Accepted Manuscript

Multi-scale Modeling Approach to Predict the Nonlinear

Behavior of CNT-reinforced Concrete Columns Subjected to
Service Loading

Mehdi Eftekhari, Ali Karrech, Mohamed Elchalakani, Hakan

Basarir

PII: S2352-0124(18)30040-7
DOI: doi:10.1016/j.istruc.2018.03.015
Reference: ISTRUC 271
To appear in: Structures
Received date: 1 February 2018
Revised date: 26 March 2018
Accepted date: 27 March 2018

Please cite this article as: Mehdi Eftekhari, Ali Karrech, Mohamed Elchalakani, Hakan
Basarir , Multi-scale Modeling Approach to Predict the Nonlinear Behavior of CNT-
reinforced Concrete Columns Subjected to Service Loading. The address for the
corresponding author was captured as affiliation for all authors. Please check if
appropriate. Istruc(2017), doi:10.1016/j.istruc.2018.03.015

This is a PDF file of an unedited manuscript that has been accepted for publication. As
a service to our customers we are providing this early version of the manuscript. The
manuscript will undergo copyediting, typesetting, and review of the resulting proof before
it is published in its final form. Please note that during the production process errors may
be discovered which could affect the content, and all legal disclaimers that apply to the
journal pertain.
 ACCEPTED MANUSCRIPT

Multi-scale modeling approach to predict the nonlinear behavior of CNT-

reinforced concrete columns subjected to service loading

Mehdi Eftekharia, Ali Karrecha, Mohamed Elchalakania, Hakan Basarira

a
School of Civil, Environmental and Mining Engineering, The University of Western Australia, 35

Stirling Highway, Perth, WA 6009, Australia

PT
Abstract

RI
Due to the susceptibility of concrete structures to severe damage under earthquake loading,

SC
an increasing demand on high-strength and high performance concrete materials has been

arising recently. Among several new materials, carbon nanotubes (CNTs) with their superior
NU
mechanical properties are being widely used in construction industry. In this study, a multi-

scale method is proposed to identify the mechanical properties of CNT-reinforced concrete

MA

columns under monotonic loading. At the nano-scale, the mechanical properties of CNTs are

obtained using a molecular dynamics approach. These properties are fed into a larger scale
D

framework to estimate the ultimate tensile and compressive strengths of CNT-reinforced

E

cement paste based on the finite element simulation. Furthermore, the mechanical properties
PT

of a concrete sample, including three major phases, cement, aggregate and interfacial
CE

transition zone is derived at the meso-scale using finite element simulations. Ultimately, the

monotonic behavior of two concrete columns containing ordinary concrete and CNT-
AC

reinforced concrete is investigated by OpenSees software. A comparision between the

responses of the two columns indicates that, adding CNTs to the concrete columns can

increase the key mechanical properties of concrete such as ultimate moment capacity, energy

absorption capacity and curvature ductility.

Keyword: Multi-scale modeling; Carbon nanotubes; Finite element analysis (FEA);

Mechanical properties; Nano composites.

1
 ACCEPTED MANUSCRIPT

1. Introduction

Severe damages can be imposed to concrete structures during earthquakes [1-3].

Consequently, researchers and structural engineers are having increasing interest in

producing efficient and effective types of concrete materials with enhanced mechanical

properties to mitigate earthquake-triggered mechanical damage. Among different concrete

PT
members, the role of columns at the first story of the buildings is considerably critical for

proper seismic behavior of concrete structures during an earthquake. Hence, they should

RI
tolerate high ductility, load-carrying capacity and large plastic deformation while their

SC
strength remains almost at the same level. Yet, the strength of the concrete is degraded in the

plastic hinges of the concrete columns due to low tensile strength and brittleness, which leads
NU
to overall instability of concrete structure under severe loading conditions. As a result, it is

desirable to produce new concrete material with high-energy absorption capacity without loss
MA

of strength.
D

In the last few years, researchers proposed several methods to increase the overall ductility,
E

damage tolerance and strength of concrete material during cyclic loading, which lead to
PT

different types of fiber reinforced concretes (FRC) [4, 5]. FRC can improve the ductility of
CE

concrete during reverse cyclic loading, in addition to their positive impact on crack bridging

and energy dissipation.

AC

Fischer and Li [6] studied experimentally the tensile and bending behavior of FRC members

under cyclic loading and showed that the interaction among cement and fibers governs the

behavior of FRCs. In addition, steel fiber can reduce the sliding between concrete and steel

rebars and can increase the ultimate strength and ductility of the concrete members.

Furthermore, FRCs can prevent the generation of cracks at the macro scale by delaying the

formation of defects at the micro scale. Enhancing the cyclic behavior of concrete columns

2
 ACCEPTED MANUSCRIPT

under earthquake loading by incorporating steel fiber reinforced concrete (SFRC) has been

investigated by Lee [7]. Their results indicate that SFRCs can eliminate the brittle shear

fracture behavior of concrete, provided that they are uniformly distributed. This means that

under cyclic loading, the longitudinal bar yield sooner than shear reinforcement.

Nano reinforcing material called, carbon nanotubes (CNTs), are attracting a considerable

PT
interest as a new class of concrete reinforcement material. CNTs were first invented by Iijima

[8] in 1991. They are made of carbon atoms, which form a hollow cylindrical structure at the

RI
nanoscale. They have high mechanical properties such as tensile and compressive strength

SC
(around 100GPa) and elastic modulus (around 1000GPa), which makes them superior

reinforcing materials. The main issue in using CNTs in concrete is their uniform dispersion
NU
within the concrete mix, as they tend to agglomerate rather than disperse uniformly. To

address this problem, the so called cement hybrid material (CHM) is commonly used [9].
MA

CHM can be easily mixed with Portland cement and leads in uniform dispersion of CNTs all

over the concrete volume [9].

E D

In this study, by means of finite element simulation, the effect of addition of new type of
PT

reinforcement at the nanoscale, called carbon nanotube (CNT), on the behavior of CNT-
CE

reinforced concrete column is investigated.

To the author’s best knowledge, no studies have been performed on the mechanical properties
AC

of CNT-reinforced concrete columns under monotonic loading. In this paper a hierarchical

multi-scaling approach is adopted to investigate the mechanical properties of CNT-reinforced

concrete embedded in service structures. In Section 2, molecular dynamics simulations of

CNTs are presented at the nanoscale. Afterwards, by means of finite element simulation,

tensile and compressive behavior of CNT-reinforced cement and concrete are derived in

Sections 3 and 4, respectively. At the end, two concrete columns, made of ordinary concrete

3
 ACCEPTED MANUSCRIPT

and CNT-reinforced concrete, are simulated at the macro scale and their mechanical behavior

is compared. A summary of the simulation framework is depicted in Figure 1.

PT
RI
SC
NU
MA
D

Figure 1: Schematic representation for the multi-scale simulation of the monotonic behavior of
E
PT

a CNT-reinforced concrete column.

2. Molecular dynamics simulation of CNTs

CE

The Large-scale Atomic/Molecular Massively Parallel Simulator (LAMMPS[10]) is used to

AC

simulate the behavior of carbon nanotubes under different tensile and compressive loadings.

This tool, which uses the molecular dynamics (MD) approach, is used to obtain the

mechanical properties of carbon nanotubes. The simulations were performed at room

temperature to be close to the real condition (300K). The modified Tersoff interatomic

potential function [11] is adopted to simulate the repulsive/attractive interaction among

carbon atoms. This potential is a 3-body potential which includes stretching and angle-

4
 ACCEPTED MANUSCRIPT

bending among carbon atoms. This potential does not contain dihedral component of

potential energy. In this potential the total energy of the system is described as:

𝑇𝑒𝑟𝑠𝑜𝑓𝑓 1 (1)
𝐸𝑖𝑗 = 𝑓𝐶 (𝑟𝑖𝑗 )[𝑓𝑅 (𝑟𝑖𝑗 ) + 𝑏𝑖𝑗 𝑓𝐴 (𝑟𝑖𝑗 )]
2

𝑓𝑅 (𝑟𝑖𝑗 ) = 𝐴𝑖𝑗 𝑒𝑥𝑝(−𝜆𝑖𝑗 𝑟𝑖𝑗 ) , 𝑓𝐴 (𝑟𝑖𝑗 ) = −𝐵𝑖𝑗 𝑒𝑥𝑝(−𝜇𝑖𝑗 𝑟𝑖𝑗 ) (2)

1, 𝑟𝑖𝑗 ≤ 𝑅𝑖𝑗 (3)

PT
1 1
𝑓𝐶 (𝑟𝑖𝑗 ) = + 𝑐𝑜𝑠[𝜋(𝑟𝑖𝑗 − 𝑅𝑖𝑗 )/(𝑆𝑖𝑗 − 𝑅𝑖𝑗 )] , 𝑅𝑖𝑗 < 𝑟𝑖𝑗 < 𝑆𝑖𝑗
2 2
{0, 𝑟𝑖𝑗 ≥ 𝑆𝑖𝑗

RI
Where 𝑟𝑖𝑗 is the interatomic distance between atoms i and j, functions 𝑓𝑅 (𝑟𝑖𝑗 ) and 𝑓𝐴 (𝑟𝑖𝑗 ) are

SC
the repulsive and attractive interactions, respectively, and 𝑓𝐶 (𝑟𝑖𝑗 ) is a smooth cut-off
NU
function. Further detail can be found in [11]. This potential provides reliable results as shown

in previous research [12-15].

MA

The structural geometry of CNTs can be defined by a vector (n,m), where if n and m are

equal, the CNT is called an armchair, while in the case of m=0, the CNT is called zigzag
D

(Figure 2).
E
PT
CE
AC

5
 ACCEPTED MANUSCRIPT

Figure 2: The (n,m) nanotube naming scheme can be thought of as a vector (𝑪𝒉 ) in an infinite

graphene sheet that describes how to "roll up" the graphene sheet to make the nanotube. T

denotes the tube axis, and 𝒂𝟏 and 𝒂𝟐 are the unit vectors of graphene in real space.

For verification, a perfect (10,0) zigzag SWCNT which was previously simulated by Poelma

et al. [16] is considered. The length and the diameter of the SWCNT are 52.62 and 7.83Å,

respectively. The results of the simulation illustrates a reasonable agreement between the two

PT
models (Figure 3).

RI
SC
NU
MA
D

Figure 3: Force-strain curve for (10,0) zigzag under compression

E
PT

Different configurations of CNTs were simulated, including single and multi-walled carbon

nanotubes (SWCNT and MWCNT) with equal lengths of 100Å. The detailed geometrical
CE

properties of CNTs are described in Table 1. The CNTs are restrained at one end, while the
AC

proper displacement is applied to the other and the corresponding stress strain is extracted at

each time step during the simulation.

Table 1 : Geometrical properties of used CNTs.

Type Diameter (Å) Length/Diameter

Armchair (10,10) 6.78 14.75
Zigzag (17,0) 6.65 15.04
Armchair (10,10)/(15,15) 10.17 9.83
Zigzag (17,0)/(26,0) 10.18 9.82

6
 ACCEPTED MANUSCRIPT

The mechanical properties of the tensile and compressive simulation are presented in Figure

4, and Tables 2 and 3. The average value of the results of simulation are used as the input for

simulation at the micro scale. The results indicate that the compressive elastic moduli of

CNTs are larger than tensile moduli, while their tensile strengths are much higher than their

compressive strengths. A very important point regarding their mechanical properties is that

their ultimate tensile strain is around 0.15 while their ultimate compressive strain is about

PT
0.035.

RI
SC
NU
a)
MA
E D
PT
CE

b)
AC

Figure 4: Stress-strain curves for CNTs under a) tension, b) compression.

Table 2: Tensile properties of CNTs.

Type E (GPa) ν Ft (GPa) Gft (N/mm)

7
 ACCEPTED MANUSCRIPT

Armchair (10,10) 1130 0.14 128.68 13180.964

Zigzag (17,0) 1124 0.13 102.29 8021.783

Armchair (10,10)/(15,15) 1087 0.14 125.84 12264.589

Zigzag (17,0)/(26,0) 1068 0.13 99.88 7732.911

PT
Table 3: Compressive properties of CNTs.

Type E (GPa) Buckling stress (GPa) Gfc (N/m)

RI
Armchair (10,10) 1280.8 43.55 1910229

SC
Zigzag (17,0) 1283.6 NU 45.54 1974698

Armchair (10,10)/(15,15) 1247 43.35 2015685

Zigzag (17,0)/(26,0) 1222 40.38 1923232

MA

An interesting point regarding the compressive behavior of CNTs is that the local shell
D

buckling mode happens in short CNTs while Euler buckling is dominant in long CNTs, as
E
PT

illustrated in Figure 5 [14].

CE
AC

Figure 5: Deformation of CNT in a) Tension, b) Compression.

3. Micromechanical simulation of CNT-reinforced cement

8
 ACCEPTED MANUSCRIPT

At the micro scale, we created a cement hydration model of dimensions of 50×50×1μm

embedding CNTs with random lengths and orientations distributed on the surfaces of cement

particles. The algorithm of Matsumoto and Nishimura [17] is used to generate the random

distribution of CNTs,. The cement hydration model contains three major phases, called

hydration products, unhydrated products and porosity (Figure 6). Researchers used different

methods to implement CNTs on the surface of cement model using the finite element

PT
simulation [18, 19], including the truss model, which is adopted in the current study.

RI
SC
NU
MA
E D
PT

Figure 6: Schematic mesh generation of the cement paste reinforced by CNTs

CE

A nonlinear plane stress simulation is performed in 1000 steps with the Newton-Raphson

solver. The simulation is performed by the open source finite element package OOFEM [20],
AC

which is a C++ open-source code. The model is incrementally loaded by the applied strain

and the corresponding stress is calculated.

An isotropic damage model for tensile failure with liner softening [21] is used to simulate the

behavior of cement and concrete at micro and meso scales,. This model assumes that the

stiffness degradation is isotropic, i.e., stiffness moduli corresponding to different directions

decrease proportionally and independently of the loading direction. The damaged stiffness

9
 ACCEPTED MANUSCRIPT

tensor is expressed as 𝐷 = (1 − 𝜔)𝐷𝑒 where 𝜔 is a scalar damage variable and 𝐷𝑒 is the

elastic stiffness tensor. The damage evolution law is postulated in an explicit form, relating

the damage variable ω to the largest previously reached equivalent strain level, κ. The Mazars

equivalent strain,(𝜀̃), is a scalar measure derived from the strain tensor:

𝜀̃ = √∑3𝐼=1〈𝜀𝐼 〉2 (4)

PT
Where 〈𝜀𝐼 〉 are positive parts of principal values of the strain tensor 𝜀. In this material model

RI
the stress is defined as:

SC
𝜎 = (1 − 𝜔)𝐸𝜀 (5)
Where, 𝜎 is stress and E is elastic modulus. The scalar damage variable is defined as:
NU
𝜀𝑓 𝜀 (6)
𝜔= (1 − 𝜀 )
MA

𝜀𝑓 −𝜀0 0
Where 𝜀0 and 𝜀𝑓 is equivalent strains at peak stress and failure, respectively.
E D
PT
CE
AC

Figure 7 : Schematic isotropic damage model for tensile failure with liner softening [21].

Smilauer et al. [22] performed the experimental test on the CNT-based cement material called

CHM (Cement Hybrid Material). In CHM, CNTs are directly synthesized on the surface of

cement grain. Their cement includes CNTs with the length of 3μm and diameter of 30nm. In

order to verify the micromechanical model, the experimental results of Smilauer et al. [9],

which contains 3.47% volume fraction of 3𝜇𝑚 − 𝐶𝑁𝑇 (CEM86%+CHM14%) is simulated.

10
 ACCEPTED MANUSCRIPT

The results indicated that the fracture energy of the experimental sample were around

15 − 25𝑁/𝑚 while this value for the micromechanical simulation were around21.7𝑁/𝑚 ,

which have good agreement (Figure 6).

PT
RI
SC
NU
MA

Figure 8: Results of the micromechanical model and the experimental data.

D

A pure cement model and three CNT-reinforced cement models with CNT volume fractions
E

of 1%, 2% and 3% and CNT length of 5μm, called CNT-1%, CNT-2% and CNT-3%,
PT

respectively, are considered in this study. The results of the tensile and compressive
CE

numerical tests are presented in Table 4, Figure 9 and Figure 10, where 𝐺𝑓𝑡 and 𝐺𝑓𝑐 are the

tensile and compressive fracture energies, E is the elastic modulus and 𝑓𝑡 and 𝑓𝑐 are the
AC

tensile and compressive strengths, respectively. The results of simulations revealed that in

CNT-reinforced cements more elements are damaged than pure cement, indicating that CNT-

reinforced cement can dissipate more energy compared with pure cement (Figure 9).

Furthermore, the addition of CNTs does not increase the elastic modulus of CNT-reinforced

cement significantly, while the tensile and compressive strength of CNT-3% are enhanced by

23% and 19%, respectively. The fracture energy of CNT-reinforced cement at tension and

compressive are also enhanced considerably: The tensile fracture energy increases from

11
 ACCEPTED MANUSCRIPT

5.89N/m to 46.52N/m and the compressive fracture energy increased from 825N/m to

1038N/m. More information regarding the simulation procedure and detailed results can be

found in [12, 23]. The results of micro-scale simulation are up-scaled to the meso-scale

model will be discussed in the next section.

Table 4: Mechanical properties of pure and CNT-reinforced cements.

PT
Type E (MPa) ft (MPa) fc (MPa) Gft (N/mm) Gfc (N/mm)

Cement 18496 2.29 28.47 0.00589 0.825

RI
CNT-1% 18568 2.46 29.05 0.00932 0.850

SC
CNT-2% 18620 2.45 30.95
NU 0.01991 1.032

CNT- 3% 18811 2.95 33.97 0.04652 1.038

MA
E D
PT
CE
AC

Figure 9: Damage distribution contour for the cement and CNT-reinforced cement with 3%

volume fraction and 5μm CNT; (a) Tension; (b) Compression.

12
 ACCEPTED MANUSCRIPT

a)

PT
RI
b)
SC
NU
MA

Figure 10: Stress-strain curve of CNT-reinforced cement by different volume fraction of CNT;
D

a) Tension, b) Compression.
E
PT

4. Finite element simulation of CNT-reinforced concrete at the meso scale

CE

For the meso-scale simulation, a 2D concrete sample with dimension of 100×100mm is

simulated. The sample contains three major phases, including, aggregates, cement and an
AC

interface transition zone (ITZ). As researchers showed that the effect of small aggregates is

negligible [24], only large aggregates are simulated in this research. Aggregates with

different sizes and volume fractions, including dimensions of 10mm, 5mm, 2.5mm and

volume fraction ratio of 5:3:2 are considered. The effects of aggregate shapes have been

investigated by other researchers [25] whose results indicate that the mechanical properties of

hardened concrete is independent of aggregate shapes. As a result, only aggregates with

13
 ACCEPTED MANUSCRIPT

circulars shapes are presented in this paper. The models including pure cement and CNT-

reinforced cement with CNT volume fraction of 1%, 2% and 3%, are simulated under tensile

and compressive loading in a plane stress condition.

The material behavior of cement is assumed to include isotropic damage model with a linear

softening. In addition, the material behavior of ITZ is assumed to be the same as cement with

PT
a reduction factor of 70% and a thickness value of 200µm [25]. Since aggregates usually

remain intact during loading, a linear elastic behavior with a Young’s modulus E=30GPa and

RI
a Poisson’s ratio ν=0.2 is considered for them [26].

SC
Different volume fractions of aggregates are considered to evaluate the mechanical behavior

of concrete, including 53% and 45% [26, 27]. Hence, in this study, 50% volume fraction is
NU
considered. The samples were restrained at left edge and subjected to a uniform displacement
MA

on the right edge (Figure 11). The average dimension of the triangular-shaped elements is

3mm.
E D
PT
CE
AC

Figure 11 : The concrete sample under uniaxial loading and the boundary conditions.

Figure 12 depicts the damage distribution within the specimen. It can be seen that in samples

under the tensile loading, the damaged area is perpendicular to the loading direction, while in

compression, the damaged area is extended in two inclined lines with 45° with respect to

horizontal direction which are perpendicular to each other. As the weakest part of concrete is

ITZ region, the damage is initiated in this region and then extended to the cement phase.

14
 ACCEPTED MANUSCRIPT

Tension Compression

PT
RI
Figure 12: Damage distribution in CNT-3% sample under tension and compression.

SC
Figure 13 illustrates the results of tensile and compressive loading in the form of stress-strain

curves. The tensile strength of the concrete sample is 1.86MPa, while adding CNT with the
NU
volume fraction of 3% increases the tensile strength of CNT-reinforced concrete by 31% to

2.44MPa. Furthermore, adding CNT to the concrete enhances significantly its ductility and
MA

shows a more stable behavior with higher energy absorption capacity. On the other hand, the

compressive strength of plain concrete is around 18.41MPa, while the compressive strength
D

of CNT-reinforced concrete with 3% volume fraction of CNT is about 24.81MPa. It means

E
PT

35% increase in compressive strength is observed and that ductility is enhanced.

CE
AC

a)

15
 ACCEPTED MANUSCRIPT

b)

PT
RI
Figure 13: Stress-strain curves for the concrete sample; a) tension, b) compression.

SC
5. Monotonic and cyclic analysis of CNT-reinforced concrete column
NU
In this section, a monotonic semi-static loading is applied to investigate the mechanical

behavior of a CNT-reinforced column (CNRC) and an ordinary reinforced concrete column

MA

(RC) and the corresponding results are compared. The numerical model is constructed and

simulated using Open Systems for Earthquake Engineering Simulation (OpenSees 2.4.4)
D

[28]. A “Concrete02” [29], (Table 5), model is used to implement the stress-strain curve
E
PT

obtained at the meso-scale (Figure 13) to the upper scale at the macro-scale. The behavior of

unconfined concrete should be converted to a confined behavior of concrete within the

CE

central core of the column (Figure 14). In this model, the concrete compressive strength at 28

days is 𝑓𝑐′ , the concrete strain at maximum strength is 𝜀𝑐0 , the concrete crushing strength is
AC

′
𝑓𝑐𝑢 , the concrete strain at crushing strength is 𝜀𝑢 and the tensile strength is 𝑓𝑡 .

A damage model with linear softening regime defines the material behavior of concrete at

tension and compression. The mechanical behavior of unconfined concrete is converted to a

confined one using the method of Mander et al. [30]. The compressive behavior of concrete is

affected only by confinement and the tensile behavior remains the same. In Table 5, the

mechanical properties of concrete materials are presented. Furthermore, the uniaxial behavior

16
 ACCEPTED MANUSCRIPT

of steel rebar is simulated by Steel02 material model [28] with isotropic strain hardening and

yield strength of σy=437MPa [31].

PT
RI
SC
NU
Figure 14: Stress strain behaviour of confined and unconfined concrete02 model.

Table 5 : Mechanical properties of concrete models.

MA

𝒇′𝒄 (𝑴𝑷𝒂) 𝒇′𝒄𝒖 (𝑴𝑷𝒂)

Material 𝜺𝒄𝟎 𝜺𝒖 𝒇𝒕 (𝑴𝑷𝒂) 𝑬(𝑴𝑷𝒂)
D

Concrete 18.41 0.000813 0 0.007 1.86 26151

E

CNT-1% 19.86 0.001016 0 0.010 2.00 26310

PT

CNT-2% 20.41 0.001372 0 0.020 2.01 24795

CE

CNT-3% 24.81 0.020830 0 0.040 2.44 25248

Concrete (Confined) 27.95 0.002920 22.34 0.014 1.86 26151

AC

CNT-1% (Confined) 29.58 0.003480 23.65 0.020 2.00 26310

CNT-2% (Confined) 30.13 0.004630 24.06 0.030 2.01 24795

CNT-3% (Confined) 34.75 0.006264 27.86 0.050 2.44 25248

As the stress-strain behavior of concrete depends on confinement, the fiber-element method

is adopted in this study to simulate the mechanical behavior of columns under monotonic

loading [32]. In this approach, the column is divided into several two-node elements and the

17
 ACCEPTED MANUSCRIPT

cross section of each element is divided into multiple steel and concrete fiber elements. In

this fiber elements, by obtaining the strain of each point, one can extract the corresponding

stresses [32]. By using this approach, the plastic behavior of the section can be obtained. In

addition, a zero-length element with the bond-slip01 material model is adopted to investigate

the effects of sliding between concrete and reinforcement at the bottom of the column (the

most critical section) [33]. This material model is for fully anchored steel reinforcement bars

PT
that experience bond slip along a portion of the anchorage length due to strain penetration

RI
effects, which are usually the case for column and wall longitudinal bars anchored into

SC
footings or bridge joints.

The experimental results of the cyclic displacement-controlled test of a concrete column

NU
(U6), which was studied in [34], is used to verify the simulation method in this study. In the

study of Saatcioglu and Ozcebe, the concrete compressive strength and the steel yield
MA

strength are 37.3MPa and 437MPa, respectively. The spacing between stirrups are 65mm.

The height of the column is 1000mm and its cross section is 350×350mm with eight
E D

longitudinal bars of diameter of 25mm. The column consisted of five nonlinear displacement-
PT

control beam-column elements and one zero-length element at bottom of the column to take

into account the probable sliding between concrete and rebar. A zero-length section element
CE

is a fiber discretization of the cross-section of a structural member. In a section analysis, the

AC

concrete and steel fiber strains are calculated for a given curvature using the plain-section

assumption. The fiber forces, obtained using the stress-strain relationship of fibers, are

integrated across the section to obtain the corresponding moment. The overall configuration

of the column and its boundary conditions are depicted Figure 15. The cross section of the

column is divided into a confined core and unconfined cover area, and different material

models can be assigned into these regions.

18
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
Figure 15: Schematic fiber element model of the column.
MA

A constant vertical force of magnitude 600kN is applied on the top of the column and a cyclic

displacement with magnitudes 11, 23, 46, 69 and 91mm are imposed to the top of the column.
D

Figure 16 illustrates a comparison between the results of numerical and experimental

E

simulation, which indicates that the results are in a very good agreement.
PT
CE
AC

Figure 16 : Comparison of numerical and experimental [34] results.

19
 ACCEPTED MANUSCRIPT

After verification, four concrete columns similar to column of Saatcioglu and Ozcebe [34],

with different materials including, ordinary concrete (RC) and three CNT-reinforced concrete

column with different volume fractions (CNRC-1%, CNRC-2%, CNRC-3%), are studied.

Their corresponding moment-curvatures are shown in Figure 9. The analysis indicates that

the nominal compressive strength of RC column is 4300kN. Considering this value, column

sections are subjected to three different axial loads (representing 8%, 13% and 20% the

PT
nominal compressive capacity) and the corresponding responses are depicted in Figure 10.

RI
From the moment-curvature diagrams illustrated in Figure 17, it can be seen that columns

SC
with higher volume fraction of CNT can undergo larger curvature and lateral displacement.
NU
MA

a)
E D
PT
CE
AC

b)

20
 ACCEPTED MANUSCRIPT

c)

PT
RI
Figure 17 : Moment-curvature diagrams for different axial loadings.

SC
The curvature ductility is defined as the ratio of ultimate curvature to yield curvature of the
NU
column. This value is calculated for all columns and the results are presented in Figure 18 for

different CNT volume fractions. As can be seen, the curvature ductility of the columns
MA

decreases when the axial force increases. When the CNT volume fraction increases, the

difference between curvature ductility of the column increases. The minimum curvature
D

ductility belongs to RC column with 860kN axial force equal to 18.53 while CNRC-3% with
E

axial force of 340kN has the maximum curvature ductility around 85.86.
PT
CE
AC

Figure 18 : Curvature ductlity for different axial loadings.

21
 ACCEPTED MANUSCRIPT

The amount of dissipated energy per length of concrete column can be estimated by

calculating the enclosed area under the monotonic moment-curvature curve. Figure 19

illustrates the ultimate moment capacity and the energy dissipation of each column. The

dissipated energy decreases with the increase of column axial force. In addition, higher CNT

volume fractions improve the energy absorption capacity. The highest energy absorption

capacity belongs to CNRC-3% with 340kN axial force equal to 140.8kN per length while RC

PT
column with 860kN axial force possess the minimum energy absorption capacity per length

RI
about 24.97kN.

SC
As the amount of axial force decreases, the moment capacity of the column increases,

although the difference is negligible (Figure 19.a). RC column has an average moment
NU
capacity of 255kN.m in comparison to CNRC-3%, which has 320kN.m with a rise of more

than 25%. It can be concluded that CNTs can decrease the extension of damage in concrete
MA

and enhance its strength and stability.

E D
PT
CE

a)
AC

22
 ACCEPTED MANUSCRIPT

b)

PT
RI
SC
Figure 19 : a) Dissipated energy; b) ultimate momonnt of the column sections with different

axial loading.
NU
Two models, including ordinary concrete and CNRC-3%, are considered for comparison

between the cyclic behaviors of columns. The columns are subjected to 2500𝐾𝑁 axial
MA

compressive force (58% of the nominal axial compressive capacity). After imposing this

constant compressive force, a lateral quasi-static cyclic displacement is applied on the top of
D

the column. The pseudo-static lateral load is applied under the displacement control
E
PT

condition.

The hysteresis behavior of the models are depicted in Figure 20, which clearly shows the
CE

pinching behavior in both models. The area enclosed among the hysteresis force-
AC

displacement curve of the CNRC-3% column is more than the RC one, which is an indication

of larger amount of energy absorption. CNRC-3% column can sustain 307𝐾𝑁 lateral force,

while in the RC column this value decreases to 213𝐾𝑁 , which means that the CNRC-3%

column can bear 44% more lateral load than the RC column. This result indicates that CNTs

can limit the crack propagation rate and consequently increase the stability and strength of

concrete under cyclic loadings.

23
 ACCEPTED MANUSCRIPT

PT
RI
SC
Figure 20 : Force-displacement behavior of the concrete columns under quasi-static cyclic
loading.
NU
The energy dissipation capacity for each cycle is estimated based on the area enclosed by the

force-displacement hysteresis curve of each cycle. The cumulative energy dissipation in each
MA

step is also obtained by summing up the energy dissipation in consecutive loops [4]. The

cumulative energy dissipation with respect to the column top displacement is illustrated in
D

Figure 21. Cleary, the cumulative energy dissipation consistently increases and in each cycle
E
PT

for CNRC-3% column is more than the RC one. In addition, by increasing the displacement

level, the difference between the cumulative energy dissipation curves increases. This can be
CE

attributed to the fact that CNTs bridge the cracks at the nanoscale and delay the crack
AC

propagation, which results in more dissipation energy. The cumulative energy dissipation at

the end of loading for RC and CNRC-3% columns are equal to 242.8𝑘𝑁.𝑚 and 304.9𝑘𝑁.𝑚 ,

respectively, showing an increase of 26%.

24
 ACCEPTED MANUSCRIPT

PT
RI
SC
Figure 21: The cumulative energy dissipation vs. the column maximum top displacement in
each cycle.
NU
6. Conclusion
MA

In this research, a hierarchical multi-scaling approach, including nano-, micro-, meso- and

macro-scales, is introduced to investigate the mechanical properties and ductility of carbon

nanotube (CNT)-reinforced concrete columns under service loading. The results of the
E D

simulations at the nano-scale indicate that the mechanical properties of CNTs in tension are
PT

superior to compression. At the micro- and meso-scales, the results indicate that increasing

the volume fraction of CNTs can increase significantly the fracture energy, tensile and
CE

compressive strength of CNT-reinforced cement and concrete. Finally, concrete columns

AC

made of different CNT volume fractions are analyzed under monotonic loading and the

results reveal that CNRC sections with higher volume fraction of CNTs can sustain larger

amount of curvature and lateral displacement. In addition, by increasing the axial force of the

columns, the amount of dissipated energy and curvature ductility decreases. Furthermore,

adding CNTs to the concrete columns can increase significantly their energy absorption

capacity.

Conflict of interest

25
 ACCEPTED MANUSCRIPT

The authors declare that they have no conflict of interest

References

[1] Ramirez CM, Liel AB, Mitrani-Reiser J, Haselton CB, Spear AD, Steiner J et al. Expected earthquake
damage and repair costs in reinforced concrete frame buildings. Earthquake Engineering &
Structural Dynamics. 2012;41:1455-75.
[2] Ates S, Kahya V, Yurdakul M, Adanur S. Damages on reinforced concrete buildings due to
consecutive earthquakes in Van. Soil Dynamics and Earthquake Engineering. 2013;53:109-18.
[3] Baran E, Mertol HC, Gunes B. Damage in Reinforced-Concrete Buildings during the 2011 Van,

PT
Turkey, Earthquakes. Journal of Performance of Constructed Facilities. 2014;28:466-79.
[4] Ganesan N, Indira P, Sabeena M. Behaviour of hybrid fibre reinforced concrete beam–column
joints under reverse cyclic loads. Materials & Design. 2014;54:686-93.

RI
[5] Zhang Y-y, Harries KA, Yuan W-c. Experimental and numerical investigation of the seismic
performance of hollow rectangular bridge piers constructed with and without steel fiber reinforced
concrete. Engineering Structures. 2013;48:255-65.

SC
[6] Fischer G, Li VC. Effect of fiber reinforcement on the response of structural members. Engineering
Fracture Mechanics. 2007;74:258-72.
[7] Lee H-H. Shear strength and behavior of steel fiber reinforced concrete columns under seismic
NU
loading. Engineering structures. 2007;29:1253-62.
[8] Iijima S. Helical microtubules of graphitic carbon. Nature. 1991;354:56-8.
[9] Šmilauer V, Hlaváček P, Padevět P. Micromechanical Analysis of Cement Paste with Carbon
MA

Nanotubes. Acta Polytechnica. 2012;52.

[10] Plimpton S. Fast Parallel Algorithms for Short-Range Molecular Dynamics. Journal of
Computational Physics. 1995;117:1-19.
[11] Tersoff J. Modeling solid-state chemistry: Interatomic potentials for multicomponent systems.
Phys Rev B. 1989;39:5566-8.
D

[12] Eftekhari M, Hatefi Ardakani S, Mohammadi S. An XFEM multiscale approach for fracture
analysis of carbon nanotube reinforced concrete. Theoretical and Applied Fracture Mechanics.
E

2014;72:64-75.
PT

[13] Eftekhari M, Mohammadi S. Molecular dynamics simulation of the nonlinear behavior of the
CNT-reinforced calcium silicate hydrate (C–S–H) composite. Composites Part A: Applied Science and
Manufacturing. 2016;82:78-87.
CE

[14] Eftekhari M, Mohammadi S, Khoei AR. Effect of defects on the local shell buckling and post-
buckling behavior of single and multi-walled carbon nanotubes. Computational Materials Science.
2013;79:736-44.
[15] Hessam Y, Kianoosh H, Mehdi E. Mechanical properties of single-walled carbon nanotubes: a
AC

comprehensive molecular dynamics study. Materials Research Express. 2017;4:055015.

[16] Poelma RH, Sadeghian H, Koh S, Zhang GQ. Effects of single vacancy defect position on the
stability of carbon nanotubes. Microelectron Reliab. 2012;52:1279-84.
[17] Matsumoto M, Nishimura T. Mersenne twister: a 623-dimensionally equidistributed uniform
pseudo-random number generator. ACM Transactions on Modeling and Computer Simulation
(TOMACS). 1998;8:3-30.
[18] Wernik J, Meguid S. Multiscale modeling of the nonlinear response of nano-reinforced
polymers. Acta Mechanica. 2011;217:1-16.
[19] Kulkarni M, Carnahan D, Kulkarni K, Qian D, Abot JL. Elastic response of a carbon nanotube fiber
reinforced polymeric composite: a numerical and experimental study. Composites Part B:
Engineering. 2010;41:414-21.
[20] Patzák B, Rypl D. Object-oriented, parallel finite element framework with dynamic load
balancing. Advances in Engineering Software. 2012;47:35-50.

26
 ACCEPTED MANUSCRIPT

[21] Mazars J. A description of micro- and macroscale damage of concrete structures. Engineering
Fracture Mechanics. 1986;25:729-37.
[22] Šmilauer V, Hlaváček P, Padevět P. Micromechanical Analysis of Cement Paste with Carbon
Nanotubes. Acta Polytechnica. 2012;52:22-8.
[23] Eftekhari M, Mohammadi S. Multiscale dynamic fracture behavior of the carbon nanotube
reinforced concrete under impact loading. International Journal of Impact Engineering. 2016;87:55-
64.
[24] Du X, Jin L, Ma G. Numerical simulation of dynamic tensile-failure of concrete at meso-scale.
International Journal of Impact Engineering. 2014;66:5-17.
[25] Kim S-M, Abu Al-Rub RK. Meso-scale computational modeling of the plastic-damage response of
cementitious composites. Cement and Concrete Research. 2011;41:339-58.

PT
[26] Nguyen VP, Stroeven M, Sluys LJ. Multiscale failure modeling of concrete: Micromechanical
modeling, discontinuous homogenization and parallel computations. Computer Methods in Applied
Mechanics and Engineering. 2012;201–204:139-56.

RI
[27] Qian Z. Multiscale modeling of fracture processes in cementitious materials: PhD thesis, Delft
University of Technology; 2012.
[28] Mazzoni S, McKenna F, Scott MH, Fenves GL. OpenSees command language manual. Pacific

SC
Earthquake Engineering Research (PEER) Center. 2006.
[29] B. D. Scott RP, Priestley MJN. Stress-Strain Behavior of Concrete Confined by Overlapping Hoops
at Low and High Strain Rates. Journal Proceedings. 1982;79.
NU
[30] Mander J, Priestley M, Park R. Observed stress-strain behavior of confined concrete. Journal of
Structural Engineering. 1988;114:1827-49.
[31] Wang Z, Dueñas-Osorio L, Padgett JE. Influence of scour effects on the seismic response of
reinforced concrete bridges. Engineering Structures. 2014;76:202-14.
MA

[32] Nizette M, Goriely A. Towards a classification of Euler–Kirchhoff filaments. Journal of

Mathematical Physics. 1999;40:2830-66.
[33] Zhao J, Sritharan S. Modeling of Strain Penetration Effects in Fiber-Based Analysis of Reinforced
Concrete StructuresConcrete Structures. ACI structural journal. 2007;104.
D

[34] Saatcioglu M, Ozcebe G. Response of reinforced concrete columns to simulated seismic loading.
ACI Structural Journal. 1989;86.
E
PT
CE
AC

27