1401
Geometallurgical Data Mining of Ore Processing Behavior and
the Integrated Value Chain
Lynnette Hutson, Isabel Barton
University of Arizona
Logan Hill, William Stavast
Freeport McMoRan
ABSTRACT: Geometallurgical characteristics of an orebody strongly influence processing outcomes. Explicitly
tracking and quantifying that impact is difficult. However, it is becoming increasingly feasible as modern
information systems can facilitate in-depth analytical models. Machine learning algorithms can then examine
relationships between the mineralogical qualities of the ore and the ensuing process performance.
As a test case, we investigate connections between in situ ore and downstream processing outcomes for a large
copper porphyry, using a combination of custom Python, discrete event simulation (DES), and data mining.
The focus is on performance variations with pyrite content and swelling clay content.
INTRODUCTION
In any mining and processing operation, the ultimate pro-
duction outcomes are in part a function of the geologic
quality of the orebody. The emerging field of geometallurgy
is tasked with elucidating the specific connections between
the geologic characteristics of the orebody and the pro-
duction outcomes such as recovery, throughput, and cost,
including the cost of processing-related consumables such
as acid, lime, or other reagents. Specific geologic attributes
which impact downstream processing outcomes include
details of the ore mineralogy, such as which minerals are
present and the relative abundance of each mineral, as well
as information on the grain size and texture for each ore
mineral. Other geologic attributes of interest include assay
results or other estimates of elemental abundances, such
as multi-element laboratory analyses. Additional geologic
information may also include details like the formation
name, lithology, and alteration style(s) present in the ore.
Of course, each of these characteristics will vary, and must
be regularly sampled throughout the deposit.
Conventionally, the geologic data corresponding to a
particular orebody is often operationally siloed from the
production information, so it is difficult to determine
exactly when a particular mined portion of the deposit is
processed, given the time lag between mining and when
the commodity of interest is recovered. In operations with
significant blending requirements, the material contained
in each shovel load of ore will be diffused and mixed with
ore from other portions of the deposit during processing,
thus making an exact linkage between the actual, observed
production outcome and the in-situ orebody difficult to
achieve. However, given the rate of technological change
and the explosion of ‘big data’-related industry applica-
tions, computationally intensive approaches which may
have been infeasible in the past are now within reach.
Such analyses have the potential to predict operational
Geometallurgical Data Mining of Ore Processing Behavior and
the Integrated Value Chain
Lynnette Hutson, Isabel Barton
University of Arizona
Logan Hill, William Stavast
Freeport McMoRan
ABSTRACT: Geometallurgical characteristics of an orebody strongly influence processing outcomes. Explicitly
tracking and quantifying that impact is difficult. However, it is becoming increasingly feasible as modern
information systems can facilitate in-depth analytical models. Machine learning algorithms can then examine
relationships between the mineralogical qualities of the ore and the ensuing process performance.
As a test case, we investigate connections between in situ ore and downstream processing outcomes for a large
copper porphyry, using a combination of custom Python, discrete event simulation (DES), and data mining.
The focus is on performance variations with pyrite content and swelling clay content.
INTRODUCTION
In any mining and processing operation, the ultimate pro-
duction outcomes are in part a function of the geologic
quality of the orebody. The emerging field of geometallurgy
is tasked with elucidating the specific connections between
the geologic characteristics of the orebody and the pro-
duction outcomes such as recovery, throughput, and cost,
including the cost of processing-related consumables such
as acid, lime, or other reagents. Specific geologic attributes
which impact downstream processing outcomes include
details of the ore mineralogy, such as which minerals are
present and the relative abundance of each mineral, as well
as information on the grain size and texture for each ore
mineral. Other geologic attributes of interest include assay
results or other estimates of elemental abundances, such
as multi-element laboratory analyses. Additional geologic
information may also include details like the formation
name, lithology, and alteration style(s) present in the ore.
Of course, each of these characteristics will vary, and must
be regularly sampled throughout the deposit.
Conventionally, the geologic data corresponding to a
particular orebody is often operationally siloed from the
production information, so it is difficult to determine
exactly when a particular mined portion of the deposit is
processed, given the time lag between mining and when
the commodity of interest is recovered. In operations with
significant blending requirements, the material contained
in each shovel load of ore will be diffused and mixed with
ore from other portions of the deposit during processing,
thus making an exact linkage between the actual, observed
production outcome and the in-situ orebody difficult to
achieve. However, given the rate of technological change
and the explosion of ‘big data’-related industry applica-
tions, computationally intensive approaches which may
have been infeasible in the past are now within reach.
Such analyses have the potential to predict operational
