1076 XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3
dollar per tonne of product made for a defined ore source.
By adding ESG metrics to the constraints assessment, a
mining business’s corporate profile can be raised to meet
global standards.
One of the challenges in the current climate-focused
initiatives is which option to adopt for the changing goal
or objective (Figure 8). The traditional mining value-driven
approach maximises the mined ore’s productivity (mini-
mising $/t). The analysis method adopted here is more
applications-driven focus—i.e., ‘there is a solution to every
problem’—because the key performance indicators to drive
profitability are clear and easily quantifiable. Introducing
environmental metrics such as Greenhouse Gas Emissions
(GHGe), the performance metric (tonne of CO2) can be
qualitative, depending on the data source at Scope 1,2 and
3 assessment levels. Herein lies the alignment problem
how will a system enabled by AI know which path to take
and which path is optimal based on the productivity and
environmental protection requirements?
The Alignment Problem
In his book on the application of AI, Brian Christian
coined the term ‘The alignment problem’, particularly
in defining the value for humans using machine learning
(Christian, 2020). Just as the ‘value’ changes based on the
focused optimisation approach, alignment is required based
on the preferred approach. The definition of value changes
depending on the objective of a mining business (Table 1).
For a process plant, value is based on the measurement
of data using highly precise sensor data. For a developing
mine, the value is less accurate but inferred using statistical
analysis of drill-hole assay data, and for the community and
environment surrounding or in the vicinity of a mine, the
value here is subjective based on the feedback from the gov-
erning body protecting the environment and community.
Each data source is pertinent to developing and operating a
mine while supporting the community and protecting the
environment.
Applying the correct solution leveraging AI in mining
across the entire value chain is essential. The correct goals
or objectives must be clearly distinguished from preferences
as part of the initiative. Preference is a more subjective view,
whereas the objective or goal is more definitive—this will
be critical as the software that runs the predictive models
requires a clear emphasis on these definitions or weighting
incorrect terminology can be catastrophic to the mining
business and environment.
How much value can AI bring to the mining industry?
For a fully automated, data-driven mining operation,
incorporating AI systems into the mining value chain can
redefine the business’s and its customers’ value. Figure 9
presents the mine-to-market vision where the entire value
chain is linked—from the mine to the process plant defines
a typical mine-to-mill approach, a concept initially devel-
oped at the Julius Kruttschintt Mineral Research Center
in 2001 (Valery et al., 2001). Extending the value chain
from the mill (process plant) to the market requires a sys-
tem that analyses the metals market, driven by supply and
demand. To date, a commercially available value optimizer
Figure 8. Defining the objective in mining (Chandramohan, 2023)
dollar per tonne of product made for a defined ore source.
By adding ESG metrics to the constraints assessment, a
mining business’s corporate profile can be raised to meet
global standards.
One of the challenges in the current climate-focused
initiatives is which option to adopt for the changing goal
or objective (Figure 8). The traditional mining value-driven
approach maximises the mined ore’s productivity (mini-
mising $/t). The analysis method adopted here is more
applications-driven focus—i.e., ‘there is a solution to every
problem’—because the key performance indicators to drive
profitability are clear and easily quantifiable. Introducing
environmental metrics such as Greenhouse Gas Emissions
(GHGe), the performance metric (tonne of CO2) can be
qualitative, depending on the data source at Scope 1,2 and
3 assessment levels. Herein lies the alignment problem
how will a system enabled by AI know which path to take
and which path is optimal based on the productivity and
environmental protection requirements?
The Alignment Problem
In his book on the application of AI, Brian Christian
coined the term ‘The alignment problem’, particularly
in defining the value for humans using machine learning
(Christian, 2020). Just as the ‘value’ changes based on the
focused optimisation approach, alignment is required based
on the preferred approach. The definition of value changes
depending on the objective of a mining business (Table 1).
For a process plant, value is based on the measurement
of data using highly precise sensor data. For a developing
mine, the value is less accurate but inferred using statistical
analysis of drill-hole assay data, and for the community and
environment surrounding or in the vicinity of a mine, the
value here is subjective based on the feedback from the gov-
erning body protecting the environment and community.
Each data source is pertinent to developing and operating a
mine while supporting the community and protecting the
environment.
Applying the correct solution leveraging AI in mining
across the entire value chain is essential. The correct goals
or objectives must be clearly distinguished from preferences
as part of the initiative. Preference is a more subjective view,
whereas the objective or goal is more definitive—this will
be critical as the software that runs the predictive models
requires a clear emphasis on these definitions or weighting
incorrect terminology can be catastrophic to the mining
business and environment.
How much value can AI bring to the mining industry?
For a fully automated, data-driven mining operation,
incorporating AI systems into the mining value chain can
redefine the business’s and its customers’ value. Figure 9
presents the mine-to-market vision where the entire value
chain is linked—from the mine to the process plant defines
a typical mine-to-mill approach, a concept initially devel-
oped at the Julius Kruttschintt Mineral Research Center
in 2001 (Valery et al., 2001). Extending the value chain
from the mill (process plant) to the market requires a sys-
tem that analyses the metals market, driven by supply and
demand. To date, a commercially available value optimizer
Figure 8. Defining the objective in mining (Chandramohan, 2023)