You are on page 1of 11

SIMULATION OF STRUCTURAL CONCRETE UNDER CYCLIC LOAD

By W.-P. Kwan1 and S. L. Billington,2 Associate Member, ASCE

ABSTRACT: The selection and evaluation of finite-element modeling approaches for predicting cyclic behavior
of structural concrete is presented. Nonlinear material models for structural concrete are evaluated through
comparison of simulations with one set of experiments. A set of models and parameters is selected and modi-
fications are implemented by the writers. This set of models is further evaluated through comparison with
experiments of cyclically loaded structural concrete components with varying geometry, loading conditions, and
hysteretic response. It is found that representation of shear transfer across crack faces strongly influences hys-
teretic behavior. Modeling compression strain-softening and the Bauschinger effect in reinforcing steel improves
predicted cyclic response relative to experiments. The selected set of material models are evaluated and found
Downloaded from ascelibrary.org by Li. Co.Sa 8181901/mi/155985 on 03/24/20. Copyright ASCE. For personal use only; all rights reserved.

to capture flexure-dominated hysteretic behavior, indicate potential shear-dominated failure, predict local influ-
ences of bonded prestressing through joints, and capture the influences of unbonded vertical prestressing in
bridge columns. Modeling bondslip, concrete shear distortion, and steel buckling are necessary for full failure
mode prediction. This research uses models available in commercial finite-element codes and is intended for
researchers and practitioners interested in using nonlinear finite-element analysis to predict cyclic response of
new and existing structural concrete designs.

INTRODUCTION tempts. With the advancement of modeling techniques and


computational power, more successful cyclic loading analyses
The challenge in numerical modeling of structural concrete of reinforced concrete have been conducted in the 1990s.
arises from the material’s composite nature. Structural con- These include simulations of shear test panels designed to have
crete’s nonlinear behavior includes cracking, crushing, tension a uniform stress state for the verification of constitutive models
stiffening, compression softening, and bondslip. Cyclic load- (Schnobrich et al. 1991; Stevens et al. 1991; Rose et al. 1999),
ing introduces further complexities, such as stiffness degra- and simulations of structural members (Okamura and Mae-
dation in concrete, the Bauschinger effect in reinforcing steel, kawa 1991; Parker and Dameron 1994; Sittipunt and Wood
and bond degradation between concrete and reinforcement. 1995; Elmorsi et al. 1998; Lee et al. 1999; Vecchio 1999). A
The accuracy and reliability of numerical modeling of rein- summary of a number of finite-element simulations of cyclic
forced concrete is governed by the abilities of the underlying loading experiments on reinforced concrete members in the
constitutive relations to capture different types of nonlinear last decade is shown in Table 1.
behavior. This paper focuses on the selection of nonlinear ma- Most of the cyclic analyses in Table 1 were carried out using
terial models to predict component behavior under cyclic load. finite-element codes developed for research. Existing com-
A selected set of models for structural concrete is evaluated mercial finite-element codes often have limitations in repre-
by using it to simulate a variety of cyclically loaded compo- senting cyclic behavior due to idealizations in material models.
nent experiments that exhibit different hysteretic responses. To contribute to bridging the gap between research and prac-
The focus of these simulations is on finite-element analyses to tice in nonlinear finite-element analysis, the objective of this
facilitate the evaluation of new and existing structural designs research is (1) to evaluate the effects of various nonlinear ma-
when only material and geometric properties are known, as terial models and their associated parameters on the cyclic re-
opposed to knowing section (hysteretic) behavior from exper- sponse of structural concrete members; and (2) to verify the
iments. ability of a set of models (selected based on calibration to one
Since the first use of nonlinear finite-element analysis to set of experiments) to represent different types of cyclic be-
simulate reinforced concrete (Ngo and Scordelis 1967), sig- havior exhibited by a variety of structural components. A brief
nificant developments have followed for more realistic mod- background is presented in the section entitled Material Mod-
eling of structural concrete behavior. A large number of so-
els for Structural Concrete. Results of the first objective are
phisticated constitutive models that can successfully capture
addressed under Selection of Material Models, and the results
behavior at the material level have been developed, as sum-
of the second objective are presented under Model Evaluation.
marized in state-of-the-art reports by ASCE-ACI (1982, 1993)
Emphasis is placed on evaluating material models available
and CEB (1996). Less has been reported on the successful
in commercial, nonlinear finite-element codes or models that
application of these material models to predict the behavior of
structural components under cyclic loading. may be easily implemented or modified within such codes.
A comprehensive review of the early attempts to simulate The research presented here focuses more on the ability of the
cyclic loading experiments of reinforced concrete using finite- models to capture global hysteretic behavior, rather than de-
element analysis is given in Noguchi (1985). Noguchi points tailed quantitative evaluation and prediction of local failure
out that the pinching in hysteretic behavior and the energy modes.
dissipation are not represented sufficiently in these early at-
1
MATERIAL MODELS FOR STRUCTURAL CONCRETE
Struct. Engr., Leslie E. Robertson Assoc., 30 Broad St., 47/F, New
York, NY 10004.
2
Asst. Prof., School of Civ. and Envir. Engrg., Cornell Univ., Ithaca, Within the different constitutive relation frameworks, many
NY 14853. models exist or can be implemented by a user (in this case,
Note. Associate Editor: John Wallace. Discussion open until May 1, the writers) to simulate the nonlinear behavior of structural
2002. To extend the closing date one month, a written request must be concrete. These models vary, for instance, in terms of tension
filed with the ASCE Manager of Journals. The manuscript for this paper and compression unloading behavior, shear transfer across
was submitted for review and possible publication on January 11, 2000;
revised July 3, 2001. This paper is part of the Journal of Structural crack faces, compression strain softening, and the Bauschinger
Engineering, Vol. 127, No. 12, December, 2001. 䉷ASCE, ISSN 0733- effect in steel. Descriptions of various nonlinear models are
9445/01/0012-1391–1401/$8.00 ⫹ $.50 per page. Paper No. 22229. presented in this section.
JOURNAL OF STRUCTURAL ENGINEERING / DECEMBER 2001 / 1391

J. Struct. Eng., 2001, 127(12): 1391-1401


TABLE 1. Selected Aspects of FE Analyses of Reinforced Concrete Subjected to Cyclic Loading in 1990s
Normal stress
Subject of Framework for Tension model transfer between Shear transfer Bondslip
Researchers simulation Element types constitutive relations (cracking) cracks across crack faces representation
Okamura and Shear Concrete: 8-node isopara- Orthotropic, equiva- Smeared, Tension stiffen- Cyclic shear Joint element
Maekawa walls metric plate elements. lent uniaxial fixed or- ing transfer model between
(1991) Reinforcement: smeared model thogonal concrete
cracks elements
Parker and Bridge Concrete: 8-node 3D brick Plasticity model for Smeared, No tension Degrading shear Modify steel
Dameron column elements. Reinforce- compression fixed or- stiffening stiffness unloading
(1994) ment: subelements in thogonal
concrete. Steel jacket: cracks
shell elements
Sittipunt and Shear Concrete: 4-node isopara- Orthotropic, equiva- Smeared, Tension stiffen- Cyclic shear Not consid-
Wood (1995) walls metric plane-stress ele- lent uniaxial fixed or- ing. 8-param- transfer func- ered
Downloaded from ascelibrary.org by Li. Co.Sa 8181901/mi/155985 on 03/24/20. Copyright ASCE. For personal use only; all rights reserved.

ments. Reinforcement: model thogonal eter model tion. (9-param-


2-node truss elements cracks for cyclic eter function)
normal stress
function
Elmorsi et al. Shear Concrete: 12-node quadri- Orthotropic, equiva- Smeared, Tension stiffen- Cyclic shear Not consid-
(1998) walls lateral plane stress ele- lent uniaxial fixed or- ing. 3-param- transfer func- ered
ments. Reinforcement: model thogonal eter cyclic tion
smeared cracks normal stress
function
Vecchio (1999) Shear Concrete: 4-node isopara- Orthotropic equiva- Smeared, ro- Considered Not applicable Not consid-
wall metric plane-stress ele- lent uniaxial tating cases both due to use of ered
and test ments model with plas- cracks with and rotating cracks
panels tic offset concept without ten-
for compression sion stiffen-
ing
Lee et al. Column Concrete: 4-node quadri- Plastic-damage Tensile dam- Normal strength Governed by Linear and
(1999) lateral isoparametric el- model age variable degrades with yield surface exponential
ements. Reinforcement: for smeared increasing similar to functions
truss elements. Bond- cracking damage gov- Drucker-Prager for load-
slip: bond link element erned by evo- (with degrad- ing; in-
lution rules ing cohesion cludes slip
in terms of and friction an- in unload-
principal gle) ing
strains

FIG. 1. Uniaxial Stress-Strain Relationships for Concrete Using Total Strain Formulation with Adopted Secant Unloading/Reloading

Constitutive Relation Framework compression curve, and compression in the transverse direc-
tion will lead to an increase in strength and ductility.
Three main approaches are popular among researchers in In this research, an elasticity-based model with a total strain
establishing the constitutive relations of concrete. They are the formulation (Feenstra et al. 1998) is selected (Fig. 1). The
elasticity-based approach, the plasticity-based approach, and influence of lateral stresses is accounted for by using the mod-
the damage-based approach (CEB 1996). The most common ified compression field theory (Vecchio and Collins 1986,
model of the elasticity-based approach is the orthotropic, 1993). A preliminary study was conducted comparing a plas-
equivalent uniaxial model based on the ‘‘equivalent uniaxial ticity-based model with the adopted elasticity-based model.
strain’’ concept, originally proposed by Darwin and Pecknold Problems arose with the robustness of the plasticity-based
(1976) and extended by researchers such as Stevens et al. model for capturing compression softening in cyclically loaded
(1991). As summarized in Table 1, the equivalent uniaxial components, and it was decided to proceed with the elasticity-
strain approach has met considerable success in terms of ac- based model. A recent comparison of elasticity- and plasticity-
curate representation of experiments on structural concrete un- based models for cyclic simulations can be found in Feenstra
der cyclic load. This success is attributed to the fact that em- and Rots (2000).
pirical data on concrete’s complex behavior can be more
directly integrated into the constitutive relations (using phe- Tension Model (Cracking)
nomenological expressions) than other approaches. The con-
stitutive relations are based on the uniaxial behavior of con- Two common approaches for numerical modeling of crack-
crete, while the influence of stress from other directions is ing in concrete are discrete cracks and smeared cracks. Dis-
taken into account by modifying the uniaxial stress-strain re- crete cracks represent a crack as a geometrical discontinuity,
lations. For instance, cracking in the transverse direction will such as through the use of interface elements or combined
lead to a decrease in the peak stress and strain in the uniaxial finite and boundary elements with topological data structures
1392 / JOURNAL OF STRUCTURAL ENGINEERING / DECEMBER 2001

J. Struct. Eng., 2001, 127(12): 1391-1401


(Carter et al. 2000). With smeared cracks, the cracked material Under cyclic loading, the hysteretic behavior of steel exhib-
is treated as a continuum with reduced stiffness properties its the Bauschinger effect, which refers to the softening of the
(Rots and Blaauwendraad 1989). The smeared crack approach steel modulus upon unloading and reloading. Typical elasto-
is predominantly adopted by researchers for the simulation of plastic models for steel available in commercial finite-element
structural concrete under cyclic load (Table 1), as extensive codes cannot capture this effect, yet the effect can influence
cracking is usually expected. The smeared crack approach is energy dissipation at the structural scale. To investigate
also adopted in this research. the influence of the Bauschinger effect on the hysteretic re-
sponse, a steel model based primarily on Su and Zhu (1994)
Tension Stiffening has been developed and implemented by the writers with a
user-supplied subroutine. The developed subroutine extends
Physically, once a crack forms, force is still transferred in Su and Zhu’s model to include rules for partial loading and
the concrete between crack locations. The contribution of the unloading. The steel model is verified and calibrated against
stiffness of this concrete between cracks to the stiffness of the cyclic experiments on reinforcing bars (Kwan 2001).
Downloaded from ascelibrary.org by Li. Co.Sa 8181901/mi/155985 on 03/24/20. Copyright ASCE. For personal use only; all rights reserved.

reinforced composite is referred to as tension stiffening. In this The bond between steel and concrete can be accounted for
research, the tension stiffening model chosen uses a linear de- explicitly by using interface elements between discrete (steel)
scending branch in the tension-softening curve of concrete, and continuum (concrete) elements. Many cyclic stress-slip re-
which represents the gradual loss in strength of the concrete lations for interface elements have been proposed, as sum-
after a crack forms. The ultimate crack strain is approximated marized in CEB (1996). Incorporation of these models in the
as a portion of the yield strain. simulation of cyclic behavior of reinforced concrete members
is not yet common. As shown in Table 1, bondslip is often not
Shear Retention considered, or it is represented by other means, such as mod-
Shear stress is physically transferred across crack faces due ification in the steel stress-strain curve (Parker and Dameron
to aggregate interlock and dowel action. There is little con- 1994) or the use of joint elements between structural compo-
sensus among researchers on the best representation of shear nents (Okamura and Maekawa 1991). Bondslip is not explic-
transfer across crack faces (Table 1). Adopted approaches itly modeled in the studies presented here.
range from the fixed crack model using constant shear reten-
tion factors or more advanced shear transfer functions, to the SELECTION OF MATERIAL MODELS
rotating crack model, which does not require explicit repre-
sentation of shear transfer, as the crack directions are aligned The influence of material models and their associated pa-
with the principal stress directions (Rots and Blaauwendraad rameters on the response of cyclically loaded reinforced con-
1989). In this study, both the rotating crack model and the crete components is investigated. To do this, a series of cycli-
fixed crack model with a constant shear retention factor are cally loaded physical experiments of concrete bridge bent cap
adopted and their influence on structural response is compared. beam-to-column connections performed by Sritharan et al.
(1996) are simulated. Based on the investigations presented in
Compression Model (Crushing) this section, a set of models and parameters is selected for
Microcracking in concrete under compression leads to strain further study. This selected set is then evaluated for its ability
softening, or the degradation of the concrete’s compressive to predict different types of cyclic response in the next section.
strength and stiffness. The Thorenfeldt model, which is appli- Fig. 2 shows the experimental setup for one of Sritharan et
cable to both normal and high strength concrete, is adopted in al.’s connections (referred to as IC1) and the corresponding
this research to represent strain softening of plain concrete finite-element mesh. Material properties and reinforcement ra-
(Thorenfeldt et al. 1987). tios for IC1 (along with two other connections discussed later)
Passive confinement provided by transverse reinforcement are shown in Tables 2 and 3. The concrete is modeled with
influences strain softening by increasing the strength and the eight-noded plane stress elements. The longitudinal and shear
ductility of the concrete. The confined concrete model pro- reinforcement is modeled as embedded bar elements. The
posed by Mander et al. (1988) is implemented by the writers thickness of the concrete elements is varied to approximate the
as a user-supplied subroutine to account for this influence. circular shape. The finite-element model is subjected to an
Simulated hysteretic response of structural members using this axial load of 400 kN and displacement cycles to a ductility
model is compared with the response using unconfined con- level of 4 (drift ratio of about 3%). The effects of the selected
crete and using concrete idealized as an elastic, perfectly plas- material models and model parameters on the hysteretic be-
tic material. havior of this connection are shown in Fig. 3.

Unloading/Reloading of Concrete Influence of Compression Unloading


The cyclic compressive behavior of plain concrete physi-
cally exhibits stiffness degradation (in unloading and reload- The load-deflection responses with elastic compression un-
ing) as well as hysteresis loops from the different unloading loading and secant compression unloading are compared in
and reloading paths. Idealization of both elastic and secant Figs. 3(a and b). The difference in concrete unloading does
unloading-reloading in compression is adopted and compared not have a very significant influence on the component load-
herein, serving as the bounds of stiffness degradation for true deflection behavior. This observation is consistent with results
unloading-reloading behavior. Tensile cyclic behavior is rep- on analyses of other structural members (Kwan 2001). The
resented by secant unloading-reloading. difference in the overall stiffness of the simulated response is
seen immediately after a reversal in the load direction, but the
Reinforcing Steel effect on the overall energy dissipation capacity is small. The
elastic unloading of the steel dominates the unloading behavior
Reinforcing steel can be modeled as embedded bars within and the energy dissipation in these simulations. The residual
concrete elements, an equivalent uniaxial material distributed displacements are slightly lower when secant compression un-
throughout the concrete elements, or discrete elements. In this loading is used for the concrete; however, the difference is
research, embedded bars and discrete bars are used. minimal.
JOURNAL OF STRUCTURAL ENGINEERING / DECEMBER 2001 / 1393

J. Struct. Eng., 2001, 127(12): 1391-1401


Downloaded from ascelibrary.org by Li. Co.Sa 8181901/mi/155985 on 03/24/20. Copyright ASCE. For personal use only; all rights reserved.

FIG. 2. Cap Beam-To-Column Connection (IC1) [from Sritharan et al. (1996)] Corresponding Finite-Element Model

TABLE 2. Material Properties of Cap Beam-to-Column Connections Tested by Sritharan et al. (1996)
f ⬘c (MPa) fy (MPa)
Cap beam Column Beam Column Joint Beam Prestressing
Specimen Column and joint longitudinal longitudinal spiral spiral stirrups bars
IC1 (ordinarily reinforced) 31.4 39.7 448 433 431 411 439 —
IC2 (partially prestressed) 34.6 40.5 448 477 431 439 434 988
IC3 (fully prestressed) 33.2 36.1 461 426 434 440 452 889

TABLE 3. Summary of Reinforcement of Cap Beam-to-Column Connections Tested by Sritharan et al. (1996)
Column Column Beam Beam Beam ties Cap beam Joint Joint
longitudinal steel transverse steel longitudinal steel transverse steel adjacent to joint prestressing spiral stirrups
Specimen (%) (%) (%) (%) (%) (MPa) (%) (%)
IC1 (ordinarily reinforced) 1.86 0.52 1.30 0.25 0.41 0 0.87 0.20
IC2 (partially prestressed) 1.86 0.52 0.54 0.25 0.25 4.00 0.41 0.20
IC3 (fully prestressed) 1.86 1.79 0 0.25 0.25 7.03 0.41 0.20

Influence of Shear Transfer in Crack Models crack models occurs when open cracks exist in the compres-
sion face of the column. These cracks, initially in tension
Figs. 3(b and c) compare the simulated load-deflection re- (Quadrant I in Fig. 4), remain open for some period of time
sponse of the connection using a fixed crack model with a after load reversal (marked in Fig. 4) due to the plastic strain
shear retention factor of 0.2 and a rotating crack model, re- of the reinforcement. The degree of shear transfer between
spectively. The rotating crack model predicts the experimental these cracks as dictated by the different crack models directly
response well, while the fixed crack model predicts a stiffer affects the load-deflection response of the column.
response. The fixed crack model tends to overestimate the de-
gree of shear transfer across cracks at large crack strains. The
implicit shear stiffness in the rotating crack model better cap- Influence of Compression Strain-Softening
tures the degree of shear transfer across cracks under cyclic
loading. This finding is supported, for instance, in the work of For the analyses shown in Figs. 3(a–c), the concrete is ide-
Rots (1988), wherein it is shown that the implicit shear stiff- alized as an elastic, perfectly plastic material in compression
ness term in the rotating crack model is equivalent to a de- with the peak stress adjusted for the effect of confinement.
grading shear retention factor under monotonic loading. It is Fig. 3(d) shows the simulated load-deflection response of the
noted that, while the rotating crack model has sometimes been connection using Thorenfeldt’s compression strain-softening
considered inadequate in modeling shear transfer across crack model, which results in a decrease in the ultimate strength and
faces, Vecchio (2000) recently demonstrated a successful ap- a reduction in the unloading stiffness. No mesh dependency
plication of the model for predicting response of lightly rein- was observed when using the Thorenfeldt model for this ap-
forced and shear-critical members. plication. Fig. 3(d) also indicates that strength degradation in
The tendency of the fixed crack model to overestimate shear the simulation occurs too early in the load history, and the
transfer across crack faces is also observed in analyses of a concrete crushes prematurely. This premature strength degra-
generic cantilever using the fixed crack model with shear re- dation is attributed to the Thorenfeldt model, which does not
tention factors of 0.2 as well as 0.05 (Fig. 4). Here, the pre- account for the increase in strength and ductility provided
dicted energy dissipation capacity using the rotating crack by the passive confinement of the transverse reinforcement.
model and the fixed crack model (␤ = 0.2) differed by roughly Mander’s confined concrete model was then adopted to ac-
18%. Convergence problems arose with the use of a very low count for the confining effect of the transverse reinforcement,
shear retention factor (␤ < 0.05) due to the near-singularity of and, as seen in Fig. 3(e), crushing of the concrete no longer
the stiffness matrix. occurs prematurely. Using the confined concrete model also
The main difference in hysteretic response caused by the results in a better prediction of component strength.
1394 / JOURNAL OF STRUCTURAL ENGINEERING / DECEMBER 2001

J. Struct. Eng., 2001, 127(12): 1391-1401


Downloaded from ascelibrary.org by Li. Co.Sa 8181901/mi/155985 on 03/24/20. Copyright ASCE. For personal use only; all rights reserved.

FIG. 3. Influence of Materials Models on Load-Deflection Behavior of Beam-Column Connection IC1 [Experiment by Sritharan et al. (1996)]

FIG. 4. Influence of Crack Models on Load-Deflection Behavior of Idealized Reinforced Concrete Cantilever Beam under Cyclic Load

Influence of Bauschinger Effect a large degree, caused by the Bauschinger effect in the steel
rather than bondslip. However, it is recognized that a certain
As seen in Fig. 3(e), there is a marked change in stiffness degree of bondslip may have also contributed to the pinching
in the experimental load-deflection curve at the zero load level, in the hysteretic response.
which is not captured in the simulation. Including the Bau-
schinger effect in the reinforcing steel model results in a sim- Summary of Model Selection
ulated hysteretic response that more closely resembles the ex-
perimental response [Fig. 3(f )]. At these low ductility levels, Simulation results of the experiment by Sritharan et al. us-
it is likely that the pinching in the hysteretic response was, to ing various models and model parameters point to the selection
JOURNAL OF STRUCTURAL ENGINEERING / DECEMBER 2001 / 1395

J. Struct. Eng., 2001, 127(12): 1391-1401


of one set of models with associated parameters for further
study. The adopted set of models consists of: (1) an elasticity-
based concrete model with rotating smeared cracking in ten-
sion and strain-softening including effects of confinement, ac-
cording to Mander et al. (1988), in compression; (2) secant
unloading and reloading for concrete in both tension and com-
pression; and (3) the writers’ modified steel model based on
Su and Zhu (1994), incorporating the Bauschinger effect to
represent reinforcing steel.

MODEL EVALUATION THROUGH SIMULATION OF


EXPERIMENTS
The set of models selected based on calibration to one set
Downloaded from ascelibrary.org by Li. Co.Sa 8181901/mi/155985 on 03/24/20. Copyright ASCE. For personal use only; all rights reserved.

of experiments is further evaluated to verify its ability to pre-


dict different types of hysteretic response. Finite-element sim-
ulations are performed on three sets of experiments: (1) col-
FIG. 6. Concrete Stress-Strain Curves for Specimens U2 and U7 Using
umns with both shear and flexure-dominated behavior; (2) Mander’s Confined Concrete Model (Mander et al. 1988).
bridge connections with horizontal prestressing across the cap
beam; and (3) bridge columns with vertical unbonded postten-
sioning. cantilevers, since the column footing was posttensioned to
the reaction floor in the experiment to prevent rotation in the
Columns with Shear and Flexure-Dominated column-footing joint (Saatcioglu and Ozcebe 1989). This ide-
Behavior alization does not take into account possible foundation move-
ment within the joint region, which may result in stiffer re-
In the experiments of Saatcioglu and Ozcebe (1989), full- sponse predictions. The same number and sequence of cycles
scale reinforced concrete columns of varying transverse rein- in the experiment are used in the simulations with one addi-
forcement ratios were subjected to cyclic loading under dis- tional cycle at the end for the simulation of U2.
placement control. The writers analyzed two of the specimens.
The first was designed to fail in shear (U2) and the second Simulation Results
was designed to fail in flexure (U7). The geometry and ma-
terial properties of the test specimens are shown in Fig. 5(a). The load-deflection response from the experiments and from
Both specimens are subjected to an axial load of 600 kN and the simulations is shown in Fig. 7. The simulations capture
unidirectional displacement-controlled load reversals in mul- the influence of different shear reinforcement ratios in terms
tiples of the yield displacement. of the general shape of the load-deflection curve, the deformed
In the experiment, specimen U2 failed by the disintegration shape of the specimen, and the maximum lateral load capacity.
of core concrete above the joint region, resulting from a lack With a low transverse reinforcement ratio, U2 exhibits shear-
of concrete confinement. The experimental results for U7 dominated behavior with pinched hysteresis loops and strength
showed the specimen maintaining its peak load in a stable degradation. Specimen U7, with its higher transverse rein-
manner up to the end of the applied load history (Saatcioglu forcement ratio, exhibits flexure-dominated behavior with sta-
and Ozcebe 1989). ble hysteresis loops showing only minor degradation. The dif-
The finite-element model uses eight-noded plane-stress el- ference between the hysteretic behavior of U2 and U7 (in both
ements for the concrete and embedded bar elements for the experiments and simulations) is largely due to the different
longitudinal and transverse reinforcement [Fig. 5(b)]. The amounts of concrete confinement, which is represented in the
adopted confined concrete model provides different compres- simulations through the concrete compression model adopted
sive stress-strain behavior for the specimens (Fig. 6), dictated (Fig. 6). Indication of the shear-dominated response of U2 as
by the different amounts of transverse reinforcement in the opposed to the flexure-dominated response of U7 is also seen
specimens. The column specimens are idealized as fixed-end in the deformed shapes of these columns (Fig. 8).

FIG. 5. Column Specimen with Varying Amounts of Transverse Reinforcement

1396 / JOURNAL OF STRUCTURAL ENGINEERING / DECEMBER 2001

J. Struct. Eng., 2001, 127(12): 1391-1401


1996). Three specimens are studied. The first has only mild
reinforcement (IC1), the second has partial prestressing across
the cap beam (IC2), and the third has full prestressing con-
necting precast cap beam segments to a cast-in-place column
(IC3). The test specimens for IC2 and IC3 are similar to that
of IC1, as shown in Fig. 2. Material properties and reinforce-
ment ratios are given in Tables 2 and 3.
The finite-element models for IC2 and IC3 are similar to
that of IC1 (also shown in Fig. 2), where eight-noded plane
stress elements are used for concrete and embedded bar ele-
ments for reinforcement. Connection response is simulated for
the entire displacement history from the experiments. Repeated
cycles for the same ductility level as seen in the experiment
Downloaded from ascelibrary.org by Li. Co.Sa 8181901/mi/155985 on 03/24/20. Copyright ASCE. For personal use only; all rights reserved.

are omitted to reduce computational time. However, it is noted


that the degradation observed experimentally with repeated
loading to the same ductility level can also be captured with
the adopted set of material models (Kwan 2001).

Simulation Results
Fig. 9 compares the load-deflection response of the beam-
column connections from the experiments and the simulations.
Indicated on the figure are the ductility levels, ␮, defined as
the ratio of the displacement level to the yield displacement.
There is good general agreement between the experimental
FIG. 7. Load-Deflection Response for U2 and U7: (a) Experiment and the simulated response. The difference in the experimental
(Saatcioglu and Ozcebe 1989); (b) Finite-Element Simulation and simulated ultimate load capacities for all of the units is
within 12%. The experimental and simulated residual displace-
ments and unloading stiffnesses are very similar. In the sim-
ulation, the overall shape of the hysteresis curves is well pre-
dicted, except in the case of IC1 at a ductility level of 6 and
beyond. For IC1, neither the increased degree of pinching of
the hysteresis curve nor the drop in the load level at the duc-
tility level of 8 is captured in the simulation.
The inability of the simulation to capture the pinching of

FIG. 8. Simulated Deformed Column Shape at Ductility 3 (Magnifi-


cation = 1.3)

Strength degradation in repeated cycles is well-represented


in the simulated flexure-dominated response (specimen U7).
For the shear-dominated specimen (U2), the simulation indi-
cates severe strength degradation, but the accuracy of the mag-
nitude of degradation is not as satisfactorily captured. Since
the behavior of U7 is dominated by flexure, the steel material
model controls the overall hysteretic response more than the
concrete material model. Higher demands are imposed on the
concrete material model in the simulation of U2. These de-
mands appear to amplify the effects of the idealizations in the
model, including the unloading/reloading behavior, cyclic de-
terioration, and the representation of cracking by a smeared
rotating crack approach.

Bridge Bent Connections with Horizontal


Prestressing
Simulations are performed on cyclic experiments of bridge
bent cap beam-to-column connections, which have varying de- FIG. 9. Load-Deflection Response of Bridge Bent Cap Beam-to-Col-
grees of posttensioning across the cap beam (Sritharan et al. umn Connections

JOURNAL OF STRUCTURAL ENGINEERING / DECEMBER 2001 / 1397

J. Struct. Eng., 2001, 127(12): 1391-1401


the hysteresis curves in IC1 is attributed to the lack of repre- For the bridge connection experiments with prestressing in
sentation of bondslip as well as limitations in the adopted con- the cap beam (IC2 and IC3), the confinement introduced by
crete model to accurately represent shear distortion. The oc- the prestress results in better anchorage of the longitudinal
currence of bondslip was evidenced by physical observation column bars. The improved anchorage reduces bond deterio-
of substantial joint deformation near the end of the experiment ration and deters the slippage of the reinforcing bars. Conse-
(Sritharan et al. 1996). In simulating the response of IC1, the quently, the load-deflection behavior of IC2 and IC3 is cap-
lack of modeling bondslip (and more accurate concrete shear tured in the simulations quite accurately without modeling
distortion) first becomes an issue after a ductility level of 6. bondslip.
Prior to this, modeling the Bauschinger effect in the reinforc- The influence of the cap beam prestressing on the force
ing steel captures the slight pinching of the hysteretic response transfer mechanism within the joint is demonstrated by the
(Fig. 9). principal compressive stress contours (Fig. 10). A diagonal
The inability of the simulation to capture the strength deg- strut (compression) develops in each of the connections, with
radation in IC1 is attributed to the lack of representation of the broadest strut in the connection with the fully prestressed
Downloaded from ascelibrary.org by Li. Co.Sa 8181901/mi/155985 on 03/24/20. Copyright ASCE. For personal use only; all rights reserved.

buckling of the longitudinal steel. The buckling of the rein- cap beam (IC3) due to the high prestressing force. The broader
forcement observed near the end of the experiment, at a duc- strut enables better anchorage of the column longitudinal re-
tility level of 8 (Sritharan et al. 1996), also led to further inforcement in the joint region, which in turn enhances the
strength and stiffness degradation, which the simulation with joint performance. Such information is useful for the detailing
the adopted set of models cannot capture. of components. For instance, in the nonprestressed specimen

FIG. 10. Simulated Principal Compressive Stresses at Ductility 6

FIG. 11. Simulated Crack Patterns at Ductility 6

FIG. 12. Bridge Column Tests with Unbonded Vertical Posttensioning

1398 / JOURNAL OF STRUCTURAL ENGINEERING / DECEMBER 2001

J. Struct. Eng., 2001, 127(12): 1391-1401


TABLE 4. Material and Geometric Properties of Columns Tested by Ito et al. (1997)

f ⬘c fy, mild steel Prestress steel Prestress Shear reinforcement


Specimen (MPa) Mild steel (MPa) ( fy = 1,250 MPa) (MPa) ( fy = 354 MPa)
R1 35.6 16—␾13 mm 388 — 0 Base: ␾6 mm @ 75 mm;
P1 35.6 12—␾10 mm 409 2—␾17 mm 3.5 elsewhere: ␾6 mm @ 100 mm.
P2 35.6 8—␾6 mm 354 4—␾17 mm 7.0

(IC1) in Fig. 10, the anchorage of the extreme column tension


reinforcing bars would not be fully confined by this strut. Ev-
idence of this lesser confinement was seen in the experiment
through bondslip and pinching of the load-deflection response.
The reduction of cracking in the cap beam due to prestress-
Downloaded from ascelibrary.org by Li. Co.Sa 8181901/mi/155985 on 03/24/20. Copyright ASCE. For personal use only; all rights reserved.

ing is also well represented in the finite-element simulation


(Fig. 11). The simulated crack patterns are in good agreement
with the observed crack patterns in the experiment in terms of
their location and orientation.

Bridge Columns with Vertical Unbonded


Posttensioning
Simulations are performed on experiments of bridge col-
umns with vertical unbonded posttensioning conducted by Ito
et al. (1997) [also in Ikeda (1998)]. The experimental program
involved three column specimens. Specimen R1 had no ver-
tical prestressing and specimens P1 and P2 had unbonded ver-
tical prestressing to carry 10 and 40% of the ultimate load,
respectively. The elevation and cross sections of the test spec-
imens are shown in Fig. 12(a). The geometry and material
properties of the columns are summarized in Table 4. The
specimens were subjected to an external axial stress of 1.0
MPa and displacement cycles in multiples of the yield dis-
placement.
The finite-element model consists of eight-noded plane-
stress elements for the concrete and embedded bar elements
for the mild steel [Fig. 12(b)]. The unbonded tendons are mod-
eled with single truss elements with end nodes attached to
concrete element nodes at the anchorage locations. This ide-
alization captures the unbonded posttensioning characteristics
of steel strain remaining constant along the length of the ten-
don and steel stress being determined by the change in length FIG. 13. Load Deflection Behavior of Bridge Columns from: (a) Ex-
of tendons between anchorages. periments (Ito et al. 1996); (b) Simulations (Arrows Indicate where Buck-
The column is idealized as a fixed-end cantilever, since the ling of Longitudinal Mild Reinforcement Was Observed)
footing was posttensioned to the reaction floor in the experi-
ments to prevent joint rotations. The column-footing connec- cant strength deterioration in the experiments corresponds to
tion region is included in the model to represent the full length the buckling of the mild steel (marked by arrows on Fig. 13),
of the unbonded tendons. Pin supports are applied along the which is not represented in the finite-element models.
left and right sides of the base region to represent the con-
straint of the posttensioned footing. Effective prestress is ap- SUMMARY AND CONCLUSIONS
plied in the unbonded tendons, and the external axial load is
applied as a distributed force at the top of the column. Cyclic Finite-element simulations of structural concrete subjected
lateral displacement is then applied as in the experiment. to cyclic loading are performed to evaluate the influence of
material models and parameters on structural response. A set
Simulation Results of models is selected, based on calibration with one experi-
ment, and is used to predict the response of other structural
Load-deflection response from the experiments and from the components. The set of models selected in this process is able
simulations for the three specimens is compared in Fig. 13. In to represent flexure-dominated hysteretic behavior, indicate
the experiments, the use of unbonded posttensioning resulted potential shear dominated failure, capture local influences of
in decreased energy dissipation capacity and reduced residual bonded horizontal prestressing through joint connections, and
displacements. These effects are also observed in the simula- capture the influence of unbonded vertical prestressing in a
tions. The maximum load and the value of residual displace- bridge column. Specific conclusions include
ment in the simulations are in good agreement with the ex-
perimental results. In addition, the unbonded tendons remained • Hysteretic behavior is strongly influenced by the choice
elastic throughout the simulations, as was the case in the ex- of the crack model and the representation of shear transfer
periments. across crack faces. Using the fixed crack model with a
A feature of the experimental response that the simulations constant shear retention factor may lead to an overesti-
do not capture adequately in these experiments is the extent mation of the stiffness of a component’s cyclic response.
of strength degradation in specimens R1 and P1. The signifi- Use of a rotating crack model can result in a hysteretic
JOURNAL OF STRUCTURAL ENGINEERING / DECEMBER 2001 / 1399

J. Struct. Eng., 2001, 127(12): 1391-1401


response more representative of physical behavior. How- REFERENCES
ever, improvements in modeling cyclic shear deterioration
ASCE. (1982). State-of-the-art report on finite element analysis of rein-
in concrete constitutive models are still necessary, partic- forced concrete, New York.
ularly for capturing shear-dominated behavior. ASCE. (1993). Finite Element Anal. of Reinforced Concrete Struct. II:
• Proper representation of compression strain-softening Proc., Int. Workshop, New York.
considering passive confinement from transverse rein- Bazant, Z. P. (1983). ‘‘Comment on orthotropic models for concrete and
forcement is necessary for predicting strength degradation geomaterials.’’ J. Engrg. Mech., ASCE, 109(3), 849–865.
of structural concrete components. Carter, B. J., Wawrzynek, P. A., and Ingraffea, A. R. (2000). ‘‘Automated
3-D crack growth simulation.’’ Int. J. Numer. Methods in Engrg., 47(1–
• For the considered reinforced concrete members, the 3), 229–253.
global load-deflection response is not significantly af- Comité Euro-International Du Beton (CEB). (1996). RC elements under
fected by the idealization of the concrete compression un- cyclic loading—state of the art report, London.
loading and reloading as either elastic or secant. This is Chang, G. A., and Mander, J. B. (1994). ‘‘Seismic energy based fatigue
due to the domination of the steel behavior. This phenom- damage analysis of bridge columns. Part I: Evaluation of seismic ca-
Downloaded from ascelibrary.org by Li. Co.Sa 8181901/mi/155985 on 03/24/20. Copyright ASCE. For personal use only; all rights reserved.

enon is expected in structural concrete members with typ- pacity.’’ Tech. Rep. NCEER-94-0006, National Center for Earthquake
Engineering Research, Buffalo, N.Y.
ical reinforcement ratios. Darwin, D., and Pecknold, D. A. (1976). ‘‘Analysis of RC shear panels
• The material model of the reinforcing steel can have a under cyclic loading.’’ J. Struct. Div., ASCE, 102(2), 355–369.
significant influence on simulated hysteretic response. Not Elmorsi, M., Kianoush, M. R., and Tso, W. K. (1998). ‘‘Nonlinear anal-
representing the Bauschinger effect in the steel constitu- ysis of cyclically loaded reinforced concrete structures.’’ ACI Struct.
tive model leads to a response that is too stiff, thus over- J., 95(6), 725–739.
estimating energy dissipation capacity. It is noted that, at Feenstra, P. H., and Rots, J. G. (2000). ‘‘Comparison of concrete models
for cyclic loading.’’ ASCE Workshop Proc., ASCE, Reston, Va., 38–
high ductility levels, effects of pinching may be domi- 55.
nated by bondslip. Feenstra, P. H., Rots, J. G., Arnesen, A., Teigen, J. G., and Hoiseth, K. V.
• The adopted material models are able to represent flexure- (1998). ‘‘A 3D constitutive model for concrete based on a co-rotational
dominated behavior well and can indicate when shear- concept.’’ Proc., Euro-C 1998 Conf. on Computational Modelling of
dominated hysteretic behavior is likely. Concrete Struct., Balkema, The Netherlands, 13–22.
• Modeling the buckling of steel reinforcement is necessary Ikeda, S. (1998). ‘‘Seismic behavior of reinforced concrete columns and
improvement by vertical prestressing.’’ Proc., 13th FIP Congr. on
to fully capture the cyclic failure modes in many structural Challenges for Concrete in the Next Millennium, Balkema, The Neth-
components, in particular ones with high axial compres- erlands, 1, 879–884.
sion load and wide spacing of transverse reinforcement. Ito, T., Yamaguchi, T., and Ikeda, S. (1997). ‘‘Seismic performance of
• The influence of horizontal prestressing through a beam- concrete piers prestressed in vertical direction.’’ Proc., Japan Concrete
column joint can be simulated, capturing effects such as Inst., 19(2), 1197–1202 (in Japanese).
a decrease in beam and joint damage. The broadening of Kang, H. D., Rose, B., Shing, P. B., Spacone, E., and William, K. J.
(1998). ‘‘Seismic performance of reinforced concrete bridge piers.’’
the struts in the joint region can also be predicted, indi- Proc., Euro-C 1998 Conf. on Computational Modelling of Concrete
cating the degree of confinement for anchorage of longi- Struct., Austria, 2, 673–683.
tudinal column steel. Kwan, W.-P. (2001). ‘‘Seismic analysis, behavior and performance-based
• The influence of vertical prestressing in a bridge column, design of unbonded post-tensioned concrete substructure systems.’’
which includes a reduction in the energy dissipation ca- PhD thesis, School of Civ. and Envir. Engrg., Cornell University,
pacity and a decrease in the residual displacement, can be Ithaca, N.Y.
Lee, J., Filippou, F. C., and Fenves, G. L. (1999). ‘‘Simulation of hys-
successfully predicted. teretic behavior of reinforced concrete members.’’ Proc., ASCE Struct.
• Based on input of only material and geometric properties, Congr., ASCE, Reston, Va., 199–202.
simulations using the adopted models can generate valu- Ma, S.-Y. M., Betero, V. V., and Popov, E. P. (1976). ‘‘Experimental and
able information on global hysteretic behavior and local analytical studies on the hysteretic behavior of reinforced concrete rec-
damage. This work focused more on representing global tangular and T-beams.’’ EERC Rep. 76-02, Earthquake Engrg. Res. Ctr.,
hysteretic response. Detailed quantitative studies of local University of California, Berkeley, Calif.
Mander, J. B., Priestley, M. J. N., and Park, R. (1988). ‘‘Theoretical
damage are warranted. stress-strain model for confined concrete.’’ J. Struct. Engrg., ASCE,
• The range of behavior that finite-element models can rep- 114(8), 1804–1826.
resent as presented in this paper demonstrates that simu- Ngo, D., and Scordelis, A. C. (1967). ‘‘Finite element analysis of rein-
lation can be an efficient tool to evaluate existing struc- forced concrete beams.’’ ACI J., 82(2), 162–169.
tures and investigate new designs. Noguchi, H. (1985). ‘‘Analytical models for cyclic loading of RC mem-
bers.’’ Finite element analysis of reinforced concrete structures, Proc.,
Seminar Sponsored by the Japan Society for the Promotion of Science
This work serves as a basis for ongoing work to study the and the U.S. National Science Foundation, Tokyo, C. Meyer and H.
influence of local behavior on global structural response for Okamura, eds., ASCE, New York, 486–506.
new structural systems in particular ones with nontraditional Okamura, H., and Maekawa, K. (1991). Nonlinear analysis and consti-
hysteretic response, for which there is limited experimental tutive models of reinforced concrete, Tokyo Gihodo, Japan.
data. The material models studied here have been used by the Parker, D. R., and Dameron, R. A. (1994). ‘‘Pretest prediction analysis
of a 1/3 scale model of a China Basin viaduct foundation.’’ Proc., 3rd
writers to study the seismic behavior of precast concrete bridge
Annu. Seismic Res. Workshop, Caltrans, Sacramento, Calif.
pier systems with unbonded posttensioning at both the com- Rose, B., Shing, P. B., and Spacone, E. (1999). ‘‘Modeling of shear be-
ponent and structural system scales (Kwan 2001). havior of RC columns.’’ Proc., ASCE Struct. Congr., ASCE, Reston,
Va., 166–169.
ACKNOWLEDGMENTS Rots, J. G. (1988). ‘‘Computational modeling of concrete fracture.’’ PhD
thesis, Delft University of Technology, Delft, The Netherlands.
This research was conducted as part of Dr. W.-P. Kwan’s PhD disser- Rots, J. G., and Blaauwendraad, J. (1989). ‘‘Crack models for concrete:
tation. The writers gratefully acknowledge Prof. Ikeda of Yokohama Na- Discrete or smeared? Fixed, multi-directional or rotating?’’ HERON,
tional University and Prof. Sritharan of Iowa State University for provid- 34(1), 1–59.
ing information on their experimental work. Consultation with Dr. Peter Saatcioglu, M., and Ozcebe, G. (1989). ‘‘Response of reinforced concrete
Feenstra of the Cornell Theory Center is also gratefully acknowledged. columns to simulated seismic loading.’’ ACI Struct. J., 86(1), 3–12.
This work was partially funded by the Multi-Disciplinary Center for Schnobrich, W. C., Xu, C., and Chesi, C. (1992). ‘‘Computation of RC
Earthquake Engineering Research. The DIANA finite-element program, wall response.’’ Concrete shear in earthquake, Elsevier, New York,
Release 7.1 (developed by TNO Building and Construction Research in 320–332.
The Netherlands), was used in this research. Sittipunt, C., and Wood, S. L. (1995). ‘‘Influence of web reinforcement

1400 / JOURNAL OF STRUCTURAL ENGINEERING / DECEMBER 2001

J. Struct. Eng., 2001, 127(12): 1391-1401


on the cyclic response of structural walls.’’ ACI Struct. J., 92(6), 745– Vecchio, F. J., and Collins, M. P. (1986). ‘‘The modified compression-
756. field theory for reinforced concrete elements subjected to shear.’’ ACI
Spacone, E., Filippou, F. C., and Taucer, F. F. (1996). ‘‘Fibre beam- J., 83(2), 219–231.
column model for nonlinear analysis of R/C Frames. Part II: Applica- Vecchio, F. J., and Collins, M. P. (1993). ‘‘Compression response of
tion.’’ Earthquake Engrg. and Struct. Dyn., 25(7), 711–725. cracked reinforced concrete.’’ J. Struct. Engrg., ASCE, 119(12), 3590–
Sritharan, S., Priestley, M. J. N., and Seible, F. (1996). ‘‘Seismic response 3610.
of column/cap beam tee connections with cap beam prestressing.’’ Rep.
No. SSRP-96/09, Div. of Struct. Engrg., University of California at San
Diego, La Jolla, Calif. NOTATION
Stevens, N. J., Uzumeri, S. M., and Collins, M. P. (1991). ‘‘Reinforced
The following symbols are used in this paper:
concrete subjected to reversed cyclic shear—experiments and consti-
tutive model.’’ ACI Struct. J., 88(2), 135–146.
Su, X., and Zhu, B. (1994). ‘‘Algorithm for hysteresis analysis of pres- f c⬘ =
compressive strength of concrete;
tressed-concrete frames.’’ J. Struct. Engrg., ASCE, 120(6), 1732–1744. fy =
yield strength of steel;
Thorenfeldt, E., Thomaszewic, A., and Jensen, J. J. (1987). ‘‘Mechanical ␤ =
shear retention factor;
properties of high-strength concrete and applications in design.’’ Proc., εn,s
Downloaded from ascelibrary.org by Li. Co.Sa 8181901/mi/155985 on 03/24/20. Copyright ASCE. For personal use only; all rights reserved.

=
concrete strain in principal coordinate system;
Symp. on Utilization of High-Strength Concrete, Tapir Publishers, ␮ =
ductility measured at ratio of ultimate displacement to
Stavanger, Norway, 149–159.
Vecchio, F. J. (1999). ‘‘Towards cyclic load modeling of reinforced con- yield displacement;
crete.’’ ACI Struct. J., 96(2), 193–202. ␴n,s = concrete stress in the principal coordinate system; and
Vecchio, F. J. (2000). ‘‘Analysis of shear-critical reinforced concrete ␶␾ = angle between global coordinate system and principal co-
beams.’’ ACI Struct. J., 97(1), 102–110. ordinate system.

JOURNAL OF STRUCTURAL ENGINEERING / DECEMBER 2001 / 1401

J. Struct. Eng., 2001, 127(12): 1391-1401

You might also like