Engineering Structures 88 (2015) 176–188

Contents lists available at ScienceDirect

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

Experimental study of a self-centering prestressed concrete frame

subassembly
Liang-Long Song, Tong Guo ⇑, Yu Gu, Zhi-Liang Cao
Key Laboratory of Concrete and Prestressed Concrete Structures of the Ministry of Education, Southeast University, Nanjing 210096, PR China

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

Article history: This paper presents the experimental investigation of the behavior of a 0.5-scale, one-bay, one-story
Received 5 May 2014 self-centering (SC) moment resisting concrete frame subassembly under cyclic loading. The SC frame
Revised 20 January 2015 utilizes post-tensioned (PT) beam–column connections and column–foundation connections to provide
Accepted 22 January 2015
the re-centering capabilities after earthquakes. Beam web friction devices are included in the beam–
Available online 14 February 2015
column connections to provide the adjustable energy dissipation capacity, and steel jackets are used at
the beam ends and column bases to avoid concrete crushing at the contact interfaces. The influence of var-
Keywords:
ious parameters (i.e. initial PT forces in the beam and columns, friction forces, etc.) on the seismic behavior
Seismic testing
Self-centering
of the SC frame is experimentally examined, and it is observed that the frame subassembly performed well
Concrete frame up to a 2.5% drift with only minor damage in concrete cover of the beam, and the column remained dam-
Web friction device age free. To avoid the stiffness reduction upon unloading and to achieve desirable energy dissipation, the
Post-tensioning stiffness of the friction devices and connecting bolts should be properly enhanced in future study.
Energy dissipation Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction ability [1–3]. However, one major concern for unbonded post-ten-
sioned PC-MRFs in seismic regions is that their lateral displace-
Reinforced concrete (RC) moment-resisting frames (MRFs) have ments during earthquakes may be larger than acceptable due to
been widely used as the primary lateral load-resisting system of small energy dissipation. Therefore, it is necessary to adopt addi-
buildings during the past decades in seismic areas. Conventional tional energy dissipators to reduce the displacement demands in
RC-MRFs designed under the prevailing capacity design philosophy earthquakes. Early form of dissipators provided energy dissipation
are expected to be a ductile system under moderate to strong through metallic yielding of mild steel bars grouted in ducts across
earthquakes, capable of developing inelastic deformations in the beam-to-column joints (‘‘hybrid’’ connection) [4], which was
well-detailed regions, so as to protect the structure from unfavor- developed in the US-PRESSS Program (PREcast Seismic Structural
able instability or collapse. Recent earthquakes, e.g. the 1994 System) [5]. Subsequent research on ‘‘hybrid’’ precast structural
Northridge earthquake, the 2011 Canterbury earthquake, etc., have systems brought other forms of dissipators, including tension-only
validated the adequacy of such a design philosophy in preventing mild steel devices [6], external mounted mild steel devices [7,8],
structural collapse and loss of lives. However, it is also observed lead-based damping devices [9,10], and friction-based devices on
after these earthquakes that the buildings may experience exten- the top and bottom surfaces of the beam ends [11]. More recently,
sive structural damage and residual deformations, resulting in con- Chou and Chen [12] conducted full-scale cyclic tests on a one-story
siderable direct and indirect losses due to operation disruption and two-bay post-tensioned reinforced concrete steel (RCS) frame sub-
repair. assembly with reduced flange plates and demonstrated the excel-
To reduce the residual deformations and eliminate the need for lent self-centering capacity of the PT frames. Shake table tests [13]
the post-earthquake retrofit, unbonded post-tensioned precast on a 0.4-scale one-story, two-by-two bay post-tensioned concrete-
concrete (PC) MRFs have been developed and investigated. filled tube (CFT) frame subassembly further confirmed the large
Previous study shows that unbonded post-tensioned PC-MRFs deformation and self-centering ability of the PT frame systems.
can have adequate strength, ductility and undergo large nonlinear Along with the above developments in precast concrete (PC) MRFs,
displacements while maintain an excellent self-centering (SC) unbonded post-tensioning (PT) technology has also been applied to
bridge piers [14,15], walls [16], bracing members [17] and beam–
column connections of steel frames [18] to reduce the residual
⇑ Corresponding author. Tel./fax: +86 25 83795054. deformations of seismic resistant systems.
E-mail address: guotong77@gmail.com (T. Guo).

http://dx.doi.org/10.1016/j.engstruct.2015.01.040
0141-0296/Ó 2015 Elsevier Ltd. All rights reserved.
 L.-L. Song et al. / Engineering Structures 88 (2015) 176–188 177

Recently, a new concrete PT connection, i.e. the self-centering plates) was sandwiched between the inner surface of the steel
prestressed concrete (SCPC) beam–column connection, has been channel and the steel jacket. One of the two surfaces of the friction
proposed by Song et al. [19]. The novel connection uses the plate was roughened and adhered to the steel channel using epoxy
unbonded PT strands to provide the self-centering (SC) capacity glue. Four 20-mm diameter high strength bolts passed through the
and bolted web friction devices to dissipate earthquake energy. channels, brass friction plates, and beam and produced normal
In addition, steel jackets are fabricated at the beam ends to prevent force on the friction surfaces using a torque wrench. The diameter
the concrete from damage. Previous tests on the beam–column of the oversized circular bolt ducts in the steel jacket and concrete
connection have shown that, the SCPC connection can not only beam was 50 mm, which was determined considering that the fric-
provide desirable energy dissipation capability but also prevent tion bolts should not bear against the bolt ducts in the beam at a
local concrete failure at the beam–column interfaces, without connection relative rotation of 3.5% rad.
diminishing the self-centering capacity.
In order to further examine the performance of the MRFs with
the proposed SCPC connection, six low cyclic tests were conducted
2.2. Test setup and instrumentation
on a 0.5-scale, one-bay, one-story frame subassembly. Post-ten-
sioned column–foundation connections are also used in the frame
A photograph and an elevation of the test setup are shown in
system to eliminate the damage at the bases of columns and pro-
Fig. 3(a) and (b), respectively. First, the foundations were posi-
vide additional re-centering capacity. Primary design parameters
tioned to the strong floor. The precast columns were then tempo-
(i.e. initial PT forces in the beam and columns, friction forces,
rarily post-tensioned to the foundation. The precast beam was
etc.) were changed in these tests, so as to evaluate their influence
then lifted and the strands were placed in the duct of the beam
on the seismic behavior of the SC frame.
and anchored on the exterior surfaces of columns. After the beams
had been installed, the PT strands in columns were post-tensioned
to the designed stress level. Finally, the web friction devices were
2. Low cyclic tests
bolted to the column and the friction bolts were tensioned to their
design force level using a calibrated torque wrench.
2.1. Test specimens
The lateral load was applied on the east column (EC) through a
MTS servo-hydraulic actuator, at the story height (centerline
To evaluate the static cyclic performance of a self-centering
dimension) of 1933 mm above the top of the foundation. The dis-
concrete moment-resisting frame with PT beam–column and col-
placement of the east column at the loading point was used as
umn–foundation connections, a 0.5-scale frame subassembly was
the target displacement, and the inter-story drift was defined as
constructed and tested at the Key Laboratory of Concrete and Pre-
the lateral displacement at this loading point divided by the story
stressed Concrete Structures of the Ministry of Education at South-
height.
east University, China. The test frame was designed according to
The force and displacement of the actuator were recorded auto-
the first story of a prototype self-centering concrete moment
matically by the loading system. Vibration string force transducers
resisting frame designed by Guo and Song [20]. Fig. 1 shows the
were installed on the top of columns and outside surface of the
reinforcements and dimensions of the frame subassembly. The col-
beam–column connection to measure the PT forces in the columns
umn had a height of 2300 mm and a cross section of
and beam. The effective length of the PT strands in the beam and
350 mm  350 mm. The clear span of the beam was 2650 mm,
columns were 3.56 m and 2.86 m, respectively. As shown in
and the beam was 200 mm wide by 350 mm deep. The beam and
Fig. 3(c), the relative rotation of the PT beam–column connection
columns were expected to experience no damage in the tests and
was obtained from two displacement transducers located at the
were repeatedly used. A tendon duct with a diameter of 70 mm
top and bottom of the beam at the beam–column interface and
was cast into the beam along its longitudinal axis and two tendon
the distance between the two transducers. Similarly, two displace-
ducts with a diameter of 40 mm were cast in each column, as
ment transducers were placed on the east and west sides of the
shown in Fig. 1(b) and (c). The beam was post-tensioned to the col-
column at the column–foundation interface to measure the PT col-
umns using two 15.2-mm diameter high strength PT strands,
umn–foundation connection rotation. Resistance strain gauges
which passed along the mid-depth of the beam and were anchored
were adhered long the midline of the steel jackets to measure
outside of the exterior columns. The columns were connected to
the strains of steel jackets in the beam and columns, as shown in
the foundations using two 15.2-mm diameter strands through
Fig. 3(c). These strain gauges had a spacing of 3 cm, and were
posttensioning. To protect the concrete from damage during rock-
labeled as Gauge 1 to Gauge 7 (from the beam–column or col-
ing, the ends of the precast beam and columns were armored with
umn–foundation interfaces toward the far end of the jacket).
steel jackets and steel plates were embedded in the concrete col-
umn and foundations, as shown in Fig. 1(a). The thickness of the
embedded steel plates and steel jacket was 8 mm. Two 8-mm thick
shim plates were welded on the steel plate embedded in the col- 2.3. Material properties
umn, as shown in Fig. 2(a), to ensure good contact between the
beam and column [19]. Similarly, two shim plates were used at The mean cubic compressive strength of the concrete was
the column–foundation interfaces to maintain good contact 35 MPa. The nominal diameter of the tendons was 15.2 mm, corre-
between the column and foundation. Due to the existence of steel sponding to a cross-sectional area of 140 mm2. The tendons had a
jackets, no shear reinforcements were used in the jacketed modulus of elasticity (Est) of 1.95  105 MPa and the measured
segments of the beam and columns. As shown in Fig. 2, web fric- yield strength was 1730 MPa (with the yield strain of 1%). Accord-
tion devices were installed at the beam ends to increase the energy ing to the typical double shear test results, the coefficient of fric-
dissipation capacity of the frame system. The web fiction device tion between the brass and steel was around 0.33 [19]. The steel
consisted of a steel channel and a steel connection plate and was plates were Q235 steel with nominal yield strength of 235 MPa.
bolted through the connection plate to the column using two 18- The longitudinal reinforcing bars were HRB335 steel with the nom-
mm diameter high strength bolts, as shown in Fig. 2(a). Four bolt inal yield strength of 335 MPa. The transverse reinforcement and
ducts, running perpendicular to the beam axis, were built in the cross ties were HPB235 steel with the nominal yield strength of
beam, as shown in Fig. 2(b). A friction plate (i.e., 2-mm thick brass 235 MPa [21].
178 L.-L. Song et al. / Engineering Structures 88 (2015) 176–188

PT φ50 Duct φ70 Duct

Strands

Steel Steel Jacket

Plate

φ40 Duct

Steel Steel Jacket

Plate
φ80 Duct

(a) Profile

(b) Section 1 (c) Section 2 (d) Section 3

(e) Section 4 (f) Section 5

Fig. 1. Configuration of the frame without web friction devices (dimensions in millimeters).

2.4. Test matrix the hysteretic behavior of the frame specimen with different lev-
els of PT forces in beam and columns. Higher level of friction
Six tests were conducted on the frame subassembly to inves- bolt forces was used in Test 5 to evaluate the influence of fric-
tigate the influence of various parameters on the hysteretic tion bolt force on the hysteretic behavior. Test 6 was intended
behavior. Various initial PT forces in the beam and columns to study the PT force versus drift relationships of the PT
and the normal friction bolt forces (N0) were considered in the beam–column and column–foundation connections. Note that
tests, as shown in Table 1, where T0 denotes the sum of the the PT strands in the beam and columns were designed to
PT forces in the beam or columns, R0 is the initial stress of each remain elastic during the tests in order to be reused in multiple
tendon normalized by the yield stress of the tendon (i.e., tests.
1730 MPa), and N0 denotes the total clamping force of the four In this study, the upper bound of the initial PT force in the beam
friction bolts. Test 1 was conducted without any web friction was determined from the prototype self-centering frame designed
devices to ensure that the bare PT frame exhibited the desired by Guo and Song [20] based on a force scaling ratio of 0.25, and
nonlinear but elastic response. Tests 2, 3, and 4 investigated lower PT beam forces were used for comparison.
 L.-L. Song et al. / Engineering Structures 88 (2015) 176–188 179

Shim
Plate Steel Jacket

φ18 Bolt
φ20 Bolt Steel jacket
Tendon duct

Channel
φ20 Duct

Steel
Plate
Steel channel

Friction bolt
Steel Jacket ducts Connection
plate
Steel Top surface
Plate Shim of foundation Brass plate
Plate

(a) PT connection detail (b) Photographs of steel jacket and steel channel

φ25 Duct

φ20 Duct

Connection
Plate

(c) Configuration of steel channel

Fig. 2. Configuration of PT connections with web friction device (dimensions in millimeters).

The friction forces were determined according to the design To ensure that the web friction devices do not move relative to
energy dissipation ratio, bE, for the PT beam–column connections. the column during gap opening of the PT beam–column connec-
According to Song et al. [19], bE = MFf/MIGO, where MIGO corresponds tions, the design values of the connecting bolt forces were set to
to the imminent gap opening moment of the PT beam–column be larger than the design friction forces on the web friction devices.
connection and is satisfied by MPT and MFf, representing the It should be noted that, due to the inadequate screw strength and
moment resistance of PT forces and friction forces, respectively. diameter of the connection bolts used in the tests, the pre-stresses
The design values of bE for the PT beam–column connection of applied to the bolts are lower than expected, resulting in that the
the test specimens are given in Table 1 and these values fall well connecting bolt forces in Test 5 are smaller than the friction forces
within the ranges recommended by Song et al. [19]. For each test and small gaps were observed between the column and the
specimen, the normal friction bolt forces N0 was determined as connection plate of the friction device (as will be discussed subse-
N0 = Ff/2l, where Ff = design friction force, and l = coefficient of quently in Section 3.1).
friction, assumed to be 0.33 from the tests by Song et al. [19]. It The frame subassembly was tested using the displacement-
should be noted that, in order to ensure the self-centering capacity based loading protocol. Gradually increasing reversed cyclic dis-
of the PT beam–column connection, the low bound of initial PT placements were applied to the specimen. The displacement cycles
force in the beam should be larger than the maximum friction force had drift amplitudes of ±0.375%, ±0.5%, ±0.75%, ±1.0%, ±1.5%, ±2.0%
on the web friction devices. and ±2.5%, as shown in Fig. 4. Note that the negative displacement
To investigate the influence of PT forces on the performance of or drift corresponds to the actuator pushing from the east to the
SC frames, two different levels of initial PT forces were adopted for west.
comparison. A PT force equivalent to about 8% of the axial strength
of the column was used (Tests 3–6) as the upper bound of the ini-
tial PT force in the column, and a lower PT column force (Tests 1 3. Results and discussion
and 2) was used for comparison. It is worth noting that the maxi-
mum axial load ratio of the column in the test specimen was sim- 3.1. General performance and hysteretic behavior
ilar to that used in the cyclic tests of a post-tensioned RCS
(reinforced concrete steel) frame subassembly conducted by Chou Fig. 5 shows the force versus drift responses of the Tests 1–5, in
and Chen [12] (equivalent to about 9% of the axial strength of the which Test 1 was conducted with no friction forces applied to eval-
column). uate the hysteretic response without energy dissipating devices. As
180 L.-L. Song et al. / Engineering Structures 88 (2015) 176–188

West East
SC Frame

Actuator

(a) Photograph of the test frame (imaged by the authors)

Negative Positive
Direction Direction
PT Strands

Actuator
Beam

West Column East Column

Reaction Wall

Foundation

(b) Experimental setup

Force
Transducer
Strain Gauge Beam
1 2 3 45 6 7

Strain Gauge
Strain
1 2 3 4 5 6 7 Gauge
Column Displacement Transducer

Strain Gauge Top Surface of

Top surface plate of steel West surface plate of steel
Foundation
jacket at the beam end jacket at the column end

(c) Typical locations of strain gauges and transducers

Fig. 3. Test setup (dimensions in millimeters).

shown in Fig. 5(a), the specimen responded approximately linearly beam–column connection. The small amount of energy dissipation
elastic until gap opening occurs in the PT beam–column and was attributed to the inherent friction between the assembled
column–base connections, which lead to reduction in the lateral members of the frame.
stiffness of system. To be more specific, gap opening at the PT The influence of web friction devices on the energy dissipation
beam–column interface was observed near 0.3% drift. The speci- is demonstrated through Test 2, in which PT forces in the beam and
men maintained its initial stiffness up to drift ratio of 2.5% and columns were similar to those in Test 1, while a friction bolt force
no degradation in the capacity occurred with the increasing drift of 25 kN was applied in each bolt. As shown in Fig. 5(a), the frame
demands. Overall, the frame without friction devices showed a specimen in Test 2 had a flag-shaped hysteretic response and
bilinear elastic response that is typical of a post-tensioned rocking showed desirable self-centering and relatively good energy
 L.-L. Song et al. / Engineering Structures 88 (2015) 176–188 181

Table 1 the test specimen and the analytical model. Note that the real com-
Test matrix. pression stiffness of beam–column and column–foundation con-
Test number Beam Column N0 (kN) bE tact region is smaller than the theoretical values, as reported in
T0 (kN) R0 T0 (kN) R0 [23].
In this study, the influence of gap opening behavior at the col-
1 171 0.33 167 0.32 / /
2 177 0.34 167 0.32 100 0.27
umn–foundation interfaces on the lateral force versus drift
3 188 0.36 261 0.51 100 0.26 response was demonstrated by comparing the hysteretic responses
4 243 0.47 272 0.53 100 0.21 in Tests 2 and 3 for example, where different initial PT column
5 182 0.35 256 0.50 200 0.42 forces and the same friction bolt forces were used. Fig. 5(c) pre-
6 176 0.34 259 0.50 100 0.27
sents the lateral force versus drift response for the test specimen
in Test 3. Compared with the specimen in Test 2, the initial PT col-
umn forces in Test 3 was increased by 56.3% to 261 kN. As a result,
3 the initial stiffness and lateral force capacity of the specimen in
Test 3 was increased to 16,075 kN/m and 152 kN, which were
2 57.7% and 15.2% larger than those of the specimen in Test 2. How-
ever, according to the lateral force versus drift responses, the post-
1 gap-opening stiffness of the specimens in the two tests were sim-
ilar, due to the same amount of PT tendons used in the two tests.
Drift (%)

0 The shape of the hysteresis loops and the area enclosed by hyster-
esis loops of specimen in Test 2 were almost the same as those in
-1 Test 3, since the two tests had the same friction bolt forces. Similar
observation can be made by comparing the hysteretic responses in
-2 Tests 3 and 4, which had the same friction bolt forces and different
PT forces in the beam. It can be found that the two tests had similar
-3 energy dissipation capacity while Test 4 had larger initial stiffness
0 1000 2000 3000
and lateral force capacity due to the higher PT forces in the beam.
Time (s) Therefore, the PT forces can affect the initial stiffness and the lat-
Fig. 4. Loading protocol.
eral force capacity of the frame system.
The influence of friction bolt forces on the hysteretic behavior
was further evaluated in Test 5, which had higher friction forces
than those in the previous tests. As shown in Fig. 5(e), the spec-
dissipation characteristics, as compared with the specimen in Test imen in Test 5 provided the largest hysteretic energy dissipation
1. The initial stiffness was 10,195 kN/m. At about 0.5% drift, the capacity. Compared with specimen in Test 3, which had the same
stiffness of the frame decreased to 1144 kN/m, due to gap opening PT forces and low friction bolt forces, the specimen in Test 5 had
of the PT beam–column and column–foundation connections. larger energy dissipation capacity and similar initial stiffness.
When the load reached zero, the beam–column and column–foun- However, it is noted that the specimen in Test 5 sustained more
dation interfaces closed under the PT forces. During load reversal, a stiffness reductions upon unloading, especially during the
stiffness reduction phenomenon was observed (labeled as ‘‘Stiff- unloading process of the +2.5% drift cycle, which is undesirable
ness reduction’’ in Fig. 5(a)), which can be attributed to the flexibil- and was more severe than that in the previous tests on the
ity of the steel channels [19]. Comparing the hysteretic responses beam–column connections [19]. This phenomenon can be attrib-
in Tests 1 and 2, it can be found that the friction devices increased uted to the inadequate stiffness of the connecting elements
both the lateral force and energy dissipation capacities of the frame which bolted the channels to the columns. With the friction bolt
system. No damage was observed in the frame specimen after the forces increased, small gaps were observed between the column
test. and the connection plate of the friction devices, as shown in
Two analytical bending force diagrams, comparing the test Fig. 5(f), and the bolt forces of the connecting bolts experienced
without any friction device (i.e. Test 1) and the test which included some losses during the cyclic loading. To mitigate this effect, an
them (i.e. Test 2 for example), are shown in Fig. 5(b). These analyt- 8-mm thick steel plate was added to increase the stiffness of
ical hysteretic responses were obtained by conducting elastic finite the connection plate in the test; however, the gap could not be
element analyses using the finite element program OpenSees [22]. avoided under the high friction, which was due to the inadequate
The numerical model consists of elastic beam–column elements to diameter of the connection bolts (i.e. 18 mm) and limited pre-
model the beam and columns, pairs of zeroLength compression- stresses that can be applied to these bolts through the torque
only elements to model the gap opening/closing behavior at the wrench. Unfortunately, due to the limitation of the diameter of
rocking interface, elastic truss elements for the unbonded PT ten- the embedded bolt ducts (i.e. 20 mm), no larger bolts could be
dons and zeroLengthSection elements for the friction forces. Note used in these tests.
that a more complicated rocking connection model, which consid- In addition, at around 1.5% drift, some minor spalling of the
ers the material nonlinearity, tendon yield and rapture, flexibility cover concrete developed at the transition region of the steel jacket
of panel zone, and the contact region flexibility at the rocking and beam (see Fig. 6) in Test 5, which led to some small strength
interfaces can be found in Song et al. [19]. According to Fig. 5(b), degradation. It is worth noting that the concrete spalling was never
the test results in Fig. 5(a) in general agree with the analytical ones observed in the previous tests of SCPC beam–column connection
in Fig. 5(b). The existence of friction forces increases the lateral specimens [19]. In the previous beam–column connection tests,
force and energy dissipation capacities of the specimen, while it the concrete in the steel jackets was cast with the open surface
does not affect the lateral stiffness of the specimen. However, of the jackets upward, which can ensure better contact between
though the tests results in general agree with the analytical ones, the concrete and the steel jackets. In addition, due to the scaling
the initial stiffness of the analytical model is greater than that of of the steel jacket from the prototype structure, the length of the
the test specimen, which can be mainly attributed to the difference jacket may be insufficient to spread the contact forces to the con-
in the contact region flexibility at the rocking interface between crete without causing any damage.
182 L.-L. Song et al. / Engineering Structures 88 (2015) 176–188

Displacement (m) Displacement (m)

-0.06 -0.04 -0.02 0.00 0.02 0.04 0.06 -0.06 -0.04 -0.02 0.00 0.02 0.04 0.0 6
200 200
Test 1
Beam: T0=171 kN, R0=0.33 Test 1
Column: T0=167 kN, R 0=0.32 Test 2
100 Test 2 100

Lateral Force (kN)

Lateral Force (kN)

Beam: T0=177 kN, R0=0.34

Column: T0=167 kN, R 0=0.32
N0=100 kN
0 0
Stiffness reduction

-100 -100
k1=10195kN/m
k2=1144kN/m
-200 -200
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3

Drift (%) Drift (%)

(a) Test 1 & Test 2 (Experimental) (b) Test 1 & Test 2 (Analytical)

Displacement (m) Displacement (m)

-0.06 -0.04 -0.02 0.00 0.02 0.04 0.06 -0.06 -0.04 -0.02 0.00 0.02 0.04 0.06
200 200

Beam: T0=188 kN, R0=0.36 Beam: T0=243 kN, R 0=0.47

Column: T0=261 kN, R0=0.51 Column: T0=272 kN, R 0=0.53
100 100
Lateral Force (kN)

Lateral Force (kN)

N0=100 kN N0=100 kN

0 0

-100 -100
k1=16075kN/m k1=24103kN/m

k2=1162kN/m k2=1076kN/m
-200 -200
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3
Drift (%) Drift (%)
(c) Test 3 (Experimental) (d) Test 4 (Experimental)

Displacement (m)
-0.06 -0.04 -0.02 0.00 0.02 0.04 0.06
200
Gap opening at
Beam: T0=182 kN, R0=0.35 the channel-
Column: T0=256 kN, R0=0.50 column interface
100
Lateral Force (kN)

N0=200 kN

0
Added
plate
-100
k1=16075kN/m
Bolt
k2=1162kN/m
-200
-3 -2 -1 0 1 2 3
Drift (%) (f) Gap opening at the channel-column interface
(e) Test 5 (Experimental)
Fig. 5. Lateral force versus drift relationships.

3.2. Hysteretic energy dissipation Both cumulative energy dissipation and the energy dissipated per
cycle are calculated and plotted in Fig. 7. The cumulative energy
The energy dissipation in a given cycle was quantified by inte- dissipated is calculated by summing up the energy dissipated in
grating the area enclosed inside the hysteresis loop of that cycle. the consecutive cycles throughout the test. Fig. 7(a) shows that
 L.-L. Song et al. / Engineering Structures 88 (2015) 176–188 183

(a) West end of the beam (b) East end of the beam
Fig. 6. Damage on the beam after Test 5.

16 6
Cumulative Energy Dissipated (kN.m)

Test 1 Test 1
Test 2 Test 2
Energy Dissipated (kN.m)

12 Test 3 Test 3
Test 4 4 Test 4
Test 5 Test 5
8

2
4

0 0
0 1 2 3 0 1 2 3
Drift (%) Drift (%)
(a) Cumulative dissipated energy (b) Energy dissipated per cycle
Fig. 7. Energy dissipation in the tests.

the total energy dissipated by the bare PT frame through the intrin- when the rotation was small. However, as the rotation became lar-
sic system friction was 3.35 kN m. The specimen in Test 5 dissi- ger, the tensile forces in connection bolts became larger and grad-
pated much more cumulative energy (i.e. 14.0 kN m), as ually exceeded the friction forces in the friction devices. Thereafter,
compared with the bare PT frame specimen. The total energy dis- better friction and energy dissipation was realized at the channel-
sipated in Tests 2, 3 and 4 was similar and was about 30% less than jacket interfaces.
that in Test 5.
Fig. 7(b) shows that the specimen with low friction forces (Tests 3.3. Damping ratio
2, 3 and 4) dissipated a more stable amount of energy dissipation
per cycle than the specimen with high friction forces (Test 5). To further analyze the influence of friction devices on the seis-
The energy dissipated per cycle in Test 2, 3 and 4 increased approx- mic performance of the proposed system, the equivalent viscous
imately linearly with the drift. For the specimen in Test 5, it is damping ratio, feff, is calculated for each drift cycle. The estimation
observed that in the 1.0% drift cycle, the dissipated energy is of the equivalent viscous damping is an important step in the dis-
266% greater than the energy dissipated in the 0.5% drift cycle, placement-based design, in which the nonlinear hysteretic system
whereas the increment becomes 685%, 1225% and 1667% in the is represented by an equivalent elastic system with equivalent
drift cycles of 1.5%, 2.0% and 2.5%, respectively. Note that the gap stiffness and equivalent viscous damping. In this study, the equiv-
opening of the beam-to-column interfaces and energy dissipation alent viscous damping ratio is calculated using the method
of the system were noticed at a drift of about 0.5%. The slightly detailed in Filiatrault et al. [24], as shown in Eq. (1),
nonlinear energy dissipation versus drift relationship in Test 5
ED
may also be attributed to the flexibility the connection elements feff ¼ ð1Þ
of the friction devices. Under large friction forces, synchronous 2pK s u2m
rotation of the steel channels and jacket toward the column where ED is the total energy dissipated per cycle of stabilized force–
occurred, resulting in some gaps between the connection plate of displacement response; Ks is the secant stiffness; and um is the max-
the web friction device and the column, and the friction between imum displacement value of the cycle. For a given cycle, Ks and um
the steel jackets and web friction devices may not be fully realized are defined as
184 L.-L. Song et al. / Engineering Structures 88 (2015) 176–188

F þm  F m 15
Ks ¼ ð2Þ
uþm  um Test 1
Test 2
Test 3
uþm  um Test 4
um ¼ ð3Þ

Damping Ratio (%)

2 Test 5
10
where F þ 
m and F m are the positive and negative maximum forces,
respectively; and uþ 
m and um are the positive and negative maximum
displacements, respectively, as shown in Fig. 8.
Fig. 9 shows the equivalent viscous damping ratio versus drift
5
relationships in the tests. For the three tests with the same level
of friction forces (Tests 2, 3 and 4), the feff values tended to slightly
decease with the amplitude of the drift cycles and were around
7.5% at a drift ratio of 2.5%. It is observed that when the bolt fric- 0.5 1.0 1.5 2.0 2.5 3.0
tion forces remained constant, increasing the amount of initial PT Drift (%)
forces in the beam or columns reduced the damping in the frame
system. However, the effect of PT forces on the equivalent viscous Fig. 9. Damping ratios in the tests.
damping ratio was relatively small. For the test with larger friction
force (Test 5), the feff values slightly exceeded 7.5% at drift ratios
Fig. 10 provides a comparison of the RSE versus drift relation-
smaller than 0.75%. For drift ratios exceeding 0.75%, the feff values
ships of the tests. It is shown that the RSE values increased with
for Test 5 increased approximately linearly with the amplitude of
the amplitude of the drift cycles for all the tests. The test with no
the drift cycles and had a value of 11% at a drift ratio of 2.0%. This
friction devices had the largest RSE values for all the drift levels
increase was primarily due to the nonlinearly increasing energy
and thus had the best re-centering performance. In general, the
dissipation with respect to drift, as shown in Fig. 7(b). Beyond that
frame specimen in Test 5, which had the largest friction forces,
drift ratio, due to the more significant stiffness reductions upon
experienced the smallest RSE values. This can be explained that
unloading (see Fig. 5(e)), the feff value ceased to increase and
the application of friction devices to the frame can introduce the
appeared to be constant. For the test with no friction devices (Test
locked-in moment in the friction devices due to friction after the
1), the feff value exponentially decreased with the amplitude of the
unloading, resulting in larger residual deformation. The specimens
drift cycles until a value of approximately 3% was reached at a drift
in the Tests 2 and 3 had nearly identical RSE values, indicating that
ratio of 2.5%.
increasing the PT forces in the columns had little influence on the
re-centering capacity. However, increasing the PT forces in the
3.4. Recentering capabilities beam improved the re-centering performance, which can be
explained that increasing the PT forces in the beam can reduce
The self-centering capabilities were quantified through the rel- the effect of the lock-in moment in the friction devices. The spec-
ative self-centering efficiency (RSE) [25], which measures the imen in Test 4 had larger RSE values than the specimens in Tests
capacity of a system to re-center itself after cyclic loading. The 2 and 3 for drift ratios smaller than 1.5%. For drift ratios exceeding
RSE is defined as 1.5%, all the three tests (Tests 2, 3 and 4) had approximately the
same RSE values.
uþres  ures
RSE ¼ 1  ð4Þ In addition to the above RSE ratios, which provide an evaluation
uþm  um
on the relative amounts of residual displacement of the tests under
where uþ  both negative and positive loading, the residual drift of the frame
res and ures are the positive and negative residual displace-
ments, respectively, as shown in Fig. 8. The residual displacement after the 2.5% drift cycle are also selected for comparison, as shown
is defined as the lateral displacement at the zero lateral force at in Table 2. It is of interest to note that ATC 2009 [26] allows an ini-
the end of each loading cycle. A RSE value of one corresponds to per- tial story drift of 0.2% as the out-of-plumb limits in construction for
fectly self-centering systems (i.e., systems with zero residual defor- new buildings. Some researchers used this value as the residual
mations), while a RSE value of zero corresponds to systems without drift limits for self-centering structural systems [27]. It can be seen
any re-centering capacity (i.e., the Coulomb friction systems). from Table 2 that, expect for the Test 5, the residual drifts for all the

200 1.00
Relative self-centering efficiency

+
Fm
100 0.95
Lateral Force (kN)

K
S

- -
um ures
0 + 0.90
+ um Test 1
ures
Test 2
-100 Test 3
0.85
Test 4
- Test 5
Fm
-200 0.80
-0.06 -0.04 -0.02 0.00 0.02 0.04 0.06 0.0 0.5 1.0 1.5 2.0 2.5
Lateral Displacement (m) Drift (%)

Fig. 8. Hysteretic behavior in Test 5 at 2.5% drift cycle. Fig. 10. RSE versus drift relationships.
 L.-L. Song et al. / Engineering Structures 88 (2015) 176–188 185

Table 2 2.0
Residual drifts at the end of tests.

Test number Residual drift ratio (%)

Normalized PT Force
Negative drift Positive drift Column (East)
Beam
1 0.11 0.06
1.5
2 0.15 0.07
3 0.16 0.06
4 0.11 0.12
5 0.13 0.22

1.0

other tests were less than 0.2%. The residual drift for the Test 5 in
the positive direction was only slightly larger than 0.2% after the
-3 -2 -1 0 1 2 3
2.5% drift cycle; however, the residual drift in the negative direc-
tion of loading was much less than 0.2%. Drift (%)
Note that in Fig. 5, the residual displacements in the five tests
Fig. 11. PT force versus drift relationships of the beam and the east column in Test
after the first cycle (i.e., 0.375% drift cycle) are 0.068%, 0.095%, 6.
0.099%, 0.027% and 0.037% rad, respectively. Larger initial PT forces
in columns and beams could not significantly reduce these residual
displacements, and these displacements may be due to the differ- be explained that, due to the ‘‘expansion’’ of the SC frame [12] dur-
ence in contact tightness on interfaces between the assembled ing gap opening of the PT connections and the actuator loading
members. Higher manufacture precision and better assembly con- applied on the east column, the displacement of the west column
trol might reduce this difference, so as to mitigate the residual dis- (D1 ) was larger than that of the east column (D2 ) during the nega-
placements after the first cycle. tive drift loading, while was smaller than that of the east column
during the positive drift loading, as shown in Fig. 13. It should be
noted that the above phenomenon can introduce some asymmetry
3.5. PT forces and gap opening to the hysteresis behaviors of the test specimens, as shown in
Fig. 5, where the negative maximum lateral forces during each drift
Test 6 was conducted to investigate the cyclic response of the cycle were larger than the positive ones.
PT strands and the loss in PT forces during the tests. According to
the measurements on PT forces prior and after the tests, the loss 3.6. Strain measurements on steel jacket
in PT forces during the tests is negligible. Fig. 11 shows the normal-
ized PT force versus drift relationships of the PT stands in the beam To investigate the performance of the steel jackets and quantify
and east column (EC). The PT forces were measured directly using the compression strain concentration area on the steel jackets dur-
the load cells, and were normalized by the initial post-tensioning ing the gap opening of the PT connections, seven strain gauges
forces. It can be seen that when the story drifts were close to zero, were mounted on the steel jacket to measure the strain along the
the PT forces increased slowly. This is because there were elastic longitudinal axis of the member, as shown in Fig. 3(c). Fig. 14
deformations of the frame prior to the gap-opening occurred at shows the maximum compressive strain during each drift cycle
the beam–column and column–foundation interfaces, and these in Test 5. It is observed that the strain increments for the Gauge
elastic deformations did not result in significant increase in the 1 on the east column were significant (but did not increase linearly
PT forces. After the gap-opening occurred, PT forces increased lin- with the drift increment) with the increasing drifts; however, the
early with increased story drifts. It is also observed that PT forces in strain increments were not very large for Gauges 2–7. Therefore,
the beam increased faster than those in the columns. Note that the the local compression strains in the steel jacket of east column
PT strands in the beam run through two PT beam–column connec- were most significant within a distance of 65 mm away from the
tions, while the PT strands in the column run through only one PT column–foundation interface. Fig. 14(b) shows that the maximum
connection. This resulted in larger gap opening displacements and compressive strains of the gauges on the beam were much less
strain increment of the strands in the beam at a given story drift. than those of the gauges on the column. In addition, the stress dis-
Slim hysteresis loops were observed in the PT force versus drift tribution in the steel jacket of the beam was more uniform, as com-
relationships of the beam, which means that the PT forces at the pared with that in the steel jacket of the column. It can be seen that
same drift were different during loading and unloading. This result the largest compressive strains in the steel jacket at the west end of
may be due to the varied directions of friction forces in the friction the beam were within a distance of 95 mm away from the beam–
device during loading and unloading. column interface. The difference in the strain distributions on the
The gap opening behaviors of the beam–column and founda- steel jackets in the beam and column may be attributed to the fol-
tion–column connections were investigated through the results lowing reasons: (1) the existence of friction devices at the beam
of Test 5. Fig. 12 shows the gap-opening angles at the beam–col- end; (2) worse concrete casting property in the beam jacket than
umn and column–foundation interfaces in Test 5, where the gap- that in the column jacket; and (3) different contact tightness at
opening angles were calculated from the relative displacements the beam–column interface and the column–foundation interfaces.
between the two displacement transducers at each PT connection Note that there are four shim plates at the two beam ends, as
divided by the distance between the transducers. It is observed shown in Fig. 2(a); the beam is fixed if any three shim plates are
that when the frame was subject to negative drifts, the gap-open- in contact with the beam during the assembly. The other shim
ing angles at the west column (WC)–foundation connection were plate may be not in good contact with the jacket, resulting in more
larger than those at the east column (EC). During positive drift complex strain distributions on the jacket during testing. For the
loading, the gap-opening angles at west column–foundation con- column, however, there are two shim plates at each column base,
nection were smaller than those at the east column, and gap-open- and both are in good contact during the assembling.
ing angles of the beam–column connection at the east end of beam Fig. 15 shows the strain versus drift relationships of the gauges
were in general larger than those at the west end of beam. This can that experienced the largest compressive strains in Test 5 (i.e.
186 L.-L. Song et al. / Engineering Structures 88 (2015) 176–188

3.0 3.0
East Column East Column
2.5 West Column 2.5 West Column

Gap Opening Angle (%)

Gap Opening Angle (%)

Beam East End Beam East End
Beam West End Beam West End
2.0 2.0

1.5 1.5

1.0 1.0

0.5 0.5

0.0 0.0
-3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0
Drift (%) Drift (%)
(a) Subject to negative drifts (b) Subject to positive drifts
Fig. 12. Gap-opening angles at the beam–column and column–foundation connections.

1 2 1 2

g1 Negative drift Positive drift g2

Gap displacement g2 Gap displacement g1

West Column East Column West Column East Column

(a) Subject to negative drifts (b) Subject to positive drifts

Fig. 13. Deformed configurations of the tested SC frame.

Serial number of strain gauges Serial number of strain gauges

1 2 3 4 5 6 7 1 2 3 4 5 6 7
0 0

-100
-200

-200
-400
Strain (με)
Strain (με)

Drift = 0.375% -300 Drift = 0.375%

Drift = 0.50% Drift = 0.50%
-600
Drift = 0.75% -400 Drift = 0.75%
Drift = 1% Drift = 1%
Drift = 1.5% Drift = 1.5%
-800 -500
Drift = 2% Drift = 2%
Drift = 2.5% Drift = 2.5%
-1000 -600
0 50 100 150 200 250 0 50 100 150 200 250

Distance to the contact surface Distance to the contact surface

(a) Steel jacket of the east column (b) Steel jacket at the west end of the beam
Fig. 14. Maximum compressive strains in the steel jackets during each drift cycle.

Gauge 1 on the east column and Gauge 3 on the beam). A flat line is with the story drifts and slim hysteresis loops were observed
observed (i.e. no significant changes in strains with respect to during the unloading processes. Fig. 15(b) shows the hysteresis
drifts) in Fig. 15(a) in the positive drift cycles, corresponding to loops of the Gauge 3 on the beam, where a double-flag shape is
the decompression of the west side of the column end. During observed. Tensile strains occurred on the steel jacket during the
the negative loading, the strains increased approximately linearly negative direction loading, which was mainly due to the horizontal
 L.-L. Song et al. / Engineering Structures 88 (2015) 176–188 187

100
0
0
-200

Strain (με)
-100
Strain (με)
-400

-600 -200

-800 -300

-1000 -400
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3
Drift (%) Drift (%)
(a) Gauge 1 on the east column (b) Gauge 3 on the beam
Fig. 15. Strain versus drift response for the gauges in the beam and column.

component of friction forces on the web friction devices when gap (5) A maximum equivalent viscous damping ratio of 11% was
opening occurred at the PT connections. reached during the tests for the SC frame subassembly with
web friction devices. Experimental results showed that the
4. Summary and conclusions effectiveness of the friction devices was influenced by the
stiffness of the friction devices and connecting bolts, and
This paper presents the quasi-static cyclic experiments on a 0.5- better energy dissipation may be obtained from a more
scale self-centering (SC) moment resisting concrete frame subas- conservative design of these elements.
sembly. Based on the results of this study, the following findings (6) In addition, the beams and columns should be accurately
can be drawn: manufactured and carefully assembled, so as to ensure the
accurate location of the members and to provide tight and
(1) The proposed post-tensioned (PT) self-centering frame uniform contact between the beam and columns (or
system, which incorporates the PT beam–column and col- between the column and foundations), which might influ-
umn–foundation connections, demonstrated the adjustable ence the intrinsic residual deformations of the frame after
energy dissipation and desirable self-centering capacities. the testing.
The frame subassembly performed well up to drift ratio of
2.5%. Although some minor spalling of concrete cover was
observed at the transition region of steel jacket and concrete Acknowledgments
beam, the main part of the beam and the columns remained
damage free. The support from: (1) the National Natural Science Foundation
(2) It is observed that the energy dissipation capacity of the self- of China under Grants No. 51078075 and 51378107; (2) Scholar-
centering frame system was directly influenced by the ship Award for Excellent Doctoral Student of Ministry of Education
magnitude of the friction bolt forces; while the magnitude of China; and (3) Excellent Doctoral Dissertation Foundation of
of the initial PT forces affects the self-centering behavior, lat- Southeast University is gratefully acknowledged.
eral force resistance and initial stiffness of the system. In
addition, the self-centering capacity is much more sensitive
References
to the initial PT forces in the beam than those in the
columns. [1] Priestley MJN, MacRae G. Seismic tests of precast beam-to-column joint
(3) The steel jacket not only works as one part of the friction subassemblages with unbonded tendons. PCI J 1996;41(1):64–81.
device, but also improves the performance of the concrete, [2] Cheok G, Lew H. Model precast concrete beam-to-column joints subject to
cyclic loading. PCI J 1993;38(4):80–92.
by avoiding the concrete crushing at the PT beam–column [3] El-Sheikh M, Sause R, Pessiki S, Lu LW. Seismic behavior and design of
and column–foundation interfaces. Experimental results unbonded post-tensioned precast concrete frames. PCI J 1999;44(3):54–71.
showed that the steel jackets have relatively large strains [4] Stone WC, Cheok GS, Stanton JF. Performance of hybrid moment-resisting
precast beam-column concrete connection subjected to cyclic loading. ACI
near the beam–column and column–foundation interfaces Struct J 1995;92(2):229–49.
and the stress distribution in the steel jacket of the beam [5] Priestley MJN. Overview of the PRESSS research program. PCI J 1991;36:50–7.
was more uniform than that in the steel jacket of the [6] Li L, Mander JB, Dhakal RP. Bi-directional cyclic loading experiment on a 3-D
beam-column joint designed for damage avoidance. J Struct Eng
column. Further investigation is needed to better quantify 2008;134(11):1733–42.
the length of steel jacket to sufficiently spreading the bear- [7] Solberg KM, Dhakal RP, Bradley BA, Mander JB, Li L. Seismic performance of
ing forces at the contact interfaces to the concrete without damage-protected beam-column joints. ACI Struct J 2008;105(2):205–14.
[8] Bradley BA, Dhakal RP, Mander JB, Li L. Experimental multi-level seismic
causing any damage. performance assessment of 3D RC frame designed for damage avoidance.
(4) In general, the tension forces of the PT strands in the beam Earthquake Eng Struct Dynam 2008;37(1):1–20.
and columns increase linearly as the drift increases and [9] Rodgers GW, Solberg KM, Chase JG, Mander JB, Bradley BA, Dhakal RP, Li L.
Performance of a damage-protected beam-column subassembly utilizing
the PT strands exhibited a V-shaped cyclic response. Due
external HF2V energy dissipation devices. Earthquake Eng Struct Dynam
to the ‘‘expansion’’ of the SC frame and the actuator loading 2008;37(13):1549–64.
applied only on the east column, some asymmetry was [10] Rodgers GW, Solberg KM, Mander JB, Chase JG, Bradley BA, Dhakal RP. High-
observed in the amount of gap opening angles at the force-to-volume seismic dissipators embedded in a jointed precast concrete
frame. J Struct Eng 2012;138(3):375–86.
beam–column and column–foundation connections in the [11] Morgen BG, Kurama YC. A friction damper for post-tensioned precast concrete
positive and negative directions of loading. moment frames. PCI J 2004;49(4):112–32.
188 L.-L. Song et al. / Engineering Structures 88 (2015) 176–188

[12] Chou CC, Chen JH. Tests and analyses of a full-scale post-tensioned RCS frame [21] GB 50010-2002. Code for design of concrete structures. Chinese code, Beijing,
subassembly. J Constr Steel Res 2010;66(11):1354–65. China (in Chinese); 2002.
[13] Chou CC, Chen JH. Seismic design and shake table tests of a steel post- [22] OpenSees-Open system for earthquake engineering simulation. Pacific
tensioned self-centering moment frame with a slab accommodating frame earthquake engineering research center. PEER. Richmond (CA, USA). <http://
expansion. Earthquake Eng Struct Dynam 2011;40(11):1241–61. opensees.berkeley.edu/>.
[14] Roh H, Reinhorn AM. Hysteretic behavior of precast segmental bridge piers [23] Seo CY, Lin YC, Sause R, Ricles JM. Development of analytical models for 0.6
with superelastic shape memory alloy bars. Eng Struct 2010;32(10):3394–403. scale self-centering MRF with beam web friction devices. In: 6th international
[15] Chou CC, Chang HJ, Hewes JT. Two-plastic-hinge and two dimensional finite conference for steel structures in seismic area (STESSA), Philadelphia, PA,
element models for post-tensioned precast concrete segmental bridge 2009.
columns. Eng Struct 2013;46:205–17. [24] Filiatrault A, Tremblay R, Christopoulos C, Folz B, Pettinga D. Elements of
[16] Holden TJ, Restrepo JI, Mander JB. Seismic performance of precast reinforced earthquake engineering and structural dynamics. 3rd ed. Montreal: Presses
and prestressed concrete walls. J Struct Eng 2003;129(3):286–96. Internationales Polytechnique; 2013.
[17] Chou CC, Chen YC, Pham DH, Truong VM. Steel braced frames with dual-core [25] Sideris P, Aref A, Filiatrault A. Quasi-static cyclic testing of a large-scale hybrid
SCBs and sandwiched BRBs: mechanics, modeling and seismic demands. Eng sliding-rocking segmental column with slip-dominant joints. J Bridge Eng
Struct 2014;72:26–40. 2014:04014036. http://dx.doi.org/10.1061/(ASCE)BE.1943-5592.0000605.
[18] Vasdravellis G, Karavasilis TL, Uy B. Finite element models and cyclic behavior [26] Applied Technology Council (ATC). Guidelines for seismic performance
of self-centering steel post-tensioned connections with web hourglass pins. assessment of buildings: ATC-58 50% draft. Rep. No. 58, ATC, Washington,
Eng Struct 2013;52:1–16. DC; 2009.
[19] Song LL, Guo T, Chen C. Experimental and numerical study of a self-centering [27] Khoo HH, Clifton C, Butterworth J, et al. Development of the self-centering
prestressed concrete moment resisting frame connection with bolted web sliding hinge joint with friction ring springs. J Constr Steel Res
friction devices. Earthquake Eng Struct Dynam 2014;43(4):529–45. 2012;78:201–11.
[20] Guo T, Song LL. Performance-based seismic design method of self-centering
prestressed concrete frames with web friction devices. J Build Struct
2014;35(2):22–8 (in Chinese).