n s
o o n s
σf
2
= α o β o  for n ≥ n s
n 3.0
(3.18)
n s
z 3.5
; β o = 1.32  + 0.154 1 – 
b̂ 1 0.5
where α o =  ; ns
3α̃ H
1 –  + 2b 
n 2 2 n 2
n s n s
SU ( z )
is the shedding frequency =  ; B is the breadth of the building; U ( z ) is the mean
B
speed at height z; S is the Strouhal number; σ FF is the mean square value of the ﬂuctuat
2
ing acrosswind force; α̃ is the exponent term in the power law of the wind velocity pro
ﬁle; H is the height of the building; b̂ is the band width coefﬁcient = 2I ( z ) , where I(z) is
the turbulence intensity at height z; and δ = 0.9. Details of this model can be found in
Kareem (1984). This acrosswind loading model can also be represented by Eq. 3.16.
The magnitude of the transfer function of the ﬁlter given by Eq. 3.16 is shown in
Fig. 3.6 (b) for parameters a1 = 0.01, c1 =1, d1 =10 and varying b1 = 6, 10, 15 and 20.
Table 3.6 shows how the optimal parameters are inﬂuenced as the ﬁlter parameter b1
changes. As b1 increases, the assumption of purely white noise becomes valid and the
solution approaches that for the white noise case. The other parameters have been kept the
same and optimal parameters have been computed for damped and undamped cases (Table
3.7).
56
TABLE 3.6 Optimum absorber parameters for SOF for different values of b1
parameter of
SOF γopt ζ opt
b1 = 6 1.05 0.1111
b1 = 10 1.01 0.0702
b1 = 15 1.00 0.0572
b1 = 20 0.995 0.0524
(All the other parameters are kept constant a1 = 0.01, c1 =1, d1 =10, µ =0.02 and ζ s =0.05)
TABLE 3.7 Optimum absorber parameters for SOF for various mass ratios.
a1 = 0.01 Undamped
b1 = 36 primary system 1% Damping 2% Damping 5% Damping
c1=1
d1=10 γopt ζ opt γopt ζ opt γopt ζ opt γopt ζ opt
µ=0.5% 1.04 0.1510 1.04 0.1401 1.045 0.1299 1.05 0.0956
µ=1% 1.04 0.1559 1.04 0.1450 1.045 0.1350 1.05 0.1008
µ=1.5% 1.04 0.1606 1.04 0.1498 1.045 0.1399 1.05 0.106
µ=2% 1.04 0.1654 1.04 0.1546 1.045 0.1448 1.05 0.1111
µ=5% 1.04 0.1927 1.04 0.1821 1.045 0.173 1.05 0.1406
increases and increases as the mass ratio increases; however, the damping in the primary
system affects ζ opt more in this case than in the case of white noise. In addition, the tun
ing ratio slightly departs from γ =1.00 as the damping in the primary system increases.
3.3.4 EXAMPLE
The optimum parameters for a TLCD placed on an eight story structure subjected
to an earthquake excitation are determined in this example using the theory presented in
the previous section. The parameters of the building stories considered are: ﬂoor mass =
345.6 tons, elastic stiffness = 34040 kN/m and internal damping coefﬁcient = 2937 tons/
sec, which corresponds to a 2% damping for each vibrational mode of the structure. The
57
computed natural frequencies are 5.79, 17.18, 27.98, 37.82, 46.38, 53.36, 58.53 and 61.69
rad/sec. The base excitation is modeled by the KanaiTajimi spectrum given in Eq. 3.17
with the parameters ω g = 10.5 rad/sec and ζ g = 0.317. The parameters of the general sec
mass of the damper has been taken as 2% of the ﬁrst generalized mass of the structure. In
Table 3.8, the optimum design damper parameters for the TMD have been compared with
TLCD parameters, both under the white noise and the SOF excitations. It is noted that
there are signiﬁcant differences in the optimum absorber parameters, justifying the inclu
sion of the anticipated loading in the optimization process for the damper design.
or distributed spatially. In this system, the natural frequencies of the TLCDs are distrib
uted over a range of frequencies. The advantages of a distributed system is that it is more
robust and effective for excitation frequencies distributed over a wide frequency band. In
and the secondary system, in this case, is the system of MTLCDs. The equations of
motion of the SDOFMTLCD system (Fig. 3.7) can be written in a matrix notation as:
58
T
m̃ s m f Ẋ˙s Cs 0 X˙ s Ks 0 Xs F e(t )
+ + = (3.19)
mf m ẋ˙fn 0 c eqn ẋ fn 0 k eqn x fn 0
∑ m fn
T
where m̃ s = M s + ; mf = α m f 1 m f 2 … m fN ;
n=1
mf1 0 0 0
0 mf2 0 0
m = ; c eqn and k eqn are (n, n) diagonal matrices similar to m .
0 0 … …
0 0 … m fN
The transfer function of the primary system is obtained by nondimensionalising Eq. 3.19,
1
HX ( ω ) = 
SF N N
2 µ fn
– ω 1 + ∑ µ fn + 2ζ s ω s ( iω ) + ω s + α ω ∑ 
2 2 4
n=1
2
n = 1 [ – ω + 2ζ fn ω fn ( iω ) + ω fn ]
2
αiω H X S F ( ω )
2
H x fn F ( ω ) = 
 ; n=1..N
[ – ω + 2ζ fn ω fn ( iω ) + ω fn ]
2 2
(∆ω)
range of MTLCDs
ωf1 ωfi
ωfN
. . . . . . . . . .
Ks
Fe(t)
Ms
Xs
Cs
The analysis of MTLCDs is similar to MMDs (multiple mass dampers), where the
important design parameters are the frequency range and damping ratio of the dampers
(Kareem and Kline, 1995). The frequency range is deﬁned as the total frequency span of
59
the MTLCDs given as ∆ω = ω fN – ω f 1 . The central damper (n = (N+1)/2) is tuned
exactly to the natural frequency of the primary system. It is assumed that N is an odd num
ber in this analysis. The frequency of each damper can be written as,
∆ω N +1
ω fn = ω s – n ; 1 ≤ n < 
N 2
N +1
= ωs ; n = 
2
∆ω N +1
= ω s + n ; N ≥ n > 
N 2
A numerical study has been conducted to examine the effects of the number of dampers,
frequency range and damping ratio of the dampers. Optimum values of these parameters
From Fig. 3.8, one can observe the ﬂattening action of MTLCDs as compared to
the double peaked response due to an STLCD. The effect of increasing dampers is similar
to that of adding damping: i.e., ﬂattening of the frequency response function. However, it
is also noted that the frequency response due to 5, 11 and 21 TLCD groups, for the partic
ular frequency range of 0.2, are very similar. This suggests that a large number of TLCDs
do not necessarily mean better performance, limiting the advantage of utilizing large num
The damping ratio of MTLCDs is studied for a group of eleven dampers with a
ﬁxed frequency range of 0.2 (Fig. 3.9). It is noted that at low damping ratios, the ampli
tude of the response function is spiked. As the damping ratio is increased, the response
60
function slowly becomes smoother and the amplitude decreases. After an optimal damp
ing ratio for the dampers is reached, any further increase in the damping ratio results in an
increase in the amplitude. This suggests that there exists an optimum damping ratio for a
Figure 3.10 shows the effect of changing the frequency range on the frequency
response function. It is can be seen from the plots that there is an optimum range where
the curve ﬂattens out over a range of frequencies. The frequency response functions of an
STLCD and a MTLCD with a low frequency range (0.02 and smaller) are similar. If the
range is smaller than the optimum, the frequency response of the MTLCD resembles that
of an STLCD, and so in a way, the MTLCD loses its effectiveness. This is intuitive
because there is a practical limit to which one can distribute the MTLCDs over a given fre
quency range. As this range becomes very small, MTLCDs act almost like an STLCD.
MTLCD conﬁguration (to control single mode of the structure) and MDOFMTLCD con
ﬁguration (to control multiple modes). The time frequency analysis of several earthquake
ground motion records utilizing wavelets has revealed the presence of higher frequency
components in the initial stages of the event, e.g., ElCentro (Gurley and Kareem, 1994).
In such cases, the presence of a TLCD or MTLCD tuned to the higher modes will be
Table 3.9 tabulates the optimum parameters of the different MTLCD system. One
can note that the optimum damping ratio decreases drastically for MTLCD groups as com
pared to an STLCD.
61
40
N=1
35 N=5
N=11
25
20
15
10
−5
−10
0.1 0.15 0.2 0.25
Frequency Hz
, ,
40
damping ratio ζfn
zs=0.0005
35
....... 0.0005
zs=0.005
−−−− 0.005zs=0.05
Magnitude of Transfer function (dB)
No damper
−.−.− 0.05 zs=0.5
30 _____ 0.5
25
20
15
10
−5
−10
0.1 0.15 0.2 0.25
Frequency Hz
Figure 3.9 Effect of damping ratio of the dampers on the frequency response of
SDOFMTLCD system
62
, ,
40
frequency range (∆ω)
range=0.02
35
......range=0.05
0.02
..range=0.1
0.05
25
20
15
10
−5
−10
0.1 0.15 0.2 0.25
Frequency Hz
using a simpliﬁed solution to the integral occurring in the estimation of the mean square
response, has been presented. SDOF systems subjected to the white noise and ﬁltered
white noise excitations utilizing ﬁrst and second order ﬁlters have been analyzed, and the
optimum absorber parameters for TLCDs have been determined numerically based on the
63
minimization of the RMS displacement of the primary system. This work can be extended
to MDOF systems for which a state space approach can be used and the response covari
ance matrix in the case of white noise can be obtained by solving the Lyapunov equation.
In the case of FWN excitations, the procedure remains the same except that the primary
Explicit expressions for optimal parameters are only feasible for a simple
undamped primary system subjected to white noise. As the systems and forcing functions
become more complex, numerical solutions are needed to evaluate the optimal parameters.
It has been seen that for lightly damped systems, the optimal damping coefﬁcient
of the absorber does not depend on the damping coefﬁcient of the primary system when
the excitation is purely white noise. However, for the ﬁrst and second order FWN cases, it
is affected by the primary system damping. This suggests that the damping in the primary
system plays a role in determining the optimum damping coefﬁcient of the TLCD.
Although the undamped case may yield an approximate value of the optimal parameters,
the primary system damping and knowledge of the excitation must be included for accu
rate estimates.
Optimal absorber parameters have been determined in the case of multiple TLCDs.
These parameters include the number of TLCDs, the frequency range and the damping
ratio of each damper. It is seen that there is an upper limit on the number of TLCDs,
beyond which additional TLCDs in the MTLCD conﬁguration do not enhance the perfor
mance. MTLCDs are more robust as compared to an STLCD and the smaller value of the
optimal damping makes them more attractive for liquid dampers which have a limited
range of damping. The small size of individual TLCDs in a MTLCD conﬁguration offers
structureliquid damper systems. Transfer of energy takes place in the coupled system
which could induce vibrations in the primary structure instead of suppressing them. This
Numerical and experimental results are presented in this chapter to elucidate the beat phe
4.1 Introduction
The beat phenomenon has been discussed in many classical texts on vibration
(e.g., Den Hartog, 1956). Figure 4.1 shows coupling present in different mechanical and
electrical systems. It is well known that beats occur when two frequencies are close
together. This usually occurs when the coupling is very soft in comparison to the main
“springs”. In an electrical analogue, this means larger capacitance of the coupling than the
main capacitances. Transfer of energy takes place in the coupled system which could
provided insightful understanding into the behavior of liquid damper systems. The motiva
tion of this paper is portrayed in Figs. 4.2 (a) and (b), which show the free vibration decay
65
of a combined structureTLD and TLCD in the laboratory. The controlled response
(a) (b)
(d)
(c)
Figure 4.1 Different coupled system (a) Vibration absorber (b) Coupled penduli
system (c) Electrical system (d) Fluid coupling within two cylinders
However, beyond a certain level of damping in the TLCD, this beat phenomenon
ceases and the structural response resembles a SDOF decay. Of course, as a limiting case
one might expect this to happen because when the damping is very high in the secondary
system, the combined system essentially behaves as a SDOF system. However, the critical
66
This chapter delves into better understanding the beat phenomenon for the combined
structureTLCD system.
Uncontrolled
0.8 TLD TSD
Controlled with
Response of Structure
0.6
0.4
0.2
−0.2
−0.4
−0.6
−0.8
−1
0 1 2 3 4 5 6 7 8 9 10
Time (sec)
(a)
Response of Structure
1.5
Uncontrolled
controlled with TLCD
1
0.5
−0.5
−1
−1.5
0 1 2 3 4 5 6 7 8 9 10
Time (sec)
(b)
Figure 4.2 Uncontrolled and Controlled response of a structure combined with (a)
TLD (b) TLCD
67
4.2 Behavior of SDOF system with TLCD
In this section, three different cases are considered as shown in Fig. 4.3. These are
undamped combined system; damped primary system with undamped secondary system;
and damped primary and secondary system. We will look at each case in detail. In order to
keep the discussion general, the subscripts 1 and 2 are introduced instead of s for structure
x2 x2 x2
m2 k2 m2 m2 c2 k2
k1 k1 k2 k1
m1 m 1 m 1
x1 x1 x1
c1 c1
The coupled equations of motion without damping in the primary and secondary system
(Fig 4.3 (a)) can be obtained from Eq. 3.6 by setting damping in each system equal to zero,
1 + µ αµ ẋ˙1 + ω 1 0 x 1 = 0
2
(4.1)
α 1 ẋ˙2 0 ω2 x2
2 0
ω1 + ω2 ( 1 + µ ) ± Π
2 2
ϖ 1, 2 = 
 (4.2)
2(1 + µ – α µ)
2
68
2
where Π = ( ω 1 – ω 2 ( 1 + µ ) ) + 4ω 1 ω 2 α µ
2 2 2 2 2 2
It is obvious from Eq. 4.2 that, for an uncoupled system (i.e., for α=0), the eigenvalues
reduce to:
ω1
ϖ 1 = 
 ; ϖ2 = ω2 (4.3)
1+µ
The coupling parameter α in the mass matrix is responsible for the beat phenomenon.
Figure 4.4 shows the phase plane portraits for the primary system for different values of α.
Unless mentioned otherwise, all units of displacements, frequencies and velocities are m,
rad/sec and m/sec, respectively. The ﬁrst portrait shows that with no coupling there is only
one frequency at which the structure responds, and as the coupling parameter increases
there is interference between the two states of the primary system, namely, x 1 and ẋ 1 .
0.1 0.1
α=0 α=0.1
0.05 0.05
dx /dt
dx /dt
0 0
1
−0.05 −0.05
−0.1 −0.1
−0.01 −0.005 0 0.005 0.01 −0.01 −0.005 0 0.005 0.01
x x
1 1
0.1 0.1
α=0.6 α=0.9
0.05 0.05
dx /dt
dx /dt
0 0
1
−0.05 −0.05
−0.1 −0.1
−0.01 −0.005 0 0.005 0.01 −0.01 −0.005 0 0.005 0.01
x x
1 1
69
For all simulations in this chapter, the following parameters have been kept constant, ω1=1
Hz, µ=0.01 and ω2=0.99 Hz. Figure 4.5 shows the time histories of the displacement of
the undamped primary system for α=0 and α=0.6. When coupling is present between the
0.005
1
0
x
−0.005
−0.01
−0.015
0 2 4 6 8 10 12 14 16 18 20
t
0.015
α=0.6
0.01
0.005
1
0
x
−0.005
−0.01
−0.015
0 2 4 6 8 10 12 14 16 18 20
t
Time (sec)
Figure 4.5 Time histories of primary system displacement for α=0 and α=0.6
To understand this phenomenon better, one can consider the solution of the system of
equations given in Eq. 4.1. After some mathematical manipulation the displacement of the
ẋ 2 ( 0 ) = 0 , is given by:
ωBt ωAt
x 1 ( t ) = x 0 cos  cos  (4.4)
2 2
70
where ω A = ϖ 1 + ϖ 2 and ω B = ϖ 2 – ϖ 1 , which means that the resulting function is an
varying with a frequency of ω A . This undamped combined system case has been exam
In this section, a linearly damped primary system with undamped secondary system as
shown in Fig. 4.3(b) is considered. Accordingly, the equations of motion are given by:
1 + µ αµ ẋ˙1 + 2ω 1 ζ 1 0 ẋ 1 + ω 1 0 x 1 = 0
2
(4.5)
α 1 ẋ˙2 0 0 ẋ 2 0 ω2 x2
2 0
2 2
λ 1, 2 = – ϖ 1 ζ̃ 1 ± i ϖ 1 1 – ζ˜1 and λ 3, 4 = – ϖ 2 ζ̃ 2 ± i ϖ 2 1 – ζ˜2 ,
where ϖ 1, 2 are the modal frequencies and ζ̃ 1, 2 are the modal damping ratios. The aver
age frequency and the beat frequency are plotted in Fig. 4.6 for different damping ratios of
the primary system. At α = 0, the beat frequency (i.e. the difference in modal frequencies)
tends to be zero. As the coupling is increased, there is an increase in the beat frequency
which causes the beat phenomenon. From this analysis, one can conclude that there is no
beat phenomenon when the difference in the modal frequencies approaches zero. Figure
4.6 also shows the effect of introducing damping in the primary system. At high levels of
damping ratio, there is a wider range of coupling term α which results in the beat fre
quency being equal to zero. This means that, over this range of the coupling term, there is
71
hardly any beat phenomenon. For α = 0.3, beat phenomenon is present when the damping
ratio in the primary system is 0.005, but it disappears when the damping ratio is 0.05. Fig
ure 4.7 shows the effect of damping in the primary system on the response of the primary
system. As the damping ratio increases, the response dies out in an exponential decay.
However, the beat phenomenon still exists. This poses difﬁculty in the estimation of sys
0.7
0.6 ζ1 = 0
ω ,
B
ζ1 = 0.005
Beat Frequency
0.5
0.4 ζ1 = 0.05
0.3
0.2
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
α
Coupling parameter
12.56
ω ,
A
Average Frequency
12.54
12.52
12.5
72
0.01
ζ1 = 0.005
1 0.005
0
x
−0.005
−0.01
0 5 10 15 20 25 30 35 40
t
0.01
ζ1 = 0.05
0.005
1
0
x
−0.005
−0.01
0 5 10 15 20 25 30 35 40
Time(sec)
At this stage, the effect of a decrease in beat frequency on the response signal can
be further examined. Figure 4.8 shows that as ωB approaches zero, TB (the time period of
the beat frequency) becomes very large. The parameter inﬂuencing the decay function is
Ψ (for a SDOF system, Ψ = ζ 1 ω 1 ). As a result, due to the damping in the primary sys
tem, the response dies out before the next peak of the beat cycle arises. Therefore, the
73
exp(Ψt)
X ωA t/2)
cos(
ωB t/2)
cos(
=
x 1 = exp(Ψt) ωA t/2)
cos( ωB t/2)
cos(
Time (sec)
In this section, the system represented by Fig 4.3 (c) is considered, where now an oriﬁce in
the middle of the Utube imparts damping to the system. In this case, the following equa
1 + µ αµ ẋ˙1 + 2ω 1 ζ 1 ω1 0 x1
2
0 ẋ 1
+ = 0 (4.6)
α 1 ẋ˙2 0 ω 2 ξ ẋ 2 ⁄ 4g ẋ 2
2
0 ω2 x2
2 0
1
where ξ is the headloss coefﬁcient and c 2 =  ρAξ . Equation 4.6 is numerically integrated
2
at different levels of the headloss coefﬁcient and setting ζ 1 = 0.001 and α=0.3 (Fig 4.9).
The ﬁgure shows an interesting behavior of the liquid damper system. In the previous sec
74
tion, the damping simply caused an exponential decay of the beat response. However, in
this case, the beat phenomenon disappears after a certain level of the headloss coefﬁcient.
Since an analytical solution is not convenient for this equation due to the quadratic nonlin
earity in the damping associated with the secondary system, a linearized version (see sec
tion 3.2.1) of this system is generally considered. Therefore, Eq. 4.6 is recast as:
1 + µ αµ ẋ˙1 + 2ω 1 ζ 1 0 ω 0 x1
2
ẋ 1
+ 1 = 0 (4.7)
α 1 ẋ˙2 0 2ω 2 ζ 2 ẋ 2 0 ω2 x2
2 0
0.01
ξ = 0.2
0.005
1
0
x
−0.005
−0.01
0 5 10 15 20 25 30 35 40
t
0.01
ξ = 2
0.005
1
0
x
−0.005
−0.01
0 5 10 15 20 25 30 35 40
t
0.01
0.005 ξ = 50
1
0
x
−0.005
−0.01
0 5 10 15 20 25 30 35 40
t
Time (sec)
75
The modal frequencies and damping ratios of the system deﬁned in Eq. 4.7 are plotted in
Fig. 4.10 as a function of equivalent damping ratio, ζ 2 . Figure 4.10 explains the disap
pearance of the beat phenomenon due to coalescing of the modal frequencies after a cer
tain value of the equivalent damping ratio. As seen in the previous chapter, this change in
equivalent damping ratio is realized through changing of the headloss coefﬁcient. The
resulting beat frequency approaches zero and hence beat phenomenon ceases to exist. This
is similar to a previous case where there was no beat phenomenon for coupling term α = 0,
1.02
ζ 1 = 0.001
Modal frequencies
1.01 α = 0.3
µ = 0.01
1
ϖ1,2
0.99
0.98
0.97
0 0.01 0.02 0.03 0.04 0.05 0.06
ζ
Equivalent damping ratio,
2
0.06
Modal damping ratio
0.05
0.04
~
ζ 0.03
1,2
0.02
0.01
0
0 0.01 0.02 0.03 0.04 0.05 0.06
ζ
Equivalent damping ratio,
2
Figure 4.10 Modal frequencies and modal damping ratios of combined system as a
function of the damping ratio of the TLCD
76
Figure 4.11 shows the three dimensional plots of state space portraits as a function of time.
Figure 4.11(a) shows the evolution for an uncoupled system in which the amplitude of
response is constant. Figures 4.11(b) and (c) show the cases discussed in sections 4.2.1
and 4.2.2. The ﬁnal plot, Fig. 4.11(d), shows case 3 in which no beat phenomenon occurs
No coupling CASE 1
20 20
time (sec)
time (sec)
10 10
0 0
0.1 0.1
0.01 0.01
0 0
0 0
dx1/dt −0.1 −0.01 x dx1/dt −0.1 −0.01 x
1 1
(a) (b)
CASE 2 CASE 3
20 20
time (sec)
time (sec)
10 10
0 0
0.1 0.1
0.01 0.01
0 0
0 0
dx /dt −0.1 −0.01 x dx /dt −0.1 −0.01 x
1 1 1 1
(c) (d)
Figure 4.11 Phaseplane 3D plots (a) uncoupled system (b) case 1: undamped
system (c) case 2: system with damping in primary system only (d) case 3: system
with damping in both primary and secondary systems
77
4.3 Experimental Veriﬁcation
In order to further validate the observations made in section 4.2, a simple experi
ment was conducted using the experimental setup shown in Fig. 4.12. A TLCD is mounted
on a SDOF structure. The TLCD was designed with a variable oriﬁce, to effectively
change the headloss coefﬁcient. At θ = 0 degrees, the valve is fully opened and the head
loss is increased with an increase in the angle of rotation, θ. In Fig. 4.13, one can note the
presence of a beat pattern for low headloss coefﬁcients. However, as the headloss coefﬁ
cient is increased, the beat phenomenon disappears and an exponentially decaying signa
ture is obtained. A similar observation was made in Fig. 4.9 for simulated time histories.
78
θ is the angle of
valve rotation
θ =0 degrees
0.5
0
x
1
−0.5
0 5 10 15
Time
0.5
θ =15 degrees
x 0
1
−0.5
0 5 10 15
Time
θ =60 degrees
0.5
x 0
1
−0.5
0 5 10 15
Time (sec)
Figure 4.13 Experimental free vibration response with different oriﬁce openings
(θ = 0 fully open)
exhibits the beat phenomenon due to the coupling term that appears in the mass matrix of
the combined system. The free vibration structural response resembles an amplitude mod
ulated signal. The beat frequency of the modulated signature is given by the difference in
the modal frequencies of the coupled system. However, beyond a certain level of damping
in the secondary system (liquid damper), the beat phenomenon ceases to exist. This is
79
attributed to the coalescing of the modal frequencies of the combined system to a common
80
CHAPTER 5
SEMIACTIVE SYSTEMS AND APPLICATIONS
If you wish to control the future, study the past...
 Confucius
This chapter describes different semiactive strategies developed for optimal func
tioning of TLCDs. These strategies include gainscheduling and clipped optimal schemes
with continuouslyvarying and onoff control. It is shown that such systems provide a sig
niﬁcant improvement over the performance of a passive system. Numerous examples and
5.1 Introduction
Hrovat et al. (1983). In other ﬁelds such as automotive vibration control, considerable
research has been done on semiactive systems (Ivers and Miller, 1991; Karnopp, 1990). A
number of devices are currently being studied in the area of structural control, namely the
variable stiffness devices, controllable ﬂuid dampers, friction control devices, ﬂuid vis
cous devices, etc. Recent papers in this area provide a stateoftheart review of semi
active control devices for vibration control of structures (Spencer and Sain, 1997; Symans
Optimization studies discussed in chapter 3 show that there exist optimal damping
and tuning ratio, which lead to high performance of TLCDs. One of the main features of
these dampers is that the damping is nonlinearly dependent on the amplitude of excitation.
81
This chapter proposes two strategies which can improve over the performance of passive
systems. One of them involves gainscheduling of the damping based on the feedforward
information of the disturbance. The other is a clipped optimal system with continuously
varying and onoff control, which involves a continuos changing of the damping based on
This section discusses a semiactive system which is useful for disturbances which
are of long duration and slowly varying (e.g., wind excitations) and where steadystate
response is the controlling objective. The optimal head loss coefﬁcient as a function of the
loading intensity is described as a lookup table. As the loading intensity changes, the
headloss coefﬁcient of the TLCD is changed in realtime in accordance with this lookup
ear regulator whose parameters are changed as a function of the operating conditions in a
preprogrammed manner. As shown in Fig. 5.1, the regulator is tuned for each operating
consuming to design, its regulator parameters can be changed very quickly in response to
system changes. This kind of control is more commonly used in aerospace and process
82
operating
condition
regulator Gain
parameters Schedule
Command
output
signal control
Process
Regulator
The procedure for estimating the optimum damping coefﬁcient, ζ opt , for TLCDs
under a host of loading conditions was outlined in chapter 3. In this section, methods to
determine the optimal headloss coefﬁcient (ξopt) is presented. This is the parameter
responsible for introducing damping in the liquid column of the TLCD. The statistical lin
earization method gives the following expression for the equivalent damping (assuming
2
cf =  ρAξσ ẋ f (5.1)
π
Equation 5.1 suggests that since σ ẋ f increases as the loading increases, therefore, in order
to maintain the optimal damping, ξ must decrease. Hence, there exists an optimal head
loss coefﬁcient at each loading intensity. These variations deﬁne the damping characteris
tics of the oriﬁce needed at different excitation levels. An iterative method has been used
in previous studies, (Balendra et al. 1995) since the damping term depends on σ ẋ f which
83
is not known a priori. An alternative, which is a direct method is developed in this study.
This involves evaluation of ζopt following the procedure outlined in the previous sections.
This value is then substituted into Eq. 3.8 to obtain σ ẋ f . One can then determine ξopt using
Eq. 5.1. Figure 5.2 provides a step by step ﬂowchart for the two methods. Figure 5.3 (a)
shows a typical iterative method for an SDOFTLCD system subjected to white noise
excitation, where σ X s and σ ẋ f are calculated by Eqs. 3.8 and 3.9. This is repeated for a
Figure 5.2 Flowchart of the two algorithms (a) iterative method (b) direct method
white noise excitation with tuning ratio close to unity, can be obtained. The optimum value
of the damping coefﬁcient for this case reduces to the expression given in Eq. 3.12. After
84
2 3⁄2
(1 + µ – α µ) µ + α
2
ξ opt = µ   glω d µ (5.2)
S0 1+µ
0.2 0.012
0.18
 RMS structure’s displacement 0.0115
.. RMS liquid velocity
0.16 __ ζf
0.0105
α = 0.9
0.12 ωs=1 rad/s
0.01
l =19.6 m
0.1
0.0095
0.08
0.009
0.06
0.0085
0.04
0.008
0.02 0 10 20 30 40 50 60 70 80 90 100
Coefficient of headloss
Optimum value = 37.5
Same value is obtaine
0
0 2 4 6 8 10 12 14 16 18 20 Direct method.
iterations
Figure 5.3 Iterative method (a) convergence of response quantities (b) optimum
headloss coefﬁcient
For tuning ratios not equal to unity, one can obtain similar expressions. However,
they are cumbersome and can be obtained numerically. It is noteworthy from Eq. 5.2 that
the optimum headloss coefﬁcient is indirectly proportional to the square root of the inten
sity of white noise. Using some representative values, it can be shown that the direct (Eq.
5.2) and the iterative methods yield the same values (Fig. 5.3 (b)). However, the direct
method is computationally superior, since it does not require iterations, making it more
Figure 5.4 shows the variation in the optimum headloss coefﬁcient for various
mass ratios of an SDOFTLCD system under white noise excitation case. It is noted from
these curves that at high loading intensities, very low headloss coefﬁcients are needed. For
85
typical oriﬁce characteristics, this corresponds to a hundred percent oriﬁce opening ratio,
i.e., the oriﬁce should be fully open. At high amplitudes of excitation, it is, therefore, bet
ter to keep the oriﬁce fully open and let the damping be provided by the liquid velocity.
For low amplitudes of excitation, the liquid velocity is inadequate, therefore, the oriﬁce
opening should be decreased (thereby increasing ξ). The relationship between the oriﬁce
opening ratio and the headloss coefﬁcient for standard oriﬁces can be found in the litera
Mass ratio 5%
Mass ratio 2%
80 Mass ratio 1%
Optimum Coefficient of Headloss
70
Parameters :
60
ωs=1 rad/sec
α=0.9
50 ζs=1 %
l =19.6 m
40
30
20
10
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Spectral loading intensity So −3
x 10
Figure 5.4 Variation of optimum headloss coefﬁcient with loading intensity: white
noise excitation
5.3 Applications
section. The ﬁrst example is for an SDOFTLCD under random white noise excitation.
The second example discusses the application of these dampers to an offshore structure.
The efﬁciency of the gainscheduled control can be seen from Fig. 5.5. The look
up table deﬁned in Fig. 5.4 is used to introduce the semiactive control law. The parame
86
ters of this system are as shown in Fig. 5.4. The efﬁciency of the passive TLCD is
improved as the intensity of the white noise excitation changes from So = 106 m2/sec3/Hz
to So =104 m2/sec3/Hz (Table 5.1). Note that in the ﬁrst segment of the loading, the per
formance of the semiactive and the passive system coincide with each other.
0.1
So=1e06 So=1e04
0.08
0.06
0.04
displacement (m)
0.02
−0.02
ξ1
ξ2
−0.04
−0.06
Uncontrolled 90
Semi−Active Control
Optimum Coefficient of Headloss
70
Parameters :
−0.1 60
0 20 40 60 80 100 120 140 160 ω 180
=1 rad/sec
s
α=0.9
200
ζs=1 %
time 50
l =19.6 m
40
30
20
10
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Spectral loading intensity So −3
x 10
(Numbers in brackets indicate improvement of each control strategy over uncontrolled case)
87
5.3.2 Example 2: Application to Offshore Structure
The forces acting on most offshore structures are due to wind, waves and ocean
currents. The motion of offshore structures is highly undesirable as it causes fatigue and
shutdown of operations. In this section, a TLCD is proposed for control of offshore struc
tures. The offshore structure has been idealized as a SDOF system as shown in Fig. 5.6(a).
ocean waves acquire additional mass and damping referred to as added mass and hydrody
namic damping. The mass, stiffness and damping can be written as (Brebbia et al. 1975):
1 1
M = l c ρ c A c ∫ [ f ( z ) ] dz + C M l c ∫ [ f ( z ) ] dz + M c
2 2
(5.3)
0 0
EI 1 ∂
2 2
K = 3 ∫ 2 f ( z ) dz (5.4)
lc 0 ∂ z
K
ωs =  (5.5)
M
8 1
C = C s + C D  ∫ σ V˙ [ f ( z ) ] dz
2
(5.6)
π 0
∞ ∞ 2
ω  S η̃η̃ ( ω ) dω
2 cosh kz
σ V˙ = ∫0 S V˙ V˙ ( ω ) dω = ∫0
2
(5.7)
sinh kD
z
where z =  , lc is the length of the column, k = ω ⁄ g , g is the acceleration due to grav
2
lc
ity, ω is the frequency, f ( z ) is the assumed shape of the column, EI is the equivalent
stiffness of the column, Ac is the equivalent area of the column, ρ is the density of water,
Mc is the mass of the platform, CD, CM and CA are the drag and inertia coefﬁcients, and
88
S η̃η̃ ( ω ) is the spectra of wave elevation. The forcing function under the action of linear
η̃ ( C M + C A )ω D
2
F ( ω, t ) =  ∫ cosh ( kz ) f ( z ) dz (5.8)
sinh ( kD ) 0
8 ω D
+ η̃C D   ∫ cosh ( kz )σ V˙ f ( z ) dz
π sinh ( kD ) 0
The shape of the deﬂected platform is approximated as f ( z ) = z and hence the mass of
2
the system is calculated using Eq. 5.3 as M = 7.72 X 106 Kg and stiffness, K = 9 X 106 N/
m using Eq. 5.4. This results in a natural frequency of the structure, ωs = 1.07 rad/s. The
total damping ratio of the structure is evaluated using Eq. 5.6 which is equal to 6%. The
drag and inertia coefﬁcients for the equivalent column are: C D = c d ρD ⁄ 2 = 5000 Kg/
Mc
TLCD
l
(a) (b)
Figure 5.6 (a) Single degree of freedom idealization of an offshore structure (b)
Concept of Liquid Dampers in TLPs
89
TABLE 5.2 Numerical parameters used: Example 2
Numerical Numerical
Parameter Value Parameter Value
Depth of water, D 75 m EI value 2250 X109 Nm2
Mass of Platform, Mc 2 X106 Kg Density of water, ρ 1000 Kg/m3
Length of Structure, lc 100 m length of liquid damper, l 18 m
Cross sectional Area, Ac 28 m2 Area of damper (with µ=2%), A 8.8 m2
Total Volume of water displaced 78 m3/m Density of Concrete, ρc 2500 Kg/m3
per unit length, VW
The wave spectrum used in this study is the Pierson and Moskowitz (PM) spectrum,
α1 g
2
g 4
– β 
S η̃η̃ ( ω ) = 
 exp  (5.9)
1 ωU
ω
5
where U is the wind speed at 10 meters above the sea surface and α1 , β1 are dimension
less parameters which determine the shape of the spectrum. For the North Sea, the value of
α1 = 0.0081 and β1 = 0.74. In the frequency domain, the expression for the forcing func
tion can be derived from Eq. 5.8, which can be written as,
(C M + C A) ω D
2 4
2
2 ∫0
S FF ( ω ) = S η̃η̃ ( ω )  cosh ( kz ) f ( z ) d z (5.10)
sinh ( kD )
8C D ω
2 2
2
+ 2 ∫ cosh ( kz )σ V˙ f ( z ) dz
D
π sinh ( kD ) 0
Figure 5.6 (b) shows a schematic of the possible design of liquid dampers func
tioning as pontoon water tanks of the Tension Leg platform (TLP). The wave forcing func
tion on such platforms may not be ideally described by Eq. 5.8. This is because the size of
the platform in comparison with the wave length of approaching waves is large, which
results in diffraction of waves. Therefore, in this case the ﬁrst component of the forcing
90
Optimal parameters are obtained using numerical optimization, as done previously
ratio of RMS structural accelerations with and without the damper). As shown from Fig.
5.7, there exists optimum damper parameters, which are found to be independent of the
loading conditions (i.e., different U10). Therefore, under all loading conditions, these
parameters must be maintained at their optimal values, otherwise the performance of the
1 1
U =50 m/s U =20 m/s
10 10
U =40 m/s U 10=30 m/s
10
U =30 m/s U 10=40 m/s
10
U =20 m/s U =50
10
m/s
10
0.95 0.95
Absorber Efficiency
Absorber Efficiency
0.9
0.9
0.85
0.85
0.8
0.8
Optimum Damping Optimal Tuning
ratio ratio
Next, one can easily apply the gainscheduled law described in the previous sec
tions for semiactive control. The lookup table can be generated as shown in Fig. 5.8 (a)
for different loading conditions. Figure 5.8(b) shows the spectra of structural acceleration
as the headloss coefﬁcient is changed. The mass ratio of the damper mass to the main mass
is 2%. The space is very limited on a typical offshore structure and therefore, the pontoon
tanks ﬁlled up with water can also be utilized as water supply for occupants. However, this
91
may not be always possible as water is used to ballast a platform and is restricted from
90 0.03
β1=10.0
70
0.02
60
ξ = 1
50 0.015
ξ = 50
40
0.01
ξ = 15
(optimal)
30
0.005
20
0
10 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
15 20 25 30 35 40 45 50
U (m/s) Frequnecy (rad/s)
10
(a) (b)
Figure 5.8 (a) Variation of Optimal Headloss Coefﬁcient with loading conditions
for different wave spectra (b) Spectra of structural acceleration at U10=20 m/s for
different ξ.
with negligible valve dynamics whose coefﬁcient of headloss can be changed rapidly by
applying a command voltage (Fig.5.9). This type of semiactive control is more suitable
for excitations which are transient in nature, for e.g., sudden wind gusts or earthquakes.
M s + m f αm f Ẋ˙ s + C s 0 Ẋ s + K s 0 Xs Fe ( t )
= + 0 u(t ) (5.11)
αm f mf ẋ˙f 0 0 ẋ f 0 kf xf 0 1
92
where the bold face denotes matrix notation and u(t) is the control force given by:
– ρ Aξ ( t ) ẋ f
u ( t ) =  ẋ f (5.12)
2
Semiactive
TLCD
Controllable Valve
Ks
F(t)
M s
Fe(t)
Cs
Primary Mass
ing the oriﬁce area of the valve. In the case of a passive system, this headloss coefﬁcient is
ξV
2
h l =  (5.13)
2g
where V is the velocity of the liquid in the tube. The coefﬁcients of headloss for different
valve openings are well documented for different types of valves (Lyons, 1982). The rela
tionship between the headloss coefﬁcient (ξ) and the valve conductance (CV) is derived in
Appendix A.3.
93
The damping force of a semiactive TLCD can be written as:
ρAξ ( Λ, t )
F d ( t ) =  ẋ f ( t ) ẋ f ( t ) (5.14)
2
needed to control valve opening, at a given time t. Equation 5.14 can be rewritten as,
F d ( t ) = C̃ ( Λ, t ) V V (5.15)
ρAξ ( Λ, t )
where C̃ ( Λ, t ) =  and V = ẋ f ( t ) . In this format, this damper system can be
2
compared to typical variable damping ﬂuid dampers. Semiactive ﬂuid viscous dampers
have been studied among others by Symans et al. 1997 and Patten et al. (1998). The
F d ( t ) = C ( Λ, t )V (5.16)
˜
where C ( Λ ) is the damping coefﬁcient which is a function of the command voltage Λ
and V is the velocity of the piston head. The damping coefﬁcient is bounded by a maxi
˜
mum and a minimum value and may take any value between these bounds.
Comparing Eqs. 5.15 and 5.16, one can some similarity in the fundamental work
ing of these dampers. However, there are basic differences in the two physical systems. In
variable oriﬁce dampers, the ﬂuid is viscous, usually some siliconebased material, which
is oriﬁced by a piston. In the TLCD case, the liquid is usually water and is under atmo
quadratic in nature, whereas the damping imparted by a ﬂuid damper is linear (Kareem
94
5.4.1 Control Strategies
introduced by the device in use. Therefore, a great deal of research is based on developing
examples are sliding mode control and nonlinear H ∞ strategies (Yoshida et al. 1998).
damper system are reported in Kurino and Kobori, 1998. Other researchers have used
fuzzy control theories to effectively implement semiactive control (e.g., Sun and Goto,
The strategy considered in this study is based on the linear optimal control theory.
The negative sign in Eq. 5.12 ensures that the control force is always acting in a direction
opposite to the liquid velocity. In case, the liquid velocity and the desired control force are
of the same sign, then Eq. 5.12, implies that ξ is negative. Since it is not practical to have
a negative coefﬁcient of headloss, the control strategy sets it to a minimum for ξ, i.e.,
ξ min . The control force is regulated by varying the coefﬁcient of headloss in accordance
In a fully active control system, one needs an actuator to supply the desired control
force. In such a case, the control force is not constrained to be in a direction opposite to the
damper velocity. Therefore, the linear control theory is readily applicable to active control
systems. In case of semiactive systems, however, the proposed control law is a clipped
95
optimal control law since it emulates a fully active system only when the desired control
force is dissipative (Karnopp et al. 1974; Dyke et al. 1996). Moreover, the actual control
force that can be introduced is dependent on the physical limitations of the valve used and
the maximum coefﬁcient of the headloss it can supply, which implies bounds on the con
– ρ Aξ min ẋ f
 – ρ Aξ max ẋ f
ẋ f ≤ u ( t ) ≤  ẋ f (5.18)
2 2
monly used onoff control. Most valve manufactures supply valves which operate in a bi
state: fully open or fully closed. These valves require a twostage solenoid valve. On the
other hand, the continuouslyvarying control requires a variable damper which utilizes a
servovalve. This servovalve is driven by a high response motor and contains a spool posi
tion feedback system, and therefore is more expensive and difﬁcult to control than a sole
ξmin can be taken as zero because this corresponds to the fully opened valve. It can be
expected that a small value of ξmax will result in a lower level of response reduction.
In order to formulate the system in a state space format, Eq. 5.11 is recast as,
Ẋ = AX + Bu + EW (5.21)
96
x 0 I 0 0
where X = ; A = ;B = –1
; and E = –1
and
–1 –1
ẋ –M K –M C M B1 M E1
E 1 and B 1 are the control effect and loading effect matrices, respectively. The states of
the system are the displacements and velocities of each lumped mass of the structure and
the displacement and velocity of the liquid in the TLCD. Measurements of the structural
Y = CX + Du + FW (5.22)
optimal control force is generated by solving the standard Linear Quadratic Regulator
(LQR) problem. The main idea in LQR problem is to formulate a feedback control law
T
which would minimize the cost function given as J = lim E ∫ ( Z QZ + u Ru ) dt ,
T T
T →∞
0
where Q and R are the control matrices for the LQR strategy. The control force is obtained
by,
u = –K g X (5.23)
–1 T
Kg = R B P (5.24)
and P is the Riccati matrix obtained by solving the matrix Riccati equation:
–1 T T
PA – PB ( R B P ) + A P+Q=0 (5.25)
97
The control performance of each strategy is evaluated based on a prescribed criterion. For
accelerations of the structure 〈 Ẋ˙s〉 , and the effective control force 〈 u〉 are deﬁned below:
where subscripts unco and co are used to distinguish between uncontrolled and controlled
cases.
W Z
Ẋ = AX + Bu + EW Y
u Y = CX + Du + FW
Plant Feedback
Feedforward
X
Kg
SemiActive
u=KgX Observer
Strategy
In actual practice, it is more realistic to consider a few noisy measurements which are then
used to estimate the system states. In this situation, the standard stochastic Linear Qua
dratic Gaussian (LQG) framework is used for estimation (Maciejowski, 1989). In a sto
98
Y = CX + Du + FW+ν̃ (5.27)
where ν̃ is the measurement (sensor) noise which is invariably present in all measure
ments. The LQG problem is solved using the seperation principle which states that ﬁrst an
T
optimal estimate X̂ of the states X (optimal in the sense that E { ( X – X̂ ) ( X – X̂ ) } is
minimized) is obtained, and then this is used as if it were an exact measurement to solve
the determinstic LQR problem discussed earlier. From the measurements, the states of the
˙ˆ
X = AX̂ + Bu + L ( Y – CX̂ – Du ) (5.28)
where L is determined using standard KalmanBucy ﬁlter estimator techniques. The opti
u = – K g X̂ (5.29)
where Kg is the optimal control gain matrix obtained by solving the standard LQR prob
high rise building subjected to alongwind aerodynamic loading. The building dimensions
are 31 m X 31 m in plan and 183 m tall. The structural system is lumped at ﬁve levels and
natural frequencies of this building are: 0.2, 0.583, 0.921, 1.182, and 1.348 Hz. The corre
sponding modal damping ratios are 1%, 1.57%, 2.14%, 2.52% and 2.9%. The description
of the wind loading and the structural system matrices for mass, stiffness and damping are
99
W 1
W 2
W 3
W 4
W 5
Figure 5.11 Schematic of 5DOF building with semiactive TLCD on top story
The TLCD is designed such that the ratio of the mass of liquid in TLCD to the ﬁrst
generalized mass of the building was 1%, the length ratio, α = 0.9 and ξ max =15. Using a
multivariate simulation approach (Li and Kareem, 1993), wind loads were simulated at
the ﬁve levels, as shown in Fig. 5.12. Two types of semiactive strategies, namely the con
tinuouslyvarying and the onoff type were examined. The LQR method, as described in
the earlier section, was used to determine the control gains. It was assumed that all states
The results are summarized in Fig. 5.13 and Table 5.3. As seen from Table 5.3, the
semiactive strategies provide an additional 1015% reduction over passive systems. Table
5.3 also shows how the two semiactive strategies deviate from the optimal control force.
100
One can observe the suboptimal performance of these schemes, which leads to a lower
response reduction than the active case. In a semiactive system, the applied control force
is generated using a controllable valve which can be operated using a small energy source
such as a battery.
80
1st Floor
nd
2 Floor
rd
3 Floor
60 th
4 Floor
th
5 Floor
40
Wind Load (kN)
20
−20
−40
−60
20 40 60 80 100 120 140 160 180 200
Time (sec)
101
Displacement of Structure
20
Displacement (cm) 10
−10
−20
0 20 40 60 80 100 120 140 160 180 200
) (cm/s
Time (sec)
Uncontrolled
Acceleration of Structure
30 Passive
2
Continuously−varying
20 On−off
Acceleration (cm/s2)
Active
10
−10
−20
−30
0 20 40 60 80 100 120 140 160 180 200
Time) (sec)
Figure 5.13 Displacements and Acceleration of Top Level using various control
strategies
again, but under harmonic loading. This example is taken from Soong (1991). The lumped
mass on each ﬂoor is 131338.6 tons and the damping ratio is assumed to be 3% in each
mode. The natural frequencies are computed to be 0.23, 0.35, 0.42, 0.49 and 0.56 Hz. A
102
where ω = 1.47 rad/s (= ﬁrst natural frequency of the structure), and the values of a, b, c
and d and the stiffness matrix of the structure are given in Appendix A.2. The excitation
acts at a frequency equal to the ﬁrst natural frequency of the structure. The semiactive
TLCD is placed on the top ﬂoor of the building with similar parameters as in Example 3.
Two cases of control strategies are considered: (a) full state feedback, and (b) acceleration
The ﬁrst strategy assumes that all states are available for feedback (total of 12
measurements). The control gains are calculated using Eq. 5.24. Figure 5.14 shows the
parametric variation of J1, J2 and Ju as a function of ξmax. There are small reductions in
the response after a certain value of ξmax is reached. This can be explained by Eq. 5.18 in
which it is implied that the applied control force is constrained by ξmax. This means that
satisfactory control results can be achieved by choosing a valve which may have a limited
Figure 5.15 shows the response of the top ﬂoor of the structure using various con
trol strategies. It is noteworthy that the continuously varying and onoff strategies give
similar reduction in response. This can be explained by the results in Fig. 5.16. The pro
ﬁles of variation in the headloss coefﬁcient as a function of time are similar for the two
strategies. The continuously varying control gives ﬂexibility in the headloss coefﬁcient.
However, the saturation bound introduces a clipping effect similar to onoff control and
therefore in this case, the advantage of continuouslyvarying control strategy is lost. Fig
ure 5.17 shows the RMS displacement of the ﬂoor displacements and accelerations, maxi
mum story shear and maximum interstory displacements using various control strategies.
103
85
80
J (%)
1
75
70
0 10 20 30 40 50 60 70 80 90 100
85
80
J (%)
75
Continuously varying
3
On−off
70
65
0 10 20 30 40 50 60 70 80 90 100
5
x 10
10
J (%)
5
u
0
0 10 20 30 40 50 60 70 80 90 100
ξm ax
0.25
Uncontrolled
0.2 Passive
uncontrolled
continuously variabl
0.15 On−Off
passive Active
displacement (m)
0.1
0.05
−0.05
−0.1
−0.15
−0.2
−0.25
0 20 40 60 80 100 120
time (sec)
104
ξ(t) continuously−varyin
20
15
10
−5
0 20 40 60 80 100 120
Time (sec)
20
15
ξ(t) On−off
10
−5
0 20 40 60 80 100 120
Time (sec)
4 4
Story Number
Story Number
3 3
2 2
1 1
0 0
0 0.05 0.1 0.15 0 0.1 0.2 0.3
RMS displacements (m) 2
RMS accelerations (m/s
)
5 5
Story Number
Story Number
4 4
Uncontrolled
3 3 Active
Continuously variable
On−off
2 2 Passive
1 1
0 2 4 6 0 0.05 0.1 0.15
Maximum Story Shear (N) 7
x 10 Maximum story displacements (m)
105
Observerbased LQG strategy
In the previous case, it was conveniently assumed that all the states were available
for feedback. However, in practice only a limited number of measurements are feasible. In
this case, we assumed that the ﬂoor accelerations and the liquid level (displacement of the
liquid) were measured. This implied that there were a total of six measurements (ﬁve
accelerations and one liquid displacement). The measurement noise was modeled as Gaus
sian rectangular pulse processes with a pulse width of 0.002 seconds and a spectral inten
LQG control is presented in Table 5.4. The response reduction is similar to the results
Two types of semiactive systems were presented in this chapter. The ﬁrst was
based on a gainscheduled feedforward type of control which utilized a lookup table for
control action. The second was a clippedoptimal feedback control system with continu
106
Numerical examples and applications were presented for the gainscheduled con
trol. This type of semiactive system leads to 1525% improvement over a passive system.
Next, the clippedoptimal control was discussed. The efﬁciency of the statefeed
back and observerbased control strategy was compared. Numerical examples showed that
semiactive strategies provide better response reduction than the passive system for both
random and harmonic excitations. In the case of harmonic loading, the improvement was
about 2530% while for the random excitation, the improvement was about 1015%. It
was also noted that continuouslyvarying semiactive control algorithm did not provide a
substantial improvement in response reduction over the relatively simple onoff control
algorithm.
107
CHAPTER 6
TLD EXPERIMENTS
It is a capital mistake to theorize before you have all the evidence...
It biases the judgment
Sherlock Holmes (Sir Arthur Conan Doyle)
Liquid Dampers (TLDs) were introduced in chapter 2. This chapter focusses on experi
mental studies conducted on TLDs. Shaking table experiments are conducted to obtain the
pressures due to sloshing are also measured along the height of the container wall. This
6.1 Introduction
plines due to the complexity of the problem and the difﬁculty in developing an analytical
model. Some of the relevant work done in the area of liquid dampers is brieﬂy reported
here. The earliest experimental studies on TLDs are reported by Modi and Welt, 1987 and
Fujino et al. 1988. A series of experimental studies, summarized in Modi et al. 1995, were
conducted using nutation dampers. These dampers covered different geometries like a tor
oidal ring, rectangular or circular crosssection cylinders, and in some situations may
include bafﬂes, screens, particle suspensions to manage liquid sloshing. Damper charac
108
teristics were determined by varying the amplitude and frequency of excitation. Fujino et
al. 1988 carried out parametric studies of cylindrical containers by freeoscillation experi
ments. Effects of liquid viscosity, roughness of container bottom, air gap between the liq
uid and tank roof, and container size on the overall TLD damping were studied.
Experimental studies have been carried out for rectangular TLDs in the region of
relatively small to medium vibration amplitudes, where breaking of a wave does not occur,
and the results have been found to be in good agreement with analytical results obtained
by the shallow water theory (Fujino et al. 1992; Sun and Fujino, 1994; Sun et al. 1995).
Similar experiments were done by Koh et al. (1994) who considered earthquake type exci
excitations, which are more representative of earthquakes, were also investigated through
similar shaking table tests and numerical modeling by Reed et al. (1998). Experimental
investigations of TLDs with submerged nets and other ﬂow dampening devices were stud
ied by Fediw et al. 1993 and Warnitchai and Pinkaew (1998). Chung and Gu, 1999 carried
active TLD systems have been conducted by Chang et al. 1997 and Natani (1998). A com
prehensive review of various analytical and experimental studies for sloshing dynamics is
As mentioned earlier, theoretical analyses are not able to predict sloshing pres
sures and forces in the neighborhood of resonance for large amplitude excitations. In
chapter 2, it was shown that the impact component is an important component of the over
all sloshing force. Therefore, experimental studies are conducted to better understand the
nature of the liquid impact on the container walls. Previous experimental studies have
109
been conducted, most notably in ship engineering (Bass et al. 1980) and marine engineer
ing applications (Schmidt et al. 1992; Hattori et al. 1994). However, speciﬁc studies of
impact pressures and their relation to the TLD performance have not been studied previ
ously. The present chapter presents experimental studies conducted on shallow water
TLDs, which shed more light into the nature of sloshingslamming caused at large ampli
tude excitations.
experiments were conducted on a rectangular TLD, shown in Fig. 6.1(a). The tank had the
following dimensions: length a = 25.4 cm, width w = 10.64 cm and a liquid height h = 3
cm.
PVC Tank
Baffles Pressure
sensors
6 DOF
Load Cell 7
6
command signal
5
to shaking table 4
Shaking Table
DAC 3
Board 2
1
Signal ADC
(a) Conditioner Board
(b)
Figure 6.1 (a) Schematic of the experimental setup (b) pressure sensor locations
From the linear wave theory, one can compute the natural frequency of the ﬁrst sloshing
mode as,
110
ω f =   tanh π 
1 gπ h
(6.1)
2π a a
Using Eq. 6.1, the ﬁrst sloshing mode frequency is computed to be 1.05 Hz. The total mass
of water is m = ρawh =0.8 kg. The linear damping is calculated from an expression
νf
ζf = 
3⁄2
 (6.2)
a g
where νf is the kinematic viscosity of water, a is the length of the tank in the direction of
the excitation, and g is the gravitational constant. Based on representative values for the
parameters in Eq. 6.2, ζf was found to be equal to 0.004 (0.4%). The water depth ratio is
0.12 which satisﬁes the shallow water assumption (h/a < 0.15). The excitation amplitudes
considered in this study range from 0.1 to 2 cm, which correspond to Ae/a ratio of 0.004 to
0.08. The excitation frequency ratio ( γ f = ω e ⁄ ω f ) in the sinesweep tests was in the
range 0.851.3.
A six degree of freedom load cell was utilized to measure the base shear due to liq
uid sloshing. A calibration matrix was used to determine the net shear force in the xdirec
tion. An accelerometer with a gain equal to the mass of the empty tank estimates the
contribution of the inertial component of the shear force due to the empty tank. This was
veriﬁed in the laboratory by testing the tank without water and comparing the value of the
base shear force and the accelerometer reading. The net sloshing force, Fb(t), due to the
liquid sloshing alone is obtained by subtracting the inertial contribution of the empty tank
from the total shear force. Finally, the shear force was expressed in a non dimensional
form as,
111
Fb
F b' = 
2
(6.3)
mω e A e
Pressure sensors were also mounted along the wall of the TLD to monitor the
impact pressures generated due to the liquid sloshing. The experimental setup is shown in
Fig. 6.1(b), wherein seven holes at 1.5 cm intervals are made on the side of the tank wall.
The pressures sensors used in this study were piezoelectric transducers with a range of 2
psi and a frequency response of 10,000 Hz. The sensitivity of these sensors is of the order
of 0.15 mV/psi. The sensors were specially fabricated with a silicon gel coating in order to
remove the possibility of any zeroshift problems associated with the change of media the
sensor comes in contact with. In the absence of such a layer, periodic artiﬁcial spikes due
to the unbalancing of the bridge resistance are observed which contaminate measure
ments. The sensor performs this way due to the response of the bridge elements to the
cooling effect of water. Although water is at room temperature, it cools the diaphragm due
to its higher thermal conductivity (Souter and Krachman). Alternating exposure to air and
water during sloshing causes this difﬁculty, which if not ameliorated can affect measure
ments signiﬁcantly.
Timehistories of the nondimensional base shear force are plotted in Fig. 6.2 for
Ae = 0.3 cm and 2.0 cm. As noted from the ﬁgure, the resonant condition occurs at differ
ent frequency ratios for the two cases, e.g., γ f = 1.10 at Ae =0.3 cm and γ f = 1.20 at Ae
=2.0 cm. Sinesweep studies were conducted in order to construct the frequency response
curves.
112
0.04 0.04
γ = 1.10,=0.3
A cm γf = 1.10,e=2.0
A cm
A ) f e
0.02 0.02
e e
Fb/(mw2
0 0
−0.02 −0.02
−0.04 −0.04
0 2 4 6 8 10 0 2 4 6 8 10
time (sec) time (sec)
0.02 0.02 γ = 1.15,A
γf = 1.15,A
e
=0.3cm f
=2.0cm
e
A )
0.01 0.01
e e
Fb/(mw2
0 0
−0.01 −0.01
−0.02 −0.02
0 2 4 6 8 10 0 2 4 6 8 10
time (sec) time (sec)
0.02 0.02
γ = 1.20,=0.3cm
A γf = 1.20,e=2.0cm
A
f e
A )
0.01 0.01
e e
Fb/(mw2
0 0
−0.01 −0.01
−0.02 −0.02
0 2 4 6 8 10 0 2 4 6 8 10
time (sec) time (sec)
Figure 6.2 Sample timehistories of the shear force at Ae = 0.3 cm and 2.0 cm
A nonlinear identiﬁcation scheme was utilized to determine the parameters for the
nonlinear impact characteristics of the TLD. The algorithm used was a nonlinear least
1992). The objective function evaluates the square of the error between the experimental
113
data and the simulated data using the assumed values of the unknown parameters. The
ﬂowchart of the optimization scheme is shown in Fig. 6.3. Figure 6.4 shows the variation
function of the non dimensional amplitude of excitation. After optimization, the following
2.3
η ≈ 2 ; ϕ ( A e ) ≈  ; ς ( A e ) ≈ 1.78 ( A e ⁄ a )
0.68
 (6.4)
( Ae ⁄ a )
0.78
assume initial
guess
Yo
114
200
Data
Non−linear Fit
150
φ(Ae)
Parameter
100
50
0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
non−dim. amplitude
e
/a A
0.35
Data
0.3
ς(A )
Non−linear Fit
effective damping
e
0.25
0.2
0.15
0.1
0.05
0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
non−dim. amplitude
/a A
e
Equation 6.4 implies that the damping due to inherent liquid (ζf = 0.4% calculated
using Eq. 6.2) is negligible compared to the total damping, ( ς ( A e ) ) , induced in the TLD
due to sloshing at higher amplitudes. The results of the identiﬁcation can be seen in Fig.
6.5 where the experimental non dimensional shear force and the analytical shear force
plotted as a function of the frequency ratio are compared. The analytical model success
fully captures the jump phenomenon and the widening of the frequency band very well.
quency ratio of 0.96 which is not reﬂected by the nonlinear model. However, this reso
nance though present at low amplitudes is more pronounced at some medium amplitudes
and is suppressed at high amplitudes. The current analytical model does not contain these
115
secondorder effects. More complex models which include higher order nonlinearities can
model this effect. However, this is not pursued in this study. Figure 6.5 suggests that even
15 15
Experimental Analytical
Ae=0.1 cm
Ae=0.25 cm
Ae=0.5 cm
Ae=1.0 cm
Ae=2.0 cm
Non−dimensional Sloshing Force
10 10
5 5
0 0
0.9 1 1.1 1.2 1.3 0.9 1 1.1 1.2 1.3
ω /ω
Frequency ratio ω /ω
Frequency ratio
e f e f
the symbols have been deﬁned earlier. The mass ratio, µ = m f ⁄ M s is equal to 0.01 and
116
the tuning ratio γ = ω f ⁄ ω s is equal to 0.99. Solving the equations of motion given in
Eqs. 6.5 and 6.6 numerically and plotting the nondimensional displacement of the struc
ture (Xs/Ae) as a function of the frequency, the transfer functions as shown in Fig. 6.6 are
obtained.
40
A e=0.01 cm
A e=1 cm
Non−dimensional displacement of Structure
35
30
25
20
15
10
0
0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2 1.25
ωe/ωs
unlike a tuned mass damper, is excitation amplitude dependent. The increased damping
(introduced by wave breaking and slamming) causes the frequency response function to
change in frequency response has also been observed experimentally, e.g., Sun and Fujino,
1994.
117
6.4 Impact Pressure Studies
The shallow water theory leads to a hydrostatic pressure description for loads on
the sloshing container walls. This is appropriate when standing waves or small travelling
waves are excited. However, as soon as impacts are recorded at the walls, the pressure dis
tribution appears very different due to the presence of the impulsive peaks. At this time,
the pressure distribution at the vertical walls is far from hydrostatic. In this section, the
local pressures on the walls of the TLDs arising due to the sloshing impacts of the liquid
Seven measurement taps were drilled in the side of the tank for pressure sensors at
intervals of 1.5 cm (Fig. 6.1(b)). Sensor 1 is at 1.5 cm from the bottom of the tank, sensor
2 is at 3.0 cm (static liquid level) and so on. The sampling frequency of the data acquisi
tion system was maintained at 1000 Hz. This was found to be adequate since the duration
of the peak impact in the resonant pressure trace was found to be of the order of 1520
milliseconds. Data acquisition for each case was carried out for about 30 sec which corre
sponded roughly to 30 cycles of data. The average value of the peak pressure over N cycles
is calculated as follows:
∑ Pi, peak
[ P peak ] = 
i=1
(6.7)
N
The pressure peak coefﬁcient at a certain height z on the vertical wall is deﬁned as:
[ P peak ]
K Pz =  (6.8)
ρga
118
6.4.1 Singlepoint pressure measurement
Figure 6.7 shows typical pressure traces at different frequency ratios including res
onant and nonresonant cases, i.e., γf = 0.7, 1.1 and 2.0. As seen from the plots, the impact
peak pulses are present only at the resonant condition. As we know from base shear
results, this resonant condition does not occur at γf = 1.0, but at 1.1 due to the hardening
0.015 0.18
γf = 1.1
0.16
0.01
0.14
0.005
Pressure (psi)
0.12
0
0.1
0.08
Pressure (psi)
0.02
γf = 1.4 0.06
0.015
0.04
0.01
0.02
0.005
0
0
Figure 6.7 Pressure time histories for various frequency ratios (Ae = 1.0 cm).
It has been observed that these typical pressure time histories are neither harmonic
nor periodic since the magnitude and duration of the peaks vary from cycle to cycle. This
is true even though the excitation experienced by the tank is harmonic. Figure 6.8 shows
the histogram of peak impact pressure for 100 cycles of pressure pulses for sensor at loca
tion 2. A statistical analysis of the pressure time records was conducted and the data was
119
Normal Probability Plot
35
lognormal distribution 0.997
0.99
30 0.98
0.95
Probability distribution
25 0.90
0.75
Probability
20
0.50
15
0.25
10 0.10
0.05
5 0.02
0.01
0.003
0
0 0.2 0.4 0.6 0.8 −2.5 −2 −1.5 −1
pressure (psi) Log of Data
Figure 6.9(a) shows the anatomy of a single pressure proﬁle as it evolves over time
along with corresponding visual photographs of wave sloshing. It is noteworthy that the
impulsive peak is observed at 15 msec which suggests the high frequency slamming
nature of the pressure pulse. After the initial impact caused by the wave, the full sloshing
action of the wave is developed, which can be seen as a second peak of lower magnitude
and longer duration. A wavelet scalogram (using Morlet wavelet) was utilized to study the
timefrequency ﬂuctuations of the pressure timehistory. For more details on this tech
nique, one can refer to Gurley and Kareem (1999). A scalogram is a plot wherein the
square of the coefﬁcients obtained by continuos wavelet transform (CWT) are plotted as a
measure of the signal energy in the timefrequency domain. The scalogram of the pressure
signal reveals the presence of high frequency components at the time of occurrence of the
impulsive peak (Fig. 6.9b). The energy in regular sloshing is concentrated at lower fre
quencies which occurs after a certain timelag following the initial impact.
120
0.14
Point A
= 15 mSec
0.12
0.1
Pressure (psi)
0.08
0.06
Point B Point C
0.04 = 200 mSec
= 37 mSec
0.02
Point D
= 285 mSec
0
Time (mSec)
(a)
(b)
Figure 6.9 (a) Anatomy of a single pressure pulse (b) wavelet scalogram of the
pressure signal
121
6.4.2 Multiplepoint pressure measurements
Next, four sensors were recorded simultaneously to observe the timelag as the
pulse travels along the tank height and the spatial distribution of the impulsive peak to the
overall slosh pulse. Figure 6.10 shows the simultaneous pressure pulse traces for a single
cycle. The timelag is measured with respect to sensor 2 (which is at the mean water
A
IIF = i (6.9)
At
where Ai is the area under the impulsive peak in the pressure timehistory and At is the
total area under the sloshing/slamming trace (including the impulse component). It is
observed that at levels above the water level, the contribution is entirely due to impulsive
slamming. On the other hand, the contribution of slamming at sensor 1, which is below the
water level is only about 10%. This corroborates with topology of wave slamming because
the slamming action is more prevalent in the region above the mean water level. The roll
ing convective mass of water, which is responsible for the slamming action, is primarily
effective at these locations. The timelag and IIF for the four locations are documented in
Table 6.1.
Like the scalogram, it is useful in revealing time varying pockets of high and low correla
tion in different frequency bands. It is obtained by plotting the product of the wavelet coef
ﬁcients of two signals as a function of time and frequency. The coscalograms in Fig.
6.10(b) are plotted with reference to sensor 2. The light patches in the coscalograms help
identify areas of correlation. It is noted that the maximum correlation between each sen
122
sors is near the low frequency sloshing component of the pressure signals. The correla
tions in the high frequency slamming portion is maximum in the sensor 21 coscalogram
and drops off progressively in the 23 and 24 coscalograms due to the time lag of peaks
over a number of measurements, yields that the contribution of the impulsive peak is
around 2030% of the total contribution of the pulse. This is a substantial contribution
which is neglected by most numerical simulations. Moreover, the peak pressures obtained
due to slamming are 510 times higher than those obtained from regular sloshing as
emphasizes the importance of estimating the effect of the liquid slamming on the overall
system response. Similar concerns have been expressed in the study of impact loading of
vertical structures in the offshore community, where the impact pressures were assumed to
be not important and hence were not considered in the design. However, Schmidt et al.
TABLE 6.1 Time lag and impact inﬂuence factor for different sensor locations
2nd sensor 0 21
3rd sensor 14 30
4th sensor 42 85
123
0.2
0.18
0.16 4
3
2
Pressure (psi)
0.14 1
0.06
0.04
0.02
0
2 2.05 2.1 2.15 2.2 2.25 2.3 2.35 2.4 2.45 2.5
time (sec)
(a)
Figure 6.10 (a) Pressure pulses at different locations on the wall (b) Wavelet
coscalograms with sensor 2 as reference
124
6.4.3 Shallow water versus deep water sloshing
Until now, the results presented were for the shallow water case (h/a < 0.15).
In this case, sloshing at high amplitudes is characterized by travelling waves and hydraulic
jumps (Fig. 6.11a). For deep water cases, i.e, h/a > 0.15, large standing waves are usually
formed at resonance. Figure 6.11(b) shows the difference between the shallow water (h/a
=0.12) slosh pressure trace and deep water (h/a = 0.25) pressure traces for the pressure tap
locations at the mean water level. In the case of shallow water TLD, the pressure is maxi
mum at the mean liquid level, while for the case of deep water TLD, impact pressures are
also observed in a large part of the ceiling. The impulsive peak is more pronounced in the
shallow water case and reaches peak value at 15 msec as opposed to the deep water case
Point A Point A
= 15 mSec = 50 mSec
0.12 0.06
Point C
= 220 mSec
0.1 0.05
pressure (psi)
Pressure (psi)
0.08 0.04
Point B
0.06 0.03 = 90 mSec
Point B Point C
0.04 = 200 mSec 0.02
= 37 mSec
0.02 0.01
Point D Point D
= 285 mSec
= 300 mSec
0 0
Time (mSec) Time (mSec)
(a) (b)
Figure 6.11 Typical sloshing wave with pressure pulse and wave mechanism
schematic for (a) shallow water (h/a =0.12) and (b) deep water (h/a = 0.25) case
125
6.4.4 Pressure variation along the tank height
The pressure distribution over the tank walls is important for establishing integral
load effects due to slamming and design considerations of walls under sloshing/slamming
induced loads. Bass et al. 1980 have provided an idealized distribution for vertical tank
walls based on their experiments in terms of a pressure coefﬁcient which was described by
5π ( z – h )
K Pz =  K Pmax 1 + cos  ;
1 H H
h –  ≤ z ≤ h +  (6.10)
2 H 5 5
where KPz is the peak pressure coefﬁcient, KPmax is the maximum pressure coefﬁcient
(which occurs at the mean water level for the shallow water case), z = distance from tank
bottom, h = liquid ﬁlling height, H = tank height. As seen from Fig. 6.12, where the maxi
mum pressure coefﬁcients at Ae = 2.0 cm are plotted along the height of the wall. One can
note that the curve described by Eq. 6.10 envelopes the maximum pressure peak coefﬁ
1
K π(z−h)/H)]/2
= [K (1+cos(5
Pz Pm ax
z/h
0.5
−0.5
−1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
K Pz/K Pmax
Figure 6.12 Variation of the peak pressure coefﬁcient with height of the tank wall
126
6.5 Hardwareintheloop Simulation
One of the main areas of investigation in the design of TLDs is the actual perfor
technique in which some of the system components are numerically simulated while oth
ers are physically modeled with appropriate interface conditions. Usually, there are real
hardware characteristics that are unknown or too complex to model in pure simulations. In
tions (Isermann, 1999). HIL is routinely used in aerospace, automotive control and
for product development. Its application in structural testing of damping systems has been
rather limited.
One can specify the external loading and model the structure by appropriate equations,
which are solved in realtime to obtain the displacement response. This displacement is
used to drive the shakingtable on which the damper is mounted. The baseshear force due
to sloshing liquid in the damper is simultaneously measured and feedback into the com
puter where it is used in the fore mentioned numerical equations. Thus, a realtime
Some of the advantages of HIL simulation over conventional testing methods are
the cost and time savings in repeated simulations. Figure 6.13 shows the difference in
scale and the associated costs one can achieve with HIL testing for combined structure
damper experiments. The dynamic testing of structural systems with nonlinear append
127
ages require considerable infrastructure involving structural system model, actuators,
reaction wall system, and instrumentation. Often the actuators are limited in their dynamic
capability which restricts these tests to a pseudodynamic level. While, in HIL simulation,
one can build a virtual structure in a computer model and the nonlinear structural ele
ments, such as dampers, hysteretic elements and baseisolators, can be included in the
physical structural model. Moreover, one needs a smaller shaking table for component
testing. One of the most useful aspects of the HIL testing is that the user can perform on
theﬂy tuning of important structural and excitation parameters. This can help in identify
ing important parametric relations between the two systems. A computer controlled sys
tem, which is standard in most dynamic testing laboratories and an essential component
for implementing controllers for the shaking table, is needed to conduct the test in real
time. Some of the main issues for the success of this test is the speed of the computer con
trol system. The disadvantage of the test is that a good system model is needed for the
sensors
On the fly
tuning
of parameters for the
virtual
structure
Ms
ζs
ωs
Fe(t)
Structure
xs(t)
Fe(t)
128
6.5.1 Experimental study
Figure 6.14 shows a schematic of the experimental setup for veriﬁcation of the
discussed earlier, the net sloshing force, Fb(t), due to the liquid sloshing alone was
obtained by subtracting the inertial contribution of the empty tank from the total shear
force. For the combined structuredamper system, the equation of motion of the structure
M s Ẋ˙s + C s X˙ s + K s X s = F e ( t ) + F b ( t ) (6.11)
The displacement of the structure was calculated using the ﬁnitedifference version of Eq.
6.11 and the displacement signal was sent back as a voltage to the shaking table. In this
way a realtime experiment of the combined dynamics of the structure and the damper was
conducted.
PVC Tank
Baffles
6 DOF
Load Cell Combined equations
of motion solved
in the computer
Ftotal
Shaking Table command signal
to shaking table
Signal
conditioner xs
DAC
Board
Sloshing Force Fb
+ ADC
Board
Signal Fs
conditioner m tank 
Figure 6.14 Schematic of the experimental setup for the HIL simulation
129
An important aspect of the HIL simulation is the realtime integration algorithm.
For realtime simulation one should use ﬁxedtime steps and should require inputs for
derivative calculations that occur at the current time step and earlier. This means that
order algorithm has poor characteristics. The AdamsBasforth second order algorithm
seems to provide much better accuracy yet it is suitable for realtime use. The displace
ment of the structure for the next time step tj+1 is calculated from displacements and
= X s, j + ∆t  X˙ s, j –  X˙ s,
3 1
X s, (6.12)
j+1 2 2 j – 1
0.8
0.6
(t) (cm)
0.4
External Excitation, F
0.2
e
−0.2
−0.4
−0.6
−0.8
−1
0 10 20 30 40 50 60
Time (sec)
2.5
2 Uncontrolled
Controlled with TLD
1.5
(cm) x
Displacement of Structure,
1
s
0.5
−0.5
−1
−1.5
−2
−2.5
0 10 20 30 40 50 60
Time (sec)
130
In the current experiment, a ﬁxed time step of 0.005 (sampling frequency of 200
Hz) was utilized. This was suitable for our application as the frequency range of interest
was less than 5 Hz. The parameters used in the simulation are ω s =1.1 Hz, ζ s = 3% and µ
= 10%. Figure 6.15 shows the excitation time history used which is a random white noise
signal. The ﬁgure also shows a comparison of the uncontrolled response and the controlled
response by including the sloshing force due to the TLD. The total reduction in RMS
The model parameters can be obtained from experimental data obtained by an instru
mented sloshing tank placed on the shaking table. Impact pressure distributions were also
measured along the height of the container. It was noted that the slamming action is
present in shallow water TLDs and has a signiﬁcant contribution to the overall sloshing
force. These impact pressure studies also indicate the nature of sloshingslamming along
the height of the container, for e.g., at levels below the static liquid level, the pressure is
dominated by the sloshing component while at levels above the static liquid level, it is
governed by the slamming action. Finally, a new technique, namely the hardwareinthe
loop testing technique was presented for testing structureliquid damper systems. This
actual structure, its scale model or large highspeed dynamic actuators to induce dynamic
loading.
131
CHAPTER 7
TLCD EXPERIMENTS
In theory, there is no difference between theory and practice.
But, in practice, there is.
 Jan L.A. Van de Snepscheut
along with a prototype semiactive TLCD are presented. First, the dynamic characteristics
of the combined structuredamper system were compared with previously obtained analyt
ical results reported in Chapter 3. Next, a gainscheduled control law based on a pre
scribed lookup table was experimentally veriﬁed for achieving the optimum damping
7.1 Introduction
Experimental studies using tuned liquid column dampers (TLCDs) for evaluating
their control performance have been limited to passive systems. Sakai et al. (1991) veriﬁed
bridge tower. Balendra et al. (1995) conducted shaking table tests using TLCDs and stud
ied the effect of different oriﬁce opening ratios on the liquid motion. Experimental studies
have also been reported by Hitchcock et al. (1997) using passive TLCDs with no oriﬁce,
termed as liquid column vibration absorbers (LCVAs). Recently Xue et al. (2000) pre
sented experimental studies on the application of a passive TLCD in suppressing the pitch
ing motion of structures and conducted experiments to delineate the inﬂuence of different
(LCVA) on a 67m steel frame communications tower has been reported by Hitchcock et
al. (1999). This device does not have an oriﬁce/valve in the Utube and hence, it is not
possible to control the damping in the LCVA. The authors also acknowledge that due to
the lack of oriﬁce, the damping ratio of the LCVA was not expected to be optimum. The
authors observed that the LCVA did not perform optimally at all wind speeds. Response
reduction of almost 50% was noted, however, nonoptimal performance of the damper was
noted above and below the design wind speed. This observation reafﬁrms the fact that
passive liquid damper systems are inadequate in performing optimally at all levels of exci
Although researchers have studied the semiactive version of TLCD theoretically (Haroun
et al. 1994; Kareem, 1994; Abe et al. 1996; Yalla et al. 1998), there has been no reported
conducted using scale models of structures along with a prototype semiactive TLCD. The
law for achieving the optimum damping based on a prescribed lookup table was veriﬁed
experimentally.
structure model attached to a TLCD. The TLCD consists of a Ushaped tube made of PVC
133
material with an electropneumatic actuator driving a ball valve attached at the center of
the tube.
(a)
80 psi Pneumatic Airline
position transmitter
shown
in detail
SigLab
Spectrum
analyzer
420 mA
signal
signal
conditioner
ADC
channels
Accelerometer encoder ouput Computer
420 mA
to positioner
Building signal
Model conditioner DAC
channels
command signal
to shaking table
Shaking Table (b)
The Utube has a circular crosssection with an inner diameter of 3.8 cm and a
horizontal length of 35.5 cm and a total length of 81 cm. The valve used in this study is a
134
ball valve of 3.8 cm (1.5 inches) diameter. A command voltage changes the valve opening
angle (θ), which effectively changes the oriﬁce area of the valve. The details of the valve
characteristics are presented in Appendix A.3, where the valve opening angle is related to
Transfer function measurements were obtained by exciting the shaking table with a
bandlimited random white noise (cutoff frequency fc = 2 Hz), at different levels of excita
tion amplitudes and the acceleration was measured at the top of the structure. The excita
value of excitation (in volts). The range of feasible RMS excitation displacement ampli
tudes of the shaking table without spilling water out of the Utube was varied between
0.050.3 volts.
The model structure without the damper is a linear system, which was conﬁrmed
effect of the pneumatic actuator used to drive the valve in the TLCD on the dynamics of
the structure was found to be negligible. This was done by comparing the transfer func
tions with and without the airsupply to the pneumatic actuator. All transfer function mea
surements were obtained using SigLabTM spectrum analyzer using the average of 15
measurements. From the transfer function and free vibration decay curves, the natural fre
quency and damping ratio of the uncontrolled building was determined to be 0.92 Hz and
0.6%, respectively. The mass ratio (ratio of the liquid mass in the damper to the ﬁrst modal
mass of the structure) is kept approximately 10% of the total mass of the structure.
135
7.2.1 Effect of tuning ratio
The tuning ratio (γ ) is deﬁned as the ratio of the natural frequency of the damper
tuning ratio, liquid columns of different lengths were considered. Figure 7.2 (a) shows the
transfer function with different tuning ratios. The H 2 norm was used as a measure of
ωb
≈ ∫ H Ẋ˙ ẋ˙ ( ω ) dω
2
H 2 (7.1)
s g
ωa
where Ẋ˙s is the acceleration of the structure, ẋ˙g is the ground acceleration of the shaking
table, ω a =0.5 Hz and ω b =1.5 Hz. The range of frequencies were limited to 0.51.5 Hz
because below 0.5 Hz there was a lot of noise in the system and above 1.5 Hz, there is neg
Figure 7.2 (b) shows the variation of the H2 norm as a function of the tuning ratio.
A fourth order polynomial ﬁt was used to determine the optimum tuning ratio as equal to
0.953. This corresponds to a liquid length of 25 inches (63.5 cm). One can observe that the
two peaks in the transfer function are almost equal in height at the optimum tuning. This is
136
2 0.22
γ=1.02
γ=1.00
1.8 γ=0.98
0.215
γ=0.96
γ=0.94
1.6
0.21
Transfer function (Mag)
1.4
0.205
1.2
H norm
1 0.2
2
0.8
0.195
0.6
0.19 curve−fitted
0.4 using 4th order
polynomial
0.185
0.2
0 0.18
0.5 1 1.5 0.92 0.94 0.96 0.98 1 1.02 1.04
frequency (Hz) γ
Tuning ratio,
(a) (b)
Figure 7.2 (a) Transfer functions for different tuning ratios (b) Variation of H2
norm with tuning ratio
The effective damping in the TLCD is obtained through changing the oriﬁce open
ing of the valve. As noted in previous chapters, the effective damping of the TLCD is an
important parameter for optimum absorber performance. The damping is varied by chang
ing the valve angle, where θ = 0 corresponds to fullyopen valve and θ = 90 degrees corre
sponds to fullyclosed valve. In the fullyclosed position, no liquid oscillations take place
and the system becomes a SDOF system. An upper limit of θ = 60 degrees is used in this
study. At this position, the valve is almost fully closed. Figure 7.3 shows how the transfer
137
2.5
0 deg
10 deg
20 deg
35 deg
40 deg
45 deg
1.5
0.5
0
0.5 1 1.5
frequency (Hz)
It is well known that the damping introduced by valves and oriﬁces is quadratic in
nature. This has been studied experimentally for passive TLCDs (Sakai and Takaeda,
1989; Balendra et al. 1995). The damping force is dependent on the liquid velocity,
F d = c ẋ f ẋ f (7.2)
This implies that the damping introduced by the valve is nonlinear and changes as a func
tion of the amplitude of excitation. Figure 7.4 shows the transfer functions of the com
bined system at two different excitation levels, i.e., S0 = 0.1 and 0.3 V with different valve
opening angles. The transfer functions at θ = 0 degrees (fullyopen) are virtually identical
138
as no nonlinearity is introduced due to the valve. At other valve opening, however, the
2 2 2
Transfer function (Mag dB)
1 1 1
0 0 0
0.5 1 1.5 0.5 1 1.5 0.5 1 1.5
frequency (Hz) frequency (Hz) frequency (Hz)
From Fig. 7.4, one can note the change in effective damping as the excitation
amplitude is varied. Therefore, for the damper to perform optimally at all levels, one needs
to determine the optimum damping required at each amplitude of excitation and organized
in the form of a lookup table. The main idea of a lookup table is to determine the angle of
opening which minimizes the H 2 norm of the structural response. This corresponds to
the optimal valve opening for a particular amplitude of excitation, as shown in Fig. 7.5(a)
for S0 = 0.1 V and S0 = 0.3 V. This procedure is repeated for a wide range of amplitudes of
excitation. Using these optimal values, one can construct a nonlinear lookup table as
139
0.4 0.38
70
0.3 V
θ 1 = 30 degrees at So=0.3 V
0.1 V
H θ 2 = 40 degrees at So=0.1 V
50
0.38 0.37
2 norm
40
30
0.36 0.36
20
10
0
0.34 0.35 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
0 10 20 30 40 50 60
θ (degrees)
0
Φ Valve opening,
(a) Excitation Amplitude, So (Volts
(b)
Figure 7.5 (a) Optimization of H2 norm (b) Lookup table for semiactive control
The equivalent damping of the TLCD is a function of the excitation amplitude and
the valve opening. An MATLABTM program was used to curveﬁt the experimental trans
fer function by minimizing the norm of the error function. The equivalent damping was
found to range from 2% (for fully open, θ = 0 deg) to 30% (for almost closed, θ = 60 deg).
7.6(a)). Figure 7.6 (b) shows the transfer function with nonoptimal damping (about 30%)
Closedform equations for the case of white noise excitation applied to the primary
known that the optimum absorber parameters that minimize the RMS accelerations of the
primary system for a white noise base excitation are the same as those that minimize the
140
RMS displacements for a white noise excitation applied to the primary system. Therefore,
in this study the equations derived in Chapter 3 are used. In the case of an undamped pri
mary system, one can write the expressions for optimal damping and tuning ratio as,
2µ
µ 1 + µ – α  α
2
α 4 1 + µ 1 – 
ζ opt 2
=  
 ; γ opt =  (7.3)
2 α
2
µ 1+µ
( 1 + µ ) 1 + µ – 
2
In the case of µ = 0.1 and α = 0.56 and ζ s ≈ 0 , optimum values of the absorber
parameters obtained from Eq. 7.3 are: ζ opt = 8.9% and γ opt = 0.95, which are close to the
3 3
Experimental Data Experimental Data
Simulated Simulated
So = 0.1 V So = 0.1 V
2.5 2.5
(Mag)
Transfer Function (Mag)
2 uncontrolled 2
Function
uncontrolled
1.5 1.5
Transfer
controlled
controlled
1 1
0.5 0.5
0 0
0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
frequency (Hz) frequency (Hz)
(a) (b)
Figure 7.6 (a) Comparison of transfer functions: (a) θ =40 deg, ζ f = 9 % (optimal
damping) (b) θ = 60 deg, ζ f = 30% (nonoptimal damping)
141
Figure 7.7 (a) shows the 3D plot of the magnitude of the experimental transfer
function as a function of the valve opening angle (and effective damping) and the fre
quency at S0 = 0.1 Volts. One can observe that the double peaked transfer function changes
to a single peak curve as the valve opening angle is increased. Figure 7.7 (b) shows the
simulated 3D transfer function as a function of frequency and equivalent damping ratio.
A similar curve is obtained by solving the actual nonlinear equations of the TLCD and
plotting the dynamic magniﬁcation ratio as a function of frequency and the headloss coef
ﬁcient (for e.g., see Haroun and Pires, 1994). The effect of coalescing of the modal fre
quencies, from a double peaked curve to a single peaked curve, was also described in
chapter 4 while examining the beat phenomenon of the combined structureTLCD system.
2.5
2.5
(mag)
2
2
Transfer Function (mag)
Function
1.5 1.5
Transfer
1
1
0.5
0.5
0
0
ζf θ0
0 0 0.1
20
0.1 0.2
ζf 1.5
40 1.3 1.4
0.2 1 1.1 1.2
1.5 0.9
0.3 0.7 0.8
1 0.5 0.6
0.3 60 ζd frequency,
Φ 0.5 frequency,HzHz
frequency, Hz
(a) (b)
Figure 7.7 3D plot of transfer function as a function of effective damping and
frequency (a) experimental results (b) simulation results.
142
The experimental results show that the effective damping is a function of the amplitude of
ζ f = f ( S 0, θ ) (7.4)
In section 3.2.1, the expression for the equivalent damping was obtained as:
ξσ ẋ f
ζ f = 
 ≡ f ( σ ẋ f , ξ ) (7.5)
2 πgl
From the Appendix A.3, one can note that the headloss coefﬁcient is a function of the
ξ = f (θ) (7.6)
while the standard deviation of the liquid velocity is related to the amplitude of excitation
by Eq. 3.9,
σ ẋ f = f ( S 0 ) (7.7)
ζ f = f ( S 0, θ ) ≡ f ( ξ, σ ẋ f ) (7.8)
Chapter 5. The main idea was to benchmark the performance of the semiactive system to
a purely passive system. In the case of a passive system, the headloss coefﬁcient was kept
constant. For the semiactive case, the valve opening was changed according to the look
143
Two different loading timehistories were chosen. The ﬁrst time history, referred to
as case 1, comprised of segments of 20 sec each in length of 0.1 and 0.3 V RMS excita
tions, while the second time history (case 2) comprises of segments 40 sec each in length
of 0.1 and 0.3 V RMS excitations. The underlying objective was to show that the semi
active TLCD, which changes the headloss coefﬁcient in response to changes in external
0.5
S (Volts)
0
0
−0.5
−1
0 5 10 15 20 25 30 35 40
time (sec)
50
Angle of Valve,
40
θ
30
20
10
0
0 5 10 15 20 25 30 35 40
time (sec)
Uncontrolled
Passive System
Semi−active System
1
0.5
Acceleration 2(m/s
)
−0.5
−1
0 5 10 15 20 25 30 35 40
Time (sec)
Figure 7.8 Excitation time history, valve angle variations and the resulting
accelerations for uncontrolled, passive and semiactive systems for case 1.
144
1
0.5
S (Volts) 0
0
−0.5
−1
0 10 20 30 40 50 60 70 80
time (sec)
50
Angle of Valve,
40
θ
30
20
10
0
0 10 20 30 40 50 60 70 80
time (sec)
2.5
Uncontrolled
2 Passive System
Semi−active System
1.5
1
Acceleration 2(m/s
)
0.5
−0.5
−1
−1.5
−2
−2.5
0 10 20 30 40 50 60 70 80
Time (sec)
Figure 7.9 Excitation time history, valve angle variations and the resulting
accelerations for uncontrolled, passive and semiactive systems for case 2.
From Figs. 7.8, 7.9 and Table 7.1, one can note that for 0.3 V, there is hardly any
response reduction for the case 1, while there is a 76% reduction for case 2. This is
because case 2 record is of a longer duration and hence the steadystate of the response is
established. This increases the liquid damper effectiveness as liquid oscillations are fully
developed. One can also see that at higher levels of excitation, the optimum damping is
close to the passive system damping, therefore the improvement of semiactive system is
145
not substantial (about 13% improvement over passive system). On the other hand, for
lower levels of excitation, the improvement is more drastic (about 27% improvement over
passive system). The overall RMS response reduction of semiactive system over passive
It is noteworthy that the response reduction of 76% is high. This is because the
mass ratio of the damper considered in the scaleddown experiment was 10%. This is a
very high mass ratio since in most typical buildings, a mass ratio of approximately 1% can
be accommodated due to the weight and space requirements. However, in this study, a
Numerical studies indicate, however, that a 1% mass ratio would provide about 45%
5.3.1), where an improvement of 20% was noted for a semiactive system over a passive
146
7.4 Concluding Remarks
An experimental investigation to determine the optimal absorber parameters of the
combined structureTLCD system was presented. The experimental results were com
pared to the previously obtained analytical results. A control strategy based on a gain
scheduled lookup table was veriﬁed experimentally. It was observed that at low ampli
tudes of excitation, the TLCD damping was enhanced by constricting the oriﬁce and at
higher amplitudes by dilating the oriﬁce to supply the optimal damping. Experimental
studies have shown that the semiactive TLCD can boost the performance of the passive
TLCD by an additional 1525% and maintains the optimal damping at all levels of excita
tion. This justiﬁes the additional costs of using sensors and controllable valves in the semi
active system. A more detailed cost and implementation comparison is discussed in Chap
ter 8.
147
CHAPTER 8
DESIGN, IMPLEMENTATION AND RELIABILITY ISSUES
I strive for structural simplicity.... the technical man must not be lost in his own
technology.
 Dr. Fazlur Khan
details, cost analysis and reliability issues of liquid dampers are discussed. First, compari
sons are made among different types of dynamic vibration absorbers (DVAs) in terms of
their implementation and cost. Next, a riskbased decision analysis framework is pre
sented to measure the risk of unserviceability in tall buildings and to provide a basis for
choosing the optimal decision. Finally, some design guidelines for technology transfer are
laid out in accordance with the research conducted and documented in earlier chapters.
8.1 Introduction
tion on scale models have been discussed. However, the actual implementation of these
constraints. Furthermore, various players including the building owners, designers, archi
tects and engineers need to be cognizant of the risks and related costs involved regarding
various choices available to them for improving the serviceability of structures due to high
winds and other loading conditions. This chapter addresses the design and implementation
issues and also quantitatively justiﬁes the use of the dampers within a riskbased decision
analysis framework.
148
The fullscale implementation of liquid dampers in airport control towers and
ers, bridge towers and offshore structures would require their integration into the overall
system. Moreover, the adoption of semiactive TLCDs requires additional equipment and
a more sophisticated setup as compared to a passive system. Figure 8.1 show some of the
(a) (b)
Figure 8.1 Implementation ideas for tuned liquid dampers (a) bridge towers (b)
tall buildings.
There are various factors which inﬂuence the selection of a dynamic vibration
absorber (DVA) for structures, namely: efﬁciency, size and compactness, capital cost,
149
operating cost, maintenance, safety, and reliability. In this section, a comparison among
three different types of DVAs, namely, the TMD, TLD and TLCD is made.
The TMD system installed in the Citicorp building is a sophisticated system with a
linear gas spring, pressure balance system, control actuator, power supply and electronic
TMD include in addition to the mass, gravity support system, and the spring system: a
ature sensors; an electronic control system which turns TMD on and off automatically.
Other parts of the TMD include restraint systems for TMDs including antiyaw torque
box, overtravel snubber system with reaction guides, and directional guides so that the
A TMD system needs to be designed in the face of several practical restraints. One
of the main disadvantages in the TMD operation is that although it is theoretically a pas
sive device, it needs electricity to operate. This is a problem since power could be lost dur
ing a high wind storm, a time when the TMD is expected to be operational (ENR, 1977).
Figure 8.2 shows the actual TMD system installed in the Citicorp building in New York.
150
Figure 8.2 TMD system installed in the Citicorp Building, New York City (taken
from Wiesner, 1979)
Pendulumtype TMDs with single and multistage suspensions have been devised. These
are advantageous for buildings with low frequency as the length of suspension can be
quite large for singlestage pendulumtype TMD as shown in Fig. 8.3(a, b) (Yamazaki et
al. 1992). Pendulumtype TMDs are usually augmented with coil springs for ﬁne tuning.
Mechanically guided slide tables, hydrostatic bearings, and laminated rubber bearings are
used to provide low friction platforms. For TMDs with laminated rubber bearings, the
bearings act as horizontal springs which eliminates the need for spring system. This type
of system is shown in Fig. 8.3 (c). Innovative methods for integrating TMDs into existing
buildings have been proposed by researchers. Mita and Feng (1994) proposed a megasub
151
control system which utilize substructures in a megastructure conﬁguration to act as
vibration absorbers. Similarly, researchers are considering the concept of a roof isolation
system in which the top ﬂoor or roof of a structure act as mass dampers.
Recent notable TMD applications include the skybridge in the Petronas towers,
Kaula Lumpur, Malaysia, where the legs of the bridge were found to be highly sensitive to
TMD systems for reduction of wind response in structures is provided by Kwok and
Figure 8.3 (a) Singlestage (b) multistage Pendulumtype TMDs (c) TMDs with
laminated rubber bearings (taken from Yamazaki et al. 1992)
complicated, TLDs are the most convenient to install and maintain due to the simplicity of
the device. Furthermore, maintenance costs of these dampers are practically nonexistent.
152
Due to their inherent simplicity, TLDs may be added to existing buildings as retroﬁt solu
tions, even for temporary use if desired, e.g., during construction phases of a structure. A
typical TLD may be designed in a variety of conﬁgurations ranging from rectangular tanks
The biggest advantage of liquid dampers is apparent in the case of tall buildings.
In most commercial buildings, water supply is needed for daytoday usage and for sprin
kler tanks used for ﬁreﬁghting purposes. The maintaining of water pressure can be effec
tively done by placing water reservoir tanks on roof tops, where the water ﬂows into
plumbing with its own gravity. So, instead of maintaining a high water level using special
ized pumping equipment, a water tank is an ideal costeffective solution. On the other
Due to the nature of the system, a small error may be expected when measuring the
still water level, which is the parameter that controls the fundamental sloshing frequency.
However, an important advantage that the liquid damper has over a TMD is that for wide
range of amplitudes of oscillation, particularly at higher levels, the system is not very sen
sitive to the actual frequency ratio between the primary and secondary systems. Another
major advantage of liquid dampers is that no activation mechanism is needed for their
operation. TMDs, for e.g., are designed to be activated at a certain threshold acceleration.
Note that for small and medium amplitudes of oscillation, proper tuning of the sys
tem may considerably inﬂuence the response. Some installations of TLDs include bafﬂes
and/or metallic balls to dissipate energy. However, the exact amount of damping cannot be
ascertained with these systems. Moreover, nonlinear frequency and damping characteris
tics inherent to these systems make them unsuitable for functioning as optimal devices.
153
Tuned Liquid Column Damper (TLCDs)
1. The damping in the TLCD can be controlled through the oriﬁce. The oriﬁce opening
ratio affects the headloss coefﬁcient which in turn affects the effective damping of the
liquid damper. Proportional valves can be actuated by a voltage signal obtained from a
battery to obtain the required damping without the use of heavy power.
2. The TLCD can be tuned by changing its frequency by way of adjusting the liquid col
umn in the tube. This is an attractive feature in case retuning becomes desirable in case
3. A mathematical model, which accurately describes the dynamics of the TLCD, can be
TLCD has the advantage of convenient mathematical formulation, but suffers from
the need for an appropriate tube length to satisfy the required frequency of oscillation.
Therefore, it may be in conﬂict with the available space allocated to house it. One way of
avoiding this is to introduce multiple TLCDs as discussed in Chapter 3. Figure 8.4 shows
the schematic of an actual TLCD implementation similar to the prototype studied in the
laboratory. Additional details are water level control system which has been introduced for
tuning control. This means that changes in structural frequency can be compensated by
154
Sensor Readings from Structure
Battery
Power TLCD Capacitance
CONTROL liquid level
CONSOLE
AIR
SUPPLY positioner
80 psi signal
420 mA
control
signal
Water
Pneumatic
Actuated liquid level cont
Control Unit Valve unit
Tiedowns
Bearing Surface
traditional approaches of increasing structural mass and stiffness. The Citicorp building’s
TMD cost was about $1.5 million (costs in 1977, in 2001 this is roughly $5.0 million);
however, it saved an overall structural cost of $4.0 million dollars that would have been
spent to add some 28,000 tons of structural steel to add lateral stiffness to the frame and
additional ﬂoor concrete to increase the mass of the structure (ENR, 1977). Typically, the
capital cost of a conventional TMD system is in the vicinity of 1% of the total building
cost. Table 8.1 lists some of the different components used in various systems. A prelimi
nary analysis of the cost of a fully functional TLCD system has been estimated to be
155
roughly 1/10 times the cost of an equivalent TMD system with similar performance in
response reduction.
Additional None, easy installation None, easy installation Local strengthening needed to
Construction during construction during construction support large amounts of spring
stages stages and actuator forces
Needs an overtravel snubber
system
Mechanical None Manual/actuated Valve Nitrogen Springs/ laminated rub
components Water level selftuning ber bearings/ Hydraulic bearings
control system Servovalve hydraulic actuators
Antiyaw torque box, linear
guideways
Pendulumtype TMDs
Electronic None Computer control sys Computer control system needed
components tem needed
156
8.3 Riskbased Decision Analysis
loading. There are primarily two types of adverse serviceability conditions caused by
strong winds. The ﬁrst is that excessive wind may cause large deﬂections in the structure
causing architectural damage to nonstructural members, for e.g., panels, cladding, etc.,
and affect elevator operation. The second is that the oscillatory motion may cause occu
pant discomfort or even panic. It is generally accepted that acceleration and the rate of
change of acceleration (commonly known as jerk) are the main causes of human discom
calculated assuming that failure occurs when the deﬂection or acceleration exceeds a cer
The example considered in this chapter is merely for illustration purposes. How
ever, the framework presented is quite general and could be applied to any system. The
building considered is a 60 story, 183 m tall building with a square base of 31 X 31 m. The
spectral characteristics of wind loads are deﬁned in Li and Kareem (1990). In this exam
ple, designers and building owners are considering the option of adding liquid dampers for
increasing the serviceability of this building under winds. Two types of TLCDs are con
sidered for application in the alongwind direction. The ﬁrst is a passive system with the
frequency of oscillation of liquid tuned to the ﬁrst mode frequency of the building while
the damping is optimized for design level wind speed. The second is a semiactive system,
In the case of passive system, the damping is assumed to be arising due to the fric
tion in the tube. The headloss coefﬁcient in this case is assumed to be equal to 1, which is
typical of such a system. In the case of semiactive system, the optimal damping ratio of
157
4.5% is maintained at all levels of excitation by means of a controllable oriﬁce using a
gainscheduled law as outlined in Chapters 5 and 7. The mass ratio (µ) is 1% and the tun
ing ratio (γ) is 0.99, which corresponds to a total mass of 280 tons and liquid column
date the total weight of the damper and these may be distributed on the building roof.
The RMS acceleration response of the uncontrolled and controlled response using
passive and semiactive systems is plotted as a function of the mean wind velocity at 10
meters height, U10 (Fig. 8.5). It can be seen from Table 8.2 that the dampers are effective
in reducing the structural accelerations and displacements. In this analysis, the effect of
bracing the structure is also examined. It has been assumed that the superstructure stiff
ness can be increased by a particular bracing system by 20%. Table 8.2 shows that the
bracing system is quite effective in reducing displacement but not equally effective in
reducing acceleration. Moreover, the bracing system increases signiﬁcantly the overall
From Table 8.2, it can be noted that there is an improvement of 1025% in RMS
acceleration response over the entire range of wind velocities using a semiactive system.
The semiactive system realizes a 45% improvement over the uncontrolled system. This
improvement justiﬁes small additional cost associated with a semiactive system, for e.g.,
sensors, controllable valves, etc. This analysis is based on the assumption that all the sys
tem parameters are known with certainty. The parametric uncertainty and the resulting
reliability of structural and loading parameters are treated in the following section.
158
20
Uncontrolled
Braced Structure
18 Passive Conrol
Semi−Active Control
16
Annoyance Threshold
10
Maximum permissable
RMS accelerations
8
14 16 18 20 22 24 26
Mean wind velocity at 10m height,
m/s U
10
Figure 8.5 Variation of RMS accelerations of the top ﬂoor with increasing wind
velocity
The decision making framework, shown in Fig. 8.6, is commonly composed of the follow
and associated probabilities and consequences. Each element of the analysis framework is
tion(s) to be clearly deﬁned. In our present example, the objective could be minimizing the
Decision variables: These could be the various decision alternatives available to the deci
sion maker. In our example, these could be the following alternatives available to the
building owners:
2. Invest in traditional bracing/outrigger systems to increase the lateral stiffness. The net
increase in the effective stiffness of the resulting structure due to the addition of bracing
is given by a factor kf deﬁned as the ratio of the stiffness of the structure with added
160
3. Install passive liquid dampers with optimal tuning ratio and optimal damping at design
wind speed. This is a suboptimal conﬁguration of the TLCD since the damping is pri
marily due to friction in the tube and a ﬁxed oriﬁce which cannot be controlled.
4. Install semiactive TLCD system which maintains the optimal damping at all levels of
response.
Decision outcomes: The various decision alternatives described above may have the fol
lowing outcomes:
2. Bracing systems and outrigger systems are expensive and are not as effective in reduc
ing acceleration which is the primary metric used to assess serviceability problems.
3. The passive liquid damper devices are effective in reducing displacement and accelera
tion responses, however they perform optimally only at the design wind speed.
4. Semiactive system is more effective than the passive system, however, there are addi
tional costs for controllable valves, computer control system, sensors and maintenance.
estimate the probabilities of failure and the associated costs/utility values of each decision
are examined. Finally these are integrated into a riskbased decision analysis tree. The risk
Risk = ∑ pi ( H , C i )U ( C i ) (8.1)
i
161
where p i ( H , C i ) is the probability of failure, H is the hazard, U ( C i ) is the utility func
tion and Ci are the consequences. The impact of risk can be improved by either reducing
the occurrence probability through system/component changes (which in our case refers to
matical functions of a combination of random variables that describe whether the structure
performs satisfactorily for the speciﬁc criteria it has been designed for. The design of
damping systems needs to consider the model and physical uncertainties, for e.g., struc
tural mass changes, damage to structure, hardening of concrete, loss of stiffness due to
corrosion and fracture, stiffness changes in foundation, etc. Changes could also be inher
ent in the loading, for e.g., wind climate, change in surface roughness, etc. The damper is
also not free from uncertainties, for e.g., decrease in its performance due to equipment
wear and tear. Therefore, all these variables need to be considered in probabilistic terms
For ultimate strength limit states, one is concerned about structural load and resis
tance, while for serviceability, the limit state represents the evaluation of a performance
criteria. For design of very tall and slender structures under winds, it is usually the service
ability limit state which often governs the design. The limit state function is usually writ
ten as,
Z = g ( X 1, X 2, …, X n ) (8.2)
and the probability of failure Pf for the component is deﬁned as,
162
Pf = ∫ f X ( X )dX (8.4)
g( X ) < 0
which describes the vector of random variables. In this case, the limit state function is a
hypersurface in the ndimensional space and separates the fail and safe regions. Usually,
standard reliability techniques, for e.g., First and secondorder reliability (FORM and
SORM) methods are used, wherein the limit state is linearized at the design point on the
failure surface (Ditlevsen, 1999). This procedure involves transformation of the variables
in the limit state equation to reduced normal variates which yields a new limit state equa
tion in the reduced space. The probability of failure is then determined from the reliability
index ( β̃ ), which is deﬁned as the shortest distance from the origin to the failure surface
P f = Φ ( – β̃ ) (8.5)
The limit state equation for drift serviceability is commonly written as:
where ∆ all is the allowable deﬂection, usually taken as = H b ⁄ 400 where H b is the height
Similarly, for comfort serviceability, the limit state equation is written as,
Z = σ ma – σ ẋ˙ (8.7)
where σ ma is the maximum allowable RMS accelerations, which lies between 510 mg in
the perception threshold range and 1015 mg in the annoyance level range. In this study
12 mg have been considered. Random variables used in the analysis are listed in Table 8.3.
163
The distribution of wind velocity for a well behaved wind climate can be adequately mod
eled by a Type 1 extreme value distribution. The other variables along with their statistical
characteristics, i.e., probability distribution, and mean and coefﬁcient of variation (COV)
can be found in Rojiani (1978) and Kareem (1990). The probability of failure for the dif
ferent systems under different mean wind velocities and different σ ma is tabulated in
Table 8.4.
Probability
Type #. Random Variable Mean COV
Distribution
Structural 1 Mass matrix multiplier, m̃ Normal 1.0 0.1
Parameters (nondimensional)
2 Normal 1.0 0.25
Stiffness matrix multiplier, k̃
(nondimensional)
3 1st mode damping, ζs Log Normal 1% 0.35
Wind Load 4 Air density, ρa Log Normal 1.25kg/m3 0.05
Parameters 5 Drag coefﬁcient, Cd Log Normal 1.2 0.17
6 Power law exponent, α̃ Log Normal 0.3 0.1
7 Wind Velocity, U10 Extreme Value 18, 20 m/s 0.1
Type 1
Liquid Damper 8 Tuning ratio, γ Normal 0.9870 0.1
Parameters 9 Coefﬁcient of Headloss, ξ Normal 1 0.1
10 Optimal Damping, ζf Log Normal 5.5 % 0.05
164
8.3.3 Cost and Utility Analysis
A generalized total expected cost function (for a period of T years) can be written as:
T T
C t = C s + C d + ∫ C m ( t ) dt + ∫ C f ( t ) dt (8.8)
0 0
where Cs is the initial ﬁxed cost of the structure, Cd is the initial ﬁxed cost of the damper,
Cm is the maintenance cost per unit year and Cf is the repair/business interruption cost per
unit year. The estimation of these cost functions requires a detailed analysis of the system
several factors, e.g., local market value and real estate demand. For a simpliﬁed analysis,
C f = T P f C(E) (8.9)
where C(E) is the cost of repair/ business interruption/ decreased employee productivity
when an event E occurs. In this analysis, C(E) has been assumed to be equal to 10. Table
8.5 tabulates some general costs and utilities of a typical tall building. Most of these values
are arrived at in an empirical way, however, the framework for more market value based
165
8.3.4 Riskbased Decision Analysis
Figure 8.7 shows a typical decision tree used to examine the given problem in a
systematic format. The decision tree includes decision and chance nodes. The decision
nodes are followed by possible actions which the decision maker takes. The chance nodes
are followed by outcomes that are beyond the control of the decision maker. The total
expected utility for each branch is computed and the decision is selected such that the
expected total utility function is minimized. As seen from Table 8.6, when the probabili
ties of failure are low, choosing semiactive dampers over passive dampers is not cost
fail
Cf*Pf
Cs*(1Pf)
} CA
Safe
C1
C2 } CB
}
C3
Decision
Node
C4 CC
Fixed
Costs
}
Chance
Nodes
CD
166
8.4 Design of Dampers
Usually water is the preferred liquid used in TLDs and TLCDs. It has been noted by
Fujino et al. 1988 that the use of high viscosity liquids do not offer any advantage. This is
because, for liquid dampers, there is an optimal level of damping that will provide the
desired level of response reduction, therefore, higher liquid viscosity is not always effec
tive.
The mass ratio is dictated by the efﬁciency (deﬁned as the ratio of response with
control system to response of uncontrolled structure) of the dampers needed. For e.g., if an
more than 1% mass ratio is possible to be placed on the top of tall buildings. For example,
TMD mass weighing up to 400 tons was installed in Citicorp Building. In case of TLDs
and TLCDs, this implies more space requirement, therefore innovative schemes to inte
grate these into water storage tanks and ﬁresprinkler tanks need to be designed.
The length ratio determines the horizontal to total length of the liquid column. The
length ratio also needs to be determined from an architectural point of view. For increasing
length ratio, the efﬁciency of the damper increases. However, two things need to be con
sidered. The vertical length of the tube should be high enough so that water does not spill
out of the tube. Secondly, water should remain in the vertical portion of the Utube at all
167
times to provide continuity in the water column in the horizontal segment. This can be
(l – b)
max { x f } ≤  (8.10)
2
Typically, auxiliary devices are tuned to the ﬁrst modal frequency of the structure.
An acceptable design is obtained by ensuring a tuning ratio of almost unity for mass ratio
of 1%. Exact values are provided for a variety of cases in chapter 3. In case the natural fre
quency of the structure changes by ∆ω s , the length of the water column in the Utube
–4 g
∆l = 3 ∆ω s (8.11)
( γ opt ω s )
This is the damping ratio of the liquid damper. For a regular TMD, this represents
the linear damping ratio. However, for liquid dampers the damping varies nonlinearly with
amplitude. Based on design curves obtained in Chapter 3, a damping ratio of about 4.5%
Number of Dampers
The number of dampers depends on various factors such as the available space,
shape and sizing of the damper units. In case of multiple dampers, it was shown in Chapter
3 that by increasing the number of dampers does not necessarily improve better perfor
168
mance concomitantly. A typical number of 5 units is usually adequate. Kareem and Kline
(1995) conducted numerical studies on multiple dampers with nonuniform mass distribu
tion and nonuniform frequency spacing. They concluded that such systems did not offer
any useful advantage over systems with uniform mass distribution and frequency spacing.
For structures with different fundamental frequencies in the two major directions,
tuning may be accomplished by using rectangular tanks or TLCDs. With proper design of
the damper dimensions, fundamental frequencies in both directions may be tuned. This is
important since the theory is based on tanks subjected to only a unidirectional excitation.
For structures with the same fundamental frequency in the two principal directions, a cir
taining optimal damping in TLCDs. Sensors on the buildings (accelerometers, liquid level
sensor, or anemometer) estimate the excitation level, which is used to adjust the headloss
Comparing Fig. 5.1 and Fig. 8.8, one can draw analogies wherein the lookup table
is the gainscheduler, the controllable valve of the TLCD is the regulator, and the head loss
coefﬁcient is the parameter being changed. The external environment is the wind loading
acting on the structure and the process represented by the combined structureTLCD sys
tem.
169
ξ)
change headloss coefficient (
Lookup Table
ξ = f(So)
Accelerometer/
Anemometer
Estimate Excitation
and loading intensity
U 10,S0
The ﬁrst step in the design of the dampers is to gather adequate knowledge of the
natural modes and damping of the structure being considered for control. The structural
characteristics are determined either at the design stage by analysis or for existing build
ings by monitoring fullscale data or a combination of both techniques. The ﬁrst method
involves a FEM analysis of the structural system. The second relies on analyzing fullscale
measurements from instrumented buildings. The response power spectral density provides
an estimate of the natural frequency and damping in the structure. Usually, it is advisable
to conduct full scale testing in order to obtain ambient or forced building response before
installing dampers. This is because FEM models usually not reliable for accurate esti
170
Loading Characteristics
site characteristics and hazard maps. Wind tunnel experiments are also needed for critical
projects to investigate the characteristics of wind force acting on the building and to esti
mate the structural response. This analysis is done during the design stages of the struc
ture. In this section, we will discuss alongwind loading only, although the acrosswind and
nS vv ( z, n ) 200 f

2
= 

5
(8.12)
uo 
( 1 + 50 f )
3
z – zd 10 – z d
= 2.5u o ln  ; u o = U 10 ⁄ 2.5 ln  ; zo =
nz
where f =  ; U ( z )
U ( z) z > 10m zo zo
surface roughness length; zd = zero plane displacement; U10 = mean wind velocity at 10m
1
2 2

–n [ C v ( z1 – z2 ) + C h ( x1 – x2 ) ]
2 2 2
coh = exp  (8.13)
1
 [ U ( z 1 ) + U ( z 2 ) ]
2
where (x1, z1) and (x2, z2) are the coordinates of the nodes, Cv and Ch are the coherence
decay coefﬁcients in the vertical and horizontal directions. The multiplepoint representa
tion may be simpliﬁed for linelike structures, e.g., buildings, towers, in which the spatial
variation of wind ﬂuctuations are only implemented for one spatial dimension. The wind
171
F j ( t ) = 0.5ρ a A bj C Dj ( U ( z j ) + v j )
2
(8.14)
where Abj is the tributary area exposed to wind, CDj is the drag coefﬁcient at the jth ﬂoor
and ρ a is the air density. From Eq. 8.14, one can also obtain the spectra of the loading,
given as: S FF ( ω, z ) = ( ρ a A bj C Dj U ( z ) ) S vv ( ω, z ) .
2
In the last section, the gainscheduled control was derived for different loading
intensities. In order to extend it to wind excited structures, one needs to ﬁnd relationship
between the wind force spectra, S FF ( ω ) , and an “equivalent” white noise excitation. For
small values of ζ s , one can approximate S FF ( ω ) by a equivalent white noise So, which is
the value of S FF ( ω ) at the natural frequency of the structure (Lutes and Sarkani, 1997).
This is shown schematically in Fig. 8.9(a) where using the following relationship:
S o ( U 10 ) = S FF ( ω s ) (8.15)
The equivalent white noise for an example case where ω s = 1 Hz is given in Fig. 8.9 (b).
0.4 1600
0.35 1400
H (ω)2
x
Magnitude of Transfer Function
s) S
0.3 1200
(lbf
Equivalent Loading intensity
2
0.25 1000
0.2 800
0.05 200
0 0
0.5 1 1.5 2 2.5 3 3.5 4 55 60 65 70 75 80 85 90 95 100
ω Frequency (Hz) U 10 (ft/s)
s (a) (b)
Figure 8.9 (a) Equivalent white noise concept (b) Variation of equivalent white
noise with wind velocity.
172
Damper Sizing
Once the structural and loading characteristics have been determined, the designer
can begin design of the damper. The optimum design parameters are discussed in Chapter
3. All symbols, unless explained here, refer to the earlier notations. The length of the
l = 2g ⁄ ω f
2
(8.16)
where ω f = γ opt ω s .
µM
A = 1 (8.17)
ρl
and for a spatially distributed single TLCD,
µM
A i = 1 (8.18)
Nρl
where N is the number of units and M1 is the generalized ﬁrst modal mass of the structure.
In case of multiple TLCDs, the length of liquid column and the cross sectional area of
l i = 2g ⁄ ω fi
2
(8.19)
µM
A i = 1 (8.20)
Nρl i
Next, from the wind loading excitation information, the headloss coefﬁcient can be deter
mined as follows,
2ζ opt glπ
ξ opt = 
 (8.21)
σ ẋ f
173
∞
σ ẋ f = S 0 ( U 10 ) ∫ H ẋ f F ( ω ) dω
2
(8.22)
0
The valve sizing should be selected such that the entire range of desired values of ξ
can be covered. This can be ensured by relating the headloss coefﬁcient to the valve con
ductance, CV for different angles of valve opening (see Appendix A.3). Typically, for most
8.4.4 Technology
Actuated Valves
modulation of the valve. The headloss characteristics for the valve are described in Appen
dix A.3.
174
Tubing Systems
Clear PVC piping systems are the best choice for the TLCD tube construction.
This is because they are rugged and durable, yet allow easy maintenance and visibility of
the liquid.
Sensors
A capacitance type liquid level sensor is needed to determine the liquid level in the
TLCD. This is important for tuning the TLCD to the building frequency. This needs to be
done on a regular basis because changes in structural frequency may take place due to
the system. Additionally, accelerometers and anemometers for estimating the loading
characteristics are needed. These are commercially available from a variety of vendors. It
should be noted that accelerometers chosen should have good frequency characteristics in
the low frequency region (< 1 Hz). This is because the response of tall buildings is prima
system running on auxiliary power is quite affordable these days. A typical computer run
ning a data acquisition and control implementation software can be set up very cheaply.
The system can also be conﬁgured to include remote control using TCP/IP system which
enables offsite users to monitor the system, which eliminates the need for an on site oper
ator.
175
8.5 Concluding Remarks
This chapter discussed the design consideration and implementation details of liq
uid dampers. Different dynamic vibration absorbers, namely TMDs, TLDs and TLCDs are
compared in terms of implementation and costs. Next, a probabilistic framework for deci
sion analysis concerning the serviceability of a building has been presented. Both deter
ministic and reliabilitybased analyses conﬁrm the attractiveness of the passive and semi
active liquid dampers in reducing acceleration response and the associated probabilities of
failure. The decision analysis framework presented here would facilitate building owners/
viewpoint at a minimum cost. Finally, some design guidelines for technology transfer are
176
CHAPTER 9
CONCLUSIONS
What we call the beginning is often the end
And to make an end is to make a beginning.
The end is where we start from.
 T.S.Elliot
This research focussed on the development of the next generation of liquid damp
ers for mitigation of structural response. Two type of liquid dampers, namely the sloshing
dampers (TLDs) and the liquid column dampers (TLCDs) were considered. Firstly, a new
sloshingslamming analogy was presented for sloshing type dampers. It was noted that the
existing models neglect the effect of impact of liquid on the container walls. The ﬁrst
This involves modeling the TLD as a linear system augmented with an impact subsystem.
This analogy captures the essence of the underlying physics behind the complexity of the
sloshing phenomenon at higher amplitudes. The second approach uses certain nonlinear
functions, described as impact characteristic functions, which can succinctly describe the
phenomenological behavior of the TLD. The parameters of this model are derived from
experimental studies. Experiments were also conducted to study the local pressures on the
walls of the container and to better understand the nature of the impact process. It was
observed that the peak pressures occur at the static liquid height. The pressuretime inte
gration shows that the contribution of the impact pulse to the overall sloshing pulse is
177
approximately 2025%. This feature may play an important role in future modeling stud
ies on TLDs.
Next, analytical modeling of tuned liquid column dampers (TLCDs) was consid
ered. Optimum absorber parameters (i.e., tuning ratio and damping ratio) were derived for
a variety of loading cases ranging from white noise excitation to ﬁltered white noise cases.
The theoretically obtained optimum absorber parameters were compared with experimen
tal results and the match was found to very be good. The optimum absorber parameters
were also determined for the case of multiple TLCDs (MTLCDs). These parameters
include number of TLCDs, the frequency range and the damping ratio of each damper.
MTLCDs are more robust as compared to single TLCDs with respect to changes in the
primary system frequency. Moreover, the smaller size of MTLCDs offers convenient port
systems. This involves transfer of energy from one system to another and in some
instances could be harmful to the structure. It has been observed that beyond a certain
level of damping in the secondary system (i.e., the damper), the beat phenomenon ceases
to exist. A mathematical and experimental study of the beat phenomenon was conducted.
It was noted that the disappearance of the beat phenomenon is attributed to the coalescing
of the modal frequencies of the combined system. Experimental validation of the beat phe
TLCDs. These include gainscheduling and clipped optimal system with continuously
varying and onoff control. Gainscheduled control is useful for disturbances which are of
longduration and slowlyvarying (e.g., wind excitation) and where the steadystate
178
response is the control objective. The headloss coefﬁcient is changed adaptively in accor
dance with a lookup table by changing the valve/oriﬁce opening. This type of semiactive
system leads to 1525% improvement over a passive system. The application of these sys
tems for offshore structures was also considered. Experimental validation of the gain
scheduled system was done in the laboratory using a prototype TLCD equipped with a
ble valve with negligible valve dynamics and whose coefﬁcient of headloss can be
changed by applying a command voltage. This type of control is more suited for excita
tions which are transient in nature, for e.g., sudden wind gusts or earthquakes. The efﬁ
response reduction than the passive systems for both random and harmonic excitations. In
the case of harmonic loading, the improvement was about 2530% while for random exci
tation, the improvement was about 1015% over a passive system. It was also noted that
ment in response reduction over the relatively simple onoff control algorithm.
developed for testing liquid dampers. The main advantages, namely the cost effectiveness
and repeatability of the test, is realized due to the fact that a virtual structure simulated in
Finally, the design, implementation, cost and riskbased decision analysis for the
use of liquid dampers in structural vibration control was laid out. Comparisons were made
between different dynamic vibration absorbers (DVAs), namely the TMDs, TLDs and
179
TLCDs. It was estimated that the cost of a fully functional TLCD system has been esti
mated to be 1/10 times the cost of a TMD system with a similar level of performance. The
designers to ensure adequate lifecycle reliability of the building from serviceability view
point at a minimum cost. It was concluded that when the probabilities of failure are low,
choosing semiactive dampers over passive dampers is not cost effective. However, in crit
ically unserviceable structures, the semiactive scheme delivers better cost/utility beneﬁts.
1. In the sloshingslamming analogy of TLDs, the mass exchange parameter was deter
mined from empirical relationships obtained through experiments, which relate the
could be further reﬁned should it be possible to quantify more accurately the mass
exchange between the sloshing and slamming modes from theoretical considerations.
2. The sloshing pressures and forces obtained during experiments should be compared to
numerical sloshing studies which incorporate the slamming/impact action of the liquid.
test of the structuredamper system and then verifying it using a HIL simulation.
type excitations in order to provide proof of concept for the damping schemes. A more
TLCD is needed before installing these dampers on actual structural systems. This will
180
REFERENCES
[1] Kareem, A. and Yalla, S.K. (1997), “Liquid Dampers: Recent Developments and
Applications’, submitted to the monograph on Structural Control.
[2] Yalla S.K., Kareem, A. and Kantor, J.C. (1998), “SemiActive Control Strategies for
Tuned Liquid Column Dampers to Reduce Wind and Seismic Response of Structures,”
2nd World Conference on Structural Control, Kyoto, John Wiley and Sons, 559568.
[3] Kareem, A., Kabat, S., Haan, F. Jr., Mei, G. and Yalla, S.K. (1998) “Modeling and
Control of Wind Induced Response of a TV Tower,” 2nd World Conference on Struc
tural Control, Kyoto, John Wiley and Sons, 24212430.
[4] Yalla, S.K. and Kareem, A. (1999) “Modeling of TLDs as SloshingSlamming Damp
ers,” Wind Engineering into the 21st century: Proc. 10th Int. Conf. on Wind Engng.,
Copenhagen, Balkema Press, 15691575.
[5] Yalla, S.K and Kareem, A. (2000a) “Optimum Absorber Parameters for Tuned Liquid
Column Dampers ,” ASCE Journal of Structural Engineering, 125(8), 906915.
[6] Yalla, S.K., Kareem, A. and Kantor, J.C. (2000b), “SemiActive Variable Damping
Tuned Liquid Column Dampers,” Proc. of the 7th SPIE Conf. on Smart Sructures and
Materials, Newport Beach, CA.
[7] Yalla, S.K. and Kareem, A. (2000c), “On the Beat Phenomenon in Coupled Systems,”
Proc. of the 8th ASCE Speciality Conf. on Probabilistic Mechanics and Structural Reli
ability, University of Notre Dame, CDROM.
[8] Yalla, S.K., Kareem A. and Abdelrazaq, A.K. (2000d), “Riskbased Decision Analysis
for the Building Serviceability,” Proc. of the 8th ASCE Speciality Conf. on Probabilis
tic Mechanics and Structural Reliability, University of Notre Dame, CDROM.
[9] Yalla, S.K. and Kareem, A. (2001a), “Beat Phenomenon in Combined StructureLiq
uid Damper Systems,” 23(6), Engineering Structures, 622630.
[10] Yalla, S.K. and Kareem, A. (2001b), “Hardwareintheloop Simulation: A case study
for Liquid Dampers,” Proceedings of the Mechanics and Materials Summer Confer
ence, San Diego, CA.
[11] Yalla, S.K., Kareem, A. and Kantor, J.C. (2001c), “SemiActive Tuned Liquid Col
umn Dampers for vibration control of structures,” Engineering Structures (in press).
184
[12] Yalla, S.K. and Kareem, A. (2001d), “SloshingSlamming S2 Damper Analogy for
Tuned Liquid Dampers,” ASCE Journal of Engineering Mechanics (in press).
[13] Yalla, S.K. and Kareem, A. (2001e), “Semiactive Tuned Liquid Column Dampers
for mitigation of wind induced vibrations: Experiments,” submitted to ASCE Journal of
Structural Engineering, special issue on Semiactive control.
[14] Yalla, S.K. and Kareem, A. (2001f), “Modeling TLDs using Impact Characteristics:
Experiments and System Identiﬁcation,” to be submitted to Earthquake Engineering
and Structural Dynamics.
Bibliography
[1]Abe, M., Kimura, S. and Fujino, Y. (1996), “Control Laws for SemiActive Tuned Liq
uid Column Damper with Variable Oriﬁce Opening”, 2nd Int. Workshop on Structural
Control, Hong Kong.
[2] Abramson, H.N. (ed.) (1966), “The Dynamic Behavior of Liquids in Moving Contain
ers”, NASA, SP106.
[3] Astrom, K.J and Wittenmark, B. “Adaptive Control”, Addison Wesley, 1989.
[4] Babitsky, V.I. (1998), Theory of Vibroimpact Systems and Applications, Springer, N.Y.
[5] Balendra, T., Wang, C.M. and Cheong, H.F. (1995) “Effectiveness of Tuned Liquid
Column Dampers for Vibration Control of Towers”, Engineering Structures, 17(9),
668675.
[6] Bass, R.L., Bowles, E.B. and Cox, P.A. (1980), “Liquid Dynamic Loads in LNG Cargo
Tanks,” SNAME Transactions, 88, 103126.
[7] Bauer, H.F. (1972), “On the destabilizing effect of liquids in various vehicles Part 1,”
Vehicle System Dynamics, 1, 227260.
[8] Bergman, L.A., Mc Farland, D.M., Hall, J.K., Johnson, E.A. and Kareem, A. (1989),
“Optimal Distribution of Tuned Mass Dampers in Wind Sensitive Structures,” Proceed
ings of 5th ICOSSAR, New York.
[9] Blevins, R.D. (1984), Applied Fluid Dynamics Handbook, Van Nostrand Reinhold.
[10] Brancaleoni, F. (1992), “The Construction phase and its Aerodynamic issues,”Aero
dynamics of Large Bridges Proceedings of the First International Symposium on
Aerodynamics of large Bridges, Copenhagen, Denmark.
[11] Brebbia et al. (1975), Vibrations of Engineering Structures, Lecture Notes in Engi
neering (10), SpringerVerlag.
[12] Breukelman, B., Irwin, P., Gamble, S. and Stone, G. (1998), “The practical applica
tion of vibration absorbers in controlling wind serviceability and fatigue problems,”
Proceedings of Structural engineers World Congress, San Francisco, July.
185
[13] Caughey, T.K. (1963), “Equivalent Linearization Techniques,” J. Acoust. Soc. Am.,
35, 17061711.
[14] Chang, C.C. and Hsu, C.T. (1999), “Control performance of Liquid column vibration
absorber,” Engineering Structures, 20(7), 580586.
[15] Chang, P.M., Lou, J.Y.K. and Lutes, L.D. (1997), “Model Identiﬁcation and Control
of a tuned liquid damper,” Engineering Structures.
[16] Chen, W., Haroun, M. A. and Liu, F. (1996), “Large amplitude Liquid Sloshing in
Seismically Excited Tanks,” Earthquake Engineering and Structural Dynamics, 25,
653669.
[17] Chung, C.C. and Gu, M. (1999), “Suppression of vortexexcited vibration of build
ings using tuned liquid dampers,” Journal of Wind Engg. and Ind. Aerodynamics, 83,
225237.
[18] Delrieu, J.L.(1994), “Cost Effective Deepwater Platform Dedicated to West Africa
and Brazil”, Proc. 7th Int. Conf. on the behavior of Offshore Structures, 2, 805812.
[19] Den Hartog, J.P. (1956), Mechanical Vibrations, 4th Ed, McGrawHill.
[20] Ditlevsen, O. (1999), Structural Reliability Methods, John Wiley and Sons.
[21] Dyke, S.J., Spencer, B.F. Jr., Sain, M.K. and Carlson, J.D. (1996) “Modeling and
Control of Magnetorheological Dampers for Seismic Response Reduction,” Smart
Materials and Structures, 5, 565575.
[22] Ehlers, J. (1987), “Active and SemiActive Control Methods in WaveStructure Inter
action,” (incomplete reference)
[23] ENR magazine (1977), “Tuned mass dampers steady sway of skyscrapers in Wind”.
[24] Faltinson, O. M. (1978), “A Numerical Nonlinear Method of Sloshing in Tanks with
TwoDimensional Flow,” Journal of Ship Research, 22.
[25] Faltinsen, O.M. and Rognebakke, O.F. (1999), “Sloshing and Slamming in Tanks,”
Hydronav’99Manouvering’99, GdanskOstrada, Poland.
[26] Fediw, A.A., Isyumov, N. and Vickery, B.J. (1993), “Performance of a onedimen
sional Tuned sloshing water damper,” Wind Engineering, London, 247256.
[27] Fujino, Y., Pacheco, B.M., Chaiseri, P. and Sun, L.M. (1988), “Parametric studies on
TLD using circular containers by freeoscillation experiments,” Struc. Engng and
Earthquake Engng., 5(2), 381391.
[28] Fujino,Y., Sun, L.M., Paceno, B. and Chaiseri, P. (1992), “Tuned Liquid Damper
(TLD) for Suppressing Horizontal Motion of Structures,” ASCE Journal of Engineer
ing Mechanics, 118(10), 20172030.
[29] Fujino, Y. and Sun, L.M. (1993), “Vibration Control by multiple tuned liquid damp
ers,” Journal of Structural Engineering, ASCE, 119(12).
[30] Gao, H., Kwok, K.C.S. and Samali, B. (1997),“Optimization of Tuned Liquid Col
umn Dampers”, Engineering Structures, 19, 476486.
186
[31] Gao, H., Kwok, K.C.S. and Samali, B. (1999),“Charcteristics of multiple Tuned Liq
uid Column Dampers in suppressing Structural Vibration,” Engineering Structures, 21,
316331.
[32] Grace, A. (1992), MATLAB Optimization Toolbox User’s Guide, MathWorks, Inc.
[33] Graham, E.W. and Rodriguez, A.M. (1962), “The Characteristics of Fuel Motion
which affect Airplane Dynamics,” Journal of Applied Mechanics, ASME, 123(9), 381
388.
[34] Gurley, K. and Kareem, A. (1994), “On the Analysis and Simulation of Random Pro
cesses Utilizing Higher Order Spectra and Wavelet Transforms,” Proc. of 2nd Int. Conf.
on Computational Stochastic Mechanics, Athens, Greece, Balkema Publishers.
[35] Hattori, M., Atsushi, A. and Yui, T. (1994), “Wave impact pressures on vertical walls
under breaking waves of various types,” Coastal Engineering, 22, 79114.
[36] Haroun, M.A. and Pires J.A. (1994), “Active oriﬁce control in Hybrid liquid column
dampers,” Proceedings of the First World Conference on Wind Engineering, Vol.I, Los
Angeles.
[37] Hitchcock, P.A., Kwok, K.C.S., Watkins, R.D. and Samali, B. (1997), “Characteris
tics of Liquid column vibration absorbers (LCVA) I and II”, Engineering Structures,
19(2).
[38] Hitchcock, P.A., Glanville, M.J., Kwok, K.C.S., Watkins, R.D. and Samali, B. (1999),
“Damping properties and windinduced response of a steel frame tower ﬁtted with liq
uid column vibration absorbers,” Journal of Wind Engineering and Industrial Aerody
namics, (83), 183196
[39] Housner, G.W., Bergman, L.A., Caughey, T.K., Chassiakos, A.G., Claus, R.O., Masri,
S.F., Skelton, R.E., Soong, T.T., Spencer, B.F. and Yao, J.T.P. (1997), “Structural Con
trol: Past, Present, and Future,” Journal of Engineering Mechanics, 123(9), 897971.
[40] Honkanen, M.G. (1990), “Heel and Roll Control by Water Tank”, Naval Architect,
215216.
[41] Hrovat, D., Barak, P. and Rabins, M. (1983), “Semiactive versus Passive or Active
Tuned Mass Dampers for Structural Control,” Journal of Engineering Mechanics,
ASCE, 109(3), 691705.
[42] Hsieh, C.C., Kareem. A. and Williams, A.N. (1988), “Wave Phase Effects on
Dynamic Response of Offshore Platforms,” Proc. of the Offshore Mechanics and Arctic
Engineering , ASME, Houston, TX.
[43] Hunt, K.H. and Crossley, F.R.E. (1975), “Coefﬁcient of restitution interpreted as
Damping in Vibroimpact,” Transactions of the ASME, Journal of Applied Mechanics,
440444.
[44] Huse, E. (1987), “Free Surface Damping Tanks to reduce Motions of Offshore Struc
tures,” Proc. of 6th Int. Symp. on Offshore Mechanics and Arctic Engineering, 313
324.
187
[45] Ibrahim, R.I., Gau, J.S. and Soundarajan, A. (1988), “Parametric and autoparametric
vibrations of an elevated water tower Part I: Parametric response,” Journal of Sound
and Vibration, 225(5), 857885.
[46] Ibrahim, R.A., Pilipchuk, V.N. and Ikeda, T. (2001), “Recent Advances in Liquid
Sloshing Dynamics,” Applied Mechanics Reviews, 54(2), 133199.
[47] Ioi, T. and Ikeda, K. (1978) “On the Dynamic Vibration Damped Absorber of the
Vibration System”, Bull. of Japanese Society of Mech. Engineers, 21 (151), 6471.
[48] Iserman, R., Schaffnit, J. and Sinsel, S. (1999), “Hardwareintheloop simulation for
the design and testing of enginecontrol systems,” Control Engineering Practice, 7,
643653.
[49] Ivers, D.E. and Miller L.R. (1991), “Semiactive Suspension Technology: an evolu
tionary view, Advanced Automotive Technologies, ASME, 32746.
[50] Kanai, K. (1961), “An Empirical Formula for the Spectrum of Strong Earthquake
Motions,” Bull. Earthquake Research Inst., Univ. of Tokyo, Japan, 39.
[51] Kareem, A. and Sun, W.J. (1987), “Stochastic Response of Structures with Fluid
Containing Appendages,” Journal of Sound and Vibration, 119(3).
[52] Kareem, A. and Li, Y. (1988), “Stochastic response of a TLP to wind and wave
ﬁelds,” Department of Civil Engineering, University of Houston, Tech. report
UHCE8818.
[53] Kareem, A. (1990), “Reliability analysis of Windsensitive structures,” Journal of
Wind Engg. and Ind. Aerodynamics, 23, 495514.
[54] Kareem, A. (1993), “Liquid Tuned Mass Dampers: Past, Present and Future,” Pro
ceedings of the Seventh U.S. National Conference on Wind Engineering, Vol. I, Los
Angeles.
[55] Kareem, A. (1994) “The next generation of Tuned liquid dampers,” Proceedings of
the First World Conference on Structural Control, Vol.I, Los Angeles.
[56] Kareem, A. and Kline, S. (1995), “Performance of Multiple Mass Dampers under
Random loading,” Journal of Structural Engineering, ASCE, 121(2), 348361.
[57] Kareem, A. and Gurley, K. (1996), “Damping in Structures: Its evaluation and Treat
ment of Uncertainity,” Journal of Wind Engineering and Structural Aerodynamics, 59,
131157.
[58] Kareem, A., Kijewski, T. and Tamura, Y. (1999), “Mitigation of motion of Tall build
ings with recent applications,” Wind and Structures, 2(3), 201251.
[59] Karnopp, D., Crosby, M.J. and Harwood, R.A. (1974), “Vibration Control using
SemiActive Force Generators,” ASME Journal of Engineering for Industry, 96(2),
619626.
[60] Karnopp, D. (1990), “Design Principles for vibration control systems using Semi
active dampers,” Journal of Dynamic systems, Measurement and Control, 112, 44855.
188
[61] Koh, C.G., Mahatma, S. and Wang, C.M. (1994) “Theoretical and experimental stud
ies on rectangular liquid dampers under arbitrary excitations,” Earthquake Engng. and
Struc. Dynamics. 23, 1731.
[62] Kim, Y. (2001), “Numerical simulation of Sloshing ﬂows with impact loads,” Applied
Ocean Research, 23, 5362.
[63] Kurino, H. and Kobori, T. (1998), “SemiActive Structural response control by opti
mizing the Forcedeformation loop of Variable Damper,” Proceedings of Second World
Conference on Structural Control, Kyoto, John Wiley and Sons, 407417.
[64] Kwok, K.C.S. and Samali, B. (1995), “Performance of Tuned Mass Dampers under
wind loads,” Engineering Structures, 17(9), 655667.
[65] Lepelletier, T.G. and Raichlen, F. (1988), “Nonlinear Oscillations in Rectangular
Tanks,” Journal of Engineering Mechanics, ASCE, 114(1).
[66] Li, Y. and Kareem, A. (1990), “Recursive Modeling of Dynamic systems”, Journal of
Engineering Mechanics, ASCE, 116, 660679.
[67] Li, Y. and Kareem, A. (1993), “Simulation of MultiVariate Random Processes:
Hybrid DFT and Digital Filtering Approach,” Journal of Engineering Mechanics,
ASCE, 119(5), 10781098.
[68] Lou, Y.K., Lutes, L.D. and Li, J.J. (1994), “Active Tuned Liquid Damper for Struc
tural Control,” Proceedings of the First World Conference on Wind Engineering, Vol.I,
Los Angeles.
[69] Lutes, L.D. and Sarkani, S. (1997), Stochastic Analysis of Structural and Mechanical
Vibrations, PrenticeHall.
[70] Lyons, J.L. (1982), Lyon’s Valve Designers Handbook, Van Nostrand Reinhold Co.
[71] Maciejowski, J.M. (1989), Multivariable Feedback Design, AddisonWesley, Inc.
[72] Masri, S.F. and Caughey, T.K. (1966), “On the Stability of the Impact Damper,” Jour
nal of Applied Mechanics, ASME, 586592.
[73] McNamara, R. J. (1977), “Tuned Mass Dampers for Buildings,” Journal of Structural
Division, ASCE, 103, 9,17851789.
[74] Mita, A. and Feng, M.Q. (1994), “Response control strategy for tall buildings using
interaction between Mega and Substructures,” Proc. Int. workshop on Civil Infrastruc
ture systems, Taipei, Taiwan, PRC, 329341.
[75] Modi, V.J. and Welt F. (1987), “Vibration control using nutation dampers,” Proceed
ings Int. conf. on Flowinduced vibrations, England, 369376.
[76] Modi, V.J., Welt, F. and Seto, M.L. (1995), “Control of WindInduced Instabilities
though application of Nutation Dampers: A brief overview,” Engineering Structures,
17(9), 626638.
[77] Natani, S. (1998), “Active Sloshing Damping System with Inverse Sloshing Genera
tion,” Proceedings of 2nd World conf. on Struc. Control, Kyoto, John Wiley and Sons.
189
[78] Okomoto, T. and Kawahara, M. (1990), “Twodimensional Sloshing analysis by
Lagrangian Finite element method,” Int. Journal of Numerical methods in Engng., 453
477.
[79] Patel, M. H. and Witz, J.A., (1985),“On Improvements to the Design of Tensioned
Buoyant Platforms”, Behavior of Offshore Structures, 563573.
[80] Pilipchuk, V.N. and Ibrahim, R.A. (1997), “The Dynamics of a nonlinear system sim
ulating liquid sloshing impact in moving structures,” Journal of Sound and Vibration,
205(5), 593615.
[81] Rana, R. and Soong, T.T. (1998), “Parametric Study and Simpliﬁed Design of Tuned
Mass Dampers,” Engineering Structures, 20(3), 193204.
[82] Randall, S.E., Halsted, D.M. and Taylor, D.L. (1981), “Optimum Vibration Absorbers
for Linear Damped Systems,” Journal of Mechanical Design, ASME, 103, 908913.
[83] Reed, D., Yu, J., Yeh, H. and Gardarsson, S. (1998), “An Investigation of Tuned Liq
uid Dampers under large amplitude excitation,” Journal of Engineering Mechanics,
ASCE, 124, 405413.
[84] Roberson, R.E. (1952), “Synthesis of a Nonlinear Dynamic Vibration Absorber,”
Journal of Franklin Institute, 205220.
[85] Roberts, J.B. and Spanos, P.D. (1990), Random Vibration and Statistical Lineariza
tion, Wiley, New York.
[86] Rojiani, K.B. (1978), “Evaluation of Steel Buildings to Wind Loadings,” Ph.D. The
sis, University of Illinois, UrbanaChampaign.
[87] Sadek, F., Mohraz, B. and Lew, H.S. (1998) “Single and MultipleTuned Liquid Col
umn Dampers for Seismic Applications,” Earthquake Engng. and Struc. Dyn., 27, 439
463.
[88] Sakai, F. et al. (1989), “Tuned Liquid Column Damper  New Type Device for Sup
pression of Building Vibrations,” Proc. Int. Conf. on High Rise Buildings, Nanjing,
China, March 2527.
[89] Sakai, F.,and Takaeda. S. (1991), “Tuned Liquid Column Damper (TLCD) for cable
stayed bridges,” Innovation in Cablestayed Bridges, Fukonova, Japan.
[90] Schmidt, R., Oumeraci, H. and Partenscky, H.W. (1992), “Impact loads induced by
plunging breakwaters on vertical structures,” Coastal Engineering, 15451558.
[91] Sellars, F. H., and Martin, P.M. (1992), “Selection and Evaluation of Ship Roll Stabi
lization Systems,” Marine Technology, 29(2), 84101.
[92] Semercigil, S.E., Lammers, D. and Ying, Z. (1992), “A new Tuned Vibration
Absorber for Wideband Excitations,” Journal of Sound and Vibration, 156 (3), 445
459.
[93] Shimizu, K. and Teramura, A. (1994), “Development of vibration control system
using Ushaped tank”, Proceedings of the Ist International Workshop and Seminar on
Behavior of Steel Structures in Seismic Areas, Timisoara, Romania, 7.257.34.
190
[94] Soong, T.T. (1991), Active Structural ControlTheory and Practice, Longman, Lon
don & Wiley, New York.
[95] Soong, T.T. and Dargush, G.F. (1997), Passive Energy Dissipation Systems in Struc
tural Engineering, Wiley, New York.
[96] Spencer, B.F., Jr. and Sain, M.K. (1997), “Controlling Buildings: A New Frontier in
Feedback,” IEEE Control Systems Magazine: Special Issue on Emerging Technolo
gies,”17( 6), 1935.
[97] Sudjic, D. (1993), “Their love keeps lifting us higher,” Telegraph magazine, May, 17
25.
[98] Suhardjo, J. and Kareem, A. (1997), “Structural Control of Offshore Platforms,” Pro
ceedings of the 7th International Offshore and Polar Engineering Conference IOSPE
97, Honolulu.
[99] Suhardjo, J., Spencer, Jr., B.F. and Kareem, A. (1992a), “Active Control of Wind
excited Buildings: A Frequency Domain based Design Approach,” Journal of Wind
Engineering and Industrial Aerodynamics, 4144.
[100] Suhardjo, J., Spencer, Jr., B.F. and Kareem, A. (1992b), “Frequency Domain Opti
mal Control of Wind Excited Buildings,” Journal of Engineering Mechanics, ASCE,
118(12).
[101] Suhardjo, J. and Kareem, A. (2001), “Feedbackfeedforward control of offshore
platforms under random waves,” Earthquake Engng Struct. Dyn., 30, 213235.
[102] Sun, L.M., Fujino, Y., Paceno, B. and Chaiseri, P. (1991), “Modeling Tuned Liquid
Damper,” Proc. of the 8th International Conference on Wind Engineering, Elsevier.
[103] Sun, L.M. and Fujino, Y. (1994), “A Semianalytical model for Tuned Liquid
Damper (TLD) with wave breaking,” Journal of Fluids and Structures, 8, 471488.
[104] Sun, L. M., Fujino,Y., Paceno, B. and Chaiseri, P. (1995), “The properties of Tuned
Liquid Dampers using a TMD analogy,” Earthquake Engng and Struc. Dynamics, 24,
967976.
[105] Symans, M.D. and Constantinou, M.C. (1999), “Semiactive control systems for
seismic protection of structures: a stateoftheart review,” Engineering Structures (21),
469487.
[106] Symans, M.D. and Constantinou, M.C. (1997), “Experimental testing and Analyti
cal modeling of Semiactive ﬂuid dampers for seismic protection, “ Journal of Intelli
gent Material Systems and Structures, 8(8), 644657.
[107] Symans, M.D. and Kelly, S.W. (1999), “Fuzzy Logic Control of Bridge Structures
using Intelligent SemiActive Seismic Isolation Systems,” Earthquake Engineering and
Structural Dynamics, 28, 3760.
[108] Tajimi, H. (1960), “A Statistical Method of Determining the Maximum Response of
a Building Structure During an Earthquake,” Proceedings 2nd World Conference on
Earthquake Engng, Vol. II, Tokyo and Kyoto, Japan, 781798.
191
[109] Tamura, Y., Fujii, K., Ohtsuki, T., Wakahara, T. and Kohsaka, R. (1995), “Effective
ness of Tuned Liquid Dampers under Wind Excitation,” Engineering Structures, 17 (9),
609621.
[110] Tokarcyzk, B.L. (1997), “The Mathematical Modeling of a Tuned Liquid Damper,”
M.S. Thesis, Department of Civil Engineering, Texas A&M University, College Sta
tion, TX.
[111] Ueda. T., Nakagaki, R. and Koshida, K. (1992) “Suppression of windinduced vibra
tion by dynamic dampers in towerlike structures,” J. of Wind Engg. and Ind. Aerody
namics, 4144, 19071918.
[112] Vandiver, J. K. and Mitome, S. (1978), “Effect of Liquid Storage Tanks on the
Dynamic Response of Offshore Platforms,” Dynamic Analysis of Offshore Structures:
Recent Developments.
[113] Venugopal, M. (1990), “A Toroidal Hydrodynamic Absorber for Damping Low Fre
quency motions of Fixed and Floating Offshore Platforms,” Marine Technology, 27(1),
4246.
[114] Warburton, G.B. and Ayorinde, E.O. (1980), “Optimum Absorber Parameters for
Simple Systems”, Earthquake Engineers and Structural Dynamics, 8, 197217.
[115] Warburton, G.B. (1982), “Optimal Absorber Parameters for Various Combination of
Response and Excitation parameters,” Earthquake Engineering and Structural Dynam
ics, 10, 381401.
[116] Warnitchai, P. and Pinkaew, T. (1998), “Modeling of Liquid Sloshing in Rectangular
Tanks with Flow Dampening Devices,” Engineering Structures, 20(7), 593600.
[117] Weisner, K.B. (1979), “Tuned Mass Dampers to reduce building sway,” presented at
the ASCE Boston Convention, April.
[118] Won, A.Y.J., Pires, J.A. and Haroun, M.A. (1996) “Stochastic seismic performance
evaluation of tuned liquid column dampers”, Earthquake Engng & Structural Dynam
ics, 25, 12591274.
[119] Wu, G.X., Ma, Q.W. and Taylor, E.R. (1998), “Numerical Simulation of Sloshing
Waves in a 3D tank based on a Finite Element Method,” Applied Ocean Research, 20,
337355.
[120] Xu, Y.L, Samali, B. and Kwok, K.C.S. (1992), “Control of Alongwind Response of
Structures by Mass and Liquid Dampers,” Journal of Engineering Mechanics, 118(1),
2039.
[121] Xue, S.M., Ko, J.M. and Xu, Y.L. (2000), “Tuned Liquid Column Damper for sup
pressing pitching motion of structures,” Engineering Structures, 23, 15381551.
[122] Yamaguchi, H. and Harnpornchai, N. (1993), “Fundamental Characteristics of Mul
tiple Tuned Mass Dampers for Suppressing Harmonically Forced Oscillation,” Journal
of Earthquake Engineering and Structural Dynamics, 22.
192
[123] Yamazaki, S., Nagata, N. and Abiru, H. (1992), “Tuned active dampers installed in
the Minato Mirai (MM) 21 Landmark tower in Yokohama,” Journal of Wind Engng and
Industrial Aerodynamics 4144, 19371948.
[124] Yang, J.N., Akbarpour, A. and Ghaemmaghami, P. (1987), “New Control algorithms
for Structural Control,” ASCE Journal of Engineering Mechanics, 113, 136986.
[125] Yao, J.T.P. (1972), “Concept of Structural Control,” ASCE, Journal of Structural
Division, 98, 156774.
[126] Yoshida, K., Yoshida, S. and Takeda, Y. (1998), “Semiactive control of Base Isola
tion using Feedforward information of Disturbance,” Proceedings of Second World
Conference on Structural Control, Kyoto, John Wiley and Sons, 377386.
[127] Yu, J., Wakahara, T. and Reed, D.A. (1999), “A Nonlinear Numerical Model of the
Tuned Liquid Damper,” Earthquake Engng and Structural Dynamics, 28, 671686.
193
APPENDIX
tions with rational power spectra, the integrals are of the following form,
∞
Ξ n ( ω ) dω
In ≡ ∫ Λ
n ( – iω )Λ n ( iω )
 (A. 1)
–∞
where
Ξn ( ω ) = χn – 1 ω + χn – 2 ω + … + χ 0 and
2n – 2 2n – 4
n n–1
Λ n ( iω ) = λ n ( iω ) + λ n – 1 ( iω ) + … + λ0
This integral can be written in a matrix form as (Roberts and Spanos, 1990),
χm – 1 χm – 2 … … … χ0
–λm λm – 2 –λm – 4 λm – 6 … …
0 –λm – 1 λm – 3 –λm – 5 … …
… 0 … … … …
π 0 0 … … –λ2 λ0
I n =   (A. 2)
λn
λm – 1 –λm – 3 λm – 5 –λm – 7 … …
–λm λm – 2 –λm – 4 λm – 6 … …
0 –λm – 1 λm – 3 –λm – 5 … …
… 0 … … … …
0 0 … … –λ2 λ0
181
A.2 Building and Excitation Parameters (Example 4 in Chapter 5)
2000 – 1000 0 0 0
– 1000 4800 – 1400 0 0
4.5
K=  0 – 1400 6000 – 1600 0 kN/m
0.0254
0 0 – 1600 6600 – 1700
0 0 0 – 1700 7400
Most valve suppliers provide a different measure of ﬂow characteristic than the
headloss coefﬁcient (ξ) used thoroughout this dissertation. The commonly used measure is
the valve conductance which is deﬁned as the mass ﬂow of liquid through the valve, given
by,
Q = C V ρ ( ∆p ) (A. 3)
where Q is the mass ﬂow (Kg/s); CV is the valve conductance (m2); ρ is the speciﬁc den
sity of the liquid (Kg/m3); ∆p is the pressure drop across the valve (Pa).The valve conduc
tance is usually supplied in British rather than S.I. units. The parameter C̃ V in gall/min/
C V = 2.3837 × 10 C̃ V
–5
(A. 4)
182
A 1.5 inch ball valve has been used for the experimental study described in chapter 7. The
valve manufacturer provided the valve conductance values as a function of the valve open
ing angle (Fig. A.1 (a)). The headloss across a valve/oriﬁce can be written as,
ρξV
2
∆p =  (A. 5)
2
Equation A.5 can be rewritten as follows:
2
Q
∆p = 2 (A. 6)
ρC V
πρD
2
Q = ρAV =  V (A. 7)
4
Comparing Eqs. A.3 and A.7, we obtain:
π D
2 4
ξ = 
2
 (A. 8)
8C V
Equation A.8 has been plotted for the 1.5 inch ball valve as a function of the angle of valve
opening.
90 50
θ = 0 deg
80 45
40
70
C values (gal/min/psi
Headloss Coefficient
35
60
1/2
)
30
ξ= f (θ )
50
25
40
θ = 25 deg 20
v
30
15
20
10
10 5
θ = 90 deg
0 0
0 20 40 60 80 100 0 20 40 60
Φ
Angle of valve opening, θ θΦ
Angle of valve opening,
Figure A.1 (a) Variation of Valve Conductance (b) Variation of headloss coefﬁcient
with the angle of valve opening
183
Bien plus que des documents.
Découvrez tout ce que Scribd a à offrir, dont les livres et les livres audio des principaux éditeurs.
Annulez à tout moment.