Characterizing Mine Tailings For Geotechnical Design

Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

INTERNATIONAL SOCIETY FOR

SOIL MECHANICS AND


GEOTECHNICAL ENGINEERING

This paper was downloaded from the Online Library of


the International Society for Soil Mechanics and
Geotechnical Engineering (ISSMGE). The library is
available here:

https://www.issmge.org/publications/online-library

This is an open-access database that archives thousands


of papers published under the Auspices of the ISSMGE and
maintained by the Innovation and Development
Committee of ISSMGE.
Geotechnical and Geophysical Site Characterisation 5 – Lehane, Acosta-Martínez & Kelly (Eds)
© 2016 Australian Geomechanics Society, Sydney, Australia, ISBN 978-0-9946261-1-0

Characterizing mine tailings for geotechnical design


K. Been
Golder Associates, Canada

ABSTRACT: Mine tailings are ground up rock and generally consist of sand and silt size particles, without clay
minerals. Tailings “storage” (actually disposal) facilities are some of the largest constructed works, with seismic
design an integral component. High value mines are frequently in earthquake prone areas and tailings liquefac-
tion is an ever-present concern. A screening level liquefaction assessment based on the CPTu is a needed first
step, but cannot be accurate because the fines content of tailings can be high, and engineering behaviour is only
loosely related to fines content. An engineering mechanics based procedure is laid out in this paper which is
applicable regardless of the fines content. It is anchored to the eighty year old principles of critical state soil
mechanics originated by the Corp of Engineers. The state parameter provides the practical engineer an entry to
this framework, as well as insight into a simple stability principle that sets the strategy for a tailings character-
ization project aimed at the analyses that a tailings engineer needs. The approach requires systematic in situ and
laboratory testing to determine the soil mechanics properties of the tailings.

1 INTRODUCTION Characterization is influenced by “what has failed


and why”, as there are no industry-standard or inter-
Characterization of mine tailings is a difficult and un- nationally-recognized codes of practice for tailings
derestimated problem in geotechnical engineering impoundment engineering. Earthquake resistant de-
(even neglecting geochemical aspects, with acid sign, largely driven by the significance of damage in
drainage a pervasive and real issue). Much of the dif- the Niigata and Lower San Fernando events, became
ficulty arises because tailings are ground up rock a focus of what were the appropriate procedures after
waste from the mineral extraction process and gener- 1965. But static failure has actually been one of the
ally consist of sand and silt size particles, without clay most damaging failure modes for tailings facilities:
minerals although cohesion may be present. They of- Aberfan (1968, UK), 144 dead from static liquefac-
ten fall into a “transitional material” category, some- tion caused by rising ground water pressure;
where between idealized sand-like or clay-like behav- Aznalcollar (1998, Spain), brittle foundation failure
iour (idealizations controlled by void ratio or over- causing dam breach and wide-spread contamination
consolidation respectively). The pressure to minimize downstream; Kingston fly ash (2008, USA), un-
costs of investigation without a coherent approach to drained static failure of fine-grained fly ash, causing
evaluate transitional materials means that empirical significant environmental contamination; Mt Polley
methods derived from research work on sands or (2014, Canada), rapid foundation failure resulting in
clays is often applied inappropriately (or at least with dam breach and one of the biggest environmental dis-
a significant uncertainty). asters in modern Canadian history; and, Fundao
Tailings deposits need to be characterized for (2015, Brazil) static liquefaction with at least 17 dead
many reasons: centerline or upstream dyke raises may and estimated damages of billions of dollars
be built on recent tailings, sometimes facility expan- (www.fundaoinvestigation.com). A guide to what
sion occurs over historical tailings deposits and there needs considering is the ICOLD bulletin on dam fail-
is a need to confirm the design assumptions for new ures. This paper focuses on the subset of geotechnical
facilities. This latter requirement is particularly im- considerations including both static and earthquake-
portant for thickened tailings where the final geome- induced liquefaction.
try includes a slope of weak silts extending above the A good practice approach, based on engineering
crest of containment structures. mechanics, is laid out in this paper. This means meas-
uring the engineering parameters needed for de-

41
sign/assessment: soil strength (drained and un- silts found in western Canada. The same testing is-
drained), soil stiffness, and cyclic softening. Assump- sues arise.
tions that engineering behaviour can be satisfactorily Silts, whether tailings or natural rock flour soils,
related to fines content are avoided as tailings may can be sampled successfully using a thin-wall un-
consist of 100% fines. Underlying the
0.9

Critical State Locus Loose, contractive


(CSL) (liquefaction likely)
1
 0.7

Void ratio
Void ratio, e

ec In situ void ratio shown as


open point
 = e ‐ ec
(negative) As tested void ratio in CSS
0.5
shown as solid point

Current void ratio of


the soil, e Note: plotted stress levels slightly adjusted to
Dense, dilatant show each test individually
(liquefaction unlikely) 0.3
0 200 400 600 800
Mean effective stress, p' Effec�ve ver�cal stress, σ′vo (kPa)
Figure 1. Definition of state parameter swaged Shelby tube, much like clays are sampled. Of-
ten, it is also possible to recover sandy silts using
Figure 2. Effect of post sampling process on void ratio of undis-
turbed silt samples
approach in this paper is generalized critical state soil
mechanics (CSSM), implemented through the state
similar samplers. But, in either case, subsequent pro-
parameter ψ (defined on Figure 1). In many ways, this
cesses, i.e sample transportation, extrusion, trim-
is the thread of characterizing soil behaviour that
ming, and setting up in the laboratory apparatus (e.g.
started in 1935 with the world’s first liquefaction-re-
reconsolidation to the in situ stress level in a triaxial
sistant dam (Franklin Falls, NH) and with contribu-
cell prior to loading), cause substantial densification
tions by various workers over the past 80 years now
between the as recovered void ratio in the sample tube
making it the dominant framework for understanding
and the as tested void ratio. The effect of these post-
all soil stress-strain behaviour (drained or undrained,
sampling processes is illustrated on Figure 2 for a nat-
and all stress paths). It is not difficult to use, as will
ural silt (from Vancouver Island, Canada) that was
be shown in this paper.
sampled and handled with extreme care (Mohajeri &
Ghafghazi, 2012). A densification of about e ≈ 0.15
was found, which is about a third of the void ratio
2 PARTICULAR ASPECTS OF SILTS
range between the loosest possible soil and the dens-
est possible soil (analogous to emax and emin, although
Tailings start as a slurry of sand and silt particles,
of course the ASTM procedure does not apply to silt-
which are discharged into an impoundment to sedi-
dominated soils).
ment and accumulate – almost a perfect example of
This densification means what is tested is much
‘normally consolidated’ geology. If simply spigoted,
stronger than what is in situ. For perspective, testing
the tailings hydraulically segregate. In any particular
of a natural silt using careful reconstitution found that
tailings storage facility (TSF), you should anticipate
the number of cycles to liquefaction doubles for e =
at least two material types with the sand fraction drop-
0.05 densification. This densification of silts from in
ping out quickly to form beaches or similar features
situ to laboratory is a serious engineering challenge,
with sandy silts and pure silts (slimes) moving further
making it dangerous to use a clay-like approach
down the impoundment and usually accumulating un-
where laboratory strengths are taken at face value.
derwater. These silt-dominated soils comprise a large
What to do ?
portion of the impoundment, so tailings characteriza-
There are two alternative, and complementary, ap-
tion becomes “how do we determine silt behavior”?
proaches to characterizing silts: a) measure strengths
A recent trend in the mining industry has been to
in the laboratory and correct the results for the known
increase the slurry density (and recover more water)
densification; and/or b) treat silts as sand-like, meas-
before discharge, referred to as thickening. Thickened
uring  in situ and using that to compute in situ
tailings do not segregate, so in this instance the mate-
strengths from appropriate test results.
rial being dealt with is commonly a silt containing
In either approach, how silt behaviour changes
around 10-15% sand size particles and the character-
with void ratio must be quantified and, today, that re-
ization question remains the same. Tailings silts are
quires critical state soil mechanics.
also not that different from rock-flour derived natural

42
3 CRITICAL STATE SOIL MECHANICS The trends in sand behaviour reported by Been &
Jefferies resurrected CSSM. The basic theoretical in-
The critical void ratio was identified from the practi- sight that followed from Figure 3 was that yield sur-
cal concern of avoiding static liquefaction failures of faces (the limits of elastic behaviour) in general do
hydraulic fill dams in the late 19th and early 20th cen- not lie on the CSL but rather evolve to the CSL with
turies. “Critical” really meant what it said – it was the shear strain. This insight produced two strands of
criterion of a safe density in constructed engineering
48
work, with the practical concern to avoid sudden tran-
sitioning of drained construction with no excess pore
44
pressure, into an undrained liquefaction failure. The
Tailings: φc = 35.5°
critical void ratio developed from engineering by the
40

Friction Angle, 'tc


Corps of Engineers for Franklin Falls dam, with
Casagrande’s famous experiments at Harvard. This
36
was rapidly followed by Taylor at MIT (in connection Sands [Eq 1]
with the Fort Peck dam failure) who showed that the φ = φc ‐ 45 ψ
φc = 31.5°
critical void ratio depended on confining stress, or 32

what is today called the critical state line (CSL).


Theoretical soil mechanics has always sought to 28
‐0.30 ‐0.20 ‐0.10 0.00 0.10
explain observed soil behaviour in terms of simple State Parameter, o
physical principles and corresponding mathematics. Figure 3: Friction angle as a function of state parameter. Origi-
Since particulate materials can exist over a range of nal data for sands (Jefferies & Been, 2015) updated with tailings,
void ratios at any confining stress, depending on the shown as coloured symbols
packing of the particles, the key challenge has always
been to understand (and quantify) how void ratio af- constitutive models: those strictly following the
fects soil strength and stiffness. The stress-strain be- Drucker at al. (1957) framework (e.g. Jefferies, 1993)
haviour of all engineering materials is dominated by and developments of bounding surface plasticity (e.g.
the theory of plasticity, and for various theoretical Manzari & Dafalias, 1997). These models are “good”
reasons the CSL was adopted as a central idealization in the sense that they closely predict the effect of void
in the Original Cam Clay and Modified Cam Clay ratio and confining stress on soil behaviour and only
models. Seemingly because of the title of the use a few (generally familiar) soil properties. There
Schofield & Wroth (1968) book setting out these are now about a dozen good models, and they extend
ideas, critical state soil mechanics (CSSM) became past their origins in triaxial p,q space to general stress
(erroneously) synonymous with Cam Clay. states and all loading paths. These models all have
Cam Clay is both simple and mathematically ele- common features:
gant, theoretically quantifying how void ratio affects 1 a CSL (not necessarily the familiar semi-log ideal-
soil stiffness and strength. Some (i.e. the British) ization);
view these models as “the greatest 20th century con- 2 a critical friction ratio (or angle);
tribution to soil mechanics”. Others view Cam Clay 3 dilation and strength controlled by ;
(either variant) as excessively idealized and utterly 4 plastic hardening that partially scales with the
useless for real soils in engineering practice (delight- slope of the CSL; and,
fully summed up in the critique of CSSM on Wikipe- 5 isotropic elasticity with stress-level dependence.
dia). The problem has become the linkage of Cam
Clay with CSSM because of the book’s title, with the The models differ in rather small details, and it is
inadequacies of Cam Clay being used to throw out the largely a matter of taste as to which is used. They all
real engineering value of the thread that started with give comparable stress-strain behaviour and show
the Corps of Engineers in 1935. how void ratio affects that behaviour. Figure 4 illus-
A key contribution in returning to the value of trates the match between one of these good models
CSSM was Been & Jefferies (1985). This was driven, (NorSand) and test data. An excellent fit to a spec-
much like the original work at Franklin Falls, with trum of drained and undrained soil behaviour is ap-
concerns about static liquefaction of hydraulically parent. The test data shown is a construction sand
placed sand. It was found that a large number of used in a large hydraulic fill that suddenly failed near
drained strength tests on various sands, with fines full-height (Nerlerk).
contents ranging from 0 to 18%, showed a simple and A recurrent critique of CSSM has been that the
systematic trend: CSL is not unique and changes with stress path to
reach CSL. This critique was thoroughly addressed
 = c –   for laboratory tests on sands in Been et al. (1991)
where the coefficient 45 in (1) appeared constant while Jefferies & Shuttle (2005) showed that
from one soil to another. An updated version of this NorSand, which has an explicit unique CSL, captured
trend is shown on Figure 3.

43
800 600
DRAINED TESTS UNDRAINED

Deviator stress, q (kPa)


Deviator stress, q (kPa)

600
400
NorSand

400 CIU_G103

NorSand CIU_G105

CID_G151 200 CIU_G107

200
CID_G156

CID_G157

0 0
0 5 10 15 0 5 10
Axial strain (%) Axial strain (%)
2 600
CSL
Mtx = 1.27
0 φ′ = 31.6
Volumetric strain (%)

400
‐2 q (kPa)

‐4
200
ψo
‐6 Tes t
CID‐G151
p 0 (kPa )
201
e0
0.694 ‐0.050
CID‐G156 201 0.640 ‐0.104
CID‐G157 204 0.572 ‐0.172
‐8
CIU‐G103 500 0.793 0.068
CIU‐G105 500 0.757 0.032
0
CIU‐G107 700 0.727 0.009 0 200 400 600 800
‐10 p' (kPa)

Figure 4. Match between CSSM model for sand (NorSand) and triaxial test data from Nerlerk hydraulic fill. Results are
compared at a range of ψ for drained (dense) and undrained (loose) specimens.

only during construction. This was the concern of the


the range of soil behaviour used to assert non-unique- Corps of Engineers in 1935, and it is a concern that
ness. ought to have applied at Fundao eighty years later.
Using CSSM in engineering practice is simple and What is the soil behaviour causing the rapid instabil-
involves just three steps which are always part of the ity and transition from drained to undrained behav-
geotechnical engineering process: iour?
 determine the state of the soil (tailings) in situ; Figure 5 shows some soil stress-strain curves and
 measure the engineering parameters as a func- the associated stress paths under constant horizontal
tion of state, and; plane-strain compression. The data annotated as A is
 carry out the appropriate analyses. a purely undrained test of loose soil, with the familiar
These steps are discussed shortly, but first we need to associated static liquefaction once the undrained-
return to 1935 and the concerns about hydraulic fill strength is reached. The curve annotated as B com-
construction to identify the key problem. prises soil with the same , but now first loaded to
point B1 drained. The soil is then subject to change to
undrained loading, but this change is not dramatic as
4 A BASIC STABILITY PRINCIPLE the stress at the change to undrained loading is less
that its undrained strength (B2). The data shown as
A feature of staged construction of earthworks is that ‘C’ is a further progression of drained loading of soil
it is done slowly to allow consolidation during fill at the same  but now with the drained loading ex-
placement. This aspect is often formalized in up- ceeding the undrained strength at B2 before switching
stream construction of tailings dams with piezometric to undrained conditions. As can be seen on the figure,
measurement to confirm drained or consolidated con- instant brittle liquefaction develops despite there be-
ditions. Yet, these dams can fail suddenly and not

44
ing zero excess pore pressure at the instant of lique- effective stress is constant (temporarily) at peak
faction. The reason for the instantaneous transition is strength. In undrained loading of loose soils, the mean
that the potential rate of excess pore pressure genera- effective stress is changing at its greatest rate because
tion is infinite under load-controlled situations if the of pore pressure changes at peak strength. This effect
current undrained strength of the soil is exceeded of rapidly changing mean effective stress on soil be-
even though the loading path to that limit is perfectly haviour is intrinsic to the good soil models mentioned
drained. This undrained strength is called the “insta- earlier, with good models accurately capturing both
bility limit” in effective stress terms. drained and undrained soil behaviour using the same
The path from point B1 to point B2 on Figure 5 properties.
corresponds to a rapid loading event, which could be The idea of triggering undrained instability is
a quickly placed berm raise, an earthquake, or even sometimes difficult to appreciate as it involves the
simply an increase in phreatic level within the tail- balance between drainage time and strain rates, with

Figure 5: Transition from drained to undrained behaviour lead-


ing to potential rapid liquefaction after undrained perturbation
Figure 6: Instability limit (effective stress ratio at peak un-
drained strength) in relation to projected limiting stress ratio
ings unrelated to construction processes. Provided the from drained tests on dense specimens (from Figure 3)
stresses are less than the instability limit, the soil has
sufficient strength reserve to withstand the perturba-
tion and the expected consolidation can develop. But, 0.6
Grey and black El Pachon P10 Silt 42/56
if the perturbation stresses exceed the instability limit symbols are sands,
TCS Sand
then monitoring construction and/or onset of failure coloured symbols
are tailings TCB Silt
with piezometers in conjunction with drained shear
Shear strength ratio su/po'

0.4 Hilton Mines


strengths is an unsafe engineering approach. There Oil Sands Tailings
will be no warning of the liquefaction failure, appar- Guindon A 120/34.5

ently the situation at Fundao and a century earlier at Guindon B 50/67

other hydraulic fill dams. Let us now consider this in- 0.2

stability limit a little further.


Figure 3 showed a large data base of drained triax- Sands
ial tests, and there is a comparable data base of un- 0.0
drained tests (that is how the various CSLs were 0.00 0.10
State parameter 
0.20

measured to determine the  plotted on Figure 3). So


that we compare like with like, the peak undrained Figure 7: Undrained shear strength ratio for sands and silts (a
total stress version of the instability limit on Figure 6)
strength can be reduced to a mobilized effective stress
ratio L = q/pʹ and the peak strength in drained com- internal load transfers also affecting the mechanics. It
pression likewise to an effective mobilized stress ra- may be simpler to think in the same way as stageload-
tio. Figure 6 plots the L at peak undrained strength ing on soft clays and always keep the embank-ment
versus  at that peak strength for the undrained tests. stable using undrained strengths in analyses. Thus, ra-
A comparable trend to Figure 3 is observed, with an ther than use Figure 6, the same data is shown on Fig-
interesting feature. The top trend line annotated on ure 7 as a familiar undrained strength ratio. Both the
Figure 6 is the average trend for the drained tests of effective instability limit and the undrained shear
Figure 3 (but extrapolated to positive state parameters show the same thing in a different way but neither is
from the dense states of Figure 3). This is sensibly an fundamental. The instability depends on the soil prop-
upper bound to the undrained data. However, the erties, such as the ratio of elastic and plastic modulus,
best-fit trend through the undrained data is parallel to critical friction ratio, compressibility and loading
the drained trendline and the two are offset by  = path. The scatter in Figures 6 & 7 reflect these unre-
0.06 as indicated. This offset develops because the ported factors, not testing errors.
conditions are different. In drained loading, the mean

45
Figure 8: Example CPTu in a TSF showing segregation of coarse and fine layers. The tailings are TCS and TCB materials used
for illustration of the CSSM techniques in the paper

Practically, it may be simpler to appreciate that the


“critical” decision point that concerned the Corps of 5 MEASURING IN SITU STATE OF TAILINGS
Engineers 80 years ago is the offset  < 0.06. This -
0.06 in void ratio or  is a generally conservative Geotechnical engineers have several options for a pa-
number, as can be seen on Figure 6, where some soils rameter to characterize the in situ state of tailings.
are systematically better since undrained behaviour Amongst them are void ratio, relative density, dila-
depends on soil compressibility as well as intrinsic tancy index, gamma ray absorption from downhole
frictional properties. logging, etc. All of these methods are technically
What should a practical engineer take away from challenging and really require an accurate calibration
these considerations? The basic instability principle to what actually occurs in situ.
is this: What we should be measuring is the in situ state
 if  <0.06 then undrained strengths will be parameter (ψ) with a CPTu, as rigorously as possible.
greater than drained; The degree of rigour will determine the cost of
 if  is looser, then stability depends on the the investigation. Firstly, we recognize that we cannot
available undrained strength even though cur- easily obtain undisturbed samples of tailings materi-
rent loading may be drained als, but the CPTu gives us an accurate profile through
our deposit. Second, we know that the value of ψ is
There is a corollary to this principle if the second sit- the major control on behaviour of soils, and this ap-
uation prevails, and that invariably arises with thick- plies to the CPTu resistance, friction angle and cyclic
ened tailings, which is whether to use peak undrained stress (although each behaviour has other factors that
strength with a decent factor of safety (to allow for are also important). Third, if we could accurately
brittle load transfer of overstressed soil) or to use re- measure void ratio or density, we would still be left
sidual undrained (minimum assured) strength with a with a problem of converting the void ratio back to ψ
much lower factor of safety. There are various views and the engineering behaviour because of the range
on this issue, with a preference for the second ap- of CSLs of the soil in situ.
proach in the Author’s company.
So, with this principle established as a background
to engineering at a TSF site, let us turn our attention
to characterizing the in situ state of the tailings and
the associated strengths to use for engineering.

46
Tip resistance, qt (MPa) Pore Pressure (MPa) Friction ratio, F (%) Bq State parameter, 
0 2 4 6 8 0 0.5 1 1.5 0 1 2 3 4 5 6 0.0 0.4 0.8 1.2 ‐0.2 0.0 0.2
0

Sub‐aerial
thickened
tailings
Depth below ground (m)

10 Earlier sub‐
aerial sand
beaches

Sub‐aqueous
silt (slimes)

20
State parameter
profile from soil
uo defined by Note high Bq specific calibration
dissipation tests values of CPTu

on a soil behaviour type chart rather than a plot with


depth, Figure 10. This chart was contoured

5.1 Examples of CPTu in tailings


Variations of material type in the TSF will show up 1000
on the CPTu and it is important to recognize them. Demarcation between strain softening and strain
However, the differences in ψ in situ affect the CPTu hardening behaviour following initial liquefaction
Dimensionless penetration resistance, Q (1‐ Bq) + 1

measurements in approximately the same way for (Shuttle & Cunning, 2008)
most materials and we are able to deal with that if we
know the properties of each material.
Gravelly
This aspect is illustrated in Figure 8, a CPTu show- Sub‐aerially deposited
ing there are two interbedded materials within the 100 beach materials

tailings storage facility (TSF). One material is a TCS sand


 = ‐0.20
coarser silty sand tailings (which we will call TCS Sands to
Sand segregated
sand) deposited near the spigots as it drops out first sand some silt
from taillings
 = ‐0.15
from the slurry, while the second is a finer (sandy) silt
tailings (TCB silt) deposited when the spigot was fur- TCB silt
ther away from the location. The TCB silt tailings Silty sands to
10  = ‐0.10
have a soil behaviour type index of Ic ~ 2.7 while TCS sandy silts
sand shows Ic ~ 2.0. Interpretation of the CPTu for Thickened tailings deposited in
tailings will be addressed using these tailings for il- layers with drying between lifts
 = ‐0.05
lustration, but first we will examine another example Clayey silts
and useful screening tool.
 = 0.0 (CSL)
Figure 9: Example CPTu in copper-zinc tailings. It is a massive silt deposit withClays
relicand other features from earlier hydraulic segrega-
beach Silts, deposited
tion and more recent thickened tailings. sensitive soils
1 sub‐aqueously

Figure 9 shows conditions in a different TSF with 0.1 1 10


Friction ratio F (%)
massive sub- aqueous silt (slimes) deposits, some rel- Figure 10: Dimensionless classification chart for soil behaviour
ict sandy beach features and more recent thickened type for identified materials on CPTu examples in Figure 8 and
tailings deposited sub-aerially. There is clearly a very Figure 9.
wide range of CPTu signatures that occur in tailings
facilities. Some insight is gained by viewing the data

47
 Shuttle & Cunning (2007) next examined what
for the state parameter by Shuttle & Cunning (2008) would happen with undrained behaviour in tail-
using their effective stress cavity expansion simula- ings, bringing in the effective stress dimension-
tions, discussion with Robertson (2008) and the less parameter grouping to tip resistance intro-
Plewes et al. (1992) equations, which give a guide to duced by Houlsby (1988). Besides their method
soil state as well as the soil behaviour type. The ex- of analysis, they entered a discussion (Shuttle &
ample sounding has the beached sands plotting only Cunning, 2008) with Robertson (2008) and iden-
slightly contractive (-0.05 < ψ < 0.0) and thus prone tified the contractive/dilatant boundary for all
to liquefaction. The sub-aqeous silts (ψ > +0.05, and soils (= - 0.05, which is essentially equivalent
very low CPTu resistance) are exceedingly loose and to offset between drained behaviour and the insta-
raise interesting questions about their potential bility limit shown on Figure 6; see Figure 10).
strength loss following liquefaction. Thickened tail-  Robertson (2009, 2010, 2012) has continued to
ings in this instance (0.0 < ψ < +0.05) would require publish more information, pulling together what
careful engineering to avoid liquefaction flow failures he sees as the state-of-the-practice.
on slopes.
For comparison, TCS sand (from 28.5 to 36m 5.2.1 Data reduction in sand
depth) and TCB silt strata (36 to 49.5 m depth) are In the case of sands, which are fully drained sound-
also shown on Figure 10, and maybe not surprisingly ings, the relationship between ψ and the CPTu re-
the three sub-aerially deposited tailings show up as sistance was originally determined from calibration
the same soil behaviour type and in situ state. chamber tests. Such tests, which are in effect a giant
triaxial cell, involve pushing a CPTu into a sand of
5.2 Interpreting in situ ψ for tailings known void ratio and under known stress. Repeating
the procedure with many samples provides a mapping
Interpretation of ψ from CPTu data has been in the between tip resistance, soil state, and stress level.
geotechnical literature for almost 30 years now, start- Most in the in situ testing community regard chamber
ing with Been et al. (1986 & 1987). A more recent calibration of the CPTu as the gold standard to be
publication specifically for tailings deposits is in used. The only catch in this is that each chamber test
Been et al. (2012). However, there are numerous ref- involves careful placement of 2 tonnes of sand so is
erences leading from 1986 to the present. An issue not commercially viable for most projects. Chamber
was that the early work was based on large calibration tests tend to be research programs by universities.
chamber tests of the CPTu in sand, and Sladen (1989) The starting point for determining the state param-
identified a potentially significant stress-level bias.
eter from the CPTu was this worldwide set of cham-
Shuttle & Jefferies (1998) eventually modeled this
ber tests, with samples of the sands then tested to de-
bias and showed how the elastic zone outside of the
termine their CSL. This process showed that (Been et
plastic zone caused this effect. Evaluation of CPTu
al, 1986):
data needs to include the in situ Gmax. In the mean-
time, various engineers had been working on other Qp = k exp (-m ψ) or in its correct inverted form
ways to progress the understanding and the methods ψ = - ln (Qp / k) / m (2)
adopted are rather varied. Some of the main contribu-
tions and their features are: where Qp= (qt - p)/pʹ and is the dimensionless CPTu
resistance (note use of mean, not vertical, stress). The
 Been et al. (1987) methods are appropriate if there coefficients k, m are soil-specific, depending on the
is indeed a calibration chamber test program, for soil properties. It was acknowledged that the CPTu
which there is good data. This is not common. resistance depended on soil compressibility, which
 Plewes et at. (1992) developed a screening level Been et al. (1987) formalized by relating k, m to
method to provide the in situ state of tailings, in (the slope of the CSL). The method recovers  with
which λ (the slope of the CSL) is related to the a precision ±0.05 across the range of available cali-
CPTu friction ratio F%. This method is still fairly bration chamber results at the time. They were mainly
robust, as expounded by Reid (2015) showing clean sands, and nearly all quartz sands, which is a
how λ10 and F compare on Figure 11. big limitation for tailings.
 Shuttle & Jefferies (1998) carried out detailed,
large strain, numerical analyses of the CPTu in
sand. They identified both the elastic and plastic
components of the resistance, and identified a
simplified method to determine ψ (of course, only
after knowing the CSL and the complete set of
soil properties including Gmax)

48
1 CPTu soundings in silts are undrained, often with ra-
Original data
ther large excess pore pressures (Figure 9). What
Original proposed correlation
should be done?
New data ‐ discrete value
In the case of clays, undrained CPTu are evaluated
New data ‐ with uncertainty range on the basis of total, not effective, stress by compar-
ing the undrained strength su inferred from the CPTu
to some reference strength to establish the calibration
λ10

(commonly a triaxial strength using one of several


0.1 testing protocols or in situ vane shear). But silts offer
a challenge as their densification from an as sampled
void ratio to the tested void ratio precludes a sensible
reference laboratory procedure. It is not evident what
might form a reference in situ test.
On the other hand, good models using CSSM have
no trouble capturing the stress-strain behaviour of
0.01 silts, drained and undrained. Thus the state parameter
0.1 1 Fr (%) 10 must be the basis for characterizing silt-dominated
Figure 11: Slope of λ10 as a function of F% (after Reid 2015, soils. It then follows (Been et al, 1988) that the frame-
Reid 2012 and Plewes et al 1992) work for evaluating CPTu in undrained soundings
should be analogous to (2):
ψ = - ln (Qʹ / kʹ) / mʹ (3)
Strictly, the coefficients k, m depend on all the soil
properties and also on the in situ Gmax. This was ex- where Qʹ = Q(1-Bq)+1 with kʹ, mʹ being the soil-spe-
plored using the familiar cavity-expansion analogue cific coefficients for undrained soundings. The Shut-
for the CPT, scaled to the chamber calibration data, tle & Cunning (2007) work determined these coeffi-
using large-displacement finite element simulations cients kʹ, mʹ using the same cavity expansion
with NorSand (Shuttle & Jefferies, 1998). The exten- approach of the Shuttle & Jefferies (1998) study by
sive parametric simulations were fitted using simple simply switching the analysis to undrained condi-
functions relating k,m to the soil properties discussed tions. This is reasonable as the NorSand constitutive
earlier. This is easy enough to use in practice and re- model fits silts, drained or undrained, just as well as
covers  with a precision ±0.02. The method explic- it fits sands and the numerical procedures do not de-
itly considers aspects like the critical stress ratio Mtx pend on the boundary conditions. The soil properties
(or constant volume friction angle) being quite differ- (e.g. Mtx, ) remain identical although obviously the
ent in tailings compared to natural soils, for example. numerical values change from one soil to another, but
Of course, many practical engineers still prefer a Shuttle & Cunning did not report extensive paramet-
physical calibration, and it is a little surprising that ric studies, and there are no convenient expressions
one sees so few calibration chamber studies in the for engineering practice to compute the state parame-
context of billion dollar liabilities for many projects. ter in undrained soundings. What is available is their
Lower cost physical calibrations have been achieved public-domain finite element code, downloadable at
by doing CPTu tests in a centrifuge (Bolton et al., www.crcpress.com/Soil-Liquefaction.
1999), with the advantage of a smaller sample and the One further step in using (3) is that it is based on
fact that each test covers a range of effective stresses the cavity expansion analogue for the CPT, which
at the same void ratio. Scaling of the soil to the CPTu then suggests that it should use the pore pressure at
diameter is a complication and it is typical to check the u1 location, not the accepted standard procedure
the five or so centrifuge tests with a couple of full of measuring at the shoulder u2 location. Thus, a map-
scale calibration chamber tests. ping needs to be introduced for u1/u2. At present the
Other researchers are investigating whether an Peuchen et al. (2010) relationship is used.
even smaller CPTu test in a triaxial specimen with
good control of boundary stresses might be an alter- 5.2.3 Improved precision: pore pressure and shear
native, but this is a relatively recent development modulus
(Damavandi-Monfared & Sadrekarimi, 2015). Tailings may be under-consolidated because of the
placement rate. It is also common to find the TSF to
5.2.2 Data reduction in silt be under-drained with a consequent downward hy-
Although sands are encountered in tailings, nearly all draulic gradient (at least in the longer term). These
TSF will have some silts and, possibly, even be dom- two factors dictate that the piezometric regime must
inated by silts. However there are no properly carried be measured during any field investigation. A con-
out calibration chamber studies for the CPTu in silt. venient way of doing this is to carry out dissipation
tests at a few rod-changes for each CPTu sounding, in

49
point for any earthquake ground response study as
Gra Sand size
Medium Fine Silt Clay well.
100 vel Coarse

80 5.3 Application to TCS sand and TCB silt


Percentage finer than

TCS sand (180/22) Representative samples of both of the tailings materi-


60 tailings
als were tested to determine their properties for the
TCB silt (70/51)
tailings CPTu interpretation, essentially k, m or kʹ, mʹ. The
40
Silt / slimes (Fig 9 and
general scope and procedures for laboratory testing
10) are left to the following section. Figure 12 shows that
20
Thickened tailings (Fig TCS sand is indeed a silty sand with a D50 of 180m
9 and 10) and 22% fines, while TCB silt contains more than
0
50% silt (D50 = 70m and 51% fines). There is excel-
10.00 0.10 1.00
0.01 0.00
Grain size (mm) lent definition of the CSL in both cases (Figure 13),
Figure 12: Grain size distributions of TCS Sand (180/22) tailings with the usual semi-log idealization being adequate
and TCB silt (70/51) tailings, as well as thickened tailings and for confining stresses less than about 2 MPa.
slimes on Figure 10. An interesting aspect of these tests is the properties
of these CSL, with 10=0.115 for TCS sand (22%
0.8
TCS Sand (180/22) fines) and 10=0.086 for TCB silty sample (51%
Tailings fines). The CSL slope 10 is directly comparable to
Γ1 = 0.914
0.7 λ10 = 0.115
the familiar compression index Cc used with clays. It
then follows that the higher-fines TCB silt is actually
Void ratio

Mtx = 1.45
less compressible than the TCS sand, in the case of
0.6 these two tailings gradations. This is an important ob-
servation that illustrates the significant limitations of
0.5 relying on fines content as a proxy for compressibility
in the interpretation of CPTu results. The additional
fines content fills up more of the pore space providing
0.4 less opportunity for particle movement during shear
0.7 TCB Silt (note that the CSL of the TCB silt is also at a lower
(70/51) Tailings void ratio than the TCS sand with 22% fines).
Γ1 = 0.713
0.6 λ10 = 0.086
Table 2 summarizes the parameters developed for
Mtx = 1.44 the full interpretation of the in situ state. The average
Void ratio

0.5
in situ state of both materials based on the CPTu turns
out to be about = 0.02. Now we have measured
the in situ  we only need the soil behaviour at that
0.4 CSL
state.
CS (end of undrained tests)
CS (from drained tests)
Figure 14 shows the corresponding stress-strain
0.3 behaviour of the drained and undrained specimens
10 100 1000 10000 closest to this state for each material. Their behaviour
Mean effective stress, p', kPa is comparable, as it should be at the same . There is
some strain softening, and large strains, but maybe
Figure 13: Critical state lines for TCS sand (180/22) and TCB not the very brittle behaviour that may be expected in
sandy silt (70/51) tailings. Black lines represent state paths – sands without fines.
horizontal lines are undrained tests while hooked lines are
drained tests.
5.4 Soil samples and laboratory testing
essence measuring both horizontal coefficient of con- Moving ahead to a full assessment of the behaviour
solidation (ch) in situ and the current pore pressure of the soil for both static and earthquake loading
(uo) using the CPTu pore pressure sensor. needs samples in addition to those needed for soil in-
The elastic shear modulus Gmax was discussed ear- dex testing and CSL determination. These do not
lier in the context of k, m values in (2) but it is also a have to be undisturbed as soil properties can be meas-
sensitive indicator of fabric as it captures how the par- ured on reconstituted samples. Tests on undisturbed
ticle contacts can transmit elastic shear waves. We would be misleading as to in situ behaviour because
should always include at least some seismic CPTu they are practically impossible with silts (see Figure
soundings in any field program; about 1 test in 5 2).
seems sufficient to greatly add to the understanding Most CPTu equipment is now available with a
of any site. A Gmax profile is, of course, the starting form of piston sampler (MOSTAP or similar) that is
deployed using the CPTu system itself. The sampler

50
1500 1500
Silty sand (180/22)
In situ ψ approx. ‐0.02
Test CID T2
CID T2 (silty sand 180/22)
= 0.617
p'o = 345 kPa
Deviator stress , q (kPa)

ψ = ‐0.005 Sandy silt (70/51) CIU T3 (silty sand 180/22)


1000 Test CID T7 1000
eo = 0.446 CID T7 (sandy silt 70/51)

q, (kPa)
p'o = 240 kPa
ψ = ‐0.062 CIU T7 (sandy silt 70/51)

CSL
500 500
Sandy silt (70/51) Silty sand (180/22)
Test CIU T7 Test CIU T3
eo = 0.477 eo = 0.615
p'o = 1016 kPa p'o = 825 kPa
ψ = 0.023 ψ = + 0.036
0 0
0 5 10 15 20 0 500 1000 1500 2000
Mean effective stress, p' (kPa)
Axial strain (%)
0.70
2.0
0.65

1.5 0.60
Stress ratio, η

Void ratio
0.55
180/22
1.0
0.50 CID T2 (silty sand 180/22)
CIU T3 (silty sand 180/22)
0.45 CID T7 (sandy silt 70/51)
0.5 CIU T7 (sandy silt 70/51) 70/51

0.40 CSL

0.0 0.35
0 5 10 15 20 1 10 100 1000 10000
Mean effective stress, p' (kPa)
Axial strain (%)

Figure 14: Stress strain curves for drained and undrained triaxial tests on TCS sand and TCB silt tailings at the estimated
average in situ state ψ ~ -0.02
is pushed to the test depth and then opened to recover 2 Data on the cyclic behaviour as it seems inevitable
the soil in the target zone. Sampling does not need to that high-value mines will be in earthquake prone
be at every CPTu sounding, but once the characteris- area; and,
tic soils have been identified, sufficient samples need 3 Index tests to document the soils tested.
to be recovered to characterize the variability of the
This testing is really quite modest (and certainly in
materials (grain size). Then you need to look at CPTu
the context of the recent tailings failures) but is suffi-
in conjunction with the grain sizes. How many mate-
cient to allow us to define the material behaviour as a
rials do you really have? Often you can boil it down
function of the in situ state ψ.
to just two or three characteristic gradations.
Assuming a semi-log CSL is used (and this will
This provides the basis for laboratory testing on
often be adequate) there are only five soil properties:
disturbed, blended samples that provide a base of
 which define the CSL; M, N, which define the
measuring engineering properties that support the in
stress-dilatancy behaviour; and  which defines the
situ test interpretation of the CPTu, with distinction
effect of  on strength. ( defines the slope of the
between different material types. trend line in Figure 3, which varies slightly from one
So, what testing is needed to support the charac- soil to another. This was not considered in 1985).
terization of tailings behaviour? These properties are simply obtained by plotting test
data in the appropriate form. The properties are all di-
mensionless (although  has a convention of being
6 ENGINEERING PROPERTIES defined at 1 kPa) and do not imply any particular con-
6.1 Laboratory test program stitutive model. They capture the fundamental behav-
iour of any particulate material. Table 1 indicates
Characterizing the soils in a TSF requires enough
which tests provide which properties.
testing on each representative material. Table 1 pro-
The discussion so far has been directed to strength,
vides a basic minimum set of tests with the purpose
but generally we will need to know tailings consoli-
of measuring:
dation behaviour. First, consolidation affects the ca-
1 Static strengths, brittleness and so forth and with pacity of the TSF, a business consideration. Second,
sufficient data to calibrate any of the good soil because mine life is commonly about 25 years there
models;

51
during the saturation step and that densification is dif-
Table 1. Suggested laboratory testing program per representa- ficult to measure accurately. This issue of accurate
tive sample of tailings void ratio measurement has plagued CSSM since the
Test type No of Purpose early work of Casagrande in 1935 until it was re-
Tests solved in the mid 1980s during work in the Canadian
Particle size 20 Define heterogeneity of material, Arctic.
distribution identify representative materials
Specific grav- 2 Basic property to calculate void The required technique for accurate void ratio
ity ratio measurement is to shut the drainage and pore pressure
Max. & min. 2 Not part of CSL framework, but measurement lines on the specimen immediately
density helpful to laboratory technicians shearing is terminated. The cell is then depressurized
for sample preparation before moving the entire specimen at its end-of-test
Triaxial 5-8 Define CSL, undrained strength,
consolidated brittleness ()
water content to a freezer. A few hours of freezing is
undrained sufficient to allow the specimen to be demounted
Triaxial 5–8 Define CSL, stress-dilatancy,, from the test equipment which had to be frozen
consolidated state-dilatancy (M, N,). Also (mainly the platens) without loss of any water. The
drained provide the basic stress-strain water content of the entire sample is then measured
data for calibrating constitutive by standard oven drying, and converted to void ratio
models.
Oedome- 3-5 Consolidation behaviour (Cc, cv) using the measured Gs (the void ratio is accurate,
ter/Rowe Cell since the test procedure was to saturate the sample
Bender element 2 sets Measure Gmax as function of with back pressure at the start of the test).
tests of 8 stress level at 2 initial void ratios. The costs and effort of this freezing technique is
Measure consolidation curves minimal. It is simply a question of taking care to make
Cyclic simple 8 Two sets of 4 tests. Each set at
shear tests same ψ. Include post-liquefaction
sure it is done.
settlement if possible.
Resonant col- 2 Optional, but avoids reliance on
umn testing published curves for G and Dr
6.3 Soil properties from laboratory tests
The soil properties  are independent of
any constitutive model but, on their own, will usually
will be substantial strength gain in this time. Thus not be quite sufficient to capture soil behaviour. The
some oedometer or Rowe Cell tests are needed. missing property is some measure of plastic shear
The elastic soil stiffness is also important, both for stiffness analogous to Gmax for elasticity. The concept
improved precision in determining  from CPTu and of plastic shear hardening may be common, but there
for seismic response assessments. We have found it is no agreed standard soil property for plastic modu-
is helpful to use bender elements within a set of triax- lus and each constitutive model generally defines its
ial consolidation tests. This defines the relationship own property. The approach we use to determine
between Gmax, void ratio and stress level better than plastic modulus from the triaxial testing is Iterative
the field measurement of Gmax, since we have a better Forward Modelling (IFM) in which the soil behav-
measurement of void ratio in the laboratory sample iour in a test is computed using the chosen constitu-
than in the field. It also defines consolidation proper- tive model and estimated properties, with the com-
ties of the tailings. puted behaviour being visually compared to that
In the case of static strength and stiffness there are measured. Then, the plastic hardening is revised to
several good models that will closely predict the improve the fit and the process repeats until a good fit
soil’s behaviour from the tests discussed. This is, as is obtained. This can all be done in Excel as any good
yet, not true for cyclic softening during earthquakes, model is readily programmed using the VBA envi-
thus cyclic simple shear tests are needed for seismic ronment that is included with Excel. The modelling
work on tailings, and these are discussed in more de- also has the advantage of rapidly showing up ques-
tail in Section 6.4. tionable laboratory results as well as confirming the
appropriate plastic hardening. The earlier Figure 4
was developed just this way.
6.2 Accurate void ratio measurement
In the case of the NorSand model, plastic harden-
Geotechnical laboratories are familiar with accurate ing is represented by the dimensionless modulus H.
measurement of stress and strain. However, the es- This modulus is proportional to the plastic hardening
sence of CSSM is quantifying the effect of void ratio of Cam Clay with H ~ 1/(). The IFM procedure
on that stress-strain behaviour, which requires accu- generates the best-value of H for the entire test, and it
rate measurement of void ratio during laboratory tri- is common to find an effect of  so that H = a + b
axial testing. And this causes a problem as the stand- where a, b become the true soil properties. It is also
ard procedures of most laboratories are inadequate for common to find that H scales with Gmax/p, which is
accurate void ratio measurement despite ISO 9000 analogous to a constant  ratio often used in Cam
accreditation. The problem is that samples densify Clay.

52
shear specimens of tailings silts in Shelby tubes, in-
Table 2. Results of laboratory testing for CPTu interpretation cluding TCS sand and TCBy silt data
q shown( )
and soil behaviour
TCS sand TCB silt
D50 (mm) 180 70
Fines (%) 22 51
Γ1 0.914 0.713
λ10 0.115 0.086
Mtx 1.45 1.44
χtc 3.5 3.5
N 0.2 0.2
H 50 140
Gmax 50 32
S 160 160
n 0.65 0.65
v 0.25 0.25
H/Ir (Ir = G/pʹ) 0.74 3.22
Note: Ko = 0.7 assumed for calculation of mean effective stress

it is common to find an effect of  so that H = a + Figure 15: Cyclic resistance ratio in simple shear for tailings
b where a, b become the true soil properties. It is
also common to find that H scales with Gmax/p, which
is analogous to a constant  ratio often used in Cam on Figure 15. Once we get these samples into a labor-
Clay. atory, extruded and reconsolidated, we like to think
Table 2 shows a full set of properties measured on we have a sample at the in situ state with a semblance
TCS sand and TCB silt materials identified from the of in situ fabric. However, the void ratio will likely be
CPTu sounding. rather different, and we cannot assume the laboratory
sample will be the same as in situ, because that would
tend to overestimate the cyclic strength. We therefore
6.4 Cyclic Resistance Curves have two options:
A cyclic shear resistance curve is needed for a total 1 Do not try to obtain undisturbed samples, and pro-
stress analysis of seismic loading (whereas a pore duce strength curves from reconstituted samples at
pressure generation function is needed for an effec- the correct range of in situ states. This will likely
tive stress analysis, which is not covered here). For a be a low estimate of the cyclic strength because
cyclic resistance curve, the engineer needs data over field fabric would be lost.
the range of in situ states encountered for each iden- 2 Test the undisturbed samples consolidated back to
tified material. This likely means at least two curves the in situ stress level, but then make an adjustment
of cyclic shear stress ratio (τ/σʹv) versus number of to account for the post-sampling void ratio change.
cycles to liquefaction (NL). A liquefaction strength
(stress ratio) can be picked for a number of cycles of In order to make the correction to the strength of the
loading from each curve to provide a “strength vs undisturbed sample, you need two pieces of infor-
state” relationship. mation. First information is an estimate of the reduc-
Cyclic simple shear tests are preferred, as they are tion in void ratio during sample handling and consol-
most representative of seismic loading. (If only cyclic idation. The second piece of information is how much
triaxial testing is available, a reduction of about 0.65 this void ratio reduction affects the value of ψ and
on the cyclic shear stress ratio may be needed to “cor- how this change in ψ affects the cyclic resistance. The
rect” for simple shear conditions.) Figure 15 shows actual procedure used is not important, but it does
cyclic resistance data for five tailings materials, in- need to be done.
cluding TCS sand, TCB silt and a couple of pure silts. The approach used for TCS sand and TCS silt is
The results cover a wide range of conditions, but in- illustrated on Figure 16. We first used the available
terestingly fall within a rather narrow band, which at data to show CSR10 (τ/σʹv at which liquefaction oc-
this stage is likely coincidental rather than a reliable curs in 10 cycles) variation with ψ. This needed cyclic
trend. triaxial data from earlier work, adjusted to simple
The key to cyclic simple shear testing is to have shear conditions. Our simple shear tests on undis-
tests at the in situ state, or at least to be able to adjust turbed samples were at ψ = -0.06 with CSR10 of
the cyclic resistance to the in situ state. We have been 0.145. The characteristic in situ state to use for lique-
able to achieve something close to undisturbed simple faction is ψ = 0 (this is looser than the average in situ

53
ψ) so the adjusted CSR10 is 0.125, as illustrated on cyclic behaviour and stability of the existing and fu-
Figure 16. ture geometries.
All of the above looks satisfactory. Yet, as an in-
dustry, we have seen three tailings dam failures in the
last decade each with huge environmental and human
Golder CSR10 ‐ TCS 180/22 tailings consequences. The news media likes to portray these
0.3 dam failures as careless (verging on negligent) man-
Cyclic Txl adjusted to CSS (old data)
agement or shareholder greed. But, each of the three
Compacted adjusted cyclic txl (old data) big failures had geotechnical engineers involved and
0.2 at least two of the three had engineering review. How
CSR10

is a mine manager supposed to know who is compe-


tent? Each firm proposing on a TSF design or assess-
0.1 ment will claim to be “world class”. The problem is
In situ characteristic
not the mine manager or shareholders. The problem
state ψk ~ 0, adjusted is the standard of engineering.
CSR10 = 0.125 An issue is the continued propagation of ideas
0.0 based on “fines content corrections”. It should be
‐0.1 0.0 0.1 0.2 self-evident that fines content is nonsense when deal-
State parameter 
ing with tailings, as pretty much every TSF will have
Figure 16: Correction to cyclic stress ratio for difference be-
tween state in laboratory tests and estimated characteristic in situ fines content higher than found in research sand, and
state. many of the tailings may comprise just fines. The
properties used for engineering could be entirely “the
correction” rather than “the data”. Further, it will be-
7 ANALYSES come quickly apparent when tailings are tested, as il-
lustrated with the example of this paper, that mechan-
This paper is not about the analyses, except to make ical properties do not correlate well with “fines”. No
the point that the groundwork has been laid for what- such trouble exists at all if the tailings behaviour is
ever analyses you want to do. You have done a thor- viewed within the framework of CSSM. The CSLs
ough investigation of the TSF, and based on the CPTu are measurable, other properties are familiar, and the
and index testing have a sound knowledge of how details of the stress-strain behaviour can be modelled
many different materials there are. Thereafter, you in a spreadsheet. The required in situ  can be reliably
carried out a series of laboratory tests (mainly testing computed from CPTu data. All the aspects for sound
to determine the CSL and other properties) on each engineering exist.
material type so that you can determine the in situ ψ We have already drawn attention to the parallels of
of your key materials. You may also have measured Fundao with the concerns of the Corps of Engineers
the in situ Gmax to enhance the CPTu analysis. eighty years earlier. It is the lack of teaching of criti-
Once you have characterized the in situ ψ, your la- cal state concepts to geotechnical engineers and ap-
boratory testing then focused on the engineering be- plication of rigorous mechanical approaches that has
haviour you need. This is generally the static strength caused this problem. If you look at the Wikipedia cri-
and the cyclic resistance ratio curve, in simple shear, tique of CSSM you will find the statement “Prof.
over the range of ψ (meaning density and stress level) Alan Bishop at Imperial College used to routinely
expected in situ. One warning, though, is that you demonstrate the inability of these theories to match
need to consider the effects of fabric and/or sample the stress-strain curves of real soils”. This is a true
disturbance on your results. Supplementary ap- statement, but what is not stated is that Bishop
proaches may include pore pressure models, shear brought Prof. Peter Wroth to Imperial College to en-
modulus reduction and post-liquefaction settlements. sure each M.Sc. course gained a thorough grounding
in the principles of CSSM. Bishop was well-aware
that CSSM was the only coherent framework for soil
8 CLOSURE behaviour (and indeed CSSM is based on work at Im-
perial College as much as Cambridge and elsewhere).
The paper has briefly overviewed determining the in The objections summarized on Wikipedia reflect
situ state of the TSF with a CPTu, recognizing that 1975 and they have since faded into the background,
there is often more than one way to make the inter- as can be seen by the excellent match of critical state
pretation. We also need to think carefully how many theory to the spectrum of real soil shown on Figure 4.
material types there are for us to consider. Once we What remains a concern is that the situation stated
have the mechanical parameters for the TSF, in this in Wikipedia probably does reflect current geo-tech-
case for two materials, we carry out the careful anal- nical teaching attitudes in North America and else-
yses for the whole facility. That includes the overall where. It is why instances such as Fundao will con-
tinue to occur. On the bright side, Hugh Golder, who

54
was one of the founders of Geotechnique, instilled and Calibration, 204-236, eds. J.A. Yamamuro and V.N.
strong values of sharing within Golder Associates. Kaliakin.
Jefferies, M. & Been, K. (2015). Soil liquefaction: a critical
The Golder Foundation has been established to reach state approach, 2nd ed. CRC Press.
outside the company as well. We make our data and Manzari, M.T., & Dafalias, Y.F. (1997). A critical state two-
models freely available under the GNU Open Soft- surface plasticity model for sands. Géotechnique 47, 255-
ware license. Do visit www.golderfoundation.org/ 272.
and then do get up to speed with CSSM. Of course Mohajeri, M. & Ghafghazi, M. (2012). Ground sampling and la-
reading “Soil Liquefaction: a critical state approach” boratory testing on low plasticity clays. Proc. 15th World
Conference on Earthquake Engineering, Lisbon.
may also help. Plewes H.D., Davies M.P., & Jefferies M.G. (1992). CPT based
screening procedure for evaluating liquefaction susceptibil-
ity. Proc 45th Canadian Geotechnical Conference, Toronto.
9 ACKNOWLEDGEMENT Peuchen, J, Vanden Berghe, J.F, & Coulais, C. (2010). Estima-
tion of u1/u2 conversion factor for piezocone. CPT'10, 2nd
International Symposium on Cone Penetration Testing, Pa-
Mike Jefferies has been a tremendous help in pulling per 1-03.
this paper together with me. Many of our colleagues Reid, D. (2015). Estimating slope of critical state line from cone
at Golder Associates have used Mike and my brain penetration test – an update. Canadian Geotechnical Jour-
waves on CSSM, the state parameter and tailings be- nal, 52, 46 - 57
haviour over the years. The engineers we have Robertson, P.K. (2008). Discussion: Liquefaction potential of
worked with, and who have contributed significantly silts from CPTu. Canadian Geotechnical Journal, 45, 1,
140-141.
in their own right, are too numerous to recognize in- Robertson, P.K. (2009). Interpretation of cone penetration tests
dividually but I acknowledgement all their input. Da- – a unified approach. Canadian Geotechnical Journal, 46,
vid Reid specifically has carried out good research on 1337-1355
the issues and provided input and review of this pa- Robertson, P.K. (2010). Evaluation of flow liquefaction and liq-
per. uefied strength using the cone penetration test. Journal of
Geotechnical and Geoenvironmental Engineering, ASCE,
136, 6, 842-853.
Robertson, P.K. (2012). Evaluating flow (static) liquefaction us-
10 REFERENCES ing the CPT: an update. Proceedings Tailings and Mine
Waste ’12, 14–17 October, Keystone, Co., USA.
Been, K. & Jefferies, M.G. (1985). A state parameter for sands. Schofield, A. & Wroth, C.P. (1968). Critical State Soil Mechan-
Géotechnique, 35, 2, 99-112. ics. London, McGraw-Hill.
Been, K, Crooks, J.H.A., Becker, D.E & Jefferies M.G. (1986). Shuttle, D.A., and Jefferies, M.G. (1998). Dimensionless and
The cone penetration test in sands: Part 1, State parameter unbiased CPT interpretation in sand. International Journal
interpretation. Géotechnique, 36, 2, 239-249. of Numerical and Analytical Methods in Geomechanics, 22,
Been K., Jefferies M.G., Crooks J.H.A. & Rothenberg L. (1987). 351-391.
The cone penetration test in sands: Part II, general inference Shuttle, D.A. & Cunning, J. (2007). Liquefaction Potential of
of state. Géotechnique, 37, 3, 285-299. Silts from CPTu. Canadian Geotechnical Journal, 44, 1, 1-
Been K., Crooks J.H.A. & Jefferies M.G. (1988). Interpretation 19.
of material state from CPT in sands and clays. Proc. Conf. Shuttle, D.A. & Cunning, J. (2008). Reply to discussion: Lique-
Penetration Testing, 215-218. ISBN 0 7277 1377 9, pbl faction Potential of Silts from CPTu. Canadian Geotech-
Thomas Telford. nical Journal, 45, 1, 142-145.
Been, K., Jefferies, M.G. & Hachey, J.E. (1991). The critical Sladen, J.A. (1989). Problems with interpretation of sand state
state of sands. Géotechnique, 41, 3, 365-381. from cone penetration test. Géotechnique, 39, 2, 323-332.
Been, K, Romero, S., Obermeyer, J and Hebeler, G. (2012). De-
termining the in situ state of silt and sand tailings with the
CPT. Proc. Tailings and Mine Waste ’12, 14–17 October,
Keystone, Co., USA. pp. 325-333
Bolton, M.D., Gui, M. W., Garnier, J., Corte, J. F., Bagge, G.,
Laue, J. & Renzi, R. (1999). Centrifuge cone penetration
tests in sand. Geotechnique, 49, 4, 543-552.
Damavandi-Monfared, S. & Sadrekarimi, A. (2015). Develop-
ment of a miniature cone penetrometer for calibration cham-
ber testing. Geotechnical Testing Journal, 38, 6, 878-892.
Drucker, D.C., Gibson, R.E. & Henkel, D.J. (1957). Soil me-
chanics and work hardening theories of plasticity. Transac-
tions American Society of Civil Engineers, 122, 338-346.
Houlsby, G.T. (1988). Session Leader’s Introduction, Proc.
Conference on Penetration Testing in the U.K., Birmingham,
Thomas Telford, 141-146.
Jefferies, M.G. (1993). NorSand: a simple critical state model
for sand. Géotechnique 43, 91-103.
Jefferies, M.G. & D.A. Shuttle (2005). NorSand: features, cali-
bration and use. ASCE Geotechnical Special Publication
No. 128, Soil Constitutive Models: Evaluation, Selection,

55

You might also like