Sei sulla pagina 1di 14

Constitutive Model for Reinforcing Steel under

Cyclic Loading
Se Hyung Kim 1 and Ioannis Koutromanos 2

Abstract: The hysteretic stress-strain response of steel reinforcement can significantly affect the performance of reinforced concrete (RC)
structures. Additionally, the possibility of inelastic buckling and fracture of longitudinal bars can have a negative impact on the ductility of
Downloaded from ascelibrary.org by Heriot-Watt University on 07/20/16. Copyright ASCE. For personal use only; all rights reserved.

RC flexural members. This paper presents a uniaxial material model to describe the inelastic behavior of reinforcing steel under cyclic
loading. As a starting point, an existing hysteretic law is employed, which is capable of capturing the material behavior in the absence of
buckling and rupture. One issue with this existing law is the need for an iterative stress calculation. A series of enhancements to the hysteretic
law are established, including the formulation and implementation of a noniterative stress update procedure and the capability to efficiently
account for local inelastic buckling and rupture due to low-cycle fatigue. The proposed material model is validated using the results of
experimental tests. DOI: 10.1061/(ASCE)ST.1943-541X.0001593. © 2016 American Society of Civil Engineers.
Author keywords: Reinforcing steel; Constitutive model; Buckling; Low-cycle fatigue; Rupture; Analysis and computation.

Introduction and Mander (1994), which has been used in simulations with very
good results, also includes the capability to account for rupture due
Reinforced concrete (RC) structures are common in earthquake- to low-cycle fatigue. The material law by Kunnath et al. (2009a)
prone areas, where cyclic loading can lead to extensive cracking uses the same hysteretic law as the one by Chang and Mander
of structural components and inelastic strains in the reinforcing (1994). The model by Hoehler and Stanton (2006) has also been
bars. The hysteretic response of RC structures is greatly affected shown to be capable of satisfactorily reproducing the results of ex-
by the cyclic stress-strain behavior of the reinforcement. Further- perimental tests. This model is formulated in the engineering stress-
more, the loss of the cover concrete and the inelastic deformations strain space and cannot account for the unsymmetric monotonic
can eventually lead to buckling of the reinforcing bars. Buckling response of the material in compression and tension (Dodd and
can significantly affect the ductility of flexural RC members such Restrepo-Posada 1995). Hoehler and Stanton (2006) found that
as columns and beams designed in accordance with modern design the use of the engineering stress-strain space did not negatively af-
standards. There are also instances where the magnitude of the in- fect the accuracy of their model for cyclic loading.
elastic strains and the repeated cyclic loading lead to rupture of Models based on the MP equations can become inaccurate
bars due to low-cycle fatigue. The analytical determination of the in several situations, due to their propensity for “overshooting”
performance of RC structures under earthquake loading requires and “undershooting,” as explained in Filippou et al. (1983) and
the use of material models which can accurately capture all the Kunnath et al. (2009a). Chang and Mander (1994) and Kunnath
aforementioned aspects of cyclic hysteretic response and failure et al. (2009a) have made modifications to the MP model to alleviate
of reinforcing steel bars. this issue. The only means to completely remove the possibility
Many of the previously proposed material models for the cyclic of inaccurate results is to allow an arbitrary number of reversal
response of steel reinforcement use variations of the Menegotto– branches to be kept to memory (Kunnath et al. 2009a). Such re-
Pinto (MP) equations (Chang and Mander 1994; Balan et al. quirement is hard to satisfy for a general-purpose analysis program.
1998; Dhakal and Maekawa 2002; Hoehler and Stanton 2006; Kunnath et al. (2009a) found that allowing the storage of up to 16
Kunnath et al. 2009a) to describe the hysteretic curves. Models reversal branches in their model was adequate for a set of analyses
based on the MP equations are computationally efficient and can that they had conducted. Still, the use of a predefined maximum
capture some of the features of the nonlinear hysteretic behavior number of stored reversal branches does not guarantee that “over-
of the material. Several models of this kind (Balan et al. 1998; shooting” will be entirely prevented. Hoehler and Stanton (2006)
Kunnath et al. 2009a) have established the material response in the have suggested a fix to the “overshooting” issue that only requires
natural stress-strain space, for which the monotonic curves in ten- the storage of two additional points in the stress-strain space. How-
sion and compression practically coincide. The model by Chang ever, this fix was characterized as “pragmatic,” and no validation
has been provided for its efficiency. Additionally, as explained in
1
Engineer, HDR Inc., 3025 Chemical Rd. Suite 110, Plymouth Meeting, Hoehler and Stanton (2006), this fix leads to an undesirable “relax-
PA 19462. ation” phenomenon, i.e., a reduction of stress values obtained at a
2
Assistant Professor, Dept. of Civil and Environmental Engineering, specific strain value, for repeated cyclic loading of fixed amplitude.
Virginia Polytechnic Institute and State Univ., 200 Patton Hall, 750 Dodd and Restrepo-Posada (1995) formulated a material model
Drillfield Dr., Blacksburg, VA 24061 (corresponding author). E-mail:
relying on the natural stress-strain space. An advantage of the spe-
ikoutrom@vt.edu
Note. This manuscript was submitted on August 17, 2015; approved on
cific formulation, compared to the models by Chang and Mander
April 27, 2016; published online on July 18, 2016. Discussion period open (1994) and Kunnath et al. (2009a), is that no more than three
until December 18, 2016; separate discussions must be submitted for in- reversal branches need to be stored in memory. This is accom-
dividual papers. This paper is part of the Journal of Structural Engineer- plished by making a distinction into major, minor, and simple re-
ing, © ASCE, ISSN 0733-9445. versal branches. A disadvantage of the specific model is that the

© ASCE 04016133-1 J. Struct. Eng.

J. Struct. Eng., 04016133


stress update procedure at the reversal branches requires iterations. behavior of steel reinforcement. This paper presents the formu-
The need for an iterative stress update in a uniaxial model is inef- lation of a phenomenological, uniaxial constitutive model for
ficient, especially for the analysis of large structural components reinforcing steel. The model enhances the formulation by Dodd
and systems. It is worth mentioning that the models by Chang and and Restrepo-Posada (1995) by removing the need for iterations
Mander (1994) and Kunnath et al. (2009a) also involve an iterative in the stress update. The curved shapes of the reversal branches
process to initialize the shape of each reversal branch. An addi- are described using nonuniform rational b-splines (NURBS). The
tional difficulty which characterizes the model by Dodd and proposed material law also includes a conceptually simple, physi-
Restrepo-Posada (1995) is that a closed-form expression for the cally meaningful, and algorithmically efficient approach to account
material tangent stiffness in the natural stress-strain space cannot for the onset of local buckling, i.e., buckling between consecutive
be established. Dodd and Restrepo-Posada (1995) proposed an transverse ties. After buckling has occurred, its effect is accounted
approximate expression to circumvent this difficulty. for by multiplying the stress obtained from the core hysteretic law by
Previous work has identified the effect of steel reinforcement a reduction coefficient. Finally, the model can account for rupture by
buckling as very important for RC members subjected to cyclic using a criterion based on the accumulation of a continuous, inelastic
loading (e.g., Monti and Nuti 1992; Rodriguez et al. 1999; Gomes work-related quantity. The specific criterion eliminates the need for a
Downloaded from ascelibrary.org by Heriot-Watt University on 07/20/16. Copyright ASCE. For personal use only; all rights reserved.

and Appleton 1997; Dhakal and Maekawa 2002; Bae et al. 2005; cycle-counting algorithm and can be used for rupture under both
Kunnath et al. 2009a; Urmson and Mander 2012). Several studies monotonic and cyclic loading. A set of validation analyses is con-
(Bae et al. 2005; Massone and Moroder 2009; Urmson and Mander ducted and presented to evaluate the accuracy of the model and the
2012) have used beam models with a layered section to analytically efficiency of the noniterative stress update algorithm. Finally, a
investigate the buckling behavior. While these studies provided discussion section provides a qualitative comparison of the core
useful insights, they were primarily limited to monotonic loading hysteretic law with other existing uniaxial models and a description
scenarios. Buckling for cyclic loading can be captured by modeling of how the proposed buckling criterion can be used in the analysis
the reinforcing bars with beam elements which also account for of RC members, where buckling can involve a bar segment whose
geometric nonlinearities (e.g., Maekawa et al. 2003). Such an ap- length exceeds the spacing of the transverse reinforcement.
proach may entail a prohibitive computational cost for the analysis
of large structural systems. For this reason, most existing models
accounting for buckling are phenomenological in nature, describ- Description of Model
ing the buckling effect through a modification of the stress-strain
law. This is the case for a modified version of the model by The following sections describe the proposed material model. The
Kunnath et al. (2009a), which has been implemented in the analysis salient features of the previously established uniaxial stress-strain
program OpenSees. This version can account for buckling using the law by Dodd and Restrepo-Posada (1995), which forms the basis of
expression by Gomes and Appleton (1997), which is physically the proposed formulation, are briefly presented. Subsequently, the
meaningful, because it considers the bar as a beam under combined enhancements conducted to the material model, i.e., the introduc-
axial and flexural deformation. Still, this expression is highly sim- tion of a noniterative stress computation, and the capability to ac-
plified because it is based on the assumption that the flexural plastic count for inelastic local buckling and rupture due to low-cycle
fatigue, are described.
strength of the bar section does not depend on the axial force.
Kunnath et al. (2009a) considered several alternative models for
capturing buckling and concluded that the most accurate one was Base Material Model
the formulation by Dhakal and Maekawa (2002), which is also
The starting point for the model presented here is a formulation by
available in the implementation of the model in OpenSees. After
Dodd and Restrepo-Posada (1995), which can satisfactorily capture
using the specific buckling formulation in validation analyses,
the performance of reinforcing steel under cyclic loading, when
Kunnath et al. (2009a) found that its accuracy was not very satis-
no buckling or rupture occurs. A complete description of the model
factory. Their ultimate conclusion was that further work is required
with accompanying validation analyses is presented in Dodd and
for satisfactorily capturing buckling.
Restrepo-Posada (1995), but several of its features need to be pre-
Cosenza et al. (2010) proposed what they considered an im-
sented herein. The model is formulated in the so-called natural
provement of the model by Kunnath et al. (2009a) to better capture
stress-strain space, consisting of the true stress, f 0 , and the natural
the inelastic buckling behavior of smooth bars. The general validity
(logarithmic) strain, ε 0 . As explained in Dodd and Restrepo-Posada
and efficiency of this modification were questioned by Kunnath
(1995), the natural stress and strain quantities can be obtained from
et al. (2010).
the engineering stress (nominal stress), σ, and engineering strain, ε,
Several models (e.g., Chang and Mander 1994; Kunnath et al.
with the following relations:
2009a; Mendes and Castro 2014) have also accounted for low-cycle
fatigue and the associated rupture of the reinforcing steel. These ε 0 ¼ lnð1 þ εÞ ð1Þ
models critically hinge on using a variable, quantifying the number 0
f 0 ¼ σð1 þ εÞ ¼ σeε ð2Þ
of cycles or half-cycles of loading. When this variable reaches a
critical value, rupture is assumed to have occurred. The accumu- If the natural strain and stress are known, then Eqs. (1) and (2)
lation of damage which may ultimately lead to rupture is a continu- can be used for the calculation of the engineering strain and stress,
ous process and may not be adequately described by a discrete, respectively.
number-of-cycles variable. For this reason, such models may be Analysis software typically uses the engineering stress and
considered conceptually inaccurate. Furthermore, the reliance on strain quantities to obtain the material contribution to structural re-
the number of cycles entails the need to use a cycle-counting algo- sistance. Dodd and Restrepo-Posada (1995) found that the mono-
rithm. Rainflow counting algorithms (Downing and Socie 1982) tonic stress-strain curves for tension and compression differ in the
are usually employed in such cases. engineering stress-strain space, while they are practically identical
Given the aforementioned issues with existing constitutive laws, in the natural stress-strain space. Thus, formulating the hysteretic
there is still a need for material models which can describe, in a model in the natural stress-strain space can yield the unsymmetric
physically meaningful and numerically efficient fashion, the stress-strain monotonic response for tension and compression in the

© ASCE 04016133-2 J. Struct. Eng.

J. Struct. Eng., 04016133


f
reversal f
fr point control points
fu Ef
fsh1 fy fy
Eu 1 f , ff
fy 1
fa 1 Eo
Es target Slope at initial point = Eo
1 a r Slope at final point = E f
target point
y sh sh1 u , fo
o
ftarget
(a) (b) (c)

Fig. 1. Salient features of Dodd–Restrepo steel model: (a) monotonic curve; (b) reversal behavior; (c) NURBS to describe reversal curve
Downloaded from ascelibrary.org by Heriot-Watt University on 07/20/16. Copyright ASCE. For personal use only; all rights reserved.

engineering space, without a need to formulate different monotonic space (Dodd and Restrepo-Posada 1995). Besides entailing a need
curves for tension and compression. The monotonic tensile curve for iterative computation, this approach has an additional issue
of the model in the engineering stress-strain space is depicted in related to the determination of the material tangent stiffness.
Fig. 1(a). A total of seven material parameters are required for the Specifically, the iterative procedure yields the slope in the normal-
description of the monotonic stress-strain curve, namely, the initial ized coordinate space, and there is no exact formula to convert this
elastic modulus Es , the yield stress fy , the strain εsh at the onset of slope into a tangent material stiffness (i.e., slope in the natural
hardening, the ultimate strength f u , the ultimate strain εu , and the stress-strain coordinate space).
stress f sh1 and strain εsh1 for a point lying on the hardening regime
of the curve, as shown in Fig. 1(a).
The hysteretic unloading-reloading response of the material is Formulation of a Noniterative Stress Update Procedure
characterized by a reversal branch, whose starting point is the cor- A first enhancement to the base model that is deemed necessary is
responding reversal point, schematically explained in Fig. 1(b). The to describe the reversal curves in the natural stress-strain space,
end point of the reversal branch is called the target point, also i.e., the curve between points (εa0 ; fa0 ) and (εtarget
0 0
; f target ), in a way
shown in Fig. 1(b), whose coordinates in the natural stress-strain that eliminates the need for iterations in the stress update procedure
space can be obtained as explained in Dodd and Restrepo-Posada and also allows the straightforward calculation of the material tan-
(1995). Further reversals are treated differently depending on gent stiffness. Instead of using a nonlinear function and coordinate
whether they occur inside or outside a previous reversal branch. normalization like in the original model by Dodd and Restrepo-
If the magnitude of the stress difference between the new reversal Posada (1995), nonuniform rational b-splines (e.g., Piegl and Tiller
point and the previous reversal point is less than 2f y , then the new 1997) are employed to describe the curved portion of the reversal
reversal point lies inside the old reversal curve, otherwise, it lies branches.
outside the old reversal curve. Any reversal branch that starts from A description is provided here for the general case where a
the monotonic curve is called a major reversal. If a new reversal NURBS curve, with continuously reducing slope, joins two points
branch occurs inside a major reversal, then the new reversal branch in the natural stress-strain space, as schematically presented in
is called a minor reversal. Similarly, if a new reversal occurs inside Fig. 1(c). The initial point is assumed to have stress-strain coordi-
a minor reversal, then the new reversal branch is called a simple nates equal to (εo0 ; f o0 ), while the corresponding coordinates of the
reversal. The hysteretic curve shapes that are used for major rever- final point are (εf0 ; ff0 ). The slope of the curve at the initial and final
sals are different than those for minor or simple ones. Reversals that points is also given and is equal to Eo0 and Ef0 , respectively. The
occur inside the yield plateau region are always treated as minor NURBS curves adopted here include three control points (Piegl
(Dodd and Restrepo-Posada 1995). and Tiller 1997)—the initial point, the final point, and an interior
A reversal branch in the natural stress-strain space is character- control point—as shown in Fig. 1(c). The reversal curves are
ized by three points, namely, the reversal point (εr0 ; f r0 ), the point described in a normalized stress-strain space (ε 0 0 ; f 0 0 ), which is re-
(εa0 ; fa0 ) which defines an initial linear portion of the reversal branch lated to the natural stress-strain space in accordance with the fol-
[as also shown in Fig. 1(b)], and the target point (εtarget 0 0
; ftarget ). The lowing expressions:
coordinates of the target point and the slope of the stress-strain dia-
0
gram, ET;target , at that point are required to determine the curved f 0 − fo0
f00 ¼ ð4aÞ
portion of the reversal branch. The point (εa0 ; fa0 ) is defined so that f f0 − fo0
the magnitude of the difference between stress value fr0 and f a0 is
equal to fy , as shown in Fig. 1(b). As also shown in the figure, the
ε 0 − εo0
points (εr0 ; f r0 ) and (εa0 ; fa0 ) are joined by a straight line whose slope ε00 ¼ ð4bÞ
is equal to Eu0 , i.e., the unloading modulus of the material. This εf0 − εo0
modulus is given by the following expression:
The range of possible values of the normalized coordinates ε 0 0
  and f 0 0 is between 0 and 1. The variation of the normalized strain,
0 1
Eu ¼ Es 0.82 þ ð3Þ ε 0 0 , on the reversal curves is expressed as a function of a dimen-
5.55 þ 1,000εM0
sionless variable u, which also varies between 0 and 1, and of a
constant εi0 0 , which is equal to the normalized coordinate of the
where εM0 = maximum attained magnitude of strain; and Es = initial
interior control point in the NURBS curve
elastic modulus.
The stress computation in the curved portion of the reversal
branches is conducted in a normalized, dimensionless coordinate ε 0 0 ðuÞ ¼ 2uð1 − uÞεi0 0 þ u2 · 1 ð5Þ

© ASCE 04016133-3 J. Struct. Eng.

J. Struct. Eng., 04016133


 
The normalized coordinate, εi0 0 , of the interior control point is df 0 0 ff0 − f o0
given by the following equation: ET0 ¼ ð11Þ
dε 0 0 εf0 − εo0
Ef0 ð1 − εf0 þ εo0 Þ − fo0
εi0 0 ¼ ð6Þ The term in brackets in Eq. (11) is the slope in the normalized,
ðEo0 − Ef0 Þðεf0 − εo0 Þ dimensionless space, which can be calculated with the following
expression:
Eq. (5) can be inverted, to provide u as a function of the
normalized strain, ε 0 0 . Specifically, u can be obtained from the df 0 0 ð2 − 4uÞf i0 0 þ 2u
0 0 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi ð12Þ
following equation: dε 4ðεi0 0 Þ2 þ 4ε 0 0 ð1 − 2εi0 0 Þ
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
−2εi0 0 þ 4ðεi0 0 Þ2 þ 4ε 0 0 ð1 − 2εi0 0 Þ Thus, a closed-form expression can be obtained for the material
uðε 0 0 Þ ¼ ð7Þ tangent stiffness, ET0 , in the natural stress-strain space. Given ET0 ,
2ð1 − 2εi0 0 Þ
the corresponding tangent stiffness, ET , in the engineering stress-
Downloaded from ascelibrary.org by Heriot-Watt University on 07/20/16. Copyright ASCE. For personal use only; all rights reserved.

strain space can be obtained from the following equation (Dodd and
The normalized stress, f 0 0 , can also be established as a function
Restrepo-Posada 1995):
of u
ET0 − f 0
f 0 0 ðuÞ ¼ 2uð1 − uÞfi0 0 þ u2 · 1 ð8Þ ET ¼ ð13Þ
ð1 þ εÞ2
where the value fi0 0 is the ordinate of the internal control point in
The NURBS curves used in the present model have been estab-
the normalized space and is given by the following equation:
lished to provide hysteretic curves which match the corresponding
curves of the original model by Dodd and Restrepo-Posada (1995).
Eo0 ðεf0 − εo0 Þ
fi0 0 ¼ εi0 0 ð9Þ For minor and simple reversals, the initial point of the NURBS
f f0 − fo0 curve is the point (εa0 ; f a0 ) with a slope equal to Eu0 , and the final
0 0 0
point is the target point (εtarget ; f target ) with slope equal to ET;target .
Eqs. (5) and (8) correspond to quadratic NURBS interpolations For major reversals, a modified approach has been established to
using three control points with coordinates 0, εi0 0 , and 1 for Eq. (5) ensure accurate results. Specifically, given the points (εa0 ; f a0 ) and
0 0
and 0, fi0 0 , and 1 for Eq. (8). (εtarget ; f target ), the strain, stress, and slope values are determined
0
Based on the preceding, an algorithm can be established to cal- for a point termed the characteristic point. The strain, εchar , of the
culate the natural stress, f 0 , of a point lying on the NURBS curve, if characteristic point is given by the following expression:
the natural strain, ε 0 , is known for that point. Given the coordinates 0 0
and slope of the initial point and final point defining the NURBS 0
ET;target ð1 − εtarget þ εa0 Þ − f a0 0
εchar ¼ ðε − εa0 Þ þ εa0 ð14Þ
curve, the normalized strain value is calculated using Eq. (4b), and ðEu0 − ET;target
0 0
Þðεtarget − εa0 Þ f
the parameter u can be calculated using Eq. (7). Given u, the nor-
malized stress value, f 0 0 , is calculated using Eq. (8), and the natural The calculation of the stress f char 0 , and slope, E 0 , of the char-
char
stress, f 0 , can be found by appropriate rearrangement of Eq. (4a) acteristic point requires the determination of two points, (1) and (2),
in the natural stress-strain space. As shown in Fig. 2(a), point (1) is
f 0 ¼ f 0 0 ðf f0 − f o0 Þ þ fo0 ð10Þ found by drawing a line with slope equal to Eu0 through point
(εa0 ; fa0 ), and point (2) is found by drawing a line with slope
0
Etarget 0
through point (εtarget 0
; f target ). The natural strain values of
The calculation of the slope of the stress-strain diagram, i.e., of
the material tangent stiffness, ET0 , in the natural stress-strain space, points (1) and (2), ε10 and ε20 , respectively, are calculated using
is straightforward. Specifically, a relation between ET0 and the slope the following equations:
in the normalized stress-strain space (ε 0 0 ; f 0 0 ) can be obtained using
the chain rule of differentiation ε10 ¼ ε10 0 ðεtarget
0
− εa0 Þ þ εa0 ð15aÞ

(a) (b) (c)

Fig. 2. Establishment of hysteretic curves for major reversals: (a) find auxiliary points (1) and (2); (b) find characteristic point; (c) define two NURBS
segments

© ASCE 04016133-4 J. Struct. Eng.

J. Struct. Eng., 04016133


ε20 ¼ ε20 0 ðεtarget
0 − εa0 Þ þ εa0 ð15bÞ 3. The lateral restraint provided by the concrete cover to the bar is
neglected.
The dimensionless parameters ε10 0 and ε20 0 are obtained using the 4. The bar behaves like a beam member with a circular cross
following expressions: section.
These assumptions are schematically summarized in Fig. 4(a).
0 0
ET;target ð1 − εtarget þ εa0 Þ − fa0 The onset of buckling is assumed to occur at the instant when a bar
ε10 0 ¼ βð2 − βÞ ð16aÞ
ðEu0 − ET;target
0 0
Þðεtarget − εa0 Þ segment between consecutive ties tends to deflect laterally. The lat-
eral deflection will lead to the development of curvature along the
bar segment. The variation of deflection along the bar segment is
ε20 0 ¼ minð1 − ε10 0 ; 0.5Þ ð16bÞ assumed to be given by the following equation:
  
The parameter β in Eq. (16a) is calculated in terms of a quantity 1 2π
dwðxÞ ¼ dwo 1 − cos x ð17Þ
Ω, which was defined by Dodd and Restrepo-Posada (1995) to 2 L
control the shape of the hysteretic curves. The relation between
Downloaded from ascelibrary.org by Heriot-Watt University on 07/20/16. Copyright ASCE. For personal use only; all rights reserved.

β and Ω is provided in Appendix I. The variable x in Eq. (17) assumes values between 0 and L, in
The stress and slope, fchar 0 0 , respectively, of the char-
and Echar which L is the transverse tie spacing. The deflection distribution
acteristic point are found by taking a NURBS interpolation described by Eq. (17) corresponds to the deformed shape of a beam
between points (1) and (2), as also shown in Fig. 2(b). After the with full rotational restraints at the two ends and symmetric bend-
characteristic point has been found, the actual hysteretic curve ing moment distribution. It can easily be verified that dwo is equal
for major reversals can be established. It consists of two NURBS to the deflection at the middle of the bar segment. The slope of the
segments, the first segment joining point (εa0 ; fa0 ) with point deflected reference curve is zero at the two end points and at the
0 ; f 0 ), and the second segment joining points (ε 0 ; f 0 ) and
(εchar char char char middle of the bar. The curvature dφo at the middle of the bar seg-
0
(εtarget 0
; f target ), as shown in Fig. 2(c). ment can be obtained by differentiating Eq. (17) twice and then
The algorithm for calculating the stress and material tangent setting x ¼ L=2:
stiffness on the reversal curves is summarized in a flowchart pro-
vided in Fig. 3. An obvious alternative to using NURBS curves 2π2
would have been to employ an MP-based curve for the reversal dφo ¼ −dwo
L2
branches, using the equations of Chang and Mander (1994). This
approach is not pursued herein because, as mentioned in the intro- Solving the obtained equation for dwo gives the following
duction, it requires iteration to establish the parameters which de- expression:
fine the shape of a given reversal curve.
L2
dwo ¼ −dφo ð18Þ
Accounting for Local Buckling 2π2

The enhanced material model can also account for the local buck- Thus, if the curvature dφo is known at the midpoint between
ling effect in an algorithmically efficient and numerically robust consecutive ties, the deflection at the same location is also known
fashion. The term local buckling refers to the occurrence of trans- from Eq. (18). The free-body diagram of half the bar segment
verse deflections between consecutive transverse ties. The buck- can be established as shown in Fig. 4(b). A relation can be ob-
ling model which is considered here is based on the following tained between the axial force and the bending moment, dM, which
assumptions: will develop at the middle of the segment. It is worth mentioning
1. The reinforcing bar is initially straight. that the algebraic values of both N and dM in Fig. 4(b) are neg-
2. The bar is assumed to be fixed against lateral deflection and ative. Moment equilibrium for the segment leads to the following
rotation at the tie locations. relation:

Fig. 3. Flowchart for computation of natural stress and tangent stiffness for the curved portion of the reversal curves

© ASCE 04016133-5 J. Struct. Eng.

J. Struct. Eng., 04016133


(a) (b)
Downloaded from ascelibrary.org by Heriot-Watt University on 07/20/16. Copyright ASCE. For personal use only; all rights reserved.

(c) (d)

Fig. 4. Simplified mechanical model for capturing local buckling: (a) simplified model for bar segment between ties; (b) free-body diagram for half of
the bar segment; (c) response of section A-A; (d) stress-strain response

2 · dM ¼ N · dwo dM ¼ D̂T;21 dεo þ D̂T;22 dφo ð20bÞ

Accounting for Eq. (18) eventually yields the following The sectional tangent stiffness terms are written in the following
expression: form:
L2    
dM ¼ −N · dφo ð19Þ D̂T;11 D̂T;12 γ T;11 · π · Es ðD=2Þ2 γ T;12 · π · Es ðD=2Þ3
4π2 ¼
D̂T;21 D̂T;22 γ T;21 · π · Es ðD=2Þ3 γ T;22 · π · Es ðD=2Þ4
The buckling equations are established by focusing on the
ð21Þ
incremental response of the cross section at the middle of the bar
segment between consecutive ties. Buckling occurs when an incre-
The dimensionless coefficients γ T;11 , γ T;12 , γ T;21 , and γ T;22 in
ment, dεo , in the reference axial strain of that section leads to zero
Eq. (21) have been established, by means of a curve-fitting process,
increment in the axial force, due to the spontaneous development of
as functions of the ratio η of the tangent modulus ET over Es, as
transverse deflections, which lead to a curvature dφo in the section.
explained in Appendix II.
The following derivations provide the mathematical expression for
Substituting Eq. (19) into Eq. (20b) and solving for dφo yields
this criterion. The material behavior for the buckling considerations
the following expression:
is based on the engineering stress and strain.
The response of the middle section, A-A, of the bar segment is  
D̂T;21
presented in Fig. 4(c). Right before buckling, a uniform compres- dφo ¼ − L2
dεo ð22Þ
N 4π 2 þ D̂T;22
sive stress distribution exists in the section. The same figure shows
that right after buckling, the combination of dεo and dφo will lead
to further compressive loading (i.e., increase in compressive stress) Substitution of Eq. (22) into Eq. (20a) provides a “condensed”
for a portion of the section, and to unloading (i.e., decrease in sectional relation between the increment of axial force, dN, and the
the compressive stress) for the remaining portion of the section. increment of sectional axial strain, dεo , in the presence of an infini-
Because of the infinitesimal nature of the increments, it can be tesimal transverse deflection of the bar segment
assumed that the sectional fibers which are further loaded in com-  
D̂ · D̂T;12
pression will have incrementally linear stress-strain response dN ¼ D̂T;11 − T;21 L2
dεo ð23Þ
characterized by the material tangent modulus, ET , while the fibers N 4π 2 þ D̂T;22

which unload will have an incrementally linear response charac-


terized by the material unloading modulus. This modulus is ap- Finally, a condition for buckling is obtained, using the stipula-
proximately equal to the initial elastic modulus, Es . It is worth tion that the increment in axial force is zero
mentioning that the assumption of incrementally linear response,
D̂T;21 · D̂T;12
presented in Fig. 4(d), essentially forms the basis of the well-known D̂T;11 − 2 ¼0 ð24Þ
2 þ D̂T;22
L
reduced modulus method (Bazant and Cedolin 2000) for determin- N 4π
ing the buckling load of inelastic columns.
The incremental change in the axial force, dN, and moment, dM, Because buckling can only be triggered for compression,
for section A-A will be given by the product of the sectional tangent Eq. (24) is only checked for compressive axial stresses, N < 0.
stiffness matrix times the incremental sectional deformations Additionally, the equation is only valid if the denominator in the
fraction of Eq. (24) is positive, N · L2 =ð4π2 Þ þ D̂T;22 > 0. If the
dN ¼ D̂T;11 dεo þ D̂T;12 dφo ð20aÞ magnitude of compressive axial force becomes so large that

© ASCE 04016133-6 J. Struct. Eng.

J. Struct. Eng., 04016133


the denominator becomes zero, then the following equation will reinforcing steel bars which incurred buckling. If the value εn is
apply: between 0 and 1, then the logarithm of the scaling factor Fb is
obtained by the following expression:
4π2 D̂T;22
N ¼ N cr ¼ − ð25Þ log10 Fb ¼ 2ub ð1 − ub Þ0.45M 1 þ u2b f2 ð29Þ
L2
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
The value N cr is equal to the buckling load of a beam with flexu- where ub ¼ −4.5 − 10 0.2 þ 0.1εn .
ral rigidity D̂T;22 and rotational restraints at both ends. For this rea- Eq. (29) corresponds to a quadratic NURBS approximation,
son, the proposed model assumes that buckling also occurs when with the ordinates of the three control points being equal to 0,
N ¼ N cr , even if Eq. (24) is not satisfied. 0.45M 1 , and f2 . The parameters M 1 and f2 are given by the
Because the axial force in the bar can be set equal to the product following equations:
of the axial stress times the cross-sectional area, Eqs. (24) and (25)
can be written in terms of the stress in the material. If Eq. (21) is M 1 ¼ −0.12ðL=D − 1Þ þ 0.87 ð30Þ
used, Eq. (24) yields the following expression:
Downloaded from ascelibrary.org by Heriot-Watt University on 07/20/16. Copyright ASCE. For personal use only; all rights reserved.

f 2 ¼ −0.1L=D þ 0.28 ð31Þ


γ T;21 · γ T;12 · π · Es
γ T;11 − ¼0 ð26Þ
σ π1 ðDL Þ2 þ γ T;22 · π · Es The value of M 1 obtained from Eq. (30) must lie within the
following interval:
Similarly, mathematical manipulations in Eq. (25) yield an
equation in terms of the stress, corresponding to buckling of a beam −2 ≤ M1 ≤ −0.3
with rotational restraints at both ends
 2 A different expression is used for the logarithm of Fb for values
σ ¼ σcr ¼ −π2 · γ T;22 · Es
D
ð27Þ of εn greater than 1
L
log10 Fb ¼ −0.17ðεn − 1Þ þ f 2 ð32Þ
Given the engineering stress, σ, buckling is assumed to occur in
the present model whenever either Eq. (26) or Eq. (27) is satisfied. Finally, the engineering stress, σ, obtained from the core con-
The two specific equations demonstrate that the buckling criterion stitutive model, is multiplied by the factor Fb to give the corrected
is size-independent, i.e., for a given material and ratio L=D, the stress, σbc , accounting for buckling
buckling stress will be the same for all values of bar diameter. The
σbc ¼ Fb · σ ð33Þ
dimensionless ratio L=D can be regarded as a slenderness param-
eter. The greater L=D is, i.e., the more slender the bar segment, the
The magnitude of the compressive stress after buckling is not
greater will be its propensity for buckling and the lower the stress at
allowed to become less than 0.2f y . This lower-bound magnitude
buckling will be.
has been proposed by Dhakal and Maekawa (2002). Eqs. (30)
After the stress computation of the core constitutive model is
and (31) give values of M 1 and f 2 in close agreement with those
conducted in accordance with the previous sections, knowledge
deduced from the experimental data by Bayrak and Sheikh (2001),
of the tangent material elastic modulus allows the computation
as shown in Figs. 5(a and b). The data values listed as experimental
of η and the determination of whether the buckling criterion is
in the two figures are those for which the proposed buckling model
met. If buckling is detected, the strain εb , at which the buckling
reproduces the experimental buckling data as accurately as pos-
criterion is satisfied, is stored. To account for the reduction in
sible. The variation of the reduction coefficient, Fb , obtained with
bar resistance due to buckling, the obtained stress is multiplied
the proposed model satisfactorily reproduces the experimental ob-
by a positive reduction coefficient, Fb , which cannot exceed 1.
servations by Bayrak and Sheikh (2001), as shown in Fig. 5(c).
The coefficient Fb is a function of a dimensionless parameter
If, at any stage in the analysis after the onset of buckling, the
εn , which is determined from the following equation:
strain becomes greater than εb , then the buckling effect is consid-
εb − ε ered lost and no buckling correction is made to the stresses. If the
εn ¼ ð28Þ
ε b þ εu buckling criterion, i.e., Eqs. (26) or (27), is satisfied at a later step
in the analysis, a new value εb is stored and the coefficient Fb
The dependence of Fb on εn has been established based on the is calculated using Eqs. (29)–(32). A flowchart summarizing the
results of experimental tests by Bayrak and Sheikh (2001) on computations accounting for buckling is provided in Fig. 6.

1.0
L Experiment
5
0.8 D Model
L
0.6 6 L
D 7
Fb

D
0.4
L
0.2 10 L L
D 9 8
0.0 D D
0.0 0.5 1.0
0 1.5 2.0 2.5
n b b u

(a) (b) (c)

Fig. 5. Curve fitting of buckling coefficient using experimental data by Bayrak and Sheikh (2001): (a) variation of parameter M 1 ; (b) variation of
parameter f 2 ; (c) variation of buckling reduction coefficient

© ASCE 04016133-7 J. Struct. Eng.

J. Struct. Eng., 04016133


ḟ 0
ε̇p0 ¼ ε̇ 0 − ð36Þ
Eu0
Rupture, i.e., loss of the material, occurs when D¼Dcr , where
Dcr is a material model parameter. The numerical implementation
of the fatigue and rupture law is based on a midpoint rule. It must be
emphasized that the specific low-cycle fatigue model should be per-
ceived as empirical and may not rigorously describe the causes of
bar rupture. For example, Restrepo-Posada (1993) has argued that
the actual fracture in steel reinforcement occurs in the compressive
strain region of the bar during extensive buckling and is caused by
stress concentrations at the roots of the rebar ribs.

Model Validation
Downloaded from ascelibrary.org by Heriot-Watt University on 07/20/16. Copyright ASCE. For personal use only; all rights reserved.

Fig. 6. Flowchart of computations to account for buckling effect


The proposed model formulation has been validated using the re-
sults of experimental tests. A goal of the analyses is to verify that
the proposed model can indeed provide the hysteretic response of
Equations to Account for Low-Cycle Fatigue the original, iterative version of the model by Dodd and Restrepo-
The model is also provided with a capability to account for low- Posada (1995). The material model parameters used in the simu-
cycle fatigue and the associated rupture of reinforcing steel. Con- lations are summarized in Table 1. Another goal of the analyses
trary to other models for reinforcing steel (e.g., Chang and Mander presented herein is to evaluate the predictive capabilities of the
1994), the proposed material law does not use the number of cycles material model. For this reason, the material parameters have been
as the damage variable. Thus, the need to use a cycle-counting al- set in accordance with the provided monotonic tensile stress-strain
gorithm is eliminated. The criterion that is used is a uniaxial version curve of each test, and no fine-tuning of material parameters to im-
of that presented in Huang and Mahin (2010) for structural steel prove the agreement with the experimental data is pursued.
members. The specific criterion states that rupture occurs when a
scalar damage variable, which is associated with the inelastic work Analyses in the Absence of Buckling or Rupture
accumulated under tensile stress, exceeds a critical value. The im-
A first series of validation analyses has been conducted for exper-
plementation in the context of the Dodd–Restrepo model is based
imental tests by Restrepo-Posada et al. (1994), by Kent and Park
on the natural (true) stress, f 0 , and the natural plastic strain, εp0 .
(1973), and by Ma et al. (1976). These tests involved cyclic loading
A scalar variable, D, is used to denote the accumulated damage
of reinforcing steel, primarily in the tensile strain regime, and did
in the material. The evolution of D is governed by the following
not lead to buckling or rupture. Analyses have been conducted
rate equation:
using both the material model proposed herein and the original for-
8  t mulation by Dodd and Restrepo-Posada (1995). As shown in Fig. 7,
< Y ε̇ 0 ; if f 0 > 0
the modified, noniterative formulation proposed herein gives very
Ḋ ¼ S p ð34Þ similar results to those obtained with the original, iterative version
: of the model. The overall accuracy of the material model is very
0; otherwise
good. To evaluate the increased efficiency provided by the noniter-
ative formulation, Table 2 summarizes the average number of iter-
where Y ¼ 1=2ðf 0 Þ2 =Es ; and S, t = material constants. The value ations required in the reversal curves, for each analysis using the
for S adopted herein is S ¼ 1=2ðf y0 Þ2 =Es . In this case, the evolution original model by Dodd and Restrepo-Posada (1995). The iterative
equation for D assumes the following form: version uses by default 10 Newton–Raphson iterations and, if con-
vergence is not obtained after this, the iterative procedure continues
8  0 2t
< f with a bisection method. Table 2 indicates that a rather signifi-
ε̇p0 ; if f 0 > 0
Ḋ ¼ f y0 ð35Þ cant number of iterations are required for the original formulation.
: Given that the iterative stress update of the original Dodd–Restrepo
0; otherwise model must be conducted for each location where reinforcing steel
exists and for each iteration to satisfy equilibrium at the structural
The rate of the plastic natural strain, ε̇p0 , can be obtained from level, the speed-up in computation associated with the noniterative
the following equation: version of the model is expected to be significant.

Table 1. Material Parameters Used in Validation Analyses


Specimen Es (MPa) f y (MPa) εsh εsh1 fsh1 (MPa) εsu fu (MPa)
Restrepo-Posada et al. (1994), HX15 196,000 447 0.017 0.051 552 0.158 605
Restrepo-Posada et al. (1994), HV15 199,000 482 0.018 0.051 592 0.146 641
Restrepo-Posada et al. (1994), MJSG2 197,000 319 0.022 0.039 380 0.225 476
Kent and Park (1973), Specimen 8 199,230 309 0.010 0.021 330 0.100 378
Ma et al. (1976), Specimens 1, 2, and 3 199,230 453 0.012 0.026 550 0.123 653
Mander (1983) 178,500 289 0.017 0.040 340 0.150 370
Monti and Nuti (1992), C3 179,355 428 0.007 0.040 523 0.100 584
Monti and Nuti (1992), S5, S8, S11 199,285 519 0.007 0.020 628 0.090 739
Kunnath et al. (2009b) 139,310 502 0.017 0.042 595 0.012 670

© ASCE 04016133-8 J. Struct. Eng.

J. Struct. Eng., 04016133


Downloaded from ascelibrary.org by Heriot-Watt University on 07/20/16. Copyright ASCE. For personal use only; all rights reserved.

Fig. 7. Validation of uniaxial stress-strain law and comparison with original material model by Dodd and Restrepo-Posada (1995) for cyclic loading
without buckling: (a) Specimen HX15 by Restrepo-posada et al. (1994); (b) Specimen HV15 by Restrepo-Posada et al. (1994); (c) Specimen MJSG2
by Restrepo-Posada et al. (1994); (d) Specimen 1 by Ma et al. (1976); (e) Specimen 2 by Ma et al. (1976); (f) Specimen 3 by Ma et al. (1976);
(g) Specimen 8 by Kent and Park (1973)

Table 2. Required Number of Iterations and Percentage of Steps Requiring reported in Dhakal and Maekawa (2002), and the comparison of the
Use of Bisection Method, for Use of the Original Dodd–Restrepo Model in analytically obtained stress-strain response with the corresponding
the Analyses Presented in Fig. 7 experimental observations, for different values of slenderness ratio,
Percentage is presented in Fig. 8. The agreement between the analytical pre-
Average Average of steps dictions and the experimental observations is deemed satisfactory.
number number of requiring The results of the tests conducted by Monti and Nuti (1992)
of N–R bisection bisection are used for the evaluation of the capability of the proposed model
Specimen iterations iterations method (%) to capture the effect of buckling under cyclic strain histories. The
Restrepo-Posada et al. (1994), HX15 4 14 27 comparison for four specimens, presented in Fig. 9, shows a very
Restrepo-Posada et al. (1994), HV15 5 10 25 good agreement between the analytically predicted and experimen-
Restrepo-Posada et al. (1994), MJSG2 5 8 15 tally obtained stress-strain response. The only obvious disagree-
Ma et al. (1976), Specimen 1 4 15 30 ment between analysis and experiment is observed for specimen
Ma et al. (1976), Specimen 2 5 11 33 S5, shown in Fig. 9(b). Still, this disagreement is not deemed sig-
Ma et al. (1976), Specimen 3 4 12 23 nificant, because this specimen corresponds to a slenderness ratio,
Kent and Park (1973), Specimen 8 5 10 27
L=D, of 5, which may be considered relatively low for actual RC
Note: The Newton–Raphson (N–R) method is initially used in the members.
iterations. If the N–R method does not converge after 10 iterations or if
a zero tangent slope is obtained, then the bisection method is employed.
Analyses Involving Rupture due to Low-Cycle Fatigue
A final set of analyses is conducted for the tests by Kunnath et al.
Analyses Involving Buckling
(2009b), in which rebar specimens were subjected to cyclic loading
The capability of the proposed model to capture the effect of in- until rupture due to low-cycle fatigue occurred. The proposed
elastic buckling is also evaluated with monotonic and cyclic analy- model is validated for two specimens that incurred rupture after
ses. The former are conducted for the tests by Mander (1983), as 7 and 17 cycles, respectively. The material parameters to be used

© ASCE 04016133-9 J. Struct. Eng.

J. Struct. Eng., 04016133


Fig. 8. Validation of buckling model with the monotonic buckling tests by Mander (1983), as reported in Dhakal and Maekawa (2002): (a) L=D ¼ 6;
(b) L=D ¼ 10; (c) L=D ¼ 15
Downloaded from ascelibrary.org by Heriot-Watt University on 07/20/16. Copyright ASCE. For personal use only; all rights reserved.

Fig. 9. Verification of model for experimental cyclic tests by Monti and Nuti (1992) involving buckling: (a) Specimen C3 (L=D ¼ 11); (b) Specimen
S5 (L=D ¼ 5); (c) Specimen S8 (L=D ¼ 8); (d) Specimen S11 (L=D ¼ 11)

in the simulations are so calibrated that the model can reproduce the can satisfactorily capture the occurrence of rupture for both the
monotonic tensile stress-strain curve for the specific type of steel, monotonic and cyclic tests. The analyses for cyclic loading provide
as shown in Fig. 10(a). The material parameters Dcr and t have good estimates of the experimentally obtained hysteretic response,
been set equal to 0.20 and 0.50, respectively, because these values as shown in Figs. 10(b and c). For the specimens which failed after

Fig. 10. Validation of model for experimental tests by Kunnath et al. (2009b) involving rupture: (a) monotonic loading; (b) cyclic loading—rupture
after 7 cycles; (c) cyclic loading—rupture after 17 cycles

© ASCE 04016133-10 J. Struct. Eng.

J. Struct. Eng., 04016133


Fig. 11. Sensitivity of low-cycle fatigue behavior on parameters of the material model: (a) effect of Dcr for t ¼ 0.50; (b) effect of t for Dcr ¼ 0.20
Downloaded from ascelibrary.org by Heriot-Watt University on 07/20/16. Copyright ASCE. For personal use only; all rights reserved.

Fig. 12. Results for rupture obtained using previously proposed model by Kunnath et al. (2009a): (a) monotonic loading; (b) cyclic loading—rupture
in analysis after 21 cycles; (c) cyclic loading—rupture in analysis after 50 cycles

7 and 17 cycles, the analyses predict rupture after 8 and 14 cycles, Discussion
respectively. Thus, the prediction of the fatigue-induced rupture is
deemed satisfactory. A sensitivity study is also provided, for the The core hysteretic law presented herein is deemed attractive for
specimen whose response is presented in Fig. 10(b), to determine use in analysis because it combines conceptual simplicity (by min-
the impact of parameters Dcr and t on the prediction of rupture. As imizing the calibration effort) and algorithmic efficiency (by not
shown in Fig. 11(a), a change by 25% in the value of Dcr leads to a requiring any iteration). If no buckling and rupture are considered,
change by 2 in the number of cycles to rupture. The sensitivity of then a minimal amount of calibration is required for parameters
the results on the value of t is negligible, as shown in Fig. 11(b) which pertain to the monotonic tensile stress-strain curve. These
because failure essentially occurs at the same instant for all values parameters can be easily determined from experimental tests be-
of t considered. cause the vast majority of tests on reinforcing bars are conducted
To allow a comparative evaluation of the proposed formulation for monotonic tension. Contrary to other accurate constitutive laws
for low-cycle fatigue and rupture with existing similar models, the which have been implemented in analysis programs, e.g., the origi-
analyses involving rupture have also been conducted using the nal model by Dodd and Restrepo-Posada (1995) and the model by
model by Kunnath et al. (2009a) as implemented in OpenSees. Kunnath et al. (2009a), the model presented herein does not require
The results of these analyses, which have been conducted using any iteration for the stress update. The noniterative formulation is
the calibrated fatigue parameters of Kunnath et al. (2009a), are pre- expected to significantly enhance the efficiency and robustness of
sented in Fig. 12. In their analyses, Kunnath et al. (2009a) used the model. The model by Hoehler and Stanton (2006) does not re-
two different values for one of the model parameters, namely, quire iteration either, but it does include two additional parameters
Cd . The analyses presented herein are conducted for both values which require calibration and affect the shape of the hysteretic
of this parameter. As shown in Fig. 12(a), the model by Kunnath curves. These parameters cannot be obtained from monotonic ten-
et al. (2009a) gives premature rupture for monotonic loading, if sion tests.
their proposed calibration is used for the fatigue and rupture The proposed buckling law requires the determination of the
law. Furthermore, the specific model leads to significant overesti- slenderness ratio parameter, L=D. One obvious choice would be
mation of the number of cycles to failure, as shown in Figs. 12(b to set L equal to the spacing of the transverse ties in an RC member
and c). Specifically, the analysis predicts rupture after 21 and 50 and D equal to the diameter of the bar. However, it has been argued
cycles for the cases presented in Figs. 12(b and c), respectively. (Falk and Govindjee 1999) that, even if the onset of buckling oc-
The corresponding number of cycles to failure in the actual exper- curs over a single tie spacing, the bar eventually “snaps” through a
imental tests was 7 and 17, respectively, for the two cases. It is different buckling mode in which the wavelength of the deflected
obvious that—for the specific analyses—the model proposed herein buckled bar is larger than the tie spacing. A phenomenological de-
gives more accurate results than the one by Kunnath et al. (2009a). scription of this type of buckling, hereinafter termed nonlocal buck-
There is another issue with the model by Kunnath et al. (2009a), ling, is deemed necessary for, e.g., beam models of RC members
as deduced from Fig. 12(a). Specifically, the activation of the in which each reinforcing steel bar is represented as a fiber at the
fatigue law leads to a distortion of the monotonic stress-strain cross-sectional level. The present material model can be calibrated
curve. This distortion is caused by a damage parameter which is to account for nonlocal buckling, by appropriate selection of the
used in the model by Kunnath et al. (2009a) to reduce the stress L=D parameter, using the procedure presented in Maekawa et al.
obtained from the core hysteretic law when fatigue is accounted for. (2003). This procedure uses a simple and efficient algorithm to

© ASCE 04016133-11 J. Struct. Eng.

J. Struct. Eng., 04016133


determine the effective bar slenderness ratio, L=D, accounting for physically meaningful approach. The inelastic buckling model
the possibility of nonlocal buckling. The effective L=D value de- can be used for simulations involving buckling over a length which
pends on the properties of the cover concrete and of the layout, is greater than the tie spacing, provided that an equivalent rebar
diameter, and spacing of the transverse reinforcement. After the slenderness ratio is used in accordance with previous research.
effective L=D ratio has been established, it can be provided as The low-cycle fatigue and rupture criterion is based on the accu-
the slenderness ratio in the model proposed herein and used in mulation of inelastic work under tensile axial stress. The buckling
the analysis of RC components. The impact of the cover concrete and low-cycle fatigue criteria are generic, in the sense that they can
and of the transverse reinforcement on the buckling reduction fac- be used with any hysteretic law for reinforcing steel.
tor, Fb , can be neglected as a first step. This is a potentially
conservative simplifying assumption because it may lead to
more severe bar strength degradation after buckling. Still, this Appendix I. Determination of Coefficient β in
assumption is supported by the fact that the cover and transverse Eq. (16a) for Major Reversals
reinforcement in an actual RC member must resist buckling of
multiple bars. Thus, the restraint provided by these mechanisms The first step for obtaining the value of β in Eq. (16a) is to obtain
Downloaded from ascelibrary.org by Heriot-Watt University on 07/20/16. Copyright ASCE. For personal use only; all rights reserved.

to an individual bar is less significant than for the case where a parameter Ω for the reversal, using the expression provided in
single bar buckles. Furthermore, it is worth mentioning that the spe- Dodd and Restrepo-Posada (1995). Given the value of Ω, β can
cific approach to capture buckling is inherently simplified and in- be calculated from the following equations:
accurate because it relies on the assumption that the stress in the bar
segment over the buckling length is constant. This assumption may β ¼ 0 if Ω ≤ 0.04 or Ω ≥ Ω3 ð37aÞ
be valid for local buckling considerations, but it becomes less ac-
curate for buckling of bar segments with larger length. Thus, pur- β ¼ 10.1ðΩ − 0.04Þ if 0.04 ≤ Ω ≤ Ω1 ð37bÞ
suing a refined law for Fb, accounting for the cover and transverse
steel, may not be meaningful for such a simplified model. β ¼ ð1 − uΩ Þ2 β 1 þ 2uð1 − uΩ Þβ 2 þ u2Ω β 3 if Ω1 ≤ Ω ≤ Ω2
The low-cycle fatigue model is a uniaxial version of previous
expressions, primarily employed for rupture in structural steel. ð37cÞ
Thus, while its use for reinforcing steel has given satisfactory re-
sults, it should be regarded as an empirical model. An advantage β ¼ md ðΩ − Ω3 Þ if Ω2 ≤ Ω ≤ Ω3 ð37dÞ
of the proposed rupture model compared to other existing formu-
lations is that it removes the need for a cycle-counting algorithm. The constant parameters appearing in Eqs. [37a] can be calcu-
For the analyses presented herein, the proposed low-cycle fatigue lated with the following equations:
and rupture model has been shown to be more accurate that a
previously proposed model, while avoiding potential issues such fu − fy
Ω1 ¼ 0.069 þ 0.0555 ð38Þ
as the undesired distortion of the monotonic curve by the low- fu
cycle fatigue law. Given that one of the two material parameters
(i.e., parameter t) of the low-cycle fatigue model does not sig-  
ðΩ3 − Ωj ÞðP − 1Þ Ω1
nificantly affect the obtained results, there is a single additional Ω2 ¼ þ Ωj ð39Þ
84 þ P Ωj
parameter, Dcr , that essentially requires calibration to account for
low-cycle fatigue. The fact that the same value of Dcr has given where P = parameter appearing in the equation describing the hard-
good results for both monotonic and cyclic tests indicates that Dcr ening regime of the monotonic stress-strain curve, as explained in
can be calibrated from the results of monotonic tensile tests. Dodd and Restrepo-Posada (1995)
A major advantage of the proposed buckling and low-cycle
fatigue models is that they are generic in the sense that they are fu − fy
not specific to the hysteretic model proposed in this paper. Thus, Ω3 ¼ 0.0753 þ 0.0691 ð40Þ
fu
the buckling equations and the low-cycle fatigue criterion described
herein can be combined with any other model for reinforcing steel
md ¼ 17034εy − 85.66 ð41Þ
(e.g., Chang and Mander 1994; Hoehler and Stanton 2006; or Kun-
nath et al. 2009a). where εy = equal to the ratio of f y over Es
β 1 ¼ 10.1ðΩ1 − 0.04Þ ð42Þ
Conclusions
β 2 ¼ 10.1ðΩj − 0.04Þ ð43Þ
A phenomenological uniaxial material model has been presented
for capturing the response of reinforcing steel under cyclic load-
ing. A previously proposed hysteretic law has been used and ap- β 3 ¼ md ðΩ2 − Ω3 Þ ð44Þ
propriately enhanced to eliminate the need for an iterative stress
update procedure. This has been accomplished by describing the The parameter uΩ in Eq. (37c) is obtained from the following
reversal curves with nonuniform rational b-splines (NURBS). The equation:
modified hysteretic law has also yielded a closed-form expression qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
for the material tangent stiffness, which was not feasible in the ½2ðΩ1 − Ωj Þ þ ½2ðΩj − Ω1 Þ2 − 4ðΩ2 − 2Ωj þ Ω1 ÞðΩ1 − ΩÞ
original model. Validation analyses demonstrate that the noniter- uΩ ¼
2ðΩ2 − 2Ωj þ Ω1 Þ
ative version of the model gives a practically identical stress-strain
response as the original, iterative version. The enhanced material ð45Þ
law has also been provided with the capability to efficiently The additional parameter, Ωj , in Eq. (45) is obtained using the
account for the effect of inelastic buckling using a simple, yet following expression:

© ASCE 04016133-12 J. Struct. Eng.

J. Struct. Eng., 04016133


Fig. 13. Variation of sectional tangent stiffness coefficients with tangent-over-unloading modulus ratio, η

0.404 − Ω3 md The following dimensionless coefficients are then defined:


Ωj ¼ ð46Þ
10.1 − md    
Downloaded from ascelibrary.org by Heriot-Watt University on 07/20/16. Copyright ASCE. For personal use only; all rights reserved.

γ T;11 γ T;12 D̂T;11 =½π · Es ðD=2Þ2  D̂T;12 =½π · Es ðD=2Þ3 


¼
The correlation between β and Ω has been obtained through a γ T;21 γ T;22 D̂T;21 =½π · Es ðD=2Þ3  D̂T;22 =½π · Es ðD=2Þ4 
curve-fitting procedure conducted by Kim (2014).
The values of the dimensionless coefficients have been obtained
for a range of values of the ratio η. A curve-fitting process gave the
Appendix II. Sectional Tangent Stiffness Coefficients following expressions for the sectional stiffness coefficients:
for Buckling Model pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
0.04 − 0.0016 þ 3.84η
γ T;11 ¼ ð49Þ
The four coefficients γ T;11 , γ T;12 , γ T;21 , and γ T;22 that are used in −1.92
Eq. (21) to provide the sectional tangent stiffness terms can be ob-
tained as functions of parameter η, which is the ratio of the tangent −0.546ðu12 − u212 Þ
γ T;12 ¼ γ T;21 ¼ ð50Þ
modulus, ET , over the unloading modulus of the material. The latter 0.44u212 − 0.44u12 þ 1
is assumed to be equal to the initial elastic modulus, Es . As shown
in Fig. 4(c), the sectional deformations dεo and dφo lead to a linear
where u12
incremental strain distribution over the sectional depth. The loca- pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
tion ȳ of the neutral axis, N.A., in Fig. 4(c) can be obtained by 0.44η þ 0.3432 − ð0.44η þ 0.3432Þ2 − 4ηð0.44η − 0.6568Þ
¼
enforcing the condition that the change in the sectional axial force 2ð0.44η − 0.6568Þ
is zero. This condition leads to the following equation:
ð51Þ
Z D=2−ȳ
ðȳ − yÞ
− ET · Δεc bðyÞdy
−D=2 ȳ þ D=2 γ T;22 ¼ −0.03u222 þ 0.28u22 ð52Þ
Z D=2
ðy − ȳÞ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
þ Es · Δεt bðyÞdy ¼ 0 ð47Þ 0.3 − 0.09 þ 2.8η
D=2−ȳ D=2 − ȳ where u22 ¼ ð53Þ
−1.4
If the equation ET =Es ¼ η is substituted in Eq. (47), the follow- These expressions provide excellent estimates of the actually
ing expression is obtained: obtained dimensionless coefficients over the entire range of η
Z D=2−ȳ Z D=2 values, as shown in Fig. 13. The exact values in the figure have
ðȳ − yÞ ðy − ȳÞ
− η · Δεc bðyÞdy þ Δεt bðyÞdy ¼ 0 been determined for 10,000 different values of η, uniformly spaced
−D=2 ȳ þ D=2 D=2−ȳ D=2 − ȳ
between 0 and 1.
ð48Þ

Eq. (48) provides the location ȳ for a circular section and for a References
given value of η. The sectional tangent stiffness terms can then be
obtained using the following set of equations: Bae, S., Mieses, A. M., and Bayrak, O. (2005). “Inelastic buckling of
Z D=2 reinforcing bars.” J. Struct. Eng., 10.1061/(ASCE)0733-9445(2005)
131:2(314), 314–321.
D̂T;11 ¼ ET ðyÞbðyÞdy Balan, T. A., Filippou, F. C., and Popov, E. P. (1998). “Hysteretic model of
−D=2
Z D=2−ȳ
Z D=2
ordinary and high strength reinforcing steel.” J. Struct. Eng., 10.1061/
¼ ET ðyÞbðyÞdy þ Es ðyÞbðyÞdy (ASCE)0733-9445(1998)124:3(288), 288–297.
−D=2 D=2−ȳ Bayrak, O., Sheikh, S. A. (2001). “Plastic hinge analysis.” J. Struct. Eng.,
Z Z  10.1061/(ASCE)0733-9445(2001)127:9(1092), 1092–1100.
D=2−ȳ D=2
¼ Es η · bðyÞdy þ bðyÞdy Bazant, Z., and Cedolin, L. (2000). Stability of structures, 2nd Ed., Dover,
−D=2 D=2−ȳ New York.
Chang, G. A., and Mander, J. B. (1994). “Seismic energy based fatigue
Z Z  damage analysis of bridge columns. Part I—Evaluation of seismic
D=2−ȳ D=2
D̂T;12 ¼ D̂T;21 ¼ −Es y · η · bðyÞdy þ y · bðyÞdy capacity.” Rep. No. NCEER-94-0006, National Center for Earthquake
−D=2 D=2−ȳ Engineering Research, State Univ. of New York at Buffalo, Buffalo, NY.
Cosenza, E., De Cicco, F., and Prota, A. (2010). “Discussion of ‘Nonlinear
Z Z  Uniaxial Material Model for Reinforcing Steel Bars’ by Sashi K.
D=2−ȳ D=2
D̂T;22 ¼ Es y2 · η · bðyÞdy þ y2 · bðyÞdy Kunnath, YeongAe Heo, and Jon F. Mohle.” J. Struct. Eng., 10.1061/
−D=2 D=2−ȳ (ASCE)ST.1943-541X.0000119, 917–918.

© ASCE 04016133-13 J. Struct. Eng.

J. Struct. Eng., 04016133


Dhakal, R. P., and Maekawa, K. (2002). “Path-dependent cyclic stress- Kunnath, S. K., Kanvinde, A., Xiao, Y., and Guowei, Z. (2009b). “Effects
strain relationship of reinforcing bars including buckling.” of buckling and low-cycle fatigue on seismic performance of reinforc-
Eng. Struct., 24(11), 1383–1396. ing bars and mechanical couplers for critical structural members.”
Dodd, L. L., and Restrepo-Posada, J. I. (1995). “Model for predicting cyclic Rep. No. CA/UCD-SESM-08-01, Univ. of California, Davis, CA.
behavior of reinforcing steel.” J. Struct. Eng., 10.1061/(ASCE)0733- Ma, S.-Y. M., Bertero, V. V., and Popov, E. P. (1976). “Experimental
9445(1995)121:3(433), 433–445. and analytical studies on the hysteretic behavior of reinforced concrete
Downing, S. D., and Socie, D. F. (1982). “Simple rainflow counting rectangular and T-beams.” EERC Rep. No. 76-02, Earthquake Engineer-
algorithms.” Int. J. Fatigue, 4(1), 31–40. ing Research Center, Univ. of California, Berkeley, CA.
Falk, W. M., and Govindjee, S. (1999). “Towards an understanding of Maekawa, K., Pinanmas, A., and Okamura, H. (2003). Nonlinear mechan-
the observed buckling modes of reinforcing bars in concrete columns.” ics of reinforced concrete, Spon Press, New York.
Rep. No. UCB/SEMM-1999/10, Univ. of California, Berkeley, CA. Mander, J. (1983). “Seismic design of bridge piers.” Ph.D. disser-
Filippou, F. C., Popov, E. P., and Bertero, V. V. (1983). “Effects of bond tation, Dept. of Civil Engineering, Univ. of Canterbury, Christchurch,
deterioration on hysteretic behavior of reinforced concrete joints.” New Zealand.
Rep. No. EERC 83-19, Earthquake Engineering Research Center, Univ. Massone, L. M., and Moroder, D. (2009). “Buckling modeling of reinforc-
of California, Berkeley, CA. ing bars with imperfections.” Eng. Struct., 31(3), 758–767.
Downloaded from ascelibrary.org by Heriot-Watt University on 07/20/16. Copyright ASCE. For personal use only; all rights reserved.

Gomes, A., and Appleton, J. (1997). “Nonlinear cyclic stress-strain Mendes, L. A. M., and Castro, L. M. S. S. (2014). “A simplified reinforcing
relationship of reinforcement bars including buckling.” Eng. Struct., steel model suitable for cyclic loading including ultra-low-cycle fatigue
19(10), 822–826. effects.” Eng. Struct., 68(1), 155–164.
Hoehler, M. S., and Stanton, J. F. (2006). “Simple phenomenological model Monti, G., and Nuti, C. (1992). “Nonlinear cyclic behavior of reinforcing
for reinforcing steel under arbitrary load.” J. Struct. Eng., 10.1061/ bars including buckling.” J. Struct. Eng., 10.1061/(ASCE)0733-9445
(ASCE)0733-9445(2006)132:7(1061), 1061–1069. (1992)118:12(3268), 3268–3284.
Huang, Y., and Mahin, S. A. (2010). “Simulating the inelastic seismic OpenSees version 2.4.4 [Computer software]. Pacific Earthquake Engineer-
behavior of steel braced frames including the effects of low-cycle ing Research Center, Berkeley, CA.
fatigue.” PEER Rep. No. 2010/104, Pacific Earthquake Engineering Piegl, L., and Tiller, W. (1997). The NURBS book, 2nd Ed., Springer,
Center, Univ. of California, Berkeley, CA. New York.
Kent, D., and Park, R. (1973). “Cyclic load behavior of reinforcing steel.” Restrepo-Posada, J. I. (1993). “Seismic behavior of connections between
Strain, 9(3), 98–103. precast concrete elements.” Ph.D. dissertation, Dept. of Civil Engineer-
Kim, S. (2014). “Cyclic uniaxial constitutive model for steel reinforce- ing, Univ. of Canterbury, Christchurch, New Zealand.
ment.” Master’s thesis, Virginia Polytechnic Institute and State Univ., Restrepo-Posada, J. I., Dodd, L. L., Park, R., and Cooke, N. (1994).
Blacksburg, VA. “Variables affecting cyclic behavior of reinforcing steel.” J. Struct.
Kunnath, S., Heo, Y., and Mohle, J. (2010). “Closure to ‘Nonlinear Uni- Eng., 10.1061/(ASCE)0733-9445(1994)120:11(3178), 3178–3196.
axial Material Model for Reinforcing Steel Bars’ by Sashi K. Kunnath, Rodriguez, M. E., Botero, J. C., and Villa, J. (1999). “Cyclic stress-strain
YeongAe Heo, and Jon F. Mohle.” J. Struct. Eng., 10.1061/(ASCE)ST behavior of reinforcing steel including effect of buckling.” J. Struct.
.1943-541X.0000197, 918–920. Eng., 10.1061/(ASCE)0733-9445(1999)125:6(605), 605–612.
Kunnath, S. K., Heo, Y., and Mohle, J. F. (2009a). “Nonlinear uniaxial Urmson, C. R., and Mander, J. B. (2012). “Local buckling analysis of
model for reinforcing steel bars.” J. Struct. Eng., 10.1061/(ASCE) longitudinal reinforcing bars.” J. Struct. Eng., 10.1061/(ASCE)ST
0733-9445(2009)135:4(335), 335–343. .1943-541X.0000414, 62–71.

© ASCE 04016133-14 J. Struct. Eng.

J. Struct. Eng., 04016133

Potrebbero piacerti anche