Vous êtes sur la page 1sur 15

Indraratna, B., Wijewardena, L. S. S. & Balasubramaniam, A. S. (1993) GLotechnique43, No.

1, 37-51

Large-scale triaxial testing of greywacke rockfill


This Paper describes the results of a series of Cet article d&it les rbultats d’une suite d’essais
large-scale triaxial tests conducted on grey- au triaxial zi grande &helle rCli&s sur du remblai
wacke rockfill, used in dam construction in rocheux (greywacke) &iii& pour la construction de
Southern Thailand. The tests are conducted at barrages dans le sud de la Thailande. Ces essais
low to moderate confining stresses to relate sont ex&cuti?s P des pressions de cellule faibles $
their findings to the stability of rockfill dams.
mod&es pour rapprocher les rbultats de la stabili-
Considering the current test results in conjunc-
tion with previous laboratory data, revised ti! de barrages P enrochement. En considitrant les
failure criteria for rockHI are proposed in non- rbultats existant d’essais, en conjonction avec les
dimensional form. For both low and hieh con- don&es de laboratoire pr&&dentes, des crit&es de
fining stresses, lower and upper bounds of rupture de l’enrochement sont proposb dans une
strength envelopes have been established, based forme sans dimensions. Pour des pressions de
on a wide array of granular materials. The cellule faibles et ClevCes, les enveloppes basses et
influence of the coofining stress on the shear hautes de r&stance ont Bt6 btablies, ba&s sur un
strength of rockfill is studied in depth, and the vaste ensemble de mati?riaux granuleux.
implications of a non-linear envelope at low
L’inflwnce de la pression de cellule sur la r(?sis-
normal stress levels on the stability of rockfill
dams are discussed. Although two parallel rock- tance au cisaillement est Ctudi&e en profondeur et
fill gradations for specimens compacted to les con&quences d’une enveloppe de rupture non
similar porosities are considered, the exact role linkaire d containtes normales faibles sur la stabili-
of particle size effect on shear strength is not ti! de barrages P enrochement sont disc&es. Bien
examined in detail, as the difference in que deux gradations parallkles d’enrochement pour
maximum particle sizes tested in this study is des 6chantillons compact& i des porositb sem-
not sufficiently large. blables aient Cti! consid&es, le r61e exact de I’effet
de la taille des grains sur la r&stance au cis-
aillement n‘est pas examine en d&tails car la diff&
KEYWORDS: dams; deformation; failure; laboratory rence entre les tailles maxis des grains test& pour
tests; shear strength. cette etude n’&ait pas suffisante.

INTRODUCTION nant mineral, and feldspar and mica are the other
Greywacke rockfill has been used in the construc- main constituents. The mean uniaxial compres-
tion of several dams in Thailand, including the sive strength determined using eight NX core
recently completed Chiew Larn Dam. The Elec- samples was found to be of the order of 13.5 MPa
tricity Generating Authority of Thailand has pro- for saturated specimens.
posed the construction of similar dams in several In the past, emphasis has been placed on tri-
parts of Thailand in the future. Detailed testing of axial tests conducted at high pressures, where
greywacke rockfill has not been carried out in the confining stress levels as high as 2.5-4.5 MPa
past: proper understanding of the shear strength have been applied (Marsal, 1973; Marachi, Chan
properties of this material should enable better & Seed, 1972). In most realistic situations, the
design of large rockfill dams in the future. Grey- shear strength of rockfill must be related to much
wacke found in Southern Thailand is dark grey to lower confining stresses. Charles & Watts (1980)
black in colour, and can be classified as a tough have reported that the maximum possible normal
sedimentary rock, formed under unstable geo- stress on a critical failure surface of a 50 m high
logical environments. It is generally found rockfill dam is unlikely to exceed 400 kPa. Even
without internal stratification or parting. A study in the case of the highest rockfill dams in South-
of thin sections indicated that quartz is its domi- East Asia, the maximum normal stresses are
unlikely to exceed 1 MPa (Lee, 1986). Therefore,
the scope of the current investigation has been to
Discussion on this Paper closes 1 July 1993; for further
details see p. ii.
study the strength and deformation behaviour of
* University of Wollongong, Australia. large-scale greywacke rockfill specimens at low to
t Asian Institute of Technology, Bangkok. medium confining pressures, and to relate these


strain behaviour of rockfill, specimens of two laboratory rockfill, including the uniformity coef-
rockfill gradations with maximum particle sizes of ficient C, and the coefficient of curvature C,,
38 mm and 25 mm were tested. Taking account of before testing. The sample size ratio is defined by
the principles of similitude (Fumagalli, 1969; the diameter of the triaxial specimen (300 mm)
Lowe, 1964), it may be postulated that realistic divided by the mean diameter of the maximum
results can be obtained only if the gradations at particle size. The effect of size ratio on the behav-
least show iour of rockfill specimens in triaxial testing has
been discussed in depth by several investigators.
(a) similar grain size distribution curves
Fagnoul & Bonnechere (1969) and Nitchiporo-
(b) comparable field and laboratory compaction
vitch (1969) proposed a minimum sample size
(c) similar angularity of particles (conforming
ratio of 5. Marachi (1969) concluded that a size
ratio of at least 6 must be employed in order to
(d) similar sample size ratios.
minimize size effects for rockfill specimens with
With regard to condition (a), the grain size curve less than 30% of particles in the maximum sieve
of the laboratory rockfill must be parallel to that size range. In this study, size ratios of 8 and 12
of the prototype; to satisfy condition (b) the poro- were adopted for rocklill gradations A and B
sity of the laboratory specimens must be close to respectively.
that of the compacted rockfill in the field. Condi- The test specimens were compacted within the
tion (c) is difficult to fulfil exactly, but the labor- protective membrane in several layers, each
atory particles must be similar in shape to the 50-60 mm thick, using a hand vibrator. For both
actual rockfill, although smaller in size. Fig. 2 gradations, an initial compacted porosity of the
compares the parallel gradations A and B used in order of 30% was achieved. The initial water
the testing programme with the grain size dis- content of the rockfill was -5%; after compac-
tribution of rockfill used in the Chiew Larn Dam. tion the mean dry density of the specimens was
The difference between the gradation curves determined to be - 18.5 kN/m3. During the con-
before and after testing (at u3’ = 600 kPa) indi- struction of the Chiew Larn Dam, each compacted
cates the extent of particle breakage during shear- rockfill layer varied in thickness from -0.6 to
ing. A reduction in d,, of at least 30% is observed 1.0 m in the field, producing a dry density of
for both gradations A and B at the maximum greater than 18.0 kN/m3, achieved by heavy
confining pressure of 600 kPa. At small cell pres- vibratory rollers (10 t), with a compacted porosity
sures (less than 200 kPa), the degree of grain similar to that obtained in the laboratory.
crushing is insignificant. Terzaghi (1960) noted that dam settlements
Table 1 summarizes the characteristics of the after impounding can be significant for soft rock-

US standard sieve number

200 100 52 25 14 7 4 316in 1/2in 1 m 1.5in 3in 20in

100 , I I I I I I I I ,’ /I-I
- Before testing /‘,I’
/’ 1’
---- AHer testing

E 1’ ’
o” 60
,I /,I/

E ’ /

$ , ,;~~~~~~

0.1 1 10 100
Diameter: mm

Fig. 2. Particle size distribution curves of greywacke rocktill


Table 1. Grain size characteristics of greywacke rockfill: C,, and C,

are determined from the grain size distribution carve

Gradation d,,, : mm d,,: mm C, C, Size ratio

A 38.1 4.9 6.0 0.9 8
B 25.4 3.6 6.0 0.9 12

hll (e.g. schist), whereas for harder granites such 10%) during placing of greywacke rockfill is sufh-
settlements are generally small. He related this cient to minimize settlements due to wetting.
phenomenon to the reduction in strength upon Nevertheless, for soft rockfill materials such as
saturation. In the field, wetting occurs after reef limestones, the settlements during saturation
impounding, and the resulting settlements depend can be significant (Brown, 1988).
on the initial water content of the rockfill. It may
be anticipated that the greater the placement
water content is, the smaller will be the settle- EXPERIMENTAL RESULTS AND DISCUSSION
ments after impounding. In the construction of Stress-strain behaviour
the Chiew Larn Dam, although greywacke rock- The results of the isotropically consolidated
fill was placed relatively dry with a mean mois- drained triaxial compression tests are shown in
ture content <lo%, significant settlements have Figs 3 and 4 for gradations A and B. As expected,
not been reported after impounding. In the labor- the peak deviator stress increases with increasing
atory, before loading, the initially dry rockfill applied confining stress. The gradual decrease in
specimens were saturated by applying a back deviator stress after peak indicates the ductile
pressure of 100 kPa. No significant settlements strain-softening response of rockfill. Similar
were recorded during this saturation phase. This behaviour has been reported for heavily fractured
implies that a relatively small water content (5- rock (Ito, 1983; Indraratna & Kaiser, 1990). For

2500 250C
A 03’ = 100 kPa A 03’ = 100 kPa
0 03’ = 200 kPa 0 us’ = 200 kPa
0 03’ = 300 kPa Gradation A 0 a3’ = 300 kPa Gradation Et
0 03’ = 400 kPa l ox’ = 400 kPa
2000 2000



6 6
0 5 10 15 20 0 5 10 15 20

Axial strain: % Axial strain: %

Fig. 3. Stress-strain behaviour of greywacke rockfill Fig. 4. Stress-strain behaviour of greywacke rockfill
(gradation A) (gradation B)

the rocktill samples subjected to very small con- laboratory. Therefore, the relative densities of the
fining pressures (~200 kPa) dilation is pro- compacted laboratory specimens must reflect the
nounced, whereas at higher confining pressures compaction requirements in the field.
(a,’ > 300 kPa), dilation is suppressed even at The volumetric strains measured at the peak
axial strains exceeding 15%. At low confining deviator stress not only increase almost linearly
stresses, gradation A shows more dilation than with the confining pressure, but are also indepen-
gradation B; this can be attributed to its larger dent of the particle gradation (Fig. 6). Lee (1986)
particle sizes. However, at higher rr3’, only a slight has made a similar observation, even for a greater
difference in the volumetric strain response is particle size difference. Data reported by Marachi
observed between the two gradations, suggesting et al. (1972) for other types of rockfill indicate
that the particle size effect may be of secondary that the volumetric strains at failure increase with
importance as compared with the influence of increasing a3’ at a gradually decreasing rate,
confining pressure on the volumetric strains. approaching a constant value at high c3’ (2.5
However, it is important to note that quantifica- MPa). While the results obtained in this study for
tion of the role of particle size would require much lower confining stresses (100-600 kPa) are
further testing, with a much greater difference in not strictly in conflict with Marachi et al. (1972)
grain size. they cannot be extrapolated to predict behaviour
The stress-strain response of gradations A and at very high stress levels. The axial and radial
B shows that the initial tangent stiffness increases strains at (ui’ - 03’)_ increase at a diminishing
with increasing confining pressure. The actual rate as 6s’ is increased. It is interesting to note
magnitudes of the initial deformation modulus as that the radial strains at (a,’ - Q~‘),_ approach a
a function of the confining stress are shown in constant value (6%) at confining pressures
Fig. 5 (each data point represents the mean of beyond 300 kPa. On the basis of the triaxial
three tests). The difference between the initial strain response discussed above, it may be con-
moduli of any two corresponding samples at the cluded that the confining stress (r3’ is the domi-
same confining pressure is relatively small, and nant factor controlling the deformation
diminishes with increasing oj’. Furthermore, behaviour of rockfill, while the particle size effect
comparison of Figs 3 and 4 shows that the peak seems to be of secondary importance, unless the
deviator stresses of the corresponding samples are difference between the maximum particle sizes is
almost the same. These results suggest that for considerable.
parallel gradations with adequate sample size
ratios, the influence of particle size may diminish
if the initial porosities of the test specimens are Shear strength andfailure criteria
similar. If the compacted field porosity is signifi- Well-documented studies have indicated that
cantly different from that of the test specimens, the friction angle of sands decreases with increas-
the actual deformation response cannot be simu- ing cell pressure in drained triaxial tests (Vesic &
lated, even if parallel gradations are used in the Clough, 1968; Bishop, 1966; Ponce & Bell, 1971).
A comprehensive series of large-scale triaxial tests
(1 m dia.) conducted on many types of rockfill
was described by Marsal (1967) and Marachi

c (1969). They found that the principal stress ratio

for drained tests (~i’/o~‘)r was considerably
increased at low confining stresses. Charles &
Watts (1980) verified this phenomenon with
smaller rockfill specimens (0.23 m dia.) in tests
conducted on sandstone, basalt and slate. The
experimental data are summarized in Fig. 7,
which also shows results obtained for the grey-
wacke rocklill samples (0.3 m dia.) for compari-
Although these specimens are weaker than a
number of other materials reported in the liter-
ature, greywacke is accepted in Thailand, as a
satisfactory rocktill. In contrast, the soft, low-
grade slate was considered an unsuitable rockfill
200 400 600
by Charles & Watts (1980). The relatively low
Effective confining stress: kPa
principal stress ratio of the greywacke specimens
Fig. 5. Variation of initial deformation modulus with may be attributed to their lower initial density
effective cordiniog stress and the relatively low uniaxial compressive

25 r o--O Axial strain (aradation A)

O----II Volumetric $Gain (gradation A)

M Radial strain (aradation AI
U Axial strain (gradation B) ’
H Volumetric strain (gradation B)
g *O * Radial strain (qradation B)



u) h
5 10 0
I I 1 I I I I
> 0 100 200 300 400 500 600 700
Effective confining stress: kPa

Fig. 6. Jnfluence of effective confining stress on sample strains at peak

deviator stress

Charles 8 Watts (1960)

z 5-
: 4-
g 3-
Low grade slate
(Charles 8 Watts, 1980)
2’ I I I I I I
10 30 100 300 1000 3000 10 000
Conftning stress 03’: kPa

Fig. 7. Variation of effective principal stress ratio at failure with effective confin-
ing stress for various rocktills

strength of greywacke as compared with highly The significance of the uniaxial compressive
compacted, harder rock fragments. The degree of strength cr, as a normalizing parameter was rec-
compaction and hence the initial porosity can ognized by Hoek & Brown (1980) for rock
have a major effect on the shear strength. At a samples tested in triaxial compression, and
confining stress of 200 kPa, an increase in poro- allowed evaluation of the characteristic constants
sity of 1% can reduce the angle of friction by at appropriate for both intact and fractured rocks.
least 0.5” (Marachi, 1969). Rockfill can be regarded as intensely fractured
The strength envelopes for the greywacke gra- rock, and further breakage of individual frag-
dations A and B considered in this study are ments during shearing is a function of particle
shown in Fig. 8(a) and (b). Both envelopes corre- angularity, confining pressure and the point load
sponding to the low confining stress region index related to aE of the parent rock.
clearly reveal non-linearity, and pass through the Marsal (1973) introduced the particle breakage
origin, indicating zero cohesion as expected of a index B8 to characterize rockfill behaviour by
granular material. De Mello (1977) proposed comparison of the grain size distribution curves
a non-linear failure criterion for typical rockfill before and after testing. However, the magnitude
of B, is sensitive to the applied confining stress,
r, = Uff”b (I) irrespective of rock type. Consequently the value
where the constants a and b are considered to be of B, cannot be interpreted as a material pro-
characteristic parameters obtained by curve perty. In this study, the uniaxial compressive
fitting. The physical significance of these con- strength has been incorporated in defining the
stants is not clear, because the value of a not only failure envelope of rockfill. In reality, cc can be
depends on the system of units used, but its estimated more reliably than B, for any rockfill
dimensions vary according to the value of b. In by conducting basic index tests on even an irregu-
this respect it seems that the introduction of a lar lump. It is proposed that the failure envelope
non-dimensional failure criterion is probably in a non-dimensional form can be expressed as a
useful, as a wide array of materials can then be modification of the de Mello (1977) criterion by
compared directly within the framework of simili-
tude (Indraratna, 1990). (2)

0 400 600 1200 1600 2OaJ 2400 2600
Normal stress: kPa


Fig. 8. Mobr-Coolomb failure envelopes of greywacke rockfill for: (a) gradation A;

(b) gradation B

The constants a and b are dimensionless, hence fill, including to some extent the effect of dilation
independent of the system of units used for the and particle sizes.
stresses. For linear MohrCoulomb materials, b Figure 9 shows the variation of shear strength
approaches unity and the magnitude of a is given with the normal stress for several types of rockfill
by the ratio of shear strength to the normal stress. tested at different confining stress levels. For com-
In effect, the parameter a encompasses the equiv- parison, the present results for greywacke rockfill
alent friction angle and can be regarded as an are plotted in Fig. 9(a) together with those dis-
intrinsic shear strength index. The magnitude of b cussed by Charles & Watts (1980) for other types
dictates the non-linearity of the failure envelope, of rockfill tested at low normal stresses. The
particularly at low confining stresses, and thereby failure envelopes for the two greywacke grada-
represents the deformation response of the rock- tions (A and B) coincide except at the lowermost

Greywacke (gradation A) (present study) Fig. 9. Variation of shear strength with normal streSS for
Greywacke (gradation B) (present study) various rockfill types: (a) low to medium stress ranges;
El Granero slate (Marsal, 1973) (b) medium to high stress ranges; (c) high to very high
stress ranges
Sandstone (Charles 8 Watts, 1980)
Slate (high grade) (Charles 8 Watts, 1980)

Slate (low grade) (Charles B Watts, 1980)


“0 400 800 1200

Effective normal stress: kPa


0 Malpaso conglomerate

6 San Francisco basalt B Crushed basalt

H Mica granitic gneiss v Argillite (Pyramid Dam)

A El lnfiemillo diorite V Amphibolite (Oroville Dam)

Data from Marsal (1973) Data from Marachi et a/. (1972)

I I 1 1

1 2 3 4 2 4 6 8

Effective normal stress: MPa Effective normal stress: MPa

(b) (c)

normal stress values, i.e. the effect of particle sizes clarity) with a regression coefficient r2 > 0.95.
is minimized at higher normal stresses, where The associated constants a and b (equation (2))
dilation is inhibited. It is significant that all the are given in Table 2 for all these rockfill
rockfill samples shown in Fig. 9(a) have compara- materials. For this wide range of rockfill, the
ble particle sizes and initial porosities. In this values of a and b are less than 0.6 and 0.9 respec-
stress range, the non-linearity of the failure tively, except for Malpaso conglomerate. For very
envelopes is pronounced, and a clear distinction low confining stresses (< 100 kPa), b tends to be
between the strong (hard) and weak (soft) rockfill less than 0.8; at high confining stress levels (> 1.5
can be observed. Fig. 9(b) and (c) illustrates the MPa) it approaches 0.9. Irrespective of the com-
behaviour of rockfill at higher stress levels, using pressive strength of rock, particle sizes, angu-
data from Marsal (1973) and Marachi et al. (1972) larity, initial porosity and initial water content,
respectively. Note that at elevated effective the above experimental data fall within a narrow
normal stresses, the failure envelopes approach band defined by the boundaries given in Table 3.
linearity, hence the conventional Mohr-Coulomb Alternatively, the failure of rockfill specimens
analysis is sufficient to represent the failure of may be represented by the major and minor prin-
rockfill. cipal stresses at failure crir’ and oSf’: this pro-
Figure 10 plots the normalized shear strength cedure is often adopted in rock mechanics.
against normal stress relationships on log scales. Incorporating the uniaxial compressive strength
Together with the present test results, experimen- cc, the following normalized expression is sug-
tal data are presented for a wide range of effective gested to represent failure
normal stresses from 100 kPa-8 MPa. The cor-
responding data points for each rockhll specimen
can be joined by a straight line (omitted for

Greywacke (gradation A) (present study)

Greywacke (gradation B) (present study)
Malpaso conglomerate (Marsal. 1973)
San Francisco basalt (Marsal. 1973)
Mica granitic gneiss (Marsal. 1973)
El lnfiernillo diorite (Marsal, 1973)
El Granero slate (gradation A) (Marsal, 1973)

Sandstone (Charles & Watts, 1980)

Slate (high grade) (Charles 8 Watts, 1980)
Slate (low grade) (Charles 8 Watts, 1980)
Basalt (Charles & Watts, 1980)
Crushed basalt (Marachi et a/., 1972)
Argillite (Pyramid Dam) (Marachi et al.. 1972)
Amphibolite (Oroville Dam) (Malachi et al., 1972)

‘1 0.001 0.01 0.1


Fig. 10. Normalized shear strengthormal stress relationship for various rockfills
Table 2. Coeficients of proposed failure criteria for various types of rocktill

Type of rockfill 6, : MPa Initial d,,, : mm d,, : mm Sample Range of Coefficient

porosity: o/u size ratio a3’: MPa
a b a

Greywacke (grad. A)* 136 30-32 38 4.9 6 8 0.10-0.60 0.39 0.86 1.61
Greywacke (grad. B)* 136 30-32 25 3.6 6 12 0~10-0%0 0.47 0.89 2.14
Malpaso conglomeratet 75 26-30 200 22 63 5.7 0.04-0.60 0.46 0.86 1.75
0.60-2.50 0.76 0.93 3.68
San Francisco basalt? 175 25-29 80 7 11 5.7 0.47-2.47 0.42 0.84 1.88
Mica granitic-gneisst 125 24-28 200 51 14 5.7 0.45-2.47 0.42 0.89 1.97 0.88 6
El Infiernillo dioritet 105 33-38 200 68 5 5.7 0~09-0~50 0.37 0.84 2.09 0.87
0.83 +
0.50-1.69 0.47 0.90 2.19
El Granero slatet 90 30-35 200 105 10 5.7 0.04-0.46 0.35 0.77 1.06 0.72 >
Sandstone1 120 20 38 5 6 0.03-0.70 0.14 0.68 0.34 0.57 3
Slate (high grade)$ 312 25 38 5 51 6 0.10-0.52 0.29 0.76 0.94 0.72
Slate (low grade)$ 58 32 38 4 6 0.10-0.52 0.24 0.73 0.53 0.66 $
Basalt4 360 25 38 13 13 6 0.03-0.70 0.33 0.77 1.34 0.72 c
Crushed basalt5 175 50 13 7 6 0.20-4.50 0.50 0.87 2.14 0.84 2
Argillite (Pyramid Dam& 40 50 13 7 6 0.20-4.50 0.58 0.88 2.62 0.85
0.86 $
Amphibolite (Oroville DamB 175 50 10 34 6 0.20-4.50 0.61 0.90 2.78 g

* Present study.
t Marsal (1973).
1: Charles & Watts (1980).
$ Marachi et al. (1972).

Table 3. Boundaries of experimental data applicable to Table 4. Upper and lower boundaries of peak (failure)
normal stress ranges r-m

This approach can be regarded as an extension of negligible. The relationships between the major
the empirical method discussed by Hoek & and minor principal stresses at failure for various
Brown (1980), who proposed a square-root rockfill materials normalized by their uniaxial
relationship between the principal stresses for the compressive strength are plotted in Fig. 11. Not-
failure of both intact and jointed rocks. Marsal withstanding the differences in this wide array of
(1973) and Indraratna (1990) have also discussed granulated materials, their peak (failure) response
the relevance of representing failure criteria in can be represented by a narrow band width
terms of principal stresses at failure, with particu- defined by the upper and lower bounds of CIand /I
lar reference to triaxial testing. Equation (3) given in Table 4.
emphasizes that in the absence of any confining The coefficients a and fi determined for individ-
pressure, the strength of the rockfill sample is ual rockfills are given in Table 2. For most rock-

O.! j-
Greywacke (gradation A) (present study)
Greywacke (gradation B) (present study)
Malpaso conglomerate (Marsal. 1973)
San Francisco basalt (Marsal. 1973)
Mica granihc gneiss (Marsal. 1973)
El lnfiemillo diorite (Marsal. 1973)
0. l- El Granero slate (Marsal, 1973)



Sandstone (Charles & Watts, 1980)

Slate (high grade) (Charles 8 Watts, 1980)
Slate (low grade) (Charles & Watts, 1980)
Basalt (Charles & Watts, 1980)
Crushed basalt (Marachi et a/., 1972)
Argillite (Pyramid Dam) (Marachl et al., 1972)
Amphibolite (Oroville Dam) (Marachi et al., 1972)

Fig. 11. Normal&d relationships of major and minor principal stresses at failure for various
types of rocklill

fills, although the values of b and /? are quite cle crushing is small for most rockfill (Leps, 1988).
similar, the value of CI is generally greater than While the data obtained by Marachi et al. (1972)
unity. For very low confining stresses (< 100 kPa) indicated a slight reduction of 9’ with increasing
the value of /I is less than 0.75; at high stress particle sizes, a contradicting trend was reported
levels (> 1.5 MPa) it is often greater than 0.85. As by Tombs (1969) for maximum grain sizes
a preliminary design tool in stability analysis, if varying from 10 to 75 mm. The effect of particle
the uniaxial compressive strength of the parent sizes on 4’ remains a more complex phenomenon
rock is known or determined by basic rock than the marked influence of confining stress. The
testing, then the failure envelope of the corre- role of the grain size distribution of the angle of
sponding rockfill can be estimated from one of friction cannot be verified purely on the basis of
the proposed failure criteria. At high normal the current test results. Barton & Kjaernsli (1981)
stresses (a,’ > 1 MPa), the values of b and /l proposed that the drained friction angle 4’ of
approach unity as the failure envelope becomes rockfill can be evaluated from the empirical
linear. Under these circumstances, the magnitudes expression
of a and tl represent tan & and the effective prin-
cipal stress ratio respectively for a given confining

where R is the equivalent roughness of rockfill,

Influence of confining stress on friction angle related to initial porosity of rockfill and genesis,
The variation of the drained friction angle 4’ of angularity and surface roughness of particles; s is
greywacke is plotted against the effective normal the equivalent strength of rockfill particles
stress in Fig. 12, where each point represents the expressed as a fraction of the uniaxial compres-
average of three independent tests. As the confin- sive strength of the parent rock; u,,’ is the effective
ing pressure and hence the normal stress is normal stress and &’ is the basic friction angle of
increased, the drained friction angle for rockfill smooth, planar unweathered rock surfaces.
gradations A and B decreases from 45” to 32” and Figure 12 also compares the test results of
from 43” to 33” respectively. Although the friction greywacke rockfill with the values predicted by
angle of gradation A drops faster than that of the equivalent roughness method. In determining
gradation B, as the normal stress is increased to 1 the angle of friction using this empirical method,
MPa, 4’ for both gradations attains the same the equivalent roughness R was taken as 6.5 for
value irrespective of the particle sizes. The sur- quarried (sharp to angular) greywacke rockfill of
prisingly large reduction in the angle of shearing compacted porosity 30%. The mean uniaxial
resistance at high confining pressures is probably compressive strength and the basic friction angle
associated with the significant increase in crush- for greywacke from southern Thailand have been
ing of angular particles. At low stress levels, parti- given by Cheng (1986) as 136 MPa and 25”. The
experimental results indicate good agreement
with the predicted values at low normal stresses.
M Gradation A: d,, = 39.1 mm At high normal stress levels, the laboratory fric-
M Gradation B: d,, = 25.4 mm tion angle decreases faster than is predicted by
- Predictions
the equivalent roughness method. It is important
(I%%% % Kjaernsli. 1981) to realize that the structural component of the
frictional resistance is not intended to model the
particle size effect or the potential breakage of
angular grains. It may be argued that due to the
dependence of the parameter s on the uniaxial
compressive strength uc, the term R log (s/a,‘)
would not decay as rapidly as the trend shown by
the current test results. On the basis of these
observations, it may be concluded that the
Barton & Kjaernsli (1981) approach should not be
extended to extrapolate the frictional behaviour
of rockfill at high effective normal stress levels,
particularly beyond 1 MPa. Nevertheless, the
311 I I I I 1 equivalent roughness method is adequate to esti-
100 300 500 700 900 1100 mate 4’ at low normal pressures for the tested
Effective normal stress: kPa
greywacke rockfill.
Fig. 12. Comparison of current test results with predic- Figure 13 compares the findings of the present
tions from the equivalent roughness method study with the summary of triaxial test data for

Isabella aranite 100 mm 0 lnfiernillo diorite 3

Cachuma gravel 19mm 0 lnfiernillo conglom.
200 mm
Cachuma gravel 1 8 Malpaso conglom.
Cachuma quarry 75 mm A Pinzandaran gravel J
Oroville tailings l lnfiernillo basalt
Soledad gravel 100 mm X lnfiernillo gneiss X
0 lnfiernillo gneiss Y 175 mm
60 l V Contreras gravel 1
W Santa Fe rock J
l Greywacke (A) 36 mm (present study)
0 Greywacke (6) 25 mm (present study)

40 -
Low-strength rockfill
(Leps. 1970)

35 -

proposed by Authors

Normal pressure o,,: kPa

Fig. 13. Comparison of current test data with previous studies

other rockfill presented by Leps (1970). The data For both gradations A and B of this greywacke
for greywacke rockfill fall between this lower rockfill, the corresponding relative densities were
bound and the line of ‘average rockfill’, except between 62-65%, as determined from several test
where the effective normal stresses exceed 700 specimens. The effect of varying the initial degree
kPa. Considering the test results of Contreras of compaction on the angle of shearing resistance
gravel and Santa Fe rock together with the was not investigated in detail, because the
current data on greywacke rockfill, it seems that primary objective was to evaluate the effect of
the lower boundary proposed by Leps (1970) for confining pressure (normal stress) on the shear
low density rockfill is slightly overestimated. A strength of greywacke rocktill. Nevertheless,
more conservative lower bound (3” less) as shown Brown (1988) has shown that excessive compac-
by the hatched line in Fig. 13 is suggested by the tion encourages particle breakage, although the
Authors. Note also that no experimental data are initial friction angle may be enhanced for the
available for confining pressures below 40 kPa, same confining pressure.
indicating the difftculty of testing rockflll samples Excessive compaction of wet rockfill may not
at such low confining stresses due to the lack of only induce crushing of angular fragments, but
cohesion. In the present study, a minimum con- also cause excess pore pressures if the per-
fining pressure of N 50 kPa was required in order meability is too low. In this respect, it is always
to prevent the samples from bulging during the good laboratory practice to saturate the rockfill
saturation stage. specimens before drained loading, so that the fric-
tion angles obtained in this manner are more rea-
Injuence of relative density and degree of compac- listic (conservative) than the higher values
tion encountered for dry specimens. As the down-
The initial porosities of the test specimens were stream shell of a dam would usually be unsatu-
maintained at 30-32%, in order to relate to the rated, the stability of the downstream slope may
actual field conditions of the Chiew Larn Dam. be enhanced due to the additional shear resist-

ante provided by suction, depending on the parti- serious consideration. Nevertheless, significant
cle size distribution and the relative density. crushing of naturally soft or weathered rock
during field compaction may cause non-
compliance with the permeability requirements
The friction angle corresponding to the failure (free draining), and enhance the risk of developing
envelope is the most important parameter construction pore pressures. It has been shown by
required in design for the slope stability analysis Balasubramaniam, Lee & Wijeyakulasuriya
of rockfill dams. It takes its maximum value at (1987) that any significant development of small
the least normal stress; at extremely high stress pore pressures can affect the effective shear
levels it may even approach values close to 30”. strength of rockfill. Therefore, good engineering
The non-linear strength envelope quantifies this practice should not allow any excess pore pres-
phenomenon adequately. In terms of the angle of sure development in the rockfill. In this respect,
friction of common rockfill materials, current the placement moisture content, grain size dis-
rockfill dams are constructed at much flatter tribution and degree of compaction in the field
slopes, in spite of the capabilities of modern have an important role to play. Penman (1978)
vibratory rollers in compacting rockfill to achieve has proposed that the average permeability of
field porosities lower than those obtained in the rockfill should not be less than 0.001 cm/s, so
laboratory. One reason for this, of course, is the that the amount of fines can be limited to mini-
presence of the central clay core, which influences mize the risk of excess pore pressure develop-
design leading to reduced upstream slope angles. ment.
If the appropriate friction angles are not carefully
selected according to the effective normal stress, CONCLUSIONS
the prediction of critical slip surfaces or sliding Greywacke rockfill tested in this study shows a
wedges becomes questionable. similar engineering behaviour to many other
Table 5 summarizes some important character- typical rockfills. Although the particle size and
istics of rockfill dams in Thailand. It is quite clear angularity of the rockfill also influence the stress-
that the smaller dams (~40 m high) are often strain behaviour, including dilation, the effect of
designed at steeper slopes than their larger confining stress on the shear strength is of
counterparts. For such small dams, the critical primary importance. While the degree of grain
failure surface may be predicted reasonably well crushing may be enhanced at elevated confining
by using a constant friction angle and still main- pressures, for competent rockfill subjected to rea-
tain an acceptable factor of safety, without having listic normal stresses particle breakage may not
to provide berms or external support at the toe. be a serious influential factor. The initial density
For instance, the Ubol Ratana dam has an of rockfill, however, is important as it is linked
average downstream slope of almost 40”, whereas directly to the degree of compaction. The labor-
the slopes of the highest Sri Nagarind dam are atory data on shear strength depend on the initial
less than 30”. porosity and relative density of the compacted
Some research has been conducted on particle test specimens. Therefore, while excessive field
crushing during compression. Although particle compaction is not justifiable, sufficient compac-
breakage has been quantified in the laboratory by tion with conventional machinery must be
Marsal (1973) under high confining stresses, in ensured, so that the laboratory and field poro-
practice crushing of particle edges may occur sities are similar.
during compaction of highly angular fragments. The angle of shearing resistance and the associ-
For less angular rockfill, crushing may not be a ated failure envelope of rockfill are directly related

Table 5. Details of some rockiill dams in Thailand (Thai National Committee on Dams, 1977)

Dam height: m Crest length: m

r Slope
r Type of rockfill

Upstream Downstream

Sri Nagarind 140 610 1 : 2.0 1 : 1.8 Limestone and quartzite

Chiew Lam 90 700 1 :2.0 1 : 1.8 Greywacke
Chulabhom 70 700 1 : 1.7 1 : 1.6 Sandstone
Pattani 85 422 1 :2.0 1 : 1.8 Greywacke
Nam Pung 40 1120 1 : 2.0 1 : 1.7 Conglomerate
Ubol Ratana 32 800 1 :1.3 1 : 1.3 Sandstone
Sirindhom 42 940 1 : 1.8 1 : 1.6 Siltstone

to the magnitude of confining stress. At low con- rockfill dams. J. Soil Mech. Fdn Engng Div. Am. Civ.
fining stresses (< 500 kPa), the non-linearity of Engrs 95, SMI, 313-330.
the failure envelope is pronounced. At much Hoek, E. &. Brown, E. T. (1980). Underground excava-
higher confining stress levels (> 1.5 MPa), the tions in rock. London: Institute of Mining and
assumption of the linear Mohr-Coulomb cri-
Indraratna, B. (1990). Development and applications of
terion is quite acceptable. Considering the current a synthetic material to simulate soft sedimentary
test results of greywacke rockfill together with rocks. Giotechnique 40, No. 2, 189-200.
previous experimental findings, two modified Indraratna, B. & Kaiser, P. K. (1990). Design for
failure criteria for rockfill have been proposed in grouted rock bolts based on the convergence control
non-dimensional form, incorporating the uncon- method. Int. J. Rock Mech. Sci. Geomech. Abstr. 27,
fined compressive strength of the rock type. The No. 4,269-281.
characteristic coefficients of these criteria are Ito, Y. (1983). Design and construction by NATM
independent of the system of units, and their through Chogiezawa Fault Zone for Enasan Tunnel
on Central Motorway. (In Japanese.) Tunnels &
values for a variety of rockfill types have been
Underground 14,7-14.
determined from low to very high stress ranges. Lee, Y. H. (1986). Strength and deformation character-
The upper and lower bounds of these coefficients istics of rock@. PhD thesis, Asian Institute of Tech-
proposed by the Authors provide the engineer nology, Bangkok.
with preliminary design guidelines, in the absence Leps, T. M. (1970). Review of shearing strength of rock-
of detailed laboratory testing of a given rockfill. fill. J. Soil Mech. Fdn Engng Div. Am. Sot. Ciu.
Using this approach, if the unconfined compres- Engrs 96, SM4, 1159-l 170.
sive strength of the parent rock is known, the Leps, T. M. (1988). Rockfill dam design and analysis.
shear strength envelope of the quarried rockfill Advanced dam engineering for design, construction
and rehabilitation (ed. R. B. Jansen), pp. 368-387.
can be estimated.
New York: Van Nostrand Reinhold.
The effect of confining pressure on angle of Lowe, J. (1964). Shear strength of coarse embankment
internal friction is very important in the stability dam materials. Proc. 8th Int. Congr. Large Dams 3,
analysis of rockfill slopes. A conventional analysis 745-761.
that employs a constant friction angle (average) Marachi, N. D. (1969). Strength and deformation charac-
provides an over-conservative factor of safety for teristics of rock@ materials. PhD thesis, University
shallow slip surfaces. If the actual variation of 4’ of California.
with 6,’ is incorporated in the design, most rock- Marachi, N. D., Chan, C. K. & Seed, H. B. (1972).
fill embankments can be raised with steeper Evaluation of properties of rocktil materials. J. Soil.
Mech. Fdn Engng Dia. Am. Sot. Civ. Engrs 98, SMl,
slopes while maintaining an adequate factor of
95-l 14.
safety greater than 1.5 (Wijewardena, 1991). The Marsal, R. J. (1967). Large scale testing of rockfill
use of a constant mean 4’ for deep-seated slips materials. J. Soil Mech. Fdn Engng Diu. Am. Sot.
may overestimate the factor of safety, but the dis- Civ. Engrs 93, SM2, 27-43.
parity may not be substantial. Marsal, R. J. (1973). Mechanical properties of rockfill.
Embankment dam engineering, Casagrande volume,
REFERENCES pp. 109-200. New York: Wiley.
Balasubramaniam, A. S., Lee Y. H. & Wijeyakulasuriya, Nitchiporovitch, A. A. (1969). Shearing strength of
V. (1987). Stress-strain behaviour of rockfill. Proc. coarse shell materials. Proc. 7th Int. Conf Soil Mech.
8th Asian Reg. Con& Soil Mech. Fdn Engng, Kyoto, Fdn Engng, Mexico City, 211-216.
Japan 1,21-25. Penman, A. D. M. (1978). The behaviour of rockfill
Barton, N. & Kjaernsli, B. (1981). Shear strength of dams. Geotechnical Engineering (Journal of the
rockfill. J. Geotech. Engng Div. Am. Sot. Cio. Engrs Southeast Asian Geotechnical Society) 9, 105-122.
107, GT7,873-891. Ponce, V. M. & Bell, J. M. (1971). Shear strength of
Bishop, A. W. (1966). The strength of soils as engineer- sand at extremely low pressures. J. Soil Mech. Fdn
ini materials. G&technique 16, No. 2,91-128. Engng Div. Am. Sot. Civ. Engrs 97, SM4,625-637.
Brown. A. J. (1988). Use of soft rockfill at Evretou Dam, Terzaghi, K. (1960). Discussion on Salt Springs and
Cyprus. &oteihnique 38, No. 3,333-354. Lower Bear River Dams. Trans. Am. Sot. Ciu. Engrs
Charles, J. A. & Watts, K. S. (1980). The influence of 125, 139-148.
confining pressure on the shear strength of com- Thai National Committee on Large Dams (1977). Large
pacted rockfill. Ghotechnique 30, No. 4,353-367. dams in Thailand. Bangkok, Thailand: Samarn.
Cheng, H. (1986). Shear behaoiour of greywacke rock@ Tombs, S. G. (1969). Strength and deformation character-
material. MEng thesis, Asian Institute of Tech- istics ofrockfill. PhD thesis, University of London.
nology, Bangkok. Vesic, A. S. & Clough, G. W. (1968). Behaviour of
de Mello, V. F. B. (1977). Reflections on design deci- granular materials under high stresses. J. Soil Mech.
sions of practical signifiance to embankment dams. Fdn Engng Div. Am. Sot. Civ. Engrs 94, SM3, 661-
Gtotechnique 27, No. 3, 279-355. 668.
Fagnoul, A. &. Bonnechere, F. (1969). Shear strength of Wiiewardena, L. S. S. (1991). Engineering properties of
porphiry materials. Proc. 7th Int. Con& Soil Mech. -greywacke rockJill and their effect on the dynamic sta-
Fdn Engng, Mexico, 61-65. bility analysis of the Chiew Larn Dam. MEng thesis,
Fumagalli, E. (1969). Tests on cohesionless materials for Asian Institute of Technology, Bangkok.