Actes du colloque - Volume 2 - page 594

1473
Technical Committee 203 /
Comité technique 203
physical characteristics of the tailings are likely to be
incremental, meaning that appropriate correlations can be
regularly updated.
Given that the maximum depth of testing achievable with the
PANDA is about 7m, there is a question about the applicability
of the method to large, deep tailings deposits. Although the 7m
depth restriction is certainly a limiting factor for existing
deposits deeper than 7m, the PANDA technique can be used in
conjunction with conventional compression testing to predict
the tailings state for future, deep deposits of tailings, as
explained below.
4 USING PANDA DATA TO PREDICT FUTURE
INSTABILITY RISKS
Aside from the material beneath the outer slopes, all the
tailings in a TSF will be subjected to essentially one
dimensional compression. Conventional oedometer tests can
therefore be used to simulate the likely change in void ratio that
will occur when a TSF is built to full height. If PANDA tests
are carried out on the initial layers of tailings (before they reach
a depth of 7m or more), the initial state will be well defined (as
long as appropriate correlations have been established) and
reasonable predictions may be made as to how the state will
change as the TSF is constructed.
This idea was argued in some detail by Park and Byrne
(2004), who used data from compression tests on eight different
sands and derived an expression relating relative density D
r
and
vertical effective stress σ'
v
through:
where P
a
is atmospheric pressure, D
r0
is the initial relative
density and α is a parameter that is a function of the maximum
and minimum void ratios and a sand stiffness number C that is
independent of void ratio.
As relative density is directly related to void ratio, the above
equation makes it possible to predict the void ratio at any depth
in a TSF profile at some time in the future, if accurate estimates
of starting in-situ values are available. It is suggested that these
initial values may be obtained using PANDA penetrometer
tests. During the initial phases of tailings deposition, it should
be possible to carry out a number of field testing campaigns at
particular locations, obtaining information over a number of
years. These data can be used to verify that the PANDA data
from one campaign to another are consistent with the
predictions made using oedometer data.
These tests should be complemented with triaxial or simple
shear tests to define the relationship between void ratio and
effective stress (the Steady State Line discussed by Jefferies and
Been (2006)), amongst others. Integration of these various data
will provide a consistent methodology for estimating the
liquefaction susceptibility of a TSF both at the time PANDA
tests are undertaken, and in the future.
The potential value of using conventional oedometer tests to
predict future relative density values (knowing initial in-situ
state) can be illustrated by consideration of data presented by
Park and Byrne (2004). They showed that different sandy
materials placed at the same initial relative density will not
necessarily consolidate (compress) to the same relative density,
even if subjected to the same overburden stress. This might be
counter-intuitive, as it is commonly considered that tailings near
the bottom of a TSF is much less prone to liquefaction than
tailings near the top (where the confining stress is lower).
However, data presented by Park and Byrne (2004) for Brasted
sand placed at a relative density of 50% compressed to a final
relative density value of only 57% under a vertical confining
stress of 1000kPa. A relative density threshold of around 60% is
often considered a reasonable first-pass estimate of the
boundary between potentially liquefiable and non-liquefiable
tailings, implying that the Brasted sand quoted above might still
be susceptible to liquefaction at a depth of around 50m to 60m.
Other data presented by Park and Byrne (2004) showed results
that are more consistent with current expectations. Tests on
Quiou sand placed at a relative density of 50% compressed to a
value of 80% under a vertical effective stress of 1000kpa. This
value of relative density is highly likely to render the Quiou
sand non-liquefiable at higher confining stresses, as is expected
using current concepts.
A key factor in the above discussion is the relative slopes of
the oedometer compression line and the Steady State Line for a
particular tailings. If they are equal for example, then tailings
placed at low relative density (which renders it susceptible to
liquefaction) would remain so, even under high confining
stresses. Unfortunately the work of Park and Byrne (2004) did
not investigate the response of the sands to shear (they utilised
data from the literature), so it is not possible to make these
comparisons for their data. However, data of this type is likely
available in many consultants’ internal databases, and
interogation of this data could prove extremely valuable.
5 POTENTIAL LIMITATIONS TO USE OF PANDA
PENETROMETER TESTING
Use of the PANDA penetrometer for routine testing of the state
of tailings is now commonplace in Chile, where it has in fact
effectively been written into legislastion. Recently revised
Chilean legislation governing the construction and operation of
TSFs specifically mentions the PANDA technique as one of the
preferred approaches for density control on TSFs. Given the
increasing awareness of the critical importance of correctly
controlling density, and the increasing acceptance by regulators
of the approach, it is important to consider potential limitations
of the technique before advocating its widespread adoption.
5.1 Site-specific correlations
As shown in Figure 1, the relationship between relative density
and penetration resistance varies with the material tested. This
in itself is not a major problem; it simply requires that adequate
correlations be established, with the obvious techniques being
calibration chambers or in-situ correlations in which the field
density is measured using techniques such as the sand
replacement method. A potential limitation is that the tailings
produced by a particular mine may change with time, as milling
rates change or the nature of the ore being mined changes.
Established correlations might then no longer be valid.
However, through index tests such as particle size distribution
tests, it is possible to monitor such changes in the nature of
tailings being produced, and simply carry out new correlation
studies. It will only be once experience is gained at a particular
operation that the required frequency of these re-calibrations
will become apparent.
5.2
Effect of moisture content
Results from a preliminary set of tests using a 0.5m diameter,
0.75m deep calibration chamber are summarised in Table 1. The
tests were carried out on a medium sand having a d
50
of 280μm.
Table 1. Relationship between water content and PANDA resistance for
medium sand prepared at 60% relative density.
a
Water content (%)
0
4
8
12
25
Resistance (MPa)
0.66
3.56
2.37
1.77
1.34
The very low resistance at zero water content is essentially
irrelevant for the application under discussion, as all tailings are
placed in either a fully saturated, or a moist state. All large
mining operations in Chile now utilise the downstream method
of construction, in which the tailings stream is split into a
coarser fraction (the underflow), used for construction of a
1...,584,585,586,587,588,589,590,591,592,593 595,596,597,598,599,600,601,602,603,604,...913