You are on page 1of 13

Mechanical Systems and Signal Processing 134 (2019) 106337

Contents lists available at ScienceDirect

Mechanical Systems and Signal Processing


journal homepage: www.elsevier.com/locate/ymssp

Simplified analytical solution for the optimal design of Tuned


Mass Damper Inerter for base isolated structures
A. Di Matteo a,⇑, C. Masnata a, A. Pirrotta a,b
a
Dipartimento di Ingegneria, Università degli Studi di Palermo, Viale delle Scienze, 90128 Palermo, Italy
b
Department of Mathematical Sciences, University of Liverpool, Liverpool, UK

a r t i c l e i n f o a b s t r a c t

Article history: In this paper the use of the Tuned Mass Damper Inerter (TMDI) to control the response of
Received 30 May 2019 base isolated structures under stochastic horizontal base acceleration is examined. Nota-
Received in revised form 28 August 2019 bly, the TMDI, recently introduced as a generalization of the classical Tuned Mass Damper,
Accepted 29 August 2019
allows to achieve enhanced performance compared to the other passive vibration control
Available online 26 September 2019
devices. Thus, it represents an ideal alternative for reducing displacements of base isolated
structures. To this aim, firstly a straightforward numerical approach is developed for the
Keywords:
optimal design of this device considering a white noise base excitation. Further, a simpli-
Base-isolation system
Tuned Mass Damper
fied analytical solution for the optimal design of TMDI parameters for base isolated struc-
Inerter tures is proposed minimizing the displacement variance of the corresponding undamped
Optimal design base isolated system. A thorough numerical analysis is performed and related results, in
terms of optimal parameters and control performance, are compared with pertinent data
obtained by a more computationally demanding iterative optimization procedure on the
original damped system, considering both white noise and coloured noise stationary base
excitation. Analytical and numerical results are found in good agreement, especially in
terms of control performance, thus establishing the reliability and efficiency of the pro-
posed approach. Finally, numerical analyses on a five-story benchmark base isolated struc-
ture controlled with an optimally designed TMDI are performed considering real recorded
ground motions as base excitation. It is concluded that the TMDI, properly optimized with
the proposed procedure, can effectively reduce the response of base isolated structures
even under strong earthquakes.
Ó 2019 Elsevier Ltd. All rights reserved.

1. Introduction

Base isolation is nowadays a common and effective method adopted in several type of structures for their protection from
potential damages induced by severe seismic excitations [1]. To this aim, a number of different isolators can be employed,
which all commonly possess low horizontal and high vertical stiffness, allowing the main structure to behave almost like a
rigid body when isolators are appropriately inserted between the main structure to be protected and its foundation. In fact,
in this manner the majority of the displacement occurs within the base isolation sub-system, while displacements and accel-
erations of the main structure are greatly reduced [2].

⇑ Corresponding author.
E-mail addresses: alberto.dimatteo@unipa.it (A. Di Matteo), chiara.masnata@unipa.it (C. Masnata), antonina.pirrotta@unipa.it (A. Pirrotta).

https://doi.org/10.1016/j.ymssp.2019.106337
0888-3270/Ó 2019 Elsevier Ltd. All rights reserved.
2 A. Di Matteo et al. / Mechanical Systems and Signal Processing 134 (2019) 106337

Although these beneficial features render base isolation extensively employed, some detrimental effects need to be taken
into account. Specifically, the base isolation sub-system often undergoes large and undesirable displacements which require
particular attention during the design process. In this regard, one possible strategy for the reduction of this displacement
demand is related to the use of highly damped isolators, even providing additional damping for instance resorting to external
dampers. It is worth noting, however, that this approach leads to a reduction of the base isolation sub-system displacement
at the expense of increasing accelerations and interstorey drifts of the main structure [3].
As an alternative and effective strategy, several researches have recently focused on the used of passive control systems
for the seismic response control of base isolated structures. In this regard, the majority of these studies have analysed the
control performances of the classical Tuned Mass Damper (TMD) connected to the base isolation sub-system [4,5]. Some
approaches for the optimal design of these systems have been presented for both base excitation [5,6] and wind type of load-
ing [7]. Generally, results have shown that the TMD is very effective when low damping isolators are used [8], leading to a
reduction of the displacement demand of the base isolation sub-system, simultaneously preserving small interstorey drifts of
the main structure [9]. On this base, some studies have also considered the use of other passive control devices, such as
Tuned Liquid Damper [10], and Tuned Liquid Column Damper (TLCD) [11], on base isolated structures. In this regard, recent
experimental studies [12] have also proved the TLCD control performances.
In this context, a promising novel passive control mechanism is related to the combined use of the classical TMD with the
so-called inerter [13]. Several typologies of inerter are currently under investigation [14–16], but in essence it is a two-
terminal device whose resisting force is directly proportional to the relative acceleration between its two terminals, and
its constant of proportionality, called inertance, has the dimension of a mass. When placed in parallel with the spring and
the damper of a classical TMD, it constitutes a hybrid device generally referred to as Tuned Mass Damper Inerter (TMDI).
The main characteristic of the TMDI mostly pertains to the attractive features of the inerter itself that acts as an additional,
virtual mass for the system to which it is connected. Specifically, the inerter is able of generating a sort of mass amplification
effect, making the TMDI behaving like a TMD with an apparent mass hundreds of times higher than its real physical mass, to
achieve enhanced performance compared to the classical TMD. Clearly, the original TMD can be retrieved setting the iner-
tance equal to zero.
On this base, several papers have been focused on the analysis of the dynamic response of TMDI controlled systems in
mechanical and civil engineering applications, see for instance [17–20,27–28] and references therein. The encouraging out-
comes have recently also motivated the study of TMDI in combination with base isolated structures [21–26], showing that
the use of TMDI can further reduce the displacement demand of base isolated structures with respect to the case of classical
TMD, thus enhancing its control performance.
Clearly, to this aim optimal design parameters of the TMDI need to be determined for the different possible operating
conditions. However, while for simple single-degree-of-freedom system equipped with TMDI many different strategies for
the optimal choice of these values exist, and even closed form solutions have been provided [19], for base isolated structures
only numerical procedures based on the minimization of different objective functions have been proposed (see [23] and ref-
erences therein). These procedures, although being very effective, may be computationally expensive and not easily appli-
cable during a design process.
In this paper, based on some reasonable assumptions pertaining the base isolation system, a direct optimization proce-
dure of the TMDI design parameters is performed, aiming at maximally control the seismic response of base isolated struc-
tures. Specifically, due to the stochastic nature of earthquake ground motion, the case of Gaussian white noise ground
excitation is considered, and simple formulae for the optimal TMDI parameters are found analytically. Comparison with a
more accurate and computationally demanding numerical optimization procedure is performed to show the reliability of
the proposed approach, even for more realistic ground motion modelled as non-white filtered processes. Notably the
obtained closed-form expression leads to very accurate results, without requiring any computational effort as in the case
of the other numerical procedure. Finally, to show the influence of the non-stationary nature of real ground motions, the
control performance of the TMDI on a benchmark base isolated structure is also tested, demonstrating the accuracy of
the proposed procedure for the optimal design of TMDI parameters even for real seismic excitations.

2. Problem formulation

Consider a base isolated (BI) structure subject to a horizontal ground acceleration € xg ðtÞ, as in Fig. 1(a). Let the main struc-
ture have n degrees of freedom (DOF); thus, the whole BI system has n + 1 DOF. Denote as mb , K b , C b respectively the mass,
the stiffness and damping coefficient of the base isolation sub-system, assumed to behave linearly. Further, let M i be the
mass associated to the i-th degree of freedom of the n-DOF main structure ði ¼ 1; . . . ; nÞ, whereas C i;j and K i;j are the generic
elements of the corresponding damping and stiffness matrices.
In this manner, the total mass of the system can be defined as
X
n
Mtot ¼ mb þ Mi ð1Þ
i¼1

and the response of the BI structure is governed by the following system of n + 1 equations
A. Di Matteo et al. / Mechanical Systems and Signal Processing 134 (2019) 106337 3

Fig. 1. Base isolated structures: a) MDOF base isolated shear-type frame; b) MDOF base-isolated TMDI-controlled shear-type frame.

8
> P
n
>
> M tot €xb ðtÞ þ Mi €xi ðt Þ þ C b x_ b ðtÞ þ K b xb ðt Þ ¼ Mtot €xg ðtÞ
<
i¼1
; ði ¼ 1; :::; nÞ ð2Þ
>
> P
n P
n
: M i €xb ðtÞ þ Mi €xi ðt Þ þ C i;j x_ j ðt Þ þ K i;j xj ðtÞ ¼ M i €xg ðtÞ
>
j¼1 j¼1

where xb ðt Þ is the displacement of the base isolation sub-system relative to the ground, xi ðtÞ is the displacement of the i-th
DOF of the main structure relative to xb ðt Þ, as shown in Fig. 1, and a dot over a variable stands for derivation with respect to
time.
Aiming at reducing the displacement of the base isolation sub-system xb ðtÞ, let the aforementioned BI structure be con-
nected to an additional energy dissipation mechanism. Specifically, it is assumed that a TMDI is attached to the base plate of
the base isolation sub-system, as shown in Fig. 1(b). Thus, denoting with md , kd and cd , the mass, the stiffness, and the damp-
ing of the TMDI, respectively, and with b its inertance, the equations of motion of the BI + TMDI system can be written as
8 Pn
>
> ðMtot þ md þ bÞ€xb ðtÞ þ M i €xi ðt Þ þ ðmd þ bÞ€xd ðt Þ þ C b x_ b ðtÞ þ K b xb ðt Þ ¼ ðMtot þ md Þ€xg ðt Þ
>
>
>
< i¼1
ðmd þ bÞ€xb ðt Þ þ ðmd þ bÞ€xd ðtÞ þ cd x_ d ðt Þ þ kd xd ðt Þ ¼ md €xg ðt Þ ð3Þ
>
>
>
> P
n P
n
>
: M i €xb ðt Þ þ M i €xi ðtÞ þ C i;j x_ j ðt Þ þ K i;j xj ðt Þ ¼ M i €xg ðt Þ
j¼1 j¼1

where xd ðt Þ is the displacement of the TMDI relative to the base. Note that, as it can be seen in Fig. 1(b), the inerter terminals
are connected to the moving mass of the TMDI on one side and to the ground on the opposite side. In this manner, consid-
ering the properties of the inerter device, the force which it generates is given by F b ¼ bð€ xb þ €xd Þ. Clearly, if the case of the
conventional BI + TMD is considered the parameter b can be set equal to zero.
Note that, Eqs. (3) and (4) greatly simplify if a single degree of freedom (SDOF) main structure is considered. In this
regard, Eq. (2) for the BI structure can be rewritten as
(
€xb ðtÞ þ lb €x1 ðtÞ þ 2xb fb x_ b ðtÞ þ x2b xb ðt Þ ¼ €xg ðt Þ
ð4Þ
€xb ðtÞ þ €x1 ðtÞ þ 2x1 f1 x_ 1 ðtÞ þ x21 x1 ðtÞ ¼ x€g ðt Þ
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where lb ¼ M1 =Mtot , xb ¼ kb =M tot and fb ¼ C b =2xb M tot are the mass ratio, natural frequency and damping ratio of the
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
base isolation sub-system, respectively. Further, x1 ¼ K 1 =M 1 and f1 ¼ C 1 =2x1 M 1 are the natural frequency and damping
ratio of the SDOF main structure.
Analogously, Eq. (4) for the BI + TMDI system with a SDOF main structure is particularized as
4 A. Di Matteo et al. / Mechanical Systems and Signal Processing 134 (2019) 106337

8     
€ € € _ €
< 1 þ ld þ b xb ðt Þ þ lb x1 ðtÞ þ ld þ b xd ðt Þ þ 2fb xb xb ðt Þ þ xb xb ðt Þ ¼  1 þ ld xg ðt Þ
2
>
€xb ðt Þ þ €xd ðt Þ þ 2fd xd x_ d ðt Þ þ x2d xd ðtÞ ¼  ðllþbÞ
d €xg ðt Þ ð5Þ
>
:
d

€xb ðt Þ þ €x1 ðt Þ þ 2f1 x1 x_ 1 ðt Þ þ x21 x1 ðtÞ ¼ €xg ðt Þ


pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where b ¼ b=M tot is the inertance ratio, ld ¼ md =M tot , while xd ¼ kd =ðmd þ bÞ and fd ¼ cd =2xd ðmd þ bÞ are the natural
frequency and damping ratio of the TMDI, respectively. Note that BI + TMDI frequency and damping ratio are influenced
by the presence of the inertance parameter b.

3. Optimal design parameters of TMDI system

As it can be seen in Eq. (5), the four main TMDI parameters which govern the response of the system are the inertance
ratio b, the mass ratio ld , the natural frequency xd , and the damping ratio fd . Clearly, the achievement of the best TMDI sys-
tem performances in controlling base-isolation displacements would require these four parameters to be conveniently cho-
sen. However, two of these variables, namely b and ld , are often chosen a priori due to structural constraints. Further, it is
well-known that higher values of both b and ld generally lead to higher control performances. Therefore, only fd and xd , or
equivalently the so-called frequency ratio m ¼ xd =xb , are required to be appropriately chosen through an optimization
procedure.
In this regard, considering the intrinsic random nature of the seismic excitation, the system is assumed to be subjected to
a zero-mean stationary Gaussian white noise process characterized by the one-sided Power Spectral Density (PSD) G0 . In this
manner, the optimal values of the m and fd are found minimizing the steady-state displacement response variance of the base
isolation sub-system r2X b . This can be directly expressed as
Z 1
r2Xb ¼ jHb ðxÞj2 G0 dx ð6Þ
0

where Hb ðxÞ denotes the base isolation sub-system displacement transfer function. Specifically, the Fourier transform of the
system in Eq. (5) yields
8        
> X ðxÞ x2 1 þ ld þ b þ 2ix fb xb þ x2b  x2 ld þ b X d ðxÞ  x2 lb X 1 ðxÞ ¼  1 þ ld X€ g ðxÞ
>
< b  
2 2 2 €
x X b ðxÞ þ X d ðxÞ x þ 2ix fd xd þ xd ¼ X g ðxÞ l þbld
ð7Þ
>
>  
d
: € g ð xÞ
x2 X b ðxÞ þ X 1 ðxÞ x2 þ 2ix f1 x1 þ x21 ¼ X
Therefore, Hb ðxÞ is given by
  2 2
X b ð xÞ 1 þ ld þ xcðxlÞd þ xaðxlÞb
Hb ðxÞ ¼ ¼ ð8Þ
€ g ðxÞ
X x4 ðld þbÞ
þ xaðxlÞb
4
bðxÞ þ cðxÞ

where
aðxÞ ¼ x2 þ 2ixf1 x1 þ x21
 
bðxÞ ¼ x2 1 þ ld þ b þ 2ixfb xb þ x2b ð9a — cÞ
cðxÞ ¼ x þ 2ixfd xd þ x
2 2
d

Considering the complexity of Eq. (8), the sought response variance r2Xb in Eq. (7) cannot be directly found analytically
solving the involved integral. Thus, the evaluation of the optimal parameters ðm; fd Þ in Eq. (5) would require a rather cum-
bersome numerical optimization procedure, in which the integral in Eq. (6) must be numerically solved for each specific val-
ues of the input parameters to minimize the response variance r2X b .
On the other hand, further simplification can be achieved considering that in base-isolated buildings the main structure
essentially behaves like a single rigid body [10,29], and the base isolation sub-system displacements can be significantly
higher than those of the main structure. Thus, it is feasible to simplify the system in Eq. (5), assuming that the entire base
isolated structure can be modelled as a SDOF system, as shown in Fig. 2 (b). Hence, the original system in Eq. (5) can be recast
as
(     
1 þ ld þ b X € b ðt Þ þ l þ b X € d ðt Þ þ 2fb xb X_ b ðtÞ þ x2 X b ðt Þ ¼  1 þ l X € g ðt Þ
d b d
ð10Þ
€ b ðt Þ þ X
X € d ðtÞ þ 2fd xd X_ d ðt Þ þ x2 X d ðtÞ ¼  d X l € g ðt Þ
d ld þb

where capital letters have been used since X € g ðtÞ is a white noise process and hence displacements and their derivatives are
stochastic processes too. In this manner, the procedure described in [30,31] can now be utilized for the TMDI case, as detailed
in the Appendix A, and the base isolation sub-system response displacement variance of the simplified system in Eq. (10) can
be expressed as
A. Di Matteo et al. / Mechanical Systems and Signal Processing 134 (2019) 106337 5

Fig. 2. Simplified models: a) A SDOF base-isolated TMDI-controlled shear-type frame; b) Base isolated rigid structure equipped with TMDI.

pG0
r2Xb ¼ ð11Þ
4x3b zXb
where zX b depends on both m and fd , and has the following expression
NZ
zX b ¼ ð12Þ
DZ
in which
  h    2 i
NZ ¼ f2d b þ ld m þ 4f3b fd m2 þ fb fd 1  2m2 þ 4f2d 1 þ b þ ld m2 þ 1 þ b þ ld m4
    
þ f2b b þ ld m3 þ 4f2d m þ 4f2d 1 þ b þ ld m3 ð13Þ
and
         
DZ ¼ fd þ fd 1 þ ld 4f2d 1 þ ld 1 þ b þ ld m2 þ b 1 þ ld þ 1 þ ld 2 þ 4f2b þ ld m2
 2  2 h  2   i
þ fd 1 þ ld 1 þ b þ ld m4 þ fb m l2d þ 1 þ ld b þ ld m2
 2    
þ 4fb f2d 1 þ ld m 1 þ 1 þ b þ ld m2 ð14Þ

On this base, taking into account Eqs. (11) and (12), the optimal values of m and fd which minimize r2X b can be equivalently
found directly looking for the minimum of the function
1
/ðfd ; mÞ ¼ ð15Þ
zX b
which is independent of G0 and of the natural frequency of the base isolation sub-system xb . In Fig. 3 a sample of the func-
tion /ðm; fd Þ is shown.
Note that an analytical expression for the minimum of /ðm; fd Þ could be identified solving the system of nonlinear alge-
braic equations
( @/ðf
d ;mÞ
@m
¼0
@/ðfd ;mÞ
ð16Þ
@fd
¼0
6 A. Di Matteo et al. / Mechanical Systems and Signal Processing 134 (2019) 106337

Fig. 3. Sample of the smooth function /ðm; fd Þ. Set of parameters: ld ¼ 0:05; fb ¼ 0:05; f1 ¼ 0:02; b ¼ 0:3.

leading, however, to rather unwieldy expressions which could not be easily implemented.
A simpler procedure consists of minimizing /ðm; fd Þ in Eq. (15) through well-known numerical minimization schemes
already implemented in most commercial software (for instance FindMinimum in Mathematica or fminsearch in MATLAB
environment). In this way, the optimal design parameter values mopt and fd;opt can be directly found.

3.1. Analytical expression of the optimal design parameters

As it can be seen in Eqs. (13) and (14), the function zX b depends on the damping ratio of the base isolation sub-system fb .
This parameter generally assumes non-negligible values, possibly being even higher than 10%. Therefore, a complete opti-
mization procedure, as the one described above, is required to take into account also the effect of this parameter, and the
corresponding optimal values mopt and fd;opt must be found numerically.
However, if a simple analytical expression of the optimal design parameters is sought, the approximate behavior for
fb ! 0 can be considered, as costumery in many optimization procedures for passive vibration control systems [19]. Clearly,
in this manner, approximate solutions of the optimal parameters can be found, which can be used for simple and direct use
in a design phase. Specifically, taking into account Eq. (11) the approximate base isolation sub-system response displace-
ment variance becomes
2 pG0
rXb ¼  ð17Þ
4x3b zX b
  
in which zX b ¼ N z =Dz is an approximation of the function zX b in Eq. (12), where
  
NZ ¼ f2d b þ ld m ð18Þ
and
          2  2
DZ ¼ fd þ fd 1 þ ld 4f2d 1 þ ld 1 þ b þ ld m2 þ ld ld þ b  1  b  2 m2 þ fd 1 þ ld 1 þ b þ ld m4 ð19Þ
 1
Looking for the minimum of the function z X b , as in Eqs. (15) and (16), yields the set of algebraic equations
8  2 4 h   i
>
> 2 1ld
< 3 1 þ b þ ld m þ ld þ 4fd 1 þ ld þ b  b 1þld  2 m  ð1þl Þ2 ¼ 0
2 1
d
 2 h   i ð20Þ
>
> 2 1l
: 1 þ b þ ld m4 þ ld  4fd 1 þ ld þ b  b 1þldd  2 m2 þ 2 ¼ 0
1
ð1þld Þ
Solving this system in the real domain leads to the sought analytical approximate optimal values of the design parameters
 
mopt and fd;opt , as
"   2 #  12
 2 1 þ ld 1 þ b þ ld
mopt ¼   ð21Þ
2 þ b þ ld 1  b  ld

and
A. Di Matteo et al. / Mechanical Systems and Signal Processing 134 (2019) 106337 7

" #  12
 1 1 5 þ 4b þ 5ld
f d;opt ¼ 2þ þ      ð22Þ
2 b þ ld b ld  3 þ ld  4 1 þ ld
 
In this manner, the optimal values of the TMDI design parameters mopt and fd;opt can be evaluated in a straightforward
manner. Note that, as it can be seen in Eqs. (21) and (22), their values do not directly depend on the input PSD G0 , nor on
the base isolation sub-system parameters ðxb ; fb Þ. Further, note that these equations are rather different from those
obtained considering a classical (not-isolated) SDOF system controlled by a TMDI and under white noise base excitation,
which have been reported in [19].

4. Accuracy of the optimal parameters

In order to investigate on the accuracy of the proposed analytical solution, pertinent results have been compared with the
optimal values obtained considering the complete system in Eqs. (5). To this aim a numerical search algorithm, based on the
built-in ‘‘fminsearch” function in MATLAB environment, has been adopted to find the values of ðm; fd Þ that minimize the base
isolation sub-system displacement variance r2X b of the complete system in Eq. (6). In Table 1 the comparison among the ana-
 
lytical optimal design parameters mopt and f d;opt in Eqs. (21) an (22), and those determined taking into account Eq. (6) is
reported for various values of the structural parameters, together with the corresponding percentage relative discrepancy
indexes dv and df . In this regard, a reference set of parameters, also used in [23], has been selected, and in turn each of them
has been varied in a wide range of values. Specifically, the following reference set of parameters has been used: b ¼ 0:3,
ld ¼ 0:05, fb ¼ 0:1, f1 ¼ 0:02, G0 ¼ 5  104 , xb ¼ 4p=3, and x 1 ¼ 5xb .
Table 1 also reports the normalized base isolation sub-system displacement variance eX b ¼ r2X b =r2X 0 obtained with the two
approaches, where r2X 0 is the base isolation sub-system displacement variance without TMDI (see Eqs. (4)). Further, the cor-
responding percentage relative discrepancy index de is also shown. Clearly, lower values of eX b denote higher efficiency in
terms of control performances of the TMDI device, thus this parameter can also serve as a performance control index.
Notably, several considerations can be drawn from a detailed analysis of Table 1. Specifically, as far as the optimal param-
  
eter values are concerned, analytical optimal values mopt ; f d;opt , and corresponding parameters based on the complete solu-
  
tion do not depend on G0 , since the original system is linear. Clearly, as previously stated mopt ; f d;opt are also independent of
fb and f1 , since they have been determined using Eqs. (21) and (22). However, the corresponding performance control index
eXb , obtained based on the analytical solution data, may vary with respect to fb and f1 . This is due to the fact that pertinent
  
base isolation sub-system displacement variance r2Xb has been evaluated substituting mopt ; fd;opt in Eq. (6), thus not using
Eq. (17), to appropriately compare the performance of different optimization procedures on the same structural system.

Table 1
Comparison of optimal TMDI design parameters: analytical solution (Eqs. (21) and (22)) vis-à-vis complete numerical procedure. Chosen reference set of
parameters: b ¼ 0:3; lb ¼ 5=6; ld ¼ 0:05; fb ¼ 0:1; f1 ¼ 0:02; G0 ¼ 5  104 ; xb ¼ 4p=3; x 1 ¼ 5xb .

mopt dm fd;opt df eX b de
(%) (%) (%)
Analytical Complete Analytical solution Complete Analytical Complete
 
solution mopt solution solution solution solution
f d;opt

b 0.1 0.878 0.829 5.9 0.184 0.180 2.2 0.605 0.598 1.2
0.2 0.826 0.786 5.1 0.230 0.225 2.2 0.519 0.515 0.8
0.3 0.781 0.746 4.7 0.264 0.260 1.5 0.464 0.462 0.4
0.4 0.741 0.711 4.2 0.292 0.287 1.7 0.425 0.423 0.5
ld 0.03 0.798 0.769 3.8 0.258 0.253 2.0 0.460 0.458 0.4
0.05 0.781 0.746 4.7 0.264 0.260 1.5 0.464 0.462 0.4
0.07 0.764 0.724 5.5 0.271 0.266 1.9 0.469 0.466 0.6
0.1 0.740 0.694 6.6 0.279 0.276 1.1 0.476 0.471 1.1
fb 0.05 0.781 0.762 2.5 0.264 0.259 1.9 0.280 0.279 0.4
0.1 0.781 0.746 4.7 0.264 0.260 1.5 0.464 0.462 0.4
0.15 0.781 0.729 7.1 0.264 0.260 1.5 0.593 0.589 0.7
0.2 0.781 0.712 9.7 0.264 0.261 1.1 0.687 0.680 1.0
f1 0.01 0.781 0.746 4.7 0.264 0.260 1.5 0.464 0.462 0.4
0.02 0.781 0.746 4.7 0.264 0.260 1.5 0.464 0.462 0.4
0.1 0.781 0.746 4.7 0.264 0.259 1.9 0.466 0.463 0.6
0.2 0.781 0.745 4.8 0.264 0.259 1.9 0.468 0.465 0.6
G0 1104 0.781 0.746 4.7 0.264 0.260 1.5 0.464 0.462 0.4
5104 0.781 0.746 4.7 0.264 0.260 1.5 0.464 0.462 0.4
1103 0.781 0.746 4.7 0.264 0.260 1.5 0.464 0.462 0.4
5103 0.781 0.746 4.7 0.264 0.260 1.5 0.464 0.462 0.4
8 A. Di Matteo et al. / Mechanical Systems and Signal Processing 134 (2019) 106337

Further, as far as the parameter eX b is concerned, Table 1 shows that higher control performances are achieved when
lower values of base isolation sub-system damping ratio fb are used, as also assessed in [23]. In this regard, it can be observed
that the control performance tends to increase more significantly for values of fb lower than 0.1. This characteristic could
warrant further investigation on the possible use of low damping base-isolation systems combined with passive control
devices.
Finally, as it can be seen in Table 1, very small differences between the two approaches are obtained in terms of the opti-
mal parameters, with the highest discrepancy less than 10% reached for the tuning ratio, in case of high value of fb .
Nevertheless, the optimized parameters in Eqs. (21) and (22) yield almost the same control performances as those
obtained through the complete solution procedure, namely a more computationally demanding numerical minimization
of r2X b in Eqs. (6). Specifically, the highest discrepancy de in terms of the performance control index eX b is approximately lower
than 1%. In this regard, note that the negative values of the parameter de denote that higher control is reached with the opti-
mal parameters evaluated through the numerical minimization, as expected.
Clearly, the above described procedure has been derived considering a white noise base excitation, whose properties have
allowed to find a simplified analytical expression for the optimal parameters. Therefore, further analysis should be per-
formed to show the reliability of Eqs. (21) and (22) in case of more generic broad-band earthquake excitation. To this
aim, the widely used Kanai-Tajimi stationary filtered white noise process can be adopted as a more realistic model of earth-
quake ground accelerations. This process is characterized by the following one-sided PSD [25,32]

x4g þ 4f2g x2g x2 x4


GX€ g ðxÞ ¼ G0  2  2 ð23Þ
x2g  x2 þ 4f2g x2g x2 x2f  x2 þ 4f2f x2f x2

where G0 is the constant white noise PSD whose value is related to the bedrock peak ground acceleration, while
 
xg ; fg ; xf ; ff are filter parameters whose values depend on the different soil conditions [33]. Next, taking into account
Eq. (6), the corresponding base isolation sub-system displacement variance r2X b can be given as
Z 1
2
r2Xb ¼ jHb ðxÞj GX€ g ðxÞ dx ð24Þ
0

where Hb ðxÞ is given in Eq. (8). In this manner, again a numerical minimization procedure of Eq. (24) can be used to find the
optimal values of the design parameters mopt and fd;opt also for this model. Once these optimized parameters are found, the
corresponding performance control index eX b can be evaluated. In this regard, values of eX b computed via the aforementioned
procedure (complete solution) vis-à-vis pertinent values obtained using optimal parameters in Eqs. (21) and (22) are shown
in Fig. 4 for a wide range of base isolation sub-system damping ratios fb . Note that variation of eX b with respect to other
parameters is here omitted for succinctness, since eX b is mostly influenced by fb , as also shown in Table 1.
As it can be seen in this figure, even assuming a non-white earthquake excitation the performance control index eX b
obtained using the optimal values in Eqs. (21) and (22) (black lines) closely agrees with the one obtained with the complete
solution (red dots), also for different soil conditions. Thus, considering the significant reduction in computational effort

Fig. 4. Comparison of the performance control index eX b in case of non-white excitation for different values of base isolation sub-system damping ratio:
analytical based solution (black lines) vis-à-vis numerical complete solution (red dots). (For interpretation of the references to colour in this figure legend,
the reader is referred to the web version of this article.)
A. Di Matteo et al. / Mechanical Systems and Signal Processing 134 (2019) 106337 9

Fig. 5. Earthquake records: a) Imperial Valley (USA, 09/15/1979); b) L’Aquila (Italy, 01/06/2009).

achieved with the proposed analytical solution and the small discrepancies in terms of eX b , the aforementioned approach can
be effectively regarded as a powerful and reliable tool for the optimal design parameters estimation.

5. Analysis of the control performance

The previous optimal design procedure has been formulated using stationary white noise and filtered noise processes as
base accelerations to reduce the computational complexity.
Therefore, to properly take into account also the influence of the non-stationary nature of real earthquakes, in this section
the reliability of the proposed analytical solution for the optimal design parameters is analyzed in the time domain in terms
of control performances of the BI system with the TMDI, considering selected recorded accelerograms with different features.
Specifically, the Imperial Valley (USA, 09/15/1979) (Fig. 5(a)) and L’Aquila (Italy, 04/06/2009) (Fig. 5(b)) recorded earth-
quakes have been used as base accelerations. Note that these two earthquakes records present quite different characteristics
since the latter has high impulsive content in the first instants of motion, which is known to be not an ideal condition for the
efficiency of the control system.
The benchmark structure used in the numerical analysis is a base-isolated five-story planar frame building ðn ¼ 5Þ, ana-
lyzed in several previous studies and whose geometrical and mechanical characteristics are fully described in [23]. Specif-
ically, the mass of the main structure is lamped at each floor level and equal to M i ¼ 60  103 kg for ði ¼ 1; . . . ; 4Þ, while the
mass of the fifth floor is M5 ¼ 50  103 kg. Further, the natural frequencies of the structure are
10 A. Di Matteo et al. / Mechanical Systems and Signal Processing 134 (2019) 106337

Fig. 6. Base isolation displacements relative to the ground: a) Response to the Imperial Valley earthquake; b) Response to the L’Aquila earthquake; BI with
TMDI - black solid line; BI without TMDI – red dashed line. (For interpretation of the references to colour in this figure legend, the reader is referred to the
web version of this article.)

xi ½rad=s ¼ ½12:5; 33:2; 56:1; 79:5; 112:2, and the system is assumed to be a classically damped structure with damping
ratio of each mode fi ¼ 0:02. As far as the base isolation sub-system is concerned, its mass is mb ¼ 50  103 kg, while natural
frequency and damping ratio are, respectively, xb ¼ p rad=s and fb ¼ 0:1; i.e., low-damping isolation bearings are assumed.
Finally, assuming a mass ratio ld ¼ 5% and inertance ratio b ¼ 0:3, the optimal design parameters of the TMDI device
 
mopt ¼ 0:781 and fd;opt ¼ 0:264 have been determined using Eqs. (21) and (22).
In passing, it is noted that a more detailed optimization procedure would require a complex analysis on the entire MDOF
structure, minimizing numerically the base-isolation variance of the whole structure. However, considering that base-
isolated systems are mainly dominated by the first mode of vibration, and in line with classical studies in the literature
(see [23] and references therein), the proposed optimization procedure can be directly applied once the frequency of the
base-isolation system has been determined.
In this manner, response of the base isolated structure, with and without the TMDI, and subjected to the earthquake
records in Fig. 5 can be evaluated by direct numerical integration of the equations of motion Eqs. (2) and (3). In this regard,
in Fig. 6 the base isolation displacements relative to the ground xb ðt Þ are shown for the aforementioned records. Comparison
between base isolation system with TMDI (black line) and without TMDI (red dashed line) is provided for the two cases.
As it can be seen in Fig. 6(a), the TMDI device is particularly effective in reducing the base isolation displacement for the
Imperial Valley earthquake, as expected, with a clear reduction of the peak base isolation displacement of almost 47%. On the
other hand, a different behavior is shown in Fig. 6(b) for the L’Aquila earthquake. In this case, the TMDI yields lower control
performance leading to a reduction of almost 13% for the first peak. As previously mentioned, this is a rather common feature
A. Di Matteo et al. / Mechanical Systems and Signal Processing 134 (2019) 106337 11

Fig. 7. Last floor displacement relative to the base: a) Response to the Imperial Valley earthquake; b) Response to the L’Aquila earthquake; BI with TMDI -
black solid line; BI without TMDI – red dashed line. (For interpretation of the references to colour in this figure legend, the reader is referred to the web
version of this article.)

of passive control devices, that generally have little effects on the structural responses in the first few seconds of the exci-
tation [11].
Similar results are also shown in terms of last floor displacements relative to the base x5 ðt Þ, as depicted in Fig. 7. Again,
higher control performances are obtained for the Imperial Valley record (Fig. 7(a)), with respect to the L’Aquila ground
motion (Fig. 7(b)).
On this base, it can be argued that, although being developed assuming a white noise base excitation, the proposed ana-
lytical solution yields optimal design parameters that lead to satisfactory control performances also for real earthquake
records.
Finally, it is noted that for sake of simplicity the main structure and the base isolation sub-system have been supposed to
be linear. Clearly, many real base-isolation systems may show characteristic nonlinear behavior. Nevertheless, the herein
developed analysis would be equally applicable by utilizing, for instance, an equivalent linearization technique to partially
take into account for the effect of the nonlinearity.

6. Concluding remarks

In this paper, the dynamic behavior of base isolated structures equipped with a recently introduced passive control sys-
tem, namely the Tuned Mass Damper Inerter (TMDI), has been studied. The equations of motion of base isolated multi-
degree-of-freedom systems, controlled at the base through a TMDI, have been introduced, and optimal design of the device
has been investigated. The following main concluding remarks can be drawn:
12 A. Di Matteo et al. / Mechanical Systems and Signal Processing 134 (2019) 106337

 A straightforward procedure for the optimal design of TMDI has been proposed, assuming a Gaussian white noise base
excitation, to maximally reduce the seismic response of the base-isolated structure in terms of base isolation sub-
system displacement variance. The aforementioned optimization procedure, however, requires employing numerical
minimization algorithms;
 Aiming at further reducing the computational complexity, approximate simple analytical expressions of the optimal
design parameters have been determined, under some assumptions regarding the base isolation sub-system parameters.
Notably, the proposed analytical approximate optimal values of the design parameters do not depend on the power spec-
tral density of the ground acceleration, nor on the natural frequency and damping ratio of the base isolation sub-system.
In this manner, ready-to-use design formulae have been derived for practical purposes.
 A comparison with the optimal parameters obtained by a rather elaborate numerical optimization procedure has been
carried out considering the original base isolated structure with TMDI. Data have shown very small discrepancies among
the two approaches, especially in terms of control performances, thus proving the reliability of the proposed simplified
analytical solution. Analogous results have been also obtained for broad-band earthquake excitation, modeled as Kanai-
Tajimi stationary filtered normal white noise process, to take into account additional features of the seismic excitation.
 The control performance of the TMDI device connected to the base-isolated structure has been further examined employ-
ing two recorded ground accelerations with different features, and considering a 5-story benchmark base-isolated planar
frame structure. It has been shown that the TMDI device designed with the proposed approach can effectively lead to a
reduction of the base isolation displacements, as well as other response quantities such as floor displacements, compared
to the case without TMDI, even for the case of real earthquake records.

Overall, the herein developed analytical and numerical analyses show that the proposed TMDI configuration can be
regarded as a practical and promising solution for passive vibration control of base isolated structures, and the proposed
approximate analytical approach offers a reliable and effective tool for the design of this system in several real conditions.

Acknowledgement

Authors gratefully acknowledge the support received from the Italian Ministry of University and Research, through the
PRIN 2015 funding scheme (project 2015JW9NJT – Advanced mechanical modeling of new materials and structures for
the solution of 2020 Horizon challenges).

Appendix A

In this appendix the procedure for the evaluation of the base isolation sub-system response displacement variance of the
system in Eq. (10) is outlined. In this regard, Eq. (10) can be rewritten in compact matrix form as
    
€ ðtÞ þ C Z_ ðt Þ þ K ZðtÞ ¼  M r X
MZ € g ðt Þ ðA1Þ

h iT
bð1þld Þ
where ZðtÞ ¼ ½ X b ðtÞ X d ð t Þ T , r ¼ bþld þbld
bþld
 bþld
, and


" #
 1 þ ld þ b ld þ b  2fb xb 0  x2b 0
M¼ ; C¼ ; K¼ ðA2Þ
1 1 0 2fd xd 0 x2d
Since the input is modeled as a zero-mean stationary Gaussian white noise process, the corresponding Lyapunov equation
of the evolution of the covariance matrix [31] can be written as

R_ Q ðtÞ ¼ DS RQ ðtÞþRQ ðt ÞDTs þGS GTS pG0 ðA3Þ


 T
where Q ¼ Z Z_ is the vector of the state variables, RQ ðtÞ denotes the covariance matrix containing the evolution of all
the response statistics of the system (10)
2 3
r2Xb
r2Xb Xd r2Xb X_ b r2Xb X_ d
6 7
6 r2Xd r2Xd X_ b r2Xd X_ d 7
6 7
RQ ðt Þ ¼ 6
6 sym
7 ðA4Þ
6 rX_ 2 rX_ b X_ d 7
2 2
7
4 b 5
r2X_ d
while DS and GS are given as
" #

0 I2 0
DS ¼  1   1  ; GS ¼  ðA5Þ
M K M C r

with I2 a 2  2 identity matrix.


A. Di Matteo et al. / Mechanical Systems and Signal Processing 134 (2019) 106337 13

Since only the steady-state variance must be computed, R_ Q ðtÞ can be equated to zero, and the sought variance of the base
isolation sub-system r2X b can be therefore directly determined from Eq. (A.3), as reported in Eq. (11).

References

[1] F. Naeim, J.M. Kelly, Design of Seismic Isolated Structures: From Theory to Practice, John Wiley & Sons, New York, 1999.
[2] J.M. Kelly, Base isolation: linear theory and design, J. Earthquake Spectra 6 (1990) 223–244.
[3] J.M. Kelly, The role of damping in seismic isolation, Earthq. Eng. Struct. Dyn. 28 (1999) 3–20.
[4] H.C. Tsai, The effect of tuned-mass damper on the seismic response of base-isolated structures, Int. J. Solids Struct. 32 (1995) 1195–1210.
[5] J.N. Yang, A. Danielians, S.C. Li, Aseismic hybrid control systems for building structures, J. Eng. Mech. 117 (1991) 836–853.
[6] T. Taniguchi, A. Der Kiureghian, M. Melkumyan, Effect of tuned mass damper on displacement demand of base-isolated structures, Eng. Struct. 30 (12)
(2008) 3478–3488, https://doi.org/10.1016/j.engstruct.2008.05.027.
[7] A. Kareem, Modelling of base-isolated buildings with passive dampers under winds, J. Wind. Eng. Ind. Aerodyn. 72 (1997) 323–333.
[8] B. Palazzo, L. Petti, Aspects of passive control of structural vibrations, Meccanica 32 (1997) 529–544.
[9] Y. Arfiadi, M.N.S. Hadi, Hybrid base isolation-passive mass damper systems, Eighth International Conference on Computing in Civil and Building
Engineering. Stanford, USA, 14-16 August. 2000.
[10] J.S. Love, M.J. Tait, H. Toopchi-Nezhad, A hybrid structural control system using a tuned liquid damper to reduce the wind induced motion of a base
isolated structure, Eng. Struct. 33 (2011) 738–746.
[11] A. Di Matteo, T. Furtmüller, C. Adam, A. Pirrotta, Optimal design of tuned liquid column dampers for seismic response control of base-isolated
structures, Acta Mech. 229 (2018) 437–454.
[12] T. Furtmüller, A. Di Matteo, C. Adam, A. Pirrotta, Base-isolated structure equipped with tuned liquid column damper: an experimental study, Mech.
Syst. Signal. Proc. 116 (2019) 816–831.
[13] M.C. Smith, Synthesis of mechanical networks: the Inerter, IEEE Trans. Autom. Control 47 (2002) 1648–1662.
[14] C. Papageorgiou, N.E. Houghton, M.C. Smith, Experimental testing and analysis of inerter devices, J. Dyn. Syst. Meas. Control. 131 (2009) 11001.
[15] D. De Domenico, P. Deastra, G. Ricciardi, N.D. Sims, D.J. Wagg, Novel fluid inerter based tuned mass dampers for optimised structural control of base-
isolated buildings, J. Franklin Inst. (2018), https://doi.org/10.1016/j.jfranklin.2018.11.012.
[16] X. Liu, J.Z. Jiang, B. Titurus, A. Harrison, Model identification methodology for fluid-based inerters, Mech. Syst. Signal. Proc. 106 (2018) 479–494.
[17] I.F. Lazar, S.A. Neild, D.J. Wagg, Using an inerter device for structural vibration suppression, Earthq. Eng. Struct. Dyn. 43 (2014) 1129–1147.
[18] D. Pietrosanti, M. De Angelis, M. Basili, Optimal design and performance evaluation of systems with tuned mass damper inerter (TMDI), Earthq. Eng.
Struct. Dyn. 46 (2017) 1367–1388.
[19] L. Marian, A. Giaralis, Optimal design of a novel tuned mass-damper–inerter (TMDI) passive vibration control configuration for stochastically support-
excited structural systems, Prob. Eng. Mech. 38 (2014) 156–164.
[20] H. Garrido, O. Curadelli, D. Ambrosini, Improvement of tuned mass damper by using rotational inertia through tuned viscous mass damper, Eng. Struct.
56 (2013) 2149–2153.
[21] M. Saito, On the performance of gyro-mass devices for displacement mitigation in base isolation systems, Struct. Control Health Monit. 19 (2012) 246–
259.
[22] T. Hashimoto, K. Fujita, M. Tsuji, I. Takewaki, Innovative base-isolated building with large mass-ratio TMD at basement for greater earthquake
resilience, Future Cities. Enviroment 1 (2015) 1–19.
[23] D. De Domenico, G. Ricciardi, An enhanced base isolation system equipped with optimal tuned mass damper inerter (TMDI), Earthq. Eng. Struct. Dyn.
47 (2018) 1169–1192.
[24] D. De Domenico, G. Ricciardi, Improving the dynamic performance of base-isolated structures via tuned mass damper and inerter devices: a
comparative study, Struct. Control Health Monit. 25 (2018) e2234.
[25] D. De Domenico, G. Ricciardi, Optimal design and seismic performance of tuned mass damper inerter (TMDI) for structures with nonlinear base
isolation systems, Earthq. Eng. Struct. Dyn. 47 (2018) 2539–2560.
[26] M. De Angelis, A. Giaralis, F. Petrini, D. Pietrosanti, Optimal tuning and assessment of inertial dampers with grounded inerter for vibration control of
seismically excited base-isolated systems, Eng. Struct. 196 (2019) 109250.
[27] A. Giaralis, A.A. Taflanidis, Optimal tuned mass-damper-inerter (TMDI) design for seismically excited MDOF structures with model uncertainties based
on reliability criteria, Struct. Control Health Monit. 25 (2018) e2082.
[28] A. Giaralis, F. Petrini, Wind-Induced vibration mitigation in tall buildings using the tuned mass-damper-inerter, J Struct. Eng. 143 (2017) 04017127.
[29] P. Xiang, A. Nishitani, Optimum design for more effective tuned mass damper system and its application to base-isolated buildings, Struct. Control
Health Monit. 21 (2014) 98–114.
[30] A. Di Matteo, F. Lo Iacono, G. Navarra, A. Pirrotta, Optimal tuning of tuned liquid column damper systems in random vibration by means of an
approximate formulation, Meccanica. 50 (2015) 795–808.
[31] J.B. Roberts, P.D. Spanos, Random Vibration and Statistical Linearization, Wiley, New York, 1990.
[32] R.W. Clough, J. Penzien, Dynamics of Structures, Computers and Structures Inc., Berkeley, 2003.
[33] A. Der Kiureghian, A. Neuenhofer, Response spectrum method for multi-support seismic excitations, Earthq. Eng. Struct. Dyn. 21 (1992) 713–740.

You might also like