See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/226479773

Size effect of concrete members applied with flexural compressive stresses

Article  in  International Journal of Fracture · March 2004

DOI: 10.1023/B:frac.0000025300.66782.38

CITATIONS READS

12 4,437

3 authors, including:

Jang-Ho Jay Kim Jin-Keun Kim

Yonsei University Korea Advanced Institute of Science and Technology
129 PUBLICATIONS   1,226 CITATIONS    153 PUBLICATIONS   3,578 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Concrete PBD Guideline Development View project

Effect of carbonation on the rebound number and compressive strength of concrete View project

All content following this page was uploaded by Jang-Ho Jay Kim on 26 May 2015.

The user has requested enhancement of the downloaded file.

 International Journal of Fracture 126: 79–102, 2004.
© 2004 Kluwer Academic Publishers. Printed in the Netherlands.

Size effect of concrete members applied with flexural compressive

stresses

JANG-HO JAY KIM1 , SEONG-TAE YI2,∗ and JIN-KEUN KIM3

1 Department of Civil & Environmental Engineering, Sejong University, 98, Kunja-dong, Kwangjin-gu, Seoul-si,
143-747, South Korea
2∗ Corresponding author, Department of Civil Engineering, Chung Cheong University, 330, Wolgok-ri,
Kangnae-myun, Cheongwon-kun, Chungbuk-do, 363-792, South Korea, phone: +82-43-230-2315; Fax:
+82-43-230-2319; e-mail: yist@ok.ac.kr
3 Department of Civil & Environmental Engineering, Korea Advanced Institute of Science and Technology, 373-1,
Kusong-dong, Yusong-gu, Daejon-si, 305-701, South Korea

Received 6 June 2002; accepted in revised form 11 February 2004

Abstract. In this study, two types of special experiments are carried out to understand flexural compressive
strength size effect of concrete members. The first type is an ordinary cylindrical specimen (CS) with a fully pen-
etrated and vertically standing plate type notch at the mid-height of the specimen, which is loaded in compression
at the top surface (e.g., in the parallel direction to the notch length). The second type is a general double cantilever
beam (DCB), which is compression loaded in axial direction (e.g., in the parallel direction of the notch). For CS, an
adequate notch length is taken from the experimental results obtained from the compressive strength experiment
of various initial notch lengths. The trial tests to select the effective initial notch length show that CS with an
initial notch length approximately greater than four times the maximum aggregate size fails without an additional
increased load and in stable manner under Mode I failure mechanism. Therefore, the initial notch length to the
maximum aggregate size ratio of 4.0 is used for all size specimens. For DCB, the eccentricity of loading points
with respect to the axial axis of each cantilever and the initial notch length are varied. In both specimens, the
compressive loads apply flexural compressive stresses on the crack tip region of the specimens. These two types
of specimens fail by Mode I crack opening mechanism. By testing 3 geometrically proportional size specimens
for CS and DCB, the experimental datum for flexural compression size effect of concrete are obtained. Using
the obtained flexural compressive strength size effect datum, regression analyses are performed using Levenberg-
Marquardt’s least square method (LSM) to suggest new parameters for the modified size effect law (MSEL). The
analysis results show that size effect is apparent for flexural compressive strength of specimens with an initial
notch. For CS, the effect of initial notch length on flexural compressive strength size effect is apparent. For DCB,
flexural compressive size effect is dependent on the eccentricity of loading points with respect to the axial axis
of the cantilever beam. In other words, if DCB specimen is applied with greater tensile stress at the crack tip, the
size effect of concrete becomes more distinct. The results show that the flexural compressive strength size effect
of initial notch length variation of DCB exists but directly dependent on the loading location. This is due to the
fact that the sizes of fracture process zone (FPZ) of all DCB specimens are similar regardless of the differences
in the specimen slenderness ratio, but the flexural compressive and tensile stress combinations resulting in stress
concentration at the crack tip region has direct effect on size effect of concrete members.

Key words: cylindrical specimen, double cantilever beam, eccentricity, flexural compressive strength, initial notch
length, size effect.
80 J.-H. Jay Kim et al.

List of symbols

a = length of crack band, crack length at failure

a0 = initial notch length
b = width of specimen
B, α = empirical constants
cn = coefficient introduced for convenience
d = characteristic dimension
D = depth of specimen
da = maximum aggregate size
dc = diameter of cylinder
e = eccentric length of loading points, eccentricity
h = height of cylinder
fc = ultimate strength in concrete
fc = uniaxial compressive strength of standard concrete cylinder
fc (h, dc ) = compressive strength of general cylinder
fct = splitting tensile strength of concrete cylinder
ft = direct tensile strength of concrete cylinder
l = span length
lch (= d) = characteristic dimension of the specimen
l0 = width of crack band (= λ0 da )
Pu = maximum load, or ultimate axial load
s = standard deviation
t = thickness of specimen
λ0 = approximate constant (= 2.0)
σ0 = size independent stress (= αft )
σN (d) = nominal compressive strength at failure

1. Introduction

All materials have specific inherent material properties. For each material, the properties are
considered unique when they are independent of a specimen size and shape. For design pur-
poses, the concrete compressive strength of standard cylinder (φ150 × 300 mm) is accepted
as the most basic and important material property. However, the common notion that concrete
compressive strength is a unique material property is an erroneous one since the compressive
strength of concrete changes based on specimen sizes due to its fracture characteristics.
Many pores and micro-cracks are formed inside of concrete during its casting and curing
stages. Since these micro-defects can be considered as initial notches in concrete members,
it is necessary to apply fracture mechanics to evaluate material characteristics of concrete
(Bazant, 1984; Shah et al., 1971; Karihaloo, 1992). Because ultimate tensile and compressive
failures are direct consequences of macro-cracks initiating from original pores and defects
present in concrete, fracture mechanics approach to solve the problem is an obvious choice.
 Size effect of concrete members applied with flexural compressive stresses 81

Therefore, various researches predicting the behavior and characteristics of concrete by ap-
plying fracture-based failure theory have been performed. Among many researches performed
to understand the effect of fracture characteristics on material properties of concrete, the most
effective but rigorously proven research method is the size effect research. The term size effect
means concrete material strength dependency to specimen size variation.
There have been numerous reports (Bazant, 1984; Shah et al., 1971; Carpinteri et al.,
1994) about the significance of member size difference on the tensile strength of quasi-brittle
materials. Also, these experimental results have been extended to the derivations of size effect
laws that calculate the nominal tensile strength of concrete members as function of member
sizes. Size effect is an inherent material characteristic, which exists in quasi-brittle or brittle
materials where the ultimate failure is caused by the formation of macro-cracks.
Although the failure mechanism and size effect of tensile failure have been studied extens-
ively, the behavior of compressive failure has not been sufficiently studied in comparison to
the tensile failure mechanism. Concrete is a construction material normally used to withstand
compressive force. Thus, it is logical to extend the tensile size effect research to compressive
failure research. For this reason, size effect studies on compressive failure of concrete are
currently being performed by many researchers (Ruettgers, 1933; Neville, 1956, 1966; Kim
et al., 1999, 2000, 2001; Yi et al., 2002).
Compressive strength size effect of concrete is not as distinct as in tensile, flexural, and
shear strengths. However, researchers (Bazant, 1984; Gonnerman, 1925; Blanks and Mc-
Namara, 1935; Powers, 1956) in the field accept the conclusion that the nominal compressive
strength decreases as the specimen size increases. Only recently, the studies on compressive
loading based size effect became a focus of interest among many researchers (Cotterell, 1972;
Bazant and Xiang, 1994, 1997). The problem of selecting correct compressive strength for
concrete structure design is critical since the structural member size will change significantly
based on the strength selection. Due to size effect, strength of concrete used for actual struc-
tural members will be different than ones obtained from the laboratory-size cylinder or cubical
specimens. Therefore, it is vital to use a precise concrete strength based on compressive
strength size effect law (SEL) rather than using strength of standardized cylindrical specimen
such as φ150 × 300 mm or φ100 × 200 mm.
The purpose of this study is to experimentally investigate the size dependency on com-
pressive failure strength of concrete. This study only deals with flexural compressive failure
among two representative compressive failure modes of axial compression and flexural com-
pression. In this study, two types of special experiments are carried out to understand flexural
compressive strength size effect of concrete members. The first type is an ordinary cylindrical
specimen (CS) with a fully penetrated and vertically standing plate type notch at the mid-
height of the specimen, which is loaded in compression at the top surface (e.g., in the parallel
direction to the notch length). The second type is a general double cantilever beam (DCB),
which is compression loaded in axial direction (e.g., in the parallel direction of the notch).
By testing 3 geometrically proportional size specimens for CS and DCB, the experimental
datum for flexural compressive strength size effect of concrete are obtained. Using the ob-
tained flexural compressive strength size effect datum, regression analyses are performed
using Levenberg-Marquardt’s least square method (LSM) to suggest new parameters for the
modified size effect law (MSEL).
82 J.-H. Jay Kim et al.

2. Compressive failure mechanism of concrete

There are various reasons for the presence of size effect in quasi-brittle materials. The most
significant characteristic is the formation of micro-cracks at crack front, which releases energy
into a damaged localized zone. This damaged localized zone is called a finite-size fracture
process zone (FPZ), where the size of FPZ is an inherent material property and is unique for
each material. When stress concentration at front of crack (or called stress intensity factor,
SIF) reaches the capacity of FPZ, the propagation of macro-crack occurs and the maximum
load of the specimen is reached causing failure. However, the nominal strength of a material
is generally defined as the maximum load divided by the characteristic size or area. Therefore,
the size of the specimen has direct effect on its nominal strength. For the compression loaded
concrete specimens, there are no observable cracks unlike tension-loaded specimen. This is
because concrete has nearly ten fold greater strength in compression than in tension. However,
stress is definitely concentrated at the crack tip region and micro-cracks form in the loaded
direction (Cotterell, 1972; Nemat-Nasser and Obata, 1988). As the load increases, one or
several macro-cracks are formed by the coalescence of these micro-cracks ultimately leading
to a failure. Occurrence and development of micro-cracks and the typical pattern of compress-
ive failure are described in detail in a research performed by Bazant and Xiang (1997). This
phenomenon is more apparent as the specimen size increases.
Based on the stress concentration rate due to the loading type or specimen configuration,
the intensity of size effect is determined. More specifically, a rate of stress concentration at the
front of crack tips for a member loaded in compression is significantly slower than a member
loaded in tension. Therefore, tensile member has more distinct size effect than compressive
member.
As stated before, there are two main types of concrete compressive strength: (1) axial
compressive strength and (2) flexural compressive strength. For axial and flexural compressive
specimens, the shapes of strain (or stress) gradients on a specimen cross-section are rectangu-
lar and triangular distributions (Kim et al., 1999, 2000, 2001), respectively. Because the stress
distributions of an axial compressive specimen is uniformly distributed along the cross-section
and linear varied in a flexural compressive specimen, the stress concentration of a flexural
compressive specimen is much more significant than in an axial compressive specimen. Due
to the weaker intensity of stress concentration of an axial compressive specimen than in a
flexural compressive specimen, the size effect reflected in axial compressive specimens is less.
This size effect difference is shown more vividly in a comparison of tension and compression
loaded specimens. In a tensile loaded member, the stress applied at the crack tip region is
pure tensile stress. On the other hand, a member loaded in compression applies a combination
of tensile and compressive stresses at the crack tip region. Since tensile stress most rapidly
brings about stress concentration at the crack tip region, the size effect of tension specimens
are much more distinct than in compression specimens.
The typical failure and crack patterns on cylindrical specimens and beams are shown in
Figure 1. In the case of (a) tensile failure, a propagation of a main macro-crack causes a failure.
However, in (b) axial compressive failure or in (c) flexural compressive failure, distributed
multiple micro-cracks coalesce into splitting cracks causing failure. In addition, due to the
loading direction, the size of the characteristic dimension lch of tensile failure is different than
that of compressive failure. Even though the failure mechanism of tensile failure is different
than compressive failure, the ultimate failure of both specimens is due to the propagation of
 Size effect of concrete members applied with flexural compressive stresses 83

Figure 1. Tensile failure and compressive failure of concrete: (a) tension; (b) axial compression; and (c) flexural
compression.

Figure 2. Failure Modes of compressive failure: (a) failure by Mode I; and (b) failure by Mode II.

macro-crack indicating a localized tension or Mode I failures. Therefore, it is safe to assume

that the tensile fracture-based concept can be applied to compressive failure as well.
Compressive failure depends on the shape of structure (i.e., slenderness and boundary
conditions), size, energy release characteristics, and second-order geometric effects (internal
buckling as described by Biot (1965)). The occurrence and development of a band of micro-
cracks can be characterized by two failure modes as follows:
Mode I: Axial propagation of a band of axial splitting micro-cracks that finally coalesce into
one axial splitting crack (Figure 2a).
Mode II: A transverse propagation of a band of axial splitting micro-cracks terminating with
an inclined failure surface traditionally regarded as a shear failure (Figure 2b). This type
of failure is observed when the aggregate is not symmetrically distributed within the
specimen or the load is not exactly applied on the central line of the specimen.
84 J.-H. Jay Kim et al.

Based on the experimental results of specimens subjected to axial compressive force, most
of the specimens fail by Mode I when subjected to tensile stress normal to initial notch
and cracked surface. Accordingly, in this study, the characteristics of Mode I crack opening
mechanism leading to failure are investigated.

3. Modified size effect law (MSEL)

In ductile materials such as metals and plastics, the size effect is not observed due to the
absence of fracture related failure. The failure of such materials occurs due to plastic deform-
ation without propagation of crack. However, it is a well-known fact that quasi-brittle material
such as concrete shows an effect of specimen size difference even if it has no initial crack.
Bazant (1984) derived SEL from the dimensional analysis and similitude arguments for
geometrically similar structures of different sizes with initial crack considering the energy
balance at crack propagation in concrete as shown in Equation (1).
Pu Bft
σN (d) = = , (1)
bd d
1+
λ0 da
where, σN (d) is the nominal strength; Pu is the maximum load; b is the thickness of specimen;
d is the characteristic dimension; ft is the direct tensile strength of concrete cylinder; da is
the maximum aggregate size; and B and λ0 are empirical constants.
In fact, concrete specimens without initial notch are not considered to have the same geo-
metrically proportional relative initial notch length, but rather the same initial notch length due
to the inherently existing defects or voids. It is thus necessary to take into account the influence
of a/d on the size effect for concrete specimens without initial notch. This problem can be
overcome by taking the parameter λ0 in Bazant’s equation as a function of a/d, implying
a smaller value of λ0 for increasing specimen size. As a consequence, Bazant’s size effect
law can be modified in the form of the following equation by introducing a monotonically
decreasing function f (a/d).
Bft
σN (d) =  . (2)
d
1+
f (a/d)da
To explain the concept of Equation (2), specimen series are considered as shown in Fig-
ure 3. These are three geometrically proportional specimen series denoted by A, B, and C
with different crack length/specimen depth ratios a/d. The size effect for each series follows
Bazant’s size effect law with the function f (a/d) for each series has a constant value. How-
ever, the diagonal specimen series consisting of the specimens denoted by A1 , B2 , and C3 may
show the size effect in accordance with Equation (2), since these specimens have the same
crack length but different a/d ratios.
The size effect of real concrete specimens without initial cracks can be evaluated from
such diagonal specimen series with initial cracks dissimilar to the specimen sizes. In other
words, if specimens A1 , B2 , and C3 have the same maximum aggregate size, the same initial
crack length can be assumed, because the cracks can be viewed as pre-existing notches (i.e.,
defects or voids). On the other hand, the initial crack length a0 is clearly distinguished from the
advanced crack length at failure a. However, there is a clear trend in discussing the geometrical
 Size effect of concrete members applied with flexural compressive stresses 85

Figure 3. Illustration of similar specimen series (horizontal) and dissimilar specimen series (diagonal).

Figure 4. Trend of modified size effect in concrete structures of different a/d ratio.

similarity of the specimen series in Figure 3, since the two crack lengths a0 and a can both be
expressed analogously as a function of the maximum aggregate size.
Figure 4 shows the trend of modified size effect for the diagonal specimen series. Ac-
cording to Bazant’s size effect law, nominal strengths at failure for each row specimen series
should behave as shown by dotted lines in Figure 4. The value of f (a/d) corresponding to
λ0 decreases gradually from series A (large a/d) to series C (small a/d). Thus, the real size
effect for the specimens A1 , B2 and C3 can be predicted with the modified size effect curve of
Equation (2), which provides a smooth transition from series A to series C showing a gradual
86 J.-H. Jay Kim et al.

trend in size effect due to the decreasing value of f (a/d). Unfortunately, it is very difficult to
exactly derive the function f (a/d). The crack length at failure a is also difficult to measure
experimentally. Thus, a new empirical formula is suggested for the purpose of regression
analyses within some practical range of sizes. In this new formula, there is little difference in
the strength reduction trends.
Thereafter, introducing the size independent strength σ0 (= αft ), Kim and Eo (1990) and
Kim et al. (1989) proposed a MSEL, which was also proposed by Bazant (1987, 1993) and
Bazant and Xiang (1997) in different approach, given by
Bft
σN (d) =  + αft , (3)
d
1+
λ0 da
where, α is an empirical constant less than unity.
The value of λ0 da in Equation (3) can be taken as constant, since the existing test data
show little effect on the width of micro-crack zone for the aggregate size usually used in
construction. Equation (3) can therefore be expressed as follows:
Bft
σN (d) =  + αft . (4)
d
1+
l0
For this study, tensile strength ft in Equation (3) is substituted with compressive strength fc
for the regression analysis of flexural compressive size effect.
In Jenq and Shah’s study (1991), the size-independent strength was explained differently
based on the fracture behavior of concrete. The nominal strength σN (d) was considered as
the sum of the size-independent and the size-dependent strength, namely the former for the
uncracked state and the latter for the cracked state. l0 being a characteristic length which can
also be determined from regression analyses.
It can be seen from Figure 4 that the size effect is insignificant for both very small and very
large specimen sizes with the modified formula when compared with the pronounced strength
reduction, especially for very large sizes, with Bazant’s formula. Furthermore, the size effect
for transition sizes is apparent, but slightly deduced by the modified formula. This is because
the ratio of the initial bond crack length to the specimen size gradually becomes smaller.
From the above discussion and the fact that cement-aggregate interfacial weakness can
probably cause initial cracks through concrete, it is apparent that the size effect for concrete
specimens with no initial artificially-made crack occurs as a rather milder strength reduction
than the severe one predicted by the previously derived size effect law. The validity of the
above theoretical investigations was demonstrated by regression analyses on available test
data for splitting tensile strength, shear strength, and uniaxial compressive strength (Kim and
Eo, 1990, Kim et al., 1989).
As an application of this law, several researches have been performed on cylindrical spe-
cimens subjected to uniaxial compressive force (Kim et al., 1999) and C-shaped specimens
subjected to flexural compressive force (Kim et al., 2000, 2001). In Equation (4), the width
of crack band l0 is empirically found to be related to the maximum aggregate size, e.g.,
l0 = λ0 da in which λ0 is an approximate constant with values between of 2.0 and 3.0 (Kim
et al., 1999, 2000, 2001). In the regression analysis, this constant is selected as 2.0 where
l0 = 2.0 × da = 2.6 cm.
 Size effect of concrete members applied with flexural compressive stresses 87
Table 1. Concrete mixture proportions (CS).

w/c s/a Unit weight S.P.∗∗

(%) (%) (kg/m3 ) (%)
W C S G∗

35 40 175 500 671 1027 2

∗ maximum aggregate size of 13 mm.
∗∗ super-plasticizer (high-range water-reducing admixture),
ratio of cement weight.

Table 2. Concrete mixture proportions (DCB).

w/c s/a Unit weight S.P.∗∗∗

(%) (%) (kg/m3 ) (%)
W C S G∗ S.F.∗∗

28 38 177 583 652 1085 58 2.7

∗ maximum aggregate size of 13 mm.
∗∗ silica fume.
∗∗∗ super-plasticizer (high-range water-reducing admixture), ratio of
cement weight.

In the previous study (Kim et al., 1999), Equation (5) was proposed to obtain the compress-
ive strength of cylindrical concrete specimens with various diameters and height/diameter
ratios. In Kim et al.’s paper (1999), the derivation and explanation of the characteristic length
determination method are clearly reported for the readers. For this purpose, the effects of the
maximum aggregate size on the fracture process zone (FPZ) were considered and the concept
of characteristic length was newly introduced. In this case, data numbers of specimens with
h/d = 2.0 and h/d  = 2.0 are 222 and 456, respectively.
0.4fc
fc (h, dc ) =  + 0.8fc , (5)
(h − dc )
1+
5
where, compressive strength of general cylinder fc (h, dc ) and fc are in MPa; height of cylin-
der h and diameter of cylinder dc are in cm. Equation (5) was compared to the ASTM standard
(ASTM C 42) and it is noted that the prediction values of the Equation (5) are less than those
of the ASTM standard, but the difference is minimal.

4. Test specimens and experimental program

To simulate flexural compressive failure by axial direction cracks, size effect in flexural com-
pressive failure is evaluated using modified CS and DCB as shown in Figures 5 and 7, respect-
ively. For ordinary CS, generally, a specimen without initial notch is loaded in compression at
the top surface in the parallel direction to the specimen length. However, in this study, a fully
penetrated and vertically standing plate type notch at the mid-height of the specimen is made
to apply flexural compressive stress at the crack front when loaded in compression at the top
surface. For DCB, loading is generally applied at the tip of each beam in the direction normal
88 J.-H. Jay Kim et al.
Table 3. Dimension of specimens (CS).

Specimen Specimen size, Initial notch length, Number of

No. dc × h (cm) specimens
(cm)

S-N0 0 4
S-N6 5 × 37.5 6 5
S-N12 12 5

M-N0 0 3
M-N12 10 × 75 12 3
M-N24 24 3

L-N0 0 2
L-N12 20 × 150 12 3
L-N24 24 3
L-N48 48 3

to axial axis of a member. In this study, however, the load is applied parallel to the axial axis
of DCB as shown in Figures 6b or 7 to apply flexural compressive stress at the crack front.

4.1. M IXTURE PROPORTIONING

The expected average concrete compressive strengths for CS and DCB are 50 MPa and
70 MPa, respectively. The concrete mixture proportions selected for CS and DCB are listed
in Tables 1 and 2, respectively. Type I portland cement is used in all mixtures. Crushed gravel
is used as the coarse aggregate and the maximum aggregate size da is 13 mm. In addition,
a high-range water-reducing admixture and vibrator are used to improve workability and
consolidation of concrete. CS and DCB are cast vertically and horizontally, respectively, on a
level surface.
All specimens and test cylinders are removed from the molds after 24 h. In addition, CS
are dry-cured under wet burlap/towel and DCB are wet-cured in a curing tank until testing.

4.2. D ETAILS OF TEST SPECIMENS

The dimensions, shape, and notch location of CS used in the experiments are shown in Fig-
ure 5. Cross-sectional details are also shown in this figure. In this study, pre-tests (strength
tests) are performed on specimens with various initial notch lengths to determine the effective
initial notch length that actually reflects significant size effect. In this research, the results
show that CS with an initial notch length approximately greater than four times the max-
imum aggregate size fails without additionally increased load and in stable manner. Notch
length to maximum aggregate size ratios of 0.5:1, 1:1, 2:1, 3:1, 4:1, and 5:1 are applied to
φ100×750 mm cylindrical specimens, which are used to study the effect of initial notch length
on strength variation. Specimen size and initial notch length are listed in Table 3. In the size
effect test, geometrically proportional specimens and a minimum number of three specimens
for each size are tested. A size ratio of 1:2:4 is used. In the CS test, the nomenclatures S, M,
and L represent the size of the specimens with S being the smallest and increasing accordingly.
 Size effect of concrete members applied with flexural compressive stresses 89

Figure 5. Size and shape of specimens (CS).

Figure 6. Crack pattern and modeling for flexural compressive failure (DCB): (a) crack pattern; and (b) modeling.

N0, N6, N12, N24, and N48 are serial numbers of specimens with an initial notch length for
each specimen size. In the case of size S, 4 or 5 specimens instead of 3 specimens are tested
since more data scattering is expected.
For DCB loaded in axial compression, the crack pattern of specimens applied with flexural
compressive stress at the crack front and the selection method of the specimen for modeling
of flexural compressive failure are shown in Figure 6. The dimensions, shape, notch location,
and loading points (eccentricity, e) of DCB used in the experiments are shown in Figure 7.
For a constant member size, it is predicted that size effect changes with eccentricity of loading
90 J.-H. Jay Kim et al.

Figure 7. Size and shape of specimens (DCB).

points and initial notch length. Thus, size effect is evaluated by varying the eccentricity and
the initial notch length for three different sizes.
The size effect of initial notch length variation is studied by considering the notch length to
specimen width ratios of 0.2, 0.4, and 0.6. The notch length to specimen width ratio of 0.4 is
selected due to stable crack propagation and failure mechanism. The main test variable is the
size ratio of 1:2:4 of the specimens as listed in Table 4. The width b and depth D are changed
proportionally. More specifically, a specimen width b to specimen depth D ratio of 1:2 is used.
In addition, the thickness is kept constant (t = 5.0 cm) to eliminate the out-of-plane effect of
the test results. Two point loads of equal magnitude are applied simultaneously on right and
left of the notch with same eccentricity. When the load is applied at a location in which the
eccentricity is small, size effect is negligible since compression dominated stress is located in
the proximity of the initial crack tip (i.e., similar to compression test of cylindrical specimen).
However, when the load is applied to the location in which the eccentricity is large, size effect
would be distinct. Effect of eccentricity on size effect will be evaluated in detail in the ‘Size
effect with eccentricity’ section.
In DCB tests, the nomenclatures S, M, and L represent the size of the specimens with S
being the smallest and increase accordingly. E1, E2, and E3 are serial numbers of specimens
with an eccentricity to specimen width ratio e/b. A1, A2, and A3 are serial numbers of
specimens with an initial notch length to specimen width ratio a0 /b. In the case of size S,
6 specimens are tested, because data scattering is more likely to occur. However, in Table 4,
total specimen number less than 6 means that some specimens have been eliminated due to
undesired modes of failures. The negative and positive values of e/b mean that the location
of a point load on a cantilever is inside and outside of the region from cantilever axial axis to
member axis, respectively.
 Size effect of concrete members applied with flexural compressive stresses 91
Table 4. Dimension of specimens (DCB).

Specimen D×b×t e/b a0 /b Number of

No. (cm) specimens

S-E1-A2 −0.15 0.4 4

S-E2-A1 0.2 6
S-E2-A2 20 × 10 × 5 0.0 0.4 6
S-E2-A3 0.6 5
S-E3-A2 0.15 0.4 5

M-E1-A2 −0.15 0.4 3

M-E2-A1 0.2 3
M-E2-A2 40 × 20 × 5 0.0 0.4 3
M-E2-A3 0.6 3
M-E3-A2 0.15 0.4 3

L-E1-A2 −0.15 0.4 3

L-E2-A1 0.0 0.2 3
L-E2-A2 80 × 40 × 5 0.0 0.4 3
L-E2-A3 0.6 3
L-E3-A2 0.15 0.4 3

Table 5. Experimental results on initial notch length (unit: MPa, CS).

a0 0.5da 1.0da 2.0da 3.0da 4.0da 5.0da fc

58.27 57.00 58.76 58.27 51.70 59.35

fc 57.98 59.35 56.21 56.39 – 50.17 59.84
54.01 55.39 55.79 58.46 54.61 54.61

4.3. T EST PROCEDURE

The axial compressive load shown in Figures 5 and 7 is applied using a universal testing
machine (UTM) with a capacity of 2,500 kN using a displacement control method. Control
velocity for CS is adjusted to fail within 10 min and control velocity for DCB is maintained at a
velocity of 0.003 mm/sec. This procedure is used until the specimen failed. Before testing, the
DCB are taken from a curing tank. As shown in Figure 7, two sets of a bearing plate attached
with a ball bearing are fixed at the end portions of both cantilever beams using epoxy. This
setup is utilized to apply pure axial forces at both ends of DCB.
92 J.-H. Jay Kim et al.

Figure 8. Relationship between fc (nda )/fc and initial notch length (CS).

5. Analysis of test results

5.1. C YLINDRICAL SPECIMENS (CS) WITH AN INITIAL NOTCH

5.1.1. Test results

Table 5 tabulates the experimental data of ultimate strength fc with an initial notch length a0
and concrete compressive strength at 28 days fc . Table 6 tabulates the experimental data of
ultimate load Pu , fc , and tensile strength fct for various specimen sizes dc . The experimental
data of fc and fct are obtained from the φ100 × 200 mm cylinders where concrete from a
same batch is used to cast CS for size effect tests.

5.1.2. Strength variance with an initial notch length

Strength variance with an initial notch length is shown in Figure 8. The compressive strength
tests of cylindrical specimens with an initial notch show that there is a critical initial notch
length that changes the compressive strength of the specimens and this critical notch length
is 4da . Compressive strength for the specimens with an initial notch length less than 4da ,
the strengths remain constant for all specimens. Similar to this trend, when the initial notch
length is greater than or equal to 4da , the compressive strength decreases but still remain
constant for all specimens. It is important to note that when the loading is applied, a macro-
crack initiates at the crack tip and subsequently propagating to a failure. When the notch
length to maximum aggregate size ratio is smaller than 4.0, an additional load is required for a
macro-crack development and subsequently leading to failure when macro-crack propagates.
The lateral displacement at the middle portion of the initial notch is obtained using LVDT
and the result is shown in Figure 9. Normal and parallel direction displacements to the initial
notch are shown in Figures 9a and 9b, respectively.
This test is carried out to determine the adequate notch length, which enforces Mode I
crack opening mechanism leading to failure. When a notch length to maximum aggregate size
ratio is 4.0, the normal direction displacement to the initial notch is greater than the parallel
direction displacement. Therefore, the initial notch length to the maximum aggregate size ratio
of 4.0 is used for all size specimens. Based on the foregoing discussion, it is apparent that we
can enforce Mode I failure mechanism using this specimen and experimental procedure. For
 Size effect of concrete members applied with flexural compressive stresses 93

Figure 9. Lateral displacement length (CS): (a) Normal direction to initial notch; and (b) Parallel direction to
initial notch.

this reason, in the ‘Size effect with specimen size’ section, notch length to maximum aggregate
size ratios of 9.2 and 18.5 (i.e., 12.0 cm and 24.0 cm) in φ100×750 mm cylindrical specimens
are used.

5.1.3. Size effect with specimen size

Ultimate loads corresponding to various specimen sizes with an initial notch are tabulated
in Table 6. To differentiate size effect between the cases with and without an initial notch,
tests on specimens without an initial notch are also conducted. The experimental results
are included in Table 6. To confirm whether or not a size effect is present, MSEL is used.
LSM regression analyses (IMSL program; Benjamin and Cornell, 1970) are carried out for
94 J.-H. Jay Kim et al.
Table 6. Ultimate load obtained from experiments (CS).

Pu fc fct
(kN) (MPa) (MPa)

S (dc = 5 cm) M (dc = 10 cm) L (dc = 20 cm)

N0 N6 N12 N0 N12 N24 N0 N12 N24 N48

111 383 1491 49.1 5.1

96 353 1432 47.6 4.8

103
353 47.6 5.2
94
1452
96 373 51.9 5.5
1383
99
99 373 1364 1216 50.2 4.8
95
94
1266
97 373 49.7 4.8
1275
110
1118
110 373 47.8 5.0
1403
96 343
981 52.9 4.8
92 353

Table 7. Results obtained from MSEL and LSM (CS).

Specimen series B λ0 da α s∗

N6:N12:N24 0.92 2.6 0.48 0.0590

N12:N24:N48 0.90 2.6 0.44 0.1200
N0:N0:N0 0.52 2.6 0.80 0.0535
N12:N12:N12 0.72 2.6 0.63 0.0759

∗ s is the standard deviation.

maximum stress values obtained from the tests. The results are tabulated in Table 7. In this
table, s is the standard deviation. Figure 10 shows fc (dc )/fc as a function of the diameter
dc . For comparison, all results are shown together in Figure 11. In this figure, the data points
represent the mean value of the experimental results of the same size specimens.
As shown in Figure 10, the results indicate a strong size effect. In Figure 11, it is found
that size effect between specimens with an initial notch and without an initial notch N0 show
a considerable difference. Specimen N12 with a constant initial notch length regardless of
specimen size shows that a notch length to cylinder height ratio increases with decreasing
cylinder diameter and vice versa. Accordingly, the compressive strength at failure decreases as
the specimen diameter decreases. In addition, strength reduction level decreases for increasing
diameter. Specimens (N12, N24, and N48) with the largest notch length would show more
 Size effect of concrete members applied with flexural compressive stresses 95

Figure 10. Relationship between fc (dc )/fc and specimen diameter dc (CS).

Figure 11. Comparison of model equations obtained from MSEL and experimental results (CS).
96 J.-H. Jay Kim et al.
Table 8. Ultimate load obtained from experiments (DCB).

Pu fc fct
(kN) (MPa) (MPa)

S (D = 20 cm) M (D = 40 cm) L (D = 80 cm)

A1 A2 A3 A1 A2 A3 A1 A2 A3
E2 E1 E2 E3 E2 E2 E1 E2 E3 E2 E2 E1 E2 E3 E2

45 44 47
77 80 78 148 147 97 71.5 5.5
41 43 40
42 43
36 76 80 84 139 138 140 73.5 5.5
43 43
49 38 44
88 82 82 142 135 137 71.8 5.5
44 32 45
45
42 76 91 140 139 75.3 5.9
48
35
68 91 93 149 154 74.3 5.8
40
54
44 87 90 134 134 77.4 6.2
52

apparent reduction phenomenon than other specimens, which have smaller initial notch length
than this specimen. However, Kim et al. (2001), Markeset and Hillerborg (1995), and Jansen
and Shah (1997) experimentally showed that the strength reduction is independent of the
specimen size when the specimen height/diameter or length/depth is greater than a constant
value (i.e., 2.0–2.5 and 3.0 for cylinder and C-shaped specimens, respectively). In this study,
we concluded that the compressive strength decreases to a certain level with an increase in a
notch length. However, beyond that level, it is nearly consistent with the literature. Namely,
the comparison of fc (dc )/fc for specimens N12, N24, and N48 as well as N6, N12, and N24
shows similar difference.
When test results on specimen N0 without an initial notch are compared with Equation (5)
(Kim et al., 1999), it can be seen that compressive strengths of specimens with a diameter
greater than 10.0 cm show similar values as shown in Figure 11. However, when the diameter
is approximately 5.0 cm, the results are different since maximum aggregate size is differ-
ent. More specifically, Equation (5) is obtained from tests with maximum aggregate size of
approximately 25 mm. Even though Equationb(5) has been derived from the test results of
da = 25 mm, according to the published paper by Kim et al. (1999) in ACI, Equation (5) is
independent of da and therefore it can be used for any aggregate size. Therefore, the aggregate
size of 13 mm can be used for this study.
 Size effect of concrete members applied with flexural compressive stresses 97
Table 9. Results obtained from MSEL and LSM (eccentricity, DCB).

Specimen No. B λ0 da α s

A2-E1 – – – –
A2-E2 0.27 2.6 0.14 0.0201
A2-E3 0.34 2.6 0.06 0.0221

∗ s is the standard deviation.

Figure 12. Relationship between fc (e)/fc and specimen depth D (DCB).

Figure 13. Comparison of model equations obtained from MSEL and experimental results (eccentricity, DCB).
98 J.-H. Jay Kim et al.

5.2. D OUBLE CANTILEVER BEAMS (DCB) WITH AXIAL COMPRESSIVE LOADING

5.2.1. Test results

Table 8 tabulates the experimental data for ultimate axial loads of DCB for size effect tests
and fc and fct of φ100 × 200 mm concrete cylinders.
In Equation (1), ultimate stress in concrete σN (d) (i.e., fc (d)) is calculated from Equa-
tion (6) (Bazant and Kazemi, 1990) in which t and l are specimen thickness and characteristic
dimension of the specimen (in this study, l is the span length), respectively.
Pu Pu
fc (d) = = cn , (6)
tl td
d
where, cn = .
l
5.2.2. Size effect with eccentricity
To confirm size effect with eccentricity, LSM regression analyses are carried out with ultimate
stress values obtained from the tests. The new parameters for MSEL are given in Table 9.
Figure 12 shows fc (e)/fc as a function of the depth of specimen D. For comparison, all
results are shown together in Figure 13.
Test results revealed that the location of the axial compressive load on the end surface of
DCB specimens causes either flexural compressive stress or flexural tensile stress dominating
stress concentration at the regions near the notch tip. More specifically, as shown in Figure 4,
when the load is applied outside or inside of the cantilever axial axes, flexural compression
domineering stress or flexural tension domineering stress is applied on the crack tip region
in the direction normal to the initial notch direction, respectively. When the eccentric load
is applied inside the central axes of each cantilever beam, the effective flexural compressive
stress at the crack front increased more than when applied outside of the central axes. This
trend leads to negligible size effect for specimens loaded inside of the cantilever central axes
(specimens A2-E1). In Figures 12 and 13, this pattern could also be seen. The horizontal line
represents the average value of experimental data of specimen A2-E1. In the case of specimen
A2-E3, the results indicate a strong size effect condition. Based on the foregoing discussions
and as evident from the experimental results, reduction phenomena of strength occurs rapidly
to a limit for the variation in eccentricity, which has direct effect on the eccentricity size effect.
However, larger changes in eccentricity do not have great effect on the size effect.

5.2.3. Size effect with an initial notch

To confirm size effect with an initial notch length, the same procedure as in the foregoing
section is used to obtain the results tabulated in Table 10. Figure 14 shows fc (a0 )/fc as a
function of the beam depth D. For comparison, all results are shown together in Figure 15.
The ratio of notch length to half beam width (0.5 b) can be referred to as a slenderness
ratio. When a specimen is subjected to the same axial load but with a different slenderness
ratio, the failure would show more brittle behavior with increase in slenderness ratio, which
increases size effect of flexural compressive strength.
In this study, high-strength concrete (HSC) of 70 MPa is used. Generally, HSC shows
more brittle failure behavior than low-strength concrete (LSC) or medium-strength concrete
(MSC). This means that FPZ size of HSC is smaller and size effect is more apparent than LSC
and MSC. Accordingly, the size effect of specimens made with HSC is more distinct by the
smaller FPZ size.
 Size effect of concrete members applied with flexural compressive stresses 99

Table 10. Results obtained from MSEL and LSM (initial notch
length, DCB).

Specimen No. B λ0 da α s

A1-E2 0.28 2.6 0.15 0.0159

A2-E2 0.27 2.6 0.14 0.0201
A3-E2 0.26 2.6 0.13 0.0177

∗ s is the standard deviation.

Figure 14. Relationship between fc (a0 )/fc and specimen depth D (DCB).

Figure 15. Comparison of model equations obtained from MSEL and experimental results (initial crack length,
DCB).
100 J.-H. Jay Kim et al.

As demonstrated in Figure 14, size effect exists, but the difference in compressive strength
based on various notch lengths is not distinct. This trend is opposite to an expected result since
specimen slenderness ratios are increasing and specimens are made using HSC. It is safe to
assume that the axial compressive loading did not properly apply flexural compressive stress
at the crack front region due to the stiff flexural modulus of the specimens from using HSC.
The difficulty of properly applying flexural compressive stresses is reported previously (Kim
et al., 2001; Markeset and Hillerborg, 1995; Jansen and Shah, 1997). Therefore, in this study,
we concluded that the strength reduction is related to notch length and approached a limit with
an increasing notch length in DCB.

6. Conclusions

In order to systematically study flexural compressive strength size effect of concrete, an

experimental evaluation is performed using CS with an initial notch and DCB with axial
compressive loading instead of flexural loading. The following conclusions are drawn based
on the findings of this study:
(1) Size effect for the flexural compressive strength in modified CS and DCB exist. Namely,
size effect based on the specimen size difference is present in CS. Also, size effect based
on specimen size difference and eccentricity of loading point variations is present in DCB.
(2) For CS, if the notch length to maximum aggregate size ratio is greater than 4.0, a stable
failure took place and the Mode I failure was ensured.
(3) For DCB, size effect with an eccentricity is present and strongly influenced by the type
of flexural stresses applied at the crack tip region. When the eccentric load is applied
inside and outside of the cantilever axial axes, flexural compression and flexural tension
domineering stresses are applied at the crack tip region, respectively. Therefore, the size
effect is negligible and strong in flexural compression and flexural tension domineering
stress concentration conditions, respectively.
(4) For DCB with axial compressive loading, size effect is present. However, size effect based
on different notch lengths (or slenderness ratios) is not distinct. This trend is due to the fact
that the axial compressive loading did not properly applied flexural compressive stress at
the crack front region due to the stiff flexural modulus of the specimens made with HSC.
This ultimately means that the rate of stress concentration growth within FPZ based on
various types of stress combinations at the crack tip region has direct effect on size effect
of concrete members.
(5) The results suggest that the current strength criteria based design practice should be re-
vised. In designing concrete structures, it is more desirable to use compressive strength of
concrete obtained not from standardized specimens such as φ150 × 300 mm or φ100 ×
200 mm cylinders but from adequately sized specimens based on the size effect law.

Acknowledgements

The first author wishes to thank the partial financial support from Technology Development
Program for Agriculture and Forestry, Ministry of Agriculture and Forestry, Republic of
Korea. These supports are deeply appreciated. The third author wishes to thank the partial
support from the research funds of NRL program of Year 2000 (2000-N-NL-01-C-162) from
 Size effect of concrete members applied with flexural compressive stresses 101

Ministry of Science & Technology in Korea. The contributions of Min-Wook Kim and Bong-
Jun Kim are greatly appreciated. In particular, the authors are grateful to Professor Zdenek P.
Bazant for his assistance during the experimental program.

References

Bazant, Z.P. (1984). Size effect in blunt fracture: concrete, rock, metal. Journal of Engineering Mechanics, ASCE
110(4), 518–535.
Bazant, Z.P. (1987). Fracture energy of heterogeneous material and similitude. SEM-RILEM International
Conference on Fracture of Concrete and Rock 390–402.
Bazant, Z.P. (1993). Size effect in tensile and compressive quasibrittle failures. JCI International Workshop on
Size Effect in Concrete Structures, 141–160.
Bazant, Z.P. and Kazemi, M.T. (1990). Determination of fracture energy, process zone length and brittleness
number from size effect, with application to rock and concrete. International Journal of Fracture 44, 111–131.
Bazant, Z.P. and Xiang, Y. (1994). Compression Failure of Quasibrittle Materials and Size Effect. AMD Symposium
of the American Society of Mechanical Engineers’ Applied Mechanics Division, 185, Damage Mechanics in
Composites, ASME Winter Annual Meeting, Chicago, Ill., Allen, D.H. and Ju, J.W., eds., 143–148.
Bazant, Z.P. and Xiang, Y. (1997). Size effect in compression fracture: splitting crack band propagation. Journal
of Engineering Mechanics, ASCE 123(2), 162–172.
Benjamin, J.R. and Cornell, C.A. (1970). Probability, Statistics, and Decision for Civil Engineers. McGraw-Hill,
New York, Section 4.3, 466–475.
Biot, M.A. (1965). Mechanics of Incremental Deformations. John Wiley & Sons, New York.
Blanks, R.F. and McNamara, C.C. (1935). Mass Concrete Tests in Large Cylinders. Journal of ACI, Presentation
at the 31th Annual Convention, New York, 280–303.
Carpinteri, A., Chiaia, B. and Ferro, G. (1994). Multifractral Scaling Law for the Nominal Strength Variation of
Concrete Structures. Size Effect in Concrete Structures, E &FN Spon, 409–416.
Cotterell, B. (1972). Brittle fracture in compression. International Journal of Fracture 8(2), 195–208.
Gonnerman, H.F. (1925). Effect of size and shape of test specimen on compressive strength of concrete.
Proceedings, ASTM 25, 237–250.
IMSL, Library, Edition 8, IMSL, Inc.
Jenq, Y.S. and Shah, S.P. (1991). Features of mechanics of quasi-brittle crack propagation in concrete.
International Journal of Fracture 51, 103–120.
Jansen, D.C. and Shah, S.P. (1997). Effect of length on compressive strain softening of concrete. Journal of
Engineering Mechanics, ASCE 123(1), 25–35.
Karihaloo, B.L. (1992). Failure modes of longitudinally reinforced beams. In: Application of Fracture Mechanics
to Reinforced Concrete (Edited by Carpinteri, A.), Elsevier Applied Science, London, 523–546.
Kim, J.K. and Eo, S.H. (1990). Size effect in concrete specimens with dissimilar initial cracks. Magazine of
Concrete Research 42(153), 233–238.
Kim, J.K., Eo, S.H. and Park, H.K. (1989). Size Effect in Concrete Structures without Initial Crack. Fracture
Mechanics: Application to Concrete, SP-118, ACI, Detroit, 179–196.
Kim, J.K., Yi, S.T. and Kim, J.H.J. (2001). Effect of specimen sizes on flexural compressive strength of concrete.
ACI Structural Journal 98(3), 416–424.
Kim, J.K., Yi, S.T., Park, C.K. and Eo, S.H. (1999). Size effect on compressive strength of plain and spirally
reinforced concrete cylinders. ACI Structural Journal 96(1), 88–94.
Kim, J.K., Yi, S.T. and Yang, E.I. (2000). Size effect on flexural compressive strength of concrete specimens. ACI
Structural Journal 97(2), 291–296.
Markeset, G. and Hillerborg, A. (1995). Softening of concrete in compression localization and size effects. Cement
and Concrete Research 25(4), 702–708.
Nemat-Nasser, S. and Obata, M. (1988). A microcrack model of dilatancy in brittle material. Journal of Applied
Mechanics, ASME 55, 24–35.
Neville, A.M. (1956). The influence of size of concrete test cubes on mean strength and standard deviation.
Magazine of Concrete Research 8(23), 101–110.
Neville, A.M. (1966). A general relation for strength of concrete specimens of different shape and size. Journal of
ACI 63, 1095–1110.
 102 J.-H. Jay Kim et al.

Powers, T.C. (1956). Concrete Studies at the Bull Run Dam. City of Portland, Oregon, U.S. Department of the
Interior, Bureau of Concrete Research.
Ruettgers, A. (1933). Mass concrete as affected by size of aggregate and related factors. Journal of ACI,
Proceedings 30, 27–34.
Shah, S.P. and McGarry F.J. (1971). Griffith fracture criterion and concrete. Journal of Engineering Mechanics,
ASCE 97, 1663–1676.
Yi, S.T., Kim, J.H.J. and Kim, J.K. (2002). Effect of specimen sizes on aci rectangular stress block for concrete
flexural members. ACI Structural Journal 99(5), 701–708.

View publication stats