See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/233368648

Dynamic analysis of structures with multiple tuned mass dampers

Article in Journal of Civil Engineering and Management · March 2009

DOI: 10.3846/1392-3730.2009.15.77-86

CITATIONS READS

42 1,629

2 authors, including:

Roman Lewandowski
Poznań University of Technology
107 PUBLICATIONS 1,625 CITATIONS

SEE PROFILE

All content following this page was uploaded by Roman Lewandowski on 28 January 2014.

The user has requested enhancement of the downloaded file.

 JOURNAL OF CIVIL ENGINEERING AND MANAGEMENT
2009
15(1): 77–86

DYNAMIC ANALYSIS OF STRUCTURES WITH MULTIPLE TUNED MASS DAMPERS

Roman Lewandowski1, Justyna Grzymisławska2

Institute of Structural Engineering, Poznan University of Technology,
ul. Piotrowo 5, 60-965 Poznań, Poland
E-mail: [email protected]; [email protected]
Received 16 May 2008, accepted 20 Nov 2008

Abstract. It is the purpose of this paper to analyse the possibility of reducing the vibrations of frame building structures
with the help of multiple tuned mass dampers. Structures exposed to strong winds are considered. Excitation forces, which
are functions of wind velocity fluctuations, are treated as random forces. The spectral density functions of wind velocity
fluctuations are assumed as proposed by Davenport. The correlation theory of random vibration is used and the root mean
squares of displacements and accelerations are determined. Several remarks, concerning the effectiveness of multiple
tuned or mass dampers, are formulated from the results of calculation.
Keywords: reduction of vibration, random vibration, multiple-tuned mass dampers (MTMD).

1. Introduction 2001). The MTMD were designed in such a way that they
are tuned to several modes of structure vibration. The
Mass dampers have been used for reducing the vibrations number of dampers depends on the number of vibration
of structures for many years (McNamara 1977). They modes for which dampers are tuned. The performance of
have been successfully used in reducing the vibrations of multiple mass dampers under both wind and seismic exci-
building structures subjected to strong winds and seismic tation is analysed by Kareem and Kline (1995).
excitations (Xu et al. 1992). Principally, tuned mass The effectiveness and robustness of a particular ver-
dampers (TMD) installed on top floors have been studied. sion of MTMD, called “the multiple dual tuned mass
They have been designed in such a way that they are dampers”, is analysed in the paper (Han, Li 2006). The
tuned to the fundamental mode of vibration. In the paper problem of determination of optimum properties of
by Warburton (1982), a method for optimization of vari- MTMD is considered in the papers (Li, Qu 2006; Li
ous types of excitation forces was presented. The formu- 2002). Spatial structures with MTMD are analysed in
lae given by him there have often been used to design (Guo, Chen 2007).
TMD parameters, when reduction of dynamical dis- Moreover, the possibilities of using the so-called ac-
placements and/or accelerations is required. Reduction of tive and semi-active versions of TMD are also considered
accelerations is important due to undesired influences, in a number of papers (Han, Li 2006; Li, Han 2007; Li,
exerted not only upon the building structure, but also on Zhu 2007; Lin et al. 2005).
people inside. The problems of TMD analysis and de- The practical application of TMD on an extremely
signing are still present in scientific papers. For example, high telecommunication tower is described in a paper
in his excellent paper Krenk (2005) derived a new for- (Ghorbani-Tanha et al. 2008).
mula for the TMD optimal damping coefficient. More- Up to now, reduction of vibration of structures with
over, in paper (Leung et al. 2008) used the particle swarm MTMD caused by earthquake forces are mainly investi-
optimization method to optimise the TDM parameters in gated. The analysis of such type of structures under wind
the case of non-stationary excited structures. Optimiza- loads are rare (Kareem, Kline 1995) and the dynamic
tion of TMD parameters is also the subject of paper behaviour of structures with MTMD are not fully under-
(Singh et al. 2002). stand. For this reasons, in the present paper, the possibil-
In the 90’s, studies on the application of multiple ity to reduce the vibration of a frame structure with the
tuned mass dampers (MTMD) for one-degree of freedom help of MTMD is analysed. The presented description of
systems were started (Xu, Igusa 1992; Igusa, Xu 1994). It the structure with MTMD exploits a particular form of
has been proved that MTMD with distributed natural fre- the motion equations to simplify the numerical algorithm
quencies are more effective than TMD. The studies of of the applied method of solution. The structure is under
MTMD were also developed in (Kareem, Klime 1995; the effect of dynamic forces caused by wind pressure.
Jangid 1995). Later on, structures subjected to seismic Wind velocities are treated as random and ergodic proc-
loads, treated as a multi degree of freedom structures and esses. The spectral density functions of wind velocity
with the MTMD on them were analysed in (Chen, Wu fluctuations are assumed as proposed by Davenport.

JOURNAL OF CIVIL ENGINEERING AND MANAGEMENT ISSN 1392–3730 print / ISSN 1822–3605 online 77
http:/www.jcem.vgtu.lt DOI: 10.3846/1392-3730.2009.15.77-86
78 R. Lewandowski, J. Grzymisławska. Dynamic analysis of structures with multiple tuned mass dampers

Some calculations were made for a 20-story building and degrees of freedom. The fluctuations of wind velocity
on this basis the effectiveness of MTMD was estimated. forces are a load to the frame, and these forces are ap-
The effects of detuning of structure parameters are also plied at the building floor levels (Fig. 1).
presented. In this case, the reduction of accelerations of
structures with MTMD is noticeably greater than the x (t)
N1
x (t)
N2
x
NK
(t)
structure without or with TMD.
mN1 mN2 mNK
2. Designing of multiple tuned mass dampers (MTMD)
yN(t)
...
PN(t)
The aim of designing MTMD is to tune damper parame- MN
ters to the modal parameters of selected modes of vibra- x (t)
31
x (t)
32
x (t)
3K
KN , CN
tion. It means that the natural damper frequency (or a
group of dampers) ωd must be close to the natural fre- m31 m32 m3K
quency of a selected vibration mode of structure ω s
(ωd ≈ ωs ) . Moreover, the damping factor of the damper P3(t)
...
y3 (t)
M3
must be appropriately chosen.
x (t) x (t) (t)
The optimal parameters of such a damper (or group 21 22
X
2K

of dampers) can be determined from the formulae given K3 , C 3

in a paper (Warburton 1982). The optimal frequency ratio m21 m22 m2K

is determined from: P2(t)

...
y2 (t)
M2
ω 2d 2+µ
= , (1)
ω 2s 2(1 + µ )2
x (t) x (t) X (t)
11 12 1K

where m11 m12 m1K K 2 , C2

m K k
µ = d , ω 2s = s , ω d2 = d . (2)
Ms md P1(t) y1 (t)
...
Ms
M1
Here M s and K s is the modal mass of the structure and the
modal stiffness of the s-th mode of vibration, respectively. K1 , C1
If only a single damper is tuned to the s-th mode of
vibration with frequency ω s , then m d is the mass of the
damper, and k d is the stiffness coefficient of the damper.
However, if a group of dampers are designed to tune to Fig. 1. The model of structure with MTMD
the frequency ω s , then m d and k d denote the mass and
the stiffness coefficients of the selected damper of this A set of mass dampers are mounted on the structure.
group, respectively. A model configuration of dampers is also shown in Fig. 1,
Assuming that the mass ratio µ is known, the dam- while in Fig. 2 the scheme of a typical mass damper is
per frequency ωd and the damper stiffness coefficient presented. The concept of a group of dampers is introduced
k d can be obtained from the above formulae. in this paper. Each group of dampers consists of a few
If excitation forces acting on a structure, have a ran- dampers. Each damper in a particular group of dampers
dom character and can be treated as white-noise excita- can be installed on different floors and may have different
tion, the optimal value of non-dimensional damping coef- mass, stiffness and damping parameters. However, all
ficient is determined from the formula (McNamara 1977; dampers in the group are designed in such a way that they
Warburton 1982): are tuned to a particular mode of vibration. A special nota-
µ (4 + 3µ ) tion described below and concerning dampers is introdu-
γ opt = . (3) ced. The symbols x ij (t ) , mij , k ij and c ij denote, respec-
8(1 + µ )(2 + µ )
tively, the damper displacement, damper mass, stiffness
The value of the damping coefficient cd can be cal- and damping factor of the damper which belongs to the j-th
culated from the relation group and is located on the i-th floor (Fig. 2).
cd = 2 γ opt ω d md . (4) The equation of motion of the system shown in
Fig. 1 and briefly described above can be written in the
Using the above formulae, the parameters of following form:
MTMD can be determined.
Mq&&(t ) + Cq& (t ) + Kq(t ) = P (t ) , (5)
~ ~ ~ ~
3. Equation of motion where M, C, K are the global matrices of mass, damp-
~ ~ ~
The building structure is treated as a discrete, linear elas- ing and stiffness of the considered system (i.e. the struc-
tic system. The frame in Fig. 1 is the model of the build- ture and MTMD), respectively, q(t ) = col (y (t ), x(t )) is
ing structure. The mass of it is concentrated at the level of the vector of displacements of the system, y (t ) – the
building floors and the beams of the frame are infinitely vector of horizontal displacements of frame, and x(t ) –
stiff. Horizontal displacements of floors are the dynamic the vector of horizontal displacements of dampers. More-
Journal of Civil Engineering and Management, 2009, 15(1): 77–86 79

~
over, P (t ) = col (P(t ), 0 ) and P (t ) is the vector of exci- As mentioned above, the symbol k ij denotes the stiffness
tation forces acting upon the structure. coefficient of a damper of the j-th group which is located
x (t) on the i-th floor (Fig. 2).
The k block of the matrix K is the diagonal matrix
ij ~
and in the following form:
mij k = diag[k11, k12 ,..., k1K , k 21, k 22, ...
k c
ij ij ...k 2 K , k31, k32 ,...k3K ,... .
...k N1, k N 2 ,...k NK ].
Fig. 2. Diagram of damper
The damping matrix of the system C is in a form
~
The theory presented below could be applied to the similar to that of the stiffness matrix K . The specific
~
non-proportionally damped structures. However, in the blocks of this matrix are defined below:
paper we assume that the structure is proportionally dam-
ped, i.e. the damping matrix of the structure is in the fol- ~ C + c c ∗  C11 C12 
C =  ∗T 1 = , (6c)
 c c  C 21 C 22 
lowing form: C = α M + κ K .
The M matrix of the system is in the following
~
form (Fig. 1): C1 + C 2 − C2 0 0 
 
~ M 0  M 11 0  − C2 C 2 + C3 − C3
C=
0 
M= = M 22 
, (6a)
 0 − C3 − CN 
 0 m  0 
...

,

where  0 0 − CN C N 
M = diag[M 1 , M 2 , M 3 ,..., M N ] , c1 = diag[c11 + c12 + ... + c1K , c 21 + c 22 + ... + c 2 K ,
m = diag[m11, m12 ,..., m1K , m21, m22, ... c 31 + c 32 + ... + c 3 K ,..., c N 1 + c N 2 + ....
...m2 K , m31, m32 ,...m3K ,...mN 1, m N 2 ,...mNK ]. ... + c NK ]
In the above formula, M is the mass matrix of the − c11 − c12 ... − c1K 0 0 0 0
structure and m – the mass matrix of the dampers. The  0 − c21 − c22 ... c2 K
symbol mij denotes the mass damper of the j -th group  0 0 0
c = 0
*
0 0 0 0 0 0 0
located on the i -th floor. 
 0 0 0 0 0 0 0 0
The stiffness matrix K of the considered system  0
~
0 0 0 0 0 0 0
can also be shown in the block form written below:
0 
~ K + k k ∗  K 11 K 12 
0 0 0 0 0 0 0 0
K =  ∗T 1 = , (6b) 0 0 0 0 0 0 0 0 0 
 k k  K 21 K 22 
− c31 − c32 ... − c3K 0 0 0 0 0 
where K is the stiffness matrix of the structure 
0 0 0 0 ... 0 0 0 0 
K1 + K 2 − K2 0 0  0 0 0 0 0 − cN 1 − c N 2 ... − c NK 
 −K K2 + K3 − K3 0 
,
K= 2
.
 0 − K3 ... − KN  where C i = α M i + κ K i and C i = κ K i .
 
 0 0 − KN KN  In the above formulae the symbol c ij denotes the
damping coefficient of the damper of the j -th group
The block matrices k 1 and k * are in the following form: which is located on the i -th floor (Fig. 2). The block c
of the C matrix is the diagonal matrix and has the fol-
~
k1 = diag[k11 + k12 + ... + k1K , k 21 + k 22 + ... + k 2 K ,
lowing form:
k31 + k32 + ... + k3K ,..., k N1 + k N 2 + .... + k NK ]
c = diag[c11 , c12 ,..., c1K , c 21 , c 22, ...
− k11 − k12 ... − k1K 0 0 0 0
 ...c 2 K , c31 , c32 ,...c3 K ,...
 0 0 0 0 − k 21 − k 22 ... k 2 K
...c N 1 , c N 2 ,...c NK ]
* 
k = 0 0 0 0 0 0 0 0
 Taking into account that the matrices of mass, stiff-
 0 0 0 0 0 0 0 0 ness and damping are in the form (6), the equation of
 0 motion (5) can be rewritten in the following block matrix
 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0  form:

0 0 0 0 0 0 0 0 0  M11&y&(t) + C11y& (t ) + C12x& (t ) + K11y(t )
 (7a)
− k 31 − k 32 ... − k 3K 0 0 0 0 0 . + K12x(t ) = P(t ) ,

0 0 0 0 ... 0 0 0 0  M 22&x&(t ) + C 21y& (t ) + C 22 x& (t ) +
... − k NK 
(7b)
0 0 0 0 0 − k N1 − kN2 K 21y (t ) + K 22 x(t ) = 0 .
80 R. Lewandowski, J. Grzymisławska. Dynamic analysis of structures with multiple tuned mass dampers

4. Modelling of wind loads In this paper, the spectral density function proposed
by Davenport (Simiu, Scanlan 1996; Dyrbye, Hansen
Wind speed acting on a structure consists of along-wind 1999; Holmes 1997) is used. The elements of the matrix
and cross-wind components, and it varies randomly in S u (λ ) are calculated from the formula:
time and space (Simiu, Scanlan 1996; Dyrbye, Hansen
1999; Holmes 1997). A complete wind velocity field u
Slk = Sllu (λ )S kk
u
(λ ) e −Φ = Su (λ, zl )Su (λ, z k ) e −Φ . (11)
should be modelled as a two-dimensional, multivariate
stochastic process. Usually, the wind speed is treated as a where Su (λ, zl ) and Su (λ, z k ) are the elements taken
stationary Gaussian stochastic process (Simiu, Scanlan from the main diagonal of the Su (λ ) matrix. They are
1996; Dyrbye, Hansen 1999). The wind speed U ( z , t ) is calculated with the help of the spectral density function
assumed to be the sum of a steady part U (z ) and a super- for the particular stories.
imposed random fluctuation of wind velocities u ( z , t ) , i.e. The diagonal elements of the matrix spectral density
U ( z, t ) = U ( z ) + u( z, t ) . function of wind velocity fluctuations Su (λ, zi ) are cal-
culated using the spectral density function as proposed by
The random fluctuation of wind velocity u ( z , t ) is a Davenport (Simiu, Scanlan 1996; Dtrbye, Hansen 1999;
zero-mean stationary Gaussian process with a known Holmes 1997)
correlation function.
The along wind speed described above is a stochas- 4u*2 f 2 (n)
Su (n) = , (12)
tic process that is continuous in space and time. When n[1 + f 2 (n)]4 / 3
high buildings are considered it is necessary to introduce where
some simplifications, to replace the continuous space and
1200n
time random function u ( z , t ) with a set of functions f (n) = , (13)
u i (t ) which depend on time only. The building is divided U (10)
into N section along its height. It is assumed that the wind and n denotes frequency in Hz.
speed does not vary along the section. The typical mid- The mean wind velocity acting at the level of the
point of the section is chosen at a structure storey level. It i-th floor can be calculated from formula:
means that wind velocity fluctuations u ( z , t ) can be re- z 
placed by a set of zero-mean stationary processes u i (t ) , U i ( z i ) = 2,5u * ln  i ,
 (14)
 z0 
where i =1, 2, ..., N . Thus, the wind force in the midpoint
of an arbitrarily chosen structure section can be described where
in the following way: u* = U (10) k . (15)
Pi (t ) = C A AρU i Χ 2ui (t ) , (8) In relationships (14) and (15), U (10) is the mean wind
velocity at the altitude of 10 m, k – the coefficient de-
where C A is the aerodynamic drag coefficient, A – the pended on type of area, z 0 – the roughness length and
wind-exposed area and the ρ symbol denotes the air the symbol z i denotes the altitude of the i-th floor over
density. ground.
The admittance function Χ describes the influence The Φ symbol denotes the correlation coefficient,
of the building on wind pressure forces and it is always which takes into consideration spatial correlations of the
that Χ ≤ 1 . According to Holmes (1997), the admittance fluctuations of wind velocity. According to monographs
function is connected with a correlation coefficient Φ , (Li, Zhu 2007; Dyrbye, Hansen 1999), this coefficient
which is used to determine the matrix elements of the can be determined from the formula:
spectral density function. It is troublesome to determine
the admittance function. Moreover, this value is unknown 2λ C z z l − z k
Φ= , (16)
in many cases. Therefore, quite often, and also in this U ( zl ) + U ( z k )
paper, it is assumed that Χ = 1 . where λ is the force frequency, whereas C z is the em-
For multi-degree-of-freedom systems the correlation pirical constant. The symbol U ( z l ) is the mean wind
matrix of the fluctuations of wind velocities is formulated velocity at the level of the l-th story. If it is assumed that
as: the fluctuations of wind velocity are totally correlated,
R u (τ) = E[u uT ] , (9) then e −Φ = 1 while, if the correlation is disregarded, the
matrix Su (λ ) is the diagonal one.
where u = col{u1 (t ), u2 (t ),..., ui (t ),...., u N (t )} is the vector The correlation matrix of the forces excited by the
of the fluctuations of wind velocity and the symbol E [⋅] wind pressure can be written in the following form:
denotes the expected value of [⋅] .
Using the Fourier transform, the following expres- R p ( τ ) = E[ P P T ] , (17)
sion of the spectral density function of wind velocity
fluctuations is obtained: where P = col {P1 (t ), P2 (t ), ..., Pi (t ), ...., PN (t )} is the vec-
tor of wind forces acting upon the structure.
1
∫ R u ( τ)e
+∞
S u (λ ) = − iλ τ
dτ . (10) Using relationships (8) and (9), the elements R Pi P j
2π −∞ of the R p (τ) matrix can be written as
Journal of Civil Engineering and Management, 2009, 15(1): 77–86 81

RPi Pj ( τ) = (C A A ρ Χ ) 2U iU j Rui u j ( τ) . (18) where the S P (λ) is a matrix of which the elements are
given by formula (19).
The spectral density matrix of excitation forces After inserting τ = 0 into Eq (25) the correlation
S p ( λ ) and the correlation matrix R p (τ) are interre- matrix of displacements R q (0 ) is obtained and, on this
lated in such a way that basis, the root mean square of displacements can be de-
1 +∞ termined from
∫ R p ( τ )e
− iλτ
S p (λ) = d τ,
R q (0) = ∫ S q (λ )dλ ,
+∞
2π −∞ (28)
which means that the element S Pi P j of the S p ( λ ) ma- −∞

trix can be written in the following form: where

S yy S yx  ~ ~T
S Pi P j ( τ ) = (C A A ρ Χ ) 2U iU j Sui u j ( τ) . (19) S q (λ ) =   = H (λ )S ~p (λ ) H (λ ) . (29)
S xy S xx 
This is the relationship between the spectral density The integral appearing in Eq (29) can be calculated
matrix of wind loads acting upon the structure and the numerically.
spectral density matrix of the fluctuations of wind velocity.
The calculation of the S q (λ) matrix is substantially
5. Solution to equation of motion simplified if we take into account the structure of the
~
S P~ (λ) and H (λ) matrices.
A solution to the equation of motion (5) that fulfills the ~
After writing the H (λ) matrix in the following
precondition: t = 0 , q& (t ) = 0, q(t ) = 0 , can be written in block form:
the following form:
~ H (λ) H12 (λ) 
H (λ) =  11  (30)
H 21(λ) H 22 (λ)
t ~
q(t ) = ∫ h(t − τ ) P( τ) dτ , (20)
0 and introducing Eq (27) and (30) into (29) we obtain:
where the symbol h(t − τ) denotes the matrix of impulse T
~ S yy (λ) = H11 (λ)S p (λ) H11 (λ) , (31)
transfer function and P ( τ) = col (P( τ), 0) is the vector of
excitation forces appearing in Eq (5). S xx (λ) = H 21 (λ)S p (λ) H T21 (λ) , (32)
Because random loads acting upon the structure are
stationary processes, then also dynamic responses of the S yx (λ) = H11 (λ)S p (λ) H T21 (λ) , (33)
system are a stationary process. Thus, the correlation
matrix of the structure responses can be written as: T
S xy (λ) = H 21 (λ)S p (λ) H11 (λ) . (34)
T
R q (t1, t 2 ) = E[q(t1 ), q (t2 )] . (21) It is easy to observe that the root mean square of
By substituting (20) into (21), we obtain: structure displacements and the root mean square of
dampers displacements can be calculated from the fol-
R q (t1 , t 2 ) = lowing relationships:

R y (0) = ∫ S yy (λ )dλ ,
t1 t 2 +∞
~ ~T
∫ ∫ h(t1 − τ 1) E[P( τ1)P ( τ 2 )] h (t 2 − τ 2 )dτ1dτ 2 . (35)
T
(22)
00 −∞
Taking into account that
R x (0 ) = ∫ S xx (λ )dλ ,
+∞
(36)
R ~p ( τ ) = ∫ S ~p ( λ ) e
+∞
iλτ
dλ , (23)
−∞

−∞ respectively. It means that, in fact, only the S yy (λ) and

S xx (λ ) matrices must be calculated.
~ ~
E[P(t ), P (t − τ)] = ∫ S ~p (λ ) eiλτ dτ .
+∞
(24) Now, the matrices H11(λ ) and H 21(λ ) , which are
~
−∞ blocks of matrix H (λ) , must be determined. This can be
Eq (22) can be rewritten in the form: done by assuming the excitation and the solution to
+∞~ ~ Eq (7) in the form:
R q (τ ) = ∫ H(λ ) S ~p (λ ) H T (λ ) eiλτ dτ , (25) P (t ) = I exp(iλt ) , (37)
−∞
~ y (t ) = H11(λ) exp(iλt ) , (38)
where H is the matrix conjugate to the H matrix de-
~
fined below x(t ) = H 21(λ ) exp(iλt ) . (39)
~ ~ ~
( ~ −1
H (λ ) = K − λ 2 M + i λ C . (26) ) where I denotes the identity matrix.
After introducing Eq (37–39) into Eq (7) we obtain:
Moreover,
S (λ ) 0  (K11 − λ 2M11 + iλC11) H11 +
S P~ (λ ) =  P , (27)
 0 0 (K12 + iλC12 ) H 21 = I , (40)
82 R. Lewandowski, J. Grzymisławska. Dynamic analysis of structures with multiple tuned mass dampers

(K 22 − λ 2M 22 + iλC 22 ) H 21 +
(K 21 + iλC 21 ) H11 = 0 . (41)
From Eq (41) it follows that
H 21 = G (λ ) (K 21 + iλC21) H11 , (42)
where the matrix
G = −(K 22 − λ 2M 22 + iλ C 22 ) −1 , (43)
is easy to calculate because the matrices K 22 , M 22 and
C 22 are diagonal.
After introducing Eq (42) into (40) we obtain
[
H11 = (K11 − λ 2M11 + iλC11) +
(44)
(K12 + iλC12 )G (λ) (K 21 + iλC 21 )]−1.

6. Results of exemplary calculations

In this section, the results of dynamic analysis of the ex-
emplary structure with MTMD are discussed. Addition-
ally, for comparison, results for the structure with only
one tuned mass damper (TMD), which is tuned to the
first vibration mode of structure will be presented. The
above-mentioned TMD is located on the top floor of the
structure.
The building parameters were calculated on the ba-
sis of paper (Spencer et al.) and they are given in Table 1.

Table 1. Main parameters of structure

Story Mass [kg] Stiffness [N/m]

1 2.83×10 5
3.31×108
2–4 2.76×105 1.06×109
4–7 2.76×105 6.79×108
8–10 2.76×105 6.79×108
11–13 2.76×105 5.84×108
14–16 2.76×105 3.86×108
17–19 2.76×105 3.47×108
20 2.92×10 5
2.29×108

Damper parameters were designed using Formulae

(1–4) and assuming that these parameters tune dampers to
the structure’s first three modes of vibration. The shapes of
the first, second and third mode of vibration are shown in
Fig. 3. In this case, it has been assumed that 3 groups of
dampers are installed on the structure. Each group of
dampers consists of one damper only. All dampers are
located on the top floor. The damper parameters and their
locations on the structures are given in Table 2. The total
mass of MTMD is nearly equal (by 4.4% smaller) to the
mass of TMD.
Moreover, the values of non-dimensional damping
coefficients of the first and second vibration modes are
equal to 1% of critical damping.

Fig. 3. First 3 vibration modes of structure

Journal of Civil Engineering and Management, 2009, 15(1): 77–86 83

The following values of parameters appearing in As it was mentioned previously, the results concern-
Relationships (25–28) are chosen: z 0 = 0,3 , ing accelerations were elaborated (Fig. 5). It has been
ρ = 1.226 kg/m 3 , U (10) = 30 m/s 2 , k = 12 ⋅10 −3 . observed that, when using MTMD acceleration, reduction
is bigger only below the 11th floor than when using
Table 2. Parameters of dampers TMD. Above the 11th floor, the observed reduction of
accelerations is smaller, compared with TMD. The
Number of mode/ placement Mass [kg] Stiffness [N/m] maximum root mean squares of acceleration (top floor)
are almost equal. The total sum of root mean square of
TMD acceleration is 38% for MTMD and 40% for TMD, com-
1/20 36214 472468 pared with the structure without dampers.
MTMD The sensitivity of both TMD and MTDM with re-
1/20 18107 238870 spect to change of structure parameters is also investi-
2/20 7956 722685 gated. Calculations are made for a structure for which the
values of all masses and all stiffness coefficients change
3/20 8550 2182386
by ±10% , but the parameters of TMD and MTMD are
Because the dynamic response of the structure is a kept constant. The above-mentioned changes of structure
stationary and ergodic random process, the root mean parameters reflect some possible uncertainties connected,
square of freely chosen displacement q i and acceleration for example, with determining the properties of structural
material and/or with errors which are introduced when
q&&i could be calculated from the following formulae:
the theoretical model of structure is chosen. All these
∫ Sii (λ )dλ , σ q&&i = ∫ λ Sii (λ )dλ ,
irregularities lead to the so-called detuning of dampers.
+∞ +∞
σ 2q = q 2 4 q
(45)
i
−∞ −∞

where S iiq is the diagonal element of the S q (λ) matrix.

Using the above formulae, an analysis of the struc-
ture without dampers, with installed conventional TMD
and with MTMD was made. The results of the analysis
are shown in Figs 4 and 5. In Fig. 4 the root mean square
of structure displacements is shown. It has been observed
that displacements reduction with MTMD installed is a
little smaller than in the case of TMD installed on the
structure. Compared with the structure without dampers,
the maximum reduction of root mean square of structure
displacements (top floor) is 30% for TMD and 25% for
MTMD, respectively.

root mean squares of accelerations [m/s2]

Fig. 5. Root mean squares of accelerations

Figs 6–8 illustrate the effects of such detuning of

dampers for structures of which the stiffness increases by
10% . In Fig. 6, the resonance curves are presented. The
thin solid line shows results for structures without damp-
ers, the dashed line shows the response curve for the
structure with TMD, while the thick line presents results
for structures with MTMD. In a similar way, in Figs 7, 8,
the root mean squares of displacements and accelerations
are presented, respectively. It is obvious that now MTMD
reduce both displacements and accelerations to a greater
extent than TMD. Similar trends are observed when the
root mean squares of displacements [m] structure stiffness decreases and when the mass of struc-
tures increases or decreases. The quantitative information
Fig. 4. Root mean squares of displacements concerning the effectiveness of TMD and MTMD con-
cerning the effects of detuning of structure parameters is
given in Table 3.
84 R. Lewandowski, J. Grzymisławska. Dynamic analysis of structures with multiple tuned mass dampers

Table 3. Reduction effects for structures with changed

parameters

Reduction of top displacement – mass changes

Original structure +10% MS –10% MS
TMD 18 % 20% 24%
MTMD 16 % 26% 27%
Reduction of top displacement – stiffness changes
Original structure +10% KS –10% KS
TMD 18% 24% 20%
MTMD 16% 27% 26%
Reduction of top acceleration – mass changes
Original structure +10% MS –10% MS
TMD 28% 26% 37%
MTMD 23% 32% 45%
Reduction of top acceleration – stiffness changes
Original structure +10% KS –10% KS
TMD 28% 37% 33%
MTMD 23% 43% 42% root mean squares of displacements [m]

Fig. 7. Root mean squares of displacements – structures

with changed stiffness

frequency [rad/s]

Fig. 6. The response curve of top of structure – structure

with changed stiffness root mean squares of accelerations [m/s2]

Fig. 8. Root mean squares of accelerations – structure

7. Concluding remarks with changed stiffness
The analysis of vibrations of a building structure with
MTMD installed, which are tuned to selected modes of • In the case where parameters of structures are not
vibration, has been studied in this paper. The root mean exactly known, MTMD reduce better both dynamics
squares of displacement and accelerations of a structure displacements and accelerations of structures than
with MTMD were determined. These calculations were do TMD. This is the main advantage of MTMD.
compared with the root mean squares of displacement • MTMD reduce both displacements and accelera-
and acceleration of the same structure with conventional tions of structures to a similar extent.
TMD installed. • MTMD reduce accelerations on lower floors of
The following conclusions could be formulated structures to a greater extent, compared with TMD.
from the results of calculations:
Journal of Civil Engineering and Management, 2009, 15(1): 77–86 85

These are the first results of calculation and, there- Kareem, A.; Kline, S. 1995. Performance of multiple tuned
fore, the above conclusions cannot be treated as defini- mass dampers under random loading, Journal of Struc-
tive. Generally speaking, the effectiveness of MTMD and tural Engineering 121: 348–361.
TMD are similar. However, MTMD are smaller than Krenk, K. 2005. Frequency analysis of the tuned mass dampers,
conventional TMD and they occupy a much smaller spa- Journal of Applied Mechanics, Transactions of ASME 72:
936–942.
ce for installation.
Moreover, the acceleration reduction of structures Leung, A. Y. T.; Zhang, H.; Cheng, C. C.; Lee, Y. Y. 2008.
Particle swarm optimization of TMD by non-stationary
with MTMD is noticeably greater in comparison with
base excitation during earthquake, Earthquake Engineer-
structures without or with TMD and when the values of ing and Structural Dynamics (in press).
structure parameters are not exactly known. The problem
Li, C.; Han, B. 2007. Control strategy of the lever-type active
of detuning the dampers parameters needs further investi- multiple mass dampers for structures, Wind and Struc-
gations. In particular, parameters of structures must be tures 10: 301–314.
regarded as the random quantities. Li, C.; Zhu, B. 2007. Investigation of response of systems with
active multiple tuned mass dampers, Structural Control
Acknowledgments and Health Monitoring 14: 1138–1154
Lin, P. Y.; Chung, L. L.; Loh, Ch. 2005. Semiactive control of
The authors acknowledge the financial support received
building structures with semiactive tuned mass dampers,
from the Poznan University of Technology (Grant Computer-Aided Civil and Infrastructure Engineering 20:
No. BW. 11-008/08) is connection with this work. 35–51.
Li, C.; Qu, W. 2006. Optimum properties of multiple tuned
References mass dampers for reduction of translational and torsional
Chen, G.; Wu, J. 2001. Optimal placement of multiple tune response of structures subject to ground acceleration, En-
mass dampers for seismic structures, Journal of Structural gineering Structures 28: 472–494.
Engineering 127: 1054–1062. Li, C. 2002. Optimum multiple tuned mass dampers for struc-
Chmielewski, T. 1982. Metody probabilistyczne w dynamice tures under the ground acceleration based on DDMF and
konstrukcji. Wyd. Wyższej Szkoły Inżynierskiej w Opolu. ADMF, Earthquake Engineering and Structural Dynam-
ics 31: 897–919.
Dyrbye, C.; Hansen, S. O. 1999. Wind Loads on Structures. 3rd
edition. John Wiley and Sons, Inc. McNamara, R. J. 1977. Tuned mass dampers for buildings,
Journal of the Structural Engineering. Division Proc.,
Ghorbani-Tanha, A. K.; Noorzad, A.; Rahimian, M. 2008. Miti-
ASCE 105: 1785–1798.
gation of wind-induced motion of Milad Tower by tuned
mass damper, The Structural Design of Tall and Special Singh, M. P.; Singh, S.; Moreschi, L. M. 2002. Tuned mass
Buildings (in press). dampers for response control of torsional buildings,
Earthquake Engineering and Structural Dynamics 31:
Guo, Y. Q.; Chen, W. Q. 2007. Dynamic analysis of space
749–769.
structures with multiple tuned mass dampers, Engineering
Structures 29: 3390–3403. Spencer, B. F. Jr.; Christenson, R. E.; Dyke, S. J. 1999. Next
Generation Benchmark Problem for Seismically Excited
Han, B. K.; Li, C. X. 2006. Evaluation of multiple dual tuned
Buildings. Available from Internet: <http:/cee.uiuc.edu/
mass dampers for structures under harmonic ground exci-
sstl/papers/NGbench.pdf>.
tation, International Journal of Structural Stability and
Dynamics 6: 59–75. Simiu, E.; Scanlan, R. 1996. Wind effects on structures, Fun-
damentals and applications to design. John Wiley and
Han, B.; Li, C. 2006. Seismic response of controlled structures
Sons, Inc.
with active multiple tuned mass dampers, Earthquake En-
gineering and Engineering Vibration 5: 205–213. Warburton, G. B. 1982. Optimum absorber parameters for vari-
ous combinations of response and excitation parameters,
Holmes, J. D. 1997. Equivalent time averaging in wind engi-
Earthquake Engineering and Structural Dynamics 10:
neering, Journal of Wind Engineering and Industrial
381–401.
Aerodynamics 72: 411–419.
Xu, Y. L.; Kwok, K. C. S.; Samali, B. 1992. Control of wind-
Igusa, T.; Xu, K. 1994. Vibration control using multiple tuned
induced tall building vibration by tuned mass dampers,
mass damper, Journal of Sound and Vibration 175: 491–
Journal of Wind Engineering and Industrial Aerodynam-
503.
ics 22: 833–854.
Jangid, R. S. 1995. Dynamic characteristic of structures with
Xu, K.; Igusa, T. 1992. Dynamic characteristic of multiple
multiple tuned mass dampers, Structural Engineering Me-
substructures with closely-spaced frequencies, Earth-
chanics 3: 497–509.
quake Engineering and Structural Dynamics 21: 1059–
1070.
 86 R. Lewandowski, J. Grzymisławska. Dynamic analysis of structures with multiple tuned mass dampers

KONSTRUKCIJŲ SU KELIAIS MASĖS SLOPINTUVAIS DINAMINĖ ANALIZĖ

R. Lewandowski, J. Grzymisławska
Santrauka
Nagrinėjama galimybė sumažinti stipraus vėjo veikiamų rėminių pastatų konstrukcijų svyravimus, taikant masės
slopintuvų sistemą. Vibracijas sukeliančios jėgos, priklausančios nuo vėjo greičio svyravimų, laikomos atsitik-
tiniais dydžiais. Vėjo greičio svyravimo spektro tankio funkcijos nagrinėjamos Davenport metodu. Atlikus atsitik-
tinių vibracijų regresinę analizę, nustatytos poslinkių ir pagreičių vidutinės kvadratinės paklaidos. Remiantis
skaičiavimo rezultatais, padarytos masės slopintuvų efektyvumo išvados.
Reikšminiai žodžiai: svyravimų mažinimas, atsitiktiniai svyravimai, masės slopintuvai.

Roman LEWANDOWSKI. Professor at the Poznan University of Technology, Faculty of Civil and Environ-
mental Engineering, Poznan, Poland. Member of the Polish Society for Theoretical and Applied Mechanics and
Polish Association for Computational Mechanics. Main research interests: non-linear vibration of structures, re-
duction of vibration by means of passive, active and semi-active methods.
Justyna GRZYMISŁAWSKA. Teaching and research assistant at the Poznan University of Technology, Faculty
of Civil and Environmental Engineering, Poznan, Poland. She teaches the strength of materials. Main research in-
terests: passive, active and semi-active methods of reduction of structure vibration.

View publication stats