Engineering Structures 156 (2018) 567–586

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Effect of cross-section size on the flexural failure behavior of RC cantilever T

beams under low cyclic and monotonic lateral loadings
⁎ ⁎
Liu Jin, Wenxuan Yu, Xiao Su, Shuai Zhang, Xiuli Du , Junyan Han , Dong Li
Key Laboratory of Urban Security and Disaster Engineering of Ministry of Education, Beijing University of Technology, Beijing 100124, China

A R T I C L E I N F O A B S T R A C T

Keywords: The objectives of this study are, on one hand, to present an experimental campaign on the flexural failure
RC beam behavior of RC beams under both low cyclic and monotonic lateral loadings, and on the other hand, to discover
Monotonic load the effect of cross-section size on the global mechanical properties of RC beams. Five groups of geometrically
Cyclic load similar RC cantilever beams, with cross-section sizes range from 200 mm to 1000 mm and a shear-span ratio of 4
Flexural failure
(length to shear span) were experimentally tested. Moreover, a 3D meso-scale numerical method, which could
Size effect
Meso-scale simulation
consider both the heterogeneity of concrete material and the nonlinear interaction between reinforcing bars and
surrounding concretes, for the simulation of failure of RC members was developed, and some numerical tests
were established based on the experimental campaign. The tested results demonstrate the presence of size effect
in flexural failure of RC beams under both low cyclic and monotonic lateral loadings. Specifically, the effects of
structural size on the seismic performances of the RC beams, involving the failure pattern, the ductility, the
stiffness degradation and the load carrying capacity, were extensively investigated based on the combined ex-
perimental and numerical tests. Under cyclic loading, due to the low cyclic fatigue damage behavior and for the
fact that failure always takes place in concrete which is a quasi-brittle engineering material, all the tested beams
could exhibit a more quasi-brittle failure pattern, and consequently the flexural failure of the beams presents a
stronger size effect compared with that under monotonic loading.

1. Introduction cracks. In turn, a statistical size effect is caused by spatial variability/

The size effect is understood as any dependence of nominal strength
and brittleness on structural size. Generally, the nominal strength de-
creases with increasing structural size, while the brittleness increases
with increasing structural size. Theoretically, based on the fracture
mechanics, several researchers have demonstrated that, concrete be-
comes ductile on a small scale and it becomes brittle on a large scale
[1], and numerous test observations [3–5] have proved that, the size
effect behavior of concrete is mainly due to its inherent quasi-brittle
feature which is originally caused by the inhomogeneity of the material.
From a mesoscopic point of view, concrete is full of micro-cracks, and
these micro-cracks become active and begin to propagate upon loading.
A strong stress concentration would develop around the crack tip and
the micro-cracks propagate to major macro-cracks that lead to the
eventual failure of the structure. Two size effects are of a major im-
portance: deterministic size effect and statistical size effect [2]. A de-
terministic (or energetic) size effect is caused by the formation of a
region of intense strain localization with a certain volume (also called
fracture process zone, FPZ) which always precedes discrete macro-
randomness of the local material strength. A combination of the de-
terministic size effect law with the statistical size effect law led to a
general energetic-statistical theory. The deterministic size effect occurs
for not too large structures and the statistical size effect is obtained as
the asymptotic limit for very large structures. The deterministic size
effect can also occur in RC members wherein the failure always takes
place in concrete which is a quasi-brittle engineering material. For re-
inforced concrete (RC) structures, the interaction between concrete and
reinforcement has a significant influence on the development of cracks
in concrete. Therefore, the size effect in RC members is rather a
structural mechanics problem, which includes not only the material size
effect in concrete, but also the size effect contributed by the nonlinear
interaction between concrete and rebar, than a material science pro-
blem [6]. Generally, for the fact that failure always takes place in
concrete which is a quasi-brittle engineering material, many experi-
mental results on RC columns [6–8], RC beams [9–12] and RC beam-to-
column connections [13] have demonstrated the existence of size effect
in RC elements. The present research works mainly concern geome-
trically similar RC beams that are designed to fail in flexure.

⁎
Corresponding authors.
E-mail addresses: duxiuli2015@163.com (X. Du), junyanhan@bjut.edu.cn (J. Han).

https://doi.org/10.1016/j.engstruct.2017.11.069
Received 5 May 2017; Received in revised form 29 November 2017; Accepted 30 November 2017
0141-0296/ © 2017 Elsevier Ltd. All rights reserved.
L. Jin et al.
1.1. Available test efforts on flexural failure of RC beams
The shear failure, the flexural failure and the flexural-shear failure
are three main failure patterns for RC beams under loads. Generally, the
shear behavior of concrete components exhibits a brittle failure mode,
and then a size effect may be expected for the fact that the size effect
always occurs in RC elements if failure takes place in concrete that is a
quasi-brittle engineering material [14]. In our previous work of Jin
et al. [12], five groups of geometrically similar RC beams, with cross-
section sizes range from 200 mm to 1000 mm and a shear-span ratio of
2 (length to depth), were tested under cyclic lateral loading. The test
results have demonstrated the existence of size effect in shear strength
of RC beams.
For the flexural failure in RC beams, many test efforts have been
conducted to explore the possibility of existence of size effect. For in-
stance, in the work of Weiss et al. [15], sixteen groups of RC beams
having different concrete strengths and percentages of longitudinal
steel bars were tested. They found that, the nominal moment-carrying
capacity of all beams was relatively constant, while the overall ductility
was dependent on the constant-moment zone length with larger spe-
cimens demonstrating a more brittle response. The test results pre-
sented by Alca et al. [16] also indicated that there was no obvious size
effect in the flexural compressive strength of RC beams. However, some
other test efforts [17–23] have shown the presence of size effect in
flexural strength. For instance, in the experiment of Rao et al. [19],
three groups of RC beams with cross-sectional depths of 100 mm,
200 mm and 400 mm were tested, which were designed with varying
percentages of flexural reinforcement, i.e. 0.15%, 0.30%, 0.60% and
1.0%. The RC beams were tested under four-point loading to study the
flexural behavior under uniform bending moment. Their test results
indicated that the flexural strength of RC beams decreased as the depth
increases. Bosco et al. [20] also performed tests on RC beams with
various beam depths and longitudinal steel percentages. The test results
indicated that, the brittleness of the structural member increases by
increasing the structural size and decreasing the longitudinal steel
percentages. Yi et al. [21] tested a series of RC beams with same per-
centages of flexural reinforcement subjected to four-point loading. The
test observations illustrated that, the flexural compressive strength at
failure exhibited an obvious size effect, and the corresponding peak
strain and the ultimate strain decrease as the structural size increases.
Furthermore, Kim et al. [22] and Yi et al. [23] also demonstrated the
presence of size effect in flexural compressive strength of RC beams.
The contradictories in the test results for the investigation of size
effect in flexural failure of RC beams indicate that, the presence of size
effect may strongly depend on how the failure takes place. In general,
for over-reinforced or less-reinforced beams, the failure is mainly due to
the fracture of concrete, which exhibits brittleness and leads to an ob-
vious size effect. While for appropriate reinforced beams, the failure is
usually less brittle or ductile, and the nominal strength would show no
obvious size effect.
1.2. Effect of seismic loading on the failure behavior of RC beams
Under seismic loading, the development of cracks in RC beams
changes, the bonding between concrete and rebar could be deteriorated
under low cyclic fatigue loading, and the failure always takes place in
concrete which is a quasi-brittle engineering material, consequently the
beam becomes more brittle as compared to that under monotonic
loading. Fatigue failure happens as a concrete structure fails at less than
the design load after exposure to a large number of stress cycles,
causing the degradation of stiffness and strength. It is a process of
progressive and permanent internal damage in a material subjected to
repeated loading. Lots of studies have been conducted to study the
failure behavior of RC beams under repeated or cyclic loadings, e.g. in
the work of Moretti and Tassios [24], Mahal et al. [25] and Dadi and
Agarwal [26]. However, less work has been focused on size effect.
568
Engineering Structures 156 (2018) 567–586
Recently, Barbhuiya and Choudhury’s [13] test efforts on RC beam-
column connections have indicated that, under cyclic loading, the three
categories of beam-column connections (i.e. beam weak in flexure,
beam weak in shear and column weak in shear) exhibit an obvious size
effect. In our previous efforts [12,27], the test observations have also
indicated that, under low-cyclic fatigue loadings, even for balanced-
reinforced RC specimens, because of that the failure always takes place
in concrete which is a quasi-brittle engineering material, the failure of
the RC beams or columns is still brittle, and the nominal shear strength
exhibits a pronounced size effect.
1.3. Scope of the present research work
The objectives of the present study are, on one hand, to present an
experimental campaign on the flexural failure behavior of RC beams
under both low cyclic and monotonic lateral loadings, and on the other
hand, to discover the effect of cross-section size on the global seismic
performances of RC beams. With respect to the previous work [12] in
which the size effect of shear failure of RC beams was experimentally
explored, the innovative points of the present study are to examine the
size effect of flexural failure of RC beams by experimental and nu-
merical approaches.
Five groups of geometrically similar RC cantilever beams with cross-
section heights ranging from 200 mm to 1000 mm and a shear-span
ratio of 4 (depth to shear span) were experimentally tested. Moreover, a
3D meso-scale numerical method, which could consider both the het-
erogeneity of concrete and the nonlinear interaction between reinfor-
cing bars and surrounding concretes, for the simulation of failure of RC
members was developed, and some numerical tests were conducted
based on the experimental campaign. Finally, the 3D meso-scale si-
mulation method was extended to examine the size effect in larger-sized
RC beams with a maximum cross-sectional height of 2000 mm.
2. Test results
2.1. Specimen description and materials properties
Five groups of geometrically similar RC cantilever beams with dif-
ferent structural sizes, subjected to monotonic and cyclic loading were
tested. The dimensions, shape, loading point locations, specimen label,
and reinforcement details of specimens used in the experiment are
shown in Fig. 1 and Table 1. The cross-sectional depth of the beams
tested ranged from 200 mm to 1000 mm. The tested RC beams pre-
sented in Fig. 1 were designed to fail as flexural failure, and a shear-
span ratio (length to depth) of 4 was utilized. There were three speci-
mens designed for every group of tested specimens having the same
structural sizes, in which one of them was subjected to monotonic
loading, and the other two were designed to bear cyclic loading. The
specimens under monotonic loading were named as MB, and the other
two specimens under cyclic loading were named as CB. Herein, the
letter “M” means the “Monotonic loading”, “C” means the “Cyclic
loading”, and “B” denotes “Beam”. A total of 15 RC beams were thus
tested.
The mixture proportions of the concrete utilized in the present tests
are given in Table 2. Medium sands are considered as the fine aggregate
(i.e. the average diameter is less than 5 mm), and crushed pebble stones
as the coarse aggregate (i.e. the average diameter ranges from 5 mm to
30 mm). The compressive and tensile strengths of concrete, which are
measured by uniaxial compressive and splitting tensile tests based on
the currently utilized standard codes of China, are also presented in
Table 2. Moreover, the detailed information about the parameters of the
reinforcing steel bars are listed in Table 3, including the diameters, the
yield strength and the ultimate strength, etc. In the tests, the strength of
concrete, reinforcement ratio, stirrup ratio, and shear span ratio were
kept the same for all RC beams.
It is to be noted that, the aggregate sizes were not scaled to the
L. Jin et al.
Fig. 1. Details of the tested RC cantilever beams.
characteristic structure size. This is because scaling the aggregate size
would complicate evaluation of the test results and prevent isolating the
structural size effect from other influences [6].
2.2. Experimental setup and loading arrangement
Fig. 2(a) presents a typical failure photograph of the beam CB-3. The
specimen was mounted on a solid component vertically, and the top of
the cantilever beam was held by a servo hydraulic dynamic actuator.
Monotonic and cyclic loading were employed to the specimen in the
horizontal direction with the aid of the servo hydraulic dynamic
569
Engineering Structures 156 (2018) 567–586
actuator. The schematic diagram is presented in Fig. 2(b). In the tests,
the collected data involved the horizontal displacement at the loading
point measured by Linear Variable Differential Transformer (i.e. LVDT),
the reaction force at the loading point, concrete compressive strain, and
the tensile strain of longitudinal reinforcement.
The force-displacement hybrid control method was utilized. The
detailed loading scheme can be illustrated as follows:
(1) Monotonic loading: firstly, the force control was used until the RC
beams yielded. The yielding displacement defined as the displace-
ment when the RC beams yielded was then adopted as the level
L. Jin et al. Engineering Structures 156 (2018) 567–586

Table 1
Physical parameters of the tested RC beams.

Specimen Name MB-1/CB-1 MB-2/CB-2 MB-3/CB-3 MB-4/CB-4 MB-5/CB-5

Total number 3 3 3 3 3
Cross-section size (b × h) (mm2) 80 × 200 160 × 400 240 × 600 320 × 800 400 × 1000
Longitudinal reinforcement ratio ρ 1.26% 1.23% 1.31% 1.22% 1.23%
Hoop reinforcement ratio ρsv 0.14% 0.14% 0.14% 0.14% 0.14%
Shear-span ratio λ 4.0 4.0 4.0 4.0 4.0
Effective height of Beam h0 (mm) 170 345 540 735 930
Beam effective length a = λ * h0 (mm) 6800 1380 2160 2940 3720
Total length of beam l (mm) 830 1630 2510 3390 4220
Concrete tensile/compressive strengths [MPa] 2.3/42.8 2.3/42.8 2.3/42.8 2.3/42.8 2.3/42.8

Table 2
Concrete mixture proportions.

Aggregate Aggregate Aggregate Cement Water Water-reducing Silica Sum w/c ratio Compressive Splitting tensile
0–5 mm (kg/m3) 5–15 mm (kg/ 15–30 mm (kg/ (kg/m3) (kg/m3) admixtures (kg/ fume (kg/ (kg/ strength (MPa) strength (MPa)
m3 ) m3) m3) m3) m3)

1090 185 436 460 184 4.0 75 2434 0.4 42.8 2.3

difference for the loading. It was utilized until the bearing capacity [21,11,17,18]. Fig. 4(b) shows the final failure pattern of the five
of the beams dropped to 85% peak bearing capacity or significant groups of RC cantilever beams under cyclic loading. On the whole, the
damage occurred in the beams. failure patterns of the ten beams are closely similar. The failure of all
(2) Cyclic fatigue loading: the typical force and displacement history beams occurred at the bottom part where the concrete crushing gen-
applied to specimen is shown in Fig. 3. The applied force at the erated. At this time, the bond lose efficacy at the interface between the
horizontal direction was loaded and only one cycle was applied at longitudinal steel bars and the surrounding concretes. Moreover, it can
each level of loading. After the RC beams yielded, the multiple of be seen from Fig. 4(b) that much more cracks generated as the struc-
yielding displacement was take as the loading control, and three tural size (cross-sectional height) of the beams increases.
cycles were applied at each level. Similarly, the experiment was In addition, the detailed information regarding crack patterns (in-
stopped at the stage when the load came down in the range of 85% cluding crack inclination and crack spacing) are described in
of the ultimate load carrying capacity or significant damage oc- Fig. 5(a) and (b) respectively for the beams under monotonic and cyclic
curred. It is to be noted that, it cost about 10 mins for each cyclic to loading. The concrete spalling areas (i.e. the grey region) can also be
ensure the pseudo-static loading. observed in Fig. 5.
Compared Fig. 4(b) with Fig. 4(a), it can be noted that the failure
It is to be noted that, fatigue loading is usually divided into two patterns of the beams under monotonic and cyclic loading are sig-
categories, including low-cyclic (e.g. structures subjected to earth- nificantly distinct. Under low cyclic fatigue loading, both the tensile/
quakes with 1–102 loading cycles) and high cyclic loading [28]. compressive strengths of the concrete and the yield strength of re-
inforcing bars degrade. Moreover, the bearing capacity (or the shear
capacity) at the interface of steel bars and surrounding concretes would
2.3. Flexural failure patterns
also decrease due to the low cyclic fatigue behavior. That is to say, the
failure definitely takes place in concrete that is a quasi-brittle en-
Under the monotonic loading, the final failure patterns of the five
gineering material. While for monotonically loaded RC beams, the re-
RC cantilever beams having different structural sizes are presented in
inforcement could make the failure of the structure more ductile. This
Fig. 4(a). It can be found that, the failure patterns of the five groups of
finally leads to the fact that, the failure of the beams subjected to cyclic
beams are closely similar. For the smaller-sized specimens (e.g. MB-1),
loading is less ductile than the one subjected to monotonic loading.
the widths of the main tensile cracks were small, while for the larger-
sized specimens (e.g. MB-5), the corresponding crack widths were much
larger, as also observed in the work of Bažant and Kim [9], the most 2.4. Moment-displacement hysteretic behavior and envelopes
important consequence of wider cracks is diminished residual tensile
stress. In general, the larger the size of the RC beam is, the larger crack The recorded cyclic hysteretic load-displacement curves (i.e. the
width and more cracks the specimens possess when the beam failed solid black lines) and the corresponding skeleton curves (i.e. the blue1
dotted lines) of the ten RC beams under low cyclic horizontal loading
Table 3 are plotted in Fig. 6. Moreover, the moment-displacement curves (i.e.
Mechanical parameters of the reinforcing bars in the test. the solid red lines) for the five beams under monotonic loading are also
Diameter of Grade of Elastic Yield Ultimate
presented in Fig. 6 for comparison.
rebar [mm] steel bar modulus strength strength [MPa] In general, for the ten RC beams subjected to cyclic loading, the shapes
[GPa] [MPa] of the recorded hysteretic curves are all plump, as can be observed in
Fig. 6. As the applied loading continued, the RC beams under cyclic
4 HPB300 210 297.89 425.74
6 HRB335 196 405.97 529.55
loading gradually entered the elastic-plastic stage, and unrecoverable
10 HRB335 198 392.10 564.63 plastic deformation occurred. At this moment, the residual deformation of
16 HRB335 202 399.13 530.93
20 HRB335 193 402.00 542.51
1
25 HRB335 195 374.84 536.29 For interpretation of color in Fig. 6, the reader is referred to the web version of this
article.

570
L. Jin et al.
Servo hydraulic
dynamic actuator
RC beam
Loading apparatus
(a)
Fig. 2. Loading apparatus and schematic diagram.
Displacement
Three times
Yield
Controlled by force Controlled by displacement
Fig. 3. Scheme of cyclic loading.
the beams after loading became larger and larger, the cracks developed
quickly, making that the stiffness degraded quickly. In the last stage, larger
plastic deformation generated, and concrete crushing and buckling of re-
inforcing bars happened. The corresponding skeleton curves obtained
exhibit a reversed ‘S-shape’, as shown in Fig. 6. Under the low cyclic fa-
tigue loadings, the cracks within the RC beams opened and closed con-
tinually, resulting into the decrease in the flexural capacity of the RC
beams. The larger the flexural deformation is, the more seriously the
flexural stiffness degrades. In addition, one can note from Fig. 6 that, the
softening curve becomes steeper with increasing the beam size, indicating
that the brittleness of the specimen increases with the increase of the
structural dimension. This can be clearly observed from the relationship
between the softening rate and structural size presented in Fig. 7. Herein,
the slope k is utilized to describe the softening rate or the brittleness, as
shown in Fig. 7. One can note from Fig. 7 that, the slope k increases with
increasing the cross-sectional height h.
By comparing with the skeleton curves of the beams under mono-
tonic and cyclic loading, one can clearly find that the bearing capacity
of the large-sized beams under monotonic loading is obviously larger
than that under cyclic loading. While for the small-sized beams (CB-1-i
and CB-1-ii), the bearing capacities are almost the same under mono-
tonic and cyclic loading. Moreover, from the skeleton curves, one can
note that under monotonic loading the RC beams exhibit a long yield
platform. That is to say, the deformation capacity of the RC beams
under cyclic loading is significantly weaker than that under monotonic
loading. In reality, compared with that under monotonic loading, not
Engineering Structures 156 (2018) 567–586
Servo hydraulic dynamic actuator
LVDT
LVDT
RC beam
Loading direction
Shear wall
Base
(b)
only the tensile/compressive strengths of concrete meso components
and yield strength of the steel bars degrade, but also the bonding ca-
pacity at the interface of concrete and steel bars weakens due to the
damage accumulation within the RC beams under cyclic loading.
2.5. Ductility capacity
Ductility is often used to describe the deformation capacity of RC
members in the seismic design. Similar with that in the work of
Rodrigues et al. [29], the displacement ductility coefficient μ defined as
μ = Δu/Δy, was adopted to assess the seismic performance of the RC
beams with various cross-section sizes. Here, Δu means the displace-
ment corresponding to ultimate moment defined as 85% bearing ca-
pacity in the decline period of moment-displacement curves [30], and
the yielding displacement Δy characterizes the displacement corre-
sponding to yield moment. The detailed information is described in
Fig. 8.
The two points corresponding to yield moment and ultimate mo-
ment are presented in Fig. 6, and they are marked by pink and green
circles, respectively. The obtained ductility coefficients of the RC can-
tilever beams under both monotonic and cyclic loading are plotted
against the cross-sectional height h in Fig. 9. One can see that the
ductility coefficient increases obviously with increasing the cross-sec-
tional height of the specimens under monotonic loading, while in-
creases slightly for the beams under cyclic loading. In addition, it can be
clearly seen that the ductility coefficients for the beams under cyclic
loading are much smaller than the one under monotonic loading. This
definitely is caused by the fact that under cyclic fatigue loading, the
failure of the beams becomes less ductile. In reality, under low cyclic
fatigue loading, the bonding between the concrete and the reinforce-
ment would be deteriorated, and the crushing of concretes at the end of
the specimen becomes the main reason for the failure of the specimens.
Under monotonic loading, for the present balanced-reinforced struc-
ture, after the beam yielded, many cracks generated at the flexural-
tensile side of the beam, and the longitudinal reinforcements would
consume most energy. It is to be noted that, in the present tests, there
were almost no concrete crushing at failure for the five beams under
monotonic loading, as can be observed from Fig. 4(a). Therefore, the
failure of the beams would be more ductile (or less brittle) as compared
to that of cyclically loaded RC beams.
571
L. Jin et al. Engineering Structures 156 (2018) 567–586

ȡ=1.26%,1.23%,1.31%,1.22%,1.23%
Concrete strength: fcu = 42.8 MPa
Shear-span ratio: Ȝ = 4

MB-1 MB-2 MB-3 MB-4 MB-5

400×1000×4220

320×800×3390

240×600×2510
1000 mm

160×400×1630

80×200×830 800 mm
660 mm

CB-1-i CB-2-i CB-3-i CB-4-i CB-5-i

Fig. 4. Typical failure patterns for the normal-strength RC beams, subjected to (a) monotonic loading, and (b) cyclic loading.

2.6. Stiffness degradation negative direction, +fi the displacement in the positive direction, and
-fi the displacement in the negative direction.
The stiffness degradation of the specimen plays an important role in Fig. 10 presents the relation for normalized stiffness K/K0 with the
the energy dissipation capacity [12]. The secant stiffness K declined number of cycle. Herein, K is the secant stiffness at different cycles, and
with the increase of displacement of the RC beam under the cyclic K0 is the initial tangent stiffness. It can be noted from Fig. 10 that at the
loading can be calculated by early stage of the test, some micro-cracks should generate within the
concrete beam. However, the damage was not heavy since there was no
| + Pi | + |−Pi |
K= macro-cracks generated within the concrete beam, resulting into a
| + fi | + |−fi | (1) slight degradation of global stiffness at the early stage. As the cracks
developed quickly within the RC beams, some macro-cracks occurred
in which, +Pi is the load in the positive direction, -Pi the load in the

572
L. Jin et al. Engineering Structures 156 (2018) 567–586

199 257 171 200 201

Splitting crack

156 121 120 125

160 137 138 205

29 40 41

91
57 102
114

274

360

320
Spalling area

286
34
MB-1 MB-2 MB-3 MB-4 MB-5
Unit: (mm)

358 147 326 168

294 274 273
228 227 211
189

354
219
78

86

287
173
278

177 193 194 202 320

82

139 126
95

126 158 120 259

101
176

261 295
495

432
80

86

127 177
152

347
242 273
65

46
177

Splitting crack

CB-1-i CB-2-i CB-3-i CB-4-i CB-5-i

Spalling area
59
42

41

45

41
46
39
32

21

Unit: (mm)
Fig. 5. Sketching of the crack distribution and the spalling area of different sizes obtained from the test for the RC beams, subjected to (a) monotonic loading, and (b) cyclic loading.

due to the flexural-tensile action, and therefore the stiffness of the
beams degraded quickly. One can see from Fig. 10 that the normalized
stiffness degradations for the ten beams having different structural sizes
are similar, indicating that the specimen sizes have a negligible influ-
ence on the stiffness degradation.
2.7. Energy dissipation
The energy dissipation capacity of the five groups of RC beams with
different structural sizes was examined. The load versus deflection re-
lationship can be described in Fig. 11. The area of the hysteretic loop is
a very important factor for evaluating the elastoplastic energy dis-
sipation capacity. The larger the hysteretic loop is, the better the energy
dissipation capacity will be. The equivalent damping coefficient ζe can
be calculated by Wang et al. [31].
1 (SABC + SCDA)
ζe =
2π (SOBE + SODF )
Eq. (2) can be utilized to calculate the equivalent damping coeffi-
cient (ζe). Herein, SABC and SCDA are the areas under the curves of ABC
and CDA, as shown in Fig. 11, respectively. SOBE and SODF are the areas
within triangles of OBE and ODF, as shown in Fig. 11, respectively.
The dissipated energy Wpl and the cumulative dissipated energy
ΣWpl versus the number of cycles n for the tested RC beams under cyclic
loading are plotted in Fig. 10(a) and (b). From Fig. 12(a) it can be seen
that the capacity of dissipated energy becomes larger with the increase
of the number of cycles n. This indicates that energy dissipation of the
RC beams occurred mostly at the elastic-plastic stage of specimens. At a
common level of displacement loading, with increasing the number of
cycles, the capacity of dissipated energy becomes to be weaker. In ad-
dition, it can be found from Fig. 12(b) that with the increase of the
beams size, the cumulated dissipated energy increases.
As mentioned previously, the larger the size of the RC beam is, the
larger crack width and more cracks the specimens possess when the
beam failed [21,11,17,18]. Simply to say, as the structural size of beam
increases, the crack width turns to be larger and the crack length is also
longer when the beam fails, and therefore damage regions becomes to
be larger. This finally leads to more energy dissipation of the beam. The
work of Jin et al. [12] also reported the similar test observations on the
shear failure of RC beams.
(2) The relationship between the hysteretic damping factor ζ for each of
the loop and the number of cycles n for the ten RC beams under cyclic
loading is presented in Fig. 13. One can see from Fig. 13 that the
equivalent hysteretic damping factor for each loop gets to be larger
first, and then it becomes to be smaller as the number of cycles n in-
creases. This is because the progressive damage generated and

573
L. Jin et al. Engineering Structures 156 (2018) 567–586

(a) Fig. 6. Tested hysteretic curves and the corre-

CB-1-i CB-1-ii sponding skeleton curves of the 15 RC canti-
lever beams subjected to monotonic and low
Yield point
cyclic loading.
The point of 85% capacity

Monotonic load Monotonic load

Cyclic load Cyclic load

(b)
CB-2-i CB-2-ii

Monotonic load Monotonic load

Cyclic load Cyclic load

(c)
CB-3-i CB-3-ii

Monotonic load Monotonic load

Cyclic load Cyclic load

(d)
CB-4-i CB-4-ii

Monotonic load Monotonic load

Cyclic load Cyclic load

(e)
CB-5-i CB-5-ii

Monotonic load Monotonic load

Cyclic load Cyclic load

574
L. Jin et al. Engineering Structures 156 (2018) 567–586

6 1.2

Normalized stiffness [K/K0]

Mu: Peak moment CB-1-i
5 1
CB-1-ii
Slope: CB-2-i
0.8 CB-2-ii
4
CB-3-i
0.6
Slope k

CB-3-ii
3 Mu
CB-4-i
0.4 CB-4-ii
2 CB-5-i
0.85Mu
0.2 CB-5-ii
Cyclic loading-i
1 Cyclic loading-ii
0
Monotonic loading u f
0 6 12 18 24
0
0 500 1000 1500 Number of cycles n
Cross-setional height h [mm] Fig. 10. Stiffness degradation curves of the RC cantilever beams subjected to cyclic
loading.
Fig. 7. Effect of size on softening rate.

P
Pu
0.85Pu
0.75Pu

ǻy ǻp ǻu ǻ

ȝ=ǻu/ǻy
ȝ: Displacement ductility
ǻu: Displacement at 15% strength degradation from peak
ǻy: Displacement at yield load
Pu: Peak load
ǻp: Displacement at peak load

Fig. 8. Definition of the displacement ductility factor μ.

Fig. 11. Definition of the dissipated energy Wpl and the equivalent hysteretic damping
factor ζ.
30
Monotonic loading
behavior. In this Section, the existence of size effect in the flexural
25 Cyclic loading
strength was examined.
Ductility coefficient ȝ

20 Herein this work, we collected some calculation formulae for the

design of flexural capacity from several codes, e.g. EN 1992-1-1, ACI
15 318-2011 and GB50010-2010 of China. The comparison of the present
tested data and design values calculated by these codes is made and
10 plotted in Fig. 14. It can be clearly seen from Fig. 14 that all design
values are lower than the test values, and the flexural capacity M in-
5 creases with increasing the structural size (i.e. cross-sectional height h).
In addition, one can note that the calculation results obtained from the
0 three different codes are closely similar.For pure flexural specimens, the
0 200 400 600 800 1000 1200 normal stress can be calculated by σ = M / W . W can be evaluated by
Cross-section height h [mm] W = Iz/y, where y is the distance between the calculated point and
Fig. 9. Displacement ductility coefficients of the RC beams having different structural neutral axis, and Iz is the inertial moment of cross-section to the neutral
sizes. axis. For rectangular cross-section, W has the form as W = bh02/6. The
maximum nominal flexural stress (i.e. nominal flexural strength) σNu
can be easily determined by
developed gradually in the beams as the loading turns out to be larger.
Moreover, it can be noted from Fig. 13 that the variations of the hys- 6Mmax 6P l
teretic damping factor for all tested beams are similar. σNu = = max2
bh02 bh0 (3)

3. Size effect on the tested nominal flexural strength in which Mmax is the maximum moment generated at the fixed end,
Mmax = Pmax l. The length l is the distance from the loading point to the
As stated and discussed earlier, the cyclic loading can make the RC fixed end, Pmax is the maximum applied loading, b is the beam width
beams exhibit brittle failure behavior because of the low cycle fatigue and h0 is the effective cross-sectional height h0 of the beam. The

575
L. Jin et al. Engineering Structures 156 (2018) 567–586

Cumulated dissipated energy

CB-1-i
Dissipated energy
CB-1-i
Wpl[KN*mm] CB-1-ii
CB-1-ii

™Wpl [kN*mm]
CB-2-i
CB-2-i
CB-2-ii
CB-2-ii
CB-3-i
CB-3-i
CB-3-ii CB-3-ii
CB-4-i CB-4-i
CB-4-ii CB-4-ii
CB-5-i CB-5-i
CB-5-ii CB-5-ii

0 5 10 15 20 25 30 0 5 10 15 20 25 30
Number of cycles n Number of cycles n
(a) Dissipated energy of every cyclic (b) Cumulated dissipated energy
Fig. 12. The energy consumption of the RC beams subjected to cyclic loading.

0.43 effect, as shown in Fig. 15(b). It is to be noted in Fig. 15(b) that the
0.41 nominal strengths for the cross-sectional heights of 200 mm and
Hysteretic damping factor ȗ

0.39 CB-1-i
600 mm are slight larger than those of others due to the larger long-
CB-1-ii itudinal reinforcement ratio.
0.37
CB-2-i For the purpose of statistical regression of data, the Size Effect Law
0.35 CB-2-ii (SEL) proposed by Bažant and Planas [4] was used. This law can be
0.33 CB-3-i
expressed as
CB-3-ii
0.31 CB-4-i Bft
0.29 CB-4-ii σNu =
CB-5-i 1 + D / D0 (4)
0.27
CB-5-ii
0.25 in which, the unknown constants B and D0 can be determined by sta-
0 5 10 15 20 25 30 tistical regression analysis, and D is the structural size (i.e. the cross-
Number of cycles n sectional height h). Herein, the value of tensile strength ft of the con-
crete utilized was taken as 2.3 MPa. The SEL in Eq. (4) was derived by
Fig. 13. The hysteretic damping factor ζ of the RC beams subjected to cyclic loading.
simple energy release analysis and later by several other ways, espe-
cially by asymptotic matching based on the asymptotic power scaling
laws for very large and very small D [32]. Herein the study, a linear
2000 regression was conducted by Eq. (4). By squaring Eq. (4) and rearran-
GB50010-2010 ging, one can get
1600 EN1992-1-1 2
ACI 318-2011 ⎛ ft ⎞ = D + 1
⎜ ⎟

⎝ σNu ⎠ D0 B2 B2 (5)
1200 Test-Monotonic
M [kN*m]

Test-Cyclic This equation can be written in the following form:

800
400
0
0 200 400 600 800
Cross-sectional height h [mm]
Fig. 14. Relationship between the peak moment M and cross-sectional height h.
parameters of the five groups of RC beams with different structural
sizes, under both monotonic and cyclic loadings are provided in
Table 4.
The tested nominal flexural strengths of the tested RC beams sub-
jected to monotonic and cyclic loading are plotted in Fig. 15(a), and the
calculated nominal strengths corresponding to the calculated flexural
moments plotted in Fig. 14 are presented in Fig. 15(b). One can clearly
note from Fig. 15(a) that the obtained flexural strength of the RC beams
under cyclic loading is smaller than that under monotonic loading.
Moreover, the flexural strength of the RC beams under cyclic loading
decreases much faster as the structural size increases. That is to say, the
RC beams under cyclic loading presents a stronger size effect than the
beams under monotonic loading. However, the utilized formulae in EN
1992-1-1, GB 50010-2010 and ACI 318-2011 could not reveal any size
Y = AX + C (6)
2 2 2
where, Y = (ft/σNu) , X = D, A = 1/D0B and C = 1/B . Therefore, one
can see that D0 should be equal to C/A, and B equals to 1/ C .
The value of stress and other parameters necessary to perform the
regression analysis and to draw bi-logarithmic plot for various types of
1000 1200 RC specimens under monotonic and cyclic loading were calculated.
According to the tested data, the results of regression analysis for the
tested nominal flexural strength of RC beams with different structural
sizes, under monotonic and cyclic loadings, are respectively presented
in Fig. 14(a) and (b). Then A and C can be calculated from Fig. 16. The
two parameters of B and D0 can be fixed. Using these magnitudes, the
bi-logarithmic plot was drawn with log (D/D0) in the X-axis and log (σNu
/Bft’) in the Y-axis, as shown in Fig. 17. Moreover, the size effect law
(SEL) proposed by Bazant [3], the strength criterion without con-
sidering the influence of size effect (for plastic materials) and the classic
linear elastic fracture mechanics (i.e., the LEFM) for pure brittle ma-
terials are also plotted in Fig. 17 for comparison.
Furthermore, the correlation coefficient R2 for the present tested
data and Bazant’s SEL [3] was also calculated. The correlation coeffi-
cient R2 under monotonic and cyclic loading are respectively 0.92 and
0.93. This means that Bažant’s SEL curve fits the present tested data
satisfactorily. According to the above analyses, it can be concluded that
the plot shows the presence of size effect on the flexural strength of RC
beams in accordance with Bazant’s SEL.

576
 L. Jin et al. Engineering Structures 156 (2018) 567–586

Table 4
Ultimate load carrying capacity and corresponding nominal flexural strength.

Specimen Name Cross-sectional width b Effective cross-sectional Tested pmax [kN] (Positive/ Tested Mmax [kN*m] Nominal flexural strength σNu [MPa]
[mm] height h0 [mm] negative) (Positive/negative) (Positive/Negative)

MB-1 80 170 32.82 23.32 60.45

MB-2 160 365 94.75 130.75 41.18
MB-3 240 540 217.71 470.25 40.31
MB-4 320 735 350.49 1030.44 35.76
MB-5 400 930 505.45 1880.26 32.61
CB-1-i 80 170 33.26/34.62 22.62/23.54 58.70/61.09
CB-1-ii 80 170 35.29/31.56 22.62/23.54 62.28/55.69
CB-2-i 160 365 83.63/80.30 115.41/110.81 36.36/34.91
CB-2-ii 160 365 84.47/83.64 116.57/115.43 36.73/36.36
CB-3-i 240 540 170.94/179.84 369.23/388.46 31.66/33.30
CB-3-ii 240 540 179.84/160.16 388.46/365.38 33.31/31.33
CB-4-i 320 735 305.82/303.87 894.13/893.39 31.21/31.01
CB-4-ii 320 735 305.63/305.17 898.56/897.20 31.19/31.14
CB-5-i 400 930 402.98/399.11 1487.58/1484.68 26.00/25.75
CB-5-ii 400 930 410.73/400.68 1492.63/1490.53 26.50/25.85

It can be seen from Fig. 17 that compared with the one under size effect could be reliably accounted for in the numerical analysis of
monotonic loading, the test data for the RC beams under cyclic loading RC members [43,33,34].
lies much closer to the inclined asymptote of slope -1/2, which corre- In computational material science, concrete is characterized as a multi-
sponds to the LEFM. The cyclic loading makes the failure of beams more phase material with several different representative scales. At macroscopic
brittle, which resulting into a stronger size effect in the flexural strength scale, concrete could be regarded as a homogeneous material, while at
of RC beams. mesoscopic scale it is treated as consisting of coarse aggregates and mortar
It should be noted that, the size effect generates due to structural matrix [35]. Many concrete nonlinear fracture models, such as Jirasek and
geometry but not cyclic loading. However, compared with monotonic Zimmermann [36,37], Bažant and Jirasek [38], Pijaudier-Cabot et al.
loading, the low cyclic fatigue loading changes the failure features. [39], Majewski et al. [40], Skarżyński et al. [41], Syroka-Korol [42,43],
Namely, cyclic loading makes RC beams fail more brittle, resulting into based on the theory of fracture mechanics and the non-local theory, have
a more obvious size effect. been established at macro-scale. The macroscopic computational models
in the literatures are able to describe the nonlinear mechanical properties
and the size effect behavior of concrete by including a characteristic length
4. Supplementary tests by means of 3D meso-scale numerical
of micro-structure. However, the macroscopic mechanical properties and
method
the failure patterns of concrete are closely associated with the meso-
structure of concrete [44,45]. Mesoscopic models have been proven to be
Because of the limitation of testing equipment and economic cost, it
the most practicable and useful approach for studying the influence of the
is hard to do many tests to study the influencing factors on the size
concrete composition on the macroscopic properties, and also for gaining
effect of RC members. The finite element method has become an ef-
insights into the origin and nature of the nonlinear behavior of concrete,
fective method to model the mechanical behavior of engineering
such as some advanced mesoscopic continuum models for concrete pre-
structures, and finite element analyses are essential for supplementing
sented by Ren et al. [46], Wang et al. [47], Skarżyński et al. [48]. Fur-
experimental research in providing insights into structural behavior
thermore, the size effect and the process of softening are all related to the
[32]. From a macroscopic of view, a deterministic size effect is caused
heterogeneity of the material in the domain within the structural dimen-
by the formation of a region of intense strain localization with a certain
sion [47]. A mesoscopic mechanical model is composed by diverse meso-
area that is not negligible to the element cross-section dimensions and is
scale components with different material properties, and therefore the
large enough to cause a significant stress redistribution in the structure
heterogeneity of concrete can be explicitly described. Herein this section, a
[2]. From a mesoscopic point of view, the sources of size effect in RC
3-D meso-scale numerical method would be developed to explore the
members are mainly attributed to the following two reasons [33,34]:
failure behavior of RC beams, in which the concrete heterogeneities and
(1) the heterogeneity and nonlinearity of concrete itself; (2) the com-
the steel/concrete interactions were explicitly taken into account.
plex nonlinear interaction between steel and concrete. Only when these
two aspects were correctly modeled in the numerical simulations, the

70 Fig. 15. Nominal flexural strengths σNu vs. cross-sec-

70
Cyclic loading GB50010-2010 tional height h for the RC beams, (a) tested strength,
Calculated strength ıNu [MPa]

60 60 EN 1992-1-1 and (b) calculated strength by different codes.

Tested strength ıNu [MPa]

Monotonic loading
50 50 ACI 318-2011

40 40
30 30
20 20
10 10
Tested results
0 0
0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200
Cross-sectional height h [mm] Cross-sectional height h [mm]

577
L. Jin et al.
0.008
0.007
y = 3E-06x + 0.0012
0.006
R² = 0.867
0.005
Y=(ft/ıNu)2
Y=(ft/ıNu)2
0.004
0.003
0.002
0.001
0 0
0 200 400 600 800 1000 1200
X=D [mm]
(a) Monotonic loading
Fig. 17. The comparison of nominal flexural strength fitted by Bazant’s SEL.
4.1. Generation of the RC beams at meso-scale
Based on the test results of aforementioned experimental
campaign, three more groups of numerical specimens with some ex-
tensive structural sizes, i.e. 480 mm × 1200 mm×4900 mm,
640 mm × 1600 mm×6500 mm and 800 mm × 2000 mm×7700 mm,
are established for supplementing and extending the experimental
research. The seismic performances of cantilever RC beams with dif-
ferent structural sizes could therefore be analyzed by a combined
experimental and numerical method.
At meso-scale, concrete is often regarded as a three-phase hetero-
geneous composite involving aggregate particles, mortar matrix and
the interfacial transition zones (ITZ). The evaluation of the composite
behavior of concrete at mesoscopic level requires the generation of a
random aggregate structure in which the shape, size and distribution of
the aggregate particles closely resemble real concrete in the statistical
sense [35]. Moreover, aggregates, including both fine and coarse ag-
gregates, generally occupy 60–80% of the volume of concrete and
greatly influence the properties, mix proportions and economy. Herein
this study, for the sake of simplicity, the fine aggregates (corresponding
to the aggregates sized range from 0 to 5 mm) and the cement paste are
both merged into the mortar matrix, and the coarse aggregates were
modeled as spheres with random distribution. Similar with that in the
experiments, the coarse aggregate represents around 40–50% of con-
crete volume. The classic Fuller’s curve was utilized to describe the
aggregate size distribution, and all RC beams were defined as two-
graded concrete specimen according to the Chinese code [49]. The sizes
of aggregates were therefore equivalent to two specific values. All small
aggregates have the diameter of 10 mm, and all large aggregates have
the diameter of 25 mm. A total of 8 groups of cantilever beams were
built for the numerical tests. Additionally, a group of
150 mm × 150 mm×150 mm cubic specimen was set up for the de-
termination of the mechanical properties of the meso components. The
Engineering Structures 156 (2018) 567–586
0.008 Fig. 16. Regression analysis of the test results of the
beams under different loading.
0.007 y = 6E-06x + 0.0009
0.006 R² = 0.885
0.005
0.004
0.003
0.002
0.001
0 200 400 600 800 1000 1200
X=D [mm]
(b) Cyclic loading
aggregate particles were put into the mortar matrix by using the “Take
and Place” method which was also utilized in other efforts [35,50]. It is
based on a random sampling principle of Monte Carlo’s simulation
method applied by taking samples of aggregate particles from a source
whose size distribution follows a certain given grading curve and pla-
cing the aggregate particles one by one into the concrete in such a way
that there is no overlapping with particles already placed. Herein this
study, we adopted the algorithm for generating a realistic aggregate
structure proposed in [35,51]. The thin layers around the aggregate
particles were set as ITZs.
The steel reinforcement cage is set up individually after the gen-
eration of the concrete partition, and then it is simply inserted into the
concrete partition regardless of the meso components. Specifically, the
interactions between the aggregates and the rebar are not taken into
consideration during the numerical tests, but uniformly set a nonlinear
spring between the concrete partition and the rebar partition. This may
lead to an uncertain mutual effect between the concrete and the re-
inforcement, which may be a drawback for this meso-scale method, and
it should be modified and revised in the future work.
A typical 3D mesoscopic numerical model of RC beam is shown in
Fig. 18. As presented in Fig. 18, each phase has its own color and also its
distinguished mechanical properties. The eight-node hexahedron ele-
ments were employed to divide the concrete meso-components, and the
beam-elements were utilized for the discretization of the reinforcing
bars. The Newton-Raphson method was used to solve the nonlinear
equations.
4.2. Constitutive model
In recent years, many coupled plasticity damage models have been
proposed to describe the mechanical behavior of cementitious materials
[44,32]. Among these efforts, the concrete damage plasticity model was
first proposed by Lubliner et al. [52] for monotonic loading, and it was
developed later by Lee and Fenves [53] to consider the dynamic and
cyclic loadings. This model, which has been included in the ABAQUS
package, has been widely used for the description of the static and
dynamic mechanical behaviors of concrete-like materials.
4.2.1. Isotropic damage model
The stress-strain relation for the general three-dimensional multi-
axial condition is given by the scalar damage elasticity equation
σ = (1−d ) D0el : (ε−ε pl ) (7)
in which D0el is the initial elasticity matrix, d means the damage variable
and ε pl is the plastic strain tensor. Tensile damage and compressive
damage are characterized independently by two hardening parameters,
ε tpl̃ and ε cpl̃ (subscripts ‘t’ and ‘c’ stand for tension and compression,
respectively), which are referred to as equivalent plastic strains in
tension and compression. The evolution equations of the hardening
578
L. Jin et al. Engineering Structures 156 (2018) 567–586

Fig. 18. Meso-scale computational model of RC

cantilever beams.

Loading
RC specimen

Reinforcement
ITZ
Mortar matrix
Aggregate
Reinforcement

Fixed bottom

σt This representation, although somewhat simplified, captures the

ft main features of concrete behavior. Sometimes, in order to avoid un-
reasonable mesh sensitive results, the tensile post-failure behavior is
E0 given in terms of a fracture energy cracking criterion by specifying a
E0 (1 − d c )(1 − d t ) stress-displacement curve instead of a stress-strain curve. Under multi-
E0 (1 − d t ) axial loading conditions, the evolution equation of the hardening
parameters for uniaxial loading should be extended.
ε According to Lee and Fenves’s work [53], tensile and compressive
E0 (1 − d c ) E0 crack closure
equivalent plastic strains have the following forms

ε tpl̃ = r (σ )̂ ε max
̂
pl
(10a)
fc
σc ε cpl̃ = ̂ ε min
−(1−r (σ )) ̂
pl
(10b)
Fig. 19. Unaxial loading path of the plasticity damage model. ̂ and
pl
in which ε max ̂
pl
are the maximum and minimum plastic strain
ε min
tensors, respectively; r (σ )̂ means a multi-axial stress weight factor and
variables, ε tpl̃ and ε cpl̃ , are conveniently formulated by considering is defined as
uniaxial loading conditions first and then extended to multi-axial con- 3

ditions. The model assumes that the uniaxial tensile and compressive ∑ 〈σi 〉̂
responses of concrete are characterized by damaged plasticity, as shown r (σ )̂ = i=1
3
0 ⩽ r (σ )̂ ⩽ 1
in Fig. 19. It is assumed that the stress-strain response follows a linear ∑ |σi |̂
elastic relationship under uniaxial tension until the value of the failure i=1 (11)
stress, ft, is reached. The failure stress corresponds to the onset of micro-
in which σi ̂ are principal stress components (i = 1, 2 for 2-D problem
cracking in the concrete material. Beyond the failure stress the forma-
and i = 1, 2, 3 for 3-D problem), the Macauley bracket “〈·〉” is defined
tion of micro-cracks is often represented macroscopically with a soft-
by 〈x 〉 = 0.5 × (|x| + x ) .
ening stress-strain response, which induces strain localization in the
concrete structure. The uniaxial stress-strain curves can be converted
4.2.2. Plasticity yield surface
into stress versus plastic strain curves of the form as
The model makes use of the yield function of Lubliner et al. [32],
σt = σt ( ε tpl̃ ,θ,fi ) (8a) with the modifications proposed by Lee and Fenves [53] to account for
different evolution of strength under tension and compression. The
σc = σc ( ε cpl̃ ,θ,fi ) (8b) evolution of the yield surface is controlled by the two hardening vari-
ables, ε tpl̃ and ε cpl̃ . In terms of effective stress, the yield function F used
where θ is the temperature and fi (i = 1,2,…) denotes other predefined in the constitutive model takes the following form
field variables.
1
As shown in Fig. 19, when the specimen is unloaded from any point F = F (σ , ε pl̃ ) = [q −3αp + β ( ε pl̃ ) 〈σmax 〉−γ〈−σmax 〉]−σc (εcpl ) = 0
1−α
on the strain softening branch of the stress-strain curves, the unloading
response is weakened: the elastic stiffness of the material appears to be (12)
1
degraded. Herein, the degradation of the elastic stiffness is character- in which p = − 3 σ : I is the effective hydrostatic pressure stress; I is a
ized by two damage variables, tensile damage factor dt and compressive stress invariant; q =
3
S: S means the Mises equivalent effective
2
damage factor dc, and the stress-strain relationships under tension and
stress, and S is the effective stress deviator, defined as S = p I + σ ; σmax
compression are respectively
is the maximum principal effective stress; and α , β ( ε pl̃ ) and γ respec-
σt = (1−d t ) E0 (εt− ε tpl̃ ) (9a) tively has the following forms
σcc/ σc−1
σc = (1−d c ) E0 (εc− ε cpl̃ ) (9b) α=
2σcc/ σc−1 (13a)

579
L. Jin et al. Engineering Structures 156 (2018) 567–586

uniaxial tension σ2 Table 5

Mechanical parameters of the three meso-components of concrete and reinforcing bars
σt utilized.

σ1 Mechanical Aggregate Motar ITZ Longitudinal bar Stirrups

uniaxial compression parameters matrix
biaxial tension
* ^
Compressive 45.3 37.2
strength σc/MPa
Tensile strength σt/ *
3.8 ^
3.3
MPa
* ^ * *
Elastic modulus E/ *70.1 30.8 24.7 196 210
GPa
Poisson’s ratio ν *0.2 *
0.2 ^
0.2 *
0.3 *
0.3
Fracture energy Gc 60 50 30
(σ cc , σ cc ) biaxial compression
σc [J/m2]
Shear - Dilative 18 15
angle ψ (°)
* *
Fig. 20. Yield surface for plane stress conditions. Yield strength fy/ 405 298
MPa

σc ( ε cpl̃ ) fracture energy approach provides mesh-independent results for load-

β ( ε pl̃ ) = (1−α )−(1 + α )
σt ( ε tpl̃ )
3(1−K c )
γ=
2K c−1
in which σcc/ σc means the ratio of initial equivalent-biaxial compressive
yield stress to initial uniaxial compressive yield stress; Kc is the ratio of
the second stress invariant on the tensile meridian. Typical yield surface
is shown in Fig. 20 for plane stress conditions.
4.2.3. Plastic flow rule
The present damaged plasticity model assumes non-associated po-
tential plastic flow. The flow potential G used for this model is the
Drucker-Prager hyperbolic function
G= (ησt tanψ)2 + q 2 −p tanψ
where ψ is the dilation angle measured in the p-q plane at high con-
fining pressure; σt is the uniaxial tensile stress at failure; η is a para-
meter, referred to as the eccentricity, that defines the rate at which the
function approaches the asymptote (the flow potential tends to a
straight line as the eccentricity tends to zero).
For more details about the constitutive model, refers to the efforts
[33,54]. It is understandable that damaged plasticity model was uti-
lized to describe the mechanical responses of concrete. As for the ITZ
and the mortar matrix, they both can be regarded as cementitious
materials. The only differences between these two meso components
and a concrete is the particle size distribution and the porosity of the
material. Actually, both Grote et al. [55] and Park et al. [56] have
found that the dynamic mechanical behavior of mortar matrix is similar
to that of concrete, and the plastic damage model can be used to de-
scribe the dynamic mechanical behavior of mortar. The ITZ, in fact, is a
kind of mortar matrix material with high porosity [57], thus the me-
chanical parameters of the ITZ can be characterized by weakening the
ones of mortar matrix [50]. Therefore, herein this study, for the mortar
matrix and the ITZ phases, the damaged plasticity model was utilized to
describe their mechanical behavior. These treatments are similar with
those in [44,32,54].
According to Majewski et al. [58], to preserve the well-posedness of
the boundary value problem, to obtain mesh-independent results and to
include a characteristic length of micro-structure for simulations of a
deterministic size effect, a non-local theory should be used as a reg-
ularization technique. Herein the utilized constitutive model, it has to
be stressed that a characteristic length is lacking. To avoid or relax
unreasonable mesh sensitive results, the tensile post-failure behavior is
given in terms of a fracture energy cracking criterion by specifying a
stress-displacement curve instead of a stress-strain curve, e.g. in the
efforts [44,32,54]. Nevertheless, it is to be noted that, the use of a
(13b) displacement diagrams only.
Moreover, aggregate particles were set as elastic due to their high
(er) tensile and compressive strengths. Similar with that in [58], the
(13c) mechanical behavior of steel bars was described by an ideal elastic-
plastic model.
4.3. Mechanical parameters
The main mechanical parameters for the three meso-components
and the reinforcing bars, including elastic modulus E, tensile/com-
pressive strengths σt/σc, and fracture energy Gc, etc., are listed in
Table 5. The mechanical parameters of aggregate and mortar matrix are
real test ones (labeled by “∗”), and the magnitudes of the fracture en-
ergy for the three meso-components are quoted from Yılmaz and Mo-
(14)
linari’s [45] work. The mechanical parameters of ITZs (labeled by “^”)
are determined by repeated trials, because it is uneasy to measure by
test. In reality, the ITZs are composed by cement mortar with relatively
a larger amount of initial defects (i.e. voids and micro-cracks). The
mechanical properties of the ITZs (especially the strength and modulus
of elasticity) were chosen to be less than that of the mortar matrix by a
constant factor. Similar with that of Jin et al. [33], some repeated
uniaxial compressive tests on square concrete specimens sized of
150 mm × 150 mm × 150 mm were conducted to obtain reasonable
parameters for ITZs (see Fig. 21). That is to say, different groups of
mechanical parameters for the ITZ were first assumed, and then the
assumed parameters were utilized to evaluate the macro-mechanical
properties of concrete. The assumed parameters that could obtain ac-
curate mechanical properties (especially for strength and elastic mod-
ulus) of concrete were set as the right parameters. It is to be noted that,
the value of this constant is related to the selected thickness of the ITZs.
The actual thickness of the ITZs is about 20–50 μm [59,60]. However, it
is almost impossible to use this tiny thickness for the consideration of
computational amount [61]. Herein this study, similar with that in Jin
et al.’s work [33], the selected ITZs thickness was set as 1 mm. With the
parameters listed in Table 5, one can obtain that the simulated uniaxial
compressive strength of concrete is 42.6 MPa (close to the tested one,
i.e. 42.8 MPa). Therefore, the selected parameters for the ITZs are re-
garded to be reasonable. Moreover, on the effects of ITZ thickness and
ITZ mechanical parameters on the global mechanical properties of
concrete one can refer to Kim and Al-Rub’s [50] and the previous si-
mulation efforts [61].
Additionally, a series of numerical tests for the evaluation of mesh
sensitivity are conducted by setting different mesh sizes in the cubic
concrete specimen sized by 150 mm × 150 mm × 150 mm. A total of
four groups of mesh sizes, i.e. 2 mm, 5 mm, 8 mm and 12 mm, were
adopted. Fig. 21(a) presents the generation of the four mesh sized

580
L. Jin et al. Engineering Structures 156 (2018) 567–586

2.4
54
Compressive stress ı [MPa]

48 2mm 2.0
5mm
Tensile stress ı [MPa]

42
8mm
36 12mm 1.6
30
1.2
24
18 0.8
12
0.4
6
0 0
0 0.002 0.0030
0.0015 0.004 0.0045
0.006 0.0060
0.008 0.0075
0.01 0 0.00005 0.0001 0.00015
Nominal compressive strain Nominal tensile strain

Fig. 21. Evaluation of the mesh sensitivity, (a) the cube concrete model having different mesh sizes; (b) typical uniaxial compressive and tensile failure patterns; and (c) stress-strain
curves under uniaxial compression and tension

specimen, Fig. 21(b) shows the typical failure patterns of concrete, and
Fig. 21(c) plots the comparison of tested and numerical stress-strain
curves of concrete under uniaxial compression and tension. One can see
from Fig. 21(c) that the numerical results are consistent with the tested
ones. Moreover, it can be seen from Fig. 21(c) that, when the mesh size
is smaller than about 5 mm, the numerical results are almost identical
and more stable. Therefore, considering of both accuracy and efficiency
of the numerical tests, the average size of the finite elements in calcu-
lations was set to be 5 mm for all RC beams.
Due to the complexity of the interaction between steel bars and the
surrounding concrete, there is no universal bond-slip law for now.
However, the bond between concrete and the steel bars plays a crucial
role for the structural behavior. Accordingly, it embraces three major
mechanisms: adhesion and friction between concrete and steel surface,
as well as the bearing of reinforcement ribs against concrete [58].
Herein this study, the numerical calculations were carried out with a
bond-slip law (assuming a relative displacement between concrete and
reinforcement) adopted by the “Code for design of concrete structures”
of China [49]. In order to describe the bond-slip behavior between steel
and concrete, a nonlinear spring element was set between concrete
elements and steel elements, as shown in Fig. 22 (a). The tangential
bond stress τ and slip s between reinforcement and concrete were de-
scribed by the shear stress and slip (τ-s) curve that recommended by the
concrete structure design code of China [49], as presented in Fig. 22(b).
The related parameters between steel/concrete is shown in Table 6.
4.4. Effects of related parameters on concrete behavior
The macro-mechanical properties of concrete are closely related to
the meso-/micro-structure of concrete [35,44,47,50,51]. Therefore, the
parameters involving the volume fraction of aggregate particles, ag-
gregate size, aggregate shape, aggregate distribution, ITZ thickness and
ITZ strength should have significant influences on the mechanical be-
havior of concrete.
Actually, the effects of the aforementioned meso-scale parameters
on the macroscopic mechanical properties of concrete have been

581
 L. Jin et al. Engineering Structures 156 (2018) 567–586

IJ Fig. 22. Bond-slip model for the interaction between

steel bars and surrounding concrete.

IJ u
i
u

IJ cr
cr

r
IJ
r

o s
s s s
cr u r

(a) Spring element (b) Relation between shear stress IJ and slip s

Table 6
Parameters utilized in the bond-slip model.
Inflection point Splitting (cr)
Stress τ (MPa) τcr = 2.5ft
Slip s (mm) scr,l = 0.025d
Note: “d” means the diameter of steel; “ft” means the cracking strength of concrete.
studied systemically in our previous work [61,62]. Generally, the
macroscopic elastic modulus and the strength of concrete under tension
decrease as aggregate size increases, and at a fixed value of total ag-
gregate fraction, a larger average size of aggregate particles corre-
sponds to a larger size of fracture process zone; aggregate shape has
little effect on the macroscopic modulus of elasticity and the averaged
tensile strength as well as the failure patterns of concrete; the spatial
distribution of aggregate has a negligible influence on the overall me-
chanical properties of concrete; the ITZ thickness and ITZ strength have
a great influence on both the fracture process and the global mechanical
properties of concrete. The effects of these parameters listed above have
also been discussed in the work of Kim and Al-Rub [50]. To avoid re-
peated simulations, the effects of the parameters were not presented in
some detail. Herein this study, a brief investigation on the effect of
aggregate distribution was conducted by presenting 40 groups of uni-
axial compression tests with randomly generated square specimens, as
shown previously in Fig. 23. The numerical results indicate that, the
aggregate distribution could only affect the post-peak softening beha-
vior of concrete and merely change the elastic response and the peak
value of the material. As for the influences of the aggregate shape, size
and volume fraction, the same values of the parameters concerning of
these influencing factors were utilized and therefore the influences in-
Peak (u) Remnant (r)
troduced by these materials-level parameters are approximately elimi-
τu = 3ft τr = ft nated at components-level. For more details on the influences of other
su,l = 0.04d sr,l = 0.55d parameters on the mechanical behavior of concrete, refer to the pre-
vious work [59,60].
In addition, according to the simulations, it is found that the
minimum aggregate size has a negligible influence on the macroscopic
mechanical properties of concrete. Therefore, all the fine aggregates
with size ranging from 0 to 5 mm and the cement paste are merged into
the mortar matrix in the study for the sake of computational efficiency.
Certainly, the bond between concrete and reinforcement has a sig-
nificant influence on the failure of RC beams. For more details on the
influences, refer to the efforts in [42].
4.5. Validation of the numerical method
The final simulated failure patterns of the RC cantilever beams
having different structural sizes under both monotonic and cyclic
loadings are presented in Fig. 24. It can be found clearly from Fig. 24
that the simulated failure patterns have good agreement with the actual
tested ones. Nevertheless, one can clearly note from Fig. 24 that there
are still some differences from the numerical outcomes and the ex-
perimental observations, e.g. the crack inclinations and the crack spa-
cing, especially for the RC beams subjected to monotonic loading. This
should be caused by the following reasons: (1) the adopted bond-slip
relations and the related parameters should be different from the real

48 Fig. 23. Effect of aggregate spatial distribution type

on the global mechanical properties of concrete.
Experimental
42
Nominal compressive stress [MPa]

Numerical

36

30

24

18

12

0
0 0.5 1 1.5 2 2.5 3 3.5 4
Nominal compressive strain (x10-3)

582
L. Jin et al. Engineering Structures 156 (2018) 567–586

1000mm 1000mm

800mm 800mm

600mm 600mm

400mm 400mm

200mm 200mm

Plastic strain
0 0.0033

Fig. 24. Comparison of simulated and tested failure patterns of the RC beams, (a) under cyclic loading and (b) under monotonic loading.

30 500 2000
M [kN*m]

M [kN*m]
M [kN*m]

1500
20 300
1000
10
100 500
¨ [mm] ¨ [mm] ¨ [mm]
0 0
-50 -30 -10 10 30 50 -120 -80 -40-100 0 40 80 120 -200 -100 0 100 200
-500
-10 CB-2-T CB-4-T
CB-2-S -1000 CB-4-S
-300
-20 CB-1-T CB-3-T -1500 CB-5-T
CB-1-S CB-3-S CB-5-S
-30 -500 -2000

30
500 2000
25
400
1500
20
M [kN*m]

M [kN*m]

M [kN*m]

300
15 1000
200 MB-4-T
10 MB-2-T
MB-2-S 500 MB-4-S
5 MB-1-T 100
MB-3-T MB-5-T
MB-1-S MB-5-S
MB-3-S
0 0 0
0 15 30 45 60 0 50 100 150 200 250 300 0 100 200 300 400 500
¨ [mm] ¨ [mm] ¨ [mm]

Fig. 25. Comparison of the skeleton curves obtained by numerical and experimental results, (a) under cyclic loading, and (b) under monotonic loading.

583
L. Jin et al. Engineering Structures 156 (2018) 567–586

1800 Table 8
Physical parameters of the larger sized RC beams simulated.
1500
Specimen name CB-6/MB-6 CB-7/MB-7 CB-8/MB-8
1200 T
M [kN*m]

Cross-section size (b × h) (mm2) 480 × 1200 640 × 1600 800 × 2000

Longitudinal reinforcement ratio ρ 1.26% 1.24% 1.24%
900 Hoop reinforcement ratio ρsv 0.14% 0.14% 0.14%
S Effective height of Beam h0 (mm) 1120 1510 1900
600 Beam effective length a = λ * h0 (mm) 4480 6040 7200
Total length of beam l (mm) 4900 6500 7700
300

0 2000 mm (CB-8/MB-8). The detailed parameters of the beams are listed

0 1 2 3 4 5 6 in Table 8.
İ/10-3
Fig. 26. Simulated and tested tensile strains of the longitudinal bars at the fixed bottom. 5.1. Simulated failure of larger sized beams

The failure modes of the three larger-sized RC beams (the cross-

section heights are 1200 mm, 1600 mm and 2000 mm) and the Mises
tested ones; (2) in the simulations, the constitutive models for the
stresses distribution within the steel reinforcements are presented in
concrete meso components cannot avoid the mesh sensitivity totally.
Fig. 27. It can be clearly found that the failure patterns of the three
The above existed problems should be solved in the future work.
larger-sized RC beams are closely similar with those of RC beams
Fig. 25 shows the comparison between the simulated and tested
having smaller sizes, under both monotonic and cyclic loadings. It can
bearing capacity under different loading. In Fig. 25, “T” means the
also be noticed from Fig. 27 that, for larger-sized beams, the horizontal
tested results, and “S” means the simulated results. One can see that in
cracks propagate towards the bottom of the specimens, and oblique
the elastic stage, the simulated results agree well with the experimental
cracks generate, resulting to the flexural-shear failure pattern.
results. In the yield stage and the softening stage, the bearing capacity
Moreover, from Fig. 27 one can note that the Mises stresses contour
between simulated result and the test result has somewhat different, but
of the beams with cross-sectional height of 2000 mm that, the long-
the errors can be accepted excepted for the beam of CB-1.
itudinal steel bars yielded (reached the average yield strength of
Fig. 26 plots the comparison of the simulated results and the tested
400 MPa) at the end of the specimen under both monotonic and cyclic
results on the strain ε of the longitudinal steel bars at the end of the
loadings.
beams under monotonic load. One can note that the simulation results
Fig. 28(a) and (b) plot the skeleton curves (moment M versus de-
are consistent with the tested ones. At the initial stage, the beams were
flection Δ) of the three larger-sized RC beams subjected to cyclic and
all in elastic stage. As the loading continued, the steel strains increased
monotonic loadings, respectively. Again, one can note from the com-
linearly, and the strains increased significantly after the longitudinal
parison of Fig. 28(a) and (b) that, (1) the peak moments M of the beams
steel bars reached their yield strength.
under cyclic loading are obviously smaller than those under monotonic
Table 7 presents the comparison of the tested and simulated results
loading; (2) the deformation capacity or the ductility capacity of the
on the nominal strength of the RC beams. One can clearly note that, for
beams under cyclic loading are weaker than those under monotonic
both series CB and MB, the simulated nominal strengths are closely
loading. In summary, the mechanical behavior of the larger-sized RC
similar with the tested nominal strengths. This demonstrates the rea-
beams is basically similar with that of the smaller-sized RC beams
sonability of the meso-scale simulation method.
tested.

5. Numerical test results of larger sized beams 5.2. Size effect in larger-sized beams

In this section, the above validated meso-scale simulation method The relationships between the simulated flexural strengths and the
was extended to explore the failure behavior of larger-sized geome- structural size D (i.e. cross-sectional height h) of the RC beams under
trically similar RC beams. The cross-sectional heights of the simulated both monotonic and cyclic loadings are all plotted in Fig. 29. It can be
beams were 1200 mm (CB-6/MB-6), 1600 mm (CB-7/MB-7) and seen that the simulated flexural strengths of the geometrically similar
RC beams decrease as the cross-sectional height increases, indicating a
size effect. Moreover, it can be noted from Fig. 29 that as the structural
Table 7 size increases, the decrease trend of the obtained flexural strengths
Comparison of nominal strength obtained from tests and simulations.
slows down. It is believed that as the structural size becomes larger and
Specimen label Tested strength (MPa) Simulated strength (MPa) Error (%) larger the flexural strength should be close to a constant.
Similarly, the bi-logarithmic plot was also drawn with log (D/D0) in
CB-1-i 58.70 51.45 12.35 the X-axis and log (σNu /Bft’) in the Y-axis, as shown in Fig. 30. The
CB-1-ii 62.28 17.39
correlation coefficient R2 for the simulated data and Bažant’s SEL was
CB-2-i 36.36 37.35 2.72
CB-2-ii 36.73 1.69 evaluated to be 0.975. This means that, for the simulated larger-sized
CB-3-i 31.66 33.27 5.10 RC beams, the nominal flexural strength is also in accordance with
CB-3-ii 33.31 0.12 Bažant’s SEL. Moreover, compared with that for monotonic loading, the
CB-4-i 31.21 32.17 3.01 simulated data for cyclic loading lies closer to the LEFM. This indicates
CB-4-ii 31.19 3.14
CB-5-i 26.00 26.24 0.92
a more pronounced size effect for the beams under cyclic loading.
CB-5-ii 26.50 0.98
MB-1 60.45 54.58 9.71 6. Conclusions
MB-2 41.18 45.86 11.37
MB-3 40.31 43.07 6.85
A series of experimental tests on the flexural failure behavior of
MB-4 35.76 34.76 2.80
MB-5 32.61 32.72 0.34 geometrically similar RC cantilever beams subjected to both monotonic
and cyclic loadings were conducted. The effects of structural size and

584
L. Jin et al. Engineering Structures 156 (2018) 567–586

Fig. 27. Failure patterns of larger-sized RC beams subjected to (a) cyclic loading, and (b) monotonic loading.

12000 14000 Fig. 28. The skeleton curve of larger-sized RC

M [kN*m]

beams subjected to (a) cyclic loading, and (b)

12000 monotonic loading.
8000
10000
M [kN*m]

4000 8000
¨ [mm]
6000
0
-300 -200 -100 0 100 200 300 4000 MB-6-S
-4000 MB-7-S
2000
MB-8-S
CB-6-S 0
-8000
CB-7-S 0 200 400 600 800 1000
CB-8-S
-12000 ¨ [mm]

60 loading type on the flexural failure of RC cantilever beams were sys-

Simulation results tematically explored. In particular, the size effect in the nominal flex-
Simulated strength ıNu [MPa]

50
ural strength was studied. Moreover, a 3D meso-scale numerical
40 method considering concrete heterogeneities and the nonlinear bond-
slip relations was established to model the failure of RC cantilever
30 beams, and good agreement between the simulation results and the
tested results proved the efficiency of the meso-scale approach. Finally,
20 the failure behavior and the size effect of larger-sized RC beams were
investigated based on the meso-scale method.
10
Monotonic Cyclic According to the analyses, the following conclusions can be drawn:
0
0 500 1000 1500 2000 2500 (1) Under both monotonic and cyclic loadings, all RC beams tested in
Cross-sectional height h [mm] this work exhibit a quasi-brittle failure pattern, and the measured
nominal flexural strength decreases as the structural size increases.
Fig. 29. Relationship between the simulated nominal flexural strength and structure size.
(2) The nominal flexural strength of the RC beams follows closely the
size effect law proposed by Bažant in all cases tested.
(3) Under low cyclic fatigue loading, the strength of concrete and steel
0.1
bars degrades, and the bonding capacity between steel bars and
Strength criterion
0 surrounding concrete weakens due the accumulation of damage.
(4) As the structural size of the beams increases, the ductility capacity
-0.1
R2 = 0.975 improves for the beams under monotonic load, while it does not
log(ıNu/Bft´)

-0.2 Simulation results change obviously for the beams under cyclic loading. In addition,
the stiffness degradation capacity does not have an obvious change
-0.3 Bazant's SEL
with increasing the structural size of RC beams.
Strength criterion
-0.4 (5) Compared to monotonic loading, cyclic loading makes the failure of
LEFM
-0.5 Cyclic RC beams less ductile. This leads to the fact that the size effect in
Monotonic the flexural strength of the RC beams under cyclic loading is much
-0.6 more obvious than that under monotonic loading.
-2.0 -1.6 -1.2 -0.8 -0.4 0.0 0.4 0.8 1.2
(6) Size effects in the flexural failure of RC beams are closely associated
log(D/D0)
with loading mode, namely they are a function of loading in ad-
Fig. 30. Logarithmic curve of the simulated nominal flexural strength vs. structure size. dition to geometry.

585
L. Jin et al. Engineering Structures 156 (2018) 567–586

Acknowledgements reinforced concrete column behaviour under biaxial cyclic loading. Earthq Eng
Struct Dyn 2013;42(2):239–59.
[30] Lu Y, Li N, Li S, Liang H. Behavior of steel fiber reinforced concrete-filled steel tube
This research was supported by the National Key Basic Research and columns under axial compression. Constr Build Mater 2015;95:74–85.
Development Program of China (No. 2016YFC0701100), the Science [31] Wang B, Wang J, Gong X, Liu B. Experimental studies on circular CFST frames with
Fund for Creative Research Groups of the National Natural Science ALC walls under cyclic loadings. Int J Steel Struct 2014;14(4):755–68.
[32] Genikomsou AS, Polak MA. Finite element analysis of punching shear of concrete
Foundation of China (No. 51421005), and the Science Fund for slabs using damaged plasticity model in ABAQUS. Eng Struct 2015;98:38–48.
Innovation Team of the Ministry of Education of China (No. IRT13044). [33] Jin L, Li D, Du X, Lu A, Ding ZX. Experimental and numerical study on size effect in
The support is gratefully acknowledged. eccentrically loaded stocky RC columns. ASCE J Struct Eng 2017; 04016170.
[34] Jin L, Du M, Li D, Du X, Xu H. Effects of cross section size and transverse rebar on
the behavior of short squared RC columns under axial compression. Eng Struct
References 2017;142:223–39.
[35] Wriggers P, Moftah SO. Mesoscale models for concrete: homogenisation and da-
mage behaviour. Finite Elem Anal Des 2006;42(7):623–36.
[1] Carpinteri A. Decrease of apparent tensile and bending strength with specimen size:
[36] Jirásek M, Zimmermann T. Embedded crack model: I. Basic formulation. Int J
two different explanations based on fracture mechanics. Int J Solids Struct
Numer Method Eng 2015;50(6):1269–90.
1989;25(4):407–29.
[37] Jirásek M, Zimmermann T. Embedded crack model. Part II: combination with
[2] Syroka-Korol E, Tejchman J, Mróz Z. FE calculations of a deterministic and statis-
smeared cracks. Int J Numer Method Eng 2015;50(6):1291–305.
tical size effect in concrete under bending within stochastic elasto-plasticity and
[38] Bazant ZP, Jirasek M. Nonlocal integral formulations of plasticity and damage:
non-local softening. Eng Struct 2013;48:205–19.
Survey of progress. ASCE J Eng Mech 2002;128(11):1119–49.
[3] Bažant ZP. Size effect in blunt fracture: concrete, rock, metal. ASCE J Eng Mech
[39] Gilles P, Khalil H. Non-local damage model with evolving internal length. Int J
1984;110(4):518–35.
Numer Anal Method Geomech 2004;28(7/8):633–52.
[4] Bažant ZP, Planas J. Fracture and size effect in concrete and other quasi-brittle
[40] Majewski T, Bobinski J, Tejchman J. FE analysis of failure behaviour of reinforced
materials. NewYork: CRC Press; 1998.
concrete columns under eccentric compression. Steel Constr 2008;30(2):300–17.
[5] Hoover CG, Bažant ZP. Comprehensive concrete fracture tests: size effects of types 1
[41] Skarżyński L, Syroka E, Tejchman J. Measurements and calculations of the width of
& 2, crack length effect and post peak. Eng Fract Mech 2013;110:281–9.
the fracture process zones on the surface of notched concrete beams. Strain
[6] Bažant ZP, Kwon YW. Failure of slender and stocky reinforced concrete columns:
2011;47:319–32.
tests of size effect. Mater Struct 1994;27(2):79–90.
[42] Syroka-Korol E, Tejchman J, Mróz Z. FE analysis of size effects in reinforced con-
[7] Şener S, Barr BIG, Abusiaf HF. Size effect in axially loaded reinforced concrete
crete beams without shear reinforcement based on stochastic elasto-plasticity with
columns. ASCE J Struct Eng 2004;130(4):662–70.
non-local softening. Finite Elem Anal Des 2014;88:25–41.
[8] Li D, Jin L, Du XL, Fu J. Size effect tests of normal-strength and high-strength RC
[43] Syroka-Korol E, Tejchman J, Mróz Z. FE investigations of the effect of fluctuating
columns subjected to axial compressive loading. Eng Struct 2016;109:43–60.
local tensile strength on coupled energetic–statistical size effect in concrete beams.
[9] Bažant ZP, Kim JK. Size effect in shear failure of longitudinally reinforced beams.
Eng Struct 2015;103:239–59.
ACI J Proceed 1984;81(5):456–68.
[44] Du X, Jin L, Ma G. A meso-scale analysis method for the simulation of nonlinear
[10] Zararis PD, Papadakis GC. Diagonal shear failure and size effect in RC beams
damage and failure behavior of reinforced concrete members. Int J Damage Mech
without web reinforcement. ASCE J Struct Eng 2001;127(7):733–42.
2013;22(6):878–904.
[11] Syroka-Korol E, Tejchman J. Experimental investigations of size effect in reinforced
[45] Yılmaz O, Molinari JF. A mesoscale fracture model for concrete. Cem Concr Res
concrete beams failing by shear. Eng Struct 2014;58:63–78.
2017;97:84–94.
[12] Jin L, Du XL, Li D, Su X. Seismic behavior of RC cantilever beams under low cyclic
[46] Ren W, Yang Z, Sharma R, et al. Two-dimensional X-ray CT image based meso-scale
loading and size effect on shear strength: an experimental characterization. Eng
fracture modelling of concrete. Eng Fract Mech 2015;133:24–39.
Struct 2016;122:93–107.
[47] Wang X, Yang Z, Jivkov AP. Monte Carlo simulations of mesoscale fracture of
[13] Barbhuiya S, Choudhury AM. A study on the size effect of RC beam-column con-
concrete with random aggregates and pores: a size effect study. Constr Build Mater
nections under cyclic loading. Eng Struct 2015;95:1–7.
2015;80:262–72.
[14] Bažant ZP, Xiang Y. Size effect in compression fracture: splitting crack band pro-
[48] Skarżyński Ł, Nitka M, Tejchman J. Modelling of concrete fracture at aggregate
pagation. ASCE J Eng Mech 1997;123(2):162–72.
level using FEM and DEM based on X-ray CT images of internal structure. Eng Fract
[15] Weiss WJ, Guler K, Shah SP. Localization and size-dependent response of reinforced
Mech 2015;147:13–35.
concrete beams. ACI Struct J 2001;98(5):686–95.
[49] Ministry of Construction of the People's Republic of China GB 50010–2010 Code for
[16] Alca N, Alexander SDB, MacGregor JG. Effect of size on flexural behavior of high-
design of concrete structures. Beijing: China Architecture & Building Press 2010.
strength concrete beams. ACI Struct J 1997;94(1):59–67.
[50] Kim SM, Al-Rub RKA. Meso-scale computational modeling of the plastic-damage
[17] Che Y, Zhang XF, Wang JJ. Size effect on flexural behavior of reinforced high
response of cementitious composites. Cem Concr Res 2011;41(3):339–58.
strength concrete beams subjected to monotonic loading. J Build Struct
[51] Unger JF, Eckardt S. Multiscale modeling of concrete. Arch Comput Meth Eng
2012;33(6):96–102. [in Chinese].
2011;18(3):341–93.
[18] Che Y, Zhang XF, Wang JJ, Song YP. Experimental study on size effect of flexural
[52] Lubliner J, Oliver J, Oller S, Oñate E. A plastic-damage model for concrete. Int J
behavior of reinforced high-strength concrete beams subjected to cyclic loading. J
Solids Struct 1989;25(3):299–326.
Build Struct 2013;34(8):100–6. [in Chinese].
[53] Lee J, Fenves GL. Plastic-damage model for cyclic loading of concrete structures.
[19] Rao GA, Vijayanand I, Eligehausen R. Studies on ductility and evaluation of
ASCE J Eng Mech 1998;124(8):892–900.
minimum flexural reinforcement in RC beams. Mater Struct 2008;41(4):759–71.
[54] Huang YJ, Yang ZJ, Chen XW, et al. Monte Carlo simulations of meso-scale dynamic
[20] Bosco C, Carpinteri A, Debernardi PG. Minimum reinforcement in high-strength
compressive behavior of concrete based on X-ray computed tomography images. Int
concrete. ASCE J Struct Eng 1990;116(2):427–37.
J Impact Eng 2016;97:102–15.
[21] Yi ST, Kim MS, Kim JK, et al. Effect of specimen size on flexural compressive
[55] Grote DL, Park SW, Zhou M. Dynamic behavior of concrete at high strain rates and
strength of reinforced concrete members. Cem Concr Comp 2007;29(3):230–40.
pressures: I. experimental characterization. Int J Impact Eng 2001;25(9):869–86.
[22] Kim JHJ, Yi ST, Kim JK. Size effect of concrete members applied with flexural
[56] Park SW, Xia Q, Zhou M. Dynamic behavior of concrete at high strain rates and
compressive stresses. Int J Fract 2004;126(1):79–102.
pressures: II. numerical simulation. Int J Impact Eng 2001;25(9):887–910.
[23] Yi ST, Kim MS, Kim JK. Effect of specimen sizes on flexural compressive strength of
[57] Garboczi EJ, Bentz DP. Digital simulation of the aggregate–cement paste interfacial
concrete. ACI Struct J 2001;98(3):416–24.
zone in concrete. J Mater Res 1991;6(1):196–201.
[24] Moretti M, Tassios TP. Behaviour of short columns subjected to cyclic shear dis-
[58] Majewski T, Bobinski J, Tejchman J. FE analysis of failure behaviour of reinforced
placements: experimental results. Eng Struct 2007;29(8):2018–29.
concrete columns under eccentric compression. Eng Struct 2008;30(2):300–17.
[25] Mahal M, Taljsten B, Blanksvard T. Experimental performance of RC beams
[59] Ollivier JP, Maso JC, Bourdette B. Interfacial transition zone in concrete. Advn Cem
strengthened with FRP materials under monotonic and fatigue loads. Constr Build
Bas Mat 1995;2:30–8.
Mater 2016;122:126–39.
[60] Grondin F, Matallah M. How to consider the Interfacial Transition Zones in the
[26] Dadi VVSSK, Agarwal P. Effect of types of reinforcement on plastic hinge rotation
finite element modelling of concrete? Cem Concr Res 2014;58:67–75.
parameters of RC beams under pushover and cyclic loading. Earthq Eng Eng Vibrat
[61] Du XL, Jin L, Ma GW. Numerical modeling tensile failure behavior of concrete at
2015;14(3):503–16.
meso-scale using extended finite element method. Int J Damage Mech
[27] Jin L, Li D, Du X. Mechanical behavior and size effect of moderate high-strength RC
2014;23(7):872–98.
columns under monotonic and cyclic axial compression. Eng Struct
[62] Jin L, Ding ZX, Li D, Du XL. Experimental and numerical investigations on the size
2016;124:269–85.
effect of moderate high-strength reinforced concrete columns under small-eccentric
[28] Lee MK, Barr BIG. An overview of the fatigue behavior of plain and fibre reinforced
compression. Int J Damage Mech 2017. doi: 1056789517699054.
concrete. Cem Concr Compos 2004;26:299–305.
[29] Rodrigues H, Arêde A, Varum H, Costa AG. Experimental evaluation of rectangular

586