Uniaxial Testing in Rock Mechanics Laboratories
Uniaxial Testing in Rock Mechanics Laboratories
Uniaxial Testing in Rock Mechanics Laboratories
U N I A X I A L T E S T I N G IN R O C K M E C H A N I C S LABORATORIES
SUMMARY
INTRODUCTION
tension, which are special cases of the triaxial test, are by far the most common
and widespread direct tests for rock properties.
Although enormous use has been made of uniaxial tests, and despite many
attempts to clarify the controlling factors, there are still no generally accepted
standards for equipment and technique. Consequently, it is difficult to make
meaningful comparison of results obtained in different laboratories. In the related
field of concrete testing, for example, SIGVALDASON(1964) gives evidence that the
results of uniaxial compression tests on identical specimens made in eight different
laboratories had wide discrepancies of magnitude and variance, even though all
tests were made on machines conforming to the appropriate British Standards and
A.S.T.M. standards.
Standardization is clearly desirable, but it should be based on a thorough
understanding of the behaviour of the test material and of the detailed mechanics
of the test. The test should also be designed to yield information which can be
applied to research and engineering problems through the medium of theoretical
concepts of deformation and fracture. In spite of the need, premature standardiza-
tion would be inadvisable; improper standards would lead to confusion, and
enforced conformity would inhibit development of sound technique.
For many years some shortcomings of typical test techniques have been
recognized, and there have been numerous studies on particular aspects of test
technique. These studies have highlighted certain problems, but they have not been
fully successful in dispelling controversy. It now appears that interaction of some
of the complicating factors necessitates a broader approach and an overall critical
review embracing the composition, condition, and preparation of the test material,
the theoretical background of testing, and the detailed mechanics of the test.
The following review, which includes original contributions, is offered as a con-
tribution to the reevaluation of uniaxial testing in rock mechanics.
TEST MATERIALS
Rocks. The term "rock" may embrace almost all solid earth materials. There
is often no clear demarcation between "rocks" and "soils", and in rock mechanics
any naturally occurring earth material which has sufficient cohesion to enable it to
be loaded uniaxially can be considered rock. There are many ways of classifying
rocks, but most are based primarily on geological origin or chemical composition,
and are not generally suitable for engineering purposes, where the emphasis is on
mechanical properties.
HANDIN (1966) has suggested that it is possible to categorize rocks on the
basis of their mechanical properties, as follows: (1) the unfoliated igneous and
metamorphic rocks and silica-cemented sandstone; (2)schist, slate and highly
indurated and fissile shale; (3) dolomite and anhydrite; (4) moderately well
cemented sandstone; (5) limestone; (6) poorly fissile shale, mudstone and siltstone;
and (7) salt and gypsum.
Typical values of mechanical constants for these rock types are given by
Handin.
I0 / I0
, -- / / / a
6 / /- 6
/
4 __ f Q ~ _
, ,,, 4 )
= == o.s
0
>
I I ~"l<<'+ I' ..,
~',,
#+ I
o.a g
- / / +<><>' - o,
o., / I ,,4
o., :/ / o.,
/
I I I I I II , I I I
0.2 I z 4 s e I0 so 40 so 0,5
( I b f / I n 2 ) x 10- 3
, , ,| = i = = I = = i =i J = |
0.1 0.2 0.4 0.6 0.8 I 2 4
( k g f / c m t ) ~ I 0 "3
UNIAXIAL COMPRESSIVE STRENGTH
Fig.l. Diagram for classification of rocks on the basis of uniaxial mechanical properties.
(After DEERE and MILLER, 1966.)
TABLE I
1. Strength classification
D~va~ (1966) also makes two suggestions for quantifying the description
of massive rock, one based on core recovery percentages in drilling operations,
the other based on the ratio of field seismic velocity to sonic velocity measured
on intact laboratory specimens.
A classification scheme for use in rock mechanics was proposed by COATES
(1964), and was later modified (COATESand PARSONS, 1966). The Coates scheme,
embodying the 1966 modifications, is given in Table II.
Items 4 and 5 of Table II were criticised by Bt~RTON(1965) who suggested
replacements (see note below Table II). These suggested changes were not adopted
by Coates.
TABLE II
4. Gross homogeneity
(1) massive
(2) layered (i.e., generally including sedimentary and schistose, as well as any other, layering
effects which would produce parallel lines of weakness)
4. Gross homogeneity
(1) homogeneous
(2) heterogeneous
TABLE II[
CLASSIFICATION OF ROCK MATERIALS BASED ON UNCONFINED COMPRESSIVE STRENGTH 1
(After STAPLEDON,1968)
Range of U.C.S. (dry samples~) Range of strength of some common
Term abbreviation lbf/sq.inch kgf/cm 2 rock materials
x Samples of fresh rock material tested to Australian Standards. For rocks showing planar
anisotropy, the long axis of the samples is normal to the fabric planes.
2 To be defined.
a Some overlap in strength with very strong cohesive soils, e.g., hard dessicated clays. The
distinction can be made usually by soaking in water, when soils can be remoulded.
The major factors which influence mechanical behaviour, apart from those
revealed during the usual mineralogical and petrofabrie analysis, should be
recognized, and appropriate descriptions should be furnished if maximum value
is to be obtained from test results.
Density and porosity. For rock of a given type and composition, e.g. sand-
stone, limestone or gypsum, porosity appears to correlate with strength and
elastic moduli (SCmLLER, 1958; KOWALSKI, 1966; MORGLrNSTERNand PHUKAN,
1966; SIgIL~'S, 1966). Relationships between porosity and strength have also been
proposed for other brittle materials, such as ceramics (e.g., KNUDSON, 1959;
BROWN et al., 1964). Hence, porosity is a useful index property to report in con-
junction with strength and deformation data.
The simplest and most direct way to determine dry bulk density of a rock
is to weigh an accurately machined cylinder on a precision balance while it is still
hot from oven-drying1, and to measure with a micrometer its linear dimensions
after cooling to room temperature. As an alternative for irregular samples, bulk
density can be found by weighing the sample in air both in the dry and saturated
states, and also weighing the saturated sample immersed in water.
True grain density and true porosity are difficult to measure if the rock
contains completely sealed pores. One possibility is to grind the rock and measure
grain density in a pycnometer. Another suggested method is to compress the rock
hydrostatically and obtain a stress-strain curve showing the effects of pore closure
and elastic compression after pore closure (VI/ALSH,1965a, b, c). Extrapolation of
an approximately linear section of stress-strain curve, which is supposed to re-
present elastic compression of grains after pore closure, gives a strain intercept
for zero stress, and this gives a measure of porosity. Other methods exist, but they
require sophisticated apparatus and technique.
For many purposes the "effective porosity", which gives the volume of
interconnected pores, is of more interest than true porosity. This is usually deter-
mined by measuring the maximum volume of water which can be absorbed by
unit volume of the rock. For this measurement, the rock should be saturated by
first evacuating in the oven-dry state, then admitting distilled degassed water
under vacuum, and finally soaking under vacuum for 24 hours. The volume of the
pore water can then be found by surface-drying the saturated sample, weighing
it, and deducting the dry weight for the sample of known volume. Alternatively,
the saturated sample can be weighed while immersed in water; deduction of the
dry weight in air yields the grain and pore volumes.
As an alternative to water saturation, effective grain density and effective
Shape and size of grains and pores. In fracture theory the length of the in-
herent cracks, or other defect structures, in a material is one of the primary para-
meters. Both BRACE (1961) and SKINNER (1959) have identified the length of the
controlling defect structure ("Griffith crack") with the maximum grain size in
rocks, which suggests that this "crack" is either within the grain or at the grain
boundary. Under uniaxial stress conditions it seems likely that grain boundaries
and pores are the source of the controlling defects, so it is of interest to record
grain and pore sizes.
It is customary to measure grain size by direct scanning of a thin section
with a graticule under a low magnification petrographic microscope. This gives
the broad order of grain size, but special statistical techniques are required in
order to produce detailed grain-size distributions, since not all of the grains are
sectioned along their mid-planes. Pore sizes are also measured from thin sections
by direct scanning, and similar limitations apply to the results. Grains are some-
times "lost" from thin sections, leaving apparent pores.
The shape of grains and pores can be observed in thin sections cut orthog-
onally from the rock. Series sectioning on a surface grinder, with sequential photo-
graphy, has also been used to give a three dimensional picture of grain and pore
structure.
Pore size distribution can be obtained by the mercury penetration method or
by water expulsion in a pressure membrane apparatus. Low temperature gas
adsorption and desorption methods are also available. Some evidence on the
shape of pores can be deduced from the mercury penetration method.
Surface area. The surface area of a rock provides a measure of its internal
crack and pore structure, and since it represents the surface available for adsorption
of pore fluids, it indicates the probable sensitivity of the rock to water, atmospheric
moisture, or other fluids. The significance of this parameter does not appear to
have been widely appreciated in rock mechanics.
The most sensitive and precise method for measuring surface area is low
temperature gas adsorption, following the classical B.E.T. technique (BRUNAUER,
et al., 1938). The common adsorbate is nitrogen. Surface area may also be calculated
from adsorption isotherms for water vapour.
Surface areas for some rocks studied by the writers are given in Table IV.
Rocks such as shales are expected to have greater surface areas than the examples
given, perhaps N 10 m2/g.
TABLE IV
Sample history. The properties of a rock sample may alter with time following
its removal from the parent rock mass, so that its history has a bearing on test
results. Items of particular concern include time-dependent strain relaxation after
removal from a stressed rock mass, exposure to contaminants, fluctuations of
water content, and thermal history (e.g., exposure to high temperatures or freezing
conditions).
EMERY (1964) and FRIEDMAN (1967b) have shown that high recoverable
strains can exist within individual rock grains. Relaxation of these strains with
time can give rise to internal microcracking (EMERY, 1964), which may reduce the
strength of samples and lead to large scale cracking or actual disintegration.
The growing use of "rock softeners" (chemical additives to drilling water)
increases the danger of receiving contaminated specimens whose mechanical
properties have been artificially modified. However, a more likely source of
contamination is the sample preparation shop itself, where the test material may
be subjected to cutting oils, solvents, and detergents.
In general, deterioration of certain rock types as a result of wetting and
drying cycles or extreme temperature changes is well appreciated.
Preparation of specimens
In order to obtain valid results from tests on brittle materials, careful and
precise specimen preparation is imperative.
Collection and storage. Test material is collected from the field in the form of
rough blocks, dressed blocks, or drilled cores. Field sampling procedures should
be rational and systematic, and the material should be marked to indicate its
original position and orientation relative to identifiable boundaries of the parent
rock mass.
Samples intended to be representative of fresh, undisturbed rock should
not be collected from material which has been modified by blasting, contamination
or weathering.
Ideally, samples should be moisture-proofed immediately after collection,
either by waxing, spraying, or packing in polyethylene bags or sheet. They should
be transported and stored under cover, and generally protected from excessive
changes in humidity and temperature. Freezing and thawing during storage should
be avoided unless the rock is completely dry.
Finally, rock which is to be used for precise testing should be handled as a
fragile material, for it is possible to introduce internal cracks by rough handling.
Rough cutting. Large blocks can be reduced to manageable size and shape by
splitting with mason's wedges, which are inserted in holes drilled with a hand-held
power drill and carbide-tipped masonry bit.
Sawing. For heavy sawing, a slabbing saw with 15-18 inch diameter diamond
wheel is adequate for most purposes. The standard cutting fluid is clean water
from a mains supply.
For exact sawing, a precision cut-off machine is used. It has a diamond
abrasive wheel about 8 inch diameter and a table with two-way screw traversing
and provision for rotation. The speed of the wheel is usually fixed, but the feed
rate of the wheel through the work can be controlled. Clean water, either direct
from mains supply or recirculated through a settling tank, is the standard cutting
and cooling fluid. For cross-cutting, core should be clamped in a vee-block slotted
to permit passage of the wheel. By supporting the core on bot~ sides of the cut,
the problem of spalling and lip formation at the end of the cut is largely avoided.
Coring. Virtually all laboratory coring is done with thin wall diamond rotary bits 1,
which may be detachable or, more commonly, integral to the core barrel. "Whole
stone" bits find universal application; they wear well and are capable of producing
high quality core. Curf width is usually about 0.125 inches. Sintered-tip diamond
abrasive core drills are used for cutting glass and rocks; they have very thin walls,
cut a narrow curf, and produce a very smooth finish.
The usual size range for laboratory core drills is from 6 inch diameter down
to 0.75 inch outside diameter (0.5 inch core). Smaller core drills down to 0.0625
inches are available for piercing samples but with these drills the core is intended
to disintegrate. Typical sample diameters for uniaxial testing are 1-2.125 inches.
Drilling machines range from small quarry drills to modified machine shop
drill presses. Almost any kind of drill can be adapted for rock work by fitting a
water swivel, but a heavy, rigid machine is desirable in order to assure consistent
production of high quality core. The work block must be d a m p e d tightly to a
strong base or table by at least two steel straps so as to prevent any tilting,
oscillation, or other shifting. To avoid unnecessary unclamping and rearrangement
of the work block, it is desirable to have provision for traversing the drill head
or the work block. Traversing devices must lock securely to eliminate any play
between drill and work. The drill travel should be sufficient to permit continuous
runs of at least 6 inches and preferably of 10-12 inches, without need for stopping
the machine. Finally, the drill should have some provision for automatic feed.
Optimum drilling speeds vary with bit size and rock type, and to some extent
with condition of the bit and the characteristics of the machine. The general
trend is that drill speed increases as drill diameter decreases; also, higher drill
speeds are sometimes used on softer rocks. The broad range of drill speeds lies
mainly between 200 and 2,000 c/min. No hard and fast rules can be given, but an
experienced operator can easily choose a suitable speed by trial. In one laboratory
used by the writers, a heavy machine shop press is run at 1,500 c/min on 1 inch
diameter and smaller work, and at 500 c/min on 2.125 inches and larger cores.
The minimum speed used on this machine is 350 c/min and the maximum is
2,100 c/min. In another shop, a smaller hand-feed drill press works in the speed
range 300-1,000 c/min, but the operator would prefer an even lower speed, say
200 c/min, for hard rocks. HARDY et al. (1966) drill at 90 c/min. The main objections
to high drilling speeds seem to be "chatter" and vibrations in the machine, although
there are also problems in matching rotational speed and feed rate.
Some coring drills have hand feed. Usually, a piece of pipe is used as a lever
extension on the feed handle and force is applied by "feel", but this cannot be
relied upon to produce core of consistently high quality, as variations and dis-
continuities in feed rate tend to produce ridges in the core. The ideal feed arrange-
ment is a constant force hydraulic feed, which can be set for each bit size and rock
type, but such machines are quite rare. Constant force feed can be improvised by
means of a weight and pulley arrangement rigged to the feed handle. On adapted
metal-working drill presses, the automatic feed usually provides constant feed
rate rather than constant force and, in general, the gear boxes which control feed
rate as a multiple of rotational speed give feed rates which are too high for coring
rock. On one press used by the writers, the standard gear box gives feed rates
ranging from 0.004-0.01 inches of feed per revolution of the bit, which is too fast
for hard rocks. A speed reducer has been fitted to cut these feed rates by a factor
of 4, and most coring is done with feed rates in the range 0.001--0.0025 inch feed
per bit revolution.
With constant feed rate there is a danger of damaging the machine or the
core barrel if too high a feed rate is used, and an electrical overload breaker
should be provided. A suitable feed rate for a given drill size and rock type can
be selected on the basis of hand-feed tests and also by switching off the drive motor
while drilling with the automatic feed engaged: if the feed rate is too high the drill
will stop immediately, but if the rate is too low it will continue to spin for a few
seconds after power has been cut off. Cutting the motor with feed engaged is a
good way to interrupt a coring run for the purpose of extending the drill travel.
While excessive feed rates are obviously dangerous, unduly small feed rates
tend to polish diamond bits. When bits do become polished, they can often be
dressed by running a few holes into an abrasive sandstone at an aggressive feed rate.
Clean water from a mains supply is the standard fluid for flushing and
cooling coring drills. Rate of flow is regulated by experience; the flow should be
just sufficient to carry away the cuttings without the water getting hot. Some
makers of diamond abrasive core drills recommend use of cutting oils when
drilling hard rocks, but the adverse experimental consequences of this procedure
have already been mentioned. For some materials, e.g., moisture-sensitive rocks,
or water soluble substances such as rock salt, water cannot be used and compressed
air is used instead. Refrigerated compressed air has been used for coring frozen
rocks and soils without thawing them in the process.
Lathe grinding. Medium size machine shop lathes are satisfactory for rock working
but small bench-top lathes are neither large enough nor robust enough to handle
rock cores above 1 inch diameter.
Most work on the lathe can be done dry, without use of any cutting or
cooling liquids, and the only modification required on the lathe, apart from toolpost
fittings, is provision of a dust extractor. Also, the lathe should be cleaned regularly
with a vacuum cleaner and compressed air.
For edge grinding, core samples may be put into the lathe directly, in the
way metal is mounted, or they can be held in steel end cups under an axial force
exerted by the tailstock. They are rotated fairly slowly, say about 300 c/min for
1 inch diameter to 200 c/min for NX (with high speeds there is danger of over-
heating and fusing the surface under the tool). The actual grinding is performed
either by a toolpost grinder or a stationary single point diamond. The grinder,
which runs about 6,000-8,000 c/min with 5-3 inch diameter wheels respec-
tively, can be used on almost any type of rock, taking off up to 0.003 inch of rock
on each pass of the wheel. For finishing cuts, less than 0.001 inch should be ground
off at each pass. For the preparation of dumbbell-shaped samples, a diamond form
wheel is desirable; a single point diamond tool pivoted on the toolpost is satisfactory
only on softer rocks.
The lathe can also be used for end-grinding cylindrical samples. A sample
is held directly in the chuck, rotated at 200-300 c/min, and the grinding wheel,
its axis inclined some 15 o to the sample axis, is passed across the end of the sample
while rotating at 6,000-8,000 c/min. The "bite" ranges from about 0.003 inch
maximum to less than 0.001 inch for finishing, and the grinding wheel is passed
across the sample end at about 0.5 inch/min. The lathe is quicker than the surface
grinder for end-grinding.
Surface grinding. Many surface grinders used in rock mechanics are standard
metal working machines, with a dust extractor added so that the wheels can be run
dry. Real need for a surface grinder arises when broad surfaces or prismatic samples
have to be prepared to close tolerances; surface grinding on small cylindrical
samples can be done quite adequately on a lathe with a toolpost grinder, and on a
lapping wheel.
Lapping. Lapping puts a final smooth finish on end-ground samples, and it provides
an alternative to end grinding in the lathe or the surface grinder.
The two broad kinds of lapping machines are the simple rotating iron disc,
with a minimum of attachments, and the automatic lapping machine, which can
handle several samples simultaneously. Some automatic lapping machines are
intended to be used with oil-carried abrasives, but the use of these oily substances
is not advisable. It is suggested that only water-borne abrasives should be used,
even if this means working with the simple lapping wheel, which is not greatly
affected by rusting.
To end-grind on the lap, a cylindrical specimen is placed in a steel carrying
tube, which is machined to accept core with a clearance of about 0.002 inch. At
the lower end of this tube is a steel collar which rests on the lapping wheel. The
Fig.3. Carbon paper imprints of specimen ends. A. Lapped end. B. End after surface
grinding. C. Saw-cut.
grinding compound is usually silicon carbide (about 120 grade) and aluminum
oxide, carried in water.
DEFORMATIONAND FRACTURE
E
K-
3(1 - 2v)
E
G-
2(1 + v)
vE
2 =
(l+v)(1 - 2v)
deviatoric stress
shear viscosity =
deviatoric strain rate
When considering large viscous deformations it is assumed that flow occurs
without change of volume, and so the bulk viscosity is infinite and the shear
viscosity remains the only finite constant.
where K is a constant, and 11 and J2 are first and second invariants of the stress
tensors:
and:
I Elasr;c 2. VlKous
[No.kol [N--~,I [s,;,,-v.,,,tl ~l
[M,xwettl o- =constant nt
relaxat~
t|on
Oe 1 &r 1
- + -- a (3)
Ot E Ot t/
or:
c = --
E
+ --
i
0
adt (4)
de
cr = Ee + r / - - (5)
0t
This again implies that the stress at any given strain will be increased by increasing
the strain rate. It would also give a non-linear characteristic for conditions typically
imposed by non-stiff constant-speed testing machines, viz., ae/at = f(tr), where
fig) is of the form ktr ~/n with n >i 2. Taken by itself, however, the Kelvin-Voigt
model gives the unrealistic result that stress at a given strain can increase indefinitely
with increasing strain rate.
The general equation describing the rheologic response of the Burgers model
is more complicated. It is:
a27
~
at (EM EM Ek ) t~tr EMEk ~t "~
~2/~ EMEk ~e
+ - + + -- + tr = E M - r ~ + (6)
~M F]k ~]k ~t r]M~k r]k ~t
where E M and//M are modulus and viscosity for the Maxwell unit, and E k and r/k
are modulus and viscosity for the Kelvin-Voigt unit. Solutions of special interest
are those for abrupt application or relaxation of a constant stress, and for applica-
tion of constant strain rate. For constant stress o applied at t = O, e = O, the
required solution is:
The final two terms define the transient creep which occurs upon application of
load and relaxation of load; the reciprocal of the coefficient in the exponential
term (r/k/Ek) is the relaxation time, i.e., the time required for strain to decay to
1/e of its initial value, or (1 - l/e) of its final value when load is removed. For
constant strain rate de/dt = K, applied at t = 0, tr = 0, the solution is:
- 4( EMEz ) i t / 2 (8)
\ ~M~k /J
Eq.7 gives a strain-time relation, or creep curve, for the standard creep test,
which is made by applying constant uniaxial stress to a specimen and recording
strain as a function of time. This curve shows instantaneous elastic strain, followed
by decelerating (primary) creep, which in turn is succeeded by creep at constant
rate (secondary creep).
Under sufficiently high stress, most rocks show "Burgers body" creep
followed by tertiary creep to rupture.
Axial stress-strain relations. The most common test for determining quasi-elastic
moduli of rocks is the uniaxial compressive test, in which a cylindrical sample is
loaded axially at a constant strain rate while load, axial deformation, and lateral
deformation are recorded.
The most obvious departure from ideal behaviour to be shown by this
test is non-linearity of the stress-strain characteristic. The first part of a typical
plot of stress (ordinate) against strain (abscissa) for a high-strength crystalline
rock curves so as to increase the slope with increasing stress (Stage I, Fig.5A).
This curvature gradually ceases, until at the mid-portion of the plot there is
approximate linear proportionality between stress and strain (Stage II, Fig.5A).
The initial curvature in Stage I, which implies "stiffening" of the rock with
increasing stress, is attributed to progressive closure of cracks and pores under
stress, and the effect has been analysed in some detail (BRACE, 1965; WALSH, 1965a,
b, C; WALSH and BRACE, 1966a, b). The linear relation of Stage II is taken to
STRESS = BI
;7 \ \
2t' \
D C
STRAIN
B \\
\
STRESS // \
/ / \\\
/(f/ \""'-...
L
B STRAIN
represent elastic straining of the constituent grains after pore closure has reached
a limit for the particular stress system. In soft or highly visco-elastic rocks, the
initial curvature of Stage I is not always measurable. WALSH (1965a, b, c) and
WALSH and BRACE (1966a, b) have derived "effective" moduli for rocks in terms
of the pore geometry and the true elastic moduli for the rock matrix. At stresses
approaching the uniaxial strength of the rock, the slope of the stress-strain curve
decreases (Stage III, Fig.5A). This effect is associated with formation of micro-
cracks, which progressively destroy the load-bearing capability of the rock and
permit irreversible strain to occur. The test culminates when the slope of the curve
approaches zero and the unconfined compressive strength of the rock is reached.
At this point a test made in a typical machine terminates with the abrupt and
violent structural collapse of the specimen, but this event often reflects inadequacy
of technique rather than an inherent property of the rock. By contrast, if the test
is made in a "stiff" machine (i.e., one which can maintain constant strain rate as
sample reaction decreases) the sample does not "blow up", but instead it continues
to deteriorate by internal cracking, yielding a continuation of the stress-strain
curve of the form shown by a broken line in Fig.5A.
Very little information on the stress-strain characteristics of rock in direct
tension is available from the literature. Some data on axial stress-strain charac-
teristics are given by BURSHTEIN (1967) and BIENIAWSKI (1967) gives a lateral
strain plot.
Fig. 5B shows the general form of the axial stress-strain relation which has
been recorded by the writers for granite, limestone, and sandstone in tension. It
is non-linear, with curvature negative throughout the normal "incomplete" curve
(solid line in Fig.5B). By using a stiff testing machine, HUGHES and CHAPMAN
(1966) were able to obtain complete tensile stress-strain curves for concrete; on the
assumption that rock behaves in the same way as concrete, the form of the complete
stress-strain curve is indicated by a broken line in Fig.5B.
Lateral and volumetric strains in uniaxial tests. In an ideal elastic material, lateral
strain in a uniaxial test is directly proportional to axial strain. This is not the case
in most rocks, and there are three distinct stages in the relationship between axial
stress and lateral strain, corresponding to the three stages described previously for
the relationship between axial stress and axial strain. In the initial phases of
loading (Stage I, Fig.6), when cracks or pores are closing, there is little lateral
strain and thus the curve of stress against lateral strain rises steeply. At the onset
of "linear" compression (Stage II, Fig.6) the slope of the curve decreases, and a
steady slope is maintained throughout linear compression. As internal cracking
commences (Stage III, Fig.6) elastic compression of the rock grains is countered
by crack formation, and lateral strain increases rapidly as the bulk density of the
L . _ U .
(-ve strain)
AXIAL
STRESS
STRAIN
Fig.6. Typical lateral, volumetric and axial strain diagrams for rock, in compression.
rock starts to decrease. This volumetric expansion of the rock by internal cracking
is termed dilatancy (BRACE et al., 1966; BRACE and BYERLEE, 1967).
The plot of stress versus volumetric strain (given by the sum of the three
principal strains) is regarded as a valuable indicator of the deformation and
fracture processes (PAULDING, 1966; BIENIAWSKI,1967). Departure from linearity
between Stages II and III (Fig.6) is taken to indicate the onset of internal cracking,
while the volumetric strain maximum (point A, Fig.6) is taken as an indicator of
onset of unstable fracture propagation.
STR
80
40
STRESS 0
Ibf/in =
-40
-80
Fig.7. Stress-strain diagrams for successive uniaxial compression and tension load cycles
(Berea Sandstone).
If, instead of cycling through stress of one sign, stress is alternated sym-
metrically between compression and tension, it appears from tests carried out by
the writers that the residual strain produced at the end of the first half-cycle is
eliminated by the second half-cycle, in which stress is reversed in sign. Fig.7
shows this phenomenon for a sandstone. The rock had previously been subjected
to load cycling and therefore the initial compression loading curve in runs 1 and 2
shows no evidence of curvature towards the stress axis. After the first run the
hysteresis loop closes almost perfectly. This indicates that the mechanism respon-
sible for the residual inelastic strain is reversible. So far there are insufficient data
to determine the rate dependence of energy dissipation in the hysteresis loop, or
to explore changes of energy dissipation after large numbers of cycles. It is inter-
esting to note that, with a stiff machine, elastic load relaxation loops can be
obtained for the broken section of the curve in Fig.5A, which represents the
limiting stress-strain behaviour of rock which is extensively fractured.
Effect of strain rate variations. Tests at varying loading rates, cyclic loading tests,
and creep tests show quite clearly that rocks in general are subject to time-depend-
ence in stress and strain. This time-dependence is such that stress-strain charac-
teristics determined by conventional uniaxial testing typically tend to become
steeper and more linear as strain rate or loading rate is increased (REDDY, 1966;
Young's modulus and Poisson's ratio. For a perfectly elastic material the axial
stress-strain relation obtained from a conventional uniaxial test would be a straight
line with constant slope E CYoung's modulus). As already explained, such a graph
for most rocks is not linear; instead, the slope of the curve varies with the stress
level. Furthermore, slope varies with rate of loading, and it differs according to
whether load is being applied or removed. In these circumstances it is not strictly
proper to use the term Young's modulus, but for practical purposes it is convenient
to regard the slope of the curve, either at a specific point or averaged over a certain
section (Fig.8), as Young's modulus. With this interpretation, E becomes a
function of stress (or strain), and certain definitions are necessary in order to
indicate the stress level or stress range for which E is given.
The slope of the stress-strain curve at some specified stress or strain gives
the tangent modulus for that stress or strain, while the average slope between two
specified points, defined by the chord joining them, gives the chord modulus for a
MODULUS
M = INITIAL TANGENT
P = SECANT
Q = TANGENT
R = CHORD
0 R
STRESS
STRAIN
Fig.8. Definition of moduli for a non-linear stress-strain diagram.
given range of stress or strain. The tangent modulus at the origin is the initial
tangent modulus, and the chord modulus between the origin and some other point
on the curve is termed the secant modulus. These moduli must also be defined in
relation to the direction of loading (application or relaxation), and to the loading
rate.
When the term "modulus" is used without qualification, it is usually taken
to be the tangent modulus at 50~o of the ultimate stress. The type of modulus
used in analysis of a problem depends upon the nature of the problem (HAWKES,
1966).
Poisson's ratio v is correspondingly awkward to define for a non-linear
material. Since v is the ratio of the total lateral strain to the total axial strain at any
given stress level, it may be seen from the foregoing notes on axial and lateral
strain that v will be a function of stress or of strain. At very low loads v may be
close to zero, whereas at high (Stage III) loads internal cracking and consequent
dilation may cause v to exceed the theoretical maximum value of 0.5. Again the
value taken for v must depend on the problem under consideration, but for purposes
of broad comparison of rocks it is suggested that v should be taken as the ratio
of the total lateral strain to the total axial strain at 5 0 ~ of the ultimate stress.
Creep under constant stress. While some of the inelastic properties of rock can be
deduced from the results of conventional short-duration tests and from load
cycling tests, clear and unambiguous data on creep properties are best obtained
from direct creep tests.
The standard creep test is made by applying constant uniaxial stress to a
specimen and recording strain as a function of time. The resulting creep curve for
a wide range of solid materials shows instantaneous elastic strain, followed by
decelerating (primary) creep, which in turn is succeeded by creep at constant rate
(secondary creep). Eventually, creep rate may accelerate (tertiary creep) and destroy
the sample. Prior to the onset of tertiary creep, this type of behaviour corresponds
qualitatively to the rheologic response of the Burgers model, given earlier by eq.7.
Under high deviator stress many rocks display the classic stages of "Burgers
body" creep, which eventually is followed by tertiary creep to rupture; under low
stress there are many rocks which either do not creep, or creep to a limiting strain
(Kelvin-Voigt behaviour). Some rocks will flow continuously even under very low
stress. In the creep of "brittle" rocks, where the creep mechanism is one of micro-
cracking (SCHOLZ, 1968a, b, c) the stress level which separates Kelvin-Voigt
behaviour (creep to a stand-still) from Burgers behaviour (continuous creep to
failure) can perhaps be identified with the stress for transition from stable crack
propagation to unstable crack propagation (BIENIAWSKI,1967).
A detailed review of rheological models applied to creep problems in rock,
together with a large amount of experimental data, is given by KIDYBINSKI(1966).
HARDY (1967) has tabulated experimental values of the Burgers constants for
representative rock types, and other data on apparent viscosity are given by
HANDIN (1966).
Data from creep tests are used to express strain as a function of time and to
determine strain rates for various stages of creep. Eq.7 gives one example of the
strain-time function; several other empirical and analytical expressions have been
used to describe the creep of rocks (e.g., PARSONSand HEDLEY, 1966) and other
solids (e.g., KENNEDY, 1962). Strain rate at some stage of creep, particularly
secondary creep, may be related to stress by a simple power relation of the form
= k.trn; experimental values for the exponent n are generally in the range
1--4, with a tendency for n to increase with increasing stress, as is the case for some
other solid materials. This tendency for strain rate to be controlled by increasingly
high powers of stress, which represents a transition from linear (Newtonian)
viscosity to plastic (Saint-Venant) behaviour, is sometimes represented by poly-
nomial or hyperbolic sine relations between strain rate and stress in the fields of
metallurgy and ice physics.
Analytical methods for determining energy storage and energy dissipation
properties of rocks 'from results of quasi-static creep tests have been given by
KONDNER (I 966).
clear definition of their meaning. In testing work, confused and imprecise termi-
nology is obviously unacceptable.
Fracture. Fracture is the process by which cracks are formed or extended. New
surface is created and strain energy is absorbed in the supply of surface energy
to the fresh surfaces. In a material which is completely uncracked initially, it is
conceivable that cracks might form under stress by some kind of "pile-up" of
dislocations (COTTRELL, 1959; PUGH, 1967). However, virtually all rocks contain
obvious inherent flaws in the form of cracks, pores, and grain boundaries; all of
these can be considered as stress-raisers equivalent to cracks for purposes of fracture
initiation. When stress reaches a sufficient level, cracks form and extend from
existing defect structures. In most granular and crystalline rocks, it appears that
many discrete cracks form or extend in succession at randomly distributed loca-
tions; they usually form abruptly, often producing microseisms (BROWN and SINGH,
1966; KNILL et al., 1968), but propagation is checked when stress concentration
around the crack tip falls or when the cracks encounter pores or grain boundaries.
Cracking frequency increases with increasing stress. In glassy or fine-grained rocks
there may be only one or two significant cracks which initially grow steadily in
response to increasing stress. BIENIAWSKI(1967) distinguishes two separate stages
in crack growth: stable crack propagation, in which crack growth is a function
of the applied loading, and which can be halted, and unstable crack propagation,
in which cracks grow uncontrollably with no input of energy from external
sources.
Fracture may also be described as "brittle" or "ductile". Brittle fracture is
usually considered to be fracture which is preceded by little or no permanent
deformation, i.e., behaviour prior to fracture is elastic. Ductile fracture is preceded
by appreciable viscous or plastic deformation, and a continuous input of energy
is required to produce fracture, even after peak stress has been reached.
BIENIAWSKI(1967) properly points out that the terms "brittle" and "ductile"
should be used to characterize processes or behaviour, rather than materials. It
may be taken as axiomatic that all real materials possess all rheological properties
under appropriate conditions. There is no universally accepted measure of relative
ductility, but HANDIN (1966) has suggested that descriptive terms might be defined
on the basis of the total strain which occurs before peak stress is reached. These
are:
Very brittle < 1~o strain up to peak stress
Brittle 1 - 5 ~ strain up to peak stress
Moderately brittle (transitional) 2 - 8 ~ strain up to peak stress
Moderately ductile 5-10Yo strain up to peak stress
Ductile > 10~o strain up to peak stress
Strength. Strength is another vague term which has different meanings in different
contexts. In general, it denotes a stress level at which some permanent and detri-
mental change occurs. This may involve a change from elastic to visco-plastic
behaviour, or it may involve extensive fracture or rupture. In rock mechanics,
strength usually means the stress at which the rock fractures, and it is generally
taken as the maximum stress reached before the specimen collapses or separates.
When the applied stress is uniaxial, this value is referred to as the compressive or
tensile strength, and it is measured by dividing the maximum load by the cross-
sectional area. Other terms are sometimes used, such as "crushing" strength
(compressive strength) and "shear" strength. This latter term is subject to diverse
interpretations (EVERLING, 1964) but where the uniaxial compressive test is
concerned, it should be interpreted as one-half of the compressive strength value.
Tensile strength is the most fundamental parameter involved in considerations
of rock fracture, and since it cannot be deduced directly from atomistic considera-
tions of the rock's constituent minerals, it is defined as a bulk property using the
uniaxial tensile test as the basis of definition.
In uniaxial testing, internal cracking of rock in compression, and probably
in tension too, begins at stresses well below the maximum stress which causes
structural collapse, and this leads to marked time-dependence and rate-dependence
for the measured strength.
rocks. Bieniawski associates the long term strength, which is also the minimum
stress level for creep to failure, with the stress for onset of unstable fracture
propagation, mentioned earlier. This same effect must obviously lead to rate-
dependence of apparent strength in conventional uniaxial tests; higher strain rates
or loading rates, which decrease test duration, give higher values for the apparent
strength (e.g., GREEN and PERKINS, 1968; KUMAR, 1968; STOWE and AINSWORTH,
1968). These facts point up the need for standardization of loading rates in uniaxial
testing.
Empirical and theoretical criteria. In order to interpret test results and apply them
to problems involving fracture and failure in complex stress fields, it is necessary
to have some kind of understanding of the failure mechanism and, in particular,
to have some kind of relationship between the various components of stress for
the failure conditions. The most important requirement is a failure criterion, i.e.,
a functional relationship between the principal stresses for the limiting failure
condition. There are two broad classes of failure criteria: those which are phenom-
enological or empirical, and those which are developed theoretically from physical
models.
Several empirical criteria have been proposed as conditions for failure by
fracture or plastic yielding in various types of materials. These typically postulate
that, at failure, there is constancy of such quantities as maximum principal stress,
maximum shear stress, maximum strain energy, or maximum shear strain energy.
Some of these prove satisfactory in describing failure by plastic flow in metals, and
they may serve similar functions for certain rocks. Two noteworthy ones are the
Tresca criterion, which takes the maximum shear stress as constant at yielding, and
the Von Mises criterion, which takes the sum of squares of reduced principal
stresses as constant at yielding. For failure of rocks by fracture, however, conditions
at failure are influenced by the bulk stress, and consequently a criterion similar in
form to eq.1 is required.
A comprehensive review of failure criteria for rocks is given by JAEGER
(1967); some of the more important criteria which are relevant to uniaxial testing
are outlined below.
The most widely used empirical criterion for failure by fracture in rock
mechanics is the two-dimensional Coulomb-Navier criterion, or a more general
variant, the Coulomb-Mohr criterion. To Coulomb's original hypothesis of failure
at constant maximum shear stress, Navier added the condition that normal stress
on the failure plane produces an extra "frictional" resistance, giving as the final
criterion:
(an - aa) (al + aa) sin = C cos (9)
2 2
Eng. Geol., 4 (1970) 177-285
UNIAXIAL TESTING IN ROCK MECHANICS 211
where C is an intrinsic shear strength of the material for zero normal stress on the
failure plane, and/~ are the angle and coefficient of friction respectively, and
0.1 > 0-2 > 0.3. The intermediate principal stress 0"2 is not considered to influence
the failure. In principle, the linear failure envelope can be determined from
uniaxial tests in tension and compression. A similar experimental criterion de-
veloped by Mohr takes as the rupture surface (in two-dimensional stress space)
the envelope of Mohr circles plotted in the well-known manner with ordinate
zm = 1/2 (0-1 - 0"3) and abscissa 0.m 1/2 (0.1 + 0.3). This envelope is not
=
where/~ is the coefficient of friction of the crack faces. This criterion is identical
to the Coulomb-Navier criterion (eq.10) with C = 2a r. However, whereas the
Coulomb-Navier criterion and other empirical criteria treat the condition for
complete failure, the criteria based on Griffith theory relate to onset of cracking,
as will be discussed in the next section.
or:
fl = n/2 if at + 3tr3 < 0 (16)
t g tends to be treated as a curve-fitting parameter, i.e., its value is chosen so as to give best fit
between experimental data and the theoretical functions.
The location of the point on the crack boundary at which the maximum tensile
stress occurs is shown in Fig.9. For the most dangerously oriented crack, the
tangent to the crack boundary (assumed elliptic) at the point of maximum tensile
stress is inclined at angle 6 to the major axis of the crack, where:
7~
6 = 2fl - - - (17)
2
It may be seen that, in general, these critical points are not at the exact tips of the
crack, except in the special case of uniaxial tensile stress. The crack therefore tends
to grow normal to the surface of the original flaw, as shown by broken lines in
Fig.9, and the original crack does not extend along its major axis.
Fig. 10A shows similar development of multiple tensile cracks in the direction
of applied compression, starting from a single initial closed crack in a glass slab.
This is in agreement with experimental work by HOEK and BIENIAWSKI(1966) and
BOMBALAKIS(1964). When a crack has grown to a certain length, the original stress
concentration which started its growth is relieved, and the crack will only continue
to grow as long as the stress concentration produced by its extension is sufficient
to overcome interatomic forces. It will be noted that the new crack is no longer
oriented at the optimum angle given by eq. 17, and growth normally ceases as the
new crack curves and becomes aligned with the direction of major compressive
principal stress. As the general stress level increases, other cracks grow, until the
material is cracked extensively throughout its volume.
The effect of multiple internal cracking has been reported by many investiga-
tors. Fig. 11 illustrates the development of such a crack structure in ice, where the
process is easy to observe. In clear ice the flaws from which cracks propagate are
associated with grain boundaries. It can be seen that the cracks are aligned
mainly in the direction of principal compressive stress, and in the initial stages of
" II
M T
A II
Fig.9. Crack growth from an elliptical flaw. A. Theoretical elliptical flaw. B. Crack
growth to failure in a flawed glass slab under uniaxial compression.
Fig.t0. Crack growth in glass under uniaxial compression. A. Inclined crack--low stress.
B. Inclinect ~erack high stress. C. Vertically aligned crack.
failure they do not propagate very far, nor do they interact to form continuous
cracks.
Validity of Griffith theory. Griffith theory was applied originally to materials such
as glass, in which there are no obvious flaws. There is, however, nothing hypo-
thetical about the flaws in rocks; they exist in the form of cracks, pores and grain
boundaries, and there can be no doubt that they act as stress raisers. By this token
the basic approach of Griffith theory is sound.
In recent years there have been a number of sweeping criticisms of Griffith
theory, based largely on observations that rock does not fail when the first crack
occurs, and that the experimental ratio of uniaxial compressive strength to uniaxial
tensile strength is different from that predicted by theory.
It is now abundantly clear that the onset of cracking in a compressive stress
field can occur at stresses far below the maximum stress which the rock can carry
before it ruptures or disintegrates. This being so, Griffith theories cannot be used
as criteria for complete structural failure of rock. They must be reinterpreted as
criteria for the onset of cracking, which, while representing irreversible deterioration
of the rock, does not necessarily represent failure.
In a uniform uniaxial tensile field there is still some uncertainty about the
fracture process. It may be that the first crack to start will propagate continuously
until the rock fractures, but the non-linearity of tensile stress-strain relations, the
Statistical nature of strength. Simple failure theory assumes that there is a generalised
stress, e., which is a function of the principal stresses given by the failure criterion,
such that failure occurs when it reaches a definite critical value S at some point
in the body. In terms of uniaxial tests this is equivalent to saying that the axial
stress at which a sample breaks is a characteristic property of the material. How-
ever, from the results of careful tests it is found that measured strength shows a
statistical distribution and, furthermore, there tend to be systematic shifts of
apparent strength as size, shape and stress field are varied. Thus it is inferred that
strength is not a unique material constant, but rather a statistical quantity. This
view is in harmony with theoretical concepts of strength, which ascribe to defects
the control of strength. For example, from eq.11 it is seen that Griffith theory
predicts that strength is inversely proportional to the square root of the length
of critically oriented cracks, which implies that the longest crack of critical orienta-
tion controls the strength, provided the stress concentration is maintained. Gritiith
suggested that reduction of specimen size would lead to increase of strength, since
the maximum defect size would be limited in very small specimens, and tests on
glass filaments did, in fact, bear out this prediction.
The best known contributor to the study of statistical effects in failure of
materials is WEIBULL (1939, 1951, 1952), who applied probability theory to the
failure of a body made up of numerous elements, the properties of which vary
according to some characteristic distribution. In the simplest form of this theory,
the so-called "weakest link theory", it is assumed that failure of one element deter-
mines the failure of the system. I f P is the probability that tr. < tr in an elementary
volume of the material dV at the moment of failure, then:
P = 1 - exp [ - f(a) dV] (18)
and the probability that a. < cr in the complete volume of the specimen V =
SdV is:
in which a o and m are material constants expressing the flaw characteristics, and
au is a limiting stress below which the probability of failure is zero.
Strength is usually expressed as the arithmetic mean of measured values
am. Weibull points out that since:
a m = ~adP (21)
its value for a given material can be influenced either by changing V (size and
shape of the specimen) or by changing a. (the type of stress system). The effect of
sample size is discussed in more detail later. The possibility that strength may
differ for different systems of stress raises fundamental questions in testing philos-
ophy.
A persuasive case for the Weibull approach to failure is made by HUDSON
and FAmHURST(1969), who illustrate its application to problems in rock mechanics.
It is important to note that the weakest link theory identifies failure with the
failure of the weakest element. In this respect it is similar to Griffith theory, which
assumes that the material fails when the first crack is formed.. Thus both statistical
theory and Griffith theory treat the onset of cracking, and not the structural
collapse or separation of a specimen, which generally occurs at higher stress levels.
Strength data relating to structural collapse, when a multitude of cracks have
formed and many "links" have already failed, do not provide direct tests for either
Griffith theory or Weibull theory.
Fig.12. Modes of failure of cylindrical rock test specimens under uniaxial compression.
A. Shear failure. B. Cataclasis. C. Cataclasis (Granulated rock not shown). D. Cataclasis, under
high loading rate, Solenhofen Limestone. (After FRIEDMAN et al., 1969). E. Cataclasis. Berea
Sandstone saturated with paraffin wax. F. Combined cataclasis/cleavage--marble. G. Combined
cataclasis/cleavage--granite. H. Axial cleavage--ice. I. Axial cleavage--granite.
Fig.14. Mode of final collapse of highly cracked ice test specimen under uniaxial com-
pression. A. Developingend cones. B. Developingshear plane.
is that crack propagation in the loading direction can take place at lower stress
than that required for crack initiation from flaw structures. It has been found by
the writers that even when many cracks are present in glass plates, only one
actually propagates to the boundaries (Fig.10B). For the example shown in Fig. 10C,
a compressive stress of over 80,000 lbf/sq.inch was required to propagate the crack
to the boundaries, while the tensile stress required to propagate a similar crack in
similar glass was only about 1,200 lbf/sq.inch.
The tentative conclusion to be drawn from presently available information
is that both cataclasis and axial cleavage are valid modes of failure. It will be
appreciated that a specimen which has failed by axial cleavage may have lost little
of its load bearing capacity in the axial direction. Continued application of load
to such a specimen during a test may cause accelerated cleavages in the separate
pieces until collapse occurs, the net result being that apart from the absence of end
cones there may be little difference between a specimen which has failed by
cataclasis or by axial cleavage. Such a result, for example, often occurs when failing
glass test specimens.
UNIAXIAL TESTS
Boundary conditions. The boundary conditions for a test specimen express the
state of stress, strain, or displacement at the boundaries of the specimen. For a
typical specimen in the form of a right circular cylinder the boundaries are the
cylindrical outer surface and the two plane ends. The uniaxial test on such a
specimen is performed by applying forces or displacements to the end planes,
while keeping the cylindrical surface free from normal restraint.
For an ideal test, the stress boundary conditions for the specimen end planes
demand that the normal stress, which is a function of time, should be uniform
across the plane at any given time. They also require that there should be no shear
stress at the boundary. The cylindrical surface should also be free of normal stress.
With a perfectly homogeneous isotropic material, the stress boundary con-
ditions for the end planes could equally well be replaced by an equivalent set of
displacement boundary conditions. These require that, at any given time, the
axial displacement of the end planes should be uniform, and there should be no
shear distortion at the end planes. These conditions imply that there should be no
restriction of lateral strain in the end planes, no net lateral displacement of the
end planes relative to the centre of the specimen, and no bending or twisting of
the specimen.
While the stress and displacement boundary conditions are equivalent to
each other in the initial stages of loading for many rocks, they cease to be neces-
sarily equivalent when internal cracking of the specimen begins. Since there is a
probabilistic aspect to the occurrence of internal microcracks, the spatial distribu-
tion of cracks at any given time is not necessarily symmetrical with respect to the
specimen axis, and therefore the specimen may not be effectively homogeneous.
In these circumstances, if the ideal stress boundary condition is maintained, the
strain field will be perturbed, whereas if the ideal displacement boundary conditions
are maintained, the stress field will be perturbed. In simple terms, with hydrostatic
pressure or a flexible platen on the sample ends (stress boundary conditions), the
sample is free to deform in a non-symmetric manner, whereas with rigid, non-
rotating platens (displacement boundary conditions), it is constrained to deform
symmetrically and stress tends to be transferred from weaker to stronger parts of
the specimen. This distinction does not appear to have received explicit recognition
in the literature, although it is known (e.g., TARRAIST, 1954a, b) that test results
may differ for non-rotating platens (locked ball seats) and rotating platens (free
ball seats).
The general trend in uniaxial compression testing has been to attempt to
satisfy the displacement boundary conditions rather than the stress boundary
conditions at the loading planes. This is not for theoretical reasons, but for practical
expediency. It has been found that the simplest way to obtain consistent test
results is to load a specimen by direct application of rigid platens, which are
initially free to align themselves with the specimen ends but later lock under load.
By contrast, in the uniaxial tensile test it is usual to permit rotation of the end
planes about all three axes, as this is the simplest way to avoid unwanted flexural
and torsional stresses.
The alternative of meeting stress boundary conditions in the compression
test involves, in practice, hydrostatic loading of the end planes through membranes
or flexible platens. This technique has not found wide favour so far.
Actually, practical testing techniques are rarely established with explicit
reference to the theoretical boundary conditions, as there tends to be a gap between
theory and experiment in the testing field.
Contact problems
End flatness. For analytical purposes it is usually assumed that loading platens
and end surfaces of test specimens are perfectly plane, so that intimate contact
occurs when they are brought together in parallel alignment. Actually, it is difficult
in practice to induce intimate contact, as can be shown experimentally by loading
the "plane" surfaces of photoelastic materials, such as glass or certain plastics,
through ground steel platens. Asperities in the surfaces, foreign particles, and
broad departures from flatness can induce non-uniform pressure distribution at
the contact. Resulting perturbations of the stress field may be quite localized, and
it might be argued, invoking the Saint-Venant principle, that their effects in the
mid-section are negligible. However, there is the possibility that contact stresses
will produce cracking in the contact zone, and that these cracks will propagate
to cause premature failure in certain kinds of rock.
To obtain some idea of the magnitude of the tensile stresses which might
be generated at non-flat surfaces, some simple calculations can be made.
Consider a single irregularity on the loading surface of a test specimen
(Fig.15) and assume that its profile is spherical with a high point at N. The
irregularity may be a general "doming" of the sample end, or may be an isolated
dome-shaped "bump", but it is assumed that the platen makes no contact with
the rest of the surface until the bump is flattened, and so the surrounding material
is essentially unrestrained. The loading platen is assumed to be perfectly plane,
.-...
Tensile ~ vr
V//////////~_//_l//////,a Crocks
"Or
I ~ Rock / ~1
A B
Fig.15. Effect of non-flat specimen ends in uniaxial compression.
TABLE V
and
2
Ks _ 1 - vs (24)
Es
and the subscripts p and s denote moduli for platen and specimen respectively.
The maximum compressive stress ao occurs at the surface in the centre of the
contact area, and is given by:
3P
ao - (25)
2ha 2
The stresses at the periphery of the contact area (B, Fig.15) are:
(I - 2v,)
~rZ = 0; a, = - o'0 = 0. (26)
3
i.e., the periphery is in a state of pure shear, and the maximum tensile stress there
is in the radial direction, with magnitude:
P 0 - 2v~) (27)
(o,), =a - 2ha2
p=5.16 3 2[ 1 2 1-v] 2
ao R - vp + (28)
Ep E~ 3
Assuming that vp = vs = 0.25, E~ = 30 l0 s lbf/sq.inch and E~ = 5 106 lbf/
sq.inch eq.28 becomes:
P = 0.215. 10 - 1 2 0.3 R 2 lbf (29)
0.0 P
0.r . . . . . . (30)
6 4ha 2
In Fig.16, 0.0 and 0.x are plotted against P for various parametric values of
R, following eq.29. The figure shows that 0.0 and 0.r rise sharply as P is applied,
and then more gradually as P increases further. Taking a numerical example,
consider a specimen with a domed end: D = 2A = 2 inches, ~ --- 0.0025 inch,
and E s = 5" 106 lbf/sq, inch. F r o m Table V the b u m p radius is seen to be 200 inches.
As load is applied, 0.r and 0. increase quite rapidly up to P ~ 15,000 lbf (Point X),
at which stage 0.a- = 2,000 lbf/sq.inch and a = 12,000 lbf/sq.inch. The radius of
the contact area, a given by eq.30 is 0.77 inch, and if the tensile strength of the
specimen is 2,000 lbf/sq.inch circumferential cracks will start to form at this radius.
As P increases further, aT cannot increase and circumferential cracking will
_ 32,000
5OOO
IOO0
30(
4 0 0 0 - 24,0OO
1600
300
1200
.oo
,b ' 3'o ' .'o ' 6'o ' .'o ' C.,o. o
APPLIED LOAD. tbf
I | , I ! nl , I , , I I , , ,
0 IvO 200 3 . 0 x 103
APPLI[D LOAD, N
Fig.16. Theoretical relationships between applied load and stress for various bump
dimensions on domed-end samples.
probably hasten the flattening of the bump. As the bump is flattened, radial
restraints are mobilized at an unlubricated contact, as will be discussed in a later
section, and propagation of the cracks may be inhibited until the end load reaches
a considerably higher level. If, however, the contact is lubricated, or radial restraint
is reduced by platen matching, the cracks formed by lack of flatness may well
propagate through the body of the specimen at a comparatively low load. Point Y
indicates the stage at which the sample is completely flattened; if av at this point
had been less than the tensile strength of the rock, the bump would have been
flattened without cracking the rock.
Flatness tolerances cannot be deduced directly from the foregoing since the
surface irregularities are rarely known in such detail. However, some idea of
magnitude can be gained from consideration of the domed-end sample. Fig.17
gives the maximum tensile stresses induced in domed-end samples of various
diameters. Replacing the values of R from Fig.16 by equivalent values of ~ from
Table V, Fig.17 gives the relation between bump height ct and maximum tensile
stress for rock having E s = 5 106 lbf/sq.inch and v = 0.25. If it Is assumed that
for this particular rock type the tensile strength is about 1,500 lbf/sq.inch, the
flatness tolerances for complete freedom from cracking are approximately 0.0005
inch for a 1 inch diameter specimen and approximately 0.0015 inch for a 3 inch
diameter specimen.
/ ES: 5 xlOelbf/inZ /
:5 1//===0"25 D= I in//"
Ibf/in2 2i
Squareness and parallelism. When specimens are loaded by rigid platens, the
platen face and the specimen end plane must be parallel when contact is first made.
HosmNs and HORINO (1968) investigated the effects of lack of end squareness,
or lack of parallelism between ends, for rocks ranging from granite (45,000 lbf/
sq.inch compressive strength) to limestone (7,000 lbf/sq.inch compressive strength).
For a non-adjustable loading head, i.e. one lacking a spherical swivel, a departure
from normality or parallelism of up to 0.13 (0.0023 rad., or 0.005 inch in a 2 inch-
diameter specimen) could be tolerated without any noticeable effect on the
measured strength. When a spherical swivel was used, there was no detectable
effect on the results for departures from squareness or parallelism up to 0.25
(0.0044 rad., or 0.009 inch in a 2 inch-diameter specimen), the limit investigated
in the experiment. A maximum tolerance for squareness or parallelism of 0.25
was proposed for strength tests in which a spherical seat is used. However, the
effects on strain fields were not studied, and since 0.25 is a coarse tolerance for
Hydrostatic loading. The contact problems arising from lack of flatness, squareness,
or parallelism could largely be avoided if the end of the test specimen were to be
loaded hydrostatically. The most practical way to apply hydrostatic pressure is
probably to confine a readily plasticised solid in a cylinder which has a bore
slightly larger than the test specimen, and then to insert the end of the specimen,
using a flexible but non-yielding membrane to prevent intrusion of the "fluid"
into the rock (Fig. 18C). There are problems associated with this technique, which
will be discussed later in connection with lateral end restraints.
tenvlll.ll.~,i/lllI//////////////I////////~
i Plo ll.,/,llll~l
A B
c D
Fig.18. End loading arrangements for compression specimens.
and its platens, produces radial constraint at the end planes of the specimen. This
is due to mis-match of lateral strain in the specimen and the platen. Various
attempts have been made to eliminate or decrease the frictional restraint in the
compression test by "platen matching", or by the use of lubricants or deformable
interracial layers.
Complete radial restraint. The hardened steel platens of a loading machine are
usually wider than the test sample, and their modulus is higher than that of any
rock. If there is direct unlubricated contact between the platen steel and the rock,
frictional forces are sufficient to provide effective radial restraint. Under these
circumstances the boundary conditions for the end planes of the specimen differ
from the ideal displacement boundary conditions given earlier. At any given time,
the axial displacement will still be uniform across the ends of the specimen, but
the radial and circumferential strains will be zero. Furthermore, there will be
finite gradients of radial and circumferential strain in the axial direction. The
distribution of stresses and displacements for an elastic cylinder with these
boundary conditions has been obtained theoretically by a number of investigators,
including FILON (1902), PICKETT(1944), D'APPOLONIA and NEWMARK (1951), and
BALLA (1960). Balla's results are of special interest, because they permit the effects
of varying friction between platen and rock to be studied and also because they
include relevant numerical results.
Table VI gives Balla's numerical results for stresses in a cylinder of length/
diameter ratio 2.0. The specimen is assumed to be elastic with a Poisson's ratio
of 0.33, there is intimate contact between platen and sample, and there is no slip
at the interface. As a matter of interest, Balla's results for uniaxial compression
, , , , i-o L , , , , J
,o ,o .............
. . . . . . . . ,,,
. . . . . . + ~~ ~ ".-"" . i ''L . . . . . . .. . ~ . . ~,
-o.o
i ~p. . . . . . . . . i/I +I~ "'''" "
Fig.19. Comparison of results obtained by different investigators for the stress distribu-
tion in a restrained cylinder (length/diameter ratio = 1) under uniaxial compression.
are compared in Fig.19 with those of earlier investigators for a cylinder with
length/diameter ratio of 1.0, the ratio favoured in early testing work.
TABLE VI
z/H o, ~ r/R
t Stresses normalized with respect to the average axial stress, a~ = P/A; compressive positive;
sample length L = 2//; L / D = 2.0; v = ~r.
reaches a value equal to the uniaxial tensile strength aT. For a specimen under
gradually increasing load, cracking will therefore commence at those points where
C is greatest, and ultimately crack density will tend to be highest in the zones
where C is highest.
Table VII gives the values of the greatest and least principal stresses al
and a 3, relative to the nominal axial stress (load/area), calculated from the values
given for the stress components in Table VI. Values of C corresponding to these
~ N \ N ' , ,N\N"X \ \ \ \ \ \ \ ~ \ \ \ \ \
===============================================
:::::::::::::::::::::::::::::::::::::::::::::
I
I ~ - : : . ' : - : ~ ~
x
II \\
~~\\\\\',,~\\\\\\\\\\\\\
Fig.20. Contours of the McClintock-Walsh parameter, C, for uniaxial compression with
radial restraint at the sample ends. The numbers give the magnitude of C as a multiple of the mean
axial stress, shaded areas show the most critically stressed zones, and the broken lines indicate
a probable pattern for final collapse of the specimen.
TABLE VII
GREATEST AND SMALLEST PRINCIPAL STRESSES AND MCCLINTOCKWALSH FUNCTION FOR RESTRAINED-
END CYLINDER IN UNIAXIAL COMPRESSION 1
z/H a, C r/R
0 0.2 0.4 0.6 0.8 1.0
pairs of principal stresses have also been calculated and tabulated assuming a
value of 0.7 for #. The distribution of C over a diametral plane is illustrated in
Fig.20, from which the probable sequence of internal crack formation and the
relative crack density in the early stages of fracture can be inferred. It can be seen
that cracking is most unlikely in the dome-shaped regions abutting the loading
platens; the most critically stressed zones lie between these regions of low crack
probability and the central cross-section, and also around the perimeter of the
platen contact, although the numerical results are not too reliable for this latter
region. Fig.20 does not necessarily predict the final mode of failure of a test
specimen, since the stress field will be modified as internal cracks are formed and
deformation is localized. It is, however, consistent with the failure mode in which
the specimen collapses in a radially symmetric pattern, leaving relatively uncracked
end cones which have intruded into cracked regions (Fig. 12C).
It is interesting to note that lateral strain measurements made at various
points along the length of a test specimen after several load cycles by SELDENaATH
and GRAMBERG(1958) show slight variation which can perhaps be associated with
these regions of higher cracking tendency (GRAMB~RG, 1965).
The condition for complete frictional restraint of the end plane has not been
firmly established. MURRELL(1958) suggests that an "average" coefficient of fric-
tion between platen and rock of 0.6 will be sufficient to prevent radial movement,
whereas CHAra~VARTV(1963) gives lower values, ~ 0.2, for Darley Dale Sandstone.
It can be deduced by plotting BaUa's results from Table VI that the shear
stress z,z across the interface between the rock and platen (z/H = 1) for rock with
a Poisson's ratio v = 0.33 is given by:
r
(z,z)z= H = 0.450 R ap (32)
/7 el, ~ R2 =
;f
0 0
0.450
R
ap dr dO (33)
Solution of this equation gives a value for the coefficient of friction of 0.3.
Actually it is fairly obvious that the friction coefficient required to prevent
radial slip will be some function of Poisson's ratio, and in Appendix 1 a simple
calculation is carried out, leading to an expression which relates the average
minimum friction coefficient for no slip, ~, to the Poisson's ratio of the rock, v:
"V
/7 - (34)
2(1 - v)
i.e.,ifv = 0.33,/7 = 0.246; ifv = 0.2,/7 = 0.125.
Complete radial restraint at the end planes represents a violation of the
ideal boundary conditions for the test, and there have been many attempts to
eliminate it. However, it is by no means certain that elimination of radial end
restraint is desirable in practical testing. It can be seen from Fig.20 that, since
gradients of C are quite small in the mid-section of an adequately long specimen,
end restraints should not greatly affect the onset of internal cracking. They do
compensate to some extent for imperfections in end surface preparation, and they
assure consistency of boundary conditions for all types of rock. The main argument
against complete restraint is that it could prevent the formation of axial cleavage
fractures which might otherwise occur (Gv,AMBEam, 1965). The practical problem,
outlined later, is that attempts to relieve end restraint can result in the opposite
effect of the end planes being dragged radially outwards (negative restraint),
which induces axial cleavage when it would not otherwise occur.
Radial freedom. Theoretically there are three ways in which radial freedom of the
loading planes can be achieved. The first is for the radial strain of the loading
platen to match exactly the radial strain of the test specimen, the second is for the
coefficient of friction at every point of the interface to be less than the value which
prevents slip, and the third is to load the end planes hydrostatically:
(1) Platen matching. One way of matching the radial strains of the platen
and the rock is to make the platen diameter equal to the specimen diameter and
then to choose a platen material with certain properties. Using subscripts p and s
to denote platen and rock, the radial strains e under axial stress tr are:
"~p Vs
% = - tr; es = - --tr (35)
Ep Es
I f lateral strains in the platen and the specimen are equal, 8p = e~ and the platen
matching condition is:
Ep _ E s (36)
Vp Vs
TABLE VIII
Material E/~
lbf/sq.inch bar
The problem is to find, for a given rock type, a platen material with a
suitable ratio E/v which will not yield under the highest stresses likely to be
imposed by the test. The ratio E/v for typical rocks lies in the range 4" 106 to 40.106
lbf/sq.inch; these ratios are not easy to match with materials of sufficient strength
for use in platens, particularly at the low end of the range. Table V I I I lists the
E/v ratios for some c o m m o n materials which might be considered as platen
materials.
The drawback to platen matching is that different platens are required for
different rock types, and the reduction of restraint is not consistent from one rock
to another. Also, unless the platen material is very hard it will need resurfacing
from time to time.
Another piece of the same rock is probably the only material that can match
the specimen perfectly and, to some extent, a "rock platen" is provided in dumbbell
shaped specimens, which flare out to greater diameter at the end sections (Fig. 18B,
25). This shape of specimen, used by CrIAKRAVAgTY(1963), BRACE (1964) and
PAULDING (1966), is ground from core sample on a lathe using a toolpost grinder
with a form wheel or, in soft rock, using a stationary single point diamond tool
pivoted on the toolpost. The fillet introduces stress concentrations, which vary with
the fillet radius (Fig.25). PAULDING(1966), who used rather abrupt fillets giving a
stress concentration factor of 1.25, found it necessary to clamp steel rings around
the end sections of his samples to avoid premature splitting in uniaxial compression.
CHAKgAVAgTY(1963) did not experience this difficulty, and after testing specimens
of various shapes advocated a dumbbell with fillet radius 1.4D where D is neck
diameter (Fig.25). The writers have used dumbbell specimens for uniaxial com-
pressive and tensile tests on ice, which presents special problems in end preparation.
Preparation of dumbbell samples is still too expensive and time-consuming
for quantity production in most laboratories, and the simple right cylinder is
likely to remain the most common sample shape for some time to come.
MOGI (1966) has used a technique in which the ends of the test specimens
are bonded with epoxy cement to steel platens which have a slightly larger dia-
meter than the specimen; extra cement is used to form a fillet between the platen
and rock (Fig. 18D). This technique, which attempts to simulate the end conditions
of a dumbbell specimen, ensures intimate restrained contact between platen and
rock, and reduces the high stress concentrations which normally occur at the
periphery of the contact zone. The drawback to the technique is the need for false
platens and the additional time and trouble associated with the bonding procedures.
(2) Friction reduction. Lubrication of the platen interface is intended to
allow the laterally expanding rock to slip past the more rigid platen, so that
frictional forces on the end plane are held to a minimum. Two techniques have
been used: interfacial lubricants, and interposed layers of low friction material or
deformable material (Fig.18C).
There is very little available information relating to the coefficient of friction
between rock and other materials, or to the effects of lubricants on the coefficient
of friction. Some tests carried out by the writers are summarized in Appendix 2.
The main argument against the use of lubricants is lack of control over the degree
of restraint. The coefficient of friction is extremely variable: it is influenced by
the nature of the rock and platen surfaces, and also by the amount and type of
lubricant used. For example, from Appendix 2 it may be noted that for a sandstone-
steel interface the coefficient of friction can vary from 0.47 to 0.17 when lubricants
are used. It does appear, however, that the use of lubricants is unlikely to induce
axial cleavage fractures by extrusion (negative restraint).
The alternative technique for reducing rock-platen friction is to interpose a
layer of low friction polymeric solid (e.g., PTFE/Teflon) or a readily deformable
solid such as rubber, lead or copper. If such materials are extruded during loading,
then theoretically the shear stress on the end planes will be equal to one-half the
axial yield stress of the material (PRANDTL, 1923) and the specimen will fail by
axial cleavage induced by the radial forces. Approximate axial stresses for plastic
yielding (measured by loading the circular end plane of a glass cylinder against a
disc of the material until an impression was made) are: lead: 1600 lbf/sq.inch;
copper: 13,000 lbf/sq.inch; soft brass: 15,000 lbf/sq.inch; hard brass: 20,000
lbf/sq.inch; mild steel: 36,000 lbf/sq.inch. Theoretically, therefore, thick pads
(> 0.01 inch) of these materials should only be used with rock types having a
lower compressive strength than the plastic yield strength of the interposed
material. In such a case the pads have no effect other than platen matching, which
was discussed earlier.
SKINr~R (1959) suggested that a possible method of reducing end restraint
would be to have interposed between the platen and rock a material which has a
plastic yield stress just less than the compressive strength of the rock. He believed
that plastic yielding would relieve the end restraints at the moment of failure, while
strain hardening would prevent extrusion and the development of high outward
radial shear stresses. This is obviously quite impractical.
The question of the effects of thin layers ( < 0.005 inch) of soft materials
interposed between a hard platen and the rock is more complicated, and to date
it has not been resolved. It is unlikely that a very thin layer with thickness of the
order of the rock surface roughness would extrude, and if the coefficient of friction
between such material and the steel platen were sufficiently low, then platen restraint
might be eliminated. For example, using a PTFE/Teflon layer 0.005 inch thick
between sandstone and a lapped steel platen, the writers found that the coefficient
of friction could be reduced to around 0.2 (Appendix 2), which under certain
circumstances would be sufficient to relieve platen restraint (see section on complete
platen restraint). CHAKRAVARTY(1963) was apparently able to achieve a reduction
in platen restraint for tests on Darley Dale Sandstone using a single sheet of
PTFE/Teflon 0.01 inch thick, but he induced axial cleavage as a result of extrusion
when the layer thickness was increased to 0.02 inch. Hsu (1967) found that he
could eliminate end constraint for tests on copper by using a sheet of PTFE/Teflon
only 0.004 inch thick, but found it necessary to interrupt the test eleven times to
keep renewing the sheet.
There appears to be little merit in using thin metallic sheets between the
rock and platen other than as "platen cushions", as the coefficient of friction would
always be too high to allow slippage.
(3) Hydrostatic loading. In principle, hydrostatic loading permits full lateral
freedom at the end of the specimen. In practice, the situation is complicated by the
need to interpose a membrane or diaphragm between the specimen end and the
pressure medium.
SELDENRATHand GRAMBERG(1958) devised a technique, shown schematically
in Fig.18C, in which the specimen is loaded through thin metal plate by a plug
of soft rubber confined in a steel cylinder. HARDY (1959a, b) has used a similar
technique for creep tests and modulus measurement. Although this technique has
been criticised on the grounds that it will induce radially outward tangential forces,
and thereby cause axial tensile splitting, it should be recognised that there are
significant differences between it and the extruding interfacial layer, or the thin
sheets confined between rock and platen just discussed. With full lateral confinement
the plug of deformable material cannot extrude, and any shear forces which may
develop are transmitted to the metal diaphragm and not directly to the rock. Thus
the radial restraints are determined by the response of the diaphragm to loading.
SELDENRATH and GRAMBERG(1958) and GRAMBERG(1965) have discussed
the problem and concluded, on the basis of lateral strain measurements, that their
technique actually maintains some positive lateral restraint of the sample end up
to the point of failure. This conclusion is based on tests in which the load was
cycled 5 or 6 times, but if only the first cycle is considered (fig.10 and 13C of the
paper by SELDENRATHand GRAMBERG,1958) the opposite conclusion could be
drawn. Seldenrath and Gramberg use a fairly thick diaphragm (0.008 inch), and
it seems likely that as long as the metal of the diaphragm remains elastic it must
provide some restraint. However, a new situation arises if the yield stress of the
metal is reached. If the diaphragm yields plastically and flows into the narrow gap
at the rim of the specimen, then radially outward shear stress will be transmitted
to the end of the sample. As already mentioned in connection with interfacial
layers, this shear stress is theoretically equal to one half the axial yield stress of the
metal. Another problem associated with thin soft diaphragms loaded hydro-
statically is the possibility of their being forced into cracks or pores to form
"intrusive failures" (BRACE, 1964).
In view of the complications introduced by plastic flow, it is suggested that
diaphragms should have yield stresses considerably higher than the stresses im-
posed by testing, i.e. they should be made of steel. If the benefits of hydrostatic
loading are to be reaped, diaphragms should also be flexible enough to conform
with the contours of the sample end.
A _ - < - - - -~ro
L (1 + v) (39)
5 E
Taking ao/E = 2 10-3 at failure and v = 0.25:
fl=l/2tan_l{~ E 1 } (41)
~ro ( 1 + v
With A/L = 5" 10-4, tro/E = 2" 10-3 and v = 0.25, the direction of the principal
stress in the sample mid-section will change by 0.57; it is not known whether this
rotation is important.
Spherical seats which are heavily greased so that they do not lock under
load, and which are not loaded at the centre of curvature (as discussed in Appen-
dix 3) will automatically tend to cause lateral platen translation if there is any
unsymmetrical strain in the sample. Referring to Appendix 3, eq. 69 and Fig.35,
the lateral displacement A for small angular rotation 0 is:
A ~hO (42)
For some ball seats in current use, h (measured in the opposite sense from h
in Fig.35) is approx. - 3 inches. If there is a 10~o differential in axial strain across
the width of the sample just before failure, angular rotation of the ball seat will
be approx. 0.03 (approx. 5 10-4 rad.) and this type of ball seat will tend to
displace the platen laterally by more than 0.001 inch.
The other case of lateral platen translation which must be considered is free
translation, which may arise when platens are not fully restrained from sideways
movement. (BIENIAWS~I et al., 1969, state that in conventional testing machines
lateral stiffness is less than longitudinal stiffness by a factor of about 100.) In
this case irregular deflection of the specimen at any stage of loading may cause
the platens to translate, leading to misalignment of loading and instability of the
test. The degree of freedom for lateral translation clearly influences the final mode
of failure of the specimen. If there is complete lateral rigidity and no possibility of
relative lateral translation between platens, and if there is no possibility of platen
rotation during loading (locked ball seat), then it seems most unlikely that a
specimen could fail along a single oblique shear plane. In some recent compressive
tests the writers noticed that failure occurred by "coning" when load was applied
directly by the platens of a 300,000 lbf machine, but when a slender extension
column was used to transmit load into an environmental chamber, "shear plane"
fractures occurred quite frequently. More compelling evidence on this point is
provided by HORINO (1968), who found that specimens loaded between heavily
Fig.21. Axial cracking in acompression specimen with fully restrained ends. Although
there were oblique saw cuts through the specimen, shear displacement along them was prevented
by the end restraints. Note that the specimen has attempted to rotate about the lower left corner.
(After HOR1NO,1968.)
constraining platens refused to fail along a single oblique shear plane even when
there were oblique saw cuts through the specimen (Fig.21).
BmNIAWSKI et al. (1969) have determined that rocks which fail by a typical
shear failure mode in soft testing machines (low longitudinal stiffness) fail by
axial cleavage when loaded in a stiff testing machine (high longitudinal stiffness).
As it can be inferred that high lateral stiffness will usually be associated with high
longitudinal stiffness, this is a very significant finding.
Platen rotation
Two types of platen rotation should be recognised: rotation about the loading
axis, i.e., in the r-O plane, and rotation about an axis normal to the loading axis,
i.e., in the r-z plane.
Rotation about the loading axis. This effect is most likely to arise in the tensile
test when the pulling system tends to twist the specimen (e.g., certain types of
stranded steel cable twist under load). If the twisting moment is Mr, a shear stress
ao~ is induced in the specimen:
16Mr (43)
O-#z -- 7tO3
and the magnitude and direction of the principal stresses are thus changed from
their torque-free values. If the magnitude of the major principal stress ao is not
to change by more than lYo then:
- 1/2 + (1 + (2ao,/uo)2} ~ < 0.01 (44)
i.e.:
PD
Mt < - - (45)
40
where P is the applied axial load. The torque induced by a pulling system can be
measured by hanging deadweight from one half of the system and measuring the
torque necessary to prevent rotation.
To avoid torque in tensile tests, various devices have been used. Pulling
systems have been fitted with thrust bearings or ball-and-socket joints, and non-
twist cable or roller drive chains have been used to apply the loads.
Rotation about an axis normal to the loading axis. This effect can arise when the
load is applied to the platen at a point, as in the typical tensile test or in a compres-
sion test when a small steel ball is used instead of a spherical seating. It can also
arise when using a heavily greased spherical seating, or when using the technique
recommended by Mo3I (1966), in which the platen is loaded at its centre by a
"load equalizer". These conditions involve free rotation in response to unsym-
and the sample will probably break where the excess stress acts. The error introduced
by eccentric loading is:
between the machine platen and a false platen on the sample. It probably fulfills the
primary requirement of compensating for misalignments, but it seems likely that
it will permit some rotation.
The surfaces of seatings and other platens which contact the rock specimen
should be hardened to prevent indentation and pitting during loading. HOSKINS
and HORINO(1968) found that platens harder than Rockwell C30 were satisfactory,
although some A.S.T.M. requirements for concrete testing stipulate a Rockwell
hardness greater than or equal to C58. Fig.22 illustrates how a high strength
specimen, in this case a glass prism (~ 105 lbf/sq.inch), can penetrate a soft
(Rockwell C18) steel platen.
The flatness and smoothness of the platens should be within 30 micro-
inches, which is easily met by normal grinding tolerances.
To assure consistent test conditions from one rock type to another it is advisable
to use platens which are of greater diameter than the specimen.
In Appendix 3 a design for a spherical seat is offered. It is believed that this
design is preferable to the one given in A.S.T.M. E9-67, to comparable designs
shown in A.S.T.M. C93-67 and C133-55, and to the arrangement illustrated by
OBERT and DUVALL (1967). One version of the spherical seat used by the writers
is shown in Fig.23: a Lucite ring for easy centring of the seat on the sample has
Fig.23. Independent ball seat on a compression specimen. The ball seat is centred on the
specimen by a Lucite ring.
been added to facilitate set-up in low temperature tests. The lower end of the
sample is also positioned on the platen by a template. The spherical bearing is
roughly aligned by bringing the upper platen to bear with a few pounds of load;
the loading platen is then backed off and brought into fresh contact for the test.
induces forces and displacements in the loading mechanism, so that there is inter-
action between specimen and testing machine. One consequence of this interaction
is that, with a machine which is insufficiently "stiff", elastic strain energy stored
in the machine is released abruptly and uncontrollably when the test specimen
reaches the limit of its resistance. The loading platens accelerate violently, and the
specimen is destroyed with explosive swiftness. Another possible consequence of
insufficient stiffness or rigidity in the loading machine is elastic distortion, leading
to lack of parallelism or axiality of travel in the loading platens. In order to define
the failure characteristics of the rock fully by a "complete stress-strain curve",
the loading machine should be capable of imposing precise, externally controlled
displacements on the sample, almost irrespective of how the sample reacts. Such a
device is termed a "stiff testing machine".
When the stress in a specimen loaded in a "stiff machine" has reached its
maximum value and the specimen begins to collapse, the rate of energy released
from the machine into the specimen is less than that which can be absorbed by the
specimen in propagating cracks. Failure or collapse can therefore proceed only as
energy continues to be fed into the testing machine from its driving source. Whether
or not the release of energy leads to violent rupture after the maximum stress
(compressive strength) point has been reached therefore depends upon the load-
Mochine
(Perfecfly RIg.ld) I
Spring
stiffness K
Specimen
[ (Perfectly Rl(lld ) I
A
~* ..-xA--
f ~1~F
F
Force
E~
L, Displacement
deformation characteristic of the testing machine relative to that of the rock in the
negative (load decreasing) region.
Referring to Fig.24, the machine can be considered as equivalent to a spring
of stiffness k loaded in series with the specimen. If it is assumed that the complete
envelope of the load-displacement behaviour of the specimen is as shown in
Fig.24, then the slope of the load-displacement characteristic (stiffness) of the
machine k (line ab, Fig.24) must be greater than the slope of the negative portion
of the curve AF/AL at every point. A full discussion of the principle of stiff testing
machines has been recently given by BIENIAWSKIet al. (1969).
The stiffness requirements for testing machines intended to give complete
stress-strain curves for rocks are stringent. Typical testing machines have
stiffness of the order of 0.5 106 lbf/inch, whereas "stiff" machines exceed 5 106
lbf/inch. The practical upper limit at present is around 10. 10 6 lbf/ineh 1. Within
certain limits the stiffness of the testing machine has little effect on the observed
stress-strain characteristics of typical rocks for the "positive" portion of the curve,
although with a soft machine a rock having a significantly non-linear stress-strain
characteristic will not be subjected to a constant strain rate. The amount of the
negative portion of the stress-strain curve that can be obtained will, however, be
related to the testing machine stiffness (BIENIAWSKIet al., 1969).
It is reported (WAWERSIK, 1968) that for some rocks it is necessary to reverse
the direction of displacement in order to maintain stability during progressive
collapse, so that part of the unloading portion of the curve lies beneath the
loading portion, giving double or triple values of stress for certain strains. The
implication here is that strain energy must be released from the sample to prevent
spontaneous crack propagation. However, BIENIAWSKI(1967) suggests that once
the peak of the stress-strain curve has been reached, the rate of specific deformation
is a function of the terminal crack velocity provided that energy can be fed into
the specimen during the deformation. As the terminal velocity is very fast ( ~ 5,000
ft./see for Norite) it seems most unlikely that an externally applied load could be
removed by backing off the testing machine at a sufficiently high rate to be able
to follow the negative portion of the stress-strain curve. When the machine and
specimen are in equilibrium it is, of course, possible to unload the specimen on the
negative side of the stress-strain curve and get a hysteresis loop.
There are several techniques for stiffening commercial testing machines.
Essentially these all require a stiffening element to be put in parallel with the test
specimen, so that the load is shared between the specimen and the stiffening
element. COOK and HOJEM (1966), and BIENIAWSKIet al. (1969) used steel bars,
and WAWERS~K(1968) used hydraulic jacks. In the technique devised by Cook
and Hojem, load is transferred from the bars to the test specimen by thermal
contraction produced by pumping a cooling fluid through them.
1 Stiff loading frames are easily acquired, but actuators and load cells are relatively compliant.
Until the question can be explored in more detail and rational guidelines
laid down, it is suggested that loading rates should be selected on the basis of a
logical consideration of the test material and the application intended. In engineering
tests this may lead to choice of very low rates where test data are to be applied
to structural problems, or high rates where test data are to be applied to problems
of rapid loading, e.g., drilling, crushing, blasting. To facilitate comparison of data,
it might be desirable to choose a loading rate at either the top or bottom of the
available range.
At the research level, it is clearly necessary to explore the effects of loading
rate, strain rate, and "time-to-failure" if the mechanical properties of a rock are
to be defined thoroughly.
R
/
>2D
Fig.25. Proportions of a dumbbell test specimen after BRACE(1964) with stress concentra-
tion factors for two fillet radii. For radius R = 8.0" D the S.C.F. (Stress Concentration Factor) =
1.02 (PAULDING,1966). For radius R = 1.4 D the S.C.F. = 1.05 (CnAKRAVARrV,1963).
linear dimension of the sample, and hence the smallest permissible diameter for a
particular reek type sets the lower limit of absolute size for the sample.
Since the test is usually intended to measure bulk properties of the rock,
the sample must obviously be big enough to be representative of the bulk material,
which suggests a lower limit of Did approximately equal to 10 where D and d are
sample diameter and maximum grain diameter respectively. This also means that
the diameter will be an order of magnitude larger than the size of the controlling
defect structure, which is assumed (BRACE, 1961, and others) to be approximately
equal to the grain size in intact rock. Looking at the question more analytically,
it is seen that the proportion of the sample's cross-sectional area made up of surface
grains becomes smaller as Did increases. The consideration is akin to a specific
surface consideration and, in fact, there is some similarity between the unbounded
conditions of surface grains in a test sample and surface molecules in a liquid or
solid.
In a material made up of randomly oriented grains, the area of influence of
an individual grain extends outwards to a radius of approximately two grain
diameters (Low, 1953) from which it might be argued that the stress field inside the
test specimen is unaffected by the free surface to within two grain diameters of
that surface. It then follows that the ratio of the area unaffected by proximity to the
surface, A', to the total cross-sectional area, A, is:
AA - [ 1- 4 i )]2 (49)
1.2 I t I
A' 0.0
"K
0.4
20 40 60 80
%
Fig.26. Area of specimen cross section which is free from surface effects, as a function
of specimen diameter. Area is normalized with respect to total cross-section area. Specimen
diameter is normalized with respect to specimen grain diameter.
This relation is shown graphically in Fig.26. While the graph gives no direct indica-
tion of how Did affects the stress field, it does suggest that it may be desirable to
have Did > 20.
Specimen length. Once the diameter of a specimen has been chosen and found
adequate in terms of grain size, the length of the specimen can be expressed as a
multiple of the diameter:
(1) Compression. The A.S.T.M. standard for uniaxial compressive tests on
natural building stone (C170-50) specifies that the samples, which may be cubes,
square prisms, or cylinders, should be at least 2 inches high with a height to width
ratio L/D 1> 1.0. Results from tests in which L/D > 1.0 are adjusted by means of
the formula:
5 x 104
_=
-3
!
I
I I I I
0 I 2 3 4
%
Fig.27. Influence of length/diameter ratio (L/D) on uniaxial compressive strength.
1 = Westerly Granite; 2 = Dunham Dolomite; 3 = Muzo Trachyte; 4 = Pennant Sandstone;
5 = Kirkby Siltstone; 6 = Ormond Sandstone and Siltstone; 7 = Darley Dale Sandstone;
8 = Berea Sandstone; 9 = Saturated Granite. References. 1, 2, 3: MoGI (1966); 4, 5, 6: Hones
(1964), 7; CnAKRAVARTY(1963); 8, 9: MELLOn(unpublished).
whereas other data for various rocks, including hard rocks, show it occurring
at LID '~ 2.0--2.5 (MoGI, 1966; GREEN and PERI~INS, 1968; MELLOR and HAWKES,
1969). Theoretical studies (see Fig.20) indicate that rough and rigid platens cause
significant perturbation o f the stress field to a distance o f D/2 f r o m each end.
A n o t h e r noticeable consequence o f varying specimen length is an apparent
dependence o f "fracture angle" on LID for very short samples. With rough
platens, most rocks which are not glassy or very fine-grained collapse in uniaxial
compression to leave conical or wedge-shaped end fragments. In short specimens
(L/D < 1) the half-angle of the cone or wedge, 0, is a function of LID. MOGI (1966)
shows that 0 ~ cot -1 (L/D); unpublished observations on sandstone by the
writers confirm this relation, which can be expressed more simply by saying that
the height o f the cone or wedge is half the height o f the test specimen. Mogi f o u n d
that the dependence of 0 on L/D ceases for LID > 2,5, and this finding is confirmed
by the present writers for sandstone. However, the writers have found that very
short granite samples tend to break leaving only one cone, which is approximately
the height of the specimen. Fracture angle in this granite becomes independent of
L/D for LID > 1. Mogi also suggests, on the basis of limited experimental evidence,
that the critical lower limit for LID depends on the intrinsic fracture angle of the
rock, i.e. on the limiting fracture angle for LID > 2.5. Theoretically the rock
property which influences stress distribution in the vicinity of restraining platens is
Poisson's ratio.
FAIRHURST (1961) notes that eq.50 implies approximate constancy of
strength for L/D > 2.5, and rationalises the observation on the basis of Griffith
theory, pointing out that a fracture inclined 30 to the major principal direction
requires that LID = 2.5 in order to traverse the length of sample which is free
from end restraint (assumed to be 1/6 of the sample length at each end).
COATES and GYENGE (1966) recommend L/D = 2.0, but consider 1.0 <
LID < 2.0 acceptable. However, in another A.S.T.M. publication NEWMANand
LACHANCE (1964) reached the firm conclusion that L/D should be I> 2.5. The
present writers, taking into account all the information available to them, have
adopted a standard of L/D = 2.5 for current testing work, and they consider L/D
= 2.0 to be the minimum acceptable ratio. Theoretical studies by Balla, discussed
in another section of this paper, give ample grounds for rejecting L/D = 1.
(2) Tension. Specimens for tensile tests should also be long enough to
provide a significant volume of material free from "end effect" perturbations of
the stress field. A butt-jointed tensile specimen is essentially similar to a compres-
sion sample loaded by "rough" platens, and the stress field should be the same for
both cases, with only a change of sign. Thus "end effects" are likely to perturb
the stress field significantly to a distance of D/2 from each end, and an absolute
lower limit of L/D = 1 is indicated.
Since the tensile test is inherently stable, there is no upper limit for LID
comparable to the buckling limit in the compressive test, but in practice samples
which are very long and slender are too fragile for handling and machining. It is
recommended that the test specimen, or the neck of a dumbbell specimen, should
have LID between 2.5 and 3.0. If the butt-jointing method is used, it might be
desirable to adopt LID = 2.5 for both compressive and tensile tests so as to maintain
geometric similarity between the two tests.
Specimen volume and size effects. The preceding considerations set certain limits,
but do not standardize specimen size, and absolute volume may vary considerably.
For example, the volume of a 2.125 inch diameter sample is about ten times as
great as the volume of a geometrically similar sample of 1 inch diameter, while a
4.5 inch diameter sample has almost 100 times the volume of a 1 inch diameter
sample. It is an experimental fact that, for most solid materials, strength decreases
Pb = p(b/a)~ (51)
where a and b are characteristic linear dimensions of the two samples (e.g., dia-
meter, or side length of cube) and 0t is a constant with the value 1, 2 or 3, depending
on whether the strength-controlling factor is distributed with length, area or volume
respectively.
From cube crushing tests on coal, EVANSand POMEROY(1958) found ~t ~ 1.
They also found experimentally that the mean compressive strength was inversely
proportional to a power of the specimen size:
where a = side length of cube specimen, and ]~ is a constant, which had values of
approximately 3 and 6 for the two coal types tested.
At this stage it becomes necessary to consider the form of the strength
distribution curve. A basic argument of the weakest link theory is that, as sample
size increases, the probability of finding an element weaker than any of those
found in small samples increases. The statistical problem is concerned with the
probability of finding, in a sample of given size chosen at random from a mass
containing a population of elements, a single element which will fail at a stress X
which is less than or equal to a stipulated value x. This involves a distribution
function F(x), which gives the probability P that an individual chosen at random
from a parent population will have a value X .N< x. The cumulative distribution
function F(x) can be written in the form:
Some information on the form of the distribution functions F(x) and ~b(x)
has been obtained experimentally. Evans and Pomeroy found that their strength
results for coal followed approximately a normal (Gaussian) distribution. SrdNr~ER
(1959) obtained similar distributions for tests on anhydrite, but suggested that the
probability of occurrence of a flaw was likely to decrease with increase of flaw size.
Drawing attention to the connection between flaw size and crystal size, and ob-
serving that the distribution of crystal size in his anhydrite was of Laplacian form,
he proposed a Laplacian distribution for flaws:
Taking this as the distribution for critically oriented flaws, and assuming that these
flaws control strength according to the Griffith theory (eq.ll), he deduced that
modal strength (most probable value of strength) S v for a specimen would be
related to its volume V by:
k
Sv - - - (57)
where k is a constant and p is the flaw density (number of flaws per unit volume).
WEmULL (1951) reasoned that ~(x) for a wide range of phenomena could
be represented by a power function, giving:
X -- Xulm
F(x) = 1 - exp (58)
1
log Sv = K - log V (59)
m
or, comparing strengths for two geometrically similar samples of different volume,
denoted by subscripts 1 and 2:
so2 ,,m
Svl \-~2/ (60)
on fine-grained and glassy materials, they apply only indirectly to final collapse of
most compression specimens, in which multiple internal cracking precedes collapse.
However, if there is a reasonably constant ratio of stress for onset of cracking
to stress for specimen collapse, then the volume effect on "ultimate", or collapse,
strength will be of the same form.
Stress-strain measurements
It is desirable to have accurate stress-strain records as a routine output from
uniaxial tensile and compressive tests. Recording load is no problem; most modern
testing machines have adequate provision for load recording built in, and when
necessary it is easy to add an external load cell and record its output on a potentio-
metric device or an oscilloscope. Recording strain or displacement in the test
section of the specimen is more difficult.
Strain distribution in test specimens. Under the ideal displacement boundary con-
ditions axial, radial and circumferential strains would each be uniform throughout
the specimen. In particular, uniformity of lateral strain would permit the specimen
to retain its cylindrical shape during loading. In reality, the original cylindrical
shape of the specimen tends to become distorted as a result of end constraints
on the specimen.
When a specimen is pressed between a pair of rough and rigid platens, so
that its ends are completely restrained radially, it tends to take on a barrel shape
under load. With a short specimen (LID = 1) the profile of lateral deformation
may show a true barrel shape, with a continuous smooth increase of deformation
from the ends to the mid-section. With a longer specimen (LID = 2.5) the profile
o f lateral deformation is more likely to show approximately uniform deformation
along the mid-portion of the specimen (Fig.28).
Jl J II ~1 I i
I I I 1 I
I I I I !
I I j ~1
I
I '1
Ii I I I I I
A
I !
/ \
"//Z "//////////7 V//////////7/~ ///////////////////
Fig.28. Schematic profiles of lateral deformation in uniaxial compression test. A. Perfect
radial freedom at the specimen ends. B. Complete radial restraint at the specimen ends. C. Ex-
trusion of an interracial material at the platen/rock contact.
Strain measurement. Provisions are made in many testing machines for recording
crosshead displacement as a function of load. In general, however, crosshead
displacement tends to exceed sample deformation, sometimes to a considerable
degree, due mainly to the bedding-in deformation as asperities on the end planes
of the specimens are crushed. Also, as the displacement recording equipment on the
machine is relatively insensitive to the small displacements associated with the
deformation of rock samples it is rarely, if ever, used for precise strain measurement.
In the past, most strain records have been obtained from electrical resistance
strain gages bonded onto the sample in the axial and circumferential directions.
These give excellent results, with strain resolution better than 5 10-6 when they
are installed properly. For precise work it is common to have three axial strain
gages spaced 120 apart in the middle third of the sample, with one or two
circumferential gages. For slow loading tests a simple bridge readout can be used,
but for higher rates of loading a potentiometric recorder is needed. Bonded strain
gages are not very suitable for tests on saturated or frozen rocks.
HAWKES (1966) has described a very simple photoelastic strain gage, in
which the strain readout is in terms of photoelastic interference fringes for strain
measurements on rock cores. The gage has a strain resolution better than 10-s
and appears to have several advantages over the electrical resistance strain gage,
including savings in time and expense.
The more obvious alternatives to bonded strain gages are demountable
extensometers. There are two main classes of these: mechanical, in which the
readout is by dial gages, scale pointers, or light beams; and electrical, which
incorporate electrical resistance strain gages, linear variable differential transformer
transducers (L.V.D.T.'s), or capacitance gages, giving voltage readout.
A simple direct-reading mechanical extensometer is described in A.S.T.M.
standard C469-65. Axial strain is obtained by measuring the convergence of two
yokes clamped to the sample on a dial micrometer. To obtain the specified sen-
sitivity of 5 1 0 - 6 inch per inch of gage length, the gage length would have to be
,-~ 10 inches, since the best dial gages are graduated down to 10-4 inch, and thus
the device is unsuitable for tests on small rock samples. Diametral strain is measured
by a third hinged yoke, from which two diametrically opposite steel points are
screwed into the face of the sample. Relative movement of the measuring points
is indicated by a dial micrometer at the free end of the yoke. This middle yoke is
supported on the lower yoke, which means that it is forced to rotate slightly as the
sample strains axially; this shortcoming occurs in other similar gages described
in the literature. Improved sensitivity can be obtained with mechanical or optical
magnification, but it is difficult to arrange for the automatic recording which is
needed in rapid tests.
The sensitivity of demountable electrical extensometers can exceed 10-7
inch resolution with a sensitive (1/~V) readout, and it is a simple matter to record
their outputs automatically on X-Y plotters or potentiometric strip-chart recorders,
which are able to trace input voltage changes down to approximately 10/tV, with
maximum chart pen sensitivities in the range 1.0 to 0.1 mV/inch.
It is beyond the scope of this review to discuss the many different types of
extensometers available (see, for example, LEBOW, 1966), but from the experience
of the writers it seems that the following factors should be borne in mind when
designing extensometers or bonded strain gage systems for strain measurements
on rock test specimens:
(1) The axial gage length should be symmetrical about the mid-section of
the specimen; it should be as long as possible, but not less than 5 times the maximum
grain diameter of the rock (ROCHA, 1965). The maximum gage length is set by the
necessity of keeping clear of the anomalous end zones of the specimen; referring
to Fig.20, it is suggested that the measuring points should not be closer to the
platens or end connections than D/2, where D is the specimen diameter. Almost
without exception, extensometers designed for Poisson's ratio measurements are
mounted across a diameter of the test specimen. When making measurements of
the negative portion of the complete stress-strain characteristic it is not usually
practical to mount strain gages or extensometers on the test specimens due to the
general breakup of the specimen. In this case, there may be no practical alternative
to mounting gages between the loading platens.
(2) The system of support and attachment for extensometers should be
light, compact and easy to handle. When testing to failure it is also essential that
the extensometer be capable of withstanding explosive disintegration of the test
specimen, particularly when testing in compression in non-stiff machines (L~EMAN
and GROBBLAAR, 1957). A special problem is the clamping of the extensometers
at the measuring points. Set screws sometimes used for this purpose have a tendency
to bite into the rock surface and create stress raisers which can cause premature
failure in tensile specimens.
(3) It is important that the strain measuring system be sensitive, robust and
stable. The total strains to be measured will depend upon the nature of the test
and the type of rock, and will vary from 20 to 20,000 microstrains (2.10 -5 to 2.10-3).
Fig.29 illustrates a system which attempts to embody the latest techniques
for continuous strain recording on rock test specimens. Axial deformation is
Fig.29. Arrangement of L.V.D.T. gages for measurement of axial and radial deformation
in uniaxial tension. (The heavy cable to the radial gage was later replaced by a lightweight
construction.)
screws, while the transformer core rods are held by another clamp near the other
end of the sample. A simple split annular spacer is used to mount the clamps
symmetrically and parallel to each other on the test piece. Diametral deformation
is sensed by a third L.V.D.T. mounted at the open end of a prestrained U-yoke,
which is clamped to the mid-section of the sample by two diametrically opposed
set screws. The yoke is supported solely by its measuring screws, but is prevented
from rocking by lightly strained rubber bands attached to the upper clamping
frame. Elastic bands are also used to couple the upper and lower clamps, so that
in tensile tests to failure the specimen is held together. L.V.D.T. transducers are
particularly suitable for tests to failure, as they are very robust and there is no
mechanical connection between the transformer and probe elements.
Signals from the L.V.D.T.'s are fed into two X-Y recorders: output from the
axial gages is recorded against the load cell output to give a continuous load-axial
deformation curve, while output from the diametral gage is fed into the second
recorder with the output from the axial gages to give a continuous plot of lateral
deformation against axial deformation.
Calibration of the complete system is carried out on a cylinder of aluminum,
for which Young's modulus and Poisson's ratio are known.
Test results
The effort expended on precise testing may be largely wasted if the standards
of recording, reducing and reporting data are inadequate. The first requirement is
for a complete, clear and accessible permanent record of all test data, including
details of the equipment and technique, and a full quantitative description of the
specimen and its condition. The second requirement is for condensation and
presentation of the results in conformance with accepted statistical and graphical
conventions.
In condensing the results of multiple tests or replications (OBERT et al.,
1946, recommend at least ten repetitions for any given sample size and rock type
to get a representative result) it is usual to give the arithmetic mean of each group
of identical tests. However, for some purposes it may be useful to give the mode
(most probable value) as well as the mean, for the distributions of actual test
results tend to be skewed. The usual measure of variability for a group of tests
is the standard deviation or the variance, but as a general rule the extreme values
of a group should also be given. In graphical presentations the convention of
plotting the dependent variable as ordinate and the independent variable as
abscissa should be followed; this helps to avoid confusion when lines or curves
are fitted to the data by regression analyses. When empirical relationships are
obtained by graphical linearization of data, e.g., log-log or log-linear plots, the
implications of the resulting relationships should be checked for physical plausibility
by examining the boundary conditions, i.e., the values of the dependent variable
and its derivatives for extreme values of the independent variable. When apparent
change of one variable with respect to another is small, a statistical test for
significance should be applied before any conclusions are drawn.
The percentage standard deviation of test results provides an index of re-
producibility, and indirectly an index of quality for the test. OBERT and DUVALL
(1967) give values of standard deviation which they consider to be near the
attainable minimum for uniaxial compression tests. These values range from
3.5~o to 10~o for various rocks. HOSKINSand HORINO(1968) give results for tests
on very carefully prepared compression specimens with standard deviations from
2.4% to 3.0?/0 for various rocks. When the same testing procedure was used on
specimens whose ends were in the condition left by the saw, the standard deviations
went up, ranging from 7.5% to 27?/0.
DEKLOTZ(1967) quotes values of coefficient of variance for concrete testing.
They range from more than 20 %, which is classified as "poor", to less than 5 %,
classified as "probably attainable only in well controlled laboratory tests". He
suggests that in rock testing a coefficient of variance of around 3% should be
possible.
Specimen preparation. Samples should be cut and prepared using clean methods
(see p. 190). Specimen diameter should be not less than ten times the maximum
grain size of the rock, and preferably more than twenty times the grain size (see
p.252). The ratio of length to diameter should be not less than 2.0, and preferably
2.5 (see p.254). Specimen ends should be flat to within 0.0001-0.0005 inches,
depending on the strength and modulus of the rock and the specimen diameter
(see p.230). They should be parallel to each other within 0.001 D, where D is
specimen diameter, and square to within 0.06 or 0.001 rad. (see p.231). Generators
of the cylindrical surface should be straight to within 0.001 inch, and the diameter
of the specimens should not vary by more than + 0.001 inch over the length of the
sample. Dimensions and tolerances should be checked with a eomparator (see
p.195). The composition and condition of the best specimen should be fully
described (see sections beginning on pp.180 and 184) and quantitative measure-
ments of index properties (see p. 184) should match the precision of the mechanical
tests.
Spherical seats. A spherical seat designed in accordance with the section on platens
and spherical seatings (p.245) and Appendix 3 should be placed on the upper end
of the specimen. It should be lightly lubricated with mineral oil so that it locks
after the deadweight of the crosshead has been picked up. Both specimen and
spherical seat should be accurately centered with respect to the loading machine.
Testing machine. The capacity of the testing machine should be sufficient to assure
reasonable longitudinal stiffness (preferably > 106 lbf/inch) for the size of
sample tested, either directly or by means of stiffening blocks (p.247). Lateral
stiffness should be sufficient to prevent relative lateral displacement of the platens
from exceeding about 0.001 inch (see p.240). Free play, or backlash, in the cross-
head should be less than 0.2% of the distance between the machine columns
(see p.243). Unless the integral spherical seating conforms with the conditions
outlined in the section on platens and spherical seatings and Appendix 3, it should
be removed and replaced with a rigid platen and an independent spherical seat of
approved design. The testing machine should provide adequate control of head
speed or loading rate over a wide range (2-3 orders of magnitude or more), and
maximum head speed should be greater than one inch/min.
Loading rate. Loading rate, or strain rate, should be selected to suit the particular
purpose of the test (see p.250), taking into account the rheological properties of the
rock under test (see pp. 197, 200). The chosen rate should be maintained constant
throughout the test.
Stress-strain measurements (see p.259). For all but the slowest or the simplest
kinds of tests, automatic recording of load and displacement is desirable, so that
independent load cells and electrical strain gages are likely to be required. The gage
length for axial strain measurements should be as long as possible, and not less
than five times the maximum grain diameter of the rock. It should not encroach
within D/2 of the specimen ends, where D is diameter. Radial or circumferential
strain should be measured in the mid-section of the sample, not closer than D/2
to the ends. All gages should be placed symmetrically with respect to the mid-
point, and they should bear directly onto the sample and not onto the platens.
The required gage sensitivity is of the order of 5 10-6 inch or better. Demountable
electrical gages with remote automatic readout are desirable, but they must be
protected against explosive disintegration of the specimen.
Test results (see p.263). Results should be reported in full. When they are condensed
for summary presentation, accepted statistical and graphical conventions should
be followed. Standard deviation, which ought to be of the order of 5 ~ for good
tests, should be given.
used for attaching the sample to the pulling device, with varying degrees of success.
(1) Gripping the specimen mechanically. The straightforward solution to
the problem is to grip the cylindrical end of the rock specimen mechanically in
some form of clamp, collet or chuck. With a suitably designed grip this method
may be acceptable for holding dumbbell samples, provided that accurate centring
can be assured. It is, however, quite unsatisfactory for simple right cylinders, as the
stresses induced by clamping influence the failure. GROSVENOR(1961) reports that
clamped specimens often broke at the grips.
(2) Cementing the specimen into a metal collar. OBERTet al. (1946) inserted
rock cores into metal collars or cups, and filled the annular space between core
and collar with an adhesive compound. They concluded that the method was
unsatisfactory for general use. The writers have used a more refined method for
attaching metal collars, and judge it to be satisfactory for some purposes. The
method is described later in this section.
(3) Cementing the specimen to a metal cap by direct butt-jointing. Modern
high-strength adhesives permit high tensile stresses to be transmitted across a plane
cemented joint. Thus, the squared ends of rock cylinders can be butted directly
against the end faces of connector plugs, cemented in place, and pulled. The
method is described in detail later. For very strong rocks it may be difficult to
develop sufficient bond strength. When the tensile test is performed by the butt-
jointing method, the end conditions become similar to those prevailing in the
conventional uniaxial compressive test with complete end restraint.
(4) Casting the specimen into plugs of some other material. As an alter-
native to grinding a rock cylinder into a dumbbell shape, it may be possible to
cast a simple right cylinder of rock into shaped end plugs of another material
which has elastic properties comparable to those of the rock. Possible substances
include sand/cement grout, stone plaster, sulphur, epoxy adhesives, and casting
resins. The resulting dumbbell could be gripped mechanically. A simple but
unsuccessful application of this method is described by GROSVENOR(1961).
Pulling systems. The mechanism used for pulling the test specimen must be such
that it cannot introduce any significant bending or torsional stresses into the sample.
To avoid bending stresses caused by misalignment of crosshead connections,
most testing machines have universal joint couplings provided for tensile testing.
While these are desirable, it seems advisable to pull rock specimens through
additional flexible connectors such as cables or chains. Stranded steel cables
provide excellent flexibility over a short length, but they stretch (allowing strain
rate to vary when machine speed is constant) and they have a tendency to twist
(minimized in non-twist cable). Roller drive chains have been used as an alter-
native to cables; the roller axes of upper and lower chains are set mutually per-
pendicular to give full flexural freedom. A tendency towards twisting in the pulling
cables or chains, which introduces torque into the specimen, can be avoided to
some extent by using ball and socket joints, but a thrust bearing in the system is
probably better (air bearings are sometimes used when testing metals).
The actual connection between the pulling system and the rock sample
must assure coincidence of the line of action of the applied force and the axis of
the test specimen. For specimens of typical size the acceptable tolerance for
eccentricity is approximately 0.001 inch (see section beginning on p.245).
The butt-jointing method. For ideal consistency the uniaxial tensile test ought to be
identical in form to the uniaxial compressive test, with only a change in the sign
of the major principal stress. To achieve this, a bond has to be developed between
the platen and the specimen, and end conditions corresponding to those of the
compression test have to be imposed.
The required bond can be formed with modern adhesives, but in practice
it is very difficult to assure complete absence of bending stresses unless the platens
are free to rotate. Thus the current compromise is a method in which the specimen
is butt-jointed to platens which are free to rotate 1. This kind of arrangement has
been described by FAIRHURST (1961) and GROSVENOR (1961) for rocks, and by
HUGHES and CHAPMAN (1965) for concrete. The following notes summarize
Fairhurst's version of the method.
A cylindrical sample (1 inch diameter x 2 inch long) with its ends lapped
flat and square is cemented to cylindrical steel end caps of the same diameter. The
end caps are grooved on the face which contacts the rock to retain epoxy adhesive
and improve the bond. Their opposite ends are drilled and tapped in the exact
centre to permit attachment of pulling cables. The rock cylinder and the end
caps are assembled and cemented in a special jig designed to align them accurately.
The specimen is pulled by stranded steel cables (0.125 inch diameter) screwed into
the end caps; one of the cables attaches to the loading device through a thrust
bearing to eliminate torque from cable twisting. Fig.30 illustrates the set-up.
This method is quite satisfactory for tests on many rocks. The stress field
probably corresponds closely to the stress field developed in the typical compressive
test with rough and rigid platens, and observations on photoelastic models (Fig.31A)
indicate that there are no serious stress concentrations near the end planes. For
tests on high strength rocks, the bonds developed by cold-setting epoxies tend to
be inadequate. High temperature curing is inadvisable, as it may affect the proper-
ties of the test specimen, and so modified techniques have to be used for high
strength rocks.
The collar method. To obtain an adequate bond for tensile tests on high strength
rock it is usually necessary to resort to the use of collars. However, bonded collars
1 HUGHESand CHAPMAN(1965) used a serf-alignment ball race, which appears to obviate the
need for platen rotation.
I Cable
Bearing
p End)
;crew
nnector
End Cop
Ill or Aluminum)
Epoxy
Joint
Rock Specimen
bds Fiat and Square)
Fig.30. Butt-joint method for uniaxial tensile tests. (After FAmHtngST, 1961.)
Fig.31. Photoclastic study of stresses generated by: A. Butt jointed specimen, and by:
B. Collar mounted specimen. Numbers give isochromatic fringe orders. Note the stress concen-
tration in collar mounted specimen.
tend to introduce serious stress concentrations in the specimen (Fig.31B). The most
favourable type of collar is the rock collar represented by the flared end of a
dumbbell specimen, but even in a dumbbell specimen with long radius fillets there
are stress concentrations which would cause the specimen always to break at the
base of the fillet in an otherwise perfect test.
As an expedient alternative to dumbbells, the writers have used aluminum
collars designed to approximate the effect of rock fillets. A length of core (1 inch
diameter x 4.75 inches long) is cemented into chamfered aluminum collars
(Fig.32A, B). The collars are machined to allow a clearance of approximately
0.003 inch (total) between rock and aluminum, and they are slotted longitudinally
n-twist]
Connector
less steel)
to minimize hoop stresses and provide some radial and circumferential strain free-
dom. Before assembly the bore of the collar and the 1 inch length of core to be
inserted into it are smeared with high-viscosity epoxy adhesive; the surplus epoxy
extrudes as a rim when the rock is inserted, and this rim is then wiped to form a
smooth fillet between the specimen and the chamfer of the collar. After assembly
the specimen is laid in a vee-block, where the collars are weighted to maintain
alignment during setting. The specimen is finally trued by grinding it lightly in a
lathe. A dummy pair of cable connectors, identical to those on the testing machine,
provide running centres on the lathe to assure complete axiality, and the same
connectors are used in the comparator for checking dimensions. Finished samples,
which have an eccentricity with respect to the pulling points of less than 0.0005 inch,
are pulled by non-twist stainless steel aircraft cables (0.22 inch diameter).
These are coupled to the specimen connectors by swaged stainless steel balls,
which are lapped into spherical seats in the screw connectors (Fig.32A, B). To
avoid undue stretching during loading, the cables are short (4.5 inch each) and
their allowable load limit is more than twice the expected maximum service load.
After testing, the aluminum collars are recovered: the epoxy is softened by heat,
rock fragments are withdrawn, and the collars are cleaned by soaking in trichlor-
ethylene.
This method yields stress-strain data up to failure, but there is an un-
avoidable tendency for fracture to occur at the collar.
ACKNOWLEDGEMENTS
REFERENCES
ATrmRTON, M. J., 1965. Some experiences with a commercial compression testing machine.
Mag. Concrete Res., 17(50): 45--46.
BAU.A, A., 1960. Stress conditions in triaxial compression. J. Soil Mech. Found. Div., Am. Soc.
Cir. Engrs., 86(SM86): 57-84.
BmNIAWSKI,Z. T., 1967. Mechanism of brittle fracture of rock. Intern. J. Rock Mech. Mining
ScL, 4: 395-430.
B1ENIAWSKI,Z. T., 1968. The effect of specimen size on compressive strength of coal. Intern. J.
Rock Mech. Mining Sci., 5(4): 325-335.
BmNIAWSKI,Z: T., DENKHAUS,H. G. and VOGLER,U. W., 1969. Failure of fractured rock. Intern.
J. Rock Mech. Mining Sci., 6(3): 323-341.
BOMBALAKIS,E. G., 1964. Photoelastic investigation of brittle crack growth within a field of
uniaxial compression. Tectonophysics, 1(4): 343-351.
BOOZER,G. D., HILLER,K. H. and SERDSNOECTI,S., 1963. Effects of pore fluids on the deforma-
tion behaviour of rocks subjected to triaxial compression. Proc. Symp. Rock Mech., 5th,
Univ. Minnesota, Minneapolis, Minn., 1963: 579-624.
BRACE, W. F., 1961. Dependence of fracture strength of rocks on grain size. Proc. Symp. Rock
Mech., 4th, Penn. State Univ., Pa., 1961: 99-103.
BRACE, W. F., 1964. Brittle fracture of rocks. In: W. R. JUDD (Editor), State of Stress in the
Earth's Crust. Elsevier, New York, N.Y., pp.110-178.
BRACE, W. F., 1965. Some new measurements of linear compressibility of rocks. J. Geophys.
Res., 70(2): 391-398.
BRACE, W. F. and BOMBALAgaS,E. G., 1963. A note on brittle rock in compression. J. Geophys.
Res., 68(12): 3709-3713.
BRACE, W. F. and BYERLEE,J. D., 1967. Recent experimental studies of brittle fracture in rocks.
Proc. Symp. Rock Mech., 8th, Univ. Minnesota, Minneapolis, Minn., 1967: 58-81.
BRACE, W. F., PAULVING, B. W. and SCHOLZ, C., 1966. Dilatancy in the fracture of crystaline
rocks. J. Geophys. Res., 71(16): 3939-3953.
BROWN, S. D., BIDI)ULPH, R. B. and WILCOX, P. D., 1964. A strength-porosity relation involving
different pore geometry and orientation. J. Am. Ceram. Soc., 47(7): 320-322.
BROWN, J. W. and SINGH, M. M., 1966. An investigation of microseismic activity in rock under
tension. Trans. Soc. Mining Engrs., 235: 255-265.
BRUNAUER,S., EMMETT, P. H. and TELLER, E., 1938. The adsorption of gases in multimolecular
layers. J. Am. Chem. Soc., 60: 309-316.
BURSFITEIN, L. S., 1967. Tension and compression diagrams for sandstone. Fiz.-Tech. Probl.
Razrabotki lskop., 1967(1): 24-29.
BURTON, A. N., 1965. Classification of rocks letter. Intern. J. Rock Mech. Mining Sci., 2(1): 105.
CHAK~VARTY, P. K., 1963. Application of the Photoelastic Technique to the Problems of Rock
Mechanics. Ph. D. Thesis, Sheffield Univ., unpublished.
CHARLES, I., 1959. The strength of silicate glasses and some crystalline oxides. In: A. AVERBACK
et al. (Editors), Fracture. Wiley, Chapman and Hall, London, pp.225-249.
COATES, D. F., 1964. Classification of rocks for rock mechanics. Intern. J. Rock Mech. Mining
Sci., 1(3): 421-429.
COATES, D. F., 1965. Principles of Rock Mechanics. Can. Dept. Mines Tech. Surv., Ottawa,
Mines Branch Monograph, 874:330 pp.
COATES, D. F. and GYENGE, M., 1966. Plate-load testing on rock for deformation and strength
properties. In: Testing Techniques for Rock Mechanics--Am. Soc. Testing Mater., Spec.
Techn. Publ., 402: 19-40.
COATES,D. F. and PARSONS,R. C., 1966. Experimental criteria for classification of rock substances.
Intern. J. Rock Mech. Mining Sci., 3(3): 181-189.
COLBACK, P. S. B. and WnD, B. L., 1965. The influence of moisture content on the compressive
strength rock. Proc. Can. Symp. Rock Mech., 3rd, Toronto, Ont., 1965: 65-83.
COOK, N. G. W., 1965. The failure of rock. Intern. J. Rock Mech. Mining Sci., 2(4): 389-403.
COOK, N. G. W. and HODGSON,K., 1965. Some detailed stress-strain curves for rock. J. Geophys.
Res., 70: 2882-2888.
CooK, N. (3. W. and HOJEM, K., 1966. A rigid 50-ton compression and tension testing machine.
S. African Mech. Engr., 15(11): 389-403.
COTTRELL,A. H., 1959. Theoretical aspects of Fracture. In: A. AWRBACKet al. (Editors), Fracture.
Wiley, Chapman and Hall, London, pp.20-53.
D'APPOLONIA, E. and NEWMARK,N. M., 1951. A method for solution of the restrained cylinder
under axial compression. Proc. U.S. Natl. Conf. Appl. Mech., 1st, Chicago, IlL--Am. Soc.
Mech. Engrs., 1961 : 217-226.
DEEI~a~, D. U., 1966. Discussion on rock classification. Proc. Congr. Intern. Soc. Rock Mech.,
1st, Lisbon, 3: 156-158.
DEEI~, D. U. and MmLER, R. P., 1966. Engineering classification and index properties for intact
rock. Air Force Weapons Lab., Techn. Rept. AFWL-TR-65-116, Kirtland A.F.B.,
New Mex., 308 pp.
DEKLOTZ, E., 1967. Discussion on properties of rocks and rock masses. Proc. Congr. Intern. Soc.
Rock Mech., 1st, Lisbon, 3: 267-268.
DRUCKER, D. C. and PRAGER,W., 1952. Soil mechanics and plastic analysis or limit design.
Quart. AppL Math., 10(2): 157-166.
DURELLI, A. V. and PARKS, V. I., 1967. Influence of size and shape on the tensile strength of
brittle materials. Brit. J. AppL Phys., 18: 387-388.
EMERY, C. L., 1964. Strain energy in rocks. In: W. R. JUDD (Editor), State of Stress in the Earth's
Crust. Elsevier, New York, N.Y., pp.235-260.
EVANS, I. and POMEROY,C. D., 1958. The strength of cubes of coal in uniaxial compression.
In: W. H. WALTON (Editor), Mechanical Properties of Non-Metallic Brittle Materials.
Butterworths, London, pp.5-25.
EVERLING,G., 1964. Comments upon the definition of shear strength. Intern. J. Rock Mech.
Mining Sci., 1: 145-154.
FAIRHURST,C., 1961. Laboratory measurements of some physical properties of rocks. Proc.
Symp. Rock Mech., 4th, Penn. State Univ., Pa., pp.105-118.
FAmHURST, C. and CooK, N. G. W., 1966. The phenomenon of rock splitting parallel to the
direction of maximum compression in the neighbourhood of a surface. Proc. Congr.
Intern. Soc. Rock Mech., 1st, Lisbon, pp.6874/91.
FILON, L. N. G., 1902, On the elastic equilibrium of circular cylinders under certain practical
systems of load. Phil. Trans. Roy. Soc., London, Ser. A, 198: 147-233.
FRIEDMAN,M., 1967a. Discussion on description of rocks and rock masses. Proc. Congr. Intern.
Soc. Rock Mech., 1st, Lisbon, pp.182-197.
FRIEDMAN, M., 1967b. Measurement of the state of residual elastic strain on rocks by X-ray
diffractometry. Norelco Reporter, 14: 7-9.
FRIEDMAN,M., PERKINS,R. D. and GREEN,S. J., 1969. Observation of brittle-deformation features
at the maximum stress of Westerly Granite and Solenhofen Limestone. Intern. J. Rock
Mech. Mining Sci., in press.
GONNERMAN, H. F., 1925. Effect of end condition of cylinder in compression tests of concrete.
Proc. Am. Soc. Testing Mater., 24(2): 1036-1065.
GRAMaERG, J., 1965. The axial cleavage fracture. Eng. GeoL, 1(1): 31-72.
GREEN, S. J. and PERKINS, R. D., 1968. Uniaxial compression tests at strain rates from 10-4/sec to
104/sec on three geologic materials. Proc. Symp. Rock Mech., lOth, Austin, Tex., in
preparation.
GRIFHTH, A. A., 1921. The phenomena of flow and rupture in solids. Phil. Trans. Roy. Soc.
London, Ser. A, 221: 163-198.
GRtVHII-I, A. A., 1924. Theory of rupture. Intern. Congr. AppL Mech., 1st, Delft, 1924: 55-63.
GROSVENOR, N. E., 1961. A new method for determining the tensile strength of a rock. Trans.
Am. Inst. Mining Met. Petrol. Engrs., 220: 447-449.
GROSVENOR,N. E., 1963. Specimen proportion--key to better compressive strength tests. Mining
Eng., 15: 31-33.
HANDIN, J., 1966. Strength and ductility. In: S. P. CLARK(Editor), Handbook of Physical Contacts.
Geol. Soc. Am., New York, N.Y., pp. 223-289.
HANSEN,H., KIELLAND,A., NIELSEN,K. E. C. and THAULOW,S., 1962. Compressive strength of
concrete--cube or cylinder. Bull. Reunion Intern. Lab. Essais Rech. Mater. Constr., 17:
22-30.
HARDY,H. R., 1959a. Standard procedures for the determination of the physical properties of
mine rock under short period uniaxial compression. Mines Branch, Dept. Mines Tech.
Surv., Ottawa Tech. Bull., TB8:108 pp.
HARDY, H. R., 1959b. Time-dependent deformation and failure of geologic materials. Quart.
Colo. School Mines, 54(3): 36.
HARDY, H. R., 1967. Analysis of the inelastic deformation of geologic materials in terms of
mechanical models. Proc. Soc. ExptL Stress Analysis, Spring Meeting, Ottawa, Ont., in
press.
HARDY, H. R., OKULICH, P. J. and KAPELLER, F., 1966. Preparation of small cylindrical test
specimens of geologic materials. Can. Dept. Mines Tech. Surv., Fields Mining Pract.
Div., Ottawa, Ont., Report FMP 66/67-P, May 1966, unpublished.
HAWKES, I., 1966. Moduli measurements on rock cores. Proc. Congr. Intern. Soc. Rock Mech.,
1st, Lisbon, pp.655--660.
Hom3s, D. W., 1964. Rock compressive strength. Colliery Eng., 41: 287-292.
HOEK, E. and BIE~WSKt, Z. T., 1966. Fracture propagation mechanism in hard rock. Proc.
Congr. Intern. Soc. Rock Mech., 1st, Lisbon, pp.243-246.
HORrNO, F. G., 1968. Effects of planes of weakness on uniaxial compressive strength of model
mine pillars. U.S. Bur. Mines Rept. Invest., 7155:24 pp.
HOSKINS, J. R. and HORINO, F. G., 1968. Effect of end conditions on determining compressive
strength of rock samples. U.S. Bur. Mines Rept. Invest., 7171:22 pp.
HOSKINS, J. R. and HORINO, F. G., 1969. Influence of spherical head size and specimen diameters
on the uniaxial compressive strength of rocks. U.S. Bur. Mines Rept. Invest., 7234:16 pp.
Hsu, T. C., 1967. A study of the compression test for ductile materials. Am. Soc. Mech. Engrs.,
67-WA/MET-11, 16 pp.
HUDSON, J. A. and FAImtURST, C., 1969. Tensile strength, Weibull's theory and a general
statistical approach to rock failure. Proc. Intern. Conf. Struct., Solid Mech. Eng. Design
Civil Eng. Mater., Univ. Southampton, Southampton, in press.
Huoi-ms, B. P. and CHAPMAN,G. P., 1965. Direct tensile test for concrete using modem adhesives.
Bulletin Reunion Intern. Lab. Essais Rech. Mater. Constr., 26: 77-80.
HUGHES, B. P. and CMAPMAN,G. P., 1966. The complete stress strain curve for concrete in direct
tension. Bulletin Reunion Intern. Lab. Essais Rech. Mater. Constr., 30: 95-97.
JAEaER, J. C., 1967. Brittle fracture of rocks. Proc. Syrup. Rock Mech., 8th, Univ. Minnesota,
Minneapolis, Minn., pp.3-57.
JOHNSON, J. W., 1943. Effect of height of test specimen on compressive strength of concrete.
Am. Soc. Testing Mater., Bull., 120: 19-21.
JLrmKIS, A. R., 1966. Some engineering aspects of Brunswick Shale. Proc. Congr. Intern. Soc.
Rock Mech., 1st, Lisbon, pp.99-102.
KENNEDY, A. J., 1962. Processes of Creep and Fatigue in Metals. Oliver and Boyd, Edinburgh,
480 pp.
KESSLER,D. W., INSLEY,I"I. and SLIGH, W. H., 1940. Physical, mineralogical and durability study
on the building and monumental granites of the United States. Natl. Bur. Standards Res.
Papers, 1320: 161-206.
KIDYBINSKI, A., 1966. Rheological models of Upper Silesian carboniferous rocks. Intern. J.
Rock Mech. Mining Sci., 3(4): 279-306.
KNILL, J. L., FRANKLIN, J. A. and MALONE, A. W., 1968. A study of acoustic emission from
stressed rock. Intern. J. Rock Mech. Mining Sci., 5(1): 87-121.
KNUDSON, F. P., 1959. Dependence of mechanical strength of brittle polycrystalline specimens
on porosity and grain size. J. Am. Ceram. Soc., 42(8): 376-387.
KONONER, R. L., 1966. Energy storage and dissipation properties of rocks from creep test
response. Proc. Congr. Intern. Soc. Rock Mech., 1st, Lisbon, pp.273-275.
KOWALSKI, W. C., 1966. The interdependence between the strength and voids ratio of limestone
and marls in connection with their water saturating and anisotrophy. Proc. Congr. Intern.
Soc. Rock Mech., 1st, Lisbon, pp.143-144.
KROKOSKY, E. i . and HISAK, k., 1968. Strength characteristics of basalt rock in ultra high
vacuum. J. Geophys. Res., 73(6): 2237-2247.
KUMAR, A., 1968. The effect of stress rate and temperature on the strength of basalt and granite.
Geophysics, 33(3): 501-510.
LEBOW, M. J., 1966. Extensometers. Exptl. Mech., 6(6): 21A-27A.
LEEMAN,E. R. and GROBBLAAR,C., 1957. A compressometer for obtaining stress-strain curves of
rock specimens up to fracture. J. Sci. Instr., 34: 279-280.
Low, J. R., 1963. The relation of microstructure to brittle fracture, in relation to microstructure.
Natl. Metal Congr. Exposition, 35th, Cleveland, Ohio, pp.163-179.
LtmDnORG, N., 1967. The strength-size relation of granite. Intern. J. Rock Mech. Mining Sci.,
4(3): 269-272.
McCLINTOCK, F. A. and WALSH, J. B., 1963. Friction on Gritiith cracks in rock under pressure.
Proc. U.S. Congr. Appl. Mech., 4th, Berkeley, Calif., pp.1015-1021.
MELLOR, M. and RAINEY, R., 1968. Tensile strength of rocks at low temperatures--preliminary
data report. U.S. Army Terrest. Sci. Center, Hanover, New Hampshire, Techn. Note,
34 pp., unpublished.
MELLOR,M. and RAIS~Y, R., 1969. Effect of low temperature on compressive strength of rocks.
U.S. Army Cold Reg. Res. Eng. Lab., Hanover, New Hampshire, Techn. Note, unpublished.
MISRA, A. K. and MURP.JSLL,S. A. F., 1965. An experimental study of the effect of temperature
and stress on the creep of rocks. Geophys. J., 9(5): 509-535.
MoGI, K., 1966. Some precise measurements of fracture strength of rocks under uniform com-
pressive stress. Rock Mech. Eng. Geol., 4(1): 41-55.
MOROENSa~RN, N. R. and tMUKAN, A. L. T., 1966. Non-linear deformation of a sandstone.
Proc. Congr. Intern. Soc. Rock Mech., 1st, Lisbon, pp.543-548.
MURRELL,S. A. F., 1958. Discussion of paper by Seldenrath and Gramberg. In: W. H. WALTON
(Editor), Mechanical Properties of Non-metallic Brittle Materials. Butterworths, London,
pp.103-104.
NEWMAN, K. and LACHANCE,L., 1964. The testing of brittle materials under uniform uniaxial
compressive stress. Proc. Am. Soc. Testing Mater., 64: 1044-1067.
OBERT, L. and DtrVALL, W., 1967. Rock Mechanics and the Design of Structures in Rock. Wiley,
New York, N.Y., 650 pp.
OBERT, L., WINDES,S. L. and DUVALL,W. I., 1946. Standardised tests for determining the physical
properties of mine rocks. U.S. Bur. Mines Rept. lnvest., 3891:67 pp.
PARSONS, R. C. and HEALEY,D. G. F., 1966. The analysis of the viscous property of rocks for
classification. Intern. J. Rock Mech. Mining Sci., 3(4): 325-335.
PAULDrNG, B. W., 1966. Techniques used in studying the fracture mechanics of rock. In: Testing
Techniques for Rock Mechanics--Am. Soc. Testing Mater., Spec. Tech. PubL, 402, pp.73-84.
I~mCE, F. T., 1926. Tensile tests for cotton yams, 5. "The weakest link" theorems on the strength
of long and of composite specimens. J. Textile Inst., 17: 355.
PII~Tr, G., 1944. Application of the Fourier method to the solution of certain boundary
problems in the theory of elasticity. J. AppL Mech., 2: 176--189.
PRA~n3TL, L., 1923. Anwendungsbeispiele zu einen Henckyschen Satz des plastische Gleich-
gewicht. Z. Angew. Math. Mech., 3: 401--406.
PaOTODYAKONOV, M. M., 1962. Methods of studying the strength of rocks used in the U.S.S.R.
In: G. R. CLARg (Editor), Mining Research. Pergamon, New York, N.Y., pp.649-668.
PUGH, S. F., 1967. The fracture of brittle materials. Brit. J. AppL Phys., 18: 129-161.
REDDY, K. J. M., 1966. A Study of the Elastic Behaviour of Rocks under Uniaxial Stress. Thesis,
Univ. Sheffield, England.
Rocr~, M., 1965. In-situ strain and stress measurements. In: O. C. Zmrcrmwicz and G. S. HOLIS-
TER (Editors), Stress Analysis. Wiley, New York, N.Y., pp.425--461.
Rutz, M. D., 1966. Some technological characteristics of twenty-six Brazilian rock types. Proc.
Congr. Intern. Soc. Rock Mech., 1st, Lisbon, pp.l15-119.
SCHILLER,K. K., 1958. Porosity and strength of brittle solids (with particular reference to gypsum).
In: W. H. WALTON (Editor), Mechanical Properties of Non-metallic Brittle Materials.
Butterworths, London, pp.35-49.
Se8OLZ, C. H., 1968a. Microfracturing and the inelastic deformation of rock in compression.
J. Geophys. Res., 73(4): 1417-1432.
SCHOLZ, C. H., 1968b. Experimental study of the fracturing process in brittle rock. J. Geophys.
Res., 73(4): 1447-1454.
SCnOLZ, C. H., 1968c. Mechanism of creeP in brittle rock. J. Geophys. Res., 73(10): 3295-3302.
SELDENRATH,TH. R. and GRAMBERG,J., 1958. Stress-strain relations on breakage of rocks.
In: W. H. WALTON (Editor), Mechanical Properties of Non-metallic Brittle Materials.
Butterworths, London, pp.79-102.
SERDENGECTI,S. and BOOZER,G, D., 1961. The effects of strain rate and temperature on the
behaviour of rocks subjected to triaxial compression. Proc. Syrup. Rock Mech., 4th,
Penn. State Univ., Pa., pp.83-97.
SI~VALDASON,O. T., 1964. The influence of the testing machine on the compressive strength of
concrete. Proc. Symp. Concrete Control, 1st, Imperial College, London, pp.62-171.
SIRmYS, P. M., 1966. Porosit6, degr6 de saturation et lois de comportement des roches. Proc.
Congr. Intern. Soc. Rock Mech., 1st, Lisbon, pp.471-475.
SKINNER, W. J., 1959. Experiments on the compressive strength of anhydrite. Engineer, 207:
255-259; 288-292.
SPETLA, Z. and KADI.,I/CEK,V., 1967. How size and shape of specimens affect the direct tensile
strength of concrete. Tech. Dig., 12: 865-872.
STAPLEDON, D. H., 1968. Classification of rock substances-discussion. Intern. d. Rock Mech.
Mining Sci., 5(4): 371-373.
STOWE, B. L. and AINSWORTn, D. L., 1968. Effect of rate of loading on strength and Young's
modulus of elasticity of rock. Proc. Symp. Rock Mech., lOth, Austin, Tex., in press.
STREET, N. and WANG, F. D., 1966. Surface potentials and rock strength. Proc. Congr. Intern.
Soc. Rock Mech., 1st, Lisbon, 1: 451-456.
TARRArcr, A. G., 1954a. Measurement of friction at very low speeds. Engineer, 198(2): 262-263.
TAartA~rr, A. G., 1954b. Frictional difficulty in concrete testing. Engineer, 198(2): 801-802.
THAULOW, S., 1962. Apparent compressive strength of concrete as affected by height of test
specimen and friction between the loading surfaces. Bulletin Reunion Intern. Lab. Essais
Rech. Mater. Constr., 17: 31-33.
TIMOSrmNKO, S. and GOoDmg, J. N., 1951. Theory of Elasticity. McGraw-Hill, New York, N.Y.,
372 pp.
WAI.SH, J. B., 1965a. The effect of cracks on the compressibility of rock. J. Geophys. Res., 70(2):
381-389.
WAts~, J. B., 1965b. The effect of cracks on the uniaxial elastic compression of rocks. J. Geophys.
Res., 70(2): 399-411.
WALSH, J. B., 1965c. The effect of cracks in rocks on Poisson's ratio, d. Geophys. Res., 70:
5249-5257.
WALSH, J. B. and BgAC~, W. F., 1966a. Cracks and pores in rock. Proc. Congr. Intern. Soc. Rock
Mech., 1st, Lisbon, 1: 643--646.
WALSH, J. B. and BRACE, W. F., 1966b. Elasticity of rock: a review of some recent theoretical
studies. Rock Mech. Eng. GeoL, 4(4): 283-297.
WAWERSm, W. R., 1968. Detailed Analysis of Rock Failure in Laboratory Compression Tests.
Thesis, Univ. of Minnesota.
WEmULL, W., 1939. A statistical theory of the strength of materials. Proc. Roy. Swed. Inst.
Eng., 151: 1-45.
WEmULL, W., 1951. A statistical distribution function of wide applicability. J. Appl. Mech., 18:
293-297.
WEmULL, W., 1952. A survey of statistical effects in the field of material failure. Appl. Mech. Rev.,
5(11): 449-451.
WRmnT, P. J. F., 1957. Compression testing machines for concrete. Engineer, 203: 639-641.
WtmRKER, R. G., 1955. Measuring the tensile strength of rocks. Trans. A.LM.E., 202: 157.
APPENDIX1
With reference to Fig.33 the frictional force on the interface between the rock and the
platen must equal the average shearing force:
D20 Po
rx - /7 (61)
8 2~
To obtain the shearing forces mathematically is complicated, but the problem can be
approached another way by assuming that the average shearing force can be equated to a hydro-
static pressure acting around the periphery of the specimen. Considering a segment of the
interface (Fig.33B):
D20 D
= Ox (62)
~,x ---if--- P 2
Platen ~ TrK
0 0.20"~ 0.4 o'x
//I'/////// // / / / / / / / A n I J _IN_ _ i
I
Rock l 0.2
J ~ I:=
Fig.33. Effect of radial restraint by platen friction in the uniaxial compression test.
where x is the distance along the specimen over which the pressure p must be applied to prevent
movement at the interface:
nDp
= x (63)
P
The lateral strain of the platen under an applied load P is given by:
vv 4P Vp
p -- _ _ -- _ _ O" x (64)
Ep ~D 2 Ep
(subscript p - platen).
For no sfippage to occur, this must equal the lateral strain of the rock:
8~ = ep (66)
(subscript r - rock).
The hydrostatic pressure to prevent differential platen/rock movement can now be deter-
mined:
vp 1
~x = {p - vr (p + ~x)}
i.e.:
{ vpE~_ + vr}
Ep
p= o"x
1 - vr
4x/ vp E r + Vr }
/7 = [ Ep (67)
o ( 1 - v3
The problem now is to determine x in terms of the specimen diameter D.
Fig.33C shows the variation of shear stress from the interface into the specimen according
to Balla's results (Table VI). Shear stress falls off very rapidly from the interface and varies
across the width. If it is assumed that an average value could be represented by the line NM, then
it may also be assumed that the maximum depth of influence is to point M, where M = 0.25 D.
Since shear stress is assumed to decrease linearly, the value of x is given by:
M D
2 8
Substituting for this value in eq.67"
,68>
2 0 - v,)
For all but the strongest rocks, the term vpEr/Ep is very small and can be ignored, i.e.
the platen can be considered rigid. Eq.68 simplifies to:
-- Vr
7
2 (1 - v~)
Assuming a Poisson's ratio of 0.3 for the rock, the minimum coefficient of friction which
will just prevent sliding is approximately 0.2.
APPENDIX2
Since little relevant data could be found in the literature, measurements were made in
order to determine approximate magnitudes of inteffacial friction between rock and platen
under various conditions.
A hollow cylinder of Darley Dale Sandstone, 6 inch O.D. 3 inch I.D. 4 inches
long was cemented into a machined pipe flange and then fac.~l-off with a diamond tool in a lathe.
The flange was bolted to the lower platform of a testing machine, and a cylindrical steel block,
6 inch diameter 4 inch long, was laid concentrically on top of the rock cylinder, giving an
annular area of contact. Axial load was applied to the steel block by the testing machine through
a 0.75 inch diameter steel ball set in greased seats. Torque was applied to the steel cylinder by a
spring balance attached to the end of a 36 inch long lever arm. Test components are shown in
Fig.34. Axial load was applied to a predetermined level, a force of 60 lbf was applied at the end
of the lever arm, and the axial load was then slowly relaxed until rotational slip between the rock
and steel occurred. Actual contact area was found by examining imprints and scratch patterns.
The resuRant frictional force was assumed to act at the "equal area" radius.
An additional test was made to find the effect of friction at the loading ball by applying
torque to the steel cylinder while it was held between two identical steel balls. This ball friction
proved to be negligibly small at the highest axial loads used in the tests. (Contact pressures ranged
from 150 to 700 lbf/sq, inch).
Test results for various contact conditions are given in Table IX.
TABLE IX
APPENDIX3
The purpose of a ball seating is to ensure intimate contact between the testing machine
crossheads and the ends of the test specimen, so that uniform strain is induced in the specimen
as the crossheads move together. Provided that sufficient care is taken with the initial alignments,
and the seating seizes early in the test, almost any design of seating can be used for tmiaxial
compression testing (HosKrNs and HoRrso, 1968, 1969). There are, however, certain basic design
features for ball seatings which can be investigated in relation to the possible errors induced by
seatings.
I
!
I
I
I
I
I
I
I
I I
--7,
l ,
A [3
Fig.35. Ball seat geometry.
Prior to load application an independent ball seating is only stable when the ball component
is set above the cup, as illustrated in Fig.35A. For ball seatings built into the testing machine
crosshead the position is reversed but the same conditions apply, as in both cases it is the ball
column which rotates and the cup which is fixed.
Initially, as the testing machine crosshead is brought to bear it will contact the bail
column at a point on its edge N (Fig.35A). Further movement of the crosshead will rotate the
ball in its seating, and in doing so will either displace the point N a distance AL or displace the
cup platen, with or without the test specimen, an equivalent amount. When the ball column is in
close contact with the crosshead, further rotation of the ball seat (produced, for example, by
tilting the crosshead as it is picked up) will tend to cause lateral displacement of the specimen.
In a correctly designed ball seating these lateral movements will be at a minimum. Referring to
Fig.35B, as the ball column rotates in its seating the locus of a general point on the periphery
of the column An is a sphere with centre O and radius O/In where O is the centre of curvature of
the seat surfaces. A planar displacement through 0 moves An to An', necessitating a horizontal
movement AL.
From geometrical considerations:
where 0 is the angle between the plane of the crosshead and the plane of the bail column; h is
AoAn (Fig.35B); ro is the radius of the ball column OAo (not the ball seat).
It is interesting to note that the radius of the ball seating has no direct effect on the
magnitude of the lateral translation of the ball column.
F r o m eq.69, therefore, the minimum lateral translation for a fixed ball column radius
occurs when h = 0, i.e., when the upper surface in contact with the machine crosshead passes
through the centre of curvature of the ball seating. Another equally important consideration is the
force available to rotate the ball column on its seating.
Referring to Fig.35B, balancing the moments in the system gives:
where P is the applied vertical force and F i s the frictional force to be overcome in the ball seating.
Again from eq.70 it will be noted that for a given value of ro the moment acting to overcome
the friction in the bail seat is at a maximum when h = 0. In this connection it must also be
remembered that the higher the value of P, the greater will be the frictional forces between the
crosshead and the edge of the ball column. The rim of the ball column should be rounded to
reduce contact friction.
In principle, the ball seat may be placed above or below the sample. If the seat is placed
above the sample the height of the ball column (BO, Fig.35B) should equal the radius of curvature
of the seat (R, Fig.35B). If the seat is placed beneath the sample, the height of the ball column
plus the height of the sample should equal R.
The considerations which determine the radius of the ball seating R and the radius of the
ball column ro are not so precisely defined.
To keep lateral displacements to a minimum it can be seen from eq,69 that ro should be
as small as possible. However, if ro is reduced much below the radius of the test sample there is a
danger of platen rotation about the contact region between the bail column and the crosshead.
For this reason it is recommended that the ball column radius ro should be approximately equal
to that of the specimen being tested.
The radius of the ball seating R must be determined from frictional considerations. To
determine the friction developed in lightly lubricated spherical seatings of different radii, two
experiments were made. Fig.36 illustrates the equipment. In the first set of tests a 2.5 inch radius
ball was squeezed between a pair of cups, and in the second a plate with concave seats on both
faces was fitted with two 0.5 inch radius bails. In each set of tests the contact area subtended an
arc of approximately 50 at the centre of curvature.
Two test procedures were followed for each case. In the first, axial load was increased
; I ' I ' I
52
linch Boll
24
Unloading
from
60,000 tbl
LOAD
x I 0 0 0 Ibf Loading
16
5 inch Boll
8
Unloading
from ~ A . f _ _ ~ a" ~ -
60,000 Ibs ,~...-~o "~'- -
~~l'~.'~'~Loadi ng
V'I I I I I I i
0 400 800 1200 600
TORQUE, Ib in
Fig.37. Results of friction tests on spherical seats lubricated with light mineral oil.
incrementally and the torque necessary to move the seat was measured; in the second, a high
axial load was applied, a fixed torque was exerted, and the axial load was slowly reduced until the
seat was released. The resuRs obtained are given in Fig.37. It will be noted that torque and load
are proportional and that the frictional resistance to rotation increases with the radius of the
seating.
Using these results it is possible to examine the probability of a lubricated ball seat
unlocking during the course of a test as a result of unsymmetrical strain in the specimen. The
seat will unlock when the line of action of the resultant sample reaction P moves off centre from
the sample axis to a radius r~ where:
T
rl --
P
in which T is the torque to unlock the seat under load P. From the results shown in Fig.37 it
appears that a 0.5 inch radius spherical seat will unlock when the resultant reaction moves off
centre by a distance of about 0.05 inch, and with a 2.5 inch radius seat by a distance of about
0.2 inches.
These results would undoubtedly be altered significantly by changing the lubricant.
TARRANT (1954a, b) found that a spherical seat had an effective cocffcient of friction of 0.6 when
lubricated with mineral oil, 0.15 when lubricated with graphite grease, and as low as 0.04 when
lubricated with a grease containing free fatty acids with polar molecules. HOSKrNS and HORINO
(1968) found that for a wide range of ball sizes with different lubricants and surface finishes there
was no evidence in the test results of ball scat movements during testing.
Since the ideal displacement boundary conditions stipulate that there should be neither
rotation nor lateral displacement of the platen during a test, it seems reasonable that ball seats
should be designed so as to always lock under load, irrespective of the distribution of forces
within the test specimens. For this reason the ball seat diameter R should be as large as possible.
However, weight and handling considerations play a part, particularly where centring is con-
cerned, and the recommendation of HOSKINS and HORINO (1969) that the radius of the ball
seating be around twice that of the specimen radius seems very reasonable.
The cup component of the ball seat must be sufficiently thick to prevent any flexural
distortion under load. It is suggested that the minimum thickness at the base of the cup should be
1.25 to.
It is important to accurately centre a test specimen in relation to the centre of curvature
of the ball scat. Any error in centring automatically shifts the line of action of the resultant
force in the specimen by an equivalent distance and gives rise to bending forces, as discussed in
the section on eccentric loading.
To sum up the foregoing, it is recommended that:
R ~ 2R~
r o ~ Rs
Rc ~ 1.25 R s
h =R,,~2Rs
t /> 0 . 5 R s
where Rs is specimen radius, R is radius of the spherical surfaces, re is radius of the ball column,
Re is the plan (or platen) radius of the cup element which bears onto the specimen, h is the height
of the ball column and t is the minimum thickness of the cup element, i.e., the distance from the
bottom of the cup to the platen face.
APPENDIX4
Up to the present time there has been no uniform system of units for physical quantities
in rock mechanics. Some investigators have used the traditional metric system, in various forms,
while others have worked with English units. This situation is likely to change soon.
In 1960 a refined and extended form of the metric system (Syst~me International d'Unit6s)
was approved by an international body, the Conf6rence G6n6rale des Poids et Mesures. This
system, generally known as "SI Units" has now been formally adopted by many countries, which
are working towards its universal application as the sole legal system. Scientific and technical
journals are now requiring conformance to SI, and new educational texts are using the system.
The SI system must obviously become standard in rock mechanics over the long term,
but universal acceptance is likely to be slow for, quite apart from innate conservatism or logical
objection on the part of individuals, measurements in many parts of the world must continue to be
made with equipment graduated in English units. There is also the undeniable fact that for many
older men, work in the traditional system of units is more efficient, since deep familiarity with
dimensions and physical constants in that system makes easier the recognition of errors and
inconsistencies. Thus it seems reasonable to aim for a gradual transition to SI, in the early
stages of which instrument readings are recorded as read, together with an appropriate conversion
factor, and published data are given in traditional units followed by SI equivalents in parentheses.
This is in keeping with the 1968 recommendations of the Royal Society Conference of Editors.
In this review paper it has seemed inappropriate to give parenthetic SI equivalents after
each numerical value. Many of the dimensions given are rounded to one or two significant
figures in English units, and worked numerical examples become unduly cluttered when SI
equivalents are quoted throughout. Thus the text has been left in English units and the following
conversions have been appended.