Applied Mathematical Modelling 34 (2010) 2562–2583

Contents lists available at ScienceDirect

Applied Mathematical Modelling

journal homepage: www.elsevier.com/locate/apm

Nonlinear finite element analysis and modeling of a precast hybrid

beam–column connection subjected to cyclic loads
R.A. Hawileh a,*, A. Rahman b,1, H. Tabatabai c,2
a
Department of Civil Engineering, American University of Sharjah, P.O. Box 26666, United Arab Emirates
b
Department Chair, Department of Civil Engineering and Mechanics, University of Wisconsin-Milwaukee, WI 53211, United States
c
Department of Civil Engineering and Mechanics, University of Wisconsin-Milwaukee, WI 53211, USA

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

Article history: In this work, a detailed three-dimensional (3D) nonlinear finite element model is devel-
Received 19 August 2008 oped to study the response and predict the behavior of precast hybrid beam–column con-
Received in revised form 18 November 2009 nection subjected to cyclic loads that was tested at the National Institute of Standards and
Accepted 24 November 2009
Technology (NIST) laboratory. The precast joint is modeled using 3D solid elements and
Available online 3 December 2009
surface-to-surface contact elements between the beam/column faces and interface grout
in the vicinity of the connection. The model takes into account the pre-tension effect in
Keywords:
the post-tensioning strand and the nonlinear material behavior of concrete. The model
Computational mechanics
Finite element analysis
response is compared with experimental test results and yielded good agreement at all
Precast concrete connections stages of loading. Fracture of the mild-steel bars resulted in the failure of the connection.
Hybrid frame connection In order to predict this failure mode, stress and strain fields in the mild-steel bars at the
Mild-steel bars beam–column interface were generated from the analyzed model. Such fields of stresses
Cyclic loads and strains are hard to measure in experimental testing. In addition, the magnitude of
Contact elements the force developed in the post-tensioning steel tendon was also monitored and it was
observed that it did not yield during the entire loading history. Successful finite element
modeling will provide a practical and economical tool to investigate the behavior of such
connections.
Ó 2009 Elsevier Inc. All rights reserved.

1. Introduction

Precast concrete systems for buildings are cost-efficient, structural systems that provide speed and ease of erection. They
allow for improved quality control in the precast plants, and freedom in the architectural form of the members. The in-
creased usage of such members is attributed to an increased interest on the part of contractors and engineers in order to
find competitive alternatives to cast-in-place concrete elements.
In the late eighties, there was a growing awareness and need in the United States for developing design methodologies for
precast, seismic, resistant structural systems based on a sound research background. For this reason, the US Precast Seismic
Structural Systems program (US-PRESSS) was initiated in 1989 to meet the needs for developing effective Precast Seismic
Structural Systems and corresponding design recommendations [1].
An overview of the PRESSS program was provided by Priestley [1]. Priestley stated that the fundamental objectives of the
program were to develop effective seismic structural systems for precast buildings, new materials, concepts, technologies,

* Corresponding author. Tel.: +971 6 515 2496; fax: +971 6 515 2979.
E-mail address: [email protected] (R.A. Hawileh).
1
Tel.: +1 414 229 2860; fax: +1 414 229 6958.
2
Tel.: +1 414 229 5166; fax: +1 414 229 6958.

0307-904X/$ - see front matter Ó 2009 Elsevier Inc. All rights reserved.
doi:10.1016/j.apm.2009.11.020
 R.A. Hawileh et al. / Applied Mathematical Modelling 34 (2010) 2562–2583 2563

and comprehensive and rational seismic design recommendations based on fundamental and basic research data. This was
intended to emphasize the viability of precast construction in the various seismic zones, and was to be incorporated into the
model building codes.
Despite many advantages of precast concrete, it is not widely used throughout the United States, especially in regions of
high seismic risk. According to Priestley [2], the reason behind this is a lack of confidence and knowledge base about their
performance in seismic regions as well as the absence of rational seismic design provisions in major model building codes.
Several precast parking structures performed poorly in the 1994 Northridge earthquake, due to incorrect detailing.
In the report presented by the Precast Seismic Structural Systems [3], a total of five different seismic structural systems
made from precast concrete elements were proposed. These systems formed various parts of the structural framing of the
experimental building in the PRESSS Phase III tests at the University of California in San Diego. The un-bonded, post-ten-
sioned frame with damping (hybrid frame) performed very well in the PRESSS evaluation. The frame offers an important fea-
ture that re-centers the structural frame (eliminate residual drift) after an earthquake. This feature is not available in the
typical framing systems recognized in the 1997 Uniform Building Code (UBC).
The hybrid frames consist of precast elements, un-bonded post-tensioning (PT) steel, and bonded reinforcement bars. The
mild-steel is used to dissipate energy by yielding in tension and compression, and the post-tensioning steel is used to restore
the frame to its original configuration and maintain continuity between the beam and the column after earthquake. In other
words, the post-tensioning steel is used to clamp the beam against the column, providing the shear resistance through fric-
tion developed at the beam–column interface. The hybrid system does not require corbels and thus offers a clean and eco-
nomical architectural appearance.
All beams at the same level in a hybrid frame are post-tensioned for the entire length of the frame with un-bonded post-
tensioned (PT) tendons placed in ducts at the beam’s mid height. The PT tendon is placed in the center of the beam to reduce
the deformations and strains caused by beam rotation. It is very important to maintain the post-tensioning steel in the elas-
tic range. This is achieved by debonding the post-tensioning steel so that once the gap opens at the beam–column interface,
the strand deformation is distributed over a length that is long enough to maintain the total strain in the PT strands below
yield. The prestressing steel offers little opportunity for energy dissipation. The energy dissipation is only provided by the bar
reinforcement. The deformed bars are placed in ducts and grouted at the top and bottom of the beam, which extends through
the columns. They are placed at the top and bottom of the beams, because this would maximize the dissipation of energy
through cyclic yielding in tension and in compression. In order to protect the mild-steel from fracture and to reduce its strain
level at high story drifts, they are debonded at the beam–column interface region for a short length (few inches) as shown in
Fig. 1.
The construction joint between the beam and column is typically made of a fiber-reinforced grout. The grout provides for
necessary erection tolerances and reduces potential damage to the concrete beam itself. Cheok et al. [4] recommended that
the beam–column interface be completely grouted with a non-shrink grout. The non-shrink grout should have good work-
ability and toughness to gain the maximum friction resistance and to avoid crushing of the grout under high compressive
forces.
An experimental program was conducted at the laboratories of the National Institute of Standards and Technology (NIST)
in Gaithersburg, Maryland [5] to help develop design guidelines for precast concrete frames with hybrid connections.
The objective of this research is to develop a finite element model (FEM) using ANSYS 8.0 [6] to predict the response and
behavior of precast hybrid beam–column connections subjected to cyclic loads. In a previous study, the authors [7] devel-
oped a non-dimensional design procedure for such type of precast/prestressed hybrid connections. The developed model
takes into consideration the nonlinear behavior of concrete, contact between surfaces, and pre-tensioning. The results of
the FE model were validated with experimental test results that were conducted at NIST [5]. Experiments can be ideal to
study the behavior and failure of structures. However, they can be costly and time consuming. They require adequate

Reinforcing bar debonded

Mild Reinforcing (A706) locally
(grouted)

Post-tensioning
Tendon (unbonded)
Fiber-reinforced
Grout

Fig. 1. Elements of hybrid frame [3].

2564 R.A. Hawileh et al. / Applied Mathematical Modelling 34 (2010) 2562–2583

facilities, space, setup, and manpower. If done properly, computational models could be used as an attractive alternative to
the expensive experimental investigation.

2. Mathematical model

The mathematical model of the structure with precast hybrid beam–column joints to be analyzed is discretized by divid-
ing it into a mesh of finite elements. Thus the fully continuous field of the precast hybrid joint structure is represented by an
equivalent piecewise continuous field by a finite number of nodal quantities with simple interpolation within each element.
The finite element analysis starts with the subdivision of the physical system into discrete elements. The formulation is
based on a variational principle related to virtual work. Applying the virtual work principle to a finite element of volume V
bounded by a surface S, the following relationship can be obtained:

dU ðeÞ ¼ dW ðeÞ ; ð1Þ

(e) (e)
where dU is the virtual strain energy due to internal stresses and dW is the virtual work of external forces on the element.
dU(e) and dW(e) can be expressed using matrix notation as follows:
Z Z Z
dU ðeÞ ¼ deT rdV; ð2Þ
V

and
Z Z
dW ðeÞ ¼ dwTs TdS; ð3Þ
S

where de is the vector of strains produced by the virtual displacement vector dws. The symbol d has the same meaning as d for
differential, but by convention d is used when the displacements are virtual. r is the stress vector and T is the surface force
per unit area matrix. The displacement functions can be related to nodal displacements d by shape functions as:

ws ¼ N s d; ð4Þ
where Ns is the shape function matrix evaluated on the surface S where traction T occurs. In addition, strains are related to
nodal displacements as:

e ¼ Bd; ð5Þ
and stresses are related to strains by using constitutive equation as:

r ¼ De; ð6Þ
where B is the strain displacement matrix and D is the element material matrix. Hence substituting Eqs. (4)–(6) into Eq. 1 the
following equation can be derived:
0 1
Z Z Z Z Z
T T
dd @ BT DBdV Ad ¼ dd N Ts TdS; ð7Þ
S
V
RRR T RR T
where V
B DBdV is the element stiffness matrix K and S
N s TdS is the vector of surface loads fs. Eq. 7 is the basic equa-
tion for the finite element discretization and can be rewritten as follows:

Kd ¼ f s : ð8Þ
More details of the formulation are given in Ref. [6]. Commercial finite element modeling (FEM) software such as ANSYS [6]
can be used to solve Eq. (8) for the unknown displacement vector d. The ANSYS [6] FEM code includes full complement of
linear and nonlinear elements, material models with different constitutive laws, and several sets of solvers. It can also handle
more complex assemblies especially those involving nonlinear contact and is the ideal choice for determining full fields of
stress, strain, and displacement at any location within the entire precast hybrid joint structure. The standard nonlinear anal-
ysis procedure was followed in which load will be applied incrementally, linear approximation and correction for each load
increment will be carried out and iteratively the final solution will be reached at a specified convergence [6].

3. 3D finite element model development

The experimental work relevant to this research is labeled as M-P-Z4 (the designation given in the PRESSS program). De-
tails of this experiment in terms of geometry of beams, columns, post-tensioned tendons, mild-steel bars, concrete reinforce-
ment, interface grout, and duct grout are generated and described below [5].
 R.A. Hawileh et al. / Applied Mathematical Modelling 34 (2010) 2562–2583 2565

4. The geometry

The finite element model is developed to accurately represent the geometrical configuration and dimensions of the test
specimen (M-P-Z4). The hybrid connection is modeled with dimensions and reinforcement details as illustrated in Figs. 2 and
3 and Table 1. The post-tensioning (PT) strands were modeled as one solid cylinder with an equivalent diameter representing
the total area of all PT strands. The PT steel is un-bonded with the surrounding concrete over a length of 1385 mm (54 in.) in
the vicinity of the connection as shown in Fig. 3. Each mild steel bar was modeled using a solid cylinder as shown in Fig. 3.
The mild-steel is un-bonded over a short length of 25.4 mm (1 in.) on either side of the beam–column interface to reduce the
strain when subjected to cyclic loading. The interface grout is 8.5 mm (1/3 in.) thick and the duct grout was modeled as a
hollow cylinder with an outer radius equal to the radius of the duct hole of 28.5 mm (1.125 in.) and an inner radius equal
to the radius of the PT steel of 9.7 mm (0.3825 in.).

5. Types of elements

Five types of elements are used in the modeling of beams, column, interface grout, duct grout, PT steel, A706 mild-steel,
and regular A615 reinforcing steel. These elements are: 3D solid elements, 3D solid concrete element, 3D spar elements, con-
tact elements, and pre-tensioning elements. A description of the different element types used in this analysis will be pro-
vided in the following subsections. The meshed structure is shown in Fig. 4. The total number of elements used in the
model is 60,810.
It should be noted that the primary reinforcements (PT and A706 mild steel reinforcement bars) are modeled using 3D
solid elements while the regular beam and column reinforcement are discritized using the discrete spar elements. The major

Fig. 2. Dimensions of test models [5]. Note: dimensions are in inches, 1 in. = 25.4 mm.
2566 R.A. Hawileh et al. / Applied Mathematical Modelling 34 (2010) 2562–2583

Major Rebars
SOLID185
Elements

Rebars Debonded

Unbonded PT Bonded PT

Rebars Bonded

Fig. 3. Isoparametric view of the structure and connection vicinity.

Table 1
Reinforcement details of test specimen (M-P-Z4) [5].

Specimen Bars (top and Area of bar Strands Area of strand PT steel distance from beam’s extreme top fiber fpe
bottom) (in.2) (in.2) (in.) (ksi)a
M-P-Z4 2#3 Grade 60 0.11 3–1/ 0.153 8 (concentric) 120
2 in.
a
1 in. = 25.4 mm, 1 ksi = 6.895 MPa.

Fig. 4. Connection isometric view with internal reinforcement.

reinforcement is modeled using solid elements because they are debonded from the adjacent concrete surfaces in the vicin-
ity of the connection as shown in Fig. 1. On the other hand, spar elements were used to model the regular beam and column
reinforcement (A615 bars) that is assumed to be fully bonded to the adjacent concrete surfaces as shown in Fig. 4.
ANSYS [6] solid elements (SOLID185) and (SOLID65) are used in this model to simulate the major reinforcement (PT steel
and A706 mild-steel bars) and the precast concrete members (beams and column), respectively, of the hybrid joint shown in
Fig. 3. Other reinforcements were modeled using LINK8 elements.
The (SOLID185) element is a 3D hexahedral element defined by eight nodes as shown in Fig. 5. The element has three
translational degrees of freedom at each node in the nodal x, y, and z directions. The element has capability of plastic defor-
mation, hyperelasticity, stress stiffening, creep, large deflection, and large strain. In addition, the element has mixed nonlin-
ear formulation that can be used for simulating deformations of nearly incompressible elastoplastic materials and fully
 R.A. Hawileh et al. / Applied Mathematical Modelling 34 (2010) 2562–2583 2567

Fig. 5. SOLID185, 3D eight node structural solid [6].

incompressible hyperelastic materials. The A706 mild-steel bars are debonded locally at the beam–column interface to dis-
sipate energy by yielding in tension and compression and the PT steel is un-bonded to distribute its deformation over a suf-
ficient length upon the opening of the gap at the beam–column interface, thus maintaining the strain in the PT tendon below
yield.
The (SOLID65) element is a 3D hexahedral element defined by eight nodes with three translational degrees of freedom in
the nodal x, y, and z directions at each node. The geometry, node locations, and the coordinate system of this element are
shown in Fig. 6. The most important aspect of this element is the treatment of nonlinear material properties. The (SOLID65)
element has the capability of plastic deformation, creep, cracking in three orthogonal directions, and crushing in
compression.
Two nodes are required to define the LINK8 [6] element. Each node has three translational degrees of freedom in the nodal
x, y, and z directions. The geometry, node locations, and the coordinate system of this element are shown in Fig. 7. The ele-
ment is also capable of plastic deformation, stress stiffening, and large deflection. The cross-sectional area of the different
steel bars used in this model is tabulated in Table 2. The orientation and location of the reinforcement is also displayed

Fig. 6. SOLID65, 3D eight node reinforced concrete solid [6].

2568 R.A. Hawileh et al. / Applied Mathematical Modelling 34 (2010) 2562–2583

Fig. 7. LINK8, 3D 2 node spar element [6].

Table 2
LINK8 bars real constants data [5].

Location Bars Cross-sectional area (in.2)a

Beams # 3 bars 0.11
Column # 6 bars 0.44
Stirrups for beams and column W3.5 wire 0.035
Stirrups for beams and column W5.5 wire 0.055
a
1 in. = 25.4 mm.

graphically in Fig. 4. Perfect bond is assumed between the reinforcing steel and concrete elements. This is achieved by con-
necting the nodes of the (LINK8) to those of adjacent (SOLID65) elements by sharing the same nodes.
The post-tensioning was modeled via a pre-tension force in the PT steel tendon elements corresponding to the tendon
initial tensile force (stress) in a preliminary load stage. The pre-tension element (PRETS179) has one translation degree of
freedom along a defined load direction. The geometry, node location and the coordinate system of this element is shown
in Fig. 8. Finite element analysis (FEA) divides the tendon into two parts separated by a pre-tension section which consists
of pre-tensioning elements. The pre-tension node should be constrained to prevent rigid body motion. Thus, perfect bond
connectivity is maintained between the pre-tensioning elements and PT steel elements.
Normal forces play the main role in transferring forces between the surfaces. Contact elements can account for friction.
Gravity and seismic shear forces are resisted by friction across the beam–column interface by the action of PT steel tendon
that provides a permanent clamping force. In this structure, the beam and column faces are in contact with the interface
grout, and there is also contact between the mild-steel bars and the grout in the vicinity of the connection. Fig. 9 and Table
3 show contact at several locations within the structure.
Two element types (CONTA174 and TARGE170) are used for the contact and target surfaces since the contact is between
two different surfaces. Surfaces with finer mesh were designated as contact surface while surfaces with coarser meshes were
considered target surfaces. The contact element has eight nodes as shown in Fig. 9.
For this model, standard unilateral contact [6] behavior with normal sticking/sliding friction behavior is used since there
is sliding with contact closing and opening behavior between connection’s surfaces. Accordingly, finite element analysis
(FEA) sets the normal pressure to zero if separation occurs between the surfaces in contact. The different values of friction
between surfaces used in this model will be discussed in the following section.

5.1. Concrete failure criteria

The concrete constitutive material model [6] used here to define the failure of concrete is based on William and Warnke
[8] formulation. An isotropic multi-linear compressive stress–strain curve for concrete is used to define the plastic behavior
of concrete. Fig. 10 represents a general 3D failure surface for concrete materials [6]. ‘‘The 3D failure surface for the states of
stress is biaxial or nearly biaxial and the most significant nonzero principle stresses are in the x and y directions. When the
principal stresses in the x and y directions are both negative (compressive), the failure mode is a function of the sign of
the principal stress in the z direction, and three failure surfaces are possible as shown in Fig. 10, with the principal stresses in
the z direction slightly greater than zero (cracking), equal to zero (crushing), and slightly less than zero (crushing). When one
or both the principal stresses in x and y directions are positive (tensile) the failure mode is cracking as shown in Fig. 10 [6]”.
 R.A. Hawileh et al. / Applied Mathematical Modelling 34 (2010) 2562–2583 2569

Fig. 8. PRETS179 Pre-tension load and element nodes [6].

‘‘Cracking typically occurs in a concrete element when the principle stress in any direction lies outside the failure surface.
Accordingly, the modulus of elasticity of the concrete element is set to zero in the direction parallel to the principle tensile
stress direction [6]. Similarly, crushing occurs when all principle stresses are compressive and lie outside the failure surface
[6]. Subsequently, the elastic modulus is set to zero in all directions [6], and the element is effectively removed. As a result,
the element local stiffness becomes zero resulting in large deformation and thus divergence in the solution.”
It was found that if the crushing (spalling) capability of the concrete is turned on, the finite element model would fail
prematurely. Crushing of the concrete started to develop in elements located directly under the loads. Subsequently, adja-
cent concrete elements crushed within several load steps as well, significantly reducing the local stiffness. Finally, the model
showed a large displacement, and the solution diverged. In this study, the crushing capability of the element was turned off
and thus yielding of the reinforcing bars and cracking of the concrete elements controlled the failure of the developed finite
element model.

6. Materials properties, meshing, boundary conditions, and loads

6.1. Material properties

To analyze and obtain an accurate prediction of the response of this structure, accurate material properties [5] and rel-
evant coefficients should be defined properly in the FE model. Components of this structure consist of the following mate-
rials: concrete, interface grout, duct grout, PT steel tendon Grade 270, A706 Grade 60 mild-steel reinforcement, and Grade 60
welded wire fabric steel reinforcement. The mechanical properties of the materials are listed in Table 4 where,
2570 R.A. Hawileh et al. / Applied Mathematical Modelling 34 (2010) 2562–2583

Fig. 9. CONTA174 contact element [6] and location of contact element within the model.

Table 3
Contacts’ locations.

Contact # Location
1 Beam and interface grout
2 Column and interface grout
3 PT steel tendon and duct holes
4 Mild-steel bars and duct holes

fc0 ¼ ultimate concrete compressive strength; in ksiðMPaÞ,pffiffiffiffi pffiffiffiffi

fr = modulus of rupture of concrete assumed equal to 7:5 fc0 in psi (0:62 fc0 in MPa), [9].
E = modulus of elasticity, in ksi (MPa).
Fig. 11 shows a typical stress–strain curve for normal weight concrete [10]. The uniaxial stress–strain relationship for
concrete in compression is simulated with multi-linear isotropic hardening material. Numerical expressions [11], Eqs. 9
and 10 were used along with Eq. (11) [12] to construct the uniaxial compressive stress–strain curve for concrete in this
study.
 R.A. Hawileh et al. / Applied Mathematical Modelling 34 (2010) 2562–2583 2571

Fig. 10. 3D failure surface for concrete [6].

Table 4
Materials’ properties.

Material # Material name fc0 (ksi)a ft (ksi)a Modulus of elasticity (E in ksi) Stress–strain curve

1 Concrete (beam and column) 7.43 0.5 4500 Fig. 12

2 PT steel – – 28,500 Fig. 15
3 Mild-steel – – 29,000 Fig. 16
4 Interface grout 11 0.6 3460 Fig. 13
5 Duct grout 8.8 0.5 2820 –
6 Beam reinforcing bars – – 29,100 –
7 Column reinforcing bars – – 29,000 –
8 5 mm Wire – – 29,000 –
9 7 mm Wire – – 29,000 –
a
1 ksi = 6.895 MPa.

Ec e
f ¼  2 ; ð9Þ
1 þ ee0

and,

2fc0
e0 ¼ ; ð10Þ
Ec
f
Ec ¼ ; ð11Þ
e
where,

f = stress at any strain e, in ksi (MPa),

e = strain at stress f,
e0 = strain at the ultimate compressive strength fc0 .

Figs. 12 and 13 display the adopted nonlinear compressive stress–strain curves for both the concrete and grout materials
based on the discussed concrete constitutive material model.
Additional concrete material data, such as the shear transfer coefficient bt for open cracks and bc for closed cracks are also
needed for the concrete constitutive material data table. The shear transfer coefficients bt and bc control the amount of shear
transfer across the cracks. Typical shear transfer coefficients range from 0.0 to 1.0, with 0.0 representing a smooth crack
(complete loss of shear transfer) and 1.0 representing a rough crack (no loss of shear transfer). A value of 0.3 and 0.5 was
used in the FE model for bt and bc, respectively.
2572 R.A. Hawileh et al. / Applied Mathematical Modelling 34 (2010) 2562–2583

Fig. 11. Uniaxial compressive and tensile stress–strain curve for concrete [10].

Fig. 12. Idealized stress–strain curve for concrete.

Fig. 13. Idealized stress–strain curve for interface grout.

 R.A. Hawileh et al. / Applied Mathematical Modelling 34 (2010) 2562–2583 2573

Fig. 14 presents the tensile stress–strain curve [6] used for the concrete and grout materials. The tensile relaxation (soft-
ening) was presented by a sudden reduction of the tensile strength to 0.6fr [6] upon reaching the tensile cracking strain ecr.
After this point, the tensile response decreases linearly to zero stress at a strain of 6ecr[6] as shown in Fig. 14.
The adopted stress–strain curves for the PT and A706 mild-steel materials in this FE model are shown in Figs. 15 and 16.
These figures were taken from monotonic tests conducted by NIST [5] as part of their experimental program. In addition the
following constants were used in the developed model:
Coefficient of friction: This factor indicates the roughness of the surfaces in contact. A coefficient of friction of 0.50 is used
at the beam–column interface according to ACI 318-02 code [9] and PRESSS Report No. 01/03/09 [3]. In addition, a very low
coefficient of friction of 0.05 is used between the A706 mild-steel bars and duct hole surfaces as well as between the PT steel
and duct hole surfaces.
Density (d): The densities of steel (mild-steel and PT steel) and concrete were assumed to be 490 lb/ft3 and 150 lb/ft3,
respectively.
Poisson’s ratio (m): The Poisson’s ratio (m) was assumed to be 0.3 for steel and 0.2 for concrete, respectively.

6.1.1. Meshing
The connection region where all the inelastic behavior occurs in this model is finely meshed, elsewhere a coarser mesh is
used. Fig. 4 shows the FE mesh used. The total number of elements used is 60,810 elements which is an adequate refinement
for the constructed 3D FE model.

6.1.2. Boundary conditions

In the experimental setup, the boundary conditions were as follows: pinned at the bottom of the column and roller sup-
ported at the top of the column and beam’s ends as shown in Fig. 17. These boundary conditions are prescribed to represent
the experimental setup shown in Fig. 17. Nodes at the bottom of the column are restrained in all translational degrees of
freedom, and the beam’s edge is restrained in the vertical y and lateral z directions. The lateral z direction is restrained since
beam’s weight and joint’s friction prevent any displacement in this direction.
An anchor Plate (4.5  4.5 in. by 1.75 in. thick) was used in the experiment at the beam’s ends around the PT steel. This
was modeled by preventing the displacements of the nodes in that area in the horizontal x, vertical y, and lateral z directions
as shown in Fig. 17.
During the analysis, it was found that the results are sensitive to the boundary conditions at the top face of the column.
Special refinement of the boundary conditions at the top face of the column was performed and the analysis was repeated
several times until stable and substantial improvement of the FE results was observed where maximum deviations from the
experimental results is always within 15% everywhere. The final boundary conditions applied to the top face of the column
are shown in Fig. 17. The nodes at the two edge lines of the column are coupled in the vertical y direction to ensure all
the nodes associated with this line move together. The nodes located at the middle face of the column along the direction
of the applied displacement are coupled in the horizontal x direction. It should be noted that the boundary conditions were
the same for all loading cycles.

6.1.3. Loads
The loading in the experiments was applied as follows:

Fig. 14. Tensile stress–strain curve for concrete [6].

2574 R.A. Hawileh et al. / Applied Mathematical Modelling 34 (2010) 2562–2583

Fig. 15. Stress–strain curve of PT tendon, Grade 270 [5].

Fig. 16. Stress–strain curve of Grade 60 mild-steel [5].

1. A pre-tensioning force of 44.5 KN (50 kips) was applied to the PT steel to develop an initial stress of 830 MPa (120 ksi).
2. Beam gravity load, approximately 20 KN (4.5 kips) was applied to the beams at 3.5 in. from the column’s face as shown in
Fig. 17. An axial load equal to 1200 KN (270 kips) was applied to the top face of the column. This is represented in the FE
model by applying a pressure of 13 MPa (1.875 ksi) on the upper elements of a total surface area of 144 in.2 as shown in
Fig. 17.
3. The loading history for the specimens is based on story drift and is the one recommended for use in the PRESSS Program
[5]. The basic loading history was three cycles at a particular drift level followed by an elastic cycle as shown in Fig. 18. It
should be noted that in the elastic cycle, the connection was loaded to approximately 30% of the peak load in the previous
cycle. This is represented in the finite element model by applying the displacement in cycles of loading and unloading
(load steps) at the top edge of the column.

6.2. Nonlinear solution and convergence criteria

The applied cyclic displacements were divided into a series of increments called load steps and load substeps. At the end
of each incremental solution, the program [6] adjusts the stiffness matrix of the model before proceeding to the next incre-
mental load step in order to reflect the nonlinear changes in the model’s structural stiffness. To predict and control load step
size increments, the automatic time stepping option [6] is turned on in this study. The model stiffness is updated in ANSYS
[6] by Newton–Raphson equilibrium iterations which default to 30 equilibrium equations.
Due to the highly nonlinear behavior of the model during cyclic loading, the convergence behavior of the model depends
on the nonlinear response of the reinforced concrete due to crack generation and propagation, and loss of stiffness in certain
 R.A. Hawileh et al. / Applied Mathematical Modelling 34 (2010) 2562–2583 2575

Fig. 17. Experimental setup and boundary conditions.

elements after energy dissipation. This might cause divergence and failure that will lead to the termination of the analysis. In
this study, failure of the FE model is defined when the solution for a 0.001 in. (0.0254 mm) displacement increment doesn’t
converge. Accordingly, ANSYS [6] will give a message specifying divergence of the analysis due to large deflection exceeding
the displacement limitation of the program.
The convergence criteria in this study for the reinforced concrete elements were based on force, displacement, and the
tolerance limits of the convergence criteria [6]. ANSYS [6] convergence tolerance default values of 0.5% for force checking
and 5% for displacement checking were initially selected. It was found that convergence was difficult to achieve using the
default values due to the highly nonlinear behavior of the reinforced concrete elements and the associated large deflections.
Thus, in order to obtain convergence of the equilibrium iterations, the convergence tolerance limits were increased to 2.5%
for the force checking criterion and up to 10% for the displacement checking criterion.
2576 R.A. Hawileh et al. / Applied Mathematical Modelling 34 (2010) 2562–2583

Fig. 18. Lateral loading history for specimen M-P-Z4 [5]: (a) displacement, (b) story drift.

7. Results and discussions

7.1. Comparison of numerical and experimental results

The finite element analysis and experimental results are compared to examine the validity and predictability of the devel-
oped FE model. The test was conducted under cyclic lateral displacement loading. The lateral reaction response for the entire
load history of the tested hybrid precast joint and the predicted response from the FE are presented and compared in Fig. 19.
Instead of comparing results of every load cycle, average response envelopes were constructed by connecting peak positive
and negative reaction loads versus the applied lateral story drift as shown in Figs. 20 and 21. In addition, the experimental
lateral load-displacement (hysteresis) loops for the precast joint is compared with those obtained from the finite analysis for
the entire loading history as shown in Fig. 22.
It should be noted that the story drift is equal to the applied lateral displacement at the top of the column face divided by
the column’s height of hc = 1320 mm (hc = 52 in.). The vertical axis in Fig. 19 represents the lateral reaction load obtained
from the load cell at the top face of the column. The horizontal axis in Figs. 20 and 21 shows the applied lateral story drift
and the vertical axis in these graphs represents the measured and predicted positive/negative peak reaction loads. For the
load-displacement hysteresis loops (Fig. 22), the vertical axis is the lateral reaction load and the horizontal axis is the applied
lateral displacement at the top face of the column. The area under the load-displacement curve is equal to the energy dis-
sipation by the precast hybrid joint connection.
Results show that the response envelope from the finite element analysis correlated fairly well with experimental results.
Good correlation exists in all stages of lateral cycling loading. It is clear from Figs. 20 and 21 that the FE results for the peak
loads is less than those of the experimental results (within 15%) in the early stages of the cyclic loading. The FE and exper-
imental results are almost the same in the middle stages of cyclic loading, and the FE results are greater than those of the
experimental results (within 13%) in the last two stages of loading cycles. In other words, the FE model behaved softer and
dissipated more energy in the early stages of loading cycles, behaved almost the same as the experimental results at about
1% drift, and behaved stiffer and dissipated less energy than the experimental tested precast joint when the applied story
 R.A. Hawileh et al. / Applied Mathematical Modelling 34 (2010) 2562–2583 2577

Fig. 19. Measured and predicted lateral reaction load for model M-P-Z4.

Fig. 20. Average envelope response curve for positive peak loads versus applied lateral peak story drifts.

Fig. 21. Average envelope response curve for negative peak loads versus applied lateral peak story drifts.
2578 R.A. Hawileh et al. / Applied Mathematical Modelling 34 (2010) 2562–2583

Fig. 22. Hysteresis curve for model M-P-Z4.

drift was greater than about 1%. The difference between the FE and experimental results at different stages of loading can be
attributed to mesh refinement, idealized boundary conditions in the FE model, material nonlinearity, and the specified coef-
ficient of friction between contact surfaces at the beam–column interface.
According to the hysteresis experimental results shown in Fig. 22, a significant drop in the recorded load was observed
when the applied displacement was reversed at the end of each cycle. This behavior occurred due to the fact that the test was
stopped for about 30 min at the end of each cycle to mark cracks, record observations, and take photos. The drop-off in the
load during this time could be a result of relaxation of the hydraulic jack and/or movements in the pins/joints in the system.
Thus, in order to compare the experimental load-displacement results with those obtained from the finite element analysis,
the drop off in the load was subtracted when the applied displacement was reversed as shown in Fig. 23a–c for different
applied story drift cycles. It is clear from Fig. 23 that the energy dissipation (area within hysteresis loops) obtained from
the FE model correlates fairly well with the experimental results if the drop-off load is taken out from the experimental re-
sults at the end of each cycle.
Sources of errors could also be related to the experimental setup. In general, the comparison between the FE and exper-
imental results is better than expected.
Additional results generated by the finite element are discussed in the following sections. These include the predicted
stress and total strain in the mild-steel bars at the beam–column interface as well as stress levels in the PT steel tendon
for the entire loading history.

7.2. Behavior of the connection

Having a reliable FE model is crucial since FEA gives full stress and strain fields throughout the connection for the exper-
imental duration. This is advantageous compared to experimental results, which are limited to discrete strain gauges, LVDT,
and load cells measurements at few points. The FE model can provide a variety of results at any location within the model.
Prediction of the failure modes and the structure’s behavior of the experiment are also possible using FEA results.
Fig. 24 shows the deflected shape of the model and the interface gap opening at an applied story drift of 2.4%. Fig. 25 dis-
plays concrete cracking pattern, interface grout cracks, and interface crack closure.

7.3. Fracture and low-cycle fatigue of mild-steel bars

Fracture of the mild-steel bars resulted in the failure of test specimen due to the high levels of plastic strains developed in
the mild-steel bars. Once the gap at the beam–column interface opens, relatively high levels of repetitive plastic strains de-
velop in the mild-steel bars. The bar fractures are due to low-cycle fatigue (LCF). Low-cycle fatigue is associated with high
strain ranges and low numbers of cycles to failure. In low-cycle fatigue, significant plastic deformations occur during each
cycle. The relatively high inelastic strains in the FE model show the potential vulnerability of mild-steel bars to low-cycle
fatigue failure.
FEA is clearly showing the development of plastic strains in the mild-steel bars at the beam–column interface due to gap
openings. At this level of plastic strains, the mild-steel bars are expected to fracture due to low-cycle fatigue. Hawileh et al.
[13] experimentally evaluated the low-cycle fatigue behavior of ASTM A706 and A615 Grade 60 deformed reinforcing steel
bars that are used in precast hybrid frame connections. Based on the experimental results, Hawileh et al. [13] proposed low-
 R.A. Hawileh et al. / Applied Mathematical Modelling 34 (2010) 2562–2583 2579

Fig. 23. Sample hysteresis curves at different at applied story drift cycles.

Fig. 24. Deflected shape and interface gap opening (exaggerated scale).

cycle fatigue relationships for both types of reinforcement steel bars in which maximum permissible design strains can be
calculated based on the number of cycles to failure.
It can also be noted that the un-bonded portion of the mild-steel bar at the beam–column interface is under combined
loading, mainly axial and bending stresses. High stress concentration occurs at the corners of the un-bonded segment of
2580 R.A. Hawileh et al. / Applied Mathematical Modelling 34 (2010) 2562–2583

Fig. 25. Concrete cracks, interface grout cracks, and interface crack closure distribution of model.

the mild-steel bar. Figs. 26 and 27 show the stress and total strain distribution for the bar’s end and mid-points of the un-
bonded segment of the mild-steel bar. Fig. 28 shows the stress vs. total strain in the mild-steel bar at both the end and mid-
points of the un-bonded segment of the mild-steel bar. Figs. 26 through 28 indicate that the values of the plastic strains in-
crease due to bending.

7.4. Behavior of post-tensioning (pt) tendon

Fig. 29 shows the predicted stresses in the PT steel tendon for the entire loading history. The yield stress in the PT steel
(shown in Fig. 15) was approximately equal to 0.9fpu (243 ksi). Fig. 29 clearly indicates that the PT steel tendon did not yield
 R.A. Hawileh et al. / Applied Mathematical Modelling 34 (2010) 2562–2583 2581

Fig. 26. Bar end point (solid) and bar mid-point (dashed) stress distribution.

Fig. 27. Bar end point (solid) and bar mid-point (dashed) total strain distribution.

Fig. 28. Corner (solid) and bar mid-node (dashed) stress–strain loops.

(an important design objective) during the entire cyclic loading history and thus it remained in the elastic range. Therefore
the elastic clamping force provided from the PT steel tendons to resist gravity loads was maintained throughout the entire
test.
2582 R.A. Hawileh et al. / Applied Mathematical Modelling 34 (2010) 2562–2583

Fig. 29. Predicted PT axial stress for the entire loading history.

7.5. Summary and conclusions

A detailed nonlinear FE model was developed to accurately predict the behavior and performance of a hybrid frame
beam–column connection subjected to cyclic loading. The model has been validated and verified against data from an exper-
imental test conducted by NIST. A highly detailed model was developed to represent all the experimental components. Spe-
cial attention was given to material properties, mesh refinement, contact surfaces, pre-tensioning, and boundary conditions
to simulate the experimental setup. The following observations and conclusions can be made from this investigation based
on the finite element and experimental results. It should be noted that the NIST test utilized here is only comprehensive test-
ing program for the hybrid connection that is found in the literature. Although the number of experiments evaluated is
therefore limited, it is believed that the available test data provide a reasonable basis for the conclusions presented below
due to the comprehensive nature of the test.

 The FE model proved to be effective in producing results that were in good agreement with the experimental results in the
elastic and plastic ranges.
 The response envelope from the finite element analysis correlated fairly well with experimental results. Good correlation
exists in all stages of cycling loading. Specifically, the FE results for the peak loads are smaller than the experimental
results (within 15%) in the early stages of cyclic loading, about the same in the middle stages cyclic loading, and larger
than the experimental results (within 13%) in the last two stages of cyclic loading.
 The energy dissipation levels obtained from the FE model correlated fairly well with the experimental results when the
experimental drop-off load is removed from the experimental results at the end of each cycle
 Fracture of the mild-steel bars resulted in the failure of the test specimen. Similarly, the FE model showed the develop-
ment of relatively high levels of inelastic plastic strains in the mild-steel bars that could cause low-cycle fatigue.
 The stresses in the PT steel tendon remained below yield throughout the entire test. Thus the clamping force provided
from the PT steel tendons to resist gravity loads was maintained.
 The development of the FE model allows for additional investigation of such a connection under other loading conditions
and different design parameters.

For future research, it is recommended that additional FE models of hybrid connections be developed if and when addi-
tional comprehensive tests become available. This would provide further validation of the current developed model, and
would generate more information about the cyclic performance of precast hybrid joints. In addition, the following issues
should be experimentally and analytically investigated to better understand the behavior of precast hybrid frame
connections:

 Impact of different material properties, and boundary conditions.

 Effect of changing the location of the PT steel tendons within the cross-section.
 Effect of the duct grout on the stress in the mild-steel bars.

Conversion factors:

1 mm = 0.039 in.
1 mm2 = 0.00152 in2
1 kN = 0.2248 kips
 R.A. Hawileh et al. / Applied Mathematical Modelling 34 (2010) 2562–2583 2583

1 MPa = 145 psi

1 in. = 25.4 mm
1 sq in. = 0.000645 m2
1 ksi = 6.895 MPa
1 ft = 0.3048 m
1 kip = 4.45 KN

Acknowledgement

The authors of this paper would to thank Ms. Gerardine Cheok from the National Institute of Standard and Technology
(NIST) in providing the experimental data and results for the analyzed model. In addition, we would like to acknowledge
the support of the National Center for Supercomputing Applications (NCSA) at the University of Illinois, Champaign for pro-
viding support in using the supercomputer facility.

References

[1] M.J.N. Priestley, Overview of PRESSS research program, PCI J. 36 (4) (1991) 50–57.
[2] M.J.N. Priestley, J.R. Tao, Seismic response of precast prestressed concrete frames with partially debonded tendons, PCI J. 38 (1) (1993) 58–69.
[3] J.F. Stanton, S. Nakaki, Design Guidelines for Precast Concrete Seismic Structural Systems, PRESSS Report No. 01/03-09, Seattle, WA, 2002.
[4] G.S. Cheok, W.C. Stone, S.D. Nakaki, Simplified Design Procedure for Hybrid Precast Connections, Report No. NISTIR 5765, NIST, Gaithersburg, MD, 1996.
[5] G.S. Cheok and W.C. Stone, Performance of 1/3 Scale Model Precast Concrete Beam–Column Connections Subjected to Cyclic Inelastic Loads, Report No.
NISTIR 5436, NIST, Gaithersburg, MD, 1994.
[6] ANSYS Version 8.0, User’s and Theory Reference Manual, 2004.
[7] R. Hawileh, H. Tabatabai, A. Rahman, A. Amro, Non-dimensional design procedures for precast, prestressed concrete hybrid frames, PCI J. 51 (5) (2006)
110–130.
[8] K.J. Willam, E.D. Warnke, Constitutive model for the triaxial behavior of concrete, in: Proceedings, International Association for Bridge and Structural
Engineering, vol. 19, ISMES, Bergamo, Italy, 1975, p. 174.
[9] American Concrete Institute (ACI), ACI 318-02 Building Code Requirements for Structural Concrete and Commentary, Farmington Hills, MI, 2002.
[10] M.Y.H. Bangash, Concrete and Concrete Structures: Numerical Modeling and Applications, Elsevier Science Publishers Ltd., London, England, 1989.
[11] P. Desayi, S. Krishnan, Equation for the stress–strain curve of concrete, J. Am. Concrete Inst. 61 (1964) 345–350.
[12] J.M. Gere, S.P. Timoshenko, Mechanics of Materials, PWS Publishing Company, Boston, Massachusetts, 1997.
[13] R. Hawileh, H. Tabatabai, A. Rahman, Evaluation of the low-cycle fatigue life in ASTM A706 and A615 Grade 60 steel reinforcing bars, J. Mater. Civil Eng.
ASCE, in press.