Engineering Structures 105 (2015) 99–122

Contents lists available at ScienceDirect

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

Nonlinear finite element modeling of reinforced concrete haunched

beams designed to develop a shear failure
Eber Alberto Godínez-Domínguez a,⇑, Arturo Tena-Colunga b,1, Gelacio Juárez-Luna b,1
a
Facultad de Ingeniería, Universidad Autónoma de Chiapas, Campus-I, Blvd. Belisario Domínguez, kilómetro 1081, S/N, Col. Terán, 29050, Tuxtla Gutiérrez, Chiapas, Mexico
b
Departamento de Materiales, Universidad Autónoma Metropolitana, Edificio P4, Último Piso, Av. San Pablo 180, Col. Reynosa Tamaulipas, 02200 México, D.F., Mexico

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

Article history: The results of different nonlinear finite element models for eight simply supported reinforced concrete
Received 25 December 2014 haunched beams designed to develop a shear failure under static loading are presented and discussed
Revised 11 August 2015 in this paper. Simplified nonlinear models in which the participation of the longitudinal steel
Accepted 17 September 2015
reinforcement and stirrups is indirectly included were assessed using SAP2000. More complex nonlinear
finite element models were assessed with ANSYS, in which longitudinal steel reinforcement and stirrups
were modeled as built. Softening of concrete due to deformation was taken into account in the selected
Keywords:
constitutive models using a failure surface with different peak compressive and tension stresses. Strain
Nonlinear finite elements
RC haunched beams
hardening for the steel reinforcement was considered using the Von Mises yield criterion. Perfect bond
Shear failure between concrete and steel was assumed. Shear–displacement curves for a specific section located at
midspan of the beams were obtained from the finite element models and compared to those obtained
from experimental testing. Also, crack patterns associated to different loads steps were obtained from
ANSYS finite element models. It can be concluded that it is possible to obtain a reasonable correlation
between analytical and experimental load–deformation curves and the main developed arch mechanism
for RCHBs failing in shear using both simplified and detailed finite element models, which for practical
purposes is more than acceptable. However, only a medium correlation between cracking patterns
numerically obtained with detailed finite element models and those experimentally identified were
observed, particularly for beams with shear reinforcement.
Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction
Reinforced concrete haunched beams (RCHBs) have been
traditionally used around the world for the design of bridges and
buildings (Fig. 1).
In order to insure the ductile behavior of RCHBs from a concep-
tual and capacity design viewpoint, it is necessary to understand
how RCHBs resist shear forces under static and cyclic loadings in
order to prevent potential shear failures, and then understand
and insure ductile flexural behavior under static and cyclic
loadings. For this reason, and as a first step, an experimental study
on simply supported RCHBs designed to fail in shear was carried
out using the geometries and practices observed in Mexico [1,2].
The geometry of prototype RCHB was defined according to a survey
conducted in buildings of recent construction in Mexico City [1,2].
⇑ Corresponding author. Tel.: +52 (961) 61 50322; fax: +52 (961) 61 50527.
E-mail addresses: eber.godinez@unach.mx (E.A. Godínez-Domínguez), atc@
correo.azc.uam.mx (A. Tena-Colunga), gjl@correo.azc.uam.mx (G. Juárez-Luna).
1
Tel.: +52 (55) 5318 9460; fax: +52 (55) 5318 9085.
In such buildings, the most common dimensions were: (a) the
haunched length was one-third the effective span of the beam
(L), that is, L/3 and, (b) haunched angles varied from 6° to 12°.
Therefore, in this paper the first eight simply supported RCHBs
reported to fail in shear under monotonic loading [1] were
modeled using nonlinear finite elements. Four of the eight studied
specimens do not have shear reinforcement whereas the remaining
ones, identical in geometry, had minimum shear reinforcement.
The considered angles of slope of the haunch from horizontal
(or haunched angle, a) were 3.07°, 6.12°, 9.13° and 12.10°. The
haunched length at both beam ends was one-third the effective
span for the beam (L/3).
This study is focused on the numerical modeling of simply
supported RCHBs designed to develop a shear failure under static
loading. One of the principal purposes was to assess, based on a
comparison of numerical and experimental results, the ability
and limitations of simple and complex nonlinear modeling to pre-
dict the experimental behavior of the tested reinforced concrete
haunched beams (RCHBs). For this purpose, simple models in
which the failure is modeled using shear plastic hinges were

http://dx.doi.org/10.1016/j.engstruct.2015.09.023
0141-0296/Ó 2015 Elsevier Ltd. All rights reserved.
100 E.A. Godínez-Domínguez et al. / Engineering Structures 105 (2015) 99–122
(a) Bridges
(b) Buildings
Fig. 1. Structures with RCHBs used within the Metropolitan Area of Mexico City.
tested, as they may be appealing to design practice. More complex
nonlinear finite element models were also assessed, where the
concrete failure is distributed over the volume of the finite ele-
ments and steel plasticity is distributed in one-dimensional
elements.
2. Description of the experimental study
2.1. Geometry, loads and boundary conditions
The geometry, loads and boundary conditions used in the
experimental research program [2,3] are shown in Fig. 2, and were
also used for the finite element models. All RCHBs elements have a
total length L = 330 cm. The effective span for all RCHBs was
L = 2.80 m and the width was b = 22 cm. The considered angles of
slope of the haunch from horizontal were: 3.07°, 6.12°, 9.13° and
12.10°. The haunched length at both beam ends was one-third
the effective span of the beam (Lh = L/3
93.3 cm). The bearing
length at both beam ends was 25 cm. The linear tapering was
obtained by keeping a constant depth hmax = 45 cm at the beam
ends while varying the depth of the beam at the central third from
45 cm (prismatic) to 25 cm, that is, hmin = 45, 40, 35, 30 and 25 cm.
Beams were simply supported and tested under monotonic
loads (V) that were applied 10 cm (3.937 in.) from the vertex
formed by the intersection of tapered sections with the prismatic
section, as depicted in Fig. 2.
The cryptogram used for the identification for the RCHBs corre-
sponds to the originally proposed by Archundia [3], TASCai-Rj,
where i is an index that indicates the considered haunched angle:
i = 0 = 0°, i = 1 = 3.07°, i = 2 = 6.12°, i = 3 = 9.13° and i = 4 = 12.10°; j
is an index that identifies the shear reinforcement: j = 0 indicates
the absence of shear reinforcement whereas j = 1 indicates the
use of minimum shear reinforcement as requested in NTCC-04
[4] guidelines.
Complete details on how the specimens were designed to insure
that they failed in shear while following general NTCC-04 guideli-
nes, as well as construction details process of all beams used to
develop the numerical models in this study can be found else-
where [1,3].
2.2. Flexural and shear steel reinforcement
Flexural and shear reinforcement details for each beam consid-
ered in this study are shown in Fig. 3. Also, the corresponding typ-
ical cross sections are shown in Fig. 4. As commented, the studied
specimens were classified into two groups: (1) R0 elements, in
which no shear reinforcement is used along the haunched length
(Fig. 3a–d) and, (2) R1 elements, where minimum shear reinforce-
ment, equal to the one required by NTCC-04 [4] for prismatic
beams is provided along the haunched length (Fig. 3e–h).
3. Simplified modeling using SAP2000
As a first step, nonlinear static analyses were performed using
SAP2000 [5]. To do this, simple models using tapered beam-
elements were developed. A lumped plasticity modeling was used
through shear plastic hinges. Average mechanical properties for
the reinforced concrete were used for the tapered elements.
3.1. Modeling considerations
Force–displacement relationships were defined and assigned
for each plastic shear hinges, which were obtained from
experimental results. These curves were defined based upon the
experimental data recorded. The purpose of this modeling was
to explore the usefulness of using a simple analytical approach to
represent the behavior of RCHBs failing in shear by taking into
account the experimental information that was already available.
The procedures to define the shear plastic hinges, as well as the
obtaining of the pushover curve, are illustrated in Fig. 5.
The location of shear plastic hinges (Lcrit), measured from the
supports, it is associated to the effective equivalent depth at the
critical haunched section (dcrit, Figs. 6 and 7) and computed using
Eq. (1), in which dcrit represents the depth that must be used to
compute the shear resistance in RCHBs according to Tena-
Colunga et al. [1], which can be obtained using Eq. (2):
dmax  dcrit
Lcrit ¼
 ð1Þ
hmax hmin
lc
" 2
! #
hmax hmin  hmax
dcrit ¼ dmin ð1 þ 1:35 tan aÞ 6 þ hmax r ð2Þ
2lc
where hmax and hmin have been defined previously, and dmax =
(hmax  r), dmin = (hmin  r), lc is the haunched length and r is the
concrete cover for the longitudinal reinforcement.
The location of plastic shear hinges used in the simplified
RCHBs modeling with SAP2000 is schematically shown in
Figs. 5a and 7. This modeling was used based upon the observed
damage during experimental tests [1,3]. All variables used to
define the location of plastic shear hinges are shown in Table 1.
3.2. Results of nonlinear static analysis
In order to assess the usefulness of simplified analytical
models to reproduce the experimental behavior of RCHBs, applied
shear force versus displacement curves at midspan obtained
 E.A. Godínez-Domínguez et al. / Engineering Structures 105 (2015) 99–122 101

V V
a = 108.3
10

(45,40,35,30 and 25)

hmax = 45

hmin = variable
5
= 0°, 3.06°, 6.12°, 9.13° and 12.10°
[cm]

25 93.3 93.3 93.3 25

25 280 25

330

Fig. 2. Geometry, loads and boundary conditions. Dimensions in cm (adapted from Tena-Colunga et al. [1]).

2 E # 2.5 @ 110 8 E # 2.5 @ 18.5

3#8 3#8

4#8 4#8

Section 1 Section 2 Section 3 (cm.) Section 1 Section 2 Section 3 (cm.)

25 93.3 93.3 93.3 25 25 93.3 93.3 93.3 25
330 330
(a) TASC 1-R0 (e) TASC 1-R1

3 E # 2.5 @ 14
2 E # 2.5 @ 110 8 E # 2.5 @ 18.5

3#8 3#8

4#8 4#8

Section 1 Section 2 Section 3 (cm.) Section 1 Section 2 Section 3 (cm.)

25 93.3 93.3 93.3 25 25 93.3 93.3 93.3 25
330 330
(b) TASC 2-R0 (f) TASC 2-R1

3 E # 2.5 @ 7.5
2 E # 2.5 @ 110 8 E # 2.5 @ 18.5

3#8 3#8
4#8 4#8

Section 1 Section 2 Section 3 (cm.) Section 1 Section 2 Section 3 (cm.)

25 93.3 93.3 93.3 25 25 93.3 93.3 93.3 25
330 330

(c) TASC 3-R0 (g) TASC 3-R1

3 E # 2.5 @ 7.5
2 E # 2.5 @ 110 8 E # 2.5 @ 18.5

3#8 3#8
4#8 4#8

Section 1 Section 2 Section 3 (cm.) Section 1 Section 2 Section 3 (cm.)

25 93.3 93.3 93.3 25 25 93.3 93.3 93.3 25
330 330

(d) TASC 4-R0 (h) TASC 4-R1

Fig. 3. Steel reinforcement for specimens tested by Tena-Colunga et al. [1].

using pushover analyses are compared with those obtained directly from cylinder testing subjected to axial compression
experimentally (‘‘Approach 1”, Figs. 8 and 9). (Table 2). Subsequently, a second analysis in which only a fraction
It is worth noting that for each studied model, a first analysis of the experimental modulus of elasticity was performed. This
was done using the modulus of elasticity for the concrete obtained fraction of the modulus of elasticity was obtained by matching
102 E.A. Godínez-Domínguez et al. / Engineering Structures 105 (2015) 99–122
4
3#8
variable
45
45 40
E # 2.5
35
30
25
4#8
4 4
(cm.)
22
SECTION 1
Fig. 4. Typical cross sections for specimens tested by Tena-Colunga et al. [1].
the slopes of the initial analysis with the corresponding experi-
mentally ones. The average factor used to match initial experimen-
tal and analytical slopes was 0.40.
Experimental and analytical V–D curves for beams without (R0)
and with (R1) shear reinforcement are compared in Fig. 8a and b,
where V is the applied shear force and D is the measured or com-
puted deflection at midspan. In each figure, three points are
depicted using blue circles, green rectangles, and red2 diamonds,
corresponding respectively to the first diagonal cracking, failure
and collapse, according to the reported experimental values [1]. As
it can be observed from the graphs, the point associated to collapse
often cannot be accurately defined in the experimental curve.
It can be observed from Fig. 8 that analytical and experimental
curves tend to converge as the haunched angle increases. Taking
into account that the experimental shear forces were used as data
for the shear springs, it is worth noting that analytical displace-
ments at failure state are reasonably estimated, although in most
cases is overestimated, primarily for small haunched angles.
Alternatively, in order to obtain the required force–displace-
ment relationships to define the behavior of nonlinear shear
springs, the analytical results obtained by Archundia [3] were used
(‘‘Approach 2”, Fig. 9), which in turn were based on the analytical
stiffness formulation for nonprismatic beam elements proposed
by Tena-Colunga [6]. Shear deformations and variations in the
properties of the cross section along a continuous centroidal axis
are taken into account in this formulation. The analytical displace-
ments obtained for a section located at the center of the haunched
zone and their comparisons with the experimental results are
shown in Table 3. It is observed that for all cases, the analytical val-
ues are smaller than the experimental ones. In this case, modeling
of RCHBs was performed considering that it is not necessary to pre-
viously have experimental results.
V–D curves for beams without shear reinforcement are shown
in Fig. 9. Two different considerations were made for the definition
of the nonlinear shear springs properties: (a) experimental data are
used (curves identified as ‘‘Measured”), (b) analytical stiffness
formulation for nonprismatic beam elements proposed by
Tena-Colunga [6] are used (curves identified as ‘‘Analytical”). Also,
the experimental results (curves identified as ‘‘Experimental”) are
shown in Fig. 9. It can be observed from Fig. 9 that the two criteria
used to define the nonlinear shear springs properties lead to
reasonable correlations between analytical and experimental
V–D curves.
2 the observed failure mode of the experimental T-beam, (d) the
For interpretation of color in Fig. 8, the reader is referred to the web version of
this article.
4
3#8
variable
40
E # 2.5 40
35
30
25
4#8
(cm.)
22
SECTION 2
4. Nonlinear modeling techniques used for brittle RC members
Nonlinear finite element analyses have been widely used to
study the behavior of reinforced concrete members and structures,
including those deficient in shear.
Cervenka [7] presented the smeared crack model implementa-
tion for the case of shear failure analysis. The behavior of cracks
in concrete was described with an exponential cohesive fictitious
crack-opening law and with a refined definition of the crack band,
which takes into account the finite element size and orientation
with respect to crack direction. The effectiveness of the model
was demonstrated on the modeling of a reinforced concrete beam
under shear failure. The numerical analyses were made with pro-
gram SBETA. The presented model guarantees objective results in
features such as: load–displacement diagram, ultimate load, crack
pattern, crack width and failure mode. It is important to point out
that the definition of the crack band allows knowing the crack
widths into each damaged finite element.
Hawileh et al. [8], developed a detailed three-dimensional (3D)
nonlinear finite element model to study the response and predict
the behavior of precast hybrid beam–column connection subjected
to cyclic loads that was tested at the National Institute of Standards
and Technology (NIST) laboratory. The precast joint was modeled
using 3D solid elements and 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 the post-tensioning strand and the nonlinear material
behavior for the concrete. The model response was compared with
experimental test results and yielded good agreement at all stages
of loading.
Also, Hawileh et al. [9] developed a finite element (FE) model for
predicting the nonlinear response and behavior of a reinforced con-
crete T-beam deficient in shear under cyclic loading. Cracking loads,
failure loads, response hysteresis envelopes and crack patterns
were used as benchmark for comparison between experimental
and FE results. The authors carried out a parametric study to predict
the optimum combination of the open and close crack shear
transfer coefficients (bt and bc) of the constitutive material model
for concrete. They concluded that: (a) when both shear transfer
coefficients are equal to 0.2, the FE results gave the best correlation
with the experimental results, (b) in both the experiment and finite
element analysis, the shear failure modes and mechanisms of the
T-beam were similar to each other, (c) the crack patterns at the
final load from finite element model corresponded well with
value of shear transfer coefficients, bt and bc, have variable effect
 E.A. Godínez-Domínguez et al. / Engineering Structures 105 (2015) 99–122 103

(a) Rigid-plastic used to model the shear plastic hinges in SAP2000

Fig. 5. Simplified modeling in SAP2000.
104 E.A. Godínez-Domínguez et al. / Engineering Structures 105 (2015) 99–122

(b) Pushover curve obtained in SAP2000

Fig. 5 (continued)

Maximum critical depth
hmax hmin
2 for the reinforcement. Perfect bonding between steel and concrete
hmax
d=h-r
lC
Fig. 6. Definition of variables used for the calculation of critical depth (adapted
from Archundia [3]).
on the free end displacement of the T-beam and on shear strain,
(e) when both shear transfer coefficients are equal to 0.2, the best
correlations with respect to experimental results are obtained
with finite elements for both beams irrespective of their geometry
(T-beam, R-beam) or loading cycles and, (f) variation of concrete
compressive strength has little effect on the deflection. Both studies
developed by Hawileh et al. [8,9] were performed using ANSYS
program.
As commented by Belletti et al. [10], if nonlinear finite element
analyses are used, appropriate checks must be done to verify the
reliability of the results obtained. It is well known indeed that
the results of nonlinear finite element analyses strongly depend
on the analysts modeling choices, e.g. type and dimension of the
elements, mechanical model, constitutive model etc. Belletti et al.
[10] reported a study in which numerical simulations of different
reinforced and prestressed beams taken from experimental
programs were performed in order to assess and criticize the finite
element approaches. Three different failure modes were
considered (bending failure, diagonal-tension failure and shear-
compression failure). Eight-node quadrilateral isoparametric plane
stress elements based on quadratic interpolation and 3  3 Gauss
integration were used for concrete. Embedded elements were used
was assumed in these simulations. The authors showed that the
behavior of reinforced and pre-stressed beams can be in detail
investigated through nonlinear finite element models if the choices
made during the modeling are appropriated and well calibrated.
According to the authors, NLFE-analyses are sensitive to many
aspects, such as: (a) the way in which the reduction of the com-
pressive strength due to lateral cracking is considered, (b) the
reduction of the peak strain, (c) the Poisson’s ratio, (d) the crack
model adopted (fixed crack model and rotating crack model), (e)
the convergence criteria, and (f) the constitutive model for the con-
crete. Some useful guidelines for the NLFEA of reinforced and pre-
stressed beams are described by the authors [10].
As it is well known, the solution of a specific problem can be
obtained using different techniques. In this case, the numerical
study of brittle structures can be carried out not only using nonlin-
ear finite element analysis (NLFEA), but using other numerical
techniques. As an alternative method, sequentially linear analysis
(SLA) could be used. In SLA, the incremental-iterative solution pro-
cedure is replaced by a series of linear analyses as a sequence of
‘‘events”, circumventing the difficulties related to controlling the
solution process [11]. In a recent study, Slobbe et al. [11] used
the SLA solution procedure to study a shear critical prismatic beam
(without shear reinforcement) that typically has a brittle failure.
The results obtained by the authors showed a general good
agreement with the load–deflection responses and crack patterns
of the experimental test and corresponding NLFE-analyses that
used an incremental–iterative solution procedure. Also, they
 E.A. Godínez-Domínguez et al. / Engineering Structures 105 (2015) 99–122
V V
a = 108.30
10
Critical depth
Lcrit
Shear plastic hinges
25 93.3
Fig. 7. Schematically location of plastic shear hinges.
Table 1
Location of plastic shear hinges.
Models a dmin, cm (in.)
TASCa1-R0 3.07° 36 (14.17)
TASCa1-R1
TASCa2-R0 6.12° 31 (12.20)
TASCa2-R1
TASCa3-R0 9.13° 26 (10.24)
TASCa3-R1
TASCa4-R0 12.10° 21 (8.27)
TASCa4-R1
showed that SLA could be able to trigger the non-symmetric failure
crack pattern automatically and it is possible to investigate the
post-peak behavior. Finally, they showed that the results of
the SL-analyses are sensitive to many parameters, such as: (a) the
adopted amount of material fracture energy, (b) the adopted
number of damage increments in the discretized stress–strain
softening relation, (c) mesh-size, and (d) mesh-alignment.
Finally, very recently Albegmprli et al. [12] reported a study in
which stochastic and reliability analyses of the ultimate shear
capacity of RC haunched and prismatic beams were performed,
based on nonlinear finite element analyses. The authors investi-
gated the influence of uncertainty in material properties and
geometry parameters on the mechanical performance and shear
capacity of RC haunched beams. Firstly, 65 experimentally tested
RC haunched beams and prismatic beams were analyzed via deter-
ministic nonlinear finite element method using 2D ATENA program
to verify the efficiency of utilized numerical models, the shear
capacity and the crack pattern. Good results were obtained by
comparing the results of nonlinear finite element and experiments.
All the experimental results were obtained from the literature
review, and include the RC haunched beams studied by Tena-
Colunga et al. [1]. Afterward, in order to obtain statistical parame-
ters, stochastic analyses were performed for each beam where the
RC material properties and geometry parameters are assigned to
take probabilistic values using Latin hypercube sampling simulat-
ing technique. The material parameters that were considered to
investigate the behavior of the RC haunched beams were the elas-
tic modulus, the compressive strength, the tensile strength and the
fracture energy for concrete. Also, the elastic modulus and the
yielding strength of steel were considered as uncertain parameters.
On the other hand, the geometry factors considered were the beam
width, effective depth, longitudinal reinforcement area and shear
reinforcement area. The statistical parameters were obtained for
resistance bias factor and the coefficient of variation. Finally, relia-
bility analyses were accomplished using the limit state functions of
105
Critical depth
(cm.)
93.3 93.3 25
330
dmax, cm (in.) dcrit, cm (in.) Lcrit, cm (in.)
41 (16.14) 38.64 (15.21) 43.98 (17.31)
41 (16.14) 35.55 (14.00) 50.83 (20.01)
41 (16.14) 31.73 (12.49) 57.68 (22.71)
41 (16.14) 27.17 (10.70) 64.53 (25.41)
ACI-318 and ASCE-7 depending on the calculated statistical param-
eters. The authors showed that RC haunched beams have higher
sensitivity and riskiness than RC prismatic beams.
5. ANSYS modeling
Finite element analyses using ANSYS [13] have been widely
used to study the behavior of reinforced concrete members and
structures for both original design and retrofit purposes. Among
the main components studied are: beams, columns, piers, walls,
slabs, beam–column connections and slab–column connections.
In general, plausible analytical results have been obtained for such
elements when compared to experimental results, mostly when
compared to global load–deformation curves, crack patterns and
stress distributions [8,9,14–22]. For the specific topic studied in
this paper, in which shear behavior of concrete elements is mod-
eled in ANSYS, there exist previous related research studies [8,9].
In this study, the material parameters that are taken to investi-
gate the behavior of the RC haunched beams are the elastic modu-
lus Ec, the compressive strength fc, the tensile strength ft and the
fracture energy Gf for concrete. Also, the elastic modulus Es and
the yielding strength fy of steel are considered as uncertain param-
eters. On the other hand, the geometry factors considered are the
beam width, effective depth, longitudinal reinforcement area and
shear reinforcement area.
5.1. Techniques used for finite element modeling of reinforced concrete
members
As commented by Tavárez [23], finite element models for rein-
forced concrete structures have generally been based on replacing
the composite continuum by an assembly of elements representing
the concrete and the steel reinforcement. Currently, three
alternative techniques are mainly used for modeling reinforcement
106 E.A. Godínez-Domínguez et al. / Engineering Structures 105 (2015) 99–122

10 10

TASCα1-R0 TASCα2-R0
8 8

6 6

V (Ton)

V (Ton)
4 4

2 2

0 0
0 5 10 15 20 25 0 5 10 15 20 25

δ (mm) δ (mm)

10 10

TASCα3-R0 TASCα4-R0
8 8

6 6
V (Ton)

V (Ton)
4 4

2 2

0 0
0 5 10 15 20 25 0 5 10 15 20 25
δ (mm) δ (mm)

Experimental Analytical
First cracking Failure Collapse
(a) R0 beams
25 25

TASCα1-R1 TASCα2-R1
20 20

15 15
V (Ton)

V (Ton)

10 10

5 5

0 0
0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45
δ (mm) δ (mm)

25 25

TASCα3-R1 TASCα4-R1
20 20

15 15
V (Ton)

V (Ton)

10 10

5 5

0 0
0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45
δ (mm) δ (mm)

Experimental Analytical
First cracking Failure Collapse

(b) R1 beams
Fig. 8. Comparison between experimental and analytical V–D curves.
 E.A. Godínez-Domínguez et al. / Engineering Structures 105 (2015) 99–122 107

10 10

TASCα1-R0 TASCα2-R0
8 8

6 6

V (Ton)
V (Ton)
4 4

Experimental Experimental
2 Measured 2
Measured
Analytical Analytical

0 0
0 5 10 15 20 25 0 5 10 15 20 25

δ (mm) δ (mm)

10 10

TASCα3-R0 TASCα4-R0
8 Experimental 8 Experimental
Measured Measured
Analytical Analytical
6 6

V (Ton)
V (Ton)

4 4

2 2

0 0
0 5 10 15 20 25 0 5 10 15 20 25
δ (mm) δ (mm)

Fig. 9. Comparison between experimental and analytical V–D curves using different considerations for the definition of the nonlinear shear springs properties.

Table 2
Properties for the concrete used in the preliminary simplified analysis [3].
0 0
Element Cylinder f c , Mpa (kg/cm2 or psi) f c prom , Mpa (kg/cm2 or psi) E, GPa (kg/cm2 or ksi) pEffiffiffi0
fc

TASCa1-R0 1 27.56 (281 or 3995.82) – – –

2 35.40 (361 or 5133.42) 31.48 (321 or 4564.62) 14.52 (148,049 or 2105.25) 2588
TASCa2-R0 1 30.30 (309 or 4393.98) – – –
2 27.56 (281 or 3995.82) 28.93 (295 or 4194.9) 14.96 (152,503 or 2168.59) 2781
TASCa3-R0 1 25.20 (257 or 3654.54) – – –
2 21.09 (215 or 3057.30) 23.14 (236 or 3355.92) 14.45 (147,345 or 2095.24) 3004
TASCa4-R0 1 26.77 (273 or 3882.06) – – –
2 28.34 (289 or 4109.58) 27.56 (281 or 3995.82) 13.75 (140,186 or 1993.44) 2619
TASCa1-R1 1 29.81 (304 or 4322.88) – – –
2 22.95 (234 or 3327.48) 26.38 (269 or 3825.18) 15.88 (161,903 or 2302.26) 3091
TASCa2-R1 1 29.32 (299 or 4251.78) – – –
2 27.95 (285 or 4052.70) 28.64 (292 or 4152.24) 14.06 (143,347 or 2038.39) 2627
TASCa3-R1 1 27.66 (282 or 4010.04) – – –
2 28.83 (294 or 4180.68) 28.24 (288 or 4095.36) 13.79 (140,613 or 1999.52) 2595
TASCa4-R1 1 21.58 (220 or 3128.40) – – –
2 20.01 (204 or 2900.88) 20.79 (212 or 3104.64) 11.74 (119,700 or 1702.13) 2575
Average – 27.85 (284 or 4038.48) – – 2735
r – 3.96 – – 205
C.V. – 14% – – 8%

in a three-dimensional finite element model of a concrete struc-
ture: (a) the discrete model, (b) the embedded model and, (c) the
smeared-crack model.
In the discrete model, which is used in this study, reinforcement
is modeled using bar or beam elements connected to the concrete
mesh nodes. As a result, there are ‘‘shared nodes” between the con-
crete mesh and the reinforcement mesh, as shown in Fig. 10a. Also,
since the reinforcement is superimposed in the concrete mesh,
concrete exists in the same regions occupied by the reinforcement.
The drawback of using the discrete model is that the concrete mesh
is restricted by the location of the reinforcement. For simplicity,
and as a first approximation, perfect bond is assumed between
the reinforcement and the concrete. The above is a priori a known
limitation, because from the experimental testing, cracking, crush-
ing of concrete and loss of concrete cover due to bond-slip failures
(debonding) of the inclined reinforcement were observed [1,3].
To overcome mesh dependency in the discrete model, the
embedded formulation allows independent choice of concrete
108 E.A. Godínez-Domínguez et al. / Engineering Structures 105 (2015) 99–122

Table 3
Computed displacements using an analytical stiffness formulation for nonprismatic beam elements [3].

Element Load condition V, t (kip) Displacement, mm (in.)

Experimental Analytical
TASCa1-R0 First cracking 5.75 (12.68) 7.42 (0.292) 2.90 (0.114)
Failure 6.75 (14.88) 8.88 (0.358) 4.81 (0.189)
TASCa2-R0 First cracking 5.00 (11.03) 7.82 (0.308) 3.51 (0.138)
Failure 6.00 (13.23) 12.73 (0.501) 5.92 (0.233)
TASCa3-R0 First cracking 2.75 (6.06) 4.74 (0.187) 3.12 (0.123)
Failure 3.75 (8.27) 7.85 (0.309) 5.76 (0.227)
TASCa4-R0 First cracking 2.50 (5.51) 4.28 (0.169) 4.90 (0.193)
Failure 3.00 (6.62) 12.46 (0.491) 7.54 (0.297)
TASCa1-R1 First cracking 11.00 (24.26) 16.09 (0.633) 4.98 (0.196)
Failure 20.00 (44.10) 29.50 (1.161) 13.21 (0.520)
TASCa2-R1 First cracking 8.75 (19.29) 15.00 (0.591) 6.42 (0.253)
Failure 17.00 (37.49) 30.79 (1.212) 17.24 (0.679)
TASCa3-R1 First cracking 4.00 (8.82) 7.71 (0.304) 4.65 (0.183)
Failure 12.00 (26.46) 30.21 (1.189) 18.58 (0.731)
TASCa4-R1 First cracking 4.00 (8.82) 9.21 (0.363) 8.97 (0.353)
Failure 8.00 (17.64) 38.34 (1.509) 22.17 (0.873)

Concrete element Concrete element Concrete element

Concrete node Concrete node Concrete node

Compatible desplacements Smeared properties of steel

"Shared node" between between concrete and in concrete elements
concrete and reinforcement
reinforcement elements

Reinforcement node
Reinforcement element

(a) Discrete model (b) Embedded model (c) Smeared-crack model

Fig. 10. Techniques used for modeling reinforcement in a concrete structure (adapted from Tavárez [23]).

mesh, as shown in Fig. 10b. In this approach, the stiffness of the

reinforcing elements is evaluated independently from the concrete
elements, but the element is built into the concrete mesh in such a
way that its displacements are compatible with those of surround-
σcu Peak compressive strength
ing concrete elements. That is, the concrete elements and their Eo
intersection points with each reinforcement segment are identified
and used to establish the nodal locations of the reinforcement ele-
ments. Finally, in the smeared-crack model, the reinforcement is Softening
assumed to be uniformly distributed over the concrete elements, Compression
as shown in Fig. 10c. As a result, the properties of the material
model in the element are constructed from individual properties
of concrete and reinforcement using composite theory. This tech- Strain at maximum stress
nique is usually applied for large structural models, where rein- +ε ε0 σyp -ε
forcement details are not essential to capture the overall
response of the structure [23]. For these reasons, this modeling σtu = Maximum tensile strength of concrete
was not selected in this study. Tension
The desired technique for the modeling of the reinforcement
is chosen depending on the application and the degree of detail +σ
in which the model needs to be developed. However, most of
Fig. 11. Typical uniaxial compressive and tensile stress–strain curve for concrete
the difficulties in modeling reinforced concrete behavior rely
(adapted from Kachlakev et al. [16]).
in the development of an effective and realistic concrete mate-
rial formulation, and not in the modeling of the reinforcement
[23].
this point, the stress increases gradually up to the maximum
5.2. Material modeling assumptions compressive strength. After it reaches the maximum compressive
strength rcu, the curve descends into a softening region, and even-
5.2.1. Concrete tually crushing failure occurs at an ultimate strain ecu. In tension,
Development of a model for the behavior of concrete is a chal- the stress–strain curve for concrete is approximately linearly
lenging task. Concrete is a quasi-brittle material and has different elastic up to the maximum tensile strength. After this point, the
behavior in compression and tension. As shown in Fig. 11, in com- concrete cracks and the strength decreases gradually to zero [16].
pression, the stress–strain curve for concrete is linearly elastic up Shear transfer coefficients for open and closed cracks (bt and bc),
0
to about 30% of the maximum compressive strength (rcu). Above ultimate uniaxial compressive strength (f c ), ultimate uniaxial
 E.A. Godínez-Domínguez et al. / Engineering Structures 105 (2015) 99–122 109

Table 4
Material properties to define the behavior of concrete in ANSYS.

Member fc0 , Mpa (kg/cm2 or psi) ft, Mpa (kg/cm2 or psi) E, Gpa (kg/cm2 or psi) m bt bc
TASCa1-R0 31.48 (321 or 4564.62) 3.15 (32.1 or 456.4) 11.28 (115,022 or 1635.61) 0.2 0.25 0.30
TASCa2-R0 28.93 (295 or 4194.90) 2.89 (29.5 or 419.49) 11.73 (119,654 or 1701.48) 0.2 0.50 0.30
TASCa3-R0 23.14 (236 or 3355.92) 2.31 (23.6 or 335.59) 12.70 (129,531 or 1841.93) 0.2 0.20 0.25
TASCa4-R0 27.56 (281 or 3995.82) 2.76 (28.1 or 399.58) 13.75 (140,186 or 1993.44) 0.2 0.20 0.30
TASCa1-R1 26.38 (269 or 3825.18) 2.64 (26.9 or 382.52) 13.11 (133,731 or 1901.65) 0.2 0.50 0.30
TASCa2-R1 28.64 (292 or 4152.24) 2.86 (29.2 or 415.22) 11.38 (116,039 or 1650.07) 0.2 0.25 0.30
TASCa3-R1 28.24 (288 or 4095.36) 2.82 (28.8 or 409.54) 11.45 (116,709 or 1659.60) 0.2 0.10 0.40
TASCa4-R1 20.79 (212 or 3014.64) 2.08 (21.2 or 301.46) 11.74 (119,700 or 1702.13) 0.2 0.20 0.25

σyp
Cracking fr
f c´
zp >0 (Cracking)
zp =0 (Crushing)
zp <0 (Crushing)
Fig. 12. Failure surface for concrete considered in ANSYS [13].
tensile strength (modulus of rupture, fr), elastic modulus (E),
Poisson’s ratio (m) and, reduction stiffness factor for cracked tensile
condition are required to define the behavior of concrete in ANSYS.
In all cases a tensile strength of concrete of 10% of the compressive
strength was considered.
The shear transfer coefficient, bt, represents conditions of the
crack face. The value of bt ranges from 0.0 to 1.0, with 0.0 repre-
senting a smooth crack (complete loss of shear transfer) and 1.0
representing a rough crack (no loss of shear transfer [13]). There
is a significant number of studies regarding to modeling of rein-
forced concrete structures in which different values of bt and bc
are used [8,9,16,18,21,22]. From the cited studies, the coefficient
bt usually varies between 0.05 and 0.65, and the coefficient bc usu-
ally varies between 0.25 and 0.90 (a wide range of values). In pre-
vious studies of reinforced concrete beams designed to develop a
flexural failure [16,18], it has been pointed out that convergence
problems occurred when the shear transfer coefficient for the open
crack bt dropped below 0.2. Therefore, and since the failure mech-
anism of the RCHBs not correspond to most previous studies (shear
failure), a number of preliminary analyses were carried out to
assess the influence of bt and bc on the model behavior and,
thereby, the shear transfer coefficients used in the analyses were
defined. The values considered in each numerical simulation to
define the behavior of concrete in ANSYS are shown in Table 4.
For all cases a reduction stiffness factor for cracked tensile condi-
tion equal to 0.9 was used. For the definition of the reduction stiff-
ness factor for cracked tensile condition, there are some research
studies in which once the concrete material reaches its tensile peak
stress, a tensile stiffness multiplier of 0.6 is used to simulate a sud-
den drop of the tensile stress to 60% of the rupture stress; followed
by a linearly descending curve to zero stress [8,9]. Nevertheless,
the literature review shows that the reduction stiffness factor for
cracked tensile condition has been considered using different val-
Cracking ues. For example, Padmarajaiah and Ramaswamy [24] used values
σxp ranging from 0.45 to 0.95 for the finite element assessment of flex-
fr ural strength of prestressed concrete beams with fiber reinforce-
ment. Also, there are some research studies that consider a very
different value for this factor, even a value of 0.0 was considered
for the analyses [18,21]. Finally, there are some research studies
where the criteria used for the definition of reduction stiffness fac-
tor for cracked tensile condition is not mentioned within the paper
Cracking
[16,17]. Based upon the above, and considering the good results
obtained in a previous analytical research in which the behavior
of concrete slabs were studied using ANSYS models considering a
f c´ reduction stiffness factor for cracked tensile condition equal to
0.9 [22], the authors considered also appropriate to use 0.9 for
the tensile stiffness multiplier.
The considered model for the behavior of concrete in ANSYS is
capable of predicting failure for concrete materials. Both cracking
and crushing failure modes are accounted for. Ultimate uniaxial
tensile and compressive strengths are needed to define a failure
surface for the concrete. A three-dimensional failure surface for
concrete proposed by Willam and Warnke in 1975 is shown in
Fig. 12, in which a multiaxial stress state is considered [16].
The most significant nonzero principal stresses are in the x and
y directions, represented by rxp and ryp, respectively. Three failure
surfaces are shown as projections on the rxp–ryp plane. The mode
of failure is a function of the sign of rzp (principal stress in the z
direction). For example, if rxp and ryp are both negative (compres-
sive) and rzp is slightly positive (tensile), cracking would be pre-
dicted in a direction perpendicular to rzp. However, if rzp is zero
or slightly negative, the material is assumed to crush [13].
In a concrete element, cracking occurs when the principal ten-
sile stress in any direction lies outside the failure surface. After
cracking, the elastic modulus for the concrete element is set to zero
in the direction parallel to the principal tensile stress direction.
Crushing occurs when all principal stresses are compressive and
lie outside the failure surface. Subsequently, the elastic modulus
is set to zero in all directions, and the element effectively disap-
pears [16].
During this study, as previously reported by Kachlakev et al.
[16] and Wolanski [18], it was found that if the crushing capability
for the concrete is set in, the reinforced concrete haunched beam
finite element models fail prematurely. Crushing of the concrete
started to develop in elements located directly under the applied
loads. Subsequently, adjacent 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.
Due to aforementioned, and taking into account that a pure
compression failure for the concrete is not congruent to the
observed failure mode of RCHBs during testing, the recommenda-
tion made by Kachlakev et al. [16] and Wolanski [18] was consid-
ered. Therefore, the crushing capability was turned off and
110 E.A. Godínez-Domínguez et al. / Engineering Structures 105 (2015) 99–122

Table 5
Steel reinforcement properties used in the finite element models.

Specimen fy, Mpa (kg/cm2 or ksi) ey esh fsu, Mpa (kg/cm2 or ksi) esu
No. 8 bars 426.41 (4348 or 61.83) 0.00237 0.0086 755.83 (7707 or 109.59) –
Reference 448.87 (4577 or 65.08) – 0.0088 734.64 (7491 or 106.52) 0.1171
No. 2.5 bars 450.34 (4592 or 65.30) 0.00235 0.0074 730.43 (7448 or 105.91) –
Reference 452.00 (4609 or 65.54) – 0.006 729.25 (7436 or 105.74) 0.142

j
4
4 p
P 5
5
o
O K
6
6
m
M i n
N
2 y
2 l
L z
3
3
Z k
K x
Z y
i
Y I 1
1 x
X Y j
J
X Surface coordinate system

(a) Solid65 element (b) Link8 element (c) Solid45 element

Fig. 13. Element types used for modeling RCHBs in ANSYS [13].

cracking for the concrete then controlled the failure of the finite
element models.
5.2.2. Steel reinforcement, steel loading and bearing plates
The properties for flexural and shear steel reinforcement used
in the finite element models are shown in Table 5, and were
obtained from the experimental program [3].
Bilinear isotropic material based on the Von Mises failure crite-
ria identical in tension and compression was used to model all the
steel reinforcement. Poisson’s ratio m = 0.3 was used for the steel
reinforcement in this study. For defining the tangent modulus, as
ultimate steel deformations were not reported in the experimental
study [3], the information derived from a statistical study reported
by Rodriguez and Botero [25] for different diameter steel bars used
in Mexico was considered, and is specified as ‘‘Reference” in
Table 5. As it can be observed from Table 5, the yield and ultimate
stresses, as well as yield deformations reported by Archundia [3]
are consistent with those reported by Rodriguez and Botero [25].
For all studied specimens steel bearing plates at loading points
and supports on the beam were modeled. Linear behavior was con-
sidered for the steel bearing plates. Bearing plates were modeled as
they were used in order to avoid stress concentrations in these
areas. A modulus of elasticity E = 210 GPa (2,100,000 kg/cm2 or
147.84 ksi) and a Poisson’s ratio of m = 0.3 were used for the steel
plates.
5.3. Element types used for modeling
The Solid65 element [13] was used to model the concrete ele-
ments. This element has eight nodes with three degrees of freedom
per node (translations in the nodal x, y, and z directions) and it is
capable of plastic deformation, cracking and crushing in three
orthogonal directions (Fig. 13a).
In order to use a discrete model, the Link8 element [13] was
used to model the steel reinforcement (Fig. 13b). This element is
a 3D bar element and it has two nodes with three degrees of free-
dom per node (translations in the x, y, and z directions). This ele-
ment is also capable of plastic deformation.
An eight-nodded solid element (Solid45) [13] was used to
model the bearing steel plates located at loading and supports
points (Fig. 13c). The element is defined with eight nodes having
three degrees of freedom per node (translations in the x, y, and z
directions).
5.4. Finite element modeling of RCHBs
5.4.1. Calibration models
Before carrying out the finite element modeling of RCHBs in
ANSYS [13], two calibration models were used. First, a simple
model of a reinforced concrete prism with a steel bar subjected
to tension was reproduced. This model was used by Lowes [15]
for the study of the bond zones in reinforced concrete bridges
connections. Subsequently, a second model in which a simply
supported prismatic reinforced concrete beam subjected to two
concentrated loads at one third the clear span was studied. This
model was used by Foley and Buckhouse [14]. For the two
calibration finite element models, shear–deformation curves
(Figs. 14 and 15) and crack patterns were obtained and compared
to those reported previously, and a good correlation was observed.
The results of these calibrations models are discussed in detail
elsewhere [26].
In addition, for all the RCHBs finite element models, the selec-
tion of the mesh density was studied prior to the performance of
the final analyses. In this case the strain energy criterion was used
to establish the mesh density. The mesh density for each RCHB
model was selected when convergence of results were obtained
(when an increase in the mesh density has a negligible effect on
the results).
5.4.2. General considerations
In order to reduce the computational effort, symmetry
conditions for the studied RCHB were taken into account, which
allowed one quarter of the beam to be modeled, as shown in
 E.A. Godínez-Domínguez et al. / Engineering Structures 105 (2015) 99–122 111

[ ]

(a) Experimental results obtained by Foley

y Buckhouse in 199818

18000

16000

14000

Avg. Load, P (lbs.) 12000

10000

8000

6000

4000

2000

0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5

Avg. Centerline Deflection (in.)

Fig. 14. Results for the first calibration model.
Fig. 16. Therefore, the corresponding length for the haunched beam
section modeled is 165 cm (66 in.) with a width of 11 cm (4.4 in.),
as depicted in Fig. 16.
It is worth noting that the meshing for each RCHB varies,
because for each beam in which shear reinforcement is needed
(R1 series), additional stirrups were placed at different spacing
from the vertex formed by the intersection of tapered sections with
the prismatic section, to account for the vertical force introduced
by the inclined longitudinal steel reinforcement in tension due to
the change of direction (Figs. 17 and 18).
The finite element model for the RCHB identified as TASCa4-R1
is schematically shown in Fig. 19. The concrete beam and the load-
ing and support plates are identified in Fig. 19a. Also, the meshing
corresponding to the flexural and shear reinforcement is shown in
Fig. 19b and c, respectively.
For all models, the top flexural steel reinforcement area at the
plane of symmetry is one half the normal area for a #8 bar because
one half of the bar is cut off in the meshing. As it can be observed
from Fig. 19c, flexural and shear steel reinforcement have some
shared nodes. For the bottom flexural reinforcement, formed by
two bar package, 2#8 bars for each package, only one package is
modeled because of the symmetry of the beam, using an equiva-
lent area element. Transverse steel reinforcement (stirrups) is
modeled throughout the beam. Only half of the stirrup is modeled
because of the symmetry of the beam (Fig. 19c).
(b) Results of the second calibration model
obtained using ANSYS
Fig. 15. Results for the second calibration model.
5.4.3. Loads and boundary conditions
As it is well known, adequate displacement boundary
conditions are needed to constrain the model to get an equilibrium
solution. Because a quarter of the entire beams were used for the
finite element models, planes of symmetry were required at
the internal faces. At a plane of symmetry, the displacement in
the direction perpendicular to the plane was held at zero, as
schematically shown in Fig. 20.
In order to simulate support conditions for the RCHBs used dur-
ing experimental tests (simply supported), supports were modeled
in such a way that a roller was created. A single line of nodes on the
plate were given constraint in the y and z directions. This was done
because when the loaded beam starts to displace downward, rota-
tion of the plate should be permitted. Excessive cracking for the
concrete elements above the steel plate could be developed if
rotation of the steel plate is not permitted [16]. The line of nodes
on the plate is located in the same place that was considered in
the experimental program (5 cm or 2 in. from the vertex).
In order to obtain force–displacement curves, displacements on
the global negative Y direction on a specific line of nodes located at
the loading plate were imposed. In order to obtain the corresponding
forces, reactions in these nodes were obtained for each step. Final
112 E.A. Godínez-Domínguez et al. / Engineering Structures 105 (2015) 99–122

11 cm

165 cm

Y
Z
X

Fig. 16. Haunched beam section modeled (continuous line).

T sen
M
T (1- cos )
T

Fig. 17. Vertical force introduced by the inclined longitudinal steel reinforcement in tension due to the change of direction (adapted from Archundia [3]).

Stirrup at vertex

scd

Additional stirrups near

of the vertex point

Fig. 18. Additional stirrups near of the vertex point due to the change of direction of longitudinal steel reinforcement in tension (adapted from Archundia [3]).

displacements were obtained as the average of the displacements
of the nodes located at midspan, in this case, the nodes located
at the right end of the model.
Quasi-static analyses considering small displacements were
done. As discussed in the ANSYS user manual [13], considering
large displacements in models where the Solid65 element is used
may lead to obtain unrealistic results.
5.5. Comparison between numerical and experimental results
In this section, comparisons between the experimental data and
the results from ANSYS models are shown. The main reported
results are: (1) shear force–deflection curves at midspan and (2)
crack patterns associated to different loads levels (first cracking,
failure and collapse).
5.5.1. Shear force–deflection curves
The applied shear force vs midspan deflection curves numeri-
cally obtained are compared in Figs. 21 and 22 with those obtained
experimentally. In each figure three points corresponding to the
first diagonal cracking, failure and collapse, according to the exper-
imental values reported [1] are depicted. As it can be observed
from the graphs, the point associated to collapse often cannot be
accurately defined in the experimental curve (Figs. 21 and 22).
 E.A. Godínez-Domínguez et al. / Engineering Structures 105 (2015) 99–122 113

Concrete haunched beam

Steel loading
bearing plate

Steel support
bearing plate

(a) Concrete beam and loading and support bearing plates

(b) Mesh for the concrete, and loading and support bearing plates

1 # 8 bar at the
plane of symmetry
Stirrups #2.5 due to the change
of direction of longitudinal steel
Stirrups # 2.5
reinforcement intension

Equivalent area Shared nodes

element (2#8 bars)

(c) Detail for the flexural and shear reinforcement

Fig. 19. Finite element modeling for TASCa4-R1 in ANSYS.

As it can be observed in Fig. 21, a plausible correlation between
experimental and analytical curves exists for most models without
shear reinforcement. The only exception is for model TASCa3-R0
(Fig. 21c), in which the notoriously larger experimental collapse
displacement (with respect to the other tested specimens, Tena-
Colunga et al. [1]) could not be reproduced numerically. An abrupt
change in the slope can be observed from the experimental curve
starting at the point where failure is predicted using finite
elements. It is observed by comparing experimental and analytical
curves that shear forces and displacements associated with the
first diagonal cracking and failure have a reasonable correlation,
although the collapse point is not always well represented.
It can be observed from Fig. 22 that for RCHBs with shear rein-
forcement (R1 series), a reasonable good correlation between
numerical and experimental curves also exists, except for
TASCa4-R1 (Fig. 22d). A better correlation is observed for the R1
series (Fig. 22) than for the R0 series (Fig. 21) in the estimates
for the points associated to first cracking, failure and collapse. As
it can be observed from Fig. 22d, for model TASCa4-R1, the
experimental collapse displacement could not be reproduced
numerically using the described finite element modeling. Again,
as noted for the TASCa3-R0 model, an abrupt change in the slope
can be observed from the experimental curve for TASCa4-R1, start-
ing at the point where failure is predicted using finite elements.
5.5.2. Cracking patterns
Cracking patterns were obtained for the following three load
steps: (a) first diagonal cracking, (b) failure and, (c) collapse.
Details for all the observed experimental cracking patterns can
be found elsewhere [1,3].
In order to ease the understanding and interpretation of crack-
ing patterns shown in this section, typical cracking signs to identify
114 E.A. Godínez-Domínguez et al. / Engineering Structures 105 (2015) 99–122

Y Constraint in the Z direction

Z
X

Support roller condition

to allow rotation

Constraint in the X direction

Fig. 20. Boundary conditions for planes of symmetry and supports modeling.

10 10
TASCα1-R0 TASCα2-R0

8 8

6 6
V (Ton)

V (Ton)

4 4

2 2

0 0
0 5 10 15 20 25 0 5 10 15 20 25 30 35 40

δ (mm) δ (mm)

(a) TASCα1-R0 (b) TASCα2-R0

10 5
TASCα3-R0 TASCα4-R0

8 4

6 3
V (Ton)

V (Ton)

4 2

2 1

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25

δ (mm) δ (mm)

(c) TASCα3-R0 (d) TASCα4-R0

Experimental ANSYS

First cracking Failure Collapse

Fig. 21. Experimental vs numerical shear force–displacement curves for models without shear reinforcement.
 E.A. Godínez-Domínguez et al. / Engineering Structures 105 (2015) 99–122 115

25 25
TASCα1-R1 TASCα2-R1

20 20

15 15

V (Ton)

V (Ton)
10 10

5 5

0 0
0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45

δ (mm) δ (mm)

(a) TASCα1-R1 (b) TASCα2-R1

25 10
TASCα3-R1 TASCα4-R1

20 8

15 6
V (Ton)

V (Ton)

10 4

5 2

0 0
0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45

δ (mm) δ (mm)

(c) TASCα3-R1 (d) TASCα4-R1

Experimental ANSYS
First cracking Failure Collapse
Fig. 22. Experimental vs numerical shear–displacement curves for models with shear reinforcement.

Compressive cracks Diagonal tensile

Flexural cracks
cracks

Fig. 23. Cracking signs used in ANSYS models.

flexural, compressive and diagonal tensile cracks in an ANSYS
model are shown in Fig. 23. Flexural cracks are identified as
vertical straight lines occurring at the integration points for the
concrete solid elements. Compressive cracks are identified using
circles that appear perpendicular to the principal tensile strains
in the z direction at integration points in the concrete elements
(loads in the y direction result in tensile strains in the z direction
by Poisson’s effect). Diagonal tensile cracks are identified as
inclined circles appearing as straight lines perpendicular to the
directions for the principal stresses at integration points for the
concrete elements.
5.5.2.1. RCHBs without shear reinforcement. In order to show the
actual damage identified during experimental tests, photographs
corresponding to the final damaged state for RCHBs without shear
reinforcement are shown in Fig. 24. Cracking patterns obtained
116 E.A. Godínez-Domínguez et al. / Engineering Structures 105 (2015) 99–122

(a) TASCα1-R0 (b) TASCα2-R0

(c) TASCα3-R0 (d) TASCα4-R0

Fig. 24. Final damaged state for RCHBs without shear reinforcement [3].

(a) First diagonal cracking

(b) Failure

(c) Collapse
Fig. 25. Experimental and numerical crack patterns for model TASCa1-R0.

from ANSYS finite element models are compared with those
identified during the experimental testing and are shown in
Figs. 25–28.
It is worth noting that, for a better interpretation of numerical
results, in ANSYS circles are deployed at integration points that
are cracked or crushed. Cracks are plotted in ANSYS regardless they
have either large or small open size. This is a limitation of the
analytical formulation implemented in ANSYS, since it is no
possible to get an estimation of the width and length for the crack.
A crack is represented by a circle in the direction of the cracking
plane, while crushing is indicated by an octahedron. If the crack
is opened and then closed, an ‘‘X” inside the circle is displayed.
Each integration point can represent cracking in three different
planes, where the first crack is displayed with a red circle, the sec-
ond one with a green circle and the third one with a blue circle.
Therefore, main cracks are normally indicated in blue and/or green,
because are those involving two or three different planes.
From the observation of Figs. 25–28, it could be concluded that
a low to medium correlation exist between cracking patterns
numerically obtained with those experimentally identified. In all
cases, cracking patterns numerically obtained are larger than those
experimentally identified (even if regions indicated by red circles
are neglected). From the numerical models, it is observed that on
the prismatic zone, flexural cracks propagate almost to midspan
of the beam section. Nevertheless, during experimental testing
flexural cracks were not always identified in the prismatic zone.
In fact, for RCHBs TASCa3-R0 (Fig. 27) and TASCa4-R0 (Fig. 28)
flexural cracks were not identified at all. The greater intensity
obtained for the numerical cracking may be due to: (a) the cracking
modeling used, in which material degradation is distributed over
 E.A. Godínez-Domínguez et al. / Engineering Structures 105 (2015) 99–122 117

(a) First diagonal cracking

(b) Failure

(c) Collapse
Fig. 26. Experimental and numerical crack patterns for model TASCa2-R0.

(a) First diagonal cracking

(b) Failure

(c) Collapse
Fig. 27. Experimental and numerical crack patterns for model TASCa3-R0.

all the volume of the element and, (b) perfect bond was assumed Then, flexural cracking becomes more important at the vertex zone
between the reinforcement and the concrete, causing high stress (concentration of highly localized and intense blue circles at the
concentrations on the concrete at the vertex zone due to the vertex area) for higher haunched angles (Figs. 27 and 28), respect
change of direction of longitudinal steel reinforcement in tension. to those with low haunched angles (Figs. 25 and 26). It is worth
118 E.A. Godínez-Domínguez et al. / Engineering Structures 105 (2015) 99–122

(a) First diagonal cracking

(b) Failure

(c) Collapse
Fig. 28. Experimental and numerical crack patterns for model TASCa4-R0.

(a) TASCα1-R1 (b) TASCα2-R1

(c) TASCα3-R1 (d) TASCα4-R1

Fig. 29. Final cracking state for RCHBs with shear reinforcement [3].

noting that debonding cracks were observed during experimental
testing (Fig. 24), which developed particularly near to failure and
between failure and collapse limit states (Figs. 25–28).
From the interpretation of experimental results, it was con-
firmed that RCHBs tend to develop an arch mechanism which
allows the damage to be distributed in terms of several fissures
along the haunched length before the main diagonal crack devel-
ops, then causing a failure mechanism that is less fragile than the
typical sudden shear failure observed in prismatic beams [1]. From
the study of the numerical crack patterns (mainly for those indi-
cated in green and blue), it is clear that for RCHBs without shear
reinforcement, this arch mechanism is reasonably captured,
mainly for models TASCa2-R0 and TASCa4-R0 (Figs. 26 and 28).
5.5.2.2. RCHBs with shear reinforcement. Photographs for the final
cracking states for beams with shear reinforcement are shown in
 E.A. Godínez-Domínguez et al. / Engineering Structures 105 (2015) 99–122 119

(a) First diagonal cracking

(b) Failure

(c) Collapse
Fig. 30. Experimental and numerical crack patterns for model TASCa1-R1.

(a) First diagonal cracking

(b) Failure

(c) Collapse
Fig. 31. Experimental and numerical crack patterns for model TASCa2-R1.

Fig. 29. Also, crack patterns obtained from the ANSYS models are
compared with those experimentally obtained in Figs. 30–33. A
better correlation between experimental and analytical crack
patterns was observed for RCHBs with shear reinforcement,
primarily regarding to flexural cracks located at the prismatic zone,
since in this case the finite element models captures better this
effect. As observed, the evolution of diagonal tensile cracks from
the loading area toward the support area and the formation of
compression struts at haunched zones are reasonably captured.
6. Concluding remarks
The presented study focused on the numerical modeling using
nonlinear finite elements of simply supported reinforced concrete
120 E.A. Godínez-Domínguez et al. / Engineering Structures 105 (2015) 99–122

(a) First diagonal cracking

(b) Failure

(c) Collapse
Fig. 32. Experimental and numerical crack patterns for model TASCa3-R1.

(a) First diagonal cracking

(b) Failure

(c) Collapse
Fig. 33. Experimental and numerical crack patterns for model TASCa4-R1.

haunched beams designed to develop a shear failure under
static loading. The purpose of this study was to assess the ability
and limitations of simple and complex analytical models to predict
the structural behavior of reinforced concrete haunched beams
(RCHBs) failing in shear.
As a first step, nonlinear static analyses were performed using
SAP2000. Simple models using beam-elements with variable sec-
tion were used. Lumped plasticity models using plastic shear
hinges and the average mechanical properties for the reinforced
concrete were evaluated. Subsequently, more complex finite
 E.A. Godínez-Domínguez et al. / Engineering Structures 105 (2015) 99–122
element models were used in ANSYS, in which the longitudinal and
transverse steel reinforcement were included as built.
From the finite element models, shear force vs displacement
curves for a specific section located at midspan for the beams were
obtained. Also, from the ANSYS models, crack patterns associated
to different loads levels were obtained.
It was observed from simplest models, in which the failure is
modeled using shear plastic hinges, that it is possible to obtain
plausible estimates of the shear–displacement curve if the ultimate
shear could be reasonably assessed using expressions available in
the literature [1] and a reasonable assessment of the initial elastic
stiffness [6] is obtained. However, in most cases, the ultimate dis-
placement is often underestimated, mainly for RCHBs without
shear reinforcement. Nevertheless, for those cases, the apparent
failure point could be reasonably estimated.
From the results of ANSYS finite element models, a reasonable
correlation between analytical and experimental shear force vs dis-
placement curves was observed. For these cases, the shear and dis-
placements associated with first diagonal cracking, failure and
collapse limit states had a reasonably good correlation, especially
for RCHBs with shear reinforcement.
For practical purposes, it could be concluded that a medium
correlation between cracking patterns numerically obtained with
ANSYS and those experimentally identified were observed, partic-
ularly for beams with shear reinforcement. The arch mechanism
observed experimentally in RCHBs was reasonably captured with
the analytical models. Also, the evolution of diagonal tensile cracks
from the loading area toward the support area was moderately
reproduced; and the formation of compression struts at haunched
zones was captured.
It is worth noting that for all cases under study, the extent of
cracking numerically obtained was larger than the experimentally
identified, and sometimes it was extended to areas where cracking
was not physically observed. Extensive flexural cracks are pre-
dicted by the finite element modeling at the prismatic zone,
whereas flexural cracks were not always identified on the pris-
matic zone of RCHBs during experimental testing. The greater
intensity obtained for the numerical cracking may be due to: (a)
the cracking modeling used, in which material degradation is dis-
tributed over all the volume of the element and (b) perfect bond
was assumed between the reinforcement and the concrete, causing
high stress concentrations on the concrete at the vertex zone due
to the change of direction of longitudinal steel reinforcement in
tension. It is worth noting that cracks are plotted in ANSYS regard-
less they have either large or small opening size. This is a limitation
of the analytical formulation implemented in ANSYS, since it is no
possible to get an estimation of the width and the length for the
crack.
It is worth mentioning that for simplicity, and as a first approx-
imation, in all cases perfect bonding between the steel reinforce-
ment and the concrete was assumed. This was a known
limitation a priori, because cracking, crushing and loss of cover
for the concrete due to bond-slip (debonding) failures of the
inclined longitudinal reinforcement were observed during experi-
mental testing, mainly associated to collapse loads [1,3]. It is well
known that stress redistributions obtained using perfect bond are
different from those obtained considering more complex models
in which the bond-slip between the concrete and the steel rein-
forcement is modeled. For this reason, additional finite element
analyses where formulations capable of including nonlinearities
due to bond-slip are planned. These studies would allow improving
the modeling and then assessing if such modeling refinement
would contribute to obtain a better correlation between the crack-
ing patterns obtained with nonlinear finite elements with respect
to those identified during experimental tests. It is worth noting
that in general, the nonlinear finite element models used in this
121
study were reasonably capable to reproduce global load–deforma-
tion curves and the main developed arch mechanism, which for
practical purposes is more than acceptable.
Acknowledgements
The first two authors gratefully acknowledged financial support
of CONACYT as part of the Basic Science Research Project No. 79878
‘‘Comportamiento cíclico a flexión y cortante de trabes acarteladas
de concreto reforzado continuas” (in Spanish), in which the first
author participated during his postdoctoral stay at UAM-A. Con-
structive comments from anonymous reviewers for this paper
were very helpful to improve it and are gratefully acknowledged.
References
[1] Tena-Colunga A, Archundia-Aranda HI, González-Cuevas OM. Behavior of
reinforced concrete haunched beams subjected to static shear loading. Eng
Struct 2008;30:478–92.
[2] Archundia-Aranda HI, Tena-Colunga A, Grande-Vega A. Behavior of reinforced
concrete haunched beams subjected to cyclic shear. Eng Struct 2013;49:27–42.
http://dx.doi.org/10.1016/j.engstruct.2012.10.037.
[3] Archundia HI. Comportamiento a cortante de trabes acarteladas de concreto
reforzado. MSc thesis. Universidad Nacional Autónoma de México, México;
May 2004 [in Spanish].
[4] NTCC-04. Normas Técnicas Complementarias para Diseño de Estructuras de
Concreto. Gaceta Oficial del Distrito Federal, Tomo II, No. 103-BIS; October
2004 [in Spanish].
[5] Wilson. SAP 2000 Nonlinear Version 14 Structural analysis program. Berkeley
(California): Computer and Structures Inc.; 2009.
[6] Tena-Colunga A. Stiffness formulation for nonprismatic beam elements. ASCE J
Struct Eng 1996;122(12):1484–9.
[7] Çervenka V. Simulation of shear failure modes of RC structures. In: de Borst R,
Bicanic N, Mang H, Meschke G, editors. Computational modelling of concrete
structures (Euro-C 98). Rotterdam (The Netherlands): A.A. Balkema; 1998. p.
833–8.
[8] Hawileh RA, Rahman A, Tabatabai H. Nonlinear finite element analysis and
modeling of a precast hybrid beam–column connection subjected to cyclic
loads. Appl Math Model 2010;34:2562–83.
[9] Hawileh RA, Abdalla JA, Tanarslan MH. Modeling of nonlinear response of R/C
shear deficient t-beam subjected to cyclic loading. Comput Concr 2012;10
(4):413–28.
[10] Belletti B, Damoni C, Hendriks MAN. Development of guidelines for nonlinear
finite element analyses of existing reinforced and prestressed beams. Eur J
Environ Civ Eng 2011;15(9):1361–84.
[11] Slobbe AT, Hendriks MAN, Rots JG. Sequentially linear analysis of shear critical
reinforced concrete beams without shear reinforcement. Finite Elem Anal Des
2012;50(9):108–24.
[12] Albegmprli HM, Çevik A, Gülsßan ME, Kurtoglu AE. Reliability analysis of
reinforced concrete haunched beams shear capacity based on stochastic
nonlinear FE analysis. Comput Concr 2015;15(2):259–77.
[13] ANSYS. ANSYS User’s Manual version 12.0. ANSYS, Inc., Canonsburg,
Pennsylvania; 2009.
[14] Foley CM, Buckhouse ER. Strengthening existing reinforced concrete beams for
flexure using bolted external structural steel channels. Report MUST-98-1.
Marquette University, Department of Civil and Environmental Engineering;
January 1998.
[15] Lowes LN. Finite element modeling of reinforced concrete beam-column
bridge connections. PhD thesis. University of California, Berkeley; 1999.
[16] Kachlakev D, Miller T, Yim S, Chansawat K, Potisuk T. Finite element modeling
of reinforced concrete structures strengthened with FRP laminates. Report SPR
316, California Polytechnic State University, San Luis Obispo, CA and Oregon
State University, Corvallis, OR for Oregon Department of Transportation; May
2001.
[17] Santhakumar R, Chandrasekaran E. Analysis of retrofitted reinforced concrete
shear beams using carbon fiber composites. Elect J Struct Eng 2004;4:66–74.
[18] Wolanski AJ. Flexural behavior of reinforced and prestressed concrete beams
using finite element analysis. MSc thesis. Marquette University, Milwaukee,
Wisconsin, USA; 2004.
[19] Santhakumar R, Dhanaraj R, Chandrasekaran E. Behaviour of retrofitted
reinforced concrete beams under combined bending and torsion: a
numerical study. Elect J Struct Eng 2007;7:2–7.
[20] Purushotham B, Alagusundaramoorthy P, Sundaravadivelu R. Finite element
modeling of RC piles strengthened with GFRP composites. In: Proceedings,
international conference on mechanical engineering, Dhaka, Bangladesh;
2007.
[21] Baetu S, Ciongradi I. Nonlinear finite element analysis of reinforced concrete
slit walls with ANSYS (I). Buletinul Institutului Politehnic Din Iasßi, Tomul LVII
(LXI), Fasc. 1; 2011.
[22] Caballero O, Juárez G, Casillas J. Determinación de coeficientes de diseño y
trayectorias de agrietamiento de losas aisladas circulares, elípticas y
122 E.A. Godínez-Domínguez et al. / Engineering Structures 105 (2015) 99–122

triangulares. In: Proceedings, congress on numerical methods in engineering
2011, Coimbra, Portugal; June 2011. CDROM, Paper IDE. 358, p. 1–20 [in Spanish].
[23] Tavárez FA. Simulation of behavior of composite grid reinforced concrete
beams using explicit finite element methods. MSc thesis. University of
Wisconsin, Madison, Wisconsin; 2001.
[24] Padmarajaiah SK, Ramaswamy A. A finite element assessment of flexural
strength of prestressed concrete beams with fiber reinforcement. Cement
Concr Compos 2002;24:229–41.
[25] Rodríguez M, Botero JC. Comportamiento sísmico de estructuras considerando
propiedades mecánicas de aceros de refuerzo mexicanos. Revista de Ingeniería
Sísmica, SMIS, No. 49; 1995. p. 39–50 [in Spanish].
[26] Godínez EA, Tena A. Modelado numérico de trabes acarteladas de concreto
reforzado. Informe de Investigación presentado a CONACYT. Departamento de
Materiales, Universidad Autónoma Metropolitana Azcapotzalco; August 2011
[in Spanish].