Engineering Structures 138 (2017) 170–180

Contents lists available at ScienceDirect

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

Moment and energy dissipation capacities of post-tensioned precast

concrete connections employing a friction device
Takeaki Koshikawa
Division of Architectural and Structural Design, Faculty of Engineering, Hokkaido University, Kita 13, Nishi 8, Kita-ku, Sapporo 060-8628, Japan

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

Article history: This paper describes an analytical investigation into the design of unbonded post-tensioned (PT) precast
Received 26 May 2016 concrete beam-column connections that employ a friction device. Design formulas related to the moment
Revised 9 December 2016 capacities, hysteretic energy dissipation capacities, and self-centering capabilities of the PT connections
Accepted 3 February 2017
are developed in normalized form. The validation of the design formulas and the effect of normalized
Available online 16 February 2017
parameters such as friction force, initial post-tensioning force and the location of the friction device on
the moment and energy dissipation capacities are investigated. A design procedure to achieve desired
Keywords:
design moment and energy dissipation capacities for the PT connections is also developed using the
Design
Unbonded post-tensioning
design formulas. Contour plots for the linear limit moment and hysteretic energy dissipation ratio are
Energy dissipation provided for determining the required values of the normalized parameters. The investigation is based
Self-centering on the analytical responses from a nonlinear section analysis method for a number of PT connections
Precast concrete with various sets of normalized parameters. The results confirm the adequacy of the design formulas
Section analysis and design procedure.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction moment capacities, hysteretic energy dissipation capacities, and

Precast concrete beam-column connections that use unbonded
post-tensioning between the beams and columns can resist seismic
loads without incurring significant residual deformation and struc-
tural damage, but they have low inherent damping. In order to
enhance their capacity to dissipate energy and thus reduce dis-
placement due to earthquakes, the use of energy dissipation
devices placed across the beam-column interfaces of the connec-
tions has been proposed, starting with the PRESSS Program [1,2].
The energy dissipation devices provide supplemental energy dissi-
pation for connections through displacements that occur as a result
of opening and closing gaps at the beam-column interfaces during
earthquakes. Various kinds of energy dissipation devices such as
steel yielding devices [1–3], lead extrusion damping devices
[4,5], and friction damping devices [6–9] have been suggested
and their application has been experimentally validated in tests
of their components and in beam-column connection tests.
The most attractive features of unbonded post-tensioned (PT)
energy dissipating precast concrete beam-column connections
are their self-centering behavior and their adequate hysteretic
energy dissipation (flag-shaped hysteresis). To perform satisfacto-
rily, the PT connections need to be designed to achieve the required
E-mail address: takeaki@eng.hokudai.ac.jp
self-centering capabilities. Although these requirements can be
satisfied by any one of an infinite number of combinations of
unbonded post-tensioning tendons and energy dissipation devices,
to date there is limited research on selecting suitable combinations
to achieve design purposes. Morgen and Kurama [10] provided a
seismic design procedure to satisfy the design beam-end moment
demand and energy dissipation requirements for an unbonded PT
precast concrete beam-column connection that had friction
devices placed on both the top and bottom surfaces of beam. To
develop design formulas, they decomposed the nominal moment
capacity into moment resistances due to the post-tensioning forces
and friction forces based on the equilibrium of the forces at the
beam end. They assumed that the friction forces acting in opposite
directions to each other were equal for the two friction devices,
and that the compression force in the equivalent compression
stress block was equal to the post-tensioning force. The validity
of the design procedure was confirmed by nonlinear analyses of
PT beam-column subassembly models with various design
parameters.
In recent years, there has been some research on PT connections
employing friction devices with different configurations aiming to
eliminate the possibility of interfering with energy dissipation
devices in the floor slab. Sugiura et al. [7] proposed a PT connection
with a beam bottom friction device (BFD), where a friction device is
placed on only the bottom surface of the beam. The results of cyclic

http://dx.doi.org/10.1016/j.engstruct.2017.02.012
0141-0296/Ó 2017 Elsevier Ltd. All rights reserved.
 T. Koshikawa / Engineering Structures 138 (2017) 170–180
(a) friction device (BFD)
Fig. 1. Test configuration for a PT connection with a BFD (adapted from Sugiura et al. [7]).
tests on the PT connection specimens shown in Fig. 1 demonstrate
that the BFD can provide satisfactory energy dissipation to PT con-
nections. Song et al. [8] developed an armored precast concrete
beam-column connection with a friction device located at the
web of the beam, called a beam web friction device (WFD). They
conducted experimental and numerical studies on its hysteretic
behavior, and found the connection to have good seismic perfor-
mance with respect to energy dissipation, self-centering, and pre-
venting local damage. Further studies on the same connection
were conducted by Song et al. [9], where the performance of PT
connections with WFDs was investigated experimentally through
a series of cyclic tests on a 0.5-scale, one-bay, one-story frame sub-
assembly. Research and development of similar connections for
steel frames has also been conducted by several other researchers
(Tsai et al. [11], Iyama et al. [12], Wolski et al. [13], Guo et al. [14],
and Lin et al. [15]).
Unlike PT connections with friction devices on both the top and
bottom of the beam, the moment resistances due to the post-
tensioning force and the friction force correlate to each other in
these PT connections because they are in equilibrium with the
compression force. In addition, the location of the friction device
may have a significant effect on the moment and energy dissipa-
tion capacities of PT connections. This paper presents an analytical
investigation focusing on this issue in the design of PT connections
having a friction device. Design formulas related to the moment
capacities, hysteretic energy dissipation capacities, and self-
centering capabilities are developed using normalized parameters
suitable for such PT connections. Results obtained using the design
formulas for the capacities are compared with results from nonlin-
ear analyses using a section analysis method that can predict the
relationship between the moment and relative rotation at the
beam-column interface under cyclic loading. The effect that using
normalized parameters has on the capacities of the PT connections
is also examined. Based on the design formulas, a design procedure
to determine the parameter values necessary to satisfy the
required moment and energy dissipation capacities is presented
and analytically validated.
(a) PT connection with a WFD
Fig. 2. Schematic diagram of unbonded PT precast concrete connections with friction devices.
171
(b) specimen
2. Design formulas for unbonded PT connections with a friction
device
The PT connections that employ either a WFD or a BFD studied
in this paper are shown schematically in Fig. 2. The precast beam is
joined to the column by an axial post-tensioning force provided by
a group of unbonded post-tensioning tendons that run straight at
the beam centroid and through the beam-column connection.
The friction device, placed on the outside of the beam web (Fig. 2
(a)) or on the bottom surface of the beam (Fig. 2(b)), is connected
to the beam end and the column face. When relative displacement
occurs between the beam end and the column face, a friction force
develops in the friction device and provides moment resistance
together with the post-tensioning force.
In a properly designed PT connection, the behavior is primarily
dominated by the opening and closing of a gap that occurs at the
beam-column interface. Fig. 3 shows the idealized moment-
rotation behavior for PT connections with WFDs and BFDs. The
term M represents the moment in the beam at the beam-column
interface and h represents the chord rotation of the beam. A posi-
tive rotation corresponds to the gap opening at the bottom of the
beam-column interface. As shown in Fig. 3, properly designed PT
connections have flag-shaped hysteresis curves under both posi-
tive and negative rotations, where the unloading curve goes back
to the origin as the energy corresponding to the area enclosed by
the loading and unloading curves is dissipated. The moments M llþ
and Mll that occur during the loading curves are the linear limit
moments [16,17] that are taken as the nominal moment capacities
under positive and negative rotations. The moments when the
unloading curves rejoin the loading curves are denoted as Mlrþ
and M lr . Since the positive and negative moment resistances
may be different depending on the location of the friction device,
the symbols for these moments are given separately for positive
and negative rotations using + and  indices. The design formulas
for the PT connections are based on idealized moment-rotation
behavior. PT connections with friction devices on both the top
and bottom of the beam are not covered by the design formulas.
(b) PT connection with a BFD
172 T. Koshikawa / Engineering Structures 138 (2017) 170–180

2.1. Moment capacity formulas positive and negative linear limit moments M llþ and M ll for each
PT connection can be expressed as follows:
The moment capacity of a PT connection can be obtained by


D  ap  af D  ap  af
summing the contributions from the unbonded post-tensioning Mllþ ¼ T po þ F fs þ df ð1Þ
2 2
tendon and the friction device. Fig. 4 shows beam-column interface
equilibria for PT connections at the positive and negative linear

D  ap  af
limit moments. The compression force in the concrete Cc is Mll ¼ T po
assumed to act at mid-depth of the equivalent compression stress
2


block, and to be in equilibrium with the initial post-tensioning D  ap  af
 F fs  df PT connection with a WFD
force in the unbonded post-tensioning tendon Tpo and the friction 2
force in the friction device Ffs. The main difference between the ð2aÞ
forces in a PT connection with a WFD (Fig. 4(a)) and a BFD (Fig. 4


(b)) is the direction of the friction force under negative rotation, D  ap þ af
which causes differences in the magnitude and location of the Mll ¼ T po
2
compression force in the concrete Cc, and hence the length of the

D  ap þ af
moment arm. Assuming, for simplicity, that the friction force acts þ F fs  df PT connection with a BFD ð2bÞ
2
parallel to the beam at the location of the friction device, the

(a) PT connection with a WFD (b) PT connection with a BFD

Fig. 3. Idealized moment-rotation behavior for PT connections with friction devices.

(a) PT connection with a WFD

(b) PT connection with a BFD

Fig. 4. Equilibrium of PT beam-column interface at linear limit moment.
 T. Koshikawa / Engineering Structures 138 (2017) 170–180 173

 
where D is the beam depth and df is the distance between the mid- 1 1
mlr ¼  qp 1 þ pf  qp ð1 þ pf Þ2  pf r f
depth of the beam and the location of the friction device. Expres- 2 0:85
sions in parentheses account for the lengths of the moment arms PT connection with a BFD ð8bÞ
of the initial post-tensioning force Tpo and the friction force Ffs. ap
and af are the components of the depth of the equivalent compres- where mlrþ and mlr are the moments Mlrþ and Mlr normalized by
0
sion stress block contributed by Tpo and Ffs, given by BD2 f c , respectively.
T po F fs
ap ¼ 0 ; af ¼ 0 ð3Þ 2.2. Formulas for energy dissipation capacities
0:85Bf c 0:85Bf c
0
where B is the beam width and f c is the compressive strength of the The energy dissipation capacities of PT connections with flag-
beam concrete. Note that although the first and second terms on the shaped hysteresis can be quantified using the hysteretic energy
right side of Eqs. (1), (2a) and (2b) can be considered to be the dissipation ratio bE, which is defined as the ratio of the hysteresis
moment contributions from Tpo and Ffs, respectively, both Tpo and loop area for the PT connection to the area of a bilinear elastic-
Ffs are related to the lengths of the moment arms, indicating that plastic connection with the same capacity [18]. According to Seo
the relation between Tpo and Ffs should be considered during design. and Sause [18], bE, which is one of the design parameters that con-
The linear limit moments given by Eqs. (1), (2a) and (2b) can be trols the ductility demands on unbonded PT systems, can range
put into normalized form by introducing the following normalized from 0 to 0.5 for a PT connection that self-centers.
parameters related to the friction force, the initial post-tensioning For the idealized moment-rotation behavior of each PT connec-
force and the location of the friction device: tion in Fig. 3, bE can be given by
F fs T po 2df M llþ  Mlrþ  ðMll  M lr Þ
pf ¼ ; qp ¼ 0 ; rf ¼ ð4Þ bE ¼ ð9Þ
T po BDf c D 2ðM llþ  M ll Þ
where pf is the ratio of the friction force to the initial post- Substituting Eqs. (5), (6a) and (6b), (7), (8a) and (8b) into Eq. (9), the
tensioning force. qp is the initial post-tensioning force normalized hysteretic energy dissipation ratio bE for each PT connection can be
0
by BDf c , and also represents the initial stress induced in the beam also written using the normalized parameters as follows:
concrete normalized by the concrete compressive strength. rf is
pf ð1  0:85
2
qp Þ
the distance from the friction device to the mid-depth of the beam bE ¼
normalized by one-half the beam depth. Note that the values of pf 1 1
q ð1
0:85 p
þ pf 2 Þ þ pf ð1  0:85
2
qp Þ
and qp are greater than 0 for PT connections. The value of rf is equal PT connection with a WFD ð10aÞ
to 0 for a PT connection with a WFD located at the beam mid-depth,
while it is greater than 1 for a PT connection with a BFD. By dividing pf r f
0 bE ¼ PT connection with a BFD
both sides of Eqs. (1), (2a) and (2b) by BD2 f c , substituting Eq. (4) 1  0:85
1
qp ð1 þ pf 2 Þ þ pf r f
into them and rearranging, the positive and negative normalized ð10bÞ
linear limit moments mllþ and mll can be derived as functions of
the normalized parameters as follows: As can be seen from Eqs. (10a) and (10b), the only difference
  between bE for the PT connections in the two cases is that
1 1  
mllþ ¼ q 1 þ pf  q ð1 þ pf Þ2 þ pf r f ð5Þ 1  0:85
2
qp in Eq. (10a) is replaced by rf in Eq. (10b). The value of
2 p 0:85 p  
rf in Eq. (10b) is larger than 1  0:85
2
qp in Eq. (10a) because rf is
   
1 1 greater than 1 for PT connections with BFDs and 1  0:85 2
qp is less
mll ¼  qp 1 þ pf  qp ð1 þ pf Þ2  pf r f than 1, so bE for PT connections with BFDs is larger than for PT con-
2 0:85
nections with WFDs with the same pf and qp values.
PT connection with a WFD ð6aÞ
  2.3. Formulas for self-centering capability
1 1
mll ¼  qp 1  pf  q ð1  pf Þ2 þ pf r f
2 0:85 p
To ensure the self-centering capability of PT connections, M lrþ in
PT connection with a BFD ð6bÞ Fig. 3 must be greater than or equal to 0. This requirement can be
where mllþ and mll are the positive and negative linear limit expressed in terms of the normalized moment mlrþ as
0
moments, respectively, normalized by BD2 f c . mlrþ P 0 ð11Þ
The moments M lrþ and M lr in Fig. 3, where the unloading
curves rejoin the loading curves, can also be expressed in a form Substituting Eq. (7) into Eq. (11) and solving for pf leads to the
similar to Eqs. (1), (2a) and (2b) by changing the signs of Ffs and following condition, where pf is greater than 0 for the PT connections:
af in these equations because the friction force changes direction, rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi!
0:85 1 4
reversing between loading and unloading. It should be noted that pf 6 1  1 þ rf  ð1 þ r f Þ2  q rf ð12Þ
2 qp 0:85 p
the reversal of the friction force also affects the lengths of the
moment arms of the initial post-tensioning force Tpo and the fric- Eq. (12) expresses the upper bound of pf for PT connection design.
tion force Ffs. The normalized form of these moments thus can be Fig. 5 shows the calculated upper bound of pf from Eq. (12)
written using the normalized parameters as follows: against rf values ranging from 0 to 2 and qp values of 0.05, 0.10,
 
1 1 0.15, and 0.20. In every case, the upper bound of pf is 1 when rf
mlrþ ¼ qp 1  pf  qp ð1  pf Þ2  pf r f ð7Þ is 0, and decreases as rf increases. For a given value of rf greater
2 0:85
than 0, increasing qp decreases the upper bound of pf, but the upper
 
1 1 bound of pf varies only slightly for different values of qp. These
mlr ¼  qp 1  pf  qp ð1  pf Þ2 þ pf rf results suggest that the upper bound of pf is determined roughly
2 0:85
by rf. In design, this can be used to estimate an upper bound for
PT connection with a WFD ð8aÞ
the friction force Ffs in a PT connection for a known initial post-
174 T. Koshikawa / Engineering Structures 138 (2017) 170–180
Fig. 5. Upper bound of pf.
tensioning force Tpo. For example, for a PT connection with a WFD
located at beam mid-depth (i.e. rf = 0), the upper bound of pf is 1, so
the value of Ffs can be set up to the value of Tpo. For a PT connection
with a BFD located at rf = 1.5, the upper bound of pf is less than
about 0.4, so the value of Ffs should not exceed 40% of the value
of Tpo. Note that this estimate is based on the idealized behavior
shown in Fig. 3, so some loss of stiffness and strength in real con-
nections’ structural members should be assumed in practice.
3. Section analysis method
A nonlinear section analysis method [19,20] to predict the
behavior of PT connections under reversed cyclic loading is pre-
sented in this section and shown in Fig. 6. This method uses an
incremental iterative approach to obtain the moment-relative
rotation hysteretic responses at PT connections’ beam-column
interfaces. The gap opening and closing behavior at the beam-
column interface is represented by the relative rotation hr and rel-
ative axial displacement ur, as shown in Fig. 6(b). The relative rota-
tion hr is imposed and fixed during each incremental step, while
the relative axial displacement ur is assumed and updated until
convergence.
(a) Geometry of PT connection (b) Cross section and displacement at beam-column interface
(c) Material properties
Fig. 6. Section analysis method.
In this section analysis method, the beam-end cross section at
the beam-column interface is modeled using a discrete number
of concrete fibers, unbonded post-tensioning tendons, and friction
devices. Each component in the cross section has a uniaxial consti-
tutive relationship for its material: the stress-strain (rc  ec) rela-
tionship for concrete in compression, the stress-strain (rp  ep)
relationship for post-tensioning tendons in tension, and the
force-deformation (Ff  df) relationship for friction devices, as
shown in Fig. 6(c). The axial deformations in each component
due to the gap opening and closing behavior are calculated as func-
tions of the relative rotation hr and relative axial displacement ur.
The strains in the concrete fibers and post-tensioning tendons
are calculated by assuming constant strain over the length of the
compressive fracture zone lc (assumed to be one half of the beam
depth [21]) and the unbonded length of the post-tensioning ten-
dons lp. The stresses and forces in each component are calculated
from the uniaxial constitutive relationships. These calculations
are iterated by updating the value of the relative axial displace-
ment ur until the section equilibrium condition is satisfied. The
moment M corresponding to a relative rotation hr is derived once
the equilibrium condition is satisfied, and the resulting forces
and deformations in the beam-column subassemblies can be deter-
mined by considering the geometrical and mechanical properties.
The proposed section analysis method was validated using
experimental results by analyzing 0.5-scale PT connection speci-
mens tested by Sugiura et al. [7] (see Fig. 1). Two specimens, one
with and one without a friction device, were analyzed. The geom-
etry and dimensions of these specimens were the same except for
the presence or absence of a friction device. The precast concrete
beam specimen, oriented in a vertical configuration, had a cross
section of 300 mm  500 mm, and was post-tensioned to a stub
using four unbonded post-tensioning steel bars with a total initial
post-tensioning force of 1600 kN. The friction device, which con-
structed of steel members and high friction materials used for
brake pads in trucks, was located on one side of the beam speci-
men, and had a design friction force of 100 kN. A distance of
440 mm between the friction device and the mid-depth of the
beam was used in the analysis. The measured concrete compres-
sive strength of the beam was 51.0 N/mm2 with a friction device
and 52.9 N/mm2 without one. Because no spiral or additional
 T. Koshikawa / Engineering Structures 138 (2017) 170–180
(a) Specimen with a friction device
Fig. 7. Comparison of hysteretic behavior from analytical responses and experimental results.
transverse reinforcements that could substantially confine the con-
crete at the beam end, the compressive strength of each concrete
fiber was set to the same value in the analysis. Fig. 7 compares
the analytical responses with the experimental results for the
load-beam chord rotation response of specimens with and without
a friction device. In the analyses, the load is calculated as the
moment at the beam-stub interface divided by the measured dis-
tance from the applied load to the interface. Also, in calculating
the beam chord rotation, the contribution of the beam’s elastic
flexural deformation to the deflection is considered together with
the relative rotation at the beam-stub interface. The analytical
responses agree reasonably well with the experimental results,
including the differences between the responses of specimens with
and without a friction device. The slight discrepancies in the max-
imum load and the unloading behavior when a friction device is
present may be related to the difference between the analytical
and actual force-deformation relationship of the friction device.
According to Song et al. [8] and Kim and Christopoulos [22], the
actual relationship is influenced by the flexibility of the connection
elements (i.e. the steel members and connection bolts of the fric-
tion device in the specimen) and the positions of the clamping ele-
ments (i.e. the prestressing bars of the friction device in the
specimen) that can cause reduction and gradual changes in the
stiffness during unloading. The discrepancies may be improved
by considering the above influence in the analysis method if test
data are available or an evaluation method has been established.
4. Analytical validation of design formulas
The design formulas described in Section 2 were validated using
analytical responses of a number of PT connections with various
sets of normalized parameters. Table 1 lists the complete set of
normalized parameters pf, qp, and rf considered in this analytical
validation. The range of pf values is determined based on Eq. (12)
and Fig. 5 for PT connections with BFDs, and the same values are
used for PT connections with WFDs for comparison. The upper
bound of qp is set at 0.2 to comply with the recommendation in
the Japanese guidelines [23]. The two rf values are for a PT connec-
tion with a WFD (rf = 0.0) and with a BFD (rf = 1.5). Combinations of
these parameter values result in 32 different PT connections.
Table 1
Normalized parameter values used in analytical validation.
0
pf(=Ffs/Tpo) qp ð¼ T po =ðBDf c ÞÞ
0.00 0.05
0.10 0.10
0.20 0.15
0.30 0.20
175
(b) Specimen without a friction device
The dimensions and material properties used in the analysis are
identical for all cases: beam width B = 0.6 m, beam depth D = 1.0 m,
beam length l = 4.0 m, the unbonded length of the post-tensioning
tendon lp = 5.0 m, the compressive strength of the beam concrete
0
f c ¼ 50 N=mm2 , and the yield strength of the tendon
f py ¼ 1500 N=mm2 . For all PT connections, the initial stress of the
unbonded post-tensioning tendon is set to 40% of the yield
strength of the tendon fpy so that there is no tendon yielding. Thus,
the tendon area is adjusted depending on the initial post-
tensioning force Tpo in each PT connection.
Monotonic and cyclic load analyses of the PT connections under
vertical loads V applied at the tip of the beam were conducted
using the section analysis method shown in Fig. 6. The monotonic
load analyses for upward and downward loading were done sepa-
rately for each PT connection. The displacement history of the cyc-
lic load analyses consists of two cycles of sets of six beam chord
rotations h: 0.005, 0.01, 0.015, 0.02, 0.025, and 0.03 rad. The analyt-
ical results and validation of the design formulas for PT connec-
tions are presented below.
As examples of the analytical results, Figs. 8(a) and (b) show the
analytical moment-rotation responses corresponding to rf = 0.0
and 1.5, respectively for a PT connection with pf = 0.20 and
qp = 0.10. The normalized beam end moment m is the moment in
0
the beam at the beam-column interface normalized by BD2 f c . The
beam chord rotation h is calculated as the vertical displacement
at the tip of the beam due to the opening and closing of the gap
at the beam-column interface combined with elastic beam deflec-
tion divided by the beam length l. The thick curves represent the
response under monotonic loading, while the thin curves represent
the response under cyclic loading.
The analytical results show essentially bilinear behavior with a
gradual transition from the initial to the second stiffness under
monotonic loading. Upon unloading, the moment decreases due
to the reversal of the direction of the friction force in the friction
device. The differences in moment between the loading curve
and the unloading curve are roughly constant in each cycle of pos-
itive and negative loadings.
Fig. 9 shows the effect of the normalized parameters on the nor-
malized linear limit moments of the PT connections in Fig. 8. In
Figs. 9(a) and (b), the pf values vary, and in Figs. 9(c) and (d), the
qp values vary. The results in Fig. 9 show that the normalized linear
limit moments are mainly affected by the qp values, and are not so
dependent on the pf values. An increased pf value results in a slight
increase in the normalized linear limit moment. Part of the differ-
rf(=2df/D)
ence between the normalized linear limit moments for PT connec-
0.0 tions with rf = 0.0 and PT connections with rf = 1.5 can be observed
1.5 most clearly under positive loading. For PT connections with the
same pf and qp values, the normalized linear limit moments for
PT connections with rf = 0.0 are smaller than for those with rf = 1.5.
176 T. Koshikawa / Engineering Structures 138 (2017) 170–180

(a) PT connection with = 0.0 (b) PT connection with = 1.5

Fig. 8. Analytical moment-rotation response.

(a) for PT connection with =0.0 (b) for PT connection with =1.5

(c) for PT connection with =0.0 (d) for PT connection with =1.5

Fig. 9. Effect of normalized parameters on normalized linear limit moment.

The dashed lines in Figs. 8 and 9 show the estimated values of
the normalized moments mllþ ; mll ; mlrþ ; and mlr from Eqs. (5),
(6a) and (6b), (7), (8a) and (8b). The estimated normalized linear
limit moments mllþ and mll , the differences in positive moment
between mllþ and mlrþ , and the differences in negative moment
between mll and mlr can all be seen to agree reasonably well with
the analytical results.
Fig. 10 plots the hysteretic energy dissipation ratios bE for PT
connections with rf = 0.0 and 1.5 for each beam chord rotation
cycle. These are calculated from the second cycle of each ana-
lytical response under cyclic loading (indicated by markers) as
well as from Eqs. (10a) and (10b) (indicated by dotted lines).
As in Fig. 9, the bE values for PT connections with varying pf
values are shown in Figs. 10(a) and (b), and the bE values for
PT connections with varying qp values are shown in Figs. 10
(c) and (d). In calculations based on the analytical response,
the bE value for a given cycle is assumed to be the area of that
cycle’s analytical hysteresis loop divided by the area of the cor-
responding bilinear elastic-plastic loop, which is defined by the
linear limit moments from Eqs. (1), (2a) and (2b) and the elastic
rotational stiffness of the beam.
The bE values from the analytical response for each PT connec-
tion are nearly identical regardless of the amplitude of the beam
chord rotation cycles, except for those with rf = 0.0 and larger qp
 T. Koshikawa / Engineering Structures 138 (2017) 170–180
(a) for PT connection with = 0.0
(c) for PT connection with
Fig. 10. Effect of normalized parameters on hysteretic energy dissipation ratio.
values. The exceptions shown in Fig. 10(c) may be due to progres-
sive damage in the beam concrete during the larger rotation cycles
that led to a shortening of the moment arm of the friction force
under both positive and negative loadings. Figs. 10(a) and (b) show
that the bE increases significantly as the pf value increases. Also,
Figs. 10(c) and (d) show that changes in qp have a smaller effect
on the bE values, and the effect is different for PT connections with
rf = 0.0 and rf = 1.5. An increased qp value leads to a slight decrease
in bE for PT connections with rf = 0.0, but a slight increase when
rf = 1.5. In addition, for PT connections with the same pf and qp val-
ues, bE for connections with rf = 0.0 is smaller than for those with
rf = 1.5.
As can be seen in Fig. 10, the bE values estimated by Eqs. (10a)
and (10b) have trends similar to those from the analytical response
described above, but there are some differences that depend on the
various normalized parameters. In the analytical response, the PT
connections dissipate energy through not only the friction device
but also by the beam concrete’s compression, which is not consid-
ered in Eqs. (10a) and (10b). Thus, for PT connections with pf = 0.0
(i.e., with no friction device), the bE values from the analytical
response are positive whereas the estimated values are zero, as
shown in Figs. 10(a) and (b). This also affects the differences
between the analytical and estimated bE values for PT connections
with larger pf values. The differences decrease as the pf values
increase when rf = 0.0 (Fig. 10(a)), and are roughly constant when
rf = 1.5 (Fig. 10(b)), indicating that desirable, stable energy dissipa-
tion can be better achieved when the friction device is located on
the bottom surface of the beam than when it is on the beam’s
web. Moreover, Figs. 10(c) and (d) show that the differences
increase as qp values increase because increasing the initial post-
tensioning force leads to the beam concrete dissipating more
energy.
177
(b) for PT connection with = 1.5
(d) for PT connection with
Similar agreement was observed between the estimated values
and the analytical results for moment and energy dissipation
capacities for the other PT connections used in the analytical vali-
dation. The results of the analytical validation indicate that both PT
connection capacities can be estimated using the normalized
parameters introduced in the design formulas. It is also worth not-
ing that the normalized moment-rotation responses are not seri-
ously affected by the dimensions and material properties if the
PT connections have the same set of normalized parameters (see
Fig. 11, where the normalized parameters are the same as those
of the PT connections in Fig. 8). Thus, the estimates based on the
design formulas in normalized form can be applied to PT connec-
tions with different dimensions and material properties.
5. Design procedure
This section describes a design procedure to achieve desired
design moment and energy dissipation capacities for PT connec-
tions. The procedure is developed using the design formulas vali-
dated above to determine the values of the normalized
parameters necessary to achieve design targets for the positive
normalized linear limit moment mllþ and hysteretic energy dissipa-
tion ratio bE. In this procedure, the negative or average normalized
linear limit moment can also be taken as the target design value in
the same manner if necessary.
Fig. 12 shows contour plots for mllþ and bE used in the design
procedure. Both mllþ and bE are described as functions of the nor-
malized parameters pf and qp for PT connections with rf = 0.0 in
Fig. 12(a) and rf = 1.5 in Fig. 12(b). Solid lines in the figures show
the contour lines for three mllþ values (0.04, 0.06 and 0.08) calcu-
lated from Eq. (5). Dashed lines show the contours of three bE val-
178 T. Koshikawa / Engineering Structures 138 (2017) 170–180
(a) PT connection with
Fig. 11. Analytical responses of PT connections with different dimensions and material properties.
(a) PT connections with
Fig. 12. Contour plots for normalized linear limit moment and hysteretic energy dissipation ratio.
ues (0.125, 0.250 and 0.375) calculated from Eqs. (10a) and (10b).
Note that the contour line for bE = 0.375 does not appear in the
range of Fig. 12(a). The shaded area in Fig. 12(b) represents the
area beyond the upper bound of pf from Eq. (12). As mentioned
previously, the upper bound of pf is 1 for PT connections with
rf = 0.0, so the shaded area does not appear in Fig. 12(a).
In Fig. 12, it can be seen that the contour plots for certain values
of mllþ and bE intersect. The normalized parameter values pf and qp
at the intersection can be assumed to be the required values for the
design targets for mllþ and bE. As examples, Figs. 13 and 14 show
the analytical cyclic responses of PT connections for target design
values of mllþ ¼ 0:4 and 0:8 and bE ¼ 0:125 and 0:250. In this anal-
ysis, the pf and qp values at the intersections indicated by dots in
Fig. 12 are set for PT connections that have the same dimensions
and material properties used in Section 4. The responses for PT
connections with rf = 0.0 and the pf and qp values from Fig. 12(a)
are shown in Figs. 13(a)–(d). Similarly, the responses for PT con-
nections with rf = 1.5 and the pf and qp values from Fig. 12(b) are
shown in Figs. 14(a)–(d). Along with the analytical moment-
rotation response, the average value of the hysteretic energy dissi-
pation ratio bEave, calculated from all of the second beam chord
rotation cycles of the analytical response, is also shown in each
figure.
Figs. 13 and 14 show that these PT connections have moment
capacities that satisfy the design targets for mllþ . The calculated
bEave values satisfy the design targets for bE = 0.125 for PT connec-
tions with rf = 0.0 (Figs. 13(a) and (c)), and are more satisfactory for
(b) PT connection with
(b) PT connections with
PT connections with rf = 1.5 (Fig. 14). However, they are smaller
than the design targets for bE = 0.250 for PT connections with
rf = 0.0 (Figs. 13(b) and (d)), indicating that the energy dissipation
provided by a friction device located on the beam web is lower
than expected in the design formulas when the friction force is
large relative to the initial post-tensioning force. This can also be
seen in Fig. 10(a) and may be due to the reversal of the friction
force at the beginning of each half-rotation cycle. The analytical
response of the friction force variation during a single rotation
cycle is shown in Fig. 15(a) for the PT connection in Fig. 13(b). In
the half-rotation cycles, the friction force acts in the negative direc-
tion at the start of the rotation and then transitions gradually to
the positive direction. The gradual transition delays the develop-
ment of the full friction force in the positive direction, so this
may lead to a loss of energy dissipation. This is not observed in
PT connections with rf = 1.5 (see Fig. 15(b)), and is not considered
in the design formulas. Thus, for PT connections with rf = 0.0, it
may be desirable for pf values to be set slightly larger than the val-
ues at the intersections in Fig. 12(a) to achieve the design energy
dissipation capacity while achieving the design target mllþ values.
From Figs. 12–14, it can be seen that PT connections with rf = 0.0
and rf = 1.5 require different combinations of pf and qp values for
the same design target mllþ and bE values. A comparison of these
PT connections, targeting the same mllþ and bE values, shows that
the pf and qp values for PT connections with rf = 1.5 are smaller
than those for PT connections with rf = 0.0, and the differences
are amplified as the target mllþ and bE values grow.
 T. Koshikawa / Engineering Structures 138 (2017) 170–180 179

(a) Targeting = 0.04 and = 0.125 (b) Targeting =0.04 and = 0.250

(c) Targeting = 0.08 and (d) Targeting =0.08 and

Fig. 13. Application of design procedure for PT connections with rf = 0.0.

(a) Targeting = 0.04 and = 0.125 (b) Targeting = 0.04 and = 0.250

(c) Targeting = 0.08 and (d) Targeting = 0.08 and

Fig. 14. Application of design procedure for PT connections with rf = 1.5.

180 T. Koshikawa / Engineering Structures 138 (2017) 170–180
(a) PT connection with
Fig. 15. Analytical friction force variation during second cycle of 0.01 rad beam chord rotation.
6. Conclusions
The following conclusions can be drawn from this study:
1. The moment capacities, hysteretic energy dissipation capaci-
ties, and self-centering capabilities of PT connections with
WFDs or BFDs can be expressed as functions of normalized
parameters pf, qp, and rf, which are related to the friction force,
initial post-tensioning force, and location of the friction device,
respectively. The design formulas suggested that the upper
bound of pf for ensuring self-centering capability in PT connec-
tions can be determined roughly using rf values.
2. The proposed nonlinear section analysis method is able to pre-
dict the behavior of PT connections under cyclic loading. The
analytical results of a number of PT connections with various
sets of normalized parameters showed that their moment
capacities were mainly affected by qp values and the hysteretic
energy dissipation capacities were mainly affected by pf values.
For PT connections with the same pf and qp values, both capac-
ities for PT connections with WFDs (rf = 0.0) were smaller than
those of PT connections with BFDs (rf = 1.5). Comparisons
between the estimated values from design formulas and the
analytical results showed generally reasonable agreement for
moment and energy dissipation capacities.
3. The analytical results also confirmed the adequacy of the design
procedure, but for PT connections with rf = 0.0, slightly larger pf
values than those that arise from the design procedure might be
needed to achieve design energy dissipation capacity. For PT
connections aiming to provide the same moment and energy
dissipation capacities, the pf and qp values for PT connections
with BFDs (rf = 1.5) were smaller than for connections with
WFDs (rf = 0.0), and the differences were amplified as the target
capacities grew.
Acknowledgments
Support for this research from JSPS KAKENHI Grant Number
15K06282 is gratefully acknowledged.
References
[1] Stone WC, Cheok GS, Stanton JF. Performance of hybrid moment-resisting
precast beam-column concrete connections subjected to cyclic loading. ACI
Struct J 1995;92(2):229–49.
[2] Priestley MJN, Sritharan S, Conley JR, Pampanin S. Preliminary results and
conclusions from the PRESSS five-story precast concrete test building. PCI J
1999;44(6):42–67.
(b) PT connection with
[3] Bradley BA, Dhakal RP, Mander JB, Li L. Experimental multi-level seismic
performance assessment of 3D RC frame designed for damage avoidance.
Earthquake Eng Struct Dyn 2008;37(1):1–20.
[4] Rodgers GW, Solberg KM, Chase JG, Mander JB, Bradley BA, Dhakal RP, et al.
Performance of a damage-protected beam-column subassembly utilizing
external HF2V energy dissipation devices. Earthquake Eng Struct Dyn
2008;37(13):1549–64.
[5] Rodgers GW, Solberg KM, Mander JB, Chase JG, Bradley BA, Dhakal RP. High-
force-to-volume seismic dissipators embedded in a jointed precast concrete
frame. J Struct Eng 2012;138(3):375–86.
[6] Morgen BG, Kurama YC. A friction damper for post-tensioned precast concrete
moment frames. PCI J 2004;49(4):112–33.
[7] Sugiura H, Sashima Y, Sanada Y. Performance evaluation tests of unbonded
prestressed concrete beams with/without a friction damper. In: 20th
Symposium on Developments in Prestressed Concrete, Hakodate, Japan;
2011 [in Japanese].
[8] Song LL, Guo T, Chen C. Experimental and numerical study of a self-centering
prestressed concrete moment resisting frame connection with bolted web
friction devices. Earthquake Eng Struct Dyn 2014;43(4):529–45.
[9] Song LL, Guo T, Gu Y, Cao ZL. Experimental study of a self-centering prestressed
concrete frame subassembly. Eng Struct 2015;88:176–88.
[10] Morgen B, Kurama Y. Seismic design of friction-damped precast concrete
frame structures. J Struct Eng 2007;133(11):1501–11.
[11] Tsai KC, Chou CC, Lin CL, Chen PC, Jhang SJ. Seismic self-centering steel beam-
to-column moment connections using bolted friction devices. Earthquake Eng
Struct Dyn 2008;37(4):627–45.
[12] Iyama J, Seo CY, Ricles JM, Sause R. Self-centering MRFs with bottom flange
friction devices under earthquake loading. J Constr Steel Res 2009;65
(2):314–25.
[13] Wolski M, Ricles JM, Sause R. Experimental study of a self-centering beam–
column connection with bottom flange friction device. J Struct Eng 2009;135
(5):479–88.
[14] Guo T, Song L, Zhang G. Numerical simulation of the seismic behavior of self-
centering steel beam-column connections with bottom flange friction devices.
Earthquake Eng Eng Vibr 2011;10(2):229–38.
[15] Lin Y, Sause R, Ricles J. Seismic Performance of a large-scale steel self-centering
moment resisting frame: MCE hybrid simulations and quasi-static pushover
tests. J Struct Eng 2013;139(7):1227–36.
[16] El-Sheikh M, Pessiki S, Sause R, Lu LW. Moment rotation behavior of unbonded
post-tensioned precast concrete beam-column connections. ACI Struct J
2000;97(1):122–32.
[17] Perez FJ, Sause R, Pessiki S. Analytical and experimental lateral load behavior
of unbonded post tensioned precast concrete walls. J Struct Eng 2007;133
(11):1531–40.
[18] Seo CY, Sause R. Ductility demands on self-centering systems under
earthquake loading. ACI Struct J 2005;102(2):275–85.
[19] Koshikawa T, Kadowaki H, Matsumora M. Energy dissipation efficiency of
unbonded post-tensioned precast concrete beam-column connections with
beam-end dampers. In: 15th World Conference on Earthquake Engineering,
Lisbon, Portugal; 2012.
[20] Koshikawa T. Energy dissipation in post-tensioned self-centering precast
concrete connections with a friction device. In: 8th International Structural
Engineering and Construction Conference, Sydney, Australia; 2015.
[21] Matsumora M, Koshikawa T, Kikuchi M. Evaluation of ultimate strength and
rotation angle for unbonded post-tensioned precast concrete beams by using
section analysis. J Struct Constr Eng 2014;79(701):1005–13 [in Japanese].
[22] Kim H-J, Christopoulos C. Friction damped post tensioned self-centering steel
moment-resisting frames. J Struct Eng 2008;134(11):1768–79.
[23] Architecture Institute of Japan, Guidelines for structural design and
construction of prestressed concrete buildings based on performance
evaluation concept (draft), Tokyo, Japan; 2015 [in Japanese].