XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3 867
volume sieve analysis with an RMSE of approximately 5%
and mass flow correlates well (R2 =0.964) with a nuclear
scale (Faucher et al., 2015). A single calibration is performed
on installation and recalibration is only needed if crushing
parameters change. Outputs from the system include:
• Full particle size distribution (%),cumulative pass-
ing fraction graphed
• Coarse particle fraction (%)
• Fine particle fraction (%)
• P50 (mm), configurable
• P80 (mm), configurable
• Conveyor speed (m/s)
• Volume flow rate (m3/h)
• Indicative mass flow rate (tons/h) based on assumed
bulk density
Data from the system is used for fragmentation monitoring
to optimize blasting, crusher gap control, empty belt detec-
tion, oversize detection to prevent damage to downstream
processes and equipment (e.g., HPGR), and to confirm suf-
ficient larger particles are present in SAG and AG mill feed.
PSD data can be used as an indicator of ore hardness and
mineralogy on mill feed. This data is not commonly used in
conjunction with data from other sensors.
Sensor fusion
Digitalization strategies can include the combination of
data from different sensors to derive information unavail-
able from any single sensor. For example, ore hardness
could be verified by elemental concentrations or ratios in
combination with PSD data. Clay content could be veri-
fied by combining elemental data with moisture data and
PSD to see if clumping is occurring, indicating potential
material handling problems. Moisture data can be used to
improve mass flow calculations from the PSD/volume/belt
speed analyzer.
Data from discussed technologies can be combined
with qualitative information from other sensors, such as
surface mineralogy measurement, to determine potential
effects on process performance. Geometallurgical under-
standing can be utilized more effectively in real time quality
control and improvement when the derived parameters are
included in process control considerations. Sensor fusion
may not help determine ore texture (as an example) which
is difficult to derive from any current sensor technologies
(elemental or mineralogical) unless there are measurable
coincident parameters.
DATA INTEGRATION
Sensor data is transmitted to plant control systems for anal-
ysis and determining responses to improve relevant opera-
tional processes (Figure 7). The importance of accurate,
representative and timely data cannot be under-estimated
when demonstrated benefits are valued in the millions or
tens of millions of dollars per year. The term garbage in,
garbage out is appropriate when considering sensor data
quality. Poor quality data can be worse than no data, as
changes made in response to poor data may cause poorer
outcomes than where there is no change to a process. An
example of this is in diversion decisions for ore and waste
where poorer measurement increases misallocation of ore
and waste causing increased metal losses, increased ore dilu-
tion and lower process feed quality.
Advanced sensors may contain internal verification
methods to determine if measurements are reliable and out-
put an error signal if malfunctioning to prevent otherwise
unreliable data being used in decision-making. Machine
learning and artificial intelligence techniques can be used
to improve data quality, however, data that is calculated
Source: Scantech International Pty Ltd
Figure 6. 3D Infrared camera PSD analyzer with photograph (for comparison) and analyzer video output of flow
volume sieve analysis with an RMSE of approximately 5%
and mass flow correlates well (R2 =0.964) with a nuclear
scale (Faucher et al., 2015). A single calibration is performed
on installation and recalibration is only needed if crushing
parameters change. Outputs from the system include:
• Full particle size distribution (%),cumulative pass-
ing fraction graphed
• Coarse particle fraction (%)
• Fine particle fraction (%)
• P50 (mm), configurable
• P80 (mm), configurable
• Conveyor speed (m/s)
• Volume flow rate (m3/h)
• Indicative mass flow rate (tons/h) based on assumed
bulk density
Data from the system is used for fragmentation monitoring
to optimize blasting, crusher gap control, empty belt detec-
tion, oversize detection to prevent damage to downstream
processes and equipment (e.g., HPGR), and to confirm suf-
ficient larger particles are present in SAG and AG mill feed.
PSD data can be used as an indicator of ore hardness and
mineralogy on mill feed. This data is not commonly used in
conjunction with data from other sensors.
Sensor fusion
Digitalization strategies can include the combination of
data from different sensors to derive information unavail-
able from any single sensor. For example, ore hardness
could be verified by elemental concentrations or ratios in
combination with PSD data. Clay content could be veri-
fied by combining elemental data with moisture data and
PSD to see if clumping is occurring, indicating potential
material handling problems. Moisture data can be used to
improve mass flow calculations from the PSD/volume/belt
speed analyzer.
Data from discussed technologies can be combined
with qualitative information from other sensors, such as
surface mineralogy measurement, to determine potential
effects on process performance. Geometallurgical under-
standing can be utilized more effectively in real time quality
control and improvement when the derived parameters are
included in process control considerations. Sensor fusion
may not help determine ore texture (as an example) which
is difficult to derive from any current sensor technologies
(elemental or mineralogical) unless there are measurable
coincident parameters.
DATA INTEGRATION
Sensor data is transmitted to plant control systems for anal-
ysis and determining responses to improve relevant opera-
tional processes (Figure 7). The importance of accurate,
representative and timely data cannot be under-estimated
when demonstrated benefits are valued in the millions or
tens of millions of dollars per year. The term garbage in,
garbage out is appropriate when considering sensor data
quality. Poor quality data can be worse than no data, as
changes made in response to poor data may cause poorer
outcomes than where there is no change to a process. An
example of this is in diversion decisions for ore and waste
where poorer measurement increases misallocation of ore
and waste causing increased metal losses, increased ore dilu-
tion and lower process feed quality.
Advanced sensors may contain internal verification
methods to determine if measurements are reliable and out-
put an error signal if malfunctioning to prevent otherwise
unreliable data being used in decision-making. Machine
learning and artificial intelligence techniques can be used
to improve data quality, however, data that is calculated
Source: Scantech International Pty Ltd
Figure 6. 3D Infrared camera PSD analyzer with photograph (for comparison) and analyzer video output of flow