You are on page 1of 14

Engineering Structures 284 (2023) 115947

Contents lists available at ScienceDirect

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

An improved analytical model for two-step corbels in a precast


concrete system
Ygor Moriel Neuberger a, Daniel de Lima Araújo b, *
a
Universidade Federal de Goiás, Escola de Engenharia Civil e Ambiental
b
Universidade Federal de Goiás, Escola de Engenharia Civil e Ambiental, Rua Universitária, n◦ 1488, Qd 86, Setor Universitário, Goiânia/GO, CEP: 74605-220, Brazil

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

Keywords: Corbels are a common element in the beam–column connections of precast concrete structures. This paper an­
Concrete corbels alyses the failure of a two-step corbels system using computational modeling. A finite element model is devel­
Precast concrete oped and calibrated using the test results for two-step corbels without secondary reinforcement but with steel
Two–step corbel
fiber–reinforced concrete. A database is developed from the parametric analyses of two-step corbels, varying the
Computational modeling
Analytical model
a/d ratio (0.34–0.76), the rate of the main reinforcement (0.19–0.50%), the compressive strength of the concrete
(30–70 MPa) and the amount of steel fibers (1.0% and 1.5%). An improved analytical model, using the developed
database, is proposed to design the two-step corbels. The reliability of the proposed analytical model is evaluated
using a second database, developed in this paper, and a strength reduction factor of 0.66 for failure by yielding of
the main reinforcement and 0.56 for failure by crushing of the concrete strut, were obtained. Therefore, the
improved analytical model obtained for the design of two-step corbels represents an advance, in terms of
structural safety and economy, when compared to the previous analytical models recommended in the literature.

1. Introduction two-step corbels. The monolithic corbel results provide inconsistent and
unsafe results, especially when the failure occurs by crushing of the
Two–step corbels are common in precast concrete construction, as concrete strut.
this solution improves the production of precast concrete columns used Therefore, Araújo et al. [2], using the analytical model suggested by
in the industry. Certain types of corbels can be used in the precast Campione [6], proposed an analytical model for two-step corbels with
concrete system, such as the column insert, which is a steel section steel fiber–reinforced concrete and a loop–shaped interface reinforce­
embedded into the precast column. However, this paper analyzes the ment. This reinforcement was used to control the opening of the joint at
traditional two-step corbel system. In step one, the column is cast the interface between the column and the corbel. Due to the small
without corbels. Fully anchored threaded couplers are placed in the database available for the testing of two-step corbels, the analytical
column mold to receive the main reinforcement of the corbel. In step model developed by Araújo et al. [2] had low precision and resulted in
two, the corbel reinforcement is inserted into the threaded couplers with very low strength reduction factors. Therefore, this analytical model is
stirrups, and then the corbel is cast. The interface between the column uneconomic for design applications.
and the corbel is the point of fragility of this solution because of the This paper aims to expand the experimental database of the two-step
concrete joint. Further details and a literature review about this corbel corbels tested by Araújo et al. [1]. Thereby, a methodology based on
type can be found in Araújo et al. [1]. computational modeling using the finite element method (FEM) is used.
Araújo et al. [2] analyzed short and very short two-step corbels, with Subsequently, an improved analytical model, based on the model rec­
an a/d ratio varying from 0.625 to 0.400. The authors concluded that, ommended by Araújo et al. [2], is presented and its accuracy is evaluated
through comparison of the ultimate load (Pu) and the yield strength (Py) from a database developed using the finite element model, validated by
with the design models in the Brazilian standard NBR 9062 [3], the PCI experimental results. Finally, strength reduction factors are proposed, to
design handbook [4] and the European standard EC2-1–1 [5], the design evaluate the design strength of the two-step corbels with the improved
models for monolithic corbels are inadequate to describe the failure of analytical model.

* Corresponding author.
E-mail addresses: ygor.neuberger@discente.ufg.br (Y.M. Neuberger), dlaraujo@ufg.br (D.L. Araújo).

https://doi.org/10.1016/j.engstruct.2023.115947
Received 16 October 2022; Received in revised form 8 February 2023; Accepted 3 March 2023
Available online 15 March 2023
0141-0296/© 2023 Elsevier Ltd. All rights reserved.
Y.M. Neuberger and D.L. Araújo Engineering Structures 284 (2023) 115947

2. Available experimental results between the main reinforcement and the concrete. Fig. 2 shows the
details of the 3D finite element model representative of the S3-L-1-B
The experimental results for two-step corbels, which are available in specimen, highlighting the finite elements used.
the literature, are detailed in the paper by Araújo et al. [1]. The authors
analyzed several types of treatments of the corbel–column interface; 3.2. Constitutive models
however, in this paper, only the two-step corbels with a 25 mm deep
recess forming a shear key at the interface were used. Therefore, of the 3.2.1. Concrete
twenty-six corbels tested by these authors, only eight specimens were Concrete without steel fibers was used in all columns and in some
used, whose properties are shown in Table 1. These corbels had two a/ two-step corbels. Table 2 shows the constitutive model and the input
d ratios (either 0.625 or 0.400) two fiber contents (0% or 1%) and the data of the concrete used in the DIANA FEA 10.2 software. All properties
loop–shaped reinforcement at the interface, as shown in Fig. 1. The steel were based on the compressive strength of concrete determined by
fibers had hooked ends, a length Lf of 35 mm, a diameter Df of 0.54 mm Araújo et al. [1].
and an aspect ratio of 65; they were only added to the concrete of the The concrete behavior was represented by the total strain crack
corbel. model; the uniaxial stress–strain diagram proposed by Lim and Ozbak­
The corbels were simply supported; that is, they had one fixed and kaloglu [8] was used. These authors suggested analytical equations that
one moving support, to avoid horizontal external loads; the width of the considered the influence of the confinement stresses, but this feature
support was reduced to 50 mm [1]. In Table 1, the yield strength of the was not used. Fig. 3 shows the stress–strain diagram of the concrete
corbel (Py,exp) is defined as the load sustained by the corbel at the time without steel fibers, or plain concrete, used in the modeling, with
when the main reinforcement reaches yield. It was determined in the compressive strengths of 30, 50 and 70 MPa. The confinement and
tests when the average strain recorded by strain gauges of tie rein­ lateral expansion effects were represented by Selby and Vecchio’s
forcement reached the yield strength of steel. The ultimate load of the constitutive model [9], available in the DIANA FEA 10.2 software for
corbel (Pu,exp) was defined as the load at which the corbel fails, that is, three-dimensional analysis with total strain crack models. Selby and
the maximum force recorded during the test. Vecchio’s constitutive model is based on an orthotropic nonlinear elastic
model that acknowledges the asymmetric response of concrete under
3. Three-dimensional finite element model multiaxial stress conditions. The formulation considers concrete
strength enhancement due to confinement, concrete strength degrada­
The corbels were modeled using the DIANA FEA 10.2 software [7]. tion due to transverse cracking, tension stiffening, and crack slip.
The 3D finite element model was developed to accurately represent the The uniaxial compressive stress–strain diagram of steel fiber-
test results, especially the corbel’s strength, the strain of the main reinforced concrete was represented by the analytical equations pro­
reinforcement and the joint opening at the corbel–column interface. posed by Ruiz et al. [10], as shown in Fig. 4. Ruiz et al. [10] recom­
mended analytical equations within the following limits: compressive
strength of steel fiber-reinforced concrete (fcf) between 29.4 MPa and
3.1. Finite element mesh 93.5 MPa; maximum diameter of the aggregate (dm) between 10 mm and
25 mm; steel fiber content (vf) between 0.24% and 3.00%; fiber length
The computational model was developed using cubic finite elements (Lf) between 10 mm and 80 mm; fiber diameter (Df) between 0.2 mm
with linear interpolation (HX24L and TP18L), which have three degrees and 1.2 mm; and an aspect ratio of the fibers between 20 and 107. The
of freedom of displacement per node. The element size was approxi­ equations that define the stress–strain diagram shown in Fig. 4 are
mately 25 mm in all models, which was determined from mesh refine­ presented in the supplementary material.
ment analysis based on the monotonic convergence of displacement The fracture energy of the steel fiber-reinforced concrete (GFf ) was
using a linear elastic model. The loading and support plates were also determined using Eq. (1), proposed by Barros and Figueiras [11], in
modeled using the HX24L finite element. which vf is the content of steel fibers as a percentage of the cementitious
The main reinforcement, the loop reinforcements and the threaded material and Gf is the fracture energy of the plain concrete obtained from
coupler system were represented by linear truss elements (L6TRU). The the fib Model Code 2010 [12]. The other properties of plain concrete and
longitudinal and transverse reinforcements of the column were repre­ steel fiber–reinforced concrete are shown in Table 2.
sented by embedded reinforcement, assuming a perfect bond between [ ]
the steel and concrete. GFf = 19.953 + 10.506vf Gf (1)
Two different interface elements were used in the model. The Q24IF
surface elements were used to represent the corbel–column interface
and the L12IF linear elements were used to represent the interface

Table 1
Properties of the two-step corbels tested by Araújo et al. [1].
Specimen a/d Volume of steel Main Concrete compressive Concrete compressive Yield strength of the Ultimate load of the
a,b,c
name fibers (vf) reinforcement strength of the column (fc, strength of the corbel (fc, corbel (Py,exp, kN) corbel (Pu,exp, kN)
MPa) MPa)

S3-L-0-A 0.625 0% 2ϕs 12.5 46.69 47.06 — 278.00


S3-L-0-B 0.625 0% 2ϕs 12.5 48.96 47.82 189.03 211.98
S3-L-1-A 0.625 1% 2ϕs 12.5 62.43 55.86 284.71 359.91
S3-L-1-B 0.625 1% 2ϕs 12.5 57.31 52.28 300.25 357.57
S4-L-0-A 0.400 0% 2ϕs 12.5 55.02 47.18 284.13 298.21
S4-L-0-B 0.400 0% 2ϕs 12.5 51.09 43.46 — 343.42
S4-L-1-A 0.400 1% 2ϕs 12.5 55.87 55.09 352.48 397.50
d d
S4-L-1-B 0.400 1% 2ϕs 12.5 52.65 56.34 440.00 440.00
a
Yield strength of the main reinforcement steel - fy = 541.11 MPa;
b
Ultimate strength of the main reinforcement steel - fu = 634.58 MPa;
c
Concrete cover - c = 25 mm;
d
The limit of the actuator was reached, without failure of the specimen.

2
Y.M. Neuberger and D.L. Araújo Engineering Structures 284 (2023) 115947

Fig. 1. Dimensions (mm) of the two-step corbel tested by Araújo et al. [1].

3.2.2. Steel However, the coefficient βt, which is constant and equal to 0.6 for plain
Steel reinforcement was not deemed to be a determinant for the concrete, and the tensile strength were changed for steel fiber-reinforced
failure of the specimen, such as the column reinforcements, the sleeve concrete. In this paper, Eq. (2) was used to evaluate the coefficient βt for
used in the splice of the main reinforcement and the support plates, was steel fiber-reinforced concrete, as was the case in Araújo et al. [2]. This
represented as a linear elastic material. The corbel reinforcements, such equation was proposed by Oliveira et al. [14], who tested reinforced
as the loop and main reinforcement, were represented by an elasto­ concrete ties with a steel fiber content range from 0.00 to 1.50% and
plastic constitutive model with isotropic hardening. Table 3 shows the concrete compressive strength range from 40 to 50 MPa. The strain εs is
input parameters used to define the constitutive model of the loop the reinforcement strain and RI is the fiber reinforcement index, given by
reinforcement and for the region of main reinforcement into the column. Eq. (3).
For these reinforcements, linear hardening was allowed after the yield
βt = 1 + (767.5RI − 742.7)εs , for 0.6⩽βt ⩽1.0 (2)
strength of steel.
The main reinforcement and the loop reinforcement influence the Lf
opening of the joint at the corbel–column interface, which is also RI = vf (3)
Df
influenced by the bond between the concretes of the corbel and the
column. The clearances present in the threaded coupler system used to The tensile strength of steel fiber-reinforced concrete fctm,f is calcu­
splice the main reinforcement (Fig. 5) also influence the opening of the lated by Eq. (4). In this equation, ρf is the rate of steel fibers in the
joint. These clearances, however, were not measured in the tests carried concrete, given by Eq. (5), which depends on the fiber orientation factor
out by Araújo et al. [1]. Therefore, its influence was determined by in­ [15,16]. The parameter nf is the ratio between the modulus of elasticity
verse analysis of the joint opening observed during the tests. The best fit of the fibers Esf and the modulus of elasticity of the concrete Ec, given by
of the corbel load–crack width curves were obtained by using a modulus Eq. (6).
of elasticity for the sleeve of 25 GPa, for corbels with a/d = 0.625. For (
1 + nρ + nf ρf
)
the corbels with a/d = 0.400, the best fit was obtained when using an fctm,f = fctm (4)
1 + nρ
elastic modulus of 80 GPa for the sleeve. Computational tests were
performed and it was observed that the yield strength and the ultimate ρf = 0.619 vf (5)
load of the corbel were not influenced by changing the value of the
modulus of elasticity of the sleeve; it only affects the opening of the joint Esf
of the corbel–column interface in the initial loading stage. nf = (6)
Ec
The region of the main reinforcement into the corbels was modeled
using a modified stress–strain diagram in tension, to consider the in­ The stress–strain curve of the reinforced concrete tie member with
fluence of the tension stiffening effect. Due to the use of truss elements in steel fiber-reinforced concrete is obtained by the same equations used
the finite element model (to represent the main reinforcement), the for the concrete without fibers, and is presented in the supplementary
bond between the steel rebar and the concrete was not represented. material. Fig. 8 shows the stress–strain curves of reinforced concrete tie
Therefore, the tension stiffening effect was considered to represent the members with different steel fiber contents and concrete with a
influence of the cracking of surrounding concrete on the main rein­ compressive strength of 40 MPa.
forcement. Fig. 6 shows the geometry of the tie member, which is
formed by the main bars with diameter ϕs and by the surrounding 3.2.3. Interfaces
concrete with concrete cover c. The expressions for the centrally rein­ The interface between the corbel and the column was represented
forced tie presented in the fib Model Code 2010 [12], as shown in Fig. 7, using a structural interface finite element, available in DIANA FEA.
were used for concrete without fibers. The expressions that define this Discrete cracking with brittle behavior was adopted to represent the
diagram are shown in the supplementary material. constitutive law of the structural interface, as presented in Table 4. The
For steel fiber-reinforced concrete corbels, the stress–strain curve for shear and normal stiffnesses of the interface were higher than the ratio
the reinforced concrete tie member is similar to that of plain concrete. of the highest value of the modulus of elasticity of concrete to the mean

3
Y.M. Neuberger and D.L. Araújo Engineering Structures 284 (2023) 115947

(a) Types of elements in the finite element model (b) Reinforcement of the finite element model

(c) Interface elements between column and corbel


(d) 3D finite element model

Fig. 2. Finite element mesh for specimen S3-L-1-B with two-step corbels tested by Araújo et al. [1].

length of the mesh to avoid penetration and sliding along the interface relationship, which represent two failure modes, i.e., pull-out of the bar
between corbel and column. The tensile strength used in the discrete or concrete splitting. For pull–out failure, Harajli [17] recommended
cracking model of the interface was obtained by Araújo et al. [2] using that the same bond stress-slip relationship be used for the plain concrete
inverse analysis of the crack width of the column–corbel interface of the as for the steel fiber-reinforced concrete. Therefore, the addition of steel
tested specimens. The closest proximity to experimental results was fibers has no influence on the failure by the pull-out of the bar. Harajli
reached by using about 10% of the tensile strength of concrete. The [17] proposed Eq. (7), with the limit values of stress and slip established
lower tensile strength is explained by the smooth interface between the in Eq. (8) and (9), for the ascending part of the diagram.
column and corbel. ( )0.3
Structural interface elements were also introduced to represent the τb = τbmáx
s
, if 0⩽s⩽s1 (7)
bond-slip between the concrete and main or loop reinforcement, since s1
they were modeled using truss elements. The local bond stress–slip
model for ribbed bars present in the fib Model Code 2010 [12] was used
to represent the adherence between the main reinforcement and the τbmáx = 2.57fc0.5 , if s1 ⩽s⩽s2 (8)
concrete without fibers in the column. Failure by pull–out in a good
bond condition was used in this case. The bond stress–slip relationship τbf = 0.35τbmáx , if s > s3 (9)
and the bond stresses between the reinforcing bar and plain concrete are
shown in Fig. 9a. The analytical equations are given in the supplemen­ For splitting failure in steel fiber-reinforced concrete, the peak value
tary material. of bond strength τbs, the post cracking stress τbps and the residual stress
The bond stress–slip model proposed by Harajli [17] was used to τbfs are given by Eq. (10), (11) and (12), respectively.
represent the bond stress between the main reinforcement and the steel √̅̅̅̅
(
c
)
fiber–reinforced concrete of the corbels. Fig. 9b presents the analytical τbs = 0.15 fc 3 + 3.5 ⩽ τbmáx (10)
ϕs

4
Y.M. Neuberger and D.L. Araújo Engineering Structures 284 (2023) 115947

Table 2
Constitutive model of concrete used in the finite element model.
Linear Material Properties
Modulus of elasticity (Ec) a ( )1/3
21500 fcm 10
/

Poisson’s ratio (ν) 0.2


Nonlinear Material Properties
Total strain crack model Fixed crack
Nonlinear tension softening Hordijk [13] b
Axial tensile strength of plain concrete and fct = 0.3f2/3
ck → fck ≤ 50 MPa
SFRC (fct) c fct = 2.12 ln (1 + 0.1fc) → fck > 50
MPa
Fracture energy Plain concrete: Gf = 73 fcm0.18N/m a
SFRC (GFf ): Eq. (1)
Crack bandwidth specification Rots

Residual tensile reduction Null e


Reduction Model No Reduction e
Compressive behavior diagram Multilinear:
For plain concrete: Lim and
Ozbakkaloglu [8]
For SFRC: Ruiz et al. [10] Fig. 4. Dimensionless compressive stress–strain diagram for steel fiber-
Cylinder compressive strength d fcm reinforced concrete [10].
Lateral cracking No Reduction
Lateral confinement Selby and Vecchio [9]
Shear retention of plain concrete and SFRC Constant e
Shear retention factor (β) 0.01 e Table 3
Constitutive model of the steel used in the finite element model.
a
Recommendation of the fib Model Code 2010 [12];
b Constitutive model von Mises plasticity a
Model available in the DIANA FEA 10.2 and defined in Hordijk [13];
c Linear Material Properties
Recommendation of the fib Model Code 2010, with fck = fcm − 8 MPa; Modulus of elasticity (Es) 200 GPa b
d
Data obtained from tests carried out by Araújo et al. [1] Modulus of elasticity of sleeve c a/d = 0.625: Es = 25 GPa
e
Default of the DIANA FEA 10.2. a/d = 0.400: Es = 80 GPa
Poisson’s ratio (ν) 0.3
Nonlinear Material Properties
Ultimate strain εu = 0.18 d
Plastic hardening
Strain-Stress diagram e Naked steel bar showed in Fig. 8
Hardening hypothesis Strain hardening
Hardening type Isotropic hardening
a
Model available in the DIANA FEA 10.2;
b
Theoretical value adopted for all steels, except the sleeve;
c
Value obtained from the inverse analysis of the test results reported in Araújo
et al. [1] for each a/d ratio;
d
Value adopted;
e
Yield strength and ultimate strength of the reinforcing steel used in the
modeling are shown in Table 1.

failure mode was concrete splitting for part of the main reinforcement
into the corbels.
Some loop reinforcements in test specimens were not completely
linear before the corbel was cast, and the length inside the corbel,
measured from the face of the column, was only 90 mm (corresponding
to 14 times the bar diameter, as explained by Araújo et al. [1]). There­
Fig. 3. Compressive stress–strain diagram for plain concrete [8]. fore, the bond stress between the loop reinforcement and concrete was
assumed to be close to zero to release slip in this short embedment
( ) length. The end of the loop reinforcement was constrained inside the
√̅̅̅̅ vf Lf c
τbps = 0.6 fc (11) corbel to represent its mechanical anchorage to concrete. This procedure
Df ϕs was adopted to correctly represent the crack width in the corbel–column
interface, given that tested specimens showed crack width even for low
τbfs = 0.3τbps (12) applied load values. The supplementary material shows the load–strain
Table 5 presents the constitutive model for the interfaces between relationship of loop reinforcements, demonstrating the efficiency of this
concrete and main or loop reinforcement. The same stiffness values were procedure in correctly representing the strains observed in tests. The
used in the linear properties of all interface elements. The normal threaded coupler system used for the splicing of the main reinforcement
stiffness of the interface was chosen to prevent penetration of the truss was considered to be without adherence to the concrete, given that its
elements in the concrete. Reduced values of shear stiffness were adopted external surface was smooth.
so that the tangential behavior of the interface was governed by the
bond stress–slip relationships. In part of the main reinforcement into the 3.3. Loading and analysis
column, the relationship presented on Fig. 9a was used. In part of the
main reinforcement into the corbel, Eqs. (7) and (8) were used for cor­ The finite element analysis was conducted by applying displacement
bels with plain concrete and Eq. (7) to (12) were used for corbels with control to the upper face of the column (Fig. 2d). The total imposed
steel fiber-reinforced concrete. Therefore, it was assumed that the displacement was 5 mm, which was reached by applying 100 loading

5
Y.M. Neuberger and D.L. Araújo Engineering Structures 284 (2023) 115947

Fig. 5. Threaded coupler system for the main reinforcement in the two-step corbel specimens tested by Araújo et al. [1].

Fig. 6. Geometry of the reinforced concrete tie.

Fig. 8. Stress–strain diagram of the naked steel bar and centrally reinforced tie
with steel fibers and concrete, with a compressive strength of 40 MPa.

Table 4
Constitutive model for interface between the corbel and the column used in
the finite element model.
Class Interface elements
Constitutive model Discrete Cracking a
Linear Material Properties
a
Type 3D Surface Interface
Normal stiffness modulus-z 10000 N/mm3
Shear stiffness modulus-x 10000 N/mm3 b
Shear stiffness modulus-y 10000 N/mm3 b
Nonlinear Material Properties
Discrete Cracking
Tensile strength 0.7515 N/mm2 c
Tensile behavior Brittle a
a
Shear behavior Zero shear traction
a
Model available in DIANA FEA 10.2;
b
Fig. 7. Simplified stress–strain relation for a centrally reinforced tie according Due to the shear key, the shear stiffness was provided for the inclined
to fib Model Code 2010 [12]. region of the interface. In the vertical region, the shear stiffness is zero (see
Fig. 2c);
c
Adopted value as 0.1fct.
increments of 0.05 mm each. A tolerance of 0.1% was used for the en­
ergy norm convergence criterion.
paper can accurately represent the results of the tests carried out by
4. Validation of the finite element model Araújo et al. [1]. Moreover, the finite element model is more accurate
than the analytical model suggested by Araújo et al. [2] for predicting
The yield strength and the ultimate load of the two-step corbels, as the ultimate load of two-step corbels.
obtained from the experimental analysis, are compared with the yield The relationship between load and strain in reinforcements, the
strength and the ultimate load obtained from the finite element width of cracks at the corbel-column interface, and the cracking pattern
modeling in Table 6. In the finite element model, the main reinforce­ used for validation of the finite element model are presented in the
ment only reaches the yield strength of the steel before the ultimate load supplementary material.
in the corbels with steel fibers and an a/d ratio of 0.625 (specimens S3-L-
1-A and S3-L-1-B). The average value of the ratio Py,comp/Py,exp in this 5. Parametric analysis
case was 1.15. The ratio Pu,comp/Pu,exp for all two-step corbels was 1.13,
with a standard deviation of 0.15 and a CoV of 13.6%. When the A parametric analysis was performed, using the finite element model
analytical model proposed by Araújo et al. [2] was applied to these same developed, to assemble two databases of two-step corbels. The first
two-step corbels, the resulting Pu,ana/Pu,exp ratio had a mean value of database had two-step corbels with four a/d ratios, three rates for the
0.85, a standard deviation of 0.22, and a coefficient of variation of 27%. main reinforcement, three concrete compressive strength grades, and
Therefore, it is concluded that the finite element model developed in this two volumes of steel fibers. The second database was assembled using

6
Y.M. Neuberger and D.L. Araújo Engineering Structures 284 (2023) 115947

(b) Steel fiber-reinforced concrete; adapted from [17]


(a) Plain concrete; adapted from [12]
Fig. 9. Bond stress–slip relationship between the main reinforcement and the concrete.

similar to that of the specimens tested by Araújo et al. [1], as shown in


Table 5 Fig. 1. The only difference is the length of the corbels, which are 350
Constitutive model for interfaces between concrete and main or loop rein­
mm. The rate of the main reinforcement was varied, to achieve two-step
forcement used in the finite element model.
corbels with failure by crushing of the concrete strut and without the
Class Interface elements yield of the main reinforcement or two-step corbels with failure by the
Constitutive model Bond slip a
yield of the main reinforcement. The concrete compressive strength of
Linear Material Properties
Type 3D Line Interface (2 normal, 1 shear) a the two-step corbel was varied within the range defined by the analytical
Normal stiffness 1550 N/mm3 b equations of Ruiz et al. [10]. The concrete compressive strength of the
modulus reinforced concrete column, without fibers, was always the same as the
Shear stiffness 15.5 N/mm3 c
two-step corbel. The amount of steel fibers used in the corbel was in the
modulus
Nonlinear Material Properties range that Oliveira et al. [14] used to make the tension–stiffening model
Bond-slip Model Multilinear for steel fiber-reinforced concrete.
Bond stress–slip Main reinforcement into the column: pull-out failure using Another modification, relating to the specimens tested by Araújo
relationship the relationship in Fig. 9a et al. [1], was the use of main reinforcement into the column with a
Main reinforcement into the corbel: splitting failure using
constant diameter of 20 mm. This was done to avoid the yield of the
the relationship in Fig. 9b
Loop reinforcement: bond stress (τbmax) of 0.01 MPa was main reinforcement in the column region when its diameter was
used to release the slip reduced. The modulus of elasticity of the sleeve was obtained by linear
a interpolation of the values shown in Table 3, using the expression E = −
Model available in DIANA FEA 10.2;
b
Approximate value for the ratio of the modulus of elasticity of the concrete 242.5 (a/d) + 177 (GPa). The lower limit of 25 GPa, obtained from the
by the size of the finite element; inverse analysis of the tests by Araújo et al. [1], was assumed in the
c
Adopted value as 1% of normal stiffness. parametric analysis.
The mesh dimension used in the parametric analysis was approxi­
mately 25 mm as was used in the validation analysis. The constitutive
Table 6 model of the steel (reinforcement and support plates) is also the same, as
Yield strength and ultimate load of two-step corbels obtained from the testing by shown in Table 3. The constitutive models of the concrete, the cor­
Araújo et al. [1] and the finite element model. bel–column interface and the main reinforcement–concrete interface
Specimen Py,exp (kN) Pu,exp (kN) Py,comp (kN) Pu,comp (kN) were adjusted for each specimen, depending on the compressive
M3-L-0-A -a 278.21 -a 264.26 strength of the concrete and the volume of steel fibers. The constitutive
M3-L-0-B 189.03 211.98 -a 266.34 model of the concrete was defined using the parameters shown in
M3-L-1-A 284.71 359.91 338.59 359.16 Table 2. The constitutive model of the interface between the column and
M3-L-1-B 300.25 357.57 334.99 352.70
the corbels and between the main reinforcement and the concrete was
M4-L-0-A 284.13 298.13 -a 392.45
M4-L-0-B -a 347.85 -a 388.83 defined by the parameters shown in Table 4 and Table 5, respectively.
M4-L-1-A 352.48 397.50 -a 506.88 The relationship between the yield strength obtained from the
M4-L-1-B > 440b > 440b -a 497.89 computational modeling (Py,comp) of the two databases and the analytical
a
The main reinforcement did not reach the yield strength of the steel before model (Py,ana) proposed by Araújo et al. [2] is shown in Fig. 11a. The
the ultimate load; individual results of each two-step corbel are shown in the supplemen­
b
It reached the limit of the actuator before the ultimate load was reached. tary material. The mean value of the ratio Py,ana/Py,comp for the first
database with 54 corbels, in which the main reinforcement reached the
three other a/d ratios and three other compressive strengths of the yield strength of the steel, was 0.81, with a standard deviation of 7% and
concrete, maintaining the values of the rate of main reinforcement and a coefficient of variation (CoV) of 8%. The results were similar, with a
the volume of steel fibers in the corbels. Fig. 10 shows the variables of mean of 0.81 and a CoV of 8.3%, using 41 two-step corbels in the second
the first and second databases, which are made up of 72 corbels and 54 database, where the main reinforcement reached the yield strength of
corbels, respectively. the steel. This result indicates that the analytical model has low
The geometry of the two-step corbels in the parametric analysis is dispersion when evaluating the yield strength of the two-step corbels,

7
Y.M. Neuberger and D.L. Araújo Engineering Structures 284 (2023) 115947

Fig. 10. Parametric analysis of the two-step corbels.

(a) Failure by yield strength of the main (b) Failure by crushing of the concrete strut
reinforcement
Fig. 11. Comparison of the finite element model with the analytical model recommended by Araújo et al. [2].

although the analytical yield strength is, on average, 20% lower than the results were similar for the second database, with 54 corbels, a mean
yield strength obtained from the finite element model. A similar result ratio of 1.18 and CoV of 30%. The high dispersion indicates that the
was obtained by Araújo et al. [2] when they validated their analytical analytical model suggested by Araújo et al. [2] does not accurately
model with the testing of two-step corbels. predict the ultimate load of the two-step corbels when failure is in the
The relationship between the ultimate load obtained from the form of crushing of the concrete strut.
computational modeling (Pu,comp) and the analytical model (Pu,ana) Based on the crack width at the column-corbel interface under ser­
proposed by Araújo et al. [2] is shown in Fig. 11b. All 72 corbels of the vice load, defined as the lowest value between 50% of the yield strength
first database were used and the mean value of the ratio Pu,ana/Pu,comp and 50% of the ultimate load of the corbel, it is concluded that it exceeds
was 1.18, with a standard deviation of 31.8% and a CoV of 26%. The 0.5 mm from the a/d ratio of 0.625. For these corbels, it is recommended

8
Y.M. Neuberger and D.L. Araújo Engineering Structures 284 (2023) 115947

that the diameter of the loop reinforcement is increased to reduce the


crack width in the corbel–column interface.

6. Proposed analytical model

The analytical model for two-step molded corbels, recommended by


Araújo et al. [2], was modified to better represent the results obtained
from the parametric analysis, especially for failure by crushing of the
concrete strut. The equilibrium and compatibility conditions for the
cross-section located on the corbel–column interface is shown in Fig. 12.
The analytical model depends, among other variables, on the depth
of the neutral axis (xc). The results of the finite element model showed
that the neutral axis depth varies very little between the yield strength
and the ultimate load at the crushing of the concrete strut. This confirms
the hypothesis adopted by Araújo et al. [2] that there is no significant
variation in the neutral axis depth after the yielding of the main rein­
forcement. Therefore, the internal arm (z) between the tensile force in
the main reinforcement and the compression force in the concrete can be
calculated by Eq. (13), both to the yield strength and to the ultimate load
of the two-step corbel.
z = d − 0.3xc (13)
The depth of the neutral axis (xc) is obtained directly from the ge­
ometry of the corbel and the deep recess, as shown in Fig. 13. Line 1
Fig. 13. Estimated depth of the neutral axis from the geometry of the deep
develops from the center of the bearing plate to the deep recess, whereas recess in a two-step corbel.
line 2 is on the external face of the column, next to the corbel. Eqs. (14)
and (15) provide the depth of the neutral axis from the geometry, where
L is the depth of the deep recess, adopted as 25 mm in all analyzed ϕs
(16)

z1 = d − e − d −
corbels, and h is the height of the corbel, next to the column. 2
L
(14)
fct
xc = L + + 0.003
tanθ e = xc Ect (17)
0.003

tan θ =
a+L
(15) The tensile strength of the steel fiber-reinforced concrete influences
h− L the flexural strength of the corbel–column interface section. Araújo et al.
The fiber contribution to the strength of the corbel in the tensile zone [2] proposed using Eq. (18), obtained from Oliveira et al. [14], for the
z1 is obtained by Eq. (16). It depends on the distance between the most post-cracking tension stiffening of fibrous concrete (fr,ts), to calculate the
compressed fiber of the transverse cross-section and the fiber in which force of the equivalent reinforced concrete tie (Fig. 6). In the rest of the
the maximum tensile strength in concrete is reached (e), which is ob­ tensioned cross-section, below the reinforced concrete tie, Araújo et al.
tained by Eq. (17). In these expressions, d’ is the distance between the [2] and Campione [6] suggested using the residual strength of the steel
main reinforcement and the most stressed fiber in tension, ϕs is the fiber-reinforced concrete (fr) evaluated by Eq. (19). In these expressions,
nominal diameter of the main reinforcing bar reinforcement, fct is the εy is the tensile strain at which the yield strength of the reinforcing steel
tensile strength of the concrete of the corbels, and Ect is the modulus of is reached and λ is a shape factor for the steel fiber, assumed to be 1.0 for
elasticity of concrete in tension, adopted as half of the modulus of hooked end fibers. Eq. (18) was obtained for steel fibers with an aspect
elasticity in compression. ratio (Lf/Df) between 65 and 80, and a volume (vf) less than 1.5%.

Fig. 12. Truss model and analysis of the corbel–column interface section.

9
Y.M. Neuberger and D.L. Araújo Engineering Structures 284 (2023) 115947

( ( ) )
Lf the number of independent variables, to a number of points. The sum­
fr,ts = 1.0 + 767.5 vf − 742.7 εy fct (18)
Df mary statistics of the multiple linear regression are shown in Table 7.
Table 8 depicts the standard error and confidence band for the co­
√̅̅̅̅ Lf efficients of Eq. (25).
fr = 0.2 fc vf λ (19)
Df a
ξu = 2.052 − 1.554 + 0.289ρfy − 0.262vf − 0.01fc (25)
The cross-section of the reinforced concrete tie (Aeff
c ) was determined
d
by Araújo et al. [2], as shown in Fig. 6. In this paper, the cross-section of Fig. 14 compares the proposed coefficient ξu with the equations
the reinforced concrete tie was changed and obtained by using Eq. (20). proposed by Russo et al. [18] and Campione [6], which are used in the
This change is justified because the proposal by Araújo et al. [2] does not monolithic corbels without steel fibers. A greater range of values for the
consider the contribution of the concrete region between the equivalent coefficient ξu is observed in Eq. (25), when compared to other equations
ties, when evaluating the tensile force on the cross-section (Fig. 12). in the literature. In addition, the coefficient ξu from Eq. (25) can assume
values greater than 1.0 for corbels with a/d < 0.5 and fc < 60 MPa,
Aeff
c = b(2c + ϕs ) (20)
which represents the confinement of the concrete node at the base of the
Using the internal arm, from Eq. (13), the angle between the strut corbels.
and the main reinforcement (α) is determined by Eq. (21). The area of The proposed analytical model was applied to the two-step corbels of
the compression strut is determined by Eq. (22). the second database developed herein. The main reinforcement reached
(z) the yield strain of the steel in the 41 corbels from the 54 specimens of the
α = arctan (21) database. The relationship between the yield strength obtained from the
a
finite element model (Py,comp) and the yield strength predicted by the
Aeff = b xc cos α (22) proposed analytical model (Py,ana) is shown in Fig. 15a. The mean value
of the Py,ana/Py,comp ratio was 0.92, with a standard deviation of 8.5%
The yield strength of the corbel (Py) is obtained from the truss
and CoV of 9.2%. Therefore, the proposed analytical model in this paper
equilibrium, as shown in Fig. 12, resulting in Eq. (23). The contribution
is more accurate when predicting the yield strength of two-step corbels
of the steel fibers at the depth z1 is neglected, to obtain the yield strength
than the model proposed by Araújo et al. [2], with a similar coefficient of
Py, which agrees with Araújo et al. [2]. From the equilibrium of the
variation.
cross–section, the ultimate load due to the crushing of concrete strut
The relationship between the ultimate load obtained from the finite
force (Pu), is determined by Eq. (24); this was also proposed by Cam­
element model (Pu,comp) and the ultimate load predicted by the proposed
pione [6].
analytical model (Pu,ana) is shown in Fig. 15b. The mean value of the Pu,
( )
Py = Aeffc fr,ts + fy As tanα (23) ana/Pu,comp ratio was 0.97, with a standard deviation of 6.6% and CoV of
6.8% for the improved analytical model. The analytical model proposed
1 2 by Araújo et al. [2] obtained a ratio of 1.18, with a standard deviation of
Pu = ξu fc b xc sen 2α + fr b z1 (24)
2 sen 2α 35.3% and CoV of 29.9%. Therefore, the analytical model proposed in
this paper is more accurate for evaluating the ultimate load of two-step
The coefficient ξu presented in Eq. (24) is the efficiency factor
corbels with crushing failure of concrete strut than the model proposed
adopted in the strut-and-tie models, to consider the biaxial state of the
by Araújo et al. [2], providing a significantly lower coefficient of
stresses (compression–tension) in the concrete strut. Therefore, the
variation.
compressive strength of the concrete in the strut may be lower than the
compressive strength of the concrete obtained from the cylinder tests, fc.
7. Evaluation of the uncertainty of the analytical model
Some researchers have suggested empirical expressions for this term,
based on the concrete compressive strength [18] or the geometry of the
The analysis of the resistance model’s uncertainty partial safety
corbel [6]. In this paper, an equation (Eq. (25)) is proposed to explain
factor (γRd) was carried out using the 54 two-step corbels of the second
the relationship of the coefficient ξu as a function of corbel geometry (a/
database obtained from the finite element model. For this purpose, the
d), reinforcement rate of the main reinforcement (ρfy, MPa), steel fiber
principles presented in the Araújo et al. [2], fib Model Code 2010 [12]
volume (vf, %), and concrete compressive strength (fc, MPa). Eq. (25)
and Castaldo et al. [19] were used. For this analysis, the model error
was obtained through multiple linear regression using IBM SPSS®
(ME) was defined by Eq. (26), in which Panalytical is the resistance
software applied to the 72 two-step corbels of the first database. The
calculated by the proposed analytical model and Pcomputational is the
purpose of multiple regression is to analyze the relationship between
resistance obtained from the finite element model.
several independent or predictor variables and a dependent or criterion
variable. The computational problem that needs to be solved in multiple Panalytical
ME = (26)
regression analysis is to fit a plane in an n-dimensional space, where n is Pcomputacional

Fig. 16 presents the cumulative distribution functions plotted on


normal probability paper for the ME. For this analysis, 41 two-step
Table 7
Summary statistics of multiple linear regression for coef­ corbels were used to evaluate the yield strength (Fig. 16a) and 54
ficient ξu. two-step corbels were used to evaluate the ultimate strength (Fig. 16b).
Parameter Value

R 0.949 Table 8
R2 0.901
Standard error and confidence band for coefficients of Eq. (25).
Adjusted R2 0.895
Standard error 0.113 Non-standardized coefficients 95% Confidence band
Durbin –Watson statistic 1.305
Parameter Value Standard Error Lower limit Upper limit
Residual Statistics
Minimum − 0.36 Constant 2.052 0.121 1.809 2.294
Maximum 0.24 a/d − 1.554 0.086 − 1.725 − 1.383
Mean 0 vf − 0.262 0.053 − 0.368 − 0.156
Standard Error 0.11 ρff − 0.289 0.019 0.251 0.328
Number of points 72 fc − 0.01 0.002 − 0.013 − 0.006

10
Y.M. Neuberger and D.L. Araújo Engineering Structures 284 (2023) 115947

(a) Proposed coefficient and coefficient Russo et al. (b) Proposed coefficient and coefficient of Campione
[18] [6]
Fig. 14. Comparison of the proposed coefficient ξu with those reported by Russo et al. [18] and Campione [6].

(a) Yield strength according to the proposed (b) Ultimate load according to the proposed
analytical model analytical model

(c) Yield strength according to the model of Araújo (d) Ultimate load according to the model of Araújo et
et al. [2] al. [2]
Fig. 15. Comparison of the proposed analytical model and that of Araújo et al. [2] using data from the second database obtained from the finite element model.

An approximately linear distribution was observed in both figures, with


1
R2 = 0.992 for yield strength and R2 = 0.966 for ultimate load, indi­ CoVm = (27)
b μR
cating a log-normal distribution. The Kolmogorov-Smirnov test was
performed and the hypothesis that ln(ME) has a normal distribution for
μR = exp( − a/b) (28)
both cases can be accepted with a probability p > 0.95.
The coefficient of variation of the model error function, CoVm, can be Eq. (27) only estimates the error in the analytical model, to predict
estimated using Eq. (27). In this equation, μR is the bias factor of the the resistance of the finite element model. However, the finite element
function ln(ME), which can be obtained from Eq. (28), using the co­ model also presents an error to predict the resistance of the physical
efficients a and b from the linear regression shown in Fig. 16. model. Therefore, the total variability of the model error CoVME is

11
Y.M. Neuberger and D.L. Araújo Engineering Structures 284 (2023) 115947

(a) Yield strength (b) Ultimate load


Fig. 16. Normal probability plot showing model error (ME) for resistance of two-step corbels.

determined by Eq. (29). In this equation, CoVm is the coefficient of


Pm
variation between the finite element model and predicted strengths in Pd = (31)
γR γ Rd
the proposed analytical model; CoVcomp is the coefficient of variation
between the resistance obtained from testing and from the finite element
γ R = exp (αR β CoVR ) (32)
model. A CoVcomp of 0.136 was adopted, which represents the greatest
variability observed by comparing the ultimate load of the finite element Assuming that the probability for the characteristic value is 5%, the
model and of the tests performed by Araújo et al. [1], as shown in coefficient of variation can be expressed by Eq. (33), according to the fib
Table 6. Model Code 2010 [12].
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ ( )
1 Pm
CoVME = CoV 2m + CoV 2comp (29) CoVR = ln (33)
1.65 Pk
The resistance model’s uncertainty partial safety factor (γ Rd) is In Eq. (33), Pk is the characteristic strength of the two-step corbel,
defined by Eq. (30), as presented in Castaldo et al. [19]. In this equation, which was obtained from corbels of the second database, assuming that:
μR is the mean value of the ratio between the resistance achieved i) the characteristic yield strength of the main reinforcement is fyk = 500
through the proposed analytical model and the resistance obtained MPa; ii) the characteristic compressive strength of concrete (fck) was
through the finite element model (herein it is the bias factor reported in obtained by Eq. (34), where fc is the compressive strength of concrete
Table 9). used in the database, and Sd is the standard deviation of the concrete
γ Rd = μR exp(αR β CoVME ) (30) production, adopted as 3.5 MPa for prefabricated structures [3]; iii) the
characteristic resistance of the residual strength of the steel fiber-
The first order-reliability method (FORM) sensitivity factor for the reinforced concrete is not known, so frk = frm/1.5 was adopted, which
variables (αR) in Eq. (30) can be assumed to be 0.32 or 0.80 [19], ac­ corresponds to an estimated coefficient of variation of 0.20. With these
counting for the hypothesis of non-dominant and dominant variables, hypotheses, the average value of Pm/Pk for the corbels in the second
respectively. The non-dominant hypothesis can be adopted if the coef­ database is 1.17 and 1.26 for the yield strength and for the ultimate load,
ficient of variation (CoVME) reported in Table 9 is less than the coeffi­ respectively.
cient of variation associated with the compressive strength of concrete,
which can be set as 0.15 [12,19]. Otherwise, the dominant hypothesis fck = fc − 1.645 Sd (34)
can be adopted. Finally, the term β denotes the reliability index, which is The global safety factor (γR) and the resistance model’s uncertainty
adopted herein as β = 3.8, i.e., a failure probability of 7x10-5 [12]. partial safety factor (γRd) can be adjusted in the form of the strength
The safety of structures can be evaluated by the global resistance reduction factor (ϕ), as presented in Eq. (35). In this equation, values of
format (γR) [12]. It is important to recognize that, in this formulation, αR = 0.8 and β = 3.8 can be adopted according to fib Model Code 2010
the global safety factor is related to the mean variable. Uncertainty due [12].
to the model formulation must be treated by a separate safety factor for
model uncertainty (γRd). Therefore, the design resistance (Pd) takes the ϕ = exp(− αR β CoVR )
1
(35)
form shown in Eq. (31). The global safety factor (γ R) can be determined γ Rd
from the method of estimation of a coefficient of variation of resistance, Table 9 shows the summary statistics for the model error function ln
by using Eq. (32). In this expression, the values of αR = 0.8 and β = 3.8 (ME), and Table 10 shows the values of the strength reduction factor (ϕ),
can be admitted [12]. calculated by the proposed analytical model. Adopting the non-
dominant hypothesis, the strength reduction factor is 0.66 for predict­
ing the yield strength and 0.56 for predicting the ultimate strength of the
Table 9
two-step corbels. The non-dominant hypothesis is more adequate in this
Statistical parameters for model errors in the proposed analytical model.
analysis, since CoVME is close to 0.15 (Table 9). As a comparison, Araújo
Estimated resistance Second database Statistical parameters et al. [2] obtained a strength reduction factor of 0.57 and 0.34, for
Mean (μ) CoV Bias factor (μR) CoVm CoVME predicting the yield strength and the ultimate strength, respectively.
Yield strength (Py) 0.92 0.093 0.92 0.10 0.17 These lower values are due to the dominant hypothesis adopted by these
Ultimate load (Pu) 0.97 0.069 0.97 0.07 0.15 authors, since their analytical model was less accurate and had a greater

12
Y.M. Neuberger and D.L. Araújo Engineering Structures 284 (2023) 115947

Table 10 d) Adopting a log-normal probabilistic distribution to represent the


Uncertainty and strength reduction factor for the proposed analytical model. resistance model uncertainties, the statistical parameters of the new
Estimated FORM factor β Pm/ CoVR γRd γR ϕ recommended analytical model for the bias factor and the CoVME are
Resistance αR Pk equal to 0.92 and 0.17 (to estimate the yield strength) and 0.97 and 0.15
Yield Strength Non- 3.8 1.17 0.095 1.13 1.33 0.66 (to estimate the ultimate load). When model uncertainties are not
(Py) dominant dominant,and assuming a reliability index of 3.8, strength reduction
0.32 factors equal to 0.66 and 0.56 are suggested as safe assumptions to
Dominant 0.8 3.8 1.17 0.095 1.54 1.33 0.49 calculate the design resistance of the yield strength and ultimate load of
Ultimate load Non- 3.8 1.26 0.141 1.17 1.53 0.56
(Pu) dominant
two-step corbels. Therefore, the analytical model proposed in this paper
0.32 represents an improvement over the analytical model proposed by
Dominant 0.8 3.8 1.26 0.141 1.54 1.53 0.42 Araújo et al. [2], for the design of two-step corbels without secondary
reinforcement and with steel fiber-reinforced concrete.

value for the coefficient of variation CoVME. Therefore, it was concluded


that the analytical model proposed herein is more accurate and adequate CRediT authorship contribution statement
to be used for the design of two-step corbels, than the one proposed by
Araújo et al. [2]. Ygor Moriel Neuberger: Methodology, Formal analysis, Investiga­
The strength reduction factor obtained for the proposed analytical tion. Daniel de Lima Araújo: Supervision, Project administration,
model in this paper is lower than the strength reduction factor recom­ Funding acquisition, Methodology, Formal analysis.
mended by ACI 318–19 [20], for the design of reinforced concrete
monolithic corbels, i.e., ϕ = 0.75. This reduction is due to the greater
value obtained for the coefficient of variation CoVR, which resulted in a Declaration of Competing Interest
higher value for the global safety factor (γR). As CoVR depends on the
ratio Pm/Pk, a strength reduction factor close to 0.75 was obtained when The authors declare that they have no known competing financial
CoVR is lower than 0.06. In the two-step corbels, the CoVR is higher than interests or personal relationships that could have appeared to influence
this value due to more variability of the residual strength of the steel the work reported in this paper.
fiber-reinforced concrete that contributed to the resistance of the corbels
in the ultimate limit state. Data availability

8. Conclusion Data will be made available on request.

Analytical models available in the literature are scarce and do not Acknowledgements
provide accurate estimates of the design strength of two-step corbels,
especially for failure by crushing of concrete strut. Araújo et al. [2] The authors would like to thank the FAPEG (The Foundation for
recommended an analytical model for these corbels, but only a few Research Support of the State of Goiás, grant ID 2017.10.267000.513)
samples and variables were tested. Therefore, the strength reduction and CNPq (The National Council for Scientific and Technological
factors recommended by these authors are too conservative. This paper Development) for their support and funding of this research project.
developed a three-dimensional finite element model of a two-step corbel Their support was instrumental in providing the necessary resources for
that considers the nonlinearity of the materials, the slip between the this research.
main reinforcement and the concrete and the effect of tension stiffening
of the steel fiber-reinforced concrete. References
Based on the results reported above, the following conclusions can be
presented: [1] Araújo D de L, Oliveira EM de, Silva EMO, Coelho SA, El Debs MK. Experimental
analysis of a modified two-step corbel for precast concrete system. Eng Struct 2021;
a) The three-dimensional finite element model showed good corre­ 242:112585. https://doi.org/10.1016/j.engstruct.2021.112585.
lation with the results of the two-step corbels tested by Araújo et al. [1], [2] Araújo D de L, Coelho SA, Almeida SRM de, El Debs MK. Computational modelling
both for ultimate load due to the crushing of concrete strut and for crack and analytical model for two-step corbel for precast concrete system. Eng Struct
2021;244:112699. https://doi.org/10.1016/j.engstruct.2021.112699.
width in the interface between corbel and column. The ultimate load
[3] ABNT-Associação Brasileira de Normas Técnicas. Design and execution of precast
obtained from the three-dimensional finite element model differs less concrete structures. ABNT NBR 9062 (in Portuguese). Rio de Janeiro: ABNT; 2017.
than 13% from the test results, with a coefficient of variation of 13.6%. [4] PCI. PCI Design Handbook. 7th ed. Chicago: PCI – Precast/Prestressed Concrete
b) Using the finite element model, a database was developed with 72 Institute; 2010.
[5] CEN – Comité Européen de Normalisation. Eurocode 2: Design of concrete
corbels. The variables of this database are a/d ratio (0.35 ≤ a/d ≤ 0.77), structures - Part 1-1 : General rules and rules for buildings. EN 1992-1-1. Brussels:
main reinforcement rate (0.2% ≤ ρ ≤ 0.5%), compressive strength of the 2004.
concrete (40 MPa ≤ fc ≤ 60 MPa) and content of steel fibers in the corbel [6] Campione G. Flexural response of FRC corbels. Cem Concr Compos 2009;31:
204–10. https://doi.org/10.1016/j.cemconcomp.2009.01.006.
(1% and 1.5%). From this database, the analytical model proposed by [7] TNO Building and Construction Research. DIANA User’s Manual – Release 10.2
Araújo et al. [2] was modified and a new efficiency factor for considering 2017.
the biaxial state of stresses (compression–tension) in the concrete strut [8] Lim JC, Ozbakkaloglu T. Stress-strain model for normal and light-weight concretes
under uniaxial and triaxial compression. Constr Build Mater 2014;71:492–509.
was proposed. https://doi.org/10.1016/j.conbuildmat.2014.08.050.
c) The new recommended analytical model was applied to a second [9] Selby RG, Vecchio FJ. A constitutive model for analysis of reinforced concrete
database with 54 corbels, developed using the finite element model. The solids. Can J Civ Eng 1997;24:460–70. https://doi.org/10.1139/l96-135.
[10] Ruiz G, de la Rosa Á, Wolf S, Poveda E. Model for the compressive stress–strain
variables of this database are the a/d ratio (0.42 ≤ a/d ≤ 0.69) and the relationship of steel fiber-reinforced concrete for non-linear structural analysis.
compressive strength of the concrete (30 MPa ≤ fc ≤ 70 MPa). The other Hormigón y Acero 2018;69:75–80. https://doi.org/10.1016/j.hya.2018.10.001.
variables were the same as those used in the first database. The new [11] Barros J.A.O., Figueiras J.A. Flexural Behavior of SFRC: Testing and Modeling. J
Mater Civ Eng 1999;11:331–339. https://doi.org/10.1061/(ASCE)0899-1561
recommended analytical model showed a good correlation with the
(1999)11:4(331).
yield strength (a difference of less than 8%, with a coefficient of varia­ [12] fib Model Code for Concrete Structures 2010 (MC-10). Berlin: Wilhelm Ernst &
tion of 9%) and the ultimate load of the corbels of the second database Sohn; 2013. https://doi.org/10.1002/9783433604090.
(difference less than 3%, with a coefficient of variation of 7%). [13] Hordijk DA. Local approach to fatigue of concrete. Technische Universiteit Delft,
1991.

13
Y.M. Neuberger and D.L. Araújo Engineering Structures 284 (2023) 115947

[14] Oliveira Júnior LÁ, Araújo D de L, Toledo Filho RD, Fairbairn EMR, Andrade MAS. [18] Russo G, Venir R, Pauletta M, Somma G. Reinforced Concrete Corbels - Shear
Tension stiffening of steel-fiber-reinforced concrete. Acta Sci Technol 2016;38: Strength Model and Design Formula. ACI Struct J 2006;103:3–10. https://doi.org/
455–63. https://doi.org/10.4025/actascitechnol.v38i4.28077. 10.14359/15080.
[15] Soroushian P, Lee C-D. Distribution and orientation of fibers in steel fiber [19] Castaldo P, Gino D, Bertagnoli G, Mancini G. Resistance model uncertainty in non-
reinforced concrete. ACI Mater J 1990;87:433–9. https://doi.org/10.14359/1803. linear finite element analyses of cyclically loaded reinforced concrete systems. Eng
[16] Lee S, Cho J, Vecchio FJ. Diverse Embedment Model for Steel Fiber-Reinforced Struct 2020;211:110496. https://doi.org/10.1016/j.engstruct.2020.110496.
Concrete in Tension : Model Verification. ACI Mater J 2011;108:526–35. https:// [20] ACI Committee 318. 318-19. Building Code Requirements for Structural Concrete
doi.org/10.14359/51683262. and Commentary. Farmington Hills. American Concrete Institute; 2019.
[17] Harajli MH. Development/splice strength of reinforcing bars embedded in plain
and fiber reinforced concrete. ACI Struct J 1994;91:511–20. https://doi.org/
10.14359/4163.

14

You might also like