Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Engineering Structures 32 (2010) 1113–1122

Contents lists available at ScienceDirect

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

A brace-type seismic damper based on yielding the walls of hollow


structural sections
Amadeo Benavent-Climent ∗
Department of Structural Mechanics, University of Granada, Polytechnic Building of Fuentenueva Campus. Granada 18071, Spain

article info abstract


Article history: This paper proposes a new brace-type seismic damper designed for earthquake protection of structures.
Received 5 August 2009 It consists of a tube-in-tube assemblage of two commonly available hollow structural sections. The outer
Received in revised form hollow section of the Tube-in-Tube Damper (TTD) has a series of strips created by cutting a series of
8 November 2009
slits though the wall, and it is welded to the inner hollow section in such a way that when the brace
Accepted 22 December 2009
Available online 12 January 2010
damper is subjected to forced displacements in the direction of its axis, the strips dissipate energy through
flexural/shear yielding. In comparison with other existing brace-type dampers such as the popular
Keywords:
buckling restrained brace, the TTD has the advantages of being much simpler – thus entailing a lower cost
Energy dissipation – and the yielding part of the damper can be easily inspected after an earthquake. The performance of the
Hysteretic damper proposed damper was assessed experimentally by an initial series of four cyclic tests, and compared with
Metallic damper that of conventional steel plates with slits in another series of five tests. The experimental results show
Passive control that the damper possesses very stable hysteric characteristics and excellent energy dissipation capacity,
Earthquake resistant structure both similar to the conventional steel plates with slits already in use for seismic applications. Based on
the test results, a hysteric model and a procedure for predicting the ultimate energy dissipation capacity
and failure of the new damper are proposed.
© 2009 Elsevier Ltd. All rights reserved.

1. Introduction increasingly taken into consideration in USA. Likewise, since the


Kobe (1995) earthquake in Japan, more buildings have been
Interest in the earthquake risk mitigation of buildings through designed to include dampers [1,2]. A number of mechanisms have
passive energy dissipation has greatly augmented in the last been used for passive energy dissipation, including metal yielding,
decades with a rapid increase of implementations starting in the phase transformation of metals, friction sliding, fluid orificing, and
mid-1990s. In contrast to the traditional seismic design approach deformation of viscoelastic solids or liquids. The yielding of metals
that relies on the inelastic deformation of particular zones of the is one of the most popular mechanisms and numerous metallic
structure to dissipate most of the energy input by the earthquake dampers with different yielding schemes have been proposed and
(commonly, beam-ends and column-ends on moment-resisting installed [3–5]. Fig. 1 shows a selection of the most popular ones.
frames), in the passive control systems this energy is delivered to The well-known hourglass shape ADAS damper [6] or its vari-
special devices called seismic dampers. This has many advantages: ant, with a triangular shape, the TADAS damper [7], makes use
(i) the inelastic deformations are concentrated in the seismic of the flexural deformation of metallic plates under out-of-plane
dampers and the damage in the parent structure can be drastically bending as shown in Fig. 1(a) and (b). In the honeycomb damper [8]
reduced or eliminated; (ii) the addition of damping reduces or the slit damper [9,10], respectively shown in Fig. 1(c) and (d), a
the lateral displacements of the structure, which also reduces metallic plate with a number of openings is subjected to in-plane
damage to non-structural elements; (iii) by strategically locating shear deformations and energy is dissipated through flexural/shear
yielding of the steel strips between the openings. Other devices
the seismic dampers, their inspection, repair and/or replacement
utilize energy dissipation through plastic shear deformations of
following an earthquake can be carried out with minimal cost
metallic panels (with or without ribs) welded within to an en-
and without interrupting occupancy. Passive energy dissipation
closing steel frame that provides support along its boundaries as
systems are now recognized as an effective and inexpensive way
shown in Fig. 1(e) [11]. One widely used metallic dampers is the
to mitigate earthquake risks to structures. Since the Northridge
so-called Buckling Restrained Brace (BRB) [12–15], based on yield-
earthquake (1994), passive damping technologies have been
ing a steel brace with a cruciform section that is surrounded by a
stiff steel tube as shown in Fig. 1(f). The BRB is mounted diagonally
within a structural frame as a conventional brace or brace-type
∗ Tel.: +34 958 246109; fax: +34 958 246109. seismic damper. The central steel core of the BRB is capable of with-
E-mail address: benavent@ugr.es. standing high axial loads, and the steel tube – casing – filled with
0141-0296/$ – see front matter © 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2009.12.037
1114 A. Benavent-Climent / Engineering Structures 32 (2010) 1113–1122

a b c d e f
Steel core that yeilds
Buckling
restraining
tube
Concrete
Yielding
panel

Inclosing frame Unbond material

Fig. 1. Types of metallic dampers: (a) ADAS; (b) TADAS; (c) honeycomb damper; (d) slit damper; (e) shear panel damper; (f) buckling restrained brace.

concrete or mortar keeps it from buckling under axial compressive a brace damper b brace damper
loads. As the BRB yields properly under tensile and compressive
loads, the loads must be borne by the steel core as opposed to the
casing. A slip interface, or ‘‘unbonding’’ layer, is therefore provided
between the steel core and the surrounding concrete to decouple
the steel core from the concrete and ensure that compression and
tensile loads are carried only by the steel core.
The BRB was originally developed and deployed in Japan, and
now is one of the most popular metallic dampers, very widely used
in the United States and other countries. One of the advantages of
the BRB is that their strength and stiffness can be adjusted indepen-
dently. Another is that it can be easily installed in a frame just like
a conventional diagonal bar; that is, it is not necessary to resort to Fig. 2. Installation schemes of the seismic damper.
auxiliary substructures, such as the typical chevron-brace arrange-
ments in ADAS and TADAS dampers, to assemble the energy dissi- 2. Proposed seismic damper
pating device to the frame. However, the BRB has also important
disadvantages. One is that it is impossible to inspect the steel core The proposed seismic damper has the form of a conventional
without having to destroy the braced frame itself [15]. This may
brace and it is intended to be installed in a framed structure as
imply having to replace the seismic dampers of the entire build-
a standard diagonal bar, as shown in Fig. 2. It is constructed by
ing after a moderate earthquake, even when BRB has left enough
assembling two standard hollow structural rectangular sections,
energy dissipation capacity to protect the main structure under ad-
one into the other, forming a tube-in-tube type bracing member as
ditional earthquakes. The second disadvantage is that the materi-
shown in Fig. 3. In the walls of the outer hollow section, a number
als and geometry of the slip interface or ‘‘unbonding’’ layer must
of slits are cut leaving a number of strips between the slits. The two
be very carefully designed and constructed to allow relative move-
tubes are joined by fillet and plug welds in specific points as shown
ments between the steel core and the concrete due to shearing and
Poisson’s effect, while simultaneously inhibiting local buckling of in Fig. 3. Under relative displacements of the ends of the brace
the core as it yields in compression [15]. In fact, placement of the damper in the direction of its axis, the strips behave as a series of
unbonding material, and pouring and curing of the concrete fill, fixed-ended beams and deform in double curvature. The tube-in-
contribute largely to the manufacturing costs of the BRB and make tube configuration and the overlapping length of one tube into the
it unaffordable for many conventional structures. In fact, the pos- other increase the buckling capacity. The position of the slits along
sibility of eliminating the unbonding material/pouring/curing of one of the tubes is not predetermined; they can be situated at the
the concrete fill has been examined by several researchers in the ends, or in the middle part to facilitate inspection after a seismic
past [16–18]; but the proposed solutions require very tight fabri- event.
cation tolerances that are difficult to fulfill in practice and translate The axial yielding load Qy and apparent maximum load QB of
as less stable hysteretic performance and lower energy dissipation the strips can be readily determined from their geometry, and the
in comparison to the concrete filled tube design. steel properties from fundamental engineering principles, giving:
This paper proposes a new seismic damper based on the yield-
fy tb2 fB tb2
   
ing of metals that keeps the advantages of BRB – i.e. it can be easily 2fy tb 2fB tb
Qy = min n ,n √ ; QB = min n ,n √ . (1)
installed like a conventional brace, and strength and stiffness can 2h0 3 3 2h0 3 3
be adjusted independently – while avoiding several of its short-
comings. In contrast to the BRB the yielding part of the proposed Here b and t are the strip width and thickness, respectively, and r
damper can be easily inspected after an earthquake, and costs may is the radius of the end part of the strip as shown in Fig. 4(a); n is
be reduced since the manufacturing process is much simpler. The the total number of strips in the damper; fy and fB are the yielding
proposed damper uses as the source of energy dissipation the very and maximum tensile stress of the steel; and h0 = h + [2r 2 /(h +
same principle that the steel plate with slits damper; yet in con- 2r )] is the height of an equivalent strip defined in Fig. 4(b),
trast, auxiliary substructures such as chevron braces or walls are where h is the height of the actual strip. The factor 2/3 present
not required for assembling the damper to the structure. The hys- in the shear strength term of Eq. (1) acknowledges the relation
teretic behavior and the energy-dissipative characteristics of the between the average shear stress and the maximum shear stress
proposed damper are investigated here via component tests, and in rectangular cross-sections. In Eq. (1) the interaction between
compared with those of the conventional steel plate with slits. shear and flexure has been neglected for the sake of simplicity.
Based on the test results, a simple hysteretic model for represent- When flexural and shear yielding limit states have similar values,
ing the load–displacement curve of the new seismic damper and the strength reduction due to this interaction is below 20%.
a procedure for predicting its ultimate energy dissipation capacity In this equivalent strip, for simplicity, the round-shape ends
and failure under arbitrarily changing cyclic load are put forth. have been replaced by straight lines, and height h0 was determined
A. Benavent-Climent / Engineering Structures 32 (2010) 1113–1122 1115

SECTION A-A' 1st tube strip sli t A' 2nd tube axis of the brace damper

B B'
fillet weld plug weld
A

SECTION B-B'

Fig. 3. Proposed seismic damper.

A
a b

2r 2

h+2r
h+2r
h'=h+

h
h+2r
r

h
b r b r
Section A - A ' A' thickness t b A ctual trip Equivalent strip

Fig. 4. Proposed damper: (a) detail of the energy-dissipative part; (b) equivalent strip.

through numerical analysis using finite element models, apply- specimens the outer tube was a 160×160×8 mm (width×height×
ing the condition that actual and equivalent strips have similar thickness) hollow section, while the inner tube dimensions were
strength and stiffness in the axial direction of the damper. In Eq. (1), 140 × 140 × 8 mm. Slits were cut in two opposite walls of the outer
min indicates that the minimum of the two terms in parenthesis tube. The differences among the specimens were the width b and
must be adopted (one corresponds to the flexural yielding and the the total number of strips n. Two specimens with b = 15 mm and
other to the shear yielding condition). By using the equivalent strip n = 20, and other two specimens with b = 20 mm and n = 12,
approximation of Fig. 4(b), and by integrating the curvatures and were made. All outer tubes were cut from a single segment of a
the shear strains along the total height h + 2r, the yielding displace- structural hollow section, whose mechanical properties averaged
ment of the strips, δy , can be expressed by: from two standard coupon tests were: fy = 449 MPa, fB = 512 MPa
Qy h03 3Qy h0 and E = 200626 MPa. The specimens were subjected to load cycles
   
h + 2r h + 2r
δy = 1 + 3 ln + 1 + ln . (2) of increasing amplitude; the amplitude increase for each consecu-
nEtb3 h0 2ntbG h0
tive cycle 1δ normalized by δy , i.e. the φ = 1δ/δy , was varied in
Here E is Young’s modulus and G the shear modulus of the mate- the experiments. A wide range of values of φ was selected in order
rial. In Eq. (2), the shear deformation of the strip is included since to obtain different patterns of energy consumption between the
its contribution can be large for small ratios h/b. The total axial dis- so-called ‘‘skeleton part’’ and the ‘‘Bauschinger part’’, as explained
placement at yielding of the proposed brace-type seismic damper
later in Section 4.1. Combining b, n and φ , the four tests were con-
is readily obtained by adding to Eq. (2) the elastic axial deformation
ducted, referred to as TTD15/20/5, TTD15/20/20, TTD20/20/5 and
of the tubes.
TTD20/20/20 in Table 1.
3. Experimental investigation In the test series USP, five specimens similar to those shown in
Fig. 5(b) were made. Each specimen consisted of an arrangement of
To investigate the hysteretic behavior and energy dissipation two identical steel plates with slits. Each plate was welded along
capacity of the proposed damper, a series of tests were conducted. one edge to the web of an H-shape steel member, and along the
In these tests – hereafter referred to as the series TTD (Tube-in- opposite edge to a rectangular steel plate — referred to as the link
Tube Damper) – full-scale specimens representative of the energy- element in Fig. 5(b). All steel plates with slits were cut from a sin-
dissipative part of the damper were subjected to cyclic loadings gle plate t = 12 mm thick, whose mechanical properties averaged
until failure. It should be underlined that in the proposed damper from two standard coupon tests were: fy = 349 MPa, fB = 509 MPa
the relative displacement of the ends of the strips in the direction and E = 210000 MPa. The width b and the total number of strips
perpendicular to the axis of the brace (see Fig. 3) is constrained by was the same in all specimens, b = 20 mm and n = 20, the only
the rest of the tube. Therefore, the strips will be subjected to some difference being the height h. The specimens were also subjected to
force perpendicular to the direction of the applied displacement cyclic forced displacements of increasing amplitude with different
which might influence their cyclic performance and ultimate values for φ . Combining h and φ the five tests labeled USP20/40/4,
energy dissipation capacity. To clarify this point, the results of the USP20/20/35, USP20/20/9 USP20/80/9 and USP20/80/23 in Table 1
TTD test series were compared with a second series of cyclic tests
were carried out. The values of Qy and δy in Table 1 for the test
conducted on specimens for which slit plates were arranged so that
specimens USP correspond to 5 strips.
the strips could freely deform in the direction perpendicular to the
applied force. This second test series will be referred to as series
USP (Unconstrained Slit Plate) hereafter. 3.2. Test set-up and instrumentation

3.1. Specimens and loading history Fig. 5(c) shows the set-up used for testing the nine specimens.
Thick end plates were welded to the tubes (in case of test series
The test series TTD included four specimens, each consisting of TTD) or to the H-shaped members (in case of test series USP)
a tube-in-tube arrangement similar to that shown in Fig. 5(a). In all and clamped solidly by bolts to the loading head and to the base
1116 A. Benavent-Climent / Engineering Structures 32 (2010) 1113–1122

Table 1
Geometry of test specimens, loading amplitudes and energy dissipation capacity.
Specimen t (mm) b (mm) h (mm) n φ Qy kN δy (mm) ep η S η B η η
TTD15/20/5 8 15 20 20 5 323 0.17 61 73 572 645
TTD15/20/20 8 15 20 20 20 323 0.17 107 135 364 500
TTD20/20/5 8 20 20 12 5 332 0.16 79 96 450 546
TTD20/20/20 8 20 20 12 20 332 0.16 160 309 155 464
USP20/40/4 12 20 40 20 4 97 0.24 109 181 593 774
USP20/20/35 12 20 20 20 35 162 0.16 176 256 151 407
USP20/20/9 12 20 20 20 9 162 0.16 128 194 431 625
USP20/80/10 12 20 80 20 10 51 0.52 140 245 459 704
USP20/80/23 12 20 80 20 23 51 0.52 149 262 234 496

a b c
displacement
transducer

A A'

A A'
LOADING HEAD

4500 mm
welding
test dispace.
specimen transducer
H - shape
member
ADJUSTBALE BASE

Link
element

Section A–A' Section A–A'


Specimens TTD Specimens USP 1925 mm

Fig. 5. Test specimens for (a) series TTD, (b) series USP, and (c) test set-up.

of a displacement-controlled servopulser machine. Forced vertical (Fig. 6(c) to (d)) reveals that the two exhibit very similar character-
displacements were applied quasi-statically by the loading head of istics: hysteretic loops approximately rectangular, free of undesir-
the servopulser. A load cell installed in the servopulser measured able pinching effects. In the test series USD, the maximum strength
the applied force and several displacement transducers measured attained by the steel plates with slits due to strain hardening effects
the relative displacements of the ends of the strips in the direction was between 1.5 to 2.0 times the yield strength, while in the test se-
of the applied load. The tests were carried out until complete ries TTD it is about 1.25Qy . This difference can be attributed, at least
failure of the specimens. Failure was assumed to occur when the partially, to the fact that although the fB of the steel used in both
applied force began to decrease steadily under increasing forced test series was similar, the ratio fB /fy was smaller (about 25%) in the
displacements. TTD test series than in USD. From these results, it can be concluded
that the effect of the spurious force acting on the strips perpendic-
3.3. Overall test results ularly to the direction of the axis of the proposed TTD damper is of
minor importance. A quantitative comparison between test series
TTD and USD in terms of dissipated energy is described in the next
Fig. 6 shows four typical load–displacement relationships, Q –δ ,
section.
obtained from the tests. Fig. 6(a) and (b) correspond to test series
TTD, and Fig. 6(c) and (d) to test series USD. The Q –δ relationships
4. Analysis of test results and discussion
of the rest of specimens showed very similar patterns. All curves
have been normalized dividing the abscissa and ordinate by the
yielding displacement δ y and by the yielding force Qy , respectively, 4.1. Energy dissipation capacity
given by Eqs. (1) and (2). As can be seen in Fig. 6(a) and (b), the pro-
posed damper has very stable hysteretic behavior and the shape 4.1.1. Decomposition of the load–displacement curves
of the loops – close to a rectangle – indicates an excellent ability The ultimate energy dissipation capacity of metallic dampers
to dissipate energy. Qualitative comparison of the cyclic responses depends greatly upon the loading pattern applied. One way to rep-
of the tube-in-tube arrangement of slit plates as proposed in this resent this dependency [19] calls for first splitting the total energy
paper for the new seismic damper (Fig. 6(a) and (b)) with those dissipated by the damping device into the so-called skeleton part
obtained for the conventional unrestrained steel plates with slits and Bauschinger part, as shown in Fig. 7. Fig. 7(a) shows a typical
A. Benavent-Climent / Engineering Structures 32 (2010) 1113–1122 1117

a 2.0
Q /Qy b 2.
Q/Q y
1.5 1.5

1.0 1.0

0.5 0.5
δ /δ y δ /δ y
0.0 0.0
-40 -2.0 0. 20 40 -60 -40 -2 0 0 20 40
-0.5 - 0.5

-1.0 - 1.0

-1. 5 - 1.5

-2.0 - 2.0

c 2.0 Q /Qy 2.0


d Q/ Q y
1.5 1.5

1.0 1.0

0.5 0.5
δ /δ y δ /δ y
0.0 0.0
-40 -20 0 20 40 - -40 - 20 0 20 40
- 0.5 -0.5

-1.0 -1.0

-1 .5 -1.5

- 2.0 -2.0

Fig. 6. Normalized Q –δ curve obtained for test specimen: (a)TTD15/20/5; (b)TTD15/20/20; (c) USD20/40/4; (d) USD20/20/9.

load–displacement, Q –δ , curve obtained from a cyclic test un- dissipated by the damping element that is consumed on the skele-
til failure. We may decompose the total amount of plastic strain ton part, and it will be expressed as S Wu+ and S Wu− .
energy dissipated by the damping device as follows: segments Segments 4–5, 10–11, and 16–17 in the positive domain, and
0–1, 5–6, 11–12, and 17–18 in the positive domain, and 2–3, 8–9, 7–8, plus 13–14 in the negative domain of loading, begin at Q = 0
14–15 in the negative domain of loading, are the paths that ex- and terminate at the maximum load level as attained in preceding
ceed the load level attained by the preceding cycle in the same cycles in the same loading domain. These paths, softened by the
domain of loading. By connecting these segments sequentially as Bauschinger effect, will be referred to as ‘‘Bauschinger segments’’.
shown in Fig. 7(b), two curves are obtained (herein referred to as They are depicted in Fig. 7(c). For each domain of loading, the sum
‘‘skeleton curves’’). As verified experimentally by Kato et al. [20], of the areas enveloped by each Bauschinger segment, by the un-
under any arbitrarily changing history of deformation, these skele- loading path passing through the point of maximum load of the
ton curves will approximately coincide with the Q –δ relationship segment, and by the horizontal axis, will represent the Bauschinger
obtained under monotonic loading. The skeleton curves can be ap- portion of total plastic strain energy dissipated by the steel compo-
proximated by the trilinear curve shown with dash and dot lines in nent; it will be referred to as B Wu+ and B Wu− .
Fig. 7(b), defined by Qy , δy , the first and second plastic stiffness Kp1
By splitting the Q –δ curve as explained above, the total plas-
and Kp2 (Kp1 > Kp2 ), and the load QB , which determines the transi-
tic strain energy dissipated by the damping device in each domain
tion point from Kp1 to Kp2 . Segments 1–2, 6–7, 12–13, 18–19, 3–4,
of loading until failure can be decomposed into the skeleton part
9–10, and 15–16 correspond to the unloading paths, and their slope
(S Wu+ , S Wu− ) and the Bauschinger part (B Wu+ , B Wu− ). Then, S Wu+ ,
roughly coincides with the initial elastic stiffness Ke (=Qy /δy ). In
S Wu , B Wu , B Wu , S δu and S δu can be expressed non-dimensionally
− + − + −
Fig. 7(b), S δu+ and S δu− denote the plastic deformation accumulated
as follows,
in each skeleton curve when the damper fails, and S δB is the plas-
tic deformation accumulated in the approximate trilinear skeleton + − +
S Wu S Wu B Wu
curve at Q = QB . For each domain of loading then, the area en- S η̄+ = ; S η̄− = ; B η̄+ = ;
veloped in Fig. 7(b) by the skeleton curve, by the unloading path Qy δy Qy δy Qy δy
(3)
passing through the point of maximum load, and by the horizon- −
B Wu δ+ S u δ− S u
B η̄ =

; ep η̄ =
+
; ep η̄ =

,
tal axis would represent the portion of total plastic strain energy Qy δy δy δy
1118 A. Benavent-Climent / Engineering Structures 32 (2010) 1113–1122

a b Q K p2
12,17
18
QB K p 1 6,11
Qy 1,5
sδ u
– Ke Ke Ke Ke Ke

sδ 3 sδ 2
– sδ 1

0,2
+ + sδ
δy s δ2 sδ 3 sδ 4
Q 17 18 Ke Ke Ke +
- sδ u
Q 12 sδ 1 sδ B
6 11 6 3,8
5 1 9,14 Approximate skeleton curve
δ 15
16 10 4 0 2 7 13 19 +
+ =
δ max δ max c Q 17
11
3 8 5
9
15 14 Ke Ke Ke

– –
Bδ 2 13 B δ 1 4
7 10 16 +
Bδ 2 Bδ3
Ke Ke Bδ 1

8
14

Fig. 7. Decomposition Q –δ curve: (a) overall curve; (b) skeleton part; (c) Bauschinger part.

where Qy and δy are given by Eqs. (1) and (2). For convenience, we response consists of steady-state harmonic displacements, and
define the following new ratios: Manson–Coffin’s formula and Miner’s rule can be easily applied.
In the case of seismic excitations, however, the vibration on the
η = S η̄+ + S η̄− ; η = B η̄+ + B η̄− ;

S B structure is of a random nature and the inelastic response of the
(4)
ep η = ep η̄
+ ep η̄ ; η = S η + Bη structural members is characterized by non-steady development
+ −

of plastic deformations, which can be hardly expressed in terms
epη will be referred to as the apparent ultimate cumulative plastic of the number and amplitude of deformation cycles. Expressing
deformation ratio on the skeleton part. S η is the ultimate cumu- the inelastic response in this way would require the use of count-
lative plastic deformation ratio on the skeleton part. B η is the ulti- ing methods to convert the arbitrary individual excursions from a
mate cumulative plastic deformation ratio on the Bauschinger part. seismic response history into a sequence of closed cycles of con-
η is the total ultimate cumulative plastic deformation ratio which stant amplitude – such as the rain-flow or the range-pair method
represents in non-dimensional form the total energy dissipated by – which are suitable only for strain histories with few reversals to
the damper until failure Wu (=S Wu+ + S Wu− + B Wu+ + B Wu− ). failure.
Proceeding in this way, the Q –δ curves obtained for each test So as to derive analytical expressions that quantify accurately
were decomposed into the skeleton part and the Bauschinger part, the ultimate energy dissipation capacity of the proposed seismic
and the ratios defined in Eq. (4) were computed. The results are damper under earthquake-induced cyclic loads accounting for the
shown in Table 1. As noted above, Table 1 clearly shows that the dependency on the loading path, the results of the tests have been
total energy dissipated by the damping device, as expressed by organized and analyzed below.
η, is highly dependent on the loading pattern or loading path. If
the damping device is subjected to cycles with small increases
4.1.2. Estimation of the energy dissipated on the Bauschinger part
in amplitude φ (i.e. test specimens TTD15/20/5, TTD20/20/5,
USP20/40/4, USP20/20/9, USP20/80/10), η is larger than when To study the energy dissipated on the Bauschinger part, the re-
cycles with large φ are applied. In the latter case – i.e. large φ – lation between ep η and B η is plotted in Fig. 8(a). The solid symbols
the apparent cumulative plastic deformation ratio on the skeleton correspond to the test series TTD and the open symbols to the test
part ep η is greater than in the former, which considerably reduces series USD. A nearly linear relationship is seen between ep η–B η that
the overall energy dissipation capacity η as discussed now in can be formally expressed by:
Sections 4.1.2 and 4.1.3. B η = aep η + b. (5)
We should also point out that the decomposition shown in Fig. 7
can be applied to any arbitrarily varying pattern of cyclic load. The value of the slope a and the ordinate of the intersection point
Therefore, the procedure presented here to evaluate the ultimate with the vertical axis b calculated from a linear regression analysis
energy dissipation capacity of the damper can be easily applied gives approximately a = −4 and b = 900, shown with a dotted
to random loading histories such those induced by earthquakes. line in Fig. 8(a). For the purposes of designing, adoption of the lower
This is a distinctive aspect of the proposed procedure as opposed bound defined by a = −4 and b = 800 is proposed, shown with a
to approaches based on the Manson and Coffin formula [21,22] or dashed line in the figure, which coincides approximately with the
Miner’s rule [23]. The Manson–Coffin formula predicts the num- results of the test series TTD.
ber of cycles that cause the failure of a steel element at one given
constant amplitude. When the steel element is subjected to load- 4.1.3. Estimation of the energy dissipated on the skeleton part
ing cycles of a different amplitudes, Miner’s linear-accumulative- To study the energy dissipated on the skeleton part, the skeleton
damage criterion predicts failure in terms of the numbers of curves obtained from the test series TTD in the positive and
cycles applied divided by the number to produce failure at a given negative domains were taken in absolute values, i.e. |Q |–|S δ|, and
displacement amplitude. In the case, for example, of vibrations normalized by Qy and δy given by Eqs. (1) and (2); the results are
induced by rotating machinery on the supporting structure, the shown in Fig. 9. Also drawn in Fig. 9 is the approximated trilinear
A. Benavent-Climent / Engineering Structures 32 (2010) 1113–1122 1119

a b c
Bη Sη TEST η TEST
TEST
1000 APPROXIMATION 1000 APPROXIMAT IO N 1000 APPROXIMAT ION

800 800 800

600 600 600

400 400 400

200 200 200


ep η
ep η ep η
0 0 0
0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250

Fig. 8. Ultimate energy dissipation capacity: (a) Bauschinger part; (b) skeleton part; (c) total.

1.50
Trilinear approximation η is expressed with the following equations, which are drawn in
\Q\/Q y Tests k p2 Fig. 8(c):

1.25 2(τB − 1)(1 − kp1 )


for ep η ≤ :
τB kp1 kp1
1.00 kp1
η = 0.25ep η2 + ep η(1 + a) + b (8)
1 − kp1
0.75
2(τB − 1)(1 − kp1 ) (τB2 − 1)(1 − kp1 )
for: ep η > : η=
kp1 kp1
0.50
η (τB − 1)(1 − kp1 )
 
ep
+ −
0.25 2 kp1
\ sδ \ / δy
η (τB − 1)(1 − kp1 )
  
kp2 ep
× 2τB + −
0.00 (1 − kp2 ) 2 kp1
0 5 10 15 20 25 30 35
+ aep η + b. (9)
Fig. 9. Experimental, numerical and approximate skeleton curves.
Fig. 8 shows as well that the ultimate energy dissipation ca-
pacity – in terms of B η, S η, η – of the proposed seismic damper
skeleton curve whose normalized plastic stiffnesses are kp1 =
(i.e. the solid symbols corresponding to test series TTD) is, in some
1/25 and kp2 = 1/125.
cases, slightly smaller than that obtained for the unconstrained slit
From the above normalized trilinear skeleton curve defined by
plates (i.e. the open symbols corresponding to test series USD). This
τB = fB /fy , kp1 , kp2 , the energy dissipated on the skeleton part
difference might be partially attributed to the force acting on the
expressed in non-dimensional form by S η can be related to ep η,
strips perpendicularly to the direction of the applied load in the
assuming that the skeleton curves in the positive and negative
TTD damper, and partially to the different ratios fB /fy of the steels
domains of loading are equal and that |ep η̄+ | = |ep η̄− | = ep η̄ =
used. In any case, the difference is minor and from the point of view
0.5ep η. From these two assumptions it follows that S η̄+ = S η̄− =
of the author it is negligible for design purposes.
S η̄ = 0.5S η . Based on these considerations and the formulae de-
veloped in Appendix, the following expressions are obtained:
4.2. Hysteretic model and prediction of failure
2(τB − 1)(1 − kp1 ) kp1
for ep η ≤ : S η = 0.25ep η 2
+ ep η (6)
kp1 1 − kp1 In previous research on the hysteretic characterization of
metallic dampers, the simple bilinear hysteretic model [24], or the
2(τB − 1)(1 − kp1 ) (τ − 1)(1 − kp1 )
2
for: ep η > : η= S
B more sophisticated Bouc-Wen [25] or Ramberg–Osgood models
kp1 kp1 [26] are used, which are able to capture the smooth transition from
η (τB − 1)(1 − kp1 ) the elastic to inelastic regime. However, with these models it is not
 
ep
+ − possible to account for the key influence of the loading path on the
2 kp1
assessment of the state of damage of the damper, and consequently
η (τB − 1)(1 − kp1 )
  
kp2 ep on the prediction of its failure [19]. In this section, a simple hys-
× 2τB + − . (7) teretic model for constructing the load–displacement relationship
(1 − kp2 ) 2 kp1
of the proposed seismic damper under an arbitrary cyclic loading
Eqs. (6) and (7) are drawn in Fig. 8(b) together with the results of pattern is presented, along with a procedure for predicting its fail-
the tests. These equations can be seen to estimate reasonably well ure. To define the hysteretic model, the shape of the skeleton part
the ultimate energy dissipating capacity of the damping device in and the shape of each segment of the Bauschinger part must first
the skeleton part. be determined. The shape of the skeleton part, as examined (Fig. 9)
in the previous section, would suggest a trilinear idealization. The
4.1.4. Ultimate energy dissipation capacity shape of the Bauschinger parts is studied next.
Finally, by adding Eq. (5) to Eqs. (6) and (7), the ultimate Returning to Fig. 7(a), consider for example that under a
energy dissipation capacity in terms of the non-dimensional ratio given loading history the damping device has covered the path
1120 A. Benavent-Climent / Engineering Structures 32 (2010) 1113–1122

a b c
600 20
Q Test α Q (kN) δ (mm)
m B
Approximation 500
Q =Q
m 6 15
= 11
400 α =0.80 β =0.8
αQ
6
300 10
Ke Ke 200
5
Bδ 100
Qm (kN) Σ δ (mm)
10 +
16 S
Bδ2
0 0
0 100 200 300 400 500 600 0 5 10 15 20

Fig. 10. Modelization of the Bauschinger part: (a) typical Bauschinger segment; (b) values of α obtained from the tests; (c) values of β obtained from the tests.

0–1–2–3–4–5–6–7–8–9–10, and that an approximation for the path, the Q –δ curve will initially follow a line of stiffness Ke until
Bauschinger segment 10–11 located in the positive loading domain the point of ordinate α Q1 is reached, after which the Q –δ curve will
is sought. The ordinate at the origin of the segment (i.e. point 10) is follow a line oriented to a point determined by B δ1+ and Q1 . Follow-
Q = 0, and the ordinate at the end (i.e. point 11) is likewise known ing this rule, the complete hysteretic curve under arbitrary loading
because it coincides with the maximum force Qm attained in the can be constructed until failure.
skeleton part of previous deformation cycles in the same domain As for predicting failure of the damping device, the following
of loading (i.e. the ordinate at point 6 referred to as Qm = Q6 in procedure is proposed. In each step i of the loading process, the
Fig. 7(a)). The Bauschinger segment 10–11 is redrawn for clarity in deformation accumulated on the skeleton curve, 6 S δi , and the to-
Fig. 10(a). It can be approximated with the bilinear curve, shown tal energy dissipated by the damping device up to this step, Wi , are
with dashed lines, determined by three parameters: (i) the stiff- monitored. The corresponding ratios of apparent cumulative plas-
ness of the first line; (ii) the displacement B δ of the segment (i.e. tic deformation ep ηi and total cumulative plastic deformation ηi are
B δ2 in Figs. 7(c) and 10(a) for the example of segment 10–11); and
+
computed as follows: ep ηi = 6 S δi /δy and ηi = Wi /(Qy δy ). Finally,
(iii) the ordinate of the intersection point of the two lines, which the point of coordinates (ep ηi , ηi ) is plotted in Fig. 8(c) and failure is
will be expressed hereafter as a fraction α of the maximum force assumed to occur when this point reaches the solid line. Deserving
attained in the skeleton part in previous deformation cycles in the emphasis here is the fact that the proposed criterion for anticipat-
same domain of loading α Qm (i.e. α Q6 for sample segment 10–11). ing failure is based not only on deformation and energy considera-
In this study, it is proposed: (i) to take the first stiffness equal to tions, as other damage indexes proposed in the literature including
the initial elastic stiffness Ke ; (ii) to estimate B δ as P
a fraction β of the well-known Park and Ang model [27]; it also takes into account
the displacement accumulated in the skeleton part S δ up to the the influence of the loading path on the ultimate energy dissipation
start of the Bauschinger segment under consideration (i.e. for the capacity of the damper.
example segment 10–11 B δ2+ = β{|S δ1+ | + |S δ1− | + |S δ2+ | + |S δ2− |}); Fig. 11(a) shows the Q –δ curve predicted with the proposed
and (iii) to determine the value of α so that the area (i.e. the dis- hysteretic model for the test specimen TTD15/20/20, up to failure
sipated energy) under the Bauschinger segment obtained from the anticipated using the procedure suggested above. Superimposed
test and the area under the approximated bilinear curve are equal, on Fig. 11(a) are the experimental results. Good agreement is seen
as indicated by the shaded areas in Fig. 10(a). between the test and the numerical simulation. For this test speci-
Each Bauschinger segment obtained by decomposing the exper- men TTD15/20/20, Fig. 11(b) compares the load path in the ep ηi −ηi
imental Q –δ curves of test series TTD as indicated in Fig. 7 was space obtained from the test and that derived from the numerical
approximated by a bilinear curve as explained above, and the val- simulation. Both paths are quite close, indicating that the proposed
ues of parameters α and β were determined. They are shown in procedure traces the experimental energy consumption path rea-
Fig. 10(b) and (c), respectively. For modeling purposes the follow- sonably well while also anticipating the failure of the damper rea-
ing values are proposed: α = 0.8 and β = 0.8. sonably well.
Once the shape of the skeleton part and the Bauschinger part are
determined by the parameters Qy , QB , δy , kp1 , kp2 , α and β , the hys- 5. Conclusions
teretic curve of the damping device when it is subjected to an arbi-
trary history of displacements can be easily constructed as follows. This paper presents a new type of seismic damper that is
The first time the damper is loaded in a given domain (positive or mounted diagonally within a structural frame as a conventional
negative), the proposed model considers that the Q –δ relationship brace. It is based on yielding the walls of common steel hollow
follows the skeleton curve. Accordingly, with reference to Fig. 7(a), structural sections. The new brace damper consists of a tube-in-
consider for example that first – starting from point 0 – the damper tube assemblage of two hollow sections. In the outer hollow sec-
is deformed an amount S δ1+ in the positive domain; it will follow tion a series of strips are created by cutting a series of slits through
the skeleton curve moving from point 0 to 1. At point 1 the de- the wall. The inner hollow section is inserted into the outer one,
formation accumulated on the skeleton part is 6 S δ = |S δ1+ | and then fixed through welding at several points so that when the brace
the corresponding force attained will be called Q1 . Second, if the damper is subjected to forced displacements in the direction of
damper is unloaded at point 1 the Q –δ curve will follow a line of its axis, the strips dissipate energy through flexural/shear yield-
stiffness Ke (=Qy /δy ) until point 2. Third, now loading the damper ing. The proposed brace damper features advantages over other
in the opposite domain of loading, the Q –δ curve will follow the existing brace-type dampers such as the well-known buckling re-
skeleton curve moving from point 2 to 3. The deformation accumu- strained brace: (i) the yielding part of the damper can be easily
lated on the skeleton part at point 3 is 6 S δ = |S δ1+ | + |S δ1− |. Fourth, inspected; (ii) it is simpler, since unbonding layers and pouring
if the damper is unloaded at point 3 it will again follow a line of concrete are not required; and (iii) its production is not subjected
stiffness Ke until point 4. Fifth, reloading the damper in the posi- to tight fabrication tolerances. As a result of (ii) and (iii), the pro-
tive domain, the Q –δ curve will follow a Bauschinger path whose duction cost can be reduced notably, which makes it viable for mas-
amplitude B δ is B δ1+ = β{|S δ1+ | + |S δ1− |}. Within this Bauschinger sive use, even in developing countries.
A. Benavent-Climent / Engineering Structures 32 (2010) 1113–1122 1121

a b
Model 2.0
Test Q/Q y 1000 η Failure condition line
Failure points 1.5 Test
800 Model
1.0 Failure point
0.5 600
δ /δ y
0.0
-6 0 - 40 -20 0 20 40 60 400
-0.5
-1.0 20 0
-1.5 epη
0
-2.0 0 50 100 150 200 250

Fig. 11. Comparison between test and numerical simulation for specimen TTD15/20/20: (a) Q –δ curves; (b) energy consumption paths until failure.

a b
Q/ Q Q/Q
y kp2 y kp2
x'
τB x' - τB

x k p1 x' - τB
kp1
1 x -1
1 1
x -1 η (x' - τB)/ kp2
s
1 (x -1) /k p1 1
1 s
η
S
δ/δ y δ/δ y
S

ηB ep
ηB
ep
η ep
ep
η

Fig. A.1. Relation between ep η̄ and S η̄ for (a) ep η̄ ≤ η̄ and for (b) ep η̄ > ep η̄B .
ep B ,

To assess the hysteretic behavior and the ultimate energy dissi- established as follows. For ep η̄ ≤ ep η̄B (see Fig. A.1(a)), the relation
pation capacity of the new damper, and in order to compare these between a given ordinate x and ep η̄ is:
features with those exhibited by conventional steel plates with
slits, two series of tests were conducted. The first test series in- (x − 1)
volved four specimens representing the dissipative part of the pro- ep η̄ = − (x − 1), (A.1)
kp1
posed damper, subjected to cyclic loads until failure. In the second
test series, five specimens were tested, each one consisting of an while solving for x, Eq. (A.1) gives:
assemblage of several conventional steel plates with slits, likewise
subjected to cyclic loading. The results of the tests reveal that: (i) kp1ep η̄ + (1 − kp1 )
x= . (A.2)
the proposed damper has very stable cyclic behavior and the shape (1 − kp1 )
of the hysteretic loops is very close to a rectangle; (ii) the increase
of strength due to strain hardening is about 1.25 times the yield The shaded area in Fig. A.1(a) can be expressed by:
strength; and (iii) there are no significant differences between the
proposed brace damper and the conventional steel plates with slits ep η̄(1 + x)
S η̄ = . (A.3)
in terms of hysteretic behavior and ultimate energy dissipation ca- 2
pacity. The proposed brace damper exhibits stable hysteretic re-
Substituting x given by Eq. (A.2) in Eq. (A.3) and operating gives:
sponse up to large ductility levels – from 30 to 60 times the yield
displacement in the tests conducted for this study. Further, a large kp1
plastic deformation capacity can be attained – i.e. brace elonga- for ep η̄ ≤ η̄ :
ep B S η̄ = ep η̄ + ep η̄2 . (A.4)
tions up to about 120 mm—by choosing an appropriate geometry 2(1 − kp1 )
for the steel strips. Based on the results of the tests, a hysteretic Similarly, if ep η̄ > ep η̄B (Fig. A.1(b)), the relation between a given
model and a procedure for predicting the ultimate energy dissipa-
ordinate x0 and ep η̄ is
tion capacity and failure of the new seismic damper are put forth.
(x0 − τB )
Acknowledgements (ep η̄ − ep η̄B ) = − (x0 − τB ), (A.5)
kp2
This research was funded by the Spanish Ministry of Education
and solving for x gives
and Science (projects BIA2005-00591 and BIA2008-00050) and by
the European Union (Fonds Européen de Dévelopment Régional). kp2 (ep η̄ − ep η̄B ) + τB (1 − kp2 )
x0 = . (A.6)
(1 − kp2 )
Appendix
The shaded area in Fig. A.1(b) can be expressed by:
Fig. A.1 shows a typical trilinear skeleton curve in the non-
dimensional space Q /Qy and s δ/δy . The relation between the ab- ep B η̄ (1 + τB ) (ep η̄ − ep η̄B )(τB + x0 )
scissa ep η̄ = S δ/δy and the shaded area which represents S η̄ can be S η̄ = + . (A.7)
2 2
1122 A. Benavent-Climent / Engineering Structures 32 (2010) 1113–1122

Substituting x0 given by Eq. (A.6) in Eq. (A.7) gives: [11] Nakashima M, Iwai S, Iwata M, Takeuchi T, Konomi S, Akazawa T, et al. Energy
dissipation behaviors of shear panels made of low yield steel. Earthq Eng Struct
ep η̄B (1 + τB ) Dyn 1994;23:1299–313.
for ep η̄ > η̄ :
ep B η̄ =
S [12] Iwata M, Kato T, Wada A. Performance evaluation of buckling-restrained
2
braces in damage-controlled structures. In: Mazzolani F, editor. Behavior of
(ep η̄ − ep η̄B ) kp2 (ep η̄ − ep η̄B )
 
steel structures in seismic areas STESSA. 2003. p. 37–43.
+ 2τB + . (A.8) [13] Sabelli R, Mahin S, Chang C. Seismic demands on steel braced frame buildings
2 (1 − kp2 ) with buckling-restrained braces. Eng Struct 2003;25(5):655–66.
[14] Iwata M, Murai M. Buckling-restrained brace using steel mortar planks:
Finally, making x = τB in Eq. (A.2) and solving, the value of ep η̄B is Performance evaluation as a hysteretic damper. Earthq Eng Struct Dyn 2006;
obtained: 35(14):1807–26.
[15] Tremblay R, Bolduc P, Neville R, Devall R. Seismic testing and performance of
(τB − 1)(1 − kp1 )
η̄ =
ep B . (A.9) buckling-restrained bracing systems. Canad J Civil Eng 2006;33(2):183–98.
[16] Morino S, Kawaguchi J, Shimokawa H. Hysteretic behavior of flat-bar braces.
kp1
In: Proceedings of the international conference on advances in steel structures.
Hong Kong: Pergamon Press; 1996. p. 1127–32.
References [17] Iami K, Yasui N, Umezu Y. Development of tube-in-tube type FLD bracing
member (force-limiting device) and its impulsive analysis. In: Proceedings of
the SSRC annual technical session and meeting. 1997 p. 515–33.
[1] Wada A, Huang YH, Iwata M. Passive damping technology for building in Japan.
[18] Iwata M, Kato T, Wada A. Buckling-restrained braces as hysteric dampers.
Prog Struct Eng Mater 2000;2(3):335–50.
In: Mazzolani F, Tremblay R, editors. Behavior of steel structures in seismic
[2] Symans MD, Charney FA, Whittaker AS, Constantinou MC, Kircher CA,
areas STESSA. 2000. p. 33–8.
John MW, et al. Energy dissipation systems for seismic applications: Current
[19] Benavent-Climent A. An energy-based damage model for seismic response of
practice and recent developments. J Struct Eng 2008;134(1):3–21.
[3] Boardman PR, Wood BJ, Carr AJ. Union House–A cross braced structure with steel structures. Earthq Engng Struct Dyn 2007;36:1049–64.
energy dissipaters. Bull New Zealand Nat Soc Earthq Eng 1983;16(2):63–75. [20] Kato B, Akiyama H, Yamanouchi H. Predictable properties of structural steels
[4] Martines-Romero E. Experiences on the use of supplemental energy dissipa- subjected to incremental cyclic loading. In: IABSE symposium on resistance
ters on building structures. Earthq Spectra 1993;9(3):581–625. and ultimate deformability of structures acted on by well defined loads. 1973,
[5] Perry CL, Fierro EA, Sedarat H, Scholl RE. Seismic upgrade in San Francisco using p. 119–24.
energy dissipation devices. Earthq Spectra 1993;9(3):559–79. [21] Manson SS. Behavior of materials under conditions of thermal stress. In: Heat
[6] Bergman DM, Goel SC. Evaluation of cyclic testing of steel plate devices for Transfer Symposium. 1953, p. 9–15.
added damping and stiffness. Report no. UMCE87-10, Ann Arbor (MI, USA), [22] Coffin LF. A study on the effect of cyclic thermal stresses in ductile metals.
The University of Michigan; 1987. Trans Amer Soc Mech Eng 1954;76:931–50.
[7] Tsai K, Chen H, Hong C, Su Y. Design of steel triangular plate energy absorbers [23] Miner MA. Cumulative damage in fatigue. J Appl Mech 1945;12:159–64.
for seismic-resistant construction. Earthq Spectra 1993;9(3):505–28. [24] Xia C, Hanson RD. Influence of ADAS element parameters on building seismic
[8] Kobori T, Miura Y, Fukusawa E, Yamada T, Arita T, Takenake Y. et al. response. J Struct Eng 1992;118(6):1903–18.
Development and application of hysteresis steel dampers. In: Proceedings of [25] de la Llera J, Esguerra C, Almazan JL. Earthquake behavior of structures with
11th world conference on earthquake engineering; 1992. p. 2341–6. cooper energy dissipaters. Earthq Eng Struct Dyn 2004;33(3):329–58.
[9] Chan RWK, Albermani F. Experimental study of steel slit damper for passive [26] Nakashima M, Akazawa T, Tsuji B. Strain-hardening behavior of shear panels
energy dissipation. Eng Struct 2008;30(4):1058–66. made of low-yield steel II Model. J Struct Eng 1995;121(12):1750–7.
[10] Oh SH, Kim YJ, Ryuet HS. Seismic performance of steel structures with slit [27] Park YJ, Ang AHS. Mechanistic seismic damage model for reinforced concrete.
dampers. Eng Struct 2009;31(9):1997–2008. J Struct Eng 1985;111(4):722–39.

You might also like