Você está na página 1de 26

MECHANICS OF COHESIVE-FRICTIONAL MATERIALS

Mech. of Cohes.-Frict. Mater. 3, 155180 (1998)

Development of a plasticity bond model for steel reinforcement

James V. Cox1* and Leonard R. Herrmann2


1Department of Civil Engineering, Whiting School of Engineering, Johns Hopkins University, Baltimore, MD 21218, U.S.A.
2Department of Civil and Environmental Engineering, University of California, Davis, CA 95616, U.S.A.

SUMMARY
The development of an interface, non-associative, plasticity model for bond between ribbed, steel bars and
concrete is discussed. The model relates average local slip and radial dilation to average bond shear stress
and radial confinement stress. The model partially accounts for the response of the damaged, finite-thickness
region around the barthe bond zone. The model is developed for standard steel bars that are initially
unstrained. With simplifying assumptions, data for the components of a plasticity law are extracted from
a key set of experimental results. In this paper, we emphasize the development of an expression for the yield
surface for monotonic loading. While the forms of the models components are empirically derived, they
qualitatively reflect the mechanics of the mechanical interaction of ribbed bars with the adjacent concrete.
A characteristic length, related to the rib pattern, helps quantify this interaction. The mechanics of the bond
are difficult to characterize in a simple form, but the calibrated model only requires four physical properties
and reproduces with acceptable accuracy experimental results with various levels of radial confinement
stress. Model refinements are suggested for future work. ( 1998 John Wiley & Sons, Ltd.
KEY WORDS: reinforced concrete; bond model; plasticity; interface; yield surface; modelling

1. INTRODUCTION
The motivation and application of bond modelling are multifold. Two of the most important
applications of bond models are in (1) interpreting and understanding experimental test
results and (2) finite element modelling of reinforced concrete members and structures
to characterize the behaviour of interface finite elements used to connect reinforcement to
concrete. The work reported herein should find its most immediate application in the first
category as related to a phenomenological understanding of pullout tests of single reinforcing
bars embedded in concrete specimens. However, it is anticipated that with further development
it will find its greatest utility in the second area (see the preliminary work by Mello and
Herrmann1).
By design, bars are used to reinforce concrete (a quasi-brittle matrix) to both prevent and
bridge cracks. As such, bond behaviour is important in determining the nature of localized
failures and the amount of energy dissipated by reinforced concrete components. Structural
models that totally disregard the effects of bond, can fail to predict localized cracking which can
have global ramifications. The corresponding stiffness degradation increases the period of

* Correspondence to: James V. Cox, Department of Civil Engineering, Whiting School of Engineering, John Hopkins
University, Baltimore, MD 21218, U.S.A. E-mail: James.Cox@jhu.edu.

CCC 1082-5010/98/02015526$17.50 Received 30 December 1996


( 1998 John Wiley & Sons, Ltd. Revised 15 July 1997
156 J. V. COX AND L. R. HERRMANN

vibration, decreases the energy dissipation capacity, and can result in a significant redistribution
of internal forces (see e.g., Bertero2).
A very brief overview of the mechanics of bond under monotonic loading is given below; for
more detailed explanations see, e.g., References 37.

1.1. Mechanics of bond


For plain bars (without ribs), adhesion and friction are the principal mechanisms of bond. The
apparent adhesion has contributions from chemical bonding and the effect of shrinkage stresses
that develop during curing. The failure of this interaction is characterized by the initiation and
propagation of an interfacial crack. (Except for this case, the use of the term bond is an accepted
misnomer which commonly is used to describe the total interaction of the reinforcement with the
concrete.) For ribbed-bars, these mechanisms are secondary (by design) to the mechanical
interaction of the ribs with the surrounding concrete. Adhesion breaks down relatively early in
the bond response, subsequently, the bond force is transferred by friction and the mechanical
interaction of the ribs with the adjacent concrete.
With increasing bar force, the mechanical interaction dominates the transfer of force which is
now concentrated near the rib faces. At this point, bond stress usually refers to an average force
per unit area. Increased loading will begin to fail the concrete near the ribs in two ways: crushing
of concrete adjacent to the contact area and transverse cracking (typically cone-shaped
cracksalso called secondary or bond cracking) that initiate at the ribs. The extent of these
cracks conceptually defines the size of the bond zone.
Researchers often emphasize one mode of failure more than the other in their studies.8,9 For
a given specimen, the principal source of slip depends upon the state of the concrete near the rib.
Since the model presented here is based upon experimental data, the collective effect of all the
bond mechanisms is included, but the explanation of the bond model puts more emphasis upon
the crushing mechanism. Other researchers have explained their models in terms of transverse
cracking alone. With stable propagation of transverse cracks, the concrete near the bar appears to
form inclined struts5 which are sometimes referred to as compression cones or prisms when
idealized without friction.10 The bond stiffness is then characterized by the stiffness of these
inclined struts.
As the loading progresses, radial splitting forces can develop. In terms of transverse cracking,
this is explained by a rotation of the inclined struts which produces a larger radial component of
contact force. In terms of crushing, the increase of radial forces is explained by the deposition of
crushed concrete on the face of the rib which increases the effective contact angle (between the rib
and concrete), the wedging action of the ribs, and the corresponding radial component of the
contact tractions. The net effect is the same; without adequate constraint, a splitting failure can
occur. Thus, the effects of bond can extend outside of the bond zone.
As with a dilatational material, the bond zone dilation (which produces the longitudinal
cracking) occurs near the maximum of the bond stress. The subsequent softening response,
a characteristic of bond slip behaviour, is often explained in terms of a progressive shear failure of
the concrete between the ribs. Under very low confinement stress, geometric changes in the
contact between the rib and concrete can contribute to softening as well. Both of these mecha-
nisms are promoted by a reduction in confinement stress, which can occur with the propagation
of longitudinal cracks. In either case, softening reflects the development of a strong discontinuity
between the bar and the concrete.

( 1998 John Wiley & Sons, Ltd. MECH. COHES.-FRICT. MATER., VOL. 3: 155180 (1998)
PLASTICITY BOND MODEL 157

1.2. Bond models


The cumulative effect of these mechanisms has been modelled as several scales. The following
discussion will use the terminology of unit cell which is a fundamental building block from which
the body of interest can be considered to be constructed. The concept of unit cell is often used in
the construction of material models and will also be used to describe the bond models discussed
here. It should be noted that when such models are used to supply constitutive properties for
finite element analyses of complete structures that the elements used are almost always much
larger than the underlying unit cell.
Three descriptive scale names (shown in Figure 1) are adopted here to classify bond models.
Rib-scale analyses, usually include an explicit discretization of the ribs on the bars. Among the
difficulties in modelling bond at this scale is that the scale is significantly smaller than the unit cell
for concrete, i.e., much of the aggregate is larger than the rib so the homogeneous concrete
assumption is idealistic. In addition, rib-scale bond models are no better than the concrete
models ability to model fracture and crushing failures. While analyses at this scale have not
provided a general analysis capability, they have provided much insight to the mechanics of
bondespecially in the early response. To model bond for large slips at the rib scale would be
numerically challenging since damaged material can change state (i.e. become a powder) and be
deposited on the rib face or into crack openings. Interesting examples of rib-scale analyses include
the studies of Hungspreug,6 Ingraffea et al.,8 Reinhardt, Blaauwendraad and Vos,11 Rots,12
Ozbolt and Eligehausen,13 and Brown, Darwin and McCabe.14

Figure 1. Scales of bond analysis

( 1998 John Wiley & Sons, Ltd. MECH. COHES.-FRICT. MATER., VOL. 3: 155180 (1998)
158 J. V. COX AND L. R. HERRMANN

At the other extreme, the member scale, either a structural component or complete structure is
modelled; this requires a careful selection of the discretization and element types. The reinforcing
is usually treated with a discrete, embedded or smeared model,15 but other novel approaches to
modelling components have also been developed (see e.g. Reference 16). Typically at the member
scale the reinforcement is treated as a one-dimensional element, and bond laws have been limited
to single-stress models (i.e. relating bond shear stress to slip).
At an intermediate scalethe bar scalethe bar and concrete are both treated as continuums,
and the mechanical interaction of the ribs is homogenized and often further idealized as an
interface phenomenon. Phenomenological models at this scale can have the following benefits
over one-dimensional characterizations: (1) ability to predict splitting failures, and (2) greater
sensitivity to the concrete stress statei.e., a greater measure of generality.
These three scales reflect the trilemma one faces in many modelling problems. The practical
problems are orders of magnitude larger than the scale at which the basic mechanics occur, and
an intermediate scale model is a compromise between two extremes.
The objective of this on-going research is to develop and validate a model that will qualitatively
reflect the mechanisms of bond while having greater potential for application than rib-scale
models (e.g., to model structural components). This objective motivates the use of a bar-scale
model which will have a phenomenological nature. The model should require only one calib-
ration for standard steel bars and regular concrete, and be capable of reproducing (with an
accuracy comparable to experimental scatter) bond tests results from several independent
research groups.
The first bar-scale model of a bond specimen was apparently presented by Bresler and
Bertero.17 They introduced a homogenized boundary layer with reduced elastic properties to
account for the damage that occurs near the bar. Since then, other bar-scale models that treat the
bond zone completely as a continuum (i.e. without explicitly including a strong discontinuity
between the bar and concrete) have been proposed; e.g. Pijaudier-Cabot, Mazars and
Pulikowski18 modelled the concrete with an isotropic, non-local damage law and reproduced the
size effects experimentally observed by Bazant and Sener.19
Many researchers have used one-dimensional bond laws for the tangential response of
interface elements, some of which involve numerous material parameters (e.g. Reference 20).
However, these approaches do not characterize the wedging effect of the ribs, so two-dimensional
models, which relate the normal and longitudinal components of the interface tractions to their
conjugate relative displacements, are addressed in this study. Among the existing two-dimen-
sional models are those of De Groot, Kusters and Monnier,10 Morita and Fujii,21 Zhiming,
Hueizhong and Jinping,22 and Mainz, Stockl and Kupfer.23 De Groot, Kusters and Monnier10
and Zhiming, Hueizhong and Jinping22 explicitly modelled the inclined strut. Morita and Fujii21
used inclined link elements, and Mainz, Stockl and Kupfer23 used inclined interface elements to
introduce the wedging effect.
Bond behaviour typified by the classic experiments of Eligehausen, Popov and Bertero7
motivated the use of elastoplasticity as a mathematical framework for the model developed in this
study. This approach is distinctly different than applying plasticity theory to model the constitut-
ive behaviour of a material. For a bond model, the potential slip plane is singular, predefined, and
macroscopic. Note that subsequent to the development of the bond model presented in this study,
the authors found examples from other areas of application where interface models were
formulated within the mathematical framework of plasticity theory (see e.g., References 24
and 25).

( 1998 John Wiley & Sons, Ltd. MECH. COHES.-FRICT. MATER., VOL. 3: 155180 (1998)
PLASTICITY BOND MODEL 159

In the following sections a bar-scale, plasticity model is developed that includes, in addition to
the tangential stress and slip, the radial normal stress and the conjugate relative displacement.
The emphasis of this paper is on the development of a monotonic bond model. Complete
descriptions of a model development for monotonic and cyclic applications are presented in
References 26 and 27, respectively.
The key components of a plasticity model are definitions of the generalized stresses and strains,
elastic moduli, yield criterion and flow rule. Each component is briefly discussed here, but the
emphasis of this paper is the development of the yield criterion. In this paper some calibration
and validation results are briefly presented; additional validation results are available in the
literature, and a detailed examination of the validation results is forthcoming.

2. GENERALIZED STRESSES AND STRAINS


Many researchers have defined a cylindrical region around the barthe bond zonethe radius
of which bounds the crushing and transverse cracks produced by the mechanical interaction. This
conceptual region does not contain: (1) all longitudinal cracks, some of which can split a test
specimen or the cover of a flexural member, or (2) transverse cracks that produce a local failure
such as a pull-out cone. For the model presented here, these large-scale cracks must be modelled
by the global representation of the structural component. The growth of transverse cracks during
a progressive failure of bond suggests that the size of the bond zone evolves, so most definitions of
the bond zone are for the maximum radius of the region. Researchers have suggested different
radii for the bond zone. For example, Gambarova, Rosati and Zasso28 experimentally deter-
mined that the microcracking extended the thickness of the maximum aggregate size.
Bar-scale bond models are phenomenological characterizations that represent all or part of
the behaviour of the bond zone. In the context of finite element modelling, for a bond model to
represent all of the behaviour of the bond zone it would have to be applied over the complete
volume of the bond zone. It is more common to characterize the bond phenomena as if they are
constrained to an interface. Since the thickness of the bond zone is significant compared to the
bars diameter this is clearly a modelling idealization. Gerstle and Ingraffea29 emphasize this by
suggesting that, bond slip may not exist in the sense in which it has been previously described in
the literature. This is also suggested by many rib-scale analyses that reflect the importance of
transverse cracking in bond (see e.g. References 6, 11 and 12). Interface descriptions of bond are
a pragmatic idealization of a very complicated problem, but this idealization is potentially very
useful in larger-scale analyses.
In addition to the interface idealization, the version of the model presented here assumes that
the bond zone response can be approximated as axisymmetric. This does not suggest that the
response outside of the bond zone is necessarily approximately axisymmetric, but it does preclude
application of the model to some problems (e.g. where significant doweling forces exist). This
assumption also implies that the detailed effects of: (1) non-axisymmetric rib patterns (i.e., all rib
patterns except those that have ribs normal to the axis of the bar), (2) non-axisymmetric cracking
of the concrete (e.g. where a small number of dominant longitudinal cracks occur), and (3)
material heterogeneity are implicitly averaged in defining the generalized stresses and strains for
the model.
The interface idealization of bond can be described in terms of two simplifications. The first,
common to all bar-scale models, is the homogenization of the barconcrete interface traction
distribution and the simplification of the interface geometry. Consider a state where no adhesion

( 1998 John Wiley & Sons, Ltd. MECH. COHES.-FRICT. MATER., VOL. 3: 155180 (1998)
160 J. V. COX AND L. R. HERRMANN

remains between the bar and the concrete (i.e. mechanical interaction dominates the bond
behaviour). Figure 2(a) presents a schematic of the horizontal component of the interface traction
(shown on the concrete) between the bar and concrete for a unit surface element. The unit surface
element consists of the complete bar surface (due to the axisymmetry assumption) for one cycle of
the rib geometry. In this case, the ribs are assumed to be perpendicular to the bars axis, so the
length of the unit surface element is the rib spacing (s ). (A cylindrical co-ordinate system is
3
assumed with the z-axis corresponding to the axis of the bar.) Figure 2(b) presents a schematic of
the horizontal component of the interface traction between the bar and concrete for a bar-scale
idealization. The uniform distribution of q shown in Figure 2(b) (macroscopically homogeneous)
would occur when the actual traction distribution for adjacent unit surface elements is identical to
that shown in Figure 2(a). In conjunction with the simplification of the traction distribution, the
actual surface geometry is also idealized as a cylindrical surface. Though not depicted in Figure 2,
the actual interface geometry changes with concrete crushing and cracking. The homogenized
stress is defined so that the average of each traction component in a cylindrical co-ordinate
system is the same; e.g.

s3
2nr
P0 q(z) dz"!PA z (z, h) dA (1)

where A denotes the area of the unit surface area, and denotes the z-component of the actual
z
interface traction. (The q depicted in Figure 1 is negative.) These definitions of homogenized
tractions are consistent with the values sought experimentally; when global measures such as the
bar end force are used to measure traction components only the average values over the full
embedment length are obtained.
The second simplification, unique to interface idealizations, addresses the kinematics of a unit
cell of the bond zone. Figure 3 gives a schematic of the deformation of the actual unit cell versus
the deformation of the same unit cell in the bar-scale model. For the two unit cells, note the
difference in the distribution of the horizontal displacement (e.g., along the left side) relative to the
top of the cell. The elimination of bars ribs and the corresponding traction concentrationsthe
first simplificationproduce a different response in the concrete matrix, even for a perfect
concrete model. While transverse cracking can still occur in the bar-scale model, the omission of
the ribs will tend to produce less cracking and in a more uniformly distributed pattern. Defining
the bar-scale model so that it will produce similar bar displacements requires a concentration of
relative displacement at the interface. That is, reduced shear deformation of the concrete in the
bar-scale model is supplemented by an increase in the strong discontinuity of displacement at the

Figure 2. Idealized distribution of the horizontal traction component over the unit surface element: (a) actual, and (b)
bar-scale model

( 1998 John Wiley & Sons, Ltd. MECH. COHES.-FRICT. MATER., VOL. 3: 155180 (1998)
PLASTICITY BOND MODEL 161

Figure 3. Idealized deformation of the bond zone: (a) actual, and (b) bar-scale model

barconcrete interface. Due to this idealization, the displacement of a point on the bar relative to
a point within the bond zone may not be accurately predicted, even when the absolute displace-
ment of the point on the bar is accurately predicted.
The generalized strains for the bar-scale model are defined to be the tangent and normal
displacements (d and d , respectively) of the concrete relative to that of the bar interface
5 /
non-dimensionalized by the bar diameter (D ). The generalized stresses and strains are thus
"
defined as

1
QT"(q p) and qT" (d d ) (2)
D 5 /
"
where a simple tangentnormal co-ordinate system (the tangent direction corresponding to the
direction of the bars axis) is adopted, and the ds are defined for points on the interface that are
coincident in the undeformed state. (Figure 3(b) depicts a deformation where d (0 and d '0.)
5 /
The generalized stresses and strains are work conjugate within a multiplicative constant, 1/D .
"
Though this is not a unique strain definition, this particular form has effectively incorporated
the diameter of the bar for a variety of test results. Morita and Fujii21 normalized their definition
of slip in this manner for their uniaxial model and demonstrated its appropriateness for their test
results. The strain measure is the first of two definitions that includes a characteristic length
related to the rib geometry. Unlike the case of smooth bars where the propagation of an
interfacial crack is principally a function of the material properties, for ribbed bars the bond
response is dominated by the mechanical interaction which the rib geometry affects. Character-
istic lengths for the rib pattern in the tangent and normal directions are the rib spacing and
height, respectively. For standard steel bars, both of these lengths are nearly proportional to the
bar diameter, and thus a single parameter of bar diameter was adopted for this study. Other rib
geometry parameters (e.g. rib anglethe angle of the rib with respect to the bars axis) were also
considered.27
A thorough discussion of the potential limitations of bar-scale models is beyond the scope of
this paper, however, a few of the limitations that relate to the deformation of the bond zone
should be noted. Comparisons between models and experiments are usually based upon global
responses such as bar force and end slipspartly because it is difficult to measure local quantities.

( 1998 John Wiley & Sons, Ltd. MECH. COHES.-FRICT. MATER., VOL. 3: 155180 (1998)
162 J. V. COX AND L. R. HERRMANN

Any slip measurement that is defined as a relative displacement between the surface of the bar and
a reference point within the bond zone may be predicted inaccurately by a bar-scale model. The
potential inaccuracy is due to the difference in the deformation of the actual bond zone versus
that of the bar-scale model. However, the same model can accurately predict slip based upon
a reference point outside of the bond zone. The influence of the deformation of the concrete
matrix might require the bond model to be recalibrated for different concrete models. Further-
more, detailed rib-scale behaviours that produce the arrest and propagation of different trans-
verse cracks (see e.g. Reference 12) is beyond the scope of bar-scale models. These and other
limitations are currently under investigation.

3. ELASTIC MODULI
The elastic response deviates from elastic behaviour since it accounts for the underlying contact
problem between the ribs and adjacent concrete after the breakdown of adhesion. Analysis of the
experimental data of Malvar30,31 suggest that very little elastic coupling occurs between the
tangential and normal response. This is consistent with previous experimental observations (e.g.
Reference 4) that indicated significant wedging action of the ribs occurs only after preliminary
crushing near the rib face. Experimental results also indicate the presence of elastoplastic
coupling which is not accounted for in the current model.
For the current model,26 the elastic response under monotonic loading is characterized by the
following incremental moduli:

C D
0)1 !0)0012 sgn(d )
D%"E 5 (3)
# !0)0012 sgn(d ) 0)04
5
where Q0 "D%q5 %, sgn(0),0, and E is the elastic modulus of concrete. The sgn(d ) factor reflects
# 5
the contact character of the interface between the inclined rib surface and the adjacent concrete,
accounting for: the slip direction and the contact symmetry at zero slip (i.e. an incremental
decrease in d does not produce a generalized shear stress when the slip is zero). The
/
implementation accounts for finite precision and treats d as zero if D d D(e, where e denotes a very
5 5
small tolerance.
Magnitudes of the individual moduli were obtained principally from the experimental data of
Malvar,30,31 Gambarova, Rosati and Zasso,28 and Eligehausen, Popov and Bertero.7 As noted
previously for the total response, the elastic response in the model is the difference between the
elastic response of the actual unit cell and the response of the concrete in the bar-scale unit cell
(Figures 2 and 3). Thus, the response is attributed principally to the local interaction of the ribs
and adjacent concrete which is consistent with the observed variation of the elastic response with
rib geometry. The elastic modulus, E , provides a simple measure of the concretes elastic
#
properties upon the elastic response of the model.

4. YIELD SURFACE
In this section we present: (1) the data analysis (with simplifying assumptions) and (2) the resulting
yield criterion of the modelthe yield surface and its evolution. Trends in the data and
corresponding features of the model are discussed in terms of bond mechanics.

( 1998 John Wiley & Sons, Ltd. MECH. COHES.-FRICT. MATER., VOL. 3: 155180 (1998)
PLASTICITY BOND MODEL 163

4.1. Experimental data


Most bond experiments give some measure of q and d , but few experiments provide measures
5
of p and d . The experimental works of Gambarova, Rosati and Zasso28 and Malvar30,31 are
/
among the exceptions; they give some measure of all the generalized stresses and strains of
equations (2) and thus served as an experimental basis for developing this model. Both specimens
are shown schematically (to scale) in Figure 4. Solid arrows represent applied forces while the
dashed arrows represent reaction forces.
The experiments of Gambarova, Rosati and Zasso28 consisted of three test seriesA, B and C;
only series C is considered in this study. All specimens and loadings had a somewhat
two-dimensional nature. The specimens were fabricated with a splitting crack through their
widths. The crack openings were monitored by mechanical gauges near the points labeled a and
b in Figure 4. A series of four tests was performed where the crack opening was manually
controlled by confinement forces at the top and bottom of the specimen to maintain constant
openings of 0, 0)1, 0)2 and 0)3 mm. The confinement forces applied to maintain the constant crack
openings gave a measure of p, and the crack opening gave a measure of d . Assuming the tractions
/
were axisymmetric and were uniformly distributed over the surface of the bar, with respect to
z and h, the traction components can be easily determined from equilibrium. The assumption of
an axisymmetric, uniform traction distribution is a first-order approximation to a very
complicated distribution that depends upon the crack opening. A simple model for the geometric
effect of the crack opening is presented in Reference 27.
The experiments of Malvar30,31 reverse the boundary conditions, controlling the applied
radial traction and measuring the change in circumference of the specimen. Unlike the tests of
Gambarova, Rosati and Zasso,28 the specimen is cylindrical, and thus higher normal stresses can
be applied. The reported confinement stress for these tests represents the average magnitude of
the normal traction component (!p) at the barconcrete interface assuming the specimen carries
no hoop stress (i.e. assuming several splitting cracks are open). These small specimens have
diameters of 7)62 cm (3 in), lengths of 10)16 cm (4 in), bar rib spacings (s ) ranging from 12)2 to
3
12)8 mm, embedment lengths of 5s , and compressive strengths ranging from 38)4 to 40)2 MPa.
3
Malvar30,31 performed four series of tests; in this paper, the second and third series are
emphasized. (These two series differ principally in the bar rib patterns, with rib angles of 68 and
90, respectively.) Malvars goal was to quantitatively demonstrate the dependence of bond

Figure 4. Schematics of the experimental specimens considered in the preliminary model development: (a) Malvar30,31;
and (b) Gambarova, Rosati and Zasso28

( 1998 John Wiley & Sons, Ltd. MECH. COHES.-FRICT. MATER., VOL. 3: 155180 (1998)
164 J. V. COX AND L. R. HERRMANN

Figure 5. Bond stress versus slip results of Malvar30,31 series 2 tests

stress-slip behaviour on the value of confinement stress. The specimens were loaded to splitting
and then unloaded prior to measuring the bond stress-slip relationship during reloading.
The initial development of the yield surface focused on the series C data of Gambarova, Rosati
and Zasso.28 The stress paths of their four tests were almost linear up to the maximum bond
stress. Motivated by simplicity, a model with a Coulomb yield surface and corresponding
evolution law was initially considered. For a given calibration, the model would only accurately
reproduce the calibration data itself. Clearly, a single stress path only gives a single point on the
yield surface for each internal state (assuming plastic response) and thus does not indicate the
shape of the yield surface. To attain a better understanding of the yield surface shape, the data of
Malvar30,31 was examined.
Malvar30,31 designed the geometry of his specimens to represent a small segment of bond zone
(i.e. comprised almost entirely of 5 unit cells of the bond zone). In addition, his tests were
performed at various levels of p,s thus if a simple measure of the internal state of the material is
assumed, these data can be used to construct a yield criterion. Figure 5 shows the graphs of bond
stress versus slip for test series 2, where the confinement stress for tests 15 was 3)45, 10)3, 17)2,
24)1 and 31)0 N/mm2, respectively. Each test was performed just once, so the only measure of
scatter is the consistency in the response trends (with increasing !p). Notice that at the scale
shown the initial elastic response does not vary significantly with p.
With a few simplifying assumptions, the experimental relationship between bond stress and
plastic slip can be determined. The initial estimate of D% , based on the data of Malvar,30,31
11
Gambarova, Rosati and Zasso,28 and Eligehausen, Popov and Bertero,7 was 0)04E . Consider
#
the loading history of the specimens after the initial cracking phase: When the external
confinement stress was applied the bond stress and slip were both zero, but !p (at the bar
surface) did not reach its maximum until the splitting cracks were reopened. Even though the
history of !p is unknown, if we assume that D% is very diagonally dominant and that
elastoplastic coupling is not significant in this context, then we can easily approximate the plastic

s The assumption of p representing the generalized stress at the bar surface will be discussed further.

( 1998 John Wiley & Sons, Ltd. MECH. COHES.-FRICT. MATER., VOL. 3: 155180 (1998)
PLASTICITY BOND MODEL 165

Figure 6. Bond stress versus plastic slip estimates for Malvar30,31 series 2 tests

slip (q1 ) as a function of q as


1
q
q1 "q ! #q10 (4)
1 1 D% 1
11
where q10 is the plastic slip which occurred during the initial splitting phase of the loading.31
1
Figure 6 shows bond stress versus plastic slip for the series 2 tests. For brevity, the series
3 results are not shown; these tests, which have ribs normal to the bar axis: (1) exhibit a higher
value of D% , (2) require less plastic slip to split the specimen, and (3) generally attain higher bond
11
stresses. Despite these differences, the initial version of the model does not include the rib angle as
a model parameter.
Once the slip progresses the length of a unit cell, the concrete between the ribs has failed by
a combination of crushing and shear or has been wedged radially outward by the ribs; thus only
frictional behaviour should remain. Many researchers have observed this behavior
experimentally (see e.g. Reference 7). Eligehausen, Popov and Bertero7 introduced a damage
factor, d, that determines the size of their monotonic envelope; this damage factor is a function of
the total energy dissipated by the bond mechanisms less half of the energy dissipated by friction
(which is assumed to be dissipated as heat). This suggests a work hardening law might be useful
for a plasticity bond model,t but for simplicity a strain hardening law is adopted here. Assuming
the plastic radial dilation is significantly smaller than the plastic slip, essentially E q1 E is used as
=
a measure of the plastic strain. Based on experimental observations, we also assume that plastic
slip greater than the rib spacing (s ) causes no further evolution of the yield surface. The bond
3
zone damage is then defined as

A B
d1
d"min 5,1 (5)
s
3

t Early versions of the model presented here incorporated work hardening laws. They displayed no obvious advantages so
the theme of simplicity prevailed.

( 1998 John Wiley & Sons, Ltd. MECH. COHES.-FRICT. MATER., VOL. 3: 155180 (1998)
166 J. V. COX AND L. R. HERRMANN

where d1"q1 D . Though the term damage is used both in this study and in that of Reference 7, its
5 1 "
definition and application differ. The use of the term damage does not imply that continuum
damage mechanics is used in the formulation. Plasticity theory is adopted as the mathematical
framework, but the model addresses the behaviour of a damaged regionthe bond zone.
By their definitions q1 and d are linearly related (for a given bar). Thus for a given q1 , Figure 6
1 1
gives points in stress space that have the same internal state (i.e. value of d). Actually, p at the bar
surface is not known accurately because the specimen can carry hoop stress at two stages: (1)
initially until the wedging action of the ribs opens the existing splitting cracks and (2) near the end
of the test when the cracks may have closed. For now, assume the cracks are open and p is known.
Assuming that (1) the effect of the different rib patterns (for the different test series) can
be neglected and (2) d alone is sufficient to characterize the internal state of the material,
Figure 7 presents loci of points on the yield surface at various stages of evolution. Neither of these
assumptions are true, but they are considered to be sufficiently accurate for the initial version of
the model. (These figures also show the bond models yield surface, as a solid curve, for the same
stages of evolution.) The overall scatter of the data is greater for smaller values of damagewhere

Figure 7. Yield surface data extracted from the results of Malvar30,31 at: (a) plasticity initiation; (b) q ; (c) q1 "0)3; and
.!9 1
(d) the frictional stage

( 1998 John Wiley & Sons, Ltd. MECH. COHES.-FRICT. MATER., VOL. 3: 155180 (1998)
PLASTICITY BOND MODEL 167

Figure 7. Continued

fracture mechanisms play a more significant role. The stresses are non-dimensionalized by f ,
5
a measure of the tensile strength of the concrete. For this study, the tensile and compressive
strengths were related by f "0)3 f 2@3 N/mm2 (Reference 7).
5 #
The plastic initiation surface (Figure 7(a) contains the most subjective data. It is difficult to
define when plastic slip actually begins. We assumed the initial plastic response coincides with the
proportional limit. The values for p"0 were obtained from two uncracked specimens of series
1 and 2; obviously p was compressive, and the actual data would be shifted to the right. The
intersect of the models curve with the horizontal axis was initially obtained from the solution of
a thick-walled cylinder subjected to a tensile traction on the outer surface (see Reference 26 for
details), but for an interface model, tensile stresses are not justified after adhesion fails. Most of the
remaining data points are for specimens that were presplit so their !p values overestimate the
actual values until the splitting cracks have opened. As plotted, all of the points ignore the effect of
the hoop stress in the specimen.

The exceptions are the preliminary tests of series 1.

( 1998 John Wiley & Sons, Ltd. MECH. COHES.-FRICT. MATER., VOL. 3: 155180 (1998)
168 J. V. COX AND L. R. HERRMANN

Figure 7(b) shows the locus of points corresponding to q (actually the maximum bond
.!9
stresses do not occur at the same internal state; the state of the model shown here corresponds to
the state of maximum isotropic hardening). Note that for high confinement stresses, the bond
stresses attained by the series 3 bars are greater than those of the series 2 bars.
For the remaining states considered, the bond stresses gradually decrease until frictional
behaviour dominates the response. In this transition the data suggest that the yield surface
changes shape. As the plastic slip increases, the data suggest that the yield surface might form
a closed curve for high confinement stresses; i.e., at high confinement stress the slope starts
positive (with respect to !p/f ) and may become negative. All of the data suggest a convex yield
5
surface.

4.2. Yield surface model


During the course of this research, many different analytical forms of yield criteria were
considered. This data clearly show that the linear form of the Coulomb criterion is not accurate
over a large range of confinement stresses. Most of the yield criteria considered could be expresed
as linear combinations of the following two criteria (the exponential criterion and power
criterion, respectively):

KK
q
"C (d) [1!e!a% (!p/f5#pL (d))] (6a)
f %
5

KK K K A B
q !p a1 !p
"C (d) #pL (d) sgn #pL (d) (6b)
f 1 f f
5 5 5
where
C , C 3C1(0, 1)isotropic hardening and softening functions;
1 %
pL 3C1(0, 1]kinematic softening function; and
a and a calibration parameters.
% 1
Consistent with the assumptions made in extracting the yield surface data, the evolution of the
criterion is expressed in terms of d alone. The isotropic and kinematic hardening/softening
functions modify the size and positions of the yield surfaces, respectively.
The stresses are non-dimensionalized by a measure of the tensile strength of the concrete. Other
researchers have scaled the stresses by various powers of the compressive strength (see e.g.
References 7 and 28) which often correlate well with the tensile strength.
The exponential criterion, equation (6a), can be calibrated to produce a reasonable representa-
tion of the experimental data for large values of damage (d'0)5, see e.g. Figure 7(d)). The data for
large d suggest a function: (1) with a finite initial (i.e. for small !p) slope, and (2) that
asymptotically approaches a constant value. For the case of large d and small !p, Coulombs
criterion can be interpreted as a Taylor series approximation (retaining only the linear term) to
this criterion. Physically this implies that friction is the sole mechanism only at relatively low
confinement stresses. For higher confinement stress, the effective coefficient of friction (the secant
to the point of the yield surface) is less. This reduction in the effective coefficient of friction can be
attributed to the omitted hoop stress. (This will be discussed further in the next section.)
The exponential criterion can also be calibrated to give a good representation of the data for
smaller values of damage, but to do so requires the kinematic softening to occur very slowly. In
addition, the apparent form of the kinematic softening is relatively difficult to express analytically.

( 1998 John Wiley & Sons, Ltd. MECH. COHES.-FRICT. MATER., VOL. 3: 155180 (1998)
PLASTICITY BOND MODEL 169

We associate the kinematic softening principally with the geometric effect of the rib wedging
under the concrete. This relative motion increases the effective angle of contact (referencing the
bar axis) at the ribconcrete interface. This mechanism produces a rapid decrease in bond stress
under low confinement stress. For the kinematic softening function to reflect this behaviour, it
must rapidly translate the yield surface but not significantly affect the response at higher !p.
Thus, the surface must have a high initial slope that subsequently reduces with increasing
!pthe motivation for the power criterion, equation (6b).
The power criterion, equation (6b), can be calibrated to produce a reasonable representation of
the experimental data for small values of damage (d(0)4, see e.g. Figures 7(a) and 7(b)); it models
the data well in this range because it has the desired shape characteristics mentioned in the
previous paragraph (for a (1). Note that a "1 the criterion reduces to Coulombs criterion.
1 1
Though the power criterion models the data well for small d, it does not model the data well for
large d. Its infinite initial slope causes an over prediction of the bond stress at low !p for large
d (i.e. when the behaviour should be dominated by friction).
It is not surprising that one form of the yield surface does not model the data well over the full
range of damage. For relatively small values of damage, the plastic behaviour (i.e. permanent slip
and radial dilation) is attributed mostly to the mechanical interaction of the ribs. For relatively
large values of damage, friction becomes an important component of the plastic behaviour since
most of the slip is attributed to the real, strong discontinuity. The progressive change in the form
of the yield surface data reflects the change in the dominant failure mechanisms. Thus, the
adopted yield criterion transitions from a power criterion to an exponential criterion as plastic
loading progresses, i.e.,

KK G K K A BH
a
q !p 1 !p
"C(d) (d)[1!e!a% (!p/f5#pL (d))]#M(1! (d)) #pL (d) sgn #pL (d) (7)
f % % f f
5 5 5
where
3C1[0, 1]weighting function for the exponential component, also called the transition
%
function; and
Ma calibration constant.
To exploit the strengths of each criterion, the weighting function should change from 0 to 1.
Choosing a simple cubic function for the transition gives as
%
0, d3[0, d ]
0

A B A B
d!d 2 d!d 3
(d)" 3 0 !2 0 , d3(d , d ) (8)
% d !d d !d 0 1
1 0 1 0
1, d3[d , 1]
1
where the interval (d , d ) is obtained by calibration. For the current calibration (d , d )"(0)38,
0 1 0 1
0)53). While a more gradual transition could be defined, the effect of the transition was minimized
by calibrating the exponential and power components of the yield criterion to be in good

This geometric effect will be referred to as geometric dilation in subsequent discussions. Use of the term dilation here
reflects a bar-scale characterization of the rib-scale contact problem. We are specifically addressing the instability
associated with the rigid-body change of configuration between the rib face and adjacent concrete (assuming the concrete
in the bond zone is divided by numerous longitudinal cracks).

( 1998 John Wiley & Sons, Ltd. MECH. COHES.-FRICT. MATER., VOL. 3: 155180 (1998)
170 J. V. COX AND L. R. HERRMANN

agreement in the interval (d , d ); the largest disparity occurs at small confinement stresses since
0 1
the two components have such different initial slopes.
As previously discussed, kinematic softening is associated principally with geometric dilation,
which occurs very rapidly. As such, this function is intended to capture the brittle nature of the
bond mechanisms that dominate under low confinement stress. The following function represents
the kinematic softening:

G A B
d !d b
pL 2 , d3[0, d )
pL (d)" 0 d
2
2 (9)
0, d3[d , 1]
2

where
d the damage value at which kinematic softening stops;
2
ba calibration constant typically greater than 1; and
pL the intersect of the yield surface with the p/f -axis when d"0.
0 5
To determine a reasonable value for d , consider a very simplified model of the wedging action
2
of the ribs that only addresses geometric dilation. Assume that the face of the rib has the shape of
a 90 arc and that the maximum radial expansion of the bond zone occurs when the point of
contact between the concrete and the rib is at the top of the arc-shaped face. For this model the
ability of the rib to transmit transverse force would reach a minimum when the slip is equal to the
rib height (i.e. when the normal to the contact surface reaches 90); thus a reasonable value for
d would be the ratio of the rib height to the rib spacing (h /s ). For typical steel reinforcement,
2 3 3
this ratio is O(1/10), so the calibrated value of d is set to 0)1.
2
Two parameters must still be determined to fully define the kinematic softening. As previously
mentioned, a thick-walled cylinder analysis initially gave pL "2)25, however, p'0 is not
0
physically meaningful after the breakdown of adhesion. Thus, pL must be interpreted as a fitting
0
parameter that defines the magnitude of the kinematic softening. For the proposed form of the
kinematic softening function, C1(0, 1) is only maintained when b'1. The experimental data of
Malvar30,31 had maximum bond stresses at an average damage value of d+0)058. Calibrating
the model at maximum bond stress gave pL (0)058)+0)02; solving equation (9) for b gave b+2)7.
The kinematic softening function produces a variation in the slip at which q occurs. Figure
.!9
5 shows that the slip at which q occurs increases with !p. Figure 6 indicates that while much
.!9
of this slip increase is elastic, the plastic slip increase is significant too. Despite the above use of an
average damage value to determine b, it is apparent from Figure 6 that the damage value for
q increases with increased confinement stress; i.e., the bond behaviour appears less brittle as the
.!9
confinement stress increases. The interaction of the kinematic softening and isotropic hardening
can capture this behaviour. Another important factor in this behaviour is the change in
confinement stress (i.e. the bond stress can increase even during isotropic softening when !p is
increasing).
As previously discussed, the power and exponential forms of the yield criterion, equations (6),
provide reasonable models for small and large values of damage, respectively. The transition
interval defined in equation (8) was easily determined while calibrating the isotropic harden-
ing/softening function. Each criterion was fit to the experimental data over the full range of
damage. The damage interval of (0)38, 0)53) was selected for transition. The quantities a and
%
a were assumed to be constants (0)27 and 0)4, respectively) and were subjectively selected based
1
( 1998 John Wiley & Sons, Ltd. MECH. COHES.-FRICT. MATER., VOL. 3: 155180 (1998)
PLASTICITY BOND MODEL 171

on linear least-squares fits (with respect to C(d)) of the functions to the experimental data over the
applicable ranges of damage. The parameter M was assigned the value of 1)38; this parameter
scales the power component so that a single isotropic hardening/softening function can be used
(i.e., instead of requiring both C and C ). Linear least-squares fits of the model to the experi-
% 1
mental data (given in Figures 7) yield values of C(d) as shown in Figure 8.
The beginning of pure frictional behaviour (Figure 6) is not distinct. Thus, in Figure 8 a range
of damage is given for this sliding behaviour. With the simple definition of damage adopted in
this study, pure frictional behaviour occurs at a unique value of d (and thus of q1 also) that is
1
independent of p.
The hardening portion of C occurs over a relatively small range of damage. It is attributed to
two mechanismscrushing and fracture. Rib-scale models11 predict high compressive stresses
near the face of the rib. Lutz and Gergely4 concluded that experimental results32,33 indicate
initial crushing causes a compaction of porous concrete paste near the rib face. Their conclusion
was supported both by the inspection of tested specimens and stiffer elastic response during
reloading (which could be characterized as elastoplastic coupling in the context of this model).
Initial crushing in effect increases the contact region with the remaining porous concrete thus
increasing the force necessary to produce additional crushing. Alternatively, one might attribute
hardening to the stable growth of transverse cracks.
For low confinement stresses, the effects of isotropic hardening should be less significant since
the stresses near the face of the rib are reduced. Data presented in the next section indicate that for
low !p the ribs tend to wedge under the concrete (geometric dilation) and thus induce less
extensive crushing and transverse cracking. It is the previously mentioned interaction of the
isotropic hardening and kinematic softening that captures this more brittle behaviour at low
confinement stresses. In this context, brittle refers to the bar-scale response not the rib-scale
mechanics.
As one would expect, relative to the rib spacing, the isotropic hardening is short-lived. The
dominant behaviour is that the force necessary to produce plastic slip progressively reduces with
plastic slipsoftening. The mechanisms include crushing, fracture and friction; the latter of which
dominates in the end.
For the calibration presented in Figure 8, C is defined as
C(d)"!3)3 exp(!10 d3@5)#6 exp(!2)6d6@5)#1)7 (10)
This definition has two important properties with respect to convergence of implicit finite element
analyses: (1) C3C1(0, 1) and (2) as dP0`, dC/ddPR. These properties complete the necessary
conditions for Q3C1(q). (The definition of d, equation (5), does imply that C has a slope
discontinuity with respect to plastic slip at d"1.) The infinite initial slope implies a C1 response
in the transition from elastic to elastoplastic behaviour, which is consistent with the lack of
distinct yield points in experimental results.
Figure 9 shows the graphs of pL and with values of d aligned with Figure 8. These figures
%
illustrate that: (1) isotropic hardening and kinematic softening occur over nearly the same range
of d, and (2) by the time the yield surface attains its exponential shape (i.e., "1) further change
%
in C(d) is relatively small, which is consistent with friction being the dominant mechanism at this
point. The reduction in C(d) that does occur can be attributed, in part, to a reduction in the
effective coefficient of friction as the damaged interface as polished.
With the calibration of the yield criterion complete, Figure 10 shows the yield surface at several
states of internal damagedepicting its evolution. The geometric effects of each component

( 1998 John Wiley & Sons, Ltd. MECH. COHES.-FRICT. MATER., VOL. 3: 155180 (1998)
172 J. V. COX AND L. R. HERRMANN

Figure 8. Comparison of C to experimental data

Figure 9. (a) pL versus d; and (b) versus d


%

Figure 10. Yield surface evolution

( 1998 John Wiley & Sons, Ltd. MECH. COHES.-FRICT. MATER., VOL. 3: 155180 (1998)
PLASTICITY BOND MODEL 173

function of the yield criterion is readily apparent: pL readily translates the yield surface to the
origin (having a significant effect on states where p is small and compressive). C expands (during
hardening) then partially contracts (during softening) the surface to a final size for friction, and
transitions the shape of the surface from the power form to the exponential form (which is
%
most apparent in the initial slope of the surface).
For values of !p less than approximately f , the bond stress at initial yielding exceeds the
5
bond stress when C is a maximum; this implies that for a fixed confinement stress in this range, the
maximum bond stress is attained prior to the maximum expansion of the yield surface. Thus, as
the confinement stress decreases, the maximum bond stress is attained with less plastic slip, and
the maximum bond stress decreases; i.e., the model exhibits a more brittle behaviour. This aspect
of the model and the variation in normal stress during the loading history cause the slip at which
the maximum bond stress occurs to vary. Furthermore, softening in the q versus d response of the
5
model can be produced by evolution of the yield surface (isotropic or kinematic softening) and by
a reduction in the confinement stress; this latter source of softening behaviour can occur with
rapid growth of longitudinal cracks.

5. FLOW RULE
For non-associative plasticity, the flow rule is obtained not from the yield function but from
a potential function. The flow rule developed for this bond model is based upon very little
experimental data, so its current form must be regarded as tentative. For brevity a full description
of the flow rules development is omitted. For additional details see Reference 26.
The experimental data of Malvar30,31 and Gambarova, Rosati and Zasso28 were examined in
developing the flow rule. Analysis of these data, with a homogenized view of the bond response,
indicates that a non-associative flow rule is necessary; among the associated liabilities is
a non-symmetric tangent stiffness matrix. Since friction contributes to bond behaviour, the need
for a non-associative flow rule was expected.
The wedging action of the ribs at the rib scale is interpreted as radial dilation of the interface at
the bar scale. Figure 11 presents bond stress versus radial dilation (measured at the specimens
surface) for a few of the series 3 tests of Malvar.30,31 As with frictional material behaviour, most of
the bond dilation occurs near the ultimate strength. Generally, the data show a reduction in
radial dilation with increasing confinement stress. This behaviour is easy to understand in terms
of work. For low confinement stress, the work associated with geometric dilation is less than that
required to crush material in front of the ribs or propagate transverse cracks. With high
confinement stress the converse is true. For a strictly convex yield surface, if radial dilation is
included in the model it must be dependent upon the normal stress to guarantee that the
incremental plastic work remains positive.26 The calibration of the yield surface and the flow rule
is carried out so as to insure this to be true.
The following form was adopted to simplify the flow rule description:

G H
sgn(q)
MqR 1N"jQ (11)
g

where j0 denotes the consistency parameter, which is positive during plastic loading and zero
otherwise. Note that g denotes the rate of plastic dilation with respect to plastic slip.

( 1998 John Wiley & Sons, Ltd. MECH. COHES.-FRICT. MATER., VOL. 3: 155180 (1998)
174 J. V. COX AND L. R. HERRMANN

Figure 11. Bond stress versus average radial dilation of Malvar30,31 series 3 tests

To obtain an approximation for g, limited data presenting radial dilation versus slip30,31 were
analysed. Consistent with assumptions made in obtaining the yield surface, we assumed that: (1)
g is only dependent upon !p and d; (2) q1 can be accurately estimated from equation (4); (3) the
1
!p value reported by Malvar30,31 is the normal stress at the bar; and (4) the radial elastic
contraction due to the applied confinement stress is not significant. With these assumptions,
approximations to g were determined from derivative estimates obtained from the experimental
data.
Figure 12 presents experimental and model approximations of g for fixed values of p. The
model is defined over four stages: (1) radial dilation, (2) constant radius, (3) radial contraction and
(4) sliding. As reflected in the models response, the shape of the graph of g is similar for each value
of !p. For the model, the limit and description of each stage are, generally, confinement stress
dependent.
While the yield surface and flow rule are independently defined, their combination
must characterize the mechanics of the bond behaviour, for the model to have a measure of
generality. The combined behaviour is important for most bond specimens. Typically,
radial dilation is constrained by concrete cover or secondary reinforcement, so the radial dilation
increases the normal stress which affects the bond stress. This is significantly different than
a model which merely scales the bond stress as a function of the normal stress; with radial dilation
incorporated in the flow rule the model actively participates in determining the local normal
stress state. As such, the model can potentially predict splitting as well as pull-out modes of
failure. The cost of this benefit, is that the model needs accurate normal stress data to give the
correct response.
The form of the model also gives insight into the mechanics of bond. Figure 11 shows that the
radial dilation of the bond zone is greater than the subsequent radial contraction.E At high
confinement stress, there are two likely sources for this permanent dilation: (1) an increase in

E
This is equivalent to noting that the area of the graph of g (with respect to d) during the radial dilation stage is larger than
the area during the radial contraction stage.

( 1998 John Wiley & Sons, Ltd. MECH. COHES.-FRICT. MATER., VOL. 3: 155180 (1998)
PLASTICITY BOND MODEL 175

Figure 12. Models flow rule, g, compared with experimental data

volume of the damaged bond-zone material and (2) deposition of damaged material into adjacent
cracks (longitudinal and transverse) which prevents their closure. In either case the bond zone
would develop hoop stress; this explanation is consistent with the observed reduction in the
effective coefficient of friction. It is also consistent with the models yield surface shape for large d.
At low !p, geometric dilation dominates the radial response, and thus the material has little
damage. In this case, the Coulomb surface is sufficiently accurate. At high !p, more material is
damaged in the bond zone, thus during contraction hoop stress develops and reduces the actual
normal stress at the bar. Since the hoop stress is not included in the interface characterization, the
actual !p is less than the models value, so the predicted bond stress must be less than a Columb
surface would predict. This is true even if finite elements are used to model the bond specimen,
because the geometric discontinuities (ribs) which cause much of the damage in the bond zone
have been eliminated by the homogenization process.

6. CALIBRATION RESULTS
Figure 13 compares simulation results using the bond model with experimental results from three
calibration tests. For the results shown, the effect of the specimen behaviour is neglected; i.e.,
consistent with the simplifications made in deriving the model, Malvars specimen is idealized as
being sufficiently small that (1) the bond behaviour can be considered constant with respect to
z (the position along the length of the bar), and (2) the actual value of p can be taken as the value
reported by Malvar (as if no hoop stress is carried by the specimen). The effects of this last
simplification were examined in some detail.26,34
The effects of the kinematic softening and transition function are only significant in the low
confinement stress case, in the region where the experiments exhibit similar behaviour. The
combination of kinematic softening and isotropic hardening effectively capture the change in
ductility with confinement stress. The flow rule predicts the radial dilation with reasonable
accuracy; as expected the model agrees best with the series 3 results, from which the flow rule
model was developed. While the bond response differs noticeably with the rib pattern, we sought
an averaged prediction in the first model.

( 1998 John Wiley & Sons, Ltd. MECH. COHES.-FRICT. MATER., VOL. 3: 155180 (1998)
176 J. V. COX AND L. R. HERRMANN

Figure 13. Calibration results for 3 confinement stresses

7. MODEL APPLICATION
Though the thrust of this paper is on model development, a brief discussion on validation,
implementation, and application of the model are provided for completeness.
The calibration results presented here do not validate the model; they simply show the
consistency between the model and the data from which it was derived. The model has been able

( 1998 John Wiley & Sons, Ltd. MECH. COHES.-FRICT. MATER., VOL. 3: 155180 (1998)
PLASTICITY BOND MODEL 177

to reproduce the results from six different experimental studies with acceptable accuracy (about
20%) using the calibration presented here. Typical results for the specimens of Gambarova,
Rosati and Zasso28 and Eligehausen, Popov and Bertero7 are presented in Figure 14. (Point
models refer to the use of simplified strength of material models for the specimens.) All of the
validation tests are for pull-out specimens since the monotonic version of the model is not directly
applicable to tensile specimens that can exhibit a reversal in bond stress with the development of
primary cracks. A forthcoming paper will present the results of the validation studies26 (a subset
of which have been presented at conferences35,36) and further discuss the strengths and
weaknesses of the model.
The model was initially implemented26 with the cutting-plane algorithm.37 The use of this
stress-point algorithm was motivated by its simple explicit updates of internal variables and
plastic strains, which did not require second derivatives to be determined during the model
development. For problems where computational efficiency is more important, variations on the
closest-point projection algorithm38 have been adopted; these algorithms allow the
algorithmically consistent elastoplastic moduli39 to be exploited. Usually E q% E;E q1 E , so the
initial elastic predictor of these return algorithms can predict stress states that significantly violate
consistency. In cases where the plastic corrector does not converge to a consistent stress state,
proportional strain subincrements have been used to attain convergence.
Probably the most feasible application of the bond model to actual reinforced concrete
structures is via a composite modeling approach. Mello and Herrmann1 give the details and

Figure 14. Typical validation results for the experiments of (a) Gambarova, Rosati and Zasso28 and (b) Eligehausen
Popov and Bertero7

( 1998 John Wiley & Sons, Ltd. MECH. COHES.-FRICT. MATER., VOL. 3: 155180 (1998)
178 J. V. COX AND L. R. HERRMANN

results of such an application. In their approach a special composite model is formulated for the
reinforcing bars, the bond between the reinforcement and the concrete, and the concrete
immediately surrounding the reinforcement. The h (hoop direction) dependence of the local
deformations and stresses in the unit cell containing one reinforcing bar (in a finite element
application utilizing this composite model, each finite element in general would contain a number
of such unit cells) are represented by the first three terms in a general Fourier series expansion.
The assumed axisymmetric bond stresses in the bond model are set equal to the average of the
h-dependent stresses in the composite unit cell (this averaging was used with good success26 in the
interpretation of the non-axisymmetric bond tests of Gambarova, Rosati and Zasso28). This
composite model assumes general three-dimensional conditions and can be easily simplified to
approximately model beams and girders in a two-dimensional plane stress setting. In the
applications to date, Mello and Herrmann,1 have obtained very good agreement with the results
of tests of reinforced concrete beams.

8. CONCLUSIONS
By making some key assumptions, a series of bond tests with different levels of normal stress (p)
were used to derive a two-dimensional bond model which includes the normal stress and the work
conjugate relative displacement (bond zone dilation), d . The form of each model component was
/
empirically derived and qualitatively reflects the underlying mechanics which is dominated by the
mechanical interaction of the ribs and adjacent concrete. The generalized strains and internal
variable include characteristic length measures related to the rib pattern. Plasticity theory
provides an effective framework for bond models applicable to reinforcement with a significant
local surface structure.
The model reproduces well the response (both bond versus slip and bond versus dilation) of
calibration tests for three levels of confinement stress. Though not demonstrated here, the radial
dilation response indicates the models potential to lead to finite element predictions of splitting
as well as pull-out failuresan important capability for predicting general bond behaviour.40,41
Several refinements in the model should be considered, including: (1) modification of the yield
surface to eliminate the region with p'0 while still maintaining the same shape for compressive
confinement stresses (see e.g., Reference 26), (2) elastoplastic coupling to produce change in the
elastic moduli (see the preliminary study of Cox34), and (3) a reformulation of the flow rule
description, with a possible dependence on the yield surface. In addition to these refinements, the
limitations inherent to bar-scale idealizations of bond need to be better defined.

ACKNOWLEDGEMENTS

This work was funded by the Naval Civil Engineering Laboratory, under the ONR sponsored
project Nonlinear Structual Modeling (project engineer: Dr. Ted Shugar).

REFERENCES
1. J. D. Mello and L. R. Herrmann, Implementation of a bond model, including dilatation, for reinforced materials in
finite element analysis, Contract Report CR 96.010, Naval Facilities Engineering Service Center, Port Hueneme, CA,
1996.
2. V. V. Bertero, Seismic behavior of structural concrete linear elements (beams and columns) and their connections,
Comite Euro-International due Beton (CEB), Bulletin DInformation no. 131, Paris, France, 1979.

( 1998 John Wiley & Sons, Ltd. MECH. COHES.-FRICT. MATER., VOL. 3: 155180 (1998)
PLASTICITY BOND MODEL 179

3. G. Rehm, he Basic Principle of Bond between Steel and Concrete, Deutscher Ausschuss fur Stahlbeton, no. 138,
Wilhelm Ernest and Sohn, Berlin (C & CA Library Translation no. 134, 1968), 1961.
4. L. A. Lutz and P. Gergely, Mechanics of bond and slip of deformed bars in concrete, ACI J., 64, 711721 (1967).
5. Y. Goto, Cracks formed in concrete around tension bars, ACI J., 68(4), 244251 (1971).
6. S. Hungspreug, Local bond between a steel bar and concrete under high intensity cyclic loading, Ph.D. Dissertation,
Cornell University, Ithaca, NY, 1981.
7. R. Eligehausen, E. P. Popov and V. V. Bertero, Local bond stress-slip relations of deformed bars under generalized
excitations, Report UCB/EERC-83/23, University of California, Berkeley, CA, 1983.
8. A. R. Ingraffea, W. H. Gerstle, P. Gergely and V. Saouma, Fracture mechanics of bond in reinforced concrete,
J. Struct. Engng. ASCE, 110(4), 871890 (1984).
9. P. G. Gambarova and E. Giuriani, Discussion of Fracture mechanics of bond in reinforced concrete, by Ingraffea
et al., J. Struct. Engng. ASCE, 111(5), 11611164 (1984).
10. A. K. De Groot, G. M. A. Kusters and T. Monnier, Numerical Modeling of Bond-Slip Behavior, Heron, Vol. 26-1b,
I.B.B.C. Institute TNO, Delft, Netherlands, 1981.
11. H. W. Reinhardt, J. Blaauwendraad and E. Vos, Prediction of bond between steel and concrete by numerical
analysis, Mater. Struct., 17(100), 311320 (1984).
12. J. G. Rots, Computational modeling of concrete fracture, Ph.D. Dissertation, Delft University of Technology, Delft,
1988.
13. J. Ozbolt and R. Eligehausen, Numerical simulation of cycling bond-slip behavior, in Bond in Concrete, Proc. Int.
Conf., CEB, pp. 12.2712.33, 1992.
14. C. J. Brown, D. Darwin and S. L. McCabe, Finite element fracture analysis of steelconcrete bond, SM Report no. 36,
Department of Civil Engineering, Univerity of Kansas, Lawrence, KS 1993.
15. ASCE Task Committee of Finite Element Analysis of Reinforced Concrete Strucures, State-of-the-Art Report of Finite
Element Analysis of Reinforced Concrete, ASCE Special Publications, 1982.
16. F. C. Filippou, A simple model for reinforcing bar anchorages under cyclic excitations, J. Struct. Engng. ASCE,
112(7), 16391659 (1986).
17. B. Bresler and V. V. Bertero, Behavior of Reinforced Concrete Under Repeated Load, J. Struct. Div. ASCE, 94(ST6),
15671590 (1968).
18. G. Pijaudier-Cabot, J. Mazars, J. Pulikowski, Steel-concrete bond analysis with nonlocal continuous demage, J.
Struct. Engng. ASCE, 117(3) 862882 (1991).
19. Z. P. Bazant and S. Sener, Size effect in pull-out tests, ACI Mater. J., 85, 347351 (1988).
20. G. Mehlhorn and M. Keuser, Isoparametric contact elements for analysis of reinforced concrete structures, in Proc.
Japan.S. Seminar on Finite Element Analysis of Reinforced Concrete Strucures, ASCE, pp. 329347, 1985.
21. S. Morita and S. Fujii, Bond-slip models in finite element analysis, in Proc. Japan.S. Seminar on Finite Element
Analysis of Reinforced Concrete Structures, ASCE, pp. 348363 1985.
22. T. Zhiming, L. Hueizhong, and Z. Jinping, A new bond model for finite element analysis of structures, in Bond in
Concrete, Proc. Int. Conf., CEB, pp. 12.912.16, 1992.
23. J. Mainz, S. Stockl and H. Kupfer, FE-calculations concerning the bond behavior of deformed bars in concrete, in
Bond in Concrete, Proc. Int. Conf., CEB, pp. 12.1712.26, 1992.
24. M. E. Plesha, Constitutive models for rock discontinuities with dilatancy and surface degradation, Int. J. Numer.
Anal. Meth. Geomech., 11(4), 345362 (1987).
25. T. Stankowski, K. Runesson and S. Sture, Fracture and slip of interfaces in cementitious composites. I:
Characteristics, J. Engng. Mech. ASCE, 119(2), 292327 (1993).
26. J. V. Cox, Development of a plasticity bond model for reinforced concrete theory and validation for monotonic
applications, echnical Report TR-2036-SHR, Naval Facilities Engineering Service Center, Port Hueneme, CA, 1994.
27. L. R. Herrmann and J. V. Cox, Development of a plasticity bond model for reinforced concrete preliminary
calibration and cyclic applications, Contract Report CR 94.001-SHR, Naval Facilities Engineering Service Center,
Port Hueneme, CA, 1994.
28. P. G. Gambarova, G. P. Rosati and B. Zasso, Steel-to-concrete bond after concrete splitting: test results, Mater.
Struct., 22(127), 3547 (1989).
29. W. H. Gerstle and A. R. Ingraffea, Does bond-slip exist?, in Micromechanics of Failure of Quasi-Brittle Materials,
Proc. Int. Conf., Albuquerque, NM, pp. 407416 1990.
30. L. J. Malvar, Bond of reinforcement under controlled confinement, echnical Note 1833, Naval Civil Engineering
Laboratory, Port Hueneme, CA, 1991.
31. L. J. Malvar, Bond of reinforcement under controlled confinement, ACI Mater. J., 89(6), 593601 (1992).
32. L. A. Lutz, P. Gergely and G. Winter, The mechanics of bond and slip of deformed reinforcing bars in concrete,
Structural Engineering Report No. 324, Cornell University, Ithaca, NY (1966).
33. G. Rehm, The fundamental law of bond, in Proc. Symp. on Bond and Crack Formation in Reinforced Concrete,
RILEM, Vol. 2, pp. 491498, 1957.
34. J. V. Cox, Elastic moduli of a bond model for reinforced cencrete, in Proc. 11th ASCE Engineering Mechanics
Specialty Conf., pp. 8487, 1996.

( 1998 John Wiley & Sons, Ltd. MECH. COHES.-FRICT. MATER., VOL. 3: 155180 (1998)
180 J. V. COX AND L. R. HERRMANN

35. J. V. Cox and L. R. Herrmann, Confinement stress dependent bond behavior, part II: A two degree of freedom
plasticity model, in Bond in Concrete, Proc. Int. Conf., CEB, pp. 11.1111.20, 1992.
36. J. V. Cox and L. R. Herrmann, A model for bond between steel bars and concrete, in Proc. 10th ASCE Engineering
Mechanics Specialty Conf., pp. 647650, 1995.
37. J. C. Simo and M. Ortiz, A unified approach to finite deformation elastoplastic analysis based on the use of
hyperelastic constitutive equations, Comput. Meth. Appl. Mech. Engng., 49, 221245 (1985).
38. M. Ortiz, P. M. Pinsky and R. L. Taylor, Operator split methods for the numerical solution of elastoplastic dynamic
problems, Comput. Meth. Appl. Mech. Engng., 39, 137157 (1983).
39. J. C. Simo, and R. L. Taylor, Consistent tangent operators for rate-independent elastoplasticity, Comput. Meth. Appl.
Mech. Engng., 48, 101118 (1985).
40. R. Tepfers, Cracking of concrete cover along anchored deformed reinforcing bars, Mag. Concrete Res., 31(106), 312
(1979).
41. R. Tepfers and P. Olsson, Ring test for evaluation of bond properties of reinforcing bars, in Bond in Concrete, Proc.
Int. Conf., CEB, pp. 1.891.99, 1992.

( 1998 John Wiley & Sons, Ltd. MECH. COHES.-FRICT. MATER., VOL. 3: 155180 (1998)

Você também pode gostar