Engineering Structures 52 (2013) 179–191

Contents lists available at SciVerse ScienceDirect

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

Cyclic failure analysis of the beam-to-column dowel

connections in precast industrial buildings
Blaž Zoubek a,⇑, Tatjana Isakovic a, Yasin Fahjan b, Matej Fischinger b
a
University of Ljubljana, Faculty of Civil and Geodetic Engineering, Jamova 2, 1000 Ljubljana, Slovenia
b
Gebze Institute of Technology, Department of Earthquake and Structural Engineering, Istanbul Caddesi 141, 41400 Gebze, Turkey

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

Article history: The dowel type of the connection is the most common in Europe. However, the knowledge about its seis-
Received 9 November 2012 mic behaviour was incomplete and poorly understood. To analyse the failure of dowel mechanism the
Revised 30 January 2013 numerical model in the FEA software ABAQUS was defined and calibrated using the results of the exper-
Accepted 19 February 2013
imental investigations. Cyclic as well as monotonic response was analyzed. The most important observa-
Available online 28 March 2013
tions are: (1) standard theory assuming that the failure mechanism is initiated by flexural yielding of the
dowel and crushing of the surrounding concrete has been confirmed, (2) the strength of the connection
Keywords:
considerably depends on the depth of the plastic hinge in the dowel, (3) in the case of the cyclic loading
Dowel connection
Precast industrial buildings
the strength is reduced due to the smaller depth of the plastic hinge, (4) neoprene bearing pad can con-
Seismic response siderably increase the strength of the connection, particularly when large relative displacements between
Failure mechanism the beam and the column are developed, and (5) in the case of large rotations between the beam and the
Failure analysis column, cyclic resistance is reduced by 15–20%, because the dowel is loaded not only in flexure but also in
tension.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction Consequently, in the classic state-of-the-art reports on the behav-

iour of precast structures [5,6] this topic was seldom addressed
Field reconnaissance reports [1] have got important impact on and incompletely covered.
the development of earthquake engineering and related seismic The lack of knowledge reflected in the seismic codes and in par-
codes. While each earthquake reconfirms the well-established ticular in the design practice. When the capacity design had be-
knowledge about the behaviour of standard structural systems, come mandatory it became obvious that the existing knowledge
individual precast systems are so specific that the information of about the behaviour of realistic dowel connections was insufficient
their behaviour in past earthquakes has been sparse or even non- to duly fulfil the requirements of the codes. Therefore the 7th EU
existent. Moreover, the information is frequently contradictive as Framework research project SAFECAST – ‘‘Performance of innova-
it obviously depends on the specifics of a precast system (first of tive mechanical connections in precast building structures under
all on its connections) and on the characteristics of earthquakes. seismic conditions (2009–2012)’’ was initiated [7] to investigate
Catastrophic behaviour of precast frame buildings was reported the behaviour of different types of connections in precast industrial
for example after the Spitak, Armenia (1988) earthquake [2]. This buildings. The results became available [7–10] just in time, when
and similar events have led to considerable mistrust against pre- the Emilia Romagna (2012) earthquake hit the highly industrial-
cast systems. Yet some events, like the Friuly (1976) earthquake ized area in Italy, where hundreds of precast buildings addressed
[3] and Montenegro (1979) earthquake [4], demonstrated accept- in this paper were affected. The authors visited the area to inspect
able performance of single-storey industrial buildings with the damage (Fig. 1).
beam-to-column dowel connections. However, these limited While there were some similar cases of damage in the recent
observations were difficult to generalize as detailed analyses of earthquakes in Turkey (1999 and 2011; [11,12]) and l’Aquila
realistic dowel connections in precast buildings were not done. Re- (2010; [13]), this was the first time in the history that such large
search was only concentrated on pure shear behaviour of dowels number of the buildings investigated in this paper was exposed
embedded into two unconfined concrete blocks (see Section 2). to the strong earthquake with a rather broad frequency spectrum.
Major problems were related to the cladding-to-structure connec-
tions (the related new EU project SAFECLADDING is just due to
⇑ Corresponding author. Tel.: +386 41494577. start on August 1st, 2012) and pure friction connections. However,
E-mail address: [email protected] (B. Zoubek). also up-to-date dowel connections in relatively new precast

0141-0296/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.engstruct.2013.02.028
180 B. Zoubek et al. / Engineering Structures 52 (2013) 179–191

If the dowel is located relatively far from the edges of the con-
nected beam and column (the distance from the edge is more than
six diameters of the dowel), it can be assumed that the strength of
the dowel is reached at simultaneous yielding of the dowel and
crushing of the surrounding concrete due to ductile behaviour of
both materials [16,20,21,23,24] (see Fig. 2).
If the concrete compressive strength, the steel yield strength
and the diameter of the dowel are known, the following expres-
sion, according to [14,15], can be used to analytically evaluate
the ultimate resistance of the dowel connection at monotonic
loading:
qffiffiffiffiffiffiffiffiffiffiffiffi
2
Ru;m ¼ F u;m ¼ 1:3  db  fcc  fy ; ð1Þ

where fcc (MPa) is the uniaxial compressive strength of concrete, fy

(MPa) is the yield strength of steel.
Several approximate assumptions were necessary to derive for-
Fig. 1. Collapsed industrial precast building after the May 2012 Emilia earthquake.
mula (1). First, the dowel was considered as a long pile in a cohe-
sive soil. When a concentrated load is imposed on the horizontal
buildings were damaged. Therefore the SAFECAST project results surface of an infinitely extending homogenous and isotropic body,
were accomplished just in time to provide effective tools for the the bearing capacity of the cohesive material is much higher than
post-earthquake analyses. its uniaxial compressive strength [14,25]. According to Broms [26],
In the SAFECAST project special attention was given to the the compressive stress imposed on the soil by a horizontally
beam-to-column connections, which represent most frequently loaded pile has a value equal to approximately 10c (where c is
used type of connections in Europe. Experiments without compar- cohesion). If we assume that concrete is a cohesive material
ison in the past were performed. Sixteen cyclic tests of realistic (fcc = 0.5c) the maximum concrete compressive stress at failure,
connections at small and large relative rotations between the beam fcc is estimated to be equal to fcc ¼ 5f cc , where fcc is concrete uniax-
and the column were tested at the University of Ljubljana ial compressive strength [15]. If the conditions are non-symmetri-
(Section 3) and shake-table tests as well as cyclic tests were per- cal (different concrete strengths in the column and in the beam)
formed at the University of Athens [10]. Improved formulas for the concrete compressive strength on the stronger side is used
the evaluation of the connection capacity were proposed for the estimation of the capacity of the connection [17].
(Section 2; [10]). However, the proposal was more or less empiri- Formula (1) was calibrated for systems that consisted of only
cally based, and although a very large number of tests was done, two concrete blocks with no reinforcement. In real precast struc-
all possible variations of such connections in realistic buildings still tures dowel connections are more complex. A neoprene bearing
greatly exceed the number of parameters actually tested. Also the pad is usually placed between the column and the beam and the
research was mainly concentrated on the global behaviour of the concrete within the connection is considerably confined. Further-
connection (its strength, deformation capacity and hysteretic more, large relative rotations between the beam and the column
behaviour), while the detailed investigation of the failure mecha- are expected in the case of strong seismic loading inducing yielding
nism on the component level was not fully accomplished and the of the column. Considerable loss in the strength capacity of the
behaviour was not fully understood. connection is associated with these large relative rotations.
Therefore a detailed (FEM) numerical tool was needed to de- Expression (1) is appropriate only for monotonic loading. In the
scribe and explain the behaviour of the individual components of case of cyclic loading, the capacity of the connection is notably
dowel connections (the confined parts of the column and beam, lower due to the cyclic degradation of concrete and steel. In [14]
the dowel itself, the steel tube around the dowel, the infill within the following formula is proposed to account for the decrease of
the steel tube and the elastomeric pad) and first of all the complex the dowel ultimate resistance in the case of cyclic loading:
degrading mechanisms of the interaction between these compo- qffiffiffiffiffiffiffiffiffiffiffiffi
2
nents. The proposed and applied model is presented in Section 4. Ru;c ¼ 0:5  Ru;m ¼ 0:65  db  fcc  fy ; ð2Þ
In Section 5 the experimentally observed behaviour (presented in
Based on the results of the experiments performed in the frame
Section 3) is successfully numerically modelled and the key param-
of the SAFECAST project (see [7,10] and Section 3), a modified for-
eters are identified. These parameters and their influence on the
mulas have been proposed, which account for cyclic behaviour of
behaviour of the connections and their failure mechanism are
the realistic beam-to-column dowel connections:
studied and explained in detail in Section 6. The authors believe
qffiffiffiffiffiffiffiffiffiffiffiffi
that the failure mechanism of realistic beam-to-column dowel con- 2
Ru;sr ¼ 1:1  db  fcc  fy ; ð3Þ
nections as affected by different construction details and parame-
ters is now adequately understood and that the effective tool qffiffiffiffiffiffiffiffiffiffiffiffi
2
needed for the analysis of the behaviour observed in the recent Ru;lr ¼ 0:9  db  fcc  fy ; ð4Þ
earthquakes (first of all in the most recent one in Emilia Romagna)
have been provided. where Ru,sr is the ultimate resistance of the connection if small rota-
tions between beam and column are expected, Ru,lr is the ultimate
2. Previous studies of the dowel failure mechanism resistance of the connection if large rotations between beam and
column are expected.
Behaviour of the precast beam-to-column connections analyzed It should be noted that the expressions (3) and (4) predict sub-
in this paper is mainly characterized by the dowel action mecha- stantially higher resistance than formula (2). However, expressions
nism for which simplified numerical models assuming idealized (3) and (4) are predominantly empirical and no detailed analysis of
conditions have already been developed and experimentally tested the failure mechanism leading to this result was done within the
in some previous studies [14–22]. SAFECAST project. Therefore the understanding of the behaviour
 B. Zoubek et al. / Engineering Structures 52 (2013) 179–191 181

Fig. 2. Failure mode of the dowel mechanism.

was incomplete and consequently the generalization of the for- The test was displacement controlled. Displacement amplitudes
mula to the cases not tested within the project was complex. For were varied between 1 mm and 22 mm. For every single ampli-
this reason some sophisticated finite element models were needed. tude, three full cycles were performed. Displacement amplitude
Only a few finite element analysis investigating dowel mecha- was gradually increased up to the failure. Firstly the amplitude
nism can be found in the past studies [27–30]. Maitra et al. [27] increment was equal to 2 d1 = 2 mm. Then it was increased to 3
performed 3D finite element analysis of the load transfer in a dow- d1 = 3 mm and finally to 4 d1 = 4 mm. This increment was then kept
el bar system in jointed concrete pavement. The authors realized the constant.
the importance of modelling the contact between the dowel and Vertical steel dowel of diameter £ = 28 mm was located at the
the surrounding concrete. Zero-length elements were used to mod- centre of the column cross-section. The dowel was anchored deep
el the interaction between the dowel and the concrete. The ele- into the body of the column (90 cm) and protruded into the steel
ments were capable of resisting only compression in the socket within the beam (Fig. 4). The empty space between the
direction normal to the contact surface and frictional force in the dowel and the socket was filled with a fine non-shrinking grout
tangential direction. (fck = 15–20 MPa). ‘‘The neoprene pad (400/220/10 mm) was
Guezouli and Lachal [28] concluded that the contact definition placed between the column and the beam in order to enable the
between the dowel and the concrete has an important impact on relative rotations between the elements. Steel used for the dowel
the response of the stud connection. They analyzed frictional con- met the requirements for the quality B500B according to the ISO
tact effect in push-out tests of the shear connection between pre- 15630-1:2010 standard [31]. The concrete of class C35/40 was pro-
fabricated concrete slab and steel girder in composite bridges. 2D vided in the beam and the column.’’
nonlinear finite element model was used in the analysis. In the pa- Two different types of columns were used in the tests (Table 1).
per, a parametric study of the influence of the friction coefficient The behaviour at small relative rotations between beams and col-
on the load-slip behaviour of the specimen and the distribution umns was studied using stiffer columns with strong reinforcement
of internal deformations and forces is presented. (S1-1 and S1-2). These columns remained elastic during the tests
Nguyen and Kim [29] analyzed large stud shear connectors and their top rotations were small. More flexible and first of all
using ABAQUS software. Tie constraint was used for the interaction very lightly reinforced column S5 was used to study the behaviour
between the stud and the concrete due to the increased analysis at large relative rotations. Such column had initially yielded prior
time if the contacts were defined. The match with the experimen- to the failure of the connection. However, due to the large relative
tal results was relatively good, however only monotonic analysis rotation the strength of the connection deteriorated in the subse-
were performed. quent cycles and the final failure occurred in the connection.
Resch and Kaliske [30] made a 3D simulation of double-shear The T-shape beams were 60 cm high and 22/50 cm wide. At the
dowel-type connections of wood. Again, the authors discovered location of the connection they were provided with a steel tube
that the suitable formulation of the contact between the wood (80/50/2 mm) which was surrounded by a number of horizontal
components and the fastener is of a great importance for the real- U-shape stirrups (£10/10 cm) (Fig. 5a). The purpose of these stir-
istic modelling of the behaviour of the connection. rups was to partly confine the dowel and first of all to provide
All of the above cited studies investigate the dowel mechanism. resistance against the splitting of the beam. For this reason, two
The findings could be useful when modelling the beam-to-column additional stirrups of larger diameter (£14) were applied at the
dowel connection in precast industrial buildings. However, this bottom of the beam (Fig. 5a). Similarly, the hoops in the beam
type of the connection has its specifics and it was therefore neces- (perpendicular to these stirrups) were closely spaced (£8/5 cm)
sary to provide new FE models. within the location of the connection at a distance of 50 cm from
the edge of a beam (Fig. 5a). Elsewhere, the transverse reinforce-
ment was equal to £8/10 cm.
3. Overview of the experiments

3.1. Experimental set-up 3.2. Observed response during the experiment

The test set-up is shown in Fig. 3. The column was fixed to the 3.2.1. The monotonic test, small rotations
ground through a special foundation, which was anchored to the The global response of the connection is presented in Fig. 6. After a
laboratory floor. On the opposite side the beam was supported detailed review it was found that the dowel deformed inside the
by a roller bearing, which allowed its horizontal movement. A ver- body of the column, approximately 5 cm below the top of the col-
tical load was applied at the mid-span of the beam by means of a umn. First, the concrete crushed around the dowel, allowing the
vertical hydraulic jack. The magnitude of the vertical load was dowel to deform (Fig. 2). Then plastic hinge formed a few centimetres
100 kN in all cases. The horizontal force was applied in the direc- inside the column. No pull-out of the dowel was observed.
tion of the beam by means of another hydraulic actuator, attached In the beam, no damage (splitting) at the surface was observed,
to the reaction steel frame. indicating that the confinement was sufficient. Steel tube in the
182 B. Zoubek et al. / Engineering Structures 52 (2013) 179–191

Fig. 3. Scheme of the experimental set-up.

Fig. 4. Closer look at the beam-to-column connection.

Table 1
Summary of the tests performed.

Label Type of the test Column cross-section (cm) Column long. reinforcement Average drift of the column
S1-1 Monotonic 50/50 16£22 Small rotations
S1-2 Cyclic 50/50 16£22 Small rotations (0.4%)
S5-2 Cyclic 40/40 4£18 + 4£20 Large rotations (5%)

beam around the dowel (provided for construction purposes only) The failure of the dowel was observed in the last cycle in the
improved the behaviour of the connection. push (Fig. 3) direction at ur  22 mm (Fig. 8a). The dowel was
The black stain marked in Fig. 7 revealed that the neoprene broken at two locations, within the body of the column as well
bearing pad rubbed against the concrete surface. This subject is as within the body of the beam (Fig. 9). The distance between
further discussed in Section 6.1. the two locations of the breaks was about 8 cm.

3.2.2. The cyclic test, small rotations 3.2.3. The cyclic test, large rotations
The maximum strength (Fh = 150 kN) was much lower than in Column yielded at the force equal to approximately 125 kN
the case of the monotonic test (Figs. 8 and 6). Considerable cyclic (Fig. 10). Due to the large rotations at the plastic hinge at the base
strength deterioration was observed (8b). No considerable damage of the column also large relative rotations between the top of the
was observed in the beam. The column remained elastic through- column and the beam occurred in the connection (Fig. 11), leading
out the experiment. Hysteretic cycles registered in the connection to subsequent deterioration of the strength of the connection. The
were wide, reaching almost 50% of the area of the perfectly elasto- connection finally failed at the force of about 100 kN and relative
plastic system (Fig. 8). displacement of 24 mm (Fig. 10). Again, similar to the cyclic test
 B. Zoubek et al. / Engineering Structures 52 (2013) 179–191 183

Fig. 5. (a) Beam and (b) column reinforcement.

Fig. 6. Monotonic response – horizontal force (Fh) versus relative displacement Fig. 7. Black stain on the beam indicates that the neoprene bearing pad rubbed
between the beam and the column (ur). against the concrete surface.

with small rotations, the dowel failed at two levels, inside the col- tween structural components listed above. Four different types of
umn as well as inside the beam. interaction were identified (Fig. 12): dowel-to-concrete contact,
dowel-to-grout contact, neoprene-to-concrete contact and rein-
4. Numerical model forcement-to-concrete contact. Modelling of the contacts is ex-
plained in detail in Sections 4.1.1 and 4.1.2.
The specimen presented in Figs. 3 and 4 was modelled using
Abaqus/Standard finite element program [32]. The following struc- 4.1. Interactions between the key components of the connection
tural components were included into the model (Fig. 12): dowel,
beam, column, infill, steel tube, reinforcement and neoprene bear- 4.1.1. The dowel-to-concrete (dowel-to-grout) contact
ing pad. Modelling of each component is described in Section 4.2. The contact properties were defined in two orthogonal direc-
To adequately simulate the experimentally observed response it tions. Hard contact with allowed separation was chosen normal
was particularly important to properly model the connections be- to the surface of the dowel and the concrete. As it was observed
184 B. Zoubek et al. / Engineering Structures 52 (2013) 179–191

Fig. 8. (a) Cyclic response – horizontal force (Fh) versus relative displacement between the beam and the column (ur) in the case of small rotations; (b) detail of the hysteretic
loops – cyclic strength deterioration from the first to the third cycle at the same displacement amplitude is considerable.

during the experiment, the concrete around the dowel at the top of
the column crushed and a crater-like void was formed around the
dowel. This loss of the contact between the dowel and the concrete
was properly modelled allowing their separation.
As mentioned previously in the Section 3.2, tests showed no
pull-out of the dowel. Nevertheless, tangential behaviour was de-
scribed with friction coefficient of 0.8 to simulate bond between
the dowel and the concrete.

4.1.2. The neoprene-to-concrete contact

As long as the rubber pad is exposed to the normal pressure,
Fig. 9. At the end of the cyclic test the dowel was broken at two levels. The distance there is a friction between the concrete and the pad activated,
between the two locations was about 8 cm. and the neoprene pad contributes to the shear resistance of the
whole connection between the beam and the column (Fig. 13).
The interaction between the neoprene and the concrete surface
150
was defined as a hard contact in normal direction and with a fric-
100 tion coefficient of 0.5 [33] in the tangential direction, similar as de-
scribed in Section 4.1.1.
50

4.1.3. The reinforcement-to-concrete interaction

Fh [kN]

0
By assuming a totally rigid connection between the reinforce-
-50 ment and the surrounding concrete the slip of the reinforcement
is neglected. For reinforcement, embedded elements were used
-100
to model total fixity to the surrounding concrete.
-150
-30 -20 -10 0 10 20 30 4.2. Materials and types of elements
ur [mm]
4.2.1. The beam, the column and the infill
Fig. 10. Cyclic response – horizontal force (Fh) versus relative displacement
Material ‘‘concrete’’ as defined in ABAQUS [32] was assigned to
between the beam and the column (ur) in the case of large rotations.
the beam and column. Its properties are presented in Fig. 14.

Fig. 11. Large relative rotations between the beam and the column.
 B. Zoubek et al. / Engineering Structures 52 (2013) 179–191 185

Fig. 12. Close-up of the connection: presentation of the contact and material assignments.

Fig. 13. Activation of friction forces on the contact between the surfaces of the neoprene pad and the concrete.

Fig. 14. (a) Behaviour in uniaxial compression or tension and (b) definition of damage propagation [8] for concrete.

Material grout was assigned to the infill between the dowel and the models exhibiting softening behaviour and stiffness degradation
steel tube. For the stress–strain relationship of concrete and grout often lead to convergence problems, viscoplastic regularization
Park & Kent model was used [34]. Elastic behaviour was assumed was used [32].
until 1/5th of the compressive strength which had been obtained All three constitutive parts (the beam, the column and the infill)
from the uniaxial compression test. Nonlinearity of concrete was were modelled with standard solid continuum elements with re-
modelled by approaches based on the concepts of plasticity and duced integration C3D8R (an 8-node linear brick). Characteristic
damage by using Concrete Plasticity Damage Model (CPDM) in- size of finite elements in the regions, where nonlinear behaviour
cluded in ABAQUS [30,34–38]. The model accounts for the loss of was observed during the experiment, was approximately 2 cm. In
elastic stiffness due to the plastic straining in tension and compres- other regions, where the response was predominantly elastic, ele-
sion. For simplicity, the following assumption was adopted: the ments of 5–10 cm were used.
nonlinearity of concrete before the peak stress is due only to plas-
ticity; the strain-hardening or softening of concrete after peak 4.2.2. The dowel, reinforcement and the steel tube
stress is due only to concrete damage [37]. Damage factor with re- 4.2.2.1. The dowel and the reinforcement. For modelling steel classi-
spect to plastic deformation is presented in Fig. 14b. Since material cal metal plasticity model (included in ABAQUS) with combined
186 B. Zoubek et al. / Engineering Structures 52 (2013) 179–191

800 isotropic hardening, which uses Misses yield surfaces, was used.
Dowel stress-plastic strain relationship is presented in Fig. 15.
700
Stress and strain values were obtained from uniaxial tension test.
600 For the reinforcement, similar steel model was used. The yield
strength was 560 MPa and the ultimate strength was 630 MPa.
500
Stress [MPa]

The dowel was modelled with standard solid continuum ele-

400 ments with reduced integration C3D8R. For the reinforcement,
2-node linear 3D truss elements were used.
300

200
4.2.2.2. The steel tube. Because no tests have been performed to ob-
tain the stress–strain diagram of the steel tube, bilinear response
100 with maximum strength of 250 MPa and maximum deformation
0
of 10% has been considered.
0 2 4 6 8 10 12 14 16 18 20
Plastic strain [%] 4.2.3. Neoprene bearing pad
Neoprene bearing pad was modelled as an ideally elastic mate-
Fig. 15. Behaviour in uniaxial tension of steel used for dowels.
rial with elastic modulus E = 3 MPa and Poisson’s ratio m = 0.49.

Fig. 16. (a) Global response (horizontal force –relative displacement) observed within the monotonic test: Comparison of the experiment and the analysis. (b) Diagram of
compressive deformations in concrete in front of the dowel. (c) Diagram of longitudinal stresses on the edge of the section along the steel dowel at different stages of the
monotonic test. (d) Diagram of longitudinal plastic deformations on the edge of the section along the steel dowel at different stages of the monotonic test.

Fig. 17. Comparison between the numerical and experimental results in the case of small relative rotations – (a) horizontal force (Fh) versus relative displacement between
the beam and the column (ur) and (b) cumulative dissipated energy.
 B. Zoubek et al. / Engineering Structures 52 (2013) 179–191 187

300 and experimental relation between the horizontal force Fh in the

250 actuator and the relative displacement between the beam and
200 the column ur is presented in Fig. 16a.
150 The analysis successfully reproduced the mechanism observed
100 during the test (Section 3.2). Characteristic points of the analytical
Fh [kN]

50 response are marked with numbers 1–4 and compared to the

0
experimental results. From Fig. 15, next observations can be made:
-50
Experiment (cyclic) (1) First, the yielding of the dowel was observed, corresponding
-100
to the displacement of approximately 3 mm (characteristic
-150 Experiment (monotonic) point 1). Simultaneously the concrete around the dowel
-200
-30 -20 -10 0 10 20 30 40 crushed (see Fig. 16b) allowing the dowel to develop large
ur [mm] plastic deformation. The difference of 5% between the anal-
ysis and the experiment was observed at the moment of
Fig. 18. Comparison of the experimental cyclic and monotonic response. the yielding of the dowel (see Fig. 16a).
(2) The maximum possible stress in the dowel was almost
170
reached at the displacement of about 10 mm (see characteris-
150 tic point 2 in Fig. 16a). The corresponding difference of 4.8%
between the analysis and the experiment was identified
130
(see Fig. 16a). After that the plastic hinge formed in the dowel
Fh [kN]

110 inside the column (as anticipated by the mechanism explained

in the theoretical background – Fig. 2) at the depth of about
90
5 cm below the top of the column (Fig. 16c and d).
Analysis
70 (3) The dowel was close to the failure. No steel hardening of the
Experiment dowel was observed between the points 2 and 4 though the
50 deformation was increasing (Fig. 16c and d). This finding
5 6 7 8 9 10 11 12
demonstrates that the hardening slope between the points
ur [mm]
3 and 4 cannot be attributed to the hardening of the steel
Fig. 19. Experimentally observed cyclic deterioration was successfully captured by in the dowel as it was indicated in some previous reports
the model. [10]. There should be another mechanism contributing to
this additional resistance of the connection. See further dis-
Standard solid continuum elements with reduced integration cussion and explanation in Section 6.
C3D8R were used. (4) Some discrepancy between the analytical and experimental
response can be observed between 10 mm and 20 mm
5. Comparison of the experimental and numerical results
(characteristic points 2 and 3). In this region the largest dif-
ference (17%) between the analysis and the experiment was
In this section the efficiency of the proposed numerical model is
noted (see Fig. 16a). Some strain hardening, which can be
demonstrated. The numerical results obtained by ABAQUS are
observed in the analytical response, is due to the imprecise
compared with the test data. Good match with the experimental
modelling of the steel tube and the grout around the dowel
results was achieved on the global level as well as in all significant
in the beam (caused by the lack of the precise data about the
details. These results will be further analyzed and discussed in Sec-
quality of the steel of the tube).
tion 6 to provide the background for the explanation of the failure
(5) The strain hardening after the displacement of 20 mm was
mechanism of the connection in various situations (monotonic,
contributed by the neoprene pad, and the total strength of
cyclic, large relative rotations).
250 kN was observed in the experimental as well as in the
5.1. Monotonic test analytical response (characteristic point 4). The difference
between the analysis and the experiment was 2.4% (see
Monotonic test was performed only for the case of small rela- Fig. 16a). Further discussion and explanation of this mecha-
tive rotations (specimen S1-1). The comparison of the numerical nism is provided in Section 6.

Fig. 20. (a) Formation of plastic hinges approximately at the same level as observed during the experiment and (b) failure of the dowel as predicted by the numerical analysis.
188 B. Zoubek et al. / Engineering Structures 52 (2013) 179–191

Fig. 21. Comparison between the numerical and experimental results in the case of large relative rotations – (a) horizontal force (Fh) versus relative displacement between
the beam and the column (ur) and (b) cumulative dissipated energy.

Fig. 22. (a) Contributions of the dowel and the neoprene for the monotonic response obtained from the analysis. (b) Neoprene is compressed due to the bending moments in
the connection.

5.2. Cyclic test connection in the monotonic test (Fig. 22a). Important strain hard-
ening, which can be observed after the displacement of 20 mm
5.2.1. Small relative rotations between the beam and the column (Fig. 22a), is contributed only by the rubber pad. In this region
within the connection there is no strain hardening in the dowel at all (see dashed line
Again, quite good match between the experimental and analyti- in Fig. 22a). In previous, predominantly empirical studies, this ef-
cal results can be seen in Fig. 17a and b. In the case of the cyclic re- fect was not recognized.
sponse no hardening (typical for the monotonic response – Fig. 18) The neoprene pad contributes to the shear strength of the
was observed, neither in the analysis nor in the experiment. After whole connection as long as it is exposed to the sufficient vertical
yielding of the dowel, the cyclic resistance remained practically con- pressure, which ensures a contact between the neoprene and the
stant. Consequently the final difference of the monotonic and cyclic concrete (prevents sliding of the pad and concrete).
resistance was about 40%. The dowel connection is not perfectly hinged connection. The
The numerical model also captured well the experimentally ob- bending moments as large as 20% of the flexural strength of the
served cyclic strength deterioration (Fig. 19) and the energy dissi- column were observed in some cases. These moments are trans-
pated in the connection (Fig. 17b). mitted between the beam and the column by the tension force in
In the case of small rotations, analysis predicted that the plastic the dowel and the compression force acting on the part of the
hinge formed around 4 cm inside the column (Fig. 20) which is less neoprene pad. Thus, the increase of the moment in the connection
than in the case of monotonic test, where this depth was around 5 cm.

5.2.2. Large relative rotations between the beam and the column
within the connection
Taking into account the complexity of the problem, the match
with the experiment is relatively good (Fig. 21). Considering large
rotations between the column and the beam the resistance of the
connection is approximately 20% smaller than in the case of small
rotations (Figs. 17a and 21).

6. Discussion

6.1. Influence of the neoprene bearing pad

The analysis demonstrated that the neoprene pad could consid- Fig. 23. Comparison of the maximum plastic deformations along the edge of the
erably (for about 20% or 50 kN) increase the total strength of the dowel at the maximum force in the monotonic and cyclic test.
 B. Zoubek et al. / Engineering Structures 52 (2013) 179–191 189

Fig. 24. Strength reduction due to the large relative rotations – (a) experimental results and (b) numerical results.

tion is proportional to the depth of the plastic hinge. Consequently,

the ultimate resistance in the cyclic test was smaller due to the
smaller depth of the plastic hinge.
In total the cyclic resistance was close to 40% lower than the
monotonic resistance.

6.3. Influence of large rotations

Both, experimental (Fig. 24a) and analytical (Fig. 24b) results

demonstrate 15–20% drop in the resistance of the connection due
to the damage induced by large rotations between the beam and
the column. At large rotations the dowel is loaded not only in flex-
ure but also in tension (Fig. 25). Combination of both leads to the
observed drop in the resistance.
Fig. 25. Additional tension in the dowel due to large relative rotations between the
beam and the column.
This phenomenon had not been taken into account until, based
on the results of the SAFECAST project [7,10], formula (4) was pro-
posed. In comparison with the formula (3), which is valid in the
case of small rotations, formula (4) correctly evaluates the reduc-
means also an increase of the compression force in the compressed tion of strength (1.1–0.9)/1.1  18%. Since inelastic response anal-
part of the rubber pad (see Fig. 22b). This increases also the max- yses of slender cantilever columns in precast buildings indicated
imum horizontal forces, which can be transmitted between the large horizontal displacements associated with relative rotations
concrete and the pad. up to 10%, the proposed reduction should be included in the de-
sign, unless shear walls/cores are included in the structural system.

6.2. Mechanism of failure and analytical predictions of the capacity

7. Conclusions
In the case of monotonic loading (specimen S1-1), it is conve-
nient to assume the failure mechanism presented in Fig. 2. If mean The authors participated in the reconnaissance mission after the
values for uniaxial compressive strength of the concrete fcc and 2012 Emilia-Romagna earthquake which had hit the highly indus-
yield strength of the steel fy are used, as well as the nominal diam- trialized area in Italy, where hundreds of precast buildings ad-
eter of the dowel (2.8 cm) is taken into account, the Eq. (1) gives dressed in this paper were affected. This event and some other
the ultimate resistance of the connection Ru,m = 193 kN. recent earthquakes in Turkey demonstrated and confirmed that
Analysis with the proposed numerical model (see Section 5) the connections between beams and columns in the precast indus-
indicates that the ultimate resistance of the dowel itself is trial buildings are crucial structural elements, which should be
200 kN (Fig. 22a). This is quite close to the value, estimated by properly designed to provide the adequate seismic safety of the
Eq. (1). The total strength of the connection is however larger whole building.
(250 kN) due to the contribution of the neoprene pad, described The most common connection in the European precast indus-
in Section 6.1. trial buildings is a beam-to-column dowel connection. In spite of
In the cyclic test, smaller relative rotations between the beam that, the knowledge about the failure mechanism of such connec-
and the column were imposed (22 mm compared to 32 mm in tions in the case of seismic action has been incomplete. Earlier
the monotonic test). Consequently, smaller contributions (10% of investigations were performed on simplified models, where many
the total force) of the elastomer to total shear strength was ob- important structural parameters of the connection were neglected.
served. During the cyclic test the contribution of the dowel itself Based on the most recent research in the frame of the SAFECAST
to the ultimate resistance of the connection was 25% lower than project empirical formulas were proposed to evaluate the ultimate
in the monotonic test. The failure mechanism was similar in both resistance of realistic connections. However, the failure mecha-
tests. However the depth of the plastic hinge was different and nism was still not adequately understood and explained.
consequently different strengths were obtained. The depth of the The research presented in this paper has provided a numerical
plastic hinges was about 4 cm and 5 cm in the cyclic and mono- tool, based on the ABAQUS FEA software, which is able to describe
tonic test, respectively (see Fig. 23). In the previous investigations the characteristic of inelastic seismic behaviour of dowel connec-
[14,15] it was demonstrated that the strength of the dowel connec- tions on the global and component level.
190 B. Zoubek et al. / Engineering Structures 52 (2013) 179–191

Comparisons with experimental results obtained with mono- Institute (ZAG). The specimens were constructed at Primorje d.d.
tonic and cyclic tests on realistic connections demonstrated the company.
soundness and efficiency of the proposed model. Considering the
complexity of the problem the match of the results is very good. References
First of all the model was able to explain the failure mechanism
as well as the most important features of the monotonic and cyclic [1] Fischinger M, Cerovšek T, Turk Ž. EASY – slide information system; 1997.
<http://www.ikpir.com/easy/> [accessed 25.07.12].
response on the component level. The most important observa- [2] Earthquake Engineering Research Institute, Armenia Earthquake
tions are summarized below: Reconnaissance Report, Earthq. Spectra 1989;5(S1):175.
[3] Fajfar P, Banovec J, Saje F. Behaviour of prefabricated industrial building in
Breginj during the Friuli earthquake. In: Proceedings of the 6th ECEE, vol. 3,
1. Standard theory assuming that the failure mechanism is initi-
Dubrovnik, Yugoslav Association for Earthquake Engineering; 1978. p. 493–
ated by flexural yielding of the dowel and crushing of the sur- 500.
rounding concrete was confirmed. [4] Fajfar P, Duhovnik J, Reflak J, Fischinger M, Breška Z. The behavior of buildings
and other structures during the earthquakes of 1979 in Montenegro. IKPIR
2. The strength of the connection considerably depends on the
Publication. University of Ljubljana; 1981.
depth of the plastic hinge in the dowel. The larger the depth [5] UNDP/UNIDO. Building construction under seismic conditions in the Balkan
is the larger is resistance. That is why the strength of the con- region: prefabricated/industrialised reinforced concrete buildings systems.
nection is reduced in the case of the cyclic loading, where the UNDP/UNIDO project RER/79/015; 1982.
[6] Park R et al. (Fischinger M contributor). Seismic design of precast concrete
depth of the plastic hinge is smaller compared to the monotonic structures. Bulletin – FIB state-of-art-report 27, Laussane; 2003.
loading. [7] SAFECAST. Performance of innovative mechanical connections in precast
3. Neoprene bearing pad can considerably increase the strength of building structures under seismic conditions; 2009. <http://
www.safecastproject.eu> [accessed 20.07.12].
the connection, particularly when large relative displacements [8] Kramar M, Isakovic T, Fischinger M. Experimental investigation of ‘‘pinned’’
between the beam and the column are developed. This has been beam-to-column connections in precast industrial buildings. In: Proceedings of
the case of the monotonic test, presented in the paper. the fourteenth European conference on earthquake engineering, Republic of
Macedonia, Macedonian Association for Earthquake Engineering; 2010. p. 1–8.
4. When the contribution of the neoprene pad is considerable, for- [9] Fischinger M, Zoubek B, Kramar M, Isakovic T. Cyclic response of dowel
mula (1), which has been used in the design practice to estimate connections in precast structures. In: 15th World conference on earthquake
the strength of the dowel connections, underestimates the engineering, Lisbon; 2012.
[10] Psycharis IL, Mouzakis HP, Kremmyda GD. Experimental investigation of the
capacity of the connection. behaviour of precast structures with pinned beam-to-column connections. In:
5. Both, experimental and analytical results demonstrated 15– Fardis M, Rakicevic ZT, editors. Role of seismic testing facilities in
20% drop in the resistance of the connection due to the dam- performance-based earthquake engineering, geotechnical, geological and
earthquake engineering. Netherlands: Springer; 2012.
age induced by large rotations between the beam and the
[11] AIJ. Report on the damage investigation of the 1999 Kocaeli Earthquake in
column. At large rotations the dowel is loaded not only in flex- Turkey, 2001. AIJ report. Architectural Institute of Japan, Tokyo; 2001.
ure but also in tension. Combination of both leads to the [12] EERI. The Mw 7.1 Ercisß-Van, Turkey Earthquake of October 23, 2011. EERI
observed drop in resistance. Since inelastic response analyses special earthquake report; 2012.
[13] Toniolo G, Colombo A. Precast concrete structures: the lessons learned from
of slender cantilever columns in precast buildings indicated the L’Aquila earthquake. Struct Concr 2012;13(2):73–83.
large horizontal displacements associated with relative rota- [14] Vintzeleou EN, Tassios TP. Behaviour of dowels under cyclic deformations. ACI
tions up to 10%, the observed reduction should be included Struct J 1987;84(1):18–30.
[15] Vintzeleou EN, Tassios TP. Mathematical model for dowel action under
in the design, unless shear walls/cores are included in the pre- monotonic and cyclic conditions. Mag Concr Res 1986;38:13–22.
cast structural system. [16] Dulascska H. Dowel action of reinforcement crossing cracks in concrete. J ACI
1972;69–70:754–7.
[17] Engström B. Combined effects of dowel action and friction in bolted
The proposed numerical model not only confirmed the empiri- connections. Nord Concr Res 1990;9:14–33.
cally based design formulas for the strength of the dowel connec- [18] Højlund-Rasmussen B. Betoninstöbe tvaerbelastade boltes og dornes
tions tested in the frame of the SAFECAST project, but also provided baereevne [Resistance of embedded bolts and dowels loaded in shear].
Byngninsstatiske Meddelser 1963:34.
a FEM based tool, which will enable the extrapolation of the exist- [19] Soroushian P, Obaseki K, Rojas MC, Sim J. Analysis of dowel bars acting against
ing experimental results and empirical formulas to beam–column concrete core. ACI J Proc V 1986;83(4):642–9.
dowel connections with different structural parameters (i.e. the [20] Dei Poli S, Di Prisco M, Gambarova PG. Dowel action as a means of shear
transmission in RC elements: a state of art and new test results. Studi e
diameter of the dowel, the influence of the confinement rate in
Ricerche, school for the design of R/C structures, vol. 9(87). Milan University of
the beam and column, the thickness of the bearing pad, etc.). This Technology; 1988. p. 217–303 [in Italian].
tool will be used for case studies from the 2012 Emilia-Romagna [21] Dei Poli S, Di Prisco M, Gambarova PG. Shear response, deformations, and
earthquake, where slightly different types of the connections subgrade stiffness of a dowel bar embedded in concrete. ACI Struct J
1992;89(6):665–75.
(e.g. absence of the steel tube in the beam, less efficient confine- [22] Tanaka Y, Murakoshi J. Reexamination of dowel behavior of steel bars
ment of the concrete, etc.) will be investigated. embedded in concrete. ACI Struct J 2011;108(6):659–68.
Presented results and discussion are related only to the beam- [23] Paulay T, Park R, Phillips MH. Horizontal construction joints in cast-in-place
reinforced concrete, shear in reinforced concrete. ACI 1974;42(SP):559–616.
to-column dowel connections, where the dowel is not close to [24] fib. Structural connections for precast concrete buildings. Bulletin 2008;43.
the edge of the beam or the column (the distance should be more [25] Prandtl L. Über die Härte plastischer Körper. Nachr K Wiss Gö Math-Phys Kl
than six diameters of the dowel). When the dowel is closer to the 1920;12:74–85.
[26] Broms BB. Lateral resistance of piles in cohesive soils. Proc Am Soc Civ Eng
edge of the beams or the columns different failure mechanisms are 1965;91:77–99.
developed. It is believed that the proposed numerical model can be [27] Maitra SR, Reddy KS, Ramachandra LS. Load transfer characteristics of dowel
upgraded to address this problem. bar system in jointed concrete pavement. J Transp Eng 2009;135(11):813–21.
[28] Guezouli S, Lachal A. Numerical analysis of frictional contact effects in push-
out tests. Eng Struct 2012;40:39–50.
Acknowledgements [29] Nguyen HT, Kim S. Finite element modelling of push-out tests for large stud
shear connectors. J Constr Steel Res 2009;65(10–11):1909–20.
[30] Resch E, Kaliske M. Numerical analysis and design of double-shear dowel-type
The presented research was supported by the SAFECAST Project
connections of wood. Eng Struct 2012;41:234–41.
‘‘Performance of Innovative Mechanical Connections in Precast [31] BS EN ISO 15630-1:2010. Steel for the reinforcement and prestressing of
Building Structures under Seismic Conditions’’ (Grant Agreement concrete. Test methods. Reinforcing bars, wire rod and wire.
No. 218417-2) in the framework of the Seventh Framework [32] ABAQUS Theory Manual, version 6.11-3. Dassault Systèmes; 2011.
[33] Magliulo G, Capozzi V, Fabbrocino G, Manfredi G. Neoprene–concrete friction
Programme (FP7) of the European Commission. Experiments were relationships for seismic assessment of existing precast buildings. Eng Struct
completed at the Slovenian National Building and Civil Engineering 2010;33(532):538.
 B. Zoubek et al. / Engineering Structures 52 (2013) 179–191 191

[34] Kent DC, Park R. Flexural members with confined concrete. J Struct Div [37] Yu T, Teng JG, Wong YL, Dong SL. Finite element modeling of confined
1997;7:1969–90. concrete-II: plastic-damage model. Eng Struct 2010;32(680):691.
[35] Lee J, Fenves GL. Plastic-damage model for cyclic loading of concrete [38] Jankowiak T, Lodygowski T. Identification of parameters of concrete damage
structures. J Eng Mech 1998;124(8):892–900. plasticity constitutive model. Found Civ Environ Eng 2005;6:53–69.
[36] Lubliner J, Oliver J, Oller S, Onate E. A plastic-damage model for concrete. Int J
Solid Struct 1989;25:299–329.