You are on page 1of 15

Received: 27 November 2017 Revised: 30 August 2018 Accepted: 22 October 2018

DOI: 10.1002/tal.1569

RESEARCH ARTICLE

Seismic design and assessment of structures with viscous


dampers at limit state levels: Focus on probability of damage
in devices
Ahmad Sepehri | Touraj Taghikhany | Seyed Mohammad Reza Ahmadi Namin

Department of Civil and Environmental


Engineering, Amirkabir University of Summary
Technology, Tehran, Iran During large earthquakes, the seismic demand of viscous dampers may exceed their
Correspondence
capacity. In this regard, current design codes must consider extreme conditions and
Touraj Taghikhany, Department of Civil and
Environmental Engineering, Amirkabir preserve the damper at limit state levels. Here, by adjusting the damping coefficient,
University of Technology, Tehran, Iran.
a procedure is introduced to mitigate device damages during severe earthquakes. To
Email: ttaghikhany@aut.ac.ir
assess the procedure, 15 special moment resisting frames with a different number of
stories (two, four, and eight) were designed by three methods: The recommended
novel procedure, the seismic provisions of ASCE7, and the procedure proposed by
Miyamoto et al.[1] for structures, installed with supplemental damping devices. A
series of incremental dynamic analyses were then performed by modeling the limit
state behavior of viscous dampers. Results indicated that the novel method reduces
the damage probability of dampers as well as the maximum demands on the structure
at different seismic hazard levels.

KEY W ORDS

viscous damper, fragility curves, seismic assessment, seismic design, steel structures, viscous
damper damage

1 | I N T RO D U CT I O N

To date, several experimental and numerical studies have shown how linear and nonlinear viscous dampers improve the seismic performance of
structures;[2–5] many of them have investigated the collapse assessment of structures, which incorporated viscous devices. Oesterle[6] evaluated
the effect of the damper velocity exponent on the seismic responses of nine‐story buildings, by conducting incremental dynamic analysis (IDA) and
providing the fragility curves of damped structures. Similarly, Kruep[7] performed IDA and concluded that implementation of viscous devices can
reduce interstory drift at any seismic hazard level. Further, Seo et al.[8] and Miao et al.[9] performed a series of IDA to the collapse assessment of
structures, both with and without viscous fluid dampers, indicating that using a damping device reduces the probability of collapse and increases
the Adjusted Collapse Margin Ratio (ACMR). With regard to the complexity of IDA, Hamidia et al.[10,11] assessed the sideways collapse capacity of
a building incorporating viscous fluid dampers through a simplified procedure.
In all the studies cited above, the dynamic behavior of viscous dampers was represented by the Maxwell model, at moderate or severe earth-
quakes. Within this model, regardless of earthquake intensity, the performance of the viscous damper only depends on the damping coefficient
and inte‐story velocity.
During the Sendai earthquake in 2011, the failure of oil dampers (Figure 1) indicated that during an unexpectedly large earthquake, the behav-
ior of viscous dampers exceeded the limit state level.[12] Indeed, due to the excessive interstory displacement or velocity, mechanical failures
resulted. These failure modes comprise the piston head bottomed out in the cylinder, the piston rod failure in tension, or deficiency of driver brace
in compression loads.[13] In these states, the Maxwell model cannot properly simulate the limit state behavior of dampers.

Struct Design Tall Spec Build. 2019;28:e1569. wileyonlinelibrary.com/journal/tal © 2018 John Wiley & Sons, Ltd. 1 of 15
https://doi.org/10.1002/tal.1569
2 of 15 SEPEHRI ET AL.

FIGURE 1 An oil damper; before and after


Sendai earthquake 2011.[12]

In 2010, the limit state model of viscous fluid dampers was introduced and experimentally validated by Miyamoto et al.[14] In order to validate
the model, a series of nonlinear dynamic analyses were conducted on a ten‐story moment‐resisting steel damped structure using the limit state
model of the viscous damper. It was demonstrated that for severe earthquakes, the Maxwell model produced nonconservative forces in
dampers.[15] Next, the collapse probability of different damped structures was investigated, and the effect on fragility curves of one‐story struc-
tures was determined.[16] Further, Ahmadi[17] examined the effect of limit state behavior on the seismic response of structures equipped with
toggle‐configured viscous dampers. It was concluded that greater magnification factors increased failure probability of the dampers as well as col-
lapse probability of the structure. Likewise, the collapse performance of the fluidic self‐centering system was also studied by Kitayama et al. incor-
porating the ultimate behavior of this device.[18]
The limit state phenomenon is not only restricted to analysis and should be considered in the design procedure, as well. To this end,
Miyamoto et al.[1] used the recommendation of ASCE41–06[19] when the number of dampers was limited to two devices per story. The code spec-
ified that velocity‐dependent devices shall be designed and sized to protect against the displacement and force associated with a hazard level
equal to 130% of the maximum considered earthquake (MCE). In addition, Miyamoto proposed that a damping ratio should be assigned to all sys-
tems, regardless of design category, in order to bound the maximum interstory drift ratio to 1% at the design‐basis earthquake (DBE) level. This
design procedure resulted in dampers with a greater capacity and more expensive than those recommended in ASCE7–10.[20]
Indeed, stronger dampers increased the safety factor and postponed the initiation of limit state behavior. Also, the collapse assessment on
damped structures indicated that this method results in a lower probability of damage compared with the recommendations of ASCE7.[1]
However, this does not guarantee a consistent safety margin of damage at the initiation of limit state behavior.
The new edition of ASCE7[21] recommends the same provisions as ASCE41–17[22] to control limit state behavior of a damper when the num-
ber of dampers is limited to two devices per story in each principal direction. When the number of dampers exceeds two, ASCE7 does not address
possible damage to devices.
Here, a methodology is examined to improve current seismic provisions used in practice for design of nonlinear viscous dampers under severe
ground motions. In fact, this method reduces the possibility of damage to damper devices or postpones its occurrence during large earthquakes
without using stronger dampers.

2 | D E S I G N P R O C E DU RE

Generally, the target damping ratio of the structure and the characteristics of the viscous dampers are determined in compliance with the seismic
design objective. In prescriptive seismic codes, the design objectives for damped structures are to protect human life, conserve nonstructural ele-
ments, and maintain facility functions. To this end, ASCE7–16 recommends that damping devices shall resist the forces and displacements related
to velocities associated with MCE ground level motions (sec. 18.2.4.6). This provision assumes that dampers and connections perform properly
during more intense earthquakes than those at the DBE level. In addition, the effective damping of the system in ith mode of vibration (βvi) should

FIGURE 2 (a) Damping ratio (b) Damping force variation in linear and nonlinear dampers (four‐story building)
SEPEHRI ET AL. 3 of 15

adequately limit interstory drift (Δd) at the DBE level. The damping ratio of structure can be calculated by implementing nonlinear viscous dampers
with damping coefficients Cj, satisfying Equation (1)[2,23]:

 1þα  1þα
T1 2−α λ∑ j Cj cosθj ϕj −ϕj−1
βV1 ¼ : (1)
ð2πÞ3−α A1−α ∑i mi ϕi 2

βv1is the additional effective damping ratio provided by dampers in the fundamental mode of vibration; parameter A is the roof displacement
at DBE for the structure with damping ratio equal to target value; θj is the inclination angle of the damper in the jth story; Cj and α are the damping
coefficient and velocity exponent of damper, respectively; ϕj and ϕj − 1 are the horizontal modal displacements of the jth and j − 1th story in the
fundamental mode of vibration; mi is story seismic mass; T1 is first‐mode period; and λ is a parameter that can be calculated by Equation (2):

Γ2 ð1 þ α=2Þ
λ ¼ 22þα (2)
Γð2 þ αÞ

in which Γ is the gamma function.


According to Equation (1), by increasing the seismic intensity and subsequently the roof displacement (A), the damping ratio (βv1) has a
descending rate. Figure 2a shows the variation of βv1 with roof displacement ratio, which is the displacement of the roof (relative to the base)
normalized by building height, in a four‐story building. The nonlinear dampers to be installed in it have been designed for a target damping ratio
of 25% at the DBE level. At large seismic displacements, Figure 2a demonstrates that the generated damping ratio will reduce below the target
value. In addition, Figure 2b illustrates the variation of viscous damping force with the velocity in linear and nonlinear dampers. It shows that when
the interstory velocity exceeds the design level velocity (VDBE), the force generated by the nonlinear damper (α < 1) decreases. In accordance with
Figure 2, severe ground motions cause a larger displacement and velocity, which continue to reduce the efficiency of the nonlinear viscous
dampers. This phenomenon increases the contribution to the structural system to carry the seismic load.
It is of paramount importance to point out that in ASCE7, the structural system is to be designed up to a minimum of 75% of the strength
required for an undamped structure. Thus, when damping devices fail, the collapse probability will be higher than that for the conventional struc-
tures. In order to improve the performance of damped structures, ASCE7 (2016) modified provisions for buildings when two damping devices are
placed in each principal direction of each story. The code recommends that due to a lack of redundancy in the building, the dampers should be
designed for hazarding motions, 30% higher than the MCE level ground motions.
As described in Equation (1), the damping ratio βvi depends on Cj and the interpretation of quantity A, which is the roof displacement at the
design basis level. The A parameter is roof amplitude and is determined from a time history analysis of the structure possessing no dampers,
whereas its viscous damping ratio equals βv1. This parameter is derived when the structure is subjected to a series of selected ground motions,
which are scaled to the acceleration amplitude at DBE level. In Equation (1), a larger roof displacement requires dampers with a greater damping
coefficient (Cj), to still provide the specific damping ratio βvi.
The damping coefficient (Cj) of nonlinear viscous dampers decreases at large velocities, which, in turn, reduces the performance of viscous
dampers during severe earthquakes. In order to improve the efficiency of dampers and postpone possible damage of device, output damping at large
velocities must be compensated. In order to achieve this, Miyamoto et al.[1]; [14]–[16] (and ASCE7–16) recommend specifying a damper able to with-
stand 30% more than the design forces at MCE level. Stronger dampers provide higher resistance and improve efficiency at the limit state level.
Increasing output damping does not necessarily require a stronger damper. In fact, a manufacturer can adjust the damping coefficient without
altering the design process. Output damping of nonlinear viscous dampers can be compensated with higher damping coefficients and not neces-
sarily with stronger dampers. In practice, the code‐based damping coefficient (Cj) is increased for a specific target damping ratio (βv1). Cj is derived
from Equation (1) when the amplitude, A, is determined from the ground motions scaled to the DBE level. If ground motions are instead scaled to

C MCE
Damper Force

C DBE

Velocity
VDBE VMCE

FIGURE 3 Comparing the performance of a specific nonlinear damper with different damping coefficients
4 of 15 SEPEHRI ET AL.

TABLE 1 Details of different approaches

Interstory drift limitation to Damping coefficient Damper capacities


Design method derive target damping ratio design level design level

ASCE7–16 Table 12.12–1 (ASCE7) DBE MCE1


Miyamoto 1.0% DBE 1.3 MCE
Proposed Table 12.12–1(ASCE7) MCE MCE
1
when the number of dampers are limited to two devices per story, it is 1.3 MCE. Notes. DBE: design‐basis earthquake; MCE: maximum considered
earthquake.

the MCE level, A increases, and a higher value of Cj is derived. In this paper, these specific damping coefficients are denoted as CDBE and CMCE,
respectively.
Seleemah et al.[3] showed that the damping ratio's dependency on the amplitude of A is weak, when its magnitude is large. Accordingly, calcu-
lation of A for a seismic event greater than MCE does not considerably affect the damping coefficient. Figure 3 shows that by increasing the damping
coefficient to be greater than CDBE, the degradation of the damper's force (and damping) at higher velocities can be compensated. In Table 1, the
design approach of this study at different stages is compared with ASCE 7–16 design method and the recommendations of Miyamoto et al.[1]
To verify the efficiency of the proposed method and to improve the seismic performance of damped structures at collapse prevention level
(as well as at other performance levels), three different structures were designed and investigated.

3 | ANALYTICAL MODELS

In the present study, three different steel structures (two, four, and eight stories) were designed with a square plan area and five bays of 6 m in
both directions with story height of 4 m. As shown in Figure 4, the peripheral antiseismic system of damped structures is composed of the follow-
ing two sections:

• A three‐bay moment‐resisting frame in the middle


• Two single bays implementing viscous damper systems that were installed on both sides

The nonperipheral frames were assumed to carry only the gravity load with a hinged beam‐column connection. The gravity loads were spec-
ified to represent those of a common residential building. The chosen seismic region is Los Angeles, USA, with Ss = 0.51 g and S1 = 1. The Ss and S1
are mapped MCER, 5%‐damped spectral response acceleration parameters at short periods and at a period of 1 second, respectively. The site soil
property is assumed stiff, and it shall be classified as site class D. The design procedure is based on the equivalent lateral force (ELF) proce-
dure[24,25] in which the base shear of the primary structure cannot be less than the recommended value of Equation (3) below:

 
V
V min ¼ max ; 0:75 V ; (3)
BVþI

Where V is seismic base shear, and BV + I is the numerical coefficient as set forth in Table 1 (8.7–1), ASCE7–16, for effective damping equal to
the sum of viscous damping in the fundamental mode of vibration plus inherent damping, βI. Nonlinear viscous dampers with a velocity exponent
of 0.5 are implemented to limit the maximum interstory drift to the allowable level at DBE. In ASCE7–16, allowable interstory drift ratios for Risk
Categories IV and III are 1 and 1.5%, respectively.

FIGURE 4 Elevation and plan of four‐story building


SEPEHRI ET AL. 5 of 15

First, the undamped structures are designed for base shear recommended by conventional seismic code without considering drift limitation.
Then, the additional viscous damping ratios (βv1) for the selected structures are calculated by ELF procedure to limit maximum interstory drift ratio
to the allowable values at DBE level. The structural elements and damping system are controlled and possibly redesigned until converge to target
value. The required damping ratios, which calculated for different structures have been listed in Table 2.
The St37‐grade steel was assumed as the material of structural elements, which were sized in accordance with AISC360.[26] The beam and
column sections are shown in Table 3.
The target damping ratios of Table 2 are then used to determine the characteristic of dampers (CDBE and CMCE) from Equation (1). In this equa-
tion, the roof amplitude (A) of each structure is derived by using a nonlinear time history analysis and scaled ground motions (Table 4) to that of
the DBE and MCE acceleration spectrum. The list of roof amplitude (A) values have been presented in Table 5.
By using Equation (1), CDBE and CMCE of the dampers in three different structures (two, four, and eight stories) are computed for two allow-
able drift ratios equal to 1 and 1.5% (Risk Category IV and III, respectively).
For the limit state modeling of dampers, the force capacity and maximum stroke of devices must be determined. These parameters are cal-
culated based on code provisions for damped structures. According to ASCE7–16, damping devices shall resist the forces, displacements, and
velocities of a system when subjected to ground motion at the MCE level. To estimate the required capacity of the dampers, a series of time

TABLE 2 Target required viscous damping ratio for selected structures


Story Drift limit (%) Required damping ratio

Two 1.5 0.12


Two 1 0.25
Four 1.5 0.12
Four 1 0.25
Eight 1.5 0.15
Eight 1 0.26

TABLE 3 The cross‐section properties of structural elements (special moment‐resisting frame+ gravitation load‐carrying frame) in two‐, four‐,
and eight‐story models

Model Story SMRF beams SMRFcolumns Interiorcolumns Cornercolumns Driverbrace


Two story 1 to 2 IPE500 BOX450 × 25 BOX250 × 15 BOX350 × 25 BOX200 × 10
Four story 1 to 2 IPE550 BOX500 × 30 BOX250 × 15 BOX450 × 25 BOX200 × 10
2 to 4 IPE450 BOX400 × 20 BOX200 × 10 BOX350 × 25 BOX200 × 10
Eight story 1 to 3 IPE750 BOX600 × 30 BOX300 × 15 BOX600 × 25 BOX200 × 12
4 to 6 IPE500 BOX500 × 20 BOX250 × 15 BOX450 × 25 BOX200 × 12
7 to 8 IPE450 BOX400 × 20 BOX200 × 10 BOX350 × 25 BOX200 × 12

TABLE 4 Ground motion specifications used for calculation of a


Earthquake Recording station
Soil
M Year Name Name Owner type

6.7 1994 Northridge Beverly Hills‐Mulhol USC D


6.5 1979 Imperial Valley El Centro Array #11 USGS D
6.9 1989 Loma Prieta Capitola CDMG D

TABLE 5 Roof amplitude of the selected structures under scaled ground motions to design‐basis earthquake and maximum considered
earthquake
Roof displacement Roof displacement
Model (%) DBE (m) MCE (m)
2 story (1) 0.078 0.125
2 story (1.5) 0.112 0.183
4 story (1) 0.143 0.251
4 story (1.5) 0.196 0.310
8 story (1) 0.297 0.406
8 story (1.5) 0.398 0.563
6 of 15 SEPEHRI ET AL.

history analyses are performed, using nine ground motions scaled to the MCE level, in accordance with Chapter 16 of the code. The specifications
of ground motion records have been listed in Table 6. The maximum stroke and force of dampers at nine ground motions are presented in Table 7
for the different design categories. Damping coefficients at the MCE level are approximately 25% higher than those at the DBE level.
Alternatively, the Miyamoto et al.'s[14] procedure determined the required capacity for dampers to be 30% more than the results above. The
proposed method of this paper, however, does not recommend increasing the damper capacity for the interstory drift limitations (1% and 1.5%),
and it is similar to the specifications in ASCE7–16. The only difference between the dampers in the proposed procedure and those in the code
provision is in the damping coefficient. The damping coefficient in this novel procedure (CMCE) is 25% more than that specified in the code,
and the maximum stroke is reduced by 10% as compared with ASCE7–16.
After determining the damping coefficient and maximum demands for each design procedure, viscous dampers with same characteristics were
selected from a manufacturer's catalog (Taylor Devices Inc.) and are presented in Table 8.[27] When the interstory drift ratio is limited to 1.5%, the
recommended viscous dampers of ASCE7–16 and the proposed method in two‐ and eight‐story structure were very similar. The damping coef-
ficients are not identical though, producing different limit state behavior.
For limit state numerical modeling of viscous dampers, the material properties must be acquired. This information for each damper in Table 8
was determined using the Taylor Devices Inc. catalog.

4 | N U M E R I C A L M O D E L I N G I N OP E N S E E S

In this study, the OpenSees software[28] was used to execute nonlinear analysis on a damped structure. To capture the three‐dimensional effects
of the structure in Figure 4, two single‐bay frames were appended to either side of the peripheral damped special moment‐resisting frames
(SMRFs; Figure 5). These columns (leaning columns) were modeled as elastic beam‐column elements and were connected to the damped SMRF
by simple beams modeled by Truss elements. The stiffness of the leaning columns in the appended bays was assumed to be four times that of

TABLE 6 Ground motions specifications

Earthquake
Recording Station
M Year Name Name
Near‐field records with pulse
6.5 1979 Imperial Valley‐06 El Centro Array #6
6.5 1979 Northridge‐01 Rinaldi receiving Sta
6.9 1989 Duzce, Turkey Duzce
Near‐field records without pulse
6.7 1994 Imperial Valley‐06 Bonds corner
6.5 1979 Northridge‐01 Northridge ‐ Saticoy
6.9 1989 Kocaeli, Turkey Yarimca
Far‐field records
6.7 1994 Northridge Beverly Hills‐Mulhol
6.5 1979 Imperial Valley El Centro Array #11
6.9 1989 Loma Prieta Capitola

TABLE 7 Damping coefficient and required capacities of dampers


Models Damping coefficient Factored stroke Factored force
description kN. (sec/m)0.5 demand (cm) demand (kN)

Damper characteristic ASCE7–16 2St‐1%‐ CDBE ‐SF1 979.58 6.7 692.88


2St‐1.5%‐ CDBE ‐SF1 531.26 8.0 401.71
4St‐1%‐ CDBE ‐SF1 1213.90 7.3 744.60
4St‐1.5%‐ CDBE ‐SF1 658.36 8.9 417.18
8St‐1%‐ CDBE ‐SF1 1629.30 7.6 765.71
8St‐1.5%‐ CDBE ‐SF1 1038.30 9.8 569.07
Proposed 2St‐1%‐ CMCE ‐SF1 1216.80 6.0 828.26
2St‐1.5%‐ CMCE ‐SF1 662.54 7.5 491.27
4St‐1%‐ CMCE ‐SF1 1507.90 6.7 913.25
4St‐1.5%‐ CMCE ‐SF1 821.04 8.1 512.80
8St‐1%‐ CMCE ‐SF1 2029.60 .6.9 1071.26
8St‐1.5%‐ CMCE ‐SF1 1298.30 8.9 699.13
Miyamoto Recomme‐ndation 2St‐1%‐ CDBE ‐SF1.3 979.58 8.7 900.74
4St‐1%‐ CDBE ‐SF1.3 1213.90 9.5 967.98
8St‐1%‐ CDBE ‐SF1.3 1629.30 9.9 995.42

Note. SF indicates scale factor.


SEPEHRI ET AL. 7 of 15

TABLE 8 Specifications of selected dampers

Model Damper design procedure Damper design safety factor Selected damper kN (kips) Stroke cm (inches)
2St‐1%‐ CDBE ‐SF1 ASCE7–16 1 730(165) 10.16(4)
2St‐1%‐ CMCE ‐SF1 Proposed 1 980(220) 10.16(4)
2St‐1%‐ CDBE ‐SF1.3 Miyamoto 1.3 980(220) 10.16(4)
2St‐1.5%‐ CDBE ‐SF1 ASCE7–16 1 490(110) 10.16(4)
2St‐1.5%‐ CMCE ‐SF1 Proposed 1 490(110) 10.16(4)
4St‐1%‐ CDBE ‐SF1 ASCE7–16 1 980(220) 10.16(4)
4St‐1%‐ CMCE ‐SF1 Proposed 1 980(220) 10.16(4)
4St‐1%‐ CDBE ‐SF1.3 Miyamoto 1.3 980(220) 10.16(4)
4St‐1.5%‐ CDBE ‐SF1 ASCE7–16 1 490(110) 10.16(4)
4St‐1.5%‐ CMCE ‐SF1 Proposed 1 730(165) 10.16(4)
8St‐1%‐ CDBE ‐SF1 ASCE7–16 1 980(220) 10.16(4)
8St‐1%‐ CMCE ‐SF1 Proposed 1 1470(330) 10.16(4)
8St‐1%‐ CDBE ‐SF1.3 Miyamoto 1.3 1470(330) 10.16(4)
8St‐1.5%‐ CDBE ‐SF1 ASCE7–16 1 730(165) 10.16(4)
8St‐1.5%‐ CMCE ‐SF1 Proposed 1 730(165) 10.16(4)

the interior gravity columns.[1] The flexural capacities were computed as the sum of the moment capacity of the interior columns. This technique
was used in order to simulate both P‐delta effect and the aggregate effect of all the interior gravity columns, which are not considered in a two‐
dimensional model.
All girder beams and columns of SMRF are modeled as elastic beam‐column elements with concentrated plastic hinges at either ends. The
point hinges are represented by rotational springs whose properties are obtained from engineering mechanics principles supplemented by back-
bone curve and deterioration rules of modified Ibarra–Krawinkler model[29–31] (Figure 6). The moment rotation behavior for all the elements of the
frame is modeled with a bilinear hysteretic material (Bilin Material), which is capable of capturing the cyclic deterioration modes; basic strength
deterioration, postcapping strength deterioration, unloading stiffness deterioration, and accelerated reloading stiffness deterioration.[30] A transi-
tion curve has been provided for this material in the intersection of the branches (i.e., tangent moduli) to avoid any sudden change in local stiffness
matrices and to ensure a smooth transition between the elastic, precapping and postcapping plastic regions. The concentrated plastic hinges are
represented by zero‐length elements. The deterioration parameters for beams and columns are obtained based on proposed relationships for
beams and tubular hollow square columns.[32]
In addition to frame elements, the limit states behavior of the viscous damper is simulated by the model shown in Figure 7.[13] The different
parts of this model are defined in the OpenSees using specific elements with the following characteristics:

• The dashpot was modeled using a truss element with a viscous material, a velocity exponent (α) of 0.5, and a damping coefficient (C)
dependent on the selected damper.

• The cylinder wall was modeled using a nonlinear beam‐column element with a gap/hook material definition. This element simulated the stroke
limit and stiffening effects of the damper when it bottoms out.
• The piston rod and piston undercut were modeled using a nonlinear beam‐column element with a steel material definition and 1% hardening
stiffness. An effective Young's modulus was used to account for the different section properties of the piston. The undercut yield and fracture
capacities are a function of that specified undercut diameter.

FIGURE 5 2D exterior frame (four‐story building)


8 of 15 SEPEHRI ET AL.

FIGURE 6 Backbone curve of the modified Ibarra–Krawinkler model[31]

Viscous Damper
Element

Piston+Undercut Driver Brace


Element Element
Gap/Hook
Element

FIGURE 7 The limit state model of the damper in OpenSees[13]

This model simulates the behavior of a fluid viscous damper at normal and at limit state levels. In the limit state level, the following scenarios exist:

1‐ Force limited state. Undercut yielding and fracture or driver brace buckling will happen when the piston is subjected to a force pulse.
2‐ Displacement limited state. It will occur when the piston impacts the end of the cylinder. This is also referred to as the damper bottoming out.

Now, to verify the analytical models in the OpenSees, their responses were compared at two stages, dynamic behavior of structure and limit state
behavior of dampers.
First, to ensure the adequacy of damper models in structures, their nonlinear time history responses were investigated under different ground
motions, at two seismic levels, DBE and MCE levels. The former seismic level for models with damping coefficients of CDBE, and the latter for

FIGURE 8 Comparison of the time history roof displacement of two‐, four‐, and eight‐story damped structures with same structures without
viscous dampers and with the equivalent damping ratio
SEPEHRI ET AL. 9 of 15

those with CMCE.. The behavior of models equipped with dampers was compared with the same frame with an inherent viscous damping ratio.
These comparisons were performed on structures with different stories and dampers. Figure 8 shows the roof displacement response of four
models under Loma Prieta record presented in Table 4, with different stories and drift ratios, which indicates the well accuracy of the numerical
models at the peak response. It is worth noting that the difference in response is attributed to the nature of nonlinear dampers compared with
linear viscous damping.
In the next stage, the limit states model of damper were validated. Miyamoto et al.[13] conducted a sinusoidal loading test by increasing the
amplitude to verify the numerical model in two limit states (Force and Displacement cases). To approve the numerical model, the same test was
repeated, and results have been shown in Figure 9. The time history variation of damper force and its hysteresis were compared between the two
limit state scenarios. The results confirm the accuracy of the OpenSees model for viscous dampers.

FIGURE 9 Time history variation of damper force and hysteresis behavior of viscous damper at two limit state scenarios; undercut fractures and
damper bottoms out (a) Miyamoto et al. (b) Present study
10 of 15 SEPEHRI ET AL.

5 | COLLAPSE ASSESSMENT

After modeling the structures in the OpenSees, a series of incremental dynamic analyses were conducted to assess performance at the collapse
prevention level. The far‐field ground motions (44 records) in Appendix A from FEMA‐P695[33] were used for the nonlinear dynamic analysis. A
maximum interstory drift ratio of 10%, reduction of the IDA curve slope to 20% of the elastic slope, and dynamic instability were defined as the
collapse criteria during IDA analysis.[29,34] Figure 10 illustrates the IDA curves of two‐ and four‐story damped structures under Record No. 14
loads. In this figure, maximum structure demands of three different procedures were compared when the design allowable interstory drift was
defined at 1%. Refer to Table 8 for the four‐story building, where all three described design procedures resulted in the same damper stroke.
The proposed novel procedure (1%‐CMCE‐SF1) resulted in a system with a higher collapse capacity than the other two methods.
In Figure 10, the system was designed in accordance with the ASCE7–16 procedure (1%‐CDBE‐SF1) and the related IDA curve is the same as
that based on the Miyamoto et al.'s procedure (1%‐CDBE‐SF1.3). In the two‐story building (Figure 10a), the capacity of the required damper in the
ASCE7–16 procedure is lower, such that bottoming out of the damper occurs in the same time, but the actual failure occurs earlier. Thus, its IDA
curve deviated, and the collapse level was reduced.
For the seismic collapse assessment of the designed structures, the FEMA P695 methodology was employed. The performance of the struc-
tures at the collapse level was quantified by the ACMR. ACMR was computed by multiplication of the collapse margin ratio, CMR, and the Spectral
Shape Factor (SSF). CMR is defined as the ratio of the median spectral acceleration at collapse level, b
SCT ; to the MCE ground motion spectral
demand, SMT, at the maximum permissible code period Tmax. SSF depends on fundamental period (T1), period‐based ductility (μT), and seismic
design category. All of the parameters, and ultimately the ACMRs, were calculated for the structures designed with the three different procedures.
As shown in Table 9, ACMR values for the proposed method were greater than the other two procedures. In other words, the proposed method
resulted in a higher safety margin. As mentioned in Section 3, for targeting a drift ratio of 1.5%, the size of the damper for the proposed method
and the ASCE7–16 provision was the same (no additional cost), whereas the proposed method reduced the collapse probability.

6 | L I M I T ST A T E A S S E S S M E N T

The fundamental objective of the proposed design procedure is to postpone the occurrence of limit state behavior in nonlinear viscous dampers.
The limit state behavior often happens when an excessive seismic demand causes mechanical problems such as (a) the bottoming out of a piston
head in the cylinder, and (b) the failure of the piston rod in tension or a driver deficiency in compression. The bottoming out of a piston head can
be determined when the damper's deformation exceeds its stroke limit, and the failure of the piston or the deficiency of the driver occurs when
the damper force exceeds the strength of its components.
The occurrence of each mechanical problem was investigated by comparing the maximum demand force and the deformation of the damper,
as listed in Table 8. In order to better understanding, Figure 11 illustrates the IDA curves of the two‐story building under ground motion that was
recorded on a site class D (Loma Prieta [Gilroy Array], record No. 14). It compares the first bottoming out occurrence and the first time at which
the force exceeded the damper capacity. The initiation of each mechanical problem changed the tangent of the IDA curve, particularly when the
damper bottoming out takes place earlier than failure due to force exceedance.
The results of the incremental dynamic analyses on all structures for 44 records have been used to derive the fragility curves for two different
limit states: (a) Spectral acceleration at which the first bottoming out of the damper occurs, and (b) spectral acceleration at which the damper force
exceeds its capacity.
According to Figure 12 and Figure 13, the probability of reaching the limit states of the damped structure based on the proposed procedure
(1%‐CMCE‐SF = 1) is less than that for the other procedures. As depicted, the limit state probability of both types of mechanical failures increased

FIGURE 10 Incremental dynamic analysis curves of damped structures designed with three methods, for 1% interstory drift, under record No.
14 (a) two‐story structure (b) four‐story structure
SEPEHRI ET AL. 11 of 15

TABLE 9 Adjusted collapse margin ratio of designed structures with different procedures

Story Design proc. Model SMT (g) b


SCT (g) CMR μ SSF ACMR

Two story ASCE7–16 1% ‐ CDBE ‐ SF = 1 1.4 4.42 3.16 6.89 1.37 4.32
1.5% ‐ CDBE ‐ SF = 1 1.4 3.56 2.54 7.88 1.40 3.56
Proposed 1% ‐ CMCE ‐ SF = 1 1.4 4.73 3.38 5.33 1.33 4.49
1.5% ‐ CMCE ‐ SF = 1 1.4 3.79 2.70 7.88 1.40 3.79
Miyamoto 1% ‐ CDBE ‐ SF = 1.3 1.4 4.61 3.29 5.33 1.33 4.37
Four story ASCE7–16 1% ‐ CDBE ‐ SF = 1 0.8 2.47 3.09 5.62 1.46 4.51
1.5% ‐ CDBE ‐ SF = 1 0.8 2.11 2.64 6.53 1.50 3.96
Proposed 1% ‐ CMCE ‐ SF = 1 0.8 2.61 3.26 5.62 1.46 4.76
1.5% ‐ CMCE ‐ SF = 1 0.8 2.35 2.94 6.93 1.52 4.47
Miyamoto 1% ‐ CDBE ‐ SF = 1.3 0.8 2.47 3.09 5.62 1.46 4.51
Eight story ASCE7–16 1% ‐ CDBE ‐ SF = 1 0.46 1.24 2.70 5.32 1.67 4.50
1.5% ‐ CDBE ‐ SF = 1 0.46 1.10 2.39 5.91 1.70 4.07
Proposed 1% ‐ CMCE ‐ SF = 1 0.46 1.39 3.03 4.62 1.62 4.91
1.5% ‐ CMCE ‐ SF = 1 0.46 1.17 2.55 5.91 1.70 4.33
Miyamoto 1% ‐ CDBE ‐ SF = 1.3 0.46 1.31 2.84 4.62 1.62 4.60

FIGURE 11 Incremental dynamic analysis curves of damped structures designed with three methods, for record No. 14 (two‐story structure)

from two‐story to eight‐story buildings. In other words, the efficiency of the design method increases, as the number of stories increases. It is
worth noting that for 1%‐CDBE‐SF1 and 1%‐CDBE‐SF1.3 models, the stroke fragility curves overlap because of identical damping coefficients.
Figure 14 illustrates the median spectral acceleration at which the first bottoming out occurs, as well as that at which the damper force
exceeds its capacity for the first time. The figure indicates that the seismic level occurring in a damper's limit states in two‐ and four‐story struc-
tures is higher than that in eight‐story structures. Furthermore, it confirms the greater median acceleration of the proposed method (1%‐CMCE‐

FIGURE 12 Fragility curves for two, four, and eight story at the limit state level, when the first bottom out takes place in the damper.
12 of 15 SEPEHRI ET AL.

FIGURE 13 Fragility curves for two, four, and eight story at the limit state level, when the force in the damper exceeds its capacity

SF = 1) when compared with the other methods. The proposed method postpones the limit state behavior by increasing the acceleration level of
the limit states. The relative enhancement of Sa for different procedures has been listed in Table 10. In this table, the procedure of Miyamoto (1%‐
CDBE‐SF = 1.3) has no effect on the seismic level at which the first bottom out occurs relative to the ASCE7–16 procedure. Whereas, the proposed
procedure (CMCE‐SF = 1) has increased the median Sa of the first occurrence of bottom out by an average value of 17%. In addition, this method
has improved force limit state capacity of dampers as level as Miyamoto's procedure.
As seen in Table 10, the median Sa at two limit state exceedance levels (first bottom out and first force) by increasing number of building
stories are decreased considerably in all design procedures and different interstory drift limitation.
In Table 11, the occurrence probabilities of the limit states at the MCE level have been calculated using fragility curves. This table demon-
strates that the probability of limit states behavior in structures designed by the proposed method was lower than the other design methods.
When the Miyamoto et al. design method is set to limit the interstory drift ratio to less than 1%, no reduction in the probability of reaching
the stroke limit is observed. In the proposed method, the reduction is approximately 40% for bottoming‐out events and 30% for the first force
exceedance. Similarly, for a target interstory drift ratio of 1.5%, the probability of limit state behavior was significantly reduced with no increase
in the cost of purchasing higher capacity dampers.

FIGURE 14 A comparison between the


median spectral acceleration (Sa[g]) of the first
bottomed out and first force exceedance in
two‐, four‐, and eight‐story structures
designed with different procedures
SEPEHRI ET AL. 13 of 15

TABLE 10 Relative improve of median spectral acceleration of the first bottomed out and first force exceedance

Improvement of median Sa for Improvement of median Sa for


Model first bottom out (%) first force exceedance (%)

1% interstory drift ratio limit


Two story 1% ‐ CDBE‐ SF = 1ASCE7–16 0.00 0.00
1% ‐ CDBE ‐ SF = 1.3Miyamoto 0.00 7.53
1% ‐ CMCE ‐ SF = 1proposed procedure 14.35 13.4
Four story 1% ‐ CDBE‐ SF = 1ASCE7–16 0.00 0.00
1% ‐ CDBE ‐ SF = 1.3Miyamoto 0.00 0.00
1% ‐ CMCE ‐ SF = 1proposed procedure 18.40 12.30
Eight story 1% ‐ CDBE‐ SF = 1 0.00 0.00
ASCE7–16
1% ‐ CDBE ‐ SF = 1.3Miyamoto 0.00 14.80
1% ‐ CMCE ‐ SF = 1proposed procedure 18.00 32.27
1.5% inter‐story drift ratio limit
Two story 1.5% ‐ CDBE ‐ SF = 1ASCE7–16 0.00 0.00
1.5% ‐ CMCE ‐ SF = 1proposed procedure 15.2 14.35
Four story 1.5% ‐ CDBE ‐ SF = 1 ASCE7–16 0.00 0.00
1.5% ‐ CMCE ‐ SF = 1proposed procedure 11.00 9.19
Eight story 1.5% ‐ CDBE ‐ SF = 1ASCE7–16 0.00 0.00
1.5% ‐ CMCE ‐ SF = 1proposed procedure 29.35 29.77

TABLE 11 Probability of reaching limit state at maximum considered earthquake level

Relative reduction Relative reduction


Probability of first to ASCE7–16 Probability of capacity to ASCE7–16
Model bottom out (%) procedure (%) (force) exceedance (%) procedure (%)

1% interstory drift ratio limit


Two story 1% ‐ CDBE‐ SF = 1ASCE7–16 8.10 0.00 4.90 0.00
1% ‐ CDBE ‐ SF = 1.3Miyamoto 8.10 0.00 3.60 26.53
1% ‐ CMCE ‐ SF = 1proposed procedure 4.90 39.51 2.90 40.82
Four story 1% ‐ CDBE‐ SF = 1ASCE7–16 7.70 0.00 4.60 0.00
1% ‐ CDBE ‐ SF = 1.3Miyamoto 7.70 0.00 4.60 0.00
1% ‐ CMCE ‐ SF = 1proposed procedure 4.60 40.26 3.20 30.43
Eight story 1% ‐ CDBE‐ SF = 1ASCE7–16 9.00 0.00% 6.10 0.00
1% ‐ CDBE ‐ SF = 1.3Miyamoto 9.00 0.00 3.70 39.34
1% ‐ CMCE ‐ SF = 1proposed procedure 5.30 41.11 1.90 68.85
1.5% interstory drift ratio limit
Two story 1.5% ‐ CDBE ‐ SF = 1ASCE7–16 15.50 0.00 8.10 0.00
1.5% ‐ CMCE ‐ SF = 1proposed procedure 10.00 35.48 4.90 39.51
Four story 1.5% ‐ CDBE ‐ SF = 1 ASCE7–16 12.90 0.00 7.70 0.00
1.5% ‐ CMCE ‐ SF = 1proposed procedure 9.70 24.81 6.00 22.08
Eight story 1.5% ‐ CDBE ‐ SF = 1ASCE7–16 26.00 0.00 14.30 0.00
1.5% ‐ CMCE ‐ SF = 1proposed procedure 14.50 44.00 6.80 52.45

7 | S E I S M I C R E S P O N S E A S SE S S M E N T I N DB E A N D M C E

Tables 12 and 13 present the average values of the maximum roof displacement of damped structures from the records of Table 6 and compare
this parameter at two hazard levels (DBE and MCE). The proposed method, in addition to postponing the limit state behavior that saves a damper
from damage, also reduces maximum roof displacement at the DBE and MCE levels. The average value of the roof displacement reduction in the
three different structures is approximately 10%.

TABLE 12 The average maximum roof displacement in design‐basis earthquake and maximum considered earthquake levels (1% drift limit)
Maximum roof displacement (m)
Relative reduction in maximum
1%‐CDBEASCE procedure 1%‐CMCEproposed procedure roof displacement (%)
Two story DBE 0.089 0.077 13.3
MCE 0.160 0.141 11.4
Four story DBE 0.174 0.157 10
MCE 0.297 0.268 9.9
Eight story DBE 0.288 0.267 7.5
MCE 0.488 0.406 9
14 of 15 SEPEHRI ET AL.

TABLE 13 The average maximum roof displacement in design‐basis earthquake and maximum considered earthquake levels (1.5% drift limit)

Maximum roof displacement (m)


Relative reduction in maximum
1.5%‐CDBEASCE procedure 1.5%‐CMCEproposed procedure roof displacement (%)
Two story DBE 0.114 0.104 8.5
MCE 0.189 0.175 7.7
Four story DBE 0.215 0.197 8
MCE 0.375 0.342 7.2
Eight story DBE 0.315 0.279 11.6
MCE 0.506 0.454 10.3

8 | C O N CL U S I O N S

In this paper, by adjusting the viscous damping coefficient, a proposed method was employed to mitigate the probability of damage to nonlinear
dampers and to postpone the initiation of damper limit states. To this end, three structures with a different number of stories were considered,
employing diagonal viscous dampers. Various design categories, target drift ratios, and safety factors were examined to validate the efficiency of
the proposed procedure. Limit state behavior of the viscous fluid dampers was numerically modeled in the OpenSees environment, and a series of
nonlinear time history analyses were performed on different structures. Based on the exhaustive numerical study, the following conclusion are drawn:

1. For target design, in which the interstory drift ratio equals 1%, use of the proposed procedure simultaneously diminishes the limit state prob-
ability and improves collapse margin ratio of the structure. Remarkably, this improvement was achieved with no additional cost, using
dampers with capacity similar to those of the Miyamoto procedure.
2. Using the proposed procedure significantly decreases the probability of reaching the stroke and force limit at the MCE level, whereas utilizing
a safety factor of 1.3 only influences the probability of reaching the force limit.
3. For target design, in which the interstory drift ratio equals 1.5%, the proposed method resulted in the same damper size as that of the
ASCE7–16 method. However, dynamic analysis indicated that limit state initiation probability and seismic demands of the structures were
decreased at different seismic levels.
4. The fragility curve for limit state occurrence indicated that the proposed procedure decreased the bottoming‐out probability and the damper
force exceedance as well. In other words, the failure probability of the damper device in a large earthquake is reduced, in comparison with the
ASCE7–16 method. In addition, the suggested procedure preserves dampers in a large earthquake without the employment of stronger dampers.
5. Here, the limit state behavior is defined when either the damper force exceeds its strength capacity or the first observation of bottoming out
occurs. The results indicated that the latter occurs at a lower seismic level.

6. The proposed method improved seismic performance of damped structures at DBE and MCE intensities. This improved roof displacement
reduction by 10% in comparison with ASCE7–16.
7. It was found that the median limit states spectral accelerations decreases by increasing the number of building stories,
8. As a modification to the ASCE7–16 provision, it is more effective to employ the damping coefficient of a nonlinear viscous fluid device at the
MCE level to provide the required damping ratio at the DBE level.

ORCID
Ahmad Sepehri http://orcid.org/0000-0003-4874-4300

RE FE R ENC E S
[1] H. K. Miyamoto, A. S. Gilani, A. Wada, C. Ariyaratana, J. Earthq. Spectra 2011, 27(4), 1147.
[2] M. Constantinou, M. Symans. Experimental and analytical investigation of seismic response of structures with supplemental fluid viscous dampers,
National Center for Earthquake Engineering Research (NCEER) Technical Report,1992
[3] A. Seleemah, M. Constantinou. Investigation of seismic response of buildings with linear and nonlinear fluid viscous dampers, National Center for
Earthquake Engineering Research (NCEER) Technical Report, 1997.
[4] C. Patel, R. Jangid, The IES J. part a: Civil & Struct. Eng. 2010, 3(1), 1.
[5] J. Whittle, M. Williams, T. L. Karavasilis, A. Blakeborough, J. Earthq. Eng. 2012, 16(4), 540.
[6] M. G. Oesterle. Use of incremental dynamic analysis to assess the performance of steel Moment‐resisting frames with Fluid viscous dampers. M.Sc.
Thesis, Virginia Virginia Polytechnic 2003.
[7] S. J. Kruep. Using incremental dynamic analysis to visualize the effects of viscous fluid dampers on steel moment frame drift. M.Sc. Thesis. Virginia
Virginia Polytechnic 2007.
[8] C. Y. Seo, T. L. Karavasilis, J. M. Ricles, R. Sause, J. Earthq. Eng. Struct. Dyn. 2014, 43(14), 2135.
SEPEHRI ET AL. 15 of 15

[9] Z. Miao, Q. e. Song, K. Chen, Y. Lu. Comparison of collapse‐resistance capacities of RC frames with and without viscous dampers, Presented at Int.
Conf. on Performance‐based and Life‐cycle Structural Engineering, University of Queensland, January 2015.
[10] M. Hamidia, A. Filiatrault, A. Aref, J. Struct. Eng. 2014, 141(6), 04014153.
[11] M. Hamidia, A. Filiatrault, A. Aref, J. Earthq. Eng. 2014, 18(4), 528.
[12] M. Cao, H. Tang, N. Funaki, S. Xue. Study on a real 8F steel Building with oil damper damaged during the 2011 Great East Japan Earthquake, Presented
at the 15th Int. Conf. on Earthquake Engineering, Lisbon, Portugal, 2012.
[13] H. K. Miyamoto, A. S. Gilani, A. Wada, C. Ariyaratana, J. Earthq. Eng. Struct. Dyn 2010, 39(11), 1279.
[14] H. K. Miyamoto. Probabilistic seismic risk identification of steel building with viscous, Tokyo Institute of Technology, 2010.
[15] H. K. Miyamoto, A. S. Gilani, A. Wada, C. Ariyaratana, J. Struct. Des. Tall Spec 2010, 19(4), 421.
[16] H. K. Miyamoto, A. S. Gilani, A. Wada. Collapse Hazard and Design Process of Essential Buildings with Dampers, China/USA Symp. for the Advance-
ment of Earthquake Sciences and Hazard Mitigation Practices, 2008.
[17] M. Ahmadi Namin. Collapse Risk of Steel Moment Frame Structures with Fluid Viscous Dampers subjected to Near Field Pulse like Earthquakes. M.Sc.
Thesis, Amirkabir University of Technology, Tehrean, Iran 2013.
[18] S. Kitayama, M. C. Constantinou, J. Earthq. Eng. Struct. Dyn. 2016, 45(12), 1335.
[19] ASCE, Seismic rehabilitation of existing buildings (ASCE/SEI 41–06). American society of civil engineers, American Society of Civil Engineers (ASCE,
Reston, Virginia 2006.
[20] ASCE, Minimum design loads for buildings and other structures (ASCE/SEI 7‐10). American society of civil engineers, American Society of Civil Engineers
(ASCE, Reston, Virginia 2010.
[21] ASCE, Minimum design loads for buildings and other structures (ASCE/SEI 7‐16). American society of civil engineers, American Society of Civil Engineers
(ASCE, Reston, Virginia 2016.
[22] ASCE, Seismic rehabilitation of existing buildings (ASCE/SEI 41–17). American society of civil engineers, American Society of Civil Engineers
(ASCE, Reston, Virginia 2017.
[23] J.‐S. Hwang, Y.‐N. Huang, S.‐L. Yi, S.‐Y. Ho, J. Struct. Eng. 2008, 134(1), 22.
[24] O. M. Ramirez, M. C. Constantinou, C. A. Kircher, A. S. Whittaker, et al. Development and evaluation of simplified procedures for the analysis and
design of buildings with passive energy dissipation systems. Ph.D. Thesis, State University of New York at Buffalo 2001.
[25] F. Y. Cheng, H. Jiang, K. Lou, Smart structures: innovative systems for seismic response control, CRC Press 2008.
[26] American Institute of Steel Construction (AISC). Specification for Structural Steel Buildings. AISC 360‐16, Chicago, IL, USA. 2016.
[27] Fluid Viscous Dampers & Lock‐up Devices Catalogue, Taylor Devices Inc., http://taylordevices.com/pdf/2016/01Jan/DamperLUDsizesPDF2.pdf
(accessed: September 2016).
[28] S. Mazzoni, F. McKenna, M. H. Scott, G. L. Fenves, OpenSees command language manual, Pacific Earthq. Eng. Res. Cent 2007, 451.
[29] L. F. Ibarra, H. Krawinkler, Global Collapse of Frame Structures under Seismic Excitations, California, Berkeley 2005.
[30] D. G. Lignos, H. Krawinkler, J. Struct. Eng. 2010, 137(11), 1291.
[31] D. Lignos, H. Krawinkler, A. Whittaker, J. Earthq. Eng. & Struct. Dyn 2011, 40(7), 807.
[32] D. Lignos. Sidesway collapse of deteriorating structural systems under seismic excitations. Ph.D. Thesis, Stanford university 2008.
[33] FEMA. Quantification of building seismic performance factors Federal emergency management agency, (FEMA P695) Washington, D.C. 2009.
[34] FEMA. Recommended seismic design criteria for new steel moment frame buildings Federal emergency management agency, (FEMA 350) Washington,
DC. 2000.

AUTHOR BIOGRAPHI ES
Ahmad Sepehri graduated from Amirkabir University of Technology with master's degree in Earthquake Engineering. His research interests are
earthquake engineering and structural control.

Touraj Taghikhany is Associate Professor in Amirkabir University of Technology, Tehran Iran. His research interests are performance‐based
earthquake engineering and structural control.

Seyed Mohammad Reza Ahmadi Namin graduated from Amirkabir University of Technology with master's degree in Earthquake Engineering.
His research interests are earthquake engineering and structural control.

How to cite this article: Sepehri A, Taghikhany T, Ahmadi Namin SMR. Seismic design and assessment of structures with viscous dampers
at limit state levels: Focus on probability of damage in devices. Struct Design Tall Spec Build. 2019;28:e1569. https://doi.org/10.1002/
tal.1569

You might also like