Professional Documents
Culture Documents
Modeling Nonlinear Bond-Slip Behavior For Finite Element Analyses of Reinforced Concrete Structures
Modeling Nonlinear Bond-Slip Behavior For Finite Element Analyses of Reinforced Concrete Structures
Finite element analyses of reinforced concrete structures should concrete and reinforcement. The two solutions are brought
allow some representation of bond stiffness, particularly if there together by a rapidly converging iterative process yielding
are areas of high shear stress. This paper reviews research on bond the bond stresses that are the link between the steel and the
stiffness and then brings together earlier work on an efficient algo- concrete. In this paper, the method is extended to represent
rithm for incorporating bond stiffness in a finite element analysis
the bond-slip behavior by a realistic nonlinear model based
and a nonlinear bond stiffness model developed from experimental
and theoretical studies. The resulting new method is used to on the experimental observations of many researchers.
analyze pullout tests and the results are compared with experi-
mental data. Good agreement is obtained without resorting to the RESEARCH SIGNIFICANCE
unlikely notion of “position-dependent” stiffnesses adopted by This research studied the use of a nonlinear bond-slip model
other researchers. in conjunction with a new method for incorporating bond
stiffness in finite element analyses of reinforced concrete
Keywords: bond-slip relationship/model; finite element analysis; nonlin- structures. Results obtained for pullout tests compare well
ear bond stiffness; pullout tests; reinforced concrete. with experimental data and show that it is not necessary to
adopt the unlikely concept of “position-dependent” bond stiff-
INTRODUCTION ness proposed by other researchers. The method is efficient in
It is common for finite element analyses of reinforced computing and could be adopted for practical application.
concrete structures to ignore the reinforcement altogether.
Even when it is included, it is usual to assume that there is no BOND-SLIP PHENOMENA
relative movement between steel and concrete, i.e., no allow- Early experimental studies of bond were concerned with
ance is made for any bond “slip” to occur. The consequence determining bond failure strengths and the influence of
of this assumption is an unrealistically rapid transfer of stress surface deformations on them. Abrams4 studied the bond of
between concrete and steel, particularly in areas of high plain and deformed bars, but a later and thorough study by
shear stress where there are inevitably high bond stresses. Glanville5 revealed an important but much neglected effect.
Fig. 1 (taken from Reference 1) compares measured rein- He demonstrated that bond failure occurs at a higher stress
forcement stresses in a beam-column junction with those for a pushout test than for the normal pullout test. The
calculated by finite element analyses assuming both infinite explanation for this is simple; the compressive axial stress
and more realistic bond stiffnesses. This clearly shows how developed in the bar during a pushout test causes an increase
the bar stresses computed with infinitely stiff bond fall off in bar diameter because of Poisson’s effect, which, in turn,
far too rapidly within the column. causes an increase in the radial pressure between bar and
concrete. Since friction is an important element in bond, this
Ngo and Scordelis2 showed that the stiffness of the bond increase in pressure leads to an increased bond strength, a
connection could be incorporated in finite element analyses factor generally ignored but included in the authors’ model
by discrete springs connecting steel and concrete together at of bond behavior described below.
discrete nodes. One spring parallel to a bar could model the Many researchers have also studied bond by performing
bond-slip stiffness, and a second spring normal to the bar experiments on strain-gaged reinforcing bars embedded in a
could model the much stiffer “dowel” action. Many authors variety of concrete specimens, e.g., Wilkins,6 Mains,7 Peattie
since have developed methods to represent continuous bond and Pope,8 Perry and Thompson,9 Nilson,10 and Scott and
stiffness along reinforcement bars. Unfortunately, this Gill.11 In these tests, the axial strain in the bars was measured
generally increases the number of degrees of freedom in the directly but bond stresses can only be determined indirectly
analysis, resulting in more costly data preparation and
computer time, particularly if a nonlinear model of bond-
slip behavior is adopted. ACI Structural Journal, V. 93, No. 5, September-October 1996.
Received July 17, 1995, and reviewed under Institute publication policies. Copy-
In a previous paper,3 the authors introduced a new method right © 1996, American Concrete Institute. All rights reserved, including the making of
for modeling linear bond stiffness that avoids these extra copies unless permission is obtained from the copyright proprietors. Pertinent discussion
will be published in the July-August 1997 ACI Structural Journal if received by
costs by performing separate finite element analyses on the March 1, 1997.
⎛ ⎧ K c 0 ⎫ ⎧ K b – C t K b ⎫⎞ D c Pc
⎜⎨ ⎬+⎨ ⎬⎟ ⋅ ⎛ ⎞ = ⎛ ⎞ (4)
⎝ ⎩ 0 K s ⎭ ⎩ – K b C K b ⎭⎠ ⎝ D s ⎠ ⎝ P s ⎠
where
Kc = assembled stiffness matrix for all concrete elements
Ks = assembled stiffness matrix for all steel elements
Fig. 2—Bond stress-slip relationship linked to radial pressure
KB = assembled matrix of all bond stiffnesses transformed
acting on reinforcing bar.
to global axes at their connections to concrete
C = assembled matrix of all of previous transformations
Kb = diagonal matrix of all bond stiffnesses in local axes
Dc,s = vectors of displacements at all concrete and steel
nodes
Pc,s = vectors of loads applied to concrete and steel nodes
An iterative solution process can be derived from the
upper and lower rows of submatrices in Eq. (4) as
i+1 i t i
Kc ⋅ Dc = Pc – ( KB ⋅ Dc – C ⋅ Kb ⋅ Ds ) (5)
i+1 i+1
( Ks + Kb ) ⋅ Ds = Ps + Kb ⋅ C ⋅ Dc (6)
APPLICATIONS
Two examples of the method are given, both pullout tests.
The first, on a plain steel bar, is included because strain gage
measurements on the bar were available. In the second test,
using a standard deformed reinforcing bar, only free-end slip
values were available.
The plain bar pullout test was studied by Parsons,22 both
experimentally and by a conventional nonlinear finite
element method using 6-noded bond elements. The bar
was 16 mm (0.63 in.) in diameter and was pulled out from
a 150-mm (6-in.) concrete cube. The parameters of the bond
model used for the FE analysis by the method described in
this paper are set out in Table 1, together with the corresponding
values for the deformed bar pullout described later. Values for
these bond parameters are not easily obtained. Published values Fig. 5—Flow chart of iterative method.
for the initial bond stress/slip slope [Ro in Eq. (3)] range from 50
to 1000 N/mm2 (0.18 to 3.68 ⋅ 106 psi/in.).
Greater agreement is found over the slip at ultimate bond will have after bond failure, i.e., when only a friction compo-
stress Δu and of the initial bond strength. The factor μ has nent will apply. For the deformed bar, the value is 1.0.
been estimated from the results of Robins and Standish.16 Fig. 6 shows how the axial load varies through the embedded
The parameter β, representing the reduction in bond strength length of the plain bar as measured experimentally and as
after bond failure, is not critical and has been set to 0.5 to computed. Axial load rather than axial stress has been
represent the likely reduction in bond strength a plain bar presented to illustrate how the pullout load is distributed
CONCLUSIONS
The applications previously shown demonstrate that the
authors’ bond model produces the effect of a lower bond
stiffness at nodes near the concrete surface as reported by
Nilson10 and Pochanart and Harmon,13 not by using a bond
stiffness that depends on location, but by adjusting the bond
stiffness to the prevailing concrete and steel stress condi-
Fig. 7—Computed bond stresses for plain 16-mm (0.63-in.)
tions. The key effect is to allow for changes in bar diameter
bar pullout test.
due to Poisson’s ratio and for the impact that that has on
bond stiffness.
through the embedded length as bond transfers the load to The analysis of a beam and column assembly designed to
the concrete. The change in that distribution as the pullout ACI 318-83 specifications was presented by the authors in
load increases is quite clear from the experimental results Reference 3 using a linear bond and concrete model. The
that are partly but not completely matched by the computed problems described in this paper have used a nonlinear bond
results. Fig. 7 shows the bond stresses from the computed model and a linear concrete model. Research continues and
results (it was not possible to derive bond stresses from the there are no reasons why the nonlinear bond model described
experimental values since the process of differentiation and a nonlinear concrete model allowing cracks to form
exaggerated the errors in the readings). The distribution of should not show the bond stress reversals that take place
these stresses clearly shows bond failure developing from between cracks across reinforcement.
the loaded end with the bond stress there being limited to
about 1.2 N/mm2 (175 psi). The limit is higher near the REFERENCES
unloaded end due to less reduction in bar diameter by 1. Allwood, R. J., “Reinforcement Stresses in a Reinforced Concrete
Poisson’s ratio effect. Beam-Column,” Magazine of Concrete Research, V. 32, No. 112, Sept.