Professional Documents
Culture Documents
by Francis A. Oluokun
The 0.5 power relation adopted by ACI Committee 318 for predict- have shown that the true tensile strength, as deter-
ing the splitting tensile strength of concrete from its compressive mined from the split cylinder test, is between 65 and 75
strength has been under scrutiny for a long time. Since its adoption,
many concrete researchers have investigated its prediction accuracy,
percent of the modulus of rupture. It has been well
that is, how well it correlates with or how well it predicts test data. established5-7 that the splitting tensile test of the cylin-
Research findings have consistently indicated that the 0.5 power drical specimen gives a more reasonable tensile strength
,relation adopted by ACI Committee 318 in the Building Code Re- estimation than the direct tensile test or the modulus of
quirements does not agree particularly well with test results. Conse- rupture test. The acceptance of the split cylinder test is
quently, concrete researchers have proposed several alternative rela-
tions. based on the fact that the stress distribution is reason-
The research for this paper investigated the validity and accuracy ably uniform along the vertical diameter of the cylin-
of these alternate relations. Tensile strength predictions from these der, which has been shown to be the plane of principal
relations were compared with test results assembled from various tensile stress for about 80 percent of its length. 8-11
sources. It was found that the splitting tensile strength is not propor-
tional to the 0.5 power of compressive strength. Although most of the HISTORICAL BACKGROUND
alternate relations appear to be good, the assembled test data re-
vealed that 0.69 is the most accurate power relation. Thus, 0.69 power
Generally, it has been accepted by concrete research-
relation is proposed as an alternative to the ACI 318 relation. ers as well as by the American Concrete Institute that
the splitting tensile strength of concrete is proportional
Keywords: compressive strength; concretes; evaluation; predictions; splitting to the 0.5 power of the cylinder compressive strength.
tensile strength.
Section 11.2 of the ACI Building Code (ACI 318-89}12
Tensile strength of concrete is a physical property gives
that has no absolute meaning; it is always expressed in fc, = 6. 7 (f: )o.s
terms of a specific test procedure. The direct tension
test, the beam (or modulus of rupture) test, and the as the empirical relationship between the splitting ten-
split cylinder (or Brazilian) test are the three kinds of sile strength and the cylinder compressive strength for
tests that have been used. In brittle materials like con- concrete with strength between 2000 and 6000 psi (13.79
crete, where there is little redistribution of stresses, the and 41.38 MPa). Historically, the splitting tensile
direct tensile test usually results in an underestimation strength test on which the ACI relation was based was
of tensile strength. The underestimation arises from the independently originated by Akazawa 13 in Japan and
difficulty in insuring that the applied load is truly ax- Carniero and Barcellos14 in Brazil. Since the invention
ial. 1 In the beam test, because of the development of and later the acceptance of this method of tensile
minute and invisible microcracks, the load at first crack strength evaluation, several relations have been. pro-
is difficult to establish. The ultimate load used in esti- posed for the splitting tensile strength relationship to
mating the modulus of rupture is therefore not the the compressive strength of concrete. Akazawa 13 rec-
same as the load at which cracking first occurred. Be-
cause of this fact and the nonlinearity of the stress- ACI Materials Journal, V. 88, No.3, May-June 1991.
Received July 6, 1990, and reviewed under Institute publication policies.
strain curve for concrete in tension, the modulus of Copyright 1991! American Co~cr~te !nstitut.e. All rights reserve~, includin_g
rupture has been found to oveFestimate the tensile the making of cop1es unless perm1ss1on IS obtamed from the copynght propn-
etors. Pertinent discussion will be published in the March-April 1992 ACI Ma-
strength of concrete. 2 3 In fact,;concrete researchers 24 terials Journal if received by Dec. I, 1991.
let = 0.69 (f; )0.735 for 1: in MPa was again proposed as a better alterna-
tive to the ACI relation for concrete up to approxi-
for 1: in psi, which interprets to mately 8300 psi.
In a comprehensive investigation of the physical
let = 0.185 (/;)0.735
properties of concrete at early ages carried out at the
University of Tennessee by Oluokun 18 and published by
Oluokun, Burdette, and Deatherage, 19 the relation be-
for 1: in MPa. The Committe Euro-International du
tween the splitting tensile strength and the compressive
Beton (CEB) 15 recommended
strength was studied. The applicability of the ACI 318-
89 equation to tensile strength prediction both at early
let = 1.43 (/;)0 '667 ages and at maturity (conventionally at 28 days) was
investigated. It was revealed in this study that the true
for 1: in psi, which is equivalent to relationship between the tensile strength and the com-
pressive strength of concrete does not conform to the
let = 0.273 (f; )0' 667 0.5 power relation adopted by ACI 318-89. Instead, a
more accurate and more data-representative relation
for 1: in MPa. was proposed as
Recent test results have indicated that the 0.5 power
relation as used in ACI 318-8912 is not necessarily valid. let = 0.584 (/;)0.79
Several alternate relations have been proposed as better
and more representative of test data for the prediction for 1: in psi, which is equal to
of tensile strength from cylinder compressive strength.
Among the most recently proposed alternative relations let = 0.206 (f; )0.79
is
for 1: in MPa for concrete strength up to about 9000 psi
let = 4.34 (/;)0.55 (62.07 MPa).
Carino and Lew 20 re-examined the relation between
for 1: in psi, which is equivalent to splitting tensile and compressive strength of normal
weight concrete. A series of statistical analyses were
let = 0.462 (/;)0' 55 performed on 124 published data points from various
sources. It was concluded that the splitting tensile
for 1: in MPa proposed by Ahmad and Shah4 for con- strength is proportional to the 0.71 power of the com-
crete strength up to 12,000 psi (84 MPa). pressive strength and
In a study of tensile strength of concrete by Ra-
phael, 16 detailed analyses of test data showed that the let = 1.15 (/;)0.71
0.5 power relationship is not representative of test data.
Consequently for 1: in psi, which is equivalent to
03000 PSI
30016000 PSI
60019000 PSI
Q..9000PSI
30 30
[ l
.,.ufc 20 .,.ufc 20
a: a:
5la: 0
a:
w lli
~ ~_,
s
c_,
10
~
c_,
~ 10
~
i" ~
0
ACI AKAZ. CARl CANIER CEll GARO OW RAPH SHAH PROP. 0
ACI AKAZ. CARl CAN1ER CEll GARO OW RAPH SHAH PROP.
AUntORS
AUntORS
Fig. 2 - Comparison of integral absolute errors Fig. 3 - Comparison of integral absolute errors (0 to
(group) 9000 psi)
high errors for all concrete compressive strength Oluokun's relation, 18 and 9.36 for Raphael's relation, 16
groupings. to a maximum for 28.29 for Carniero and Barcellos's
A thorough study of the percentage of lABs for each relation, 14 with ACI's relation having 10.12 percentage
of the relations, as shown in Table 1 and illustrated in of IAE with an overestimation of 37.63 percent. Again,
Fig. 2 for all concrete compressive strength groups, as shown in Fig. 3, Akazawa's, Carniero and Barcel-
shows that the difference in the values of lABs for each los's, and the CEB's relations gave consistently high
relation does not vary appreciably (less than 5 percent prediction errors.
maximum) with strength groupings. Therefore, apply- In determining the best power relation for predicting
ing each relation to predict concrete from 0 to approx- the splitting tensile strength from the compressive
imately 9000 psi collectively will not result in any sig- strength, the following factors were considered: (1) best
nificant difference in prediction errors from the errors test data representative power relation; (2) minimum
obtained from the group analyses just presented. prediction error as indicated by minimum IAE; and, to
From the results of the collective 566 data point some degree, (3) conservatism as expressed by per-
analyses for concrete up to approximately 9000 psi, centage of overprediction.
percentage of IAE varies from. a minimum of 7.43 for Fig. 4, a comparison of the best five prediction rela-
the proposed relation through 7.46 for Gardner's rela- tions (as indicated by their IAE values in Table 1) with
tion, 17 8.01 for Carino and Lew's relation, 20 8.43 for the ACI relation and test data shows that "the ACI re-
306 ACI Materials Journal I May-June 1991
G-9000 PSI
CNEAI'REDICT.
so+-----
40+-----
30
20
10
AUTHORS
Fig. 4 - Comparison of best five relations with ACI
and test data
Fig. 6 - Comparison of integral absolute errors and
overprediction
800
800
600
600
~
ii5
j!:
"i!i 400
~
ii5
~ j!: 400
~
ffi
I 200
~
_,
w
.~ 200
o+-~--+-~--+-~--+-~--+-~--+-~__,
COMPRESSIVE STRENGTH IN PSI 0 2000 4000 6000 8000 10000 12000