Journal of Sound and Vibration 412 (2018) 184e206

Contents lists available at ScienceDirect

Journal of Sound and Vibration

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

Theoretical and experimental investigation of position-

controlled semi-active friction damper for seismic structures
Lyan-Ywan Lu a, *, Tzu-Kang Lin b, Rong-Jie Jheng c, Hsin-Hsien Wu d
a
Department of Civil Engineering, National Cheng Kung University, 1 University Road, Tainan 701, Taiwan
b
Department of Civil Engineering, National Chiao Tung University, Taiwan
c
Department of Construction Engineering, National Kaohsiung First University of Science and Technology, Taiwan
d
Department of Civil Engineering, National Cheng Kung University, Taiwan

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

Article history: A semi-active friction damper (SAFD) can be employed for the seismic protection of
Received 8 April 2017 structural systems. The effectiveness of an SAFD in absorbing seismic energy is usually
Received in revised form 14 September 2017 superior to that of its passive counterpart, since its slip force can be altered in real time
Accepted 24 September 2017
according to structural response and excitation. Most existing SAFDs are controlled by
adjusting the clamping force applied on the friction interface. Thus, the implementation of
SAFDs in practice requires precision control of the clamping force, which is usually sub-
Keywords:
stantially larger than the slip force. This may increase the implementation complexity and
Semi-active control
Friction damper
cost of SAFDs. To avoid this problem, this study proposes a novel position-controlled SAFD,
Position control named the leverage-type controllable friction damper (LCFD). The LCFD system combines a
Variable damper traditional passive friction damper and a leverage mechanism with a movable central
Leverage mechanism pivot. By simply controlling the pivot position, the damping force generated by the LCFD
Energy dissipation device system can be adjusted in real time. In order to verify the feasibility of the proposed SAFD,
Adaptive system a prototype LCFD was tested by using a shaking table. The test results demonstrate that the
equivalent friction force and hysteresis loop of the LCFD can be regulated by controlling the
pivot position. By considering 16 ground motions with two different intensities, the
adaptive feature of the LCFD for seismic structural control is further demonstrated
numerically.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction

Numerous studies have demonstrated that incorporating energy dissipation devices into a seismic structure can effec-
tively enhance the seismic resistance capacity of a structure [1e4]. Friction dampers, a kind of energy dissipation device, have
been successfully applied in numerous civil engineering structures for seismic protection [5e7]. Design methods and
guidelines for structures with friction dampers have been developed and incorporated in some seismic provisions [8e12]. A
typical friction damper generally consists of one or multiple friction interfaces and a constant normal force (clamping force) is
applied at these interfaces. The frictional interfaces slip when the applied seismic load exceeds the maximum static friction
force (slip force) of the damper, and thus provide the structure to be protected an additional energy dissipation source
through friction behavior [1,2]. This type of passive friction damper (PFD) does not require additional control energy to

* Corresponding author.
E-mail address: lylu@mail.ncku.edu.tw (L.-Y. Lu).

https://doi.org/10.1016/j.jsv.2017.09.029
0022-460X/© 2017 Elsevier Ltd. All rights reserved.
 L.-Y. Lu et al. / Journal of Sound and Vibration 412 (2018) 184e206 185

operate and is thus more reliable and also easier to implement and maintain compared with active control devices. Never-
theless, the slip force of a PFD is usually a fixed value, which is predetermined according to a given seismic load specified by
design codes in practice, and once the device is designed and manufactured, the slip force may no longer be altered. PFDs may
thus not be able to meet the seismic demand of a structure subjected to an earthquake with an intensity or characteristics
different from those specified in the design code. For instance, a PFD may not be activated in a moderate earthquake whose
intensity is lower than that of the design earthquake, or a PFD may not provide sufficient energy dissipation capacity under a
severe earthquake whose intensity is higher than that of the design one.
To improve the seismic control performance of friction-type dampers, semi-active friction dampers (SAFDs) have been
proposed [13e16]. Also called a variable or controllable friction damper, an SAFD is able to generate a slip force that adapts to
external excitations. A typical SAFD usually consists of one or more friction interfaces and a controllable clamping mechanism
that produces an adjustable normal contact force between these interfaces. By controlling the normal force, the slip force of
the damper can be regulated in real time. An SAFD is usually able to produce more favorable damping effects than those
produced by a PFD, since its slip force can be adjusted instantaneously according to the dynamic characteristics of seismic
loads and structural responses, while consuming substantially less control energy than that consumed by an active device,
since the control action of SAFDs is applied internally rather than directly on the controlled structural system, which usually
has large mass. The application of SAFDs generally requires the installation of a sensing system to monitor the system
response and excitation, and also a suitable control law to determine the most appropriate slip force in real time, such that the
seismic response of the controlled structure can be more effectively suppressed [17e19].
In order to make the slip force of SAFDs variable, many damper mechanisms and configurations have been proposed in the
literature. Kannan et al. [20] and Samani et al. [21] adopted a hydraulic power source, which is able to produce large actuation
forces, to regulate the clamping force of an SAFD, with a large variable slip force achieved. Laflamme et al. [22] and Cao et al.
[23] proposed and tested a reliable SAFD, termed the modified friction device (MFD), for large-scale structures. They have
demonstrated that the MFD, which consists of a variable friction mechanism based on a vehicle duo-servo drum brake, a
mechanically robust and matured technology, and is capable of producing a large damping forces, while operating on a
battery power. Later on, Downey et al. [24] incorporated a toggle brace with a variable friction device similar to the MFD,
named banded rotary friction device (BRFD) that is capable of providing variable braking torques as a linear function of an
applied force significantly amplified by the brake's positive servo effect. The toggle brace, which is commonly used in damper
systems for amplifying the motion of inter-story drift, allows the BRFD to reach a maximum frictional force faster and thus
enhances the control performance. Yang and Agrawal [25] adopted an electromagnetic device to control the clamping force of
an SAFD installed in a base isolation system to enhance the seismic safety of a building structure. Nishitani et al. [26] utilized
an oil damper with an electromagnetic relief valve to mimic a friction damper with a variable slip force level. Narasimhan and
Nagarajaiah [27] developed a variable friction device that is composed of four pairs of friction and stiffness elements arranged
in a rhombus configuration, with each arm consisting of friction and stiffness elements placed in parallel. The level of friction
force is adjusted smoothly by changing the angle of the rhombus arms via a linear electromechanical actuator.
More recently, numerous researchers have proposed various SAFD configurations with piezoelectric actuators that provide
variable clamping forces [28e34]. As an actuation source for structural vibration control, piezo-ceramic materials have many
advantages, including light weight, swift response, easy implementation, and low control power [35]. Nevertheless, the major
drawback of using piezoelectric actuators is that the actuation force generated by a piezo-ceramic material is usually limited
and very sensitive to the stiffness of the boundary or support conditions due to its stroke limitation, which is tens of mi-
crometers. To increase the level of the controllable slip force for a piezoelectric SAFD, Pardo-Varela and Llera [36] proposed a
configuration that consists of multiple friction interfaces and piezoelectric actuators placed in a very stiff clamping system.
As mentioned above, for most existing SAFD systems, the level of the slip force is regulated by controlling the clamping
force applied at the friction interfaces. In other words, the variation of the slip force is realized by using a force-controlled
mechanism that controls the clamping force. At any time instant, the variable slip force of these SAFDs is the product of
the clamping force and the friction coefficient between the contact interfaces. Therefore, for a given slip force, the required
clamping force is reversely proportional to the friction coefficient. Moreover, because of practical concerns, such as
compression capacity, wear resistance, friction stability, and durability, the selection of suitable friction materials for SAFDs is
usually very limited. The most commonly used friction materials are metals or composite materials that usually have a
frictional coefficient of 0.1e0.4 [30,35,37]. The magnitude of the controllable clamping force must be much larger than the
desired slip force, which can have an order of around one hundred metric tons for controlling a real seismic structure.
Consequently, in practice, producing and precisely controlling this huge clamping force in real time may be difficult and
costly. To avoid this problem, the present study proposes a novel SAFD called the leverage-type controllable friction damper
(LCFD). The proposed system consists of a traditional PFD with a constant slip force and a controllable leverage mechanism
that has a movable central pivot. By instantaneously adjusting the pivot position on the lever arm, the LCFD is able to convert
the passive friction force into an equivalent controllable force. Therefore, different from most existing SAFDs, which are
usually force-controlled, the proposed LCFD is essentially a position-controlled system with a cost-effective traditional PFD.
The objective of the present study is to develop an analytical method and a numerical tool for the proposed semi-active
system, and to demonstrate the system's feasibility and adaptive nature via experimental and numerical approaches. The
development of control laws is not the focus of this study. The rest of this paper is organized as follows. Using Lagrange's
equation, Section 2 derives the governing equation of motion for a structure with the LCFD in a comprehensive way, since the
concept of LCFDs is proposed for the first time. A numerical simulation method based on the derived analytical model is
186 L.-Y. Lu et al. / Journal of Sound and Vibration 412 (2018) 184e206

developed in Section 3. The results of a shaking table test conducted on a prototype LCFD are reported in Section 4 and
compared with numerical results simulated using the developed analysis method. To further highlight the adaptive nature of
the LCFD system, a numerical simulation that considers 16 seismic ground motions with different characteristics is performed
in Section 5. Finally, conclusions are given in Section 6.

2. Analytical model for structure with LCFD

2.1. Configuration of LCFD

Fig. 1 schematically illustrates a building structure equipped with an LCFD installed on the first floor. Fig. 2 illustrates the
mathematical model of the system in Fig. 1. To more clearly explain the configuration and principles of the LCFD system, the
leverage part is enlarged in Fig. 3. Fig. 2 shows that the leverage mechanism includes a lever arm with a movable central pivot
(point P). One end of the lever arm (point D) is connected to a brace that is then connected to the main structure. The effect of
the brace stiffness, denoted by kb, is considered in the study. The other end of the lever arm (point A) is connected to a PFD,
whose friction force is expressed as uf (see Fig. 2). Pivot point P can only move along the y axis, and points A and D can only
move parallel to the x axis. In the case where point P moves in the positive direction along the y axis (see Fig. 3, pivot
displacement xp > 0), the length of the lever arm at the damper side is larger than that at the structure side. As a result, the
equivalent friction force u(t) exerted on point D (the structural end) is amplified compared with the actual friction force uf
acting at point A. Conversely, if point P moves in the negative direction along the y axis (i.e., xp < 0), the equivalent friction
force exerted on the structural end is reduced, since the lever arm at the damper side is reduced. Therefore, the equivalent
friction force u(t) of the LCFD applied on the structure can be altered by controlling the position of pivot point P, so that the
desired damper force can be achieved.
From the above explanation, different from traditional SAFDs, the frictional damping force of the proposed LCFD is varied
by controlling the pivot position (i.e., position control) rather than the clamping force (i.e., force control). The clamping force
of the PFD in the LCFD remains constant. In other words, the major difference between the LCFD and a traditional SAFD is that
the former requires a passive friction device, which is a reliable and matured technology, while the latter usually requires the
precision control of a huge variable clamping force through an internal mechanism. The major purpose of the leverage
mechanism in the LCFD is to attenuate the magnitude of the output damping force of the LCFD applied on the controlled
structure, rather than acting as a force amplification mechanism, even though it amplifies the damper force in some instants.
Notably, in a low-intensity earthquake, the leverage mechanism may have to reduce the slip force of the PFD, so that the
friction interface will remain in its sliding state to provide energy dissipation capacity.

2.2. Equation of motion for LCFD-controlled structure

(1) Degrees of freedom of the system:

The equation of motion for an LCFD installed on an n-story shear-type structure can be derived based on the mathematical
model illustrated in Fig. 2, where x€g denotes the ground acceleration, xi (i ¼ 1, 2, … n) is the relative-to-the-ground

Fig. 1. Schematic diagram for a structure with LCFD system.

 L.-Y. Lu et al. / Journal of Sound and Vibration 412 (2018) 184e206 187

Fig. 2. Mathematical model for a structure with LCFD system.

Fig. 3. Geometric relationship of lever-arm with pivot displacement.

displacement of the ith story, and mi, ki, and ci are the mass, stiffness, and damping coefficient of the ith story, respectively. The
symbol Fp represents the actuation force required to move the pivot point. This force, which is a function of structural re-
sponses and the pivot displacement, is derived later. Furthermore, in addition to the floor displacement xi, the degrees of
freedom (DOFs) in Fig. 2 also include pivot displacement xp, friction damper stroke xr, and bracing deformation xb. Based on
the geometric condition of the lever arm shown in Fig. 3, the relation between xp and xr can be expressed as:
188 L.-Y. Lu et al. / Journal of Sound and Vibration 412 (2018) 184e206


0:5 L þ xp
xr ¼  ðxD Þ (1)
0:5 L  xp

where xD is the displacement of point D. Fig. 3 further reveals that xD is related to x1 and xb by:

xD ¼ x1  xb (2)

where x1 denotes the displacement of the first story. Substituting Eq. (2) into Eq. (1) indicates that the variable xr is not an
independent variable, but is related to xp , xb , and x1 . Therefore, the mathematical model illustrated in Fig. 2 contains a total of
(nþ2) DOFs, i.e., xb, xp, and xi (i ¼ 1, 2, … n).

(2) Lagrange's equation and energy formulation:

After the total number of DOFs is known, the equations of motion for the model shown in Fig. 2 can be derived using
Lagrange's equation of motion. Generally, Lagrange's equation can be expressed as:


d v£ v£
 ¼ Qk ðk ¼ 1; 2; :::; n þ 2Þ (3)
dt vq_ k vqk

where ¼ £(T e V) represents the Lagrangian, where T and V are the total kinetic energy and potential energy of the system,
respectively. Also, qk represents the kth DOF in a generalized coordinate system, and Qk represents the generalized non-
conservative force for the kth DOF. Since the structure-LCFD system shown in Fig. 2 has (nþ2) DOFs, there are a total of
(nþ2) generalized coordinates. In the following derivation, the generalized coordinates are chosen as qi ¼ xi (for i ¼ 1, 2, … n),
qnþ1 ¼ xb, and qnþ2 ¼ xp. In Eq. (3), the kinetic and potential energies of the structure-LCFD system can be respectively
expressed explicitly as follows:

Xn
1  2
T¼ mi x_i þ x_g (4)
i¼1
2

1 Xn
1 1
V ¼ k1 x21 þ ki ðxi  xi1 Þ2 þ kb x2b (5)
2 i¼2
2 2

where x_g denotes the ground velocity. Note that in the above equations, the LCFD itself is assumed to be massless and rigid, so
it has no contribution to the energy terms.

(3) Determination of non-conservative forces:

The generalized non-conservative forces Qk shown in Eq. (3) usually have to be derived by using the principle of virtual
work as shown below. First, let the virtual work done by all non-conservative forces be d Wnc , which can be written as:

X
nþ2 X
n
dWnc ¼ Qk d qk ¼ c1 x_1 d x1  ci ðx_i  x_i1 Þðd xi  d xi1 Þ  uf dxr þ Fp d xp (6)
i¼1 i¼2

where a symbol preceded by d denotes a virtual displacement in a certain DOF. In Eq. (6), since xr is related to xp , xb , and x1
through Eqs. (1) and (2), the virtual displacement d xr can be expressed as:





L þ 2xp L þ 2xp x  x1 2L
d xr ¼  d x1 þ dxb þ 2 b d xp (7)
L  2xp L  2xp L  2xp L  2xp

Then, substituting Eq. (7) into Eq. (6) yields:

X
nþ2 X
n1
dWnc ¼ Qk d qk ¼ ½  cn ðx_n  x_n1 Þd xn þ ½ciþ1 ðx_iþ1  x_i Þ  ci ðx_i  x_i1 Þd xi
i¼1 2
 
 

 
L þ 2xp L þ 2xp x  x1 2L
þ c2 ðx_2  x_1 Þ  c1 x_1 þ uf d x1 þ  uf d xb þ  2uf b þ Fp d xp
L  2xp L  2xp L  2xp L  2xp
(8)

By inspecting Eq. (8), the generalized non-conservative force corresponding to each DOF can be identified as follows:
 L.-Y. Lu et al. / Journal of Sound and Vibration 412 (2018) 184e206 189


L þ 2xp
Q1 ¼ c2 ðx_2  x_1 Þ  c1 ðx_1 Þ þ u (9)
L  2xp f

Qi ¼ ci ðx_i  x_i1 Þ þ ciþ1 ðx_iþ1  x_i Þ ðfor i ¼ 2; /; ðn  1ÞÞ (10)

Qn ¼ cn ðx_n  x_n1 Þ (11)

L þ 2xp
Qnþ1 ¼ Qxb ¼  u (12)
L  2xp f



x  x1 2L
Qnþ2 ¼ Qxp ¼ 2uf b þ Fp (13)
L  2xp L  2xp

(4) Derivation of dynamic equation for each DOF:

Next, by substituting Eqs. (9)e(13) and Eqs. (4)e(5) into Lagrange's equation (Eq. (3)) one by one, one may obtain the
equation of motion for each DOF. For instance, let the DOF q1 ¼ x1; then, substituting Eqs. (4)e(5) and Eq. (9) into Eq. (3) yields
the following equation of motion for the first floor:

L þ 2xp ðtÞ
m1 x€1 ðtÞ þ c1 x_1 ðtÞ  c2 ðx_2 ðtÞ  x_1 ðtÞÞ þ k1 x1 ðtÞ  k2 ðx2 ðtÞ  x1 ðtÞÞ ¼ m1 x€g ðtÞ þ u ðtÞ (14)
L  2xp ðtÞ f

Let qi ¼ xi ðfor i ¼ 2; /; ðn  1ÞÞ and qn ¼ xn; then, substituting Eqs. (4)e(5) and Eqs. (10) and (11) into Eq. (3) yields the
equations of motion for each middle floor xi and the top floor xn, respectively:

mi x€i ðtÞ þ ci ðx_i ðtÞ  x_i1 ðtÞÞ  ciþ1 ðx_iþ1 ðtÞ  x_i ðtÞÞ þ ki ðxi ðtÞ  xi1 ðtÞÞ  kiþ1 ðxiþ1 ðtÞ  xi ðtÞÞ ¼ mi x€g ðtÞ (15)

mn x€n ðtÞ þ cn ðx_n ðtÞ  x_n1 ðtÞÞ þ kn ðxn ðtÞ  xn1 ðtÞÞ ¼ mn x€g ðtÞ (16)

Let qnþ1 ¼ xb; then, substituting Eqs. (4)e(5) and Eq. (12) into Eq. (3) yields the equation of motion for the DOF xb:

L þ 2xp
kb xb ðtÞ ¼  u ðtÞ (17)
L  2xp f

Eq. (17) describes the equilibrium condition between the damper and the brace. Finally, let qnþ2 ¼ xp; then, substituting
Eqs. (4)e(5) and Eq. (13) into Eq. (3) yields the equation for the DOF xp as follows:
" #
xb ðtÞ  x1 ðtÞ 4 uf ðtÞ
Fp ðtÞ ¼   2 (18)
L 1  2 xp ðtÞ=L

Eq. (18) gives the driving force Fp required to achieve the desired pivot position xp. As shown, the magnitude of Fp depends
on pivot displacement xp, bracing deformation xb, first-floor displacement x1, and damper passive friction force uf. Eq. (18) also
shows that Fp is inversely proportional to the lever length L.

(5) Dynamic equation in matrix form:

Eqs. (14)e(16), which represent n dynamic equations for an n-story structure with the LCFD, can be combined and
rewritten in concise matrix form as:

€ þ Cs xðtÞ
Ms xðtÞ _ þ Ks xðtÞ ¼ Bs ð  uðtÞÞ  Es x€g ðtÞ (19)

where matrices Ms , Cs , and Ks , whose dimensions are (n  n), represent the structural mass, damping, and stiffness matrices,
respectively. Since the structure is modeled as a shear building, matrix Ms is a diagonal matrix, and Cs and Ks are typical
banded diagonal matrices [38]. Vector x(t) contains the story displacements; Bs represents the damper placement matrix; Es
represents the seismic force distribution matrix. These vectors can be written explicitly as:
190 L.-Y. Lu et al. / Journal of Sound and Vibration 412 (2018) 184e206

8 9 8 9 8 9
>
> xn ðtÞ >
> >
> 0>> >
> mn >
>
>
> >
> >
> >
> >
> >
< « = <«= < « >=
xðtÞ ¼ xi ðtÞ ; Bs ¼ « ; Es ¼ mi (20)
>
> >
> > >
> > >
> « >
> > >
> 0>>
>
>
>
> « >
>
>
: ; : ; : ;
x1 ðtÞ 1 m1

More importantly, in Eq. (19), u(t) represents the controllable semi-active force of the LCFD, and is physically equal to the
equivalent damper force exerted on the structure. From Eq. (14), this semi-active force can be written as:

uðtÞ ¼ rp ðtÞuf ðtÞ (21)

where rp(t) represents the dimensionless leverage amplification factor, which can be written as:

L þ 2xp ðtÞ 1 þ 2xp ðtÞ=L

rp ðtÞ ¼ ¼ (22)
L  2xp ðtÞ 1  2xp ðtÞ=L

Notably, Eq. (21) states that due to the presence of rp(t), the passive friction force uf (t) is converted into a controllable
damper force u(t). Fig. 3 graphically shows the relation between u(t) and uf (t). Eq. (22) states that the value of rp(t) relies on
the pivot position xp ðtÞ, whose actual value is determined using an on-line control law. Moreover, since physically xp must be
within the range of 0.5L < xp < 0.5L, rp(t) is always positive (i.e., rp ðtÞ > 0). If the sign conventions of u(t) and uf (t) are both
defined as positive for tension, Eq. (21) shows that the signs of u(t) and uf (t) are always in opposite directions.

3. Numerical analysis method

3.1. Discrete-time state-space equation of motion

To numerically simulate the seismic response of an LCFD-controlled structure, the state-space method was employed in
this study. In the field of structural control, state-space formulations are widely used for controller design or numerical
simulation, since design methods and numerical solutions based on these formulations are well developed and easily
implemented [41]. To this end, let us rewrite Eq. (19) as a first-order state-space equation as follows:

_
zðtÞ ¼ A zðtÞ þ B uðtÞ þ E x€g ðtÞ (23)

where z(t) represents the state vector, A represents the system matrix, and B and E represent the placement matrices of the
damper and the seismic force, respectively. The contents of the above matrices are:
      
_ 1
M1 K ; B ¼ M1 Bs ; E ¼ M1 Es
; A ¼ M C
xðtÞ
zðtÞ ¼ (24)
xðtÞ I 0 0 0

Next, let Dt be the time interval for numerical analysis. If Dt is sufficiently small, the variation of the semi-active force u(t)
and seismic excitation x€g ðtÞ in Eq. (23) can be assumed to be linear within each time interval. Then, the linear interpolation is
applicable for u(t) and x€g ðtÞ between two adjacent numerical steps. Based on this assumption, the discrete-time solution for
Eq. (23) can be expressed as [39]:

z½k ¼ Ad z½k  1 þ Bd1 u½k  1 þ Bd2 u½k þ Ed1 x€g ½k  1 þ Ed2 x€g ½k (25)

where a variable with ½k indicates that the quantity is evaluated at the kth time step, i.e., at time t ¼ k Dt. The coefficient
matrices in Eq. (25) can be written explicitly as:

Ad ¼ eADt (26)
   
1 1
Bd1 ¼ A1 Ad þ A2 ðI  Ad Þ B; Bd2 ¼  A1 þ A2 ðAd  IÞ B (27)
Dt Dt
   
1 1
Ed1 ¼ A1 Ad þ A2 ðI  Ad Þ E; Ed2 ¼  A1 þ A2 ðAd  IÞ E (28)
Dt Dt

The discrete-time state-space equation in Eq. (25) indicates that in order to calculate the system state z½k at the kth time
step (current step), we need the state z½k  1, semi-active force u½k  1, and ground excitation x€g ½k  1 of the previous step,
as well as the force u[k] and excitation x€g ½k of the current step. Since x€g ½k is the given excitation, u[k] is the only unknown on
the right-hand side of Eq. (25) in the beginning of the computation of the kth step. Moreover, since u[k] is related to the
 L.-Y. Lu et al. / Journal of Sound and Vibration 412 (2018) 184e206 191

passive friction force uf [k] through Eq. (21), u[k] can be determined if uf [k] is given. In this study, uf [k] is computed using the
shear-balance method proposed by Lu et al. [40]. The computational method is elaborated below.

3.2. Determination of passive friction force, uf [k]

The shear-balance method starts with the assumption that the friction interface of the PFD in the LCFD is in its sticking
state within the interval between the kth and (k1)th time steps. As a result, the damper stroke of the kth step is equal to that
of the (k1)th step, i.e.:

xr ½k ¼ xr ½k  1 (29)

Then, substituting Eqs. (1), (2) and (22) into Eq. (29) yields:

xr ½k  1  rp ½kðxb ½k  Dd z½kÞ ¼ 0 (30)

In Eq. (30), x1 ½k ¼ Dd z½k is employed, where Dd ¼ ½0/0 1 is used to extract the displacement of the first floor x1 ðtÞ from
the state vector zðtÞ. Furthermore, let u~ f ½k be the damper friction force when the damper is assumed to be in its sticking state.
Substituting Eq. (21) into Eq. (25) and letting uf ½k ¼ u ~ f ½k yields:


~ f ½k þ Ed1 x€g ½k  1 þ Ed2 x€g ½k
z½k ¼ Ad z½k  1 þ Bd1 u½k  1 þ Bd2  rp ½ku (31)

~ f ½k and obtain:

Next, by substituting xb[k] and z[k] from Eqs. (17) and (31) into Eq. (30), respectively, one may solve for u

1  
~ f ½k ¼ C1 ½krp ½k2
u C2 ½krp ½k  C3 ½k (32)

where:

C1 ½k ¼ k1
b  Dd Bd2 (33)
 
C2 ½k ¼ Dd Ad z½k  1 þ Bd1 u½k  1 þ Ed1 x€g ½k  1 þ Ed2 x€g ½k (34)

C3 ½k ¼ xr ½k  1 (35)

Notably, the value of force u~ f ½k calculated using Eq. (32) may not be the actual friction force, since it is obtained under the
assumption that the PFD damper is in its sticking state. When the computed magnitude of u ~ f ½k exceeds the maximum friction
force (slip force) of the damper, denoted by uf ;max, it signifies that the PFD should be in its slip state, rather than sticking state.
In this case, the friction force of the PFD is equal to uf ;max . Otherwise, the PFD friction force is equal to u ~ f ½k. Therefore, the
actual friction force uf ½k should be expressed as:

  

~  ~ f ½k
uf ½k ¼ min u f ½k; uf ;max sgn u (36)

where min (x, y) represents taking the smaller value of (x, y) and sgn (x) represents taking the sign of x. If the Coulomb friction
model is adopted, the slip force uf ;max of the damper can be determined as:

uf ;max ¼ m N (37)

where m and N represent the friction coefficient and the normal force (clamping force) applied at the friction interface,
respectively. If rp[k] has already been determined by a control law, Eqs. (32) and (36) can be used to calculate the actual
damper force uf ½k at any time instant. The computed uf ½k can then be used in Eq. (21) to calculate the semi-active friction
force u[k]. Finally, u[k] is substituted into Eq. (25) to compute the system response z[k] at the kth time step. The numerical
procedure of using the shear-balance method described above is summarized as a computational flow chart in Fig. 4.

4. Prototype LCFD system for shaking table test

To verify the developed LCFD theory, a prototype LCFD system was designed, fabricated, and tested dynamically using a
shaking table test. A diagram illustrating the configuration of the prototype system is given in Fig. 5. As a preliminary
experimental study, the prototype system is intended to mimic a single-DOF structural system controlled by an LCFD. This
simple structural system was chosen so that the characteristics of the LCFD system itself under seismic excitation can be
192 L.-Y. Lu et al. / Journal of Sound and Vibration 412 (2018) 184e206

Fig. 4. Flowchart of numerical simulation procedure.

observed more clearly. Fig. 6 shows the mathematical model of the tested system. The results of the test are compared with
the theoretical results in the next section.
As shown in Fig. 5, the prototype can be roughly divided into four parts, namely the (1) sliding frame, (2) leverage system,
(3) passive friction damper (PFD), and (4) driving and control system (pivot position control system). A detailed description of
each part and its function are presented below.

(1) Sliding frame: The sliding frame is primarily used to simulate a single-DOF structural system. It comprises springs,
guide rails, and a device to simulate the bracing. The springs are used to mimic the stiffness of the simulated structure.
 L.-Y. Lu et al. / Journal of Sound and Vibration 412 (2018) 184e206 193

Fig. 5. Prototype LCFD system for the shaking table test.

Several mass blocks are mounted on the platform of the sliding frame to simulate the structural mass. The low-friction
guide rails restrain the movement of the structure in the horizontal direction. The bracing device, which connects the
lever arm (at point D) and the sliding frame, is used to simulate the stiffness effect of the bracing (see Fig. 6). As shown
in the enlarged part of Fig. 5, the bracing device is composed of a pair of springs and a load cell that is utilized to
measure the transmitted semi-active force uðtÞ.
(2) Leverage system: As shown in Fig. 5, the leverage system comprises a lever arm with a movable central pivot (point P).
The lever arm is joined to points D and A at its two ends, respectively. Point D is connected to the sliding frame via the
bracing device, and point A is joined to the PFD via a linear bearing. The movement directions of points D and A are
parallel to that of the sliding frame. In addition, Fig. 5 indicates that the movement direction of point P is perpendicular
to those of the sliding frame and points D and A. The initial length L of the lever arm is 0.38 m, and the maximum
displacement of point P is ±0.087 m.
(3) PFD: The function of the PFD is to generate a constant passive slip force, which is then transferred into a controllable
damper force through the leverage system. As shown in the enlarged image in Fig. 7, the PFD primarily comprises a pre-
compression screw, friction pads, and a friction bar. The friction pads are made of brass, which has a stable friction
property. The pre-compression screw produces a normal (clamping) force between the friction pads and friction bar,
and exerts a stable friction force when relative motion occurs between the friction interfaces. The exerted friction
damping force is transferred to the sliding frame via the lever arm.
(4) Driving and control system: The driving and control system, which is used to drive pivot P to the desired position on
the lever arm, consists of a lead screw set, servo motor, motor controller box, analog-to-digital (A/D) motion card,

Fig. 6. Mathematical model of the prototype LCFD system.

194 L.-Y. Lu et al. / Journal of Sound and Vibration 412 (2018) 184e206

Fig. 7. Photograph of passive friction damper in prototype LCFD.

Fig. 8. Control block-diagram of prototype LCFD.

personal computer (PC), and sensors. The control block diagram of this system is illustrated in Fig. 8. As shown, based
on the sensor feedback signal and control law, the on-line control command b x p of the pivot displacement is first
computed by the PC and transmitted to the motor controller through the A/D motion card, and then the servo motor
and lead screw set drive pivot P to the designated position.

For practical applications, the design procedure of an LCFD may involve three major steps. In the first step, the maximum
stroke and slip force of the PFD are usually determined for a given earthquake intensity, which can be the design earthquake
level specified in a seismic design code. In the second step, the details of the leverage mechanism, including the total length
and cross-sectional dimensions of the level arm, are determined according to the range of xp or rp decided by a control law. In
the third step, the details of the driving and control system, including the power and rotary speed of the server motor and the
pitch of the lead screw, are determined based on the demand of the control law. The above design process is usually iterative
and requires the use of the numerical method proposed in Section 3.

5. Experimental verification by shaking table test

5.1. Test setup and system parameters

Fig. 9 illustrates the test setup and instrumentation of the shaking table test. The parametric values for the tested system,
i.e., a single-DOF structure with the prototype LCFD, are tabulated in Table 1. The listed values were determined from
component tests conducted prior to the shaking table test. As shown in the table, the single-DOF structure has a natural
period of 1 Hz. The total structural mass m1 is 59 kg, which is the combination of the mass of the sliding frame (38 kg) and
added mass blocks (21 kg). The friction coefficient of the guide rails is about 0.026, which leads to an equivalent structural
damping ratio of about 2.9%. In addition, the bracing stiffness is about 21.2 times the structural stiffness, i.e., kb ¼ 21:2 k1 , in
order to reflect the common structural condition.
Fig. 9 also depicts the sensor deployment. A displacement sensor (linear variable differential transformer) and an accel-
erometer were installed on the sliding frame to measure the response of the controlled structure, and a displacement sensor
 L.-Y. Lu et al. / Journal of Sound and Vibration 412 (2018) 184e206 195

Fig. 9. Setup of the shaking table test.

Table 1
Parametric values of prototype LCFD and single-DOF structure in the test.

System Parameter Valuea

Single-DOF Structural mass (m1) 59 kg
Structural system Structural stiffness (k1) 2317.21 N/m
Equivalent damping (c1) 21.67 N/(m/s)
Bracing stiffness (kb) 49143 N/m
Natural frequency ðu1 Þ 1 Hz
Equivalent damping ratio ðz1 Þ 2.9%
LCFD system Passive damper slip-force (ud,max) 0.076 W
Pivot friction force (up,max) 0.016 W
Total passive friction force (uf,max ¼ up,max þ ud,max) 0.092 W
Lengthen of lever-arm (L) 0.38 m
Range of pivot displacement xp (xp,max, xp,min) (0.21 L, 0.21 L)
PDC controller parameter Range of amplification factor rp (rp,max, rp,min) (2.45, 0.41)
Threshold story-drift, x1,max 0.125 m (for test in Sec. 5)
0.05 m (for simulation in Sec. 6)
a
Symbol W represents the weight of the controlled structure, i.e., W ¼ m1 g.

was installed on the PFD to record the movement xr of the damper. As also shown in Fig. 9, two load cells were used to
measure the passive friction force ud and the normal force N of the PFD. An additional load cell (see Fig. 5) and a displacement
meter (not shown in Fig. 9) were installed under the sliding frame to measure the variation of the semi-active friction force u
and pivot displacement xp, respectively.

5.2. Effect of pivot friction

In the test, the prototype LCFD was inevitably affected by friction forces in the leverage mechanism. These friction forces,
which primarily result from pivot rotation motion and the sliding motion of the linear bearings embedded at the three joints
(A, P, and D; see Figs. 5 and 6), are called pivot friction. To thoroughly investigate the effect of the pivot friction on LCFD
behavior, a pre-test prior to the formal shaking table test was conducted; the results are shown in Figs. 10e12 and summaried
in Table 2. Figs. 10e12 compare the measured hysteresis loops of ud ðtÞ (the passive friction of the PFD) and uðtÞ (the equivalent
semi-active friction force) for three pivot positions, namely xp ¼ 8.0, 0.0, and þ8.0 cm, which correspond to three ampli-
fication factors, namely rp ¼ 0.41, 1.00, and 2.45, respectively. In Figs. 10e12, subplots (a) and (b) show the hysteresis loops of
ud ðtÞ and uðtÞ, respectively. The cases xp ¼ 8.0 and þ8.0 cm represent the extreme pivot positions in the control law used in
the test. Notably, in the tests for Figs. 10e12, the pivot was fixed at the three pre-set positions mentioned above, and kb ¼ ∞
(i.e., the springs that mimic bracing were removed; see Fig. 5). The test results are summarized in Table 2. In the table, ud;max
denotes the average slip force (maximum friction) of ud ðtÞ measured in the test; it was estimated as:
196 L.-Y. Lu et al. / Journal of Sound and Vibration 412 (2018) 184e206

Fig. 10. Friction hysteresis loops for rp ¼ 0.41 (kb ¼ rigid, sinusoidal excitation, PGA ¼ 0.15g): (a) passive force (ud vs. xr ) and (b) semi-active force (u vs. x1 ).

Fig. 11. Friction hysteresis loops for rp ¼ 1.00 (kb ¼ rigid, sinusoidal excitation, PGA ¼ 0.20g): (a) passive force (ud vs. xr ) and (b) semi-active force (u vs. x1 ).

4  
1X  ðiÞ 
ud;max ¼ u  (38)
4 i¼1 d;max

ð1Þ ð4Þ
where ud;max … ud;max are the friction forces measured at the four corner points of the ud ðtÞ hysteresis loop in Figs. 10(a), 11(a),
and 12(a). Similarly, umax , which represents the average slip force (maximum friction) of uðtÞ, was also computed using Eq.
ðiÞ ðiÞ
(38), with ud;max replaced by umax ði ¼ 1…4Þ, which represent the friction forces measured at the four corner points of the uðtÞ

Fig. 12. Friction hysteresis loops for rp ¼ 2.45 (kb ¼ rigid, sinusoidal excitation, PGA ¼ 0.45g): (a) passive force (ud vs. xr ) and (b) semi-active force (u vs. x1 ).
 L.-Y. Lu et al. / Journal of Sound and Vibration 412 (2018) 184e206 197

Table 2
Effect of pivot friction for different pivot positions.

Pivot Theoretical amp. Ave. max. passive friction Ave. max. semi-active Exp. amp. factor Effect of pivot friction
position xp factor rp ud;max friction umax b
r p ¼ uumax
d;max
Dumax ¼ ud;max ðbr p  rp Þ
8.0 cm 0.41 57.5 N 24.2 N 0.42 0.58 N
0.0 cm 1.00 50.9 N 60.2 N 1.18 9.16 N
þ8.0 cm 2.45 53.3 N 187.5 N 3.52 57.0 N

Fig. 13. Relation of rp and x1 determined by PDC control law.

hysteresis loop in Figs. 10(b), 11(b), and 12(b). Moreover, in Table 2, symbol b r p ¼ ðumax =ud;max Þ denotes the experimental
amplification factor. As shown, b r p is always larger than the theoretical amplification factor rp, i.e., b
r p > rp . The increment of b
r p,
Drp ¼ ð r p  rp Þ, is most likely caused by the pivot friction.
b
To quantify the effect of the pivot friction, let Dumax be the increase of the slip force umax due to the effect, which can be
evaluated experimentally using the following equation and the test data of Table 2:
 
Dumax ¼ umax  rp ud;max ¼ ud;max br p  rp (39)

where ud;max and b r p are obtained from Table 2. The last column of Table 2 summarizes the calculated values of Dumax for rp ¼
0.41, 1.00, and 2.45. It is shown that the effect of the pivot friction on umax increases nonlinearly with increasing rp . However,
for simplicity, in the numerical model of this study, the pivot friction effect is assumed to be a linear function of rp , i.e.:

Du ¼ rp up (40)

where Du is the effect of the pivot friction on u at any time instant, and up is a reference pivot friction for rp ¼ 1. Notably, Eq.
(40) is applicable for the LCFD in either the sticking or sliding state. With the above assumption, up can be treated as a passive
friction force parallel to that of the PFD, as shown schematically in Fig. 6. Accordingly, the total friction force uf in Eq. (21) is
modified as:

uf ¼ ud þ up (41)

where ud is the passive friction force due to the PFD alone. As verified by Figs. 14(a) and (b), 15(a), and (b), this simplified pivot
friction model leads to satisfactory results in simulating the measured time history responses of the LCFD. This friction model
may produce a larger modeling error when rp becomes larger, but it is more convenient for engineering applications and is
applicable for most instants of time. In practice, in order to reduce the pivot friction effect, it is suggested that smaller values
are chosen for the range of rp .
As shown in Table 1, in the test, the slip forces ud and up were identified experimentally to be ud;max ¼ 0.076W and up;max ¼
0.016W, so the total slip force uf was about uf ;max ¼ ud;max þ up;max ¼ 0:092W, where W ð¼ m1 gÞ denotes the total structural
weight. Notably, only the slip force ud;max of the PFD is adjustable through the clamping force N via the equation ud;max ¼ m N,
while the slip force up;max of the pivot friction is not. The pivot friction may also increase the driving force Fp required to move
the pivot position since, as shown in Eq. (18), Fp is proportional to the friction force uf .

5.3. Control law e proportional displacement control

In order to determine the pivot position of the prototype LCFD in real time, a suitable control law is required. Since the
LCFD is general semi-active hardware, many possible control methods may be adopted. In order to focus on the objective of
198 L.-Y. Lu et al. / Journal of Sound and Vibration 412 (2018) 184e206

Fig. 14. Comparison of experimental and numerical responses of the LCFD- controlled single-DOF structure (Sinusoidal excitation, PGA ¼ 0.16g): (a) Story drift
(displacement), (b) Structural acceleration, (c) Passive hysteresis loops (ud vs. xr ) and (d) Semi-active hysteresis loops (u vs. x1 ).

this study stated previously rather than control laws, a simple controller called proportional displacement control (PDC) is
employed in the test. This controller is developed based on the observation that an increase in the story drift (relative
displacement) of the structure usually implies an increased seismic force exerted on the structure. Thus, the LCFD should
generate a larger damping force to suppress the structural responses. Conversely, a decreased story drift signifies a decreased
seismic force, and thus the level of the effective slip force should be reduced to prevent the friction damper from entering its
sticking state. Based on the above concept, the PDC controller aims to make the effective slip force of the LCFD proportional to
the structural story drift.
As indicated in Eq. (21), the semi-active control force uðtÞ of the LCFD is regulated by adjusting rp ðtÞ, i.e., the leverage
amplification factor. Therefore, to realize the PDC controller explained above, the factor rp ðtÞ is made proportional to the story
drift x1 ðtÞ, as shown in Fig. 13, which may be stated mathematically as below:
8 
> ðr  rp;min
< p;max jx1 ðtÞj þ rp;min ; for jx1 ðtÞj  x1;max
rp ðtÞ ¼ x1;max (42)
>
:
rp;max ; for jx1 ðtÞj > x1;max

where rp;max and rp;min denote the upper and lower bounds of rp ðtÞ, respectively, and x1;max represents a threshold value of the
story drift beyond which rp ðtÞ is set to its maximum value rp;max . As illustrated in Fig. 13, the PDC comprises three controller
parameters: rp;max , rp;min , and x1;max . Conceptually, the values ofrp;min and rp;max are used to mitigate the structural response
under an earthquake with lower and higher intensity, respectively. Once rp ðtÞ is determined from Eq. (42), the displacement
command of pivot P can be calculated using the following equation, which is obtained by solving Eq. (25):
 L.-Y. Lu et al. / Journal of Sound and Vibration 412 (2018) 184e206 199

 
L rp ðtÞ  1
xp ðtÞ ¼   (43)
2 rp ðtÞ þ 1

Eqs. (42) and (43) are the PDC employed to control the prototype LCFD in the test. The major advantages of the PDC is that
it only requires very simple on-board computation and one sensor measurement x1 , and thus is more easily implemented
than other controllers.
As shown in Table 1, in the experiment, the threshold displacement x1;max was taken to be 0.125 m, and the upper and
lower bounds rp;max and rp;min were chosen to be 2.45 and 0.41, respectively, which correspond to a range of xp ¼ (0.08, 0.08)
m. These values were determined from a parametric study under the physical constraints of the prototype LCFD, such as the
rotary speed of the servo motor. Eq. (42) indicates that for a given range of ðrp;min ; rp;max Þ, a smaller value of x1;max will
require a faster change of rp ðtÞ and thus a higher pivot motor speed; therefore, the selection of x1;max is restrained by the
motor speed.

5.4. Comparison of experimental and theoretical results

To verify the analysis method developed in the previous sections, in this section, the test results are discussed and
compared with the theoretical ones. Two types of ground excitation were considered in the test. One was harmonic ground
acceleration (peak ground acceleration (PGA) ¼ 0.16 g), and the other was the ground acceleration recorded during the
1940 El Centro earthquake (PGA ¼ 0.4 g). Fig. 14 compares the experimental and simulated responses of the prototype LCFD
with PDC controller subjected to the harmonic acceleration. The experimental and simulated responses due to the El Centro
earthquake are compared in Fig. 15. There are four subplots in each of Figs. 14 and 15: (a) story drift (relative displacement) of

Fig. 15. Comparison of experimental and numerical responses of the LCFD- controlled single-DOF structure (El Centro earthquake, PGA ¼ 0.4g): (a) Story drift
(displacement), (b) Structural acceleration, (c) Passive hysteresis loops (ud vs. xr ) and (d) Semi-active hysteresis loops (u vs. x1 ).
200 L.-Y. Lu et al. / Journal of Sound and Vibration 412 (2018) 184e206

Fig. 16. Comparison of command and measured pivot displacement (El Centro earthquake, PGA ¼ 0.4g).

the structure, (b) structural acceleration, (c) passive hysteresis loop (ud vs. xr ) measured at the PFD end, and (d) semi-active
hysteresis loop (u vs. x1 ) measured at the structural end. Notably, the value of ud in subplot (c) does not include the effect of
pivot friction up . The diagrams shown in Figs. 14(c) and 15(c) are called passive hysteresis loops since they represent the
relationships between the passive friction force ud of the PFD and the slip displacement xr of the PFD, whereas those in Figs.
14(d) and 15(d) are called semi-active hysteresis loops since they represent the relationships between the semi-active force u
and the story drift (relative displacement) x1 of the first story of the controlled structure.
The following observations can be drawn from Figs. 14 and 15. (1) Subplots (a) and (b) show that both the displacement
and acceleration time responses of the tested LCFD (PCD) system very well match the simulated ones, with only slight dif-
ferences in peak values. These results suggest that the proposed analytical model is able to accurately simulate the dynamic
behavior of the LCFD-controlled structure. (2) As expected, Figs. 14(c) and 15(c) show that the passive hysteresis loop
measured at the PFD end is rectangular with a constant slip force. The figures also reveal that the cyclic friction behavior of the
PFD is relatively stable. (3) Unlike the loop in subplots (c), Figs. 14(d) and 15(d) demonstrate that the semi-active hysteresis
loop measured at the structural end is no longer rectangular and becomes two trapezoids. As expected, the slip force in these
dual-trapezoidal hysteresis loops increases proportionally with increasing structural displacement. This also indicates that
the pivot position of the LCFD was shifted by the PDC controller, so that the semi-active hysteresis loop of the LCFD was
effectively altered in the desired manner. The results of Figs. 14 and 15 prove that the friction damping force of the LCFD can be
adjusted effectively by changing its pivot position; therefore, the LCFD is a feasible realization of an SAFD. The figures also
confirm that the derived analysis method is applicable for the prediction of LCFD system responses.
Since the control efficiency of the LCFD system is also influenced by the accuracy of the pivot displacement, Fig. 16
compares the measured pivot displacement with the control command under the El Centro earthquake. The figure shows
that the actual pivot displacement was able to accurately track the control command, with a slight difference in the peak
values. This error was primarily caused by the speed limitation of the servo motor when a swift movement was required by
the control command. This in turn may have also caused the errors in the system responses shown in Figs. 14 and 15.

6. Control performance of LCFD under different seismic intensities

By using the numerical method developed above and considering more earthquakes with different intensities, in this
section, the seismic responses of structures with the LCFD(PDC) will be simulated, in order to demonstrate the adaptability of
the LCFD system. In the simulation, two types of structural systems will be considered: one is a single-DOF structure and the
other is a 3-story structure with multiple-DOFs. For comparison purpose, in each structure type, the responses of the
LCFD(PDC) will be compared with those of two passive control schemes named Passive-max and Passive-min, so that the
adaptive feature of the LCFD can be further highlighted. The Passive-max system refers to the case where the rp value of the
LCFD is fixed at its upper bound (i.e., rp ¼ rp;max ), so Passive-max is equivalent to a PFD system with a higher slip force. In the
Passive-min system, rp is fixed at its lower bound (i.e., rp ¼ rp;max ), so Passive-min is equivalent to a PFD system with a lower
slip force.
As for the ground motions, the 16 earthquake records listed in Table 3 were used in the simulation in this section. These 16
ground motion records, which can be obtained from the websites of the Pacific Earthquake Engineering Research Center,
Berkeley, or Central Weather Bureau, Taipei, were carefully selected to be representative of earthquakes with different
characteristics. The average spectral accelerations Sa of these 16 ground motions are listed in the second-last column of Table
3, when the PGA of the ground motions is scaled to 1g. The ground motions are numbered in ascending order of Sa . The
averages of the long-period spectral acceleration Sa; L , which is defined as the average spectral acceleration for the structural
 L.-Y. Lu et al. / Journal of Sound and Vibration 412 (2018) 184e206 201

Table 3
List of ground motions used in numerical study.

No. Earthquake name Station Year Ave. of spectral acc.a Sa Ave. of long- period spectral acc.b Sa; L
1 Northridge Pacoima 1994 0.600g 0.385g
2 Loma Prieta Corralitos 1989 0.757g 0.334g
3 Imperial Valley El Centro 1940 0.819g 0.470g
4 Kern county Taft 1952 0.820g 0.430g
5 Northridge Sylmar 1994 0.834g 0.560g
6 Kobe JMA 1995 0.855g 0.470g
7 Cape Mendocino Petrolia 1992 0.862g 0.515g
8 Northridge Newhall 1994 0.872g 0.488g
9 Hachinohe Hachinohe 1994 0.980g 0.691g
10 Loma Prieta Oakland 1989 1.057g 0.671g
11 Loma Prieta Lexington 1989 1.134g 0.886g
12 Imperial Valley El Centro 1979 1.145g 1.030g
13 Chi-Chi TCU075 1999 1.193g 1.010g
14 Landers Yermo 1992 1.204g 0.870g
15 Loma Prieta Hollister 1989 1.221g 0.895g
16 Chi-Chi TCU102 1999 1.493g 1.420g

Note:
a
Average of spectral acceleration in the range of structural period between 0s and 4s.
b
Average of spectral acceleration for the structural period between 1s and 4s.

period range of 1e4 s, are also listed in the table. An earthquake with a higher value of Sa; L signifies that the ground motion
had higher long-period components. As shown in the table, the selected 16 ground motions have a wide range of spectral
acceleration values and thus represent earthquakes with different frequency contents.

6.1. Comparison of control performance for a single-DOF structure

The structural parameters for the single-DOF structure were adopted from Table 1, which were obtained from the
experiment. For fair comparison, the values of the system parameters for all three damping systems mentioned above, namely
LCFD(PDC), Passive-max, and Passive-min, were also adopted from Table 1. Notably, the threshold value x1;max for the
LCFD(PDC) system was taken as x1;max ¼ 0:05 m in the simulation, in order to achieve better control performance. This
threshold value is smaller than that used in the test, which adopted a larger value of x1;max ¼ 0:125 m due to the motor speed
limitation.
Fig. 17 compares the hysteresis loops of LCFD(PDC), Passive-max, and Passive-min, when the single-DOF structure is
subjected to the Northridge (Newhall) earthquake listed in Table 3. Two PGA levels, i.e., 0.1 and 0.6 g, representing earth-
quakes with small and large intensities, respectively, are considered in the figure. Fig. 18 compares the time histories of the
accumulative energy dissipated by the three control systems for the two different earthquake intensities. Note that the
amount of accumulative energy Eu ðtÞ dissipated at a given time instant t in Fig. 18 was calculated using the following
equation:

Fig. 17. Comparison of hysteresis loops of the three damping systems under two different PGA levels (Northridge-Newhall earthquake): (a) PGA ¼ 0.1g and (b)
PGA ¼ 0.6g.
202 L.-Y. Lu et al. / Journal of Sound and Vibration 412 (2018) 184e206

Fig. 18. Comparison of accumulated energy dissipated by the three damping systems under two different PGA levels (Northridge-Newhall earthquake): (a)
PGA ¼ 0.1g and (b) PGA ¼ 0.6g.

Z t
Eu ðtÞ ¼ uðtÞ dxD ðtÞ (44)
0

where xD ðtÞ denotes the displacement of point D in Fig. 1. Note that the amount of energy in Fig. 18 is negative, since it
represents the portion of structural energy dissipated by the dampers. Fig. 17(a) reveals that in a smaller earthquake
(PGA ¼ 0.1 g), the hysteresis loop of the LCFD(PDC) system is close to that of the Passive-min system. Accordingly, Fig. 18(a)
shows that the amounts of seismic energy dissipated by LCFD(PDC) and Passive-min are almost the same, and are much larger
than that of the Passive-max system, in which the friction damper was not activated for most of the earthquake duration. In a
severe earthquake (PGA ¼ 0.6 g), Fig. 17(b) shows that the hysteresis loop of the LCFD is enlarged by the PDC controller and
becomes much larger than that of Passive-min. Consequently, Fig. 18(b) shows that the amount of energy dissipated by
LCFD(PDC) at PGA ¼ 0.6 g is much higher than that of Passive-min, and is close to that of Passive-max. Figs. 17 and 18
demonstrate the excellent adaptability of the proposed LCFD(PDC) system.
For comparison of time history responses, Fig. 19(a) and (b) compare the displacement responses of LCFD(PDC) and
Passive-min under the Northridge (Newhall) earthquake with PGA values of 0.1 and 0.6 g, respectively, and Fig. 20(a) and (b)
compare the displacement responses of LCFD(PDC) and Passive-max. The response of the Northridge (Newhall) earthquake
was chosen for demonstration because the average spectral acceleration of this earthquake is around the median of the 16
ground motions considered. Fig. 19 shows that the displacement response of LCFD resembles that of Passive-min (smaller slip
force) in the low-intensity earthquake (PGA ¼ 0.1 g), but not in the high-intensity earthquake (PGA ¼ 0.6 g). In contrast, Fig. 20
shows that the responses of the LCFD (PDC) and Passive-max (higher slip force) are more alike in the high-intensity earth-
quake (PGA ¼ 0.6 g) than in the low-intensity earthquake. The above figures show how the seismic response of a structure can
be regulated by the LCFD system according to seismic intensity level.

Fig. 19. Comparison of displacement responses of LCFD and Passive-Min systems under two PGA levels (Northridge-Newhall earthquake): (a) PGA ¼ 0.1g and (b)
PGA ¼ 0.6g.
 L.-Y. Lu et al. / Journal of Sound and Vibration 412 (2018) 184e206 203

Fig. 20. Comparison of displacement responses of LCFD and Passive-Max systems under two PGA levels (Northridge-Newhall earthquake): (a) PGA ¼ 0.1g and (b)
PGA ¼ 0.6g.

6.2. Comparison of control performance for a 3-story structure

In order to investigate the control performance of the LCFD when installed in a more realistic multiple-DOF structure
system, in this subsection, a 3-story structural system is considered. The peak responses of this structure with the three
control systems, namely, the LCFD(PDC), Passive-max and Passive-min will be compared. The schematic diagram of this
structure with the LCFD can be represented by Fig. 1 (with n ¼ 3), and the structural matrices that were obtained from a
system identification test performed on a real steel frame are listed in Table 4. The frame was erected in the earthquake
simulation laboratory of the National Center for Research on Earthquake Engineering (NCREE, Taipei) and has been used for
numerous studies on advanced technology of seismic structural control [40,41]. The three natural frequencies of the 3-story
frame were identified to be 1.07, 3.48 and 5.98 Hz. The control parameters for the LCFD(PDC) used in the simulation are also
listed in the lower part of Table 4. Basically, these parameters resemble those used for the single-DOF structure. Notably, the
threshold value of the first-story displacement for the PDC controller is taken as x1;max ¼ 0:02 m, which is about one-third of
x1;max in the single-DOF structure, since the structure has three stories. The bracing stiffness for the LCFD is taken to be kb ¼
97371 kN/m that yields about the same bracing-to-structure stiffness ratio for the single-DOF structure.
To investigate the adaptive feature of the LCFD, Figs. 21 and 22 compare the average peak responses of the 3-story
structure controlled by the three damper systems subjected to the 16 ground motions (see Table 2) with the PGA scaled to
0.1g (low intensity) and 0.3g (high intensity), respectively. The averages of the peak displacement and story-shear at each
floor level are depicted in subplots (a) and (b), respectively, since these responses have practical importance in earthquake
engineering. In the figures, the average peak responses of an uncontrolled system, which refers to the structure without any
kind of damper systems, are also depicted for comparison. From Fig. 21(a) and (b), it is observed that for lower-intensity
earthquakes (PGA ¼ 0.1g), the LCFD(PDC) has the lowest displacement responses at the second and third floors, while its
story-shears are close to those of the Passive-min, which has the lowest story shears at all floor levels. As for the Passive-max,
its story shears at all floor-levels are almost equal to those of the uncontrolled system, and its top-floor displacement exceeds
that of the Passive-min, even although the Passive-max has the lowest displacement at the first floor. This implies that the
Passive-max may not be activated due to its high slip force that yields almost no control effect in earthquakes with a lower

Table 4
Structural and system parameters of the 3-story structure with LCFD [41].

Parameter Values
2 3
Mass matrix 10:914 0 0
M¼4 0 11:213 0 5 ton
0 0 11:213
2 3
Damping matrix 2:124 1:131 0:093
C ¼ 4 1:131 2:364 0:209 5 kN=ðm=sÞ
0:093 0:209 2:325
2 3
Stiffness matrix 3673:845 4555:417 1172:960
4
K ¼ 4556:112 8753:602 5451:810 5 kN=m
1172:190 5452:178 9186:159
Natural frequencies 1.07, 3.48 and 5.98 Hz
LCFD parameter (all parameters are the same as those in Table 1)
PDC controller Range of amp. factor (rp,max, rp,min) ¼ (2.45, 0.41)
Threshold story-drift, x1,max ¼ 0.02 m
204 L.-Y. Lu et al. / Journal of Sound and Vibration 412 (2018) 184e206

Fig. 21. Comparison of average peak responses at three floor levels for different damping systems (considering 16 ground motions, PGA ¼ 0.1g): (a) Peak floor
displacement, (b) Peak story shear.

Fig. 22. Comparison of average peak responses at three floor levels for different damping systems (considering 16 ground motions, PGA ¼ 0.3g): (a) Peak floor
displacement, (b) Peak story shear.

intensity. On the other hand, for earthquakes with a higher intensity (PGA ¼ 0.3g), Fig. 22(a) and (b) show that the LCFD(PDC)
has the second lowest displacement and shear responses at all floor-levels, and are very close to those of the Passive-max,
which has the lowest responses. As for the Passive-min, even though it is able to substantially reduce the responses of the
3-story structure as compared to the uncontrolled responses, the Passive-min induces the highest floor-displacement re-
sponses that are substantially higher than those of the LCFD(PDC) and Passive-max. The above observation implies that the
seismic energy dissipated by the Passive-min may be insufficient, so its control performance is unsatisfactory. Figs. 21 and 22
together demonstrate that the variable friction property has made the LCFD system adaptive to the earthquake intensity and
is advantageous for the seismic response mitigation of a multiple-DOF structural system. This finding is consistent with those
observed in the previous studies on multiple-DOF structures with SAFDs [22,26,31,41], in which more sophisticated control
laws were applied and the superiority of the control performance of SAFDs can be even more clearly observed.

7. Conclusions

The present study developed a position-controlled SAFD called LCFD that is different from most existing SAFDs, which are
realized using a force-controlled mechanism. The LCFD system primarily consists of a cost-effective PFD and a leverage
mechanism with a movable central pivot. The controllable friction force of the LCFD is achieved by adjusting the position of
the pivot in real time, rather than adjusting the clamping force in the SAFD. To provide an analytical tool, the governing
equation of motion for a structure equipped with the LCFD was derived using Lagrange's equation, and a numerical procedure
 L.-Y. Lu et al. / Journal of Sound and Vibration 412 (2018) 184e206 205

based on the derived equation was introduced. To verify the feasibility of the proposed system and the applicability of the
analytical tool, a seismic simulation test was conducted on a prototype LCFD system using a shaking table. For easy imple-
mentation, a simple PDC controller, whose control goal is to produce a friction force proportional to the story drift of the
controlled structure, was adopted in the test. Finally, in the numerical section, the adaptive feature of the LCFD with the PDC
controller was demonstrated by comparing the responses of the LCFD with those of two types of PFDs, one with a lower slip
force (Passive-min) and one with higher slip force (Passive-max), respectively. The conclusions drawn from the present study
are summarized below.

1. In the derived governing equation of motion, the controllable semi-active friction force of the LCFD is expressed explicitly
in terms of the position of the movable central pivot. This expression is advantageous for the implementation of the
proposed system, and is also a major difference from those for most existing SAFDs, which are usually controlled by
varying the clamping force applied at friction interfaces.
2. The results of the shaking table test demonstrate that the semi-active friction force and hysteresis loop of the prototype
LCFD can be attenuated by the leverage mechanism in the desired manner. The test results also show consistency between
the numerical and experimental responses of the LCFD, and thus confirm the accuracy of the developed analysis method
for the LCFD.
3. By a comparison with two PFD systems with a higher and a lower constant slip force, respectively, the numerical study
showed that the hysteresis loop and energy dissipation capacity of the LCFD with the PDC controller resemble those of the
Passive-min system in a lower-intensity earthquake, and resemble those of the Passive-max system in a higher-intensity
earthquake. This demonstrates that the friction behavior of the LCFD system is adaptive to seismic excitations, so that
better energy dissipation capacity can be achieved for earthquakes of different intensities.
4. Based on the simulated results of a LCFD-controlled 3-story structure subjected to 16 different ground motions scaled to a
lower and a higher PGA levels, it was observed that for the ground motions with the lower PGA, the average peak story-
shears at all floor levels due to the LCFD system are close to those of the Passive-min system that retains the lowest story
shears among the three control cases, while the LCFD retains the lowest top-floor displacement. Conversely, for the ground
motions with the higher PGA, where the Passive-min possesses the largest responses among the three control cases, the
peak floor-displacements and story-shears of the LCFD are significantly reduced and close to those of the Passive-max
system that retains the lowest responses. This again confirms the adaptiveness of the LCFD system.

Acknowledgement

This research was supported in part by the Ministry of Science and Technology, Taiwan, through grant MOST103-2625-M-
006-010. The authors are grateful to the Department of Construction Engineering, National Kaohsiung First University of
Science and Technology, Kaohsiung, for their technical support in conducting the shaking table test, and also to Mr. Wei-Hao
Ke at National Chiao Tung University for preparing some of numerical simulation results.

References

[1] T.T. Soong, B.F. Spencer, Supplemental energy dissipation: state-of-the-art and state-of-the practice, Eng. Struct. 24 (2002) 243e259.
[2] M.D. Symans, F.A. Charney, A.S. Whittaker, M.C. Constantinou, C.A. Kircher, M.W. Johnson, R.J. McNamara, Energy dissipation systems for seismic
applications: current practice and recent developments, J. Struct. Eng. 134 (2008) 3e21.
[3] F. Mazza, Nonlinear seismic analysis of unsymmetric-plan structures retrofitted by hysteretic damped braces, Bull. Earthq. Eng. 14 (2016) 1311e1331.
[4] F. Mazza, E. Pedace, F. Del Favero, Effectiveness of damped braces to mitigate seismic torsional response of unsymmetric-plan buildings, Mech. Syst.
Signal Process. 85 (2017) 610e624.
[5] C. Pasquin, H. Charania, R. Steele, R. Pall, A. Pall, Friction-dampers for seismic control of Selkirk Water Front Offices, Victoria, in: Proceedings of the 6th
U.S. National Conference on Earthquake Engineering, Seattle, Washington, 1998. May.
[6] R.T. Pall, G. Gauthier, S. Delisle, A. Pall, Friction-dampers for seismic upgrade police headquarters, Montreal, in: Proceedings of the 12th World
Conferences on Earthquake Engineering, Auckland, New Zealand, 2000, 2014, January.
[7] I. Mualla, B. Belev, Overview of recent projects implementing rotational friction dampers, in: Proceedings of the 16th World Conference on Earthquake
Engineering, Santiago, Chile, 2016, 1667, January.
[8] R. Hanson, T.T. Soong, Seismic Design with Supplemental Energy Dissipation Devices, Monograph Series MNO-8, EERI, Oakland, California, 2001.
[9] ASCE 41-13, Seismic Evaluation and Retrofit of Existing Buildings, ASCE Standard ASCE/SEI 41e13, American Society of Civil Engineers, Reston, Virginia,
2014.
[10] K.W. Min, J.Y. Seong, J. Kim, Simple design procedure of a friction damper for reducing seismic responses of a single-story structure, Eng. Struct. 32
(2010) 3539e3547.
[11] F.F.L. Miguel, F.F.L. Miguel, R.H. Lopez, Robust design optimization of friction dampers for structural response control, Struct. Control Health Monit. 21
(2014) 1240e1251.
[12] N. Nabid, I. Hajirasouliha, M. Petkovski, A practical method for the design of friction-based passive control wall damper for RC frame buildings, in:
Proceedings of the 16th World Conference on Earthquake Engineering, Santiago, Chile, 2016, 1111, January.
[13] Z. Akbay, H.M. Aktan, Abating earthquake effects on buildings by active slip brace devices, Shock Vib. 2 (1995) 133e142.
[14] M.D. Symans, M.C. Constantinou, Semi-active control systems for seismic protection of structures: a state-of-the-art review, Eng. Struct. 21 (1999)
469e487.
[15] B.F. Spencer, S. Nagarajaiah, State of the art of structural control, J. Struct. Eng. 129 (2003) 845e856.
[16] T.E. Saaed, G. Nikolakopoulos, J.E. Jonasson, H. Hedlund, A state-of-the-art review of structural control systems, J. Vib. Control 21 (2015) 919e937.
[17] J.A. Inaudi, Modulated homogeneous friction: a semi-active damping strategy, Earthq. Eng. Struct. Dyn. 26 (1997) 361e376.
[18] C.L. Ng, Y.L. Xu, Semi-active control of a building complex with variable friction dampers, Eng. Struct. 29 (2007) 1209e1225.
206 L.-Y. Lu et al. / Journal of Sound and Vibration 412 (2018) 184e206

[19] L.Y. Lu, G.L. Lin, C.Y. Lin, Experiment of an ABS-type control strategy for semi-active friction isolation systems, Smart Struct. Syst. 8 (2011) 501e524.
[20] S. Kannan, H.M. Uras, H.M. Aktan, Active control of building seismic response by energy dissipation, Earthq. Eng. Struct. Dyn. 24 (1995) 747e759.
[21] H.R. Samani, M. Mirtaheri, A.P. Zandi, Experimental and numerical study of a new adjustable frictional damper, J. Constr. Steel Res. 112 (2015)
354e362.
[22] M. Laflamme, D. Taylor, M.A. Maane, J.J. Connor, Modified friction device for control of large-scale systems, Struct. Control Health Monit. 19 (2012)
548e564.
[23] L. Cao, A. Downey, S. Laflamme, D. Taylor, J. Ricles, Variable friction device for structural control based on duo-servo vehicle brake: modeling and
experimental validation, J. Sound Vib. 348 (2015) 41e56.
[24] A. Downey, L. Cao, S. Laflamme, D. Taylor, J. Ricles, High capacity variable friction damper based on band brake technology, Eng. Struct. 113 (2016)
287e298.
[25] J.N. Yang, A.K. Agrawal, Semi-active hybrid control systems for nonlinear building against near-fault earthquakes, Eng. Struct. 24 (2002) 271e280.
[26] A. Nishitani, Y. Nitta, Y. Ikeda, Semi-active structural control based on variable slip-force level dampers, J. Struct. Eng. 129 (2003) 933e940.
[27] S. Narasimhan, S. Nagarajaiah, Smart base isolated buildings with variable friction systems: H∞ controller and SAIVF device, Earthq. Eng. Struct. Dyn.
35 (2006) 921e942.
[28] Y.L. Xu, W.L. Qu, Z.H. Chen, Control of wind-excited truss tower using semi-active friction damper, J. Struct. Eng. 127 (2001) 861e868.
[29] C. Chen, G. Chen, Shaking table test of quarter-scale building model with piezoelectric friction dampers, Struct. Control Health Monit. 11 (2004)
239e257.
[30] L.Y. Lu, C.C. Lin, G.L. Lin, C.Y. Lin, Experiment and analysis of a fuzzy-controlled piezoelectric seismic isolation system, J. Sound Vib. 329 (2010)
1992e2014.
[31] O.E. Ozbulut, M. Bitaraf, S. Hurlebaus, Adaptive control of base-isolated structures against near-field earthquakes using variable friction dampers, Eng.
Struct. 33 (2011) 3143e3154.
[32] G.L. Lin, C.C. Lin, L.Y. Lu, Y.B. Ho, Experimental verification of seismic vibration control using semi-active friction tuned mass damper, Earthq. Eng.
Struct. Dyn. 41 (2012) 813e830.
[33] Y.M. Park, K.J. Kim, Semi-active vibration control of space truss structures by friction damper for maximization of modal damping ratio, J. Sound Vib.
332 (2013) 4817e4828.
[34] H. Garrido, O. Curadelli, D. Ambrosini, Semi-active friction tendons for vibration control of space structures, J. Sound Vib. 333 (2014) 5657e5679.
[35] G. Song, V. Sethi, H.N. Li, Vibration control of civil structures using piezoceramic smart materials: a review, Eng. Struct. 28 (2006) 1513e1524.
[36] J. Pardo-Varela, J.C. Llera, A Semi-active piezoelectric friction damper, Earthq. Eng. Struct. Dyn. 44 (2015) 333e534.
[37] I. Mualla, Experimental evaluation of new friction damper device, in: Proceedings of the 12th World Conference on Earthquake Engineering, Auckland,
New Zealand, 2000, 1048, January.
[38] A.K. Chopra, Dynamics of Structures, fourth ed., Peason Education Limited, Harlow, England, 2014.
[39] L. Meirovitch, Dynamics and Control of Structures, John Wiley & Sons, Hoboken, New Jersey, 1990.
[40] L.Y. Lu, L.L. Chung, L.Y. Wu, G.L. Lin, Dynamic analysis of structures with friction devices using discrete-time state-space formulation, Comput. Struct.
84 (2006) 1049e1071.
[41] L.Y. Lu, Predictive control of seismic structures with semi-active friction dampers, Earthq. Eng. Struct. Dyn. 33 (2004) 647e668.