You are on page 1of 8

ACI JOURNAL ~ . - _- ~.

TECHNICAL PAPER
Title No. 81-17

Tensile Strength of Concrete

by Jerome M. Raphael

A limiting factor in the safety of mass concrete structures, such as element analysis of stresses in concrete dams during
concrete arch dams under seismic loadings, is the tensile strength of earthquakes. In research sponsored by the U.S. Engi-
the concrete. Tensile strength can be tested three ways: direct ten-
neers using two-dimensional finite element analyses,
sion, splitting tension, and flexural tests. Results of these tests differ,
and results of tests made on cores taken in the field differ from re- Chopra 1 studied the dynamic response of Koyna Dam,
sults made on laboratory specimens. Some 12,000 individual test re- treating the water as a compressible fluid. Three-di-
sults were examined to find reason for these discrepancies. Low ten- mensional finite element analyses of arch dams of the
sile strength of cores from dams was found to be caused by drying Los Angeles Flood Control District began after 1971
shrinkage and surface cracking. Some tests were discarded because of
flaws in testing technique. A theoretical relationship was found be-
with reanalyses of the flood control dams of that dis-
tween tensile strength and modulus of rupture. Values are recom- trict. 2 The state of the art in 1983 was real-time three-
mended for true and apparent tensile strength for a wide range of dimensional analysis of stresses in concrete dams, tak-
compressive strengths under static and seismic loadings. ing into account interaction of the reservoir and foun-
dation. From such studies, it is apparent that the key
Keywords: compressive strength; concrete dams; dynamic loads; earthquake
resistant structures; flexural tests; mass concrete; measurement; splitting tensile property in limiting the capacity of concrete dams dur-
strength; static loads; structural analysis; tensile strength; tension tests. ing earthquakes is the tensile strength of the concrete.
This is because while maximum compressive and tensile
In a number of recent investigations of the behavior stresses under seismic loading are roughly equal, com-
of actual concrete dams during earthquakes, it has be- pressive strength is many times the tensile strength of
come apparent that a limiting factor has been that the concrete, and the concrete will always fail in tension
tensile strength of any concrete is only a fraction of its long before it approaches compressive capacity.
compressive strength. Hence, the concrete will fail in The question is: What is the tensile strength of con-
tension long before it begins to be distressed in crete, and how should it be measured? A number of
compression. Engineers working with reinforced con- test methods have been used to evaluate this property.
crete have simply ignored the tensile strength of the In the direct test for tensile strength, the specimen is
concrete because of its low value and placed steel to gripped at its ends and pulled apart in tension; tensile
pick up the entire tensile load. Engineers working with strength is failure load divided by area. In the splitting
dams must rely on the tensile strength of concrete un- tension test, a cylinder is loaded in compression on two
der earthquake conditions as it is impracticable to diametrically opposite elements, failing in tension on
specify the enormous quantities of steel needed to resist the plane between the loaded elements. In the modulus
the tensile forces in a dam. Therefore, dams are gener- of rupture test, a rectangular beam is loaded at the
ally unreinforced. center or third points and fails in bending, with the
Literature has numerous reports of tests of tensile computed tensile stress at failure load called the mod-
strength of concrete under different test methods and ulus of rupture. Unfortunately, each method seems to
conditions, with a confusing variety of results. It is the have its own characteristic result. Many engineers as-
purpose of this paper to make order out of this confu- sume that the direct tensile strength of concrete is about
sion so that tensile strength of a given concrete can be 10 percent of its compressive strength; splitting tensile
predicted reliably. strength is about the same, perhaps one percent

State of the art in analysis


Received May 5, 1983, and reviewed under Institute publication policies.
The Koyna earthquake of December 11, 1967, and Copyright 1984, American Concrete Institute. All rights reserved, including
the San Fernando earthquake of February 9, 1971, both the making of copies unless permission is obtained from the copyright propri-
etors. Pertinent discussion will be published in the January-February 1985 ACI
resulted in significant advances in the dynamic finite JOURNAL if received by Oct. I, 1984.

158 ACI JOURNAL I March-April 1984


Jerome M. Raphe/, FACI, has been a member of ACI since 1934. He has
worked almost exclusively on large concrete dams with the U.S. Engineers and
the U.S. Bureau of Reclamation until1953 and since then as a consulting civil
engineer while on the faculty of civil engineering at the University of Califor-
nia at Berkeley. On his retirement in 1979, he was appointed professor emeri- ;;;
tus of civil engineering and continues with research and consultation on studies 1'-
of large concrete dams. He is a past chairman of ACI Committee 209, Creep w
a:
::::l
and Shrinkage in Concrete, and is a member of ACI Committee 207, Mass t-
o._
Concrete. ::::l
a:
lL.
0
(J)
::::l
stronger; and modulus of rupture is about 15 percent of __J
::::l
0
compressive strength. Examination of actual test re- 0
::.
sults from almost 12,000 specimens shows that there is a:
0
little basis for assuming a linear relationship between J:
t-
(!)
tensile and compressive strength, by whatever measure- z
w TENSION RUPTURE

ment means. a: GONNERMAN 8 SHUMAN


t-
(J) WALKER B BLOOM
w GRIEB 8 WERNER
__J
Measurements of tensile strength (J)
HOUK

z
While many results of laboratory tests for tensile w
t-
strength, modulus of rupture, and compressive strength 1200) 1300) 14001 1500) 1600)

of concrete have been published, only a very few in- COMPRESSIVE STRENGTH- psi (kg/cm 2 )
stances are available where all three properties of the
same concrete have been tested in one laboratory. Four Fig. ]-Relation between modulus of rupture, tensile
sets of data, representing tests spread over nearly 40 strength, and compressive strength of concrete
years, are described next.
Gonnerman and Shuman, 19283 , tested 1760 moist-
cured 6 in. (152 mm) diameter cylinders and 7 x 10-in. distinct from all the tensile strength points, being about
(178 x 254-mm) beams using concretes with compres- a third higher in value. Next, the Gonnerman and Shu-
sive strengths varying from 200 to 9200 psi ( 1.4 to 63 man tensile strength data have lower values than those
MPa). Tensile strength was determined on 6 x 18-in. for the remaining three investigators by about 20 per-
(152 x 457-mm) cylinders held at the ends by bolted cent. Considering that the end conditions of the Gon-
steel strap grips with leather friction surfaces. nerman and Shuman tensile test specimens combine
Walker and Bloem, 1960\ related splitting tension compression from the bolted grips with tension from
and modulus of rupture to compressive strength of 576 the testing machine, a condition shown by Ri.isch 12 to
laboratory concrete specimens using various sized ag- result in failure at less than either maximum tensile or
gregates and water-cement ratios, all moist cured and compressive strength, these specimens are eliminated
all either 6-in. (152-mm) cylinders or 6 x 6-in. (152 x from further consideration.
152-mm) beams. Now consider tests of actual concrete from dams. In
Grieb and Werner 5 reported in 1962 on tests of more the past few years, a number of dams in service have
than 600 specimens made during a 10-year period been reexamined to determine their safety in an earth-
(1951-1961) using concrete made of natural, crushed, or quake. Fig. 2 shows averaged results of over 500 6-in.
lightweight aggregates to a maximum size of 1.5 in. (38 (152-mm) cores taken from 14 concrete dams on the
mm). Compression and splitting tension specimens were West Coast. The splitting tension tests show strengths
6 x 12-in. (152 x 254-mm) cylinders, and flexural tests averaging about ten percent of compressive strength
were made on 6 x 6 x 21-in. (152 x 152 x 533-mm) and direct tensile strength about half that value. The
beams. five and ten percent line are not intended to be the ac-
In 1965, Houk 6 studied 324 specimens made from a tual functions for these data but give a basis for com-
variety of mass concrete mixes of somewhat lower parison.
strength, including a number of cements and pozzo- When it was first noted that field concrete cores
lanic admixtures, in connection with the concrete for seemed only half as strong in tension as laboratory
the construction of Dworshak Dam. His tensile speci- concrete, the test methods were suspected. The earliest
mens were 6 in. (152 mm) square prisms with the load tests at the University of California were made by
applied through Y4 in. (19 mm) diameter steel rods bonding steel plates to the ends of the cylindrical core
embedded on the center axis. Compression specimens and applying load to the plates. Considering that the
were 6 x 12-in. (152 x 254-mm) cylinders, and flex- changes in stiffness between steel and concrete might
ural specimens were 6 in. (152 mm) square prisms. somehow have introduced a concentrated stress that
For all these tests, representing a wide variety of weakened the core, a new test method was tried. Epoxy
concretes, values of tensile strength and modulus of was used to build up the ends of the core to a dumbbell
rupture have been plotted against compressive strength shape, and steel plates were bonded to the enlarged
in Fig. 1. In this figure, three distinct families of data ends. All these dumbbell specimens broke in the origi-
can be seen. First, all the modulus of rupture points are nal center section, but there was no change in the pre-
ACI JOURNAL I March-April 1984 159
BOO
<t
150) I I
700 0 2 3 4
DIRECT TENSION
5 61NCHES
IOfc'
o SPLIT TENSION
00
100
600 w
N" 1401
a:
80 :::>
E 1-
~ 500 60 0
en
=
;;; :::E
?" 130)
40 1-
I 400
t- z
(!)
z .05fc w
20 u
w
a:
t-
300
120) .. a:
w
.. . 0
(J)
0..
w
.J
iii 200
z
~ 110) .. MOISTURE GRADIENT

100

0~--~~~--~L---~L-~--~~--~--~
Fig. 3-Drying shrinkage of a 6-in. cylinder
1000 2000 3000 4000 5000 6000 7000
1100) 1200) 1300) 1400)

COMPRESSIVE STRENGTH- psi ( kg/cm 2 )


mens were stored moist and tested moist. No cracks
Fig. 2-Tensile strength versus compressive strength of could have formed, and they tended to show true ten-
concrete cores sile strength.

viously observed result of only half the tensile strength Splitting tension test
of laboratory concrete. Now consider the splitting tension test. Prior to fail-
The chief difference between the laboratory speci- ure, 10 there is a biaxial compression region immediately
mens and the field specimens was in their curing his- below the region of application of the load which, while
tory. Laboratory specimens were routinely kept in a highly stressed, has great resistance to failure because
moist atmosphere until testing. Field specimens, repre- of its confined state. For the greatest part of the loaded
senting mass concrete kept moist by virtue of its mass, axis, a nearly uniform tensile region exists, and at fail-
seemed to have been allowed to dry out at some time ure, the cylinder usually splits neatly on that axis, the
between removal and testing. The importance of this failure going through aggregate as well as mortar. Ten-
drying period is explained below. sile strength is computed as
Fig. 3 shows three stages in the drying of a 6 in. (152
mm) diameter concrete cylinder. Since drying obeys the 2P
same physical laws as cooling, these curves have been 1rLD
drawn from plots for variation of temperature with
time published by the U.S. Bureau of Reclamation/ where
using the value of 0.0001 ft 2 per day for drying diffu- fs splitting tensile stress, psi (or MPa)
sivity as suggested by Carlson. 8 After only one week of P total load at failure, lb (or N)
drying, the surface is completely dry, but drying one L length of cylinder, in. (or mm)
inch from the surface has only begun. The relation be- D diameter of cylinder, in. (or mm)
tween shrinkage and drying is linear; shrinkage is di-
rectly proportional to drying. Thus, the surface con- Any surface cracks caused by shrinkage are likely to
crete could attain 100 percent of its ultimate shrinkage be in the compression region and are not likely to af-
if it were not restrained by the moist concrete at the fect the behavior of the concrete in the tensile region.
center 4 in. (100 mm) of the cylinder. As it is, the cyl- Hence, the splitting tension test will not be affected by
inder is on the average still 80 percent moist, and the surface drying and should give the actual tensile
'l>'U'i~'d~e d\He'ient\'d\ \'i> %(} \)e:,:ce\\t (}{ the \)(}te\\t\a\ 'i.t"l:ewe,th (}{ the C.(}\\C."I:ete.
shrinkage. Since the potential for drying shrinkage
might be as much as 800 millionths for a concrete with Flexural test
a natural gravel aggregate, the surface strain could be Now consider the bending test, standardized by
640 millionths, which is equivalent to 2500 psi (17 MPa) ASTM as ASTM C 78, Standard Test Method for
tension. Since this far exceeds the tensile strength of the Flexural Strength of Concrete (Using Simple Beam
concrete, minute surface cracks will form extending in- With Third-Point Loading). 9 Looking back at Fig. 1
ward from the surface, perhaps as much as a Yz in. (12 which shows test results of modulus of rupture and
mm). Thus, it is no wonder that the dry cores exhibited tension tests plotted against the compressive strength of
low tensile strength; they were already partially concrete, the modulus of rupture is clearly seen to be 30
cracked. On the other hand, all the laboratory speci- to 50 percent higher than the tensile strength for all
160 ACI JOURNAL I March-April 1984
500

P/2 ~P/2

I
'E
0

'
01
P/2 I--L13----L~3----Lt3---l P/2
~
;;; 300
a.
I MODULUS OF RUPTURE
rn
rn
UJ
a:
\ii
~ (15)
.333d
Vi 200
zUJ 287d
1-
213 d
(10)

STRAIN DISTRIBUTION STRESS DISTRIBUTION

QL-----l..._ ___L__ __;L__ _ j __ _...L__ ___.J_ __ J Fig. 5-Flexural failure of a rectangular concrete beam
25 50 75 100 125 150
MICROSTRAINS
mode in concrete, not the elastic mode. Such an equa-
tion will now be derived.
Fig. 4-Stress-strain diagrams for tensile failure Fig. 5 shows the general conditions of the third-point
flexural loading test for modulus of rupture, and the
compressive strengths. Both tests measure tensile variation of strain over the depth of the beam in the
strength, so why the discrepancy? highly stressed midsection, straight line all the way to
The modulus of rupture is measured by a derivation failure. The stress distribution varies linearly in the
of the beam equation f = Mel I. In the case of third- compression region of the beam and reflects the stress-
point loading, which is the loading usually employed strain diagram for that particular concrete in the tensile
region. This is because the beam is failing in the tensile
region but is stressed to a fraction of its ultimate
j,. strength in the compressive region. The actual shape of
the tensile stress-strain curve reflects to some extent the
strength of the concrete, since stress-strain curves for
where higher strength concretes seem to have sharper crests
j,. modulus of rupture, psi (or MPa) than for lower strength concretes. Thus, the magnitude
p load at failure, pounds (or N) and position of the resultant tensile force will vary
L span length, in. (or mm) slightly with individual concretes.
b width of beam, in. (or mm) While many attempts have been made to quantify the
d depth of beam, in. (or mm) magniture and position of the resultant force under the
curved stress diagram, the simple rectangular stress
This is an equation derived from elastic theory, as- diagram proposed by Whitney'' in 1937 has gained
suming elastic behavior of concrete to the point of fail- widest acceptance, at least in the United States.' 5 16
ure. This is far from the actual state of affairs at the Thus, as a first approximation to evaluating the maxi-
time of the failure of the concrete beam. mum tensile stress at failure, assume that the neutral
Fig. 4 shows nine stress-strain diagrams for tensile axis remains at the centroid of the section, as is as-
failure tests of concrete cores from three dams. While sumed when computing the modulus of rupture under
the scales differ by a factor of ten, there is marked elastic conditions, and replace the curvilinear tensile
similarity in the shape of the curves to the familiar stress diagram by a simple rectangular stress diagram
stress-strain diagram for compression of concrete. No with the same dimensional constraints proposed by
diagram is linear to failure, but all are marked by grad- Whitney. The magnitude of tensile stress in the stress
ually increasing deformations above 50 percent of their block is 0.85 J;, where J; is the maximum tensile stress,
respective tensile strengths. What is needed for flexural and this block extends over only 85 percent of the ten-
failure is an equation representing the actual failure sile region. The magnitude of the tensile force can be
ACI JOURNAL I March-April 1984 161
. . .. ,
~ BOO : ;w(. ..
N' 1401 ""'
c>
..:.. . .. ""
.. ::.
E -~ (50)
.f"
0

'Jf 500 w
a:
' 700
"\..
:.:
,
: -:
.,
I
..
. . .~~
.
.. .. ... . ,..-: .-.
:::>
.
in f
f-
';"- (30) a.
I
t-
400 :::>
a: . II
~~

(.!)
z
w
lL.
0
ci 500 ...
: ~ ...
... :
.......
. ... .
a: 300
t; (20) 0
:::;: '\
a:
. .
... ..
w (30)
.J 0 400
(f)
z I
t-
w o TENSILE STRENGTH (.!)
t-
100 3/4 MODULUS OF RUPTURE
z
l;:!
t-
300
(20) .. GONNERMAN a SHUMAN
WALKER a BLOEM
(f) HOUK
0 L_--~~--~--_LL__ __L~--~--l_L__ __
w 200 GRIEB a WERNER
::::!
1000 2000 3000 4000 5000 6000 7000 (f)
(100) (200) (300) (400) z (10)
:
w
COMPRESSIVE STRENGTH- psi (kg/cm 2 ) t- 100

0 ~---L~--~--_LL__ __ L J __ _~--~----~

Fig. 6- Tensile strength versus compressive strength by 1000 2000 3000 4000 5000 6000 7000
(100) (200) (300) (400)
two types of test
COMPRESSIVE STRENGTH- psi (kg/C:TT\2)_

Fig. 7-Modulus of rupture versus compressive


evaluated as T = 0.85 ft x b x 0.425 d = 0.361 ftbd. strength
The length of arm is 0.333 d + 0.287 d = 0.620 d.
Thus, the internal moment isM = 0.361 ftbd x 0.62 d
Equations of curves
When fitting a mathematical function to experimen-
= 0.224 ftbd or ft = 4.46 mlbd
2 2
tal data with a wide distribution as in typical concrete
experiments, one has the choice of working either with
Since
the average or choosing a safe value such that a mini-
p mum number of tests will fall below the chosen equa-
L PL
M --
X - - tion. In this paper, averages are used for two reasons.
2 3 6 In setting the strength of concrete for new construc-
tion, the distribution of test values above and below the
4.46 PL
----
PL mean is taken into account by the concrete mix de-
!t 0.744
6 bd 2
bd 2 signer when he sets the average mix strength higher
than the required design strength by a factor depending
For the elastic case, as was stated above on expected quality control. Therefore, to use the safe
value here would compound safety factors. When ana-
lyzing existing structures, average strength and elastic
properties are assumed. Since most concrete structures
are highly indeterminate, any local weakness in the
In a second approximation for determining the ft!f, concrete will be reflected in lower elastic modulus and
ratio, the neutral axis was allowed to move upward un- lower stress, and vice versa, so the stresses tend to av-
til the tensile and compressive forces were in balance. erage out.
For this case, a similar analysis showed that ft = 0.73 The values of modulus of rupture are plotted against
f,. compressive strength for all four sets of data in Fig. 7.
Comparing these equations, it can be seen that ten- It can be seen that they all merge into one consistent
sile strength can be measured in a flexural test, but its family of data. A curve that fits these data can be ex-
value is about three-fourths of the modulus of rupture, pressed as
which is the value derived by the elastic analysis of the
data. With this in mind, Fig. 6 has been redrawn from f, 2.3 f/ 13 in psi
Fig. 1 with hollow dots representing tensile strength
measured by the tension test, and solid dots represent-
and
ing tensile strength as three-fourths the results of the
modulus of rupture tests. It can be seen that the two
families of data merge to a fairly consistent pattern. j, = 0. 95 f/ 13 in kg/ cm 2
162 ACI JOURNAL I March-April 1984
Fig. 8 is a plot of tensile strength versus compressive 700
strength for the Walker and Bloem, the Grieb and
Werner, and the Houk data. The Gonnerman and Shu- 600
man data were not plotted because of failure in experi- ;::; (401
E
mental techniques previously described. An equation 0
~ 500
that relates these tensile and compressive tests is ""'
;;;
a. (301
}; = I. 7 f/ in psi
13
~ 400
t-
(.!)
z
and ~ 300
t- (201 go
rn 0 6 0

w
}; = 0.7 j}13 in kg/cm2 ~ 200
z o WALKER S BLOEM
~ (101
It must, therefore, be concluded that direct tensile 6 HOUK

tests are very sensitive to technique in both testing and 100 o GRIEB S WERNER

care of the specimens. The splitting tension test is most


forgiving of slight periods of drying during preparation
1000 2000 3000 4000 5000 6000 7000
of specimens for test, seems least dependent on testing (1001 12001 (3001 (4001

technique, and gives reproducible results from a variety COMPRESSIVE STRENGTH- psi (kg/cm2)
of laboratories.
Fig. 8- Tensile strength versus compressive strength
Dynamic loading
A number of authors have shown that the apparent quake increases compressive strength an average of 31
compressive strength of concrete varies with the speed percent and increases tensile strength an average of 56
of testing; the faster the test, the greater the load re- percent. These values can be compared with the 33 per-
quired to break a concrete cylinder. Perhaps the most cent commonly allowed for earthquake loadings by
comprehensive series of tests were Hatano's 13 14 in which most American building codes. 15 16
concretes were tested to failure in tension and compres-
sion at a number of speeds ranging from a few hun- Tensile strength to be used in structural analysis
dredths of a second to hundreds of seconds. In all We now come to the heart of the matter: What ten-
cases, increasing the speed of loading resulted in in- sile strength should be ascribed to mass concrete
creases in both strength and elastic modulus. This fac- stressed in tension under a seismic load? It has been
tor seems to be more marked in tension tests than in shown that direct tension tests can be in error by as
compression tests. much as 50 percent if the cores are not treated very
When the advent of the computer made it possible to carefully and kept from drying. Splitting tension tests
perform dynamic finite element analyses of the behav- are the least subject to modification due to storage
ior of dams during earthquakes, concrete testing was problems. If we are interested in knowing the true ten-
modified to reflect the actual speed with which con- sile strength of concrete, the values from the splitting
crete in a dam could be stressed from zero to maximum tension test are the most reliable.
stress. The next question that arises is: Is this what we need
If a dam were to vibrate at, say 5 Hz, a complete to compare with the tensile stresses in the dam pro-
stress cycle would last only one-fifth of a second, and duced by the computer? To answer this question re-
concrete would reach its maximum stress in one-fourth quires considering the actual method by which the
of a cycle, say 0.05 seconds. Testing machines are maximum stresses in the dam are computed.
available that can load specimens at this rate, and re- Present day stress analyses are nearly all some var-
cording oscillographs are used to record stress-strain iant of the finite element method and, for practicabil-
diagrams for these rapid tests. Table I shows some re- ity, assume a constant modulus of elasticity. The actual
sults of tests of cores from five western dams. It can be operation of the method involves a strain analysis, since
seen that the rate of loading characteristic of an earth- compatibility of deformation of the individual ele-

Table 1 - Effect of speed of loading on strength


, Direct tensile Splitting tensile Compressive
strength, psi strength, psi strength, psi
Dam Slow Fast Factor Slow Fast Factor Slow Fast Factor
Crystal Springs 203 337 1.66 490 640 1.31 4500 5930 1.32
Big Tujunga A 139 225 1.62 440 - - 3540 4070 1.15
Big Tujunga B 217 397 1.83 5320 5970 1.12
Santa Anita 224 373 1.53 440 650 1.48 4520 5810 1.29
Juncal 462 723 1.56 4420 6740 1.52
Morris 474 694 1.46 5290 7780 1.47
Average 1.66 1.45 1.31
Grand average all tensile factors = 1.56.

ACI JOURNAL I March-April 1984 163


APPARENT
MODULUS OF TENSILE
RUPTURE STRENGTH NE {120)
A ~ 1600 3.4 fc213

="'
01
.><;

(100)
a.
TENSILE I 1200 2.6
I (80)
STRENGTH f- 2.3
(.!)
zw (60)
0:: BOO 1.7
f-
en en
en ~ (40)
w
a: ~ 400
1-
en w
f-

0 2000 4000 6000 6000 10,000

I
(100) (2001 (300) (400) (500) (600)
2
COMPRESSIVE STRENGTH- psi (kg/cm )

Fig. 10-Design chart for tensile strength

der seismic loading should be the value determined as


STRAIN modulus of rupture, augmented by the multiplier found
appropriate by dynamic tensile tests, or about 1.5.
Fig. 9-Apparent tensile strength Fig. I 0 presents four recommended plots of tensile
strength as a function of compressive strength, to be
used depending on need. The lowest plot J; = 1. 7 f/ 13
ments is the key to the analysis. When equilibrium of represents actual tensile strength under longtime or
forces and deformation is attained, the strains are con- static loading. The second plot j,. = 2.3 fc 213 is also for
verted to stress through the elastic modulus. static loading, but takes into account the nonlinearity
The important point in concrete dam analysis is that of concrete, and is to be used with finite analyses. The
a linear analysis is being used on a material that be- third plot J; = 2.6 fc 213 is the actual tensile strength of
haves nonlinearly. As shown in Fig. 9, the linear anal- the concrete under seismic loading, and the highest plot
ysis will predict stress A at failure, whereas the mate- j,. = 3.4 f/ 13 is the apparent tensile strength under seis-
rial for the same strain will only be stressed to B. The mic loading that should be used with linear finite ele-
Bureau of Reclamation, 1980, 17 postulated a conver- ment analyses.
sion factor based on observed ratios of A to B of 1.20.
This factor has been used in a number of analyses in Conclusions
setting required concrete strength from computed 1. Tensile strength of concrete is a constant quantity
maximum tensile stresses. Observed values of tensile that can be measured by three types of tests if care is
strength augmented by the factor have been termed by taken handling specimens and allowance is made for
Dungar , 1981, 18 "Apparent Tensile Strength" and the mode of failure.
compared directly with tensile stresses computed in 2. Direct tension test is difficult to accomplish and is
stress analyses. subject to large errors if the specimen is allowed to sur-
An experimental method exists by which this appar- face dry.
ent tensile strength can be measured directly and accu- 3. The splitting tension test is easiest to accomplish
rately. Use of the flexural test, together with its usual and gives the most reliable results.
linearly derived modulus of rupture, accomplishes this. 4. The modulus of rupture test gives consistent re-
This is essentially a test to failure dictated by the tensile sults in a variety of laboratories and gives the value that
strength of the concrete but with the strength com- can be used directly with results of a finite element
puted as though the material were elastic, exactly the analysis.
conditions of the finite element analysis. This direct 5. Tensile strength can be computed from modulus
measurement cuts through all the uncertainties of the of rupture by multiplying by a factor which takes into
relationship between tensile and compressive strength, account the shape of the specimen and the mode of
and the appropriate multiplier to take care of the non- failure. The factor is 3/4 for rectangular cross sections.
linearity of the material. The test can be performed on 6. All values of tensile strength should be increased
cores as well as prisms if reasonable care is taken to by 50 percent when used with seismic loadings.
prevent drying out of the cores during transportation. 7. The apparent tensile strength of concrete under
Thus, it is recommended that the tensile strength of seismic loading is twice its splitting tensile strength un-
concrete for judging the safety of a concrete dam un- der normal loading.
164 ACI JOURNAL I March-April 1984
REFERENCES 10. Carneiro, Fernando L. L. B., and Barcellos, Aguinaldo,
1. Chakrabarti, P., and Chopra, A. K., "Earthquake Analysis of "Tensile Strength of Concretes," RILEM Bulletin (Paris), No. 13,
Gravity Dams Including Hydrodynamic Interaction," International Mar. 1952, pp. 97-127.
Journal of Earthquake Engineering and Structural Dynamics, V. 2, 11. Whitney, Charles S., "Design of Reinforced Concrete Mem-
1973, pp. 143-160. bers Under Flexure or Combined Flexural and Direct Compression,"
2. Clough, R. W.; Raphael, J. M.; and Mojtahedi, S., "ADAP- ACI JouRNAL, Proceedings V. 33, No. 4, Mar.-Apr. 1937, pp. 483-
A Computer Program for Static and Dynamic Analysis of Arch 498.
Dams," Report No. EERC 73-14, Earthquake Engineering Research 12. Kupfer, Helmut; Hilsdorf, Hubert K.; and Riisch, Hubert,
Center, University of California, Berkeley, June 1973, 170 pp. "Behavior of Concrete Under Biaxial Stresses," ACI JOURNAL, Pro-
3. Gonnerman, H. F., and Shuman, E. C., "Compression, Flex- ceedings V. 66, No.8, Aug. 1969, pp. 656-666.
ural and Tension Tests of Plain Concrete," Proceedings, ASTM, V. 13. Hatano, T., and Tsutsumi, H., "Dynamic Compressive Defor-
28, Part II, 1928, pp. 527-564. mation and Failure of Concrete Under Earthquake Load," Technical
4. Walker, Stanton, and Bloem, Delmar L., "Effects of Aggregate Report No. C-5904, Technical Laboratory of the Central Research
Size on Properties of Concrete," ACI JOURNAL, Proceedings V. 57, Institute of the Electric Power Industry, Tokyo, Sept. 30, 1959, 30
No. 3, Sept. 1960, pp. 283-298. pp.
5. Grieb, W. E., and Werner, G., "Comparison of the Splitting
14. Hatano, T., "Dynamic Behavior of Concrete Under Impulsive
Tensile Strength of Concrete with Flexural and Compressive
Tensile Load," Technical Report No. C-6002, Technical Laboratory
Strengths," Public Roads, V. 32, No.5, Dec. 1962, pp. 97-106.
of the Central Research Institute of Electric Power Industry, Tokyo,
6. "Concrete Aggregate and Concrete Properties Investigations,
Nov. 5, 1960, 15 pp.
Dworshak Dam and Reservoir," Design Memorandum No. 16, U.S.
Army Engineer District, Walla Walla, 1965, pp. 203-212. 15. Uniform Building Code, International Conference of Building
7. "Cooling of Concrete Dams," Bulletin No. 3, Final Reports, Officials, Whittier, 1982, p. 125.
Boulder Canyon Project-Part VII, Cement and Concrete Investiga- 16. ACI Committee 318, "Building Code Requirements for Rein-
tions, U.S. Bureau of Reclamation, Denver, 1949, p. 73. forced Concrete (ACI 318-77), American Concrete Institute, Detroit,
8. Carlson, Roy W., "Drying Shrinkage of Large Concrete Mem- 1977, p. 29.
bers," ACI JouRNAL, Proceedings V. 33, No.3, Jan.-Feb. 1937, pp. 17. "Feasibility Design Summary, Auburn Dam, Concrete Curved-
327-336. Gravity Dam Alternative (CG-3)," 'Water and Power Resources Ser-
9. "Standard Test Method for Flexural Strength of Concrete (Us- vice, U.S. Bureau of Reclamation, Denver, Aug. 1980, p. 26.
ing Simple Beam With Third-Point Loading)," (ASTM C 78-75), 18. Dungar, R., "El Cajon Hydroelectric Power Plant, Arch Dam
1981 Annual Book of ASTM Standards, Part 14, American Society Final Design, Static and Dynamic Stress Analysis," Motor Colum-
for Testing and Materials, Philadelphia, pp. 38-44. bus Consulting Engineers, Baden, May 1981, p. 25.

ACI JOURNAL I March-April 1984 165

You might also like