You are on page 1of 17

Finite Elements in Analysis and Design 88 (2014) 2541

Contents lists available at ScienceDirect

Finite Elements in Analysis and Design


journal homepage: www.elsevier.com/locate/finel

FE analysis of size effects in reinforced concrete beams without


shear reinforcement based on stochastic elasto-plasticity
with non-local softening
E. Syroka-Korol a, J. Tejchman a,n, Z. Mrz b
a
b

Faculty of Civil and Environmental Engineering, Gdask University of Technology, Poland


Institute of Fundamental Technological Research, Polish Academy of Sciences, Warsaw, Poland

art ic l e i nf o

a b s t r a c t

Article history:
Received 29 July 2013
Received in revised form
30 April 2014
Accepted 12 May 2014

The paper presents results of FE analysis of mechanical size effects in longitudinally reinforced concrete
slender beams without shear reinforcement failing in shear mode. The simulations were performed
under plane stress conditions for three beams of different sizes and a xed shape (height/length ratio).
The attention was focused on deterministic and statistical size effects related to the nominal beam shear
strength. Concrete was assumed as an isotropic elasto-plastic material exhibiting non-local softening.
The bond strength between concrete and reinforcement was assumed to depend on interface slip with
both stable and softening responses. Statistical simulations were performed for spatially correlated
Gaussian random elds of tensile strength using a stratied sampling reduction method. The FE
numerical results were compared with the respective own experimental test results.
& 2014 Elsevier B.V. All rights reserved.

Keywords:
Elasto-plasticity
Non-local softening
Random elds
Reinforced concrete beams
Size effect
Strain localization

1. Introduction
The size effect phenomenon in quasi-brittle structures is
related to a transition from a ductile behaviour of small specimens
to a totally brittle response of large ones. Thus, the nominal
strength sN decreases with increasing characteristic specimen
dimension D. The reasons for this behaviour are: (a) intense strain
localization regions with a certain volume (i.e. micro-cracked
damage regions called also fracture process zones, FPZ) which
precede discrete macro-cracks; their size related to D contributes
to a deterministic size effect and (b) a spatial variability/randomness of local material properties contributing to a statistical size
effect that becomes dominant with increasing D.
A strong size effect also occurs in reinforced concrete beams
without shear reinforcement wherein diagonal sheartensile fracture takes place in concrete. It was experimentally observed
among others by Leonhardt and Walther [1], Kani [2,3], Bhal [4],
Taylor [5], Walraven [6], Chana [7], Iguro et al. [8], Bazant and
Kazemi [9], Shioya et al. [10], Kim and Park [11], Grimm [12],
Ghannoum [13], Kawano and Watanabe [14], PodgorniakStanik [15], Yoshida [16], Angelakos et al. [17], Lubell et al. [18]
and Syroka-Korol [19]. The diagonal cracks at failure had in

Corresponding author.
E-mail addresses: esyroka@pg.gda.pl (E. Syroka-Korol),
tejchmk@pg.gda.pl (J. Tejchman), zmroz@ippt.gov.pl (Z. Mrz).
http://dx.doi.org/10.1016/j.nel.2014.05.005
0168-874X/& 2014 Elsevier B.V. All rights reserved.

experimental tests essentially similar paths and relative lengths


at the maximum load independently of the beam size. Therefore, this
size effect in such reinforced concrete beams could be described by
the analytical deterministic (energetic) size effect law (SEL) of Type II
according to Bazant [20], being valid for structures of a positive
similar geometry possessing large stress-free cracks that grow in a
stable manner up to the maximum load (Fig. 1)
o
N D p;
1
1 D=D0
Where N is the nominal strength, and o and D0 are the empirical
parameters depending on material properties, structure geometry
and structure shape [21]. They can be determined by tting Eq. (1) to
the experimental data. The parameter D0 separates the ductile failure
(D0D) from the brittle one (D0D). For very large structures (D-1),
the nominal strength approaches N-D  1/2. Assuming the residual
strength R for very large sizes (D-1) due to the strength of
reinforcement and compressed concrete, Eq. (1) becomes valid if N
is replaced by the expression N  R [21]. For small structures (D-0),
the size effect disappears. Thus, the size effect is strong only in the
limited size range. The SEL curve (Fig. 1b) in a double-logarithmic
plot represents a smooth transition from a strength (plastic) limit for
small sizes to the solution given by the Linear Elastic Fracture
Mechanics (LEFM) for large and very large sizes.
In spite of the ample experimental evidence, the physically based
size effect is not taken into account in practical design rules of
engineering structures, assuring a specied safety factor with respect

26

E. Syroka-Korol et al. / Finite Elements in Analysis and Design 88 (2014) 2541

Fig. 1. Size effect curve of type II with large cracks or notches by Bazant [20,22]
(N nominal strength, D characteristic specimen size, LEFM linear elastic
fracture mechanics); (a) linear scale and (b) loglog scale.

to the failure load [22,23]. Instead, a purely empirical approach is


sometimes considered in building codes which is doomed to yield an
incorrect formula since physical foundations are lacking.
Our objective is to provide a quantitative assessment of a size
effect and a related description of a brittle failure mode in slender
reinforced concrete beams without shear reinforcement under
bending. The nite element method based on an isotropic elastoplastic model with non-local softening enhanced by a characteristic length parameter of micro-structure was used in numerical
studies. The plane stress 2D calculations were performed. Material
parameters were calibrated with conventional laboratory tests and
code recommendations. A characteristic length of microstructure
was estimated by means of displacement measurements on the
beam surface using the non-invasive Digital Image Correlation
(DIC) technique [24]. Deterministic calculations were performed assuming a constant value of the tensile strength. In turn,
statistical analyses were carried out with spatially correlated
random elds according to the Gauss distribution reecting the
random nature of a local tensile strength. In order to reduce the
number of FE statistical simulations, a stratied sampling scheme
was used belonging to a group of variance reduced Monte Carlo
methods. This approach enabled us a signicant reduction of the
sample number without affecting the accuracy of calculations. The
present analysis constitutes the continuation of our earlier successful simulations of a combined deterministic-statistical size
effect in notched [25] and unnotched concrete beams [26].
The numerical results were compared with our laboratory experiments [19,27]. The experiments were carried out on

longitudinally reinforced concrete beams of different sizes and


xed height/length ratio: 3 small-size beams of the height of
200 mm and length 1500 mm, 3 medium-size beams of the height
of 400 mm and length 3000 mm and 3 large-size beams of the
height of 800 mm and length 6000 mm (the thickness
t200 mm). The beams were geometrically similar in 2 dimensions to avoid differences in the hydration heat effects which are
proportional to the thickness of the member [22]. They and made
from the same concrete mix with the mean aggregate diameter
equal to 9 mm. The simply supported beams were subjected to
4-point bending with the constant shear span-effective height
ratio equal to 3. To induce a sheartension failure mechanism in
concrete, the beams were over-reinforced without shear reinforcement (the reinforcement ratio was always 1%). The experimental results showed a signicant size effect on the nominal
shear strength versus the beam effective height. The mean
nominal shear strength of large-size beams was smaller by 40%
with respect to small-size beams. In all RC beams, a combined
diagonal sheartensile (signicantly more tensile) and bond failure mode dominated, characterized by the development of a
critical diagonal sheartensile crack connected with a horizontal
splitting crack along the top of the bottom longitudinal reinforcement toward the beam support (a shearcompression failure
mechanism did not occur in concrete). The failure mode proceeded
in a brittle manner in the post-critical stage.
Numerical FE analyses of slender beams without transverse
reinforcement were performed among others by Kotsovos and
Pavlovic [28,29], Vecchio and Swim [30], Sato et al. [31] and Slobbe
et al. [32]. They used a non-linear elastic-brittle model ([28,29]),
a smeared rotating crack approach [30], a smeared xed crack
approach with a sequentially linear (SL) analysis [32] and a
discrete rotating crack model [31]. In calculations, a diagonal
tensile brittle failure mode was usually obtained. A critical diagonal crack propagated towards the beam top, if the beam failure
was caused by concrete splitting in a compression zone [32].
It propagated towards the bottom, if the beam failure was caused
by bond splitting [31]. A deterministic size effect on the nominal
shear strength of beams failing by diagonal tension was studied
only by Sato et al. [31] for four virtual reinforced concrete beams
with the height ranging from 100 mm up to 1600 mm. The
numerical results were overestimated as compared to an analytical
size effect formula. According to 3D simulation results in [28,29],
the size effect in slender reinforced concrete beams without shear
reinforcement is mainly caused by non-symmetric cracking combined with the unintended out-of-plane action, the latter being
impossible to be avoided in experiments. They have also found out
that stirrups eliminate the size effect in reinforced concrete beams
(in contrast to recent outcomes by Yu and Bazant [33]).
Summarized, the novel elements in our calculations for reinforced concrete beams failing in shear are: (a) combined
deterministic-statistical FE calculations for 3 different beams by
taking strain localization and bond into account, (b) direct comparison between numerical and experimental results and (c)
application of a stratied sampling scheme to reduce the number
of statistical calculations. To our knowledge, such calculations have
so far not been performed.

2. Constitutive models
2.1. Concrete
The concrete deformational response was simulated by assuming an isotropic elasto-plastic constitutive model with a nonlocal softening which was used in our previous calculations
[19,26,3436]. This relatively simple isotropic model for concrete

E. Syroka-Korol et al. / Finite Elements in Analysis and Design 88 (2014) 2541

Fig. 2. Assumed hardening/softening curves in FE-calculations: (a) tensile stress


versus non-local parameter st f(1) and (b) compressive stress versus non-local
parameter sc f(2).

(see Appendix) consists of two yield criterions: by Rankine in


tension (Eq. (A1)) and by DruckerPrager in compression (Eq.
(A2)). The softening under tension (Fig. 2a) was characterized by
the exponential curve by Hordijk [37] (Eqs. (A7) and (A8)). In
compression, linear hardening/softening was assumed (Fig. 2b).
The concrete model requires two elastic constants: modulus of
elasticity E and Poisson's ratio , two plastic constants: internal
friction angle and dilatancy angle , one tensile yield function st
f(1) and one compressive yield function sc f(2). The disadvantages
of the model are the following: the shape of the compressive failure
surface in a principal stress space is linear (not paraboloidal as in
reality). In deviatoric planes, the shape is circular (in compression
states) and triangular (in tension states); thus it does not gradually
change from a curvilinear triangle with smoothly rounded corners to
nearly circular with increasing pressure. The effect of third stress
deviator invariant is not taken into account. The strength is similar
for triaxial compression and extension, and the stiffness degradation
due to strain localization and non-linear volume changes during
loading are not taken into account.
A non-local theory was used as a regularization technique
[4042]. In this approach, the principle of a local action does not
take place any more. Polizzotto et al. [43] laid down a thermodynamically consistent formulation of non-local plasticity. In the
calculations, the softening parameters i (i 1, 2) were assumed to
be non-local (independently for both yield surfaces fi) [44]
R
x  i d
with i 1; 2;
2
i x 1  mi x m VR
V x  d

27

Fig. 3. Bond stress-slip relationship (shear stress against slip displacement )


between concrete and reinforcement according to Do
rr [51] (a) and CEB-FIP Code
[39] (b).

Fig. 4. Bond relationship: radial normal stress sr,rs and bond slip versus radial
normal strain r,rs between concrete and reinforcement during splitting failure [51].

where i (x) are the non-local softening parameters, V denotes the


body volume, x is the coordinate vector of the considered point,
is the coordinate vector of the surrounding points, denotes the
weighting function and m is the additional non-local parameter
controlling the size of the localized plastic zone. As a weighting

28

E. Syroka-Korol et al. / Finite Elements in Analysis and Design 88 (2014) 2541

function , the Gauss distribution function was used


2
1
r pe  r=lc
lc

where lc is a characteristic length of micro-structure and the


parameter r is the distance between material points. The averaging
in Eq. (3) was restricted to a small representative area around each
material point (the inuence of points at the distance of r 3lc
was only of 0.01%). The softening non-local parameters near
boundaries and at both sides of a localized zone were always
calculated on the basis of Eqs. (1) and (2) (which satisfy the
normalizing condition) [36]. Other approaches were proposed by
Polizzotto [45] and Jirsek et al. [46]. The reinforcement was not
taken into account when calculating a non-local parameter in
concrete [36]. To simplify the calculations, non-local rates were
replaced by their approximations calculated with known total
strain increments [42].
The non-local model was implemented in the commercial nite
element code ABAQUS [47] with the aid of subroutine UMAT (user
constitutive law denition) and UEL (user element denition) for
efcient computations [42]. The calculations were carried out
using a large-displacement analysis available in the ABAQUS
code [47]. According to this method, the current conguration of
the body was taken into account. The Jaumann rate of the Cauchy
stress was taken. The conjugate strain rate was the rate of
deformation. The rotation of the stress and strain tensor was
calculated with the HughesWinget method [48]. It is known that
the stress and strain rate are not work conjugate in this formulation and erroneous results occur, in particular for highly compressible materials [49]. Our calculations showed however that the
effect of large displacements on the failure force in reinforced
beams was negligible (less than 1%). Thus the effect of a work
conjugacy error was not important. The non-local averaging was
performed in a current conguration. This choice was governed by
the fact that element areas in this conguration were automatically calculated by ABAQUS [47]. The edge and vertex in the
Rankine yield function were taken into account by the interpolation of 23 plastic multipliers according to Koiter's rule [50]. The
same procedure was adopted in the case of combined tension
(Rankine criterion) and compression (DruckerPrager criterion).
2.2. Reinforcement
To simulate the reinforcement behaviour, an elastic-perfectly
plastic material with the von Misses criterion was assumed
f s sij q f y ;

where fy is the yield stress for steel bars.


2.3. Bond between concrete and reinforcement
To describe the interaction between concrete and reinforcement, a bond relationship was dened. In general, two different
bond-failure mechanisms may appear connected to a pull-out or
splitting mode. The pullout bond failure takes place when the bar
anchorage length is insufcient to carry tensile stresses in the

longitudinal reinforcement, whereas the splitting bond failure


occurs when the concrete cover thickness is insufcient to resist
radial cracks caused by local forces transmitted through bar ribs.
To consider bond model, an interface with a zero thickness was
assumed along a contact surface, Since the bond model was crucial
for describing numerically the experimental failure mechanism,
three different bond stressslip denitions were tested. Initially,
the simple and well-known models proposed by Dorr [51] and
CEB-FIP Model Code [39] were used which described the pull-out
failure by a relationship between the bond shear stress and slip
displacement . The rst bond-slip model neglects softening and
assumes a yield plateau when the pullout failure begins (Fig. 3a).
The model needs solely 2 parameters: the tensile strength ft and
slip displacement u at which perfect slip occurs (usually
u 0.06 mm). In turn, the model in CEB-FIP Model Code [39]
assumes softening and residual yield (Fig. 3b). The 6 model
parameters 1, 2, 3, , max and f depend on the concrete
connement and bond conditions.
In order to describe the splitting bond failure along reinforcement, the model by Akkerman [52] was also used, being a
modication of the original model proposed by den Uijl and Bigaj
[53]. The model takes into account the evolution of the radial
stress sr,rs versus the radial strain r,rs and is divided into 3 phases
(Fig. 4). The rst phase (0 r r,rs rr,rs,max) characterizes a nonlinear material behaviour caused by cracks, the second one
(r,rs,max or,rs rr,rs,res) includes linear softening and the third
one (r,rs,res or,rs) described the residual behaviour
8
k  2
>
s
>
>
< r;rs; max 1 k  2

sr;rs r;rs sr;rs; max 1  1  b r;rs  r;rs: max
r;rs  r;rs: max
>
>
>
: sr;rs;res

0 r r;rs r r;rs: max


r;rs: max o r;rs r r;rs:res
r;rs:res o r;rs

5
with
k

Er r;rs; max

sr;rs; max

and

r;rs
:
r;rs; max

The empirical parameter b determines the ratio between the


maximum and residual stress (b sr,rs,res/sr,rs,max); it was assumed
as b 0.2. The initial radial stiffness (Eq. (6)) is
!1
cef f 0:5b 2 0:252b

;
7
Er E
cef f 0:5b 2  0:252b
where b is the bar diameter, ceff is the effective concrete cover and
E and are the modulus of elasticity of concrete and Poisson's
ratio, respectively.
The limit radial normal stresses and strains were dened as
follows:




cef f 0:88
cef f 1:08
f
sr;rs; max 2f t
; r;rs; max 4:2 t
8
b
E
b

sr;rs;res b sr;rs; max ;

r;rs;res

ft
E


2cef f c0
;

b
b

where c0 is the empirical parameter inuencing softening in the


second phase (r,rs,max or,rs rr,rs,res). The radial strains were

Fig. 5. Random eld areas included in calculations of sampling parameter.

E. Syroka-Korol et al. / Finite Elements in Analysis and Design 88 (2014) 2541

29

related to the radial normal stress sr,rs by the ctitious friction rule

computed as follows:

tan b ; ;
r:rs
0:5b

10

where b is the cone angle between the bar axis and cone-shaped
cracking surface starting from ribs. The ctitious value was
assumed as b[o] 0.1fc [MPa]. In turn, the bond stress was

sr;rs cot b

11

with cot b 1. The model needs 8 parameters: 3 depending on


concrete properties E, and ft, 2 depending on the specimen and
reinforcement geometry ceff and b and 3 empirical ones b, c0,
and b.
3. Random elds in statistical calculations
In the rst step, the material tensile strength ft was the sole
random parameter in our statistical analyses. The spatially correlated random elds of the tensile strength were described by a
Gauss distribution function and the homogenous squared exponential auto-covariance correlation function C
!
x1  x2 2
Cx1 ; x2 exp 
;
12
2
lcor
where x1 and x2 are the co-ordinate points and l is the correlation
length. The auto-covariance function had the following spectral
decomposition:
1

Cx1 ; x2 i f i x1 f i x2 ;

13

i1

Fig. 6. Stratied sampling scheme used in statistical FE analyses (cdf cumulative


probability function, sampling parameter, Pi probability intervals).

Table 1
Geometry of reinforced concrete slender beams of Figs. 7 and 8.
Dimension Small-size beam
SL20

Medium-size beam
SL40

Large-size beam
SL80

L [mm]
Leff [mm]
H [mm]
D [mm]
a [mm]
b [mm]

3000
2700
400
360
1080
540

6000
5620
800
750
2250
1120

1500
1200
200
160
480
240

where the eigenvalues i and eigenfunctions fi(x) were the solution


of the Fredholm integral equation of the second kind
Z
Cx1 ; x2 f i x1 dx1 i f i x2
14
D

The spatially correlated random elds H(x,) (here with the zero
mean and unit variance) were dened according to the Karhunen
Love expansion [54,55] by an innite linear combination of
orthogonal functions with random coefcients
1 p
Hx;
15
i i f i x;
i1

where i() is the vector of uncorrelated random variables sampled


^
from N(0,1) distribution. The approximated solution Hx;
was
obtained by truncating the series in Eq. (17) after M terms. The

Fig. 7. Cross-section of slender reinforced concrete beams with horizontal bars used in calculations and experiments: (a) small-size SL20, (b) medium-size SL40 and
(c) large-size SL80.

30

E. Syroka-Korol et al. / Finite Elements in Analysis and Design 88 (2014) 2541

Fig. 8. Geometry and loading scheme of reinforced concrete slender beams without stirrups.

Fig. 9. Experimental crack pattern on front (solid lines) and back (dashed lines) side of medium-size reinforced concrete beam.

Fig. 10. FE mesh for small-size beam SL20 (a) and large-size beam SL80 (b) (note that beams are not proportionally scaled).

Haar wavelets was created from a so called mother function


8
x A 0; 0:5
>
<1
x A 0:5; 1
x  1
16
>
:
0
otherwise
by scaling and shifting it according to formula
j;k x j 2j x k;

j; k A N;

17

where j and k are the positive integer constants responsible for


scaling and shifting and j is the function amplitude. In the current
study j 1, which forms an orthogonal basis. The extended set of
i, which creates the complete set of orthogonal functions over the
domain [0, 1], was dened after [58] as
0 x 1
Fig. 11. Calculated vertical force P versus deection u with various bond denitions
compared with experiments (small-size reinforced concrete beam SL20).

i x 2j x  k;

18

where i 2j k; j 0, 1,m  1, k 0, 1,,2j 1 (m the wavelet


level). Each eigenfunction was approximated by a truncated series
of the N Haar wavelets
N 1

resulting error was minimized by sorting eigenvalues in a decreasing order and controlling the sum of i for i 1,, M.
An analytical solution of Eq. (17) is available with e.g. an
exponential autocorrelation function (Ghanen and Spanos [56]).
In other cases, Eq. (17) has to be solved numerically. In our study,
the wavelet-Galerkin method proposed by Phoon et al [57,58] was
used wherein conventional bases (e.g. trigonometric or polynomials) were replaced by Haar wavelets. A family of the orthogonal

and

f k x di i x T xDk ;
k

19

i0

where N 2m and dki are the wavelet coefcients.


In order to reduce the number of statistical simulations, the
so-called stratied sampling scheme originally proposed by
Neyman [59] was used [60,61]. Initially 2000 random elds of
the local tensile strength were generated for each beam size. Next,
the generated samples were classied according to the sampling

E. Syroka-Korol et al. / Finite Elements in Analysis and Design 88 (2014) 2541

31

Fig. 12. Contours of non-local tensile softening parameter 1 in small-size reinforced concrete beam SL20 at failure with various bond-slip models: (a) perfect bond, (b) by
Do rr [51], (c) by CEB-FIP [39] and (d) by Akkermann [52] from FE analyses compared to experimental crack pattern (single lines).

parameter, chosen as the mean tensile strength in two shear


regions of beams (Fig. 5) according to their failure mode. Thus,
we assumed that the effect of the beam mid-region on the
statistical failure force was negligible (the effect of the sampling
region size on results merits further investigations). For the left
and right shear region of Fig. 5, the mean value was separately
calculated and afterwards the lower value was used. Next, the
random elds were classied and arranged in an increasing order
according to the calculated mean value. Based on the sampling
parameter, the cumulative probability of an initially generated set
of the random elds was calculated (Fig. 6). Next, the cumulative
probability function was divided into a nite number n of the
equal intervals (Pi, i1,,n), where n corresponded to the number
of samples. From each subset of samples included in the intervals
of the equal probability, only one sample (nearest to the midpoint) was chosen for further FE analyses. Thus, the chosen
representative set of n samples included random elds of the
mean value in a shear region from the lowest to the highest one.
We used the sampling procedure after the random elds were
generated instead of sampling of on random vectors (Eq. (15))
during a random eld generation [62] because of the inuence of
the input random vector on a random distribution. The sampling
on output random elds enabled us to choose the samples in a
critical localized zone of the beam shear region, being approximately representative for the beam failure force. Our method
allows for a fast convergence to the mean failure force [60,61].

4. FE input data
The FE-analyses of longitudinally reinforced concrete slender
beams without shear reinforcement were performed with 3 different beam sizes of a similar geometry from experiments by Syroka
Korol [19]. The beams had the same dimensions H  L as in
the experiments: 200  1500 mm2 (small-size beam SL20),

400  3000 mm2 (medium-size beam SL40) and 800  6000 mm2
(large-size beam SL80) (Table 1). All specimens had the constant
thickness t200 mm and constant reinforcement percentage
1% (Fig. 7). The beams were subjected to 4-point bending
(Fig. 8) under the constant shear span to the effective depth ratio
a/D3. The numerical calculations were carried out under plane
stress conditions since experimental crack patterns were very
similar on both sides of all beams (Fig. 9 shows these patterns in
a medium-size beam).
The specimens were cast from concrete C35/45 of a maximum
aggregate size 32 mm (the characteristic compressive strength
fck 35 MPa, the characteristic tensile strength fctk 2.2 MPa,
Young modulus E 34 GPa and Poisson's ratio 0.2). The internal
friction angle was 141 with rs
bc 1.2 (Eq. (A2) and (A4)) and the
dilatancy angle 81 [34]. The calculations were carried out with
the characteristic length lc 5 mm based on the experimentally
measured mean width of localized tensile zones wloc by means of
the DIC technique equal to 20 mm. The non-locality parameter
was m 2 [42]. The ultimate non-local softening parameter in
tension was mainly u1 0.0207 (Gf gf4lc E 180 N/m) and in compression u2 0.0057 (Gc gc4lc E2700 N/m) based on initial calculations. The steel behaviour was specied by the yield stress
fy 500 MPa, the elastic modulus Es 200 GPa and the Poisson's
ratio s 0.3.
The bar diameters were b 10 mm, 16 mm and 20 mm in
the beams SL20, SL40 and SL80, respectively, and b 10 mm for
the second reinforcement layer in the large-size beam SL80. In the
model by Do rr [51] Eq. (7), the slip displacement was u 0.06 mm
(as originally proposed by Do
rr [51]). When using the bond model
by CEB-FIP [39], for the unconned concrete and good bond
conditions case, the material parameters were: 1 0.6 mm,
2 0.6 mm, 3 1 mm, 0.4, max 2.0fck 11.83 MPa and
f 0.15max 1.77 MPa. In the bond model by Akkermann [52]
(Eqs. (5)(11)), we assumed ceff 22 mm, b 0.2 [39] and
c0 0.18 mm (based on own initial FE simulations).

32

E. Syroka-Korol et al. / Finite Elements in Analysis and Design 88 (2014) 2541

Fig. 13. Evolution of non-local tensile softening parameter 1 in small-size reinforced concrete beam SL20 with bond-slip denition by Akkermann [52] from FE analysis at
different vertical resultant force: (a) P 30 kN, (b) P 50 kN, (c) P 85 kN, (d) Pmax 91.4 kN and experimental crack pattern after failure [19] (e). (For interpretation of the
references to color in this gure, the reader is referred to the web version of this article.)

FE meshes of quadrilateral elements composed of four diagonally crossed triangles were used (Fig. 10). The nite elements
had mainly the area of 7.5  10 mm2 (width  height) which was
equal to (1.52)  lc to obtain mesh-independent results [24,26].
In the case of the large-size beam, the ne mesh covered the beam
mid-part of 4410 mm only. Totally, 130 6001690 344 nite elements
were used depending upon the beam size. The steel bars were
modelled as 2D elements with the width of 1.5 mm and height of
2.02.5 mm depending on the bar diameter b. Since our FE
analyses were two-dimensional, the bar height was taken as
0.5  b to obtain a similar contact surface as in experiments
(2  0.5b). The inuence of the mesh size on FE results was
checked by comparing 2 different meshes in the small-size
reinforced concrete beam SL20 assuming the bond model by
Akkermann [52].

5. Effect of bond-slip model


Fig. 11 shows the calculated load-displacement diagrams with
different bond models of Section 3 for the small size-beam SL20,
and Figs. 12 and 13 the calculated evolution of the non-local
tensile softening parameter 1 .
Using the model by Akkermann [52] only, a satisfactory
agreement was achieved with experiments (Figs. 12 and 13). The
ultimate vertical force from deterministic simulations was 91.4 kN
which was by 1% lower only than the average failure force from

experiments (91.9 kN). With the remaining bond models, the


maximum vertical force was too high due to the different main
failure mechanism assumed that was not consistent with the
experimental failure mode (steel yielding instead of diagonal
tension). The number of localized zones varied from 5 with the
model by Do
rr (1980) [51] and by CEB-FIP [39] (Fig. 13c) up to
8 with the perfect bond (Fig. 12a). Thus, the pattern of localized
zones differed from the experimental one.
The evolution of localized zones was similar to the experimentally observed cracks with the bond model by Akkermann [52]
(Fig. 13). First, vertical tensile bending localized zones appeared
at the mid-region, next inclined shear localized zones developed
in the shear span region, and nally a diagonal sheartensile
zone occurred connected to the splitting failure along reinforcement, where bond radial stresses sr,rs reached their residual
value (marked as dashed line in Fig. 13e). A critical inclined
localized zone occurred at the distance of dc 270 mm from
the support with the ratio of dc/a 0.56 being identical as the
average experimental value from 3 tests. In total, 9 localized
zones were calculated whose average spacing was 87 mm
(the average experimental crack spacing was also similar
 91 mm).
We investigated also the effect of plastic parameters , , Gf and
Gc on FE results. The effect of (with rs
bc 1.151.25) and (with
812o) was negligible. The lower Gf, less localized zones appeared
and a critical diagonal localized zone occurred at a higher distance
from the support. In turn, the effect of Gc on a localized zones

E. Syroka-Korol et al. / Finite Elements in Analysis and Design 88 (2014) 2541

33

Fig. 14. Comparative deterministic FE results with coarse (element size 15  20 mm2) and ne mesh (element size 7.5  10 mm2) for small-size reinforced concrete beam
SL20: (A) load-deection diagram and (B) contours of non-local tensile softening parameter 1 at failure (a) coarse mesh and (b) ne mesh.

Fig. 15. Evolution of diagonal sheartensile localized zone with marked points: (a) at P 76 kN, (b) at P 86 kN, (c) at peak and (d) at failure (small-size reinforced concrete beam).

pattern and load capacity was negligible. The assumed softening


curves of Fig. 2 with Gf 180 N/m and Gc 2700 N/m provided the
most realistic FE results with respect to experiments.
The effect of the mesh size on the results is demonstrated in
Fig. 14. The results show that the FE results with a ne and a coarse

mesh are similar (Fig. 14A and B). The failure force for the coarse
mesh (element size 15  20 mm2) was 93 kN while for the ne
mesh (element size 7.5  10 mm2) was 91 kN (Fig. 14A). Based on
these results, we assumed that there was no need to decrease
more the size of nite elements (less than 7.5  10 mm2).

34

E. Syroka-Korol et al. / Finite Elements in Analysis and Design 88 (2014) 2541

region, some inclined localized zones started to develop in a shear


region at the support which gradually propagated along a beam
height during loading until they reached a compressive zone. The
critical inclined localized zones (preceding failure), which
occurred at the nearest distance x to the support (x Ea/2), were
the last ones which occurred in the shear region. They initiated
horizontal bond splitting failure zones along the reinforcement
propagating simultaneously towards both supports (Figs. 1719). A
similar failure mode happened during laboratory tests, however a
single horizontal splitting crack always appeared there.
Table 2 presents the calculated local normal (un) and tangential (ut) displacement increments across a diagonal localized
zone during its propagation path (Fig. 15) in the small beam
SL20. The dominant mechanism in a diagonal localized zone was
normal extension. Both displacement increments un and ut
gradually increased during deformation (e.g. un increased from
0.02 mm up to 0.39 mm, whereas ut increased from 0.13 mm up
to 0.25 mm across EF). However, the tangential displacement
increment ut reected mainly the beam deection.
The evolution of the vertical force versus the deection was
consistent with the experimental measurements for all beam sizes
(Fig. 16). The calculated maximum vertical forces were: 91.4 kN
(small-size beam), 167.9 kN (medium-size beam) and 263.9 kN
(large-size beam). Hence, the nominal shear strength, expressed
by N V/tD with V 0.5Pmax, was N 1.43 MPa, N 1.17 MPa and
N 0.88 MPa, respectively it decreased strongly with the beam
height. The calculated width of all localized zones was about
wloc 4lc, i.e. was similar as in our other FE simulations of
reinforced concrete elements [34,36,63]. It was also similar as in
our experiments (measured with the DIC technique using the
length resolution of 100 pixel/mm), which was 16.519.6 mm [19].
A direct comparison between the experimental and numerical
fracture pattern is demonstrated in Fig. 20. The critical diagonal
localized zone occurred at the distance dc 458 mm from
the support (dc/a 0.424) in the beam SL40 and dc 1161 m
(dc/a 0.523) in the beam SL80. The corresponding experimental
average values of the ratio dc/a were 0.58 (SL40) and 0.54 (SL80).
Thus, the numerical outcome was too small in the beam SL40 and
realistic in the beam SL80. The average spacing of vertical localized
zones s at the mid-region was equal to: 87 mm (SL20), 150 mm
(SL40) and 324 mm (SL80), i.e. it was close to the experimental
one, which was 76 mm, 155 mm and 300 mm.
The localized zones at failure in the beam SL20 covered 86% of
its height (0.86H), while in beams SL40 and SL80: 0.773H and
0.784H, respectively. The mean cracked region height in the
experiment was 0.864H (SL20), 0.816H (SL40), 0.836H (SL80),
indicating a good agreement with numerical results with the
small discrepancy of 6% in the beam SL40 and SL80.
6.2. Statistical calculations

Fig. 16. Vertical force P deection u diagrams from deterministic simulations


compared to mean experimental results from 3 tests: (a) small-size beam SL20,
(b) medium-size beam SL40 and (c) large-size beam SL80.

6. Calculated size effect on shear strength


6.1. Deterministic calculations
The deterministic simulations were carried out with the constant tensile strength ft 2.2 MPa. The results showed a signicant
effect of a beam effective depth on the nominal shear strength.
In analyses, the failure mechanism was similar in all beam sizes:
after appearance of several localized zones in a pure bending

The effect of a variability of the local concrete tensile strength ft


on the nominal shear strength was investigated using spatially
correlated random elds. The random elds were generated with
the mean value of ft 2.2 MPa and variation coefcient covft 0.12
(covft sft/ft) using the Gauss distribution [26]. The range of
correlation was assumed to be lcor 100 mm 5wloc (Eq. (16))
based on calculations in unnotched concrete beams [26]. Using
the stratied sampling method (Section 4), 12 samples were solely
chosen for FE numerical analyses. Fig. 21 presents one exemplary
random distribution of the local tensile strength for each
beams size.
In the calculations, a diagonal sheartensile failure mechanism
always occurred (as in deterministic simulations). The 12 vertical
forcedeection curves from statistical calculations compared to
the deterministic curve are shown in Fig. 22. The mean failure

E. Syroka-Korol et al. / Finite Elements in Analysis and Design 88 (2014) 2541

35

Fig. 17. Contours of non-local tensile softening parameter 1 (a) compared to experimental crack pattern (b) and its distribution along beam bottom (c) and along mid-height
(d) from deterministic simulations with small-size reinforced concrete beam.

vertical forces from statistical calculations were as follows:


83.0 kN (7 4.0 kN), 146.0 kN (76.1 kN) and 234.5 kN (7 12.7 kN)
in the small-size, medium-size and large-size beams, respectively.
The corresponding mean nominal shear strengths N V/tD
(V 0.5Pmax) were: 1.30 MPa, 1.01 MPa and 0.78 MPa, respectively.
They were solely by 9%, 13% and 11% lower than the deterministic
outcomes. Thus, a statistical size effect was practically negligible
that was in agreement with the theory by Bazant and Planas [22].
The contours of a non-local tensile softening parameter (with
the marked sections where the bond radial strain was higher than
the limit value r,rs Zr,rs,max) are shown in Figs. 2325.
The statistical width of localized zones and statistical height of
vertical localized zones in a constant bending moment region was
similar to the deterministic values, whereas the statistical height
of a critical diagonal localized zone was slightly higher than in
deterministic calculations: 0.92H (small-size beam SL20), 0.87H
(medium-size beam SL40) and 0.83H (large-size beam SL80),
respectively. The calculated normalized position of a critical
inclined localized zone dc/a in all beams during statistical simulations was higher on average than in corresponding deterministic
analyses and in experiments (i.e. the localized zone developed at a
larger distance from the support). It was dc/a 0.58 (Fig. 23a)
0.64 (Fig. 23c), dc/a 0.46 (Fig. 24a) 0.79 (Fig. 24c) and dc/a 0.43
(Fig. 25a) 0.70 (Fig. 25c) in the beams SL20, SL40 and SL80,
respectively. There exists a clear relationship between the beam
bearing capacity and the position of a critical inclined localized
zone from which the bond failure developed. Generally, when a
critical inclined zone appeared at a larger distance from the
loading point (i.e. at a smaller distance from the support), the
beam bearing capacity was higher. It was, in particular, visible in

large-size beams (Fig. 25) where the maximum vertical force


strongly varied. The highest vertical force (Fig. 25a) was at a
critical inclined localized zone at the distance of 1.28 m from the
loading point, and the lowest one (Fig. 25c) at the distance of
0.68 m. The average spacing of vertical tensile zones in the smallsize beams was s 94 mm (86105 mm), in the medium-size
beams s 170 mm (137198 mm) and in the large-size beams
s355 mm (291463 mm).
In Fig. 26, a comparison of the maximum shear strength,
expressed in analogy to an elastic beam theory as N 1.5 V/(tD),
between FE deterministic and experimental outcomes together
with the calibrated SEL curve by Bazant (Eq. (1)) and two solutions
for a rigid-perfectly plastic body is given.
The original plastic solution for the shear failure mechanism with
a diagonal crack given by Eq. (20) [64,65] assumes an inclined
straight yield line starting from the support, whereas the improved
plastic solution (Eqs. (21) and (22)) by Zhang [66] considers an
inclined straight yield line starting at a certain distance from the
support x (x0.74D(a/D-2). The shear strengths by Nielsen and
Brstrup [64,65] and by Zhang [66], which create the upper bound
of the load bearing capacity, based on the equation of the internal
and external work, are calculated as follows:
"r
#
"r
#
 a 2 a
 a 2 a
0:5f
1

1

0:5c f ck
D
D
D
D
n

and
0:5f

nn

"r
#
"r
#
a  x2 a  x
a  x2 a  x
1

1

0:5s c f ck
D
D
D
D

20

21

36

E. Syroka-Korol et al. / Finite Elements in Analysis and Design 88 (2014) 2541

Fig. 18. Contours of non-local tensile softening parameter 1 (a) compared to experimental crack pattern (b) and its distribution along beam bottom (c) and along mid-height
(d) from deterministic simulations with reinforced concrete medium-size beam.

Fig. 19. Contours of non-local tensile softening parameter 1 (a) compared to experimental crack pattern (b) and its distribution of along beam bottom (c) and along midheight (d) from deterministic simulations with large-size beam.

E. Syroka-Korol et al. / Finite Elements in Analysis and Design 88 (2014) 2541

37

Table 2
Calculated normal un and tangential ut displacement increments between points of Fig. 15 across diagonal localized zone during its evolution (small-size reinforced
concrete beam).
Vertical force

u [mm]

Distance ab

Distance cd

Distance ef

Distance gh

Distance ik

Distance ln

Distance mo

Distance pr

76 kN

un
ut

0.01
0.12

0.01
0.12

0.02
0.13

0.04
0.13

0.09
0.11

0.01
0.09

0.02
0.12

0.15
0.12

86 kN

un
ut

0.03
0.15

0.05
0.15

0.10
0.17

0.15
0.18

0.22
0.14

0.11
0.07

0.04
0.13

0.27
0.12

At peak

un
ut

0.09
0.16

0.14
0.17

0.23
0.21

0.31
0.22

0.40
0.18

0.29
0.01

0.16
0.11

0.40
0.09

After failure

un
ut

0.16
0.19

0.26
0.19

0.39
0.25

0.49
0.28

0.61
0.20

0.46
0.06

0.29
0.11

0.48
0.07

Fig. 20. Contours of non-local tensile softening parameter 1 from deterministic simulations compared to experimental crack pattern (lines) with: (a) small-size,
(b) medium-size and (c) large-size beam.

Fig. 21. Exemplary random elds of tensile strength ft in: (a) small-size, (b) medium-size and (c) large-size beam used in FE analyses (note that beams are not proportionally
scaled).

with



3:5
1
c 1:6p0:27 1 p 0:15 0:58; ;
D
f ck

22

where cfck denotes the effective compressive strength of un-cracked


concrete whereas scfck is the effective compressive strength of

cracked concrete (by taking into account the decreasing concrete


cohesion, expressed through the additional reduction factor s E0.6).
The FE deterministic nominal shear strengths match very well
the experimental ndings and experimentally calibrated SEL curve
by Bazant (with o 2.87 MPa and D0 213 mm in Eq. (1) using the
LevenbergMarquardt non-linear regression method). Both FE and

38

E. Syroka-Korol et al. / Finite Elements in Analysis and Design 88 (2014) 2541

Considering statistical simulations, all mean values underestimate the experimental shear strength by 10% on average (this
result is probably caused by the assumption of a deterministic
bond model in statistical computations or a sampling region
limited to beam shear regions). The mean shear strength reduction
in statistical computations with respect to deterministic results
was close to the assumed coefcient of variation covft of the tensile
strength. Substituting into the experimentally calibrated SEL formula (Eq. (1)) the reduced value of o, expressed by o(1  covft)
2.87(1  0.12) 2.53 MPa, a perfect match between the new SEL
curve and the statistical results was obtained (Fig. 27). Therein, the
brittleness number dened by the ratio D/D0 was not modied.
This indicates that the size effect on the nominal shear strength is
predominantly of a deterministic type.
With respect to FE numerical results on a size effect by Sato
et al. [31], our deterministic results indicate a stronger reduction of
the nominal shear strength with similar increasing beam size (38%
between the small- and large-size beam against 20%).

7. Concluding remarks
The following conclusions can be drawn from our plane stress
non-linear FE analyses with a constant and spatially correlated
random distribution of the local tensile strength in geometrically
similar longitudinally reinforced concrete beams without shear
reinforcement failing in shear:

 An isotropic elasto-plastic constitutive model with non-

Fig. 22. Vertical force P deection u diagrams (12 curves) for reinforced concrete
beams from statistical FE simulations (solid lines) compared to one deterministic
curve (dashed line): (a) small-size beam, (b) medium-size beam, and (c) largesize beam.

experimental results lie obviously below solutions predicted by


the upper bound plastic theory (Eqs. (20) and (21)) and above the
lower elastic bound ( 1). The improved plastic solution (Eq. (21))
is closer to experimental results. The upper horizontal asymptotes
for small structures by Eqs. (20) and (21) are at 3.59 MPa and
3.13 MPa, respectively, whereas for the experimentally calibrated Eq. (1) at 2.87 MPa (Fig. 26).

local softening enhanced by a characteristic length of microstructure was able to realistically reproduce both a very
complex experimental failure mode in beams (combined
diagonal sheartensile failure and horizontal bond splitting
failure) and a pronounced reduction of the nominal shear
strength with increasing beam size by taking a bond model
into account.
The size effect on the nominal shear strength of slender RC
beams without shear reinforcement was of a deterministic type
only. The nominal shear strength from deterministic simulations strongly decreased with increasing specimen size. The
deterministic size effect was caused by localized zones with a
constant width and a linearly varying height. It was in agreement with the energetic size effect by Bazant. Thus, it was not
caused by non-symmetric cracking combined with the unintended out-of-plane beam action. The mean shear strengths in
statistical computations were solely lower by 10% than corresponding deterministic outcomes. The mean statistical shear
strength was proportional to the mean tensile strength reduced
by the standard deviation.
The effect of the bond denition between concrete and reinforcement was pronounced and led to the different failure
mode, load capacity, beam stiffness and number of localized
zones. The application of the bond model by Akkermann [52]
contributed to a sheardiagonal tensile failure mechanism in
numerical simulations as in experiments.
A position of a critical diagonal localized zone in deterministic
calculations was close to experimental outcomes. In statistical
calculations, a critical diagonal localized zone was located at a
higher distance from the beam support, hence the mean
bearing capacity was lower than the corresponding deterministic one. The height of a cracked section in experiment was
successfully reproduced by both deterministic and statistical
analyses. The height of a critical diagonal localized zone was
slightly larger in statistical calculations than in deterministic
ones. The width of localized zones was similar in deterministic
and statistical calculations.

E. Syroka-Korol et al. / Finite Elements in Analysis and Design 88 (2014) 2541

39

Fig. 23. Contours of non-local tensile softening parameter 1 from statistical simulations of small-size reinforced concrete beam corresponding to: (a) highest failure force,
(b) mean failure force, and (c) lowest failure force.

Fig. 24. Contours of non-local tensile softening parameter 1 from statistical simulations of medium-size reinforced concrete beam corresponding to: (a) highest failure
force, (b) mean failure force, and (c) lowest failure force.

Fig. 25. Contours of non-local tensile softening parameter 1 from statistical simulations of large-size reinforced concrete beam corresponding to: (a) highest failure force,
(b) mean failure force, and (c) lowest failure force.

40

E. Syroka-Korol et al. / Finite Elements in Analysis and Design 88 (2014) 2541



1
f 2 sij ; 2 q  p tan  c 2 q  p tan  1  tan sc 2 r 0
3

A2
r
3
s s ;
q
2 ij ij
tan

Fig. 26. FE results of nominal shear strength 1.5 V/(tD) against effective beam
depth D from deterministic analyses for reinforced concrete beams without shear
reinforcement as compared with experiments, size effect law (Eq. (1)) and upper
bound plastic theory results (Eqs. (20) and (21)).

1
p  skk ;
3

A3

31 r sbc
;
1  2r sbc

A4

g1 f 1

A5

g 2 q  p tan

A6

where, si the principal stress (i 1, 2, 3), st the uniaxial tensile


yield stress, 1 the softening parameter equal to the maximum
principal plastic strain 1p, q the Mises equivalent deviatoric
stress, p the mean stress, the internal friction angle in the
meridional stress plane (pq plane), c the cohesion related to
uniaxial compression strength, sij the deviator of the stress
tensor sij, (sij sij ijp) sc the uniaxial compression yield stress,
2 the hardening/softening parameter corresponding to the
plastic vertical normal strain during uniaxial compression, gi
ow potential function, rbcs the ratio between the biaxial
compressive strength and uniaxial compressive strength
(rbcs E1.2) and the dilatancy angle (a). The last term in
Eq. (A3) results from the yield condition q  p tan  c 0 for
uniaxial compression with q sc and p 1/3sc.
The exponential function by Hordijk [37] was always taken as a
non-linear softening function in tension (Fig. 4a)

st 1 f t 1 A1 1 3 exp  A2 1  A3 1 

A7

with
A1

Fig. 27. Mean statistical nominal shear strength 1.5 V/(tD) against effective
beam depth D from FE analyses for reinforced concrete beams without shear
reinforcement compared to modied size effect law SEL of type II (Eq. (1)) by
expression o(1-covft).

Acknowledgements
The research work has been carried out within the project:
Innovative ways and effective methods of safety improvement
and durability of buildings and transport infrastructure in the
sustainable development nanced by the European Union
(POIG.01.01.02-10-106/09-01) and the project Experimental and
numerical analysis of coupled deterministic-statistical size effect
in brittle materials nanced by National Research Centre NCN
(UMO-2013/09/B/ST8/03598).

Appendix
In order to describe the concrete behaviour, two failure criteria
were assumed. In tension regime, the Rankine criterion was used
with the yield function f1 using isotropic softening and associated
ow rule and in compression, the DruckerPrager yield surface f2
with isotropic hardening/softening and non-associated ow rule
was used [36]
f 1 si ; 1 maxfs1 ; s2 ; s3 ;g  st 1 r 0;

A1

c1
;
u

A2

c2
u

and

A3

1
1 c31 exp c2 ;
u

A8

wherein ft is the uniaxial tensile strength, u is the ultimate


softening parameter and ci are the empirical parameters (c1 3
and c2 6.93).
The compressive strength of concrete fc was assumed to be
equal to the characteristic compressive strength fck [38] and tensile
strength ft to the characteristic tensile strength fctk [28], respectively
q
3
2
f ctk 0:7f ctm 0:7 f ck with b 0:3
A9
The elastic modulus was Ec 9500(fck 8)1/3 [38] and the Poisson0 s
ratio 0.2. The ultimate value of the softening parameter under
tension u was estimated from the assumed concrete fracture
energy Gf gfwloc (gf area under the softening curve st(1),
wloc E4lc the width of a localized zone), wherein Gf was
calculated following CEB-FIP [39] as
Gf Gf o f cm =f cmo 0:7 ;

A10

where fcm fck fck is the mean compressive strength with


fck 8 MPa, the reference value is fcmo 10 MPa and Gfo denotes
the parameter depending on the maximum aggregate size
(Gfo 0.058 N/mm with the aggregate diameter of 32 mm). The
ultimate softening parameter under compression 2 was calculated
similarly as in tension (Gc E 4lcgc) from the compressive fracture
energy based on initial computations [35].
References
[1] F. Leonhardt, R. Walther, Schubversuche an einfeldrigen Stahlbeton-Balken
mit und ohne Schubbewehrung zur Ermittlung der Schubtragfhigkeit und
der oberen Schubspannungsgrenze, DAfStb, Heft, 151, W. Ernst and Sohn,
Verlag, Berlin, 1962.
[2] G.N.J. Kani, The basic facts concerning shear failure, ACI J. 63 (1966) 675692.

E. Syroka-Korol et al. / Finite Elements in Analysis and Design 88 (2014) 2541

[3] G.N.J. Kani, How safe are our large reinforced concrete beams? ACI J. 64 (1967)
128142.
[4] N.S. Bhal, ber den Einuss der Balkenho
he auf die Schubtragfhigkeit von
einfeldrigen Stahlbetonbalken mit und ohne Schubbewehrung (Dissertation),
Stuttgart Universitt, 1968.
[5] H.P.J. Taylor, Shear strength of large beams, J. Struct. Div., ASCE 98 (1972)
24732489.
[6] J.C. Walraven, The Inuence of Depth on the Shear Strength of Lightweight
Concrete Beams without Shear Reinforcement, Stevin Laboratory Report no.
5-78-4, Delft University of Technology, 1978.
[7] P.S. Chana, Some aspects of modelling the behavior of reinforced concrete
under shear loading, Tech. Rep., Cement and Concrete Assoc., Wexham
Springs, 1981.
[8] M. Iguro, T. Shioya, Y. Nojiri, H. Akiyama, Experimental studies on shear
strength of large reinforced concrete beams under uniformly distributed load,
Concr. Library Int. JSCE 5 (1985) 137146.
[9] Z. Bazant, M.T. Kazemi, Size effect on diagonal shear failure of beams without
stirrups, ACI Struct. J. 88 (1991) 268276.
[10] T. Shioya, H. Akiyama, Application to design of size effect in reinforced
concrete structures, in: H. Mihashi, H. Okamura, Z.P. Bazant (Eds.), Size Effect
in Concrete Structures, Spon, London, 1994, pp. 409416.
[11] J.K. Kim, Y.D. Park, Shear strength of reinforced high-strength concrete beams
without web reinforcement, Mag. Concr. Res. 46 (1994) 716.
[12] R. Grimm, Einu bruchmechanischer Kenngro
en auf das Biege- und
Schubtragverhalten hochfester Betone (Ph.D. thesis), Fachb. Konstr. Ingenieurbau der TH Darmstadt und DafStb H.477, Beuth Verlag GmbH, Berlin, 1997.
[13] W.M. Ghannoum, Size effect on shear strength of reinforced concrete beams
(M.A.Sc thesis), MC Gill University, Department of Civil Engineering and
Applied Mechanics, Montreal, Canada, 1998.
[14] H. Kawano, H. Watanabe, Shear strength of reinforced concrete columns
effect of specimen size and load reversal, in: Proceedings of the 2nd Italy
Japan Workshop on Seismic Design and Retrot of Bridges, Rome, Italy, 1997,
pp. 141154.
[15] B.A. Podgorniak-Stanik, The inuence of concrete strength, distribution of
longitudinal reinforcement, amount of transverse reinforcement and member
size on shear strength of reinforced concrete members (M.A.Sc. thesis), Dept.
of Civil Engineering, University of Toronto, 1998.
[16] Y. Yoshida, Shear reinforcement for large lightly reinforced concrete members
(M.A.Sc. thesis), Department of Civil Engineering, University of Toronto, 2000.
[17] D. Angelakos, E.C. Bentz, M.P. Collins, Effect of concrete strength and minimum
stirrups on shear strength of large members, ACI Struct. J. 98 (2001) 290300.
[18] A. Lubell, T. Sherwood, E. Bentz, M. Collins, Safe shear design of large, wide
beams, Concr. Int. 26 (2004) 6678.
[19] E. Syroka-Korol, Theoretical and experimental study on size effect in concrete
beams reinforced with steel or basalt bars (Ph.D. thesis), Gdansk University of
Technology, 2012.
[20] Z. Bazant, Size effect in blunt fracture: concrete, rock, metal, J. Eng. Mech.
ASCE, 110, 1984518535.
[21] Z. Bazant, Q. Yu, Designing against size effect on shear strength of reinforced
concrete beams without stirrups: I. Formulation, J. Struct. Eng. 131 (2005)
18771885.
[22] Z. Bazant, J. Planas, Fracture and Size Effect in Concrete and Other Quasi-brittle
Materials, CRC Press LLC, Boca Raton, US, 1998.
[23] Q. Yu, Size effect and design safety in concrete structures under shear (Ph.D.
thesis), Northwestern University, 2007.
[24] L. Skarynski, E. Syroka, J. Tejchman, Measurements and calculations of the
width of the fracture process zones on the surface of notched concrete beams,
Strain 47 (s1) (2011) e319e332.
[25] J. Bobiski, J. Tejchman, J. Grski, Notched concrete beams under bending
-calculations of size effects within stochastic elasto-plasticity within non-local
softening, Arch. Mech. 61 (2009) 125.
[26] E. Syroka-Korol, J. Tejchman, Z. Mrz, FE calculations of a deterministic and
statistical size effect in concrete under bending within stochastic elastoplasticity and non-local softening, Eng. Struct. 48 (2013) 205219.
[27] E. Syroka-Korol, J. Tejchman, Experimental investigations of size effect in
reinforced concrete beams failing by shear, Eng. Struct. 58 (2013) (2014)
6378.
[28] M.D. Kotsovos, M.N. Pavlovic, A possible explanation for size effects in
structural concrete, Arch. Civ. Eng., Pol. Acad. Sci. XL (1994) 243261.
[29] M.D. Kotsovos, M.N. Pavlovic, Size effects in structural concrete: a numerical
experiment, Comput. Struct. 64 (1997) 285295.
[30] F.J. Vecchio, W. Shim, Experimental and analytical re-examination of classic
concrete beam testsJ. Struct. Eng., ASCE (2004)460469.
[31] Y. Sato, T. Tadokoro, T. Ueda, Diagonal tensile mechanism of reinforced
concrete beams, J. Adv. Concr. Technol. 2 (2004) 327341.
[32] A. Slobbe, A.M. Hendriks, J. Rots, Sequentially linear analysis of shear critical
reinforced concrete beams without shear reinforcement, Finite Elem. Anal.
Des. 50 (2012) 108124.
[33] Q. Yu, Z. Bazant, Can stirrups suppress the size effect on shear strength of RC
beams? J. Struct. Eng., ASCE 137 (2011) 607617.

41

[34] I. Marzec, J. Bobinski, J. Tejchman, Simulations of crack spacing in reinforced


concrete beams using elastic-plasticity and damage with non-local softening,
Comput. Concr. 4 (2007) 377403.
[35] E. Syroka, J. Bobiski, J. Tejchman, FE analysis of reinforced concrete corbels
with enhanced continuum models, Finite Elem. Anal. Des. 47 (2011)
10661078.
[36] J. Tejchman, J. Bobiski, Continuous and Discontinuous Modelling of Fracture
in Concrete using FEM, Springer, Berlin-Heidelberg, 2012.
[37] D.A. Hordijk, Local approach to fatigue of concrete (Ph.D. thesis), Delft
University of Technology, 1991.
[38] EN-1992-1-1: 2004 Eurocode 2: Design of concrete structures Part 1-1:
General rules and rules for buildings, 2008.
[39] CEBFIB Model code 1990 for concrete structures, 1990.
[40] G. Pijaudier-Cabot, Z. Bazant, Nonlocal damage theory, ASCE J. Eng. Mech. 113
(1987) 15121533.
[41] Z. Bazant, M. Jirasek, Nonlocal integral formulations of plasticity and damage:
survey of progress, J. Eng. Mech. 128 (2002) 11191149.
[42] J. Bobinski, J. Tejchman, Numerical simulations of localization of deformation
in quasi-brittle materials within non-local softening plasticity, Comput. Concr.
4 (2004) 433455.
[43] C. Polizzotto, G. Borino, P. Fuschi, A thermodynamic consistent formulation of
nonlocal and gradient plasticity, Mech. Res. Commun. 25 (1998) 7582.
[44] R.B.J. Brinkgreve, Geomaterial models and numerical analysis of softening
(Ph.D. thesis), Delft University of Technology, 1994.
[45] C. Polizzotto, Remarks on some aspects of non-local theories in solid
mechanics, in: Proceedings of the 6th National Congress SIMAI, Chia Laguna,
Italy, CD-ROM, 2002.
[46] M. Jirsek, S. Rolshoven, P. Grassl, Size effect on fracture energy induced by
nonlocality, Int. J. Numer. Anal. Methods Geomech. 28 (78) (2004) 653670
(2004b).
[47] ABAQUS, Theory Manual, Version 5.8, Hibbit, Karlsson & Sorensen Inc., 1998.
[48] T.J.R. Hughes, J. Winget, Finite rotations effects in numerical integration of rate
constitutive equations arising in large deformation analysis, Int. J. Numer.
Methods Eng. 15 (1980) 18621867.
[49] Z. Bazant, M. Gattu, J. Vorel, Work conjugacy error in commercial niteelement codes: its magnitude and how to compensate for it, Proc. R. Soc. A
468 (2012) 30473058.
[50] E. Pramono, Numerical simulations of distributed and localized failure in
concrete (Ph.D. thesis), University of Colorado-Boulder, 1988.
[51] K. Do
rr, Ein Beitrag zur Berechnung von Stahlbetonscheiben unter besonderer
Bercksichtigung des Verbundverhaltens (Ph.D. thesis), Darmstadt Universitt, 1980.
[52] J. Akkermann, Rotationsverhalten von Stahlbeton-Rahmendecken (Ph.D. thesis), Karlsruhe Universitt, 2000.
[53] J.A. den Uijl, A. Bigaj, A bond model for ribbed bars based on concrete
connement, Heron 41 (1996) 201226.
[54] K. Karhunen, ber lineare Methoden in der Wahrscheinlichkeitsrechnung,
Ann. Acad. Sci. Fennicae Ser. A. I. Math. Phys. 37 (1947) 179.
[55] M. Love, Functions aleatoires du second ordre. Supplement to P. Levy,
Processus Stochastic et Mouvment Brownien, Paris, Gauthier Villares, 1948.
[56] R.G. Ghanem, P.D. Spanos, Stochastic Finite Elements A Spectral Approach,
Springer Verlag, New york, US, 1991.
[57] K.K. Phoon, S.P. Huang, S.T. Quek, Simulation of second-order processes using
KarhunenLoeve expansion, Comput. Struct. 80 (2002) 10491060.
[58] K.K. Phoon, S.P. Huang, S.T. Quek, Implementation of KarhunenLoeve expansion for simulation using a wavelet-Galerkin scheme, Probab. Eng. Mech. 17
(2002) 293303.
[59] J. Neyman, On the two different aspects of the representative method: the
method of stratied sampling and the method of purposive selection, J. R. Stat.
Soc. 97 (1934) 558625.
[60] J. Tejchman, J. Grski, Computations of size effects in granular bodies within
micro-polar hypoplasticity during plane strain compression, Int. J. Solids
Struct. 45 (6) (2007) 15461569.
[61] J. Tejchman, J. Grski, Deterministic and statistical size effect during shearing
of granular layer within a micro-polar hypoplasticity, Int. J. Numer. Anal.
Methods Geomech. 32 (2008) 81107.
[62] D. Novak, W. Lawanwisut, C. Bucher, Simulation of random elds based
orthogonal transformation of covariance matrix and latin hypercube sampling, in: Proceedings of the International Conference on Monte Carlo
simulation MC 2000. Swets&Zeitlinger, Liesse, Monaco, Monte Carlo, 2001,
pp. 129136.
[63] T. Malecki, I. Marzec, J. Bobiski, J. Tejchman, Effect of a characteristic length
on crack spacing in a reinforced concrete bar under tension, Mech. Res.
Commun. 34 (2007) 460465.
[64] M.P. Nielsen, M.W. Brstrup, Plastic shear strength of reinforced concrete
beams. Report R 73 1976, Technical University of Denmark, 1967.
[65] M.P. Nielsen, M.W. Brstrup, Shear strength of prestressed concrete beams
without web reinforcement, Mag. Concr. Res. 30 (1978) 119128.
[66] J.P. Zhang, Shear strength of conventional reinforced concrete beams, deep
beams, corbels and prestressed reinforced concrete beams without shear
reinforcement (Ph.D. thesis), Department of Structural Engineering, Technical
University of Denmark, 1994.