Engineering Structures 41 (2012) 413–427

Contents lists available at SciVerse ScienceDirect

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

Shear resistance of pinned connections of precast members to monotonic

and cyclic loading
Ioannis N. Psycharis ⇑, Harris P. Mouzakis
School of Civil Engineering, National Technical University, Athens, Greece
Laboratory for Earthquake Engineering, National Technical University of Athens, 9, Heroon Polytechneiou Str., Zografos 15780, Athens, Greece

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

Article history: One of the major issues in the design of precast structures against earthquakes is the proper design of their
Received 15 September 2011 connections, mainly the beam-to-column joints. Many different types of such connections are used world-
Revised 10 January 2012 wide, as monolithic, emulative and dry pinned ones. The latter case is the most common solution in south-
Accepted 26 March 2012
ern Europe and elsewhere for single-storey or low-rise precast buildings, and is the subject of the
Available online 7 May 2012
experimental research reported in this paper. The experiments were performed at the Laboratory for
Earthquake Engineering of the National Technical University of Athens, Greece in the framework of the
Keywords:
European FP7 project, SAFECAST. Precast beam and column elements connected with dowels were tested
Shear design
Shear resistance
under monotonic and cyclic, pure shear loading and the research was focused on several design aspects, as
Design of dowels the shear ductility capacity of the connections and the effect of various parameters on their strength. The
Connections parameters examined include the diameter D of the dowels, their number, their distances d and dn from
Pinned connections the edges in the longitudinal and the transverse direction of the beam respectively, and the strength of the
Precast grout of their ducts. Improvements in the design were also proposed and tested experimentally. The
Prefabrication results show that, for small values of the ratio d/D, the strength of the connection is lower in the pull direc-
Cyclic response tion than that in the push direction for both monotonic and cyclic loading, due to the early spalling of the
Monotonic response
cover concrete that occurs. Compared to the strength for monotonic loading, the cyclic response shows
Seismic design
significantly lower resistance, less than half the monotonic one. The results for cyclic loading also show
that significant values of shear ductility can be achieved by dry pinned joints, provided that the concrete
cover of the dowels has sufficient thickness. Comparisons between the experimental results obtained for
various design parameters show that secondary effects related to the number of the dowels can occur
for large forces during monotonic loading, but are less important for cyclic response. A formula is pro-
posed for the calculation of the shear strength of pinned connections, which can be used in seismic design.
Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction been conducted on the resistance of connections of precast mem-

Precast reinforced concrete constructions have been used for
more than half of a century. Due to many advantages that they pos-
sess, as the fast and economic construction, the small amount of
the on-site required labour work and the minimized influence on
the weather conditions that affect the construction time and the
quality of the structure, their application is spreading considerably
and successfully all over the world, making prefabrication one of
the leading construction techniques.
Although the safety of precast constructions during earthquakes
has been recognized as one of the most important issues in the de-
sign, surprisingly enough, only limited experimental research has
⇑ Corresponding author at: Laboratory for Earthquake Engineering, National
Technical University of Athens, 9, Heroon Polytechneiou Str., Zografos 15780,
Athens, Greece. Tel.: +30 2107721154; fax: +30 2107721182.
E-mail address: ipsych@central.ntua.gr (I.N. Psycharis).
bers and the behaviour of the structure as a whole.
The seismic performance of precast structures greatly depends
on the type of the connections, their position into the structural sys-
tem and the type of the structural system itself. An overview of the
research and development for the design of precast framed struc-
tures can be found in the fib Bulletin No. 27 [1], in earlier reports
by Elliott [2] and Simeonov et al. [3] and in ATC-08 [4]. Recently,
fib Bulletin No. 43 [5] was published, in which considerable infor-
mation on practically every type of beam-to-column connection is
given concerning its design and behaviour. However, these guide-
lines were based merely on the resistance of the connections to
monotonic loading and do not directly reflect the influence of
strength and deformability of the connections on the overall behav-
iour of a precast structure.
The most notable effort on the experimental investigation of the
earthquake response of precast structures has been the PRESSS
(Precast Seismic Structural Systems) project [6–9] carried out in
USA and Japan. In the framework of this project, a 60% scale,

0141-0296/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.engstruct.2012.03.051
414 I.N. Psycharis, H.P. Mouzakis / Engineering Structures 41 (2012) 413–427
30 ft.  30 ft. five-storey building with several types of connections
was tested at the University of California, San Diego. The major
objective of this project was to issue design guidelines for broader
acceptance of the precast concrete construction in seismic zones
and to develop new concepts and technologies for precast systems
to be applied in regions of high seismicity; thus, it does not comply
directly to the main problematic of traditional techniques of pre-
fabrication as they usually appear around the world.
Semi-rigid connections were examined extensively within the
European project COST C1 Action [10], in which 25 countries and
more than 125 research centres were involved. A wide range of
materials and geometries were studied; the results, however,
showed significant diversion [11–16]. Significant work related to
the behaviour of several types of precast connections was reported
by many other researchers (Imai et al. [17], Watanabe [18], Priest-
ley and Tao [19], Nakaki et al. [20], Ersoy and Tankut [21], El Debs
et al. [22], Rahman et al. [23]).
Most of the above-mentioned research has been focused on the
seismic behaviour of specific types of connections, mainly moment
resisting ones, since in many countries (USA, New Zealand, Japan,
Australia, etc.), rigid connections are preferred for beam-to-column
joints. However in Europe (Italy, Greece, Spain, Portugal, Slovenia,
etc.) and elsewhere (Turkey, Armenia, etc.), simple dry pinned
beam-to-column connections are traditionally used for single-sto-
rey or low-rise precast, mainly industrial, buildings.
Little investigation has been reported on pinned connections
that are materialized by dowels, e.g. by Leong [24] who proposed
improvements of perfectly pinned dowel connections. The dowel
mechanism was investigated extensively by Vintzeleou and Tassios
[25,26], while Tsoukantas and Tassios [27] conducted an analytical
investigation on the shear resistance of connections between linear
precast elements under monotonic and cyclic loading and pro-
posed design values for the shear resistance of each connection
and its corresponding shear slip for rough and smooth surfaces.
Recently, significant experimental and numerical research on
the seismic behaviour of precast structures with dry pinned con-
nections was conducted in the framework of the ‘‘Growth’’ FP5
project of the European Commission, ‘‘Precast EC8: Seismic behav-
iour of precast concrete structures with respect to Eurocode 8 (Co-
Normative Research)’’. The project focused on the overall nonlinear
behaviour of structures with several types of connections and on
the global ductility that can be attained (Negro et al. [28], Carydis
Fig. 1. General layout of the specimens.
et al. [29], Psycharis et al. [30], Kramar et al. [31]). However, a de-
tailed investigation of the seismic response of the connections was
not performed.
Considering the limited research that has been conducted on
structures composed by linear precast elements with pinned
beam-to-column connections, the experimental research pre-
sented herein aims at the investigation of their shear resistance
to monotonic and cyclic loading and at recommendations concern-
ing their design towards the improvement of their seismic re-
sponse. This investigation was carried out at the Laboratory for
Earthquake Engineering of the National Technical University of
Athens, Greece (NTUA), within the framework of the FP7 project
of the European Commission, ‘‘SAFECAST: Performance of innova-
tive mechanical connections in precast building structures under
seismic conditions. More details on the experimental results that
are reported in this paper and on the experimental investigation
performed in other Institutions within the SAFECAST project can
be found in [32].
It is known that there are numerous alternative pinned-connec-
tion mechanisms used worldwide in precast systems and, evi-
dently, the results presented herein cannot cover all the cases.
However, the design issues that are discussed and the conclusions
that are drawn can be extended to other similar systems, beyond
the specific ones examined.
2. Experimental setup
The primary objective of this experimental research was to
investigate the effect of various design parameters on the resis-
tance of pinned beam-to-column connections under pure shear
monotonic and cyclic loading. This type of connection represents
the common solution in precast construction practice in many
countries in Europe and elsewhere, mainly for single-storey indus-
trial buildings.
The specimens represented a typical pinned connection of lin-
ear precast members and consisted of the end parts of the beam
and the column that were connected by steel dowels (Fig. 1). The
experimental setup is shown in Fig. 2a. Each specimen was sub-
jected to monotonic or cyclic, displacement-controlled loading, ap-
plied to the rear end of the beam, while the column was securely
fastened to the strong floor of the Laboratory. For the tests, a
 I.N. Psycharis, H.P. Mouzakis / Engineering Structures 41 (2012) 413–427
ACTUATOR
Fmax = 500 kN
dmax = ± 0.20 m
SPECIAL DEVICE
FOR THE UNIAXIAL
APPLICATION OF
THE EXCITATION
(a)
Fig. 2. (a) Experimental setup and (b) special device for the uniaxial application of the excitation.
500 kN capacity MTS hydraulic actuator with displacement range
of ±200 mm, connected to the reaction wall of the Laboratory,
was used. The driving force was applied exactly at the level of
the joint, in order to achieve pure shear conditions. To prevent
any out-of-plane motion, a special device was used (Fig. 2b), which
consisted of a piston and a sleeve connected to the strong floor and
which allowed only uniaxial motion of the beam. No extra vertical
load was applied to the joint.
The special device that was used prevented any rotation at the
joint and the tests were performed for pure shear behaviour of
the connection. In this way, the phenomena associated with the
shear deformation alone could be studied. It should be noted,
however, that in real structures there might be rotation at the
joints, caused mainly by the deflection of the columns. Large joint
rotations may reduce the overall shear resistance of the connec-
tions, as observed in cyclic and dynamic tests on frames per-
formed in the framework of SAFECAST project at the University
of Ljubljana and NTUA respectively [32,33]. This issue is further
discussed in Section 4.2.
2.1. Description of the specimens
The column elements were short and stiff in order to prevent
any bending deformation during the tests. The cross sections of
the column and beam elements were orthogonal with dimensions
similar to the ones usually applied in precast single-storey build-
ings (physical scale). In particular, the cross section of the column
was 0.60 m  0.40 m and the cross section of the beam was
0.40 m  0.60 m. All specimens were identical in dimensions.
For the proper sitting of the beams on the columns, elastomeric
(neoprene) pads of about 20 mm thickness are typically used in
practice. Their purpose is to eliminate any anomalies of the contact
surfaces, to allow small displacements due to thermal effects, pre-
stress of the beam, creep, etc., and to prevent the impact between
beam and column due to the rotation at the joint during seismic re-
sponse. These elastomeric pads affect the shear resistance of the
connections, because they increase the distance between the plas-
tic hinges that are developed in the dowels for large displacements.
This affects the axial strains and the axial force that are developed
in the dowels. For this reason, and in order to simulate the real
conditions, elastomeric pads of 20 mm thickness were included
in the specimens despite the fact that no rotations were allowed
at the joints.
The columns and the beams were constructed of high strength
concrete, while the steel of the dowels and the reinforcement was
of grade B500C. The mean value of the measured compression
strength of the concrete of the specimens was 30–35 MPa (see
415
SPECIMEN
(b)
Table 2). For the assembly of the specimens, Ø65 waiting ducts
were placed in the beams for the passage of the dowels during
mounting. In most cases, jagged steel ducts were used; however,
in some specimens the ducts were from smooth plastic tubes.
The ducts were grouted after assemblage with non-shrinking grout
and the dowels were bolted at their top. All the experiments, ex-
cept two, were performed 24 h after grouting. The mean compres-
sion strength of the grout at that time was measured 23 MPa,
considerably less than the strength of the concrete. The strength
of the grout was not measured for all specimens, but it is estimated
that it was close to the above-mentioned value which is similar to
the value given by the specifications for the material used.
The column elements were reinforced with 12Ø20 rebars and
Ø10/100 stirrups. The beam elements were reinforced with 3Ø18
longitudinal bars at the top and the bottom sides and Ø10/100 stir-
rups. Horizontal hooks, fully anchored in the body of the beam,
were also placed in front of the dowels, specifically: 5Ø12/50 at
the lower 0.30 m of the section and 3Ø12/100 at the rest of the sec-
tion. A mechanism made of two steel plates and four high strength
screws was embedded in the rear part of the beam (Fig. 1) for the
fastening with the force application system.
The dowels were fully anchored according to the requirements
of EC2 [34]. It is noted that large axial forces were developed in the
dowels for large shear displacements and in some monotonic tests
the dowels broke in tension. Thus, their proper anchorage was
important in order to ensure that no slippage of the bars would
occur.
2.2. Experimental programme
The experimental investigation was focused on the effect of var-
ious parameters for which decisions have to be taken during the
design process, on the shear resistance of the joint, and on possible
improvements and recommendations for a more efficient design of
pinned connections. The following parameters were investigated
during the experimental course:
 The diameter of the dowels. Dowels of three diameters were
tested, namely: Ø32, Ø25 and Ø16. In practice, dowels of Ø28
and Ø32 are commonly used. However, only few tests were per-
formed with Ø32 dowels, and only for connections with one
dowel, because the strength of connections with 2Ø32 dowels
exceeded the capacity of the actuator. Most of the tests were
performed on connections made of 2Ø25 dowels; in this case,
the resulting strength of the joint was within the limits of the
hydraulic actuator and we were able to bring the specimens
to failure or close to it. Connections with Ø16 dowels are rather
416 I.N. Psycharis, H.P. Mouzakis / Engineering Structures 41 (2012) 413–427

unusual in practice, but a few tests were performed with 2Ø16
dowels in order to investigate the response of connections of
low strength and high deformability.
 The number of dowels. Most tests were performed on connec-
tions with two dowels. However, some tests with connections
with one dowel were performed in order to investigate how
the resistance of the joint is related to the number of dowels
and whether interaction phenomena between adjacent dowels
occur. It should be noted that connections with only one dowel
are not recommended, because they may result in accidents
during construction, due to low torsional resistance and stiff-
ness of the connection to non-symmetrical loading.
 The distance d of the axis of the dowels from the beam edge in
the loading direction (see Fig. 1), which is used as a measure of
the thickness of the cover concrete. In real precast structures,
due to lack of space it is usually desirable to reduce the distance
d as much as possible. Thus, it is typical to have connections
with d being as low as 0.10 m. For this reason, most experi-
ments were performed on specimens with d = 0.10 m. However,
specimens with d = 0.15 m and d = 0.20 m were also tested in
order to investigate how the thickness of the concrete cover
of the dowels affects the resistance of the connection, especially
in the pull direction.
 The strength of the grout that is placed in the dowels’ ducts. As
mentioned above, most of the tests were performed 24 h after
grouting, when the strength of the grout was about 20 MPa.
However, two tests (one monotonic and one cyclic) were per-
formed 1 month after grouting, when the strength of the grout
was much larger, about 44 MPa according to the specifications
for the material used.
In total, 22 tests were performed as shown in Table 1. Speci-
mens for each type of connection were tested under monotonic
loading in pull and/or in push direction and under cyclic excitation.
Most of the cyclic tests started in the pull direction, but in some
cases, two cyclic tests were performed; in these cases, one test
started in the pull direction and one in the push direction.
One test (1D25-d10-CY-F) was performed with the beam spec-
imen made of fibre-reinforced concrete, which, however, did not
improve the behaviour, thus the results of this test are not reported
here.
The monotonic tests were executed with the driving force ap-
plied in the pull or in the push direction up to failure of the dowels
or when the maximum capacity of the actuator was reached. The
cyclic tests were displacement-controlled and were performed at
a rate of 0.2 mm/s. The protocol shown in Fig. 3a was followed:
three cycles were performed at each displacement level, which
was increasing in steps of dy. In some tests, a smaller step, equal
to 0.5 dy, was used during the first stages of the loading procedure.
The yield displacement, dy, was determined from the point of bend
of the force–displacement curve for monotonic pull or monotonic
push loading. In cases in which the value of dy was different in
the pull and in the push direction, the minimum value was consid-
ered. The cyclic tests stopped when significant strength degrada-
tion was observed, more than 20%.
2.3. Instrumentation setup
The horizontal load of the actuator was monitored during the
tests through a load cell and the imposed horizontal displacement
was measured using an internal displacement transducer. Addi-
tionally, two displacement transducers were placed at the top
(D1) and the bottom (D2) of the beam’s front and one at the top
of the column (D3) as shown in Fig. 3b. In general, D1 and D2
recordings were identical to the one of the internal transducer un-
til spalling occurred to the beam. The recording of D3 was practi-
cally zero. The strains that were developed in the dowels at
about 0.10 m below the joint in the column side were measured
with strain gauges that were glued on them.
3. Experimental results
3.1. General remarks
As mentioned above, all tests were performed under pure shear
conditions by imposing horizontal displacement to the beam.
However, in the strength of the connection contributed not only
the shear resistance of the dowels due to their bending (dowel ac-
tion, for details see [25]) but also their strength in tension, since
the dowels were significantly elongated during the movement of
the beam, as shown in Fig. 4a. Most of the deformation of the

Table 1
Experimental data.

Dowels d (m) d/D dn/D Specimen-test Type of test Remarks

2Ø25 0.10 4.00 4.00 2D25-d10-PSH Monotonic Push
2D25-d10-PLL Monotonic Pull
2D25-d10-CY-1 Cyclic Start in push direction
2D25-d10-CY-2 Cyclic Start in pull direction
2D25-d10-CY-PL Cyclic With anchored steel plate
0.15 6.00 4.00 2D25-d15-PLL Monotonic Pull
2D25-d15-CY Cyclic Start in pull direction
0.20 8.00 4.00 2D25-d20-PLL Monotonic Pull
2D25-d20-CY Cyclic Start in pull direction
1Ø25 0.10 4.00 8.00 1D25-d10-PSH Monotonic Push
1D25-d10-PLL Monotonic Pull
1D25-d10-PSH-G Monotonic Push With high strength grout
1D25-d10-CY Cyclic Start in pull direction
1D25-d10-CY-F Cyclic With fibre reinf. concrete
2Ø16 0.10 6.25 6.25 2D16-d10-PSH Monotonic Push
2D16-d10-PLL Monotonic Pull
2D16-d10-CY-1 Cyclic Start in push direction
2D16-d10-CY-2 Cyclic Start in pull direction
2D16-d10-CY-G Cyclic With high strength grout
1Ø32 0.20 6.25 6.25 1D32-d20-PSH Monotonic Push
1D32-d20-PLL Monotonic Pull
1D32-d20-CY Cyclic Start in pull direction
 I.N. Psycharis, H.P. Mouzakis / Engineering Structures 41 (2012) 413–427 417

Fig. 3. (a) Loading protocol for most cyclic tests. In some tests, loading at displacement levels 0.5 dy and 1.5 dy were also applied; (b) Instrumentation setup.

dowels was concentrated around the joint, in a length that ex-
tended a few centimetres inside the beam and the column where
plastic hinges were developed; in this region, the dowels were in-
clined and the grout and the concrete around the ducts and the
dowels were crushed. It should be noted that the tension that
was developed in the dowels led to a decrease in the corresponding
shear force that was induced to them, because the plastic moments
decreased [26]. However, the overall shear resistance of the con-
nection was increasing with the shear displacement, despite the
reduction in the shear force of the dowels. This happened because
the axial force of the dowels was increasing and was contributing
more to the horizontal resistance due to the larger inclination of
the deformed part of the dowels in the region of the joint.
As mentioned, significant tensile stresses were developed in the
dowels for large displacements, causing their plastic elongation
and permanent increase in their length. For monotonic loading,
this was not a problem, since the dowels could be elongated up
to their breaking point. In cyclic tests, however, the dowels had
to ‘‘absorb’’ this elongation during unloading, since the special de-
vice that was used did not allow the upward motion of the beam.
Thus they were forced to slide upwards along their duct as the dis-
placement was decreasing. This was verified from the observed up-
ward displacement of the dowels’ bolts on the top of the beam
(Fig. 4b). The opposite phenomenon was occurring during re-load-
ing in the opposite direction, when the dowels were elongated
again, thus tension forces were developed that forced them to slide
downwards again. The driving force had to overcome the friction
along the ducts and, for this reason, the obtained force–displace-
ment diagrams were not smooth but had the saw-shape shown
with thin blue line in Fig. 4c. In real structures, this phenomenon
is not expected to occur, at least to that extend, because the beams
are restricted to move upwards only by the weight of the super-
structure. In order to eliminate this effect from the experimental
results and obtain smooth curves representative of the expected
behaviour in practice, the original data were filtered by applying
a running average procedure (thick red line in Fig. 4c). All cyclic
curves that are shown in the following correspond to the filtered
data.
It is interesting to mention that it was not always possible to
determine whether the dowels had failed or not during a test. This
happened because breaking of the dowels did not necessarily lead
to loss of resistance, since the breaking point was not located at the
level of the joint, but in a small distance, close to 2D in most cases,
inside the beam or the column. Thus, a portion of the broken dow-
els extruded from the column or the beam inside the opposite ele-
ment (Fig. 5a and b) and continued to pose significant resistance
against horizontal movement. In some cases, the maximum resis-
tance was attained after the failure of the dowels. Due to this phe-
nomenon, the dowels could break again at a second point inside
the opposite element, as observed in some cases (Fig. 5c).

Fig. 4. (a) Deformation of dowels during testing; (b) upward displacement of dowels’ bolts on top of the beam, observed during cyclic tests; (c) saw-type force–displacement
diagrams for cyclic tests (thin blue line) and smoothed response after filtering (thick red line). (For interpretation of the references to colour in this figure legend, the reader is
referred to the web version of this article.)
418 I.N. Psycharis, H.P. Mouzakis / Engineering Structures 41 (2012) 413–427
It is noted that no slippage of the dowels was observed in the
tests performed. Also, although the stirrups in front of the dowels
were bent in some tests in which extensive spalling occurred, they
did not break in any of the experiments.
3.2. Monotonic loading
The force–displacement curves obtained for specimens with
2Ø25 dowels and various values of the distance d of the dowels
from the beam edge are shown in Fig. 6. The maximum force at-
tained during the pull tests was determined by the failure of the
dowels. In the push direction, the test stopped at F = 465 kN, when
the driving force was close to the limit of the hydraulic actuator. At
that point the dowels had not broken, thus the actual resistance
was somehow larger.
Significantly reduced yield force and ultimate resistance were
attained in the pull direction for d = 0.10 m (d/D = 4), compared
to those in the push direction. Compared to the corresponding
resistance for d = 0.20 m (d/D = 8), reduced values by 60% were at-
tained. For the intermediate case of d = 0.15 m (d/D = 6), the yield
force and the ultimate resistance were similar to the ones for
d = 0.20 m (d/D = 8) but they were attained at larger displacements.
This significant reduction in the resistance in the pull direction
for small values of d/D is attributed to the spalling of the cover con-
crete of the dowels, which started early in the test, when the dis-
placement of the beam was quite small (Fig. 7a). At larger
displacements, complete disorganization of the cover concrete
and the concrete in the vicinity of the lower part of the dowels oc-
curred (Fig. 7b). Such extensive damage was not observed for
thicker covers, where only minor cracking happened at the base
of the beam (Fig. 8). It is noted, however, that, even for
d = 0.20 m where the damage was quite small, the response in
the pull direction showed smaller stiffness compared to the push
direction (Fig. 6).
The spalling of the cover concrete and the extensive cracking of
the concrete in the vicinity of the dowels that occurred for small
values of d during loading in the pull direction resulted in a tempo-
rary loss of stiffness, as evident from the ‘‘plateau’’ that is observed
on the corresponding force–displacement diagrams in Fig. 6. The
length of the plateau was generally decreasing as d was increasing.
For large displacements, however, the dowels were restrained by
the horizontal stirrup reinforcement of the beam and the resis-
tance started increasing again, up to the point where the dowels
broke in tension. The horizontal stirrup reinforcement did not
break at any test, but, in many cases, the lower bar was evidently
bent due to the pushing of the dowels.
Damage was also observed at the end of the monotonic push
test. In this case, splitting of the concrete in the normal to the load-
ing direction occurred to the beam and the column (Fig. 9a). Also,
significant splitting occurred at the top of the beam, in the region
Fig. 5. (a) Broken dowels protruding from the column; (b) broken dowels protruding from the beam; (c) dowel broken at two places.
around the dowels, due to the high pressure applied by the bolts
(Fig. 9b).
Comparison of the response of connections made of 2Ø25 and
1Ø25 dowels is presented in Fig. 10. In both cases, the distance d
of the dowels from the beam edge was 0.10 m (d/D = 4). It is seen
that the resistance of the connections with two dowels was less
than two times the corresponding value for the connections with
one dowel. This must be attributed to the interaction of the two
dowels in the transverse direction and the plastic zone that was
developed in the region between them, and to the smaller concrete
cover of the dowels in the normal to the loading direction, dn (see
Fig. 1): for the specimens with two dowels, dn = 0.10 m (dn/D = 4.0),
while for the specimens with one dowel, dn = 0.20 m (dn/D = 8.0).
Note that, for loading in the pull direction, spalling of the cover
concrete occurred also to the specimen with one dowel, similarly
to the specimen with two dowels.
The resistance of joints made of 2Ø16 dowels with d = 0.10 m
and of 1Ø32 dowel with d = 0.20 m is shown in Fig. 11a and b,
respectively. In both cases, d/D = 6.25. For both specimens, the
yield force and the ultimate resistance in the pull direction were
close to the ones in the push direction but the maximum force
was attained at larger displacement and the response was not sym-
metric. A similar behaviour was observed for the connections with
2Ø25 dowels and d = 0.15 m (Fig. 6), for which the ratio d/D had a
similar value (d/D = 6.00).
Concerning the observed damage, spalling of the cover concrete
occurred at the bottom side of the beam during the pull tests,
which, however, was smaller for Ø16 because smaller forces were
developed.
For the specimen with 1Ø32 dowel, after the end of the mono-
tonic push test the load was reversed and applied in the pull direc-
tion. The response is shown in Fig. 11b with blue dashed line and is
similar to the one obtained from the direct monotonic pull test.
This shows that damage in one direction does not affect the re-
sponse in the other direction. This conclusion was verified for other
specimens, too.
In Fig. 12a, comparison is made between connections with
2Ø16 dowels (As = 402 mm2) and 1Ø25 dowel (As = 491 mm2). In
the push direction, the difference in the attained ultimate resis-
tances was practically equal to the corresponding difference in
the cross sections of the dowels (equal to 22%). This shows that
the resistance of the connection made of 2Ø16 dowels was not af-
fected by secondary phenomena, as was the case for the joint with
2Ø25 dowels. Note that, for 2Ø16 dowels, dn/D = 6.25 in the trans-
verse direction and the distance between the dowels was 12.5
times their diameter, values large enough to avoid secondary phe-
nomena. In the pull direction, the connection made of 2Ø16 dowels
showed better response than the one made of 1Ø25 dowel, despite
the fact that the cross section of the dowels was smaller, which is
attributed to the larger value of d/D (equal to 6.25 for Ø16 and to
 I.N. Psycharis, H.P. Mouzakis / Engineering Structures 41 (2012) 413–427 419

Fig. 6. Force–displacement diagrams for monotonic push and monotonic pull tests of connections with 2Ø25 dowels and various values of d.

Fig. 7. Damage to the specimen with connection made of 2Ø25 dowels with d = 0.10 m, observed during the monotonic pull test: (a) at an early stage of the test; (b) at the end
of the test.

Fig. 8. Damage to specimens with connections made of 2Ø25 dowels, observed during monotonic pull tests: (a) d = 0.15 m; (b) d = 0.20 m.

4.00 for Ø25). Specifically, the resistance was larger and the max-
imum force was attained at smaller displacement.
A similar comparison is shown in Fig. 12b between connections
made of 1Ø32 dowel (As = 804 mm2) with d = 0.20 m (d/D = 6.25)
and of 2Ø25 dowels (As = 981 mm2) with d = 0.15 m (d/D = 6.00)
under monotonic loading in the pull direction. In this case, the ra-
tio d/D was about the same in both cases; hence, the ratio of the
attained resistances was close to the ratio of the cross sections of
the dowels. However, the connection with 2Ø25 dowels had a bet-
ter overall response, since the maximum resistance was attained at
significantly smaller displacement.
In general, these comparisons show that the shear resistance in
the push direction is similar for specimens with dowels of the same
cross section, independently of their number and their diameter, as
420 I.N. Psycharis, H.P. Mouzakis / Engineering Structures 41 (2012) 413–427

Fig. 9. Damage to the specimen with connection made of 2Ø25 dowels, observed at the end of the monotonic push test.

Fig. 10. Comparison of force–displacement diagrams for connections with 1Ø25 and 2Ø25 dowels under monotonic push and monotonic pull loading (d = 0.10 m).

Fig. 11. Force–displacement diagrams for monotonic push and monotonic pull tests of connections with: (a) 2Ø16 dowels and d = 0.10 m; (b) 1Ø32 dowels and d = 0.20 m.

far as the thickness of the cover concrete in the transverse direc-
tion and the distance between the dowels are sufficiently large
to prevent secondary effects. In the pull direction, the response is
generally better for connections made of dowels of smaller diame-
ter, other critical parameters being the same.
3.3. Cyclic tests
In all cyclic tests it was observed that the strength of the con-
nection was decreasing with the number of applied cycles at each
level of displacement, which is common in this type of tests and is
 I.N. Psycharis, H.P. Mouzakis / Engineering Structures 41 (2012) 413–427 421

Fig. 12. Comparison of the response to monotonic loading of connections made of: (a) 2Ø16 dowels and 1Ø25 dowel; (b) 1Ø32 dowel and 2Ø25 dowels.

attributed to the increased local cracking of the concrete around
the dowels after each cycle. In general, the strength drop was
around 10% in the second cycle and more than 15% in the third cy-
cle, compared to the first one.
The cyclic response of the connection with 2Ø25 dowels and
d = 0.10 m (d/D = 4) is shown in Fig. 13a. The two curves corre-
spond to different starting direction of the test (push or pull) and
indicate that the behaviour is not affected by the direction of the
first application of the load. The response was not symmetric, with
the resistance in the push direction being almost double the one in
the pull direction. This behaviour was expected from the results of
the monotonic tests and is attributed to the small thickness of the
cover concrete and the resulting reduction in the resistance in the
pull direction. Significant pinching was also observed, especially in
the push direction, which was increasing with the number of ap-
plied cycles due to the disorganization of the concrete around
the dowels and their consequent loosening. The observed damage
was similar to the one for monotonic tests: spalling of the cover
concrete at the front side of the beam during pulling (Fig. 13b)
and cracks in the column in the normal to the loading direction.
It is interesting to note that no significant strength degradation
(larger than 20%) occurred even for quite large shear displace-
ments, up to 30 mm in the push direction and about 20 mm in
the pull direction. If these values are considered as ultimate
displacements, the corresponding shear ductility is ls = 5 and
ls = 3.5, respectively. The ductility capacity of pinned connections
is discussed in Section 4.3.
Comparison between monotonic and cyclic tests is shown in
Fig. 14 for the connection made of 2Ø25 dowels with d = 0.10 m.
Similar were the results for other cases examined. It is seen that
the cyclic response follows the monotonic one up to displacement
at least equal to 2dy. After that point, the cyclic resistance remains
practically constant for several more cycles, while the monotonic
resistance generally increases. For small ratios of d/D, for which
the monotonic strength in the pull direction shows a ‘‘plateau’’
after yielding, the cyclic behaviour follows the monotonic one in
the pull direction up to the point where the monotonic resistance
starts increasing again after the engagement of the horizontal stir-
rup reinforcement (Fig. 14).
The effect of the distance d of the dowels from the beam edge on
the cyclic behaviour is shown in Fig. 15a, where the response of the
connection made of 2Ø25 dowels with d = 0.10 m (thick blue line)
is compared to the one of a similar connection with d = 0.20 m
(thin red line). As expected, the connection with the thicker cover
showed much more symmetrical response and much less pinching.
It should be noted that significant spalling occurred to the speci-
men with d = 0.10 m during pulling, which, however, did not affect
the resistance in the push direction. This verifies the previously

Fig. 13. (a) Force–displacement diagrams for the cyclic tests of connections with 2Ø25 dowels and d = 0.10 m; (b) damage to the beam at the end of the test.
422 I.N. Psycharis, H.P. Mouzakis / Engineering Structures 41 (2012) 413–427
Fig. 14. Comparison of the behaviour of connections with 2Ø25 dowels and
d = 0.10 m for monotonic and cyclic loading.
mentioned observation that damage in one direction does not af-
fect the strength in the other direction.
In Fig. 15b, the cyclic behaviour of the connection made of 2Ø16
dowels (As = 402 mm2) is compared to that of the connection made
of 1Ø25 dowel (As = 491 mm2). In both cases d = 0.10 m, thus d/
D = 6.25 in the former case and d/D = 4 in the latter in which the
response was asymmetric as expected. The comparison of the ob-
tained resistances in the push direction shows that they are anal-
ogous to the cross section of the dowels, similarly to what was
observed for monotonic loading. In the pull direction, similar resis-
tance was attained in both cases despite the difference in the cross
section of the dowels, due to the reduced strength of the connec-
tion with 1Ø25 dowel, caused by the small value of d/D. It is noted
that the connection with 2Ø16 dowels showed significant strength
degradation and small shear ductility (around ls = 2) in both push
and pull directions. This somehow ‘‘strange’’ behaviour is dis-
cussed in Section 4.3. For the connection with 1Ø25 dowel, the at-
tained shear ductility was ls = 4 in the push direction and ls = 2.5
in the pull direction.
In Fig. 16a, the force–displacement curves of connections with
1Ø32 (As = 804 mm2) and 2Ø25 (As = 981 mm2) dowels are com-
pared. In both cases, d = 0.20 m and the ratio d/D was larger than
6; thus, the behaviour was practically symmetric. The ratio of the
attained resistances was equal to the ratio of the cross sections of
the dowels, which shows that there were not any evident secondary
Fig. 15. Comparison of the cyclic behaviour of connections made of: (a) 2Ø25 dowels with d = 0.10 m and d = 0.20 m; (b) 1Ø25 dowel and 2Ø16 dowels with d = 0.10 m.
effects in the case of two dowels, caused by the interaction between
them or by the small value of the ratio dn/D in the transverse direc-
tion (equal to 4). Concerning the shear ductility, values larger than
ls = 4 were attained in both cases and in both directions.
In Fig. 16b, the comparison of the behaviour of connections
made of 1Ø25 dowel and 2Ø25 dowels is presented. It is seen that
the resistance of the connection with one dowel is practically one
half the resistance of the connection with two dowels.
The comparisons made in Fig. 16a and b shows that the effect of
secondary phenomena, as the interaction of the dowels and the
influence of the small concrete cover in the normal to the loading
direction, is less important for cyclic loading than for monotonic
loading. It is reminded that, for monotonic loading, the resistance
of the connections with two dowels was less than two times the
corresponding value for one dowel (Fig. 10) due to these phenom-
ena. The smaller influence of these secondary effects in the case of
cyclic loading should be attributed to the lower forces that were
developed.
4. Design issues and recommendations
4.1. Design of pinned beam-column connections for seismic loads
Pinned connections of precast beams with the supporting col-
umns are usually made using one or two dowels that protrude
from the top of the column and enter into waiting sleeves inserted
in the beams. The sleeves are filled with no-shrinkage mortar to
ensure the anchorage of the dowels by bond, while, in many cases,
the dowels are additionally bolted on their top for extra security.
The formulae that are provided below for the calculation of the
shear resistance of such connections are valid under the assump-
tion that no slippage of the dowels occurs during deformation,
allowing thus for significant axial forces to be developed in the
dowels.
The seismic design of precast structures with pinned beam-col-
umn connections is based on the concept that the prevailing en-
ergy dissipation mechanism should be through plastic rotations
within critical regions of the columns, while the connections re-
main in the elastic region. In Eurocode 8 [35] such connections
are termed ‘‘overdesigned connections’’. Their design is based on
the capacity design rule, according to which the design shear force
Ed that is applied to the connection is derived assuming that the
ultimate flexural resistance has been developed at the base of
the column. The latter is calculated by multiplying the flexural
 I.N. Psycharis, H.P. Mouzakis / Engineering Structures 41 (2012) 413–427
Fig. 16. Comparison of the cyclic behaviour of connections made of: (a) 1Ø32 dowel and 2Ø25 dowels with d = 0.20 m; (b) 1Ø25 and 2Ø25 dowels with d = 0.10 m.
resistance MRd with the overstrength factor cRd. In EC8, cRd is set
equal to 1.20 for DCM and to 1.35 for DCH.
The ability of the connection to sustain the induced seismic
loads is verified if the design force Ed is smaller than the shear
resistance Rd, i.e. if the classical inequality of the codes holds:
E d  Rd
The estimation of the shear resistance Rd is discussed in the follow-
ing section.
4.2. Estimation of the shear resistance
For the seismic design of pinned connections in praxis, and
since specific equations for the estimation of the shear resistance
of the joints made of dowels are not available, various empirical
formulae that give the shear strength of dowels are usually applied.
Such a formula, which is often used in practice, is the one proposed
by Vintzeleou and Tassios [26], which was derived for dowels
embedded in concrete without confinement. According to this for-
mula, the shear resistance of one dowel of diameter D, with thick
enough concrete cover, specifically larger than 6–8 times its diam-
eter, is estimated by the following equations (without considering
any partial safety factors for the materials):
qffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Ru ¼ 1:30  D2  fck  fyk for monotonic loading
Rcy ¼ 0:5  Ru for cyclic loading
where fck and fyk are the characteristic strength of the concrete and
the characteristic yield stress of the steel of the dowels, respec-
tively. For smaller cover thickness, another formula is provided in
[26] based on the tensile strength of the cover concrete, which,
however, does not apply to our case because of the existence of
the stirrups.
As mentioned above, Eqs. (2) and (3) were derived for uncon-
fined concrete. The shear strength was considered equal to the
shear force that corresponds to the yield of the dowel or the crush-
ing of the concrete in front of it. Also, it was calculated for rela-
tively small displacements, before any strain hardening of the
dowels occurs, and for concrete blocks being practically in contact
(without the gap of the elastomeric bearing). Hence, they do not
reflect the real conditions of pinned connections that are encoun-
tered in typical precast structures.
Compared to the results of the experimental investigation re-
ported herein, these formulae underestimate the real resistance
of pinned connections, as shown in Table 2. In this comparison,
423
we substituted fck with fcm in Eq. (2), where fcm is the mean strength
of the concrete of each specimen, as determined from compression
tests on cylindrical samples that were taken during concreting.
Also, the measured yield strength of the steel of the dowels was
used instead of the nominal one.
In Fig. 17, the obtained experimental values of the ultimate
ð1Þ
resistance of all the connections for monotonic and cyclic loading
are plotted against the ratio d/D.
pThe resistance
ffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi is normalized with
respect to the quantity: n  D2  fcm  fyk , where n is the number of
dowels. In the push direction, d/D was calculated for d = 0.40 m,
which corresponds to the position of the dowels in the column
(see Fig. 1).
It is evident that the normalized resistance for cyclic loading is
practically constant for d/D > 6.00, equal to 1.10. It is noted that the
corresponding value according to Eq. (3) is 0.65 (shown with
dashed blue line in Fig. 17). It seems, therefore, that the experi-
mental results could be matched by Vintzeleou and Tassios for-
mula, if the coefficient 0.65 is increased to 1.10.
Based on these observations and substituting back fcm with the
characteristic strength of the concrete fck, the ultimate shear resis-
tance under cyclic loading of pinned connections with sufficient
cover thickness can be estimated by:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Ru ¼ 1:10  n  D2  fck  fyk for d=D > 6 ð4aÞ
ð2Þ
where n is the number of the dowels. For d/D = 4 it is suggested that
one half of this value is used, while linear interpretation is applied
ð3Þ
for intermediate values of d/D. Then, one can write:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Ru ¼ 1:10  ð0:25d=D  0:50Þ  n  D2  fck  fyk for 4
 d=D  6 ð4bÞ
It is not recommended to use values of d/D < 4. Eqs. (4a) and (4b)
(with fcm in place of fck) are shown in Fig. 17 with solid blue line.
In the above equations, fck and fcm refer to the compression
strength of the concrete and not of the grout. The strength of the
grout was measured at 24 h and was found about 23 MPa, i.e. less
than the strength of the concrete for the tests that are included in
Fig. 17. In this sense, these formulae are in accordance with fib Bul-
letin 43 [5], where it is proposed that the larger strength among
the concrete and the grout must be used for the calculation of
the shear resistance of the connection. This issue is further dis-
cussed in Section 4.4.
It is reminded that all the experiments were performed for pure
shear deformation of the connections, preventing any rotation at
the joints. In real structures, however, the deflection of the col-
424 I.N. Psycharis, H.P. Mouzakis / Engineering Structures 41 (2012) 413–427

Table 2
Comparison of the experimental ultimate resistance of the connections with the estimation according to the formula by Vintzeleou and Tassios [26].

Dowels d (m) fcm (kPa) fyk (kPa) Type of test Experimental resistance (kN) Formula in [26] (kN)
Pull Push Pull Push
2Ø25 0.10 34,832 580,000 Monotonic Push 465 231
Monotonic Pull 275 –
Cyclic 115 200 – 115
2Ø25 0.15 30,000 580,000 Monotonic Pull 470 214
2Ø25 0.20 30,130 580,000 Monotonic Pull 450 215
Cyclic 180 200 107 107
1Ø25 0.10 33,530 580,000 Monotonic Push 270 113
36,560 Monotonic Pull 170 –
35,200 Cyclic 75 95 – 57
2Ø16 0.10 35,042 558,000 Monotonic Push 230 95
Monotonic Pull 185 95
Cyclic 70 70 47 47
1Ø32 0.20 29,918 519,000 Monotonic Push 405 175
Monotonic Pull 400 175
Cyclic 150 150 88 88

umns during strong seismic excitations will produce rotations at
the joints, the amount of which depends on the flexibility of the
columns. In such cases, the resistance of the connections might
be reduced, due to the additional tension that will be induced to
the dowels by the joint rotation. Cyclic tests on frames performed
at the University of Ljubljana [32] and dynamic shaking table tests
performed at the NTUA [33] within the SAFECAST project showed
that the resistance of pinned connections might be reduced as
much as 20% if large rotations occur at the joints. In such cases,
it is proposed to reduce the coefficient 1.10 in Eqs. (4a) and (4b)
accordingly.
For the seismic design of structures with pinned connections
according to the aforementioned capacity concept (Eq. (1)), the
shear resistance of the connections can be calculated using Eqs.
(4a) and (4b) by substituting fck and fyk with their design values:
fcd = fck/cc and fyd = fyk/cs, respectively, where cc and cs are the par-
tial safety factors for concrete and steel. The proposed values for
these coefficients in Eurocodes 2 and 8 are: cc = 1.50 and cs = 1.15.
Based on the above observations, the shear resistance of pinned
connections to be used in the seismic design of precast structures
according to Eq. (1) can be calculated by the following equations:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffi
C0
Rd ¼  n  D2  fcd  fyd for d=D > 6:00
cR
qffiffiffiffiffiffiffiffiffiffiffiffiffiffi
C0
Rd ¼  ð0:25d=D  0:50Þ  n  D2  fcd  fyd for 4:00
cR
 d=D  6:00
For the coefficient C0, values varying from 0.90 to 1.10 are sug-
gested, depending on the magnitude of the expected joint rotations:
for flexible columns, for which large joint rotations may occur, a va-
lue of C0 around 0.90–0.95 is suggested, while for stiff columns and
walls, for which small joint rotations are expected, this coefficient
can be increased. The maximum value of C0 is 1.10 for practically
zero joint rotations, as was the case in the experiments.
In Eqs. (5a) and (5b), a general safety factor cR was added to ac-
count for the uncertainties in the experimental procedure and the
limited number of experimental data used in the derivation of
these formulae. It is suggested that cR  1.30, as proposed in [5].
It should be noted, however, that another safety factor, which con-
cerns the over-strength of the flexural resistance of the column, is
applied in the calculation of the design load Ed that is used for the
verification of the resistance of the connection according to Eq. (1).
In order to avoid being over-conservative, one must consider these
two safety factors together: if a large over-strength factor is used
for the determination of Ed, then a value of cRd less than 1.30 could
be used in Eqs. (5a) and (5b).
It must be noted that Eqs. (5a) and (5b) are valid under the fol-
lowing conditions: the dowels are properly anchored and bolted on
their top; the ducts are grouted with concrete of about the same
strength with the concrete of the precast members; adequate stir-
rup reinforcement is placed in front of the dowels; and there is a
distance of about 20 mm between beam and column at the joint,
which is achieved in practice by the elastomeric pad. Additionally,
these equations are valid assuming that no significant axial forces
are applied to the dowels by other actions, which might result in a
reduction of the resistance. Due to the flexibility of the elastomeric
pads, such axial forces can be developed in the dowels by the addi-
tional loads of the roof applied after grouting of the ducts. They can
also be developed during an earthquake due to the vertical compo-
nent of the ground motion and the seismic moment in the trans-
verse direction of the beam induced in connections with two
dowels.
4.3. Shear ductility capacity
ð5aÞ
As already mentioned, pinned connections are designed with
the capacity rule, thus they are expected to behave elastically dur-
ing earthquakes. Therefore, their shear ductility capacity is not
used in the seismic design but can be thought as a reserve in case
ð5bÞ that significant unexpected shear displacements are developed.
The shear ductility capacity, ls, of the connections was calcu-
lated from their cyclic response, considering as ultimate displace-
ment the one that corresponds to 20% drop in strength. As
reported above, if the distance of the dowels from the beam edge
was large enough (d/D P 6) quite satisfactory values of ls were at-
tained in most cases, ranging from 4 to 6. For d/D = 4, significantly
smaller values of ls were attained in the pull direction, between
2.5 and 3.5.
It must be noted, however, that low values of ls, around 2, were
observed in some cases, as for connections with 2Ø16 dowels
(Fig. 15b) and for connections with 2Ø25 dowels and a steel plate
in front of them (Fig. 18a, thin red line). This somehow ‘‘strange’’
behaviour must be attributed to the lesser cracking that occurred
in front of the dowels in these cases, due to the small diameter
of the dowels (in the former case) or the intense confinement pro-
vided by the plate (in the latter case). As a result, the dowels were
 I.N. Psycharis, H.P. Mouzakis / Engineering Structures 41 (2012) 413–427 425

Fig. 17. Normalized ultimate strength of pinned connections for monotonic (red) and cyclic (blue) loading. (For interpretation of the references to colour in this figure legend,
the reader is referred to the web version of this article.)

bent in sharper angles (see deformation of the dowels in Fig. 4a)
and increased strains were developed for the same displacement
of the beam, compared to the cases, in which more extensive
cracking of the concrete around the dowels occurred. Due to this
phenomenon, the dowels yielded earlier and the shear strength
dropped at smaller displacement leading to reduced ductility.
4.4. Design recommendations
As evident from the experimental results, provision of suffi-
ciently thick cover of the dowels in the loading direction is very
important, since the resistance in the pull direction can be signifi-
cantly reduced for small values of d/D. For this reason, it is recom-
mended to use values of d/D larger than 6. It is also recommended
that strong and dense horizontal hooks are placed in front of the
dowels, in order to restrain dowels’ motion if spalling of the cover
concrete occurs.
In many cases encountered in practice, there are limitations in
the value of d, which arise from geometric restrictions in the struc-
ture. In such cases, in which the available distance d is small and it
is not possible or desired to increase it, a steel plate, properly an-
chored, can be placed in front of the dowels, in order to restrict
their deformation in the pull direction. One cyclic test was per-
formed with a connection made of 2Ø25 dowels with d = 0.10 m
and the addition of such a plate. The response is presented in
Fig. 18a and is compared to the corresponding response of the con-
nection without the plate. It is seen that the presence of the plate
increased the resistance in the pull direction resulting in practi-
cally symmetric behaviour with less pinching. Also, no significant
damage occurred in the beam with the steel plate (Fig. 18b).
The resistance of pinned connections can also be increased if
grout of high strength is used. As mentioned above, most of the
tests were performed 24 h after grouting, when the strength of
the grout was somehow larger than 20 MPa. However, two tests
(one monotonic and one cyclic) were performed a month after gro-
uting, when the strength of the grout was much larger, about
44 MPa according to the specifications for the material used.
The results of these tests are presented in Fig. 19 and are com-
pared to the corresponding results for grout of normal strength. It
is evident that the behaviour of the connection was improved
when grout of high strength was used: in the monotonic push test,
larger resistance was attained for the same displacement; for cyclic
loading, better hysteretic loops were obtained, no pinching was ob-
served, the attained ductility was larger and the ultimate resis-
tance was slightly increased. It is noted that, in practice, the
importance of the quality of the grout is sometimes underesti-
mated and grout of low strength is used, especially if the dowels
are bolted on their top.
As mentioned above, in fib Bulletin 43 [5] it is proposed that the
larger strength among the concrete and the grout must be used for

Fig. 18. (a) Comparison of the cyclic behaviour of connections made of 2Ø25 dowels with d = 0.10 m with and without steel plate; (b) no significant damage occurred to the
specimen with the steel plate. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
426 I.N. Psycharis, H.P. Mouzakis / Engineering Structures 41 (2012) 413–427
Fig. 19. Increased resistance achieved by use of high strength grout: (a) monotonic push of connections made of 1Ø25 dowel; (b) cyclic response of connections made of 2Ø16
dowels.
the calculation of the shear resistance of the connection. The re-
sults depicted in Fig. 19 show that this proposal is generally true,
but it was not possible to verify it quantitatively, since the exact
strength of the grout was not measured. For this reason, and taking
under consideration that, usually, there are uncertainties in the ac-
tual strength of the grout, it is suggested that this overstrength
should not be considered in the design and that Eqs. (5a) and
(5b) must be applied using the compression strength of the con-
crete and not of the grout.
5. Conclusions
The experimental investigation of the shear resistance of pinned
beam-to-column connections under monotonic pull, monotonic
push and cyclic loading is reported. The main conclusions can be
summarized as follows:
 The thickness of the cover concrete of the dowels in the direc-
tion of the loading plays an important role to the response.
For dowels placed close to the edges (small values of d/D), spall-
ing of the cover concrete occurs during pulling, which reduces
the resistance and leads to asymmetric response.
 For monotonic loading in the push direction, the force–displace-
ment diagram shows an almost elastic branch up to yielding,
followed by a post elastic branch with significant stiffness,
which is attributed to the horizontal component of the axial
force of the dowels, which increases with the shear displace-
ment, since the inclination of the dowels in the area of the joint
increases.
 The resistance of the connection for cyclic response is less than
one half the monotonic one. In general, the force–displacement
curves for cyclic loading follow the monotonic ones up to two
times the yield displacement. After that point, the strength for
cyclic loading does not increase further.
 The cross section of the dowels is the main parameter that
determines the resistance of the joint. Secondary effects, related
to the normal distance of the dowels from the edges (ratio dn/D)
and the interaction between the dowels in the transverse direc-
tion of the beam seem to be important for large shear forces
only, close to failure for monotonic loading. Such effects were
not observed for cyclic loading.
 Use of high strength grout increases the resistance of the con-
nection and improves the cyclic response by decreasing pinch-
ing and increasing ductility.
 Significant values of shear ductility, around 4–6, can be
achieved by dry pinned connections, provided that the cover
of the dowels is sufficiently thick. This ductility capacity, how-
ever, is used only as a reserve of the connections for better seis-
mic performance. In the seismic design, yielding is allowed at
the columns only, while the connections are designed according
to the capacity criterion.
 Failure of the dowels does not necessarily imply loss of resis-
tance, because broken dowels usually protrude inside the oppo-
site element and resist the horizontal movement. Due to this
behaviour, the dowels may break again at a second point. This
phenomenon, however, might be downgraded in frame struc-
tures with flexible columns, because the rotation of the connec-
tion may lead to disengagement of the dowels.
 Based on the experimental results, a formula is proposed for the
estimation of the shear resistance of pinned connections that
can be used in the seismic design.
Acknowledgements
The research presented herein was conducted in the framework
of the FP7 project ‘‘SAFECAST: Performance of innovative mechan-
ical connections in precast building structures under seismic con-
ditions’’, Research for SME Associations, Grant Agreement
number 218417, which was coordinated by Dr. Antonella Colombo
from ASSOBETON, Italy. The financial support provided by the
Commission of the European Communities through this project is
greatly appreciated. Special thanks are due to Professor S. Tsoukan-
tas of NTUA for many valuable advices and suggestions. Georgia
Kremmyda, graduate students at NTUA, and Dr Lucia Karapitta
had significant contribution in the design of the specimens and
the processing of the data. The help of civil engineer Tryfon Topint-
zis with the design of the specimens was also important. Special
thanks are due to the Greek companies Interbeton S.A., Asprokat
S.A., EKKAF S.A. and Sintecno S.A. for supplying materials for the
construction of the specimens and to Chalyvourgiki S.A. for per-
forming the strength tests on the steel bars.
References
[1] Park R, et al. Seismic design of precast concrete structures – state of the art
report. fib Bulletin 27; 2003.
[2] Elliott KS. Research and development in precast concrete framed structures.
Prog Struct Eng Mater 2000;2(4):405–28.
 I.N. Psycharis, H.P. Mouzakis / Engineering Structures 41 (2012) 413–427
[3] Simeonov S, et al. Building construction under seismic conditions in the Balkan
region – design and construction of prefabricated reinforced concrete building
systems. UNDP/UNIDO, project RER/79/015; 1985.
[4] American Technology Council (ATC). Proceedings of a workshop on design of
prefabricated concrete buildings for earthquake loads. Report No. ATC-08;
1981.
[5] International Federation for Structural Concrete (fib). Structural connections
for precast concrete buildings. Bulletin 43; 2008.
[6] Priestley MJN. The PRESSS program – current status and proposed plans for
phase III. PCI J 1996;41(2):22–40.
[7] Nakaki SW, Standon J, Sritharan S. An overview of the PRESSS five story precast
test building. PCI J 1999;44(2):26–39.
[8] Stanton JF. The PRESSS program in the USA and Japan – seismic testing of
precast concrete structures. Proceedings of the COST C1 international
conference on control of the semi-rigid behaviour of civil engineering
structural connections, Liege; 1998. p. 13–24.
[9] Shiohara H, Watanabe F. The Japan PRESS precast concrete connection design.
Proceedings f the 12th WCEE, Aukland; 2000.
[10] COST C1. Control of the semi-rigid behaviour of civil engineering structural
connections. Final Report European Commission EUR 19244; 1999.
[11] Virdi K, Ragupathy R. Tests on precast concrete subframes with semi-rigid
joints. Proceedings of the 1st COST C1 state of the art workshop on semi-rigid
behaviour of civil engineering structural connections, Strasbourg; 1992. p.
120–31.
[12] Virdi K, Ragupathy R. Analysis of precast concrete frames with semi-rigid
joints. Proceedings of the 1st COST C1 state of the art workshop on semi-rigid
behaviour of civil engineering structural connections, Strasbourg; 1992. p.
296–307.
[13] Elliott KS. Semi-rigid connections in precast concrete structures and bridges.
Proceedings of the COST C1 international conference on control of the semi-
rigid behaviour of civil engineering structural connections, Liege; 1998. p. 3–
12.
[14] De Chefbedien A. Precast beam-to-column head connections. Proceedings of
the COST C1 international conference on control of the semi-rigid behaviour of
civil engineering structural connections, Liege; 1998. p. 35–44.
[15] Elliott KS, Davies G, Mahdi AA, Gorgun H, Virdi K, Ragupathy P. Precast
concrete semi-rigid beam-to-column connections in skeletal frames.
Proceedings of the COST C1 international conference on control of the semi-
rigid behaviour of civil engineering structural connections, Liege; 1998. p. 45–
54.
[16] Jolly CK, Guo M. Application of numerical analysis to connection in precast
concrete frames. Proceedings of the COST C1 international conference on
control of the semi-rigid behaviour of civil engineering structural connections,
Liege; 1998. p. 65–74.
[17] Imai H, Castro JJ, Yanez R, Yamaguchi T. A new precast system for frame
structures and its structural characteristics. In: Dhir, RK, Jones MR, editors.
Concrete 2000. London: E & N Spon.; 1993. p. 655–70.
[18] Watanabe F. Seismic design for prefabricated and prestressed concrete
moment resisting frames. Proceedings of the PCI/FHWA/FIB international
symposium on high performance concrete, Florida, USA; 2000. p. 820–9.
427
[19] Priestley MJN, Tao J. Seismic response of precast prestressed concrete frames
with partially debonded tendons. PCI 1993;38(1):58–69.
[20] Nakaki SD, Englekirk RE, Plaehn JL. Ductile connectors for a precast concrete
frame. PCI 1994;39(5):46–59.
[21] Ersoy U, Tankut T. Precast concrete members with welded plate connections
under reversed cyclic loading. PCI 1993;38(4):94–100.
[22] El Debs MK, El Debs ALH, Miotto AM. Experimental analysis of Beam-
to-Column connection with semi-rigid behaviour of precast concrete
structures. Proceedings of the 2nd fib congress, Naples, Italy; 2006.
[23] Rahman Abd AB, Ghazali AR, Hamid Abd Z. Comparative study of monolithic
and precast concrete beam-to-column connections. Malaysian Constr Res J
2008;2(1):42–55.
[24] Leong DCP. Testing of pinned beam-to-column connections of precast concrete
frames. Master thesis, University Technology, Kuala Lampur, Malaysia; 2006.
[25] Vintzeleou EN, Tassios TP. Mechanics of load transfer along interfaces in
reinforced concrete, prediction of shear force vs. shear displacement curves.
Studie de Reserche, No. 7, Corpo di Perfezionamento per le Construzioni in
Cemento Amato, Italcementi Societa per Azioni (S. p:A), Bergamo, Italy; 1985.
p. 121–61.
[26] Vintzeleou EN, Tassios TP. Behavior of dowels under cyclic deformations. ACI
1987;84(1):18–30.
[27] Tsoukantas SG, Tassios TP. Shear resistance of connections between reinforced
concrete linear precast elements. ACI 1989;86(3):242–9.
[28] Negro P, Mola E, Ferrara L, Zhao B, Magonette G, Molina J. PRECAST EC8:
seismic behaviour of precast concrete structures with respect to Eurocode 8.
Final report of the experimental activity of the Italo-Slovenian Group, Parts 1,
2, 3, FP6 Project No. G6RD-CT-2002-00857; 2007.
[29] Carydis PG, Psycharis IN, Mouzakis HP. PRECAST EC8: seismic behaviour of
precast concrete structures with respect to Eurocode 8. Final report of the
contribution of LEE/NTUA, FP5 Project No. G6RD-CT-2002-00857; 2007.
[30] Psycharis IN, Mouzakis HP, Carydis PG. Experimental investigation of the
seismic behaviour of prefabricated RC structures. Proceedings of the 2nd fib
congress, Naples, Italy; 2006.
[31] Kramar M, Fischinger M, Isaković T. Seismic vulnerability of the EC8 designed
columns in industrial buildings. Proceedings of the 2nd fib congress, Naples,
Italy; 2006.
[32] SAFECAST. Experimental behaviour of existing connections. Deliverables 2.1.,
FP7 Project No. 218417; 2012.
[33] SAFECAST. Quantification of the effects of dynamic loads. Deliverables 2.2., FP7
Project No. 218417; 2012.
[34] European Committee for Standardization (CEN). Eurocode 2: design of
concrete structures – Part 1–1: general rules and rules for buildings. EN
1992–1-1, Brussels; 2004.
[35] European Committee for Standardization (CEN). Eurocode 8: design of
structures for earthquake resistance – Part 1: general rules, seismic actions
and rules for buildings. EN 1998–1, Brussels; 2004.