You are on page 1of 6

Proceedings of Computational Engineering Conference JSCES

Vol. 23, 2018 June

Simulation of Failure Localization and Buckling Behavior of


Steel-tube Confined Concrete Columns using Coupled
RBSM-Shell FEM Analysis
Rodolfo Mendoza Jr.1) and Yoshihito Yamamoto2)
1) Candidate, Dr. Eng. (Nagoya University, E-mail: rodolfo.mendozajr.civiledu@gmail.com)
2) Dr. Eng. Associate Professor (Nagoya University, E-mail: y.yamamoto@civil.nagoya-u.ac.jp)

A coupled model is presented to simulate the failure localization of concrete and local buckling behavior
of steel tube. The coupled model combines the three-dimensional Rigid Body Spring Model and the
MITC shell FEM element to model concrete and steel tube, respectively, in a steel-tube confined
concrete member. Simulation of published test experiments of concrete-filled steel tube confirmed the
capability of the coupled RBSM-shell FEM model in simulating these behaviors.
Key Words : Localization, Local Buckling, RBSM, MITC Shell, Steel-tube Confined Concrete

1. INTRODUCTION simulate the softening and localization of concrete under steel


The use of concrete confined by thin steel tubes (also known tube confinement.
as thin-walled concrete-filled steel tubes or CFST) presents an Contrary to concrete, there is unanimity in the modeling of
economical means of improving the axial capacity of concrete steel behavior where researchers used either elastic-perfectly
by means of confinement. The analysis of such system, plastic model [3, 4] or elastic-plastic model with linear
however, poses some challenges in particular in the simulation hardening [5]. The variation in the results from the used of
of failure localization of concrete and local buckling of steel these models is, however, minimal [5]. Existing numerical
tube. When the level of confinement provided by the steel models have also shown promising results in terms of
tubes is low, the compressive post-peak behavior of concrete is simulating local buckling effects in steel tubes (e.g., Hu et al.
localized which results in the softening of the axial [4]). Realistic representations of concrete failure behavior
stress-strain curve. This behavior affects the axial ductility of within the buckled region of steel tube from these models,
the composite which—because of compression however, are still far from being acceptable.
localization—induces size (height) effect, i.e., under the same In summary, although several studies have been performed
level of confinement, long columns tend to have lower axial to produce numerical models that can capture the compressive
ductility as compared to shorter columns [1]. Similarly, due to behavior of concrete-filled steel tube, a numerical model that
local buckling of steel tube, its axial contribution and can both simulate the failure localization of concrete and local
confinement effectiveness is expected to be less which may buckling behavior of steel tube has yet to be presented. In this
also result in the reduction of composites’ axial ductility. It is, paper, a coupled numerical model is presented to address this
therefore, important to accurately simulate these behaviors in limitation.
order to properly assess the post-peak performance of this type The coupled numerical model combines the
of system. three-dimensional Rigid Body Spring Model (RBSM) and a
Conventionally, the simulation of concrete under steel tube four-node isoparametric shell element to model the concrete
confinement is performed using finite element analysis. and steel, respectively, in a steel-tube confined concrete
Concrete is commonly modeled using solid element (e.g., [2, member. The use of RBSM is aimed at modeling the
3] while the steel is modeled either using shell [2] or solid localization of deformation, strain softening, and failure
element [3]. The material modeling of concrete is, however, behavior of concrete under steel tube confinement. The shell
still far from having a consensus, as different numerical element, on the other hand, is intended to simulate the local
models have used different approaches in modeling concrete. buckling behavior of steel tube through its large displacement
In most cases, these concrete models lack flexibility as they and large rotation feature. In the succeeding sections, the
can only be applied to the presented problem at hand. In details of the coupled model development are discussed and its
particular, in the authors’ knowledge, there are still no performance in simulating the axial compressive behavior of
available numerical models in the literature that can both concrete-filled steel tube columns is presented.
Fig. 1 Constitutive models of concrete using RBSM

2. MODELING OF CONCRETE USING RBSM program follows closely the geometric nonlinear shell
In modeling concrete using RBSM, concrete is idealized as formulation of Noguchi and Hisada [7] in which an integrated
an assemblage of rigid particles interconnected by normal and Total and Updated Lagrangian Method, used in solving
shear springs. The method is based on a mesoscale, single large-displacement problems, was used in the FEM
phase averaging model of concrete. Here, we utilized the formulation. In addition, large rotation increments is also
RBSM formulated and developed by Yamamoto et al. [6]. The considered by including the quadratic term of the Taylor
novelty of this method is centered on the evaluation of expansion of a finite rotation tensor. An elastoplastic model
inter-particle interaction and the material constitutive models with linear isotropic hardening was used to model the behavior
for concrete—which are specifically designed to capture the of the steel material with failure surface following the classical
softening and localization fracture behavior of concrete [6]. von Mises yield criterion.
The constitutive models of concrete using RBSM for tension,
compression, and shear are given in Fig. 1. A complete
description of these models can be found in Yamamoto et al.
[6]. An important note about these models is the non-existence
of the phenomenological softening in the compression
constitutive model. The softening and localization behavior of
concrete, however, can be well-represented through the
combined interaction of the compression and shear constitutive
models. This ability of the model to capture these behaviors
may be explained by the fact that at the mesoscale level, the
compression behavior of concrete is a combination of tension,
compressive, and shear behavior. These developed models by
Yamamoto et al. [6] have been proven to capture these
mesoscale interactions.

3. MODELING OF STEEL TUBE USING MITC


SHELL ELEMENT
The steel tube is modeled using a four-node, isoparametric,
degenerated shell element shown in Fig. 2. The developed Fig. 2 Four-node isoparametric shell element
Fig. 3 The RBSM-Shell FEM Model

The shear locking problem commonly observed in shell desired isoparametric coordinates in terms of the Jacobian
elements was solved by introducing assumed strains based on matrix, known shell coordinates, and known contact point
Mixed Interpolation Tensorial Components (MITC) functions coordinates as given in Eq. (3). Here, the Jacobian matrix J in
[8] given in Eqs. (1) and (2). Eq. (4) consists of the components of covariant base vectors Gi.
1 1 Equation (3) is solved using Newton-Raphson iterative
1 1 (1)
2 2 solution. An important feature of this approach is that it
eliminates the mesh size ratio (ratio of RBSM element size to
1 1
1 1 (2) shell element size) dependency which was found to be present
2 2
when contact points are assigned directly to the shell nodes.
It should be noted that the selective reduced integration (SRI)
scheme in solving locking problem in shell elements was also (3)
initially employed and found that for simulations under
uniaxial compression, the MITC and SRI schemes will provide
similar predictions. The performance of the developed shell
(4)
program was validated with benchmark models for a
geometrically nonlinear shell with good agreement.

4. THE COUPLED RBSM-SHELL FEM MODEL One interface point (contact point) consists of one normal
The coupled RBSM-Shell FEM model is depicted in Fig. 3 and two shear springs. Using these springs, one-dimensional
using a circular, steel-tube confined concrete column. To constitutive models can be employed to simulate the
provide interaction between the RBSM and shell element, that concrete-steel interface interaction. Here, an elastic
is, to simulate the concrete-steel interface interaction, discrete constitutive model is adopted for normal spring while a
interface elements are used. These interface elements are bi-linear model was adopted for the shear springs. These
pre-set on the surface of RBSM in contact with shell elements constitutive models are shown in Fig. 4. The linear model of
with coordinates corresponding to the centroid of triangles the normal spring represents a constant increase in normal
which were generated by dividing the RBSM surface into sets pressure under increasing compression contact. The normal
of triangles. This is similar to the generation of integration spring, however, has no resistance to tension to allow the
points on RBSM particles’ boundary surface. With this separation of RBSM and shell element under tension loading.
approach, however, the isoparametric coordinates at contact On the other hand, the bilinear model of the shear springs
points—used in calculating the relative displacements between represents two conditions: the hardening model represents the
RBSM and shell—for the shell elements are initial unknown, increase in shear interaction between concrete and steel with
and therefore, evaluation of the displacements using shape increasing contact (increasing compression force in the normal
functions will not be directly possible. We used an inverse spring), whereas the softening model represents the decrease in
isoparametric mapping using an approximate (higher order shear interaction between concrete and steel as separation
terms neglected) Taylor Expansion rearranged to express the between elements initiates. The relation between the normal
(compression) spring and the resultant shear springs is given 2.00 D = 190 mm, f’c = 41.0 MPa, Ec = 17.81 Gpa
by the Mohr-Coulomb envelope. The coefficient of friction is Es = 204 GPa
1.60
taken as equal to 0.25. This value, although within the range of S20CS50A
friction coefficient between concrete and steel as reported in s = 4.00%

fcc / f’c
1.20
the literature [3], has minimal effect on uniaxial compression
simulation. 0.80 s = 1.80%

S10CS50A
0.40 Test
RBSM-Shell

0 0.003 0.006 0.009 0.012 0.015


Axial Strain (mm/mm)
Fig. 5 Axial stress-strain curve comparison
between test and simulation of thin and
thick-walled circular CFST

height equivalent to the height of the specimen. The


comparisons of axial stress-strain curve between test and
simulation results for the two specimens are given in Fig. 5. In
the test, the normalized stress was used to show the strength
increase in each specimen with increasing steel ratio. The
contribution of concrete in the axial strength of the composite
Fig. 4 Constitutive Model for Interface Springs
was obtained incrementally by first determining the steel
5. SIMULATION OF SOFTENING AND contribution through measured strains in the steel tube
LOCALIZATION OF CONCRETE UNDER combined with the yield failure surface obtained from the
STEEL-TUBE CONFINEMENT coupon test [9]. In the simulation, the confined concrete
To present the capability of the coupled RBSM-Shell FEM strength (fcc) was obtained directly from the output of the top
model in simulating the softening and localization of concrete rigid plate element. In both simulation and test, this confined
under steel tube confinement, we simulate the test performed concrete strength value is then divided by the unconfined
by O’Shea and Bridge [9]. In particular, we investigate the concrete strength (f’c) to obtain the normalized stress. Also,
case of a concentrically loaded thin-walled (S10CS50A) and the steel ratio (s) of each specimen (defined as the ratio of the
relatively thick-walled (S20CS50A) CFST to show how the cross-sectional area of the steel tube to the combined
effect of steel ratio can transition the axial stress-strain curve cross-sectional area of concrete and steel) are also given in Fig.
of the composite from softening into a ductile response, and 5. It can be seen from the comparison that the RBSM-Shell
simultaneously, the elimination of localized failure with an FEM model can reasonably capture the axial stress-strain
increased confinement. The properties of these specimens are curve of both thick and thin-walled CFST. In particular, the
shown in Fig. X. The two specimens basically have the same model can well-represent the transition of the post-peak
properties except for the yield strength of steel (Fy for stress-strain curve from softening into a relatively ductile
S10CS50A = 210.7 MPa; Fy for S20CS50A = 256.4 MPa). In response as the steel ratio is increased from 1.80% to 4.0%.
the test, the vertical displacements between the top and bottom To show the effect of this observed transition in the fracture
platens were measured using three displacement transducers. behavior of the specimens, the simulated deformed behaviors
The axial strains were determined by dividing the measured (at 1.2% global axial strain) of the two specimens are shown in
displacements with the platen-to-platen distance. The load, on Fig. 6. It is clear from the simulation results how an increase in
the other hand, was directly obtained from load cell recordings. steel ratio can change the failure of concrete from a localized
In the simulation, a displacement control scheme was used to failure into a more or less uniform failure. This change
apply the load with an increment of 0.005 mm/ step. Top and corresponds to the transition of the axial stress-strain curve of
bottom rigid plate elements were introduced in which the top the specimen from softening into a ductile response, that is, the
plate element was used to apply the displacement to the RBSM axial ductility of the member is increased with increasing steel
elements while for the shell elements the displacements were ratio. To date, although some numerical models based on FEM
applied directly at corresponding top shell nodes. The axial can simulate softening of the stress-strain curve of concrete
strains in the simulations were calculated considering a gauge under steel tube confinement (see Ref. [5]), existing models
Fig. 6 Simulation of deformed behavior for thin
and thick-walled circular CFST

have shown less progress in the simulation of localization of


deformation in concrete. The presented RBSM-shell FEM
model shows that it can represent these two important
behaviors.

6. SIMULATION OF LOCAL BUCKLING IN


STEEL TUBE WITH CONCRETE INFILL
Local buckling of steel tube in a CFST member is more Fig. 7 Simulation of axial load-strain curve for
pronounced in rectangular cross-section than in circular square CFST and local buckling initiation
cross-section. To show the capability of the RBSM–shell FEM
was investigated by cutting the specimen half at regions where
model to simulate this behavior, numerical simulation of test
local buckling was observed after the completion of loading.
performed by Tomii et al. [10] and Schneider [2] is presented
The comparison of the axial load-strain curve between test and
in this section. First, we present the comparison of the axial
simulation results is shown in Fig. 8. A good agreement can be
load-strain curve between test and simulation results together
generally observed in terms of the prediction of initial elastic
with the comparison of local buckling initiation. The definition
modulus, strength, and post-peak stress-strain curve although a
of axial strain and the simulation conditions used in this
slight strain hardening was observed in the simulation result
analysis are similar to those described in Section 5. Fig. 7
which can be attributed to the assumed elastic model for the
shows the axial load-strain comparison between the simulation
normal spring interface which allows a constant increase in
results and test results performed by Tomii et al. [10]. The
compression load. In Fig. 9, the comparison of deformed
properties of the specimens are given in the respective figures.
shapes at failure between test and simulation result are shown.
From the figure, it can be seen that the RBSM-shell FEM
The comparison includes photographs from test results that
model provides a reasonable prediction of the strength and
show the deformed behavior of concrete at cut regions of local
post-peak capacity of square CFST although the prediction of
buckling.
initial elastic modulus is slightly overestimated. The prediction
for the initiation of local buckling was performed by
evaluating the deformed behavior at every 10 step interval.
The gray inverted triangle marks the observed initiation of
local buckling in the test while the red inverted triangle
represents the observed initiation of local buckling in the
simulation. It can be seen that for both specimens, the
simulation results predict well the onset of local buckling in
the steel tube. The corresponding deformed behavior of
concrete at these levels is also shown. It can be seen that
concrete deformation tends to concentrate in the observed
buckled region. To confirm this observed concrete behavior in
regions of local buckling, we simulate the test (Specimen S-3) Fig. 8 Comparison of simulation and test results
for specimen S-3 by Schneider [2]
performed by Schneider [2] where the behavior of concrete
Fig. 9 Comparison of deformed shapes at failure for specimen S-3

The simulated deformed shape of concrete from RBSM coupled RBSM-shell FEM model has successfully addressed
analysis is also provided. The simulation result shows a good this limitation.
agreement in terms of predicting the local buckling observed
REFERENCES
in the test. In the test, local buckling was observed on almost
[1] Federation internationale du beton (fib): Practitioners’
all sides of the specimen and the simulation result likewise
guide to finite element modeling of reinforced concrete
shows a similar tendency. From the photographs provided in
structures. pp 86-87, 2008
the test at sections where the specimen was cut open, it can be
[2] Schneider, S.P.: Axially loaded concrete-filled steel
observed that contact between concrete and steel was
tubes, J. Struct. Eng., 124(10), pp 1125-1138, 1998
maintained and concrete appears to “flow plastically” (as
[3] Johansson, M., Gyltoft, K.: Mechanical behaviour of
described by Schneider [2]) into the buckled region. The same
circular steel-concrete composite stub columns. J. Struct.
tendency can be observed in the simulation result, where at
Eng., 128(8): 1073-1081, 2002
regions of local buckling, concrete tends to fill the shape of the
[4] Hu H.T., Huang C.S., Wu M.H., and Wu Y.M.:
buckled steel wall region. This comparison confirms the
Nonlinear analysis of axially loaded concrete-filled tube
capability of the RBSM-shell FEM model to represent this
columns with confinement effect, J. Struct. Eng., ASCE,
behavior of concrete when local buckling of steel tube occurs
129(10), pp. 1322-1329, 2003
in steel-tube confined concrete members.
[5] Tao, Z., Wang, Z.B., Yu, Q.: Finite element modeling of

7. CONCLUSION concrete-filled steel stub columns under axial

The capability of the RBSM-shell FEM model to simulate compression. J. Const. Steel R., 89, pp.121-131, 2013.

the failure localization of concrete and local buckling of steel [6] Yamamoto Y., Nakamura H., Kuroda I., Furuya N.:

tube in a steel-tube confined concrete columns was presented Analysis of compression failure of concrete by

through simulation of some experiments in the literature. It three-dimensional rigid body spring model. Doboku

was confirmed that the coupled model was effective in Gakkai Ronbunshuu, 64(4), pp. 612-630, 2008

simulating these behaviors as well as the prediction of the axial [7] Noguchi H., Hisada T.: Integrated FEM formulation for

load-strain curve which includes predictions of initial stiffness, total/updated Lagrangian Method in Geometrically

strength, and post-peak stress-strain behavior. Although some Nonlinear Problems. JSME Intl Journal, 38:1, 1995
[8] Bathe KJ. Finite Element Procedures. Prentice-Hall, Inc;
numerical models for steel-tube confined concrete members in 1996
the literature can simulate the local buckling behavior of the [9] O’Shea, M.D., Bridge, R.Q.: Design of circular
steel tube or the softening of the stress-strain curve, to the thin-walled concrete-filled steel tubes. J. Struct.
authors’ knowledge, none of the models can provide a realistic Eng.,126(11), 1295-1303, 2000
prediction of concrete failure localization. Moreover, no [10] Tomii, M., Yoshimura, K., Morishita, Y.: Experimental
models have been presented that can capture both the failure studies on concrete-filled steel tubular stub columns
localization of concrete and local buckling behavior of steel under concentric loading. Stability of Structures under
tube in a steel-tube confined concrete member. The presented Static and Dynamic Loads, ASCE, pp. 718-739, 1977

You might also like