Documente Academic
Documente Profesional
Documente Cultură
Degradation and improvement of mechanical properties of rock under triaxial compressive cyclic
loading
i) Senior Lecturer, School of Civil, Environmental and Mining Engineering, Adelaide, Australia.
ii) PhD student, School of Civil, Environmental and Mining Engineering, Adelaide, Australia.
ABSTRACT
An extensive experimental study on mechanical properties of a sandstone subjected to cyclic loadings are presented
and discussed in this study. Cyclic loading test were performed in triaxial condition at confining pressure of 4 MPa.
The mechanical properties of the rock were altered by cyclic loading; dependent upon the applied stress level during
cyclic loading the rock may experience damage or hardening as a result of cyclic loading. It was identified that a
critical maximum normalized deviator stress (i.e. deviator stress at the beginning of unloading normalized by rock
peak strength) exists between 92-95% which defines the limit between hardening and damage behavior during cyclic
loading. If the cyclic loading deviator stress is lower than this critical boundary, the peak strength my increase in post-
cyclic monotonic loading. This increase in peak strength is directly proportional to the normalized deviator stress in
unloading; the amount of increase in peak strength decreases with a decrease in cyclic loading deviator stress. In cyclic
damage tests, in general, the increase of Poisson’s ratio, v, and the degradation of tangent Young’s modulus, Etan,
indicate the progressive accumulation of damage in the rock sample over the loading cycles. In cyclic hardening tests,
however, v values for the cyclic hardening test remained fairly constant and Etan values slightly increased. Moreover,
it was found that in cyclic damage tests the amount of axial and lateral strains cumulated during cyclic testing in much
greater than those values for cyclic hardening tests.
http://doi.org/10.3208/jgssp.v05.017 71
condition. For the first time, damage and hardening without damaging the sample. As schematically shown
behaviour of rocks during cyclic loading, in terms of in Fig. 1a, the deviator stress value at this indicator point,
peak strength and deformability, is investigated by qid, always provides a constant value for the ratio of
undertaking a series of cyclic triaxial tests. qid/qf.
2 EXPERIMENTAL METHOD
2.1 Sample preparation and testing set-up
For this study the Hawkesbury sandstone which is a
medium-strong sedimentary rock was selected.
Hawkesbury sandstone is an early Middle Triassic
(Anisian) formation widely exposed in the Sydney Basin
in Australia. The rock has an average UCS of 44 MPa
and average density of 2.3 gr/cm3. Cylindrical specimens
were prepared by coring the rock blocks. The diameter
of the specimens was 42 mm. The end faces and sides of
the specimen were prepared smooth and straight
according to the ISRM standard. The aspect ratio (i.e.
length to diameter ratio) of the samples was maintained
at 2.4 for all the samples.
Tests were conducted by using a closed-loop servo-
controlled testing machine with a loading capacity of a)
250 kN. The loading rate can be adjusted from 0.0001 to
600 mm/min. The triaxial cell is equipped with a linear
variable differential transformer (LVDT) to measure
axial displacement and control axial loading. The axial
and lateral strains were measured with pairs of axial and
lateral strain gauges. To prevent the strain gauges from
being damaged inside the Hoek cell, a novel method to
protect strain gauges was developed and applied
successfully (Taheri et al. 2016b).
3.2 Loading method
As demonstrated by Bastian et al. (2014),
Hawkesbury sandstone is not completely a homogenous
rock. To investigate the effect of cyclic loading on
mechanical properties, it was essential to have identical b)
samples. To remove variations among samples they Fig. 1. Loading method; a) Use of Esec=max as an indicator point
cored all the samples in a same direction using a rock to predict peak Strength, b) Applying a pre-monotonic before
block. However, the uniaxial compressive strength of the cyclic loading
samples were varied up to 25%. In order to analyze the
effect of cyclic loading on rock mechanical properties, it Therefore, in this study, a pre-monotonic loading was
is essential to eliminate the effect of sample variation. To applied in each cyclic loading test which was used to
solve this problem, a method to predict rock peak obtain at maximum Esec, without failing the sample.
strength, qf, without failing the specimen is required to The peak strength, , was estimated by using
allow results to be confidently attributed to cyclic divided by the constant, c. Constant c value were
loading effect, rather than estimating rock strength from obtained to be 0.94 to estimate the peak strength of each
previous tested specimens under monotonic loading sample in the pre-monotonic loading under constant
assuming that all the samples are completely identical. loading rate of 0.05 mm/min. When Esec maximum point
This method is briefly explained below as it is discussed is reached, sample fully unloaded and then subjected to
in detail by Taheri et al. (2016a). a cyclic loading under a pre-defined cyclic unloading
Analysing the results of monotonic tests on the stress level, qu, and unloading amplitude, qb. As shown
sandstone samples, it was found that during monotonic by Bastian et al. (2014) variations among the
loading the secant Young’s modulus, Esec, evaluated by Hawkesbury sandstone specimens in terms of
the external LVDT, keeps increasing until the axial stress deformability parameters is negligible, therefore,
come to a point closer to the failure point, after which the normalization of deformability parameters wasn’t
Esec starts decreasing. The results showed that this point required. Fig. 1b shows the testing method adopted for
is a reliable indicative point to predict peak strength this study. If the rock specimen did not fail after 1300
cycles, a monotonic load was applied until failure in
72
order to examine the effect of long cyclic loading on qf next section), if the applied stress level in cyclic loading
value. However, to investigate effect of number of cycles is considerably lower than the monotonic peak strength,
on mechanical properties of sample which does not fail rock won’t fail during cyclic loading and it can even
during cyclic loading, in few number tests number of become stronger.
cycles before post-monotonic loading were varied
ranging between 10-10,000 cycles.
To study the effect of stress level on rock behaviour
during cyclic loading, the deviator stress at which cyclic
loading was initiated, qun, was varied. In addition, the
cyclic loading amplitude, qb, was set at 0.5qun, 0.75qun,
and 1.0qun.
4 TEST RESULTS
4.1 General behavior
Several specimens tested in this study failed as a
result of cyclic loading. The fatigue stress for those
specimens was found to be smaller than the predicted qf,
which indicates a progressive damage mechanism due to
cyclic loading. Many samples, however, didn’t fail
during cyclic loading after at least 1300 cycles.
Therefore a monotonic load was applied until failure. It
was found that in all the tests which did not fail due to
cyclic loading, the deviator stress at failure was higher
Fig. 2. Improvement in rock peak strength due to cyclic loading
than the monotonic peak strength. This indicates that the
strength of the rock specimens increased as a result of
cyclic loading. In other words, if the stress level at the
beginning of cyclic loading is not high enough to
generate a failure in sandstone due to progressive cyclic
loading damage, the rock strength, as a result of cyclic
loading for a long period of time, may be increased.
Based on these results, if the unloading stress value is
lower than a certain value (i.e. boundary) which is called
hereafter “Cyclic Damage Stress (CDS)” value, rock
experience hardening in mechanical properties as a result
of cyclic loading. Fig. 2, schematically demonstrates
how rock monotonic loading peak strength can increase
during post-monotonic test after a cyclic hardening test
if unloading stress level is lower than CDS. The
combination of damage and hardening effect of cyclic
loading on peak strength has been observed by few other
Fig. 3. Cyclic damage-hardening mechanism schematic model
researchers (Singh 1989, Burdine 1963, Ma et al. 2013)
and schematically is demonstrated in Fig. 3.
4.2 Stress-strain behavior 5 VARIATIONS OF MECHANICAL
Fig. 4 shows typical stress—strain results for PROPERTIES
specimens which failed and didn’t fail during cyclic
loading. Figs. 4a & c demonstrate deviator stress versus 5.1 Peak strength
axial and lateral strain values whereas Figs. 4b & d Figs. 5a & b shows the plot of normalized deviator
exhibit deviator stress vs. volumetric strain. The stress- unloading stress versus number of cycles at failure for
strain results in Figs. 4a & b are very dense until a stage specimens with 75% and 100% unloading amplitude
where stress-strain loop becomes loose until sample respectively, where sample is failed during cyclic
fails. Figs. 4 c & d shows the typical results for a sample loading (i.e. cyclic damage tests). It may be seen in these
which did not fail during cyclic loading. As can be seen figures that when unloading amplitude is constant, a
in these figures, the sample does not experience much higher normalized unloading stress will result in failure
axial, lateral and volumetric plastic deformation after after fewer cycles. A strong linear correlation can be
1300 cycles as compare to the cyclic damage test. As observed in the results.
explained earlier (and will be discussed in detail in the
73
were choose which failed almost after similar number of
cycles (~100) but under different unloading amplitudes
of 50%, 75% and 100%. The figure shows that a higher
unloading amplitude will results in higher normalized
unloading stress when sample failure occurs in the
similar number of cycles. Therefore, fatigue life
becomes longer with a decrease in unloading amplitude.
a)
a)
b)
b)
d)
Fig. 4.Typical stress-strain results a) & b) a cyclic damage test; c)
& d) a cyclic hardening test
74
Fig. 7 presents the amount of peak strength increase which are produced as a result of breakage of weak
observed in post-monotonic loading after a cyclic bounding. As can be seen in Fig. 4c & d, in a typical
loading in cyclic hardening tests. General trend of the hardening test, axial strain increases and volumetric
results for three series of tests undertaken under different strain doesn’t increase which means that sample
unloading amplitudes (i.e. 50%, 75% and 100%) are experience a slight shrinkage until end of cyclic loading.
presented by solid lines. There values has been This can be considered as a confirmation for this
determined through an accurate post-test analysis by statement. As a result of this, the sample experience
using the precise maximum Esec values obtained during compaction during the test and therefore, exhibit higher
each pre-monotonic loading (refer to Taheri et al. peak strength when subjected to a monotonic loading
2016a). after cyclic loading. Analyzing variations of rock
Based on results presented in Fig. 7, it is stiffness and Poisson’s ratio and comparing the results
hypothesized that there exists a value of qun/qf which lies with those obtained by cyclic damage tests confirms this
between 92-95% that indicates the critical boundary of theory.
rock strength hardening and damage under cyclic
loading. As mentioned earlier, this boundary is called the
cyclic damage stress (CDS). Fig. 7, shows that an
increase in fatigue strength is directly proportional to the
normalized deviator stress, qun/qf in unloading. In this
way, increase of peak strength with cyclic loading
decreases with a decrease in qun/qf. In addition, peak
strength increase reduces if unloading amplitude during
cyclic loading decreases.
Fig. 8 shows increase in peak strength of cyclic
damage tests undertaken under 100% unloading
amplitude after different number of cycles. As can be
seen in this figure, increase in rock strength, increases
with an increase in number of cycles and an exponential
function can be fitted to the results with good accuracy.
75
amplitudes. The stiffness and Poisson’s ratio values
resulted from cyclic hardening tests indicate that the
behaviour of the sample is significantly different from
those presented in Figs. 9a & b. Etan and v values for the
cyclic hardening test remained fairly constant with cycle
loading. Even, Etan slightly increased (from 0 to 2GPa)
between the first and last loading cycle when unloading
amplitude is 100%. This suggests that, as explained
earlier, cyclic loading causes existing pores to be closed,
the sample is experiencing compaction and therefore, its
mechanical properties is improved.
d) 10 CONCLUSIONS
Fig. 9. Variations of tangent Young’s modulus and Poisson’s ratio Mechanical behavior of a sandstone under triaxial
versus number of cycles for a) & b) cyclic damage tests; c) & d) compressive loading condition is investigated when
cyclic hardening tests. subjected to systematic cyclic loading. Effect of different
parameters including stress level at unloading, unloading
Figs. 9c and 9d presents example of results obtained amplitude, and number of cycles were evaluated.
for a cyclic hardening tests at different unloading
76
cyclic loading the rock may become weaker or stronger
as a result of cyclic loading. Variations of deformability
properties during cyclic loading confirmed this argument.
In general, it was found that the amount of increase
in peak strength decreases with a decrease in cyclic
loading deviator stress. In addition, peak strength
increase reduces if unloading amplitude during cyclic
loading decreases. In addition, it was found that in cyclic
damage tests the amount of axial and lateral strains
cumulated during cyclic testing in much greater than
those values for cyclic hardening tests. Damage and
hardening both increases with an increase in unloading
a) amplitude.
Hardening behavior of sandstone during cyclic
loading was related to inherent properties of sandstone
which is a porous rock. When sandstone is subjected to
cyclic loading at a stress level below cyclic damage
stress, weak bounding between particles breaks, which
causes pore spaces to be filled with fine material. As a
result the sample experience compaction during the
cyclic loading and its mechanical properties improves.
Contrary to this, if unloading stress is higher that
cyclic damage stress, sample experience irreversible
damages during cyclic loading and therefore, fails after
a number of cycles due to generation of unstable axial
b) and lateral macro cracks.
ACKNOWLEDGEMENTS
The authors would like to thank Ms. X. Pan, Mr. C.
Ho, Mr. Y. Zhao, Ms. A. Royle, Mr. Y. Zhao., previous
honours students, for their help in undertaking the
experiment and analyzing the results.
REFERENCES
1) Bastian, T. J., Connelly, B. J., Lazo Olivares, C. S., Yfantidis
N. and Taheri, A. (2014): Progressive damage of
Hawkesbury sandstone subjected to systematic cyclic
c) loading. Research Journal Review, AusIMM, 3:15-22.
2) Bieniawski, Z. T. (1967): Mechanism of brittle fracture of
rock, Parts I, I & Iii, International Journal of Rock Mechanics
and Mining Sciences, 4: 395-430.
3) Burdine, N.T. (1963): Rock failure under dynamic loading
conditions, Society of Petroleum Engineers Journal 3, 1-8.
4) Chen, Z. H., Tham, L. G., Yeung, M. R., and Xie, H. (2006):
Confinement effects for damage and failure of brittle rocks.
International Journal of Rock Mechanics and Mining
Sciences, 43:1262-1269.
5) Eberhardt, E., Stead, D. and Stimpson, B. (1999):
Quantifying progressive pre-peak brittle facture damage in
rock during uniaxial compression, International Journal of
Rock Mechanics and Mining Sciences, 36: 361-380.
6) Koseki, J., Indou, H. and Hayano, K. (2003): Cyclic triaxial
d)
tests on residual deformation and small strain properties on
soft rocks, Swets & Zeitlinger B.V., Lisse, The Netherlands.
Fig. 10. Variations of cumulative strain versus normalized
7) Ma, L., Liu, X., Wang, M., Xu, H., Hua, R., Fan, P., Jiang, S.,
number of cycles for a) & b) cyclic damage tests; c) & d) cyclic
Wang, G. and Yi, Q. (2013): Experimental investigation of
hardening tests.
the mechanical properties of rock salt under triaxial cyclic
loading, International Journal of Rock Mechanics and
Peak strength of the rock was altered by cyclic Mining Sciences, 62: 34-41.
loading; dependent upon the applied stress level during 8) Ray, S. K., Sarkar, M. and Singh, T. N. (1998): Effect of
77
cyclic loading and strain rate on the mechanical behaviour of
Sandstone, International Journal of Rock Mechanics and
Mining Sciences, 36: 543-549.
9) Singh, S. K. (1989): Fatigue and strain hardening behaviour
of Graywacke from the Flagstaff formation, Engineering
Geology, 26: 171-179.
10) Taheri, A., Sasaki, Y., Tatsuoka, F. and Watanabe, K. (2012):
Strength and deformation characteristics of cemented-mixed
gravelly soil in multiple-step triaxial compression, Soils and
Foundations, 52(1): 126-145
11) Taheri, A. and Tatsuoka, F. (2012): Primary stress-strain
relations inferred from multiple-step triaxial compression test
results, Soils and Foundations, 52(4): 748-766.
12) Taheri, A., and Tatsuoka, F. (2015): Small and large strain
behaviour of a cement-treated soil during various loading
histories and testing conditions, Acta Geotechnica, 10(1):
131-155.
13) Taheri A., Royle A., Yang Z. and Zhao Y. (2016a): Study on
variations of peak strength of a sandstone during cyclic
loading. Geomechanics and Geophysics for Geo-energy
and Geo-Resources, 2(1): 1-10
14) Taheri A., Yfantidis N., Lazo Olivares C.S., Connelly B.J.
and Bastian, T.J. (2016b): An experimental study on
degradation of mechanical properties of sandstone under
different cyclic loadings. Geotechnical Testing Journal,
39(4): 673-687
15) Wang, Z., Li, S., Qiao, L. and Zhao, J. (2013): Fatigue
behaviour of granite subjected to cyclic loading under triaxial
compression condition. Rock Mechanics and Rock and
Engineering, 46, 1603–1615.
16) Xiao, J. Q., Ding, D. Q., Jiang, F. L. and Xu, G. (2010):
Fatigue damage variable and evolution of rock subjected to
cyclic loading. International Journal of Rock Mechanics and
Mining Sciences, 47: 461-468.
78