Structures 35 (2022) 939–957

Contents lists available at ScienceDirect

Structures
journal homepage: www.elsevier.com/locate/structures

Cyclic behaviour of precast beam-to-column connections: An experimental

and numerical investigation
Ines Torra-Bilal a, Mustafa Mahamid a, *, Eray Baran b
a
Department of Civil, Material and Environmental Engineering, University of Illinois at Chicago, Chicago, IL 60607, USA
b
Department of Civil Engineering, Middle East Technical University, Ankara 06800, Turkey

A R T I C L E I N F O A B S T R A C T

Keywords: The main objective of this study is to examine the behavior of hybrid beam-column connections with numerous
Failure modes detailing methods to be incorporated in precast moment resisting frames under simulate reversed cyclic loading,
Nonlinear Finite Element Analysis which will lead to better understanding of their performance and safer seismic applications. To accomplish the
Precast Connections’ Detailing
research objectives, an integrated experimental and numerical study was conducted. The experimental part
Modified Concrete Damage Plasticity Model
Reversed cyclic loading
consisted of reversed cyclic loading of precast concrete beam-column specimens with six different anchorage
Kinematic Hardening details used in beam element, as well as monolithically cast companion specimens. In the numerical part, a
nonlinear finite element analysis (FEA) approach was proposed to reproduce the experimental response of a
specific type of moment resisting precast concrete beam-column connection exposed to reversed cyclic load tests.
To prove the validity of the proposed finite element approach, three-dimensional (3D) finite element models
were built using ABAQUS, a nonlinear finite element analysis tool, for a reference monolithic and various precast
hybrid beam-column connections. The models intend to lead to comprehensive investigation of concrete and
steel behavior reaching failure. A modified Concrete Damage Plasticity model (CDP) that accounts for
compression-softening and tension-stiffening effect was adopted for concrete in order to reproduce the typical
cyclic behavior of cracked reinforced concrete. Kinematic Bilinear Elasto-Plastic nonlinear model was used to
model all steel parts with the addition of bond-slip effect for all reinforcing bars embedded in concrete.
Good comparison between finite element results and tested members has been achieved, which demonstrates
the accuracy of the proposed finite element model despite the complexity of the investigated connections. All
failure modes of the precast connection components were captured from the developed finite element model and
compared with those of the corresponding monolithic connection. It was observed that simple detailing modi­
fications highly influence failure modes of the connections and may result in major improvement of the
connection performance in terms of strength and energy dissipation mechanism. This study concludes that
precast connections have the potential to perform well under cyclic loading and results from this study can be
utilized to efficiently investigate the influence of various design parameters on these connections’ performance.

1. Introduction between beam and column elements in precast concrete moment-

resisting frames, while limiting construction and production efforts.
Precast concrete structural systems have gained popularity mainly in The beam-to-column connections are identified as the fundamental
design of buildings due to their important characteristics such as cost element that influence not only the overall performance of the structure
effectiveness, improved quality control, low environmental pollution, but also the cost and construction efficiency. Additionally, joints play a
construction efficiency, and adaptability as opposed to cast-in-place significant role in the energy dissipating mechanism and redistribution
reinforced concrete and structural steel systems. Nonetheless, there is of forces as the applied load increases. Performance of precast connec­
limited application of precast concrete structural systems in structural tions during seismic events around the world questioned the safety of
and seismic design of buildings. This is essentially due to the deficiency these structural systems as compared to other traditional structural
of connection detailing schemes that can properly transmit loads systems [1,2]. Particularly, precast buildings revealed seismic

* Corresponding author.
E-mail address: [email protected] (M. Mahamid).

https://doi.org/10.1016/j.istruc.2021.11.046
Received 1 June 2021; Received in revised form 23 October 2021; Accepted 22 November 2021
Available online 7 December 2021
2352-0124/© 2021 Institution of Structural Engineers. Published by Elsevier Ltd. All rights reserved.
I. Torra-Bilal et al. Structures 35 (2022) 939–957

vulnerability due to inadequate detailing and design of connections 2. Description of connections

during 1999 Kocaeli earthquake in Turkey or 1994 Northridge earth­
quake in California. Lessons learned from past earthquakes evidence the Experimental investigation included testing of half-scale exterior
need of the investigation of beam- to-column joints’ behavior to make reference monolithic and precast beam-column connections. A total of
precast buildings safe and allow the full exploitation of their potential. thirteen connections (two monolithic and eleven precast specimens with
Considerable amount of experimental studies has been previously various detailing schemes) were considered for investigation for the
performed to comprehend the performance of precast concrete moment scope of this study. Monolithic and precast specimens’ dimensions and
connections under cyclic loading, yet numerical studies that have the reinforcement arrangement are shown in Fig. 1.
potential to accurately predict the behavior of these connections with Column elements of a typical 400 × 400 mm cross-section with
various detailing schemes along with their respective failure modes twelve 20 mm diameter longitudinal bars and 8 mm diameter stirrups at
through all load cycles remains limited. Nzabonimpa et al. (2017) [3] 100 mm spacing were used as part of the beam-column joint assembly.
performed an analytical investigation on a novel mechanical beam- The typical column height was defined to be 1410 mm.
column connection as part of a precast concrete moment frames. It Beam elements had an 1800 mm length and a 400 × 250 mm rect­
was concluded that the nonlinear finite element approach showed suf­ angular cross section with six 20 mm or 16 mm diameter longitudinal
ficient agreement with the experimental results which allowed for reinforcement, placed symmetrically at the top and bottom of the beam.
investigation of deformation of beam end plates and concrete damage at Yield strength of the 16 mm and 20 mm diameter reinforcing bars were
the joint region. Feng et al. [4] proposed a detailed finite element measured to be 525 MPa and 465 MPa, respectively. The measured
modeling approach to be used in precast reinforced concrete beam-to- concrete compressive strengths for the specimens are given below. The
column connections under cyclic loading. Various material models and monolithic specimens served as the reference specimen for the assess­
model parameters were considered for investigation and modified as ment of the performance of the investigated precast beam-column
needed to reach the most optimal agreement in simulation of precast connection details. The difference between two monolithic specimens
connections in comparison with experimental results. Hawileh et al. is the longitudinal reinforcement ratio in beam elements. M1 has three
(2009) [5] presented a nonlinear finite element study of a precast hybrid ϕ16 and M2 has three ϕ20 longitudinal reinforcement located in top and
beam-column connection exposed to cyclic loads by considering the pre- bottom part of beam cross section.
tension effect in the post-tensioning strands as well the nonlinear In order to achieve force transfer between beam and column ele­
behavior of concrete. Good agreement was achieved between experi­ ments in the precast connections, 400 mm long L120 × 120 × 12 top and
mental and analysis results which proved the reliability of the proposed bottom steel angles were implemented in the connection construction
finite element model procedure. scheme. Steel angles were attached to the column face utilizing 30 mm
Najafgholipour et al. (2017) [6] conducted a numerical investigation diameter steel threaded rods, which were pretensioned to a 30 kN load
of reinforced concrete joints subjected to simulated lateral loading to level. The threaded rods had measured yield and ultimate strengths of
examine shear at the beam-column connection region. The shear ca­ 458 MPa and 517 MPa, respectively. Steel plates with 120 mm width, 12
pacity of the joint and deformed shape were studied as part of this mm thickness and 250 length were anchored inside the beam element
research. Bahrami et al. (2017) [7] studied the behavior of two new and steel top and bottom angles were welded to the plates. The manu­
moment resisting precast beam to column connections subjected to facturer specified minimum yield strength for the plates and angles was
lateral loading using a nonlinear finite element approach. A bolted 235 MPa (grade S235), and 275 MPa (grade S275), respectively.
(inverted E) and a welded (box type) connection were investigated. Details of connection region in precast beam elements is shown in
Results considered to understand the behavior of these connections Fig. 2. Specimen D1 includes three 16 mm “Z” shape bars with a total
included lateral resistance, lateral stiffness, ductility and energy dissi­ length of 600 mm welded to each steel plate and embedded inside the
pation. Other studies that focused on developing and testing various concrete of 32.7 MPa compressive strength. Revisions to detail D1
moment resisting connections between precast beams and columns were include detail D1R1, D1R2, D1R3, D1R3R, and D1R3H. Specimen D1R1
performed at Minnesota University by French et al. [8,9]. Conclusions had an increased number of the “Z” shape bars from three to four and
were drawn on their performance based on strength and ductility. was characterized by a concrete strength of 33.5 MPa. Specimen D1R2
Additional experimental investigations that contributed to understand­ had a reduced stirrup spacing from 100 mm to 75 mm and three U-
ing the behavior of various precast connections under cyclic loading in shaped 10 mm diameter rebars with a length of 1500 mm were longi­
comparison with their monolithic counterparts were performed by Choi tudinally placed in the joint region in order to provide better confine­
et al. [10] and Parastesh et al. [11]. ment. This specimen’s concrete strength was determined to be 36.4
Due to absence of studies on the behavior of beam-column precast MPa. Specimen D1R3 had wider beam connection plates with modified
connections subjected to cyclic loading, the goal of this study to width of 150 mm instead of 120 mm previously used and its concrete
experimentally and numerically investigate the behavior of a hybrid compressive strength was specified to be 35.6 MPa. Specimen D1R3R
exterior beam-column connection to be used in precast moment resisting was similar to connection D1R3 with the absence of pre-tensioning force
frames. In the numerical analyses Finite Element Analysis (FEA) was in the rods and a compressive strength of 36 MPa. Specimen D1R3H was
implemented using the finite element software, ABAQUS [12,13] in similar to connection D1R3 with increased beam longitudinal rein­
order to predict failure modes of the connection parts throughout all forcement, where six 20 mm reinforcement bars were used. The
load cycles. Three-dimensional (3D) nonlinear finite element models compressive strength of this connection was specified to be 36.7 MPa.
were developed to compare and investigate the behavior of a typical Furthermore, five additional detailing schemes other than D1 and its
monolithic and precast hybrid beam-column connection exposed to revisions were considered for testing. Connection D2, characterized by a
reversed cyclic loading against experimental data. The models involve 45.7 MPa compressive strength, consisted of two pieces of UPN-65 steel
detailed investigation of both concrete and steel behavior along with sections with 376 mm lengths welded to the beam top and bottom plates.
their failure modes. Simulating concrete and steel nonlinearly, choosing Also, two holes were opened on the web of each of these sections and
the right nonlinear concrete material model and related parameters for two pieces of 16 mm diameter and 200 mm length anchor bars were
cyclic loading, defining contact properties between the models’ com­ passed through these holes. Specimen D3 detailing included one piece of
ponents were challenging in the simulation of precast connections on a UPN-65 steel section welded to top and bottom steel plates in weak axis
moment resisting frame under cyclic loading. FEA results show good orientation and four pieces of U-shape anchor bars having a 14 mm
agreement with experimental results, suggesting that computational diameter and 1500 mm length placed around the UPN-65 profile inside
models present attractive and cost-efficient alternative compared to the beam element. This connection’s compressive strength was specified
experiments. to be 42.7 MPa. In Specimen D4, four pieces of anchor bars with 20 mm

940
I. Torra-Bilal et al. Structures 35 (2022) 939–957

Fig. 1. Geometry of test specimens: (a) monolithic specimens; (b) precast specimens; (c) column reinforcement layout; (d) beam reinforcement layout.

diameter and 376 mm length were welded to the beam top and bottom that there is no consensus on the loading protocol to be followed for
plates. Then, four pieces of U-shape anchor bars with 14 mm diameter reversed cyclic testing of structural connections. Even though some
and 1500 mm length were positioned around the vertical anchored bars. sources specify protocols with three cycles at each drift ratio, there are
This connection’s concrete compressive strength was determined to be recently conducted studies in the literature that utilize loading protocols
47 MPa. In Specimen D5, the beam top and bottom plates were replaced with fewer displacement cycles.
by two pieces of L120x120x12 steel sections. One piece of UPN-65 steel More details on the loading mechanism, construction steps of the
section having length of 300 mm (facing out) was welded to each beam-column connection method and force transfer mechanism through
L120x120x12 section. Also, three holes were opened on the web of these the connection components can be found in Baran et al. [15].
UPN-65 sections, and three pieces of anchor bars having diameter of 20
mm and length 270 mm were passed through these holes. This specimen 4. Nonlinear finite element modeling
had a 35.8 MPa compressive strength. In Specimen D6, two pieces of
150x12 steel plates with a 376 length were welded between the beam To evaluate the performance of precast connections under cyclic
top and bottom plates. Seven pieces of anchor bars having diameter of loading, three-dimensional finite element models were built for the
16 mm and length of 700 mm were welded in a staggered orientation to reference monolithic and precast connections tested experimentally
each of these 150x12 steel plates. This specimen had a 39.8 MPa using the FEA software package ABAQUS [12,13].
compressive strength.
4.1. Element selection
3. Experimental program
A three-dimensional eight-node-brick element with reduced inte­
A testing frame was developed to perform load tests on the specimens gration and hourglass control, named as C3D8R in ABAQUS library as
as shown in Fig. 3. Overall specimen dimensions, as well as the geo­ shown in Fig. 4, was used for the entire model (concrete, steel rein­
metric details of the connection are also shown in this figure. Column forcement and other steel components). Modeling reinforcing bars using
specimen was connected to steel frame with pin connections at the top C3D8R element is an important modeling feature of this study that al­
and bottom in order to restrain the displacement in the vertical and lows for capturing analysis results of the steel reinforcement and other
horizontal directions with a specified length of 1900 mm in between the steel parts embedded in concrete, which is significant in understanding
two pins. Displacement-controlled reversed cyclic loading in pull and the seismic behavior of precast frames due to affecting the structure’s
push directions was applied toward the beam free end, at an 1800 mm ductility. The active degrees of freedom (DOFs) for this element are the
distance away from the column face with a 500 kN capacity hydraulic translational DOFs. Furthermore, the integration points were reduced to
cylinder. Displacement-controlled reversed cyclic loading was applied one point at the center of the element in order to ease solution time,
with increasing displacement amplitudes, determined based on ACI prevent elements locking at boundaries and allow for more flexibility of
374.1-05 [14] and shown in Fig. 3. A loadcell attached at the end of the the model in order to apprehend the inelastic performance of the
hydraulic cylinder was utilized to measure the applied load applied at connection. The single point reduced integration method is established
the free end of beam. Additionally, displacement transducers were considering the “uniform strain formulation”, which is defined as the
positioned at the free end of the beam in order to measure the net ver­ computed average strain distributed over the element’s volume.
tical displacement. Readings from loadcells and displacement trans­
ducers were continuously recorded during load tests until a maximum
4.2. Material properties
4% beam end drift ratio was reached. Displacement-controlled reversed
cyclic loading was applied with increasing displacement amplitudes,
4.2.1. Steel material model and properties
determined based on ACI 374.1-05 [14] and shown in Fig. 3. The cyclic
To capture the nonlinear behavior of steel up to failure, a bilinear
displacement amplitudes proposed in the ACI 374.1-05 document was
constitutive relationship was defined to model the steel components.
used without the trailing displacement cycles. It should be mentioned
The modulus of elasticity, E, and Poisson ratio (v) for steel material were

941
I. Torra-Bilal et al. Structures 35 (2022) 939–957

Fig. 2. Reinforcement detailing schemes in connection region (a) D1 group (b) other precast specimens.

set to be 200 GPa and 0.3, respectively. Steel in ABAQUS is treated as rule defines, the size of yield surface remains constant while translating
elastic up to its yield strength and as plastic from that point to the final in the direction of yielding.
strain under consideration. To predict yielding in steel material, the In order to define the kinematic hardening rule, a hardening ratio (b)
classical metal plastic model was adopted, which employs Von-Mises is used. The hardening ratio defines the rate at which the yield stress,
yield surface with associated plastic flow. The behavior after yield σ0 changes with plastic strain. A value of 0.01 was adopted based on
point follows the kinematic strain hardening rule which represents the previously published mechanics studies to calculate the ultimate
real material behavior for cyclically loaded materials. Kinematic hard­ strength of steel [4]. The hardening modulus Eh was calculated as the
ening predicts that the yielding in the reverse direction occurs when the hardening ratio (b) times the defined modulus of elasticity Es, as shown
stress change from the unloading point is twice the monolithic yield in Equation (1). The typical steel stress- strain curve is shown in Fig. 6.
stress (Δσ = 2σ) as shown in Fig. 5. Therefore, the Bauschinger effect,
Eh = b x E s (MPa) (1)
which presents early yielding in one loading direction than the other is
considered in the material model, which is very essential to represent Additionally, the bond-slip effect is an important factor that was
the behavior under cyclic loading. Moreover, as the kinematic hardening considered due to highly influencing the behavior of the beam-to-

942
I. Torra-Bilal et al. Structures 35 (2022) 939–957

Fig. 3. Specimen details, test setup and applied loading.

Fig. 4. Eight-node brick element, reduced integration (C3D8R) [8].

column connections subjected to cyclic loadings. This effect is particu­

larly significant in the joint and plastic hinge regions, and more so for
the precast beam-column connections. This is because in these regions
the reinforcing bars tend to undergo large bond-slip actions. The perfect Fig. 5. Different unloading behavior for kinematic and isotropic hard­
ening [16].
bond assumption will lead to an overestimate of the load capacity. The
bond-slip effect in modeling the reinforcement was applied as an
interface between steel and concrete defined as embedded region which parameters still remains a challenge. Among other material models
will be discussed later in this study in Section 4.3. The yield and tensile available in ABAQUS, a modified version of Concrete Damage Plasticity
strength values used for different steel components in the FEA models (CDP) that takes into consideration the softening effect for cracked
are given in Table 1. concrete under cyclic loading was selected for this study in order to
capture the complete inelastic behavior of concrete in both tension and
4.2.2. Concrete material model and properties compression. The modulus of elasticity was calculated using Equation
One of the key challenges of this study was the accurate modeling of (2) (ACI 318–14) [17]:
concrete material under reversed cyclic loading. Due to the complex √̅̅̅̅̅̅
Ec = 4700 f’c (MPa) (2)
nature of this material, numerical analysis has been used to predict its
nonlinear behavior. However, choosing the right concrete model and where f′ c: Concrete compressive strength.

943
I. Torra-Bilal et al. Structures 35 (2022) 939–957

Fig. 6. Typical steel stress–strain curve.

Fig. 7a. Typical stress–strain curves of plain and reinforced concrete in ten­
sion [18].
Table 1
Steel components properties.
Part E (MPa) ν σy σu
(MPa) (MPa)

Rebar (ϕ20) 200,000 0.3 465 590

Rebar (ϕ16) 525 670
Steel Angle (L120x120x12) 235 400
Connection rods 458 516
Stirrups (ϕ8) 275 400
Beam Plates 275 400
UPN 65 steel sections 415 500
All other steel parts (Z-shape, U-shape, 248 400
etc)

To define the stress–strain curves of concrete material in both tension

and compression under simulated reversed cyclic loading, the modified
compression field theory (MCFT) was utilized [18]. This approach
considers the compression-softening effect of reinforced concrete as well
as the tension stiffening effect. The newly modified damage plasticity
model has the ability to capture the shear behavior of typical cracked
reinforced concrete structures subjected to reversed cyclic loading, un­
like other numerical models established for plain concrete. The standard
CDP is typically used for plain concrete and the interaction between steel Fig. 7b. Typical stress–strain curves of plain and reinforced concrete in
and concrete is usually neglected. However, the properties of reinforced compression [18].
concrete are somewhat different from that of plain concrete because of
the complex interaction between concrete and steel. Reinforced concrete reinforced concrete structures. Equations 3–6 were used to derive the
can bear tensile stress even after cracking, and the corresponding tensile tensile stress–strain curve of concrete.
strain can develop at a higher magnitude due to the presence of rein­
forcement bars. This phenomenon is known as tension-stiffening. On the σ 1 = Ec ∙ε1 ε1 ≤ εcr (3)
other hand, the strength of plain concrete is insignificant at this time and
reaches zero so the tensile strain has no real meaning in the structure. In σ1 =
ft
√̅̅̅̅̅̅̅̅̅̅ forε1 > εcr (4)
the interim, the high transverse tensile strain results in the decline of the 1+ 200∙ε1
compressive strength in the perpendicular direction. This phenomenon √̅̅̅̅̅̅
is defined as compression-softening. These two effects make the prop­ ft = 0.33∙ f’c (5)
erty of reinforced concrete very different from that of plain concrete and
are the key factors that should be considered in nonlinear analysis of where
reinforced concrete structures, especially under cyclic loading. In other
words, the compression-softening of reinforced concrete means the Ec denotes the elastic modulus of concrete;
reduction of compressive strength of reinforced concrete under ten­ σ1 and ε1 denote the principal tensile stress and strain
sion–compression stress state and the failure mode usually is parallel ft and εcr denote the tensile strength and the corresponding strain
cracks normal to the tensile principal direction. Therefore, a softening
coefficient is introduced to the model. Figs. 7a and 7b show the tensile Equations (6) and (7) were used to derive the compressive stress–­
and compressive stress–strain relationship in plain vs. reinforced strain curve of concrete.
concrete. [ ( ) ( ) ]
ε2 ε2
Originally developed by Vecchio and Collins [19,20], the modified σ 2 = fcmax ∙ 2∙ − 2
(6)
ε0 ε0
compression field theory has been utilized to take into account the in­
fluence of both compression-softening and tension-stiffening in

944
I. Torra-Bilal et al. Structures 35 (2022) 939–957

fcmax 1
βMCFT = = ≤1 (7)
f ’c 0.8 + 170ε1

where

fc and ε are the compressive strength and the corresponding strain

βMCFT is the softening coefficient which is less or equal to 1.
ε1 represents all tensile strain values used in the tensile stress–strain
curve
fcmax is a value smaller than f’c which is derived based on βMCFT
values.

Since all equations are sensitive to the compressive strength of

concrete (f’c), various curves were derived for each connection based on
given compressive strength. The typical modified tensile and compres­
sive concrete stress–strain relationships are shown in Figs. 8 and 9
below.

4.2.3. Damage parameters

Damage is defined both for uniaxial tension and compression during
softening in Concrete Damage Plasticity model (CDP) [21,22]. Elastic
stiffness degradation after initial failure is taken into consideration by
Fig. 9. Stress–strain relationship with tension stiffening effect (Typical).
two damage variables, dt and dc corresponding to tensile and compres­
sive damage. These parameters are functions of tensile and compressive
• μ: viscosity parameter (neglected)
plastic strains, εpl pl
t andεc respectively and were inputted as part of the CDP
material modeling definition.
Initially, the CDP default parameters’ values were adopted from
The compressive damage parameter is derived as shown in Equation
ABAQUS manual. However, after performing a sensitivity analysis, the
(8):
upper limit of the dilation angle (55 degrees) was used. The dilation
dc = 1 −
σc
(8) angle is the volume change caused by inelastic strains. As concrete starts
Ec (εc − εpl
c ) cracking, the dilation angle controls the volume changes that can be
allowed under a given load. ABAQUS allows to choose an angle in the
The tensile damage parameter is derived as shown in Equation (9):
range between 30 and 56 degrees, with the default of 30 degrees. The
dt = 1 −
σt
(9) reason for the high dilation angle is the cyclic load and ductility of the
Ec (εc − εpl
t ) investigated structure. Dilation angle should be larger under cyclic loads
than under monotonic loads because more plastic distortions occur.
The relationship of dc vs. inelastic strain and the relationship of dt vs.
Additionally, the steel–concrete frame provides more ductile behavior
inelastic strain are shown in Fig. 10.
than concrete itself, which means that the dilation angle should be
greater than the default value of 30 degrees proposed by ABAQUS for
4.2.4. CDP model parameters
concrete. Table 2 shows the standard CDP parameters in comparison
The constitutive parameters in the CDP model per the ABAQUS
with the final calibrated values used for this study.
Manual [8] are:

• ψ: Dilation angle at increased confining pressure 4.3. Contact properties

• Kc: Ratio of the second stress invariant on the tensile meridian to that
on the compressive meridian One of the most essential stages in a finite element problem is
• ε: Eccentricity of the plastic flow interaction definition between different model parts, which is very
• σb0/σco: Ratio of biaxial to the uniaxial compressive yield stresses fundamental for accuracy of results and convergence in solution. The
definition of contact properties is mainly selected to reproduce the
experimental tests as properly as possible. Interactions between these
components were defined by using surfaces, where one surface acts as
the master surface and the other as the slave surface. The difference
between a master and a slave surface is that the former can penetrate the
latter, but the reverse cannot take place. Three interaction types were
used in this study: surface-to-surface, tie constraint and embedded re­
gion. Surface-to-Surface Interaction was used between the following
components of the tested specimens: the steel angle and the column, and
the steel rods and the angles. The column was chosen to be the master
surface during its interaction with the steel angle while the steel angle
was chosen to be the master surface during its interaction with the steel
rods. The other interacting part was set to be as slave surface. In ABA­
QUS, the term ‘‘normal behavior’’ refers to the pressure developed be­
tween the surfaces whereas the term ‘‘tangential behavior’’ describes the
extent of friction and the occurrence of slippage between the two sur­
faces caused by high shear stresses. Therefore, the ‘hard contact’ was
selected to represent the normal behavior, and the ‘penalty’ friction for
Fig. 8. Stress–strain relationship with compression softening effect (Typical). the interaction between steel and steel was set to be a coefficient of

945
I. Torra-Bilal et al. Structures 35 (2022) 939–957

Fig. 10. Compressive and tensile damage parameters.

Table 2
Standard vs. Adjusted CDP Model Parameter Values.
Parameter Name Standard Value Calibrated Value

Dilatation Angle (Degree) 36 55

Eccentricity 0.1 0.1
Fbo/Fco 1.16 1.16
K 0.667 0.667

friction equal to 0.25 whereas between steel and concrete a value of 0.35
was used. These values were chosen based on typical friction coefficient
values of these materials’ interaction as well as based on the authors’
previous research FEA studies when similarly defining friction contacts.
Tie constraint was used to define the interaction between the steel angle
and the beam and in some type of connections between the steel angle
and the embedded steel plates in the beam. The beam or the embedded
beam plates were considered as master surfaces while the angle was
considered as slave surfaces. Tie connection was used to connect each
reference point to the equivalent point nodes, to ensure complete Fig. 12. Embedded region of steel reinforcement inside the concrete (with
transfer for loads and bonded conditions. The “Embedded Region” slip effect).
command was used to simulate all reinforcement bars (flexural, shear, Z-
shape bars, U-shape bars, rods extending in the column, steel plates, etc) pinned by restraining the displacement in the vertical and horizontal
interaction with the beam and column. The beam and column were directions while the rotation was allowed. On the other hand, the beam
defined as “host” elements whereas all reinforcement bars were defined end (200 mm away from the free edge) was allowed to be free against all
as “embedded” elements. The bond-slip effect was added to the all types of rotation and against lateral displacements (x and z directions). A
defined embedded regions of reinforcement. Fig. 11a shows the surface- set of nodes was defined at the beam top edge which was tied to a
to-surface interaction between concrete column and steel angle, Fig. 11b reference point through a rigid body connection. The reversed cyclic
shows the tie connection between angles and embedded beam plates and loading was applied in the y-direction at the defined reference point. The
Fig. 12 shows the embedded region of steel reinforcement inside the aim of using a set of nodes is to avoid local damage that would affect the
concrete with slip effect. accuracy of results especially at high load cycles. Gravity was applied
with the appropriate unit weight value in the y-direction, as shown in
4.4. Boundary conditions Fig. 13.

The boundary conditions were assigned to all investigated specimens

as shown in Fig. 13. The top and bottom surfaces of the column were

Fig. 11. (a) Surface-to-surface Interaction (b) Tie Constraint.

946
I. Torra-Bilal et al. Structures 35 (2022) 939–957

select a mesh size that is not too fine but also does not influence the
results. The typical mesh size utilized in this analysis was 35 mm for
both the concrete and steel components, but it was refined in some
critical regions when necessary to better capture failure modes. This
value was chosen for the meshing elements to be larger than the
aggregate size and also not too large resulting in a coarse mesh. To
simulate the rebar cross-section, four to eight elements were used for
various connection types depending on the rebar diameter to accurately
capture the behavior of the reinforcing bars. Fig. 15 shows the meshing
arrangement of steel and concrete parts of some of the connections.
Lastly, the mesh quality was tested using Verifying Mesh option in
ABAQUS, which identifies any poorly meshed elements and gives
warnings for elements with poor aspect ratios. All models passed this
check with no warnings or errors.

5. Model validation

Fig. 13. Boundary conditions and loading including self-weight. To demonstrate the reliability of the proposed modeling methodol­
ogy, analysis results were compared with the experimental results. The
4.5. Loading stages validation of the models was done by comparing connections’
load–displacement curves, moment rotation curves, and by concrete
During the experiments, the angles used for the precast connections failure. It is important to note that experimental data captured only
were clamped on the column face by pre-tensioning the steel threaded development of cracks in concrete throughout load cycles. Load-
rods to a load level of 30 kN. All investigated connections used for displacement hysteresis curves of one of the reference monolithic con­
verification had pretensioned rods. The load type called “bolt load” was nections M1 and one of the precast connections D1R2 are shown in
applied to the surface of the rod as shown below in Fig. 14. It is essential Figs. 16 and 17 respectively.
to note that boundary conditions, rod pre-tensioning forces, and applied Figs. 18 and 19 compare the crack propagation pattern of the
external loads were applied in three different time steps for convergence monolithic and precast connection in experiments vs. FEA. A distributed
purposes. The initial time step was used to define the boundary condi­ damage, in the form of concrete cracking in the beam and column
tions of the connection. Then, in the second time step, bolts pre- element is observed in the monolithic specimen for both experimental
tensioning forces were applied on each bolt in order to initiate the and FEA results, as shown in Fig. 18. On the other hand, Fig. 19 shows
interaction between the angle and column. Finally, the external cyclic the failure of concrete (cracking) for the precast specimen, which shows
loading was applied after defining the interaction and boundary con­ that crack development mainly occurs in the beam close to the
ditions of the connection. The initial step is predefined by ABAQUS, connection region, with little to no damage in the column for both ex­
while general static step was used for the second and third steps. periments and FEA results.
Additionally, 1 s was used as the time period with 0.01,1 × 10–9 and 0.1 To ease the comparison of the results between FEA and experimental,
s as initial, minimum and maximum increment size, respectively. load versus beam end displacement ratio envelope curves were plotted
by considering the maximum load of each beam end displacement ratio
in the pull and push directions. Figs. 20a–20c show the difference in
4.6. Meshing
response for all connections along with the percent difference in results
at 2.2% drift ratio for each investigated connection in pull and push
All parts were selected as dependent parts in order to perform
directions. Percent differences are reported with respect to reference
meshing at a part level. Part partitioning in ABAQUS was utilized to
line 1 and reference line 2 for pull & push directions respectively for
incorporate model discretization and break down the model into small
each connection.
elements. All beam and column parts were modeled using a similar
As shown in the results, there is good agreement between the pro­
calibrated mesh size. Finer mesh leads to a narrower band of localization
posed FEA model and experimental results for all investigated precast
and after a while the equations fail to converge numerically. For that
connections. Maximum percent difference is 11% which is considered
reason, the model becomes mesh-size dependent, as happens with most
acceptable for this comparison. In general, the difference between the
plasticity-based models that show strain softening. The challenge is to
two curves could be related to the following aspects: uncertainty in the
nonlinear CDP material properties, uncertainty of the actual pre-
tensioning bolt force, effectiveness of the boundary conditions in the
experiments versus numerical modeling, application and quality of the
weld in experiments versus numerical modeling, etc. However, the dif­
ference between the reaction force at various cycles was within
acceptable ranges. Additionally, connection D6 shows the highest ca­
pacity compared to the other connections, including the equivalent
monolithic connection M1 as shown in Fig. 20a. Specimens D1R3 and
D1R3R also show higher capacity than the monolithic reference spec­
imen M1 as shown in Fig. 20a. Precast connection D1R3H showed lower
capacity than the equivalent monolithic connection M2. Monolithic
connection M2 showed higher capacity than the monolithic connection
M1 as also shown in Fig. 20a. The rest of group D1 connections (D1,
D1R1, D1R2) and connection D5 do not perform as well as the equiva­
lent monolithic connection M1 as shown in Fig. 20b. Representative
connections of group D2, D3, D4 also show limited capacity compared to
Fig. 14. Applied pre-tensioning force in the rod. the monolithic connection M1 and the rest of the precast connections as

947
I. Torra-Bilal et al. Structures 35 (2022) 939–957

Fig. 15. Meshed model in concrete and steel parts.

Fig. 16. Hysteresis – Load vs. Beam Displacement (M1).

Fig. 17. Hysteresis – Load vs. Beam Displacement (D1R2.

948
I. Torra-Bilal et al. Structures 35 (2022) 939–957

Fig. 18. Damage distribution of Specimen M1 at the end of testing in (a) experiments and (b) FEA.

Fig. 19. Damage distribution of Specimen D1R2 at the end of testing in (a) experiments and (b) FEA.

shown in Fig. 20c. In summary, Table 3 presents the load capacities of all As shown in Fig. 21, the rotation increases as the distance away from
investigated connections in both pull and push directions at the last the beam face increases. For instance, rotation of top and bottom nodes
loading stage (4% drift ratio for most connections). at 350 mm away from the beam face is larger than the rotation of the
Furthermore, analytically computed yield and maximum load ca­ nodes 200 mm away. Similarly, the horizontal displacement of top and
pacity of the specimens are presented in Table 4 together with the bottom nodes 500 mm away from the beam face is higher than other
experimentally determined load and energy dissipation capacities. Beam measured rotation at 200 and 350 mm away from the beam face. This
longitudinal reinforcement and additional horizontal rebars used pattern is consistent among all other connection, which verifies the
around the connection region were taken into account when computing behavior of these connections. Moment-rotation curves compare well.
the load capacities. For monotonic specimens, the measured load ca­ The only discrepancy between the measured and numerical response at
pacities are in close agreement with the computed maximum load ca­ the location of 500 mm away from the column faces, mainly shown in
pacities. For some of the specimens the maximum measured load late stages of loading, could be related to erroneous displacement
capacities remained below the computed yield load capacity, indicating measurements in the experiments. The displacement transducers used to
no beam reinforcement yielding. determine beam rotation were attached on beam element with steel
The response of precast connections was further investigated in dowels embedded inside concrete. Concrete damage around one of these
moment rotation curves and was compared with experimental results. dowels might have caused the erroneous displacement readings. This
Rotation during experiments was measured at a distance of 200, 350 and does not occur in the finite element results which is why a more stable
500 mm away from the column face by using three sets of linear variable response is observed.
differential transformers (LVDTs) and obtaining horizontal displace­ In conclusion, the comparison of the numerical results with the
ment values at top and bottom surface of the beam. Similar results were experimental data proves the validity of the proposed modeling
obtained from FEA by using a probe at the node locations similar to the approach and shows that the numerical model could accurately capture
experiments. A set of nodes was selected with the center at the same the behavior of the precast beam-to-column connections. This allows for
distances away from the column face in order to avoid local damage that further investigation of these connections’ behavior.
potentially influences results. Rotation was plotted against the
maximum moment, which was calculated based on the maximum 6. Numerical modeling results and discussion
resultant force at the beam end. Experimental vs FEA results compare
well as shown below for a sample precast specimen, D1. Similar re­ The goal of this paper is to evaluate the performance of precast
sponses were obtained from other investigated precast connections. connections with various detailing schemes in comparison with the

949
I. Torra-Bilal et al. Structures 35 (2022) 939–957

Fig. 20a. Load vs. End Displacement Ratio Envelope curves for various connections.

corresponding monolithic connection. To achieve this study’s objec­ damage caused a considerable amount of stiffness degradation espe­
tives, connection responses associated with stiffness degradation and cially at high load cycles. The additional U-shaped bars and stirrups
energy dissipation were obtained and investigated. Additionally, failure provided in connection D1R2 provided better confinement and as a
modes of connection parts, captured from the finite element models result concrete damage and rapid degradation of load-carrying capacity
were studied to determine the influence of various detailing schemes on of the specimen were limited with this revision. Providing wider beam
the overall performance of the connections, in comparison with the plates in connection D1R3 and providing increased beam longitudinal
reference equivalent monolithic connection. reinforcement ratio in D1R3H resulted in a stable response without a
significant degradation in the load resisting capability similar to the
monolithic connection. The absence of pre-tensioning effect in threaded
6.1. Stiffness degradation
rods used for fixing the beam element to the column face resulted in not
much difference in terms of stiffness. Stiffness values of Specimen
Variation in the stiffness of specimens during successive load cycles
D1R3R were less than those of Specimen D1R3 for the beam end
was determined and the stiffness degradation characteristics were
displacement ratio cycles smaller than 2.20% which is related to sepa­
compared. Stiffness of specimens in each loading cycle was taken as the
ration of the angles from the column face at early stages due to lack of
slope of the line connecting the maximum drift points in pull and push
pre-tensioning in rods. However, stiffness values of these two specimens
directions. The stiffness determined this way is termed as “secant stiff­
are very close to each other at beam end displacement ratio values
ness” or “peak-to-peak stiffness” and is derived from the
higher than 2.20 % since at higher drift ratios, the separation between
load–displacement curves presented in Figs. 20a through 20c. Figs. 22a
angles and column face also occurs in Specimen D1R3. Connection D2
and 22b show stiffness degradation vs. drift ratio curves for all investi­
resulted in low stiffness and strength due to lack of anchorage bars in the
gated connections, monolithic and precast, based on finite element
connection region. Proposed detailing of connection D3 also showed
experimental results. These results also validate the proposed finite
limited strength and stiffness, due to consisting of one piece of UPN-65
element methodology.
profile oriented in the weak axis. Strength and stiffness of proposed
As shown in Figs. 22a and 22b, both the precast beam-column
connection D4 and D5 also remained limited compared to that of the
specimens and the monolithic specimen suffered from stiffness degra­
monolithic one. Specimen D6 had high stiffness values than the mono­
dation with increasing drift ratio. The decrease in stiffness was very
lithic specimen due to the embedded bars potentially acting as longi­
rapid at the beginning of loading and the rate of change decreased at
tudinal reinforcement. Additionally, the critical region of the beam is no
higher drift ratio values. Increasing the number of anchorage bars from
longer at the column face but shifts away towards the beam due to the
three in connection D1 to four and providing stiffeners for the seat and
location of the plates in the beam which prevents concrete cracking in
top angles improved the strength and stiffness of the steel connection
the connection vicinity. Overall, connections D1R3, D1R3H and D6
components in Specimen. As a result, these connection components were
showed the highest strength and stiffness compared to other precast
observed to remain almost rigid, and the entire damage concentrated on
connections. These connections also showed stable performance and
beam element near the connection region. Such extensive localized

950
I. Torra-Bilal et al. Structures 35 (2022) 939–957

Fig. 20b. Load vs. End Displacement Ratio Envelope curves for various connections.

Fig. 20c. Load vs. End Displacement Ratio Envelope curves for various connections.

951
I. Torra-Bilal et al. Structures 35 (2022) 939–957

Table 3
Load capacity values of all connections in push and pull directions – FEA vs. Experimental (kN).
M1 M2 D1 D1R1 D1R2 D1R3 D1R3R D1R3H D2 D3 D4 D5 D6

Exp.-Pull 72.9 106.7 34.8 57.6 59.8 84.9 88.1 91.6 24.6 33 29.3 30.1 109.8
FEA-Pull 69.9 104.7 35.2 61.6 63.47 79.3 90.1 86.8 23.2 31.5 24.6 24.9 104.7
Exp-Push 72.6 101.6 28.6 50.6 58.8 82.4 82.3 80.6 21.4 30.2 33.5 25.4 99.4
FEA-Push 69.4 103.3 28.9 58.9 61.8 78.5 88.3 74.6 19.4 29.7 28.9 19.9 90.1

connection D1R3H since the same anchorage detail and same beam
Table 4
longitudinal reinforcement ratio was investigated. The capacity of pre­
Measured and computed capacities
cast connection D1R3H remained limited compared to that of mono­
Specimen Specimen Measured load Total Computed load lithic connection M2. Stiffness of Specimen D1R3H was lower than
type name capacity kN dissipated capacity kN
Specimen M2 especially in early cycles before 1.0% beam end
energy, N-m
Pull Push Yield Max. displacement ratio. However, both specimens showed similar stiffness in
dir. dir. cap. cap.
the proceeding cycles.
Monolithic M1 72.9 72.6 29.7 62.5 73.8
M2 106.7 101.6 42.1 85.6 100.6
Precast D1 34.8 28.6 7.7 62.5 73.8 6.2. Energy dissipation
D1R1 57.6 50.6 10.1 62.5 73.8
D1R2 59.8 58.8 8.3 72.5 83.1
One of the requirements for the satisfactory seismic performance of a
D1R3 84.9 82.4 15.3 72.5 83.1
D1R3R 88.1 82.3 21.4 72.5 83.1 structure is capability to sufficiently dissipate energy. It is generally
D1R3H 91.6 80.6 22.7 88.1 101.9 known that the main source of energy dissipation in reinforced concrete
D2 24.6 21.4 5.1 62.5 73.8 frame structures is through inelastic deformation at beam and column
D3 33.0 30.2 9.2 110.6 128.7 ends and beam-column connections. Similarly, in this study the amount
D4 29.3 33.5 5.9 110.6 128.7
D5 30.1 25.4 4.9 62.5 73.8
of energy being dissipated is related to the connection detailing and its
D6 109.8 99.4 19.3 111.1 131.1 overall behavior and failure modes: yielding of steel, cracking of con­
crete, frictional contact in the model, etc. The ability of specimens to
dissipate energy was evaluated based on each cycles’ cumulative dissi­
higher strength and stiffness compared to that of the monolithic pated energy. Energy dissipated during each loading cycle was deter­
connection M1. Monolithic connection M2 with increased beam longi­ mined as the area enclosed by the hysteresis loop. The dissipated energy
tudinal reinforcement ratio showed higher strength and stiffness as calculated this way on a per cycle basis was then summed up for suc­
compared to monolithic connection M1. This connection was mainly cessive cycles to get the cumulative dissipated energy. Cumulative
used as a reference specimen to compare with the behavior of dissipated energy for each specimen plotted with respect to the

Fig. 21. Moment rotation curves for D1 connection at 200, 350 and 500 mm away and FEA node sets.

952
I. Torra-Bilal et al. Structures 35 (2022) 939–957

Fig. 22a. Stiffness Degradation.

Fig. 22b. Stiffness Degradation.

953
I. Torra-Bilal et al. Structures 35 (2022) 939–957

Fig. 23a. Energy Dissipation.

Fig. 23b. Energy Dissipation.

increasing level of drift ratio is presented in Figs. 23a and 23b based on small. Some precast connections performed better than others in terms
finite element and experimental results. In addition, Table 5 presents of energy dissipation, but the plot also indicates the limited energy
cumulative dissipated energy values at 4% drift ratio for all investigated dissipation ability of the precast concrete specimens as compared to the
connections. Good agreement was obtained when comparing FEA re­ monolithic specimen. Connections D1R3, D1R3H and D1R3R showed
sults to experimental results in energy dissipation curves. higher dissipated energy than other connections in Group D1 connec­
As shown in Figs. 23a and 23b, most of the energy was dissipated by tions. This is attributed to the improved detailing of the connections that
the specimens after some level of damage has occurred in beam elements allowed the connection components to dissipate more energy while still
and steel connection components. For all specimens, the total dissipated showing good levels of strength and stiffness. However, the total energy
energy up to the beam end drift ratio of 1% remained considerably dissipated by Specimen D1R3 and D1R3R during the load test was

Table 5
Cumulative dissipated energy values of all connections in push and pull directions – FEA vs. Experimental (kJ).
M1 M2 D1 D1R1 D1R2 D1R3 D1R3R D1R3H D2 D3 D4 D5 D6

Exp. 29.7 42.1 7.7 10.1 8.3 15.3 21.4 22.7 5.1 9.2 5.9 4.9 19.3
FEA 26.6 40.1 8.3 9.1 14.5 14.22 19.8 21.5 4.7 9.0 4.9 4.1 17.5

954
I. Torra-Bilal et al. Structures 35 (2022) 939–957

approximately 70% and 72% of that of the monolithic specimen M1, connection component showed the onset of failure. Failure is mainly
respectively. Considerable difference occurred between the energy defined as the drift ratio where stresses during load cycles first exceed
dissipation capacities of Specimens M2 and D1R3H, which reached a the compressive strength (f’c) for concrete or yield stress (fy) for steel for
ratio of about 40% at the end of the test. Connections D2, D3, D4 and D5 each connection part as defined in Table 1.
showed significantly low amount of dissipated energy compared to the As shown in Table 6, the damage development in the monolithic
reference monolithic connection M1 as well as other precast connec­ specimens was as following: initial cracking of concrete followed by the
tions. Connection D6 showed higher dissipated energy than other pre­ tensile yielding of longitudinal bars and transverse reinforcement. It was
cast connections, which is due to the proposed detailing of this observed from FEA that cracking of concrete initially occurred in the
connection, yet the total energy dissipated by Specimen D6 during the beam and dispersed in the column around the column ends and
load test was approximately 73% of that of the monolithic specimen. connection region as well. The typical damage development was
Additionally, the energy dissipated by monolithic connection M2 is essentially the same for the precast connections with some differences
higher than that of M1 due to larger area enclosed by the hysteresis loops due to various provided detailing. All specimens exhibited early angle
which results from higher capacity of using higher longitudinal rein­ yielding followed by beam plates yielding, concrete cracking and ulti­
forcement ratio. Similar observations are valid for the experimentally mately other steel components yielding. Concrete damage occurred
determined energy dissipation capacity of the specimens tabulated in mainly in the beam with no effect to the column in the precast con­
Table 3. The relatively limited energy dissipation capacity of the precast nections and most of steel that yielded was localized at the beam cracks,
specimens as compared to their monolithic counterparts is due to closer to the connection region, rather than further away from the col­
pinching present in the load-drift response of the precast specimens. The umn face. In Specimen D1, which had the original connection detail,
pinched hysteresis loops of these specimens. Resulted mainly from the almost the entire damage was localized in the connection region
deformation of connection angles and connection rods. including anchorage bars, beam plates and angles. The bars yielded at
2.2% drift ratio, the angles showed damage at early cycles, 0.5% drift
ratio and beam plates yielded at 1% drift ratio. Concrete failure occurred
6.3. Failure modes at 2.2% beam end ratio which is at a later stage compared to the
monolithic connection. As a result of such localized extensive damage,
Another aim of this study is to further investigate the failure modes of the load capacity of the specimen remained limited. Specimen D1R1
each connection part throughout all drift ratios (up to 4%) and identify showed a more rigid behavior due to the additional provided anchorage
which connection components are critical and demonstrate failure bars and stiffeners at the angles’ seats. As a result, initial concrete
initiation under applied loading. Failure modes identified and investi­ cracking occurred earlier than specimen D1 (1% ratio), which led to less
gated include concrete crushing, reinforcement yielding, angle and developed stresses in the angles that yielded slightly later compared to
beam plates yielding, bolts failure, and other steel components yielding. connection D1, followed by the beam plates and the rest of the steel bars.
One of the advantages of the finite element modeling is the capability of Specimen D1R2 showed initial failure in the angles and later concrete
capturing all failure modes throughout various drift ratios (0.5%, 1%, damage due to better provided confinement. Also, no stirrups failure
2.2%, 4%). Table 6 shows the beam end drift ratio at which each

Table 6
Failure Mode Initiation - Beam End Drift Ratio (%).
M1/M2 D1 D1R1 D1R2 D1R3 D1R3R D1R3H D2 D3 D4 D5 D6

Concrete Crushing 0.5 2.2 1.0 2.2 2.2 2.2 2.2 1.0 1.0 1.0 1.0 1.0
Longitudinal Bars Yielding 1.0 2.2 2.2 2.2 2.2 2.2 2.2 2.2 2.2 2.2 2.2 2.2
Stirrups Yielding 1.0 2.2 4.0 No 2.2 2.2 2.2 No No No No No
Bolts Failure
N/A
No No No No No No No No No No No
Beam Plates
Yielding 1.0 2.2 1.0 2.2 2.2 2.2 1.0 1.0 1.0 N/A 1.0
Angles Yielding 0.5 1.0 0.5 0.5 1.0 1.0 0.5 0.5 0.5 0.5 0.5
Z-shape Bars Yielding 4.0 2.2 4.0 2.2 2.2 2.2
N/A N/A N/A N/A
N/A
U-shape Bars Yielding
N/A
N/A
4.0
2.2 2.2 2.2 No No N/A
UPN-65 Steel Sections Yielding
N/A
N/A
N/A
N/A
No No N/A No
Vertical Anchor Bars Yielding
N/A
N/A No N/A
Angles as beam plates yielding (D5)
N/A 2.2
#20 Anchor Bars Yielding (D5) No
#16 Anchor Bars yielding (D6)
N/A 2.2
Vertical plates Yielding (D6) 2.2
N/A = Not Applicable
No = No failure obtained

955
I. Torra-Bilal et al. Structures 35 (2022) 939–957

was reported for the same reason. Z-shape and U-shape bars yielded at response of these specimens is the localization of damage in beam ele­
late stages of loading. ments within the connection region. Precast connection D6 showed very
D1R3 revision led to failure of beam plates at a later cycle due to good performance in load displacement and stiffness responses as
larger provided surface that transfers the applied force to the connec­ opposed to the corresponding monolithic specimen M1. However, the
tion. The main connecting element that showed immediate failure, total energy dissipated by specimen D6 was found to be at lower levels
similar to other connections were the angles connecting the beam to (73%) than that of monolithic specimen M1. Generally, precast con­
column. Concrete cracking and most of the other steel components failed nections D1R3 and D6 performed more optimally as compared to all
at 2.2% beam end drift ratio. Similar failure modes were obtained for other precast connections and their corresponding monolithic specimen.
D1R3R connection which was characterized by the lack of pre- Response of this specimen is especially characterized by high stiffness
tensioning in the rods. The only difference was the onset of failure in and load capacity. Connection D6 exhibited the highest dissipated en­
the angles that was delayed in specimen D1R3R (1%) compared to ergy as opposed to D1R3 and all other investigated precast connections.
specimen D1R3 (0.5%). Specimen D1R3H with increased beam rein­ The results form this study and the developed finite element method­
forcement ratio placed the connection into a more severe test due to ology will be utilized in the future to further improve the performance of
larger force demands resulting from increased moment capacity of the precast connections.
concrete. As a result, lower stresses are developed in the concrete
without immediately reaching failure that got transferred to the rein­ 7. Conclusion
forcing steel. The latter one was able to withstand more loading in a
ductile manner. Consequently, this connection was capable of dissi­ A detailed nonlinear finite element model was developed to predict
pating more energy. Specimen D2 showed initial failure in the angles, the behavior and performance of precast connections with novel various
followed by the beam plates and concrete cracking and lastly, yielding of detailing schemes under cyclic loading and compare with a reference
the longitudinal bars. No failure was observed in the other connection monolithic connection. Concrete Damage Plasticity (CDP) model with
parts. This connection did not consist of anchorage bars extended in the softening effect and high dilatancy angle and Bilinear kinematic steel
beam which determined the location of the plastic hinge to be closer to model with bond-slip effect were applied to the numerical procedure to
the column face, resulting in limited strength, stiffness and energy model concrete and steel nonlinearly under reversed cyclic loading. The
dissipation. Specimen D3, D4 and D5 followed the typical damage finite element models were capable of capturing all possible failure
development of precast connections. Not much of the steel components modes in the connection parts throughout load cycles as well capture the
were able to yield which resulted in limited energy dissipation in these behavior of the connections in terms of load displacement, moment-
connections. Specimen D6 followed the typical damage development rotation, strength degradation and energy dissipation responses. Com­
but showed a desired behavior due to the location of the beam plates in parison of numerical results with experimental results in
the beam which shift away the critical region of the beam as well as the load–displacement relationships exhibited very good agreement. Some
embedded bars acting as additional longitudinal beam reinforcement. of the model characteristics that highly influenced the results in
Most of the steel components yield at 2.2% beam end ratio. Overall, no obtaining good comparison between experimental data and finite
bolt failure was obtained for all connections. element results include: element selection for both concrete and rein­
forcement modeling, mesh size, concrete softening coefficient in
6.4. Discussion on performance of different connection details modeling concrete nonlinearly, definition of damage parameters for
CDP, dilatancy angle in CDP parameters, bond-slip effect of steel ma­
The behavior of precast specimens D1, D1R1, D1R2 and D1R3 was terial, and interactions between model parts. The following conclusions
compared to monolithic specimen M1 due to the presence of ϕ16 beam can be reached from this investigation based on finite element results:
longitudinal bars. Connection D1R3 showed a performance equivalent
to that of the monolithic specimen in terms of load capacity and stiff­ 1. Concrete failure is delayed in precast connections compared to the
ness. Although load carrying capacity and stiffness values of Specimen monolithic one. The initial failure in the precast connections is
D1R3 was very close to the values of Specimen M1, the energy dissi­ localized in the connecting angles followed by the embedded beam
pation mechanism remained limited, even though it was better than plates.
other precast connections (D1, D1R1, D1R2). Connection D1R3R 2. No bolt failure was obtained for all investigated connections and bolt
showed very similar behavior in terms of load carrying capacity and pre-tensioning effect did not have a major effect on the overall
dissipated energy to connection D1R3. Connection D1R3H behavior was connection behavior but causes earlier angle failure in the
compared to monolithic connection M2 due to the presence of ϕ20 beam connection.
longitudinal bars. These two specimens showed similar behavior in 3. Most connections including the monolithic ones experience high
terms of measured load capacities and stiffness in all load cycles. stiffness degradation at earlier load cycles. Some precast connections
However, considerable difference occurred between the energy dissi­ including the last revision of group D1 (D1R3, D1R3R, D1R3H) and
pation capacities of Specimens M2 and D1R3H. Beneficial effect of the D6 showed higher strength and stiffness compared to the monolithic
revisions performed on connection detail D1 is evident in the response of specimen.
specimens. Providing additional anchor bars and stirrups inside the 4. Most of the energy was dissipated by the specimens after some level
beam element within the end region, as well as providing a longer weld of damage has occurred in beam elements and steel connection
length improved the behavior in terms of load capacity, stiffness and components (after 1% drift. ratio). The maximum achieved energy
energy dissipation. With this connection detail, the presence of pre­ dissipation in the proposed connections occurred in connection D6
tensioning force in connection rods did not result in an appreciable and was measured at approximately 73% of the corresponding
difference in the overall response. Connection detail D1 also exhibited a monolithic connection M1. Precast connection D1R3H also showed
stable response without a significant degradation in load resisting limited amount of dissipated energy compared to the reference
capability when used with a higher beam longitudinal reinforcement specimen M2.
ratio. 5. Based on the developed load–displacement ratio envelope curves,
Connection D2, D3, D4 and D5 did not demonstrate satisfactory the maximum capacity achieved in precast connections is 110 kN
behavior when compared to the corresponding monolithic specimen M1 which is about 1.6 times of that of the equivalent reference mono­
due to their detailing schemes. Specimens utilizing these connection lithic specimen (M1) reported at about 70kN. The capacity of
details exhibited limited load capacity and stiffness, which led to poor connection M2 was found to be 1.5 times higher than that of
energy dissipation capability. The main reason for the relatively poor connection M1 due to the increase in the beam longitudinal

956
I. Torra-Bilal et al. Structures 35 (2022) 939–957

reinforcement ratio. Analogously, an increase in the capacity by 20% [6] Najafgholipour MA, Dehghan SM, Dooshabi A, Niroomandi A. Finite element
analysis of reinforced concrete beam-column connections with governing joint
maximum was also obtained when increasing the beam longitudinal
shear failure mode. Latin Am J Solids Struct 2017.
reinforcement ratio in the precast specimens (D1R3H vs D1R3). [7] Bahrami S, Madhkhan M, Shirmohammadi F, Nazemi N. Behavior of two new
6. Minor detailing modifications influence the development of the moment resisting precast beam to column connections subjected to lateral
failure modes, as well the location of the critical region, hence loading”. Eng Struct 2017;132:808–21.
[8] French CW, Amu O, Tarzikhan C. Connections between precast elements-failure
affecting the performance of the connections. outside connection region. J Struct Eng 1989;115(2):316–40.
7. This study’s results can be utilized to further explore precast con­ [9] French CW, Hafner M, Jayashankar V. Connections between precast elements
nections with numerous detailing schemes to consequently enhance failure within connection region. J Struct Eng 1989;115(12):3171–92.
[10] Choi HK, Choi YC, Choi CS. Development and testing of precast concrete beam-to
their behavior. The influence of various parameters on the optimal column connections. Eng Struct 2013;56:1820–35.
performance of the precast connections will also be studied by the [11] Parastesh H, Hajirasouliha I, Ramezani R. A new ductile moment-resisting
authors in the future to conclusively develop a design procedure. connection for precast concrete frames in seismic regions: an experimental
investigation. Eng Struct 2014;70:144–57.
[12] ABAQUS (2014). Analysis user’s manual 6.14-2, Dassault Systems Simulia Corp.,
Declaration of Competing Interest Providence, RI.
[13] ABAQUS (2013), Theory Guide, Version 6.13, Dassault Systèmes solidworks corp,
Waltham, Massachusetts.
The authors declare that they have no known competing financial [14] American Concrete Institute (ACI) (2005) “Acceptance criteria for moment frames
interests or personal relationships that could have appeared to influence based on structural testing and commentary: ACI 374.1-05” Detroit (MI).
the work reported in this paper. [15] Baran E, Mahamid M, Baran M, Kurtoglu M, Torra-Bilal I. Performance of a
moment resisting beam-column connection for precast concrete construction. Eng
Struct 2021. https://doi.org/10.1016/j.engstruct.2021.113005.
References [16] Dowling NE. Mechanical behavior of materials. New Jersey: Prentice-Hal
International; 2007.
[1] Mitchell D, DeVall RH, Saatcioglu M, Simpson R, Tinawi R, Tremblay R. Damage to [17] American Concrete Institute (ACI) (2014): Building code requirements for
concrete structures due to the 1994 Northridge earthquake. Can J Civ Eng 1995;22 structural concrete and commentary.’’ ACI 318-14.
(2):361–77. [18] Feng D-C, Ren X-D, Li J. Softened damage -plasticity model for analysis of cracked
[2] Arslan MH, Korkmaz HH, Gulay FG. Damage and failure pattern of prefabricated reinforced concrete structures. J Struct Eng 2018.
structures after major earthquakes in Turkey and shortfalls of the Turkish [19] Vecchio FJ, Collins MP. The modified compression-field theory for reinforced
earthquake code. Eng Fail Anal 2006;13:537–57. concrete elements subjected to shear. ACI J Proc, 83, 219–231.
[3] Nzabonimpa JD, Hong W-K, Kim J. Nonlinear finite element model for the novel [20] Vecchio FJ, Collins MP. Compression response of cracked reinforced concrete.
mechanical beam-column joints of precast concrete-based frames. Comput Struct J Struct Eng 1993;119(12):3590–610.
2017;189:31–48. [21] Genikomsou AS, Polak MA. Finite element analysis of punching shear of concrete
[4] Feng D, Gang W, Lu Y. (2018): Finite Element modelling approach for precast slabs using damaged plasticity model in ABAQUS. Eng Struct 2015;98.
reinforced concrete beam-to-column connections under cyclic loading , Eng Struct. [22] Torra-Bilal I , Mahamid M, Bilal K (2019): Performance of precast connections on a
[5] Hawileh RA, Rahman A, Tabatabai H. Nolinear finite element analysis and moment resisiting frame under cyclic loading , Struct Congress.
modeling of a precast hybrid beam-column connection subvjected to cyclic loads.
Appl Math Model 2009;34:2563–83.

957