MECHANICAL BEHAVIOUR OF ROCKS UNDER FATIGUE

Reaction rnecanlque des roches soumises a la fatigue

Mechanisches Verhalten der Felsen unter Beanspruchung

G. BRIGHENTI
Professor Istituto di Scienze Minerarie
Universit~ di Bologna

SUMMARY:
After briefly considering the phenomenon of fatigue in rocks and illustrating the importance of
such a phenomenon in the planning of rock structures and in comminution, the main studies on'the
subject are reviewed.
Experimental results are then reported, in relation to research we carried out at the Mining
Science Institute of the University of Bologna, stressing the importance of laboratory test
standardization in order to obtain results, which can be easily applicable to pratical cases.

RESUME:
A~r~s avoir bri~vement consid~r~ 1: ph~nom~ne de la fatigue des roches et en avoir illustre Ie
role en ce qui concerne la concep,t1on des structures en roche et la comminution, l'on fait r~f~-
rence aux principales ~tudes existantes et concernant ce sujet.
Ensuite, on mentionne les r~sultats d'une recherche experimentale ex~cut~e par l'Institut des
Sciences Mini~res de l'Universit~ de Bologne, en soulignant l'importance de la normalisation
des essais de laboratoire afin d'obtenir des donnees que l'on puisse rapporter au reel.

ZUSAMMENFASSUNG:
Nach kurzer Einordnung des Ph~nomens der Beanspruchung in den Felsen und der Illustration der
Wichtigkeit dieses Ph~nomens bei der Projektierung der Strukturen im Fels und im Splitterbruch
werden die wichtigsten Studien zum Thema in Erinnerung gerufen. '
Anschliessend werden die Resultate einer im Istituto di Scienze Minerarie der Universitat Bolo-
gna durchgefUhrten experimentellen Untersuchung aufgefUhrt und die Wichtigkeit ,einer Standardi-
sierung der Laborversuche hervorgehoben urn in der Praxis leicht anwendbare Resultate zu erhalten.

1. INTRODUCTION
In many cases rocks are subjected to predict their behaviour at least in the most
periodically variable loads. In fact rock common cases, data about rocks are very scar-
constructions may be subject to vibrations ce; it seems therefore advisable to examine
due to earthquakes, explosions, traffic, w~r- the results obtained both in the laboratory
king of big machines etc .. Even rock mater1al of the Mining Science Institute of Bologna,
may be subject to cyclic loading in order to and in other laboratories, in order to see
drill or crush it. which definite results have already been ac-
In the presence of a cyclic loading, quired and on which points it is desirable
both to guarantee the stability of the struc- to concentrate research-works, considering
ture which is subject to it (tunnels, founda- practical application, that is the determina-
tions of dams or industrial structures etc.) tion of allowable loads for rock structures
and toe:iCploit the load for comminution, it and of loads and ways of application in order
is necessary to know the behaviour of rocks. to improve rock drilling and crushing as the
It is therefore necessary to study rock final objective.
fatigue. With this in mind, that is thinking of
The study of fatigue behaviour of mate- practical applications, it is very important
rials began since the beginning of the 19th to consider that the tests can usually be
century for metals; however, as mentioned by performed only in laboratory on little 'sam-
Nordby (12), non metallic materials, particu- pl~s of rock ma~erial. The possibility of
larly cement, began to be studied only about uS1ng data obta1ned from tests in planning
1900. Concrete may be considered as an arti- is questionable even in the case of metals,
ficial rock, and therefore its behaviour is which are homogeneous materials with constant
probably similar to that of some natural and well-determined characteristics. It is
rocks; yet significant studies carried out therefore evident how uncertain the applica-
on natural rocks began much later and the tion of these results is in'the study of a
first interesting results were obtained in rock mass w~ich is.n~arly always heteroge-
the last 15 years. neous and w1th fam1l1es of discontinuities,
Thus, while for metals there exists, whose presence often determines its behaviour.
if not a complete theory, a remarkable number Thus it's necessary to keep in mind
of experimental data, making it possible to that all laboratory tests are conventional

- 65-
and that their main goal is that of providing 3. PREVIOUS RESEARCH.
data to compare the various materials and,
even for fatigue tests, that of standardizing In spite of the limited number of expe-
methods which, in the opinion of the writer, rimental works on rock behaviour under fati-
should be characterized above all by sempli- gue, some interesting studies carried out in
city and easy comparibility, even at the ex- the last few years (1,4,5,7,8,9,10,11,13,14)
pense of theoretic refinements. and the most important results obtained are
illustrated below.
The tests took into consideration the
2. COMMON EXPERIMENTAL TECHNIQUES IN FATIGUE most common rock materials: sandstones, mar-
TESTS bles, limestones, dolomites, quartzites and
granites; all samples examined were homoge-
The fatigue phenomenon may be referred neous and small in size. The frequency of cy-
to as a progressive damage of a material, cles ranged from 0.25 Hz (9,10) to 10 KHz (5);
when it is submitted to repeated stresses of pulsating compression or tension cycles or
a certain extent. alternate compression-tension cycles were
The fatigue tests consist of subjecting used, which were usually sinusoidal or trian-
a sample of the material to be examined to in- gular; in pulsating cycles minimum stresses
creasing and decreasing forces whose maximum in absolute value were set slightly higher
value must be lower than the resistance of than zero, due to technical reasons; in al-
the material for a great number of times, up ternate cycles maximum values of compression
to failure. In the most common case force va- and tensile stresses were generally chosen
ries according to a sinusoidal path or a so as to be in the same relationship to their
triangular path, sometimes it is applied by respective ultimate strengths.
steps or by impulses; using more sophistica- After comparing the results it was seen
ted tests stresses vary according to random that, although tests varied from case to ca-
laws or to laws in which the trend of the se, in the rocks examined:
stresses is controlled by a magnetic tape on a - the phenomenon of fatigue failure does
which the real stress rate to which the mate- occur;
rial will be subjected is recorded. b - in the case of compression cycles only,
It should be kept in mind that in some failure, at least within the limits of
cases a variation law applies to deformations the cycles applied, varies from 80' to
rather that s t re sse s . 60' of ultimate compression strength; in
In fatigue studies the variable forces the case of tension cycles it is about
may change in sign (alternate cyc~e) accor- 60-65' of ultimate tensile strength;
ding to symmetric or asymmetric laws, or they c - in the case of tension-compression cycles
may range from maximum values of the same fatigue resistance is reduced to values
sign (pulsating cycle); in any case amplitude of around 30' of the ultimate strength;
Sa is defined as the algebraic semidifferen- d - in the case of compression cycles the
ce between maximum and minimum stresses and presence of a compression radial stress
medium stress Sm which is defined as the enhances percent resistence to fatigue.
continuous component of stresses. As for other aspects of behaviour under
The fatigue life of the material, which fatigue, on the contrary, research workers
is a function of stress, is derived experimen- obtained contrasting results.
tally and is usually graphically represented For example, only some rocks, at least
by plotting, for each sample, the point within the limits of the cycles applied, sho-
having the logarithm of the number of cycles wed a fatigue limit; furthermore, in the pre-
on the abscissa and the failure stress on the sence of such a limit, some Authors found
ordinate. The curve obtained is referred to that rocks subjected to compression cycles
as W5hler's curve. below such limit showed a remarkable increase
It should be noted that, while for me- in ultimate resistance (10), some others did
tals in these curves the value of Sa is not find considerable variations (7).
plotted on the ordinate, for rocks nearly al- The results concerning the influence of
ways maximum stress Smax is plotted; on the cycle frequency on fatigue resistance were
other hand, this does not cause ambiguity, as equally contrasting; for example, while Bur-
in this case minimum stress is either the sa- dine (4) did not find a considerable influen-
me in all tests or correlated to maximum ce in a sandstone within 15-50 Hz, Attwell
stress. and Farmer (I), examining a dolomite within
Depending on load conditions to be stu- 0.20-20 Hz, found that as frequency increases
died, rock materials are subjected to pulsa- the number of cycles which is necessary to
ting cycles of compression, tension or shear, reach failure increases according to the law:
or to alternate cycles of compression-tension; log N = 0.98 log f + 3.01.
such tests can be carried out also in the pre-
sence of confinement stresses with dry mate-
rial or material saturated with liquids, com~
pressed or not. Frequencies ranging from a 4. EXPERIMENTAL RESULTS
few fractions of Hertz to several tens of
KHertz have been considered by research wor- 4.1 In our Institute an experimental investi-
kers and applied using oleodynamic or mecha- gation on the influence of pulsating
nical equipment for the lowest frequencies loads on the resistance of rocks is in pro-
and electromagnetic or sonic equipment for gress. The equipment used and the condttions
the highest ones. of the tests have already been described in
BesidesW5hler's curves, also deforma- two preceeding notes (2,3). Only a few gene-
tions and stresses applied vs time are obtai- ral data are mentioned below: all tests have
ned using fatigue tests. been carried out using the pulsating cycle
In some cases it is possible to observe shown in fig. I, applied by means of an oleo-
the variations of the characteristics of the dynamic system developed in the laboratory of
material according to the number of cycles the Institute. Apart from apreliminary test,
applied; this can be done using destructive frequency was either 0.05 Hz or 0.1 Hz; the
or non-destructive methods, by recording acou- tests were carried out at room temperature.
stic emissions and by microscopic examina- The stress applied varied from a maximum, ex-
pressed as a percentage of static ultimate
tions.

-66-
strength (obtained by applying the load at a the three materials have fatigue limits of
speed of about 0.5 MN/m2), and a minimum that about 60-65% (A,B) and 75% (C) of static ul-
is very close to zero (1-2 MN/m2). timate strength.
Gypsum behaved in a completely different
way: it did not break, at least up to 10,000
cycLes, not even when it was subjected to a
maximum load equal to 95\ of ultimate
strength. Thus gypsum does not seem to be
subject to fatigue.
The different behaviour of gypsum, as
compared to other rocks, was confirmed also
by failure tests on samples which did not
break after 10,000 cycles; in fact, while in

the sandstones and in the limestone no appre-

s ciable variations in ultimate strength were
observed, an increase in resistance of over
10\ was noticed in gypsum.

4.3 The second study focused on a more de-

tailed analysis of the phenomenon of fa-
tigue; in order to understand this phenomenon
clearly t~e study was limited to a single,
extremely homogeneous and uniform material;
t the fatigue test consisted of pulsating com-
pression, tension and shear stresses, using
Fig. 1 both dry and water saturated material.
The rock examined is a fine-grained sand-
stone taken from a quarry in the Tuscan Appen-
4,2 In the first study, which was limited to nine, details of which are to be found in re-
compression fatigue tests on dry rocks, ference (3); its porosity is 3.4\ and resi-
we examined two sandstones, a marly limestone, stance to compression, tension and shear is
and a gypsum, ,these being i~dicatedby the shown in Table 2.
letters A,B,C and D. The maIn features of the-
se rocks are listed in Table 1; for further Table 2
data see (2).
Table 1 Static Strength(MN/m2)
Material
Compression Tension Shear
Static Strength ~t:>dulus
of
RockType Porosity to Compression Deformation Dry 122 10.5
(\) (MN/m2) (MN/m2) 29
Saturated 84 5.5 23
SandstoneA 13 53 11,282
SandstoneB 2.5 165 26,193
The first set of tests was aimed at re-
Marly vealing the possible influence of frequency of
LimestoneC 1 218 73,575 application of the load on fatigue failure wi-
thin very low frequencies ranging from 0.01 Hz
Gypsum D 7,5 101 12,753 to 0.1 Hz.
In our investigation we set maximum
stress at 80\ of the resistance of the material
and measured the number of cycles necessary for
Frequencyin these testswas 0.05 Hz. .' failure at frequencies of 0.01 - 0.02 - 0.05
As for materials A,B and C we obtaIned and 0.1 Hz.
W~hler's curves up to 10,000 cycles; such cur- Tests on compression, tension and shear
ves are shown in fig. 2. It can be seen that stresses showed that failure does not depend on
frequency in a considerable way. The next se-
100 ries of tests was aimed at determining W~hler's
curves, frequency being equal to 0.1 Hz.
These curves are shown in fig. 3, 4 and 5
for compression, tension and shear respective-
ly, both for dry (a) and water-saturated mate-
rial tb).
80 The analysis of the results showed:
c a - for the three types of stresses, when wet
material is used its resistance value per
S% cent is smaller than that found in dry ma-
terial;
b - fatigue resistance per cent decreases pas-
60 sing from compression, to tension, to
shear tests;
c - the fatigue limit is revealed in all cases
in a more or less evident fashion; ,
d - those samples which did not break after
100,000 cycles have the same initial resi-
40 stance; they did not suffer detectable da-
1 10 mages or show improvement of their
strengths.

Fig. 2

-67 -
 Some subsidiary measurements that we car- 4.4 Understanding of fatigue behaviour of
ried out during the tests can give further rock masses subjected to pulsating cy-
information on the behaviour of the material. cles is based on the knowledge of cyclic loa-
More precisely, for some samples we cal- ding effect on friction between rock mate-
culated stress-strain curves as a function of rials. In order to know this a preliminary
time: remarkable hysteresis phenomena were enquiry was made, using the ro~ks previously
noted in the first cycles; the phenomenon indicated by the letters B, C and D.
then decreases and subsequently increases
again when the sample is about to break. 100
100

80
80
S%
S% a
60
80 b

40
1 10 1~
40
1 10
N Fig. 5
Small blocks of rock, whose contact sur-
Fig. 3 faces were smoothed, rectified and kept under
This behaviour, which has already been pressure with a normal stress of 39 MN/m2,
pointed out by other Authors (7-12), is more were subjected to a pulsating load T, which
evident and gradual in compression tests. varied from a few Newtons to maximum values
Deformation curve under maximum load in between 60\ and 95\ of those necessary to
function to the number of cycles, which is cause sliding; load scheme is shown in fig.7.
shown in fig. 6, is similar to that found by
Haimson (9), who noticed an analogy with the
curves. found in creep tests. Also in this ca-
se three stages can be observed: in the first
one the deformation rate decreases; in the
second it is constant; in the third it increa-
ses rapidly up to failure.

100

80
1 2 3
S%

60 N
Fig. 6

40
1~
T
1 10
N
Fig. 4
Microscopic examination of samples where N
fatigue tests under compression had been di- N
scontinued after an ever increasing number of
cycles did not show remarkable differences as
compared to the initial material in the first
two stages, whereas in the third stage micro-
fractures in the sense of the application of
load were found, partly in the grains an par-
tly in the cement. These observations are Fig. 7
confirmed by some measurements of sound pro-
pagation velocity in those samples which were At least up to 10,000 cycles no displace-
closer to failure. ment was observed in any of the tests. Thus

-~-
it can be said that cyclic load does not re- tion of a pulsatine load; this result
duce the coefficient of friction. On the con- was obtained by laboratory tests and
trary, if we measure the force which is ne- needs further verifications.
cessary to cause such displacement at the end - 5. It is essential to standardize fatigue
of the test, we can see that it increases up tests; it is the only way to compare all
to III for rock B, 44% for rock C and 24% for results and therefore be able to reiy on
rock D. The values of these forces go back to the experience gained with the construc-
the initial values if the surfaces are sepa- tion of other works.
rated and then approached again; it would
seem that pulsating cycles cause a sort of
"glueing" between surfaces. 6. ACKNOWLEDGEMENTS
These results need further verifications,
which are being'made in our laboratory; at The studies described in this paper were
any rate, the phenomenon must be considered carried out ~n the Rock Mechanics Laboratory
carefully as instability of rock masses de- at the" Ls t i t u t o di Scienze Mi ne rari.e" of
pends very often on sliding along disconti- Bologna University.
nuous surfaces rather than on the behaviour The project was in part supported by the
of intact material. Italian Consiglio Nazionale delle Ricerche.

5. CONCLUSIONS REFERENCES / BIBLIOGRAPHIE / LITERATUR

According to what has been said in the 1. ATTEWELL, P.B. et Al - Fatigue Behaviour
previous paragraphs, the phenomenon of fati- of Rock
gue failure in rock materials appears a very Int. J. Rock. Mech. Min. Sci.
complex one. In rocks showing a predominantly Vol. 10, no. 1, 1973
elastic behaviour collapse is perhaps to be pp. 1 - 9 '
related to brittle fracture.
Following the most widely accepted inter- 2. BRIGHENTI, G. - Sui fenomeni di fatica nei
pretation, pre-existing microcracks are pre- materiall rOCClOSl
sent in rocks and new microcracks are caused L'Industrla Mlneraria
during the first cycles both in grains or.cri- Vol. 27, no. 2, 1976
stals and at ~heir borders; then a stage lS pp , 71 - 74
reached where further damages are not caused.
or rather they are not detected; finally, in 3. BRIGHENTI, G. - Contributo allo studio spe-
the third and last stage microcracks join and rimentale della resistenza a fa-
propagate within the sample according to an tlca delle rocce
accelerated and progressive prQcess up to in- Boll. Assoc. Mln. Subalpina
stability and collapse. If.we accep~ ~his Anno 15th, no. 1, 1978
hypothesis the phenomenon l~ very Slmllar to pp . 1 - 9
that illustrated by Cruden ln the case of
creep {6). 4. BURDINE, N.T. - Rock Failure Under Dynamic
The same phenomenon was partially obser- Loading Condltlons
ved in our laboratory by microscopic examina- Soc. Pet. Eng. J.
tions of a few samples; it ~hould be reminded, Vol. 3, no. 1, 1963
however that Haimson, studymg fracture for- pp . 1 - 8
mation by microseismic emission ~etection,
did not find always the progresslve process 5. CAIN, P.J. et Al - Rock Fragmentation by
described above' it might well be that the High-Fretuency Fatlgue
phenomenon of fatigue is different depending Bureau 0 Mlnes
on the material and our knowledge still in Rep. Inv. 8020, 1975
the initial phase. pp . 1 - 21
From a practical point of view, however,
a few useful conclusions can be drawn: 6. CRUDEN, D.M. - The Static Fatigue of Britt-
- 1. Almost all brittle rocks are subject to le Rock Under UnlaXlal Compres-
fatigue failure; in many o~ them.i~ is Slon
possible to identify a fatlgue llmlt Tii't:" J. Rqck . Mech . Min. Sci.
after 10 000 - 100.000 cycles. This li- Vol. 11, 1974
mit is u~ually higher than 60% of static pp , 67 - 73
ultimate strength in the case of pulsa-
ting cycles and equals 30% in the case 7. HAIMSON, B.C. et Al - Mechanical Behavior
of compression-tension cycles. of Rock Under CYCllC Fatl~ue
These values refer to dry rocks and de- "Stablll ty of Rock Slopes ,
crease if the rock is saturated with 13th Symp. on Rock Mech.
water. Urbana, Illinois, Aug 30-Sept 1
- 2. The phenomenon of fatigue can be used 1971
for comminution; a thorough study of fa- pp , 845 - 863
tigue in the case of high-frequenc~ cy-
cles, which seem to be the most sUl~able 8. HAIMSON, B.C. - Mechanical Behavior of
for practical application as less tlme Rock Under C*CllC Loadlng
is needed, will certainly prove useful. Advances lnock Mechanlcs
- 3. When planning rock structures subj7ct to ISBN 0-309-02246-0
variable loads it is necessary to lden- National Academy of Sciences
tify the loading cycle and work out cal- Washington, D.C. 1974 '
culations on the basis of resistance to pp. 373 - 378
fatigue. Obviously, as it is difficult
to apply the results obtained from a sam- 9. HAIMSON, B.C. et Al - Acoustic Emission
ple to the determination of maximum allo- and Fatigue Mechanlsm ln Rock
wable stresses in structures, it is ne- Conference on Acoustlc EmlSSlon
cessary to follow prudential criteria. in Geologic Structures and Ma-
- 4. The coefficient of friction between two terials,
rock blocks increases with the applica- The Pennsylvania State Univers.

- 69-
 June, 1975
12. NORDBY, G.M. - Fatigue of Concrete - A
Review of Research
pp. 35 - 55 Journal of the American Concrete
Inst.
10. HARDY, August 1958
pp. 191 -219

13. PENG, 5.5. et Al - The Behavior of Salem

Limestone in Cycllc Loadlng
Soc. Petro Eng. J.
11. KHAIR, A.W. - A Study of Acoustic Emission Vol. 14, no. 1, 1974
pp , 19 -24
during Laborator~ Fatlgue Tests
on Tennessee San stone
Conference on Acoustlc Emission 14. ROBERTS, A.- GeotechnOlOg~
in Geologic Structures and Mate- Pergamon Press,xford, 1977
rials,
The Pennsylvania State University
G.Brighenti - Professor, Istituto di Scienze ~tinerarie,
June 1975 Viale Risorgimento 2, 40136 BOLOGNA ( Italy )
pp. 57 - 85

-70-