Application of the Indirect Tensile Test

To Stabilized l\Iaterials
THOMAS W. KENNEDY and W. RONALD HUDSON, Center for Highway Research,
The University of Texas

The importance of the tensile characteristics of the subbase of

rigid pavements can be demonstrated both from theoretical con-
siderations and from field observations. Information on the
tensile behavior and properties of treated and untreated subbase
material is limited primarily because of the lack of a satisfactory
tensile test. On the basis of a literature review concerned with
tensile testing, it was concluded that of the currently available
tensile tests the indirect tensile test has the greatest potential
for the evaluation of the tensile properties of highway materials.
This paper discusses tensile testing, theory of the indirect
tensile test, and factors affecting the test. In addition, the re-
sults of a limited testing program to evaluate the effect of such
factors as composition and width of loading strip, testing tem-
perature, and loading rate on the indirect tensile test param-
eters of strength, vertical failure deformation, and a load-
vertical failure deformation modulus for asphalt-stabilized and
cement-treated materials are included. On the basis of the
literature review and experimental investigation, it is recom-
mended that the indirect tensile test be used for evaluating the
tensile properties of stabilized materials and that the test be
conducted utilizing a 1. 0-in. wide stainless steel loading strip,
a loading rate of 2 in./min, and a testing temperature of 77 F.

•THE importance of the tensile characteristics of the subbase of a rigid pavement can
be demonstrated from both theoretical considerations and field observations, yet little
is known about behavior and design of subbase materials.
From available evidence it is logical to assume that the cohesive or tensile charac-
teristics of the subbase significantly affect pavement performance. Unfortunately, little
information is available on the tensile behavior and properties of treated and untreated
subbase material primarily because of the lack of a satisfactory tensile test. The pur-
pose of this paper is to evaluate tensile testing and to tentatively recommend a tensile
test and a testing procedure (!)·

TYPES OF TENSILE TESTS FOR HIGHWAY MATERIALS

Various tests and modifications have been developed and used for evaluating the ten-
sile characteristics of highway materials. These tests can be classified as (a) direct
tensile tests, (b) bending tests, or (c) indirect tensile tests.

Direct Tensile Test and Bending Test

The direct tensile test, which consists of applying an axial tensile force directly to
a specimen and measuring the stress-strain characteristics of the material, is simple

Paper sponsored by Committee on Strength and Deformation Characteristics of Pavement Sections and
presented at the 47th Annual Meeting .
36
 37

in theory and principle. However, serious difficulties have been encountered in prac-
tical applications. Major problems have included the addition of bending stresses due
to eccentricity or misalignment of the load and the addition of stress concentrations at
the loading grips. Another problem concerns the evaluation of test results; it is as-
sumed that the stress is distributed uniformly across the cross section, but it has been
reported that the maximum stress on the central cross section of a figure-eight bri-
quette is about 1. 75 times the average stress.
The bending test involves the application of a bending load to a beam specimen. It is
considerably simpler to conduct than the direct tensile test, requires less care in the
preparation of the specimens, and is favored by many engineers because the loading
conditions are similar to the field loading conditions of pavement materials. Basically,
this test involves two types of loading conditions: the common flexure test is conducted
by applying a load to a simply supported beam, and the cohesiometer test involves the
application of a bending moment to a specimen through a cantilever arm.
The results of the common flexure test are normally expressed as the modulus of
rupture or by relating the modulus of rupture to the tensile strength. The modulus of
rupture, however, is calculated by the standard flexure formula which assumes a linear
stress-strain relationship. Such a relationship does not exist for most material, and
even in the more elastic materials this assumption is seriously in error at failure con-
ditions. The net effect usually produces a modulus of rupture which is much higher

Loading Strip

Specimen

Figure l. The indirect tensile test.

38

than the actual failure stress. For concrete, it has been estimated that the modulus of
rupture is equal to or greater than two times the tensile strength.
One method of utilizing the modulus of rupture is to consider it to be an index of ten-
sile strength. A second method is to establish a relationship between the modulus of
rupture and the tensile strength, but this approach has not been too satisfactory since the
relationship has generally been assumed to be linear when in reality it appears to be
curvilinear.
The cohesiometer test consists of clamping a sample in the testing device directly
over a hinge. One end of the specimen is held fixed and the other is loaded through a
cantilever arm, producing failure. The load required to cause failure is used to cal-
culate the cohesiometer value (grams per inch of width corrected to a 3-in. height).
This value is empirical and has no theoretical counterpart.
The major criticisms of both types of bending tests concern the nonuniform and un-
defined stress distribution which exists across the specimen and the fact that the max-
imum tensile stress occurs at the outer surface. This latter condition accentuates the
effect of surface irregularities and may result in low indicated values of tensile strength.

Indirect Tensile Test

The indirect tensile test was developed simultaneously but independently in Brazil
and in Japan. The test involves loading a cylindrical specimen with compressive loads
distributed along two opposite generators (Fig. 1). This condition resultsinarelatively
uniform tensile stress perpendicular to and along the diametral plane containing 1the ap-
plied load. The failure usually occurs by splitting along this loaded plane.
Previous use of this test has generally been on concrete or mortar specimens; how-
ever, Thompson (2) found the test to 'be satisfactory for the evaluation of the tensile
characteristics oflime - s oil mixtures while Messina (3) and' Breen a nd Stephens (4, 5)
used the test for the study of asphaltic concrete. In addition, Livneh and ShklarskY" (6)
used the test in the evaluation of anisotropic cohesion of asphaltic concrete. From a
review of the literature concerned with the evaluation and use of the indirect tensile test,
a number of advantages and two disadvantages were found. The main disadvantage is
that the test loading conditions do not resemble those in the field; the second is that the
theory is more complex than the theory of the direct tension test and the bending test.
The six major advantages attributed to the test are the following:
1. It is relatively simple;
2, The type of specimen and equipment are the same as that used for compression
testing;
3. Failure is not seriously affected by surface conditions;
4. Failure is initiated in a region of relatively uniform tensile stress;
5. The coefficient of variation of the test results is low; and
6. Mohr's theory is a satisfactory means of expressing failure conditions for brittle
crystalline materials such as concrete.

Choice of Test
On the basis of the review of literature, it was concluded that of the currently avail-
able tensile tests the indirect tensile test has the greatest potential for the evaluation of
the tensile properties of highway materials. The main disadvantage attributed to the
test concerns its failure to duplicate field loading conditions. Although such conditions
may be desirable, the lack is not decisive and is more than offset by the many apparent
advantages of the test, as is the secondary disadvantage, lhal the theory is more com-
plex than for the direct tensile and bending tests. Thus, the indirect tensile test has
been given priority for use as a method for evaluating the tensile properties of stabilized
highway materials.

THEORY OF INDIRECT TENSILE TEST

According to the literature, the theory for the stress distribution for the indirect
tensile test was first developed by H. Hertz. Later A, Foppl and L. Foppl, S. Timo-
shenko and J, N. Goodier, M. M. Frocht, and R. Peltier considered the theory.
 39

Theoretical Development
Usually the theory of the indirect tensile test is developed from Frocht's equations
for stresses at a point. The distributions of stresses calculated from these equations
are shown in Figure 2 for the horizontal diameter and Figure 3 for the vertical diameter.
The vertical stress cry along the horizontal diameter is compressive and the magnitude
varies from a maximum of ;t~ at the center to zero at the circumference. The hori-
zontal stress cr x along the vertical diameter is a constant tensile stress of magnitude
rr t ~ ; the vertical stress cry is compressive and varies from a minimum of rr ~ ~ at the
2

center to a maximum of infinity at the circumference beneath the loads.

Under conditions of a line load, the specimen would be expected to fail near the load
points due to compressive stresses and not in the center portion of the specimen due to
tensile stress. It has been shown, however, that these compressive stresses are
greatly reduced by distributing the load through a loading strip. In addition, the hori-
zontal tensile stress along the vertical diameter changes from tension to compression
near the points of load application.

Deviation of Test From Ideal Conditions

The preceding development is an exact solution for the idealized case considered. In
reality the actual test deviates from the assumed ideal condition. The following devia-
tions should be considered.
Heterogeneous Nature of Material Tested-The theory on which this test is based as-
sumes a homogeneous material. Stabilized materials are normally heterogeneous not
homogeneous; nevertheless, the greatest application of the test has been with concrete,

a-K, Tension= -2P

1r 1
2
d [d
~
2
-4X
d +4X
J 2

. -_-2P
a-v, Compression Jrld
2
[ di+
4d X -I
4 2
J

Figure 2. Stress distributions on x-axis.

40
y

"'x ,Tension= !fd=Constant

-CD

"'y, Compression =
-2P[2 2 I]
-;!'" d-2y + dt2y - d

- IX)

Figure 3. Stress distributions on y-axis.

which is also very heterogeneous. In addition, the test has been used for the evaluation
of asphaltic concrete (3, 6), a nonhomogeneous material, and lime-soil mixtures (2). In
all of these cases, thet es t was found to_be satisfactory although undoubtedly errors were
introduced by the heterogeneous nature of the tested materials. With regard to this
problem, it has been concluded that although the effect on the general stress distribution
cannot be determined it is probably small enough to permit the use of the test.
Distribution of Applied Load-The theory of the test assumes a point load on a thin
disk which corresponds to line loading along a generator of the cylinder. Actually the
load is distributed over an area with an appreciable width because of the practice of ap-
plying the load through a loading strip. Studies concerning the effects of a load strip
on stress distribution have shown that the magnitude of the vertical compressive stresses
is significantly reduced and that the magnitude of the horizontal stress is virtually un-
affected near the center of the specimen but is changed to compression near the edges
(Fig. 4).
A number of investigations have indicated that a semisoft material is desirable as a
loading strip. It has been recommended that the loading strip should be soft enough to
allow distribution of the load over a reasonable area and yet narrow enough to prevent
the contact area from becoming excessive. The basic requirement for selection of the
loading strip is that it produce tensile rather than compressive or shear failures.
Deviation from Hooke's Law-It is assumed in the theoretical considerations of the
test that strain is proportional to stress. This does not hold in the case of concrete,
asphaltic concrete, and stabilized materials. Probably the worst case occurs with bi-
tuminous materials. In all of these materials, the modulus of elasticity or deformation
tends to decrease with increased stress. A nonlinear stress- strain relationship such
as this tends to relieve the more highly stressed parts of the specimen. This condition,
however, w()uld tend to increase the load r equired t o cause failur e in the s pecimen and
 41
c
2 0 2 4 6 8 10 12 14 16 18 20
0

.Q..
6
c:
"'E 0
·u 3
- 2p
"'
0.
(/) 0 O"X - C ( Ff d )

.: 2
.<: 20
0. 3
c"'
50
6
0

Ten1lon ComprH,ion

Figure 4. Horizonta I stress distributions on the y-axis for loading strip width equal to d/12.

to give higher strength values. Nevertheless, there is no apparent reason to question

seriously the results obtained from indirect tensile testing of nonlinear stress-strain
materials provided the specimen fails in tension.
It is also reasonable to conclude that the test is mo1·e applicable to brittle materials
and that some consideration and test evaluation would be desirable for materials such
as asphaltic concrete and bituminous stabilized materials before the test can confidently
be used for the evaluation of these materials.

EXPERIMENTAL EVALUATION AND DEVELOPMENT

OF THE INDIRECT TENSILE TEST
Some of the characteristics of the indirect tensile test and the materials tested which
may affect the test results are the following:
1. Load-deformation characteristics of the material tested;
2. Size and dimensions of the specimen;
3. Composition and dimensions of the loading strip;
4. Rate of loading; and
5. Testing temperature.
More important than the change in strength associated with increased loading rates
and decreased temperature is the change in the character of the stress-strain relation-
ship exhibited by the material. Both a decreased testing temperature and an increased
loading rate tend to produce more brittle behavior and a more linear stress-strain re-
lationship which is advantageous according to test theory.
Previous evaluation, both theoretical and experimental, has established the influence
of some of these factors. It has been shown that the length-to-diameter ratio of the
specimen tested has little effect on the resulting strength parameter, and it has been
shown that an increase in the overall specimen size results in more uniform strength
data, but slightly reduced average strength values.
It has also been established that the composition and dimensions of the loading strips
affect strength results and type of failure. However, previous tests do not indicate the
best type of material and dimensions of the loading strips. In addition, there is little
information on the effects of testing temperature and loading rate.
Unfortunately, most of the experimental evaluation of the indirect tensile test has
been conducted on concrete and has not included deformation measurements. This fact,
along with the lack of conclusive evidence concerning the most desirable composition
and width of the loading strips and the lack of temperature and loading rate information,
42

makes it important to evaluate the indirect tensile test using materials other than con-
crete and to include deformation measurements. The findings of such an evaluation
along with previously repor ted findings will aid in establishing standard test procedures
for future studies.
The objectives of this initial phase of investigation were to develop equipment and a
technique for conducting the indirect tensile test and, as a part of this development, to
evaluate the effect of (a) composition of loading strip, (b) width of loading strip, (c)
testing temperature, and (d) loading rate on several test parameters including the in-
direct tensile strength, vertical failure deformation, and a load-vertical deformation
modulus.

Experimental Program
The primary objective of the experimental program was to evaluate the effects pro-
duced by the composition of the loading strip, width of loading strip, testing tempera-
ture, and loading rate. The pl'imary statistical parameters for the evaluation were the
standard deviation or variance and the coefficient of variation used as measures of
dispersion.
The three test series which were conducted included samples of asphaltic concrete
and cement-treated gravel. The asphaltic concrete consisted of crushed limestone and
5. 3 percent AC-10; the cement-treated gravel was a rounded gravel treated with 6 per-
cent type I portland cement. All specimens were 4. 0 in. in diameter with a nominal
height of 2. 0 in. and were compacted using the Texas automotive gyratory shear com-
pactor. Details concerned with the equipment, mix design, sample preparation, and
curing of the asphaltic concrete and cement-treated gravel are given elsewhere (1).
In these preliminary tests, the following parameters were defined and evaluated:
1. Indirect Tensile Strength-

where
Pmax = maximum total load, lb;
t = average height of specimen, in. ; and
d = nominal diameter of specimen, in.
2. Vertical Failure Deformation-vertical deformation of the specimen in inches at
maximum load including the deformation in the loading strip (corrections were made for
deformations occurring in the neoprene load strip in some parts of the analysis). This
deformation was assumed to be equ.a l to the movement of the upper platen from the point
of initial load application to the point of maximum load as measured by a DCDT and re-
corded on the load-vertical deformation plot.
3. Tangent Modulus of Vertical Deformation-slope of the load-vertical deformation
relationship prior to failure as defined by a regression analysis. Approximately 10 p0ints
between the points of initial load and maximum load were obtained from the load-vertical
deformation relationship and analyzed by the method of least squares to obtain the slope
of a line through the points.
Because of space limitations only the test results associated with strength are in-
cluded in this paper. The findings associated with the vertical failure deformation and
tangent modulus of vertical deformation are included and discussed elsewhere (1).
Evaluation of Composition and Width of Loading Strip-The first phase of testing was
concerned with t)l.e evaluation of the lype of material used for the loading strip and the
width of the loading strip. Initially, plywood loading strips were considered and were
used in testing because of previous recomme.n dations. These previous studies, how-
ever, did not involve deformation measurements. Since the measured vertical deforma-
tion included the deformation of the loading strip and since plywood strips deform ap-
preciably, it was necessary to subtract the loading-strip defo1·mation from the deforma-
tion measured in order to obtain an estimate of the vertical deformation of the specimen.
 43
TABLE 1
SUMMARY OF DATA AND ANALYSIS OF VARIANCE OF EFFECT OF STRIP TYPE
AND WIDTH ON LOG VARIANCE FOR ASPHALTIC CONCRETE

Stainless Platens
Type of Loading Strip Neoprene
Steel (no strips)

Strip width, i n. o. 5 1. 0 o. 5 1. 0 ~

Number of specimens 8 8 8 8 8

Average, psi 105 108 106 103 111

Indirect
Tensile Standard deviation, psi 7. 0 2. 0 8. 1 4. 2 9. 8
Strength
Coefficient of variation, % 6. 7 1. 9 7.6 4. 1 8. 8

Degrees of Mean Significancea

Sour ce of Variation F
Freedom Squa res Level (%)

Strip type 1 0. 418 4.49 None

Strip width 1 1. 541 16. 6 5

Indirect
Tensile Interaction 1 0. 280 3. 01 None
Strength
Error 4 0. 093

Total 8
0
1f significance level is greater than 10 percent, it is called "none."

Such corrections were difficult and probably erroneous due to the fact that (a) wood is
heterogeneous and variable, (b) wood deforms appreciably at higher stresses, and (c)
wood does not exhibit a linear stress-strain relationship. For these reasons wood was
discarded as a possible loading-strip material.
Other strip materials investigated were stainless steel and neoprene. These two ma-
terials were chosen because they were readily available, easily specified, arid represent,

TABLE 2
SUMMARY OF DATA AND ANALYSIS OF VARIANCE OF EFFECT OF STRIP TYPE
AND WIDTH ON MEAN FOR CEMENT-TREATED GRAVEL

Stainless Platens
Type of Loading Strip Neoprene (no strips)
Steel

Strip width, in. o. 5 1. 0 o. 5 1. 0 m

Number of specimens 5 5 5 5 5

Average, psi 146 178 166 177 167

Indirect
Tensile Standard deviation, psi 12. 0 16. 5 30.8 15. 4 21. l
Strength
Coefficient of variation, ( 8. 3 9. 3 18. 6 8. 7 12. 6

Degrees of Mean Slgnificance3

Source of Variation F Level ('f,)
Freedom Squares

Strip type 1 476 1. 19 None

Strip width 1 1835 4. 58 5

Indirect
Tensile Interaction 1 1038 2. 59 None
Strength
Error 16 401

Total 19
0
Jf significance level is greater than 10 percent, it is called 11 none! 1
44
to a certain degree, extremes with regard to rigidity. Strip widths of 0. 5 and 1. 0 in.
were used. An additional variable involved the application of load directly through the
platens with no loadlng strips. Ali specimens were tested at 75 F at a loading rate of
0. 5 in./min. This phase of the testing was divided into two parts. The first part in-
volved the testing of asphaltic concrete, which was considered a questionable material
since it exhibits plastic characteristics rather than purely elastic characteristics as
assumed by theory and because there was lack of information concerning the use of the
indirect tensile test for testing asphaltic materials. The second part of the testing in-
volved cement-treated gravel, a more brittle material, which more closely approximates
the behavior of an elastic material.

-. 2
"'...
~

;
u
c
~"
>"
...
0
...J

0
-10 20 50 80 110 140

Test inQ Temperature, °F

Figure 5. Effect of testing temperature on the log-variance of indirect tensile strength of asphaltic
concrete.
 45

The experimental designs for these two series of tests were full-factorial, randomized
designs involving two types of strips and two strip widths. Analyses of variance of the
log-variances and some of the means were conducted for these variables. No statistical
analysis was conducted for the variable involving the direct application of load with no
loading strip, although subjective comparisons were made.
Findings Using Asphaltic Concrete-The initial test series in the evaluation of the
effect of composition and width of loading strip was conducted on asphaltic concrete
specimens. Basic statistical par~meters and results of the analysis of variance of the
log-va1·iances are summarized in Table 1. Similar information for the parameters of
vertical failure and tangent modulus of vertical deformation is summarized and dis-
cussed elsewhere (1).
The analysis of variance of the log-variances indicated that the type and width of ma-
terial used as a loading strip had no significant effect on the standard deviations of the
vertical failure deformations and the tangent moduli of vertical deformation. In the case
of the strengths, the type of material was found to have no significant effect; however,
the width of strip did produce a significant effect (a'. = O. 05).
There is apparently a definite advantage to using the 1. 0-in. wide strip because of
the reduced dispersion for both steel and neoprene. The standard deviations for the
specimens tested with steel strips are slightly higher than those for specimens tested
with neoprene; however, the difference is small and is not statistically significant in
this experiment. The very high dispersion values associated with the use of the platens
alone precludes the possibility of eliminating the loading strip.
On the basis of this analysis, it could be recommended that future testing be con-
ducted using a 1. 0-in. wide neoprene loading strip. Nevertheless, in view of the prac-
tical advantages of using steel and the small and insignificant differences between the
dispersion of the data obtained from the steel and neoprene, it i.s felt that a 1. 0-in. wide
steel loading strip is more desirable. Results published with regard to concrete and
mortar, however, have generally recommended a softer, more flexible loading strip
material. In addition, it has been reported that the width of the loading strip has an ef-
fect on the type oi failure. On the basis of the above recommendation and the lack of
sign.i ficant advantage of one material over the other, it was desirable to investigate the
effects of both type and widtl1 of loading strip on a more brittle material.
Findings Using Cement-Treated Gravel-The second test series in the evaluation of
the effect of composition and width of loading strip was conducted on cement-treated
gravel specimens. The data and the analysis of variance of the means of the strength
parameter are summarized in Table 2. Similar information for the other parameters
are given elsewhere (1).
The analysis of vanance of the log-variances indicated that the type and width of the
load strip had no significant effect on the variances of the test parameters. Hence, the
analysis of variance of the means in this case is justified and showed a significant ef-
fect (a= 0. 05) due to strip width with the 1. 0-in. strips resulting in higher average
·strengths.
Although not statistically significant (a ·= 0. 05), the 1. 0-in. wide strips in this ex-
periment produced less scatter of the strength values than the 0. 5-in. strips. Con-
sidering only the data for the steel strips, the 1. 0-in. strips resulted in a lower stan-
dard than the O. 5-in. strips; however, the reverse was true for the neoprene loading
strips. The minimum value for dispersion occurred with 0. 5-in. neoprene and the high-
est value occurred with the 0. 5-in. steel. The standard deviations for the 1.0-in. strips
for both steel and neoprene were essentially equal.
The best strip appears to be neoprene, as it did in the case of the test series on as-
phaltic concrete. Nevertheless, this slight advantage of neoprene over steel is not
statistically significant; therefore, it is felt that the use of steel loading strips is jus-
tifiable because of the many practical advantages of using steel strips. Analyzing the
findings for only steel loading strips indicates that the 1. 0-in. steel strips are better.
Recommendation Concerning the Composition and Width of Loading Strip-It is rec-
ommended tentatively that future testing utilize a loading strip composed of stainless
steel which is 1. 0 in. in width. This recommendation is based primarily on the many
practical advantages of using steel rather than the softer, more flexible neoprene.
46
700

600

Loading Rate, in./mln .

600 0.0 5 - - 0
0.14-··--·

--·
-~ 0.5 - - - 0
Cl.
2 .0 ------ 6
40
~
60 - - - - \ }
c
~
"' 400
u;
..
......
·~
c

~
:g
300
~

..
."'
~
~

200

100

0
-10 0 20 60 80 110 140
Testin9 Temperature, °F
Figure 6. Effect of testing temperature on indirect tensile strength of asphaltic concrete.

Evaluation of the Effects of Testing Temperature and Loading Rate-The second phase
of testing was concerned with the evaluation of the effects of testing temperature and
loading rate. The evaluation was conducted on asphaltic concrete because of its tem-
perature susceptibility. Testing temperature ranged from -10 F to 140 F ± 2 F; loading
rates ranged from 0. 05 to 6. 0 in./min. A split plot type experiment design with three
blocks or replications was used in this phase of the testing.
The analysis of variance of the log-variance of strengths indicates that temperature
has a significant effect (o: = 0. 01) on the standard deviation of strength, but that there
is no significant effect associated with loading rate. The reduction in variance in the
range between 50 F and 80 F observed in Figure 5 is statistically significant (o: = O. 05).
In Figure 6, a substantial change also occurs in the slope of the strength-temperature
relationship at or slightly less than a temperature of 80 F, indicating that the effects of
temperature are much more pronounced in the lower temperature range. At the lower
temperatures, the relationships become slightly erratic.
 47

Figure 6 indicates that the effect of load rate is not as great as the temperature ef-
fect. A possible exception can be seen for the strength averages obtained at very low
loading rates. There would appear to be a substantial increase in the mean value as
the loading rate is increased at these low rates, especially at low testing temperatures.
It is recommended that future testing be conducted at room temperature (77 F) and
at a loading rate of 2. 0 in./min. This temperature was chosen because (a) it approxi-
mates the lower temperature range in which the strength and tangent-modulus param-
eters were relatively uniform and non-temperature susceptible, (b) it approximates the
lower limit of the temperature range exhibiting reasonably low dispersion, (c) it has
previously been used as a standard testing temperature, and (d) it is fairly close to the
normal temperature of air conditioned laboratories and, thus, does not require special
equipment or facilities for substantially raising or lowering the temperature. The load-
ing rate of 2. 0 in./min. was chosen primarily as a compromise. At slow loading rates
the magnitudes of the test parameters were more susceptible to loading-rate changes
than at higher rates. In addition, the theory assumes a linear-stress-strain or brittle
characteristic for the material being tested, and a more rapid loading rate tends to
produce a more brittle behavior. At the very rapid loading rates, however, the test is
more difficult to conduct. At 2. 0 in./per min. the indirect tensile test was easy to
conduct, and this loading rate is above the range in which the test parameters appeared
to be very susceptible to changes in loading rate.

CONCLUSIONS AND RECOMMENDATIONS

Conclusions
1. Review of existing information indicates that the indirect tensile test is the best
test currently available for determining the tensile properties of highway materials.
2. From this information and the results of a limited testing program, the indirect
tensile test appears to be a feasible method for evaluating the tensile characteristics of
stabilized subbase materials although previous use of this test has generally been with
concrete.
3. Primary characteristics of the indirect tensile test and the materials tested
which may affect the test results are (a) load-deformation characteristics of the ma-
terial tested, (b) size and dimensions of the specimen, (c) compositfon and dimensions
of the loading strip, (d) rate of loading, and (e) testing temperature.
4. Characteristics and properties of the material being tested are not considered
in the theoretical development of the test, except as a limiting tensile strength. The
materials are assumed to have linear-elastic stress-strain characteristics. Although
1.1any deviations from the assumed conditions exist and although the use of the simple
formula ST= (2 Pmax )/(TTt d) introduces small errors in the results, there does not ap-
pear to be any evidence that the error is significant as long as the specimen ultimately
fails in tension.
5. The indirect tensile strength has been shown both theoretically and experimen-
tally to be independent of the length-diameter ratio. It has been assumed that other
indirect tensile parameters such as failure deformations and load-deformation charac-
teristics are also independent of this ratio.
6. The indirect tensile strength is reduced slightly by an increase in the overall
size of the specimen, and the dispersion of the data is reduced.
7. On the basis of the literature review, it is concluded that the composition and
width of the loading strip have a definite effect on the stress distribution in the speci-
men, the test results, and the mode of failure.
8. Wood which has often been recommended as a loading strip was eliminated from
future use by this project because of practical difficulties associated with measuring
deformations in the specimen.
9. It is recommended that steel be used as a loading strip because of its significant
practical advantages even though experimental results presented in this report indicate
that neoprene is a slightly better loading strip material than steel.
10. A 1. 0-in. wide strip is recommended over a 1/2-in. width because of the reduced
data dispersion.
48

11. Under the conditions of the tests performed in this study, temperature had a
highly significant effect on the average test results. The parameters of strength and
load-deformation modulus were less temperature susceptible and more uniform at
testing te mperature s of 80 F and above.
12. Under the conditions of these tests, loading rate had a significant effect on the
average test results. The effect, however, was not as great as that produced by testing
temperature.

Recommendations
Based on the preceding conclusions, certain decisions concerning parameters in the
indirect tensile test have been made. These parameters will be fixed tentatively for
evaluation tests to be conducted in the project in the near future.
1. The specimen will be as large as is practical in order to obtain more uniform
test results and a better measure of the average of the test results. It is planned that
ultimately samples will be 6 in. in diameter with heights in the range of 8 to 12 in.
2. The loading strip will be stainless steel with a width of 1 in.
3. The loading rate will be 2. 0 in./min.
4. The testing temperature will be room temperature in the range of 75 to 77 F.

REFERENCES
1. Hudson, W. Ronald, and Kennedy, Thomas W. An Indirect Tensile Test for
Stabilized Materials. Research Report No. 98-1, Center for Highway Research,
Univ. of Texas, June 1967.
2. Thompson, M. R. The Split-Tensile Strength of Lime - Stabilized Soils. Lime
Stabilization. Highway Research Record 92, p. 11-23, 1965.
3. Messina, R. Split Cylinder Test for Evaluation of the Tensile Strength of Asphalt
Concrete Mixtures. Unpublished Master's thesis, Univ. of Texas, Jan. 1966.
4. Breen, J. J., and Stephens, J. E. Split Cylinder Test Applied.to Bituminous Mix-
tures at Low Temperatures. Jour. of Materials, Vol. 1, No. 1, ASTM,
March 1966.
5. Breen, J. J., and Stephens, J. E. Fatigue and Tensile Characteristics of Bitu-
minous Pavements at Low Temperatures. Report No. JHR 66-3, School of
Engineering, Univ. of Connecticut, July 1966.
6. Livneh, M., and Shklarsky, E. The Splitting Test for Determination of Bituminous
Concrete Strength. Proc. AAPT, Vol. 31, p. 457-476, 1962.