XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3 1443
of tens or hundreds of thousands of records which have
the potential to improve understanding and prediction of
metallurgical responses.
A key principle of geometallurgy is to use the geo-
logical sample database to gain leverage from the relatively
small number of high-cost metallurgical tests. Applying
geometallurgical principles to design of metallurgical sam-
pling programmes aligns sparse testwork data to abundant
geological data and ensures that the data is suitable for the
application of data science methods to derive robust predic-
tions of ore processing behaviour at the local scale.
The authors propose that the use of geometallurgical
principles and data science for metallurgical sampling and
test data analysis provides a measurable and repeatable basis
for sample-selection, driven by the geological characteris-
tics of the mineralization. It provides an alternative to the
plethora of rules of thumb sampling procedures that have
limited connection to the geological variability, nor the
variability of the metallurgical responses.
What Type Of Samples Do We Need?
Before any sampling or testwork is commissioned the pur-
pose of the testwork must be clear so that the right sample
type is selected. Drill core is a necessity for comminution
testing and coarse particle beneficiation. For mineral recov-
ery testwork (leaching, flotation, fine particle separation)
uncrushed drill core is preferred so that the sample can be
crushed in stages to produce a particle size distribution that
mimics the process plant design. Coarse rejects from core
samples crushed in a well-controlled laboratory may also be
suitable but Kormos et al (2013) note that assay rejects may
contain excessive fines, so checks on particle size distribu-
tion should be carried out to confirm suitability. Reverse
circulation drill cuttings may contain excessive fines which
may bias the test results.
From a geometallurgical perspective, drill core is
also preferred because much more geological data can be
obtained by visual logging, hyperspectral scanning, geo-
technical measurements, and non-destructive hardness
testing than can be obtained from chip samples. Drill core
data joined with metallurgical test results from the match-
ing depth intervals forms the essential input to predictive
Geometallurgical modelling.
There is an underlying and infrequently articulated
idea regarding representativity. That is, that a representative
sample can fully describe the behaviour of an ore deposit
and provide sufficient information to confirm that the
designed plant will operate in a uniform and predictable
manner. The reality is that such a design will perform in a
manner reflective of the variability of the ore feed over time
and the extremes of the ore feed will perform very differ-
ently from the representative sample. The best scenario is
a plant designed to perform within defined specifications
across all ore types.
Composite samples (typically multiple ore intercepts
from multiple drill holes) are used for developing the basic
processing and testwork flowsheet. Composites are made
from blends of the ore types expected to be mined over
the life of mine or other long production periods. Later,
large composites (bulk samples) may be required to provide
products for further testing such as concentrates or tailings.
Composite samples may establish a feasible process flow-
sheet but they provide little information about the behav-
iour of individual ore types.
To design a process plant that can accommodate all
the variation in ore characteristics, variability sampling is
required. Variability samples should be selected to char-
acterize the full range of geological ore types. The authors
recommend selection of variability samples from continu-
ous intercepts in single drill holes. This approach reduces
the risk of accidentally testing a blend of ore types and
also allows the metallurgical testwork results to be consid-
ered within their geospatial context. However, it should
be recognized that there is no intrinsic causal relationship
between geospatial location and metallurgical behaviour.
The geometallurgical principle for variability sampling
is to identify samples that cover the full range of the relevant
multivariate characteristics of the orebody. ‘Relevant’ refers
to those physical, mineralogical or chemical characteristics
that have the potential to affect ore processing responses.
We define an ore type as mineralized material of eco-
nomic interest with similar ore and gangue mineral assem-
blages, texture, and tenor (grade). In most orebodies, ore
types have a spatial arrangement that is the product of
the geological history of the mineral deposit. The term
‘domain’ is commonly applied to describe a volume of rock
with similar geological characteristics. From a geometallur-
gical perspective, it is preferable that a domain consists of a
single ore type so that, ultimately, estimates of ore process-
ing response derived from testing samples of an individual
ore type can be applied within the domain.
It is common practice to select metallurgical samples
from three-dimensional, geographic locations with the aim
of providing good spatial coverage of the deposit. This is
understandable in the absence of a good knowledge of the
spatial distribution of ore types and domains. However,
three-dimensional coverage is not a logical requirement of
metallurgical sampling. Ore processing response is intrinsi-
cally a function of geology not location. Therefore, from a
geometallurgical perspective, it is more important to achieve
of tens or hundreds of thousands of records which have
the potential to improve understanding and prediction of
metallurgical responses.
A key principle of geometallurgy is to use the geo-
logical sample database to gain leverage from the relatively
small number of high-cost metallurgical tests. Applying
geometallurgical principles to design of metallurgical sam-
pling programmes aligns sparse testwork data to abundant
geological data and ensures that the data is suitable for the
application of data science methods to derive robust predic-
tions of ore processing behaviour at the local scale.
The authors propose that the use of geometallurgical
principles and data science for metallurgical sampling and
test data analysis provides a measurable and repeatable basis
for sample-selection, driven by the geological characteris-
tics of the mineralization. It provides an alternative to the
plethora of rules of thumb sampling procedures that have
limited connection to the geological variability, nor the
variability of the metallurgical responses.
What Type Of Samples Do We Need?
Before any sampling or testwork is commissioned the pur-
pose of the testwork must be clear so that the right sample
type is selected. Drill core is a necessity for comminution
testing and coarse particle beneficiation. For mineral recov-
ery testwork (leaching, flotation, fine particle separation)
uncrushed drill core is preferred so that the sample can be
crushed in stages to produce a particle size distribution that
mimics the process plant design. Coarse rejects from core
samples crushed in a well-controlled laboratory may also be
suitable but Kormos et al (2013) note that assay rejects may
contain excessive fines, so checks on particle size distribu-
tion should be carried out to confirm suitability. Reverse
circulation drill cuttings may contain excessive fines which
may bias the test results.
From a geometallurgical perspective, drill core is
also preferred because much more geological data can be
obtained by visual logging, hyperspectral scanning, geo-
technical measurements, and non-destructive hardness
testing than can be obtained from chip samples. Drill core
data joined with metallurgical test results from the match-
ing depth intervals forms the essential input to predictive
Geometallurgical modelling.
There is an underlying and infrequently articulated
idea regarding representativity. That is, that a representative
sample can fully describe the behaviour of an ore deposit
and provide sufficient information to confirm that the
designed plant will operate in a uniform and predictable
manner. The reality is that such a design will perform in a
manner reflective of the variability of the ore feed over time
and the extremes of the ore feed will perform very differ-
ently from the representative sample. The best scenario is
a plant designed to perform within defined specifications
across all ore types.
Composite samples (typically multiple ore intercepts
from multiple drill holes) are used for developing the basic
processing and testwork flowsheet. Composites are made
from blends of the ore types expected to be mined over
the life of mine or other long production periods. Later,
large composites (bulk samples) may be required to provide
products for further testing such as concentrates or tailings.
Composite samples may establish a feasible process flow-
sheet but they provide little information about the behav-
iour of individual ore types.
To design a process plant that can accommodate all
the variation in ore characteristics, variability sampling is
required. Variability samples should be selected to char-
acterize the full range of geological ore types. The authors
recommend selection of variability samples from continu-
ous intercepts in single drill holes. This approach reduces
the risk of accidentally testing a blend of ore types and
also allows the metallurgical testwork results to be consid-
ered within their geospatial context. However, it should
be recognized that there is no intrinsic causal relationship
between geospatial location and metallurgical behaviour.
The geometallurgical principle for variability sampling
is to identify samples that cover the full range of the relevant
multivariate characteristics of the orebody. ‘Relevant’ refers
to those physical, mineralogical or chemical characteristics
that have the potential to affect ore processing responses.
We define an ore type as mineralized material of eco-
nomic interest with similar ore and gangue mineral assem-
blages, texture, and tenor (grade). In most orebodies, ore
types have a spatial arrangement that is the product of
the geological history of the mineral deposit. The term
‘domain’ is commonly applied to describe a volume of rock
with similar geological characteristics. From a geometallur-
gical perspective, it is preferable that a domain consists of a
single ore type so that, ultimately, estimates of ore process-
ing response derived from testing samples of an individual
ore type can be applied within the domain.
It is common practice to select metallurgical samples
from three-dimensional, geographic locations with the aim
of providing good spatial coverage of the deposit. This is
understandable in the absence of a good knowledge of the
spatial distribution of ore types and domains. However,
three-dimensional coverage is not a logical requirement of
metallurgical sampling. Ore processing response is intrinsi-
cally a function of geology not location. Therefore, from a
geometallurgical perspective, it is more important to achieve