H. E.

BOFINGER

M.E., A.M.I.E.Aust., Lecturer in Civil Engineering, University of Queensland

THE FATIGUE BEHAVIOUR OF SOIL-CEMENT

A number of soil-cement specimens, moulded from a heavy black clay,
were subjected to a series of load repetitions to determine the fatigue
behaviour of soil-cement under compressive, tensile and flexural stresses.
Two important conclusions were drawn from the results. Firstly, a
fatigue limit does exist irrespective of the method of stressing. While
the reduction in strength after tensile and flexural stressing is high, the
application of numerous compressive load cycles has little effect on the
strength of the soil-cement. Secondly, the tensile and flexural fatigue
limit appears to be independent of the cement content within the range
investigated.*
The author discusses the accuracy of several accepted laboratory
tests for soil-cement pavement design.

INTRODUCTION effect of repeated loadings is much more

1. A number of types of tests have important than the static strength of the
been used in the past to estimate the per- pavement material. For this reason all of
formance of cement stabilized soils in pave- the above mentioned tests are deficient and
ments but no test can be accurately corre- it is improbable that their continued use
lated with pavement performance. One of can be justified. In this paper the author
the best methods available at the present presents the results of repeated loading
time is the wet-dry test but it is not in tests on a cement-stabilized heavy 'black'
general use due to the extended period clay and discusses their significance in re-
needed for the test. The unconfined com- lation to the static strengths.
pression test at 7 days after a sample is CEMENT STABILIZED PAVEMENTS
moulded is most commonly used. In some
cases the samples are tested at their mould- 3. Pavements have been classified
into three groups: flexible, semi-rigid, and
ing moisture content and in other cases
rigid. Most engineers include cement stab-
samples are soaked for varying periods
ilized pavements in the second group with-
prior to test. The cylinder size has been
out taking account of the variation in the
varied considerably and the period of cure
properties with change in cement content.
has been increased to 14 or 28 days. How-
At low cement contents the material has
ever, the correlation between compressive
quite a low 'elastic' modulus while at higher
strength and field performance is very poor.
cement contents the elastic performance is
The CBR test also provides an inaccurate
similar to concrete. The borderline between
estimation of field performance. During the
semi-flexible pavements and rigid pave-
past few years the flexural strength has
ments is quite indistinct.
been used increasingly to indicate the load-
carrying capacity of a cement stabilized 4. The clay used in this experimental
work is a heavy 'black' clay from Ipswich,
pavement.
Queensland, which X-ray tests have shown
2. Pavements in practice are subjected to be predominantly bentonite. Routine
to a repetition of loads and therefore the classification tests gave the following values:

*The fatigue strength of a material is defined as the stress such that if a smaller stress
is applied to the material an infinite number of times it will not cause failure of the
material.

12 AUSTRALIAN ROAD RESEARCH

 BOFINGEH: THE 14ATIGUE BEHAVIOUR OF SOIL-CEMENT

Liquid limit = 116% 8. When bridging occurs over soft

Plastic limit = 44% material the pavement must be subjected
Linear shrinkage = 22% to repeated flexural loading from subse-
quent traffic. A wheel load sited directly
Particle size: over the soft spot will be taken by the pave-
Sand 2% ment and little stress will be placed in the
Silt 28% soft material. The maximum stress will be
Clay 70% induced in the pavement at the edge and
Activity 1.03 centre of the soft spot, and failure will
A series of soaked unconfined compres- occur if the tensile strength is exceeded at
sion tests at 7 days using varying cement either point.
contents gave the moduli of deformation 9. If failure occurs in tension it is
shown in TABLE I. apparent that the unconfined compression
test may give misleading results, especially
5. The author feels that a cement con-
as it is a static test. Similarly other tests
tent as high as 16 per cent is significant
in common use can lead to erroneous re-
in this case because of the very high per-
sults.
centage of clay in the soil. In fact, this soil
lies in the range of materials which are 10. The testing programme for this re-
generally considered unsuitable for stabili- search covers the following:
zation with cement. (a) Static
6. From the values given above it can (i) Unconfined compression
be seen that soil-cement pavements made (ii) Tension
from this soil will have sufficient rigidity (iii) Flexure.
to bridge over localized soft spots unless (b) Repeated loading
the pavement material is damaged by (i) Unconfined compression
shrinkage during curing. The extent of (ii) Tension
shrinkage cracks is uncertain. If the cracks (iii) Flexure.
only form in the top of the layer they will EXPERIMENTAL WORK
not prevent the pavement from acting as a
11 . The classification test results for the
slab. soil have been listed previously in this
7. An example has been reported in paper. All of the black soil was crushed till
Texas where a 9 in. pavement of soil- the maximum clod size was 1/4 in. The
cement supported highway traffic for sev- specimens were moulded at optimum mois-
eral months after a hole several feet wide ture content, dipped in wax and allowed to
had formed beneath the pavement. cure for 14 days at room temperature.
Three cement contents were used in the
TABLE I series, viz, 8, 12 and 16 per cent. The
UNCONFINED COMPRESSION TESTS AT SEVEN DAYS variation in the maximum dry density cor-
responding to the optimum moisture con-
Cement Content Modulus of Deformation (p.s.i.) tent is set out in TABLE II.
12. Compression specimens were com-
80/4) 3 x 105 pacted in a Harvard miniature mould of
12% 9 x 105 1-1),T in. diameter and 2,816 in. high.
16% 2.1 x 106
The tension specimens were moulded hori-
zontally in a cement-mortar tension mould

June, 1965 13
 BOFINGER: THE FATIGUE BEHAVIOUR OF SOIL-CEMENT

TABLE II
be very inaccurate because of the practical
difficulty of preventing moisture change in
VARIATION IN MAXIMUM DRY DENSITY WITH
DIFFERENT CEMENT CONTENTS
the sample during the extended period of
the test.
Optimum Maximum Dry
Cement Content
Moisture Content Density (p.c.f.) 15. A major aim of the research was
to determine whether a 'fatigue limit' exists
80
for a cement stabilized soil, and if so,
0 38% 86
whether it is similar for tension, compres-
12% 38% 87 sion and flexure. The static strength of
16°0 38% 88 the material was determined from the mean
of at least three tests for each type of
using a 3/4 in. diameter, non-standard ham- specimen and cement content. Repetitive
mer. Four layers were used to give good loads of decreasing magnitude were then
uniformity across the specimen and the applied and the number of cycles to cause
target density was the maximum dry density failure was recorded.
shown in TABLE II. Flexure specimens were 16. A set of tensile samples containing
compacted to the maximum dry density by 12 per cent cement was tested to ascer-
static compression. The specimen size was tain the effect of the magnitude of the
8 in. by 13/4 in. wide by 2 in. deep. minimum load on the 'fatigue' curve. Simi-
TESTING
lar tests were not conducted on compres-
compression or flexure specimens.
13. Static strengths were measured by
loading the specimens through a proving 17. Another set of tensile specimens
ring. with a 12 per cent cement content was
tested at 12 c/min instead of 6.0 c/min.
14. Repeated load tests were conducted Similar tests were not conducted on either
on a machine (Fig. 1) which was developed compression or flexure specimens.
in the Soils Laboratory of the Department
Variable eccentricity
of Civil Engineering at the University of to allow adjustment
Queensland. The load is applied by a pre- of minimum load.

determined extension of a tension spring

connected to an eccentric pin. Several ad-
justments are possible. Springs of different
flexibilities allow the maximum load to be Two loading cages which
varied continuously between 7 lb and 150 fit between the Tension
Spring and fixed base
lb with an accuracy of 2 per cent. Frequent are used for Compression
recalibrations of the springs prevent inac- and Flexure specimens.
Tension Spring
curacies due to distortion. It was essential
to maintain a minimum load on the samples
to prevent movement during testing. Adjust-
ment of the eccentricity of the pin allowed Jaws for tension
specimen.
the minimum load to be varied. The mach-
ine is driven through a multi-speed gear-
_Fixed base
box which enables the loading rate to be
reduced to almost zero c/min. The major Adjustment for
max. load.
portion of the testing was conducted at
60 c/min. Investigations at low speeds can Fig. 1—Fatigue machine

14 AUSTRALIAN ROAD RESEARCH

 BOFINGER: THE FATIGUE BEHAVIOUR OF SOIL-CEMENT

TABLE HI cycles further specimens were tested at 95

STATIC LOADING RESULTS per cent of the static strength. In several
Unconfined Compressive
cases the specimens failed within the first
Cement Content
Strength (p.s.i.) few cycles, which pointed to the fact that
these specimens were slightly inferior to the
8% 100
average specimen. Any specimen which
withstood the initial loading cycles remained
12% 155
unbroken even after prolonged testing. The
16% 278 results are shown graphically in Fig. 2.
TENSION
RESULTS Static loading
18. It should be noted that all results 21. Static loading results are shown in
are the mean of at least three tests. TABLE IV.

COMPRESSION Repeated loading

Static loading 22. The results from the tests at 60

c/min are shown in detail in Fig. 3. A
19. Static compressive loading results
`fatigue limit' exists at a tensile strength
are shown in TABLE M.
of approximately 16 p.s.i. and is independ-
Repeated loading ent of the cement content within the range
20. When a repetitive compressive load from 8 to 16 per cent. The author feels
is applied to a soil-cement specimen there justified in limiting the curves at 16 p.s.i.,
appears to be no reduction in the strength because some specimens remained unbrok-
of the soil-cement from the static value. en even after 600,000 cycles. Because of
Initially the repetitive load applied was 80 the time necessary to conduct a test it is
per cent of the static strength. As this not practicable to continue the tests be-
load did not cause failure after 500,000 yond 600,000 cycles at a rate of 60 c/min.

16 Z Cement

2
Unbrol en
Test p ecim en

co
w 12 2 Cement
cr 1
•--

Unbr oken
Test specimen
8 Z Cement
10

Unbr ken
Test specimen

N.B. Samples which took Initial Loading

remained unbroken.

a
l 10 100 1000 10 000 100 000 1 000 000
N° OF LOADING CYCLES

Fig. 2—Compression tests

June, I965 15
 TIOFINGER: THE FATIGUE BEHAVIOUR OF SOIL-CEMENT

TABLE IV used was 12 per cent. Fig. 5 shows a com-

STATIC LOADING RESULTS parison between specimen sets with differ-
Cement Content Tensile Strength (p.s.i.) ent minimum loads. There is no significant
difference between the curves, which indi-
cates that the size of the minimum load is
8% 30
not important.
12% 39
FLEXURE
16% 45 Static loading

25. The static strengths obtained for

Slow repeated loading the flexure specimens are set out in TABLE
The curves shown in Fig. 4 com- V. It should be noted that the stress has
pare the results of a series of tests con- been calculated as the extreme fibre stress
ducted at 12 c/min with those obtained at assuming a triangular stress distribution in
60 c/min. A cement content of 12 per cent the beam.
was used in both cases. The slower loading 26. Fig. 6 shows the flexural strength
rate causes failure at a lower number of of the soil-cement specimens after the re-
load cycles for the range of 1 to 10,000 peated application of loadings. A fatigue
cycles. However, for a greater number of limit does exist for soil-cement which ap-
cycles the difference between the curves is pears to be independent of the cement con-
within the limits of experimental error, and tent within the range from 8 to 16 per cent.
the fatigue limit remains at 16 p.s.i. The stress at the fatigue limit is 28 p.s.i.,
assuming that the distribution of stress
High minimum loading
through the beam is a straight line. Further
24. During normal testing the minimum comments on the bending stress distribu-
load was maintained at approximately one tion will be made during the discussion of
quarter of the maximum load. A series of these results. When the number of repe-
tension tests was completed in which the titions of load is less than 5,000 it is neces-
minimum load was raised to three-quarters sary to apply a stress almost equal to the
of the maximum load. The cement content ultimate static strength before failure will

• -...- Denote Tests which haw

ho been
e.-- }
16% Cement stopped prior to Failure.

4
12Y. - •
S TR ESS ( lb. 117 2)

8% Cement
-111111114

01

20

10

10 100 1 000 10 CM 100 000 1 0000

N° OF LOADING CYCLES

Fig. 3—Tension tests

AUSTRALIAN ROAD RESEARCH

16
 110FINGER: THE FATIGLE BEHAVIOUR OF SO1 L-CENT ENT

CEMENT CONTENT 12%

is Samples loaded at 60
5I
x do. 12 cp.rn.

in
‘N, x •
3
--....,
-........
x,_......

2 ---.4..._
_ ----* --w— —x --...- Unbro en Test
spec. ens.

0
1 10 /00 1000 10 000 100 Cal 1 000000
If? OF LOADING CYCLES

Fig. 4—Tension tests—samples loaded at 60 c/min and 12 c/min

occur. As the number of repetitions in- drop in the load necessary to cause the
creases from 5,000 to approximately failure of soil-cement in compression. How-
100,000 the load necessary to cause failure ever, a considerable reduction occurs in the
is reduced by up to 30 per cent. repetitive load necessary to cause tensile
failure. It is likely that this phenomenon is
DISCUSSION OF RESULTS a result of the formation of micro-cracks
27. Several important facts have in the soil-cement structure. If we consider
emerged from this series of tests. For the that these micro-cracks are not continuous
black clay used, a repetition of loading of through the soil-cement mass a compressive
up to 500,000 cycles causes no significant load will tend to close the cracks, allowing
CEMENT CONTENT 12%
• MINIMUM LOAD = 25% of MAXIMUM LOAD

5• do c 75%

il
l t -.........
-----........
•
I -----.....,
U)
---.........
---.....„, ----z.
U)
-___ ,......
2C

Utibrok en
Test E p ec im en

I
10 100 1 000 10 OCO 100 000 1 000 000
N? OF LOADING CYCLES

Fig. 5—Tension tests—minimum load equal to 25 per cent and 75 per cent of maximum load

June, 1965 17
 11( )I'1 LAGER: THE FATIGUE BEHAVIOUR OF SOIL-CEMENT

TABLE V 28. If the micro-cracks in the original

STATIC STRENGTH OBTAINED FOR FLEXURE structure are responsible for failure in
SPECIMENS `durability' tests such as the wet-dry test,
Cement Content Load (Ib) Stress (p.s.i.)
the results given above show why some
correlation has been obtained between the
durability test and the fatigue life of the
8% 50 43 pavement.
12% 62 53.2 29. Before considering the flexural
16% 72 i 61.7 strength tests it will be helpful if we exam-
ine the stress-strain behaviour of soil-cement
and the stress distribution in bending. At
the load to be transmitted across without
the present time the author has not devel-
necessarily causing excessive stress concen-
oped a satisfactory method of obtaining the
trations. Similarly, the strength of the sound tensile stress-strain relationship for soil-
portion of the skeleton will hold the frac- cement. An assumption has therefore been
tured sections in a position of interlock, made that the deformation behaviour for
thus allowing the shear strength of the tension is the same as for compression. Fig.
material to remain almost at its original 7 is a diagrammatic representation of the
value. When tensile stresses are applied stress-strain behaviour of a soil-cement
there will certainly be a heavy stress con- specimen moulded from black clay and sub-
centration at the extremity of certain cracks jected to a number of loading cycles. If we
causing progressive failure in the sound assume that plane strain occurs across the
material. In this case ultimate failure will section of the beam then the approximate
occur at a much lower stress. bending stress distribution at the ultimate
0

16% n.nt

• 12 % Cement

51

♦ 8 % Cement
........,

3t
•
Unbroker
Test spe imen

1 -——

0
10 100 1000 10 000 100 000 1 000 000
N° OF LOADING CYCLES

Fig. 6—Flexure tests

18 AUSTRALIAN ROAD RESEARCH

 TIOPINGER: THE FATIGUE BEHAVIOUR OF SOIL-CEMENT

Fig. 9—Triangular stress distribution across beam

under the same applied moment in Fig. 9.

It is concluded therefore that the author
has obtained a very good correlation be-
STRAIN
tween the tensile and flexural tests.
30. When the flexural beams are repeat-
Fig. 7 Typical soil-cement stress-strain curves for
several repetitions of loading edly loaded at a stress which is lower than
the ultimate strength, the failure will be
load is given in Fig. 8. By considering Fig. similar in most respects to a tensile fatigue
8 we can reconcile the tensile strengths failure. Initially, after a number of loading
obtained from the direct tension test and repetitions the stress which occurs in the
the flexural test. For a cement content of extreme fibres will cause them to crack in
8 per cent the tensile strength is 30 p.s.i. tension fatigue and increase the stress on
compared with an extreme fibre stress in the interior fibres. Progressive cracking will
flexure of 43 p.s.i. It should be remem- occur upon further repeated loading and it
bered that this flexural strength is calculated will continue until the beam collapses. It
using the assumption of a triangular stress follows that a reduction in the flexural
distribution across the beam (Fig. 9). An fatigue curves is not expected until a cer-
approximate analysis of Fig. 8 shows that tain number of load repetitions have
the maximum stress is approximately 75 occurred, whereas the tensile curves have
per cent of the maximum stress developed a significant drop even after a low number
of cycles. The ultimate fatigue limit appears
to be similar for both the tensile and flex-
ural curves.
31. Consideration must be given to the
significance of the value of the fatigue lim-
its. Both the flexural and tensile curves
tend to the same fatigue limit irrespective
of the cement content of the specimens.
It is therefore pointless to include excess
cement in soil-cement pavements if this
type of behaviour is general for all types of
Fig. 8—Stress distribution across beam assuming plane
soils and after longer periods of curing.
strain conditions Much more investigation must be completed

June, 1965 19
 BOFINGER: THE FATIGUE BEHAVIOUR OF SOIL-CEMENT

before the desirable cement content can be to cause failure in tension and flexure
estimated. 'Slab' action may therefore occur specimens. Compressive stresses which are
even though the cement content is not very lower than the ultimate strength can be ap-
high and consideration of the thickness of plied almost indefinitely without causing
the soil-cement layer may be very import- failure. The compressive strength test can-
ant. not be related to the behaviour of soil-
cement pavements because it bears no re-
CONCLUSIONS lation to the fatigue behaviour of soil-
32. For the soil used in this series of cement. The tensile fatigue limit should be
tests it has been shown that a number of used as the basis of design for soil-cement
loading cycles reduces the stress necessary pavements.

BIBLIOGRAPHY
1. Larew and Leonards, A strength criterion for repeated loads, Proc. H.R.B. (1962).
2. Influence of stabilizers on properties of soils and soil aggregate mixtures, H.R.B. Bull.
282.
3. Stabilization of soils with portland cement, H.R.B. Bull. 353.

20 AUSTRALIAN ROAD RESEARCH