Abstract
The purpose of this work is to deepen some of the features related to the
application of Liquid Dampers for vibration control of engineering structures.
To this aim, both theoretical, numerical and experimental questions are dealt
with, finally giving an overview of in progress studies by the authors in the
specific research field. In details, some methodologies applied for analysis of
the seismic behaviour of existing bottom supported storage tanks, under
predominantly horizontal seismic actions are presented. Actually the full
understanding of the device behaviour and fluid-container interactions is
necessary for reliably analyzing and designing control devices based on liquid
dissipation. Moreover the paper reports some results obtained by means of an
experimental investigation developed on rigid blocks moving on a foundation
base subjected to a horizontal forcing function and equipped with some
rudimental Sloshing Water devices. Experimental data derived by means of a
shaking table facility are devoted to compare the dynamic response of the
blocks equipped or not with the devices for various liquid levels of the tanks,
in order to produce a first check of the possible effectiveness of such devices
on rigid structural models exhibiting a non-linear behaviour under dynamic
shaking.
Introduction
The risk of occurrence of severe structural damage or failures during a catastrophic
event (eg. Earthquakes and hurricanes) can be reduced by adopting vibration control
2 Rui Carneiro de Barros and Ottavia Corbi
possible benefits deriving from the application of liquid devices for attenuating the
dynamic response of non-linear structural models with unilateral constraints [7-9].
To this regard, one should notice that the need of predicting and preventing
failures associated to rocking and overturning of rigid structures undergoing strong
ground shaking have motivated a consistent number of studies on rocking response of
rigid blocks [19]-[21]; this is basically due to two main reasons, e.g., on one side, a
wide variety of structures can be collected in this model class, and, on the other side,
dynamics of rigid blocks are pretty complex due to their intrinsic non-linearity, and
can be considered a still open research topic.
One should also consider that very few studies have been developed, to date,
addressing the problem of coupling of TLDs to rigid structures, and the evaluation of
their potential effectiveness.
(a) (b)
Solution of the BVP of the Liquid and FEM Modeling of Tank Shell
The first developed FEM approach for predicting the seismic response of storage
tanks is based on assuming small liquid displacements. A rigorous mathematical
treatment involves deriving an expression for the harmonic fluid velocity potential
function satisfying Laplaces equation of the Boundary Value Problem (BVP)
expressed in cylindrical coordinates, and also satisfying appropriate boundary
conditions at the rigid tank bottom, at the liquid interface with the elastic thin shell,
and at the free surface.
The forced vibration of the tank-fluid model is governed by the following
variational formulation of Hamiltons principle
t2
I = (Tt + Tf U t + W ) dt = 0
t1
(1)
where Tt is the kinetic energy of the tank wall, Tf is the kinetic energy of the
contained liquid, Ut is the potential strain energy stored in the tank wall, and W is the
work done by external loads. (When W=0, the variational principle governs free
vibration). The velocity potential function is used to obtain the kinetic energy function
Tf of the liquid.
Due to the axial-symmetry of the tank fluid-system, once the surface loadings (qz,
q, qr) are represented in series of harmonic components
An Overview on Some Ongoing Computational 5
(m)
qz = qz (z, t ) cos m
m =0
(m)
q = q (z, t ) sin m
m =0
(m)
q r = q r (z, t ) cos m
m =0 (2)
then the shell displacements (u,v,w) would be harmonic of similar forms, in such a
way that the mth harmonic component of the displacements is coupled only to the mth
harmonic of the loading functions.
But because fluid is assumed inviscid, for all the harmonics it is found that
(m) (m)
q z ( z, t ) = q ( z, t ) = 0
. The fluid velocity potential satisfying Laplaces equation
and boundary conditions is also harmonic, and the mth harmonic of radial loading is
(m)
qr (z, t ) = f m (R , z, t )
expressed in terms of fluid mass density by .
a (t )
For the tank subjected to any horizontal earthquake of ground acceleration g
through the rigid bottom plate, Fig.2 justifies trigonometrically the existence of the
indicated components in the radial and circumferential directions.
Then, only the first (m=1) harmonic of tank is excited, because the component
cos is already the trigonometric expansion of itself. Also only the first harmonic
term of the hydrodynamic velocity potential will be different from zero. The acting
hydrodynamic pressures on the tank walls are then p = f (r = R , t ) cos .
Uncoupling the BVP by separation of variables permits to express the potential in
a series of contributions or terms of Bessel function J1, as
6 Rui Carneiro de Barros and Ottavia Corbi
r 2R
= J 1 ( n )[ 2 + A n (z, t )]
w
n R ( n 1) J1 ( n )
(3)
where the n are the roots of the Bessel function J1 of first type and first order, and
A n (z, t ) are the contributions of the interaction between the fluid and the flexible tank
wall.
This velocity potential function may be decomposed into two parts (impulsive
and convective) that permit to describe the corresponding effects on the tanks and
their performances [14].
The impulsive response is due to the impulsive pressures associated with inertia
forces produced by impulsive accelerations of tank wall; while the convective
response is due to convective pressures produced by oscillation (sloshing) of the
liquid. Fig.3 visualizes these two types of tank responses, induced by earthquake
horizontal ground motions.
The tank wall is considered as a linear, elastic thin shell, in which the three-
dimensional generalized displacement components (u,v,w) are related to the radial,
circumferential and axial coordinates (r, ,z) of a point on the shell middle surface.
Neglecting rotational inertia, the equilibrium equations of Sanders shell theory have
already been applied [13,14], such that standard kinematics relations for membrane
strains and curvatures in flexure and torsion, expressed in terms of generalized
N , N , N z
displacement components (u,v,w), are related to the membrane actions z
M z , M , M z
and moments , as represented in Fig.4. The finite element model is
obtained by assuming for the tank wall and the contained liquid a mesh of ring-shaped
finite elements.
An Overview on Some Ongoing Computational 7
N z , N , N z M z , M , M z
Figure 4 : Membrane actions and moments in Sanders
theory.
The hydrodynamic actions on the shell tank wall depend on the response of the
shell and liquid contents to the seismic motions of the foundation soil. By formulating
the fluid velocity potential in terms of nodal displacements, the hydrodynamic
pressures, effective generalized forces, consistent mass and stiffness matrices will also
be expressed in terms of the full set of nodal displacements.
The finite element model is obtained dividing the tank wall and the contained
liquid into ring-shaped finite elements of high b, represented in Fig.5.
Figure 5 : Finite element numbering and local coordinates for individual element.
Also, the vector of amplitudes of mth displacement harmonic for finite element i is
expressed in terms of the generalized coordinates
{}(mi) by
8 Rui Carneiro de Barros and Ottavia Corbi
1 s 0 0 0 0 0 0
{A} (i )
m = D {} (i )
m = 0 0 1 s 0 0 0 0 {}(mi )
0 0 0 0 1 s s 2 s3
(5)
based on assuming that the displacements u and v are linear (in s) while the radial
displacements w(s) are a polynomial of the 3rd degree [17].
The vector of generalized nodal displacements, at bottom (s=0) and top (s=b)
element ends, is expressed in terms of the generalized coordinates by the following
expression in FEM notation for geometric matrix B, detailed in [13]:
U m (s = 0)
V (s = 0)
m
Wm (s = 0)
Wm , s (s = 0)
{X}(mi) = = B {}m
(i )
U
m (s = b )
Vm (s = b)
Wm (s = b)
W (s = b)
m,s (6)
with which the generalized strains and stresses can be determined by standard
FEM notation.
In terms of the strains-displacements matrix C and material matrix E, the (8x8)
stiffness matrix of the mth harmonic of generalized displacements of ith element is
expressed by:
[ ] [ ]
2 b
T
[K ](mi) = 1 B 1 [C m ]T [E ] [C]m ds d B 1
0 0 (8)
0
s
2
s
3
s (9)
An Overview on Some Ongoing Computational 9
since only the first harmonic of the hydrodynamic solution of the velocity potential
is non-null. In Eq.(9) the time derivative of the velocity potential is expressed by a
laborious series development in terms of the Bessel function of the first kind and first
order and other terms, in accordance with the already mentioned Eq.(3). However this
series development is conveniently truncated after the N most significant terms
(usually 5-8 terms suffices, with a truncation relative error less than 0.1%).
Also the consistent mass matrix of the tank ith element is
[ ] T
[ ]
b
[M ](i) = s B1 [D]T [D] ds B1
0 (10)
and the hydrodynamic load vector on the tank shell is expressed by the first N
terms truncated from the complete series of Bessel functions by
{Ff } = f [B 1 ] T J 1 ( n ) K 1( n ) {1 }( n )
n (11)
( n ) { }
where the hydrodynamic quantities K 1 and the vector 1 (n ) assume quite
cumbersome analytical expressions.
Assembling the finite elements contributions into the global system of equations
modeling the coupled fluid-tank shell behavior (neglecting damping), one gets
{
[K ]{V} = [M ]{V } [M f ]{V }+ {Ff } + Fs , ext } (12)
where K and M are respectively the global stiffness and consistent mass matrices
of the (metallic) tank shell wall. Mf is an effective (or virtual) mass matrix
corresponding to the added mass contributions of the tank-contained liquid due to the
presence of the fluid near the tank walls (with non-null terms associated with the
radial displacements of the generalized relative displacements vector V), and Fs , ext is
the vector of external applied loads on the tank walls.
The differential matrix equation of motion Eq.(12) is conveniently solved applying
the Runge-Kutta numerical integration method of 4th order. A computer program
(STTKEQ2D.for) developed in [13] models the above-mentioned formulation; the
application of such software is successfull for the steel tank represented in Fig.6 filled
with water and subjected to the so-called Veletsoss pulse, and also for the model
subjected to artificial synthetic earthquakes, generated under specific spectrum for the
space and temporal evolution of the instantaneous spectral density Sp(f).
0 .0 0 1 2
0 .0 0 0 8
0 .0 0 0 4
u [m ]
0 .0 0 0 0 v [m ]
-1 .4 0 0 .6 0 2 .6 0 4 .6 0 6 .6 0 8 .6 0 1 0 .6 0 1 2 .6 0 w [m ]
-0 .0 0 0 4 Delta t -
Delt a t +
-0 .0 0 0 8 Filme
-0 .0 0 1 2
z [m ]
(a) (b)
Figure 6 : Tank displacements response to Veletsoss pulse. (a) validating model and
(b) software.
10 Rui Carneiro de Barros and Ottavia Corbi
The expressions for the velocity potential and hydrodynamic pressures induced by
horizontal earthquakes, allows to evaluate the total pressures, stresses and
displacements and even free surface elevation throughout the tank.
As mentioned in the above the tank in Fig.6 was subjected to artificial synthetic
earthquakes, generated under specific spectrum for the space and temporal evolution
of the instantaneous spectral density Sp(f). The strong motion software SIMQKE-II
was used herein [18] to produce conditioned or unconditioned earthquake ground
motions, matching the target spectrum on average. Some tank responses to a generic
earthquake acceleration time-history calculated with the developed software
STTKEQ2D are represented in Figs 7,8 for certain time-steps of the response history.
The quantities represented are respectively: the instantaneous axial-shear-
membrane stresses, the instantaneous slope of the shell wall, the instantaneous surface
elevation, and the time-history of radial displacement at the level of the 4th ring finite
element.
2500
Inclinaes da Geratriz
2000 9.0E-03
1500
6.0E-03
1000
500 3.0E-03
Nz [kN/m]
0 Nq 0 [kN/m]
0 2 4 6 8 10 12 Nzq90[kN/m] 0.0E+00 dw/dz
0 2 4 6 8 10 12
-500
Delta t -
Delta t +
-1500
Filme
-6.0E-03
Filme
-2000
-9.0E-03
-2500
z [m]
z [m]
(a) (b)
0. 016
0.100
0.080 0. 012
0.060
0. 008
0.040
0. 004
0.020
Sobreel ev [ m ] 0 w [ m]
0.000
-20 -15 -10 -5 0 5 10 15 20 25 25. 5 26 26. 5 27 27. 5 28 28. 5 29 29.5 30
-0.020
Delta t + -0. 004
-0.040 D elt a t -
-0. 008
F ilm e
-0.060
-0. 012
-0.080
(a) (b)
With this essential information and with the expression for hydrodynamic
pressures, the envelopes of hydrodynamic pressure in the given reference tank are
obtained for seismic actions according to Portuguese seismic standards for soil type I
and zone A with higher seismic risk (A.S. 1-I ; A.S. 2-I), as represented in Fig.9.
Several other envelopes can be determined for different tank design variables through
the developed versatile software, very useful in ascertaining tank design under the
strength stiffness and stability viewpoints.
The significant envelope of axial stresses Nz is crucial in evaluating potential
onset of elephant-foot bulge buckling.
Figure 10 : Tank studied for earthquakes (AS-1; AS-2) and its finite element mesh
Some natural frequencies and mode shapes of the complete structure are presented
(Fig.11), as well as instantaneous deformed shapes (Fig.12.a) and envelopes of shell
stress resultants (Fig.12.b). The stress envelope can be used in conjunction with
critical buckling stress of symmetric instability of cylindrical shells in function of the
order of the harmonic, to ascertain the minimum design thickness requirements used
[15], by assessing the ultimate capacity by instability of the laminar structure of the
tank walls near the base.
(a) (b)
Figure 11 : Natural frequencies (a) 12=27.07 rad /s (b) 32=35.22 rad /s and mode
shapes of the complete tank.
An Overview on Some Ongoing Computational 13
(a) (b)
Figure 12 : (a) Deformed shape and sloshing at t=9.24s; (b) Envelope of vertical
compression of the shell along tank wall (kN/m)
(a) (b)
In order to obtain experimental data on blocks motion due to pure rocking, the
blocks are suitably fixed on the moving table and an accelerometer is placed on the
top of each block, measuring its accelerations. A special device is adopted for fixing
the blocks in such a way to avoid any rotation around zaxis and sliding along the
shaking table surface (Fig. 14.a,b).
Some prismatic aluminium blocks are suitably fixed on the moving table in such a
way to develop pure rocking (Fig. 14.c) in the direction of the table motion (no
allowed translation or torsion rotation on the table) and an is accelerometer placed on
the top of each block, for measuring blocks accelerations.
Figure 14 : Device for fixing the blocks on the shaking table: (a) plant and section,
(b) three-dimensional view. (c) Allowed motion of the blocks during experiments.
One basically refers to three prismatic blocks in aluminium with dimensions (cm):
i) 303040 cm, ii) 302040 cm, iii) 301040 cm, respectively (Fig.15).
An Overview on Some Ongoing Computational 15
For evaluating the energy dissipation due to the sloshing action of liquid in
containers, some liquid dampers (LDs) are mounted on the upper surface of the rigid
blocks.
The adopted LDs simply consist of some tanks with different shapes to be filled
with a liquid (in the specific case, water) at various levels.
In details one considers three tanks, one of which has a simply parallelepiped
shape (vertical lateral sides), as shown in Fig.16, and will be referred to as 0 tank.
The remaining two tanks have a trapezoidal shape with 30- or 45-inclined lateral
sides as shown in Fig.17; these will be referred to in the following as 30 and 45
tanks.
8cm
4cm
Figure 16 : Parallelepiped-shaped TLD with marks at two liquid levels (4cm and
8cm).
16 Rui Carneiro de Barros and Ottavia Corbi
8cm
4cm
8cm
4cm
(a) (b)
Figure 17 : Trapezoid-shaped TLDs with marks at two liquid levels (4cm and 8cm).
Inclination of the lateral sides of: (a) 30 , (b) 45.
Fig. 18 shows the resuming scheme of adopted tanks for LDs whilst Fig.19 reports
the considered water levels.
As mentioned in the above, the experiments are executed filling tanks with water
up to 4 and 8 cm of depth and comparing the results with the ones of the block with
empty tanks, which is referred to by the 0 cm notation (Fig.19).
In details the set of Figs 20 and 22 represents peak accelerations (m/s2) of the
blocks versus frequencies (of the unidirectional base acceleration) (Hz) for a span
level o 0.5 cm and for the mentioned (8cm, 4cm and 0 cm) liquid levels in the tanks:
in every graphic one has three curves relevant to the depth of the water (0, 4, 8 cm).
Figs 21 and 23 represents time histories at 5 Hz showing response attenuation when
increasing the liquid level. Comparison between the three liquid depth levels (Figs 20
and 22) reveal the potential efficiency of the liquid damper system in damping the
vibrations of the structural model. Looking at the results of the experimentation, some
basic features can be outlined.
First of all, the benefit by the liquid damper is not homogeneous on the frequency
range, but it is dependent on the frequencies at which the power of the excitation is
lumped. This is clear observing Fig.20 as well as other plots, where the benefit is
much higher in some frequency range.
18 Rui Carneiro de Barros and Ottavia Corbi
120 120
100 100
acceleration - m/s2
acceleration - m/s2
80 80
0 cm 0 cm
60 4 cm 60 4 cm
8 cm 8 cm
40 40
20 20
0 0
0 2 4 6 8 10 0 2 4 6 8 10
frequency - Hz frequency - Hz
(a) (b)
Figure 20 : Block 303040 cm with: a) 0 tank and b) 45 tanks for different water
levels.
Aluminium 30x30 - 0 tank - water 0 cm - span 5mm - 5Hz Aluminium 30x30 - 45 tank - water 0 cm - span 5mm - 5Hz
80 80
60 60
40 40
2
acceleration - m/sec 2
acceleration - m/sec
20 20
0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
-20 -20
-40 -40
-60 -60
-80 -80
time sec time sec
Aluminium 30x30 - 0 tank - water 4 cm - span 5mm - 5Hz A luminium 3 0 x3 0 - 4 5 t ank - wat er 4 cm - sp an 5mm
- 5Hz
80
60
80
40 60
2
acceleration - m/sec
40
20
20
0
0
0 1 2 3 4 5 6
-20 -20 0 1 2 3 4 5 6
-40
-40
-60
-60
-80
-80 t i m e se c
time sec
Aluminium 30x30 - 0 tank - water 8 cm - span 5mm - 5Hz Aluminium 30x30 - 45 tank - water 8 cm - span 5mm - 5Hz
80 80
60 60
40 40
2
2
acceleration - m/sec
acceleration - m/sec
20 20
0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
-20 -20
-40 -40
-60 -60
-80 -80
time sec time sec
(a) (b)
90 90
80 80
acceleration - m/s2
70
acceleration - m/s2
70
60 60
0 cm 0 cm
50 50
4 cm 4 cm
40 40
8 cm 8 cm
30 30
20 20
10 10
0 0
0 2 4 6 8 10 0 2 4 6 8 10
frequency - Hz frequency - Hz
(a) (b)
Alum inium 30x20 - 0 tank - w ater 0 cm - Alum inium 30x20 - 45 tank - w ater 0 cm -
span 5m m - 5Hz span 5m m - 5Hz
60 80
60
40
40
20
20
0 0
0 1 2 3 4 5 6 -20 0 1 2 3 4 5 6
-20
-40
-40
-60
-60 -80
t i m e se c t i m e se c
Alum inium 30x20 - 0 tank - w ater 4 cm - Alum inium 30x20 - 45 tank - w ater 4 cm -
span 5m m - 5Hz span 5m m - 5Hz
80 80
60 60
40 40
20 20
0 0
-20 0 1 2 3 4 5 6 -20 0 1 2 3 4 5 6
-40 -40
-60 -60
-80 -80
t i m e se c t i m e se c
Alum inium 30x20 - 0 tank - w ater 8 cm - Alum inium 30x20 - 45 tank - w ater 8 cm -
span 5m m - 5Hz span 5m m - 5Hz
80 80
60 60
40 40
20 20
0 0
-20 0 1 2 3 4 5 6 -20 0 1 2 3 4 5 6
-40 -40
-60 -60
-80 -80
t i m e se c t i m e se c
(a) (b)
and the shape of the tank) is determinant for the performance of the device. In all the
tests the benefit potentially appears very significant but some cases have been
encountered where the device increases the risk. The conclusion is that a theoretical
settlement cannot be ignored and the device has to be carefully designed with regard
to shape and entity of the sloshing liquid mass, keeping into account both the
excitation's character and the object's inertial properties.
Acknowledgment
This work, as part of the European Science Foundation EUROCORES Program
COVICOCEPAD, was supported by funds from the FEUP, the Italian CNR and the
EC Sixth Framework Program.
Conclusions
The paper focuses on the possibility of exploiting liquid sloshing motion for vibration
control purposes of civil structures. Both analytical, numerical and experimental
issues are introduced, giving a general overview of some in progress studies
developed by the authors on the specific topic.
As regards to liquid motion modeling at the inner of the tanks, two FEM approaches
are introduced for the numerical analysis of seismic response of liquid-filled tanks and
to determine design envelopes, through successfully validated software. The
investigation allows to find out the responses of 10-14 artificial earthquakes,
satisfying frequency requirements of the spectral density at the site, are required in
order to determine accurate design envelopes of bottom supported tanks. Moreover,
from an experimental point of view, coupling of liquid devices prototypes with rigid
structural models exhibiting non-linear behavior under dynamic motion is
investigated; although an high number of studies have been produced in the research
area, this special application has been rarely treated in literature, even if it can be
considered of basic importance because it covers a wide variety of structural
typologies.
Experimental results allow to appreciate potential effectiveness and reliability of
liquid devices also for the specific case.
References
[1] Kosaka H, Noji T, Yoshida H, Tatsumi E, Yamanaka H, Agrawal AK,
Damping Effects of a Vibration Control Damper Using Sloshing of Water,
Proc. Tenth World Conf. on Earthquake Engineering; Madrid; 4: 2435-2440,
1992.
[2] Fujino Y, Tuned Liquid Damper (TLD) for Suppressing Horizontal Motion of
Structures, Journal of Engineering Mechanics; 118, n. 10, 2017-2030, 1992.
[3] Kareem A, Next Generation of Tuned Liquid Dampers, Proc. First World
Conf. on Structural Control: International Association for Structural Control;
An Overview on Some Ongoing Computational 21