XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3 1405
to pass through the crusher and additional distributions
covering the time needed to traverse the apron feeder
and conveyors. These distributions were developed using
Arena’s input analyzer using one year’s worth of throughput
and speed data. A summary of the distributions is shown
in Table 2.
Since the ore data is loaded into the model based on a
series of sequential files, the model setup requires the cre-
ation of a dummy truckload entity, combined with a delay
module and separate module for the model to correctly
read the input data and accurately account for the timing of
the ore entering the system. To avoid including this dummy
truckload into the cumulative statistics, there is a decide
module to dispose of it after it has been used to read the
data for the next truckload from the input file. The entities
with actual production data attached are allowed to con-
tinue through the rest of the model. A dispose module sits
at the end of the system to terminate the truckload entities.
As the model is designed to work as a capsule model within
a larger system, the outputs from this model include a series
of sequential files with a timestamp of when each truckload
entity reaches the intermediate ore stockpile at the end of
the model, along with the attribute value for each of the
geological attributes included. An overview of the model is
shown in Figure 2.
Downtime events are modeled in three ways as sum-
marized in Table 3. Shift change is included as a scheduled
down, with events occurring every 12 hours. The length of
the shift change is drawn from a probability distribution of
the actual downtimes incurred during shift change. Both
planned and unplanned downs are modeled using a mean
time between failure (MTBF) and mean time to repair
(MTTR) values. These distributions were also derived
from the maintenance dataset. Although planned downs
are scheduled, the frequency is variable and encompasses
several types of planned maintenance, so the model incor-
porates this as a random series of events. The unplanned
down distribution was based on the observed distribution
of downs excluding the geology-related causes. All units are
in seconds.
A fourth category of downtime is related to the geol-
ogy of the ore. As previously shown, most of the operating
problem downtimes are related to the hardness of the ore.
If a correlation between the hardness of the ore, as indicated
in the ore control dataset and the likelihood of a geology-
related ore down can be quantified, then the probability
of a downtime in the DES model can be assigned that
explicitly considers ore hardness. Similarly, if a high clay
content also increases the chance of an unplanned down,
any correlation to that effect can be used as an input to the
DES model where the greater the incidence of high-clay
ore, the more likely an unplanned down is to occur in the
model. Therefore, it is also necessary to look for correla-
tions between the ore control/dispatch and maintenance
datasets with plausible causal connections between the two.
Table 2. Summary of probability distributions used for the
model
Model Component Distribution
Crusher 3 throughput EXPO(1.56)
Apron Feeder LOGN(1.06, 0.555)
P11 Conveyor LOGN(1.06, 0.555)
P5 Conveyor NORM(99.4, 5.84)
P6 Conveyor 100 *BETA(42.1, 0.85)
Figure 2. Overview of the DES model of the primary crusher and associated intermediate ore stockpile, as configured using
Arena software
to pass through the crusher and additional distributions
covering the time needed to traverse the apron feeder
and conveyors. These distributions were developed using
Arena’s input analyzer using one year’s worth of throughput
and speed data. A summary of the distributions is shown
in Table 2.
Since the ore data is loaded into the model based on a
series of sequential files, the model setup requires the cre-
ation of a dummy truckload entity, combined with a delay
module and separate module for the model to correctly
read the input data and accurately account for the timing of
the ore entering the system. To avoid including this dummy
truckload into the cumulative statistics, there is a decide
module to dispose of it after it has been used to read the
data for the next truckload from the input file. The entities
with actual production data attached are allowed to con-
tinue through the rest of the model. A dispose module sits
at the end of the system to terminate the truckload entities.
As the model is designed to work as a capsule model within
a larger system, the outputs from this model include a series
of sequential files with a timestamp of when each truckload
entity reaches the intermediate ore stockpile at the end of
the model, along with the attribute value for each of the
geological attributes included. An overview of the model is
shown in Figure 2.
Downtime events are modeled in three ways as sum-
marized in Table 3. Shift change is included as a scheduled
down, with events occurring every 12 hours. The length of
the shift change is drawn from a probability distribution of
the actual downtimes incurred during shift change. Both
planned and unplanned downs are modeled using a mean
time between failure (MTBF) and mean time to repair
(MTTR) values. These distributions were also derived
from the maintenance dataset. Although planned downs
are scheduled, the frequency is variable and encompasses
several types of planned maintenance, so the model incor-
porates this as a random series of events. The unplanned
down distribution was based on the observed distribution
of downs excluding the geology-related causes. All units are
in seconds.
A fourth category of downtime is related to the geol-
ogy of the ore. As previously shown, most of the operating
problem downtimes are related to the hardness of the ore.
If a correlation between the hardness of the ore, as indicated
in the ore control dataset and the likelihood of a geology-
related ore down can be quantified, then the probability
of a downtime in the DES model can be assigned that
explicitly considers ore hardness. Similarly, if a high clay
content also increases the chance of an unplanned down,
any correlation to that effect can be used as an input to the
DES model where the greater the incidence of high-clay
ore, the more likely an unplanned down is to occur in the
model. Therefore, it is also necessary to look for correla-
tions between the ore control/dispatch and maintenance
datasets with plausible causal connections between the two.
Table 2. Summary of probability distributions used for the
model
Model Component Distribution
Crusher 3 throughput EXPO(1.56)
Apron Feeder LOGN(1.06, 0.555)
P11 Conveyor LOGN(1.06, 0.555)
P5 Conveyor NORM(99.4, 5.84)
P6 Conveyor 100 *BETA(42.1, 0.85)
Figure 2. Overview of the DES model of the primary crusher and associated intermediate ore stockpile, as configured using
Arena software