XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3 1079
will likely require additional online measurement to refine
the operational decisions.
STEPS TO INCREASING AI SOLUTIONS
IN MINING
There is a close link between data availability and the appli-
cation of AI solutions for a mining operation. Table 3 pres-
ents an overview of the data status at each value driver of
an operating mine. Resource definition is the critical driver
for the entire mining business that defines value, which
also happens to be the most scarce in data resolution. With
measurement technology adapted in the mine and AI sys-
tems in place, ore-body knowledge can be increased in
accuracy.
The ultimate application of AI in the mining industry
is data-related, and the operation and extent of the automa-
tion use (Table 4). In the case of a lack of data, AI systems
can be trained to build the knowledge required to make
reasonable predictions. Gaining a robust understanding
of the ore body knowledge in the mining value chain can
provide tremendous opportunities to refine the operations.
Energy and water efficiency, ESG metrics, and commodity
Table 3. Where is the opportunity for AI-driven systems?
Value Driver Data Status Importance Definition AI Application
Resource
Definition
• Lacks high-resolution ore characteristics
data
• Ore source estimation and characteristics
defined by statistical analysis from a few
drill-hole samples
• Ore tracking is challenging unless excessive
instrumentation is used to measure
chemical proxies—which then can be used
to define the ore characteristics
• Resource definition is
critical as this defines
the revenue of the entire
mining operation
• Increased statistical analysis of
the sparse data
• Train online elemental
analyzers such as XRF or
PGNAA scanners to predict ore
characteristics
Process
Automation
• High-resolution data collected through
instrumentation
• Advanced process and model predictive
controllers are recommended solutions
for highly variable feed with vary ore
characteristics
• The objective of process
automation is to achieve
process stability for highly
variable feed
• Achieving stability is
pertinent for variable ore
characteristics
• AI is used in some APC or
MPC systems to increase the
model’s predictability for control
applications. The application
range is tightly controlled within
narrow controller limits to refine
the process stability
Asset
Diagnostics
• High-resolution data available on unit asset
performance
• Devices such as Edge Sensors are becoming
cheap/high bandwidth solutions for real-
time monitoring
• Real-time asset monitoring
is pertinent to predict
the next failure event.
Realtime asset monitoring
increases the overall asset
utilization and the overall
system availability
• AI applications are used
to predict the next failure
event. Key data inputs are
ore characteristic changes,
equipment operational change
behaviour and overall life of
consumables
Constraints
Identification
• The only data source required for constraint
identification is the resource—ore
characteristics, zones, fragmentation, etc..
• Constraints identification mimics
the system’s performance in a virtual
environment—a ‘digital twin’. Digital
twins/simulations of the mining as the
whole system do not require real-time data
to predict the constraints in the system.
Instead, simulations of the entire mining
process are modelled using first principle
models and empirical relationships to
identify critical bottlenecks.
• Constraints identification
is essential to define the
operational strategy if
certain areas of the systems
are constrained. Realtime
data from the system can
be used to update the
digital twin outcomes to
refine the strategy for the
entire system
• Digital twins enabled by AI
algorithms can run multiple
scenarios to optimize the entire
system. Operating strategies
can be refined based on the
bottlenecks in the flowsheet.
will likely require additional online measurement to refine
the operational decisions.
STEPS TO INCREASING AI SOLUTIONS
IN MINING
There is a close link between data availability and the appli-
cation of AI solutions for a mining operation. Table 3 pres-
ents an overview of the data status at each value driver of
an operating mine. Resource definition is the critical driver
for the entire mining business that defines value, which
also happens to be the most scarce in data resolution. With
measurement technology adapted in the mine and AI sys-
tems in place, ore-body knowledge can be increased in
accuracy.
The ultimate application of AI in the mining industry
is data-related, and the operation and extent of the automa-
tion use (Table 4). In the case of a lack of data, AI systems
can be trained to build the knowledge required to make
reasonable predictions. Gaining a robust understanding
of the ore body knowledge in the mining value chain can
provide tremendous opportunities to refine the operations.
Energy and water efficiency, ESG metrics, and commodity
Table 3. Where is the opportunity for AI-driven systems?
Value Driver Data Status Importance Definition AI Application
Resource
Definition
• Lacks high-resolution ore characteristics
data
• Ore source estimation and characteristics
defined by statistical analysis from a few
drill-hole samples
• Ore tracking is challenging unless excessive
instrumentation is used to measure
chemical proxies—which then can be used
to define the ore characteristics
• Resource definition is
critical as this defines
the revenue of the entire
mining operation
• Increased statistical analysis of
the sparse data
• Train online elemental
analyzers such as XRF or
PGNAA scanners to predict ore
characteristics
Process
Automation
• High-resolution data collected through
instrumentation
• Advanced process and model predictive
controllers are recommended solutions
for highly variable feed with vary ore
characteristics
• The objective of process
automation is to achieve
process stability for highly
variable feed
• Achieving stability is
pertinent for variable ore
characteristics
• AI is used in some APC or
MPC systems to increase the
model’s predictability for control
applications. The application
range is tightly controlled within
narrow controller limits to refine
the process stability
Asset
Diagnostics
• High-resolution data available on unit asset
performance
• Devices such as Edge Sensors are becoming
cheap/high bandwidth solutions for real-
time monitoring
• Real-time asset monitoring
is pertinent to predict
the next failure event.
Realtime asset monitoring
increases the overall asset
utilization and the overall
system availability
• AI applications are used
to predict the next failure
event. Key data inputs are
ore characteristic changes,
equipment operational change
behaviour and overall life of
consumables
Constraints
Identification
• The only data source required for constraint
identification is the resource—ore
characteristics, zones, fragmentation, etc..
• Constraints identification mimics
the system’s performance in a virtual
environment—a ‘digital twin’. Digital
twins/simulations of the mining as the
whole system do not require real-time data
to predict the constraints in the system.
Instead, simulations of the entire mining
process are modelled using first principle
models and empirical relationships to
identify critical bottlenecks.
• Constraints identification
is essential to define the
operational strategy if
certain areas of the systems
are constrained. Realtime
data from the system can
be used to update the
digital twin outcomes to
refine the strategy for the
entire system
• Digital twins enabled by AI
algorithms can run multiple
scenarios to optimize the entire
system. Operating strategies
can be refined based on the
bottlenecks in the flowsheet.