Engineering Structures 190 (2019) 276–284

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Reliability-based optimum design of passive friction dampers in buildings in T

seismic regions
⁎
Sergio Pastor Ontiveros-Péreza, , Letícia Fleck Fadel Miguela, Jorge Daniel Rierab
a
Department of Mechanical Engineering, Federal University of Rio Grande do Sul, Av. Sarmento Leite, 425, 2 andar, CEP 90050-170 Porto Alegre, RS, Brazil
b
Department of Civil Engineering, Federal University of Rio Grande do Sul, Av. Osvaldo Aranha, 99, 3 andar, CEP 90035-190 Porto Alegre, RS, Brazil

A R T I C LE I N FO A B S T R A C T

Keywords: This paper describes an approach to design friction damper systems for installation in buildings located in
Reliability-based optimization seismic regions using a novel optimization technique: the Search Group Algorithm (SGA). The method aims at
Passive friction dampers reducing the probability of failure of typical concrete buildings and is illustrated with a case study for Cúcuta,
Seismic hazard Colombia. To characterize seismic risk, records of ground acceleration during seismic events are used to de-
termine the expected value of the peak ground acceleration (PGA) annual exceedance rate for the site of interest.
Seismic risk was estimated by fitting probability distribution functions of the PGA for periods of one and fifty
years, with the objective of evaluating, through numerical simulation, the reliability of a concrete structure. It is
confirmed that a significant increase of the system reliability may be achieved through the use of an optimized
passive friction dampers system. Robust estimators of the reliability are determined with a database of seismic
records with markedly different characteristics. The results highlight the efficiency of passive friction dampers to
reduce the probability of failure in structures located in regions with high seismic activity.

1. Introduction multiple tuned mass dampers (MTMDs) in buildinngs subjected to

seismic excitation considering uncertainties in the structural para-
Practical applications of energy dissipation systems experienced an meters, in the dynamic load and also in the MTMDs design with the aim
explosive growth at the end of the 20th Century, leading to remarkable to obtain a robust design. Lavan and Amir [10] propose a method for
progress in methods to control structural response to seismic excitation. the optimization of the position and the parameters of viscous dampers
Passive devices have so far been more widely used in view of their low to carry out the seismic retrofitting of large-scale structures considering
cost of construction and maintenance, but also on account of their in- a deterministic input. Castoldo and De Iuliis [11] propose an integrated
dependence of an external energy source. Descriptions of passive de- seismic design of the elastic stiffness and viscoelastic properties of a
vices such as the tuned mass damper, dry friction, viscoelastic or me- dissipative bracing-damper system, within the displacement-based
tallic dampers, that demonstrate their effectiveness and suggest seismic design.
methods of design were published, for example, by Curadelli and Riera Note that, although friction dampers are undoubtedly competitive
[1], Miguel and Riera [2], Min et al. [3], Rakicevic et al. [4], Monir and in relation to other passive devices, few references focus attention, for
Zeynali [5] and Belash [6]. random excitation, on the optimization of their parameters or on their
In this context, methods to optimize damper systems design in best location within the structure of interest. The optimization studies
practical applications still constitute a topic of current research, al- of damping systems found in the literature, such as the contributions of
though most contributions aim at achieving an optimum solution for Miguel et al. ([12–16]) consider only deterministic excitation. How-
deterministic excitation, i.e., they do not account for the inherent un- ever, it must be underlined that the various uncertainties present in the
certainty of the excitation. For instance, Hadi and Arfiadi [7] resort to problem may alter the optimum solution, which justifies the present
genetic algorithms to find the optimum design of a tuned mass damper proposal to take uncertainties into account in the optimization process.
installed in a multiple degrees of freedom structure while Fadel Miguel Thus, the main contribution of the present paper is a methodology
et al. [8] propose a method for optimum design of multiple tuned mass for the simultaneous optimization of the friction forces and the posi-
dampers in buildings. Similarly, Vellar et al. [9] describe a new meth- tions of friction dampers to be installed in buildings subjected to seismic
odology for simultaneous optimization of parameters and positions of excitation, considering the random nature of earthquake motion and

⁎
Corresponding author.
E-mail address: [email protected] (S.P. Ontiveros-Pérez).

https://doi.org/10.1016/j.engstruct.2019.04.021
Received 19 December 2018; Received in revised form 4 April 2019
Available online 15 April 2019
0141-0296/ © 2019 Elsevier Ltd. All rights reserved.
S.P. Ontiveros-Pérez, et al. Engineering Structures 190 (2019) 276–284

the non-linear behavior of friction dampers, but assuming linear dy-

namic response of the structure. The last simplification results in a large
reduction of processing time, in principle, without affecting the op-
timum solution for the real system, as explained later in more detail.
It is important to note that this optimization problem is complex,
due to the following reasons: (i) it is a problem of optimization of a
dynamic system that involves the non-linear behavior of the friction
dampers; (ii) it is an optimization problem under uncertainties; (iii) the
objective function is not convex; (iv) it is an optimization problem with
mixed design variables (discrete and continuous).
Consequently, the optimization problem under uncertainties pro-
posed in this paper must be solved through optimization methods
capable of dealing with this type of problem, such as heuristic algo-
rithms. Known advantages of these algorithms include the following: (i)
they do not require gradient information and therefore can be applied Fig. 1. Comparison of the four representations of the signal function.
to problems in which the gradient is difficult to obtain or simply does
not exist; (ii) they do not get stuck in local minimums if they are Miguel and Riera [2] and Ontiveros-Pérez ([18–20]).
properly adjusted; (iii) they may be applied to non-convex or dis-
continuous functions; (iv) they provide a set of optimum solutions ra- f1 (α1, →
v (t )) = Erf(α1→
v (t ))
ther than a single solution, giving to the designer more choice to select f (α , v (t )) = tanh (α →
2 2
→ v (t ))
2
the final design and (v) they can be easily employed to solve mixed-
f3 ⎛⎜α3, →
v (t ) ⎞⎟ = ( ) ArcTan ⎛⎝α →v (t ) ⎞⎠
2
variable optimization problems. ⎜ 3 ⎟
π
A large number of heuristic algorithms can be found in the technical ⎝ ⎠
literature. Among those heuristic algorithms, the Search Group α 4→
f4 ⎛⎜α4, →
v (t ) ⎞⎟ =
v (t )
Algorithm (SGA) has shown to be very efficient, for which reason it was (1 + α 4 ∣ →
v (t ) ∣)
⎝ ⎠ (3)
chosen to solve the optimization problem proposed in this paper.
A numerical approach must be used when it is not feasible to solve
2. Problem formulation analytically the differential equations of motion. In this paper, Eq. (1) is
solved using a computational routine based on the Central Finite Dif-
The differential equation governing the motion of a multi-degree of ferences explicit method, developed by the authors using MATLAB [21]
freedom system equipped with friction dampers and subjected to a for calculating the dynamic response in the time domain of structures
seismic load can be written as: equipped with friction dampers. When applied to linear systems, the
→ → → ⎯→
⎯ Central Finite Differences method is stable for any integration interval
M z¨ (t ) + C z ̇ (t ) + K→
z (t ) = −MBy¨ (t ) − Ffn (1)
Δt smaller than a critical value, proportional to the shortest period of
in which M and K are the matrices of mass and stiffness of size n × n , vibration of the structure (Rao [22]). No similar stability criterion is
respectively, and n is the number of degrees of freedom. The damping known for non-linear systems, in which case the stability of the in-
matrix C is proportional to the matrices M and K , in which tegration procedure must be verified by periodic examination of the
C = α M + β K . The vector →z (t ) of dimension n represents the relative energy balance during the integration process. In this study, the non-
displacement with respect to the base and the differentiation with re- linear terms affect only the friction dampers, resulting in small non-
spect to the time is represented by a dot over the displacement vector linearity of the global system (Eq. (1)). No evidence of instabilities was
symbol. The matrix B of dimensions n × w contains the cosine directors found.
of the angles formed between the base motion and the direction of the
displacement of the considered degree of freedom. w is the number of
→ 3. Friction damper operation
directions of ground motion, ÿ is the vector of dimension w of the
ground acceleration due to a seismic excitation and the friction force of
⎯→
⎯ Soong and Dargush [23] reviewed early applications of passive
Coulomb is represented by the vector Ffn of dimension n. The Coulomb
dampers to reduce or control structural vibrations. Pall [24] described
friction force can be written as:
one of the first applications of Coulomb’s friction law to develop a
⎯→
⎯ ⎯→
⎯
Ffn = μ N sgn (→
v (t )) (2) damper intended to reduce seismic structural response (Pall et al. [25]).
Additional developments as well as descriptions of important applica-
⎯→
⎯
in which μ is the friction coefficient considered constant, N is the tions of friction dampers were later reported by Pall ([26,27]).
normal force vector, sgn is the signal function and → v (t ) is the relative Miguel [28] assessed experimentally a damper in which the fric-
velocity vector between the two ends of the friction damper. It should tional force occurs between two sliding brass plates. The control of
be noted that the magnitude of the friction force in Eq. (2) is constant, normal force between the two plates is achieved by two compression
but its direction is always opposite to that of the slip speed. The changes springs. Although under certain conditions normal sliding may be
in the direction of velocity cause discontinuities in the friction force, hampered by not widely known phenomena, such as stick and slip,
leading to difficulties in evaluating the response of systems subjected to properly tested friction dampers can be expected to be robust and re-
friction forces. For this reason, one of the four continuous functions liable components. Herein, it is also assumed that the bracing elements
presented in Eq. (3) and proposed by Mostaghel and Davis [17] was that connect the dampers to neighboring floors have sufficiently high
used to replace the discontinuous signal function. Each of these pro- stiffness to justify neglecting their axial flexibility or lateral instabilities
posed functions uses a parameter, αi , which controls the level of pre- in the ensuing analysis.
cision of the function representing the friction force. The comparison of The φ angle (see Fig. 2) represents the inclination of the friction
→
the variation of these functions with the slip velocity, v (t ) , for the damper with the horizontal direction of the motion, being relevant in
parameter αi = 10 (i = 1, 2, 3, 4 ) is given in Fig. 1. the calculation of the friction force of each device together with the
The continuous function implemented in this study is calculation of the difference between the friction forces of the neigh-
f2 (α2, →
v (t )) with α2 = 1 × 1050 , which was used in previoous studies by boring storeys.

277
S.P. Ontiveros-Pérez, et al. Engineering Structures 190 (2019) 276–284

Fig. 4. Type I function fitted to simulated maximum accelerations in fifty years

Cúcuta, Colombia.
Fig. 2. Typical n-storey structure equipped with friction dampers.

iterations of the optimization process, SGA tries to find promising re-

4. Assessment of seismic risk
gions in the domain (exploration), and as the iterations proceed, it re-
fines the best design in each of these promising regions (exploitation).
Based on information provided by the Colombian Association of
For this reason, the optimization process is separated into two phases:
Seismic Engineering (AIS), which specifies the expected value of the
global and local [30].
exceedance rate of the maximum annual acceleration at Cúcuta,
The optimization process starts with the randomly population P
Colombia, the expected value E(PGA) = 0.0545g and the variance
generated in the search domain. After the initial population P is gen-
Var(PGA) = 0.0025g2 of the maximum annual acceleration Amax1 at the
erated, the objective function of each individual is evaluated. After the
indicated location were determined. The parameters of a Lognormal
objective functions are evaluated, the search group R is constructed by
probability density function, shown in Fig. 3 were determined from the
selecting ng individuals from the population P . A standard tournament
statistical moments.
selection is applied to follow this step of the algorithm.
Ontiveros-Pérez [29] obtained by simulation samples of Amax50 , i.e.,
In order to increase the overall search capacity of the algorithm, the
the maximum acceleration in fifty years in Cúcuta, which led to the
search group R mutates after each iteration. This mutation strategy
expected value E(PGA) = 0.2523g and the standard deviation
consists of replacing nmut individuals of R with new individuals gen-
σ (PGA) = 0.0904g . When the parent distribution of Amax1 is Lognormal,
erated based on the statistics of the current search group. The idea is to
extreme values for large samples tend asymptotically to a Type I (also
include individuals in the search group away from the current position
known as Gumbel) probability distribution. Thus, the Type I distribution
of current members by exploring new regions of the search domain.
appears as a plausible model to describe seismic risk at the site and was
Thus, once the search group is formed, each one of its members gen-
then adopted, after verifying its performance, illustrated by Fig. 4,
erates a family. An important aspect of the algorithm is that the size of
which shows a plot of a Type I (Gumbel) probability density function
the family of a member of the search group varies according to the
fitted to the histogram of simulated maximum accelerations in fifty
quality of the objective function of said member.
years in Cúcuta.
As explained above, the SGA consists of two phases: the overall
phase and the local phase. In the first iterations itglobal
max
(global phase),
5. Optimization methodology: Search Group Algorithm (SGA) the main goal of the algorithm is to explore most of the design space.
Thus, the new search group is formed by the best member of each fa-
max
This section presents a brief explanation of the optimization algo- mily. When the number of iterations is greater than itglobal the selection
rithm used to perform the optimization proposal to be explained later. scheme is modified: the new search group is formed by the best in-
The SGA is a new meta-heuristic optimization technique, proposed by dividuals among all families. This phase is called a local because the
Gonçalves et al. [30], characterized by the way to generate new in- algorithm will tend to explore the region of the best current project. For
dividuals, allowing a proper balance between exploration and ex- more information, see Gonçalves et al. [30].
ploitation of the project domain. The basic idea is that in the first
6. Reliability based optimum design of friction dampers

According to Schuëller and Jensen [31], the basic goal in any en-
gineering discipline is to design and build systems or components that
meet certain performance objectives over their lifetime. These objec-
tives cover a wide range of possibilities, for example, control of vibra-
tions induced by wind on bridges or schemes to prevent the collapse of
buildings due to earthquakes. In any practical situation, it is not suffi-
cient to meet performance objectives deterministically, due to the in-
herently random nature of the loading conditions, structural para-
meters, and operating conditions of the structure. Thus, the
achievement of performance goals should be accomplished by prob-
abilistic means, that is, assessing the resulting reliability. In this con-
text, note that high levels of reliability are usually associated with large
economic costs. Considering that the available resources are always
limited, an adequate design must imply an adequate balance between
Fig. 3. Lognormal probability density function of the annual PGA for the city of an acceptable level of reliability and an economic design of the system.
Cúcuta, Colombia. However, few seismic structural reliability studies were found in the

278
S.P. Ontiveros-Pérez, et al. Engineering Structures 190 (2019) 276–284

literature. Castoldo et al. ([32,33]), for instance, proposed a seismic

reliability-based relationship between response factors and displace-
ment demand for nonlinear hardening and softening structures isolated
by friction pendulum system devices.
The reliability-based optimum design is a tool that allows de-
termining the best design according to some predefined criteria. This
section describes the reliability-based optimum design of a friction
dampers system for a real structure. For illustration purposes, a
building under construction in a region of Colombia with high seismic
activity was selected, and the dampers system designed to minimize the
probability of failure of the building in a fifty years period.
The formulation of the reliability-based optimum design problem
requires the identification and definition of system input variables, i.e.,
design variables and uncertain parameters, system failure events, i.e.,
violation of target performance, the constraints of the design problem
and the objective function that allows identifying the most convenient
design. Each of these items is briefly explained in the following. Fig. 5. Six-storey concrete building.

6.1. Definition of design variables Table 1

Geometrical properties of the 6-storey building.
Design variables of the reliability-based optimum design problem
Member number Area [m2 ] Ixz [m4 ] Izy [m4 ] J [m4 ]
must be selected by the engineer, aiming at optimizing the performance
of the system. In this paper, the locations of the proposed friction 1 a 12 0.18 54E-4 13E-4 67E-4
dampers in the structure and the corresponding friction forces were 13 a 36 0.15 31E-4 11E-4 42E-4
chosen as design variables. The appropriate locations for installing the 37 a 42 0.12 16E-4 9E-4 25E-4
43 a 48 0.15 31E-4 11E-4 42E-4
friction dampers in the building are represented by discrete design
85 a 96 0.12 9E-4 16E-4 25E-4
variables and the friction force (mechanical device parameter) for each 97 a 102 0.16 21E-4 21E-4 43E-4
damper are represented by continuous variables. Thus, the constraints 103 a 108 0.12 9E-4 16E-4 25E-4
of the reliability-based optimum design problem are the number of 67 70 73 76 79 82 0.06 1E-4 8E-4 9E-4
positions allowed (npfd ) in the structure for the installation of the spe- 49 a 66 0.12 9E-4 16E-4 25E-4
68 69 71 72 74 75 77 78 80 81 83 84 0.12 9E-4 16E-4 25E-4
cified maximum number of devices (nfd ) and the limits for the friction
→
forces (lowerbound ⩽ Ffn ⩽ upperbound ). P is the position vector of the
dampers, of dimension npfd , containing 0 and 1, in which 1 represents The values of dmax and z max adopted herein (Eq. (4)), employed
that there is a damper installed in that position. Thus, the maximum earlier by Miguel et al. [15], may be regarded as lower bounds for typical
→
number of 1 in the vector P is nfd . On the other hand, the friction forces reinforced concrete frame structures, since code prescriptions, numer-
Ffn of each damper are represented by continuous variables. For con- ical or experimental evidence typically suggest higher limits. The au-
venience of notation, the design variables are grouped in the design thors adopted the values indicated above to ensure that the assumption
→ ⎯⎯⎯→
vector →x = [P , Ffn]. of linear behavior of the structure is acceptable within the entire range of
the search process, ruling out concerns about the influence of nonlinear
6.2. Prediction of structure failure effects on the optimum solution. It follows, in consequence, that the
estimated failure probabilities should be regarded as upper bounds of the
As mentioned earlier, the system must meet specific performance correct values.
requirements. Violation of any of these requirements causes a system Once the structural boundary condition and the set of seismic re-
failure. It should be noted that, in this context, failure does not ne- cords for each PGA level are defined, the number of times the limit state
cessarily imply total collapse of the system, but undesirable perfor- is reached is counted. The result permits determining the so-called
mance. A failure event is defined by the so-called performance function, fragility curve, i.e., the relationship between the probability of failure of
which depends on the design variables and on the uncertainty of the the structure and the PGA level that caused failure. The conditional
parameters involved. probability of failure is estimated by the relative frequency of the
In order to conduct a seismic risk study, it is important to determine variable D, expressed in Eq. (5).
the potential damage caused by seismic events. The probability of PD = P [D = 1] = P [ dmax ⩾ 1 cm or z max ⩾ 10 cm] (5)
structural failure caused by given intensity levels of the seismic event is
described by its relation with the peak acceleration of the seismic event, The reliability analysis consists of the following steps: (i) normalize
usually known as fragility curve. each seismic record at 0.1g intervals, evaluating whether the structure
Herein, a widely used failure criterion was adopted, according to exceeds the limit state; (ii) by counting the relative number of failures
which failure occurs when the maximum relative displacement between for each PGA level, determine the fragility curve by simulation; (iii)
floors dmax exceeds a prescribed value or the maximum displacement of perform the convolution of the structural fragility curve and the fAmax50
the building z max exceeds another specified limit. A linear dynamic function representing the seismic threat in a given region.
analysis of the building for the selected random excitations was then Taking into account the relative number of structures that fail
performed. Both parameters, dmax and z max , are determined through the (observations of the random variable D), it is possible to calculate the
vector →z (t ) , which is obtained by solving Eq. (1) in the time domain statistical moments in order to fit a Lognormal probability distribution
(see Section 2). Thus, the random variable indicating failure D, which function (Eq. (6)) to the point fragilities PD obtaining the fragility curve.
represents the limit state of the structure, is defined by Eq. (4). 2
PGA 1 ⎡ 1 lnPGA − λPGA ⎞ ⎤
PD (PGA) = ∫0 exp ⎢− ⎛⎜ ⎟
⎥
= 1, If dmax ⩾ 1 cm or z max ⩾ 10 cm (fail) 2π ζPGA PGA 2⎝ ζPGA ⎠⎦ (6)
D=⎧ ⎣
⎨
⎩= 0, If dmax < 1 cm and z max < 10 cm (does not fail) (4)
The parameters of the Lognormal distribution λPGA and ζPGA may be

279
S.P. Ontiveros-Pérez, et al. Engineering Structures 190 (2019) 276–284

Fig. 6. First three vibrations modes of the concrete building.

Table 2
Selected seismic records.
Earthquake (State) Station in Cúcuta Date Epicentral Distance [km] Focal Depth PGA [g] Magnitude ML Geology

Agustin Codazzi (Cesar) Circunvalación 18/03/1999 269.24 97.5 0.004 5.8 Rock
Los Santos (Santander) Santo Domingo 10/01/2006 135.82 150.7 0.002 5.6 Rock
Los Santos (Santander) Santo Domingo 13/10/2007 134.53 155.4 0.009 4.3 Rock
Los Santos (Santander) UFPS 17/02/2008 137.97 152.6 0.005 5.7 Soil
Los Santos (Santander) Santo Domingo 28/03/2008 134.13 148.8 0.002 5.6 Rock
Quetame (Cundinamarca) Santo Domingo 24/05/2008 411.86 0.3 0.005 5.7 Rock
San Alberto (Cesar) UFPS 20/04/2009 123.77 122.9 0.004 5.1 Soil
San Alberto (Cesar) Santo Domingo 20/04/2009 95.05 117.6 0.004 5.5 Rock
Rio Negro (Santander) UFPS 21/07/2009 109.26 144 0.007 4.7 Soil
Los Santos (Santander) UFPS 10/03/2015 138.72 160 0.066 5.2 Soil

calculated using Eqs. (7) and (8). PD (PGA) function (which describes the structural fragility) and the
2
fAmax50 (PGA) function (which describes the seismic threat in a region in
λPGA = E(lnPGA) = ln (E(PGA)) − 0, 5ζ PGA (7) 50 years). The Eq. (9) presents the convolution which is numerically
integrated in the PGA domain to determine the probability of structural
2
⎡ σ (PGA) ⎞ ⎤ failure.
2
ζ PGA = Var (lnPGA) = ln ⎢1 + ⎜⎛ ⎟
⎥
⎣ ⎝ PGA) ⎠ ⎦
E( (8) PGA

Once steps (i) and (ii) of the reliability analysis are performed, the
Pf = ∫0 PD (PGA) fAmax50 (PGA) dPGA (9)
probability of failure Pf is determined by the convolution of the

280
S.P. Ontiveros-Pérez, et al. Engineering Structures 190 (2019) 276–284

Fig. 9. Structural fragility curves without dampers (red curve) and after in-
stallation of the friction dampers optimum design (blue curve) of Run 1. (For
Fig. 7. 36 Possible locations for installing friction dampers.
interpretation of the references to colour in this figure legend, the reader is
referred to the web version of this article.)
Table 3
Friction dampers system based on reliability.
Table 4
Run → Ffn [kN] Pf Comparison of the probability of failure Pf and the seismic acceleration in the
Position P
average fragility.
Without dampers Structure without control 0.4130
1 [P3 P9 P15 P22 P26 P33] [322.558; 361.730; 366.034; Without With Difference [%] Rate
232.502; 301.972; 255.209] 0.0354 control control
2 [P3 P9 P14 P22 P26 P33] [375.589; 338.007; 336.850;
Pf 0.4130 0.0354 91.43 11.67
275.198; 326.059; 263.211] 0.0380
Acceleration [g] in 0.26 0.57 −117.51 0.46
average fragility

Fig. 10. Maximum floor displacements for the East-West component of the 18/
03/1999 earthquake, before and after installation of the friction dampers.
Fig. 8. Final configuration of the frame from Fig. 7 equipped with optimum
design of friction dampers (Run 1). of:

6.3. Problem formulation of reliability-based optimum design of friction Find →

x
dampers Minimize J(→
x ) = Pf
min max
Subjected to F fn ⩽ Ffn ⩽ F fn
This section describes the formulation of the reliability-based op-
timum design of friction dampers employed herein. The objective of the Number of available positions:npfd
proposed methodology is to search for the optimum design of the Maximum number of dampers:nfd (10)
system of friction dampers, which minimizes Pf (Eq. (9)). Following the
procedure described in Section 6.2, the optimization problem consists

281
S.P. Ontiveros-Pérez, et al. Engineering Structures 190 (2019) 276–284

2.25 Hz.

7.1. Seismic excitations

In order to represent the uncertainty of the dynamic load on the

structure under study, records of earthquakes obtained by stations of
the Colombian Geological Service in Cúcuta were considered. In order
to account for the uncertainties introduced by the local soil type and by
the seismic source, ten events were selected with epicenters located in
different regions of Colombia. Table 2 presents the main characteristics
of these seismic events, such as the local magnitude (ML) and epicentral
distances.

7.2. Reduction of the probability of failure

Fig. 11. Displacement at node 56 in the component North-South before and The probability of failure Pf is determined by convolution (see Eq.
after installing optimum design of friction dampers.
(9)) of the PD (PGA) and fAmax50 (PGA) functions, for which purpose → z (t )
was determined by solving Eq. (1) using a time step equal to 0.2 ms.
Fig. 7 shows that there are 36 possible locations (npfd = 36) for the in-
stallation of the friction dampers, represented by the blue elements
located in positions P1, P2, …, P36, that is, there are 6 available spans
in the longitudinal direction in which up to a total of 6 dampers
(nfd = 6) must be positioned. The dampers are installed between
neighboring floors using supports with sufficient stiffness. In this way,
the constraints are the number of possible locations (npfd = 36 ), the
maximum number of devices (nfd = 6) to be installed and the allowable
limits for the frictional forces of the dampers: 200k ⩽ Ffn ⩽ 400kN . To
perform the optimization using the SGA, a population of npop = 30 in-
dividuals was considered, the number of iterations limited to it max = 20 ,
the percentage of it max dedicated to the global phase set equal to 30%
and the percentage of npop that makes up the search group equal to 30%.
Two independent simulations were performed, which led to the results
presented in the Table 3.
Table 3 presents the probability of failure Pf in fifty years of the
Fig. 12. Displacement at node 56 in the component East-West before and after structure equipped with the optimal design of friction dampers. The
installing optimum design of friction dampers.
positions and dampers forces of the optimum designs are considered
correct estimates, while the failure probabilities are regarded upper
7. Illustrative example and analysis of results bound estimates. It may be seen that in both simulations, the design of
the system is similar, both in terms of the friction forces of the dampers
The structure of a typical six floors building in the city of Cúcuta, and the dampers positions within the structure. Only one position dif-
Colombia is analyzed in this section. According to data from the con- fers in the two sets of simulations (from position P15 to P14). The
struction company, the building site is 9 m wide and 20 m long, the difference between the two optimum solutions in terms of the position
ground floor counts with garages while upper floors include two of a damper resulted in a decrease of 6.4% in the failure probability Pf ,
apartments per floor. The structure is a conventional reinforced con- which is not significant, but preference is given to the configuration of
crete frame with three spans in the longitudinal direction, 17.1 m high, run 1. The final configuration of the structure equipped with the op-
7 m wide and 16.55 m long. The building was modeled as a finite ele- timum design of the six friction dampers at the optimal locations is
ment 3-D frame structure consisting of 108 elements and 56 nodes for a shown in Fig. 8.
total of 336 degrees of freedom (see Fig. 5). The finite elements are 3-D As may be observed in Fig. 9, the installation of the dampers op-
beam elements with six degrees of freedom per node. The geometrical timum design increases the structural strength to the seismic loading,
properties of the structural elements are indicated in Table 1. demonstrated by the displacement of the fragility curve. The seismic
A damping ratio (ζ ) for the first and second vibration modes equal acceleration associated with the average fragility increases from 0.26 g
to 1% of critical damping (α = 0.1192/ s e β = 8.3908 × 10−4s ), con- to 0.57 g as a consequence of the installation of the dampers.
sidered representative of damping in concrete frame buildings sub- In Table 4 it is demonstrated the effectiveness of the optimum de-
jected to nondestructive vibrations, was adopted in the simulations. sign of the friction dampers, since after installation of the optimal de-
Thus, the first three natural frequencies referring to the first three vices, the probability that the building reaches the limit state in fifty
modes of vibration (shown in Fig. 6a, b and c) are: 1.88, 1.92 and years decreased by 91.43%.

Table 5
Comparison of the proposed reliability-based optimum design of dampers system with three alternative dampers distribution.
Method → Ffn [kN] Pf
Position P

Proposed [P3 P9 P15 P22 P26 P33] [322.558; 361.730; 366.034; 232.502; 301.972; 255.209] 0.0354
Alternative 1 [P4 P11 P17 P23 P27 P35] [322.558; 361.730; 366.034; 232.502; 301.972; 255.209] 0.1328
Alternative 2 [P1 P2 P3 P4 …P36] 51.111 for each damper 0.0382
Alternative 3 [P12 P16 P18 P30 P34 P36] [322.558; 361.730; 366.034; 232.502; 301.972; 255.209] 0.3271

282
S.P. Ontiveros-Pérez, et al. Engineering Structures 190 (2019) 276–284

For purposes of illustration, Fig. 10 shows the expected value of Acknowledgments

maximum displacements per floor in the East-West direction for the
earthquake of 18/03/1999, before and after the installation of the The authors acknowledge the financial support of CNPq and CAPES,
friction dampers (Run 1). Brazil.
The North-South and East-West components of the horizontal dis-
placement of node 56 for the same 18/03/1999 earthquake are shown References
in Figs. 11 and 12, respectively. Reductions of 31% and 50%, respec-
tively, of the peak displacements in both directions are achieved by [1] Curadelli R, Riera JD. Design and testing of a lead damper for seismic applications.
installation of the dampers. It should be underlined that these reduc- Proc Inst Mech Eng, Part C: J Mech Eng Sci 2007;221(2):159–64. https://doi.org/
10.1243/0954406JMES254.
tions were observed at one location, for one excitation, within an op- [2] Miguel LFF, Riera JD. Controle de vibrações de estruturas utilizando amortecedores
timization process based on reliability that aims at the minimization of por atrito. Revista Int Desastres Naturales, Accidentes e Infraestructura Civil
the probability of failure (and not directly of arbitrarily chosen dis- 2008;8(1):1–14 https://www.scipedia.com/public/Miguel_Riera_2008a .
[3] Min K-W, Seong J-Y, Kim J. Simple design procedure of a friction damper for re-
placements) in a period of fifty years. In the optimization process, the ducing seismic responses of a single-story structure. Eng Struct
uncertainty in the dynamic load was accounted by examining the re- 2010;32(11):3539–47. https://doi.org/10.1016/j.engstruct.2010.07.022.
sponse through a set of earthquakes and not through the response to a [4] Rakicevic ZT, Bogdanovic A, Jurukovski D, Nawrotzki P. Effectiveness of tune mass
damper in the reduction of the seismic response of the structure. Bull Earthq Eng
single deterministic seismic record.
2012;10(3):1049–73. https://doi.org/10.1007/s10518-012-9341-3.
Finally, in order to verify the robustness of the proposed method, [5] Monir HS, Zeynali K. A modified friction damper for diagonal bracing of structures.
the optimal solution obtained in Run 1 of Table 3 was compared with J Constr Steel Res 2013;87:17–30. https://doi.org/10.1016/j.jcsr.2013.04.004.
[6] Belash T. Dry friction dampers in quake-proof structures of buildings. Proc Eng
the performance of three design alternatives with fixed locations or
2015;117:397–403. https://doi.org/10.1016/j.proeng.2015.08.184.
properties of the dampers. The first alternative was to locate the friction [7] Hadi MNS, Arfiadi Y. Optimum design of absorber for mdof structures. J Struct Eng
dampers at randomly selected locations (but with the same frictional 1998;124(11):1272–80. https://doi.org/10.1061/(ASCE)0733-9445(1998)
forces), the second alternative to install friction dampers with equal 124:11(1272).
[8] Miguel LFF, Lopez RH, Miguel LFF, Torii AJ. A novel approach to the optimum
frictional forces at all available positions (uniformly distributed) and design of mtmds under seismic excitations. Struct Control Health Monit
the third alternative to locate the friction dampers at positions closed to 2016;23:1290–313. https://doi.org/10.1002/stc.1845.
the top of the structure. Similarly, Ribakov and Agranovich [34] de- [9] Vellar LS, Ontiveros-Pérez SP, Miguel LFF, Miguel LFF. Robust optimum design of
multiple tuned mass dampers for vibration control in buildings subjected to seismic
veloped a method for determination of damping and stiffness para- excitation. Shock Vib 2019;2019(9273714):1–9. https://doi.org/10.1155/2019/
meters of viscoelastic dampers, presenting three comparative cases: (1) 9273714.
The uncontrolled structure; (2) A structure with active controlled [10] Lavan O, Amir O. Simultaneous topology and sizing optimization of viscous dam-
pers in seismic retrofitting of 3d irregular frame structures. Earthq Eng Struct
dampers and (3) A structure with passive viscoelastic dampers selected Dynam 2014;43(9):1325–42. https://doi.org/10.1002/eqe.2399.
according to the proposed method. [11] Castaldo P, Iuliis MD. Optimal integrated seismic design of structural and viscoe-
The results of this assessment are shown in Table 5. It was verified lastic bracing-damper systems. Earthq Eng Struct Dynam 2014;43(12):1809–27.
https://doi.org/10.1002/eqe.2425.
that only the second alternative, which consists of 36 dampers dis-
[12] Miguel LFF, Miguel LFF, Lopez RH. Robust design optimization of friction dampers
tributed in the structure, achieves the objective of achieving an im- for structural response control. Struct Control Health Monit 2014;21:1240–51.
portant reduction of the probability of failure, but is still about 8% https://doi.org/10.1002/stc.1642.
[13] Miguel LFF, Miguel LFF, Lopez RH. A firefly algorithm for the design of force and
larger Pf for the optimum solution (which consists of only 6 friction
placement of friction dampers for control of man-induced vibrations in footbridges.
dampers). The lower bounds for the failure probabilities in the other Optim Eng 2015;16:633–61. https://doi.org/10.1007/s11081-014-9269-3.
two cases (alternatives 1 and 3) are somewhat smaller, but close to the [14] Miguel LFF, Miguel LFF, Lopez RH. Simultaneous optimization of force and place-
failure probabilities of the building without dampers. ment of friction dampers under seismic loading. Eng Optim 2016;48(4):582–602.
https://doi.org/10.1080/0305215X.2015.1025774.
[15] Miguel LFF, Miguel LFF, Lopez RH. Failure probability minimization of buildings
8. Conclusions through passive friction dampers. Struct Des Tall Special Build
2016;25(17):869–85. https://doi.org/10.1002/tal.1287.
[16] Miguel LFF, Miguel LFF, Lopez RH. Methodology for the simultaneous optimization
In this paper a methodology was proposed for simultaneous opti- of location and parameters of friction dampers in the frequency domain. Eng Optim
mization of friction forces and positions of friction dampers, aiming at 2018;50(12):2108–22. https://doi.org/10.1080/0305215X.2018.1428318.
minimizing the probability of failure of buildings subjected to seismic [17] Mostaghel N, Davis T. Representations of coulomb friction for dynamic analysis.
Earthq Eng Struct Dynam 1997;26(5):541–8. https://doi.org/10.1002/(SICI)1096-
risk. The approach permits the consideration of uncertainties that may 9845(199705)26:5<541::AID-EQE660>3.0.CO;2-W.
affect the design of the dampers system, mainly uncertainties in the [18] Ontiveros-Pérez SP, Miguel LFF, Miguel LFF. Optimization of location and forces of
definition of the excitation. friction dampers. REM-Int Eng J 2017;70(3):273–9. https://doi.org/10.1590/0370-
44672015700065.
A computational routine was developed to allow numerical in- [19] Ontiveros-Pérez SP, Miguel LFF, Miguel LFF. Robust simultaneous optimization of
tegration of the dynamic response of linear frame structures provided friction damper for the passive vibration control in a colombian building. Procedia
with nonlinear friction dampers. The optimization problem herein Eng 2017;199:1743–8. https://doi.org/10.1016/j.proeng.2017.09.430.
[20] Ontiveros-Pérez SP, Miguel LFF, Miguel LFF. A new assessment in the simultaneous
considered is complex, since the objective function is not convex and
optimization of friction dampers in plane and spatial civil structures. Mathe Probl
includes both discrete and continuous variables, so it must be solved by Eng 2017;2017(6040986):1–18. https://doi.org/10.1155/2017/6040986.
optimization methods capable of dealing with such problems. Thus, a [21] MATLAB®, version 8.0.0.783 (r2012b), The MathWorks Inc., Natick, Massachusetts,
modern optimization tool known as Search Group Algorithm (SGA) was 2012.
[22] Rao SS, Yap FF. Mechanical vibrations vol. 4. Upper Saddle River: Prentice Hall;
coupled to the computational routine developed previously to enable 2011.
the determination of the optimum forces and positions of friction [23] Soong TT, Dargush GF. Passive energy dissipation systems in structural engineering.
dampers. New York: Wiley; 1997.
[24] Pall AS. Limited slip bolted joints: a device to control the seismic response of large
The proposed reliability-based optimization method was used to panel structures. Ph.D. thesis, Concordia University; 1979. URL https://spectrum.
design the friction dampers system of a concrete building presently library.concordia.ca/3604/.
under construction in the city of Cúcuta, Colombia, in order to mini- [25] Pall A, Vezina S, Proulx P, Pall R. Friction-dampers for seismic control of canadian
space agency headquarters. Earthq Spectra 1993;9(3):547–57. https://doi.org/10.
mize the probability of failure of the building in a 50 years period by 1193/1.1585729.
providing it with friction dampers. The example confirms the effec- [26] Pall A. Performance-based design using pall friction dampers-an economical design
tiveness of friction dampers to increase the structural reliability of solution. In: 13th World conference on earthquake engineering, Vancouver, BC,
Canada, vol. 71; 2004. URL http://palldynamics.com/pdf/Paper001955.pdf.
buildings under seismic excitation and serves as well to illustrate the [27] Pall AS, Pall R. Friction-dampers for seismic control of buildings–a canadian ex-
implementation of the method and to discuss topics that play a role in perience. In: Eleventh world conference on earthquake engineering, Acapulco,
the formulation of the problem and deserve further research. Mexico; 1996. URL http://www.iitk.ac.in/nicee/wcee/article/11_497.PDF.

283
S.P. Ontiveros-Pérez, et al. Engineering Structures 190 (2019) 276–284

[28] Miguel LFF. Estudo teórico e experimental de amortecedores de vibração por atrito. uncertainties–an overview. Comput Methods Appl Mech Eng 2008;198(1):2–13.
Master’s dissertation, Programa de Pós-Graduação em Engenharia Civil, https://doi.org/10.1016/j.cma.2008.05.004.
Universidade Federal do Rio Grande do Sul, PPGEC/UFRGS; 2002. URL http://hdl. [32] Castaldo P, Amendola G, Palazzo B. Seismic fragility and reliability of structures
handle.net/10183/2443. isolated by friction pendulum devices: seismic reliability-based design (srbd).
[29] Ontiveros-Pérez SP. Projeto ótimo sob incertezas de amortecedores por atrito para Earthq Eng Struct Dynam 2017;46(3):425–46. https://doi.org/10.1002/eqe.2798.
controle de vibrações em edifícios submetidos à excitação sísmica. Doctoral thesis, [33] Castaldo P, Palazzo B, Alfano G, Palumbo MF. Seismic reliability-based ductility
Programa de Pós-Graduação em Engenharia Mecânica, Universidade Federal do Rio demand for hardening and softening structures isolated by friction pendulum
Grande do Sul, PROMEC/UFRGS; 2018. URL http://hdl.handle.net/10183/182336. bearings. Struct Control Health Monit 2018;25(11):e2256. https://doi.org/10.
[30] Gonçalves MS, Lopez RH, Miguel LFF. Search group algorithm: a new metaheuristic 1002/stc.2256.
method for the optimization of truss structures. Comput Struct 2015;153:165–84. [34] Ribakov Y, Agranovich G. A method for design of seismic resistant structures with
https://doi.org/10.1016/j.compstruc.2015.03.003. viscoelastic dampers. Struct Des Tall Special Build 2011;20(5):566–78. https://doi.
[31] Schuëller GI, Jensen HA. Computational methods in optimization considering org/10.1002/tal.578.

284