1446 XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3
the UMAP space are consistent with the observations of
alteration and mineralization associations observed in the
drill core. The seven groups provide a useful tool for ore
type description, sample selection, and interpretation of
sulphide domains in three dimensions.
A Data-Driven Multivariate Sample Selection
Methodology
The UMAP dimension reduction and K-means group-
ing analysis provide a multivariate view of the data with
which the question of multivariate representativity can
now be considered in a convenient and reproducible man-
ner. Figure 3 shows two possible selections of metallurgi-
cal samples plotted in UMAP space against the geological
composite data. The selections provide reasonably good
multivariate coverage of the seven identified mineralization
groups but there are notable gaps, illustrated by the nine
larger green dots.
We evaluated the composition of the gaps by statisti-
cal analysis of the drill hole samples proximal to the gaps,
identified using their UMAP coordinates. The boxplots in
Figure 4 show that the compositional ranges of the gaps are
sufficiently narrow, in most cases, to form specific, practical
targets for further metallurgical sample selection.
How Many Samples Do We Need?—Rules Of Thumb
When faced with deciding the minimum number of sam-
ples required for metallurgical test work, an important
question is “what are the results to be used for?” There are
two primary applications in the geometallurgical space.
First is for the purpose of process design and equipment
selection. The second is for the development of a geometal-
lurgical block model.
For process design and sizing of equipment, the primary
objective is to design a series of unit processes to handle
ore with average characteristics, at a specified throughput
rate. It is also necessary to make sure the full range of ore
characteristics is tested and is within the capability of the
proposed design. The design outcome is a compromise.
By way of illustration, consider the process (simplified)
to size a SAG mill. Samples are selected for testing on the
assumption they represent the range of ore hardness (or
more correctly the specific energies) to achieve the required
size reduction and throughput. The samples are tested
(drop weight tests, SAG mill comminution tests, etc) and
the 85th percentile measurement is (usually) used as the
basis of design. The outcome is a mill that performs well at
the nominated specific energy and slightly less well (lower
throughput or larger P80) at higher energies that comprise
15% or less of the ore, and slightly better at lower energies.
Figure 3. Scatter plots showing metallurgical samples and gaps in UMAP space
the UMAP space are consistent with the observations of
alteration and mineralization associations observed in the
drill core. The seven groups provide a useful tool for ore
type description, sample selection, and interpretation of
sulphide domains in three dimensions.
A Data-Driven Multivariate Sample Selection
Methodology
The UMAP dimension reduction and K-means group-
ing analysis provide a multivariate view of the data with
which the question of multivariate representativity can
now be considered in a convenient and reproducible man-
ner. Figure 3 shows two possible selections of metallurgi-
cal samples plotted in UMAP space against the geological
composite data. The selections provide reasonably good
multivariate coverage of the seven identified mineralization
groups but there are notable gaps, illustrated by the nine
larger green dots.
We evaluated the composition of the gaps by statisti-
cal analysis of the drill hole samples proximal to the gaps,
identified using their UMAP coordinates. The boxplots in
Figure 4 show that the compositional ranges of the gaps are
sufficiently narrow, in most cases, to form specific, practical
targets for further metallurgical sample selection.
How Many Samples Do We Need?—Rules Of Thumb
When faced with deciding the minimum number of sam-
ples required for metallurgical test work, an important
question is “what are the results to be used for?” There are
two primary applications in the geometallurgical space.
First is for the purpose of process design and equipment
selection. The second is for the development of a geometal-
lurgical block model.
For process design and sizing of equipment, the primary
objective is to design a series of unit processes to handle
ore with average characteristics, at a specified throughput
rate. It is also necessary to make sure the full range of ore
characteristics is tested and is within the capability of the
proposed design. The design outcome is a compromise.
By way of illustration, consider the process (simplified)
to size a SAG mill. Samples are selected for testing on the
assumption they represent the range of ore hardness (or
more correctly the specific energies) to achieve the required
size reduction and throughput. The samples are tested
(drop weight tests, SAG mill comminution tests, etc) and
the 85th percentile measurement is (usually) used as the
basis of design. The outcome is a mill that performs well at
the nominated specific energy and slightly less well (lower
throughput or larger P80) at higher energies that comprise
15% or less of the ore, and slightly better at lower energies.
Figure 3. Scatter plots showing metallurgical samples and gaps in UMAP space