You are on page 1of 13

Performance of Pendulum Tuned Mass Dampers in

Reducing the Responses of Flexible Structures


A. J. Roffel 1; S. Narasimhan, M.ASCE 2; and T. Haskett 3

Abstract: The primary objective of this paper is to study the dynamic responses of a flexible multiple degree-of-freedom structure coupled
Downloaded from ascelibrary.org by UNB - Universidade de Bras?lia on 02/18/20. Copyright ASCE. For personal use only; all rights reserved.

with the three-dimensional motions of a nonlinear pendulum tuned mass damper (PTMD). The three-dimensional motions consist of both
planar and spherical motions of the PTMD. The optimal damper parameters obtained by using the three-dimensional model are compared
with those predicted by models in which PTMD motion is linearized to the planar direction. The effect of and sensitivity to frequency and
auxiliary damping detuning are considered for various levels of primary to auxiliary mass ratios. The performance of the PTMD is evaluated
by using a finite element representation of an actual tower structure equipped with a PTMD, together with the responses obtained by using the
high frequency base balance method from a boundary layer wind tunnel. The results are compared for linearized and three-dimensional
PTMDs and the effect of directional coupling introduced by the nonlinear three-dimensional PTMD on the response estimates is studied
by using the numerical model. Finally, a procedure is presented for conducting a condition assessment of existing PTMDs. DOI: 10.1061/
(ASCE)ST.1943-541X.0000797. © 2013 American Society of Civil Engineers.
Author keywords: Tuned mass dampers; Pendulum dynamics; Wind engineering; Passive vibration control; Structural vibration control;
Parametric study; Structural control.

Introduction investigation that the authors have recently conducted on a structure


located in Canada.
Tuned mass dampers (TMDs) are commonly used to attenuate The study of TMDs in the literature has primarily focused on
vibrations in flexible structures (Abe and Fujino 1994; Kareem two key aspects: (1) to determine the optimal damper parameters,
and Kijewski 1999; Kwok and Samali 1995). TMDs consist of a i.e., the frequency and damping of the TMD (den Hartog 1956; Ioi
relatively small auxiliary mass, a stiffness element, and a damper to and Ikeda 1978; Thompson 1981; Abe and Fujino 1994; Rana and
dissipate the mechanical energy as heat. Pendulum tuned mass Soong 1998; Lee et al. 2006; Bakre and Jangid 2007; Ghosh and
dampers (PTMDs) use a suspended mass instead of a sprung mass, Basu 2007) and to estimate the resulting performance; and (2) to
with a damping element to dissipate energy. When optimally tuned enhance the efficiency and robustness of passive TMDs by adding
to the resonant mode of interest, a TMD is effective in reducing the adaptive stiffness and damping elements adjusted through feedback
dynamic response of the structure. The performance aspects of control mechanisms (Nagarajaiah and Varadarajan 2005; Roffel
TMDs are well understood (Sacks and Swallow 1993; Kwok and et al. 2011). The focus of this paper is on the first aspect. A wide-
Samali 1995; Kwok and Macdonald 1990; Kareem and Kijewski spread approach of using linear or linearized models to estimate the
1999), and optimal design equations for TMDs are readily available parameters of a TMD and their effectiveness has resulted in a large
(den Hartog 1956; Warburton and Ayorinde 1980; Rana and Soong database of information. However, systematic comparisons of
1998; Gerges and Vickery 2005; Bakre and Jangid 2007). Despite a simple and linear models for more complex and realistic models,
vast amount of literature and widespread full-scale applications of specifically for the case of pendulum TMDs, which exhibit signifi-
TMDs, PTMDs have received relatively less attention (Gerges and cant nonlinear and three-dimensional behavior, are not readily
Vickery 2005; Roffel et al. 2011). Concerns related to their mod- available in the literature.
eling, simulation, determination of optimal design parameters, and Previous studies of pendulum-type tuned mass dampers have
response analysis have not been systematically explored. This considered linearized planar pendulum models coupled with a
paper aims to study these aspects and to present results from an single degree-of-freedom (DOF) primary system, described by us-
ing modal coordinates to find the optimal parameters and to predict
responses (Gerges and Vickery 2005). Such models are adequate,
1
Staff II—Structures, Simpson Gumpertz and Heger, Inc., Waltham, for example, in analyzing the response of flexible structures in
MA; formerly, Graduate Research Assistant, Dept. of Civil and which the dominant response is believed to be primarily in one di-
Environmental Engineering, Univ. of Waterloo, Waterloo, ON, Canada rection. When the responses in both along-wind and across-wind
N2L 3G1. directions are of concern, it is important to quantify the effect of the
2
Associate Professor, Dept. of Civil and Environmental Engineering, PTMD in both directions. It will be shown that both the linearized
Univ. of Waterloo, 200 Univ. Ave. West, Waterloo, ON, Canada N2L model and a three-dimensional model produce similar results for
3G1 (corresponding author). E-mail: snarasim@uwaterloo.ca the optimal auxiliary parameters; however, the linearized planar
3
Technical Director, Rowan Williams Davies and Irwin, Inc., Guelph,
model may sometimes lead to unconservative estimates in response
ON, Canada N1K 1B8.
Note. This manuscript was submitted on July 14, 2011; approved on predictions.
December 27, 2012; published online on December 29, 2012. Discussion The performance of TMDs is generally assessed using paramet-
period open until February 13, 2014; separate discussions must be sub- ric studies of numerical models, which are excited by using
mitted for individual papers. This paper is part of the Journal of Structural harmonic inputs (Rana and Soong 1998; Pinkaew and
Engineering, © ASCE, ISSN 0733-9445/04013019(13)/$25.00. Fujino 2001), earthquake time histories (Rana and Soong 1998;

© ASCE 04013019-1 J. Struct. Eng.

J. Struct. Eng., 2013, 139(12): 04013019


Lee et al. 2006), or broadband, filtered, or band-limited white noise linearized unidirectional models. This investigation is conducted
(Bakre and Jangid 2007; Ghosh and Basu 2007). The performance by using model tests on an airport control tower located in Toronto,
of PTMDs is generally overestimated when using the aforemen- Canada, combined with a high-fidelity numerical model of the
tioned conventional approaches rather than realistic wind loads same structure.
measured from a wind tunnel study (Xu and Kwok 1994). Studies
predicting the response of structures equipped with TMDs by using
synthetic wind time histories (Xu et al. 1992; Ricciardelli et al. Dynamic Response of a Flexible Structure Coupled
2000), or by directly measuring the response of TMD-equipped with a PTMD
structures through wind tunnel studies (Tanaka and Mak 1983;
Xu and Kwok 1994), are limited in the literature. The proposed A system of equations that couples the dynamics of the flexible
approach aims at quantifying the performance of the TMD by using structure with the motion of the PTMD is developed first. Civil
time history loading information from wind tunnel studies, engineering structures are complex, and the use of overly simplified
Downloaded from ascelibrary.org by UNB - Universidade de Bras?lia on 02/18/20. Copyright ASCE. For personal use only; all rights reserved.

combined with the popular high-frequency base balance (HFBB) systems to represent their dynamic behavior may lead to significant
method. errors in the estimation of their response. Furthermore, the coupled
This work is proposed within the context of vibration nature of the nonlinear equations of motion of the PTMD impacts
occupancy comfort criteria for flexible buildings. A recent paper the responses of the structure in both lateral directions. Therefore,
on this topic (Kwok et al. 2009) reviewed an extensive database of the ensuing study seeks to model the dynamic response of a multi-
studies aimed at understanding and unifying comfort guidelines, ple degree of freedom (MDOF) structure in three dimensions,
and presented an excellent overview of current standards in use equipped with a PTMD in the time domain, by using Lagrangian
today. It is generally accepted that within the lower frequency mechanics.
range less than 1 Hz, human perception threshold reduces with The origin of the system is set up to coincide with the initial
increasing frequency. For structures with well separated lateral suspension point of the pendulum mass. The vectors u, v, and
frequencies, a PTMD designed to satisfy comfort performance w are the displacements of the suspension point in the x, y, and
requirements in one direction may have a significant impact in z directions, respectively. The angle θ is the angle of swing away
the other direction. As a consequence, it may not meet the from the vertical line passing through the origin, also known as the
overall performance objectives in a given situation. Hence, it is planar angle. The angle φ is the angle of the auxiliary mass rotating
critical that the response for structures equipped with PTMDs about the vertical line, also known as the spherical angle; φ ¼ 0
be estimated with a good degree of accuracy to ascertain its corresponds to the positive x direction. All of the aforementioned
overall performance. parameters vary with time. La is the length of the pendulum and ma
With this background, the primary objectives of this paper are is the auxiliary pendulum mass. Fig. 1 shows a schematic geometry
summarized as follows. First, the optimal parameters are deter- of the pendulum mass.
mined for a pendulum TMD by simulating the nonlinear three- A linear auxiliary viscous damper and linear spring are intro-
dimensional motions of the damper connected to a flexible duced and fixed along the suspension length of the pendulum
structure. The parameters are compared with existing values from and to the same DOF of the suspension point. The viscous damper
the literature. This is undertaken to shed new light on the impact of has damping coefficients cx in the x direction and cy in the y di-
simplified modeling assumptions. Next, the performance of the rection. The linear spring has spring constants kx in the x direction
PTMD when optimally tuned is evaluated by using wind tunnel and ky in the y direction. The damper and spring are connected to
measurement data combined with the HFBB method. The impact the pendulum length at distances hx and hy from the suspension
of simplifying assumptions on the response estimates is studied point in the x and y directions, respectively. It is assumed that
in detail. Comparisons are made to the response predictions by the auxiliary spring and damper remain horizontal.

Fig. 1. Schematic geometry of the PTMD mass with auxiliary damper and spring: (a) isometric view; (b) x direction; (c) y direction

© ASCE 04013019-2 J. Struct. Eng.

J. Struct. Eng., 2013, 139(12): 04013019


The kinetic energy of the auxiliary mass is where Δ _ ¼ f u_ v_ w_ Δ _ r gT = velocity vector of the primary
system; u,_ v,
_ and w_ = nodal velocities of the suspension point;
1 _ r = velocities of the remainder of the DOFs for the primary
T a ¼ ma vvT and Δ
2 structure. Similarly, the subscript r in the global mass matrix
1
¼ ma ½ðu_ þ La cos θ cos ϕθ_ − La sin θ sin ϕϕÞ
_ 2 represents the remainder of the rows or columns of the mass matrix.
2 The potential (strain) energy of the primary system is
þ ðv_ þ La cos θ sin ϕθ_ þ La sin θ cos ϕϕÞ
_ 2 8 9T 2 38 9
>
> u >
> kuu kuv kuw Kur > > u >>
_ 2
þ ðw_ þ La sin θθÞ 1 T 1 < = 6 kvu kvv kvw Kvr 7< v =
v 6 7
V m ¼ Δ KΔ ¼
1 2 2>> w > 4 kwu kwv kww Kwr 5> w>
_ a cos θ cos ϕθ_
¼ ma ½u_ 2 þ v_ 2 þ w_ 2 þ L2a θ_ 2 þ L2a ϕ_ 2 sin2 θ þ 2uL : > ; : >
> ;
2 Δr Kru Krv Krw Krr Δr
_ a sin θ sin ϕϕ_ þ 2vL
− 2uL _ a cos θ sin ϕθ_ þ 2_vLa sin θ cos ϕϕ_ ð5Þ
Downloaded from ascelibrary.org by UNB - Universidade de Bras?lia on 02/18/20. Copyright ASCE. For personal use only; all rights reserved.

þ 2La w_ θ_ sin θ ð1Þ where Δ ¼ f u v w Δr gT = displacement vector of the pri-


mary system; u, v, and w = nodal displacements of the suspension
The potential (strain) energy of the auxiliary mass is point; and Δr = displacements of the remainder of the DOFs for
1 1 the primary structure. Similarly, the subscript r in the global stiff-
V a ¼ kx r2p;x þ ky r2p;y þ ma gðw − La cos θÞ ness matrix represents the remainders of the rows or columns of the
2 2
stiffness matrix.
1 1
¼ kx hx sin θcos2 φ þ ky h2y sin2 θsin2 φ þ ma gðw − La cos θÞ
2 2 The Raleigh dissipation function for the primary mass is
2 2 8 9T 2 38 9
ð2Þ >
> u_ >
> cuu cuv cuw Cur > > u_ >>
1 _T _ <
1 v_ = 6 cvu cvv cvw Cvr 7< v_ =
F m ¼ Δ CΔ ¼ 6 7
A Raleigh dissipation function for the auxiliary system is intro- 2 2>
> w_ > 4 cwu cwv cww Cwr 5> w_ >
:_ > ; :_ >
> ;
duced. The auxiliary dissipation factor is Δr Cru Crv Crw Crr Δr
1 1 ð6Þ
F a ¼ cx v2p;x þ cy v2p;y
2 2
1
¼ cx ðh2x cos2 θcos2 φθ_ 2 − 2h2x cos θ cos φθ_ sin θ sin φφ_ Lagrange’s Equation
2
Lagrange’s equation (Meirovitch 1970), including the Raleigh dis-
þ h2x sin2 θsin2 φφ_ 2 Þ sipation function and the generalized forces, is
1
þ cy ðh2y cos2 θsin2 φθ_ 2 þ 2h2y cos θ cos φθ_ sin θ sin φφ_ d ∂T ∂T ∂V ∂F
2 − þ þ ¼ Qr ð7Þ
dt ∂ q_ r ∂qr ∂qr ∂ q_ r
þ h2y sin2 θcos2 φφ_ 2 Þ ð3Þ
where T , V, and F = sum of the primary and auxiliary kinetic
Consider the flexible primary MDOF structure with three trans- energy, potential energy, and Raleigh dissipation functions,
lational degrees of freedom in each orthogonal direction for each respectively; and the variables qr and q_ r = generalized coordinates
floor mass. Mn×n is the global mass matrix, Cn×n is the global and velocities of the system. For the combined primary struc-
damping matrix, and Kn×n is the global stiffness matrix, where ture and PTMD system, the general coordinates are u, v, w,
n is the number of DOFs. The upper left 3 × 3 matrix of the ma- θ, φ, and Δr and the generalized velocities are u, _ v_ , w, _ φ,
_ θ, _
trices corresponds to the DOF from which the PTMD is suspended. and Δ _ r . For the MDOF system, the generalized forces are
This is typically the top floor, although the equations can be modi- Q ¼ ½ Pu Pv Pw Pr 0 0 , where Pu , Pv , and Pw are the
fied to account for a PTMD suspended at a different location. arbitrary forces applied to the primary structure at the location
The kinetic energy of the primary system is of the suspended mass in the x, y, and z directions, respectively;
1 _T _ Pr is the arbitrary force excitation for the remainder of the
Tm¼ Δ MΔ DOF of the primary structure. Substituting the relevant quan-
2
8 9T 2 38 9 tities into Eq. (7) produces the differential equations of mo-
> u_ >
> > muu muv muw Mur > u_ >
< >
>
1 < v_ = 6 = tion for the primary MDOF structure coupled with the PTMD
6 mvu mvv mvw Mvr 7
7 v_
¼ 4 mwu 5 ð4Þ dynamics.
2>> w_ > mwv mww Mwr > w_ >
:_ > ; :_ >
> ; The equations of motion corresponding to the translational
Δr Mru Mrv Mrw Mrr Δr DOFs of the primary MDOF structure for forced vibration are

8 2 398 9 8 9 8 9
>
> ma 0 0 0 > > >
> ü >
> >
> u_ >> >
> u >
< 6 0 ma 0 0 7 = < = < = < > =
v̈ v_ v
Mþ6 4 0
7 þ C þ K ¼ ma La
>
> 0 ma 0 5> > > ẅ > > w_ > > w>
: ;> : > ; >
:_ > ; >
: > ;
0 0 0 0 Δ̈r Δr Δr
8 _ 2 _ 9 8 9 ð8Þ
> − cos θ cos φθ̈ þ sin θ cos φθ þ 2 cos θ sin φθ φ_ þ sin θ sin φφ̈ þ sin θ cos φφ_ 2 > > Pu >
>
< >
= <P >
> =
− cos θ sin φθ̈ þ sin θ sin φθ_ 2 − 2 cos θ cos φθ_ φ_ − sin θ cos φφ̈ þ sin θ sin φφ_ 2 v
× þ
>
>
: − sin θθ̈ − cos θθ_ − La
2 g >
; >
> > Pw >
: >
;
0 Pr

© ASCE 04013019-3 J. Struct. Eng.

J. Struct. Eng., 2013, 139(12): 04013019


where M, C, and K are described in Eqs. (4)–(6). The equation 1
jH j ðωÞj2 ¼   ω 2
2 2ωζ j 2 ð13Þ
of motion corresponding to the planar (θ-DOF) motion of the 1 − ωj þ ωj
pendulum is

La θ̈ − La sin θ cos θφ_ 2 þ cos θ cos φü þ cos θ sin φv̈ þ sin θẅ H j ðωÞ = mechanical admittance function; ζ j = modal damping;
K j = generalized stiffness; and Swj ðωÞ = power spectral density of
kx h2x ky h2y the generalized wind forces.
þ g sin θ þ sin θ cos θcos2 φ þ sin θ cos θsin2 φ
ma La ma La The accuracy of the HFBB method depends on the effect of
cx h2x higher order modes. The HFBB method assumes that the general-
þ ðcos2 θcos2 φθ_ − cos θ cos φ sin θ sin φφÞ
_ ized forces are proportional to the measured base moments for
ma La
structures with uncoupled mode shapes that are approximately
cy h2y
þ ðcos2 θsin2 φθ_ þ cos θ cos φ sin θ sin φφÞ
_ ¼0 ð9Þ linear. The response of the structure to the wind excitation is
ma La determined by solving the generalized form of the equations of
Downloaded from ascelibrary.org by UNB - Universidade de Bras?lia on 02/18/20. Copyright ASCE. For personal use only; all rights reserved.

motion. The primary advantage of the HFBB method is that,


The equation of motion corresponding to spherical (φ-DOF) under the simplifying assumptions of mode shape linearity
motion of the pendulum is and broadband stationary excitation, the response of the structure
can be solved in closed form by using random vibration theory.
La sin θφ̈ þ 2La cos θθ_ φ_ − sin φü þ cos φv̈ For nonlinear and coupled mode shapes, correction factors
kx h2x − ky h2y can be applied (Holmes 1987; Yip and Flay 1995; Zhou
− ðsin θ sin φ cos φÞ et al. 2002; Chen and Kareem 2005; Lam and Li 2009; Tse
ma La
et al. 2009). However, applying the modes that uncouple the
cx h2x
þ ð− cos θ cos φ sin φθ_ þ sin θsin2 φφÞ
_ primary system from the PTMD controlled system does not
ma La allow the overall system of equations to be solved in the fre-
cy h2y quency domain. Hence, the solution to the case of the structure
þ ðcos θ cos φ sin φθ_ þ sin θcos2 φφÞ
_ ¼0 ð10Þ equipped with a PTMD is pursued in the time domain, as
ma La
explained next.
Eqs. (8)–(10) are the integrated finite-element representation of The equations of motion for the flexible structure coupled
a MDOF flexible structure with coupled dynamics of a PTMD for with a planar-spherical PTMD [Eqs. (8)–(10)] are adapted in
forced vibration. If the motion of the primary structure is restricted the generalized form to include rotation about the vertical axis
to the x direction, and the PTMD begins from at-rest initial and translation in each horizontal direction. It is assumed that
conditions, then φ ¼ φ_ ¼ φ̈ ¼ 0, v ¼ v_ ¼ v̈ ¼ 0, and w ¼ w_ ¼ the mode shapes of the primary structure are not affected by
ẅ ¼ 0. For small angle displacements, sin θ and cos θ can be ap- the addition of the PTMD and that the mode shapes for the pri-
proximated by θ and 1, respectively. Neglecting the nonlinear term mary translational modes remain linear. To account for the
θθ_ 2 , the equations of motion reduce to coupling between the pendulum and the primary structure in
the physical domain, the modal responses for the primary
       
ma 0 ü u_ u structure are transformed back into the physical domain at
Mx þ þ Cx þ Kx each time step and the control force is calculated. The control
0 0 Δ̈r;x _ r;x
Δ Δr;x
force is transformed into modal coordinates and applied to the
 
Pu − ma La θ̈ structure, together with the generalized force measured by using
¼ ð11aÞ HFBB. The equations of motion in the modal domain for the jth
Pr;x
mode are

ma L2a θ̈ þ cx h2x θ_ þ ðma gLa þ kx h2x Þθ þ ma La ü ¼ 0 ð11bÞ M j ÿj ðtÞ þ Cj y_ j ðtÞ þ K j yj ðtÞ ¼ uj ðtÞ þ wj ðtÞ ð14Þ

where Mx , Cx , and Kx = mass, damping, and stiffness matrices where


when all except the terms corresponding to translation in the x 8 2 39
direction are removed. >
> ma 0 0 0 >>
< 6 0 =
6 ma 0 077 ϕj
M j ¼ ϕj M þ 4
T ð15Þ
>
> 0 0 0 0 5>
>
: ;
High Frequency Base Balance Method 0 0 0 0

The HFBB method is commonly adopted for predicting the re- The first three DOFs of the mass matrix correspond to the
sponse of structure to wind excitations. The premise of the method translation in the x and y directions and rotation about the
is that the base overturning and torsional moments measured on a z direction of the rigid diaphragm mass to which the PTMD is
lightweight rigid scale model in a wind tunnel can be used to es- fixed. The PTMD is assumed to be connected to the structure
timate the generalized forces exerted on the structure (Tschanz and by using a frictionless hinge that cannot transfer rotations about
Davenport 1983; Kareem 1992). Once the generalized wind forces the z axis. Similarly, Cj ¼ ϕTj Cϕj , where C is the proportional
are estimated from the measurements, the following expression is damping matrix of the primary system, and K j ¼ ϕTj Kϕj . The
used to calculate the SD of the modal response (for the jth mode) mass moment of inertia of the pendulum about its own axis is
(Tse et al. 2009) neglected.
Z 1=2 The generalized wind force, wj ðtÞ, is measured directly from the
1 ∞
2 HFBB model:
σj ¼ jHj ðωÞj Swj ðωÞdω ð12Þ
Kj 0
My ðtÞ M ðtÞ
wj ðtÞ ¼ X jx ϕj1 − X jy ϕj2 x þ X jzz ϕj3 Mz ðtÞ ð16Þ
where h h

© ASCE 04013019-4 J. Struct. Eng.

J. Struct. Eng., 2013, 139(12): 04013019


where M x , M y , and Mz = base moments about the x, y, and z axis, X jzz ¼ 0.7 (Tse et al. 2009) are introduced to account for nonideal
respectively; and h = height of the structure; the coefficients ϕj1 , mode shapes, specifically because the torsional mode shape is not
ϕj2 , and ϕj3 ¼ x, y, and θz components of the mode shape constant.
coefficients at the top of the building for the jth mode. The com- The control force exerted on the structure by the PTMD is trans-
monly used mode shape correction factors of X jx ¼ X jy ¼ 1 and formed into the modal domain by using

 T  
ϕ1j _2 __ _2
uj ðtÞ ¼ ma La − cos φθ̈ þ θ cos φ_θ2 þ 2 sin φ_θ φ þθ sin φφ̈ þ θ cos φφ2 ð17Þ
ϕ2j − sin φθ̈ þ θ sin φθ − 2 cos φθ φ_ −θ cos φφ̈ þ θ sin φφ_
Downloaded from ascelibrary.org by UNB - Universidade de Bras?lia on 02/18/20. Copyright ASCE. For personal use only; all rights reserved.

θ and φ and their first and second time derivatives are found by the RMS displacement response of the primary system by using
solving the corresponding equations of motion [Eqs. (9) and (10) closed-form solutions that are documented in the literature (Gerges
with ẅ ¼ 0; ü and v̈ are found by transforming the modal responses and Vickery 2005). The approach relies on the presence of fixed
back into physical coordinates at each time step with the following point frequencies, in which the transmissibility of the vibration
transformation: is independent of the auxiliary damping; for structures that exhibit
main mass damping, these frequencies no longer exist. Therefore,
X
N closed-form solutions are only possible for the special case of an
ü ¼ ϕj1 ÿ ð18aÞ undamped primary structure (Gerges and Vickery 2005; Rana and
j¼1
Soong 1998; Bakre and Jangid 2007), although close agreement
with numerical results for low to moderate main mass damping
X
N in structures has been demonstrated by approximating the fixed
v̈ ¼ ϕj2 ÿ ð18bÞ point frequencies (Ghosh and Basu 2007). A popular approach
j¼1 to determine the optimal parameters for a PTMD is to calculate
the theoretical values for equivalent linear systems and parameters
(that is, stiffness and damping). Subsequently, the equivalent pen-
These transformations are necessary because the calculation of dulum length is calculated. As will be shown in this paper, such
the control forces, according to Eq. (17), depends on the physical methods are satisfactory for the calculation of the design parame-
coordinates. At each time step, the following operations are con- ters; however, they may lead to unconservative estimates predicting
ducted. First, the primary structure dynamic analysis is conducted the response of the controlled system.
in the modal domain, because the generalized force is directly The second approach, which is necessary for structures with a
measurable from the HFBB model. Second, the directional cou- damped primary system, involves a numerical search. The optimal
pling of the PTMD is captured in physical domain by first trans- values for the parameters are determined by using simulations of
forming the acceleration responses for the suspension point (ü and the coupled response of the primary and auxiliary systems. Results
v̈) into the physical coordinates at each time step and subsequently have been demonstrated by several researchers for conventional
transforming the control force exerted by the PTMD on the primary translational TMDs (Ioi and Ikeda 1978; Warburton and Ayorinde
structure back into modal coordinates. In doing so, the HFBB 1980; Thompson 1981; Bakre and Jangid 2007) and PTMDs
method can be adapted to solve the nonlinear PTMD equations (Gerges and Vickery 2005). Even in numerical approaches, many
in the time domain. researchers have utilized linearized equations. Hence, the concerns
in accurately predicting the responses still remain.

Optimum Design Parameters


Planar Pendulum Tuned Mass Damper Model
The optimal design parameters consist of the optimal mass ratio, Consider an SDOF system with a pendulum-type TMD (Fig. 1).
μopt ; frequency ratio, f r;opt ; and auxiliary damping ratio, β a;opt . The equations of motion given in Eqs. (11a) and (11b) are simpli-
These quantities represent the ratio of the auxiliary parameters fied for a mixed two DOF system, one translational (primary) in the
to their corresponding primary structure parameters. The mass ra- modal domain and one rotational (auxiliary):
tio, μ, is a ratio of the auxiliary mass, ma , to the primary effective
modal mass, Mn , for the mode to be controlled (typically the fun- M n ẍðtÞ þ Cn x_ ðtÞ þ K n xðtÞ þ ma ẍðtÞ þ ma La θ̈ðtÞ ¼ Fn ðtÞ ð19Þ
damental mode of vibration). The frequency ratio is the ratio of the
frequency of the PTMD to the natural frequency of the primary and
system and is directly related to the pendulum length, La . Similarly,
the damping ratio, the ratio of the auxiliary damping coefficient to _ þ ðma gLa þ ks h2 ÞθðtÞ þ ma La ẍðtÞ ¼ 0
ma L2a θ̈ðtÞ þ ca h2 θðtÞ
the critical auxiliary damping, is directly related to the auxiliary ð20Þ
damping coefficients, cx and cy .
Values for the optimal frequency ratio and auxiliary system where M n , Cn , K n , and Fn ðtÞ = generalized mass, damping,
damping ratio under random main mass excitation are calculated stiffness, and force of the primary system for the mode to be con-
by using two primary approaches. The first seeks to minimize trolled. Here, the primary system is modelled as an SDOF with a

© ASCE 04013019-5 J. Struct. Eng.

J. Struct. Eng., 2013, 139(12): 04013019


translational modal coordinate xðtÞ. The auxiliary system has a ro- displacement) of the primary mass was used as the performance
tational DOF, θðtÞ; auxiliary mass, ma ; and pendulum length, La ; index, because the PTMD is usually designed to reduce this re-
and h is the distance between the suspension point and the spring/ sponse within acceptable serviceability criteria limits (Xu and
damper attachment point. The auxiliary stiffness is comprised of Kwok 1994). Using standard curve fitting techniques, the following
the inherent stiffness of the pendulum, ma gLa , and the auxiliary relationships were found for the optimal frequency and auxiliary
spring with stiffness ks (previously kx in Fig. 1); ca is the damping damping ratios:
of the auxiliary damper (previously cx in Fig. 1).
The damping ratio for a PTMD is defined as fr;opt ¼ −0.64μ − 0.23ζ p þ 0.99 ð25aÞ

ca h 2 β a;opt ¼ −4.89μ2 þ 1.89μ þ 0.018 ð25bÞ


βa ¼ ð21Þ
2ma L2a ω2

where the circular natural frequency of the auxiliary system is


Downloaded from ascelibrary.org by UNB - Universidade de Bras?lia on 02/18/20. Copyright ASCE. For personal use only; all rights reserved.

The proposed design equations for the planar PTMD are plotted
given by in Fig. 2 for 1, 3, and 5% damping ratios, together with the closed-
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi form solutions for no primary mass damping [Eqs. (23) and (24)].
ma gLa þ ks h2 The maximum error between the results from the design equations
ω2 ¼ ð22Þ
ma L2 developed by using standard curve fitting techniques and the results
pffiffiffiffiffiffiffiffiffiffi from the numerical search are also shown in Fig. 2. These results
which simplifies to ω2 ¼ g=La for the case in which there is no are deemed applicable only for the limited range of mass and damp-
auxiliary spring. Optimum tuning parameters for pendulum-type ing ratios presented here.
TMDs have been presented in the literature that minimize the
RMS displacement response of the primary system, for the case
of the undamped primary system (Gerges and Vickery 2005). Optimal Frequency and Damping Ratios for the
The optimal frequency ratio for force excited primary mass with Three-Dimensional Case
a PTMD in which the mass is assumed to be lumped at the free To determine the optimal damper parameters from the combined
end of the pendulum length is planar-spherical PTMD model, a numerical search method is em-
pffiffiffiffiffiffiffiffiffiffiμffi ployed by using an SDOF model with x, y, and z translational
ω2opt 1þ2
f r;opt ¼ ¼ ð23Þ degrees of freedom in addition to θ and φ. Several simulation
ω1 1þμ trials are performed by using a broadband white noise primary mass
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi excitation, and a cost function is evaluated each time, based on
where ω1 ¼ K n =M n = undamped circular natural frequency of the RMS acceleration response in both horizontal directions.
the primary system, The optimal auxiliary damping ratio is The optimal frequency and damping ratios are determined for
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi various primary structure damping ratios and auxiliary to primary
2
μ þ 3μ4 mass ratios.
β a;opt ¼ ð24Þ
4 þ 6μ þ 2μ2 There are a few important considerations when selecting the
auxiliary to primary mass ratio. It is desirable to reduce the aux-
Various numerical simulations were performed and averaged for iliary to primary mass ratio because this reduces the overall weight
primary mass damping ratios of ζ p ¼ 1 to 5% and auxiliary to pri- of the materials and the size of the dampers; however, higher
mary mass ratios varying from μ ¼ 0.01 to 0.15. The equations of mass ratios offer better performance. Although it is well known
motion were implemented in Simulink by using state-space repre- that the performance of a TMD increases with increasing mass
sentation and the numerical integration was performed by using ratio, the mass ratio is often determined by other design consider-
Runge-Kutta method. The RMS acceleration response (rather than ations and project constraints. Hence, the ensuing study assumes

(a) (b)

Fig. 2. Optimal frequency ratio and auxiliary damping ratio for planar PTMD: (a) frequency ratio; (b) damping ratio

© ASCE 04013019-6 J. Struct. Eng.

J. Struct. Eng., 2013, 139(12): 04013019


that the designer has the ability to select an auxiliary mass β a;opt ¼ ð−45ζ p − 2.4Þμ2 þ ð9.7ζ p þ 1.5Þμ − 0.22ζ p þ 0.05
based on such constraints and focuses on determining the
ð26bÞ
optimal auxiliary damping and pendulum length for a range of
mass ratios.
The damping in the primary structure is considered to be 1 to The maximum errors between the design equations developed
5% critical. Auxiliary to primary mass ratios range from 0.005 to by using standard curve fitting techniques and the results from the
0.125. To simplify the analysis, the auxiliary damping in the x and y numerical search are shown in Fig. 3. Once the frequency and
directions are assumed to be equal (cx ¼ cy ). Also, the damper at- damping ratios are calculated, they can be converted into the ap-
tachment point coincides with the pendulum length (hx ¼ hy ¼ L). propriate physical parameters through the following procedure.
Finally, no additional auxiliary stiffness is introduced in the form When no auxiliary springs are included (kx ¼ ky ¼ 0), the optimal
of auxiliary springs (kx ¼ ky ¼ 0). The optimal frequency ratio pendulum length is related to the frequency ratio by the following
found by using the numerical search method is shown in Fig. 3, relationship (Gerges and Vickery 2005)
Downloaded from ascelibrary.org by UNB - Universidade de Bras?lia on 02/18/20. Copyright ASCE. For personal use only; all rights reserved.

together with the closed-form solution for no primary mass damp-


g
ing [Eq. (23)]. The optimal frequency ratio results demonstrate Lopt ¼ ð27Þ
less sensitivity to mass ratio than that predicted by the closed-form f 2r;opt ω21
solutions and the numerical search of the planar PTMD model.
Furthermore, varying the primary mass damping has the effect of The optimal auxiliary damping is related to the damping ratio by
changing the sensitivity of the frequency ratio to the mass ratio. the following relationship
For the primary mass damping ratios under consideration, the sen- cd;opt ¼ 2β a;opt ma f r;opt ω1 ð28Þ
sitivity of the results indicate that the primary system damping and
auxiliary to primary mass ratio do not play significant roles in the
selection of the optimal frequency ratio. Effect of Mass Ratio on Detuning
The optimal damping ratio found by using the numerical search
method is shown in Fig. 3, together with the closed-form solution The detuning of a TMD occurs when the TMD parameters do
for no primary mass damping [Eq. (24)]. The numerical search not coincide with the corresponding optimal parameters. For
of the planar-spherical system results in slightly greater damping practical applications, it is important to understand the conditions
ratios than the closed-form optimal auxiliary damping equations. under which the effect of detuning will become important. In
For mass ratios of less than approximately 3%, optimal auxiliary this study, this concern is specifically addressed from the stand-
damping ratios decrease with increasing primary mass damping point of detuning versus the auxiliary to primary mass ratio of
ratios and increase with increasing mass ratios for mass ratios the PTMD.
greater than approximately 3%. As the mass ratios moves into To evaluate the effect of the mass ratio on the optimal auxiliary
the upper range of the values considered, the disparity between parameters f r;opt and β a;opt , a mass, free to translate in each direc-
the closed-form optimal auxiliary damping and the planar-spherical tion, coupled with a planar-spherical PTMD is considered.
prediction grows, with the latter producing greater optimal aux- Primary mass damping is fixed at 2% for illustration purposes
iliary damping. (it has previously been shown that this parameter is not critical
By using standard curve fitting techniques, the following rela- for the range of values under consideration), and mass ratios of
tionships are found for the optimal frequency and auxiliary damp- μ ¼ 0.01, 0.02, 0.03, 0.05, 0.07, and 0.09 are considered. The sys-
ing ratios: tem is excited in both horizontal directions by Gaussian white
noise. The auxiliary damping coefficient and pendulum length
are varied and the RMS of the acceleration response in the horizon-
fr;opt ¼ ð4.2ζ p − 0.15Þμ þ 0.15ζ p þ 0.99 ð26aÞ tal plane is used as the performance indicator. The results are

(a) (b)

Fig. 3. Optimal frequency and damping ratios for planar-spherical PTMD: (a) frequency ratio; (b) damping ratio

© ASCE 04013019-7 J. Struct. Eng.

J. Struct. Eng., 2013, 139(12): 04013019


averaged over many realizations of the input. Fig. 4 shows the
variation in the optimal auxiliary damping coefficient at various
mass ratios for a fixed pendulum length and the change in optimal
pendulum length at various mass ratios for a fixed level of auxiliary
damping.
Several observations can be made regarding the optimal param-
eters as the mass ratio is increased. First, the optimal pendulum
length decreases slightly with increasing mass ratio. Second, the
sensitivity to detuning reduces as the mass ratio increases, indicated
by the flattening of the curves in Fig. 4. This implies that selection
of optimal damper parameters for higher mass ratios is less critical.
The optimal pendulum length curves (Fig. 4) become flatter in both
Downloaded from ascelibrary.org by UNB - Universidade de Bras?lia on 02/18/20. Copyright ASCE. For personal use only; all rights reserved.

directions with increasing mass ratio, meaning that a similar per-


formance is expected for a PTMD that has a pendulum length away Fig. 5. Apron Tower at Pearson International Airport in Toronto,
from the optimal length, regardless of whether it is shorter or longer Ontario, Canada (photograph courtesy of Aaron J. Roffel)
than optimal length. The optimal auxiliary damping coefficient also
is less susceptible to detuning as the mass ratio increases, but only
for damping greater than the optimal level. Therefore, in selecting
the level of auxiliary damping for higher mass ratios, it is critical
that the damping coefficient is at least optimal.

Full-Scale Study: Application to the Apron Tower

To investigate the effectiveness of the proposed design equations


and simulation model, the Apron Tower at Pearson International
Airport in Toronto, Ontario, Canada, is considered as the test
bed. The Apron Tower, shown in Fig. 5, rises 49 m above the fourth
level of Terminal 1, with a total height above grade of 68.5 m. Con-
structed in 2000, the tower is a steel structure with composite steel
deck and concrete floors. The structure has 10 core levels, primarily
consisting of a scissor stairwell, elevator, and service shafts. The
upper five levels contain mechanical, electrical, and communica-
tion service equipment, office space, and areas for controllers to
work performing apron management tasks. The structure is sup- Fig. 6. PTMDs installed on the Pearson Apron Tower (photograph
ported by six primary steel columns resting on transfer girders courtesy of Aaron J. Roffel)
at the terminal roof level. Lateral loads are transferred to a
combination of braced and moment frames through rigid dia-
phragm action in the composite floors. The structure is augmented ratio, one hypothetical case of a 9,050-kg tuned mass is also se-
with two pendulum-type tuned mass dampers, each 25,000 kg lected, corresponding to mass ratios of 2.00 and 2.25% in the
(ma ¼ 50,000 kg), supported at the roof level within the steel truss two lateral modes.
roof system (Fig. 6). The corresponding mass ratios in the first two To estimate the generalized forces, a scaled model of the Apron
lateral modes are 12.4 and 11.1%. In addition to the actual mass Tower is built and tested in the Rowan Williams Davies and Irwin

(a) (b)

Fig. 4. Sensitivity of tuning parameters for auxiliary to primary mass ratio: (a) for auxiliary damping coefficient, ca ; (b) for pendulum length, La

© ASCE 04013019-8 J. Struct. Eng.

J. Struct. Eng., 2013, 139(12): 04013019


(RWDI) boundary layer wind tunnel facility in Guelph, Ontario, Table 1. Optimal Parameters for Apron Tower from Design Equations
Canada. The experimental setup is shown in Fig. 7. The generalized Optimal parameter Closed form Planar Planar-spherical
forces thus calculated are subsequently used in the previously de-
Eqs. (23) Eqs. (25a) Eqs. (26a)
scribed numerical simulations. In addition to wind tunnel tests, a
Parameter Units and (24) and (25b) and (26b)
vibration measurement program is performed to determine the
natural frequencies and primary mode shapes. The identified mode μ ¼ 12.4%
shapes are subsequently used as inputs for the HFBB method. In fr;opt 0.917 0.913 0.975
addition, the identified primary structure damping is used on β a;opt 0.169 0.178 0.204
Lopt m 0.659 0.663 0.582
prediction simulations. The instrumentation program consists of in-
cd;opt N s=m 65.1 × 103 68.4 × 103 83.5 × 103
stallation and monitoring of 12 high-sensitivity seismic accelerom- μ ¼ 2.25%
eters to measure the responses of the structure. The responses are fr;opt 0.984 0.978 0.987
processed by using a hybrid time frequency blind source separation β a;opt 0.074 0.058 0.081
Downloaded from ascelibrary.org by UNB - Universidade de Bras?lia on 02/18/20. Copyright ASCE. For personal use only; all rights reserved.

algorithm to determine the modal characteristics. The three lowest Lopt m 0.572 0.579 0.568
frequencies identified are 0.67, 0.83, and 1.44 Hz, and the cd;opt N s=m 5.57 × 103 4.33 × 103 6.05 × 103
corresponding damping ratios are 0.85, 1.92, and 2.70%, critical.
The details of the instrumentation and identification studies are
documented elsewhere (Hazra 2010); the details are not repeated
here for the sake of brevity. The mode shapes obtained from the the sensitivity to detuning is less pronounced for higher mass ratios.
finite-element model of the tower are used for the analysis and The 0° (also 360°) degree wind direction corresponds to a wind
closely correspond to the identified modes. coming from the north. The direction proceeds clockwise, so a
90° heading is an east wind. Because the varying nature of the wind
excitation, the performance of the PTMD will vary depending on
Comparison of the Optimal PTMD Parameters between the wind direction. Therefore, a numerical search is performed for
Various Methods each wind direction to determine the optimal auxiliary parameters.
To investigate the approach that best predicts the optimal auxiliary The optimal pendulum length and auxiliary damping coefficient for
parameters, f r;opt and β a;opt are evaluated by using the three ap- each wind direction are presented in Table 2. For the sake of brev-
proaches described earlier: the closed-form equations in Eqs. (23) ity, only the critical wind direction results are shown; however, the
and (24) (Gerges and Vickery 2005); the design equations in average and SD are calculated by using the data for all of the wind
Eqs. (25a) and (25b), which are found by using a numerical search directions.
approach of a planar PTMD excited by using Gaussian white noise Table 2 shows a greater variation in the damper parameters for
(planar); and the design equations in Eqs. (26a) and (26b), which the higher mass ratio. The optimal pendulum length varies by ap-
are found by a numerical search of the planar-spherical PTMD proximately 3.4 cm for the 2.25% mass ratio and 7.8 cm for the
model, also excited using Gaussian white noise (planar-spherical). 12.4% mass ratio. The difference in the average pendulum length
Table 1 shows the predicted optimal frequency ratio, f r;opt , and between both mass ratios is negligible. The planar-spherical predic-
damping ratio, β a;opt , and the optimal pendulum length and aux- tion approach best captures the negligible difference in optimal
iliary damping coefficient, found by using Eqs. (27) and (28) pendulum length for the two mass ratios.
for both the current in-service mass ratio of 12.4%, and a mass ratio The optimal auxiliary damping coefficient varies considerably
of 2.25%. for both mass ratios and for both cases of excitations in Tables 1
Each method predicts slightly varying pendulum lengths when a and 2. The planar-spherical prediction approach better predicts the
structure equipped with a PTMD is considered. For the 2.25% mass optimal auxiliary damping than the numerical search results using
ratio, the variation is a nominal 1.1 cm. For the 12.4% mass ratio, the HFBB method; the optimal damping coefficient is 12% greater
the difference is 8.1 cm; however, this is expected to have a limited than the best prediction for the 2.25% mass ratio and 15% less for
effect on the actual response because it has been demonstrated that the 12.4% mass ratio. It is hypothesized that the assumption of
equal lateral natural frequencies in the orthogonal directions results
in an uncertain estimate of the auxiliary damping coefficients. The
effect of lateral frequency ratio on the optimal parameter estimates
needs to be investigated further.

Table 2. Optimal Damper Parameters from Apron Tower Model, Excited


by Using HFBB Data
Optimal damper parameters
Reference
μ ¼ 2.25% μ ¼ 12.4%
wind
Wind velocity La ca La ca
direction (km=h) (m) (×103 N s=m) (m) (×103 N s=m)
260 129 0.553 4.84 0.539 77.4
270 129 0.543 5.69 0.523 73.7
280 126 0.537 4.74 0.523 79.7
290 122 0.557 5.28 0.532 82.1
300 116 0.556 4.32 0.536 80.4
Fig. 7. Wind tunnel study model of Pearson Apron Tower (image Average 0.554 4.90 0.555 72.9
SD 0.007 0.69 0.016 29.5
courtesy of Rowan Williams Davies & Irwin Inc.)

© ASCE 04013019-9 J. Struct. Eng.

J. Struct. Eng., 2013, 139(12): 04013019


To evaluate the performance of the parameter prediction meth- separated. The auxiliary damper parameters, particularly the pen-
ods, the responses are calculated by using the HFBB approach for dulum length, are tuned to the fundamental mode and provide little
all three auxiliary parameter settings for each wind direction. The improvement, and in some cases, may degrade the performance, in
wind time histories for each angle are scaled to a one-year mean the other lateral direction. The critical response of the uncontrolled
recurrence interval. The RMS response are provided for each wind system is the y direction response; for the controlled structure, this
direction, together with the corresponding reduction when com- shifts to the x direction. The optimally tuned results are found by
pared to the bare structural system lacking the PTMD. The results numerical search based on a cost function of reducing the combined
for the responses in the north-south (y) direction for a 2.25% tuned x and y direction RMS response; therefore, a greater x direction
mass are provided in Table 3; the responses in the east-west (x) response occurs in several the wind directions when the lateral
direction are provided in Table 4. Only the data for the critical wind responses are considered separately.
directions (260 to 300°) are shown, for the sake of brevity. A means of overcoming this inherent weakness in PTMD design
Tables 3 and 4 show that the 260° wind direction produces a is to increase the mass ratio, μ. As discussed earlier, an increased
Downloaded from ascelibrary.org by UNB - Universidade de Bras?lia on 02/18/20. Copyright ASCE. For personal use only; all rights reserved.

critical response in the y direction for the uncontrolled system. mass ratio decreases the overall RMS response of the primary sys-
For the structure equipped with the PTMD, this result shifts slightly tem. Also, according to Fig. 4, the sensitivity to detuning for higher
to the 270° wind direction. Several observations regarding the op- mass ratios is reduced. Therefore, it is expected that the x direction
timal auxiliary parameter predictions can be made. First, the planar- response will experience an improved RMS response when
spherical prediction slightly outperforms the others, whereas the equipped with the 12.4% tuned mass as a result of the reduced sen-
planar numerical search predictor performs slightly worse. Across sitivity to detuning, although the PTMD is tuned to the first lateral
all wind directions, the optimal parameters obtained from the mode in the y direction. The y direction roof RMS acceleration
planar-spherical prediction model provides very similar perfor- responses for the structure without the PTMD are compared with
mance to those obtained from the closed-form parameter prediction a structure equipped with a 12.4% tuned mass. Additionally, vari-
method. Second, the various methods are reasonably accurate in ous parameter prediction approaches, together with the optimal
predicting the actual optimal parameters (determined by numerical auxiliary parameters, are presented in Table 5. The results for
search by using the HFBB excitations). the x direction responses are given in Table 6.
In general, the optimal parameters obtained by the planar- For all wind directions, an increased reduction in the roof RMS
spherical prediction model result in marginally better performance acceleration response is experienced for y direction motion com-
for the x direction response when tuned to the y direction lateral pared with the 2.25% tuned mass. On average, this reduction is
mode. Therefore, the reduced effect of the PTMD on the x direction 22%; for the winds in the vicinity of the critical wind direction
responses shown in Table 4 is expected. The critical wind direction (260 to 300°), the average performance improvement is 26%.
for the x direction response is 280° for the structure without the There is a significant performance improvement for the 12.4%
PTMD. This shifts to the 270° wind direction for the PTMD- tuned mass in the x direction compared with the 2.25% tuned mass
equipped structure, in which degraded performance compared to (22% reduction), particularly in the vicinity of the critical wind di-
the bare structure is experienced. This highlights a critical weak- rection (260 to 300°). The performance improvement is twofold.
ness in a PTMD with a low auxiliary to primary mass ratio when First, PTMD performance increases with larger mass ratios.
the orthogonal lateral frequencies of the primary structure are well Second, the increased mass ratio results in a reduced sensitivity

Table 3. Roof RMS Acceleration Response in the y Direction and Corresponding Reduction for the Apron Tower Equipped with a PTMD Tuned to μ ¼
2.25% for the Critical Wind Directions
RMS acceleration response in y direction
With PTMD tuned by using method
Without PTMD Closed form Planar Planar-spherical Optimally tuned
Wind direction mg mg % mg % mg % mg %
260 11.53 5.56 52 5.74 50 5.52 52 5.49 52
270 11.19 5.63 50 5.86 48 5.58 50 5.51 51
280 11.06 5.49 50 5.66 49 5.43 51 5.36 52
290 9.85 4.73 52 4.89 50 4.69 52 4.64 53
300 11.25 4.10 64 4.08 64 4.13 63 4.03 64

Table 4. Roof RMS Acceleration Response in the x Direction and Corresponding Reduction for the Apron Tower Equipped with a PTMD Tuned to μ ¼
2.25% for the Critical Wind Directions
RMS acceleration response in x direction
With PTMD tuned by using method
Without PTMD Closed form Planar Planar-spherical Optimally tuned
Wind direction mg mg % mg % mg % mg %
260 6.69 5.79 13 5.92 11 5.74 14 5.80 13
270 6.83 5.88 14 6.02 12 5.83 15 5.76 16
280 6.98 6.03 14 6.18 11 5.98 14 5.99 14
290 6.10 5.71 6 5.87 4 5.65 7 5.71 6
300 6.41 5.83 9 6.03 6 5.75 10 5.95 7

© ASCE 04013019-10 J. Struct. Eng.

J. Struct. Eng., 2013, 139(12): 04013019


Table 5. Roof RMS Acceleration Response in the y Direction and Corresponding Reduction for the Apron Tower Equipped with a PTMD Tuned to μ ¼
12.4% for the Critical Wind Directions
RMS acceleration response in y direction
With PTMD tuned by using method
Without PTMD Closed form Planar Planar-spherical Optimally tuned
Wind direction mg mg % mg % mg % mg %
260 11.53 3.88 66 3.88 66 3.73 68 3.70 68
270 11.19 3.92 65 3.92 65 3.75 66 3.72 67
280 11.06 3.86 65 3.86 65 3.67 67 3.63 67
290 9.85 3.35 66 3.35 66 3.21 67 3.19 68
300 11.25 3.03 73 3.02 73 2.92 74 2.91 74
Downloaded from ascelibrary.org by UNB - Universidade de Bras?lia on 02/18/20. Copyright ASCE. For personal use only; all rights reserved.

Table 6. Roof RMS Acceleration Response in the x Direction and Corresponding Reduction for the Apron Tower Equipped with a PTMD Tuned to μ ¼
12.4% for the Critical Wind Directions
RMS acceleration response in x direction
With PTMD tuned by using method
Without PTMD Closed form Planar Planar-spherical Optimally tuned
Wind direction mg mg % mg % mg % mg %
260 6.69 4.58 32 4.53 32 4.17 38 4.11 39
270 6.83 4.56 33 4.51 34 4.18 39 4.12 40
280 6.98 4.65 33 4.60 34 4.20 40 4.04 42
290 6.10 4.05 34 4.02 34 3.67 40 3.53 42
300 6.41 3.45 46 3.44 46 3.17 51 3.06 52

Table 7. Comparison of Roof RMS Acceleration Response Predictions and Corresponding Reduction for the Apron Tower Equipped with a PTMD Tuned to
μ ¼ 2.25% Using Planar-Spherical and Planar Models for the Critical Wind Directions
Without PTMD Planar-spherical PTMD Planar PTMD
x y x y x y
Wind direction mg mg mg % mg % mg % mg %
260 6.69 11.53 5.80 13 5.49 52 5.70 15 5.41 53
270 6.83 11.19 5.76 16 5.51 51 5.72 16 5.46 51
280 6.98 11.06 5.99 14 5.36 52 5.91 15 5.35 52
290 6.10 9.85 5.71 6 4.64 53 5.46 11 4.57 54
300 6.41 11.25 5.95 7 4.03 64 5.81 9 4.01 64

to detuning; therefore, a significant improvement is realized in the x ability of the planar PTMD model to predict the structural re-
direction response, although the PTMD is tuned to the first lateral sponses when compared to the planar-spherical model described
mode in the y direction. earlier. The PTMD is tuned to the fundamental mode in the y
As observed for the 2.25% tuned mass, the evaluation of the direction.
various parameter prediction approaches finds nominal perfor- Table 7 outlines the RMS acceleration of the roof for the critical
mance improvement of the planar-spherical over the closed-form wind directions, evaluated by combining the planar PTMD model
prediction or planar prediction model for the response in the (with 2.25% tuned mass) coupled with the effective modal mass,
y direction. This is of particular interest, because the different stiffness, and damping for the primary system for each mode of
methods predict relatively different auxiliary frequency ratios vibration. The results for all wind directions are summarized in
(fr;opt ¼ 0.913 to 0.975). This underscores the fact that increasing Fig. 8. The input excitation is the generalized force for each
the mass ratio dramatically reduces the sensitivity of the system to mode, which is directly measured from the HFBB model. The
detuning. For the responses in the x direction, the planar-spherical RMS response of the planar model is compared with the previously
model better predicts the optimal auxiliary parameters than the described planar-spherical HFBB model. The auxiliary damper
2.25% mass ratio, demonstrating an increasing lateral coupling parameters are set to their optimal values, given previously in
effect for higher mass ratios. Table 2.
There is very little discrepancy in the roof RMS responses be-
tween the two methods for motion in the y direction. For motion in
Comparison of Planar and Planar-Spherical Model RMS the east-west direction, the RMS response simulated by each model
Acceleration Responses varies by as much as 9%, with the more simplistic model predicting
After investigating the effect of optimal parameters calculated by greater reductions in RMS responses (4% in the critical wind di-
using the various approaches, attention is now given to the rection range). The results for a 12.4% mass ratio are presented in

© ASCE 04013019-11 J. Struct. Eng.

J. Struct. Eng., 2013, 139(12): 04013019


Table 8 for the critical wind directions. The data are summarized in acceptance threshold for occupant comfort decreases with increas-
Fig. 8 for all wind directions. ing frequency within the 0–1 Hz range (ISO 6897) (1984). Hence, it
The performance of the planar model for predicting the RMS is important to quantify the effect of TMD tuned to the dominant
response of the structure equipped with a 12.4% tuned mass is mode on higher modes, because they may not meet the perfor-
worse than that for the 2.25% tuned mass. Unlike the 2.25% tuned mance requirements.
mass, there is also discrepancy in the results for the y direction re-
sponse. Up to 14% difference in RMS responses (8% unconserva-
tively) are reported for the y direction; within the critical wind Conclusions
direction, errors of up to 14% are reported, although the linearized
planar model is more conservative. For the response in the x direc- The following primary conclusions can be drawn from this study.
tion, the planar model predicts up to a 20% lower response than the Increasing the auxiliary to primary mass ratio results in a reduced
more complex model, with up to 16% difference in the critical wind sensitivity to detuning. The planar-spherical model predicted a neg-
Downloaded from ascelibrary.org by UNB - Universidade de Bras?lia on 02/18/20. Copyright ASCE. For personal use only; all rights reserved.

direction and with the linearized planar model predicting better per- ligible change in the optimal pendulum length for an increased
formance. For the larger auxiliary mass, the coupling effect of the mass, which was later confirmed by comparing the response to
PTMD on the primary structural response is more pronounced. wind load generated from a HFBB wind tunnel test for tuned
This is an important conclusion that has implications with respect masses of 2.25 and 12.4%. Therefore, the simple closed-form
to the design of PTMDs. Specifically, for the higher modes, the design equations reported in the literature are adequate for design
purposes.
Subsequent investigation compares the ability of the planar
PTMD model to predict the primary structural response when com-
pared to the planar-spherical model. The responses are investigated
for a finite-element representation of an actual structure and com-
pared for tuned masses of 2.25% and 12.4%. It is found that for the
lower mass ratio, the planar model accurately predicts the response
in the same direction as the fundamental lateral mode, to which the
PTMD was tuned. However, for the other direction, the planar
model has errors of 9%. For the higher mass ratios, the performance
of the planar model deteriorates in predicting the primary structure
RMS acceleration response. Errors of up to 12% in the direction of
the fundamental lateral mode, to which the PTMD was tuned, are
reported, in addition to up to 20% difference for the response in the
(a) orthogonal direction. These conclusions are applicable for the
range of mass ratios and primary structure damping studied here.
For low frequency structures, this can have a significant impact on
the design, because the occupancy comfort thresholds have been
shown to reduce with increasing frequency.

Acknowledgments

The authors are grateful to the Natural Sciences and Engineering


Research Council of Canada (NSERC) and the Ontario Centres of
Excellence (OCE) for providing the financial support to conduct
this study. The authors also thank Greater Toronto Airports Author-
(b) ity (GTAA) and Rowan Williams Davies & Irwin Inc. (RWDI) who
serve as the industrial partners in this collaborative project. Special
Fig. 8. Comparison of roof RMS acceleration response predictions by thanks to Mr. Greg Thompson, Mr. Scott Gamble, and Dr. Peter
using planar-spherical and planar PTMD models: (a) for 2.25% tuned
Irwin of RWDI for providing valuable insights into the practical
mass; (b) for 12.4% tuned mass
design aspects of TMDs.

Table 8. Comparison of Roof RMS Acceleration Response Predictions and Corresponding Reduction for the Apron Tower Equipped with a PTMD Tuned to
μ ¼ 12.4% Using Planar-Spherical and Planar Models for the Critical Wind Directions
Without PTMD Planar-spherical PTMD Planar PTMD
x y x y x y
Wind direction mg mg mg % mg % mg % mg %
260 6.69 11.53 4.11 39 3.70 68 3.95 41 4.23 63
270 6.83 11.19 4.12 40 3.72 67 3.78 45 3.99 64
280 6.98 11.06 4.04 42 3.63 67 3.74 46 3.85 65
290 6.10 9.85 3.53 42 3.19 68 3.32 46 3.32 66
300 6.41 11.25 3.06 52 2.91 74 3.00 53 2.95 74

© ASCE 04013019-12 J. Struct. Eng.

J. Struct. Eng., 2013, 139(12): 04013019


References Meirovitch, L. (1970). Methods of analytical dynamics, McGraw-Hill,
New York.
Abe, M., and Fujino, Y. (1994). “Dynamic characterization of multiple Nagarajaiah, S., and Varadarajan, N. (2005). “Semi-active control of wind
tuned mass dampers and some design formulas.” Earthquake Eng. excited building with variable stiffness TMD using short-time fourier
Struct. Dynam., 23(8), 813–835. transform.” Eng. Struct., 27(3), 431–441.
Bakre, S. V., and Jangid, R. S. (2007). “Optimum parameters of tuned mass Pinkaew, T., and Fujino, Y. (2001). “Effectiveness of semi-active tuned
damper for damped main system.” Struct. Control Health Monit., 14(3), mass dampers under harmonic excitation.” Eng. Struct., 23(7),
448–470. 850–856.
Chen, X., and Kareem, A. (2005). “Coupled dynamic analysis and equiv- Rana, R., and Soong, T. T. (1998). “Parametric study and simplified design
alent static wind loads on buildings with three-dimensional modes.” of tuned mass dampers.” Eng. Struct., 20(3), 193–204.
J. Struct. Eng., 131(7), 1071–1082.
Ricciardelli, F., Occhiuzzi, A., and Clemente, P. (2000). “Semi-active tuned
den Hartog, D. B. (1956). Mechanical vibrations, 4th Ed., McGraw-Hill,
mass damper control strategy for wind-excited structures.” J. Wind Eng.
New York.
Ind. Aerod., 88(1), 57–74.
Gerges, R. R., and Vickery, B. J. (2005). “Optimum design of pendulum-
Downloaded from ascelibrary.org by UNB - Universidade de Bras?lia on 02/18/20. Copyright ASCE. For personal use only; all rights reserved.

type tuned mass dampers.” Struct. Des. Tall Special Build., 14(4), Roffel, A. J., Lourenco, R., Narasimhan, S., and Yarusevych, S. (2011).
353–368. “Adaptive compensation for detuning in pendulum tuned mass
Ghosh, A., and Basu, B. (2007). “A closed-form optimal tuning criterion dampers.” J. Struct. Eng., 137(2), 242–251.
for TMD in damped structures.” Struct. Control Health Monit., 14(4), Sacks, M. P., and Swallow, J. C. (1993). “Tuned mass dampers for towers
681–692. and buildings.” Proc., Symp. on Structural Engineering in Natural
Hazra, B. (2010). “Hybrid time and time-frequency blind source separation Hazards Mitigation, SEI Structures Congress, Irvine, CA, 640–645.
towards ambient system identification of structures.” Ph.D. thesis, Univ. Simulink [Computer software]. MathWorks, Natick, MA.
of Waterloo, Waterloo, ON, Canada. Tanaka, H., and Mak, C. Y. (1983). “Effect of tuned mass dampers on wind
Holmes, J. D. (1987). “Mode shape corrections for dynamic response to induced response of tall buildings.” J. Wind Eng. Ind. Aerod., 14(1–3),
wind.” Eng. Struct., 9(3), 210–212. 357–368.
ISO. (1984). “Guidelines for the evaluation of the response of occupants of Thompson, A. G. (1981). “Optimum tuning and damping of a dynamic
fixed structures, especially buildings and off-shore structures, to low- vibration absorber applied to a force excited and damped primary
frequency horizontal motion (0.063 to 1 Hz).” ISO 6897, Geneva. system.” J. Sound Vib., 77(3), 403–415.
Ioi, T., and Ikeda, K. (1978). “On the dynamic vibration damped absorber Tschanz, T., and Davenport, A. G. (1983). “The base balance technique for
of the vibration system.” Bull. JSME, 21(151), 64–71. the determination of dynamic wind loads.” J. Wind Eng. Ind. Aerod.,
Kareem, A. (1992). “Dynamic response of high-rise buildings to stochastic 13(1–3), 429–439.
wind loads.” J. Wind Eng. Ind. Aerod., 42(1–3), 1101–1112. Tse, K. T., Hitchcock, P. A., and Kwok, K. C. S. (2009). “Mode shape
Kareem, A., and Kijewski, T. (1999). “Mitigation of motions of tall build- linearization for HFBB analysis of wind-excited complex tall build-
ings with specific examples of recent applications.” Wind Struct., 2(3), ings.” Eng. Struct., 31(3), 675–685.
201–252.
Warburton, G. B., and Ayorinde, E. O. (1980). “Optimum absorber param-
Kwok, K., Hitchcock, P. A., and Burton, M. D. (2009). “Perception of
eters for simple systems.” Earthquake Eng. Struct. Dynam., 8(3),
vibration and occupant comfort in wind-excited tall buildings.” J. Wind
197–217.
Eng. Ind. Aerod., 97(7–8), 368–380.
Xu, Y., Kwok, K., and Samali, B. (1992). “The effect of tuned mass damp-
Kwok, K. C. S., and Macdonald, P. A. (1990). “Full-scale measurements of
wind-induced acceleration response of Sydney Tower.” Eng. Struct., ers and liquid dampers on cross-wind response of tall/slender struc-
12(3), 153–162. tures.” J. Wind Eng. Ind. Aerod., 40(1), 33–54.
Kwok, K. C. S., and Samali, B. (1995). “Performance of tuned mass damp- Xu, Y. L., and Kwok, K. C. S. (1994). “Semianalytical method for
ers under wind loads.” Eng. Struct., 17(9), 655–667. parametric study of tuned mass dampers.” J. Struct. Eng., 120(3),
Lam, K., and Li, A. (2009). “Mode shape correction for wind-induced dy- 747–764.
namic responses of tall buildings using time-domain computation and Yip, D. Y. N., and Flay, R. G. J. (1995). “A new force balance data analysis
wind tunnel tests.” J. Sound Vib., 322(4–5), 740–755. method for wind response predictions of tall buildings.” J. Wind Eng.
Lee, C.-L., Chen, Y.-T., Chung, L.-L., and Wang, Y.-P. (2006). “Optimal Ind. Aerod., 54–55, 457–471.
design theories and applications of tuned mass dampers.” Eng. Struct., Zhou, Y., Kareem, A., and Gu, M. (2002). “Mode shape corrections for
28(1), 43–53. wind load effects.” J. Eng. Mech., 128(1), 15–23.

© ASCE 04013019-13 J. Struct. Eng.

J. Struct. Eng., 2013, 139(12): 04013019

You might also like