ACI STRUCTURAL JOURNAL TECHNICAL PAPER

Title no. 99-S72

Effect of Specimen Sizes on ACI Rectangular Stress Block

for Concrete Flexural Members
by Seong-Tae Yi, Jang-Ho Jay Kim, and Jin-Keun Kim

It is important to consider the effect of concrete member sizes

when estimating the ACI rectangular stress block of a concrete
flexural member. The experimental data and analytical results,
however, are still insufficient for a proper evaluation. For all types
of loading conditions, the trend is that the size of the ACI rectangular
stress block tends to change when the member size changes.
In this paper, the size variations of strength coefficients for ACI
rectangular stress block (α1 and β1) and the stress block parameters
(k3, k1k3, and k2) have been studied. The results from a series of
C-shaped specimens subjected to axial compressive load and
bending moment were adopted from other references.
The analysis results show the effect of specimen sizes on the
strength coefficients for ACI rectangular stress block and the stress Fig. 1—Conditions at ultimate load.
block parameters of a concrete member to be apparent. They also
show that the current strength criteria-based design practice
should be reviewed to include member size effect. More studies are
needed in the future to verify that.

Keywords: flexural strength; specimen; stress.

INTRODUCTION
The actual compressive stress distribution in the compressive
zone of concrete flexural members is extremely difficult to
measure and to adequately model. Ignoring the tension stress
carried by concrete, the stress and strain distributions in a re-
inforced concrete (RC) beam section when the compressive
concrete strain has reached 0.003 are shown in Fig. 1.
Koenen1 was the first to propose the theory of ultimate failure
capacity of flexural members. He assumed that the stress
distribution in the cross section of RC beam is linear and is Fig. 2—C-shaped specimen.
uniform across the width of the cross section. Following
Koenen’s proposal, various stress distribution shapes in the
compressive zone of RC beams have been suggested. Among Several researchers4-6 experimentally studied the flexural
the suggested stress distributions, the equivalent rectangular stress distribution in the compressive zone of RC members
stress block was found to be the most practical and simplest and showed that the compressive stress during flexural loading
model, with a satisfactory accuracy for design purposes. The increases until the maximum stress is reached and decreases
theory of equivalent rectangular stress distribution was first afterwards in the USD approach. Several researchers4-6
proposed by Emperger 2 and modified by Whitney 3 for further simplified their model by assigning the strength co-
application to Ultimate Strength Design (USD). efficients (that is, α1, β1, k3, k1 k3, and k2) for application to
design codes and practical usage.
To obtain accurate and well-controlled data on flexure-
Concrete as a quasibrittle material fails ultimately by the
compression-loaded members, a test procedure for a series of
formation and propagation of cracks induced by stresses
experiments on C-shaped concrete specimens (Fig. 2) subject-
caused by external loads or environment changes resulting in
ed to axial load and bending moment was proposed by sev-
the release of internal energy.8-12 Therefore, based on the
eral researchers.4-6 The position of neutral axis depth c was
energy concept, there is an effect of size on the nominal
kept fixed by continuously monitoring strains on one surface
strength of specimens made with quasibrittle materials such
of the C-shaped specimen and adjusting the eccentricity of
as concrete, rock, ice, ceramic, and composite materials. In a
the applied force so that the strains on the neutral surface re-
main zero. This test procedure was developed by the Port-
land Cement Association (PCA)4,5 and reported by several ACI Structural Journal, V. 99, No. 5, September-October 2002.
researchers.4-6 The results of tests carried out using this pro- MS No. 01-381 received November 6, 2001, and reviewed under Institute publica-
tion policies. Copyright © 2002, American Concrete Institute. All rights reserved,
cedure formed the basis of the rectangular stress block used including the making of copies unless permission is obtained from the copyright
proprietors. Pertinent discussion will be published in the July-August 2003 ACI Structural
in ACI 318 code.7 Journal if received by March 1, 2003.

ACI Structural Journal/September-October 2002 701

 testing of beams, there is little evidence that there exists a
Seong-Tae Yi is a senior engineer with Korea Power Engineering Company, Inc.
(KOPEC), Yongin, Korea. He received his MS and PhD from the Korea Advanced size effect in flexural member. However, the equivalent rectan-
Institute of Science and Technology, Daejon, Korea. His research interests include gular stress block coefficients are obtained based on the
fracture mechanics, size effect, early-age concrete, permeability, and behavior of flexural compression testing of C-shaped specimens. There-
concrete structures.
fore, there is a clear size effect in flexure compression when
Jang-Ho Jay Kim is an assistant professor in the Department of Civil and Environmental a C-shaped specimen is used rather than a normal beam
Engineering at Sejong University, Seoul, Korea. He received his BS from the University specimen. The reason for this trend is due to the precise con-
of California-Los Angeles, his MS from the University of California-Berkeley, and his
PhD from Northwestern University, Evanston, Ill. His research interests include trol of the neutral axis depth during the flexural testing in C-
numerical concrete structure analysis, concrete material mechanics, and nonlinear shaped specimens, whereas the neutral axis depth of a nor-
fracture mechanics.
mal beam specimen cannot be clearly specified.
ACI member Jin-Keun Kim is a professor in the Department of Civil and Environmental The purpose of this study is to investigate the effect of
Engineering, Korea Advanced Institute of Science and Technology. He received his BS member sizes on the generally used ACI rectangular stress
and MS from Seoul National University, Seoul, Korea, and his PhD from North-
western University. His research interests include the application of fracture mechanics block for structural designs and to investigate the effect of actual
to concrete structures, and inelastic analysis of concrete structures. compressive stress distributions on strength coefficients by
comparing the analysis results, which were calculated using the
practical sense, the nominal strengths of laboratory-size experimental data, with the ACI 318 design code.
specimens will differ from that of larger structural members
used in construction of real structures. Ba ant13 derived a RESEARCH SIGNIFICANCE
size effect law (SEL) from dimensional analysis and similitude Based on References 23 and 24, specimen size dependencies
arguments for geometrically similar structures of different sizes are shown in the ultimate strength and stress-strain curves of
with initial cracks, considering the energy balance at crack RC beams. As the specimen size increases, the ultimate
propagation in concrete. Kim and Eo14 and Kim, Eo, and strength value decreases. Therefore, an equivalent rectangular
Park15 proposed the modified size effect law (MSEL) by stress block and a stress distribution in the compressive zone
adding the size independent strength σo (= αft′ to the size used in the ACI 318 Code can vary depending on the specimen
efect law, which was also proposed by Ba ant16-17 and size. The focus of this research is to review the effect of
Ba ant and Xiang18in a different approach. specimen size dependency on the equivalent rectangular
Researches have focused more on pure tension and shear stress block and the stress distribution defined in ACI 318
loading conditions. Only recently, the studies19-21 on pure Code, which are vital design criteria for practicing structural
compression loading based size effect became a focus of engineers and designers.
interest among researchers. Gonnerman 22 experimentally
showed that, under pure compression loading, the ratio of REVIEW OF ACI 318 CODE AND
the compressive failure stress to the compressive strength STRENGTH COFFICIENTS
decreases as the specimen size increases. Also, the In many countries, including the United States and Korea,
studies23,24 have shown that, under flexural compression the design of an RC flexural member uses an equivalent rect-
loading, the failure strengths decrease as the sizes of the angular stress block as the compressive stress distribution.
concrete specimens increase. The ACI rectangular stress distribution shown in Fig. 1 is
A lack of studies that relate flexural compression loading defined as follows. The coefficient α1 is assumed to have a
of various size concrete beams to the compressive stress constant value of 0.85. A concrete stress of 0.85 fc ′ is assumed
profiles in a concrete compressive zone has forced designers as uniformly distributed over an equivalent compression zone
to use the compressive strength from a pure axial compres- bounded by edges of the cross section and a straight line located
sion test of standard cylindrical concrete specimens on the parallel to the neutral axis at a distance a = β1c from the fiber of
compressive zone of a flexural member design. Also, this maximum compressive strain. The distance c from the fiber of
simplified practice of design has been omnipresent because maximum strain to the neutral axis shall be measured in a
the stress-strain curves from pure compression of cylindrical direction perpendicular to that axis. In the ACI 318 Code, the
specimens and flexure-compression tests are similar until the factor β1 is taken as 0.85 for concrete strengths fc′ up to and
maximum compressive strength fc′ is reached. The relationship, including 4000 psi (27.6 MPa). For strengths above 4000 psi, β1
however, is significantly different after the peak load. shall be reduced continuously at a rate 0.05 for each 1000 psi
Presently, most design codes for concrete structures do not (6.9 MPa) of strength in excess of 4000 psi. However, in any
consider size effect. Because quasibrittle materials fail from case, β1 shall not be taken less than 0.65.
the formation of cracks, size effect has to be implemented. In In Fig. 1, the softening branch of stress distribution in the
compressive failure of quasibrittle materials, the size effect compressive zone of the cross section (shown schematically)
is quite apparent. Though the behavior of compressive failure is used as a material property. Parameters k3, k1k3, and k2 have
has been studied extensively, the flexural compressive been used in the strength-based design method to account for
behavior and its size effect have been insufficiently studied the shape of the compressive stress-strain diagram. However,
when compared to the tensile failure mechanism. Although experimental results by van Mier29 in uniaxially loaded spec-
the study of the size effect in compressive failures has not imens have indicated that the softening portion of the com-
been widely pursued, a few researchers (Kim, Yi, and Eo25 and pressive stress-strain curves depends on the length of
Kim et al.26) have made progress. Kim et al. showed that the ul- specimens. As the length of the compressive specimen in-
timate strength is directly affected by the size of specimens. creases, the slope of the decreasing or the softening branch of
Therefore, the size of compression stress block used in a design stress-strain curves becomes steeper.29 Alternatively, if stress
will be significantly affected as the size of the specimens varies. is plotted against deformation for the postpeak behavior using
However, Corley27 and Alca, Alexander, and MacGregor28 the same test results, basically the same postpeak stress-dis-
rejected the hypothesis that there is a size effect in flexure placement curves are obtained. Therefore, the stress-deforma-
based on the experimental results. From the normal flexural tion curve, instead of the stress-strain curve, should be used to

702 ACI Structural Journal/September-October 2002

Fig. 3—Representative stress-strain curve.23

describe the softening behavior of concrete. As a result, values

of parameters k3, k1k3, and k2 depend not only on the strength
value fc′ 4, but also on the size of concrete compressive zone of
a beam. Clearly, these parameters are related to the compres-
sive stress-deformation response of the concrete. The reason
behind this behavior is that tensile failures in concrete are
caused by microcrack formation and growth, and eventually
coalescence into macrocracks. More specifically, the behavior
of a concrete specimen under compression loading cannot be
captured by its strain data because the strain values are expres-
sions of an averaged compressive deformation of a continuous
specimen. Therefore, a more physical and realistic behavior
that represents localized failures of concrete with cracks can
be expressed by its deformation values. Basically, the size ef-
fect law of Ba ant was derived under tension stress condition,
but compressive failure is also related to splitting cracks due
to localized tension effect that causes the ultimate failure of
the concrete specimens. Recently, Ibrahim and MacGregor30
reported that α1 and β1 coefficients are dependent on the con-
crete compressive strength. This phenomenon can be modeled
using methods of fracture mechanics. Fig. 4—Descriptions for various size effect specimens.
The ACI 318 Code states that an equivalent rectangular stress. However, the postpeak range of the compressive stress-
stress block coefficient α1 is a constant value, but the coeffi- strain curves shows that they are significantly different. This
cient β1 changes based on concrete compressive strength. trend clearly suggests that member size differences will affect
Tests on C-shaped specimen under linear strain the stress block parameters, especially k3 value.
distribution23,24 show that it is reasonable to assume that the
coefficient α1 will change when the specimen size changes. In
DISCUSSIONS OF EFFECT OF SIZE DIFFERENCES
particular, when the length-to-depth ratio h/c (Fig. 2) is greater ON ACI STRENGTH COEFFICIENTS
than or equal to 3.0 (h/c ≥ 3.0),24 even though the failure The relationship between the effect of size differences in
strength does not change, the coefficient α1 will change, indi- compression tests to ACI equivalent stress block size and
cating a modification in equivalent stress block size. actual stress distribution are discussed. The discussion is
Figure 3 shows representative compression stress-strain re- based on the experimental data of size effect (size, length,
lationships that consider specimen size difference where the and depth variations) of C-shaped flexure compression test
thin and the thick solid lines are the flexural compressive reported by Kim, Yi, and Yang,23 and Kim, Yi, and Kim.24
stress-strain curves from C-shaped specimens, and the uniaxial The specimens used for the size, length, and depth effect
compressive stress-strain curve obtained from standard con- experiments are shown in Fig. 4. For size effect, the main
crete cylinder tests, respectively. Also, the roman numerals I, test variable was a size ratio of 1:2:4 of the specimen. The
II, and III represent the size of the specimens, with I being the height h and the depth c were changed proportionally
smallest and increasing accordingly. The figure shows that the (Fig. 4(a)). Specimen length:depth ratios of 1:1, 2:1, 3:1, and
maximum stress and the corresponding strain value increase as 4:1 were used to study the effect of length where a constant
the specimen size becomes small. Therefore, it is suspected depth (c = 10 cm) was maintained and specimen lengths
that there is a significant influence of member size in ACI were varied from 10 to 20 to 30 to 40 cm (Fig. 4(b)).
stress block coefficients and size effect must be introduced in Specimen length:depth ratios of 1:1, 2:1, and 4:1 were used to
the calculation of these coefficients. ACI 318 Code takes α1 as study the effect of depth where a constant length (h = 20 cm) was
a constant equal to 0.85. Based on the comparison of the rep- maintained and specimen depths were varied from 5 to 10
resentative experimental data of compressive stress-strain rela- to 20 cm (Fig. 4(c)). The thickness of all specimens was kept
tionships of C-shaped and cylindrical specimens, it shows that constant (b = 12.5 cm) to eliminate the out-of-plane size effect.
the curves for the two cases are nearly identical up to the peak The specimen thickness b was chosen to allow stable failure.

ACI Structural Journal/September-October 2002 703

Table 1—Test results for size, length, and depth effect
P u ,kN
No. of
specimen h, cm c, cm fc′, MPa P1 P2 β 1, ACI e cu , cm α1 β1 k3 k1k3 k2
S-I-1 272 16 1.0478 0.5618 1.1488 0.7753 1.2227 0.8906 0.3630
S-I-2 264 13 1.0061 0.5715 1.1086 0.7714 1.1920 0.8552 0.3804
S-I-3 286 15 1.0923 0.5304 1.1785 0.7878 1.2570 0.9285 0.3727
S-I-4 266 20 1.0403 0.6122 1.1710 0.7551 1.2906 0.8842 0.3244
S-I-5 10 5 265 14 1.0128 0.6335 1.1531 0.7466 † 0.8609 0.3768
S-I-6 270 17 1.0414 0.5725 1.1481 0.7710 1.2205 0.8852 0.3554
S-I-7* — — — — — — — — —
S-I-8 260 16 1.0046 0.5171 1.0766 0.7932 1.1476 0.8539 0.3518
S-I-9 278 15 1.0645 0.5340 1.1506 0.7864 1.2178 0.9048 0.3694
S-II-1 454 21 0.8627 1.2298 0.9725 0.7540 0.9808 0.7333 0.3874
S-II-2 51.80 469 17 0.8829 1.1763 0.9813 0.7647 1.0198 0.7505 0.4117
S-II-3 482 16 0.9043 1.1018 0.9859 0.7796 1.0252 0.7686 0.4187
20 10
S-II-4 480 17 0.9032 1.3445 1.0501 0.7311 1.1288 0.7677 0.4122
S-II-5 490 25 0.9360 1.3621 1.0933 0.7276 1.0825 0.7955 0.3800
S-II-6 460 21 0.8742 * * * 1.0651 0.7432 0.3874
S-III-1 861 42 0.8202 2.5898 0.9409 0.7410 0.9976 0.6972 0.3724
S-III-2 907 37 0.8575 2.0478 0.9166 0.7952 0.9761 0.7289 0.3922
S-III-3 867 42 0.8259 2.6485 0.9550 0.7352 1.0829 0.7020 0.3736
40 20
S-III-4 891 39 0.8447 2.4787 0.9547 0.7521 1.0255 0.7180 0.3845
S-III-5 913 45 0.8707 2.4093 0.9749 0.7591 1.0839 0.7400 0.3719
S-III-6 868 49 0.8330 2.7519 0.9768 0.7248 1.1715 0.7080 0.3529
L-I-1 571 35 0.9850 0.6852 0.9703 0.8630 † 0.8373 0.3551
L-I-2 10 572 31 0.9798 0.5859 0.9434 0.8828 1.3000 0.8328 0.3730
L-I-3* — — — — — — — — —
L-II-1 546 30 0.9370 0.9815 0.9910 0.8037 1.1253 0.7965 0.3706
L-II-2 20 562 23 0.9514 1.0410 1.0213 0.7918 1.1991 0.8087 0.4002
L-II-3 530 22 0.8971 1.0220 0.9585 0.7956 1.1211 0.7626 0.4009
10 57.88
L-III-1 502 37 0.8761 1.3910 1.0317 0.7218 1.1489 0.7447 0.3284
L-III-2 30 532 22 0.9010 1.3050 1.0363 0.7390 1.0655 0.7658 0.4022
L-III-3 515 23 0.8747 1.4050 1.0340 0.7190 1.0857 0.7435 0.3915
L-IV-1 515 24 0.8763 1.5150 1.0686 0.6970 1.1771 0.7448 0.3871
L-IV-2 40 526 24 0.8949 1.6270 1.1276 0.6746 † 0.7607 0.3895
L-IV-3 517 18 0.8697 1.6220 1.0942 0.6756 1.0657 0.7393 0.4170
D-I-1 278 5 0.9612 * * * 1.2088 0.8170 0.4028
D-I-2 5 270 2 0.9245 0.3147 0.8990 0.8741 1.0959 0.7858 0.4544
D-I-3 275 4 0.9492 0.3915 0.9566 0.8434 1.2380 0.8068 0.4131
D-II-1 502 29 0.9017 1.5730 1.1182 0.6854 1.0675 0.7664 0.3610
D-II-2 20 10 55.43 508 24 0.9030 1.2530 1.0243 0.7494 1.0905 0.7676 0.3889
D-II-3 494 24 0.8797 1.2770 1.0042 0.7446 1.1807 0.7478 0.3850
D-III-1† — — — — — — — — —
D-III-2 20 932 74 0.8540 4.1560 1.2421 0.5844 1.0636 0.7259 0.4263
D-III-3 919 85 0.8521 4.7170 1.3709 0.5283 1.0186 0.7242 0.4158
*S-I-7 and L-I-3 failed during testing due to failure of two end sections; D-III-1 failed during testing due to failure of rod that applies load P2.
* †
and : unable to obtain proper stress-strain curve ecu and fmax.
Note: All specimens had b = 12.5 mm.

The test results for size, length, and depth effect are shown MacGregor30 model equations of β1 values. By comparing
in Table 1. Also, the numbering of the specimen (that is, β1 values, the effect of specimen size dependence on β1 values
L-I-1) and obtained data are tabulated in Table 1. The is indicated. The other coefficient used to define the stress
specimens for size, length, and depth effect are assigned block was α1 which the ACI code takes constant at 0.85. The
with S, L, and D in the specimen names, respectively. Also, following sections will describe the specimen size effect on
the roman numerals I, II, III, and IV represent the size of the the strength coefficients α1, k3, k1k3, and k2.
specimens, with I being the smallest and increasing accordingly.
The arabic numbers 1, 2, 3 are the three specimens tested for Comparison of test results with reference data31
each specimen size. on β1
The data are compared with the experimental data reported A specimen numbered III in the test23 is exactly same as the
by Nilson and Slate31 to verify the accuracy. Also, the data specimen size used to derive the β1 value in the ACI 318 Code.
are used to check with ACI 318 code and Ibrahim and The ACI equation for β1 shows that when the compressive

704 ACI Structural Journal/September-October 2002

 Fig. 5—Variations of β1 values suggested by Reference 30 and ACI with specimen sizes.

Fig. 6—Variations of α1 values with specimen sizes.

strength fc′ is 51.8 MPa, β1 is equal to approximately 0.68. The results from Eq. (2), respectively. It is important to note that
values recommended for design are based on a lower-bound β1, test values in Eq. (1) are calculated using a constant α1
limit, which is not always safe. However, Reference 31 sug- value of 0.85. The results show that β1 values decrease as
gests that the ACI value of β1 is extremely conservative. When specimen sizes increase. The calculated values, however, are
the experimental data of Size III specimens are used to calcu- still higher than the test values based on a α1 value of 0.85.
late the β1 value, β1 is equal to approximately 0.83. It should The ACI approach of using an equivalent rectangular stress
be emphasized that this calculated value is obtained using the block width of 0.85fc′ for concrete compressive strengths
polynomial equation, which is not recommended for design. of 51.8, 57.88, and 55.43 MPa gives constant β1 values of
Also, if the ultimate load values Pu23 are substituted into Eq. 0.68, 0.65, and 0.65, respectively.
(1) for simple calculations, the β1 values are calculated as 0.82, The equation suggested by Ibrahim and MacGregor30 pre-
0.86, 0.83, 0.84, 0.87, and 0.83. Comparing these values with dicts a more accurate value to the experimental data, especially
Reference 31’s suggested value of 0.83, it can be seen that the when the specimen size is increased. Figure 5 shows that the
difference is minute and insignificant. Therefore, the result calculated β1 values indicate an effect of member size rather
shows that the accuracy of the experimental data is nearly than the constant value assumed in the ACI method. This clear
equivalent to the data used by the ACI 318 Code. size-effect trend is probably due to the member size difference
as well as the aggregate size relative difference to the specimen
Ratio of average compressive stress to maximum size used to obtain the experimental data. Therefore, a more de-
stress β1 based on α 1 value of 0.85 tailed analysis of calculating β1 values should be performed.
Figure 5 shows the relationship between specimen sizes
and β1 obtained using Eq. (1) and (2), which are based on a
α1 value of 0.85 and suggested by Ibrahim and MacGregor,30 Strength coefficients α 1 and β1 for rectangular
respectively. stress block
In this section, the effect of member size difference on the
Pu calculated values of the equivalent rectangular stress block
β 1, test = ----------------------------------
- (1) coefficients α1 and β1 is presented. ecu is the eccentricity
0.85f ′c × b × d from the centroid of the cross section above the neutral axis
due to the resultant load parallel to the member axis. ecu is
f c′ obtained using the Levenberg-Marquardt’s Least Square
β 1, ref. 30 = 0.95 – --------
- ≥ 0.70 where f ′c in MPa (2)
400 Method (LSM) regression analyses based on a cubic equation
fc = A1 + A2 εc + A3 εc2 + A4 εc3 to obtain the stress fc and
In Eq. (1), the thickness b represents the value 12.5 of the strain εc relationship on the compression zone for size,
specimen, in cm. Equation (1) is based on the calculation length, and depth effects of C-shaped specimens. The coeffi-
methodology to obtain a β1 value following the method cients α 1 and β1 are calculated based on the previously
suggested by the ACI 318 Code. In Fig. 5, the hollow circular obtained values of ecu using Eq. (3) and (4), respectively.
data points, the thick dashed line, and the thin solid line Equation (3) is formulated based on the assumption that the
represent the experimental data, the results from Eq. (1), and external load Pu is equal to the internal force of the equivalent

ACI Structural Journal/September-October 2002 705

 Fig. 7—Variations of β1 values with specimen sizes.

rectangular stress block. Equation (4) is formulated based on the The k3, k1k3, and k2 parameters are calculated using Eq. (5),
depth of the equivalent stress block. The calculated values of α1 (6), and (7), respectively. The calculated values for k3, k1k3,
and β1 for various member sizes (that is, size, length, and depth) and k2 parameters for various specimen sizes are shown in
are shown in Fig. 6 and 7, respectively. In Fig. 6 and 7, the hol- Fig. 8, 9, and 10, respectively. In these figures, the hollow
low circular data points represent the values calculated using the circular data points represent the values calculated from the
experimental data with Eq. (3) and (4); the thick dashed lines experimental data with Eq. (5), (6), and (7); the thick dashed
represent α1 value of 0.85 and β1 values of 0.68, 0.65, and 0.65, lines represent k3, k1k3, and k2 equal to 0.85, 0.85 β1, and
calculated using the ACI 318 Code with α1 equal to 0.85 for β1/2 calculated using the ACI 318 Code, respectively.
size, length, and depth effect specimens, respectively. The parameter k3, the ratio of maximum compression in
the beam to the cylinder strength of the concrete, is found by
Pu evaluating the following equation
α 1 = ---------------------------------------------------------
- (3)
( c ⁄ 2 – e cu ) × 2 × f c′ × b
k 3 = f max ⁄ f ′c (5)
( c ⁄ 2 – e cu ) × 2
β 1 = ------------------------------------
- (4)
c where fmax is the maximum compressive stress obtained
from the stress-strain relationship, and fc′ is the uniaxial
The change of α1 and β1 values with respect to three types compressive strength of a standard concrete cylinder. Figure 8
of size effect (size, length, and depth) are as follows. In shows the variation of k3 depending on the specimen sizes.
Fig. 6, the α1 values, calculated using the experimental data, Figure 8 shows that k3 decreases as the member size increases.
decrease as member size and depth increase. However, the Herein, k3 characterizes the size effect23,24 and is equal
α1 values increase when the member lengths increase. Also, to σN(c)/fc′ . If k3 has a value of unity, then k1 becomes the
Fig. 7 shows that the calculated β1 values are not sensitive to ratio of average compressive stress in the beam to the cylin-
the size and depth increase, but decreases when the member der strength fc′ .
length increases. Based on the characteristic of α1 and β1 To calculate k1, which relates the average to the maximum
values from the analysis results, it can be safely accepted stress in the beam, one must first evaluate the product of k1
that the trend that α1 and β1 values increase and decrease, and k3. By the equilibrium of forces
respectively, as the length of members increases. The reason
behind these trends is due to the concentration of stress as the Pu P1 + P2
member length increases. More specifically, the flexure- k 1 k 3 = ----------- = -----------------
- (6)
compression failure behaviors of longer and shorter specimens bcf ′c bcf ′c
are different. For shorter specimens, the flexure-compres-
sion failure occurred throughout the cross section of the In Fig. 9, k1k3 decreases as member size increases. The
specimen, whereas the flexure-compression failure of longer obvious reason is that the k1k3 value is strongly influenced
specimens initiated at a more localized zone of the cross by the value k3, a parameter, which characterizes the size
section above the neutral axis in the compressive zone. This effect. From Fig. 9, however, the k1k3 values are still larger
means that the flexure-compressive stress at the compressive than the ACI suggested values of 0.85 β1. To observe the
zone of the specimens became more localized in the case of independent characteristic of the k1 value, the experimentally
a longer specimen than in a shorter specimen. Therefore, the obtained k1k3 value can be divided by k3. The calculated
trend is represented by the width and depth of the equivalent value of k1 using the experimental data shows that k3 is a
rectangular stress block’s strength coefficients α1 and β1 domineering parameter when compared to k1. Therefore, it is
increasing and decreasing, respectively, as the specimen length safe to assume that k1 plays a minor role in the calculated
increases. This trend is reasonable where the dependence of β1 value of k1k3.
on α1 is intimate; therefore they would show inversely propor- The parameter k2 establishes the depth of the compressive
tional characteristics. resultant relative to the neutral axis depth. Again by equilib-
rium of moments
Stress block parameters k3, k1k3, and k2 for actual
stress distribution P1 a1 + P2 a2
In this section, the effect of member size on the effective k 2 = 1 – ----------------------------- (7)
stress distribution’s parameters k3, k1k3, and k2 is presented. ( P1 + P2 ) c

706 ACI Structural Journal/September-October 2002

 Fig. 8—Variations of k3 values with specimen sizes.

Fig. 9—Variations of k1k3 values with specimen sizes.

Fig. 10—Variations of k2 values with specimen sizes.

where the values a1 and a2 include both the initial values and 0.85fc′ is used as suggested by ACI, a distinct change in the
those due to deflection of the specimen. As shown in Fig. 10, β1 coefficient has been observed. If the strength coefficients
the value of k2 is not influenced greatly by the specimen size α1, β1, k3, k1k3, and k2 are calculated based on member size
differences. When the specimen size increases, the speci- differences, the strength coefficients sometimes showed a
men tends to show a brittle type of stress-strain characteristic. In significant influence of size effect, and at other times, showed no
addition, the brittle characteristic is dictated by the postpeak influence. The coefficient values, however, are always larger
behavior of the stress-strain (Fig. 3). However, k2 is a parameter than the values obtained without considering size effect.
that is significantly affected by concrete strength and the From the analyses, the following conclusions are drawn:
prepeak characteristic of the stress-strain curve. Therefore, 1. The sizes of the ACI rectangular stress block and the
k2 is not greatly effected by specimen size differences because actual stress distribution at failure should change as the sizes
the strengths of the concrete used for the experiments are similar. of the specimens change. The effect of specimen size on the
The k2 value, however, is still larger than the ACI-used k2 value strength coefficients is significant in the case of small-size
of β1/2. The ACI strength coefficients are still conservative specimens (Size I). In the case of large-size specimens
even though they do not consider the size effect. (Size III), however, the effect of specimen size on the
strength coefficients is minor and similar to ACI stress block.
CONCLUSIONS Because a design code must be able to consider all types of
The influence of member size differences on the strength specimen sizes, the size effect must be introduced into the cal-
coefficients for ACI rectangular stress block and the stress culation of ACI rectangular stress block coefficient;
block parameters presented in this paper is based on the 2. The stress block parameters that are influenced by the
flexural compressive strength experimental data (specimen stress-strain curve characteristic and the postpeak behavior
size, length, and depth effect) published previously in the are strongly dependent on member size differences. However,
ACI Structural Journal.23,24 When the stress block width of the parameter k2, which is influenced by the prepeak behavior

ACI Structural Journal/September-October 2002 707

and the strength (peak) values are not dependent on speci- 7. ACI Committee 318-99, “Building Code Requirements for Structural
men size differences; and Concrete (ACI 318-99) and Commentary (318R-99),” American Concrete
Institute, Farmington Hills, Mich., 1999, 391 pp.
3. The results suggest that further studies are required to 8. Barenblatt, G. I., “The Formation of Equilibrium Cracks during Brittle
determine the effect of member size differences on ACI Fracture—General Ideas and Hypotheses, Axially Symmetric Cracks,” Jour-
rectangular stress block. The current strength-based criteria nal of Applied Mathematics and Mechanics, V. 23, No. 3, 1959, pp. 622-636.
should be reviewed. 9. Hillerborg, A., “The Theoretical Basis of a Method to Determine the Frac-
ture Energy GF of Concrete,” Materials and Structures, V. 18, 1985, pp. 291-296.
10. Hillerborg, A., “Numerical Methods to Simulate Softening and
ACKNOWLEDGMENTS Fracture of Concrete,” Fracture Mechanics of Concrete: Structural Appli-
The authors would like to thank the Korea Institute of Science and Technol-
cation and Numerical Calculation, G. C. Sih and A. DiTomasso, eds., Mar-
ogy Evaluation and Planning (KISTEP) for the financial support of the
tinus Nijoff, Dordrecht, 1985, pp. 141-170.
National Research Laboratory (NRL). The second author also wishes to
11. Rice, J. R., “A Path Independent Integral and the Approximate Analysis
acknowledge the partial financial support of the Korea Science and Engineer-
of Strain Concentration by Notches and Cracks,” Journal of Applied Mechanics,
ing Foundation (R01-2000-00365).
ASME, V. 35, No. 6, 1968, pp. 379-386.
12. Rice, J. R., “Mathematical Analysis in the Mechanics of Fracture,” Frac-
CONVERSION FACTORS ture—An Advanced Treatise, V. 2, H. Liebowits, ed., Academic Press, New
100 mm = 3.94 in. York, 1968, pp. 191-311.
1 kN = 0.225 kips 13. Ba ant, Z. P., “Size Effect in Blunt Fracture: Concrete, Rock, and Metal,”
1 MPa = 145 psi Journal of Engineering Mechanics, ASCE, V. 110, No. 4, Apr. 1984,
1 kN-m = 0.738 kip-ft pp. 518-535.
14. Kim, J. K., and Eo, S. H., “Size Effect in Concrete Specimens with
NOTATION Dissimilar Initial Cracks,” Magazine of Concrete Research, V. 42, No. 153,
a1 = distance from neutral surface to center of member Dec. 1990, pp. 233-238.
a2 = distance from neutral surface to center of rod 15. Kim, J. K.; Eo, S. H.; and Park, H. K., “Size Effect in Concrete
As = area of steel Structures without Initial Crack,” Fracture Mechanics: Application to
b = thickness of specimen, width of section Concrete, SP-118, V. C. Li and Z. P. Ba ant, eds., American Concrete
C = compressive force Institute, Farmington Hills, Mich., 1989, pp. 179-196.
c = depth to neutral axis of critical section of C-shaped specimen 16. Ba ant, Z. P., “Fracture Energy of Heterogeneous Material and
d = effective depth, distance from compression face to centroid of Similitude,” SEM-RILEM International Conference on Fracture of Concrete
tension steel and Rock, Jun. 1987, pp. 390-402.
da = maximum aggregate size 17. Ba ant, Z. P., “Size Effect in Tensile and Compressive Quasibrittle
ecu = eccentricity of resultant load parallel to member axis measured Failures,” JCI International Workshop on Size Effect in Concrete Structures,
from centroid of cross section above neutral axis Oct. 1993, pp. 141-160.
fc = stress in concrete 18. Ba ant, Z. P., and Xiang, Y., “Size Effect in Compression Fracture:
fc′ = uniaxial compressive strength of standard concrete cylinder Splitting Crack Band Propagation,” Journal of Engineering Mechanics,
fmax = maximum compressive stress in beam ASCE, V. 123, No. 2, Feb. 1997, pp. 162-172.
ft′ = direct tensile strength 19. Cotterell, B., “Brittle Fracture in Compression,” International Journal
h = length of critical section of C-shaped specimen of Fracture, V. 8, No. 2, 1972, pp. 195-208.
k1k3 = ratio of average compression in beam to cylinder strength of concrete 20. Ba ant, Z. P., and Xiang, Y., “Compression Failure of Quasibrittle
k2 = depth of compressive resultant relative to neutral axis depth Materials and Size Effect,” AMD Symposium of the American Society of
k3 = ratio of maximum compression in beam to cylinder strength of Mechanical Engineers’ Applied Mechanics Division, 185, Damage
concrete Mechanics in Composites, ASME Winter Annual Meeting, Chicago, Ill.,
P1 = major load D. H. Allen and J. W. Ju, eds., 1994, pp.143-148.
P2 = minor load 21. Ba ant, Z. P., and Xiang, Y., “Size Effect in Compression Fracture:
Pu = ultimate axial load = P1 + P2 Splitting Crack Band Propagation,” Journal of Engineering Mechanics,
α = empirical constant defining size-independent strength ASCE, V. 123, No. 2, Feb. 1997, pp. 162-172.
α1 = width of equivalent rectangular stress block 22. Gonnerman, H. F., “Effect of Size and Shape of Test Specimen on
β1 = depth of equivalent rectangular stress block Compressive Strength of Concrete,” ASTM, Proceedings V. 25, 1925,
εc = strain in concrete pp. 237-250.
εcu = ultimate strain in concrete 23. Kim, J.-K.; Yi, S.-T.; and Yang, E.-I.; “Size Effect on Flexural Com-
σo = αft′ pressive Strength of Concrete Specimens,” ACI Structural Journal, V. 97,
= size-independent strength No. 2, Mar.-Apr. 2000, pp. 291-296.
σN (c) = nominal flexural compressive strength at failure in beam 24. Kim, J.-K.; Yi, S.-T.; and Kim, J.-H. J., “Effect of Specimen Sizes on
Flexural Compressive Strength of Concrete,” ACI Structural Journal,
REFERENCES V. 98, No. 3, May-June 2001, pp. 416-424.
1. Koenen, M., “Fur die Berechnung der Stärke der Monierschen 25. Kim, J. K.; Yi, S. T.; and Eo, S. H., “Size Effect of Concrete
Cement-platten,” Zentralblatt der Bauverwaltung (Berlin), V. 6, No. 47, Nov. Compressive Strength for the Non-Standard Cylindrical Specimens,”
1886, 462 pp. FRAMCOS-3, AEDIFICATIO Publishers, 1998, pp. 1973-1982.
2. Emperger, F. V., “Ein graphischer Nachweis der Tragfahigkeit und aller in 26. Kim, J.-K.; Yi, S.-T.; Park, C.-K.; and Eo, S.-H., “Size Effect on
einem Tragwerke aus Eisenbeton auftretenden Spannungen,” Beton und Eisen Compressive Strength of Plain and Spirally Reinforced Concrete Cylinders,”
(Vienna), V. 4, No. 5, 1904, pp. 306-320. ACI Structural Journal, V. 96, No. 1, Jan.-Feb. 1999, pp. 88-94.
3. Whitney, C. S., “Design of Reinforced Concrete Members under Flex- 27. Corley, G. W., “Rotational Capacity of Reinforced Concrete Beams,”
ure or Combined Flexure and Direct Compression,” ACI JOURNAL, Pro- Proceedings, ASCE, V. 92, ST5, Oct. 1966, pp. 121-146.
ceedings V. 33, No. 2, Mar.-Apr. 1937, pp. 483-498. 28. Alca, N.; Alexander, S. D. B.; and MacGregor, J. G., “Effect of Size
4. Hognestad, E.; Hanson, N. W.; and McHenry, D., “Concrete Stress on Flexural Behavior of High-Strength Concrete Beams,” ACI Structural
Distribution in Ultimate Strength Design,” ACI JOURNAL, Proceedings V. 27, Journal, V. 94, No. 1, Jan.-Feb. 1997, pp. 59-67.
No. 3, Nov. 1955, pp. 455-479. 29. van Mier, J. G. M., “Multiaxial Strain-Softening of Concrete—Part I:
5. Kaar, P. H.; Hanson, N. W.; and Capell H. T., “Stress-Strain Characteristics Fracture,” Materials and Structures, V. 19, No. 111, 1986, pp. 179-190.
of High-Strength Concrete,” PCA Research and Development Bulletin 30. Ibrahim, H. H. H., and MacGregor, J. G., “Modification of the ACI
RD051.01D, 1977, pp. 1-10. Rectangular Stress Block for High-Strength Concrete,” ACI Structural
6. Swartz, S. E.; Nikaeen, A.; Narayan Babu, H. D.; Periyakaruppan, N.; Journal, V. 94, No. 1, Jan.-Feb. 1997, pp. 40-48.
and Refai, T. M. E., “Structural Bending Properties of Higher Strength 31. Nilson, A. H., and Slate, F. O., “Structural Properties of Very High
Concrete,” High-Strength Concrete, SP-87, H. G. Russell, ed., American Strength Concrete,” Second Progress Report, Department of Structural
Concrete Institute, Farmington Hills, Mich., 1985, pp.147-178. Engineering, Cornell University, Ithaca, N.Y., Mar. 1979, 62 pp.

708 ACI Structural Journal/September-October 2002