1028 XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3
All these methods have merits and, when combined, can
indicate a breakage mechanism and target grind sizes based
on the breakage response and not simply grain size distribu-
tion (Little et al., 2016). However, these techniques require
significant data interpretation and cannot be applied read-
ily to different datasets.
Vizcarra (2010) carried out extensive studies on three
different ores to determine whether any change in libera-
tion was obtained using different comminution methods.
Using laboratory piston-die and hammer mill breakage
devices, the author concluded that there was no statistically
significant difference in the liberation of the product for
any of the ores. Yet, there was a difference in shape, with
the piston-die producing more angular particles. Liberation
was evaluated using the mass fraction of a given mineral
in a given size fraction for particles where the mineral of
interest comprised over 80 or 90% of the particle. Some
observed differences in one dataset (i.e., an ore from the
Northparkes mine in New South Wales, Australia) for
bornite were found not to be statistically significant, due,
in part, to the low number of bornite grains, particularly in
the coarser fractions.
One of the limitations of using only the mass fraction
in a certain liberation class is that it does not say how that
mass is distributed within that size class. For instance, the
same mass fraction could be obtained by one large particle
or several smaller particles. It is valuable, therefore, to con-
sider also the number fraction in a given liberation class.
In this paper, we present a new approach to surface
liberation interpretation. Unlike traditional liberation
analysis, we propose employing the number fraction, con-
sidering both liberated and poorly liberated particles. This
innovative technique provides a refined efficiency metric,
enabling the comparison of different comminution circuits
for better processing planning.
METHODOLOGY
In order to showcase the capabilities of this methodol-
ogy, a high-grade iron oxide ore sample was selected. Its
iron content was 58.5%. The main mineral phases were
goethite (FeO(OH)), hematite (Fe2O3), and kaolinite
(Al2Si2O5(OH)). A summary of the mineral composition
is shown in Table 1.
The ore was milled with a 174 mm height by 127 mm
diameter laboratory-scale drum mill using ceramic balls as
grinding media. The main purpose was to produce a particle
size distribution (PSD) that would mimic that of a typical
comminution plant. The PSD of the sample can be found
in Figure 1. Only the size fractions between 1–10 mm were
used for this analysis. It is also important to mention that
in this study, iron ore is considered as the valuable material
and kaolinite as the gangue. Although, this analysis can be
applied to any ore sample, regardless of its composition and
mineral content.
Mineral Identification and Micro-CT Analysis
Samples were sieved and separated into narrow size fractions,
namely +1–3, +3–4, +4–6-, and +6–10 mm. Representative
samples of about 10 g of each size fractions were collected
for analysis. X-ray fluorescence (XRF) and loss on igni-
tion (LOI) testing was used to estimate the composition of
the sample, which can be found in Table 1. Subsequently,
samples were packed into small glass containers of 30 mm
height by 15 mm diameter. A Nanotom S microtomogra-
phy scanner by Waygate Technologies was used to scan each
container at 120 keV, 310 µA, reaching a linear resolution
of 5 micron per voxel. For each sample, 2001 projections
were acquired using 1 s exposure time with a 1 mm copper
filter. Three images were combined per projection in order
to reduce image artifacts.
The main reason why micro-CT was used rather than
SEM/EDX combined with a mineralogical mapping soft-
ware (such as MLA, QEMSCAN, TIMA, among others)
was to minimise the stereological errors that arise when a
3D object is characterised using a 2D section. For instance,
minerals typically have uneven surfaces. By taking a slice off
the sample and exposing a region (as it would be the case of
SEM/EDX), the exposed surface can only be measured as a
line rather than a surface and it will not be possible to cap-
ture the roughness of such surface (see Figure 2). However,
the micro-CT technique is able to capture the unevenness
of a mineral’s surface. Its non-destructive and non-invasive
nature produces 3D representations of the internal com-
position of the sample, rendering a more accurate quanti-
fication of surface liberation (Reyes et al., 2018, Ueda and
Oki 2020).
Image Processing and Quantification Methodology
A bespoke image processing methodology* was developed
to analyse this data set and generate a quantification library
*A pseudocode for this routine can be made available upon
request.
Table 1. Mineral composition of the ore
Mineral Mass fraction (%)
Goethite 43.1
Hematite 45.0
Kaolinite 7.8
Other gangue minerals 4.1
All these methods have merits and, when combined, can
indicate a breakage mechanism and target grind sizes based
on the breakage response and not simply grain size distribu-
tion (Little et al., 2016). However, these techniques require
significant data interpretation and cannot be applied read-
ily to different datasets.
Vizcarra (2010) carried out extensive studies on three
different ores to determine whether any change in libera-
tion was obtained using different comminution methods.
Using laboratory piston-die and hammer mill breakage
devices, the author concluded that there was no statistically
significant difference in the liberation of the product for
any of the ores. Yet, there was a difference in shape, with
the piston-die producing more angular particles. Liberation
was evaluated using the mass fraction of a given mineral
in a given size fraction for particles where the mineral of
interest comprised over 80 or 90% of the particle. Some
observed differences in one dataset (i.e., an ore from the
Northparkes mine in New South Wales, Australia) for
bornite were found not to be statistically significant, due,
in part, to the low number of bornite grains, particularly in
the coarser fractions.
One of the limitations of using only the mass fraction
in a certain liberation class is that it does not say how that
mass is distributed within that size class. For instance, the
same mass fraction could be obtained by one large particle
or several smaller particles. It is valuable, therefore, to con-
sider also the number fraction in a given liberation class.
In this paper, we present a new approach to surface
liberation interpretation. Unlike traditional liberation
analysis, we propose employing the number fraction, con-
sidering both liberated and poorly liberated particles. This
innovative technique provides a refined efficiency metric,
enabling the comparison of different comminution circuits
for better processing planning.
METHODOLOGY
In order to showcase the capabilities of this methodol-
ogy, a high-grade iron oxide ore sample was selected. Its
iron content was 58.5%. The main mineral phases were
goethite (FeO(OH)), hematite (Fe2O3), and kaolinite
(Al2Si2O5(OH)). A summary of the mineral composition
is shown in Table 1.
The ore was milled with a 174 mm height by 127 mm
diameter laboratory-scale drum mill using ceramic balls as
grinding media. The main purpose was to produce a particle
size distribution (PSD) that would mimic that of a typical
comminution plant. The PSD of the sample can be found
in Figure 1. Only the size fractions between 1–10 mm were
used for this analysis. It is also important to mention that
in this study, iron ore is considered as the valuable material
and kaolinite as the gangue. Although, this analysis can be
applied to any ore sample, regardless of its composition and
mineral content.
Mineral Identification and Micro-CT Analysis
Samples were sieved and separated into narrow size fractions,
namely +1–3, +3–4, +4–6-, and +6–10 mm. Representative
samples of about 10 g of each size fractions were collected
for analysis. X-ray fluorescence (XRF) and loss on igni-
tion (LOI) testing was used to estimate the composition of
the sample, which can be found in Table 1. Subsequently,
samples were packed into small glass containers of 30 mm
height by 15 mm diameter. A Nanotom S microtomogra-
phy scanner by Waygate Technologies was used to scan each
container at 120 keV, 310 µA, reaching a linear resolution
of 5 micron per voxel. For each sample, 2001 projections
were acquired using 1 s exposure time with a 1 mm copper
filter. Three images were combined per projection in order
to reduce image artifacts.
The main reason why micro-CT was used rather than
SEM/EDX combined with a mineralogical mapping soft-
ware (such as MLA, QEMSCAN, TIMA, among others)
was to minimise the stereological errors that arise when a
3D object is characterised using a 2D section. For instance,
minerals typically have uneven surfaces. By taking a slice off
the sample and exposing a region (as it would be the case of
SEM/EDX), the exposed surface can only be measured as a
line rather than a surface and it will not be possible to cap-
ture the roughness of such surface (see Figure 2). However,
the micro-CT technique is able to capture the unevenness
of a mineral’s surface. Its non-destructive and non-invasive
nature produces 3D representations of the internal com-
position of the sample, rendering a more accurate quanti-
fication of surface liberation (Reyes et al., 2018, Ueda and
Oki 2020).
Image Processing and Quantification Methodology
A bespoke image processing methodology* was developed
to analyse this data set and generate a quantification library
*A pseudocode for this routine can be made available upon
request.
Table 1. Mineral composition of the ore
Mineral Mass fraction (%)
Goethite 43.1
Hematite 45.0
Kaolinite 7.8
Other gangue minerals 4.1